xref: /dpdk/drivers/net/bnxt/bnxt_ethdev.c (revision 8879ece56832b7f2dd5e5ee10f36e848089f09df)
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(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(type, size, 0);
590 	if (ctx->va == NULL)
591 		return -ENOMEM;
592 	rte_mem_lock_page(ctx->va);
593 	ctx->size = size;
594 	ctx->dma = rte_mem_virt2iova(ctx->va);
595 	if (ctx->dma == RTE_BAD_IOVA)
596 		return -ENOMEM;
597 
598 	return 0;
599 }
600 
601 static int bnxt_init_fc_ctx_mem(struct bnxt *bp)
602 {
603 	struct rte_pci_device *pdev = bp->pdev;
604 	char type[RTE_MEMZONE_NAMESIZE];
605 	uint16_t max_fc;
606 	int rc = 0;
607 
608 	max_fc = bp->flow_stat->max_fc;
609 
610 	sprintf(type, "bnxt_rx_fc_in_" PCI_PRI_FMT, pdev->addr.domain,
611 		pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
612 	/* 4 bytes for each counter-id */
613 	rc = bnxt_alloc_ctx_mem_buf(type,
614 				    max_fc * 4,
615 				    &bp->flow_stat->rx_fc_in_tbl);
616 	if (rc)
617 		return rc;
618 
619 	sprintf(type, "bnxt_rx_fc_out_" PCI_PRI_FMT, pdev->addr.domain,
620 		pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
621 	/* 16 bytes for each counter - 8 bytes pkt_count, 8 bytes byte_count */
622 	rc = bnxt_alloc_ctx_mem_buf(type,
623 				    max_fc * 16,
624 				    &bp->flow_stat->rx_fc_out_tbl);
625 	if (rc)
626 		return rc;
627 
628 	sprintf(type, "bnxt_tx_fc_in_" PCI_PRI_FMT, pdev->addr.domain,
629 		pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
630 	/* 4 bytes for each counter-id */
631 	rc = bnxt_alloc_ctx_mem_buf(type,
632 				    max_fc * 4,
633 				    &bp->flow_stat->tx_fc_in_tbl);
634 	if (rc)
635 		return rc;
636 
637 	sprintf(type, "bnxt_tx_fc_out_" PCI_PRI_FMT, pdev->addr.domain,
638 		pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
639 	/* 16 bytes for each counter - 8 bytes pkt_count, 8 bytes byte_count */
640 	rc = bnxt_alloc_ctx_mem_buf(type,
641 				    max_fc * 16,
642 				    &bp->flow_stat->tx_fc_out_tbl);
643 	if (rc)
644 		return rc;
645 
646 	rc = bnxt_register_fc_ctx_mem(bp);
647 
648 	return rc;
649 }
650 
651 static int bnxt_init_ctx_mem(struct bnxt *bp)
652 {
653 	int rc = 0;
654 
655 	if (!(bp->fw_cap & BNXT_FW_CAP_ADV_FLOW_COUNTERS) ||
656 	    !(BNXT_PF(bp) || BNXT_VF_IS_TRUSTED(bp)) ||
657 	    !BNXT_FLOW_XSTATS_EN(bp))
658 		return 0;
659 
660 	rc = bnxt_hwrm_cfa_counter_qcaps(bp, &bp->flow_stat->max_fc);
661 	if (rc)
662 		return rc;
663 
664 	rc = bnxt_init_fc_ctx_mem(bp);
665 
666 	return rc;
667 }
668 
669 static int bnxt_update_phy_setting(struct bnxt *bp)
670 {
671 	struct rte_eth_link new;
672 	int rc;
673 
674 	rc = bnxt_get_hwrm_link_config(bp, &new);
675 	if (rc) {
676 		PMD_DRV_LOG(ERR, "Failed to get link settings\n");
677 		return rc;
678 	}
679 
680 	/*
681 	 * On BCM957508-N2100 adapters, FW will not allow any user other
682 	 * than BMC to shutdown the port. bnxt_get_hwrm_link_config() call
683 	 * always returns link up. Force phy update always in that case.
684 	 */
685 	if (!new.link_status || IS_BNXT_DEV_957508_N2100(bp)) {
686 		rc = bnxt_set_hwrm_link_config(bp, true);
687 		if (rc) {
688 			PMD_DRV_LOG(ERR, "Failed to update PHY settings\n");
689 			return rc;
690 		}
691 	}
692 
693 	return rc;
694 }
695 
696 static void bnxt_free_prev_ring_stats(struct bnxt *bp)
697 {
698 	rte_free(bp->prev_rx_ring_stats);
699 	rte_free(bp->prev_tx_ring_stats);
700 
701 	bp->prev_rx_ring_stats = NULL;
702 	bp->prev_tx_ring_stats = NULL;
703 }
704 
705 static int bnxt_alloc_prev_ring_stats(struct bnxt *bp)
706 {
707 	bp->prev_rx_ring_stats =  rte_zmalloc("bnxt_prev_rx_ring_stats",
708 					      sizeof(struct bnxt_ring_stats) *
709 					      bp->rx_cp_nr_rings,
710 					      0);
711 	if (bp->prev_rx_ring_stats == NULL)
712 		return -ENOMEM;
713 
714 	bp->prev_tx_ring_stats = rte_zmalloc("bnxt_prev_tx_ring_stats",
715 					     sizeof(struct bnxt_ring_stats) *
716 					     bp->tx_cp_nr_rings,
717 					     0);
718 	if (bp->prev_tx_ring_stats == NULL)
719 		goto error;
720 
721 	return 0;
722 
723 error:
724 	bnxt_free_prev_ring_stats(bp);
725 	return -ENOMEM;
726 }
727 
728 static int bnxt_start_nic(struct bnxt *bp)
729 {
730 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(bp->eth_dev);
731 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
732 	uint32_t intr_vector = 0;
733 	uint32_t queue_id, base = BNXT_MISC_VEC_ID;
734 	uint32_t vec = BNXT_MISC_VEC_ID;
735 	unsigned int i, j;
736 	int rc;
737 
738 	if (bp->eth_dev->data->mtu > RTE_ETHER_MTU) {
739 		bp->eth_dev->data->dev_conf.rxmode.offloads |=
740 			DEV_RX_OFFLOAD_JUMBO_FRAME;
741 		bp->flags |= BNXT_FLAG_JUMBO;
742 	} else {
743 		bp->eth_dev->data->dev_conf.rxmode.offloads &=
744 			~DEV_RX_OFFLOAD_JUMBO_FRAME;
745 		bp->flags &= ~BNXT_FLAG_JUMBO;
746 	}
747 
748 	/* THOR does not support ring groups.
749 	 * But we will use the array to save RSS context IDs.
750 	 */
751 	if (BNXT_CHIP_P5(bp))
752 		bp->max_ring_grps = BNXT_MAX_RSS_CTXTS_P5;
753 
754 	rc = bnxt_alloc_all_hwrm_stat_ctxs(bp);
755 	if (rc) {
756 		PMD_DRV_LOG(ERR, "HWRM stat ctx alloc failure rc: %x\n", rc);
757 		goto err_out;
758 	}
759 
760 	rc = bnxt_alloc_hwrm_rings(bp);
761 	if (rc) {
762 		PMD_DRV_LOG(ERR, "HWRM ring alloc failure rc: %x\n", rc);
763 		goto err_out;
764 	}
765 
766 	rc = bnxt_alloc_all_hwrm_ring_grps(bp);
767 	if (rc) {
768 		PMD_DRV_LOG(ERR, "HWRM ring grp alloc failure: %x\n", rc);
769 		goto err_out;
770 	}
771 
772 	if (!(bp->vnic_cap_flags & BNXT_VNIC_CAP_COS_CLASSIFY))
773 		goto skip_cosq_cfg;
774 
775 	for (j = 0, i = 0; i < BNXT_COS_QUEUE_COUNT; i++) {
776 		if (bp->rx_cos_queue[i].id != 0xff) {
777 			struct bnxt_vnic_info *vnic = &bp->vnic_info[j++];
778 
779 			if (!vnic) {
780 				PMD_DRV_LOG(ERR,
781 					    "Num pools more than FW profile\n");
782 				rc = -EINVAL;
783 				goto err_out;
784 			}
785 			vnic->cos_queue_id = bp->rx_cos_queue[i].id;
786 			bp->rx_cosq_cnt++;
787 		}
788 	}
789 
790 skip_cosq_cfg:
791 	rc = bnxt_mq_rx_configure(bp);
792 	if (rc) {
793 		PMD_DRV_LOG(ERR, "MQ mode configure failure rc: %x\n", rc);
794 		goto err_out;
795 	}
796 
797 	/* default vnic 0 */
798 	rc = bnxt_setup_one_vnic(bp, 0);
799 	if (rc)
800 		goto err_out;
801 	/* VNIC configuration */
802 	if (BNXT_RFS_NEEDS_VNIC(bp)) {
803 		for (i = 1; i < bp->nr_vnics; i++) {
804 			rc = bnxt_setup_one_vnic(bp, i);
805 			if (rc)
806 				goto err_out;
807 		}
808 	}
809 
810 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, &bp->vnic_info[0], 0, NULL);
811 	if (rc) {
812 		PMD_DRV_LOG(ERR,
813 			"HWRM cfa l2 rx mask failure rc: %x\n", rc);
814 		goto err_out;
815 	}
816 
817 	/* check and configure queue intr-vector mapping */
818 	if ((rte_intr_cap_multiple(intr_handle) ||
819 	     !RTE_ETH_DEV_SRIOV(bp->eth_dev).active) &&
820 	    bp->eth_dev->data->dev_conf.intr_conf.rxq != 0) {
821 		intr_vector = bp->eth_dev->data->nb_rx_queues;
822 		PMD_DRV_LOG(DEBUG, "intr_vector = %d\n", intr_vector);
823 		if (intr_vector > bp->rx_cp_nr_rings) {
824 			PMD_DRV_LOG(ERR, "At most %d intr queues supported",
825 					bp->rx_cp_nr_rings);
826 			return -ENOTSUP;
827 		}
828 		rc = rte_intr_efd_enable(intr_handle, intr_vector);
829 		if (rc)
830 			return rc;
831 	}
832 
833 	if (rte_intr_dp_is_en(intr_handle) && !intr_handle->intr_vec) {
834 		intr_handle->intr_vec =
835 			rte_zmalloc("intr_vec",
836 				    bp->eth_dev->data->nb_rx_queues *
837 				    sizeof(int), 0);
838 		if (intr_handle->intr_vec == NULL) {
839 			PMD_DRV_LOG(ERR, "Failed to allocate %d rx_queues"
840 				" intr_vec", bp->eth_dev->data->nb_rx_queues);
841 			rc = -ENOMEM;
842 			goto err_out;
843 		}
844 		PMD_DRV_LOG(DEBUG, "intr_handle->intr_vec = %p "
845 			"intr_handle->nb_efd = %d intr_handle->max_intr = %d\n",
846 			 intr_handle->intr_vec, intr_handle->nb_efd,
847 			intr_handle->max_intr);
848 		for (queue_id = 0; queue_id < bp->eth_dev->data->nb_rx_queues;
849 		     queue_id++) {
850 			intr_handle->intr_vec[queue_id] =
851 							vec + BNXT_RX_VEC_START;
852 			if (vec < base + intr_handle->nb_efd - 1)
853 				vec++;
854 		}
855 	}
856 
857 	/* enable uio/vfio intr/eventfd mapping */
858 	rc = rte_intr_enable(intr_handle);
859 #ifndef RTE_EXEC_ENV_FREEBSD
860 	/* In FreeBSD OS, nic_uio driver does not support interrupts */
861 	if (rc)
862 		goto err_out;
863 #endif
864 
865 	rc = bnxt_update_phy_setting(bp);
866 	if (rc)
867 		goto err_out;
868 
869 	bp->mark_table = rte_zmalloc("bnxt_mark_table", BNXT_MARK_TABLE_SZ, 0);
870 	if (!bp->mark_table)
871 		PMD_DRV_LOG(ERR, "Allocation of mark table failed\n");
872 
873 	return 0;
874 
875 err_out:
876 	/* Some of the error status returned by FW may not be from errno.h */
877 	if (rc > 0)
878 		rc = -EIO;
879 
880 	return rc;
881 }
882 
883 static int bnxt_shutdown_nic(struct bnxt *bp)
884 {
885 	bnxt_free_all_hwrm_resources(bp);
886 	bnxt_free_all_filters(bp);
887 	bnxt_free_all_vnics(bp);
888 	return 0;
889 }
890 
891 /*
892  * Device configuration and status function
893  */
894 
895 uint32_t bnxt_get_speed_capabilities(struct bnxt *bp)
896 {
897 	uint32_t link_speed = 0;
898 	uint32_t speed_capa = 0;
899 
900 	if (bp->link_info == NULL)
901 		return 0;
902 
903 	link_speed = bp->link_info->support_speeds;
904 
905 	/* If PAM4 is configured, use PAM4 supported speed */
906 	if (link_speed == 0 && bp->link_info->support_pam4_speeds > 0)
907 		link_speed = bp->link_info->support_pam4_speeds;
908 
909 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_LINK_SPEED_100MB)
910 		speed_capa |= ETH_LINK_SPEED_100M;
911 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_100MBHD)
912 		speed_capa |= ETH_LINK_SPEED_100M_HD;
913 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_1GB)
914 		speed_capa |= ETH_LINK_SPEED_1G;
915 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_2_5GB)
916 		speed_capa |= ETH_LINK_SPEED_2_5G;
917 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_10GB)
918 		speed_capa |= ETH_LINK_SPEED_10G;
919 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_20GB)
920 		speed_capa |= ETH_LINK_SPEED_20G;
921 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_25GB)
922 		speed_capa |= ETH_LINK_SPEED_25G;
923 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_40GB)
924 		speed_capa |= ETH_LINK_SPEED_40G;
925 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_50GB)
926 		speed_capa |= ETH_LINK_SPEED_50G;
927 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_100GB)
928 		speed_capa |= ETH_LINK_SPEED_100G;
929 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_PAM4_SPEEDS_50G)
930 		speed_capa |= ETH_LINK_SPEED_50G;
931 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_PAM4_SPEEDS_100G)
932 		speed_capa |= ETH_LINK_SPEED_100G;
933 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_PAM4_SPEEDS_200G)
934 		speed_capa |= ETH_LINK_SPEED_200G;
935 
936 	if (bp->link_info->auto_mode ==
937 	    HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_MODE_NONE)
938 		speed_capa |= ETH_LINK_SPEED_FIXED;
939 
940 	return speed_capa;
941 }
942 
943 static int bnxt_dev_info_get_op(struct rte_eth_dev *eth_dev,
944 				struct rte_eth_dev_info *dev_info)
945 {
946 	struct rte_pci_device *pdev = RTE_DEV_TO_PCI(eth_dev->device);
947 	struct bnxt *bp = eth_dev->data->dev_private;
948 	uint16_t max_vnics, i, j, vpool, vrxq;
949 	unsigned int max_rx_rings;
950 	int rc;
951 
952 	rc = is_bnxt_in_error(bp);
953 	if (rc)
954 		return rc;
955 
956 	/* MAC Specifics */
957 	dev_info->max_mac_addrs = bp->max_l2_ctx;
958 	dev_info->max_hash_mac_addrs = 0;
959 
960 	/* PF/VF specifics */
961 	if (BNXT_PF(bp))
962 		dev_info->max_vfs = pdev->max_vfs;
963 
964 	max_rx_rings = bnxt_max_rings(bp);
965 	/* For the sake of symmetry, max_rx_queues = max_tx_queues */
966 	dev_info->max_rx_queues = max_rx_rings;
967 	dev_info->max_tx_queues = max_rx_rings;
968 	dev_info->reta_size = bnxt_rss_hash_tbl_size(bp);
969 	dev_info->hash_key_size = HW_HASH_KEY_SIZE;
970 	max_vnics = bp->max_vnics;
971 
972 	/* MTU specifics */
973 	dev_info->min_mtu = RTE_ETHER_MIN_MTU;
974 	dev_info->max_mtu = BNXT_MAX_MTU;
975 
976 	/* Fast path specifics */
977 	dev_info->min_rx_bufsize = 1;
978 	dev_info->max_rx_pktlen = BNXT_MAX_PKT_LEN;
979 
980 	dev_info->rx_offload_capa = BNXT_DEV_RX_OFFLOAD_SUPPORT;
981 	if (bp->flags & BNXT_FLAG_PTP_SUPPORTED)
982 		dev_info->rx_offload_capa |= DEV_RX_OFFLOAD_TIMESTAMP;
983 	dev_info->tx_queue_offload_capa = DEV_TX_OFFLOAD_MBUF_FAST_FREE;
984 	dev_info->tx_offload_capa = BNXT_DEV_TX_OFFLOAD_SUPPORT |
985 				    dev_info->tx_queue_offload_capa;
986 	dev_info->flow_type_rss_offloads = BNXT_ETH_RSS_SUPPORT;
987 
988 	dev_info->speed_capa = bnxt_get_speed_capabilities(bp);
989 
990 	dev_info->default_rxconf = (struct rte_eth_rxconf) {
991 		.rx_thresh = {
992 			.pthresh = 8,
993 			.hthresh = 8,
994 			.wthresh = 0,
995 		},
996 		.rx_free_thresh = 32,
997 		.rx_drop_en = BNXT_DEFAULT_RX_DROP_EN,
998 	};
999 
1000 	dev_info->default_txconf = (struct rte_eth_txconf) {
1001 		.tx_thresh = {
1002 			.pthresh = 32,
1003 			.hthresh = 0,
1004 			.wthresh = 0,
1005 		},
1006 		.tx_free_thresh = 32,
1007 		.tx_rs_thresh = 32,
1008 	};
1009 	eth_dev->data->dev_conf.intr_conf.lsc = 1;
1010 
1011 	eth_dev->data->dev_conf.intr_conf.rxq = 1;
1012 	dev_info->rx_desc_lim.nb_min = BNXT_MIN_RING_DESC;
1013 	dev_info->rx_desc_lim.nb_max = BNXT_MAX_RX_RING_DESC;
1014 	dev_info->tx_desc_lim.nb_min = BNXT_MIN_RING_DESC;
1015 	dev_info->tx_desc_lim.nb_max = BNXT_MAX_TX_RING_DESC;
1016 
1017 	if (BNXT_PF(bp) || BNXT_VF_IS_TRUSTED(bp)) {
1018 		dev_info->switch_info.name = eth_dev->device->name;
1019 		dev_info->switch_info.domain_id = bp->switch_domain_id;
1020 		dev_info->switch_info.port_id =
1021 				BNXT_PF(bp) ? BNXT_SWITCH_PORT_ID_PF :
1022 				    BNXT_SWITCH_PORT_ID_TRUSTED_VF;
1023 	}
1024 
1025 	/*
1026 	 * TODO: default_rxconf, default_txconf, rx_desc_lim, and tx_desc_lim
1027 	 *       need further investigation.
1028 	 */
1029 
1030 	/* VMDq resources */
1031 	vpool = 64; /* ETH_64_POOLS */
1032 	vrxq = 128; /* ETH_VMDQ_DCB_NUM_QUEUES */
1033 	for (i = 0; i < 4; vpool >>= 1, i++) {
1034 		if (max_vnics > vpool) {
1035 			for (j = 0; j < 5; vrxq >>= 1, j++) {
1036 				if (dev_info->max_rx_queues > vrxq) {
1037 					if (vpool > vrxq)
1038 						vpool = vrxq;
1039 					goto found;
1040 				}
1041 			}
1042 			/* Not enough resources to support VMDq */
1043 			break;
1044 		}
1045 	}
1046 	/* Not enough resources to support VMDq */
1047 	vpool = 0;
1048 	vrxq = 0;
1049 found:
1050 	dev_info->max_vmdq_pools = vpool;
1051 	dev_info->vmdq_queue_num = vrxq;
1052 
1053 	dev_info->vmdq_pool_base = 0;
1054 	dev_info->vmdq_queue_base = 0;
1055 
1056 	return 0;
1057 }
1058 
1059 /* Configure the device based on the configuration provided */
1060 static int bnxt_dev_configure_op(struct rte_eth_dev *eth_dev)
1061 {
1062 	struct bnxt *bp = eth_dev->data->dev_private;
1063 	uint64_t rx_offloads = eth_dev->data->dev_conf.rxmode.offloads;
1064 	int rc;
1065 
1066 	bp->rx_queues = (void *)eth_dev->data->rx_queues;
1067 	bp->tx_queues = (void *)eth_dev->data->tx_queues;
1068 	bp->tx_nr_rings = eth_dev->data->nb_tx_queues;
1069 	bp->rx_nr_rings = eth_dev->data->nb_rx_queues;
1070 
1071 	rc = is_bnxt_in_error(bp);
1072 	if (rc)
1073 		return rc;
1074 
1075 	if (BNXT_VF(bp) && (bp->flags & BNXT_FLAG_NEW_RM)) {
1076 		rc = bnxt_hwrm_check_vf_rings(bp);
1077 		if (rc) {
1078 			PMD_DRV_LOG(ERR, "HWRM insufficient resources\n");
1079 			return -ENOSPC;
1080 		}
1081 
1082 		/* If a resource has already been allocated - in this case
1083 		 * it is the async completion ring, free it. Reallocate it after
1084 		 * resource reservation. This will ensure the resource counts
1085 		 * are calculated correctly.
1086 		 */
1087 
1088 		pthread_mutex_lock(&bp->def_cp_lock);
1089 
1090 		if (!BNXT_HAS_NQ(bp) && bp->async_cp_ring) {
1091 			bnxt_disable_int(bp);
1092 			bnxt_free_cp_ring(bp, bp->async_cp_ring);
1093 		}
1094 
1095 		rc = bnxt_hwrm_func_reserve_vf_resc(bp, false);
1096 		if (rc) {
1097 			PMD_DRV_LOG(ERR, "HWRM resource alloc fail:%x\n", rc);
1098 			pthread_mutex_unlock(&bp->def_cp_lock);
1099 			return -ENOSPC;
1100 		}
1101 
1102 		if (!BNXT_HAS_NQ(bp) && bp->async_cp_ring) {
1103 			rc = bnxt_alloc_async_cp_ring(bp);
1104 			if (rc) {
1105 				pthread_mutex_unlock(&bp->def_cp_lock);
1106 				return rc;
1107 			}
1108 			bnxt_enable_int(bp);
1109 		}
1110 
1111 		pthread_mutex_unlock(&bp->def_cp_lock);
1112 	}
1113 
1114 	/* Inherit new configurations */
1115 	if (eth_dev->data->nb_rx_queues > bp->max_rx_rings ||
1116 	    eth_dev->data->nb_tx_queues > bp->max_tx_rings ||
1117 	    eth_dev->data->nb_rx_queues + eth_dev->data->nb_tx_queues
1118 		+ BNXT_NUM_ASYNC_CPR(bp) > bp->max_cp_rings ||
1119 	    eth_dev->data->nb_rx_queues + eth_dev->data->nb_tx_queues >
1120 	    bp->max_stat_ctx)
1121 		goto resource_error;
1122 
1123 	if (BNXT_HAS_RING_GRPS(bp) &&
1124 	    (uint32_t)(eth_dev->data->nb_rx_queues) > bp->max_ring_grps)
1125 		goto resource_error;
1126 
1127 	if (!(eth_dev->data->dev_conf.rxmode.mq_mode & ETH_MQ_RX_RSS) &&
1128 	    bp->max_vnics < eth_dev->data->nb_rx_queues)
1129 		goto resource_error;
1130 
1131 	bp->rx_cp_nr_rings = bp->rx_nr_rings;
1132 	bp->tx_cp_nr_rings = bp->tx_nr_rings;
1133 
1134 	if (eth_dev->data->dev_conf.rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG)
1135 		rx_offloads |= DEV_RX_OFFLOAD_RSS_HASH;
1136 	eth_dev->data->dev_conf.rxmode.offloads = rx_offloads;
1137 
1138 	if (rx_offloads & DEV_RX_OFFLOAD_JUMBO_FRAME) {
1139 		eth_dev->data->mtu =
1140 			eth_dev->data->dev_conf.rxmode.max_rx_pkt_len -
1141 			RTE_ETHER_HDR_LEN - RTE_ETHER_CRC_LEN - VLAN_TAG_SIZE *
1142 			BNXT_NUM_VLANS;
1143 		bnxt_mtu_set_op(eth_dev, eth_dev->data->mtu);
1144 	}
1145 	return 0;
1146 
1147 resource_error:
1148 	PMD_DRV_LOG(ERR,
1149 		    "Insufficient resources to support requested config\n");
1150 	PMD_DRV_LOG(ERR,
1151 		    "Num Queues Requested: Tx %d, Rx %d\n",
1152 		    eth_dev->data->nb_tx_queues,
1153 		    eth_dev->data->nb_rx_queues);
1154 	PMD_DRV_LOG(ERR,
1155 		    "MAX: TxQ %d, RxQ %d, CQ %d Stat %d, Grp %d, Vnic %d\n",
1156 		    bp->max_tx_rings, bp->max_rx_rings, bp->max_cp_rings,
1157 		    bp->max_stat_ctx, bp->max_ring_grps, bp->max_vnics);
1158 	return -ENOSPC;
1159 }
1160 
1161 void bnxt_print_link_info(struct rte_eth_dev *eth_dev)
1162 {
1163 	struct rte_eth_link *link = &eth_dev->data->dev_link;
1164 
1165 	if (link->link_status)
1166 		PMD_DRV_LOG(INFO, "Port %d Link Up - speed %u Mbps - %s\n",
1167 			eth_dev->data->port_id,
1168 			(uint32_t)link->link_speed,
1169 			(link->link_duplex == ETH_LINK_FULL_DUPLEX) ?
1170 			("full-duplex") : ("half-duplex\n"));
1171 	else
1172 		PMD_DRV_LOG(INFO, "Port %d Link Down\n",
1173 			eth_dev->data->port_id);
1174 }
1175 
1176 /*
1177  * Determine whether the current configuration requires support for scattered
1178  * receive; return 1 if scattered receive is required and 0 if not.
1179  */
1180 static int bnxt_scattered_rx(struct rte_eth_dev *eth_dev)
1181 {
1182 	uint16_t buf_size;
1183 	int i;
1184 
1185 	if (eth_dev->data->dev_conf.rxmode.offloads & DEV_RX_OFFLOAD_SCATTER)
1186 		return 1;
1187 
1188 	if (eth_dev->data->dev_conf.rxmode.offloads & DEV_RX_OFFLOAD_TCP_LRO)
1189 		return 1;
1190 
1191 	for (i = 0; i < eth_dev->data->nb_rx_queues; i++) {
1192 		struct bnxt_rx_queue *rxq = eth_dev->data->rx_queues[i];
1193 
1194 		buf_size = (uint16_t)(rte_pktmbuf_data_room_size(rxq->mb_pool) -
1195 				      RTE_PKTMBUF_HEADROOM);
1196 		if (eth_dev->data->dev_conf.rxmode.max_rx_pkt_len > buf_size)
1197 			return 1;
1198 	}
1199 	return 0;
1200 }
1201 
1202 static eth_rx_burst_t
1203 bnxt_receive_function(struct rte_eth_dev *eth_dev)
1204 {
1205 	struct bnxt *bp = eth_dev->data->dev_private;
1206 
1207 	/* Disable vector mode RX for Stingray2 for now */
1208 	if (BNXT_CHIP_SR2(bp)) {
1209 		bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
1210 		return bnxt_recv_pkts;
1211 	}
1212 
1213 #if (defined(RTE_ARCH_X86) || defined(RTE_ARCH_ARM64)) && \
1214 	!defined(RTE_LIBRTE_IEEE1588)
1215 
1216 	/* Vector mode receive cannot be enabled if scattered rx is in use. */
1217 	if (eth_dev->data->scattered_rx)
1218 		goto use_scalar_rx;
1219 
1220 	/*
1221 	 * Vector mode receive cannot be enabled if Truflow is enabled or if
1222 	 * asynchronous completions and receive completions can be placed in
1223 	 * the same completion ring.
1224 	 */
1225 	if (BNXT_TRUFLOW_EN(bp) || !BNXT_NUM_ASYNC_CPR(bp))
1226 		goto use_scalar_rx;
1227 
1228 	/*
1229 	 * Vector mode receive cannot be enabled if any receive offloads outside
1230 	 * a limited subset have been enabled.
1231 	 */
1232 	if (eth_dev->data->dev_conf.rxmode.offloads &
1233 		~(DEV_RX_OFFLOAD_VLAN_STRIP |
1234 		  DEV_RX_OFFLOAD_KEEP_CRC |
1235 		  DEV_RX_OFFLOAD_JUMBO_FRAME |
1236 		  DEV_RX_OFFLOAD_IPV4_CKSUM |
1237 		  DEV_RX_OFFLOAD_UDP_CKSUM |
1238 		  DEV_RX_OFFLOAD_TCP_CKSUM |
1239 		  DEV_RX_OFFLOAD_OUTER_IPV4_CKSUM |
1240 		  DEV_RX_OFFLOAD_OUTER_UDP_CKSUM |
1241 		  DEV_RX_OFFLOAD_RSS_HASH |
1242 		  DEV_RX_OFFLOAD_VLAN_FILTER))
1243 		goto use_scalar_rx;
1244 
1245 #if defined(RTE_ARCH_X86) && defined(CC_AVX2_SUPPORT)
1246 	if (rte_vect_get_max_simd_bitwidth() >= RTE_VECT_SIMD_256 &&
1247 	    rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX2) == 1) {
1248 		PMD_DRV_LOG(INFO,
1249 			    "Using AVX2 vector mode receive for port %d\n",
1250 			    eth_dev->data->port_id);
1251 		bp->flags |= BNXT_FLAG_RX_VECTOR_PKT_MODE;
1252 		return bnxt_recv_pkts_vec_avx2;
1253 	}
1254  #endif
1255 	if (rte_vect_get_max_simd_bitwidth() >= RTE_VECT_SIMD_128) {
1256 		PMD_DRV_LOG(INFO,
1257 			    "Using SSE vector mode receive for port %d\n",
1258 			    eth_dev->data->port_id);
1259 		bp->flags |= BNXT_FLAG_RX_VECTOR_PKT_MODE;
1260 		return bnxt_recv_pkts_vec;
1261 	}
1262 
1263 use_scalar_rx:
1264 	PMD_DRV_LOG(INFO, "Vector mode receive disabled for port %d\n",
1265 		    eth_dev->data->port_id);
1266 	PMD_DRV_LOG(INFO,
1267 		    "Port %d scatter: %d rx offload: %" PRIX64 "\n",
1268 		    eth_dev->data->port_id,
1269 		    eth_dev->data->scattered_rx,
1270 		    eth_dev->data->dev_conf.rxmode.offloads);
1271 #endif
1272 	bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
1273 	return bnxt_recv_pkts;
1274 }
1275 
1276 static eth_tx_burst_t
1277 bnxt_transmit_function(struct rte_eth_dev *eth_dev)
1278 {
1279 	struct bnxt *bp = eth_dev->data->dev_private;
1280 
1281 	/* Disable vector mode TX for Stingray2 for now */
1282 	if (BNXT_CHIP_SR2(bp))
1283 		return bnxt_xmit_pkts;
1284 
1285 #if defined(RTE_ARCH_X86) || defined(RTE_ARCH_ARM64) && \
1286 	!defined(RTE_LIBRTE_IEEE1588)
1287 	uint64_t offloads = eth_dev->data->dev_conf.txmode.offloads;
1288 
1289 	/*
1290 	 * Vector mode transmit can be enabled only if not using scatter rx
1291 	 * or tx offloads.
1292 	 */
1293 	if (eth_dev->data->scattered_rx ||
1294 	    (offloads & ~DEV_TX_OFFLOAD_MBUF_FAST_FREE) ||
1295 	    BNXT_TRUFLOW_EN(bp))
1296 		goto use_scalar_tx;
1297 
1298 #if defined(RTE_ARCH_X86) && defined(CC_AVX2_SUPPORT)
1299 	if (rte_vect_get_max_simd_bitwidth() >= RTE_VECT_SIMD_256 &&
1300 	    rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX2) == 1) {
1301 		PMD_DRV_LOG(INFO,
1302 			    "Using AVX2 vector mode transmit for port %d\n",
1303 			    eth_dev->data->port_id);
1304 		return bnxt_xmit_pkts_vec_avx2;
1305 	}
1306 #endif
1307 	if (rte_vect_get_max_simd_bitwidth() >= RTE_VECT_SIMD_128) {
1308 		PMD_DRV_LOG(INFO,
1309 			    "Using SSE vector mode transmit for port %d\n",
1310 			    eth_dev->data->port_id);
1311 		return bnxt_xmit_pkts_vec;
1312 	}
1313 
1314 use_scalar_tx:
1315 	PMD_DRV_LOG(INFO, "Vector mode transmit disabled for port %d\n",
1316 		    eth_dev->data->port_id);
1317 	PMD_DRV_LOG(INFO,
1318 		    "Port %d scatter: %d tx offload: %" PRIX64 "\n",
1319 		    eth_dev->data->port_id,
1320 		    eth_dev->data->scattered_rx,
1321 		    offloads);
1322 #endif
1323 	return bnxt_xmit_pkts;
1324 }
1325 
1326 static int bnxt_handle_if_change_status(struct bnxt *bp)
1327 {
1328 	int rc;
1329 
1330 	/* Since fw has undergone a reset and lost all contexts,
1331 	 * set fatal flag to not issue hwrm during cleanup
1332 	 */
1333 	bp->flags |= BNXT_FLAG_FATAL_ERROR;
1334 	bnxt_uninit_resources(bp, true);
1335 
1336 	/* clear fatal flag so that re-init happens */
1337 	bp->flags &= ~BNXT_FLAG_FATAL_ERROR;
1338 	rc = bnxt_init_resources(bp, true);
1339 
1340 	bp->flags &= ~BNXT_FLAG_IF_CHANGE_HOT_FW_RESET_DONE;
1341 
1342 	return rc;
1343 }
1344 
1345 static int bnxt_dev_set_link_up_op(struct rte_eth_dev *eth_dev)
1346 {
1347 	struct bnxt *bp = eth_dev->data->dev_private;
1348 	int rc = 0;
1349 
1350 	if (!BNXT_SINGLE_PF(bp))
1351 		return -ENOTSUP;
1352 
1353 	if (!bp->link_info->link_up)
1354 		rc = bnxt_set_hwrm_link_config(bp, true);
1355 	if (!rc)
1356 		eth_dev->data->dev_link.link_status = 1;
1357 
1358 	bnxt_print_link_info(eth_dev);
1359 	return rc;
1360 }
1361 
1362 static int bnxt_dev_set_link_down_op(struct rte_eth_dev *eth_dev)
1363 {
1364 	struct bnxt *bp = eth_dev->data->dev_private;
1365 
1366 	if (!BNXT_SINGLE_PF(bp))
1367 		return -ENOTSUP;
1368 
1369 	eth_dev->data->dev_link.link_status = 0;
1370 	bnxt_set_hwrm_link_config(bp, false);
1371 	bp->link_info->link_up = 0;
1372 
1373 	return 0;
1374 }
1375 
1376 static void bnxt_free_switch_domain(struct bnxt *bp)
1377 {
1378 	int rc = 0;
1379 
1380 	if (!(BNXT_PF(bp) || BNXT_VF_IS_TRUSTED(bp)))
1381 		return;
1382 
1383 	rc = rte_eth_switch_domain_free(bp->switch_domain_id);
1384 	if (rc)
1385 		PMD_DRV_LOG(ERR, "free switch domain:%d fail: %d\n",
1386 			    bp->switch_domain_id, rc);
1387 }
1388 
1389 static void bnxt_ptp_get_current_time(void *arg)
1390 {
1391 	struct bnxt *bp = arg;
1392 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
1393 	int rc;
1394 
1395 	rc = is_bnxt_in_error(bp);
1396 	if (rc)
1397 		return;
1398 
1399 	if (!ptp)
1400 		return;
1401 
1402 	bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_CURRENT_TIME,
1403 				&ptp->current_time);
1404 
1405 	rc = rte_eal_alarm_set(US_PER_S, bnxt_ptp_get_current_time, (void *)bp);
1406 	if (rc != 0) {
1407 		PMD_DRV_LOG(ERR, "Failed to re-schedule PTP alarm\n");
1408 		bp->flags2 &= ~BNXT_FLAGS2_PTP_ALARM_SCHEDULED;
1409 	}
1410 }
1411 
1412 static int bnxt_schedule_ptp_alarm(struct bnxt *bp)
1413 {
1414 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
1415 	int rc;
1416 
1417 	if (bp->flags2 & BNXT_FLAGS2_PTP_ALARM_SCHEDULED)
1418 		return 0;
1419 
1420 	bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_CURRENT_TIME,
1421 				&ptp->current_time);
1422 
1423 	rc = rte_eal_alarm_set(US_PER_S, bnxt_ptp_get_current_time, (void *)bp);
1424 	return rc;
1425 }
1426 
1427 static void bnxt_cancel_ptp_alarm(struct bnxt *bp)
1428 {
1429 	if (bp->flags2 & BNXT_FLAGS2_PTP_ALARM_SCHEDULED) {
1430 		rte_eal_alarm_cancel(bnxt_ptp_get_current_time, (void *)bp);
1431 		bp->flags2 &= ~BNXT_FLAGS2_PTP_ALARM_SCHEDULED;
1432 	}
1433 }
1434 
1435 static void bnxt_ptp_stop(struct bnxt *bp)
1436 {
1437 	bnxt_cancel_ptp_alarm(bp);
1438 	bp->flags2 &= ~BNXT_FLAGS2_PTP_TIMESYNC_ENABLED;
1439 }
1440 
1441 static int bnxt_ptp_start(struct bnxt *bp)
1442 {
1443 	int rc;
1444 
1445 	rc = bnxt_schedule_ptp_alarm(bp);
1446 	if (rc != 0) {
1447 		PMD_DRV_LOG(ERR, "Failed to schedule PTP alarm\n");
1448 	} else {
1449 		bp->flags2 |= BNXT_FLAGS2_PTP_TIMESYNC_ENABLED;
1450 		bp->flags2 |= BNXT_FLAGS2_PTP_ALARM_SCHEDULED;
1451 	}
1452 
1453 	return rc;
1454 }
1455 
1456 static int bnxt_dev_stop(struct rte_eth_dev *eth_dev)
1457 {
1458 	struct bnxt *bp = eth_dev->data->dev_private;
1459 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
1460 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
1461 	struct rte_eth_link link;
1462 	int ret;
1463 
1464 	eth_dev->data->dev_started = 0;
1465 	eth_dev->data->scattered_rx = 0;
1466 
1467 	/* Prevent crashes when queues are still in use */
1468 	eth_dev->rx_pkt_burst = &bnxt_dummy_recv_pkts;
1469 	eth_dev->tx_pkt_burst = &bnxt_dummy_xmit_pkts;
1470 
1471 	bnxt_disable_int(bp);
1472 
1473 	/* disable uio/vfio intr/eventfd mapping */
1474 	rte_intr_disable(intr_handle);
1475 
1476 	/* Stop the child representors for this device */
1477 	ret = bnxt_rep_stop_all(bp);
1478 	if (ret != 0)
1479 		return ret;
1480 
1481 	/* delete the bnxt ULP port details */
1482 	bnxt_ulp_port_deinit(bp);
1483 
1484 	bnxt_cancel_fw_health_check(bp);
1485 
1486 	if (BNXT_P5_PTP_TIMESYNC_ENABLED(bp))
1487 		bnxt_cancel_ptp_alarm(bp);
1488 
1489 	/* Do not bring link down during reset recovery */
1490 	if (!is_bnxt_in_error(bp)) {
1491 		bnxt_dev_set_link_down_op(eth_dev);
1492 		/* Wait for link to be reset */
1493 		if (BNXT_SINGLE_PF(bp))
1494 			rte_delay_ms(500);
1495 		/* clear the recorded link status */
1496 		memset(&link, 0, sizeof(link));
1497 		rte_eth_linkstatus_set(eth_dev, &link);
1498 	}
1499 
1500 	/* Clean queue intr-vector mapping */
1501 	rte_intr_efd_disable(intr_handle);
1502 	if (intr_handle->intr_vec != NULL) {
1503 		rte_free(intr_handle->intr_vec);
1504 		intr_handle->intr_vec = NULL;
1505 	}
1506 
1507 	bnxt_hwrm_port_clr_stats(bp);
1508 	bnxt_free_tx_mbufs(bp);
1509 	bnxt_free_rx_mbufs(bp);
1510 	/* Process any remaining notifications in default completion queue */
1511 	bnxt_int_handler(eth_dev);
1512 	bnxt_shutdown_nic(bp);
1513 	bnxt_hwrm_if_change(bp, false);
1514 
1515 	bnxt_free_prev_ring_stats(bp);
1516 	rte_free(bp->mark_table);
1517 	bp->mark_table = NULL;
1518 
1519 	bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
1520 	bp->rx_cosq_cnt = 0;
1521 	/* All filters are deleted on a port stop. */
1522 	if (BNXT_FLOW_XSTATS_EN(bp))
1523 		bp->flow_stat->flow_count = 0;
1524 
1525 	return 0;
1526 }
1527 
1528 /* Unload the driver, release resources */
1529 static int bnxt_dev_stop_op(struct rte_eth_dev *eth_dev)
1530 {
1531 	struct bnxt *bp = eth_dev->data->dev_private;
1532 
1533 	pthread_mutex_lock(&bp->err_recovery_lock);
1534 	if (bp->flags & BNXT_FLAG_FW_RESET) {
1535 		PMD_DRV_LOG(ERR,
1536 			    "Adapter recovering from error..Please retry\n");
1537 		pthread_mutex_unlock(&bp->err_recovery_lock);
1538 		return -EAGAIN;
1539 	}
1540 	pthread_mutex_unlock(&bp->err_recovery_lock);
1541 
1542 	return bnxt_dev_stop(eth_dev);
1543 }
1544 
1545 static int bnxt_dev_start_op(struct rte_eth_dev *eth_dev)
1546 {
1547 	struct bnxt *bp = eth_dev->data->dev_private;
1548 	uint64_t rx_offloads = eth_dev->data->dev_conf.rxmode.offloads;
1549 	int vlan_mask = 0;
1550 	int rc, retry_cnt = BNXT_IF_CHANGE_RETRY_COUNT;
1551 
1552 	if (!eth_dev->data->nb_tx_queues || !eth_dev->data->nb_rx_queues) {
1553 		PMD_DRV_LOG(ERR, "Queues are not configured yet!\n");
1554 		return -EINVAL;
1555 	}
1556 
1557 	if (bp->rx_cp_nr_rings > RTE_ETHDEV_QUEUE_STAT_CNTRS)
1558 		PMD_DRV_LOG(ERR,
1559 			    "RxQ cnt %d > RTE_ETHDEV_QUEUE_STAT_CNTRS %d\n",
1560 			    bp->rx_cp_nr_rings, RTE_ETHDEV_QUEUE_STAT_CNTRS);
1561 
1562 	do {
1563 		rc = bnxt_hwrm_if_change(bp, true);
1564 		if (rc == 0 || rc != -EAGAIN)
1565 			break;
1566 
1567 		rte_delay_ms(BNXT_IF_CHANGE_RETRY_INTERVAL);
1568 	} while (retry_cnt--);
1569 
1570 	if (rc)
1571 		return rc;
1572 
1573 	if (bp->flags & BNXT_FLAG_IF_CHANGE_HOT_FW_RESET_DONE) {
1574 		rc = bnxt_handle_if_change_status(bp);
1575 		if (rc)
1576 			return rc;
1577 	}
1578 
1579 	bnxt_enable_int(bp);
1580 
1581 	eth_dev->data->scattered_rx = bnxt_scattered_rx(eth_dev);
1582 
1583 	rc = bnxt_start_nic(bp);
1584 	if (rc)
1585 		goto error;
1586 
1587 	rc = bnxt_alloc_prev_ring_stats(bp);
1588 	if (rc)
1589 		goto error;
1590 
1591 	eth_dev->data->dev_started = 1;
1592 
1593 	bnxt_link_update_op(eth_dev, 1);
1594 
1595 	if (rx_offloads & DEV_RX_OFFLOAD_VLAN_FILTER)
1596 		vlan_mask |= ETH_VLAN_FILTER_MASK;
1597 	if (rx_offloads & DEV_RX_OFFLOAD_VLAN_STRIP)
1598 		vlan_mask |= ETH_VLAN_STRIP_MASK;
1599 	rc = bnxt_vlan_offload_set_op(eth_dev, vlan_mask);
1600 	if (rc)
1601 		goto error;
1602 
1603 	/* Initialize bnxt ULP port details */
1604 	rc = bnxt_ulp_port_init(bp);
1605 	if (rc)
1606 		goto error;
1607 
1608 	eth_dev->rx_pkt_burst = bnxt_receive_function(eth_dev);
1609 	eth_dev->tx_pkt_burst = bnxt_transmit_function(eth_dev);
1610 
1611 	bnxt_schedule_fw_health_check(bp);
1612 
1613 	if (BNXT_P5_PTP_TIMESYNC_ENABLED(bp))
1614 		bnxt_schedule_ptp_alarm(bp);
1615 
1616 	return 0;
1617 
1618 error:
1619 	bnxt_dev_stop(eth_dev);
1620 	return rc;
1621 }
1622 
1623 static void
1624 bnxt_uninit_locks(struct bnxt *bp)
1625 {
1626 	pthread_mutex_destroy(&bp->flow_lock);
1627 	pthread_mutex_destroy(&bp->def_cp_lock);
1628 	pthread_mutex_destroy(&bp->health_check_lock);
1629 	pthread_mutex_destroy(&bp->err_recovery_lock);
1630 	if (bp->rep_info) {
1631 		pthread_mutex_destroy(&bp->rep_info->vfr_lock);
1632 		pthread_mutex_destroy(&bp->rep_info->vfr_start_lock);
1633 	}
1634 }
1635 
1636 static void bnxt_drv_uninit(struct bnxt *bp)
1637 {
1638 	bnxt_free_leds_info(bp);
1639 	bnxt_free_cos_queues(bp);
1640 	bnxt_free_link_info(bp);
1641 	bnxt_free_parent_info(bp);
1642 	bnxt_uninit_locks(bp);
1643 
1644 	rte_memzone_free((const struct rte_memzone *)bp->tx_mem_zone);
1645 	bp->tx_mem_zone = NULL;
1646 	rte_memzone_free((const struct rte_memzone *)bp->rx_mem_zone);
1647 	bp->rx_mem_zone = NULL;
1648 
1649 	bnxt_free_vf_info(bp);
1650 	bnxt_free_pf_info(bp);
1651 
1652 	rte_free(bp->grp_info);
1653 	bp->grp_info = NULL;
1654 }
1655 
1656 static int bnxt_dev_close_op(struct rte_eth_dev *eth_dev)
1657 {
1658 	struct bnxt *bp = eth_dev->data->dev_private;
1659 	int ret = 0;
1660 
1661 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
1662 		return 0;
1663 
1664 	pthread_mutex_lock(&bp->err_recovery_lock);
1665 	if (bp->flags & BNXT_FLAG_FW_RESET) {
1666 		PMD_DRV_LOG(ERR,
1667 			    "Adapter recovering from error...Please retry\n");
1668 		pthread_mutex_unlock(&bp->err_recovery_lock);
1669 		return -EAGAIN;
1670 	}
1671 	pthread_mutex_unlock(&bp->err_recovery_lock);
1672 
1673 	/* cancel the recovery handler before remove dev */
1674 	rte_eal_alarm_cancel(bnxt_dev_reset_and_resume, (void *)bp);
1675 	rte_eal_alarm_cancel(bnxt_dev_recover, (void *)bp);
1676 	bnxt_cancel_fc_thread(bp);
1677 
1678 	if (eth_dev->data->dev_started)
1679 		ret = bnxt_dev_stop(eth_dev);
1680 
1681 	bnxt_uninit_resources(bp, false);
1682 
1683 	bnxt_drv_uninit(bp);
1684 
1685 	return ret;
1686 }
1687 
1688 static void bnxt_mac_addr_remove_op(struct rte_eth_dev *eth_dev,
1689 				    uint32_t index)
1690 {
1691 	struct bnxt *bp = eth_dev->data->dev_private;
1692 	uint64_t pool_mask = eth_dev->data->mac_pool_sel[index];
1693 	struct bnxt_vnic_info *vnic;
1694 	struct bnxt_filter_info *filter, *temp_filter;
1695 	uint32_t i;
1696 
1697 	if (is_bnxt_in_error(bp))
1698 		return;
1699 
1700 	/*
1701 	 * Loop through all VNICs from the specified filter flow pools to
1702 	 * remove the corresponding MAC addr filter
1703 	 */
1704 	for (i = 0; i < bp->nr_vnics; i++) {
1705 		if (!(pool_mask & (1ULL << i)))
1706 			continue;
1707 
1708 		vnic = &bp->vnic_info[i];
1709 		filter = STAILQ_FIRST(&vnic->filter);
1710 		while (filter) {
1711 			temp_filter = STAILQ_NEXT(filter, next);
1712 			if (filter->mac_index == index) {
1713 				STAILQ_REMOVE(&vnic->filter, filter,
1714 						bnxt_filter_info, next);
1715 				bnxt_hwrm_clear_l2_filter(bp, filter);
1716 				bnxt_free_filter(bp, filter);
1717 			}
1718 			filter = temp_filter;
1719 		}
1720 	}
1721 }
1722 
1723 static int bnxt_add_mac_filter(struct bnxt *bp, struct bnxt_vnic_info *vnic,
1724 			       struct rte_ether_addr *mac_addr, uint32_t index,
1725 			       uint32_t pool)
1726 {
1727 	struct bnxt_filter_info *filter;
1728 	int rc = 0;
1729 
1730 	/* Attach requested MAC address to the new l2_filter */
1731 	STAILQ_FOREACH(filter, &vnic->filter, next) {
1732 		if (filter->mac_index == index) {
1733 			PMD_DRV_LOG(DEBUG,
1734 				    "MAC addr already existed for pool %d\n",
1735 				    pool);
1736 			return 0;
1737 		}
1738 	}
1739 
1740 	filter = bnxt_alloc_filter(bp);
1741 	if (!filter) {
1742 		PMD_DRV_LOG(ERR, "L2 filter alloc failed\n");
1743 		return -ENODEV;
1744 	}
1745 
1746 	/* bnxt_alloc_filter copies default MAC to filter->l2_addr. So,
1747 	 * if the MAC that's been programmed now is a different one, then,
1748 	 * copy that addr to filter->l2_addr
1749 	 */
1750 	if (mac_addr)
1751 		memcpy(filter->l2_addr, mac_addr, RTE_ETHER_ADDR_LEN);
1752 	filter->flags |= HWRM_CFA_L2_FILTER_ALLOC_INPUT_FLAGS_OUTERMOST;
1753 
1754 	rc = bnxt_hwrm_set_l2_filter(bp, vnic->fw_vnic_id, filter);
1755 	if (!rc) {
1756 		filter->mac_index = index;
1757 		if (filter->mac_index == 0)
1758 			STAILQ_INSERT_HEAD(&vnic->filter, filter, next);
1759 		else
1760 			STAILQ_INSERT_TAIL(&vnic->filter, filter, next);
1761 	} else {
1762 		bnxt_free_filter(bp, filter);
1763 	}
1764 
1765 	return rc;
1766 }
1767 
1768 static int bnxt_mac_addr_add_op(struct rte_eth_dev *eth_dev,
1769 				struct rte_ether_addr *mac_addr,
1770 				uint32_t index, uint32_t pool)
1771 {
1772 	struct bnxt *bp = eth_dev->data->dev_private;
1773 	struct bnxt_vnic_info *vnic = &bp->vnic_info[pool];
1774 	int rc = 0;
1775 
1776 	rc = is_bnxt_in_error(bp);
1777 	if (rc)
1778 		return rc;
1779 
1780 	if (BNXT_VF(bp) && !BNXT_VF_IS_TRUSTED(bp)) {
1781 		PMD_DRV_LOG(ERR, "Cannot add MAC address to a VF interface\n");
1782 		return -ENOTSUP;
1783 	}
1784 
1785 	if (!vnic) {
1786 		PMD_DRV_LOG(ERR, "VNIC not found for pool %d!\n", pool);
1787 		return -EINVAL;
1788 	}
1789 
1790 	/* Filter settings will get applied when port is started */
1791 	if (!eth_dev->data->dev_started)
1792 		return 0;
1793 
1794 	rc = bnxt_add_mac_filter(bp, vnic, mac_addr, index, pool);
1795 
1796 	return rc;
1797 }
1798 
1799 int bnxt_link_update_op(struct rte_eth_dev *eth_dev, int wait_to_complete)
1800 {
1801 	int rc = 0;
1802 	struct bnxt *bp = eth_dev->data->dev_private;
1803 	struct rte_eth_link new;
1804 	int cnt = wait_to_complete ? BNXT_MAX_LINK_WAIT_CNT :
1805 			BNXT_MIN_LINK_WAIT_CNT;
1806 
1807 	rc = is_bnxt_in_error(bp);
1808 	if (rc)
1809 		return rc;
1810 
1811 	memset(&new, 0, sizeof(new));
1812 
1813 	if (bp->link_info == NULL)
1814 		goto out;
1815 
1816 	do {
1817 		/* Retrieve link info from hardware */
1818 		rc = bnxt_get_hwrm_link_config(bp, &new);
1819 		if (rc) {
1820 			new.link_speed = ETH_LINK_SPEED_100M;
1821 			new.link_duplex = ETH_LINK_FULL_DUPLEX;
1822 			PMD_DRV_LOG(ERR,
1823 				"Failed to retrieve link rc = 0x%x!\n", rc);
1824 			goto out;
1825 		}
1826 
1827 		if (!wait_to_complete || new.link_status)
1828 			break;
1829 
1830 		rte_delay_ms(BNXT_LINK_WAIT_INTERVAL);
1831 	} while (cnt--);
1832 
1833 	/* Only single function PF can bring phy down.
1834 	 * When port is stopped, report link down for VF/MH/NPAR functions.
1835 	 */
1836 	if (!BNXT_SINGLE_PF(bp) && !eth_dev->data->dev_started)
1837 		memset(&new, 0, sizeof(new));
1838 
1839 out:
1840 	/* Timed out or success */
1841 	if (new.link_status != eth_dev->data->dev_link.link_status ||
1842 	    new.link_speed != eth_dev->data->dev_link.link_speed) {
1843 		rte_eth_linkstatus_set(eth_dev, &new);
1844 
1845 		rte_eth_dev_callback_process(eth_dev,
1846 					     RTE_ETH_EVENT_INTR_LSC,
1847 					     NULL);
1848 
1849 		bnxt_print_link_info(eth_dev);
1850 	}
1851 
1852 	return rc;
1853 }
1854 
1855 static int bnxt_promiscuous_enable_op(struct rte_eth_dev *eth_dev)
1856 {
1857 	struct bnxt *bp = eth_dev->data->dev_private;
1858 	struct bnxt_vnic_info *vnic;
1859 	uint32_t old_flags;
1860 	int rc;
1861 
1862 	rc = is_bnxt_in_error(bp);
1863 	if (rc)
1864 		return rc;
1865 
1866 	/* Filter settings will get applied when port is started */
1867 	if (!eth_dev->data->dev_started)
1868 		return 0;
1869 
1870 	if (bp->vnic_info == NULL)
1871 		return 0;
1872 
1873 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
1874 
1875 	old_flags = vnic->flags;
1876 	vnic->flags |= BNXT_VNIC_INFO_PROMISC;
1877 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1878 	if (rc != 0)
1879 		vnic->flags = old_flags;
1880 
1881 	return rc;
1882 }
1883 
1884 static int bnxt_promiscuous_disable_op(struct rte_eth_dev *eth_dev)
1885 {
1886 	struct bnxt *bp = eth_dev->data->dev_private;
1887 	struct bnxt_vnic_info *vnic;
1888 	uint32_t old_flags;
1889 	int rc;
1890 
1891 	rc = is_bnxt_in_error(bp);
1892 	if (rc)
1893 		return rc;
1894 
1895 	/* Filter settings will get applied when port is started */
1896 	if (!eth_dev->data->dev_started)
1897 		return 0;
1898 
1899 	if (bp->vnic_info == NULL)
1900 		return 0;
1901 
1902 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
1903 
1904 	old_flags = vnic->flags;
1905 	vnic->flags &= ~BNXT_VNIC_INFO_PROMISC;
1906 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1907 	if (rc != 0)
1908 		vnic->flags = old_flags;
1909 
1910 	return rc;
1911 }
1912 
1913 static int bnxt_allmulticast_enable_op(struct rte_eth_dev *eth_dev)
1914 {
1915 	struct bnxt *bp = eth_dev->data->dev_private;
1916 	struct bnxt_vnic_info *vnic;
1917 	uint32_t old_flags;
1918 	int rc;
1919 
1920 	rc = is_bnxt_in_error(bp);
1921 	if (rc)
1922 		return rc;
1923 
1924 	/* Filter settings will get applied when port is started */
1925 	if (!eth_dev->data->dev_started)
1926 		return 0;
1927 
1928 	if (bp->vnic_info == NULL)
1929 		return 0;
1930 
1931 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
1932 
1933 	old_flags = vnic->flags;
1934 	vnic->flags |= BNXT_VNIC_INFO_ALLMULTI;
1935 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1936 	if (rc != 0)
1937 		vnic->flags = old_flags;
1938 
1939 	return rc;
1940 }
1941 
1942 static int bnxt_allmulticast_disable_op(struct rte_eth_dev *eth_dev)
1943 {
1944 	struct bnxt *bp = eth_dev->data->dev_private;
1945 	struct bnxt_vnic_info *vnic;
1946 	uint32_t old_flags;
1947 	int rc;
1948 
1949 	rc = is_bnxt_in_error(bp);
1950 	if (rc)
1951 		return rc;
1952 
1953 	/* Filter settings will get applied when port is started */
1954 	if (!eth_dev->data->dev_started)
1955 		return 0;
1956 
1957 	if (bp->vnic_info == NULL)
1958 		return 0;
1959 
1960 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
1961 
1962 	old_flags = vnic->flags;
1963 	vnic->flags &= ~BNXT_VNIC_INFO_ALLMULTI;
1964 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1965 	if (rc != 0)
1966 		vnic->flags = old_flags;
1967 
1968 	return rc;
1969 }
1970 
1971 /* Return bnxt_rx_queue pointer corresponding to a given rxq. */
1972 static struct bnxt_rx_queue *bnxt_qid_to_rxq(struct bnxt *bp, uint16_t qid)
1973 {
1974 	if (qid >= bp->rx_nr_rings)
1975 		return NULL;
1976 
1977 	return bp->eth_dev->data->rx_queues[qid];
1978 }
1979 
1980 /* Return rxq corresponding to a given rss table ring/group ID. */
1981 static uint16_t bnxt_rss_to_qid(struct bnxt *bp, uint16_t fwr)
1982 {
1983 	struct bnxt_rx_queue *rxq;
1984 	unsigned int i;
1985 
1986 	if (!BNXT_HAS_RING_GRPS(bp)) {
1987 		for (i = 0; i < bp->rx_nr_rings; i++) {
1988 			rxq = bp->eth_dev->data->rx_queues[i];
1989 			if (rxq->rx_ring->rx_ring_struct->fw_ring_id == fwr)
1990 				return rxq->index;
1991 		}
1992 	} else {
1993 		for (i = 0; i < bp->rx_nr_rings; i++) {
1994 			if (bp->grp_info[i].fw_grp_id == fwr)
1995 				return i;
1996 		}
1997 	}
1998 
1999 	return INVALID_HW_RING_ID;
2000 }
2001 
2002 static int bnxt_reta_update_op(struct rte_eth_dev *eth_dev,
2003 			    struct rte_eth_rss_reta_entry64 *reta_conf,
2004 			    uint16_t reta_size)
2005 {
2006 	struct bnxt *bp = eth_dev->data->dev_private;
2007 	struct rte_eth_conf *dev_conf = &bp->eth_dev->data->dev_conf;
2008 	struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
2009 	uint16_t tbl_size = bnxt_rss_hash_tbl_size(bp);
2010 	uint16_t idx, sft;
2011 	int i, rc;
2012 
2013 	rc = is_bnxt_in_error(bp);
2014 	if (rc)
2015 		return rc;
2016 
2017 	if (!vnic->rss_table)
2018 		return -EINVAL;
2019 
2020 	if (!(dev_conf->rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG))
2021 		return -EINVAL;
2022 
2023 	if (reta_size != tbl_size) {
2024 		PMD_DRV_LOG(ERR, "The configured hash table lookup size "
2025 			"(%d) must equal the size supported by the hardware "
2026 			"(%d)\n", reta_size, tbl_size);
2027 		return -EINVAL;
2028 	}
2029 
2030 	for (i = 0; i < reta_size; i++) {
2031 		struct bnxt_rx_queue *rxq;
2032 
2033 		idx = i / RTE_RETA_GROUP_SIZE;
2034 		sft = i % RTE_RETA_GROUP_SIZE;
2035 
2036 		if (!(reta_conf[idx].mask & (1ULL << sft)))
2037 			continue;
2038 
2039 		rxq = bnxt_qid_to_rxq(bp, reta_conf[idx].reta[sft]);
2040 		if (!rxq) {
2041 			PMD_DRV_LOG(ERR, "Invalid ring in reta_conf.\n");
2042 			return -EINVAL;
2043 		}
2044 
2045 		if (BNXT_CHIP_P5(bp)) {
2046 			vnic->rss_table[i * 2] =
2047 				rxq->rx_ring->rx_ring_struct->fw_ring_id;
2048 			vnic->rss_table[i * 2 + 1] =
2049 				rxq->cp_ring->cp_ring_struct->fw_ring_id;
2050 		} else {
2051 			vnic->rss_table[i] =
2052 			    vnic->fw_grp_ids[reta_conf[idx].reta[sft]];
2053 		}
2054 	}
2055 
2056 	rc = bnxt_hwrm_vnic_rss_cfg(bp, vnic);
2057 	return rc;
2058 }
2059 
2060 static int bnxt_reta_query_op(struct rte_eth_dev *eth_dev,
2061 			      struct rte_eth_rss_reta_entry64 *reta_conf,
2062 			      uint16_t reta_size)
2063 {
2064 	struct bnxt *bp = eth_dev->data->dev_private;
2065 	struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
2066 	uint16_t tbl_size = bnxt_rss_hash_tbl_size(bp);
2067 	uint16_t idx, sft, i;
2068 	int rc;
2069 
2070 	rc = is_bnxt_in_error(bp);
2071 	if (rc)
2072 		return rc;
2073 
2074 	if (!vnic)
2075 		return -EINVAL;
2076 	if (!vnic->rss_table)
2077 		return -EINVAL;
2078 
2079 	if (reta_size != tbl_size) {
2080 		PMD_DRV_LOG(ERR, "The configured hash table lookup size "
2081 			"(%d) must equal the size supported by the hardware "
2082 			"(%d)\n", reta_size, tbl_size);
2083 		return -EINVAL;
2084 	}
2085 
2086 	for (idx = 0, i = 0; i < reta_size; i++) {
2087 		idx = i / RTE_RETA_GROUP_SIZE;
2088 		sft = i % RTE_RETA_GROUP_SIZE;
2089 
2090 		if (reta_conf[idx].mask & (1ULL << sft)) {
2091 			uint16_t qid;
2092 
2093 			if (BNXT_CHIP_P5(bp))
2094 				qid = bnxt_rss_to_qid(bp,
2095 						      vnic->rss_table[i * 2]);
2096 			else
2097 				qid = bnxt_rss_to_qid(bp, vnic->rss_table[i]);
2098 
2099 			if (qid == INVALID_HW_RING_ID) {
2100 				PMD_DRV_LOG(ERR, "Inv. entry in rss table.\n");
2101 				return -EINVAL;
2102 			}
2103 			reta_conf[idx].reta[sft] = qid;
2104 		}
2105 	}
2106 
2107 	return 0;
2108 }
2109 
2110 static int bnxt_rss_hash_update_op(struct rte_eth_dev *eth_dev,
2111 				   struct rte_eth_rss_conf *rss_conf)
2112 {
2113 	struct bnxt *bp = eth_dev->data->dev_private;
2114 	struct rte_eth_conf *dev_conf = &bp->eth_dev->data->dev_conf;
2115 	struct bnxt_vnic_info *vnic;
2116 	int rc;
2117 
2118 	rc = is_bnxt_in_error(bp);
2119 	if (rc)
2120 		return rc;
2121 
2122 	/*
2123 	 * If RSS enablement were different than dev_configure,
2124 	 * then return -EINVAL
2125 	 */
2126 	if (dev_conf->rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG) {
2127 		if (!rss_conf->rss_hf)
2128 			PMD_DRV_LOG(ERR, "Hash type NONE\n");
2129 	} else {
2130 		if (rss_conf->rss_hf & BNXT_ETH_RSS_SUPPORT)
2131 			return -EINVAL;
2132 	}
2133 
2134 	bp->flags |= BNXT_FLAG_UPDATE_HASH;
2135 	memcpy(&eth_dev->data->dev_conf.rx_adv_conf.rss_conf,
2136 	       rss_conf,
2137 	       sizeof(*rss_conf));
2138 
2139 	/* Update the default RSS VNIC(s) */
2140 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
2141 	vnic->hash_type = bnxt_rte_to_hwrm_hash_types(rss_conf->rss_hf);
2142 	vnic->hash_mode =
2143 		bnxt_rte_to_hwrm_hash_level(bp, rss_conf->rss_hf,
2144 					    ETH_RSS_LEVEL(rss_conf->rss_hf));
2145 
2146 	/*
2147 	 * If hashkey is not specified, use the previously configured
2148 	 * hashkey
2149 	 */
2150 	if (!rss_conf->rss_key)
2151 		goto rss_config;
2152 
2153 	if (rss_conf->rss_key_len != HW_HASH_KEY_SIZE) {
2154 		PMD_DRV_LOG(ERR,
2155 			    "Invalid hashkey length, should be 16 bytes\n");
2156 		return -EINVAL;
2157 	}
2158 	memcpy(vnic->rss_hash_key, rss_conf->rss_key, rss_conf->rss_key_len);
2159 
2160 rss_config:
2161 	rc = bnxt_hwrm_vnic_rss_cfg(bp, vnic);
2162 	return rc;
2163 }
2164 
2165 static int bnxt_rss_hash_conf_get_op(struct rte_eth_dev *eth_dev,
2166 				     struct rte_eth_rss_conf *rss_conf)
2167 {
2168 	struct bnxt *bp = eth_dev->data->dev_private;
2169 	struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
2170 	int len, rc;
2171 	uint32_t hash_types;
2172 
2173 	rc = is_bnxt_in_error(bp);
2174 	if (rc)
2175 		return rc;
2176 
2177 	/* RSS configuration is the same for all VNICs */
2178 	if (vnic && vnic->rss_hash_key) {
2179 		if (rss_conf->rss_key) {
2180 			len = rss_conf->rss_key_len <= HW_HASH_KEY_SIZE ?
2181 			      rss_conf->rss_key_len : HW_HASH_KEY_SIZE;
2182 			memcpy(rss_conf->rss_key, vnic->rss_hash_key, len);
2183 		}
2184 
2185 		hash_types = vnic->hash_type;
2186 		rss_conf->rss_hf = 0;
2187 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV4) {
2188 			rss_conf->rss_hf |= ETH_RSS_IPV4;
2189 			hash_types &= ~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV4;
2190 		}
2191 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV4) {
2192 			rss_conf->rss_hf |= ETH_RSS_NONFRAG_IPV4_TCP;
2193 			hash_types &=
2194 				~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV4;
2195 		}
2196 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV4) {
2197 			rss_conf->rss_hf |= ETH_RSS_NONFRAG_IPV4_UDP;
2198 			hash_types &=
2199 				~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV4;
2200 		}
2201 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV6) {
2202 			rss_conf->rss_hf |= ETH_RSS_IPV6;
2203 			hash_types &= ~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV6;
2204 		}
2205 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV6) {
2206 			rss_conf->rss_hf |= ETH_RSS_NONFRAG_IPV6_TCP;
2207 			hash_types &=
2208 				~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV6;
2209 		}
2210 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV6) {
2211 			rss_conf->rss_hf |= ETH_RSS_NONFRAG_IPV6_UDP;
2212 			hash_types &=
2213 				~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV6;
2214 		}
2215 
2216 		rss_conf->rss_hf |=
2217 			bnxt_hwrm_to_rte_rss_level(bp, vnic->hash_mode);
2218 
2219 		if (hash_types) {
2220 			PMD_DRV_LOG(ERR,
2221 				"Unknown RSS config from firmware (%08x), RSS disabled",
2222 				vnic->hash_type);
2223 			return -ENOTSUP;
2224 		}
2225 	} else {
2226 		rss_conf->rss_hf = 0;
2227 	}
2228 	return 0;
2229 }
2230 
2231 static int bnxt_flow_ctrl_get_op(struct rte_eth_dev *dev,
2232 			       struct rte_eth_fc_conf *fc_conf)
2233 {
2234 	struct bnxt *bp = dev->data->dev_private;
2235 	struct rte_eth_link link_info;
2236 	int rc;
2237 
2238 	rc = is_bnxt_in_error(bp);
2239 	if (rc)
2240 		return rc;
2241 
2242 	rc = bnxt_get_hwrm_link_config(bp, &link_info);
2243 	if (rc)
2244 		return rc;
2245 
2246 	memset(fc_conf, 0, sizeof(*fc_conf));
2247 	if (bp->link_info->auto_pause)
2248 		fc_conf->autoneg = 1;
2249 	switch (bp->link_info->pause) {
2250 	case 0:
2251 		fc_conf->mode = RTE_FC_NONE;
2252 		break;
2253 	case HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_TX:
2254 		fc_conf->mode = RTE_FC_TX_PAUSE;
2255 		break;
2256 	case HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_RX:
2257 		fc_conf->mode = RTE_FC_RX_PAUSE;
2258 		break;
2259 	case (HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_TX |
2260 			HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_RX):
2261 		fc_conf->mode = RTE_FC_FULL;
2262 		break;
2263 	}
2264 	return 0;
2265 }
2266 
2267 static int bnxt_flow_ctrl_set_op(struct rte_eth_dev *dev,
2268 			       struct rte_eth_fc_conf *fc_conf)
2269 {
2270 	struct bnxt *bp = dev->data->dev_private;
2271 	int rc;
2272 
2273 	rc = is_bnxt_in_error(bp);
2274 	if (rc)
2275 		return rc;
2276 
2277 	if (!BNXT_SINGLE_PF(bp)) {
2278 		PMD_DRV_LOG(ERR,
2279 			    "Flow Control Settings cannot be modified on VF or on shared PF\n");
2280 		return -ENOTSUP;
2281 	}
2282 
2283 	switch (fc_conf->mode) {
2284 	case RTE_FC_NONE:
2285 		bp->link_info->auto_pause = 0;
2286 		bp->link_info->force_pause = 0;
2287 		break;
2288 	case RTE_FC_RX_PAUSE:
2289 		if (fc_conf->autoneg) {
2290 			bp->link_info->auto_pause =
2291 					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_RX;
2292 			bp->link_info->force_pause = 0;
2293 		} else {
2294 			bp->link_info->auto_pause = 0;
2295 			bp->link_info->force_pause =
2296 					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_RX;
2297 		}
2298 		break;
2299 	case RTE_FC_TX_PAUSE:
2300 		if (fc_conf->autoneg) {
2301 			bp->link_info->auto_pause =
2302 					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_TX;
2303 			bp->link_info->force_pause = 0;
2304 		} else {
2305 			bp->link_info->auto_pause = 0;
2306 			bp->link_info->force_pause =
2307 					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_TX;
2308 		}
2309 		break;
2310 	case RTE_FC_FULL:
2311 		if (fc_conf->autoneg) {
2312 			bp->link_info->auto_pause =
2313 					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_TX |
2314 					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_RX;
2315 			bp->link_info->force_pause = 0;
2316 		} else {
2317 			bp->link_info->auto_pause = 0;
2318 			bp->link_info->force_pause =
2319 					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_TX |
2320 					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_RX;
2321 		}
2322 		break;
2323 	}
2324 	return bnxt_set_hwrm_link_config(bp, true);
2325 }
2326 
2327 /* Add UDP tunneling port */
2328 static int
2329 bnxt_udp_tunnel_port_add_op(struct rte_eth_dev *eth_dev,
2330 			 struct rte_eth_udp_tunnel *udp_tunnel)
2331 {
2332 	struct bnxt *bp = eth_dev->data->dev_private;
2333 	uint16_t tunnel_type = 0;
2334 	int rc = 0;
2335 
2336 	rc = is_bnxt_in_error(bp);
2337 	if (rc)
2338 		return rc;
2339 
2340 	switch (udp_tunnel->prot_type) {
2341 	case RTE_TUNNEL_TYPE_VXLAN:
2342 		if (bp->vxlan_port_cnt) {
2343 			PMD_DRV_LOG(ERR, "Tunnel Port %d already programmed\n",
2344 				udp_tunnel->udp_port);
2345 			if (bp->vxlan_port != udp_tunnel->udp_port) {
2346 				PMD_DRV_LOG(ERR, "Only one port allowed\n");
2347 				return -ENOSPC;
2348 			}
2349 			bp->vxlan_port_cnt++;
2350 			return 0;
2351 		}
2352 		tunnel_type =
2353 			HWRM_TUNNEL_DST_PORT_ALLOC_INPUT_TUNNEL_TYPE_VXLAN;
2354 		bp->vxlan_port_cnt++;
2355 		break;
2356 	case RTE_TUNNEL_TYPE_GENEVE:
2357 		if (bp->geneve_port_cnt) {
2358 			PMD_DRV_LOG(ERR, "Tunnel Port %d already programmed\n",
2359 				udp_tunnel->udp_port);
2360 			if (bp->geneve_port != udp_tunnel->udp_port) {
2361 				PMD_DRV_LOG(ERR, "Only one port allowed\n");
2362 				return -ENOSPC;
2363 			}
2364 			bp->geneve_port_cnt++;
2365 			return 0;
2366 		}
2367 		tunnel_type =
2368 			HWRM_TUNNEL_DST_PORT_ALLOC_INPUT_TUNNEL_TYPE_GENEVE;
2369 		bp->geneve_port_cnt++;
2370 		break;
2371 	default:
2372 		PMD_DRV_LOG(ERR, "Tunnel type is not supported\n");
2373 		return -ENOTSUP;
2374 	}
2375 	rc = bnxt_hwrm_tunnel_dst_port_alloc(bp, udp_tunnel->udp_port,
2376 					     tunnel_type);
2377 	return rc;
2378 }
2379 
2380 static int
2381 bnxt_udp_tunnel_port_del_op(struct rte_eth_dev *eth_dev,
2382 			 struct rte_eth_udp_tunnel *udp_tunnel)
2383 {
2384 	struct bnxt *bp = eth_dev->data->dev_private;
2385 	uint16_t tunnel_type = 0;
2386 	uint16_t port = 0;
2387 	int rc = 0;
2388 
2389 	rc = is_bnxt_in_error(bp);
2390 	if (rc)
2391 		return rc;
2392 
2393 	switch (udp_tunnel->prot_type) {
2394 	case RTE_TUNNEL_TYPE_VXLAN:
2395 		if (!bp->vxlan_port_cnt) {
2396 			PMD_DRV_LOG(ERR, "No Tunnel port configured yet\n");
2397 			return -EINVAL;
2398 		}
2399 		if (bp->vxlan_port != udp_tunnel->udp_port) {
2400 			PMD_DRV_LOG(ERR, "Req Port: %d. Configured port: %d\n",
2401 				udp_tunnel->udp_port, bp->vxlan_port);
2402 			return -EINVAL;
2403 		}
2404 		if (--bp->vxlan_port_cnt)
2405 			return 0;
2406 
2407 		tunnel_type =
2408 			HWRM_TUNNEL_DST_PORT_FREE_INPUT_TUNNEL_TYPE_VXLAN;
2409 		port = bp->vxlan_fw_dst_port_id;
2410 		break;
2411 	case RTE_TUNNEL_TYPE_GENEVE:
2412 		if (!bp->geneve_port_cnt) {
2413 			PMD_DRV_LOG(ERR, "No Tunnel port configured yet\n");
2414 			return -EINVAL;
2415 		}
2416 		if (bp->geneve_port != udp_tunnel->udp_port) {
2417 			PMD_DRV_LOG(ERR, "Req Port: %d. Configured port: %d\n",
2418 				udp_tunnel->udp_port, bp->geneve_port);
2419 			return -EINVAL;
2420 		}
2421 		if (--bp->geneve_port_cnt)
2422 			return 0;
2423 
2424 		tunnel_type =
2425 			HWRM_TUNNEL_DST_PORT_FREE_INPUT_TUNNEL_TYPE_GENEVE;
2426 		port = bp->geneve_fw_dst_port_id;
2427 		break;
2428 	default:
2429 		PMD_DRV_LOG(ERR, "Tunnel type is not supported\n");
2430 		return -ENOTSUP;
2431 	}
2432 
2433 	rc = bnxt_hwrm_tunnel_dst_port_free(bp, port, tunnel_type);
2434 	return rc;
2435 }
2436 
2437 static int bnxt_del_vlan_filter(struct bnxt *bp, uint16_t vlan_id)
2438 {
2439 	struct bnxt_filter_info *filter;
2440 	struct bnxt_vnic_info *vnic;
2441 	int rc = 0;
2442 	uint32_t chk = HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN;
2443 
2444 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
2445 	filter = STAILQ_FIRST(&vnic->filter);
2446 	while (filter) {
2447 		/* Search for this matching MAC+VLAN filter */
2448 		if (bnxt_vlan_filter_exists(bp, filter, chk, vlan_id)) {
2449 			/* Delete the filter */
2450 			rc = bnxt_hwrm_clear_l2_filter(bp, filter);
2451 			if (rc)
2452 				return rc;
2453 			STAILQ_REMOVE(&vnic->filter, filter,
2454 				      bnxt_filter_info, next);
2455 			bnxt_free_filter(bp, filter);
2456 			PMD_DRV_LOG(INFO,
2457 				    "Deleted vlan filter for %d\n",
2458 				    vlan_id);
2459 			return 0;
2460 		}
2461 		filter = STAILQ_NEXT(filter, next);
2462 	}
2463 	return -ENOENT;
2464 }
2465 
2466 static int bnxt_add_vlan_filter(struct bnxt *bp, uint16_t vlan_id)
2467 {
2468 	struct bnxt_filter_info *filter;
2469 	struct bnxt_vnic_info *vnic;
2470 	int rc = 0;
2471 	uint32_t en = HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN |
2472 		HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN_MASK;
2473 	uint32_t chk = HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN;
2474 
2475 	/* Implementation notes on the use of VNIC in this command:
2476 	 *
2477 	 * By default, these filters belong to default vnic for the function.
2478 	 * Once these filters are set up, only destination VNIC can be modified.
2479 	 * If the destination VNIC is not specified in this command,
2480 	 * then the HWRM shall only create an l2 context id.
2481 	 */
2482 
2483 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
2484 	filter = STAILQ_FIRST(&vnic->filter);
2485 	/* Check if the VLAN has already been added */
2486 	while (filter) {
2487 		if (bnxt_vlan_filter_exists(bp, filter, chk, vlan_id))
2488 			return -EEXIST;
2489 
2490 		filter = STAILQ_NEXT(filter, next);
2491 	}
2492 
2493 	/* No match found. Alloc a fresh filter and issue the L2_FILTER_ALLOC
2494 	 * command to create MAC+VLAN filter with the right flags, enables set.
2495 	 */
2496 	filter = bnxt_alloc_filter(bp);
2497 	if (!filter) {
2498 		PMD_DRV_LOG(ERR,
2499 			    "MAC/VLAN filter alloc failed\n");
2500 		return -ENOMEM;
2501 	}
2502 	/* MAC + VLAN ID filter */
2503 	/* If l2_ivlan == 0 and l2_ivlan_mask != 0, only
2504 	 * untagged packets are received
2505 	 *
2506 	 * If l2_ivlan != 0 and l2_ivlan_mask != 0, untagged
2507 	 * packets and only the programmed vlan's packets are received
2508 	 */
2509 	filter->l2_ivlan = vlan_id;
2510 	filter->l2_ivlan_mask = 0x0FFF;
2511 	filter->enables |= en;
2512 	filter->flags |= HWRM_CFA_L2_FILTER_ALLOC_INPUT_FLAGS_OUTERMOST;
2513 
2514 	rc = bnxt_hwrm_set_l2_filter(bp, vnic->fw_vnic_id, filter);
2515 	if (rc) {
2516 		/* Free the newly allocated filter as we were
2517 		 * not able to create the filter in hardware.
2518 		 */
2519 		bnxt_free_filter(bp, filter);
2520 		return rc;
2521 	}
2522 
2523 	filter->mac_index = 0;
2524 	/* Add this new filter to the list */
2525 	if (vlan_id == 0)
2526 		STAILQ_INSERT_HEAD(&vnic->filter, filter, next);
2527 	else
2528 		STAILQ_INSERT_TAIL(&vnic->filter, filter, next);
2529 
2530 	PMD_DRV_LOG(INFO,
2531 		    "Added Vlan filter for %d\n", vlan_id);
2532 	return rc;
2533 }
2534 
2535 static int bnxt_vlan_filter_set_op(struct rte_eth_dev *eth_dev,
2536 		uint16_t vlan_id, int on)
2537 {
2538 	struct bnxt *bp = eth_dev->data->dev_private;
2539 	int rc;
2540 
2541 	rc = is_bnxt_in_error(bp);
2542 	if (rc)
2543 		return rc;
2544 
2545 	if (!eth_dev->data->dev_started) {
2546 		PMD_DRV_LOG(ERR, "port must be started before setting vlan\n");
2547 		return -EINVAL;
2548 	}
2549 
2550 	/* These operations apply to ALL existing MAC/VLAN filters */
2551 	if (on)
2552 		return bnxt_add_vlan_filter(bp, vlan_id);
2553 	else
2554 		return bnxt_del_vlan_filter(bp, vlan_id);
2555 }
2556 
2557 static int bnxt_del_dflt_mac_filter(struct bnxt *bp,
2558 				    struct bnxt_vnic_info *vnic)
2559 {
2560 	struct bnxt_filter_info *filter;
2561 	int rc;
2562 
2563 	filter = STAILQ_FIRST(&vnic->filter);
2564 	while (filter) {
2565 		if (filter->mac_index == 0 &&
2566 		    !memcmp(filter->l2_addr, bp->mac_addr,
2567 			    RTE_ETHER_ADDR_LEN)) {
2568 			rc = bnxt_hwrm_clear_l2_filter(bp, filter);
2569 			if (!rc) {
2570 				STAILQ_REMOVE(&vnic->filter, filter,
2571 					      bnxt_filter_info, next);
2572 				bnxt_free_filter(bp, filter);
2573 			}
2574 			return rc;
2575 		}
2576 		filter = STAILQ_NEXT(filter, next);
2577 	}
2578 	return 0;
2579 }
2580 
2581 static int
2582 bnxt_config_vlan_hw_filter(struct bnxt *bp, uint64_t rx_offloads)
2583 {
2584 	struct bnxt_vnic_info *vnic;
2585 	unsigned int i;
2586 	int rc;
2587 
2588 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
2589 	if (!(rx_offloads & DEV_RX_OFFLOAD_VLAN_FILTER)) {
2590 		/* Remove any VLAN filters programmed */
2591 		for (i = 0; i < RTE_ETHER_MAX_VLAN_ID; i++)
2592 			bnxt_del_vlan_filter(bp, i);
2593 
2594 		rc = bnxt_add_mac_filter(bp, vnic, NULL, 0, 0);
2595 		if (rc)
2596 			return rc;
2597 	} else {
2598 		/* Default filter will allow packets that match the
2599 		 * dest mac. So, it has to be deleted, otherwise, we
2600 		 * will endup receiving vlan packets for which the
2601 		 * filter is not programmed, when hw-vlan-filter
2602 		 * configuration is ON
2603 		 */
2604 		bnxt_del_dflt_mac_filter(bp, vnic);
2605 		/* This filter will allow only untagged packets */
2606 		bnxt_add_vlan_filter(bp, 0);
2607 	}
2608 	PMD_DRV_LOG(DEBUG, "VLAN Filtering: %d\n",
2609 		    !!(rx_offloads & DEV_RX_OFFLOAD_VLAN_FILTER));
2610 
2611 	return 0;
2612 }
2613 
2614 static int bnxt_free_one_vnic(struct bnxt *bp, uint16_t vnic_id)
2615 {
2616 	struct bnxt_vnic_info *vnic = &bp->vnic_info[vnic_id];
2617 	unsigned int i;
2618 	int rc;
2619 
2620 	/* Destroy vnic filters and vnic */
2621 	if (bp->eth_dev->data->dev_conf.rxmode.offloads &
2622 	    DEV_RX_OFFLOAD_VLAN_FILTER) {
2623 		for (i = 0; i < RTE_ETHER_MAX_VLAN_ID; i++)
2624 			bnxt_del_vlan_filter(bp, i);
2625 	}
2626 	bnxt_del_dflt_mac_filter(bp, vnic);
2627 
2628 	rc = bnxt_hwrm_vnic_ctx_free(bp, vnic);
2629 	if (rc)
2630 		return rc;
2631 
2632 	rc = bnxt_hwrm_vnic_free(bp, vnic);
2633 	if (rc)
2634 		return rc;
2635 
2636 	rte_free(vnic->fw_grp_ids);
2637 	vnic->fw_grp_ids = NULL;
2638 
2639 	vnic->rx_queue_cnt = 0;
2640 
2641 	return 0;
2642 }
2643 
2644 static int
2645 bnxt_config_vlan_hw_stripping(struct bnxt *bp, uint64_t rx_offloads)
2646 {
2647 	struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
2648 	int rc;
2649 
2650 	/* Destroy, recreate and reconfigure the default vnic */
2651 	rc = bnxt_free_one_vnic(bp, 0);
2652 	if (rc)
2653 		return rc;
2654 
2655 	/* default vnic 0 */
2656 	rc = bnxt_setup_one_vnic(bp, 0);
2657 	if (rc)
2658 		return rc;
2659 
2660 	if (bp->eth_dev->data->dev_conf.rxmode.offloads &
2661 	    DEV_RX_OFFLOAD_VLAN_FILTER) {
2662 		rc = bnxt_add_vlan_filter(bp, 0);
2663 		if (rc)
2664 			return rc;
2665 		rc = bnxt_restore_vlan_filters(bp);
2666 		if (rc)
2667 			return rc;
2668 	} else {
2669 		rc = bnxt_add_mac_filter(bp, vnic, NULL, 0, 0);
2670 		if (rc)
2671 			return rc;
2672 	}
2673 
2674 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
2675 	if (rc)
2676 		return rc;
2677 
2678 	PMD_DRV_LOG(DEBUG, "VLAN Strip Offload: %d\n",
2679 		    !!(rx_offloads & DEV_RX_OFFLOAD_VLAN_STRIP));
2680 
2681 	return rc;
2682 }
2683 
2684 static int
2685 bnxt_vlan_offload_set_op(struct rte_eth_dev *dev, int mask)
2686 {
2687 	uint64_t rx_offloads = dev->data->dev_conf.rxmode.offloads;
2688 	struct bnxt *bp = dev->data->dev_private;
2689 	int rc;
2690 
2691 	rc = is_bnxt_in_error(bp);
2692 	if (rc)
2693 		return rc;
2694 
2695 	/* Filter settings will get applied when port is started */
2696 	if (!dev->data->dev_started)
2697 		return 0;
2698 
2699 	if (mask & ETH_VLAN_FILTER_MASK) {
2700 		/* Enable or disable VLAN filtering */
2701 		rc = bnxt_config_vlan_hw_filter(bp, rx_offloads);
2702 		if (rc)
2703 			return rc;
2704 	}
2705 
2706 	if (mask & ETH_VLAN_STRIP_MASK) {
2707 		/* Enable or disable VLAN stripping */
2708 		rc = bnxt_config_vlan_hw_stripping(bp, rx_offloads);
2709 		if (rc)
2710 			return rc;
2711 	}
2712 
2713 	if (mask & ETH_VLAN_EXTEND_MASK) {
2714 		if (rx_offloads & DEV_RX_OFFLOAD_VLAN_EXTEND)
2715 			PMD_DRV_LOG(DEBUG, "Extend VLAN supported\n");
2716 		else
2717 			PMD_DRV_LOG(INFO, "Extend VLAN unsupported\n");
2718 	}
2719 
2720 	return 0;
2721 }
2722 
2723 static int
2724 bnxt_vlan_tpid_set_op(struct rte_eth_dev *dev, enum rte_vlan_type vlan_type,
2725 		      uint16_t tpid)
2726 {
2727 	struct bnxt *bp = dev->data->dev_private;
2728 	int qinq = dev->data->dev_conf.rxmode.offloads &
2729 		   DEV_RX_OFFLOAD_VLAN_EXTEND;
2730 
2731 	if (vlan_type != ETH_VLAN_TYPE_INNER &&
2732 	    vlan_type != ETH_VLAN_TYPE_OUTER) {
2733 		PMD_DRV_LOG(ERR,
2734 			    "Unsupported vlan type.");
2735 		return -EINVAL;
2736 	}
2737 	if (!qinq) {
2738 		PMD_DRV_LOG(ERR,
2739 			    "QinQ not enabled. Needs to be ON as we can "
2740 			    "accelerate only outer vlan\n");
2741 		return -EINVAL;
2742 	}
2743 
2744 	if (vlan_type == ETH_VLAN_TYPE_OUTER) {
2745 		switch (tpid) {
2746 		case RTE_ETHER_TYPE_QINQ:
2747 			bp->outer_tpid_bd =
2748 				TX_BD_LONG_CFA_META_VLAN_TPID_TPID88A8;
2749 				break;
2750 		case RTE_ETHER_TYPE_VLAN:
2751 			bp->outer_tpid_bd =
2752 				TX_BD_LONG_CFA_META_VLAN_TPID_TPID8100;
2753 				break;
2754 		case RTE_ETHER_TYPE_QINQ1:
2755 			bp->outer_tpid_bd =
2756 				TX_BD_LONG_CFA_META_VLAN_TPID_TPID9100;
2757 				break;
2758 		case RTE_ETHER_TYPE_QINQ2:
2759 			bp->outer_tpid_bd =
2760 				TX_BD_LONG_CFA_META_VLAN_TPID_TPID9200;
2761 				break;
2762 		case RTE_ETHER_TYPE_QINQ3:
2763 			bp->outer_tpid_bd =
2764 				 TX_BD_LONG_CFA_META_VLAN_TPID_TPID9300;
2765 				break;
2766 		default:
2767 			PMD_DRV_LOG(ERR, "Invalid TPID: %x\n", tpid);
2768 			return -EINVAL;
2769 		}
2770 		bp->outer_tpid_bd |= tpid;
2771 		PMD_DRV_LOG(INFO, "outer_tpid_bd = %x\n", bp->outer_tpid_bd);
2772 	} else if (vlan_type == ETH_VLAN_TYPE_INNER) {
2773 		PMD_DRV_LOG(ERR,
2774 			    "Can accelerate only outer vlan in QinQ\n");
2775 		return -EINVAL;
2776 	}
2777 
2778 	return 0;
2779 }
2780 
2781 static int
2782 bnxt_set_default_mac_addr_op(struct rte_eth_dev *dev,
2783 			     struct rte_ether_addr *addr)
2784 {
2785 	struct bnxt *bp = dev->data->dev_private;
2786 	/* Default Filter is tied to VNIC 0 */
2787 	struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
2788 	int rc;
2789 
2790 	rc = is_bnxt_in_error(bp);
2791 	if (rc)
2792 		return rc;
2793 
2794 	if (BNXT_VF(bp) && !BNXT_VF_IS_TRUSTED(bp))
2795 		return -EPERM;
2796 
2797 	if (rte_is_zero_ether_addr(addr))
2798 		return -EINVAL;
2799 
2800 	/* Filter settings will get applied when port is started */
2801 	if (!dev->data->dev_started)
2802 		return 0;
2803 
2804 	/* Check if the requested MAC is already added */
2805 	if (memcmp(addr, bp->mac_addr, RTE_ETHER_ADDR_LEN) == 0)
2806 		return 0;
2807 
2808 	/* Destroy filter and re-create it */
2809 	bnxt_del_dflt_mac_filter(bp, vnic);
2810 
2811 	memcpy(bp->mac_addr, addr, RTE_ETHER_ADDR_LEN);
2812 	if (dev->data->dev_conf.rxmode.offloads & DEV_RX_OFFLOAD_VLAN_FILTER) {
2813 		/* This filter will allow only untagged packets */
2814 		rc = bnxt_add_vlan_filter(bp, 0);
2815 	} else {
2816 		rc = bnxt_add_mac_filter(bp, vnic, addr, 0, 0);
2817 	}
2818 
2819 	PMD_DRV_LOG(DEBUG, "Set MAC addr\n");
2820 	return rc;
2821 }
2822 
2823 static int
2824 bnxt_dev_set_mc_addr_list_op(struct rte_eth_dev *eth_dev,
2825 			  struct rte_ether_addr *mc_addr_set,
2826 			  uint32_t nb_mc_addr)
2827 {
2828 	struct bnxt *bp = eth_dev->data->dev_private;
2829 	char *mc_addr_list = (char *)mc_addr_set;
2830 	struct bnxt_vnic_info *vnic;
2831 	uint32_t off = 0, i = 0;
2832 	int rc;
2833 
2834 	rc = is_bnxt_in_error(bp);
2835 	if (rc)
2836 		return rc;
2837 
2838 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
2839 
2840 	if (nb_mc_addr > BNXT_MAX_MC_ADDRS) {
2841 		vnic->flags |= BNXT_VNIC_INFO_ALLMULTI;
2842 		goto allmulti;
2843 	}
2844 
2845 	/* TODO Check for Duplicate mcast addresses */
2846 	vnic->flags &= ~BNXT_VNIC_INFO_ALLMULTI;
2847 	for (i = 0; i < nb_mc_addr; i++) {
2848 		memcpy(vnic->mc_list + off, &mc_addr_list[i],
2849 			RTE_ETHER_ADDR_LEN);
2850 		off += RTE_ETHER_ADDR_LEN;
2851 	}
2852 
2853 	vnic->mc_addr_cnt = i;
2854 	if (vnic->mc_addr_cnt)
2855 		vnic->flags |= BNXT_VNIC_INFO_MCAST;
2856 	else
2857 		vnic->flags &= ~BNXT_VNIC_INFO_MCAST;
2858 
2859 allmulti:
2860 	return bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
2861 }
2862 
2863 static int
2864 bnxt_fw_version_get(struct rte_eth_dev *dev, char *fw_version, size_t fw_size)
2865 {
2866 	struct bnxt *bp = dev->data->dev_private;
2867 	uint8_t fw_major = (bp->fw_ver >> 24) & 0xff;
2868 	uint8_t fw_minor = (bp->fw_ver >> 16) & 0xff;
2869 	uint8_t fw_updt = (bp->fw_ver >> 8) & 0xff;
2870 	uint8_t fw_rsvd = bp->fw_ver & 0xff;
2871 	int ret;
2872 
2873 	ret = snprintf(fw_version, fw_size, "%d.%d.%d.%d",
2874 			fw_major, fw_minor, fw_updt, fw_rsvd);
2875 	if (ret < 0)
2876 		return -EINVAL;
2877 
2878 	ret += 1; /* add the size of '\0' */
2879 	if (fw_size < (size_t)ret)
2880 		return ret;
2881 	else
2882 		return 0;
2883 }
2884 
2885 static void
2886 bnxt_rxq_info_get_op(struct rte_eth_dev *dev, uint16_t queue_id,
2887 	struct rte_eth_rxq_info *qinfo)
2888 {
2889 	struct bnxt *bp = dev->data->dev_private;
2890 	struct bnxt_rx_queue *rxq;
2891 
2892 	if (is_bnxt_in_error(bp))
2893 		return;
2894 
2895 	rxq = dev->data->rx_queues[queue_id];
2896 
2897 	qinfo->mp = rxq->mb_pool;
2898 	qinfo->scattered_rx = dev->data->scattered_rx;
2899 	qinfo->nb_desc = rxq->nb_rx_desc;
2900 
2901 	qinfo->conf.rx_free_thresh = rxq->rx_free_thresh;
2902 	qinfo->conf.rx_drop_en = rxq->drop_en;
2903 	qinfo->conf.rx_deferred_start = rxq->rx_deferred_start;
2904 	qinfo->conf.offloads = dev->data->dev_conf.rxmode.offloads;
2905 }
2906 
2907 static void
2908 bnxt_txq_info_get_op(struct rte_eth_dev *dev, uint16_t queue_id,
2909 	struct rte_eth_txq_info *qinfo)
2910 {
2911 	struct bnxt *bp = dev->data->dev_private;
2912 	struct bnxt_tx_queue *txq;
2913 
2914 	if (is_bnxt_in_error(bp))
2915 		return;
2916 
2917 	txq = dev->data->tx_queues[queue_id];
2918 
2919 	qinfo->nb_desc = txq->nb_tx_desc;
2920 
2921 	qinfo->conf.tx_thresh.pthresh = txq->pthresh;
2922 	qinfo->conf.tx_thresh.hthresh = txq->hthresh;
2923 	qinfo->conf.tx_thresh.wthresh = txq->wthresh;
2924 
2925 	qinfo->conf.tx_free_thresh = txq->tx_free_thresh;
2926 	qinfo->conf.tx_rs_thresh = 0;
2927 	qinfo->conf.tx_deferred_start = txq->tx_deferred_start;
2928 	qinfo->conf.offloads = txq->offloads;
2929 }
2930 
2931 static const struct {
2932 	eth_rx_burst_t pkt_burst;
2933 	const char *info;
2934 } bnxt_rx_burst_info[] = {
2935 	{bnxt_recv_pkts,		"Scalar"},
2936 #if defined(RTE_ARCH_X86)
2937 	{bnxt_recv_pkts_vec,		"Vector SSE"},
2938 #endif
2939 #if defined(RTE_ARCH_X86) && defined(CC_AVX2_SUPPORT)
2940 	{bnxt_recv_pkts_vec_avx2,	"Vector AVX2"},
2941 #endif
2942 #if defined(RTE_ARCH_ARM64)
2943 	{bnxt_recv_pkts_vec,		"Vector Neon"},
2944 #endif
2945 };
2946 
2947 static int
2948 bnxt_rx_burst_mode_get(struct rte_eth_dev *dev, __rte_unused uint16_t queue_id,
2949 		       struct rte_eth_burst_mode *mode)
2950 {
2951 	eth_rx_burst_t pkt_burst = dev->rx_pkt_burst;
2952 	size_t i;
2953 
2954 	for (i = 0; i < RTE_DIM(bnxt_rx_burst_info); i++) {
2955 		if (pkt_burst == bnxt_rx_burst_info[i].pkt_burst) {
2956 			snprintf(mode->info, sizeof(mode->info), "%s",
2957 				 bnxt_rx_burst_info[i].info);
2958 			return 0;
2959 		}
2960 	}
2961 
2962 	return -EINVAL;
2963 }
2964 
2965 static const struct {
2966 	eth_tx_burst_t pkt_burst;
2967 	const char *info;
2968 } bnxt_tx_burst_info[] = {
2969 	{bnxt_xmit_pkts,		"Scalar"},
2970 #if defined(RTE_ARCH_X86)
2971 	{bnxt_xmit_pkts_vec,		"Vector SSE"},
2972 #endif
2973 #if defined(RTE_ARCH_X86) && defined(CC_AVX2_SUPPORT)
2974 	{bnxt_xmit_pkts_vec_avx2,	"Vector AVX2"},
2975 #endif
2976 #if defined(RTE_ARCH_ARM64)
2977 	{bnxt_xmit_pkts_vec,		"Vector Neon"},
2978 #endif
2979 };
2980 
2981 static int
2982 bnxt_tx_burst_mode_get(struct rte_eth_dev *dev, __rte_unused uint16_t queue_id,
2983 		       struct rte_eth_burst_mode *mode)
2984 {
2985 	eth_tx_burst_t pkt_burst = dev->tx_pkt_burst;
2986 	size_t i;
2987 
2988 	for (i = 0; i < RTE_DIM(bnxt_tx_burst_info); i++) {
2989 		if (pkt_burst == bnxt_tx_burst_info[i].pkt_burst) {
2990 			snprintf(mode->info, sizeof(mode->info), "%s",
2991 				 bnxt_tx_burst_info[i].info);
2992 			return 0;
2993 		}
2994 	}
2995 
2996 	return -EINVAL;
2997 }
2998 
2999 int bnxt_mtu_set_op(struct rte_eth_dev *eth_dev, uint16_t new_mtu)
3000 {
3001 	struct bnxt *bp = eth_dev->data->dev_private;
3002 	uint32_t new_pkt_size;
3003 	uint32_t rc = 0;
3004 	uint32_t i;
3005 
3006 	rc = is_bnxt_in_error(bp);
3007 	if (rc)
3008 		return rc;
3009 
3010 	/* Exit if receive queues are not configured yet */
3011 	if (!eth_dev->data->nb_rx_queues)
3012 		return rc;
3013 
3014 	new_pkt_size = new_mtu + RTE_ETHER_HDR_LEN + RTE_ETHER_CRC_LEN +
3015 		       VLAN_TAG_SIZE * BNXT_NUM_VLANS;
3016 
3017 	/*
3018 	 * Disallow any MTU change that would require scattered receive support
3019 	 * if it is not already enabled.
3020 	 */
3021 	if (eth_dev->data->dev_started &&
3022 	    !eth_dev->data->scattered_rx &&
3023 	    (new_pkt_size >
3024 	     eth_dev->data->min_rx_buf_size - RTE_PKTMBUF_HEADROOM)) {
3025 		PMD_DRV_LOG(ERR,
3026 			    "MTU change would require scattered rx support. ");
3027 		PMD_DRV_LOG(ERR, "Stop port before changing MTU.\n");
3028 		return -EINVAL;
3029 	}
3030 
3031 	if (new_mtu > RTE_ETHER_MTU) {
3032 		bp->flags |= BNXT_FLAG_JUMBO;
3033 		bp->eth_dev->data->dev_conf.rxmode.offloads |=
3034 			DEV_RX_OFFLOAD_JUMBO_FRAME;
3035 	} else {
3036 		bp->eth_dev->data->dev_conf.rxmode.offloads &=
3037 			~DEV_RX_OFFLOAD_JUMBO_FRAME;
3038 		bp->flags &= ~BNXT_FLAG_JUMBO;
3039 	}
3040 
3041 	/* Is there a change in mtu setting? */
3042 	if (eth_dev->data->dev_conf.rxmode.max_rx_pkt_len == new_pkt_size)
3043 		return rc;
3044 
3045 	for (i = 0; i < bp->nr_vnics; i++) {
3046 		struct bnxt_vnic_info *vnic = &bp->vnic_info[i];
3047 		uint16_t size = 0;
3048 
3049 		vnic->mru = BNXT_VNIC_MRU(new_mtu);
3050 		rc = bnxt_hwrm_vnic_cfg(bp, vnic);
3051 		if (rc)
3052 			break;
3053 
3054 		size = rte_pktmbuf_data_room_size(bp->rx_queues[0]->mb_pool);
3055 		size -= RTE_PKTMBUF_HEADROOM;
3056 
3057 		if (size < new_mtu) {
3058 			rc = bnxt_hwrm_vnic_plcmode_cfg(bp, vnic);
3059 			if (rc)
3060 				return rc;
3061 		}
3062 	}
3063 
3064 	if (!rc)
3065 		eth_dev->data->dev_conf.rxmode.max_rx_pkt_len = new_pkt_size;
3066 
3067 	PMD_DRV_LOG(INFO, "New MTU is %d\n", new_mtu);
3068 
3069 	return rc;
3070 }
3071 
3072 static int
3073 bnxt_vlan_pvid_set_op(struct rte_eth_dev *dev, uint16_t pvid, int on)
3074 {
3075 	struct bnxt *bp = dev->data->dev_private;
3076 	uint16_t vlan = bp->vlan;
3077 	int rc;
3078 
3079 	rc = is_bnxt_in_error(bp);
3080 	if (rc)
3081 		return rc;
3082 
3083 	if (!BNXT_SINGLE_PF(bp)) {
3084 		PMD_DRV_LOG(ERR, "PVID cannot be modified on VF or on shared PF\n");
3085 		return -ENOTSUP;
3086 	}
3087 	bp->vlan = on ? pvid : 0;
3088 
3089 	rc = bnxt_hwrm_set_default_vlan(bp, 0, 0);
3090 	if (rc)
3091 		bp->vlan = vlan;
3092 	return rc;
3093 }
3094 
3095 static int
3096 bnxt_dev_led_on_op(struct rte_eth_dev *dev)
3097 {
3098 	struct bnxt *bp = dev->data->dev_private;
3099 	int rc;
3100 
3101 	rc = is_bnxt_in_error(bp);
3102 	if (rc)
3103 		return rc;
3104 
3105 	return bnxt_hwrm_port_led_cfg(bp, true);
3106 }
3107 
3108 static int
3109 bnxt_dev_led_off_op(struct rte_eth_dev *dev)
3110 {
3111 	struct bnxt *bp = dev->data->dev_private;
3112 	int rc;
3113 
3114 	rc = is_bnxt_in_error(bp);
3115 	if (rc)
3116 		return rc;
3117 
3118 	return bnxt_hwrm_port_led_cfg(bp, false);
3119 }
3120 
3121 static uint32_t
3122 bnxt_rx_queue_count_op(struct rte_eth_dev *dev, uint16_t rx_queue_id)
3123 {
3124 	struct bnxt *bp = (struct bnxt *)dev->data->dev_private;
3125 	struct bnxt_cp_ring_info *cpr;
3126 	uint32_t desc = 0, raw_cons;
3127 	struct bnxt_rx_queue *rxq;
3128 	struct rx_pkt_cmpl *rxcmp;
3129 	int rc;
3130 
3131 	rc = is_bnxt_in_error(bp);
3132 	if (rc)
3133 		return rc;
3134 
3135 	rxq = dev->data->rx_queues[rx_queue_id];
3136 	cpr = rxq->cp_ring;
3137 	raw_cons = cpr->cp_raw_cons;
3138 
3139 	while (1) {
3140 		uint32_t agg_cnt, cons, cmpl_type;
3141 
3142 		cons = RING_CMP(cpr->cp_ring_struct, raw_cons);
3143 		rxcmp = (struct rx_pkt_cmpl *)&cpr->cp_desc_ring[cons];
3144 
3145 		if (!CMP_VALID(rxcmp, raw_cons, cpr->cp_ring_struct))
3146 			break;
3147 
3148 		cmpl_type = CMP_TYPE(rxcmp);
3149 
3150 		switch (cmpl_type) {
3151 		case CMPL_BASE_TYPE_RX_L2:
3152 		case CMPL_BASE_TYPE_RX_L2_V2:
3153 			agg_cnt = BNXT_RX_L2_AGG_BUFS(rxcmp);
3154 			raw_cons = raw_cons + CMP_LEN(cmpl_type) + agg_cnt;
3155 			desc++;
3156 			break;
3157 
3158 		case CMPL_BASE_TYPE_RX_TPA_END:
3159 			if (BNXT_CHIP_P5(rxq->bp)) {
3160 				struct rx_tpa_v2_end_cmpl_hi *p5_tpa_end;
3161 
3162 				p5_tpa_end = (void *)rxcmp;
3163 				agg_cnt = BNXT_TPA_END_AGG_BUFS_TH(p5_tpa_end);
3164 			} else {
3165 				struct rx_tpa_end_cmpl *tpa_end;
3166 
3167 				tpa_end = (void *)rxcmp;
3168 				agg_cnt = BNXT_TPA_END_AGG_BUFS(tpa_end);
3169 			}
3170 
3171 			raw_cons = raw_cons + CMP_LEN(cmpl_type) + agg_cnt;
3172 			desc++;
3173 			break;
3174 
3175 		default:
3176 			raw_cons += CMP_LEN(cmpl_type);
3177 		}
3178 	}
3179 
3180 	return desc;
3181 }
3182 
3183 static int
3184 bnxt_rx_descriptor_status_op(void *rx_queue, uint16_t offset)
3185 {
3186 	struct bnxt_rx_queue *rxq = rx_queue;
3187 	struct bnxt_cp_ring_info *cpr;
3188 	struct bnxt_rx_ring_info *rxr;
3189 	uint32_t desc, raw_cons;
3190 	struct bnxt *bp = rxq->bp;
3191 	struct rx_pkt_cmpl *rxcmp;
3192 	int rc;
3193 
3194 	rc = is_bnxt_in_error(bp);
3195 	if (rc)
3196 		return rc;
3197 
3198 	if (offset >= rxq->nb_rx_desc)
3199 		return -EINVAL;
3200 
3201 	rxr = rxq->rx_ring;
3202 	cpr = rxq->cp_ring;
3203 
3204 	/*
3205 	 * For the vector receive case, the completion at the requested
3206 	 * offset can be indexed directly.
3207 	 */
3208 #if defined(RTE_ARCH_X86) || defined(RTE_ARCH_ARM64)
3209 	if (bp->flags & BNXT_FLAG_RX_VECTOR_PKT_MODE) {
3210 		struct rx_pkt_cmpl *rxcmp;
3211 		uint32_t cons;
3212 
3213 		/* Check status of completion descriptor. */
3214 		raw_cons = cpr->cp_raw_cons +
3215 			   offset * CMP_LEN(CMPL_BASE_TYPE_RX_L2);
3216 		cons = RING_CMP(cpr->cp_ring_struct, raw_cons);
3217 		rxcmp = (struct rx_pkt_cmpl *)&cpr->cp_desc_ring[cons];
3218 
3219 		if (CMP_VALID(rxcmp, raw_cons, cpr->cp_ring_struct))
3220 			return RTE_ETH_RX_DESC_DONE;
3221 
3222 		/* Check whether rx desc has an mbuf attached. */
3223 		cons = RING_CMP(rxr->rx_ring_struct, raw_cons / 2);
3224 		if (cons >= rxq->rxrearm_start &&
3225 		    cons < rxq->rxrearm_start + rxq->rxrearm_nb) {
3226 			return RTE_ETH_RX_DESC_UNAVAIL;
3227 		}
3228 
3229 		return RTE_ETH_RX_DESC_AVAIL;
3230 	}
3231 #endif
3232 
3233 	/*
3234 	 * For the non-vector receive case, scan the completion ring to
3235 	 * locate the completion descriptor for the requested offset.
3236 	 */
3237 	raw_cons = cpr->cp_raw_cons;
3238 	desc = 0;
3239 	while (1) {
3240 		uint32_t agg_cnt, cons, cmpl_type;
3241 
3242 		cons = RING_CMP(cpr->cp_ring_struct, raw_cons);
3243 		rxcmp = (struct rx_pkt_cmpl *)&cpr->cp_desc_ring[cons];
3244 
3245 		if (!CMP_VALID(rxcmp, raw_cons, cpr->cp_ring_struct))
3246 			break;
3247 
3248 		cmpl_type = CMP_TYPE(rxcmp);
3249 
3250 		switch (cmpl_type) {
3251 		case CMPL_BASE_TYPE_RX_L2:
3252 		case CMPL_BASE_TYPE_RX_L2_V2:
3253 			if (desc == offset) {
3254 				cons = rxcmp->opaque;
3255 				if (rxr->rx_buf_ring[cons])
3256 					return RTE_ETH_RX_DESC_DONE;
3257 				else
3258 					return RTE_ETH_RX_DESC_UNAVAIL;
3259 			}
3260 			agg_cnt = BNXT_RX_L2_AGG_BUFS(rxcmp);
3261 			raw_cons = raw_cons + CMP_LEN(cmpl_type) + agg_cnt;
3262 			desc++;
3263 			break;
3264 
3265 		case CMPL_BASE_TYPE_RX_TPA_END:
3266 			if (desc == offset)
3267 				return RTE_ETH_RX_DESC_DONE;
3268 
3269 			if (BNXT_CHIP_P5(rxq->bp)) {
3270 				struct rx_tpa_v2_end_cmpl_hi *p5_tpa_end;
3271 
3272 				p5_tpa_end = (void *)rxcmp;
3273 				agg_cnt = BNXT_TPA_END_AGG_BUFS_TH(p5_tpa_end);
3274 			} else {
3275 				struct rx_tpa_end_cmpl *tpa_end;
3276 
3277 				tpa_end = (void *)rxcmp;
3278 				agg_cnt = BNXT_TPA_END_AGG_BUFS(tpa_end);
3279 			}
3280 
3281 			raw_cons = raw_cons + CMP_LEN(cmpl_type) + agg_cnt;
3282 			desc++;
3283 			break;
3284 
3285 		default:
3286 			raw_cons += CMP_LEN(cmpl_type);
3287 		}
3288 	}
3289 
3290 	return RTE_ETH_RX_DESC_AVAIL;
3291 }
3292 
3293 static int
3294 bnxt_tx_descriptor_status_op(void *tx_queue, uint16_t offset)
3295 {
3296 	struct bnxt_tx_queue *txq = (struct bnxt_tx_queue *)tx_queue;
3297 	struct bnxt_tx_ring_info *txr;
3298 	struct bnxt_cp_ring_info *cpr;
3299 	struct rte_mbuf **tx_buf;
3300 	struct tx_pkt_cmpl *txcmp;
3301 	uint32_t cons, cp_cons;
3302 	int rc;
3303 
3304 	if (!txq)
3305 		return -EINVAL;
3306 
3307 	rc = is_bnxt_in_error(txq->bp);
3308 	if (rc)
3309 		return rc;
3310 
3311 	cpr = txq->cp_ring;
3312 	txr = txq->tx_ring;
3313 
3314 	if (offset >= txq->nb_tx_desc)
3315 		return -EINVAL;
3316 
3317 	cons = RING_CMP(cpr->cp_ring_struct, offset);
3318 	txcmp = (struct tx_pkt_cmpl *)&cpr->cp_desc_ring[cons];
3319 	cp_cons = cpr->cp_raw_cons;
3320 
3321 	if (cons > cp_cons) {
3322 		if (CMPL_VALID(txcmp, cpr->valid))
3323 			return RTE_ETH_TX_DESC_UNAVAIL;
3324 	} else {
3325 		if (CMPL_VALID(txcmp, !cpr->valid))
3326 			return RTE_ETH_TX_DESC_UNAVAIL;
3327 	}
3328 	tx_buf = &txr->tx_buf_ring[cons];
3329 	if (*tx_buf == NULL)
3330 		return RTE_ETH_TX_DESC_DONE;
3331 
3332 	return RTE_ETH_TX_DESC_FULL;
3333 }
3334 
3335 int
3336 bnxt_flow_ops_get_op(struct rte_eth_dev *dev,
3337 		     const struct rte_flow_ops **ops)
3338 {
3339 	struct bnxt *bp = dev->data->dev_private;
3340 	int ret = 0;
3341 
3342 	if (!bp)
3343 		return -EIO;
3344 
3345 	if (BNXT_ETH_DEV_IS_REPRESENTOR(dev)) {
3346 		struct bnxt_representor *vfr = dev->data->dev_private;
3347 		bp = vfr->parent_dev->data->dev_private;
3348 		/* parent is deleted while children are still valid */
3349 		if (!bp) {
3350 			PMD_DRV_LOG(DEBUG, "BNXT Port:%d VFR Error\n",
3351 				    dev->data->port_id);
3352 			return -EIO;
3353 		}
3354 	}
3355 
3356 	ret = is_bnxt_in_error(bp);
3357 	if (ret)
3358 		return ret;
3359 
3360 	/* PMD supports thread-safe flow operations.  rte_flow API
3361 	 * functions can avoid mutex for multi-thread safety.
3362 	 */
3363 	dev->data->dev_flags |= RTE_ETH_DEV_FLOW_OPS_THREAD_SAFE;
3364 
3365 	if (BNXT_TRUFLOW_EN(bp))
3366 		*ops = &bnxt_ulp_rte_flow_ops;
3367 	else
3368 		*ops = &bnxt_flow_ops;
3369 
3370 	return ret;
3371 }
3372 
3373 static const uint32_t *
3374 bnxt_dev_supported_ptypes_get_op(struct rte_eth_dev *dev)
3375 {
3376 	static const uint32_t ptypes[] = {
3377 		RTE_PTYPE_L2_ETHER_VLAN,
3378 		RTE_PTYPE_L3_IPV4_EXT_UNKNOWN,
3379 		RTE_PTYPE_L3_IPV6_EXT_UNKNOWN,
3380 		RTE_PTYPE_L4_ICMP,
3381 		RTE_PTYPE_L4_TCP,
3382 		RTE_PTYPE_L4_UDP,
3383 		RTE_PTYPE_INNER_L3_IPV4_EXT_UNKNOWN,
3384 		RTE_PTYPE_INNER_L3_IPV6_EXT_UNKNOWN,
3385 		RTE_PTYPE_INNER_L4_ICMP,
3386 		RTE_PTYPE_INNER_L4_TCP,
3387 		RTE_PTYPE_INNER_L4_UDP,
3388 		RTE_PTYPE_UNKNOWN
3389 	};
3390 
3391 	if (!dev->rx_pkt_burst)
3392 		return NULL;
3393 
3394 	return ptypes;
3395 }
3396 
3397 static int bnxt_map_regs(struct bnxt *bp, uint32_t *reg_arr, int count,
3398 			 int reg_win)
3399 {
3400 	uint32_t reg_base = *reg_arr & 0xfffff000;
3401 	uint32_t win_off;
3402 	int i;
3403 
3404 	for (i = 0; i < count; i++) {
3405 		if ((reg_arr[i] & 0xfffff000) != reg_base)
3406 			return -ERANGE;
3407 	}
3408 	win_off = BNXT_GRCPF_REG_WINDOW_BASE_OUT + (reg_win - 1) * 4;
3409 	rte_write32(reg_base, (uint8_t *)bp->bar0 + win_off);
3410 	return 0;
3411 }
3412 
3413 static int bnxt_map_ptp_regs(struct bnxt *bp)
3414 {
3415 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3416 	uint32_t *reg_arr;
3417 	int rc, i;
3418 
3419 	reg_arr = ptp->rx_regs;
3420 	rc = bnxt_map_regs(bp, reg_arr, BNXT_PTP_RX_REGS, 5);
3421 	if (rc)
3422 		return rc;
3423 
3424 	reg_arr = ptp->tx_regs;
3425 	rc = bnxt_map_regs(bp, reg_arr, BNXT_PTP_TX_REGS, 6);
3426 	if (rc)
3427 		return rc;
3428 
3429 	for (i = 0; i < BNXT_PTP_RX_REGS; i++)
3430 		ptp->rx_mapped_regs[i] = 0x5000 + (ptp->rx_regs[i] & 0xfff);
3431 
3432 	for (i = 0; i < BNXT_PTP_TX_REGS; i++)
3433 		ptp->tx_mapped_regs[i] = 0x6000 + (ptp->tx_regs[i] & 0xfff);
3434 
3435 	return 0;
3436 }
3437 
3438 static void bnxt_unmap_ptp_regs(struct bnxt *bp)
3439 {
3440 	rte_write32(0, (uint8_t *)bp->bar0 +
3441 			 BNXT_GRCPF_REG_WINDOW_BASE_OUT + 16);
3442 	rte_write32(0, (uint8_t *)bp->bar0 +
3443 			 BNXT_GRCPF_REG_WINDOW_BASE_OUT + 20);
3444 }
3445 
3446 static uint64_t bnxt_cc_read(struct bnxt *bp)
3447 {
3448 	uint64_t ns;
3449 
3450 	ns = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3451 			      BNXT_GRCPF_REG_SYNC_TIME));
3452 	ns |= (uint64_t)(rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3453 					  BNXT_GRCPF_REG_SYNC_TIME + 4))) << 32;
3454 	return ns;
3455 }
3456 
3457 static int bnxt_get_tx_ts(struct bnxt *bp, uint64_t *ts)
3458 {
3459 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3460 	uint32_t fifo;
3461 
3462 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3463 				ptp->tx_mapped_regs[BNXT_PTP_TX_FIFO]));
3464 	if (fifo & BNXT_PTP_TX_FIFO_EMPTY)
3465 		return -EAGAIN;
3466 
3467 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3468 				ptp->tx_mapped_regs[BNXT_PTP_TX_FIFO]));
3469 	*ts = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3470 				ptp->tx_mapped_regs[BNXT_PTP_TX_TS_L]));
3471 	*ts |= (uint64_t)rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3472 				ptp->tx_mapped_regs[BNXT_PTP_TX_TS_H])) << 32;
3473 	rte_read32((uint8_t *)bp->bar0 + ptp->tx_mapped_regs[BNXT_PTP_TX_SEQ]);
3474 
3475 	return 0;
3476 }
3477 
3478 static int bnxt_clr_rx_ts(struct bnxt *bp, uint64_t *last_ts)
3479 {
3480 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3481 	struct bnxt_pf_info *pf = bp->pf;
3482 	uint16_t port_id;
3483 	int i = 0;
3484 	uint32_t fifo;
3485 
3486 	if (!ptp || (bp->flags & BNXT_FLAG_CHIP_P5))
3487 		return -EINVAL;
3488 
3489 	port_id = pf->port_id;
3490 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3491 				ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO]));
3492 	while ((fifo & BNXT_PTP_RX_FIFO_PENDING) && (i < BNXT_PTP_RX_PND_CNT)) {
3493 		rte_write32(1 << port_id, (uint8_t *)bp->bar0 +
3494 			    ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO_ADV]);
3495 		fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3496 					ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO]));
3497 		*last_ts = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3498 					ptp->rx_mapped_regs[BNXT_PTP_RX_TS_L]));
3499 		*last_ts |= (uint64_t)rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3500 					ptp->rx_mapped_regs[BNXT_PTP_RX_TS_H])) << 32;
3501 		i++;
3502 	}
3503 
3504 	if (i >= BNXT_PTP_RX_PND_CNT)
3505 		return -EBUSY;
3506 
3507 	return 0;
3508 }
3509 
3510 static int bnxt_get_rx_ts(struct bnxt *bp, uint64_t *ts)
3511 {
3512 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3513 	struct bnxt_pf_info *pf = bp->pf;
3514 	uint16_t port_id;
3515 	uint32_t fifo;
3516 
3517 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3518 				ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO]));
3519 	if (!(fifo & BNXT_PTP_RX_FIFO_PENDING))
3520 		return -EAGAIN;
3521 
3522 	port_id = pf->port_id;
3523 	rte_write32(1 << port_id, (uint8_t *)bp->bar0 +
3524 	       ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO_ADV]);
3525 
3526 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3527 				   ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO]));
3528 	if (fifo & BNXT_PTP_RX_FIFO_PENDING)
3529 		return bnxt_clr_rx_ts(bp, ts);
3530 
3531 	*ts = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3532 				ptp->rx_mapped_regs[BNXT_PTP_RX_TS_L]));
3533 	*ts |= (uint64_t)rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3534 				ptp->rx_mapped_regs[BNXT_PTP_RX_TS_H])) << 32;
3535 
3536 	return 0;
3537 }
3538 
3539 static int
3540 bnxt_timesync_write_time(struct rte_eth_dev *dev, const struct timespec *ts)
3541 {
3542 	uint64_t ns;
3543 	struct bnxt *bp = dev->data->dev_private;
3544 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3545 
3546 	if (!ptp)
3547 		return -ENOTSUP;
3548 
3549 	ns = rte_timespec_to_ns(ts);
3550 	/* Set the timecounters to a new value. */
3551 	ptp->tc.nsec = ns;
3552 	ptp->tx_tstamp_tc.nsec = ns;
3553 	ptp->rx_tstamp_tc.nsec = ns;
3554 
3555 	return 0;
3556 }
3557 
3558 static int
3559 bnxt_timesync_read_time(struct rte_eth_dev *dev, struct timespec *ts)
3560 {
3561 	struct bnxt *bp = dev->data->dev_private;
3562 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3563 	uint64_t ns, systime_cycles = 0;
3564 	int rc = 0;
3565 
3566 	if (!ptp)
3567 		return -ENOTSUP;
3568 
3569 	if (BNXT_CHIP_P5(bp))
3570 		rc = bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_CURRENT_TIME,
3571 					     &systime_cycles);
3572 	else
3573 		systime_cycles = bnxt_cc_read(bp);
3574 
3575 	ns = rte_timecounter_update(&ptp->tc, systime_cycles);
3576 	*ts = rte_ns_to_timespec(ns);
3577 
3578 	return rc;
3579 }
3580 static int
3581 bnxt_timesync_enable(struct rte_eth_dev *dev)
3582 {
3583 	struct bnxt *bp = dev->data->dev_private;
3584 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3585 	uint32_t shift = 0;
3586 	int rc;
3587 
3588 	if (!ptp)
3589 		return -ENOTSUP;
3590 
3591 	ptp->rx_filter = 1;
3592 	ptp->tx_tstamp_en = 1;
3593 	ptp->rxctl = BNXT_PTP_MSG_EVENTS;
3594 
3595 	rc = bnxt_hwrm_ptp_cfg(bp);
3596 	if (rc)
3597 		return rc;
3598 
3599 	memset(&ptp->tc, 0, sizeof(struct rte_timecounter));
3600 	memset(&ptp->rx_tstamp_tc, 0, sizeof(struct rte_timecounter));
3601 	memset(&ptp->tx_tstamp_tc, 0, sizeof(struct rte_timecounter));
3602 
3603 	ptp->tc.cc_mask = BNXT_CYCLECOUNTER_MASK;
3604 	ptp->tc.cc_shift = shift;
3605 	ptp->tc.nsec_mask = (1ULL << shift) - 1;
3606 
3607 	ptp->rx_tstamp_tc.cc_mask = BNXT_CYCLECOUNTER_MASK;
3608 	ptp->rx_tstamp_tc.cc_shift = shift;
3609 	ptp->rx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
3610 
3611 	ptp->tx_tstamp_tc.cc_mask = BNXT_CYCLECOUNTER_MASK;
3612 	ptp->tx_tstamp_tc.cc_shift = shift;
3613 	ptp->tx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
3614 
3615 	if (!BNXT_CHIP_P5(bp))
3616 		bnxt_map_ptp_regs(bp);
3617 	else
3618 		rc = bnxt_ptp_start(bp);
3619 
3620 	return rc;
3621 }
3622 
3623 static int
3624 bnxt_timesync_disable(struct rte_eth_dev *dev)
3625 {
3626 	struct bnxt *bp = dev->data->dev_private;
3627 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3628 
3629 	if (!ptp)
3630 		return -ENOTSUP;
3631 
3632 	ptp->rx_filter = 0;
3633 	ptp->tx_tstamp_en = 0;
3634 	ptp->rxctl = 0;
3635 
3636 	bnxt_hwrm_ptp_cfg(bp);
3637 
3638 	if (!BNXT_CHIP_P5(bp))
3639 		bnxt_unmap_ptp_regs(bp);
3640 	else
3641 		bnxt_ptp_stop(bp);
3642 
3643 	return 0;
3644 }
3645 
3646 static int
3647 bnxt_timesync_read_rx_timestamp(struct rte_eth_dev *dev,
3648 				 struct timespec *timestamp,
3649 				 uint32_t flags __rte_unused)
3650 {
3651 	struct bnxt *bp = dev->data->dev_private;
3652 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3653 	uint64_t rx_tstamp_cycles = 0;
3654 	uint64_t ns;
3655 
3656 	if (!ptp)
3657 		return -ENOTSUP;
3658 
3659 	if (BNXT_CHIP_P5(bp))
3660 		rx_tstamp_cycles = ptp->rx_timestamp;
3661 	else
3662 		bnxt_get_rx_ts(bp, &rx_tstamp_cycles);
3663 
3664 	ns = rte_timecounter_update(&ptp->rx_tstamp_tc, rx_tstamp_cycles);
3665 	*timestamp = rte_ns_to_timespec(ns);
3666 	return  0;
3667 }
3668 
3669 static int
3670 bnxt_timesync_read_tx_timestamp(struct rte_eth_dev *dev,
3671 				 struct timespec *timestamp)
3672 {
3673 	struct bnxt *bp = dev->data->dev_private;
3674 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3675 	uint64_t tx_tstamp_cycles = 0;
3676 	uint64_t ns;
3677 	int rc = 0;
3678 
3679 	if (!ptp)
3680 		return -ENOTSUP;
3681 
3682 	if (BNXT_CHIP_P5(bp))
3683 		rc = bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_PATH_TX,
3684 					     &tx_tstamp_cycles);
3685 	else
3686 		rc = bnxt_get_tx_ts(bp, &tx_tstamp_cycles);
3687 
3688 	ns = rte_timecounter_update(&ptp->tx_tstamp_tc, tx_tstamp_cycles);
3689 	*timestamp = rte_ns_to_timespec(ns);
3690 
3691 	return rc;
3692 }
3693 
3694 static int
3695 bnxt_timesync_adjust_time(struct rte_eth_dev *dev, int64_t delta)
3696 {
3697 	struct bnxt *bp = dev->data->dev_private;
3698 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3699 
3700 	if (!ptp)
3701 		return -ENOTSUP;
3702 
3703 	ptp->tc.nsec += delta;
3704 	ptp->tx_tstamp_tc.nsec += delta;
3705 	ptp->rx_tstamp_tc.nsec += delta;
3706 
3707 	return 0;
3708 }
3709 
3710 static int
3711 bnxt_get_eeprom_length_op(struct rte_eth_dev *dev)
3712 {
3713 	struct bnxt *bp = dev->data->dev_private;
3714 	int rc;
3715 	uint32_t dir_entries;
3716 	uint32_t entry_length;
3717 
3718 	rc = is_bnxt_in_error(bp);
3719 	if (rc)
3720 		return rc;
3721 
3722 	PMD_DRV_LOG(INFO, PCI_PRI_FMT "\n",
3723 		    bp->pdev->addr.domain, bp->pdev->addr.bus,
3724 		    bp->pdev->addr.devid, bp->pdev->addr.function);
3725 
3726 	rc = bnxt_hwrm_nvm_get_dir_info(bp, &dir_entries, &entry_length);
3727 	if (rc != 0)
3728 		return rc;
3729 
3730 	return dir_entries * entry_length;
3731 }
3732 
3733 static int
3734 bnxt_get_eeprom_op(struct rte_eth_dev *dev,
3735 		struct rte_dev_eeprom_info *in_eeprom)
3736 {
3737 	struct bnxt *bp = dev->data->dev_private;
3738 	uint32_t index;
3739 	uint32_t offset;
3740 	int rc;
3741 
3742 	rc = is_bnxt_in_error(bp);
3743 	if (rc)
3744 		return rc;
3745 
3746 	PMD_DRV_LOG(INFO, PCI_PRI_FMT " in_eeprom->offset = %d len = %d\n",
3747 		    bp->pdev->addr.domain, bp->pdev->addr.bus,
3748 		    bp->pdev->addr.devid, bp->pdev->addr.function,
3749 		    in_eeprom->offset, in_eeprom->length);
3750 
3751 	if (in_eeprom->offset == 0) /* special offset value to get directory */
3752 		return bnxt_get_nvram_directory(bp, in_eeprom->length,
3753 						in_eeprom->data);
3754 
3755 	index = in_eeprom->offset >> 24;
3756 	offset = in_eeprom->offset & 0xffffff;
3757 
3758 	if (index != 0)
3759 		return bnxt_hwrm_get_nvram_item(bp, index - 1, offset,
3760 					   in_eeprom->length, in_eeprom->data);
3761 
3762 	return 0;
3763 }
3764 
3765 static bool bnxt_dir_type_is_ape_bin_format(uint16_t dir_type)
3766 {
3767 	switch (dir_type) {
3768 	case BNX_DIR_TYPE_CHIMP_PATCH:
3769 	case BNX_DIR_TYPE_BOOTCODE:
3770 	case BNX_DIR_TYPE_BOOTCODE_2:
3771 	case BNX_DIR_TYPE_APE_FW:
3772 	case BNX_DIR_TYPE_APE_PATCH:
3773 	case BNX_DIR_TYPE_KONG_FW:
3774 	case BNX_DIR_TYPE_KONG_PATCH:
3775 	case BNX_DIR_TYPE_BONO_FW:
3776 	case BNX_DIR_TYPE_BONO_PATCH:
3777 		/* FALLTHROUGH */
3778 		return true;
3779 	}
3780 
3781 	return false;
3782 }
3783 
3784 static bool bnxt_dir_type_is_other_exec_format(uint16_t dir_type)
3785 {
3786 	switch (dir_type) {
3787 	case BNX_DIR_TYPE_AVS:
3788 	case BNX_DIR_TYPE_EXP_ROM_MBA:
3789 	case BNX_DIR_TYPE_PCIE:
3790 	case BNX_DIR_TYPE_TSCF_UCODE:
3791 	case BNX_DIR_TYPE_EXT_PHY:
3792 	case BNX_DIR_TYPE_CCM:
3793 	case BNX_DIR_TYPE_ISCSI_BOOT:
3794 	case BNX_DIR_TYPE_ISCSI_BOOT_IPV6:
3795 	case BNX_DIR_TYPE_ISCSI_BOOT_IPV4N6:
3796 		/* FALLTHROUGH */
3797 		return true;
3798 	}
3799 
3800 	return false;
3801 }
3802 
3803 static bool bnxt_dir_type_is_executable(uint16_t dir_type)
3804 {
3805 	return bnxt_dir_type_is_ape_bin_format(dir_type) ||
3806 		bnxt_dir_type_is_other_exec_format(dir_type);
3807 }
3808 
3809 static int
3810 bnxt_set_eeprom_op(struct rte_eth_dev *dev,
3811 		struct rte_dev_eeprom_info *in_eeprom)
3812 {
3813 	struct bnxt *bp = dev->data->dev_private;
3814 	uint8_t index, dir_op;
3815 	uint16_t type, ext, ordinal, attr;
3816 	int rc;
3817 
3818 	rc = is_bnxt_in_error(bp);
3819 	if (rc)
3820 		return rc;
3821 
3822 	PMD_DRV_LOG(INFO, PCI_PRI_FMT " in_eeprom->offset = %d len = %d\n",
3823 		    bp->pdev->addr.domain, bp->pdev->addr.bus,
3824 		    bp->pdev->addr.devid, bp->pdev->addr.function,
3825 		    in_eeprom->offset, in_eeprom->length);
3826 
3827 	if (!BNXT_PF(bp)) {
3828 		PMD_DRV_LOG(ERR, "NVM write not supported from a VF\n");
3829 		return -EINVAL;
3830 	}
3831 
3832 	type = in_eeprom->magic >> 16;
3833 
3834 	if (type == 0xffff) { /* special value for directory operations */
3835 		index = in_eeprom->magic & 0xff;
3836 		dir_op = in_eeprom->magic >> 8;
3837 		if (index == 0)
3838 			return -EINVAL;
3839 		switch (dir_op) {
3840 		case 0x0e: /* erase */
3841 			if (in_eeprom->offset != ~in_eeprom->magic)
3842 				return -EINVAL;
3843 			return bnxt_hwrm_erase_nvram_directory(bp, index - 1);
3844 		default:
3845 			return -EINVAL;
3846 		}
3847 	}
3848 
3849 	/* Create or re-write an NVM item: */
3850 	if (bnxt_dir_type_is_executable(type) == true)
3851 		return -EOPNOTSUPP;
3852 	ext = in_eeprom->magic & 0xffff;
3853 	ordinal = in_eeprom->offset >> 16;
3854 	attr = in_eeprom->offset & 0xffff;
3855 
3856 	return bnxt_hwrm_flash_nvram(bp, type, ordinal, ext, attr,
3857 				     in_eeprom->data, in_eeprom->length);
3858 }
3859 
3860 static int bnxt_get_module_info(struct rte_eth_dev *dev,
3861 				struct rte_eth_dev_module_info *modinfo)
3862 {
3863 	uint8_t module_info[SFF_DIAG_SUPPORT_OFFSET + 1];
3864 	struct bnxt *bp = dev->data->dev_private;
3865 	int rc;
3866 
3867 	/* No point in going further if phy status indicates
3868 	 * module is not inserted or if it is powered down or
3869 	 * if it is of type 10GBase-T
3870 	 */
3871 	if (bp->link_info->module_status >
3872 	    HWRM_PORT_PHY_QCFG_OUTPUT_MODULE_STATUS_WARNINGMSG) {
3873 		PMD_DRV_LOG(NOTICE, "Port %u : Module is not inserted or is powered down\n",
3874 			    dev->data->port_id);
3875 		return -ENOTSUP;
3876 	}
3877 
3878 	/* This feature is not supported in older firmware versions */
3879 	if (bp->hwrm_spec_code < 0x10202) {
3880 		PMD_DRV_LOG(NOTICE, "Port %u : Feature is not supported in older firmware\n",
3881 			    dev->data->port_id);
3882 		return -ENOTSUP;
3883 	}
3884 
3885 	rc = bnxt_hwrm_read_sfp_module_eeprom_info(bp, I2C_DEV_ADDR_A0, 0, 0,
3886 						   SFF_DIAG_SUPPORT_OFFSET + 1,
3887 						   module_info);
3888 
3889 	if (rc)
3890 		return rc;
3891 
3892 	switch (module_info[0]) {
3893 	case SFF_MODULE_ID_SFP:
3894 		modinfo->type = RTE_ETH_MODULE_SFF_8472;
3895 		modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8472_LEN;
3896 		if (module_info[SFF_DIAG_SUPPORT_OFFSET] == 0)
3897 			modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8436_LEN;
3898 		break;
3899 	case SFF_MODULE_ID_QSFP:
3900 	case SFF_MODULE_ID_QSFP_PLUS:
3901 		modinfo->type = RTE_ETH_MODULE_SFF_8436;
3902 		modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8436_LEN;
3903 		break;
3904 	case SFF_MODULE_ID_QSFP28:
3905 		modinfo->type = RTE_ETH_MODULE_SFF_8636;
3906 		modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8636_MAX_LEN;
3907 		if (module_info[SFF8636_FLATMEM_OFFSET] & SFF8636_FLATMEM_MASK)
3908 			modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8636_LEN;
3909 		break;
3910 	default:
3911 		PMD_DRV_LOG(NOTICE, "Port %u : Unsupported module\n", dev->data->port_id);
3912 		return -ENOTSUP;
3913 	}
3914 
3915 	PMD_DRV_LOG(INFO, "Port %u : modinfo->type = %d modinfo->eeprom_len = %d\n",
3916 		    dev->data->port_id, modinfo->type, modinfo->eeprom_len);
3917 
3918 	return 0;
3919 }
3920 
3921 static int bnxt_get_module_eeprom(struct rte_eth_dev *dev,
3922 				  struct rte_dev_eeprom_info *info)
3923 {
3924 	uint8_t pg_addr[5] = { I2C_DEV_ADDR_A0, I2C_DEV_ADDR_A0 };
3925 	uint32_t offset = info->offset, length = info->length;
3926 	uint8_t module_info[SFF_DIAG_SUPPORT_OFFSET + 1];
3927 	struct bnxt *bp = dev->data->dev_private;
3928 	uint8_t *data = info->data;
3929 	uint8_t page = offset >> 7;
3930 	uint8_t max_pages = 2;
3931 	uint8_t opt_pages;
3932 	int rc;
3933 
3934 	rc = bnxt_hwrm_read_sfp_module_eeprom_info(bp, I2C_DEV_ADDR_A0, 0, 0,
3935 						   SFF_DIAG_SUPPORT_OFFSET + 1,
3936 						   module_info);
3937 	if (rc)
3938 		return rc;
3939 
3940 	switch (module_info[0]) {
3941 	case SFF_MODULE_ID_SFP:
3942 		module_info[SFF_DIAG_SUPPORT_OFFSET] = 0;
3943 		if (module_info[SFF_DIAG_SUPPORT_OFFSET]) {
3944 			pg_addr[2] = I2C_DEV_ADDR_A2;
3945 			pg_addr[3] = I2C_DEV_ADDR_A2;
3946 			max_pages = 4;
3947 		}
3948 		break;
3949 	case SFF_MODULE_ID_QSFP28:
3950 		rc = bnxt_hwrm_read_sfp_module_eeprom_info(bp, I2C_DEV_ADDR_A0, 0,
3951 							   SFF8636_OPT_PAGES_OFFSET,
3952 							   1, &opt_pages);
3953 		if (rc)
3954 			return rc;
3955 
3956 		if (opt_pages & SFF8636_PAGE1_MASK) {
3957 			pg_addr[2] = I2C_DEV_ADDR_A0;
3958 			max_pages = 3;
3959 		}
3960 		if (opt_pages & SFF8636_PAGE2_MASK) {
3961 			pg_addr[3] = I2C_DEV_ADDR_A0;
3962 			max_pages = 4;
3963 		}
3964 		if (~module_info[SFF8636_FLATMEM_OFFSET] & SFF8636_FLATMEM_MASK) {
3965 			pg_addr[4] = I2C_DEV_ADDR_A0;
3966 			max_pages = 5;
3967 		}
3968 		break;
3969 	default:
3970 		break;
3971 	}
3972 
3973 	memset(data, 0, length);
3974 
3975 	offset &= 0xff;
3976 	while (length && page < max_pages) {
3977 		uint8_t raw_page = page ? page - 1 : 0;
3978 		uint16_t chunk;
3979 
3980 		if (pg_addr[page] == I2C_DEV_ADDR_A2)
3981 			raw_page = 0;
3982 		else if (page)
3983 			offset |= 0x80;
3984 		chunk = RTE_MIN(length, 256 - offset);
3985 
3986 		if (pg_addr[page]) {
3987 			rc = bnxt_hwrm_read_sfp_module_eeprom_info(bp, pg_addr[page],
3988 								   raw_page, offset,
3989 								   chunk, data);
3990 			if (rc)
3991 				return rc;
3992 		}
3993 
3994 		data += chunk;
3995 		length -= chunk;
3996 		offset = 0;
3997 		page += 1 + (chunk > 128);
3998 	}
3999 
4000 	return length ? -EINVAL : 0;
4001 }
4002 
4003 /*
4004  * Initialization
4005  */
4006 
4007 static const struct eth_dev_ops bnxt_dev_ops = {
4008 	.dev_infos_get = bnxt_dev_info_get_op,
4009 	.dev_close = bnxt_dev_close_op,
4010 	.dev_configure = bnxt_dev_configure_op,
4011 	.dev_start = bnxt_dev_start_op,
4012 	.dev_stop = bnxt_dev_stop_op,
4013 	.dev_set_link_up = bnxt_dev_set_link_up_op,
4014 	.dev_set_link_down = bnxt_dev_set_link_down_op,
4015 	.stats_get = bnxt_stats_get_op,
4016 	.stats_reset = bnxt_stats_reset_op,
4017 	.rx_queue_setup = bnxt_rx_queue_setup_op,
4018 	.rx_queue_release = bnxt_rx_queue_release_op,
4019 	.tx_queue_setup = bnxt_tx_queue_setup_op,
4020 	.tx_queue_release = bnxt_tx_queue_release_op,
4021 	.rx_queue_intr_enable = bnxt_rx_queue_intr_enable_op,
4022 	.rx_queue_intr_disable = bnxt_rx_queue_intr_disable_op,
4023 	.reta_update = bnxt_reta_update_op,
4024 	.reta_query = bnxt_reta_query_op,
4025 	.rss_hash_update = bnxt_rss_hash_update_op,
4026 	.rss_hash_conf_get = bnxt_rss_hash_conf_get_op,
4027 	.link_update = bnxt_link_update_op,
4028 	.promiscuous_enable = bnxt_promiscuous_enable_op,
4029 	.promiscuous_disable = bnxt_promiscuous_disable_op,
4030 	.allmulticast_enable = bnxt_allmulticast_enable_op,
4031 	.allmulticast_disable = bnxt_allmulticast_disable_op,
4032 	.mac_addr_add = bnxt_mac_addr_add_op,
4033 	.mac_addr_remove = bnxt_mac_addr_remove_op,
4034 	.flow_ctrl_get = bnxt_flow_ctrl_get_op,
4035 	.flow_ctrl_set = bnxt_flow_ctrl_set_op,
4036 	.udp_tunnel_port_add  = bnxt_udp_tunnel_port_add_op,
4037 	.udp_tunnel_port_del  = bnxt_udp_tunnel_port_del_op,
4038 	.vlan_filter_set = bnxt_vlan_filter_set_op,
4039 	.vlan_offload_set = bnxt_vlan_offload_set_op,
4040 	.vlan_tpid_set = bnxt_vlan_tpid_set_op,
4041 	.vlan_pvid_set = bnxt_vlan_pvid_set_op,
4042 	.mtu_set = bnxt_mtu_set_op,
4043 	.mac_addr_set = bnxt_set_default_mac_addr_op,
4044 	.xstats_get = bnxt_dev_xstats_get_op,
4045 	.xstats_get_names = bnxt_dev_xstats_get_names_op,
4046 	.xstats_reset = bnxt_dev_xstats_reset_op,
4047 	.fw_version_get = bnxt_fw_version_get,
4048 	.set_mc_addr_list = bnxt_dev_set_mc_addr_list_op,
4049 	.rxq_info_get = bnxt_rxq_info_get_op,
4050 	.txq_info_get = bnxt_txq_info_get_op,
4051 	.rx_burst_mode_get = bnxt_rx_burst_mode_get,
4052 	.tx_burst_mode_get = bnxt_tx_burst_mode_get,
4053 	.dev_led_on = bnxt_dev_led_on_op,
4054 	.dev_led_off = bnxt_dev_led_off_op,
4055 	.rx_queue_start = bnxt_rx_queue_start,
4056 	.rx_queue_stop = bnxt_rx_queue_stop,
4057 	.tx_queue_start = bnxt_tx_queue_start,
4058 	.tx_queue_stop = bnxt_tx_queue_stop,
4059 	.flow_ops_get = bnxt_flow_ops_get_op,
4060 	.dev_supported_ptypes_get = bnxt_dev_supported_ptypes_get_op,
4061 	.get_eeprom_length    = bnxt_get_eeprom_length_op,
4062 	.get_eeprom           = bnxt_get_eeprom_op,
4063 	.set_eeprom           = bnxt_set_eeprom_op,
4064 	.get_module_info = bnxt_get_module_info,
4065 	.get_module_eeprom = bnxt_get_module_eeprom,
4066 	.timesync_enable      = bnxt_timesync_enable,
4067 	.timesync_disable     = bnxt_timesync_disable,
4068 	.timesync_read_time   = bnxt_timesync_read_time,
4069 	.timesync_write_time   = bnxt_timesync_write_time,
4070 	.timesync_adjust_time = bnxt_timesync_adjust_time,
4071 	.timesync_read_rx_timestamp = bnxt_timesync_read_rx_timestamp,
4072 	.timesync_read_tx_timestamp = bnxt_timesync_read_tx_timestamp,
4073 };
4074 
4075 static uint32_t bnxt_map_reset_regs(struct bnxt *bp, uint32_t reg)
4076 {
4077 	uint32_t offset;
4078 
4079 	/* Only pre-map the reset GRC registers using window 3 */
4080 	rte_write32(reg & 0xfffff000, (uint8_t *)bp->bar0 +
4081 		    BNXT_GRCPF_REG_WINDOW_BASE_OUT + 8);
4082 
4083 	offset = BNXT_GRCP_WINDOW_3_BASE + (reg & 0xffc);
4084 
4085 	return offset;
4086 }
4087 
4088 int bnxt_map_fw_health_status_regs(struct bnxt *bp)
4089 {
4090 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4091 	uint32_t reg_base = 0xffffffff;
4092 	int i;
4093 
4094 	/* Only pre-map the monitoring GRC registers using window 2 */
4095 	for (i = 0; i < BNXT_FW_STATUS_REG_CNT; i++) {
4096 		uint32_t reg = info->status_regs[i];
4097 
4098 		if (BNXT_FW_STATUS_REG_TYPE(reg) != BNXT_FW_STATUS_REG_TYPE_GRC)
4099 			continue;
4100 
4101 		if (reg_base == 0xffffffff)
4102 			reg_base = reg & 0xfffff000;
4103 		if ((reg & 0xfffff000) != reg_base)
4104 			return -ERANGE;
4105 
4106 		/* Use mask 0xffc as the Lower 2 bits indicates
4107 		 * address space location
4108 		 */
4109 		info->mapped_status_regs[i] = BNXT_GRCP_WINDOW_2_BASE +
4110 						(reg & 0xffc);
4111 	}
4112 
4113 	if (reg_base == 0xffffffff)
4114 		return 0;
4115 
4116 	rte_write32(reg_base, (uint8_t *)bp->bar0 +
4117 		    BNXT_GRCPF_REG_WINDOW_BASE_OUT + 4);
4118 
4119 	return 0;
4120 }
4121 
4122 static void bnxt_write_fw_reset_reg(struct bnxt *bp, uint32_t index)
4123 {
4124 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4125 	uint32_t delay = info->delay_after_reset[index];
4126 	uint32_t val = info->reset_reg_val[index];
4127 	uint32_t reg = info->reset_reg[index];
4128 	uint32_t type, offset;
4129 	int ret;
4130 
4131 	type = BNXT_FW_STATUS_REG_TYPE(reg);
4132 	offset = BNXT_FW_STATUS_REG_OFF(reg);
4133 
4134 	switch (type) {
4135 	case BNXT_FW_STATUS_REG_TYPE_CFG:
4136 		ret = rte_pci_write_config(bp->pdev, &val, sizeof(val), offset);
4137 		if (ret < 0) {
4138 			PMD_DRV_LOG(ERR, "Failed to write %#x at PCI offset %#x",
4139 				    val, offset);
4140 			return;
4141 		}
4142 		break;
4143 	case BNXT_FW_STATUS_REG_TYPE_GRC:
4144 		offset = bnxt_map_reset_regs(bp, offset);
4145 		rte_write32(val, (uint8_t *)bp->bar0 + offset);
4146 		break;
4147 	case BNXT_FW_STATUS_REG_TYPE_BAR0:
4148 		rte_write32(val, (uint8_t *)bp->bar0 + offset);
4149 		break;
4150 	}
4151 	/* wait on a specific interval of time until core reset is complete */
4152 	if (delay)
4153 		rte_delay_ms(delay);
4154 }
4155 
4156 static void bnxt_dev_cleanup(struct bnxt *bp)
4157 {
4158 	bp->eth_dev->data->dev_link.link_status = 0;
4159 	bp->link_info->link_up = 0;
4160 	if (bp->eth_dev->data->dev_started)
4161 		bnxt_dev_stop(bp->eth_dev);
4162 
4163 	bnxt_uninit_resources(bp, true);
4164 }
4165 
4166 static int
4167 bnxt_check_fw_reset_done(struct bnxt *bp)
4168 {
4169 	int timeout = bp->fw_reset_max_msecs;
4170 	uint16_t val = 0;
4171 	int rc;
4172 
4173 	do {
4174 		rc = rte_pci_read_config(bp->pdev, &val, sizeof(val), PCI_SUBSYSTEM_ID_OFFSET);
4175 		if (rc < 0) {
4176 			PMD_DRV_LOG(ERR, "Failed to read PCI offset 0x%x", PCI_SUBSYSTEM_ID_OFFSET);
4177 			return rc;
4178 		}
4179 		if (val != 0xffff)
4180 			break;
4181 		rte_delay_ms(1);
4182 	} while (timeout--);
4183 
4184 	if (val == 0xffff) {
4185 		PMD_DRV_LOG(ERR, "Firmware reset aborted, PCI config space invalid\n");
4186 		return -1;
4187 	}
4188 
4189 	return 0;
4190 }
4191 
4192 static int bnxt_restore_vlan_filters(struct bnxt *bp)
4193 {
4194 	struct rte_eth_dev *dev = bp->eth_dev;
4195 	struct rte_vlan_filter_conf *vfc;
4196 	int vidx, vbit, rc;
4197 	uint16_t vlan_id;
4198 
4199 	for (vlan_id = 1; vlan_id <= RTE_ETHER_MAX_VLAN_ID; vlan_id++) {
4200 		vfc = &dev->data->vlan_filter_conf;
4201 		vidx = vlan_id / 64;
4202 		vbit = vlan_id % 64;
4203 
4204 		/* Each bit corresponds to a VLAN id */
4205 		if (vfc->ids[vidx] & (UINT64_C(1) << vbit)) {
4206 			rc = bnxt_add_vlan_filter(bp, vlan_id);
4207 			if (rc)
4208 				return rc;
4209 		}
4210 	}
4211 
4212 	return 0;
4213 }
4214 
4215 static int bnxt_restore_mac_filters(struct bnxt *bp)
4216 {
4217 	struct rte_eth_dev *dev = bp->eth_dev;
4218 	struct rte_eth_dev_info dev_info;
4219 	struct rte_ether_addr *addr;
4220 	uint64_t pool_mask;
4221 	uint32_t pool = 0;
4222 	uint32_t i;
4223 	int rc;
4224 
4225 	if (BNXT_VF(bp) && !BNXT_VF_IS_TRUSTED(bp))
4226 		return 0;
4227 
4228 	rc = bnxt_dev_info_get_op(dev, &dev_info);
4229 	if (rc)
4230 		return rc;
4231 
4232 	/* replay MAC address configuration */
4233 	for (i = 1; i < dev_info.max_mac_addrs; i++) {
4234 		addr = &dev->data->mac_addrs[i];
4235 
4236 		/* skip zero address */
4237 		if (rte_is_zero_ether_addr(addr))
4238 			continue;
4239 
4240 		pool = 0;
4241 		pool_mask = dev->data->mac_pool_sel[i];
4242 
4243 		do {
4244 			if (pool_mask & 1ULL) {
4245 				rc = bnxt_mac_addr_add_op(dev, addr, i, pool);
4246 				if (rc)
4247 					return rc;
4248 			}
4249 			pool_mask >>= 1;
4250 			pool++;
4251 		} while (pool_mask);
4252 	}
4253 
4254 	return 0;
4255 }
4256 
4257 static int bnxt_restore_filters(struct bnxt *bp)
4258 {
4259 	struct rte_eth_dev *dev = bp->eth_dev;
4260 	int ret = 0;
4261 
4262 	if (dev->data->all_multicast) {
4263 		ret = bnxt_allmulticast_enable_op(dev);
4264 		if (ret)
4265 			return ret;
4266 	}
4267 	if (dev->data->promiscuous) {
4268 		ret = bnxt_promiscuous_enable_op(dev);
4269 		if (ret)
4270 			return ret;
4271 	}
4272 
4273 	ret = bnxt_restore_mac_filters(bp);
4274 	if (ret)
4275 		return ret;
4276 
4277 	ret = bnxt_restore_vlan_filters(bp);
4278 	/* TODO restore other filters as well */
4279 	return ret;
4280 }
4281 
4282 static int bnxt_check_fw_ready(struct bnxt *bp)
4283 {
4284 	int timeout = bp->fw_reset_max_msecs;
4285 	int rc = 0;
4286 
4287 	do {
4288 		rc = bnxt_hwrm_poll_ver_get(bp);
4289 		if (rc == 0)
4290 			break;
4291 		rte_delay_ms(BNXT_FW_READY_WAIT_INTERVAL);
4292 		timeout -= BNXT_FW_READY_WAIT_INTERVAL;
4293 	} while (rc && timeout > 0);
4294 
4295 	if (rc)
4296 		PMD_DRV_LOG(ERR, "FW is not Ready after reset\n");
4297 
4298 	return rc;
4299 }
4300 
4301 static void bnxt_dev_recover(void *arg)
4302 {
4303 	struct bnxt *bp = arg;
4304 	int rc = 0;
4305 
4306 	pthread_mutex_lock(&bp->err_recovery_lock);
4307 
4308 	if (!bp->fw_reset_min_msecs) {
4309 		rc = bnxt_check_fw_reset_done(bp);
4310 		if (rc)
4311 			goto err;
4312 	}
4313 
4314 	/* Clear Error flag so that device re-init should happen */
4315 	bp->flags &= ~BNXT_FLAG_FATAL_ERROR;
4316 
4317 	rc = bnxt_check_fw_ready(bp);
4318 	if (rc)
4319 		goto err;
4320 
4321 	rc = bnxt_init_resources(bp, true);
4322 	if (rc) {
4323 		PMD_DRV_LOG(ERR,
4324 			    "Failed to initialize resources after reset\n");
4325 		goto err;
4326 	}
4327 	/* clear reset flag as the device is initialized now */
4328 	bp->flags &= ~BNXT_FLAG_FW_RESET;
4329 
4330 	rc = bnxt_dev_start_op(bp->eth_dev);
4331 	if (rc) {
4332 		PMD_DRV_LOG(ERR, "Failed to start port after reset\n");
4333 		goto err_start;
4334 	}
4335 
4336 	rc = bnxt_restore_filters(bp);
4337 	if (rc)
4338 		goto err_start;
4339 
4340 	PMD_DRV_LOG(INFO, "Recovered from FW reset\n");
4341 	pthread_mutex_unlock(&bp->err_recovery_lock);
4342 
4343 	return;
4344 err_start:
4345 	bnxt_dev_stop(bp->eth_dev);
4346 err:
4347 	bp->flags |= BNXT_FLAG_FATAL_ERROR;
4348 	bnxt_uninit_resources(bp, false);
4349 	pthread_mutex_unlock(&bp->err_recovery_lock);
4350 	PMD_DRV_LOG(ERR, "Failed to recover from FW reset\n");
4351 }
4352 
4353 void bnxt_dev_reset_and_resume(void *arg)
4354 {
4355 	struct bnxt *bp = arg;
4356 	uint32_t us = US_PER_MS * bp->fw_reset_min_msecs;
4357 	uint16_t val = 0;
4358 	int rc;
4359 
4360 	bnxt_dev_cleanup(bp);
4361 
4362 	bnxt_wait_for_device_shutdown(bp);
4363 
4364 	/* During some fatal firmware error conditions, the PCI config space
4365 	 * register 0x2e which normally contains the subsystem ID will become
4366 	 * 0xffff. This register will revert back to the normal value after
4367 	 * the chip has completed core reset. If we detect this condition,
4368 	 * we can poll this config register immediately for the value to revert.
4369 	 */
4370 	if (bp->flags & BNXT_FLAG_FATAL_ERROR) {
4371 		rc = rte_pci_read_config(bp->pdev, &val, sizeof(val), PCI_SUBSYSTEM_ID_OFFSET);
4372 		if (rc < 0) {
4373 			PMD_DRV_LOG(ERR, "Failed to read PCI offset 0x%x", PCI_SUBSYSTEM_ID_OFFSET);
4374 			return;
4375 		}
4376 		if (val == 0xffff) {
4377 			bp->fw_reset_min_msecs = 0;
4378 			us = 1;
4379 		}
4380 	}
4381 
4382 	rc = rte_eal_alarm_set(us, bnxt_dev_recover, (void *)bp);
4383 	if (rc)
4384 		PMD_DRV_LOG(ERR, "Error setting recovery alarm");
4385 }
4386 
4387 uint32_t bnxt_read_fw_status_reg(struct bnxt *bp, uint32_t index)
4388 {
4389 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4390 	uint32_t reg = info->status_regs[index];
4391 	uint32_t type, offset, val = 0;
4392 	int ret = 0;
4393 
4394 	type = BNXT_FW_STATUS_REG_TYPE(reg);
4395 	offset = BNXT_FW_STATUS_REG_OFF(reg);
4396 
4397 	switch (type) {
4398 	case BNXT_FW_STATUS_REG_TYPE_CFG:
4399 		ret = rte_pci_read_config(bp->pdev, &val, sizeof(val), offset);
4400 		if (ret < 0)
4401 			PMD_DRV_LOG(ERR, "Failed to read PCI offset %#x",
4402 				    offset);
4403 		break;
4404 	case BNXT_FW_STATUS_REG_TYPE_GRC:
4405 		offset = info->mapped_status_regs[index];
4406 		/* FALLTHROUGH */
4407 	case BNXT_FW_STATUS_REG_TYPE_BAR0:
4408 		val = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
4409 				       offset));
4410 		break;
4411 	}
4412 
4413 	return val;
4414 }
4415 
4416 static int bnxt_fw_reset_all(struct bnxt *bp)
4417 {
4418 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4419 	uint32_t i;
4420 	int rc = 0;
4421 
4422 	if (info->flags & BNXT_FLAG_ERROR_RECOVERY_HOST) {
4423 		/* Reset through master function driver */
4424 		for (i = 0; i < info->reg_array_cnt; i++)
4425 			bnxt_write_fw_reset_reg(bp, i);
4426 		/* Wait for time specified by FW after triggering reset */
4427 		rte_delay_ms(info->master_func_wait_period_after_reset);
4428 	} else if (info->flags & BNXT_FLAG_ERROR_RECOVERY_CO_CPU) {
4429 		/* Reset with the help of Kong processor */
4430 		rc = bnxt_hwrm_fw_reset(bp);
4431 		if (rc)
4432 			PMD_DRV_LOG(ERR, "Failed to reset FW\n");
4433 	}
4434 
4435 	return rc;
4436 }
4437 
4438 static void bnxt_fw_reset_cb(void *arg)
4439 {
4440 	struct bnxt *bp = arg;
4441 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4442 	int rc = 0;
4443 
4444 	/* Only Master function can do FW reset */
4445 	if (bnxt_is_master_func(bp) &&
4446 	    bnxt_is_recovery_enabled(bp)) {
4447 		rc = bnxt_fw_reset_all(bp);
4448 		if (rc) {
4449 			PMD_DRV_LOG(ERR, "Adapter recovery failed\n");
4450 			return;
4451 		}
4452 	}
4453 
4454 	/* if recovery method is ERROR_RECOVERY_CO_CPU, KONG will send
4455 	 * EXCEPTION_FATAL_ASYNC event to all the functions
4456 	 * (including MASTER FUNC). After receiving this Async, all the active
4457 	 * drivers should treat this case as FW initiated recovery
4458 	 */
4459 	if (info->flags & BNXT_FLAG_ERROR_RECOVERY_HOST) {
4460 		bp->fw_reset_min_msecs = BNXT_MIN_FW_READY_TIMEOUT;
4461 		bp->fw_reset_max_msecs = BNXT_MAX_FW_RESET_TIMEOUT;
4462 
4463 		/* To recover from error */
4464 		rte_eal_alarm_set(US_PER_MS, bnxt_dev_reset_and_resume,
4465 				  (void *)bp);
4466 	}
4467 }
4468 
4469 /* Driver should poll FW heartbeat, reset_counter with the frequency
4470  * advertised by FW in HWRM_ERROR_RECOVERY_QCFG.
4471  * When the driver detects heartbeat stop or change in reset_counter,
4472  * it has to trigger a reset to recover from the error condition.
4473  * A “master PF” is the function who will have the privilege to
4474  * initiate the chimp reset. The master PF will be elected by the
4475  * firmware and will be notified through async message.
4476  */
4477 static void bnxt_check_fw_health(void *arg)
4478 {
4479 	struct bnxt *bp = arg;
4480 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4481 	uint32_t val = 0, wait_msec;
4482 
4483 	if (!info || !bnxt_is_recovery_enabled(bp) ||
4484 	    is_bnxt_in_error(bp))
4485 		return;
4486 
4487 	val = bnxt_read_fw_status_reg(bp, BNXT_FW_HEARTBEAT_CNT_REG);
4488 	if (val == info->last_heart_beat)
4489 		goto reset;
4490 
4491 	info->last_heart_beat = val;
4492 
4493 	val = bnxt_read_fw_status_reg(bp, BNXT_FW_RECOVERY_CNT_REG);
4494 	if (val != info->last_reset_counter)
4495 		goto reset;
4496 
4497 	info->last_reset_counter = val;
4498 
4499 	rte_eal_alarm_set(US_PER_MS * info->driver_polling_freq,
4500 			  bnxt_check_fw_health, (void *)bp);
4501 
4502 	return;
4503 reset:
4504 	/* Stop DMA to/from device */
4505 	bp->flags |= BNXT_FLAG_FATAL_ERROR;
4506 	bp->flags |= BNXT_FLAG_FW_RESET;
4507 
4508 	bnxt_stop_rxtx(bp);
4509 
4510 	PMD_DRV_LOG(ERR, "Detected FW dead condition\n");
4511 
4512 	if (bnxt_is_master_func(bp))
4513 		wait_msec = info->master_func_wait_period;
4514 	else
4515 		wait_msec = info->normal_func_wait_period;
4516 
4517 	rte_eal_alarm_set(US_PER_MS * wait_msec,
4518 			  bnxt_fw_reset_cb, (void *)bp);
4519 }
4520 
4521 void bnxt_schedule_fw_health_check(struct bnxt *bp)
4522 {
4523 	uint32_t polling_freq;
4524 
4525 	pthread_mutex_lock(&bp->health_check_lock);
4526 
4527 	if (!bnxt_is_recovery_enabled(bp))
4528 		goto done;
4529 
4530 	if (bp->flags & BNXT_FLAG_FW_HEALTH_CHECK_SCHEDULED)
4531 		goto done;
4532 
4533 	polling_freq = bp->recovery_info->driver_polling_freq;
4534 
4535 	rte_eal_alarm_set(US_PER_MS * polling_freq,
4536 			  bnxt_check_fw_health, (void *)bp);
4537 	bp->flags |= BNXT_FLAG_FW_HEALTH_CHECK_SCHEDULED;
4538 
4539 done:
4540 	pthread_mutex_unlock(&bp->health_check_lock);
4541 }
4542 
4543 static void bnxt_cancel_fw_health_check(struct bnxt *bp)
4544 {
4545 	rte_eal_alarm_cancel(bnxt_check_fw_health, (void *)bp);
4546 	bp->flags &= ~BNXT_FLAG_FW_HEALTH_CHECK_SCHEDULED;
4547 }
4548 
4549 static bool bnxt_vf_pciid(uint16_t device_id)
4550 {
4551 	switch (device_id) {
4552 	case BROADCOM_DEV_ID_57304_VF:
4553 	case BROADCOM_DEV_ID_57406_VF:
4554 	case BROADCOM_DEV_ID_5731X_VF:
4555 	case BROADCOM_DEV_ID_5741X_VF:
4556 	case BROADCOM_DEV_ID_57414_VF:
4557 	case BROADCOM_DEV_ID_STRATUS_NIC_VF1:
4558 	case BROADCOM_DEV_ID_STRATUS_NIC_VF2:
4559 	case BROADCOM_DEV_ID_58802_VF:
4560 	case BROADCOM_DEV_ID_57500_VF1:
4561 	case BROADCOM_DEV_ID_57500_VF2:
4562 	case BROADCOM_DEV_ID_58818_VF:
4563 		/* FALLTHROUGH */
4564 		return true;
4565 	default:
4566 		return false;
4567 	}
4568 }
4569 
4570 /* Phase 5 device */
4571 static bool bnxt_p5_device(uint16_t device_id)
4572 {
4573 	switch (device_id) {
4574 	case BROADCOM_DEV_ID_57508:
4575 	case BROADCOM_DEV_ID_57504:
4576 	case BROADCOM_DEV_ID_57502:
4577 	case BROADCOM_DEV_ID_57508_MF1:
4578 	case BROADCOM_DEV_ID_57504_MF1:
4579 	case BROADCOM_DEV_ID_57502_MF1:
4580 	case BROADCOM_DEV_ID_57508_MF2:
4581 	case BROADCOM_DEV_ID_57504_MF2:
4582 	case BROADCOM_DEV_ID_57502_MF2:
4583 	case BROADCOM_DEV_ID_57500_VF1:
4584 	case BROADCOM_DEV_ID_57500_VF2:
4585 	case BROADCOM_DEV_ID_58812:
4586 	case BROADCOM_DEV_ID_58814:
4587 	case BROADCOM_DEV_ID_58818:
4588 	case BROADCOM_DEV_ID_58818_VF:
4589 		/* FALLTHROUGH */
4590 		return true;
4591 	default:
4592 		return false;
4593 	}
4594 }
4595 
4596 bool bnxt_stratus_device(struct bnxt *bp)
4597 {
4598 	uint16_t device_id = bp->pdev->id.device_id;
4599 
4600 	switch (device_id) {
4601 	case BROADCOM_DEV_ID_STRATUS_NIC:
4602 	case BROADCOM_DEV_ID_STRATUS_NIC_VF1:
4603 	case BROADCOM_DEV_ID_STRATUS_NIC_VF2:
4604 		/* FALLTHROUGH */
4605 		return true;
4606 	default:
4607 		return false;
4608 	}
4609 }
4610 
4611 static int bnxt_map_pci_bars(struct rte_eth_dev *eth_dev)
4612 {
4613 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
4614 	struct bnxt *bp = eth_dev->data->dev_private;
4615 
4616 	/* enable device (incl. PCI PM wakeup), and bus-mastering */
4617 	bp->bar0 = (void *)pci_dev->mem_resource[0].addr;
4618 	bp->doorbell_base = (void *)pci_dev->mem_resource[2].addr;
4619 	if (!bp->bar0 || !bp->doorbell_base) {
4620 		PMD_DRV_LOG(ERR, "Unable to access Hardware\n");
4621 		return -ENODEV;
4622 	}
4623 
4624 	bp->eth_dev = eth_dev;
4625 	bp->pdev = pci_dev;
4626 
4627 	return 0;
4628 }
4629 
4630 static int bnxt_alloc_ctx_mem_blk(struct bnxt *bp,
4631 				  struct bnxt_ctx_pg_info *ctx_pg,
4632 				  uint32_t mem_size,
4633 				  const char *suffix,
4634 				  uint16_t idx)
4635 {
4636 	struct bnxt_ring_mem_info *rmem = &ctx_pg->ring_mem;
4637 	const struct rte_memzone *mz = NULL;
4638 	char mz_name[RTE_MEMZONE_NAMESIZE];
4639 	rte_iova_t mz_phys_addr;
4640 	uint64_t valid_bits = 0;
4641 	uint32_t sz;
4642 	int i;
4643 
4644 	if (!mem_size)
4645 		return 0;
4646 
4647 	rmem->nr_pages = RTE_ALIGN_MUL_CEIL(mem_size, BNXT_PAGE_SIZE) /
4648 			 BNXT_PAGE_SIZE;
4649 	rmem->page_size = BNXT_PAGE_SIZE;
4650 	rmem->pg_arr = ctx_pg->ctx_pg_arr;
4651 	rmem->dma_arr = ctx_pg->ctx_dma_arr;
4652 	rmem->flags = BNXT_RMEM_VALID_PTE_FLAG;
4653 
4654 	valid_bits = PTU_PTE_VALID;
4655 
4656 	if (rmem->nr_pages > 1) {
4657 		snprintf(mz_name, RTE_MEMZONE_NAMESIZE,
4658 			 "bnxt_ctx_pg_tbl%s_%x_%d",
4659 			 suffix, idx, bp->eth_dev->data->port_id);
4660 		mz_name[RTE_MEMZONE_NAMESIZE - 1] = 0;
4661 		mz = rte_memzone_lookup(mz_name);
4662 		if (!mz) {
4663 			mz = rte_memzone_reserve_aligned(mz_name,
4664 						rmem->nr_pages * 8,
4665 						SOCKET_ID_ANY,
4666 						RTE_MEMZONE_2MB |
4667 						RTE_MEMZONE_SIZE_HINT_ONLY |
4668 						RTE_MEMZONE_IOVA_CONTIG,
4669 						BNXT_PAGE_SIZE);
4670 			if (mz == NULL)
4671 				return -ENOMEM;
4672 		}
4673 
4674 		memset(mz->addr, 0, mz->len);
4675 		mz_phys_addr = mz->iova;
4676 
4677 		rmem->pg_tbl = mz->addr;
4678 		rmem->pg_tbl_map = mz_phys_addr;
4679 		rmem->pg_tbl_mz = mz;
4680 	}
4681 
4682 	snprintf(mz_name, RTE_MEMZONE_NAMESIZE, "bnxt_ctx_%s_%x_%d",
4683 		 suffix, idx, bp->eth_dev->data->port_id);
4684 	mz = rte_memzone_lookup(mz_name);
4685 	if (!mz) {
4686 		mz = rte_memzone_reserve_aligned(mz_name,
4687 						 mem_size,
4688 						 SOCKET_ID_ANY,
4689 						 RTE_MEMZONE_1GB |
4690 						 RTE_MEMZONE_SIZE_HINT_ONLY |
4691 						 RTE_MEMZONE_IOVA_CONTIG,
4692 						 BNXT_PAGE_SIZE);
4693 		if (mz == NULL)
4694 			return -ENOMEM;
4695 	}
4696 
4697 	memset(mz->addr, 0, mz->len);
4698 	mz_phys_addr = mz->iova;
4699 
4700 	for (sz = 0, i = 0; sz < mem_size; sz += BNXT_PAGE_SIZE, i++) {
4701 		rmem->pg_arr[i] = ((char *)mz->addr) + sz;
4702 		rmem->dma_arr[i] = mz_phys_addr + sz;
4703 
4704 		if (rmem->nr_pages > 1) {
4705 			if (i == rmem->nr_pages - 2 &&
4706 			    (rmem->flags & BNXT_RMEM_RING_PTE_FLAG))
4707 				valid_bits |= PTU_PTE_NEXT_TO_LAST;
4708 			else if (i == rmem->nr_pages - 1 &&
4709 				 (rmem->flags & BNXT_RMEM_RING_PTE_FLAG))
4710 				valid_bits |= PTU_PTE_LAST;
4711 
4712 			rmem->pg_tbl[i] = rte_cpu_to_le_64(rmem->dma_arr[i] |
4713 							   valid_bits);
4714 		}
4715 	}
4716 
4717 	rmem->mz = mz;
4718 	if (rmem->vmem_size)
4719 		rmem->vmem = (void **)mz->addr;
4720 	rmem->dma_arr[0] = mz_phys_addr;
4721 	return 0;
4722 }
4723 
4724 static void bnxt_free_ctx_mem(struct bnxt *bp)
4725 {
4726 	int i;
4727 
4728 	if (!bp->ctx || !(bp->ctx->flags & BNXT_CTX_FLAG_INITED))
4729 		return;
4730 
4731 	bp->ctx->flags &= ~BNXT_CTX_FLAG_INITED;
4732 	rte_memzone_free(bp->ctx->qp_mem.ring_mem.mz);
4733 	rte_memzone_free(bp->ctx->srq_mem.ring_mem.mz);
4734 	rte_memzone_free(bp->ctx->cq_mem.ring_mem.mz);
4735 	rte_memzone_free(bp->ctx->vnic_mem.ring_mem.mz);
4736 	rte_memzone_free(bp->ctx->stat_mem.ring_mem.mz);
4737 	rte_memzone_free(bp->ctx->qp_mem.ring_mem.pg_tbl_mz);
4738 	rte_memzone_free(bp->ctx->srq_mem.ring_mem.pg_tbl_mz);
4739 	rte_memzone_free(bp->ctx->cq_mem.ring_mem.pg_tbl_mz);
4740 	rte_memzone_free(bp->ctx->vnic_mem.ring_mem.pg_tbl_mz);
4741 	rte_memzone_free(bp->ctx->stat_mem.ring_mem.pg_tbl_mz);
4742 
4743 	for (i = 0; i < bp->ctx->tqm_fp_rings_count + 1; i++) {
4744 		if (bp->ctx->tqm_mem[i])
4745 			rte_memzone_free(bp->ctx->tqm_mem[i]->ring_mem.mz);
4746 	}
4747 
4748 	rte_free(bp->ctx);
4749 	bp->ctx = NULL;
4750 }
4751 
4752 #define bnxt_roundup(x, y)   ((((x) + ((y) - 1)) / (y)) * (y))
4753 
4754 #define min_t(type, x, y) ({                    \
4755 	type __min1 = (x);                      \
4756 	type __min2 = (y);                      \
4757 	__min1 < __min2 ? __min1 : __min2; })
4758 
4759 #define max_t(type, x, y) ({                    \
4760 	type __max1 = (x);                      \
4761 	type __max2 = (y);                      \
4762 	__max1 > __max2 ? __max1 : __max2; })
4763 
4764 #define clamp_t(type, _x, min, max)     min_t(type, max_t(type, _x, min), max)
4765 
4766 int bnxt_alloc_ctx_mem(struct bnxt *bp)
4767 {
4768 	struct bnxt_ctx_pg_info *ctx_pg;
4769 	struct bnxt_ctx_mem_info *ctx;
4770 	uint32_t mem_size, ena, entries;
4771 	uint32_t entries_sp, min;
4772 	int i, rc;
4773 
4774 	rc = bnxt_hwrm_func_backing_store_qcaps(bp);
4775 	if (rc) {
4776 		PMD_DRV_LOG(ERR, "Query context mem capability failed\n");
4777 		return rc;
4778 	}
4779 	ctx = bp->ctx;
4780 	if (!ctx || (ctx->flags & BNXT_CTX_FLAG_INITED))
4781 		return 0;
4782 
4783 	ctx_pg = &ctx->qp_mem;
4784 	ctx_pg->entries = ctx->qp_min_qp1_entries + ctx->qp_max_l2_entries;
4785 	if (ctx->qp_entry_size) {
4786 		mem_size = ctx->qp_entry_size * ctx_pg->entries;
4787 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "qp_mem", 0);
4788 		if (rc)
4789 			return rc;
4790 	}
4791 
4792 	ctx_pg = &ctx->srq_mem;
4793 	ctx_pg->entries = ctx->srq_max_l2_entries;
4794 	if (ctx->srq_entry_size) {
4795 		mem_size = ctx->srq_entry_size * ctx_pg->entries;
4796 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "srq_mem", 0);
4797 		if (rc)
4798 			return rc;
4799 	}
4800 
4801 	ctx_pg = &ctx->cq_mem;
4802 	ctx_pg->entries = ctx->cq_max_l2_entries;
4803 	if (ctx->cq_entry_size) {
4804 		mem_size = ctx->cq_entry_size * ctx_pg->entries;
4805 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "cq_mem", 0);
4806 		if (rc)
4807 			return rc;
4808 	}
4809 
4810 	ctx_pg = &ctx->vnic_mem;
4811 	ctx_pg->entries = ctx->vnic_max_vnic_entries +
4812 		ctx->vnic_max_ring_table_entries;
4813 	if (ctx->vnic_entry_size) {
4814 		mem_size = ctx->vnic_entry_size * ctx_pg->entries;
4815 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "vnic_mem", 0);
4816 		if (rc)
4817 			return rc;
4818 	}
4819 
4820 	ctx_pg = &ctx->stat_mem;
4821 	ctx_pg->entries = ctx->stat_max_entries;
4822 	if (ctx->stat_entry_size) {
4823 		mem_size = ctx->stat_entry_size * ctx_pg->entries;
4824 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "stat_mem", 0);
4825 		if (rc)
4826 			return rc;
4827 	}
4828 
4829 	min = ctx->tqm_min_entries_per_ring;
4830 
4831 	entries_sp = ctx->qp_max_l2_entries +
4832 		     ctx->vnic_max_vnic_entries +
4833 		     2 * ctx->qp_min_qp1_entries + min;
4834 	entries_sp = bnxt_roundup(entries_sp, ctx->tqm_entries_multiple);
4835 
4836 	entries = ctx->qp_max_l2_entries + ctx->qp_min_qp1_entries;
4837 	entries = bnxt_roundup(entries, ctx->tqm_entries_multiple);
4838 	entries = clamp_t(uint32_t, entries, min,
4839 			  ctx->tqm_max_entries_per_ring);
4840 	for (i = 0, ena = 0; i < ctx->tqm_fp_rings_count + 1; i++) {
4841 		/* i=0 is for TQM_SP. i=1 to i=8 applies to RING0 to RING7.
4842 		 * i > 8 is other ext rings.
4843 		 */
4844 		ctx_pg = ctx->tqm_mem[i];
4845 		ctx_pg->entries = i ? entries : entries_sp;
4846 		if (ctx->tqm_entry_size) {
4847 			mem_size = ctx->tqm_entry_size * ctx_pg->entries;
4848 			rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size,
4849 						    "tqm_mem", i);
4850 			if (rc)
4851 				return rc;
4852 		}
4853 		if (i < BNXT_MAX_TQM_LEGACY_RINGS)
4854 			ena |= HWRM_FUNC_BACKING_STORE_CFG_INPUT_ENABLES_TQM_SP << i;
4855 		else
4856 			ena |= HWRM_FUNC_BACKING_STORE_CFG_INPUT_ENABLES_TQM_RING8;
4857 	}
4858 
4859 	ena |= FUNC_BACKING_STORE_CFG_INPUT_DFLT_ENABLES;
4860 	rc = bnxt_hwrm_func_backing_store_cfg(bp, ena);
4861 	if (rc)
4862 		PMD_DRV_LOG(ERR,
4863 			    "Failed to configure context mem: rc = %d\n", rc);
4864 	else
4865 		ctx->flags |= BNXT_CTX_FLAG_INITED;
4866 
4867 	return rc;
4868 }
4869 
4870 static int bnxt_alloc_stats_mem(struct bnxt *bp)
4871 {
4872 	struct rte_pci_device *pci_dev = bp->pdev;
4873 	char mz_name[RTE_MEMZONE_NAMESIZE];
4874 	const struct rte_memzone *mz = NULL;
4875 	uint32_t total_alloc_len;
4876 	rte_iova_t mz_phys_addr;
4877 
4878 	if (pci_dev->id.device_id == BROADCOM_DEV_ID_NS2)
4879 		return 0;
4880 
4881 	snprintf(mz_name, RTE_MEMZONE_NAMESIZE,
4882 		 "bnxt_" PCI_PRI_FMT "-%s", pci_dev->addr.domain,
4883 		 pci_dev->addr.bus, pci_dev->addr.devid,
4884 		 pci_dev->addr.function, "rx_port_stats");
4885 	mz_name[RTE_MEMZONE_NAMESIZE - 1] = 0;
4886 	mz = rte_memzone_lookup(mz_name);
4887 	total_alloc_len =
4888 		RTE_CACHE_LINE_ROUNDUP(sizeof(struct rx_port_stats) +
4889 				       sizeof(struct rx_port_stats_ext) + 512);
4890 	if (!mz) {
4891 		mz = rte_memzone_reserve(mz_name, total_alloc_len,
4892 					 SOCKET_ID_ANY,
4893 					 RTE_MEMZONE_2MB |
4894 					 RTE_MEMZONE_SIZE_HINT_ONLY |
4895 					 RTE_MEMZONE_IOVA_CONTIG);
4896 		if (mz == NULL)
4897 			return -ENOMEM;
4898 	}
4899 	memset(mz->addr, 0, mz->len);
4900 	mz_phys_addr = mz->iova;
4901 
4902 	bp->rx_mem_zone = (const void *)mz;
4903 	bp->hw_rx_port_stats = mz->addr;
4904 	bp->hw_rx_port_stats_map = mz_phys_addr;
4905 
4906 	snprintf(mz_name, RTE_MEMZONE_NAMESIZE,
4907 		 "bnxt_" PCI_PRI_FMT "-%s", pci_dev->addr.domain,
4908 		 pci_dev->addr.bus, pci_dev->addr.devid,
4909 		 pci_dev->addr.function, "tx_port_stats");
4910 	mz_name[RTE_MEMZONE_NAMESIZE - 1] = 0;
4911 	mz = rte_memzone_lookup(mz_name);
4912 	total_alloc_len =
4913 		RTE_CACHE_LINE_ROUNDUP(sizeof(struct tx_port_stats) +
4914 				       sizeof(struct tx_port_stats_ext) + 512);
4915 	if (!mz) {
4916 		mz = rte_memzone_reserve(mz_name,
4917 					 total_alloc_len,
4918 					 SOCKET_ID_ANY,
4919 					 RTE_MEMZONE_2MB |
4920 					 RTE_MEMZONE_SIZE_HINT_ONLY |
4921 					 RTE_MEMZONE_IOVA_CONTIG);
4922 		if (mz == NULL)
4923 			return -ENOMEM;
4924 	}
4925 	memset(mz->addr, 0, mz->len);
4926 	mz_phys_addr = mz->iova;
4927 
4928 	bp->tx_mem_zone = (const void *)mz;
4929 	bp->hw_tx_port_stats = mz->addr;
4930 	bp->hw_tx_port_stats_map = mz_phys_addr;
4931 	bp->flags |= BNXT_FLAG_PORT_STATS;
4932 
4933 	/* Display extended statistics if FW supports it */
4934 	if (bp->hwrm_spec_code < HWRM_SPEC_CODE_1_8_4 ||
4935 	    bp->hwrm_spec_code == HWRM_SPEC_CODE_1_9_0 ||
4936 	    !(bp->flags & BNXT_FLAG_EXT_STATS_SUPPORTED))
4937 		return 0;
4938 
4939 	bp->hw_rx_port_stats_ext = (void *)
4940 		((uint8_t *)bp->hw_rx_port_stats +
4941 		 sizeof(struct rx_port_stats));
4942 	bp->hw_rx_port_stats_ext_map = bp->hw_rx_port_stats_map +
4943 		sizeof(struct rx_port_stats);
4944 	bp->flags |= BNXT_FLAG_EXT_RX_PORT_STATS;
4945 
4946 	if (bp->hwrm_spec_code < HWRM_SPEC_CODE_1_9_2 ||
4947 	    bp->flags & BNXT_FLAG_EXT_STATS_SUPPORTED) {
4948 		bp->hw_tx_port_stats_ext = (void *)
4949 			((uint8_t *)bp->hw_tx_port_stats +
4950 			 sizeof(struct tx_port_stats));
4951 		bp->hw_tx_port_stats_ext_map =
4952 			bp->hw_tx_port_stats_map +
4953 			sizeof(struct tx_port_stats);
4954 		bp->flags |= BNXT_FLAG_EXT_TX_PORT_STATS;
4955 	}
4956 
4957 	return 0;
4958 }
4959 
4960 static int bnxt_setup_mac_addr(struct rte_eth_dev *eth_dev)
4961 {
4962 	struct bnxt *bp = eth_dev->data->dev_private;
4963 	int rc = 0;
4964 
4965 	eth_dev->data->mac_addrs = rte_zmalloc("bnxt_mac_addr_tbl",
4966 					       RTE_ETHER_ADDR_LEN *
4967 					       bp->max_l2_ctx,
4968 					       0);
4969 	if (eth_dev->data->mac_addrs == NULL) {
4970 		PMD_DRV_LOG(ERR, "Failed to alloc MAC addr tbl\n");
4971 		return -ENOMEM;
4972 	}
4973 
4974 	if (!BNXT_HAS_DFLT_MAC_SET(bp)) {
4975 		if (BNXT_PF(bp))
4976 			return -EINVAL;
4977 
4978 		/* Generate a random MAC address, if none was assigned by PF */
4979 		PMD_DRV_LOG(INFO, "VF MAC address not assigned by Host PF\n");
4980 		bnxt_eth_hw_addr_random(bp->mac_addr);
4981 		PMD_DRV_LOG(INFO,
4982 			    "Assign random MAC:%02X:%02X:%02X:%02X:%02X:%02X\n",
4983 			    bp->mac_addr[0], bp->mac_addr[1], bp->mac_addr[2],
4984 			    bp->mac_addr[3], bp->mac_addr[4], bp->mac_addr[5]);
4985 
4986 		rc = bnxt_hwrm_set_mac(bp);
4987 		if (rc)
4988 			return rc;
4989 	}
4990 
4991 	/* Copy the permanent MAC from the FUNC_QCAPS response */
4992 	memcpy(&eth_dev->data->mac_addrs[0], bp->mac_addr, RTE_ETHER_ADDR_LEN);
4993 
4994 	return rc;
4995 }
4996 
4997 static int bnxt_restore_dflt_mac(struct bnxt *bp)
4998 {
4999 	int rc = 0;
5000 
5001 	/* MAC is already configured in FW */
5002 	if (BNXT_HAS_DFLT_MAC_SET(bp))
5003 		return 0;
5004 
5005 	/* Restore the old MAC configured */
5006 	rc = bnxt_hwrm_set_mac(bp);
5007 	if (rc)
5008 		PMD_DRV_LOG(ERR, "Failed to restore MAC address\n");
5009 
5010 	return rc;
5011 }
5012 
5013 static void bnxt_config_vf_req_fwd(struct bnxt *bp)
5014 {
5015 	if (!BNXT_PF(bp))
5016 		return;
5017 
5018 	memset(bp->pf->vf_req_fwd, 0, sizeof(bp->pf->vf_req_fwd));
5019 
5020 	if (!(bp->fw_cap & BNXT_FW_CAP_LINK_ADMIN))
5021 		BNXT_HWRM_CMD_TO_FORWARD(HWRM_PORT_PHY_QCFG);
5022 	BNXT_HWRM_CMD_TO_FORWARD(HWRM_FUNC_CFG);
5023 	BNXT_HWRM_CMD_TO_FORWARD(HWRM_FUNC_VF_CFG);
5024 	BNXT_HWRM_CMD_TO_FORWARD(HWRM_CFA_L2_FILTER_ALLOC);
5025 	BNXT_HWRM_CMD_TO_FORWARD(HWRM_OEM_CMD);
5026 }
5027 
5028 struct bnxt *
5029 bnxt_get_bp(uint16_t port)
5030 {
5031 	struct bnxt *bp;
5032 	struct rte_eth_dev *dev;
5033 
5034 	if (!rte_eth_dev_is_valid_port(port)) {
5035 		PMD_DRV_LOG(ERR, "Invalid port %d\n", port);
5036 		return NULL;
5037 	}
5038 
5039 	dev = &rte_eth_devices[port];
5040 	if (!is_bnxt_supported(dev)) {
5041 		PMD_DRV_LOG(ERR, "Device %d not supported\n", port);
5042 		return NULL;
5043 	}
5044 
5045 	bp = (struct bnxt *)dev->data->dev_private;
5046 	if (!BNXT_TRUFLOW_EN(bp)) {
5047 		PMD_DRV_LOG(ERR, "TRUFLOW not enabled\n");
5048 		return NULL;
5049 	}
5050 
5051 	return bp;
5052 }
5053 
5054 uint16_t
5055 bnxt_get_svif(uint16_t port_id, bool func_svif,
5056 	      enum bnxt_ulp_intf_type type)
5057 {
5058 	struct rte_eth_dev *eth_dev;
5059 	struct bnxt *bp;
5060 
5061 	eth_dev = &rte_eth_devices[port_id];
5062 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
5063 		struct bnxt_representor *vfr = eth_dev->data->dev_private;
5064 		if (!vfr)
5065 			return 0;
5066 
5067 		if (type == BNXT_ULP_INTF_TYPE_VF_REP)
5068 			return vfr->svif;
5069 
5070 		eth_dev = vfr->parent_dev;
5071 	}
5072 
5073 	bp = eth_dev->data->dev_private;
5074 
5075 	return func_svif ? bp->func_svif : bp->port_svif;
5076 }
5077 
5078 void
5079 bnxt_get_iface_mac(uint16_t port, enum bnxt_ulp_intf_type type,
5080 		   uint8_t *mac, uint8_t *parent_mac)
5081 {
5082 	struct rte_eth_dev *eth_dev;
5083 	struct bnxt *bp;
5084 
5085 	if (type != BNXT_ULP_INTF_TYPE_TRUSTED_VF &&
5086 	    type != BNXT_ULP_INTF_TYPE_PF)
5087 		return;
5088 
5089 	eth_dev = &rte_eth_devices[port];
5090 	bp = eth_dev->data->dev_private;
5091 	memcpy(mac, bp->mac_addr, RTE_ETHER_ADDR_LEN);
5092 
5093 	if (type == BNXT_ULP_INTF_TYPE_TRUSTED_VF)
5094 		memcpy(parent_mac, bp->parent->mac_addr, RTE_ETHER_ADDR_LEN);
5095 }
5096 
5097 uint16_t
5098 bnxt_get_parent_vnic_id(uint16_t port, enum bnxt_ulp_intf_type type)
5099 {
5100 	struct rte_eth_dev *eth_dev;
5101 	struct bnxt *bp;
5102 
5103 	if (type != BNXT_ULP_INTF_TYPE_TRUSTED_VF)
5104 		return 0;
5105 
5106 	eth_dev = &rte_eth_devices[port];
5107 	bp = eth_dev->data->dev_private;
5108 
5109 	return bp->parent->vnic;
5110 }
5111 uint16_t
5112 bnxt_get_vnic_id(uint16_t port, enum bnxt_ulp_intf_type type)
5113 {
5114 	struct rte_eth_dev *eth_dev;
5115 	struct bnxt_vnic_info *vnic;
5116 	struct bnxt *bp;
5117 
5118 	eth_dev = &rte_eth_devices[port];
5119 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
5120 		struct bnxt_representor *vfr = eth_dev->data->dev_private;
5121 		if (!vfr)
5122 			return 0;
5123 
5124 		if (type == BNXT_ULP_INTF_TYPE_VF_REP)
5125 			return vfr->dflt_vnic_id;
5126 
5127 		eth_dev = vfr->parent_dev;
5128 	}
5129 
5130 	bp = eth_dev->data->dev_private;
5131 
5132 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
5133 
5134 	return vnic->fw_vnic_id;
5135 }
5136 
5137 uint16_t
5138 bnxt_get_fw_func_id(uint16_t port, enum bnxt_ulp_intf_type type)
5139 {
5140 	struct rte_eth_dev *eth_dev;
5141 	struct bnxt *bp;
5142 
5143 	eth_dev = &rte_eth_devices[port];
5144 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
5145 		struct bnxt_representor *vfr = eth_dev->data->dev_private;
5146 		if (!vfr)
5147 			return 0;
5148 
5149 		if (type == BNXT_ULP_INTF_TYPE_VF_REP)
5150 			return vfr->fw_fid;
5151 
5152 		eth_dev = vfr->parent_dev;
5153 	}
5154 
5155 	bp = eth_dev->data->dev_private;
5156 
5157 	return bp->fw_fid;
5158 }
5159 
5160 enum bnxt_ulp_intf_type
5161 bnxt_get_interface_type(uint16_t port)
5162 {
5163 	struct rte_eth_dev *eth_dev;
5164 	struct bnxt *bp;
5165 
5166 	eth_dev = &rte_eth_devices[port];
5167 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev))
5168 		return BNXT_ULP_INTF_TYPE_VF_REP;
5169 
5170 	bp = eth_dev->data->dev_private;
5171 	if (BNXT_PF(bp))
5172 		return BNXT_ULP_INTF_TYPE_PF;
5173 	else if (BNXT_VF_IS_TRUSTED(bp))
5174 		return BNXT_ULP_INTF_TYPE_TRUSTED_VF;
5175 	else if (BNXT_VF(bp))
5176 		return BNXT_ULP_INTF_TYPE_VF;
5177 
5178 	return BNXT_ULP_INTF_TYPE_INVALID;
5179 }
5180 
5181 uint16_t
5182 bnxt_get_phy_port_id(uint16_t port_id)
5183 {
5184 	struct bnxt_representor *vfr;
5185 	struct rte_eth_dev *eth_dev;
5186 	struct bnxt *bp;
5187 
5188 	eth_dev = &rte_eth_devices[port_id];
5189 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
5190 		vfr = eth_dev->data->dev_private;
5191 		if (!vfr)
5192 			return 0;
5193 
5194 		eth_dev = vfr->parent_dev;
5195 	}
5196 
5197 	bp = eth_dev->data->dev_private;
5198 
5199 	return BNXT_PF(bp) ? bp->pf->port_id : bp->parent->port_id;
5200 }
5201 
5202 uint16_t
5203 bnxt_get_parif(uint16_t port_id, enum bnxt_ulp_intf_type type)
5204 {
5205 	struct rte_eth_dev *eth_dev;
5206 	struct bnxt *bp;
5207 
5208 	eth_dev = &rte_eth_devices[port_id];
5209 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
5210 		struct bnxt_representor *vfr = eth_dev->data->dev_private;
5211 		if (!vfr)
5212 			return 0;
5213 
5214 		if (type == BNXT_ULP_INTF_TYPE_VF_REP)
5215 			return vfr->fw_fid - 1;
5216 
5217 		eth_dev = vfr->parent_dev;
5218 	}
5219 
5220 	bp = eth_dev->data->dev_private;
5221 
5222 	return BNXT_PF(bp) ? bp->fw_fid - 1 : bp->parent->fid - 1;
5223 }
5224 
5225 uint16_t
5226 bnxt_get_vport(uint16_t port_id)
5227 {
5228 	return (1 << bnxt_get_phy_port_id(port_id));
5229 }
5230 
5231 static void bnxt_alloc_error_recovery_info(struct bnxt *bp)
5232 {
5233 	struct bnxt_error_recovery_info *info = bp->recovery_info;
5234 
5235 	if (info) {
5236 		if (!(bp->fw_cap & BNXT_FW_CAP_HCOMM_FW_STATUS))
5237 			memset(info, 0, sizeof(*info));
5238 		return;
5239 	}
5240 
5241 	if (!(bp->fw_cap & BNXT_FW_CAP_ERROR_RECOVERY))
5242 		return;
5243 
5244 	info = rte_zmalloc("bnxt_hwrm_error_recovery_qcfg",
5245 			   sizeof(*info), 0);
5246 	if (!info)
5247 		bp->fw_cap &= ~BNXT_FW_CAP_ERROR_RECOVERY;
5248 
5249 	bp->recovery_info = info;
5250 }
5251 
5252 static void bnxt_check_fw_status(struct bnxt *bp)
5253 {
5254 	uint32_t fw_status;
5255 
5256 	if (!(bp->recovery_info &&
5257 	      (bp->fw_cap & BNXT_FW_CAP_HCOMM_FW_STATUS)))
5258 		return;
5259 
5260 	fw_status = bnxt_read_fw_status_reg(bp, BNXT_FW_STATUS_REG);
5261 	if (fw_status != BNXT_FW_STATUS_HEALTHY)
5262 		PMD_DRV_LOG(ERR, "Firmware not responding, status: %#x\n",
5263 			    fw_status);
5264 }
5265 
5266 static int bnxt_map_hcomm_fw_status_reg(struct bnxt *bp)
5267 {
5268 	struct bnxt_error_recovery_info *info = bp->recovery_info;
5269 	uint32_t status_loc;
5270 	uint32_t sig_ver;
5271 
5272 	rte_write32(HCOMM_STATUS_STRUCT_LOC, (uint8_t *)bp->bar0 +
5273 		    BNXT_GRCPF_REG_WINDOW_BASE_OUT + 4);
5274 	sig_ver = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
5275 				   BNXT_GRCP_WINDOW_2_BASE +
5276 				   offsetof(struct hcomm_status,
5277 					    sig_ver)));
5278 	/* If the signature is absent, then FW does not support this feature */
5279 	if ((sig_ver & HCOMM_STATUS_SIGNATURE_MASK) !=
5280 	    HCOMM_STATUS_SIGNATURE_VAL)
5281 		return 0;
5282 
5283 	if (!info) {
5284 		info = rte_zmalloc("bnxt_hwrm_error_recovery_qcfg",
5285 				   sizeof(*info), 0);
5286 		if (!info)
5287 			return -ENOMEM;
5288 		bp->recovery_info = info;
5289 	} else {
5290 		memset(info, 0, sizeof(*info));
5291 	}
5292 
5293 	status_loc = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
5294 				      BNXT_GRCP_WINDOW_2_BASE +
5295 				      offsetof(struct hcomm_status,
5296 					       fw_status_loc)));
5297 
5298 	/* Only pre-map the FW health status GRC register */
5299 	if (BNXT_FW_STATUS_REG_TYPE(status_loc) != BNXT_FW_STATUS_REG_TYPE_GRC)
5300 		return 0;
5301 
5302 	info->status_regs[BNXT_FW_STATUS_REG] = status_loc;
5303 	info->mapped_status_regs[BNXT_FW_STATUS_REG] =
5304 		BNXT_GRCP_WINDOW_2_BASE + (status_loc & BNXT_GRCP_OFFSET_MASK);
5305 
5306 	rte_write32((status_loc & BNXT_GRCP_BASE_MASK), (uint8_t *)bp->bar0 +
5307 		    BNXT_GRCPF_REG_WINDOW_BASE_OUT + 4);
5308 
5309 	bp->fw_cap |= BNXT_FW_CAP_HCOMM_FW_STATUS;
5310 
5311 	return 0;
5312 }
5313 
5314 /* This function gets the FW version along with the
5315  * capabilities(MAX and current) of the function, vnic,
5316  * error recovery, phy and other chip related info
5317  */
5318 static int bnxt_get_config(struct bnxt *bp)
5319 {
5320 	uint16_t mtu;
5321 	int rc = 0;
5322 
5323 	bp->fw_cap = 0;
5324 
5325 	rc = bnxt_map_hcomm_fw_status_reg(bp);
5326 	if (rc)
5327 		return rc;
5328 
5329 	rc = bnxt_hwrm_ver_get(bp, DFLT_HWRM_CMD_TIMEOUT);
5330 	if (rc) {
5331 		bnxt_check_fw_status(bp);
5332 		return rc;
5333 	}
5334 
5335 	rc = bnxt_hwrm_func_reset(bp);
5336 	if (rc)
5337 		return -EIO;
5338 
5339 	rc = bnxt_hwrm_vnic_qcaps(bp);
5340 	if (rc)
5341 		return rc;
5342 
5343 	rc = bnxt_hwrm_queue_qportcfg(bp);
5344 	if (rc)
5345 		return rc;
5346 
5347 	/* Get the MAX capabilities for this function.
5348 	 * This function also allocates context memory for TQM rings and
5349 	 * informs the firmware about this allocated backing store memory.
5350 	 */
5351 	rc = bnxt_hwrm_func_qcaps(bp);
5352 	if (rc)
5353 		return rc;
5354 
5355 	rc = bnxt_hwrm_func_qcfg(bp, &mtu);
5356 	if (rc)
5357 		return rc;
5358 
5359 	rc = bnxt_hwrm_cfa_adv_flow_mgmt_qcaps(bp);
5360 	if (rc)
5361 		return rc;
5362 
5363 	bnxt_hwrm_port_mac_qcfg(bp);
5364 
5365 	bnxt_hwrm_parent_pf_qcfg(bp);
5366 
5367 	bnxt_hwrm_port_phy_qcaps(bp);
5368 
5369 	bnxt_alloc_error_recovery_info(bp);
5370 	/* Get the adapter error recovery support info */
5371 	rc = bnxt_hwrm_error_recovery_qcfg(bp);
5372 	if (rc)
5373 		bp->fw_cap &= ~BNXT_FW_CAP_ERROR_RECOVERY;
5374 
5375 	bnxt_hwrm_port_led_qcaps(bp);
5376 
5377 	return 0;
5378 }
5379 
5380 static int
5381 bnxt_init_locks(struct bnxt *bp)
5382 {
5383 	int err;
5384 
5385 	err = pthread_mutex_init(&bp->flow_lock, NULL);
5386 	if (err) {
5387 		PMD_DRV_LOG(ERR, "Unable to initialize flow_lock\n");
5388 		return err;
5389 	}
5390 
5391 	err = pthread_mutex_init(&bp->def_cp_lock, NULL);
5392 	if (err) {
5393 		PMD_DRV_LOG(ERR, "Unable to initialize def_cp_lock\n");
5394 		return err;
5395 	}
5396 
5397 	err = pthread_mutex_init(&bp->health_check_lock, NULL);
5398 	if (err) {
5399 		PMD_DRV_LOG(ERR, "Unable to initialize health_check_lock\n");
5400 		return err;
5401 	}
5402 
5403 	err = pthread_mutex_init(&bp->err_recovery_lock, NULL);
5404 	if (err)
5405 		PMD_DRV_LOG(ERR, "Unable to initialize err_recovery_lock\n");
5406 
5407 	return err;
5408 }
5409 
5410 static int bnxt_init_resources(struct bnxt *bp, bool reconfig_dev)
5411 {
5412 	int rc = 0;
5413 
5414 	rc = bnxt_get_config(bp);
5415 	if (rc)
5416 		return rc;
5417 
5418 	if (!reconfig_dev) {
5419 		rc = bnxt_setup_mac_addr(bp->eth_dev);
5420 		if (rc)
5421 			return rc;
5422 	} else {
5423 		rc = bnxt_restore_dflt_mac(bp);
5424 		if (rc)
5425 			return rc;
5426 	}
5427 
5428 	bnxt_config_vf_req_fwd(bp);
5429 
5430 	rc = bnxt_hwrm_func_driver_register(bp);
5431 	if (rc) {
5432 		PMD_DRV_LOG(ERR, "Failed to register driver");
5433 		return -EBUSY;
5434 	}
5435 
5436 	if (BNXT_PF(bp)) {
5437 		if (bp->pdev->max_vfs) {
5438 			rc = bnxt_hwrm_allocate_vfs(bp, bp->pdev->max_vfs);
5439 			if (rc) {
5440 				PMD_DRV_LOG(ERR, "Failed to allocate VFs\n");
5441 				return rc;
5442 			}
5443 		} else {
5444 			rc = bnxt_hwrm_allocate_pf_only(bp);
5445 			if (rc) {
5446 				PMD_DRV_LOG(ERR,
5447 					    "Failed to allocate PF resources");
5448 				return rc;
5449 			}
5450 		}
5451 	}
5452 
5453 	rc = bnxt_alloc_mem(bp, reconfig_dev);
5454 	if (rc)
5455 		return rc;
5456 
5457 	rc = bnxt_setup_int(bp);
5458 	if (rc)
5459 		return rc;
5460 
5461 	rc = bnxt_request_int(bp);
5462 	if (rc)
5463 		return rc;
5464 
5465 	rc = bnxt_init_ctx_mem(bp);
5466 	if (rc) {
5467 		PMD_DRV_LOG(ERR, "Failed to init adv_flow_counters\n");
5468 		return rc;
5469 	}
5470 
5471 	return 0;
5472 }
5473 
5474 static int
5475 bnxt_parse_devarg_accum_stats(__rte_unused const char *key,
5476 			      const char *value, void *opaque_arg)
5477 {
5478 	struct bnxt *bp = opaque_arg;
5479 	unsigned long accum_stats;
5480 	char *end = NULL;
5481 
5482 	if (!value || !opaque_arg) {
5483 		PMD_DRV_LOG(ERR,
5484 			    "Invalid parameter passed to accum-stats devargs.\n");
5485 		return -EINVAL;
5486 	}
5487 
5488 	accum_stats = strtoul(value, &end, 10);
5489 	if (end == NULL || *end != '\0' ||
5490 	    (accum_stats == ULONG_MAX && errno == ERANGE)) {
5491 		PMD_DRV_LOG(ERR,
5492 			    "Invalid parameter passed to accum-stats devargs.\n");
5493 		return -EINVAL;
5494 	}
5495 
5496 	if (BNXT_DEVARG_ACCUM_STATS_INVALID(accum_stats)) {
5497 		PMD_DRV_LOG(ERR,
5498 			    "Invalid value passed to accum-stats devargs.\n");
5499 		return -EINVAL;
5500 	}
5501 
5502 	if (accum_stats) {
5503 		bp->flags2 |= BNXT_FLAGS2_ACCUM_STATS_EN;
5504 		PMD_DRV_LOG(INFO, "Host-based accum-stats feature enabled.\n");
5505 	} else {
5506 		bp->flags2 &= ~BNXT_FLAGS2_ACCUM_STATS_EN;
5507 		PMD_DRV_LOG(INFO, "Host-based accum-stats feature disabled.\n");
5508 	}
5509 
5510 	return 0;
5511 }
5512 
5513 static int
5514 bnxt_parse_devarg_flow_xstat(__rte_unused const char *key,
5515 			     const char *value, void *opaque_arg)
5516 {
5517 	struct bnxt *bp = opaque_arg;
5518 	unsigned long flow_xstat;
5519 	char *end = NULL;
5520 
5521 	if (!value || !opaque_arg) {
5522 		PMD_DRV_LOG(ERR,
5523 			    "Invalid parameter passed to flow_xstat devarg.\n");
5524 		return -EINVAL;
5525 	}
5526 
5527 	flow_xstat = strtoul(value, &end, 10);
5528 	if (end == NULL || *end != '\0' ||
5529 	    (flow_xstat == ULONG_MAX && errno == ERANGE)) {
5530 		PMD_DRV_LOG(ERR,
5531 			    "Invalid parameter passed to flow_xstat devarg.\n");
5532 		return -EINVAL;
5533 	}
5534 
5535 	if (BNXT_DEVARG_FLOW_XSTAT_INVALID(flow_xstat)) {
5536 		PMD_DRV_LOG(ERR,
5537 			    "Invalid value passed to flow_xstat devarg.\n");
5538 		return -EINVAL;
5539 	}
5540 
5541 	bp->flags |= BNXT_FLAG_FLOW_XSTATS_EN;
5542 	if (BNXT_FLOW_XSTATS_EN(bp))
5543 		PMD_DRV_LOG(INFO, "flow_xstat feature enabled.\n");
5544 
5545 	return 0;
5546 }
5547 
5548 static int
5549 bnxt_parse_devarg_max_num_kflows(__rte_unused const char *key,
5550 					const char *value, void *opaque_arg)
5551 {
5552 	struct bnxt *bp = opaque_arg;
5553 	unsigned long max_num_kflows;
5554 	char *end = NULL;
5555 
5556 	if (!value || !opaque_arg) {
5557 		PMD_DRV_LOG(ERR,
5558 			"Invalid parameter passed to max_num_kflows devarg.\n");
5559 		return -EINVAL;
5560 	}
5561 
5562 	max_num_kflows = strtoul(value, &end, 10);
5563 	if (end == NULL || *end != '\0' ||
5564 		(max_num_kflows == ULONG_MAX && errno == ERANGE)) {
5565 		PMD_DRV_LOG(ERR,
5566 			"Invalid parameter passed to max_num_kflows devarg.\n");
5567 		return -EINVAL;
5568 	}
5569 
5570 	if (bnxt_devarg_max_num_kflow_invalid(max_num_kflows)) {
5571 		PMD_DRV_LOG(ERR,
5572 			"Invalid value passed to max_num_kflows devarg.\n");
5573 		return -EINVAL;
5574 	}
5575 
5576 	bp->max_num_kflows = max_num_kflows;
5577 	if (bp->max_num_kflows)
5578 		PMD_DRV_LOG(INFO, "max_num_kflows set as %ldK.\n",
5579 				max_num_kflows);
5580 
5581 	return 0;
5582 }
5583 
5584 static int
5585 bnxt_parse_devarg_app_id(__rte_unused const char *key,
5586 				 const char *value, void *opaque_arg)
5587 {
5588 	struct bnxt *bp = opaque_arg;
5589 	unsigned long app_id;
5590 	char *end = NULL;
5591 
5592 	if (!value || !opaque_arg) {
5593 		PMD_DRV_LOG(ERR,
5594 			    "Invalid parameter passed to app-id "
5595 			    "devargs.\n");
5596 		return -EINVAL;
5597 	}
5598 
5599 	app_id = strtoul(value, &end, 10);
5600 	if (end == NULL || *end != '\0' ||
5601 	    (app_id == ULONG_MAX && errno == ERANGE)) {
5602 		PMD_DRV_LOG(ERR,
5603 			    "Invalid parameter passed to app_id "
5604 			    "devargs.\n");
5605 		return -EINVAL;
5606 	}
5607 
5608 	if (BNXT_DEVARG_APP_ID_INVALID(app_id)) {
5609 		PMD_DRV_LOG(ERR, "Invalid app-id(%d) devargs.\n",
5610 			    (uint16_t)app_id);
5611 		return -EINVAL;
5612 	}
5613 
5614 	bp->app_id = app_id;
5615 	PMD_DRV_LOG(INFO, "app-id=%d feature enabled.\n", (uint16_t)app_id);
5616 
5617 	return 0;
5618 }
5619 
5620 static int
5621 bnxt_parse_devarg_rep_is_pf(__rte_unused const char *key,
5622 			    const char *value, void *opaque_arg)
5623 {
5624 	struct bnxt_representor *vfr_bp = opaque_arg;
5625 	unsigned long rep_is_pf;
5626 	char *end = NULL;
5627 
5628 	if (!value || !opaque_arg) {
5629 		PMD_DRV_LOG(ERR,
5630 			    "Invalid parameter passed to rep_is_pf devargs.\n");
5631 		return -EINVAL;
5632 	}
5633 
5634 	rep_is_pf = strtoul(value, &end, 10);
5635 	if (end == NULL || *end != '\0' ||
5636 	    (rep_is_pf == ULONG_MAX && errno == ERANGE)) {
5637 		PMD_DRV_LOG(ERR,
5638 			    "Invalid parameter passed to rep_is_pf devargs.\n");
5639 		return -EINVAL;
5640 	}
5641 
5642 	if (BNXT_DEVARG_REP_IS_PF_INVALID(rep_is_pf)) {
5643 		PMD_DRV_LOG(ERR,
5644 			    "Invalid value passed to rep_is_pf devargs.\n");
5645 		return -EINVAL;
5646 	}
5647 
5648 	vfr_bp->flags |= rep_is_pf;
5649 	if (BNXT_REP_PF(vfr_bp))
5650 		PMD_DRV_LOG(INFO, "PF representor\n");
5651 	else
5652 		PMD_DRV_LOG(INFO, "VF representor\n");
5653 
5654 	return 0;
5655 }
5656 
5657 static int
5658 bnxt_parse_devarg_rep_based_pf(__rte_unused const char *key,
5659 			       const char *value, void *opaque_arg)
5660 {
5661 	struct bnxt_representor *vfr_bp = opaque_arg;
5662 	unsigned long rep_based_pf;
5663 	char *end = NULL;
5664 
5665 	if (!value || !opaque_arg) {
5666 		PMD_DRV_LOG(ERR,
5667 			    "Invalid parameter passed to rep_based_pf "
5668 			    "devargs.\n");
5669 		return -EINVAL;
5670 	}
5671 
5672 	rep_based_pf = strtoul(value, &end, 10);
5673 	if (end == NULL || *end != '\0' ||
5674 	    (rep_based_pf == ULONG_MAX && errno == ERANGE)) {
5675 		PMD_DRV_LOG(ERR,
5676 			    "Invalid parameter passed to rep_based_pf "
5677 			    "devargs.\n");
5678 		return -EINVAL;
5679 	}
5680 
5681 	if (BNXT_DEVARG_REP_BASED_PF_INVALID(rep_based_pf)) {
5682 		PMD_DRV_LOG(ERR,
5683 			    "Invalid value passed to rep_based_pf devargs.\n");
5684 		return -EINVAL;
5685 	}
5686 
5687 	vfr_bp->rep_based_pf = rep_based_pf;
5688 	vfr_bp->flags |= BNXT_REP_BASED_PF_VALID;
5689 
5690 	PMD_DRV_LOG(INFO, "rep-based-pf = %d\n", vfr_bp->rep_based_pf);
5691 
5692 	return 0;
5693 }
5694 
5695 static int
5696 bnxt_parse_devarg_rep_q_r2f(__rte_unused const char *key,
5697 			    const char *value, void *opaque_arg)
5698 {
5699 	struct bnxt_representor *vfr_bp = opaque_arg;
5700 	unsigned long rep_q_r2f;
5701 	char *end = NULL;
5702 
5703 	if (!value || !opaque_arg) {
5704 		PMD_DRV_LOG(ERR,
5705 			    "Invalid parameter passed to rep_q_r2f "
5706 			    "devargs.\n");
5707 		return -EINVAL;
5708 	}
5709 
5710 	rep_q_r2f = strtoul(value, &end, 10);
5711 	if (end == NULL || *end != '\0' ||
5712 	    (rep_q_r2f == ULONG_MAX && errno == ERANGE)) {
5713 		PMD_DRV_LOG(ERR,
5714 			    "Invalid parameter passed to rep_q_r2f "
5715 			    "devargs.\n");
5716 		return -EINVAL;
5717 	}
5718 
5719 	if (BNXT_DEVARG_REP_Q_R2F_INVALID(rep_q_r2f)) {
5720 		PMD_DRV_LOG(ERR,
5721 			    "Invalid value passed to rep_q_r2f devargs.\n");
5722 		return -EINVAL;
5723 	}
5724 
5725 	vfr_bp->rep_q_r2f = rep_q_r2f;
5726 	vfr_bp->flags |= BNXT_REP_Q_R2F_VALID;
5727 	PMD_DRV_LOG(INFO, "rep-q-r2f = %d\n", vfr_bp->rep_q_r2f);
5728 
5729 	return 0;
5730 }
5731 
5732 static int
5733 bnxt_parse_devarg_rep_q_f2r(__rte_unused const char *key,
5734 			    const char *value, void *opaque_arg)
5735 {
5736 	struct bnxt_representor *vfr_bp = opaque_arg;
5737 	unsigned long rep_q_f2r;
5738 	char *end = NULL;
5739 
5740 	if (!value || !opaque_arg) {
5741 		PMD_DRV_LOG(ERR,
5742 			    "Invalid parameter passed to rep_q_f2r "
5743 			    "devargs.\n");
5744 		return -EINVAL;
5745 	}
5746 
5747 	rep_q_f2r = strtoul(value, &end, 10);
5748 	if (end == NULL || *end != '\0' ||
5749 	    (rep_q_f2r == ULONG_MAX && errno == ERANGE)) {
5750 		PMD_DRV_LOG(ERR,
5751 			    "Invalid parameter passed to rep_q_f2r "
5752 			    "devargs.\n");
5753 		return -EINVAL;
5754 	}
5755 
5756 	if (BNXT_DEVARG_REP_Q_F2R_INVALID(rep_q_f2r)) {
5757 		PMD_DRV_LOG(ERR,
5758 			    "Invalid value passed to rep_q_f2r devargs.\n");
5759 		return -EINVAL;
5760 	}
5761 
5762 	vfr_bp->rep_q_f2r = rep_q_f2r;
5763 	vfr_bp->flags |= BNXT_REP_Q_F2R_VALID;
5764 	PMD_DRV_LOG(INFO, "rep-q-f2r = %d\n", vfr_bp->rep_q_f2r);
5765 
5766 	return 0;
5767 }
5768 
5769 static int
5770 bnxt_parse_devarg_rep_fc_r2f(__rte_unused const char *key,
5771 			     const char *value, void *opaque_arg)
5772 {
5773 	struct bnxt_representor *vfr_bp = opaque_arg;
5774 	unsigned long rep_fc_r2f;
5775 	char *end = NULL;
5776 
5777 	if (!value || !opaque_arg) {
5778 		PMD_DRV_LOG(ERR,
5779 			    "Invalid parameter passed to rep_fc_r2f "
5780 			    "devargs.\n");
5781 		return -EINVAL;
5782 	}
5783 
5784 	rep_fc_r2f = strtoul(value, &end, 10);
5785 	if (end == NULL || *end != '\0' ||
5786 	    (rep_fc_r2f == ULONG_MAX && errno == ERANGE)) {
5787 		PMD_DRV_LOG(ERR,
5788 			    "Invalid parameter passed to rep_fc_r2f "
5789 			    "devargs.\n");
5790 		return -EINVAL;
5791 	}
5792 
5793 	if (BNXT_DEVARG_REP_FC_R2F_INVALID(rep_fc_r2f)) {
5794 		PMD_DRV_LOG(ERR,
5795 			    "Invalid value passed to rep_fc_r2f devargs.\n");
5796 		return -EINVAL;
5797 	}
5798 
5799 	vfr_bp->flags |= BNXT_REP_FC_R2F_VALID;
5800 	vfr_bp->rep_fc_r2f = rep_fc_r2f;
5801 	PMD_DRV_LOG(INFO, "rep-fc-r2f = %lu\n", rep_fc_r2f);
5802 
5803 	return 0;
5804 }
5805 
5806 static int
5807 bnxt_parse_devarg_rep_fc_f2r(__rte_unused const char *key,
5808 			     const char *value, void *opaque_arg)
5809 {
5810 	struct bnxt_representor *vfr_bp = opaque_arg;
5811 	unsigned long rep_fc_f2r;
5812 	char *end = NULL;
5813 
5814 	if (!value || !opaque_arg) {
5815 		PMD_DRV_LOG(ERR,
5816 			    "Invalid parameter passed to rep_fc_f2r "
5817 			    "devargs.\n");
5818 		return -EINVAL;
5819 	}
5820 
5821 	rep_fc_f2r = strtoul(value, &end, 10);
5822 	if (end == NULL || *end != '\0' ||
5823 	    (rep_fc_f2r == ULONG_MAX && errno == ERANGE)) {
5824 		PMD_DRV_LOG(ERR,
5825 			    "Invalid parameter passed to rep_fc_f2r "
5826 			    "devargs.\n");
5827 		return -EINVAL;
5828 	}
5829 
5830 	if (BNXT_DEVARG_REP_FC_F2R_INVALID(rep_fc_f2r)) {
5831 		PMD_DRV_LOG(ERR,
5832 			    "Invalid value passed to rep_fc_f2r devargs.\n");
5833 		return -EINVAL;
5834 	}
5835 
5836 	vfr_bp->flags |= BNXT_REP_FC_F2R_VALID;
5837 	vfr_bp->rep_fc_f2r = rep_fc_f2r;
5838 	PMD_DRV_LOG(INFO, "rep-fc-f2r = %lu\n", rep_fc_f2r);
5839 
5840 	return 0;
5841 }
5842 
5843 static int
5844 bnxt_parse_dev_args(struct bnxt *bp, struct rte_devargs *devargs)
5845 {
5846 	struct rte_kvargs *kvlist;
5847 	int ret;
5848 
5849 	if (devargs == NULL)
5850 		return 0;
5851 
5852 	kvlist = rte_kvargs_parse(devargs->args, bnxt_dev_args);
5853 	if (kvlist == NULL)
5854 		return -EINVAL;
5855 
5856 	/*
5857 	 * Handler for "flow_xstat" devarg.
5858 	 * Invoked as for ex: "-a 0000:00:0d.0,flow_xstat=1"
5859 	 */
5860 	ret = rte_kvargs_process(kvlist, BNXT_DEVARG_FLOW_XSTAT,
5861 				 bnxt_parse_devarg_flow_xstat, bp);
5862 	if (ret)
5863 		goto err;
5864 
5865 	/*
5866 	 * Handler for "accum-stats" devarg.
5867 	 * Invoked as for ex: "-a 0000:00:0d.0,accum-stats=1"
5868 	 */
5869 	rte_kvargs_process(kvlist, BNXT_DEVARG_ACCUM_STATS,
5870 			   bnxt_parse_devarg_accum_stats, bp);
5871 	/*
5872 	 * Handler for "max_num_kflows" devarg.
5873 	 * Invoked as for ex: "-a 000:00:0d.0,max_num_kflows=32"
5874 	 */
5875 	ret = rte_kvargs_process(kvlist, BNXT_DEVARG_MAX_NUM_KFLOWS,
5876 				 bnxt_parse_devarg_max_num_kflows, bp);
5877 	if (ret)
5878 		goto err;
5879 
5880 err:
5881 	/*
5882 	 * Handler for "app-id" devarg.
5883 	 * Invoked as for ex: "-a 000:00:0d.0,app-id=1"
5884 	 */
5885 	rte_kvargs_process(kvlist, BNXT_DEVARG_APP_ID,
5886 			   bnxt_parse_devarg_app_id, bp);
5887 
5888 	rte_kvargs_free(kvlist);
5889 	return ret;
5890 }
5891 
5892 static int bnxt_alloc_switch_domain(struct bnxt *bp)
5893 {
5894 	int rc = 0;
5895 
5896 	if (BNXT_PF(bp) || BNXT_VF_IS_TRUSTED(bp)) {
5897 		rc = rte_eth_switch_domain_alloc(&bp->switch_domain_id);
5898 		if (rc)
5899 			PMD_DRV_LOG(ERR,
5900 				    "Failed to alloc switch domain: %d\n", rc);
5901 		else
5902 			PMD_DRV_LOG(INFO,
5903 				    "Switch domain allocated %d\n",
5904 				    bp->switch_domain_id);
5905 	}
5906 
5907 	return rc;
5908 }
5909 
5910 /* Allocate and initialize various fields in bnxt struct that
5911  * need to be allocated/destroyed only once in the lifetime of the driver
5912  */
5913 static int bnxt_drv_init(struct rte_eth_dev *eth_dev)
5914 {
5915 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
5916 	struct bnxt *bp = eth_dev->data->dev_private;
5917 	int rc = 0;
5918 
5919 	bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
5920 
5921 	if (bnxt_vf_pciid(pci_dev->id.device_id))
5922 		bp->flags |= BNXT_FLAG_VF;
5923 
5924 	if (bnxt_p5_device(pci_dev->id.device_id))
5925 		bp->flags |= BNXT_FLAG_CHIP_P5;
5926 
5927 	if (pci_dev->id.device_id == BROADCOM_DEV_ID_58802 ||
5928 	    pci_dev->id.device_id == BROADCOM_DEV_ID_58804 ||
5929 	    pci_dev->id.device_id == BROADCOM_DEV_ID_58808 ||
5930 	    pci_dev->id.device_id == BROADCOM_DEV_ID_58802_VF)
5931 		bp->flags |= BNXT_FLAG_STINGRAY;
5932 
5933 	if (BNXT_TRUFLOW_EN(bp)) {
5934 		/* extra mbuf field is required to store CFA code from mark */
5935 		static const struct rte_mbuf_dynfield bnxt_cfa_code_dynfield_desc = {
5936 			.name = RTE_PMD_BNXT_CFA_CODE_DYNFIELD_NAME,
5937 			.size = sizeof(bnxt_cfa_code_dynfield_t),
5938 			.align = __alignof__(bnxt_cfa_code_dynfield_t),
5939 		};
5940 		bnxt_cfa_code_dynfield_offset =
5941 			rte_mbuf_dynfield_register(&bnxt_cfa_code_dynfield_desc);
5942 		if (bnxt_cfa_code_dynfield_offset < 0) {
5943 			PMD_DRV_LOG(ERR,
5944 			    "Failed to register mbuf field for TruFlow mark\n");
5945 			return -rte_errno;
5946 		}
5947 	}
5948 
5949 	rc = bnxt_map_pci_bars(eth_dev);
5950 	if (rc) {
5951 		PMD_DRV_LOG(ERR,
5952 			    "Failed to initialize board rc: %x\n", rc);
5953 		return rc;
5954 	}
5955 
5956 	rc = bnxt_alloc_pf_info(bp);
5957 	if (rc)
5958 		return rc;
5959 
5960 	rc = bnxt_alloc_link_info(bp);
5961 	if (rc)
5962 		return rc;
5963 
5964 	rc = bnxt_alloc_parent_info(bp);
5965 	if (rc)
5966 		return rc;
5967 
5968 	rc = bnxt_alloc_hwrm_resources(bp);
5969 	if (rc) {
5970 		PMD_DRV_LOG(ERR,
5971 			    "Failed to allocate response buffer rc: %x\n", rc);
5972 		return rc;
5973 	}
5974 	rc = bnxt_alloc_leds_info(bp);
5975 	if (rc)
5976 		return rc;
5977 
5978 	rc = bnxt_alloc_cos_queues(bp);
5979 	if (rc)
5980 		return rc;
5981 
5982 	rc = bnxt_init_locks(bp);
5983 	if (rc)
5984 		return rc;
5985 
5986 	rc = bnxt_alloc_switch_domain(bp);
5987 	if (rc)
5988 		return rc;
5989 
5990 	return rc;
5991 }
5992 
5993 static int
5994 bnxt_dev_init(struct rte_eth_dev *eth_dev, void *params __rte_unused)
5995 {
5996 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
5997 	static int version_printed;
5998 	struct bnxt *bp;
5999 	int rc;
6000 
6001 	if (version_printed++ == 0)
6002 		PMD_DRV_LOG(INFO, "%s\n", bnxt_version);
6003 
6004 	eth_dev->dev_ops = &bnxt_dev_ops;
6005 	eth_dev->rx_queue_count = bnxt_rx_queue_count_op;
6006 	eth_dev->rx_descriptor_status = bnxt_rx_descriptor_status_op;
6007 	eth_dev->tx_descriptor_status = bnxt_tx_descriptor_status_op;
6008 	eth_dev->rx_pkt_burst = &bnxt_recv_pkts;
6009 	eth_dev->tx_pkt_burst = &bnxt_xmit_pkts;
6010 
6011 	/*
6012 	 * For secondary processes, we don't initialise any further
6013 	 * as primary has already done this work.
6014 	 */
6015 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
6016 		return 0;
6017 
6018 	rte_eth_copy_pci_info(eth_dev, pci_dev);
6019 	eth_dev->data->dev_flags |= RTE_ETH_DEV_AUTOFILL_QUEUE_XSTATS;
6020 
6021 	bp = eth_dev->data->dev_private;
6022 
6023 	/* Parse dev arguments passed on when starting the DPDK application. */
6024 	rc = bnxt_parse_dev_args(bp, pci_dev->device.devargs);
6025 	if (rc)
6026 		goto error_free;
6027 
6028 	rc = bnxt_drv_init(eth_dev);
6029 	if (rc)
6030 		goto error_free;
6031 
6032 	rc = bnxt_init_resources(bp, false);
6033 	if (rc)
6034 		goto error_free;
6035 
6036 	rc = bnxt_alloc_stats_mem(bp);
6037 	if (rc)
6038 		goto error_free;
6039 
6040 	PMD_DRV_LOG(INFO,
6041 		    "Found %s device at mem %" PRIX64 ", node addr %pM\n",
6042 		    DRV_MODULE_NAME,
6043 		    pci_dev->mem_resource[0].phys_addr,
6044 		    pci_dev->mem_resource[0].addr);
6045 
6046 	return 0;
6047 
6048 error_free:
6049 	bnxt_dev_uninit(eth_dev);
6050 	return rc;
6051 }
6052 
6053 
6054 static void bnxt_free_ctx_mem_buf(struct bnxt_ctx_mem_buf_info *ctx)
6055 {
6056 	if (!ctx)
6057 		return;
6058 
6059 	if (ctx->va)
6060 		rte_free(ctx->va);
6061 
6062 	ctx->va = NULL;
6063 	ctx->dma = RTE_BAD_IOVA;
6064 	ctx->ctx_id = BNXT_CTX_VAL_INVAL;
6065 }
6066 
6067 static void bnxt_unregister_fc_ctx_mem(struct bnxt *bp)
6068 {
6069 	bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_RX,
6070 				  CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
6071 				  bp->flow_stat->rx_fc_out_tbl.ctx_id,
6072 				  bp->flow_stat->max_fc,
6073 				  false);
6074 
6075 	bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_TX,
6076 				  CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
6077 				  bp->flow_stat->tx_fc_out_tbl.ctx_id,
6078 				  bp->flow_stat->max_fc,
6079 				  false);
6080 
6081 	if (bp->flow_stat->rx_fc_in_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
6082 		bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->rx_fc_in_tbl.ctx_id);
6083 	bp->flow_stat->rx_fc_in_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
6084 
6085 	if (bp->flow_stat->rx_fc_out_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
6086 		bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->rx_fc_out_tbl.ctx_id);
6087 	bp->flow_stat->rx_fc_out_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
6088 
6089 	if (bp->flow_stat->tx_fc_in_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
6090 		bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->tx_fc_in_tbl.ctx_id);
6091 	bp->flow_stat->tx_fc_in_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
6092 
6093 	if (bp->flow_stat->tx_fc_out_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
6094 		bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->tx_fc_out_tbl.ctx_id);
6095 	bp->flow_stat->tx_fc_out_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
6096 }
6097 
6098 static void bnxt_uninit_fc_ctx_mem(struct bnxt *bp)
6099 {
6100 	bnxt_unregister_fc_ctx_mem(bp);
6101 
6102 	bnxt_free_ctx_mem_buf(&bp->flow_stat->rx_fc_in_tbl);
6103 	bnxt_free_ctx_mem_buf(&bp->flow_stat->rx_fc_out_tbl);
6104 	bnxt_free_ctx_mem_buf(&bp->flow_stat->tx_fc_in_tbl);
6105 	bnxt_free_ctx_mem_buf(&bp->flow_stat->tx_fc_out_tbl);
6106 }
6107 
6108 static void bnxt_uninit_ctx_mem(struct bnxt *bp)
6109 {
6110 	if (BNXT_FLOW_XSTATS_EN(bp))
6111 		bnxt_uninit_fc_ctx_mem(bp);
6112 }
6113 
6114 static void
6115 bnxt_free_error_recovery_info(struct bnxt *bp)
6116 {
6117 	rte_free(bp->recovery_info);
6118 	bp->recovery_info = NULL;
6119 	bp->fw_cap &= ~BNXT_FW_CAP_ERROR_RECOVERY;
6120 }
6121 
6122 static int
6123 bnxt_uninit_resources(struct bnxt *bp, bool reconfig_dev)
6124 {
6125 	int rc;
6126 
6127 	bnxt_free_int(bp);
6128 	bnxt_free_mem(bp, reconfig_dev);
6129 
6130 	bnxt_hwrm_func_buf_unrgtr(bp);
6131 	if (bp->pf != NULL) {
6132 		rte_free(bp->pf->vf_req_buf);
6133 		bp->pf->vf_req_buf = NULL;
6134 	}
6135 
6136 	rc = bnxt_hwrm_func_driver_unregister(bp, 0);
6137 	bp->flags &= ~BNXT_FLAG_REGISTERED;
6138 	bnxt_free_ctx_mem(bp);
6139 	if (!reconfig_dev) {
6140 		bnxt_free_hwrm_resources(bp);
6141 		bnxt_free_error_recovery_info(bp);
6142 	}
6143 
6144 	bnxt_uninit_ctx_mem(bp);
6145 
6146 	bnxt_free_flow_stats_info(bp);
6147 	if (bp->rep_info != NULL)
6148 		bnxt_free_switch_domain(bp);
6149 	bnxt_free_rep_info(bp);
6150 	rte_free(bp->ptp_cfg);
6151 	bp->ptp_cfg = NULL;
6152 	return rc;
6153 }
6154 
6155 static int
6156 bnxt_dev_uninit(struct rte_eth_dev *eth_dev)
6157 {
6158 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
6159 		return -EPERM;
6160 
6161 	PMD_DRV_LOG(DEBUG, "Calling Device uninit\n");
6162 
6163 	if (eth_dev->state != RTE_ETH_DEV_UNUSED)
6164 		bnxt_dev_close_op(eth_dev);
6165 
6166 	return 0;
6167 }
6168 
6169 static int bnxt_pci_remove_dev_with_reps(struct rte_eth_dev *eth_dev)
6170 {
6171 	struct bnxt *bp = eth_dev->data->dev_private;
6172 	struct rte_eth_dev *vf_rep_eth_dev;
6173 	int ret = 0, i;
6174 
6175 	if (!bp)
6176 		return -EINVAL;
6177 
6178 	for (i = 0; i < bp->num_reps; i++) {
6179 		vf_rep_eth_dev = bp->rep_info[i].vfr_eth_dev;
6180 		if (!vf_rep_eth_dev)
6181 			continue;
6182 		PMD_DRV_LOG(DEBUG, "BNXT Port:%d VFR pci remove\n",
6183 			    vf_rep_eth_dev->data->port_id);
6184 		rte_eth_dev_destroy(vf_rep_eth_dev, bnxt_representor_uninit);
6185 	}
6186 	PMD_DRV_LOG(DEBUG, "BNXT Port:%d pci remove\n",
6187 		    eth_dev->data->port_id);
6188 	ret = rte_eth_dev_destroy(eth_dev, bnxt_dev_uninit);
6189 
6190 	return ret;
6191 }
6192 
6193 static void bnxt_free_rep_info(struct bnxt *bp)
6194 {
6195 	rte_free(bp->rep_info);
6196 	bp->rep_info = NULL;
6197 	rte_free(bp->cfa_code_map);
6198 	bp->cfa_code_map = NULL;
6199 }
6200 
6201 static int bnxt_init_rep_info(struct bnxt *bp)
6202 {
6203 	int i = 0, rc;
6204 
6205 	if (bp->rep_info)
6206 		return 0;
6207 
6208 	bp->rep_info = rte_zmalloc("bnxt_rep_info",
6209 				   sizeof(bp->rep_info[0]) * BNXT_MAX_VF_REPS,
6210 				   0);
6211 	if (!bp->rep_info) {
6212 		PMD_DRV_LOG(ERR, "Failed to alloc memory for rep info\n");
6213 		return -ENOMEM;
6214 	}
6215 	bp->cfa_code_map = rte_zmalloc("bnxt_cfa_code_map",
6216 				       sizeof(*bp->cfa_code_map) *
6217 				       BNXT_MAX_CFA_CODE, 0);
6218 	if (!bp->cfa_code_map) {
6219 		PMD_DRV_LOG(ERR, "Failed to alloc memory for cfa_code_map\n");
6220 		bnxt_free_rep_info(bp);
6221 		return -ENOMEM;
6222 	}
6223 
6224 	for (i = 0; i < BNXT_MAX_CFA_CODE; i++)
6225 		bp->cfa_code_map[i] = BNXT_VF_IDX_INVALID;
6226 
6227 	rc = pthread_mutex_init(&bp->rep_info->vfr_lock, NULL);
6228 	if (rc) {
6229 		PMD_DRV_LOG(ERR, "Unable to initialize vfr_lock\n");
6230 		bnxt_free_rep_info(bp);
6231 		return rc;
6232 	}
6233 
6234 	rc = pthread_mutex_init(&bp->rep_info->vfr_start_lock, NULL);
6235 	if (rc) {
6236 		PMD_DRV_LOG(ERR, "Unable to initialize vfr_start_lock\n");
6237 		bnxt_free_rep_info(bp);
6238 		return rc;
6239 	}
6240 
6241 	return rc;
6242 }
6243 
6244 static int bnxt_rep_port_probe(struct rte_pci_device *pci_dev,
6245 			       struct rte_eth_devargs *eth_da,
6246 			       struct rte_eth_dev *backing_eth_dev,
6247 			       const char *dev_args)
6248 {
6249 	struct rte_eth_dev *vf_rep_eth_dev;
6250 	char name[RTE_ETH_NAME_MAX_LEN];
6251 	struct bnxt *backing_bp;
6252 	uint16_t num_rep;
6253 	int i, ret = 0;
6254 	struct rte_kvargs *kvlist = NULL;
6255 
6256 	if (eth_da->type == RTE_ETH_REPRESENTOR_NONE)
6257 		return 0;
6258 	if (eth_da->type != RTE_ETH_REPRESENTOR_VF) {
6259 		PMD_DRV_LOG(ERR, "unsupported representor type %d\n",
6260 			    eth_da->type);
6261 		return -ENOTSUP;
6262 	}
6263 	num_rep = eth_da->nb_representor_ports;
6264 	if (num_rep > BNXT_MAX_VF_REPS) {
6265 		PMD_DRV_LOG(ERR, "nb_representor_ports = %d > %d MAX VF REPS\n",
6266 			    num_rep, BNXT_MAX_VF_REPS);
6267 		return -EINVAL;
6268 	}
6269 
6270 	if (num_rep >= RTE_MAX_ETHPORTS) {
6271 		PMD_DRV_LOG(ERR,
6272 			    "nb_representor_ports = %d > %d MAX ETHPORTS\n",
6273 			    num_rep, RTE_MAX_ETHPORTS);
6274 		return -EINVAL;
6275 	}
6276 
6277 	backing_bp = backing_eth_dev->data->dev_private;
6278 
6279 	if (!(BNXT_PF(backing_bp) || BNXT_VF_IS_TRUSTED(backing_bp))) {
6280 		PMD_DRV_LOG(ERR,
6281 			    "Not a PF or trusted VF. No Representor support\n");
6282 		/* Returning an error is not an option.
6283 		 * Applications are not handling this correctly
6284 		 */
6285 		return 0;
6286 	}
6287 
6288 	if (bnxt_init_rep_info(backing_bp))
6289 		return 0;
6290 
6291 	for (i = 0; i < num_rep; i++) {
6292 		struct bnxt_representor representor = {
6293 			.vf_id = eth_da->representor_ports[i],
6294 			.switch_domain_id = backing_bp->switch_domain_id,
6295 			.parent_dev = backing_eth_dev
6296 		};
6297 
6298 		if (representor.vf_id >= BNXT_MAX_VF_REPS) {
6299 			PMD_DRV_LOG(ERR, "VF-Rep id %d >= %d MAX VF ID\n",
6300 				    representor.vf_id, BNXT_MAX_VF_REPS);
6301 			continue;
6302 		}
6303 
6304 		/* representor port net_bdf_port */
6305 		snprintf(name, sizeof(name), "net_%s_representor_%d",
6306 			 pci_dev->device.name, eth_da->representor_ports[i]);
6307 
6308 		kvlist = rte_kvargs_parse(dev_args, bnxt_dev_args);
6309 		if (kvlist) {
6310 			/*
6311 			 * Handler for "rep_is_pf" devarg.
6312 			 * Invoked as for ex: "-a 000:00:0d.0,
6313 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6314 			 */
6315 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_IS_PF,
6316 						 bnxt_parse_devarg_rep_is_pf,
6317 						 (void *)&representor);
6318 			if (ret) {
6319 				ret = -EINVAL;
6320 				goto err;
6321 			}
6322 			/*
6323 			 * Handler for "rep_based_pf" devarg.
6324 			 * Invoked as for ex: "-a 000:00:0d.0,
6325 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6326 			 */
6327 			ret = rte_kvargs_process(kvlist,
6328 						 BNXT_DEVARG_REP_BASED_PF,
6329 						 bnxt_parse_devarg_rep_based_pf,
6330 						 (void *)&representor);
6331 			if (ret) {
6332 				ret = -EINVAL;
6333 				goto err;
6334 			}
6335 			/*
6336 			 * Handler for "rep_based_pf" devarg.
6337 			 * Invoked as for ex: "-a 000:00:0d.0,
6338 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6339 			 */
6340 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_Q_R2F,
6341 						 bnxt_parse_devarg_rep_q_r2f,
6342 						 (void *)&representor);
6343 			if (ret) {
6344 				ret = -EINVAL;
6345 				goto err;
6346 			}
6347 			/*
6348 			 * Handler for "rep_based_pf" devarg.
6349 			 * Invoked as for ex: "-a 000:00:0d.0,
6350 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6351 			 */
6352 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_Q_F2R,
6353 						 bnxt_parse_devarg_rep_q_f2r,
6354 						 (void *)&representor);
6355 			if (ret) {
6356 				ret = -EINVAL;
6357 				goto err;
6358 			}
6359 			/*
6360 			 * Handler for "rep_based_pf" devarg.
6361 			 * Invoked as for ex: "-a 000:00:0d.0,
6362 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6363 			 */
6364 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_FC_R2F,
6365 						 bnxt_parse_devarg_rep_fc_r2f,
6366 						 (void *)&representor);
6367 			if (ret) {
6368 				ret = -EINVAL;
6369 				goto err;
6370 			}
6371 			/*
6372 			 * Handler for "rep_based_pf" devarg.
6373 			 * Invoked as for ex: "-a 000:00:0d.0,
6374 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6375 			 */
6376 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_FC_F2R,
6377 						 bnxt_parse_devarg_rep_fc_f2r,
6378 						 (void *)&representor);
6379 			if (ret) {
6380 				ret = -EINVAL;
6381 				goto err;
6382 			}
6383 		}
6384 
6385 		ret = rte_eth_dev_create(&pci_dev->device, name,
6386 					 sizeof(struct bnxt_representor),
6387 					 NULL, NULL,
6388 					 bnxt_representor_init,
6389 					 &representor);
6390 		if (ret) {
6391 			PMD_DRV_LOG(ERR, "failed to create bnxt vf "
6392 				    "representor %s.", name);
6393 			goto err;
6394 		}
6395 
6396 		vf_rep_eth_dev = rte_eth_dev_allocated(name);
6397 		if (!vf_rep_eth_dev) {
6398 			PMD_DRV_LOG(ERR, "Failed to find the eth_dev"
6399 				    " for VF-Rep: %s.", name);
6400 			ret = -ENODEV;
6401 			goto err;
6402 		}
6403 
6404 		PMD_DRV_LOG(DEBUG, "BNXT Port:%d VFR pci probe\n",
6405 			    backing_eth_dev->data->port_id);
6406 		backing_bp->rep_info[representor.vf_id].vfr_eth_dev =
6407 							 vf_rep_eth_dev;
6408 		backing_bp->num_reps++;
6409 
6410 	}
6411 
6412 	rte_kvargs_free(kvlist);
6413 	return 0;
6414 
6415 err:
6416 	/* If num_rep > 1, then rollback already created
6417 	 * ports, since we'll be failing the probe anyway
6418 	 */
6419 	if (num_rep > 1)
6420 		bnxt_pci_remove_dev_with_reps(backing_eth_dev);
6421 	rte_errno = -ret;
6422 	rte_kvargs_free(kvlist);
6423 
6424 	return ret;
6425 }
6426 
6427 static int bnxt_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
6428 			  struct rte_pci_device *pci_dev)
6429 {
6430 	struct rte_eth_devargs eth_da = { .nb_representor_ports = 0 };
6431 	struct rte_eth_dev *backing_eth_dev;
6432 	uint16_t num_rep;
6433 	int ret = 0;
6434 
6435 	if (pci_dev->device.devargs) {
6436 		ret = rte_eth_devargs_parse(pci_dev->device.devargs->args,
6437 					    &eth_da);
6438 		if (ret)
6439 			return ret;
6440 	}
6441 
6442 	num_rep = eth_da.nb_representor_ports;
6443 	PMD_DRV_LOG(DEBUG, "nb_representor_ports = %d\n",
6444 		    num_rep);
6445 
6446 	/* We could come here after first level of probe is already invoked
6447 	 * as part of an application bringup(OVS-DPDK vswitchd), so first check
6448 	 * for already allocated eth_dev for the backing device (PF/Trusted VF)
6449 	 */
6450 	backing_eth_dev = rte_eth_dev_allocated(pci_dev->device.name);
6451 	if (backing_eth_dev == NULL) {
6452 		ret = rte_eth_dev_create(&pci_dev->device, pci_dev->device.name,
6453 					 sizeof(struct bnxt),
6454 					 eth_dev_pci_specific_init, pci_dev,
6455 					 bnxt_dev_init, NULL);
6456 
6457 		if (ret || !num_rep)
6458 			return ret;
6459 
6460 		backing_eth_dev = rte_eth_dev_allocated(pci_dev->device.name);
6461 	}
6462 	PMD_DRV_LOG(DEBUG, "BNXT Port:%d pci probe\n",
6463 		    backing_eth_dev->data->port_id);
6464 
6465 	if (!num_rep)
6466 		return ret;
6467 
6468 	/* probe representor ports now */
6469 	ret = bnxt_rep_port_probe(pci_dev, &eth_da, backing_eth_dev,
6470 				  pci_dev->device.devargs->args);
6471 
6472 	return ret;
6473 }
6474 
6475 static int bnxt_pci_remove(struct rte_pci_device *pci_dev)
6476 {
6477 	struct rte_eth_dev *eth_dev;
6478 
6479 	eth_dev = rte_eth_dev_allocated(pci_dev->device.name);
6480 	if (!eth_dev)
6481 		return 0; /* Invoked typically only by OVS-DPDK, by the
6482 			   * time it comes here the eth_dev is already
6483 			   * deleted by rte_eth_dev_close(), so returning
6484 			   * +ve value will at least help in proper cleanup
6485 			   */
6486 
6487 	PMD_DRV_LOG(DEBUG, "BNXT Port:%d pci remove\n", eth_dev->data->port_id);
6488 	if (rte_eal_process_type() == RTE_PROC_PRIMARY) {
6489 		if (eth_dev->data->dev_flags & RTE_ETH_DEV_REPRESENTOR)
6490 			return rte_eth_dev_destroy(eth_dev,
6491 						   bnxt_representor_uninit);
6492 		else
6493 			return rte_eth_dev_destroy(eth_dev,
6494 						   bnxt_dev_uninit);
6495 	} else {
6496 		return rte_eth_dev_pci_generic_remove(pci_dev, NULL);
6497 	}
6498 }
6499 
6500 static struct rte_pci_driver bnxt_rte_pmd = {
6501 	.id_table = bnxt_pci_id_map,
6502 	.drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC |
6503 			RTE_PCI_DRV_PROBE_AGAIN, /* Needed in case of VF-REPs
6504 						  * and OVS-DPDK
6505 						  */
6506 	.probe = bnxt_pci_probe,
6507 	.remove = bnxt_pci_remove,
6508 };
6509 
6510 static bool
6511 is_device_supported(struct rte_eth_dev *dev, struct rte_pci_driver *drv)
6512 {
6513 	if (strcmp(dev->device->driver->name, drv->driver.name))
6514 		return false;
6515 
6516 	return true;
6517 }
6518 
6519 bool is_bnxt_supported(struct rte_eth_dev *dev)
6520 {
6521 	return is_device_supported(dev, &bnxt_rte_pmd);
6522 }
6523 
6524 RTE_LOG_REGISTER_SUFFIX(bnxt_logtype_driver, driver, NOTICE);
6525 RTE_PMD_REGISTER_PCI(net_bnxt, bnxt_rte_pmd);
6526 RTE_PMD_REGISTER_PCI_TABLE(net_bnxt, bnxt_pci_id_map);
6527 RTE_PMD_REGISTER_KMOD_DEP(net_bnxt, "* igb_uio | uio_pci_generic | vfio-pci");
6528