xref: /dpdk/drivers/net/bnxt/bnxt_ethdev.c (revision 10eaf41d)
1 /* SPDX-License-Identifier: BSD-3-Clause
2  * Copyright(c) 2014-2021 Broadcom
3  * All rights reserved.
4  */
5 
6 #include <inttypes.h>
7 #include <stdbool.h>
8 
9 #include <rte_dev.h>
10 #include <ethdev_driver.h>
11 #include <ethdev_pci.h>
12 #include <rte_malloc.h>
13 #include <rte_cycles.h>
14 #include <rte_alarm.h>
15 #include <rte_kvargs.h>
16 #include <rte_vect.h>
17 
18 #include "bnxt.h"
19 #include "bnxt_filter.h"
20 #include "bnxt_hwrm.h"
21 #include "bnxt_irq.h"
22 #include "bnxt_reps.h"
23 #include "bnxt_ring.h"
24 #include "bnxt_rxq.h"
25 #include "bnxt_rxr.h"
26 #include "bnxt_stats.h"
27 #include "bnxt_txq.h"
28 #include "bnxt_txr.h"
29 #include "bnxt_vnic.h"
30 #include "hsi_struct_def_dpdk.h"
31 #include "bnxt_nvm_defs.h"
32 #include "bnxt_tf_common.h"
33 #include "ulp_flow_db.h"
34 #include "rte_pmd_bnxt.h"
35 
36 #define DRV_MODULE_NAME		"bnxt"
37 static const char bnxt_version[] =
38 	"Broadcom NetXtreme driver " DRV_MODULE_NAME;
39 
40 /*
41  * The set of PCI devices this driver supports
42  */
43 static const struct rte_pci_id bnxt_pci_id_map[] = {
44 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM,
45 			 BROADCOM_DEV_ID_STRATUS_NIC_VF1) },
46 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM,
47 			 BROADCOM_DEV_ID_STRATUS_NIC_VF2) },
48 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_STRATUS_NIC) },
49 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57414_VF) },
50 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57304_VF) },
51 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_NS2) },
52 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57406_VF) },
53 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57407_MF) },
54 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_5741X_VF) },
55 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_5731X_VF) },
56 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57417_MF) },
57 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57412) },
58 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57414) },
59 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57416_RJ45) },
60 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57417_RJ45) },
61 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57412_MF) },
62 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57317_RJ45) },
63 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57417_SFP) },
64 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57416_SFP) },
65 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57317_SFP) },
66 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57414_MF) },
67 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57416_MF) },
68 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58802) },
69 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58804) },
70 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58808) },
71 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58802_VF) },
72 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57508) },
73 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57504) },
74 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57502) },
75 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57500_VF1) },
76 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57500_VF2) },
77 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57508_MF1) },
78 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57504_MF1) },
79 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57502_MF1) },
80 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57508_MF2) },
81 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57504_MF2) },
82 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57502_MF2) },
83 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58812) },
84 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58814) },
85 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58818) },
86 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58818_VF) },
87 	{ .vendor_id = 0, /* sentinel */ },
88 };
89 
90 #define	BNXT_DEVARG_ACCUM_STATS	"accum-stats"
91 #define BNXT_DEVARG_FLOW_XSTAT	"flow-xstat"
92 #define BNXT_DEVARG_MAX_NUM_KFLOWS  "max-num-kflows"
93 #define BNXT_DEVARG_REPRESENTOR	"representor"
94 #define BNXT_DEVARG_REP_BASED_PF  "rep-based-pf"
95 #define BNXT_DEVARG_REP_IS_PF  "rep-is-pf"
96 #define BNXT_DEVARG_REP_Q_R2F  "rep-q-r2f"
97 #define BNXT_DEVARG_REP_Q_F2R  "rep-q-f2r"
98 #define BNXT_DEVARG_REP_FC_R2F  "rep-fc-r2f"
99 #define BNXT_DEVARG_REP_FC_F2R  "rep-fc-f2r"
100 #define BNXT_DEVARG_APP_ID	"app-id"
101 
102 static const char *const bnxt_dev_args[] = {
103 	BNXT_DEVARG_REPRESENTOR,
104 	BNXT_DEVARG_ACCUM_STATS,
105 	BNXT_DEVARG_FLOW_XSTAT,
106 	BNXT_DEVARG_MAX_NUM_KFLOWS,
107 	BNXT_DEVARG_REP_BASED_PF,
108 	BNXT_DEVARG_REP_IS_PF,
109 	BNXT_DEVARG_REP_Q_R2F,
110 	BNXT_DEVARG_REP_Q_F2R,
111 	BNXT_DEVARG_REP_FC_R2F,
112 	BNXT_DEVARG_REP_FC_F2R,
113 	BNXT_DEVARG_APP_ID,
114 	NULL
115 };
116 
117 /*
118  * accum-stats == false to disable flow counter accumulation
119  * accum-stats == true to enable flow counter accumulation
120  */
121 #define	BNXT_DEVARG_ACCUM_STATS_INVALID(accum_stats)	((accum_stats) > 1)
122 
123 /*
124  * app-id = an non-negative 8-bit number
125  */
126 #define BNXT_DEVARG_APP_ID_INVALID(val)			((val) > 255)
127 
128 /*
129  * flow_xstat == false to disable the feature
130  * flow_xstat == true to enable the feature
131  */
132 #define	BNXT_DEVARG_FLOW_XSTAT_INVALID(flow_xstat)	((flow_xstat) > 1)
133 
134 /*
135  * rep_is_pf == false to indicate VF representor
136  * rep_is_pf == true to indicate PF representor
137  */
138 #define	BNXT_DEVARG_REP_IS_PF_INVALID(rep_is_pf)	((rep_is_pf) > 1)
139 
140 /*
141  * rep_based_pf == Physical index of the PF
142  */
143 #define	BNXT_DEVARG_REP_BASED_PF_INVALID(rep_based_pf)	((rep_based_pf) > 15)
144 /*
145  * rep_q_r2f == Logical COS Queue index for the rep to endpoint direction
146  */
147 #define	BNXT_DEVARG_REP_Q_R2F_INVALID(rep_q_r2f)	((rep_q_r2f) > 3)
148 
149 /*
150  * rep_q_f2r == Logical COS Queue index for the endpoint to rep direction
151  */
152 #define	BNXT_DEVARG_REP_Q_F2R_INVALID(rep_q_f2r)	((rep_q_f2r) > 3)
153 
154 /*
155  * rep_fc_r2f == Flow control for the representor to endpoint direction
156  */
157 #define BNXT_DEVARG_REP_FC_R2F_INVALID(rep_fc_r2f)	((rep_fc_r2f) > 1)
158 
159 /*
160  * rep_fc_f2r == Flow control for the endpoint to representor direction
161  */
162 #define BNXT_DEVARG_REP_FC_F2R_INVALID(rep_fc_f2r)	((rep_fc_f2r) > 1)
163 
164 int bnxt_cfa_code_dynfield_offset = -1;
165 
166 /*
167  * max_num_kflows must be >= 32
168  * and must be a power-of-2 supported value
169  * return: 1 -> invalid
170  *         0 -> valid
171  */
172 static int bnxt_devarg_max_num_kflow_invalid(uint16_t max_num_kflows)
173 {
174 	if (max_num_kflows < 32 || !rte_is_power_of_2(max_num_kflows))
175 		return 1;
176 	return 0;
177 }
178 
179 static int bnxt_vlan_offload_set_op(struct rte_eth_dev *dev, int mask);
180 static int bnxt_dev_uninit(struct rte_eth_dev *eth_dev);
181 static int bnxt_init_resources(struct bnxt *bp, bool reconfig_dev);
182 static int bnxt_uninit_resources(struct bnxt *bp, bool reconfig_dev);
183 static void bnxt_cancel_fw_health_check(struct bnxt *bp);
184 static int bnxt_restore_vlan_filters(struct bnxt *bp);
185 static void bnxt_dev_recover(void *arg);
186 static void bnxt_free_error_recovery_info(struct bnxt *bp);
187 static void bnxt_free_rep_info(struct bnxt *bp);
188 
189 int is_bnxt_in_error(struct bnxt *bp)
190 {
191 	if (bp->flags & BNXT_FLAG_FATAL_ERROR)
192 		return -EIO;
193 	if (bp->flags & BNXT_FLAG_FW_RESET)
194 		return -EBUSY;
195 
196 	return 0;
197 }
198 
199 /***********************/
200 
201 /*
202  * High level utility functions
203  */
204 
205 static uint16_t bnxt_rss_ctxts(const struct bnxt *bp)
206 {
207 	unsigned int num_rss_rings = RTE_MIN(bp->rx_nr_rings,
208 					     BNXT_RSS_TBL_SIZE_P5);
209 
210 	if (!BNXT_CHIP_P5(bp))
211 		return 1;
212 
213 	return RTE_ALIGN_MUL_CEIL(num_rss_rings,
214 				  BNXT_RSS_ENTRIES_PER_CTX_P5) /
215 				  BNXT_RSS_ENTRIES_PER_CTX_P5;
216 }
217 
218 uint16_t bnxt_rss_hash_tbl_size(const struct bnxt *bp)
219 {
220 	if (!BNXT_CHIP_P5(bp))
221 		return HW_HASH_INDEX_SIZE;
222 
223 	return bnxt_rss_ctxts(bp) * BNXT_RSS_ENTRIES_PER_CTX_P5;
224 }
225 
226 static void bnxt_free_parent_info(struct bnxt *bp)
227 {
228 	rte_free(bp->parent);
229 	bp->parent = NULL;
230 }
231 
232 static void bnxt_free_pf_info(struct bnxt *bp)
233 {
234 	rte_free(bp->pf);
235 	bp->pf = NULL;
236 }
237 
238 static void bnxt_free_link_info(struct bnxt *bp)
239 {
240 	rte_free(bp->link_info);
241 	bp->link_info = NULL;
242 }
243 
244 static void bnxt_free_leds_info(struct bnxt *bp)
245 {
246 	if (BNXT_VF(bp))
247 		return;
248 
249 	rte_free(bp->leds);
250 	bp->leds = NULL;
251 }
252 
253 static void bnxt_free_flow_stats_info(struct bnxt *bp)
254 {
255 	rte_free(bp->flow_stat);
256 	bp->flow_stat = NULL;
257 }
258 
259 static void bnxt_free_cos_queues(struct bnxt *bp)
260 {
261 	rte_free(bp->rx_cos_queue);
262 	bp->rx_cos_queue = NULL;
263 	rte_free(bp->tx_cos_queue);
264 	bp->tx_cos_queue = NULL;
265 }
266 
267 static void bnxt_free_mem(struct bnxt *bp, bool reconfig)
268 {
269 	bnxt_free_filter_mem(bp);
270 	bnxt_free_vnic_attributes(bp);
271 	bnxt_free_vnic_mem(bp);
272 
273 	/* tx/rx rings are configured as part of *_queue_setup callbacks.
274 	 * If the number of rings change across fw update,
275 	 * we don't have much choice except to warn the user.
276 	 */
277 	if (!reconfig) {
278 		bnxt_free_stats(bp);
279 		bnxt_free_tx_rings(bp);
280 		bnxt_free_rx_rings(bp);
281 	}
282 	bnxt_free_async_cp_ring(bp);
283 	bnxt_free_rxtx_nq_ring(bp);
284 
285 	rte_free(bp->grp_info);
286 	bp->grp_info = NULL;
287 }
288 
289 static int bnxt_alloc_parent_info(struct bnxt *bp)
290 {
291 	bp->parent = rte_zmalloc("bnxt_parent_info",
292 				 sizeof(struct bnxt_parent_info), 0);
293 	if (bp->parent == NULL)
294 		return -ENOMEM;
295 
296 	return 0;
297 }
298 
299 static int bnxt_alloc_pf_info(struct bnxt *bp)
300 {
301 	bp->pf = rte_zmalloc("bnxt_pf_info", sizeof(struct bnxt_pf_info), 0);
302 	if (bp->pf == NULL)
303 		return -ENOMEM;
304 
305 	return 0;
306 }
307 
308 static int bnxt_alloc_link_info(struct bnxt *bp)
309 {
310 	bp->link_info =
311 		rte_zmalloc("bnxt_link_info", sizeof(struct bnxt_link_info), 0);
312 	if (bp->link_info == NULL)
313 		return -ENOMEM;
314 
315 	return 0;
316 }
317 
318 static int bnxt_alloc_leds_info(struct bnxt *bp)
319 {
320 	if (BNXT_VF(bp))
321 		return 0;
322 
323 	bp->leds = rte_zmalloc("bnxt_leds",
324 			       BNXT_MAX_LED * sizeof(struct bnxt_led_info),
325 			       0);
326 	if (bp->leds == NULL)
327 		return -ENOMEM;
328 
329 	return 0;
330 }
331 
332 static int bnxt_alloc_cos_queues(struct bnxt *bp)
333 {
334 	bp->rx_cos_queue =
335 		rte_zmalloc("bnxt_rx_cosq",
336 			    BNXT_COS_QUEUE_COUNT *
337 			    sizeof(struct bnxt_cos_queue_info),
338 			    0);
339 	if (bp->rx_cos_queue == NULL)
340 		return -ENOMEM;
341 
342 	bp->tx_cos_queue =
343 		rte_zmalloc("bnxt_tx_cosq",
344 			    BNXT_COS_QUEUE_COUNT *
345 			    sizeof(struct bnxt_cos_queue_info),
346 			    0);
347 	if (bp->tx_cos_queue == NULL)
348 		return -ENOMEM;
349 
350 	return 0;
351 }
352 
353 static int bnxt_alloc_flow_stats_info(struct bnxt *bp)
354 {
355 	bp->flow_stat = rte_zmalloc("bnxt_flow_xstat",
356 				    sizeof(struct bnxt_flow_stat_info), 0);
357 	if (bp->flow_stat == NULL)
358 		return -ENOMEM;
359 
360 	return 0;
361 }
362 
363 static int bnxt_alloc_mem(struct bnxt *bp, bool reconfig)
364 {
365 	int rc;
366 
367 	rc = bnxt_alloc_ring_grps(bp);
368 	if (rc)
369 		goto alloc_mem_err;
370 
371 	rc = bnxt_alloc_async_ring_struct(bp);
372 	if (rc)
373 		goto alloc_mem_err;
374 
375 	rc = bnxt_alloc_vnic_mem(bp);
376 	if (rc)
377 		goto alloc_mem_err;
378 
379 	rc = bnxt_alloc_vnic_attributes(bp);
380 	if (rc)
381 		goto alloc_mem_err;
382 
383 	rc = bnxt_alloc_filter_mem(bp);
384 	if (rc)
385 		goto alloc_mem_err;
386 
387 	rc = bnxt_alloc_async_cp_ring(bp);
388 	if (rc)
389 		goto alloc_mem_err;
390 
391 	rc = bnxt_alloc_rxtx_nq_ring(bp);
392 	if (rc)
393 		goto alloc_mem_err;
394 
395 	if (BNXT_FLOW_XSTATS_EN(bp)) {
396 		rc = bnxt_alloc_flow_stats_info(bp);
397 		if (rc)
398 			goto alloc_mem_err;
399 	}
400 
401 	return 0;
402 
403 alloc_mem_err:
404 	bnxt_free_mem(bp, reconfig);
405 	return rc;
406 }
407 
408 static int bnxt_setup_one_vnic(struct bnxt *bp, uint16_t vnic_id)
409 {
410 	struct rte_eth_conf *dev_conf = &bp->eth_dev->data->dev_conf;
411 	struct bnxt_vnic_info *vnic = &bp->vnic_info[vnic_id];
412 	uint64_t rx_offloads = dev_conf->rxmode.offloads;
413 	struct bnxt_rx_queue *rxq;
414 	unsigned int j;
415 	int rc;
416 
417 	rc = bnxt_vnic_grp_alloc(bp, vnic);
418 	if (rc)
419 		goto err_out;
420 
421 	PMD_DRV_LOG(DEBUG, "vnic[%d] = %p vnic->fw_grp_ids = %p\n",
422 		    vnic_id, vnic, vnic->fw_grp_ids);
423 
424 	rc = bnxt_hwrm_vnic_alloc(bp, vnic);
425 	if (rc)
426 		goto err_out;
427 
428 	/* Alloc RSS context only if RSS mode is enabled */
429 	if (dev_conf->rxmode.mq_mode & ETH_MQ_RX_RSS) {
430 		int j, nr_ctxs = bnxt_rss_ctxts(bp);
431 
432 		/* RSS table size in Thor is 512.
433 		 * Cap max Rx rings to same value
434 		 */
435 		if (bp->rx_nr_rings > BNXT_RSS_TBL_SIZE_P5) {
436 			PMD_DRV_LOG(ERR, "RxQ cnt %d > reta_size %d\n",
437 				    bp->rx_nr_rings, BNXT_RSS_TBL_SIZE_P5);
438 			goto err_out;
439 		}
440 
441 		rc = 0;
442 		for (j = 0; j < nr_ctxs; j++) {
443 			rc = bnxt_hwrm_vnic_ctx_alloc(bp, vnic, j);
444 			if (rc)
445 				break;
446 		}
447 		if (rc) {
448 			PMD_DRV_LOG(ERR,
449 				    "HWRM vnic %d ctx %d alloc failure rc: %x\n",
450 				    vnic_id, j, rc);
451 			goto err_out;
452 		}
453 		vnic->num_lb_ctxts = nr_ctxs;
454 	}
455 
456 	/*
457 	 * Firmware sets pf pair in default vnic cfg. If the VLAN strip
458 	 * setting is not available at this time, it will not be
459 	 * configured correctly in the CFA.
460 	 */
461 	if (rx_offloads & DEV_RX_OFFLOAD_VLAN_STRIP)
462 		vnic->vlan_strip = true;
463 	else
464 		vnic->vlan_strip = false;
465 
466 	rc = bnxt_hwrm_vnic_cfg(bp, vnic);
467 	if (rc)
468 		goto err_out;
469 
470 	rc = bnxt_set_hwrm_vnic_filters(bp, vnic);
471 	if (rc)
472 		goto err_out;
473 
474 	for (j = 0; j < bp->rx_num_qs_per_vnic; j++) {
475 		rxq = bp->eth_dev->data->rx_queues[j];
476 
477 		PMD_DRV_LOG(DEBUG,
478 			    "rxq[%d]->vnic=%p vnic->fw_grp_ids=%p\n",
479 			    j, rxq->vnic, rxq->vnic->fw_grp_ids);
480 
481 		if (BNXT_HAS_RING_GRPS(bp) && rxq->rx_deferred_start)
482 			rxq->vnic->fw_grp_ids[j] = INVALID_HW_RING_ID;
483 		else
484 			vnic->rx_queue_cnt++;
485 	}
486 
487 	PMD_DRV_LOG(DEBUG, "vnic->rx_queue_cnt = %d\n", vnic->rx_queue_cnt);
488 
489 	rc = bnxt_vnic_rss_configure(bp, vnic);
490 	if (rc)
491 		goto err_out;
492 
493 	bnxt_hwrm_vnic_plcmode_cfg(bp, vnic);
494 
495 	rc = bnxt_hwrm_vnic_tpa_cfg(bp, vnic,
496 				    (rx_offloads & DEV_RX_OFFLOAD_TCP_LRO) ?
497 				    true : false);
498 	if (rc)
499 		goto err_out;
500 
501 	return 0;
502 err_out:
503 	PMD_DRV_LOG(ERR, "HWRM vnic %d cfg failure rc: %x\n",
504 		    vnic_id, rc);
505 	return rc;
506 }
507 
508 static int bnxt_register_fc_ctx_mem(struct bnxt *bp)
509 {
510 	int rc = 0;
511 
512 	rc = bnxt_hwrm_ctx_rgtr(bp, bp->flow_stat->rx_fc_in_tbl.dma,
513 				&bp->flow_stat->rx_fc_in_tbl.ctx_id);
514 	if (rc)
515 		return rc;
516 
517 	PMD_DRV_LOG(DEBUG,
518 		    "rx_fc_in_tbl.va = %p rx_fc_in_tbl.dma = %p"
519 		    " rx_fc_in_tbl.ctx_id = %d\n",
520 		    bp->flow_stat->rx_fc_in_tbl.va,
521 		    (void *)((uintptr_t)bp->flow_stat->rx_fc_in_tbl.dma),
522 		    bp->flow_stat->rx_fc_in_tbl.ctx_id);
523 
524 	rc = bnxt_hwrm_ctx_rgtr(bp, bp->flow_stat->rx_fc_out_tbl.dma,
525 				&bp->flow_stat->rx_fc_out_tbl.ctx_id);
526 	if (rc)
527 		return rc;
528 
529 	PMD_DRV_LOG(DEBUG,
530 		    "rx_fc_out_tbl.va = %p rx_fc_out_tbl.dma = %p"
531 		    " rx_fc_out_tbl.ctx_id = %d\n",
532 		    bp->flow_stat->rx_fc_out_tbl.va,
533 		    (void *)((uintptr_t)bp->flow_stat->rx_fc_out_tbl.dma),
534 		    bp->flow_stat->rx_fc_out_tbl.ctx_id);
535 
536 	rc = bnxt_hwrm_ctx_rgtr(bp, bp->flow_stat->tx_fc_in_tbl.dma,
537 				&bp->flow_stat->tx_fc_in_tbl.ctx_id);
538 	if (rc)
539 		return rc;
540 
541 	PMD_DRV_LOG(DEBUG,
542 		    "tx_fc_in_tbl.va = %p tx_fc_in_tbl.dma = %p"
543 		    " tx_fc_in_tbl.ctx_id = %d\n",
544 		    bp->flow_stat->tx_fc_in_tbl.va,
545 		    (void *)((uintptr_t)bp->flow_stat->tx_fc_in_tbl.dma),
546 		    bp->flow_stat->tx_fc_in_tbl.ctx_id);
547 
548 	rc = bnxt_hwrm_ctx_rgtr(bp, bp->flow_stat->tx_fc_out_tbl.dma,
549 				&bp->flow_stat->tx_fc_out_tbl.ctx_id);
550 	if (rc)
551 		return rc;
552 
553 	PMD_DRV_LOG(DEBUG,
554 		    "tx_fc_out_tbl.va = %p tx_fc_out_tbl.dma = %p"
555 		    " tx_fc_out_tbl.ctx_id = %d\n",
556 		    bp->flow_stat->tx_fc_out_tbl.va,
557 		    (void *)((uintptr_t)bp->flow_stat->tx_fc_out_tbl.dma),
558 		    bp->flow_stat->tx_fc_out_tbl.ctx_id);
559 
560 	memset(bp->flow_stat->rx_fc_out_tbl.va,
561 	       0,
562 	       bp->flow_stat->rx_fc_out_tbl.size);
563 	rc = bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_RX,
564 				       CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
565 				       bp->flow_stat->rx_fc_out_tbl.ctx_id,
566 				       bp->flow_stat->max_fc,
567 				       true);
568 	if (rc)
569 		return rc;
570 
571 	memset(bp->flow_stat->tx_fc_out_tbl.va,
572 	       0,
573 	       bp->flow_stat->tx_fc_out_tbl.size);
574 	rc = bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_TX,
575 				       CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
576 				       bp->flow_stat->tx_fc_out_tbl.ctx_id,
577 				       bp->flow_stat->max_fc,
578 				       true);
579 
580 	return rc;
581 }
582 
583 static int bnxt_alloc_ctx_mem_buf(struct bnxt *bp, char *type, size_t size,
584 				  struct bnxt_ctx_mem_buf_info *ctx)
585 {
586 	if (!ctx)
587 		return -EINVAL;
588 
589 	ctx->va = rte_zmalloc_socket(type, size, 0,
590 				     bp->eth_dev->device->numa_node);
591 	if (ctx->va == NULL)
592 		return -ENOMEM;
593 	rte_mem_lock_page(ctx->va);
594 	ctx->size = size;
595 	ctx->dma = rte_mem_virt2iova(ctx->va);
596 	if (ctx->dma == RTE_BAD_IOVA)
597 		return -ENOMEM;
598 
599 	return 0;
600 }
601 
602 static int bnxt_init_fc_ctx_mem(struct bnxt *bp)
603 {
604 	struct rte_pci_device *pdev = bp->pdev;
605 	char type[RTE_MEMZONE_NAMESIZE];
606 	uint16_t max_fc;
607 	int rc = 0;
608 
609 	max_fc = bp->flow_stat->max_fc;
610 
611 	sprintf(type, "bnxt_rx_fc_in_" PCI_PRI_FMT, pdev->addr.domain,
612 		pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
613 	/* 4 bytes for each counter-id */
614 	rc = bnxt_alloc_ctx_mem_buf(bp, type,
615 				    max_fc * 4,
616 				    &bp->flow_stat->rx_fc_in_tbl);
617 	if (rc)
618 		return rc;
619 
620 	sprintf(type, "bnxt_rx_fc_out_" PCI_PRI_FMT, pdev->addr.domain,
621 		pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
622 	/* 16 bytes for each counter - 8 bytes pkt_count, 8 bytes byte_count */
623 	rc = bnxt_alloc_ctx_mem_buf(bp, type,
624 				    max_fc * 16,
625 				    &bp->flow_stat->rx_fc_out_tbl);
626 	if (rc)
627 		return rc;
628 
629 	sprintf(type, "bnxt_tx_fc_in_" PCI_PRI_FMT, pdev->addr.domain,
630 		pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
631 	/* 4 bytes for each counter-id */
632 	rc = bnxt_alloc_ctx_mem_buf(bp, type,
633 				    max_fc * 4,
634 				    &bp->flow_stat->tx_fc_in_tbl);
635 	if (rc)
636 		return rc;
637 
638 	sprintf(type, "bnxt_tx_fc_out_" PCI_PRI_FMT, pdev->addr.domain,
639 		pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
640 	/* 16 bytes for each counter - 8 bytes pkt_count, 8 bytes byte_count */
641 	rc = bnxt_alloc_ctx_mem_buf(bp, type,
642 				    max_fc * 16,
643 				    &bp->flow_stat->tx_fc_out_tbl);
644 	if (rc)
645 		return rc;
646 
647 	rc = bnxt_register_fc_ctx_mem(bp);
648 
649 	return rc;
650 }
651 
652 static int bnxt_init_ctx_mem(struct bnxt *bp)
653 {
654 	int rc = 0;
655 
656 	if (!(bp->fw_cap & BNXT_FW_CAP_ADV_FLOW_COUNTERS) ||
657 	    !(BNXT_PF(bp) || BNXT_VF_IS_TRUSTED(bp)) ||
658 	    !BNXT_FLOW_XSTATS_EN(bp))
659 		return 0;
660 
661 	rc = bnxt_hwrm_cfa_counter_qcaps(bp, &bp->flow_stat->max_fc);
662 	if (rc)
663 		return rc;
664 
665 	rc = bnxt_init_fc_ctx_mem(bp);
666 
667 	return rc;
668 }
669 
670 static int bnxt_update_phy_setting(struct bnxt *bp)
671 {
672 	struct rte_eth_link new;
673 	int rc;
674 
675 	rc = bnxt_get_hwrm_link_config(bp, &new);
676 	if (rc) {
677 		PMD_DRV_LOG(ERR, "Failed to get link settings\n");
678 		return rc;
679 	}
680 
681 	/*
682 	 * On BCM957508-N2100 adapters, FW will not allow any user other
683 	 * than BMC to shutdown the port. bnxt_get_hwrm_link_config() call
684 	 * always returns link up. Force phy update always in that case.
685 	 */
686 	if (!new.link_status || IS_BNXT_DEV_957508_N2100(bp)) {
687 		rc = bnxt_set_hwrm_link_config(bp, true);
688 		if (rc) {
689 			PMD_DRV_LOG(ERR, "Failed to update PHY settings\n");
690 			return rc;
691 		}
692 	}
693 
694 	return rc;
695 }
696 
697 static void bnxt_free_prev_ring_stats(struct bnxt *bp)
698 {
699 	rte_free(bp->prev_rx_ring_stats);
700 	rte_free(bp->prev_tx_ring_stats);
701 
702 	bp->prev_rx_ring_stats = NULL;
703 	bp->prev_tx_ring_stats = NULL;
704 }
705 
706 static int bnxt_alloc_prev_ring_stats(struct bnxt *bp)
707 {
708 	bp->prev_rx_ring_stats =  rte_zmalloc("bnxt_prev_rx_ring_stats",
709 					      sizeof(struct bnxt_ring_stats) *
710 					      bp->rx_cp_nr_rings,
711 					      0);
712 	if (bp->prev_rx_ring_stats == NULL)
713 		return -ENOMEM;
714 
715 	bp->prev_tx_ring_stats = rte_zmalloc("bnxt_prev_tx_ring_stats",
716 					     sizeof(struct bnxt_ring_stats) *
717 					     bp->tx_cp_nr_rings,
718 					     0);
719 	if (bp->prev_tx_ring_stats == NULL)
720 		goto error;
721 
722 	return 0;
723 
724 error:
725 	bnxt_free_prev_ring_stats(bp);
726 	return -ENOMEM;
727 }
728 
729 static int bnxt_start_nic(struct bnxt *bp)
730 {
731 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(bp->eth_dev);
732 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
733 	uint32_t intr_vector = 0;
734 	uint32_t queue_id, base = BNXT_MISC_VEC_ID;
735 	uint32_t vec = BNXT_MISC_VEC_ID;
736 	unsigned int i, j;
737 	int rc;
738 
739 	if (bp->eth_dev->data->mtu > RTE_ETHER_MTU) {
740 		bp->eth_dev->data->dev_conf.rxmode.offloads |=
741 			DEV_RX_OFFLOAD_JUMBO_FRAME;
742 		bp->flags |= BNXT_FLAG_JUMBO;
743 	} else {
744 		bp->eth_dev->data->dev_conf.rxmode.offloads &=
745 			~DEV_RX_OFFLOAD_JUMBO_FRAME;
746 		bp->flags &= ~BNXT_FLAG_JUMBO;
747 	}
748 
749 	/* THOR does not support ring groups.
750 	 * But we will use the array to save RSS context IDs.
751 	 */
752 	if (BNXT_CHIP_P5(bp))
753 		bp->max_ring_grps = BNXT_MAX_RSS_CTXTS_P5;
754 
755 	rc = bnxt_alloc_all_hwrm_stat_ctxs(bp);
756 	if (rc) {
757 		PMD_DRV_LOG(ERR, "HWRM stat ctx alloc failure rc: %x\n", rc);
758 		goto err_out;
759 	}
760 
761 	rc = bnxt_alloc_hwrm_rings(bp);
762 	if (rc) {
763 		PMD_DRV_LOG(ERR, "HWRM ring alloc failure rc: %x\n", rc);
764 		goto err_out;
765 	}
766 
767 	rc = bnxt_alloc_all_hwrm_ring_grps(bp);
768 	if (rc) {
769 		PMD_DRV_LOG(ERR, "HWRM ring grp alloc failure: %x\n", rc);
770 		goto err_out;
771 	}
772 
773 	if (!(bp->vnic_cap_flags & BNXT_VNIC_CAP_COS_CLASSIFY))
774 		goto skip_cosq_cfg;
775 
776 	for (j = 0, i = 0; i < BNXT_COS_QUEUE_COUNT; i++) {
777 		if (bp->rx_cos_queue[i].id != 0xff) {
778 			struct bnxt_vnic_info *vnic = &bp->vnic_info[j++];
779 
780 			if (!vnic) {
781 				PMD_DRV_LOG(ERR,
782 					    "Num pools more than FW profile\n");
783 				rc = -EINVAL;
784 				goto err_out;
785 			}
786 			vnic->cos_queue_id = bp->rx_cos_queue[i].id;
787 			bp->rx_cosq_cnt++;
788 		}
789 	}
790 
791 skip_cosq_cfg:
792 	rc = bnxt_mq_rx_configure(bp);
793 	if (rc) {
794 		PMD_DRV_LOG(ERR, "MQ mode configure failure rc: %x\n", rc);
795 		goto err_out;
796 	}
797 
798 	/* default vnic 0 */
799 	rc = bnxt_setup_one_vnic(bp, 0);
800 	if (rc)
801 		goto err_out;
802 	/* VNIC configuration */
803 	if (BNXT_RFS_NEEDS_VNIC(bp)) {
804 		for (i = 1; i < bp->nr_vnics; i++) {
805 			rc = bnxt_setup_one_vnic(bp, i);
806 			if (rc)
807 				goto err_out;
808 		}
809 	}
810 
811 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, &bp->vnic_info[0], 0, NULL);
812 	if (rc) {
813 		PMD_DRV_LOG(ERR,
814 			"HWRM cfa l2 rx mask failure rc: %x\n", rc);
815 		goto err_out;
816 	}
817 
818 	/* check and configure queue intr-vector mapping */
819 	if ((rte_intr_cap_multiple(intr_handle) ||
820 	     !RTE_ETH_DEV_SRIOV(bp->eth_dev).active) &&
821 	    bp->eth_dev->data->dev_conf.intr_conf.rxq != 0) {
822 		intr_vector = bp->eth_dev->data->nb_rx_queues;
823 		PMD_DRV_LOG(DEBUG, "intr_vector = %d\n", intr_vector);
824 		if (intr_vector > bp->rx_cp_nr_rings) {
825 			PMD_DRV_LOG(ERR, "At most %d intr queues supported",
826 					bp->rx_cp_nr_rings);
827 			return -ENOTSUP;
828 		}
829 		rc = rte_intr_efd_enable(intr_handle, intr_vector);
830 		if (rc)
831 			return rc;
832 	}
833 
834 	if (rte_intr_dp_is_en(intr_handle) && !intr_handle->intr_vec) {
835 		intr_handle->intr_vec =
836 			rte_zmalloc("intr_vec",
837 				    bp->eth_dev->data->nb_rx_queues *
838 				    sizeof(int), 0);
839 		if (intr_handle->intr_vec == NULL) {
840 			PMD_DRV_LOG(ERR, "Failed to allocate %d rx_queues"
841 				" intr_vec", bp->eth_dev->data->nb_rx_queues);
842 			rc = -ENOMEM;
843 			goto err_out;
844 		}
845 		PMD_DRV_LOG(DEBUG, "intr_handle->intr_vec = %p "
846 			"intr_handle->nb_efd = %d intr_handle->max_intr = %d\n",
847 			 intr_handle->intr_vec, intr_handle->nb_efd,
848 			intr_handle->max_intr);
849 		for (queue_id = 0; queue_id < bp->eth_dev->data->nb_rx_queues;
850 		     queue_id++) {
851 			intr_handle->intr_vec[queue_id] =
852 							vec + BNXT_RX_VEC_START;
853 			if (vec < base + intr_handle->nb_efd - 1)
854 				vec++;
855 		}
856 	}
857 
858 	/* enable uio/vfio intr/eventfd mapping */
859 	rc = rte_intr_enable(intr_handle);
860 #ifndef RTE_EXEC_ENV_FREEBSD
861 	/* In FreeBSD OS, nic_uio driver does not support interrupts */
862 	if (rc)
863 		goto err_out;
864 #endif
865 
866 	rc = bnxt_update_phy_setting(bp);
867 	if (rc)
868 		goto err_out;
869 
870 	bp->mark_table = rte_zmalloc("bnxt_mark_table", BNXT_MARK_TABLE_SZ, 0);
871 	if (!bp->mark_table)
872 		PMD_DRV_LOG(ERR, "Allocation of mark table failed\n");
873 
874 	return 0;
875 
876 err_out:
877 	/* Some of the error status returned by FW may not be from errno.h */
878 	if (rc > 0)
879 		rc = -EIO;
880 
881 	return rc;
882 }
883 
884 static int bnxt_shutdown_nic(struct bnxt *bp)
885 {
886 	bnxt_free_all_hwrm_resources(bp);
887 	bnxt_free_all_filters(bp);
888 	bnxt_free_all_vnics(bp);
889 	return 0;
890 }
891 
892 /*
893  * Device configuration and status function
894  */
895 
896 uint32_t bnxt_get_speed_capabilities(struct bnxt *bp)
897 {
898 	uint32_t link_speed = 0;
899 	uint32_t speed_capa = 0;
900 
901 	if (bp->link_info == NULL)
902 		return 0;
903 
904 	link_speed = bp->link_info->support_speeds;
905 
906 	/* If PAM4 is configured, use PAM4 supported speed */
907 	if (link_speed == 0 && bp->link_info->support_pam4_speeds > 0)
908 		link_speed = bp->link_info->support_pam4_speeds;
909 
910 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_LINK_SPEED_100MB)
911 		speed_capa |= ETH_LINK_SPEED_100M;
912 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_100MBHD)
913 		speed_capa |= ETH_LINK_SPEED_100M_HD;
914 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_1GB)
915 		speed_capa |= ETH_LINK_SPEED_1G;
916 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_2_5GB)
917 		speed_capa |= ETH_LINK_SPEED_2_5G;
918 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_10GB)
919 		speed_capa |= ETH_LINK_SPEED_10G;
920 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_20GB)
921 		speed_capa |= ETH_LINK_SPEED_20G;
922 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_25GB)
923 		speed_capa |= ETH_LINK_SPEED_25G;
924 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_40GB)
925 		speed_capa |= ETH_LINK_SPEED_40G;
926 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_50GB)
927 		speed_capa |= ETH_LINK_SPEED_50G;
928 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_100GB)
929 		speed_capa |= ETH_LINK_SPEED_100G;
930 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_PAM4_SPEEDS_50G)
931 		speed_capa |= ETH_LINK_SPEED_50G;
932 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_PAM4_SPEEDS_100G)
933 		speed_capa |= ETH_LINK_SPEED_100G;
934 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_PAM4_SPEEDS_200G)
935 		speed_capa |= ETH_LINK_SPEED_200G;
936 
937 	if (bp->link_info->auto_mode ==
938 	    HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_MODE_NONE)
939 		speed_capa |= ETH_LINK_SPEED_FIXED;
940 
941 	return speed_capa;
942 }
943 
944 static int bnxt_dev_info_get_op(struct rte_eth_dev *eth_dev,
945 				struct rte_eth_dev_info *dev_info)
946 {
947 	struct rte_pci_device *pdev = RTE_DEV_TO_PCI(eth_dev->device);
948 	struct bnxt *bp = eth_dev->data->dev_private;
949 	uint16_t max_vnics, i, j, vpool, vrxq;
950 	unsigned int max_rx_rings;
951 	int rc;
952 
953 	rc = is_bnxt_in_error(bp);
954 	if (rc)
955 		return rc;
956 
957 	/* MAC Specifics */
958 	dev_info->max_mac_addrs = bp->max_l2_ctx;
959 	dev_info->max_hash_mac_addrs = 0;
960 
961 	/* PF/VF specifics */
962 	if (BNXT_PF(bp))
963 		dev_info->max_vfs = pdev->max_vfs;
964 
965 	max_rx_rings = bnxt_max_rings(bp);
966 	/* For the sake of symmetry, max_rx_queues = max_tx_queues */
967 	dev_info->max_rx_queues = max_rx_rings;
968 	dev_info->max_tx_queues = max_rx_rings;
969 	dev_info->reta_size = bnxt_rss_hash_tbl_size(bp);
970 	dev_info->hash_key_size = HW_HASH_KEY_SIZE;
971 	max_vnics = bp->max_vnics;
972 
973 	/* MTU specifics */
974 	dev_info->min_mtu = RTE_ETHER_MIN_MTU;
975 	dev_info->max_mtu = BNXT_MAX_MTU;
976 
977 	/* Fast path specifics */
978 	dev_info->min_rx_bufsize = 1;
979 	dev_info->max_rx_pktlen = BNXT_MAX_PKT_LEN;
980 
981 	dev_info->rx_offload_capa = BNXT_DEV_RX_OFFLOAD_SUPPORT;
982 	if (bp->flags & BNXT_FLAG_PTP_SUPPORTED)
983 		dev_info->rx_offload_capa |= DEV_RX_OFFLOAD_TIMESTAMP;
984 	dev_info->tx_queue_offload_capa = DEV_TX_OFFLOAD_MBUF_FAST_FREE;
985 	dev_info->tx_offload_capa = BNXT_DEV_TX_OFFLOAD_SUPPORT |
986 				    dev_info->tx_queue_offload_capa;
987 	dev_info->flow_type_rss_offloads = BNXT_ETH_RSS_SUPPORT;
988 
989 	dev_info->speed_capa = bnxt_get_speed_capabilities(bp);
990 	dev_info->dev_capa = RTE_ETH_DEV_CAPA_RUNTIME_RX_QUEUE_SETUP |
991 			     RTE_ETH_DEV_CAPA_RUNTIME_TX_QUEUE_SETUP;
992 
993 	dev_info->default_rxconf = (struct rte_eth_rxconf) {
994 		.rx_thresh = {
995 			.pthresh = 8,
996 			.hthresh = 8,
997 			.wthresh = 0,
998 		},
999 		.rx_free_thresh = 32,
1000 		.rx_drop_en = BNXT_DEFAULT_RX_DROP_EN,
1001 	};
1002 
1003 	dev_info->default_txconf = (struct rte_eth_txconf) {
1004 		.tx_thresh = {
1005 			.pthresh = 32,
1006 			.hthresh = 0,
1007 			.wthresh = 0,
1008 		},
1009 		.tx_free_thresh = 32,
1010 		.tx_rs_thresh = 32,
1011 	};
1012 	eth_dev->data->dev_conf.intr_conf.lsc = 1;
1013 
1014 	dev_info->rx_desc_lim.nb_min = BNXT_MIN_RING_DESC;
1015 	dev_info->rx_desc_lim.nb_max = BNXT_MAX_RX_RING_DESC;
1016 	dev_info->tx_desc_lim.nb_min = BNXT_MIN_RING_DESC;
1017 	dev_info->tx_desc_lim.nb_max = BNXT_MAX_TX_RING_DESC;
1018 
1019 	if (BNXT_PF(bp) || BNXT_VF_IS_TRUSTED(bp)) {
1020 		dev_info->switch_info.name = eth_dev->device->name;
1021 		dev_info->switch_info.domain_id = bp->switch_domain_id;
1022 		dev_info->switch_info.port_id =
1023 				BNXT_PF(bp) ? BNXT_SWITCH_PORT_ID_PF :
1024 				    BNXT_SWITCH_PORT_ID_TRUSTED_VF;
1025 	}
1026 
1027 	/*
1028 	 * TODO: default_rxconf, default_txconf, rx_desc_lim, and tx_desc_lim
1029 	 *       need further investigation.
1030 	 */
1031 
1032 	/* VMDq resources */
1033 	vpool = 64; /* ETH_64_POOLS */
1034 	vrxq = 128; /* ETH_VMDQ_DCB_NUM_QUEUES */
1035 	for (i = 0; i < 4; vpool >>= 1, i++) {
1036 		if (max_vnics > vpool) {
1037 			for (j = 0; j < 5; vrxq >>= 1, j++) {
1038 				if (dev_info->max_rx_queues > vrxq) {
1039 					if (vpool > vrxq)
1040 						vpool = vrxq;
1041 					goto found;
1042 				}
1043 			}
1044 			/* Not enough resources to support VMDq */
1045 			break;
1046 		}
1047 	}
1048 	/* Not enough resources to support VMDq */
1049 	vpool = 0;
1050 	vrxq = 0;
1051 found:
1052 	dev_info->max_vmdq_pools = vpool;
1053 	dev_info->vmdq_queue_num = vrxq;
1054 
1055 	dev_info->vmdq_pool_base = 0;
1056 	dev_info->vmdq_queue_base = 0;
1057 
1058 	return 0;
1059 }
1060 
1061 /* Configure the device based on the configuration provided */
1062 static int bnxt_dev_configure_op(struct rte_eth_dev *eth_dev)
1063 {
1064 	struct bnxt *bp = eth_dev->data->dev_private;
1065 	uint64_t rx_offloads = eth_dev->data->dev_conf.rxmode.offloads;
1066 	int rc;
1067 
1068 	bp->rx_queues = (void *)eth_dev->data->rx_queues;
1069 	bp->tx_queues = (void *)eth_dev->data->tx_queues;
1070 	bp->tx_nr_rings = eth_dev->data->nb_tx_queues;
1071 	bp->rx_nr_rings = eth_dev->data->nb_rx_queues;
1072 
1073 	rc = is_bnxt_in_error(bp);
1074 	if (rc)
1075 		return rc;
1076 
1077 	if (BNXT_VF(bp) && (bp->flags & BNXT_FLAG_NEW_RM)) {
1078 		rc = bnxt_hwrm_check_vf_rings(bp);
1079 		if (rc) {
1080 			PMD_DRV_LOG(ERR, "HWRM insufficient resources\n");
1081 			return -ENOSPC;
1082 		}
1083 
1084 		/* If a resource has already been allocated - in this case
1085 		 * it is the async completion ring, free it. Reallocate it after
1086 		 * resource reservation. This will ensure the resource counts
1087 		 * are calculated correctly.
1088 		 */
1089 
1090 		pthread_mutex_lock(&bp->def_cp_lock);
1091 
1092 		if (!BNXT_HAS_NQ(bp) && bp->async_cp_ring) {
1093 			bnxt_disable_int(bp);
1094 			bnxt_free_cp_ring(bp, bp->async_cp_ring);
1095 		}
1096 
1097 		rc = bnxt_hwrm_func_reserve_vf_resc(bp, false);
1098 		if (rc) {
1099 			PMD_DRV_LOG(ERR, "HWRM resource alloc fail:%x\n", rc);
1100 			pthread_mutex_unlock(&bp->def_cp_lock);
1101 			return -ENOSPC;
1102 		}
1103 
1104 		if (!BNXT_HAS_NQ(bp) && bp->async_cp_ring) {
1105 			rc = bnxt_alloc_async_cp_ring(bp);
1106 			if (rc) {
1107 				pthread_mutex_unlock(&bp->def_cp_lock);
1108 				return rc;
1109 			}
1110 			bnxt_enable_int(bp);
1111 		}
1112 
1113 		pthread_mutex_unlock(&bp->def_cp_lock);
1114 	}
1115 
1116 	/* Inherit new configurations */
1117 	if (eth_dev->data->nb_rx_queues > bp->max_rx_rings ||
1118 	    eth_dev->data->nb_tx_queues > bp->max_tx_rings ||
1119 	    eth_dev->data->nb_rx_queues + eth_dev->data->nb_tx_queues
1120 		+ BNXT_NUM_ASYNC_CPR(bp) > bp->max_cp_rings ||
1121 	    eth_dev->data->nb_rx_queues + eth_dev->data->nb_tx_queues >
1122 	    bp->max_stat_ctx)
1123 		goto resource_error;
1124 
1125 	if (BNXT_HAS_RING_GRPS(bp) &&
1126 	    (uint32_t)(eth_dev->data->nb_rx_queues) > bp->max_ring_grps)
1127 		goto resource_error;
1128 
1129 	if (!(eth_dev->data->dev_conf.rxmode.mq_mode & ETH_MQ_RX_RSS) &&
1130 	    bp->max_vnics < eth_dev->data->nb_rx_queues)
1131 		goto resource_error;
1132 
1133 	bp->rx_cp_nr_rings = bp->rx_nr_rings;
1134 	bp->tx_cp_nr_rings = bp->tx_nr_rings;
1135 
1136 	if (eth_dev->data->dev_conf.rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG)
1137 		rx_offloads |= DEV_RX_OFFLOAD_RSS_HASH;
1138 	eth_dev->data->dev_conf.rxmode.offloads = rx_offloads;
1139 
1140 	if (rx_offloads & DEV_RX_OFFLOAD_JUMBO_FRAME) {
1141 		eth_dev->data->mtu =
1142 			eth_dev->data->dev_conf.rxmode.max_rx_pkt_len -
1143 			RTE_ETHER_HDR_LEN - RTE_ETHER_CRC_LEN - VLAN_TAG_SIZE *
1144 			BNXT_NUM_VLANS;
1145 		bnxt_mtu_set_op(eth_dev, eth_dev->data->mtu);
1146 	}
1147 	return 0;
1148 
1149 resource_error:
1150 	PMD_DRV_LOG(ERR,
1151 		    "Insufficient resources to support requested config\n");
1152 	PMD_DRV_LOG(ERR,
1153 		    "Num Queues Requested: Tx %d, Rx %d\n",
1154 		    eth_dev->data->nb_tx_queues,
1155 		    eth_dev->data->nb_rx_queues);
1156 	PMD_DRV_LOG(ERR,
1157 		    "MAX: TxQ %d, RxQ %d, CQ %d Stat %d, Grp %d, Vnic %d\n",
1158 		    bp->max_tx_rings, bp->max_rx_rings, bp->max_cp_rings,
1159 		    bp->max_stat_ctx, bp->max_ring_grps, bp->max_vnics);
1160 	return -ENOSPC;
1161 }
1162 
1163 void bnxt_print_link_info(struct rte_eth_dev *eth_dev)
1164 {
1165 	struct rte_eth_link *link = &eth_dev->data->dev_link;
1166 
1167 	if (link->link_status)
1168 		PMD_DRV_LOG(INFO, "Port %d Link Up - speed %u Mbps - %s\n",
1169 			eth_dev->data->port_id,
1170 			(uint32_t)link->link_speed,
1171 			(link->link_duplex == ETH_LINK_FULL_DUPLEX) ?
1172 			("full-duplex") : ("half-duplex\n"));
1173 	else
1174 		PMD_DRV_LOG(INFO, "Port %d Link Down\n",
1175 			eth_dev->data->port_id);
1176 }
1177 
1178 /*
1179  * Determine whether the current configuration requires support for scattered
1180  * receive; return 1 if scattered receive is required and 0 if not.
1181  */
1182 static int bnxt_scattered_rx(struct rte_eth_dev *eth_dev)
1183 {
1184 	uint16_t buf_size;
1185 	int i;
1186 
1187 	if (eth_dev->data->dev_conf.rxmode.offloads & DEV_RX_OFFLOAD_SCATTER)
1188 		return 1;
1189 
1190 	if (eth_dev->data->dev_conf.rxmode.offloads & DEV_RX_OFFLOAD_TCP_LRO)
1191 		return 1;
1192 
1193 	for (i = 0; i < eth_dev->data->nb_rx_queues; i++) {
1194 		struct bnxt_rx_queue *rxq = eth_dev->data->rx_queues[i];
1195 
1196 		buf_size = (uint16_t)(rte_pktmbuf_data_room_size(rxq->mb_pool) -
1197 				      RTE_PKTMBUF_HEADROOM);
1198 		if (eth_dev->data->dev_conf.rxmode.max_rx_pkt_len > buf_size)
1199 			return 1;
1200 	}
1201 	return 0;
1202 }
1203 
1204 static eth_rx_burst_t
1205 bnxt_receive_function(struct rte_eth_dev *eth_dev)
1206 {
1207 	struct bnxt *bp = eth_dev->data->dev_private;
1208 
1209 	/* Disable vector mode RX for Stingray2 for now */
1210 	if (BNXT_CHIP_SR2(bp)) {
1211 		bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
1212 		return bnxt_recv_pkts;
1213 	}
1214 
1215 #if (defined(RTE_ARCH_X86) || defined(RTE_ARCH_ARM64)) && \
1216 	!defined(RTE_LIBRTE_IEEE1588)
1217 
1218 	/* Vector mode receive cannot be enabled if scattered rx is in use. */
1219 	if (eth_dev->data->scattered_rx)
1220 		goto use_scalar_rx;
1221 
1222 	/*
1223 	 * Vector mode receive cannot be enabled if Truflow is enabled or if
1224 	 * asynchronous completions and receive completions can be placed in
1225 	 * the same completion ring.
1226 	 */
1227 	if (BNXT_TRUFLOW_EN(bp) || !BNXT_NUM_ASYNC_CPR(bp))
1228 		goto use_scalar_rx;
1229 
1230 	/*
1231 	 * Vector mode receive cannot be enabled if any receive offloads outside
1232 	 * a limited subset have been enabled.
1233 	 */
1234 	if (eth_dev->data->dev_conf.rxmode.offloads &
1235 		~(DEV_RX_OFFLOAD_VLAN_STRIP |
1236 		  DEV_RX_OFFLOAD_KEEP_CRC |
1237 		  DEV_RX_OFFLOAD_JUMBO_FRAME |
1238 		  DEV_RX_OFFLOAD_IPV4_CKSUM |
1239 		  DEV_RX_OFFLOAD_UDP_CKSUM |
1240 		  DEV_RX_OFFLOAD_TCP_CKSUM |
1241 		  DEV_RX_OFFLOAD_OUTER_IPV4_CKSUM |
1242 		  DEV_RX_OFFLOAD_OUTER_UDP_CKSUM |
1243 		  DEV_RX_OFFLOAD_RSS_HASH |
1244 		  DEV_RX_OFFLOAD_VLAN_FILTER))
1245 		goto use_scalar_rx;
1246 
1247 #if defined(RTE_ARCH_X86) && defined(CC_AVX2_SUPPORT)
1248 	if (rte_vect_get_max_simd_bitwidth() >= RTE_VECT_SIMD_256 &&
1249 	    rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX2) == 1) {
1250 		PMD_DRV_LOG(INFO,
1251 			    "Using AVX2 vector mode receive for port %d\n",
1252 			    eth_dev->data->port_id);
1253 		bp->flags |= BNXT_FLAG_RX_VECTOR_PKT_MODE;
1254 		return bnxt_recv_pkts_vec_avx2;
1255 	}
1256  #endif
1257 	if (rte_vect_get_max_simd_bitwidth() >= RTE_VECT_SIMD_128) {
1258 		PMD_DRV_LOG(INFO,
1259 			    "Using SSE vector mode receive for port %d\n",
1260 			    eth_dev->data->port_id);
1261 		bp->flags |= BNXT_FLAG_RX_VECTOR_PKT_MODE;
1262 		return bnxt_recv_pkts_vec;
1263 	}
1264 
1265 use_scalar_rx:
1266 	PMD_DRV_LOG(INFO, "Vector mode receive disabled for port %d\n",
1267 		    eth_dev->data->port_id);
1268 	PMD_DRV_LOG(INFO,
1269 		    "Port %d scatter: %d rx offload: %" PRIX64 "\n",
1270 		    eth_dev->data->port_id,
1271 		    eth_dev->data->scattered_rx,
1272 		    eth_dev->data->dev_conf.rxmode.offloads);
1273 #endif
1274 	bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
1275 	return bnxt_recv_pkts;
1276 }
1277 
1278 static eth_tx_burst_t
1279 bnxt_transmit_function(struct rte_eth_dev *eth_dev)
1280 {
1281 	struct bnxt *bp = eth_dev->data->dev_private;
1282 
1283 	/* Disable vector mode TX for Stingray2 for now */
1284 	if (BNXT_CHIP_SR2(bp))
1285 		return bnxt_xmit_pkts;
1286 
1287 #if defined(RTE_ARCH_X86) || defined(RTE_ARCH_ARM64) && \
1288 	!defined(RTE_LIBRTE_IEEE1588)
1289 	uint64_t offloads = eth_dev->data->dev_conf.txmode.offloads;
1290 
1291 	/*
1292 	 * Vector mode transmit can be enabled only if not using scatter rx
1293 	 * or tx offloads.
1294 	 */
1295 	if (eth_dev->data->scattered_rx ||
1296 	    (offloads & ~DEV_TX_OFFLOAD_MBUF_FAST_FREE) ||
1297 	    BNXT_TRUFLOW_EN(bp))
1298 		goto use_scalar_tx;
1299 
1300 #if defined(RTE_ARCH_X86) && defined(CC_AVX2_SUPPORT)
1301 	if (rte_vect_get_max_simd_bitwidth() >= RTE_VECT_SIMD_256 &&
1302 	    rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX2) == 1) {
1303 		PMD_DRV_LOG(INFO,
1304 			    "Using AVX2 vector mode transmit for port %d\n",
1305 			    eth_dev->data->port_id);
1306 		return bnxt_xmit_pkts_vec_avx2;
1307 	}
1308 #endif
1309 	if (rte_vect_get_max_simd_bitwidth() >= RTE_VECT_SIMD_128) {
1310 		PMD_DRV_LOG(INFO,
1311 			    "Using SSE vector mode transmit for port %d\n",
1312 			    eth_dev->data->port_id);
1313 		return bnxt_xmit_pkts_vec;
1314 	}
1315 
1316 use_scalar_tx:
1317 	PMD_DRV_LOG(INFO, "Vector mode transmit disabled for port %d\n",
1318 		    eth_dev->data->port_id);
1319 	PMD_DRV_LOG(INFO,
1320 		    "Port %d scatter: %d tx offload: %" PRIX64 "\n",
1321 		    eth_dev->data->port_id,
1322 		    eth_dev->data->scattered_rx,
1323 		    offloads);
1324 #endif
1325 	return bnxt_xmit_pkts;
1326 }
1327 
1328 static int bnxt_handle_if_change_status(struct bnxt *bp)
1329 {
1330 	int rc;
1331 
1332 	/* Since fw has undergone a reset and lost all contexts,
1333 	 * set fatal flag to not issue hwrm during cleanup
1334 	 */
1335 	bp->flags |= BNXT_FLAG_FATAL_ERROR;
1336 	bnxt_uninit_resources(bp, true);
1337 
1338 	/* clear fatal flag so that re-init happens */
1339 	bp->flags &= ~BNXT_FLAG_FATAL_ERROR;
1340 	rc = bnxt_init_resources(bp, true);
1341 
1342 	bp->flags &= ~BNXT_FLAG_IF_CHANGE_HOT_FW_RESET_DONE;
1343 
1344 	return rc;
1345 }
1346 
1347 static int bnxt_dev_set_link_up_op(struct rte_eth_dev *eth_dev)
1348 {
1349 	struct bnxt *bp = eth_dev->data->dev_private;
1350 	int rc = 0;
1351 
1352 	if (!BNXT_SINGLE_PF(bp))
1353 		return -ENOTSUP;
1354 
1355 	if (!bp->link_info->link_up)
1356 		rc = bnxt_set_hwrm_link_config(bp, true);
1357 	if (!rc)
1358 		eth_dev->data->dev_link.link_status = 1;
1359 
1360 	bnxt_print_link_info(eth_dev);
1361 	return rc;
1362 }
1363 
1364 static int bnxt_dev_set_link_down_op(struct rte_eth_dev *eth_dev)
1365 {
1366 	struct bnxt *bp = eth_dev->data->dev_private;
1367 
1368 	if (!BNXT_SINGLE_PF(bp))
1369 		return -ENOTSUP;
1370 
1371 	eth_dev->data->dev_link.link_status = 0;
1372 	bnxt_set_hwrm_link_config(bp, false);
1373 	bp->link_info->link_up = 0;
1374 
1375 	return 0;
1376 }
1377 
1378 static void bnxt_free_switch_domain(struct bnxt *bp)
1379 {
1380 	int rc = 0;
1381 
1382 	if (!(BNXT_PF(bp) || BNXT_VF_IS_TRUSTED(bp)))
1383 		return;
1384 
1385 	rc = rte_eth_switch_domain_free(bp->switch_domain_id);
1386 	if (rc)
1387 		PMD_DRV_LOG(ERR, "free switch domain:%d fail: %d\n",
1388 			    bp->switch_domain_id, rc);
1389 }
1390 
1391 static void bnxt_ptp_get_current_time(void *arg)
1392 {
1393 	struct bnxt *bp = arg;
1394 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
1395 	int rc;
1396 
1397 	rc = is_bnxt_in_error(bp);
1398 	if (rc)
1399 		return;
1400 
1401 	if (!ptp)
1402 		return;
1403 
1404 	bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_CURRENT_TIME,
1405 				&ptp->current_time);
1406 
1407 	rc = rte_eal_alarm_set(US_PER_S, bnxt_ptp_get_current_time, (void *)bp);
1408 	if (rc != 0) {
1409 		PMD_DRV_LOG(ERR, "Failed to re-schedule PTP alarm\n");
1410 		bp->flags2 &= ~BNXT_FLAGS2_PTP_ALARM_SCHEDULED;
1411 	}
1412 }
1413 
1414 static int bnxt_schedule_ptp_alarm(struct bnxt *bp)
1415 {
1416 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
1417 	int rc;
1418 
1419 	if (bp->flags2 & BNXT_FLAGS2_PTP_ALARM_SCHEDULED)
1420 		return 0;
1421 
1422 	bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_CURRENT_TIME,
1423 				&ptp->current_time);
1424 
1425 	rc = rte_eal_alarm_set(US_PER_S, bnxt_ptp_get_current_time, (void *)bp);
1426 	return rc;
1427 }
1428 
1429 static void bnxt_cancel_ptp_alarm(struct bnxt *bp)
1430 {
1431 	if (bp->flags2 & BNXT_FLAGS2_PTP_ALARM_SCHEDULED) {
1432 		rte_eal_alarm_cancel(bnxt_ptp_get_current_time, (void *)bp);
1433 		bp->flags2 &= ~BNXT_FLAGS2_PTP_ALARM_SCHEDULED;
1434 	}
1435 }
1436 
1437 static void bnxt_ptp_stop(struct bnxt *bp)
1438 {
1439 	bnxt_cancel_ptp_alarm(bp);
1440 	bp->flags2 &= ~BNXT_FLAGS2_PTP_TIMESYNC_ENABLED;
1441 }
1442 
1443 static int bnxt_ptp_start(struct bnxt *bp)
1444 {
1445 	int rc;
1446 
1447 	rc = bnxt_schedule_ptp_alarm(bp);
1448 	if (rc != 0) {
1449 		PMD_DRV_LOG(ERR, "Failed to schedule PTP alarm\n");
1450 	} else {
1451 		bp->flags2 |= BNXT_FLAGS2_PTP_TIMESYNC_ENABLED;
1452 		bp->flags2 |= BNXT_FLAGS2_PTP_ALARM_SCHEDULED;
1453 	}
1454 
1455 	return rc;
1456 }
1457 
1458 static int bnxt_dev_stop(struct rte_eth_dev *eth_dev)
1459 {
1460 	struct bnxt *bp = eth_dev->data->dev_private;
1461 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
1462 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
1463 	struct rte_eth_link link;
1464 	int ret;
1465 
1466 	eth_dev->data->dev_started = 0;
1467 	eth_dev->data->scattered_rx = 0;
1468 
1469 	/* Prevent crashes when queues are still in use */
1470 	eth_dev->rx_pkt_burst = &bnxt_dummy_recv_pkts;
1471 	eth_dev->tx_pkt_burst = &bnxt_dummy_xmit_pkts;
1472 
1473 	bnxt_disable_int(bp);
1474 
1475 	/* disable uio/vfio intr/eventfd mapping */
1476 	rte_intr_disable(intr_handle);
1477 
1478 	/* Stop the child representors for this device */
1479 	ret = bnxt_rep_stop_all(bp);
1480 	if (ret != 0)
1481 		return ret;
1482 
1483 	/* delete the bnxt ULP port details */
1484 	bnxt_ulp_port_deinit(bp);
1485 
1486 	bnxt_cancel_fw_health_check(bp);
1487 
1488 	if (BNXT_P5_PTP_TIMESYNC_ENABLED(bp))
1489 		bnxt_cancel_ptp_alarm(bp);
1490 
1491 	/* Do not bring link down during reset recovery */
1492 	if (!is_bnxt_in_error(bp)) {
1493 		bnxt_dev_set_link_down_op(eth_dev);
1494 		/* Wait for link to be reset */
1495 		if (BNXT_SINGLE_PF(bp))
1496 			rte_delay_ms(500);
1497 		/* clear the recorded link status */
1498 		memset(&link, 0, sizeof(link));
1499 		rte_eth_linkstatus_set(eth_dev, &link);
1500 	}
1501 
1502 	/* Clean queue intr-vector mapping */
1503 	rte_intr_efd_disable(intr_handle);
1504 	if (intr_handle->intr_vec != NULL) {
1505 		rte_free(intr_handle->intr_vec);
1506 		intr_handle->intr_vec = NULL;
1507 	}
1508 
1509 	bnxt_hwrm_port_clr_stats(bp);
1510 	bnxt_free_tx_mbufs(bp);
1511 	bnxt_free_rx_mbufs(bp);
1512 	/* Process any remaining notifications in default completion queue */
1513 	bnxt_int_handler(eth_dev);
1514 	bnxt_shutdown_nic(bp);
1515 	bnxt_hwrm_if_change(bp, false);
1516 
1517 	bnxt_free_prev_ring_stats(bp);
1518 	rte_free(bp->mark_table);
1519 	bp->mark_table = NULL;
1520 
1521 	bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
1522 	bp->rx_cosq_cnt = 0;
1523 	/* All filters are deleted on a port stop. */
1524 	if (BNXT_FLOW_XSTATS_EN(bp))
1525 		bp->flow_stat->flow_count = 0;
1526 
1527 	return 0;
1528 }
1529 
1530 /* Unload the driver, release resources */
1531 static int bnxt_dev_stop_op(struct rte_eth_dev *eth_dev)
1532 {
1533 	struct bnxt *bp = eth_dev->data->dev_private;
1534 
1535 	pthread_mutex_lock(&bp->err_recovery_lock);
1536 	if (bp->flags & BNXT_FLAG_FW_RESET) {
1537 		PMD_DRV_LOG(ERR,
1538 			    "Adapter recovering from error..Please retry\n");
1539 		pthread_mutex_unlock(&bp->err_recovery_lock);
1540 		return -EAGAIN;
1541 	}
1542 	pthread_mutex_unlock(&bp->err_recovery_lock);
1543 
1544 	return bnxt_dev_stop(eth_dev);
1545 }
1546 
1547 static int bnxt_dev_start_op(struct rte_eth_dev *eth_dev)
1548 {
1549 	struct bnxt *bp = eth_dev->data->dev_private;
1550 	uint64_t rx_offloads = eth_dev->data->dev_conf.rxmode.offloads;
1551 	int vlan_mask = 0;
1552 	int rc, retry_cnt = BNXT_IF_CHANGE_RETRY_COUNT;
1553 
1554 	if (!eth_dev->data->nb_tx_queues || !eth_dev->data->nb_rx_queues) {
1555 		PMD_DRV_LOG(ERR, "Queues are not configured yet!\n");
1556 		return -EINVAL;
1557 	}
1558 
1559 	if (bp->rx_cp_nr_rings > RTE_ETHDEV_QUEUE_STAT_CNTRS)
1560 		PMD_DRV_LOG(ERR,
1561 			    "RxQ cnt %d > RTE_ETHDEV_QUEUE_STAT_CNTRS %d\n",
1562 			    bp->rx_cp_nr_rings, RTE_ETHDEV_QUEUE_STAT_CNTRS);
1563 
1564 	do {
1565 		rc = bnxt_hwrm_if_change(bp, true);
1566 		if (rc == 0 || rc != -EAGAIN)
1567 			break;
1568 
1569 		rte_delay_ms(BNXT_IF_CHANGE_RETRY_INTERVAL);
1570 	} while (retry_cnt--);
1571 
1572 	if (rc)
1573 		return rc;
1574 
1575 	if (bp->flags & BNXT_FLAG_IF_CHANGE_HOT_FW_RESET_DONE) {
1576 		rc = bnxt_handle_if_change_status(bp);
1577 		if (rc)
1578 			return rc;
1579 	}
1580 
1581 	bnxt_enable_int(bp);
1582 
1583 	eth_dev->data->scattered_rx = bnxt_scattered_rx(eth_dev);
1584 
1585 	rc = bnxt_start_nic(bp);
1586 	if (rc)
1587 		goto error;
1588 
1589 	rc = bnxt_alloc_prev_ring_stats(bp);
1590 	if (rc)
1591 		goto error;
1592 
1593 	eth_dev->data->dev_started = 1;
1594 
1595 	bnxt_link_update_op(eth_dev, 1);
1596 
1597 	if (rx_offloads & DEV_RX_OFFLOAD_VLAN_FILTER)
1598 		vlan_mask |= ETH_VLAN_FILTER_MASK;
1599 	if (rx_offloads & DEV_RX_OFFLOAD_VLAN_STRIP)
1600 		vlan_mask |= ETH_VLAN_STRIP_MASK;
1601 	rc = bnxt_vlan_offload_set_op(eth_dev, vlan_mask);
1602 	if (rc)
1603 		goto error;
1604 
1605 	/* Initialize bnxt ULP port details */
1606 	rc = bnxt_ulp_port_init(bp);
1607 	if (rc)
1608 		goto error;
1609 
1610 	eth_dev->rx_pkt_burst = bnxt_receive_function(eth_dev);
1611 	eth_dev->tx_pkt_burst = bnxt_transmit_function(eth_dev);
1612 
1613 	bnxt_schedule_fw_health_check(bp);
1614 
1615 	if (BNXT_P5_PTP_TIMESYNC_ENABLED(bp))
1616 		bnxt_schedule_ptp_alarm(bp);
1617 
1618 	return 0;
1619 
1620 error:
1621 	bnxt_dev_stop(eth_dev);
1622 	return rc;
1623 }
1624 
1625 static void
1626 bnxt_uninit_locks(struct bnxt *bp)
1627 {
1628 	pthread_mutex_destroy(&bp->flow_lock);
1629 	pthread_mutex_destroy(&bp->def_cp_lock);
1630 	pthread_mutex_destroy(&bp->health_check_lock);
1631 	pthread_mutex_destroy(&bp->err_recovery_lock);
1632 	if (bp->rep_info) {
1633 		pthread_mutex_destroy(&bp->rep_info->vfr_lock);
1634 		pthread_mutex_destroy(&bp->rep_info->vfr_start_lock);
1635 	}
1636 }
1637 
1638 static void bnxt_drv_uninit(struct bnxt *bp)
1639 {
1640 	bnxt_free_leds_info(bp);
1641 	bnxt_free_cos_queues(bp);
1642 	bnxt_free_link_info(bp);
1643 	bnxt_free_parent_info(bp);
1644 	bnxt_uninit_locks(bp);
1645 
1646 	rte_memzone_free((const struct rte_memzone *)bp->tx_mem_zone);
1647 	bp->tx_mem_zone = NULL;
1648 	rte_memzone_free((const struct rte_memzone *)bp->rx_mem_zone);
1649 	bp->rx_mem_zone = NULL;
1650 
1651 	bnxt_free_vf_info(bp);
1652 	bnxt_free_pf_info(bp);
1653 
1654 	rte_free(bp->grp_info);
1655 	bp->grp_info = NULL;
1656 }
1657 
1658 static int bnxt_dev_close_op(struct rte_eth_dev *eth_dev)
1659 {
1660 	struct bnxt *bp = eth_dev->data->dev_private;
1661 	int ret = 0;
1662 
1663 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
1664 		return 0;
1665 
1666 	pthread_mutex_lock(&bp->err_recovery_lock);
1667 	if (bp->flags & BNXT_FLAG_FW_RESET) {
1668 		PMD_DRV_LOG(ERR,
1669 			    "Adapter recovering from error...Please retry\n");
1670 		pthread_mutex_unlock(&bp->err_recovery_lock);
1671 		return -EAGAIN;
1672 	}
1673 	pthread_mutex_unlock(&bp->err_recovery_lock);
1674 
1675 	/* cancel the recovery handler before remove dev */
1676 	rte_eal_alarm_cancel(bnxt_dev_reset_and_resume, (void *)bp);
1677 	rte_eal_alarm_cancel(bnxt_dev_recover, (void *)bp);
1678 	bnxt_cancel_fc_thread(bp);
1679 
1680 	if (eth_dev->data->dev_started)
1681 		ret = bnxt_dev_stop(eth_dev);
1682 
1683 	bnxt_uninit_resources(bp, false);
1684 
1685 	bnxt_drv_uninit(bp);
1686 
1687 	return ret;
1688 }
1689 
1690 static void bnxt_mac_addr_remove_op(struct rte_eth_dev *eth_dev,
1691 				    uint32_t index)
1692 {
1693 	struct bnxt *bp = eth_dev->data->dev_private;
1694 	uint64_t pool_mask = eth_dev->data->mac_pool_sel[index];
1695 	struct bnxt_vnic_info *vnic;
1696 	struct bnxt_filter_info *filter, *temp_filter;
1697 	uint32_t i;
1698 
1699 	if (is_bnxt_in_error(bp))
1700 		return;
1701 
1702 	/*
1703 	 * Loop through all VNICs from the specified filter flow pools to
1704 	 * remove the corresponding MAC addr filter
1705 	 */
1706 	for (i = 0; i < bp->nr_vnics; i++) {
1707 		if (!(pool_mask & (1ULL << i)))
1708 			continue;
1709 
1710 		vnic = &bp->vnic_info[i];
1711 		filter = STAILQ_FIRST(&vnic->filter);
1712 		while (filter) {
1713 			temp_filter = STAILQ_NEXT(filter, next);
1714 			if (filter->mac_index == index) {
1715 				STAILQ_REMOVE(&vnic->filter, filter,
1716 						bnxt_filter_info, next);
1717 				bnxt_hwrm_clear_l2_filter(bp, filter);
1718 				bnxt_free_filter(bp, filter);
1719 			}
1720 			filter = temp_filter;
1721 		}
1722 	}
1723 }
1724 
1725 static int bnxt_add_mac_filter(struct bnxt *bp, struct bnxt_vnic_info *vnic,
1726 			       struct rte_ether_addr *mac_addr, uint32_t index,
1727 			       uint32_t pool)
1728 {
1729 	struct bnxt_filter_info *filter;
1730 	int rc = 0;
1731 
1732 	/* Attach requested MAC address to the new l2_filter */
1733 	STAILQ_FOREACH(filter, &vnic->filter, next) {
1734 		if (filter->mac_index == index) {
1735 			PMD_DRV_LOG(DEBUG,
1736 				    "MAC addr already existed for pool %d\n",
1737 				    pool);
1738 			return 0;
1739 		}
1740 	}
1741 
1742 	filter = bnxt_alloc_filter(bp);
1743 	if (!filter) {
1744 		PMD_DRV_LOG(ERR, "L2 filter alloc failed\n");
1745 		return -ENODEV;
1746 	}
1747 
1748 	/* bnxt_alloc_filter copies default MAC to filter->l2_addr. So,
1749 	 * if the MAC that's been programmed now is a different one, then,
1750 	 * copy that addr to filter->l2_addr
1751 	 */
1752 	if (mac_addr)
1753 		memcpy(filter->l2_addr, mac_addr, RTE_ETHER_ADDR_LEN);
1754 	filter->flags |= HWRM_CFA_L2_FILTER_ALLOC_INPUT_FLAGS_OUTERMOST;
1755 
1756 	rc = bnxt_hwrm_set_l2_filter(bp, vnic->fw_vnic_id, filter);
1757 	if (!rc) {
1758 		filter->mac_index = index;
1759 		if (filter->mac_index == 0)
1760 			STAILQ_INSERT_HEAD(&vnic->filter, filter, next);
1761 		else
1762 			STAILQ_INSERT_TAIL(&vnic->filter, filter, next);
1763 	} else {
1764 		bnxt_free_filter(bp, filter);
1765 	}
1766 
1767 	return rc;
1768 }
1769 
1770 static int bnxt_mac_addr_add_op(struct rte_eth_dev *eth_dev,
1771 				struct rte_ether_addr *mac_addr,
1772 				uint32_t index, uint32_t pool)
1773 {
1774 	struct bnxt *bp = eth_dev->data->dev_private;
1775 	struct bnxt_vnic_info *vnic = &bp->vnic_info[pool];
1776 	int rc = 0;
1777 
1778 	rc = is_bnxt_in_error(bp);
1779 	if (rc)
1780 		return rc;
1781 
1782 	if (BNXT_VF(bp) && !BNXT_VF_IS_TRUSTED(bp)) {
1783 		PMD_DRV_LOG(ERR, "Cannot add MAC address to a VF interface\n");
1784 		return -ENOTSUP;
1785 	}
1786 
1787 	if (!vnic) {
1788 		PMD_DRV_LOG(ERR, "VNIC not found for pool %d!\n", pool);
1789 		return -EINVAL;
1790 	}
1791 
1792 	/* Filter settings will get applied when port is started */
1793 	if (!eth_dev->data->dev_started)
1794 		return 0;
1795 
1796 	rc = bnxt_add_mac_filter(bp, vnic, mac_addr, index, pool);
1797 
1798 	return rc;
1799 }
1800 
1801 int bnxt_link_update_op(struct rte_eth_dev *eth_dev, int wait_to_complete)
1802 {
1803 	int rc = 0;
1804 	struct bnxt *bp = eth_dev->data->dev_private;
1805 	struct rte_eth_link new;
1806 	int cnt = wait_to_complete ? BNXT_MAX_LINK_WAIT_CNT :
1807 			BNXT_MIN_LINK_WAIT_CNT;
1808 
1809 	rc = is_bnxt_in_error(bp);
1810 	if (rc)
1811 		return rc;
1812 
1813 	memset(&new, 0, sizeof(new));
1814 
1815 	if (bp->link_info == NULL)
1816 		goto out;
1817 
1818 	do {
1819 		/* Retrieve link info from hardware */
1820 		rc = bnxt_get_hwrm_link_config(bp, &new);
1821 		if (rc) {
1822 			new.link_speed = ETH_LINK_SPEED_100M;
1823 			new.link_duplex = ETH_LINK_FULL_DUPLEX;
1824 			PMD_DRV_LOG(ERR,
1825 				"Failed to retrieve link rc = 0x%x!\n", rc);
1826 			goto out;
1827 		}
1828 
1829 		if (!wait_to_complete || new.link_status)
1830 			break;
1831 
1832 		rte_delay_ms(BNXT_LINK_WAIT_INTERVAL);
1833 	} while (cnt--);
1834 
1835 	/* Only single function PF can bring phy down.
1836 	 * When port is stopped, report link down for VF/MH/NPAR functions.
1837 	 */
1838 	if (!BNXT_SINGLE_PF(bp) && !eth_dev->data->dev_started)
1839 		memset(&new, 0, sizeof(new));
1840 
1841 out:
1842 	/* Timed out or success */
1843 	if (new.link_status != eth_dev->data->dev_link.link_status ||
1844 	    new.link_speed != eth_dev->data->dev_link.link_speed) {
1845 		rte_eth_linkstatus_set(eth_dev, &new);
1846 
1847 		rte_eth_dev_callback_process(eth_dev,
1848 					     RTE_ETH_EVENT_INTR_LSC,
1849 					     NULL);
1850 
1851 		bnxt_print_link_info(eth_dev);
1852 	}
1853 
1854 	return rc;
1855 }
1856 
1857 static int bnxt_promiscuous_enable_op(struct rte_eth_dev *eth_dev)
1858 {
1859 	struct bnxt *bp = eth_dev->data->dev_private;
1860 	struct bnxt_vnic_info *vnic;
1861 	uint32_t old_flags;
1862 	int rc;
1863 
1864 	rc = is_bnxt_in_error(bp);
1865 	if (rc)
1866 		return rc;
1867 
1868 	/* Filter settings will get applied when port is started */
1869 	if (!eth_dev->data->dev_started)
1870 		return 0;
1871 
1872 	if (bp->vnic_info == NULL)
1873 		return 0;
1874 
1875 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
1876 
1877 	old_flags = vnic->flags;
1878 	vnic->flags |= BNXT_VNIC_INFO_PROMISC;
1879 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1880 	if (rc != 0)
1881 		vnic->flags = old_flags;
1882 
1883 	return rc;
1884 }
1885 
1886 static int bnxt_promiscuous_disable_op(struct rte_eth_dev *eth_dev)
1887 {
1888 	struct bnxt *bp = eth_dev->data->dev_private;
1889 	struct bnxt_vnic_info *vnic;
1890 	uint32_t old_flags;
1891 	int rc;
1892 
1893 	rc = is_bnxt_in_error(bp);
1894 	if (rc)
1895 		return rc;
1896 
1897 	/* Filter settings will get applied when port is started */
1898 	if (!eth_dev->data->dev_started)
1899 		return 0;
1900 
1901 	if (bp->vnic_info == NULL)
1902 		return 0;
1903 
1904 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
1905 
1906 	old_flags = vnic->flags;
1907 	vnic->flags &= ~BNXT_VNIC_INFO_PROMISC;
1908 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1909 	if (rc != 0)
1910 		vnic->flags = old_flags;
1911 
1912 	return rc;
1913 }
1914 
1915 static int bnxt_allmulticast_enable_op(struct rte_eth_dev *eth_dev)
1916 {
1917 	struct bnxt *bp = eth_dev->data->dev_private;
1918 	struct bnxt_vnic_info *vnic;
1919 	uint32_t old_flags;
1920 	int rc;
1921 
1922 	rc = is_bnxt_in_error(bp);
1923 	if (rc)
1924 		return rc;
1925 
1926 	/* Filter settings will get applied when port is started */
1927 	if (!eth_dev->data->dev_started)
1928 		return 0;
1929 
1930 	if (bp->vnic_info == NULL)
1931 		return 0;
1932 
1933 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
1934 
1935 	old_flags = vnic->flags;
1936 	vnic->flags |= BNXT_VNIC_INFO_ALLMULTI;
1937 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1938 	if (rc != 0)
1939 		vnic->flags = old_flags;
1940 
1941 	return rc;
1942 }
1943 
1944 static int bnxt_allmulticast_disable_op(struct rte_eth_dev *eth_dev)
1945 {
1946 	struct bnxt *bp = eth_dev->data->dev_private;
1947 	struct bnxt_vnic_info *vnic;
1948 	uint32_t old_flags;
1949 	int rc;
1950 
1951 	rc = is_bnxt_in_error(bp);
1952 	if (rc)
1953 		return rc;
1954 
1955 	/* Filter settings will get applied when port is started */
1956 	if (!eth_dev->data->dev_started)
1957 		return 0;
1958 
1959 	if (bp->vnic_info == NULL)
1960 		return 0;
1961 
1962 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
1963 
1964 	old_flags = vnic->flags;
1965 	vnic->flags &= ~BNXT_VNIC_INFO_ALLMULTI;
1966 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1967 	if (rc != 0)
1968 		vnic->flags = old_flags;
1969 
1970 	return rc;
1971 }
1972 
1973 /* Return bnxt_rx_queue pointer corresponding to a given rxq. */
1974 static struct bnxt_rx_queue *bnxt_qid_to_rxq(struct bnxt *bp, uint16_t qid)
1975 {
1976 	if (qid >= bp->rx_nr_rings)
1977 		return NULL;
1978 
1979 	return bp->eth_dev->data->rx_queues[qid];
1980 }
1981 
1982 /* Return rxq corresponding to a given rss table ring/group ID. */
1983 static uint16_t bnxt_rss_to_qid(struct bnxt *bp, uint16_t fwr)
1984 {
1985 	struct bnxt_rx_queue *rxq;
1986 	unsigned int i;
1987 
1988 	if (!BNXT_HAS_RING_GRPS(bp)) {
1989 		for (i = 0; i < bp->rx_nr_rings; i++) {
1990 			rxq = bp->eth_dev->data->rx_queues[i];
1991 			if (rxq->rx_ring->rx_ring_struct->fw_ring_id == fwr)
1992 				return rxq->index;
1993 		}
1994 	} else {
1995 		for (i = 0; i < bp->rx_nr_rings; i++) {
1996 			if (bp->grp_info[i].fw_grp_id == fwr)
1997 				return i;
1998 		}
1999 	}
2000 
2001 	return INVALID_HW_RING_ID;
2002 }
2003 
2004 static int bnxt_reta_update_op(struct rte_eth_dev *eth_dev,
2005 			    struct rte_eth_rss_reta_entry64 *reta_conf,
2006 			    uint16_t reta_size)
2007 {
2008 	struct bnxt *bp = eth_dev->data->dev_private;
2009 	struct rte_eth_conf *dev_conf = &bp->eth_dev->data->dev_conf;
2010 	struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
2011 	uint16_t tbl_size = bnxt_rss_hash_tbl_size(bp);
2012 	uint16_t idx, sft;
2013 	int i, rc;
2014 
2015 	rc = is_bnxt_in_error(bp);
2016 	if (rc)
2017 		return rc;
2018 
2019 	if (!vnic->rss_table)
2020 		return -EINVAL;
2021 
2022 	if (!(dev_conf->rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG))
2023 		return -EINVAL;
2024 
2025 	if (reta_size != tbl_size) {
2026 		PMD_DRV_LOG(ERR, "The configured hash table lookup size "
2027 			"(%d) must equal the size supported by the hardware "
2028 			"(%d)\n", reta_size, tbl_size);
2029 		return -EINVAL;
2030 	}
2031 
2032 	for (i = 0; i < reta_size; i++) {
2033 		struct bnxt_rx_queue *rxq;
2034 
2035 		idx = i / RTE_RETA_GROUP_SIZE;
2036 		sft = i % RTE_RETA_GROUP_SIZE;
2037 
2038 		if (!(reta_conf[idx].mask & (1ULL << sft)))
2039 			continue;
2040 
2041 		rxq = bnxt_qid_to_rxq(bp, reta_conf[idx].reta[sft]);
2042 		if (!rxq) {
2043 			PMD_DRV_LOG(ERR, "Invalid ring in reta_conf.\n");
2044 			return -EINVAL;
2045 		}
2046 
2047 		if (BNXT_CHIP_P5(bp)) {
2048 			vnic->rss_table[i * 2] =
2049 				rxq->rx_ring->rx_ring_struct->fw_ring_id;
2050 			vnic->rss_table[i * 2 + 1] =
2051 				rxq->cp_ring->cp_ring_struct->fw_ring_id;
2052 		} else {
2053 			vnic->rss_table[i] =
2054 			    vnic->fw_grp_ids[reta_conf[idx].reta[sft]];
2055 		}
2056 	}
2057 
2058 	rc = bnxt_hwrm_vnic_rss_cfg(bp, vnic);
2059 	return rc;
2060 }
2061 
2062 static int bnxt_reta_query_op(struct rte_eth_dev *eth_dev,
2063 			      struct rte_eth_rss_reta_entry64 *reta_conf,
2064 			      uint16_t reta_size)
2065 {
2066 	struct bnxt *bp = eth_dev->data->dev_private;
2067 	struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
2068 	uint16_t tbl_size = bnxt_rss_hash_tbl_size(bp);
2069 	uint16_t idx, sft, i;
2070 	int rc;
2071 
2072 	rc = is_bnxt_in_error(bp);
2073 	if (rc)
2074 		return rc;
2075 
2076 	if (!vnic)
2077 		return -EINVAL;
2078 	if (!vnic->rss_table)
2079 		return -EINVAL;
2080 
2081 	if (reta_size != tbl_size) {
2082 		PMD_DRV_LOG(ERR, "The configured hash table lookup size "
2083 			"(%d) must equal the size supported by the hardware "
2084 			"(%d)\n", reta_size, tbl_size);
2085 		return -EINVAL;
2086 	}
2087 
2088 	for (idx = 0, i = 0; i < reta_size; i++) {
2089 		idx = i / RTE_RETA_GROUP_SIZE;
2090 		sft = i % RTE_RETA_GROUP_SIZE;
2091 
2092 		if (reta_conf[idx].mask & (1ULL << sft)) {
2093 			uint16_t qid;
2094 
2095 			if (BNXT_CHIP_P5(bp))
2096 				qid = bnxt_rss_to_qid(bp,
2097 						      vnic->rss_table[i * 2]);
2098 			else
2099 				qid = bnxt_rss_to_qid(bp, vnic->rss_table[i]);
2100 
2101 			if (qid == INVALID_HW_RING_ID) {
2102 				PMD_DRV_LOG(ERR, "Inv. entry in rss table.\n");
2103 				return -EINVAL;
2104 			}
2105 			reta_conf[idx].reta[sft] = qid;
2106 		}
2107 	}
2108 
2109 	return 0;
2110 }
2111 
2112 static int bnxt_rss_hash_update_op(struct rte_eth_dev *eth_dev,
2113 				   struct rte_eth_rss_conf *rss_conf)
2114 {
2115 	struct bnxt *bp = eth_dev->data->dev_private;
2116 	struct rte_eth_conf *dev_conf = &bp->eth_dev->data->dev_conf;
2117 	struct bnxt_vnic_info *vnic;
2118 	int rc;
2119 
2120 	rc = is_bnxt_in_error(bp);
2121 	if (rc)
2122 		return rc;
2123 
2124 	/*
2125 	 * If RSS enablement were different than dev_configure,
2126 	 * then return -EINVAL
2127 	 */
2128 	if (dev_conf->rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG) {
2129 		if (!rss_conf->rss_hf)
2130 			PMD_DRV_LOG(ERR, "Hash type NONE\n");
2131 	} else {
2132 		if (rss_conf->rss_hf & BNXT_ETH_RSS_SUPPORT)
2133 			return -EINVAL;
2134 	}
2135 
2136 	bp->flags |= BNXT_FLAG_UPDATE_HASH;
2137 	memcpy(&eth_dev->data->dev_conf.rx_adv_conf.rss_conf,
2138 	       rss_conf,
2139 	       sizeof(*rss_conf));
2140 
2141 	/* Update the default RSS VNIC(s) */
2142 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
2143 	vnic->hash_type = bnxt_rte_to_hwrm_hash_types(rss_conf->rss_hf);
2144 	vnic->hash_mode =
2145 		bnxt_rte_to_hwrm_hash_level(bp, rss_conf->rss_hf,
2146 					    ETH_RSS_LEVEL(rss_conf->rss_hf));
2147 
2148 	/*
2149 	 * If hashkey is not specified, use the previously configured
2150 	 * hashkey
2151 	 */
2152 	if (!rss_conf->rss_key)
2153 		goto rss_config;
2154 
2155 	if (rss_conf->rss_key_len != HW_HASH_KEY_SIZE) {
2156 		PMD_DRV_LOG(ERR,
2157 			    "Invalid hashkey length, should be %d bytes\n",
2158 			    HW_HASH_KEY_SIZE);
2159 		return -EINVAL;
2160 	}
2161 	memcpy(vnic->rss_hash_key, rss_conf->rss_key, rss_conf->rss_key_len);
2162 
2163 rss_config:
2164 	rc = bnxt_hwrm_vnic_rss_cfg(bp, vnic);
2165 	return rc;
2166 }
2167 
2168 static int bnxt_rss_hash_conf_get_op(struct rte_eth_dev *eth_dev,
2169 				     struct rte_eth_rss_conf *rss_conf)
2170 {
2171 	struct bnxt *bp = eth_dev->data->dev_private;
2172 	struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
2173 	int len, rc;
2174 	uint32_t hash_types;
2175 
2176 	rc = is_bnxt_in_error(bp);
2177 	if (rc)
2178 		return rc;
2179 
2180 	/* RSS configuration is the same for all VNICs */
2181 	if (vnic && vnic->rss_hash_key) {
2182 		if (rss_conf->rss_key) {
2183 			len = rss_conf->rss_key_len <= HW_HASH_KEY_SIZE ?
2184 			      rss_conf->rss_key_len : HW_HASH_KEY_SIZE;
2185 			memcpy(rss_conf->rss_key, vnic->rss_hash_key, len);
2186 		}
2187 
2188 		hash_types = vnic->hash_type;
2189 		rss_conf->rss_hf = 0;
2190 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV4) {
2191 			rss_conf->rss_hf |= ETH_RSS_IPV4;
2192 			hash_types &= ~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV4;
2193 		}
2194 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV4) {
2195 			rss_conf->rss_hf |= ETH_RSS_NONFRAG_IPV4_TCP;
2196 			hash_types &=
2197 				~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV4;
2198 		}
2199 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV4) {
2200 			rss_conf->rss_hf |= ETH_RSS_NONFRAG_IPV4_UDP;
2201 			hash_types &=
2202 				~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV4;
2203 		}
2204 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV6) {
2205 			rss_conf->rss_hf |= ETH_RSS_IPV6;
2206 			hash_types &= ~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV6;
2207 		}
2208 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV6) {
2209 			rss_conf->rss_hf |= ETH_RSS_NONFRAG_IPV6_TCP;
2210 			hash_types &=
2211 				~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV6;
2212 		}
2213 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV6) {
2214 			rss_conf->rss_hf |= ETH_RSS_NONFRAG_IPV6_UDP;
2215 			hash_types &=
2216 				~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV6;
2217 		}
2218 
2219 		rss_conf->rss_hf |=
2220 			bnxt_hwrm_to_rte_rss_level(bp, vnic->hash_mode);
2221 
2222 		if (hash_types) {
2223 			PMD_DRV_LOG(ERR,
2224 				"Unknown RSS config from firmware (%08x), RSS disabled",
2225 				vnic->hash_type);
2226 			return -ENOTSUP;
2227 		}
2228 	} else {
2229 		rss_conf->rss_hf = 0;
2230 	}
2231 	return 0;
2232 }
2233 
2234 static int bnxt_flow_ctrl_get_op(struct rte_eth_dev *dev,
2235 			       struct rte_eth_fc_conf *fc_conf)
2236 {
2237 	struct bnxt *bp = dev->data->dev_private;
2238 	struct rte_eth_link link_info;
2239 	int rc;
2240 
2241 	rc = is_bnxt_in_error(bp);
2242 	if (rc)
2243 		return rc;
2244 
2245 	rc = bnxt_get_hwrm_link_config(bp, &link_info);
2246 	if (rc)
2247 		return rc;
2248 
2249 	memset(fc_conf, 0, sizeof(*fc_conf));
2250 	if (bp->link_info->auto_pause)
2251 		fc_conf->autoneg = 1;
2252 	switch (bp->link_info->pause) {
2253 	case 0:
2254 		fc_conf->mode = RTE_FC_NONE;
2255 		break;
2256 	case HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_TX:
2257 		fc_conf->mode = RTE_FC_TX_PAUSE;
2258 		break;
2259 	case HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_RX:
2260 		fc_conf->mode = RTE_FC_RX_PAUSE;
2261 		break;
2262 	case (HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_TX |
2263 			HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_RX):
2264 		fc_conf->mode = RTE_FC_FULL;
2265 		break;
2266 	}
2267 	return 0;
2268 }
2269 
2270 static int bnxt_flow_ctrl_set_op(struct rte_eth_dev *dev,
2271 			       struct rte_eth_fc_conf *fc_conf)
2272 {
2273 	struct bnxt *bp = dev->data->dev_private;
2274 	int rc;
2275 
2276 	rc = is_bnxt_in_error(bp);
2277 	if (rc)
2278 		return rc;
2279 
2280 	if (!BNXT_SINGLE_PF(bp)) {
2281 		PMD_DRV_LOG(ERR,
2282 			    "Flow Control Settings cannot be modified on VF or on shared PF\n");
2283 		return -ENOTSUP;
2284 	}
2285 
2286 	switch (fc_conf->mode) {
2287 	case RTE_FC_NONE:
2288 		bp->link_info->auto_pause = 0;
2289 		bp->link_info->force_pause = 0;
2290 		break;
2291 	case RTE_FC_RX_PAUSE:
2292 		if (fc_conf->autoneg) {
2293 			bp->link_info->auto_pause =
2294 					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_RX;
2295 			bp->link_info->force_pause = 0;
2296 		} else {
2297 			bp->link_info->auto_pause = 0;
2298 			bp->link_info->force_pause =
2299 					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_RX;
2300 		}
2301 		break;
2302 	case RTE_FC_TX_PAUSE:
2303 		if (fc_conf->autoneg) {
2304 			bp->link_info->auto_pause =
2305 					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_TX;
2306 			bp->link_info->force_pause = 0;
2307 		} else {
2308 			bp->link_info->auto_pause = 0;
2309 			bp->link_info->force_pause =
2310 					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_TX;
2311 		}
2312 		break;
2313 	case RTE_FC_FULL:
2314 		if (fc_conf->autoneg) {
2315 			bp->link_info->auto_pause =
2316 					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_TX |
2317 					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_RX;
2318 			bp->link_info->force_pause = 0;
2319 		} else {
2320 			bp->link_info->auto_pause = 0;
2321 			bp->link_info->force_pause =
2322 					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_TX |
2323 					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_RX;
2324 		}
2325 		break;
2326 	}
2327 	return bnxt_set_hwrm_link_config(bp, true);
2328 }
2329 
2330 /* Add UDP tunneling port */
2331 static int
2332 bnxt_udp_tunnel_port_add_op(struct rte_eth_dev *eth_dev,
2333 			 struct rte_eth_udp_tunnel *udp_tunnel)
2334 {
2335 	struct bnxt *bp = eth_dev->data->dev_private;
2336 	uint16_t tunnel_type = 0;
2337 	int rc = 0;
2338 
2339 	rc = is_bnxt_in_error(bp);
2340 	if (rc)
2341 		return rc;
2342 
2343 	switch (udp_tunnel->prot_type) {
2344 	case RTE_TUNNEL_TYPE_VXLAN:
2345 		if (bp->vxlan_port_cnt) {
2346 			PMD_DRV_LOG(ERR, "Tunnel Port %d already programmed\n",
2347 				udp_tunnel->udp_port);
2348 			if (bp->vxlan_port != udp_tunnel->udp_port) {
2349 				PMD_DRV_LOG(ERR, "Only one port allowed\n");
2350 				return -ENOSPC;
2351 			}
2352 			bp->vxlan_port_cnt++;
2353 			return 0;
2354 		}
2355 		tunnel_type =
2356 			HWRM_TUNNEL_DST_PORT_ALLOC_INPUT_TUNNEL_TYPE_VXLAN;
2357 		bp->vxlan_port_cnt++;
2358 		break;
2359 	case RTE_TUNNEL_TYPE_GENEVE:
2360 		if (bp->geneve_port_cnt) {
2361 			PMD_DRV_LOG(ERR, "Tunnel Port %d already programmed\n",
2362 				udp_tunnel->udp_port);
2363 			if (bp->geneve_port != udp_tunnel->udp_port) {
2364 				PMD_DRV_LOG(ERR, "Only one port allowed\n");
2365 				return -ENOSPC;
2366 			}
2367 			bp->geneve_port_cnt++;
2368 			return 0;
2369 		}
2370 		tunnel_type =
2371 			HWRM_TUNNEL_DST_PORT_ALLOC_INPUT_TUNNEL_TYPE_GENEVE;
2372 		bp->geneve_port_cnt++;
2373 		break;
2374 	default:
2375 		PMD_DRV_LOG(ERR, "Tunnel type is not supported\n");
2376 		return -ENOTSUP;
2377 	}
2378 	rc = bnxt_hwrm_tunnel_dst_port_alloc(bp, udp_tunnel->udp_port,
2379 					     tunnel_type);
2380 	return rc;
2381 }
2382 
2383 static int
2384 bnxt_udp_tunnel_port_del_op(struct rte_eth_dev *eth_dev,
2385 			 struct rte_eth_udp_tunnel *udp_tunnel)
2386 {
2387 	struct bnxt *bp = eth_dev->data->dev_private;
2388 	uint16_t tunnel_type = 0;
2389 	uint16_t port = 0;
2390 	int rc = 0;
2391 
2392 	rc = is_bnxt_in_error(bp);
2393 	if (rc)
2394 		return rc;
2395 
2396 	switch (udp_tunnel->prot_type) {
2397 	case RTE_TUNNEL_TYPE_VXLAN:
2398 		if (!bp->vxlan_port_cnt) {
2399 			PMD_DRV_LOG(ERR, "No Tunnel port configured yet\n");
2400 			return -EINVAL;
2401 		}
2402 		if (bp->vxlan_port != udp_tunnel->udp_port) {
2403 			PMD_DRV_LOG(ERR, "Req Port: %d. Configured port: %d\n",
2404 				udp_tunnel->udp_port, bp->vxlan_port);
2405 			return -EINVAL;
2406 		}
2407 		if (--bp->vxlan_port_cnt)
2408 			return 0;
2409 
2410 		tunnel_type =
2411 			HWRM_TUNNEL_DST_PORT_FREE_INPUT_TUNNEL_TYPE_VXLAN;
2412 		port = bp->vxlan_fw_dst_port_id;
2413 		break;
2414 	case RTE_TUNNEL_TYPE_GENEVE:
2415 		if (!bp->geneve_port_cnt) {
2416 			PMD_DRV_LOG(ERR, "No Tunnel port configured yet\n");
2417 			return -EINVAL;
2418 		}
2419 		if (bp->geneve_port != udp_tunnel->udp_port) {
2420 			PMD_DRV_LOG(ERR, "Req Port: %d. Configured port: %d\n",
2421 				udp_tunnel->udp_port, bp->geneve_port);
2422 			return -EINVAL;
2423 		}
2424 		if (--bp->geneve_port_cnt)
2425 			return 0;
2426 
2427 		tunnel_type =
2428 			HWRM_TUNNEL_DST_PORT_FREE_INPUT_TUNNEL_TYPE_GENEVE;
2429 		port = bp->geneve_fw_dst_port_id;
2430 		break;
2431 	default:
2432 		PMD_DRV_LOG(ERR, "Tunnel type is not supported\n");
2433 		return -ENOTSUP;
2434 	}
2435 
2436 	rc = bnxt_hwrm_tunnel_dst_port_free(bp, port, tunnel_type);
2437 	return rc;
2438 }
2439 
2440 static int bnxt_del_vlan_filter(struct bnxt *bp, uint16_t vlan_id)
2441 {
2442 	struct bnxt_filter_info *filter;
2443 	struct bnxt_vnic_info *vnic;
2444 	int rc = 0;
2445 	uint32_t chk = HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN;
2446 
2447 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
2448 	filter = STAILQ_FIRST(&vnic->filter);
2449 	while (filter) {
2450 		/* Search for this matching MAC+VLAN filter */
2451 		if (bnxt_vlan_filter_exists(bp, filter, chk, vlan_id)) {
2452 			/* Delete the filter */
2453 			rc = bnxt_hwrm_clear_l2_filter(bp, filter);
2454 			if (rc)
2455 				return rc;
2456 			STAILQ_REMOVE(&vnic->filter, filter,
2457 				      bnxt_filter_info, next);
2458 			bnxt_free_filter(bp, filter);
2459 			PMD_DRV_LOG(INFO,
2460 				    "Deleted vlan filter for %d\n",
2461 				    vlan_id);
2462 			return 0;
2463 		}
2464 		filter = STAILQ_NEXT(filter, next);
2465 	}
2466 	return -ENOENT;
2467 }
2468 
2469 static int bnxt_add_vlan_filter(struct bnxt *bp, uint16_t vlan_id)
2470 {
2471 	struct bnxt_filter_info *filter;
2472 	struct bnxt_vnic_info *vnic;
2473 	int rc = 0;
2474 	uint32_t en = HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN |
2475 		HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN_MASK;
2476 	uint32_t chk = HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN;
2477 
2478 	/* Implementation notes on the use of VNIC in this command:
2479 	 *
2480 	 * By default, these filters belong to default vnic for the function.
2481 	 * Once these filters are set up, only destination VNIC can be modified.
2482 	 * If the destination VNIC is not specified in this command,
2483 	 * then the HWRM shall only create an l2 context id.
2484 	 */
2485 
2486 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
2487 	filter = STAILQ_FIRST(&vnic->filter);
2488 	/* Check if the VLAN has already been added */
2489 	while (filter) {
2490 		if (bnxt_vlan_filter_exists(bp, filter, chk, vlan_id))
2491 			return -EEXIST;
2492 
2493 		filter = STAILQ_NEXT(filter, next);
2494 	}
2495 
2496 	/* No match found. Alloc a fresh filter and issue the L2_FILTER_ALLOC
2497 	 * command to create MAC+VLAN filter with the right flags, enables set.
2498 	 */
2499 	filter = bnxt_alloc_filter(bp);
2500 	if (!filter) {
2501 		PMD_DRV_LOG(ERR,
2502 			    "MAC/VLAN filter alloc failed\n");
2503 		return -ENOMEM;
2504 	}
2505 	/* MAC + VLAN ID filter */
2506 	/* If l2_ivlan == 0 and l2_ivlan_mask != 0, only
2507 	 * untagged packets are received
2508 	 *
2509 	 * If l2_ivlan != 0 and l2_ivlan_mask != 0, untagged
2510 	 * packets and only the programmed vlan's packets are received
2511 	 */
2512 	filter->l2_ivlan = vlan_id;
2513 	filter->l2_ivlan_mask = 0x0FFF;
2514 	filter->enables |= en;
2515 	filter->flags |= HWRM_CFA_L2_FILTER_ALLOC_INPUT_FLAGS_OUTERMOST;
2516 
2517 	rc = bnxt_hwrm_set_l2_filter(bp, vnic->fw_vnic_id, filter);
2518 	if (rc) {
2519 		/* Free the newly allocated filter as we were
2520 		 * not able to create the filter in hardware.
2521 		 */
2522 		bnxt_free_filter(bp, filter);
2523 		return rc;
2524 	}
2525 
2526 	filter->mac_index = 0;
2527 	/* Add this new filter to the list */
2528 	if (vlan_id == 0)
2529 		STAILQ_INSERT_HEAD(&vnic->filter, filter, next);
2530 	else
2531 		STAILQ_INSERT_TAIL(&vnic->filter, filter, next);
2532 
2533 	PMD_DRV_LOG(INFO,
2534 		    "Added Vlan filter for %d\n", vlan_id);
2535 	return rc;
2536 }
2537 
2538 static int bnxt_vlan_filter_set_op(struct rte_eth_dev *eth_dev,
2539 		uint16_t vlan_id, int on)
2540 {
2541 	struct bnxt *bp = eth_dev->data->dev_private;
2542 	int rc;
2543 
2544 	rc = is_bnxt_in_error(bp);
2545 	if (rc)
2546 		return rc;
2547 
2548 	if (!eth_dev->data->dev_started) {
2549 		PMD_DRV_LOG(ERR, "port must be started before setting vlan\n");
2550 		return -EINVAL;
2551 	}
2552 
2553 	/* These operations apply to ALL existing MAC/VLAN filters */
2554 	if (on)
2555 		return bnxt_add_vlan_filter(bp, vlan_id);
2556 	else
2557 		return bnxt_del_vlan_filter(bp, vlan_id);
2558 }
2559 
2560 static int bnxt_del_dflt_mac_filter(struct bnxt *bp,
2561 				    struct bnxt_vnic_info *vnic)
2562 {
2563 	struct bnxt_filter_info *filter;
2564 	int rc;
2565 
2566 	filter = STAILQ_FIRST(&vnic->filter);
2567 	while (filter) {
2568 		if (filter->mac_index == 0 &&
2569 		    !memcmp(filter->l2_addr, bp->mac_addr,
2570 			    RTE_ETHER_ADDR_LEN)) {
2571 			rc = bnxt_hwrm_clear_l2_filter(bp, filter);
2572 			if (!rc) {
2573 				STAILQ_REMOVE(&vnic->filter, filter,
2574 					      bnxt_filter_info, next);
2575 				bnxt_free_filter(bp, filter);
2576 			}
2577 			return rc;
2578 		}
2579 		filter = STAILQ_NEXT(filter, next);
2580 	}
2581 	return 0;
2582 }
2583 
2584 static int
2585 bnxt_config_vlan_hw_filter(struct bnxt *bp, uint64_t rx_offloads)
2586 {
2587 	struct bnxt_vnic_info *vnic;
2588 	unsigned int i;
2589 	int rc;
2590 
2591 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
2592 	if (!(rx_offloads & DEV_RX_OFFLOAD_VLAN_FILTER)) {
2593 		/* Remove any VLAN filters programmed */
2594 		for (i = 0; i < RTE_ETHER_MAX_VLAN_ID; i++)
2595 			bnxt_del_vlan_filter(bp, i);
2596 
2597 		rc = bnxt_add_mac_filter(bp, vnic, NULL, 0, 0);
2598 		if (rc)
2599 			return rc;
2600 	} else {
2601 		/* Default filter will allow packets that match the
2602 		 * dest mac. So, it has to be deleted, otherwise, we
2603 		 * will endup receiving vlan packets for which the
2604 		 * filter is not programmed, when hw-vlan-filter
2605 		 * configuration is ON
2606 		 */
2607 		bnxt_del_dflt_mac_filter(bp, vnic);
2608 		/* This filter will allow only untagged packets */
2609 		bnxt_add_vlan_filter(bp, 0);
2610 	}
2611 	PMD_DRV_LOG(DEBUG, "VLAN Filtering: %d\n",
2612 		    !!(rx_offloads & DEV_RX_OFFLOAD_VLAN_FILTER));
2613 
2614 	return 0;
2615 }
2616 
2617 static int bnxt_free_one_vnic(struct bnxt *bp, uint16_t vnic_id)
2618 {
2619 	struct bnxt_vnic_info *vnic = &bp->vnic_info[vnic_id];
2620 	unsigned int i;
2621 	int rc;
2622 
2623 	/* Destroy vnic filters and vnic */
2624 	if (bp->eth_dev->data->dev_conf.rxmode.offloads &
2625 	    DEV_RX_OFFLOAD_VLAN_FILTER) {
2626 		for (i = 0; i < RTE_ETHER_MAX_VLAN_ID; i++)
2627 			bnxt_del_vlan_filter(bp, i);
2628 	}
2629 	bnxt_del_dflt_mac_filter(bp, vnic);
2630 
2631 	rc = bnxt_hwrm_vnic_ctx_free(bp, vnic);
2632 	if (rc)
2633 		return rc;
2634 
2635 	rc = bnxt_hwrm_vnic_free(bp, vnic);
2636 	if (rc)
2637 		return rc;
2638 
2639 	rte_free(vnic->fw_grp_ids);
2640 	vnic->fw_grp_ids = NULL;
2641 
2642 	vnic->rx_queue_cnt = 0;
2643 
2644 	return 0;
2645 }
2646 
2647 static int
2648 bnxt_config_vlan_hw_stripping(struct bnxt *bp, uint64_t rx_offloads)
2649 {
2650 	struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
2651 	int rc;
2652 
2653 	/* Destroy, recreate and reconfigure the default vnic */
2654 	rc = bnxt_free_one_vnic(bp, 0);
2655 	if (rc)
2656 		return rc;
2657 
2658 	/* default vnic 0 */
2659 	rc = bnxt_setup_one_vnic(bp, 0);
2660 	if (rc)
2661 		return rc;
2662 
2663 	if (bp->eth_dev->data->dev_conf.rxmode.offloads &
2664 	    DEV_RX_OFFLOAD_VLAN_FILTER) {
2665 		rc = bnxt_add_vlan_filter(bp, 0);
2666 		if (rc)
2667 			return rc;
2668 		rc = bnxt_restore_vlan_filters(bp);
2669 		if (rc)
2670 			return rc;
2671 	} else {
2672 		rc = bnxt_add_mac_filter(bp, vnic, NULL, 0, 0);
2673 		if (rc)
2674 			return rc;
2675 	}
2676 
2677 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
2678 	if (rc)
2679 		return rc;
2680 
2681 	PMD_DRV_LOG(DEBUG, "VLAN Strip Offload: %d\n",
2682 		    !!(rx_offloads & DEV_RX_OFFLOAD_VLAN_STRIP));
2683 
2684 	return rc;
2685 }
2686 
2687 static int
2688 bnxt_vlan_offload_set_op(struct rte_eth_dev *dev, int mask)
2689 {
2690 	uint64_t rx_offloads = dev->data->dev_conf.rxmode.offloads;
2691 	struct bnxt *bp = dev->data->dev_private;
2692 	int rc;
2693 
2694 	rc = is_bnxt_in_error(bp);
2695 	if (rc)
2696 		return rc;
2697 
2698 	/* Filter settings will get applied when port is started */
2699 	if (!dev->data->dev_started)
2700 		return 0;
2701 
2702 	if (mask & ETH_VLAN_FILTER_MASK) {
2703 		/* Enable or disable VLAN filtering */
2704 		rc = bnxt_config_vlan_hw_filter(bp, rx_offloads);
2705 		if (rc)
2706 			return rc;
2707 	}
2708 
2709 	if (mask & ETH_VLAN_STRIP_MASK) {
2710 		/* Enable or disable VLAN stripping */
2711 		rc = bnxt_config_vlan_hw_stripping(bp, rx_offloads);
2712 		if (rc)
2713 			return rc;
2714 	}
2715 
2716 	if (mask & ETH_VLAN_EXTEND_MASK) {
2717 		if (rx_offloads & DEV_RX_OFFLOAD_VLAN_EXTEND)
2718 			PMD_DRV_LOG(DEBUG, "Extend VLAN supported\n");
2719 		else
2720 			PMD_DRV_LOG(INFO, "Extend VLAN unsupported\n");
2721 	}
2722 
2723 	return 0;
2724 }
2725 
2726 static int
2727 bnxt_vlan_tpid_set_op(struct rte_eth_dev *dev, enum rte_vlan_type vlan_type,
2728 		      uint16_t tpid)
2729 {
2730 	struct bnxt *bp = dev->data->dev_private;
2731 	int qinq = dev->data->dev_conf.rxmode.offloads &
2732 		   DEV_RX_OFFLOAD_VLAN_EXTEND;
2733 
2734 	if (vlan_type != ETH_VLAN_TYPE_INNER &&
2735 	    vlan_type != ETH_VLAN_TYPE_OUTER) {
2736 		PMD_DRV_LOG(ERR,
2737 			    "Unsupported vlan type.");
2738 		return -EINVAL;
2739 	}
2740 	if (!qinq) {
2741 		PMD_DRV_LOG(ERR,
2742 			    "QinQ not enabled. Needs to be ON as we can "
2743 			    "accelerate only outer vlan\n");
2744 		return -EINVAL;
2745 	}
2746 
2747 	if (vlan_type == ETH_VLAN_TYPE_OUTER) {
2748 		switch (tpid) {
2749 		case RTE_ETHER_TYPE_QINQ:
2750 			bp->outer_tpid_bd =
2751 				TX_BD_LONG_CFA_META_VLAN_TPID_TPID88A8;
2752 				break;
2753 		case RTE_ETHER_TYPE_VLAN:
2754 			bp->outer_tpid_bd =
2755 				TX_BD_LONG_CFA_META_VLAN_TPID_TPID8100;
2756 				break;
2757 		case RTE_ETHER_TYPE_QINQ1:
2758 			bp->outer_tpid_bd =
2759 				TX_BD_LONG_CFA_META_VLAN_TPID_TPID9100;
2760 				break;
2761 		case RTE_ETHER_TYPE_QINQ2:
2762 			bp->outer_tpid_bd =
2763 				TX_BD_LONG_CFA_META_VLAN_TPID_TPID9200;
2764 				break;
2765 		case RTE_ETHER_TYPE_QINQ3:
2766 			bp->outer_tpid_bd =
2767 				 TX_BD_LONG_CFA_META_VLAN_TPID_TPID9300;
2768 				break;
2769 		default:
2770 			PMD_DRV_LOG(ERR, "Invalid TPID: %x\n", tpid);
2771 			return -EINVAL;
2772 		}
2773 		bp->outer_tpid_bd |= tpid;
2774 		PMD_DRV_LOG(INFO, "outer_tpid_bd = %x\n", bp->outer_tpid_bd);
2775 	} else if (vlan_type == ETH_VLAN_TYPE_INNER) {
2776 		PMD_DRV_LOG(ERR,
2777 			    "Can accelerate only outer vlan in QinQ\n");
2778 		return -EINVAL;
2779 	}
2780 
2781 	return 0;
2782 }
2783 
2784 static int
2785 bnxt_set_default_mac_addr_op(struct rte_eth_dev *dev,
2786 			     struct rte_ether_addr *addr)
2787 {
2788 	struct bnxt *bp = dev->data->dev_private;
2789 	/* Default Filter is tied to VNIC 0 */
2790 	struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
2791 	int rc;
2792 
2793 	rc = is_bnxt_in_error(bp);
2794 	if (rc)
2795 		return rc;
2796 
2797 	if (BNXT_VF(bp) && !BNXT_VF_IS_TRUSTED(bp))
2798 		return -EPERM;
2799 
2800 	if (rte_is_zero_ether_addr(addr))
2801 		return -EINVAL;
2802 
2803 	/* Filter settings will get applied when port is started */
2804 	if (!dev->data->dev_started)
2805 		return 0;
2806 
2807 	/* Check if the requested MAC is already added */
2808 	if (memcmp(addr, bp->mac_addr, RTE_ETHER_ADDR_LEN) == 0)
2809 		return 0;
2810 
2811 	/* Destroy filter and re-create it */
2812 	bnxt_del_dflt_mac_filter(bp, vnic);
2813 
2814 	memcpy(bp->mac_addr, addr, RTE_ETHER_ADDR_LEN);
2815 	if (dev->data->dev_conf.rxmode.offloads & DEV_RX_OFFLOAD_VLAN_FILTER) {
2816 		/* This filter will allow only untagged packets */
2817 		rc = bnxt_add_vlan_filter(bp, 0);
2818 	} else {
2819 		rc = bnxt_add_mac_filter(bp, vnic, addr, 0, 0);
2820 	}
2821 
2822 	PMD_DRV_LOG(DEBUG, "Set MAC addr\n");
2823 	return rc;
2824 }
2825 
2826 static int
2827 bnxt_dev_set_mc_addr_list_op(struct rte_eth_dev *eth_dev,
2828 			  struct rte_ether_addr *mc_addr_set,
2829 			  uint32_t nb_mc_addr)
2830 {
2831 	struct bnxt *bp = eth_dev->data->dev_private;
2832 	char *mc_addr_list = (char *)mc_addr_set;
2833 	struct bnxt_vnic_info *vnic;
2834 	uint32_t off = 0, i = 0;
2835 	int rc;
2836 
2837 	rc = is_bnxt_in_error(bp);
2838 	if (rc)
2839 		return rc;
2840 
2841 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
2842 
2843 	if (nb_mc_addr > BNXT_MAX_MC_ADDRS) {
2844 		vnic->flags |= BNXT_VNIC_INFO_ALLMULTI;
2845 		goto allmulti;
2846 	}
2847 
2848 	/* TODO Check for Duplicate mcast addresses */
2849 	vnic->flags &= ~BNXT_VNIC_INFO_ALLMULTI;
2850 	for (i = 0; i < nb_mc_addr; i++) {
2851 		memcpy(vnic->mc_list + off, &mc_addr_list[i],
2852 			RTE_ETHER_ADDR_LEN);
2853 		off += RTE_ETHER_ADDR_LEN;
2854 	}
2855 
2856 	vnic->mc_addr_cnt = i;
2857 	if (vnic->mc_addr_cnt)
2858 		vnic->flags |= BNXT_VNIC_INFO_MCAST;
2859 	else
2860 		vnic->flags &= ~BNXT_VNIC_INFO_MCAST;
2861 
2862 allmulti:
2863 	return bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
2864 }
2865 
2866 static int
2867 bnxt_fw_version_get(struct rte_eth_dev *dev, char *fw_version, size_t fw_size)
2868 {
2869 	struct bnxt *bp = dev->data->dev_private;
2870 	uint8_t fw_major = (bp->fw_ver >> 24) & 0xff;
2871 	uint8_t fw_minor = (bp->fw_ver >> 16) & 0xff;
2872 	uint8_t fw_updt = (bp->fw_ver >> 8) & 0xff;
2873 	uint8_t fw_rsvd = bp->fw_ver & 0xff;
2874 	int ret;
2875 
2876 	ret = snprintf(fw_version, fw_size, "%d.%d.%d.%d",
2877 			fw_major, fw_minor, fw_updt, fw_rsvd);
2878 	if (ret < 0)
2879 		return -EINVAL;
2880 
2881 	ret += 1; /* add the size of '\0' */
2882 	if (fw_size < (size_t)ret)
2883 		return ret;
2884 	else
2885 		return 0;
2886 }
2887 
2888 static void
2889 bnxt_rxq_info_get_op(struct rte_eth_dev *dev, uint16_t queue_id,
2890 	struct rte_eth_rxq_info *qinfo)
2891 {
2892 	struct bnxt *bp = dev->data->dev_private;
2893 	struct bnxt_rx_queue *rxq;
2894 
2895 	if (is_bnxt_in_error(bp))
2896 		return;
2897 
2898 	rxq = dev->data->rx_queues[queue_id];
2899 
2900 	qinfo->mp = rxq->mb_pool;
2901 	qinfo->scattered_rx = dev->data->scattered_rx;
2902 	qinfo->nb_desc = rxq->nb_rx_desc;
2903 
2904 	qinfo->conf.rx_free_thresh = rxq->rx_free_thresh;
2905 	qinfo->conf.rx_drop_en = rxq->drop_en;
2906 	qinfo->conf.rx_deferred_start = rxq->rx_deferred_start;
2907 	qinfo->conf.offloads = dev->data->dev_conf.rxmode.offloads;
2908 }
2909 
2910 static void
2911 bnxt_txq_info_get_op(struct rte_eth_dev *dev, uint16_t queue_id,
2912 	struct rte_eth_txq_info *qinfo)
2913 {
2914 	struct bnxt *bp = dev->data->dev_private;
2915 	struct bnxt_tx_queue *txq;
2916 
2917 	if (is_bnxt_in_error(bp))
2918 		return;
2919 
2920 	txq = dev->data->tx_queues[queue_id];
2921 
2922 	qinfo->nb_desc = txq->nb_tx_desc;
2923 
2924 	qinfo->conf.tx_thresh.pthresh = txq->pthresh;
2925 	qinfo->conf.tx_thresh.hthresh = txq->hthresh;
2926 	qinfo->conf.tx_thresh.wthresh = txq->wthresh;
2927 
2928 	qinfo->conf.tx_free_thresh = txq->tx_free_thresh;
2929 	qinfo->conf.tx_rs_thresh = 0;
2930 	qinfo->conf.tx_deferred_start = txq->tx_deferred_start;
2931 	qinfo->conf.offloads = txq->offloads;
2932 }
2933 
2934 static const struct {
2935 	eth_rx_burst_t pkt_burst;
2936 	const char *info;
2937 } bnxt_rx_burst_info[] = {
2938 	{bnxt_recv_pkts,		"Scalar"},
2939 #if defined(RTE_ARCH_X86)
2940 	{bnxt_recv_pkts_vec,		"Vector SSE"},
2941 #endif
2942 #if defined(RTE_ARCH_X86) && defined(CC_AVX2_SUPPORT)
2943 	{bnxt_recv_pkts_vec_avx2,	"Vector AVX2"},
2944 #endif
2945 #if defined(RTE_ARCH_ARM64)
2946 	{bnxt_recv_pkts_vec,		"Vector Neon"},
2947 #endif
2948 };
2949 
2950 static int
2951 bnxt_rx_burst_mode_get(struct rte_eth_dev *dev, __rte_unused uint16_t queue_id,
2952 		       struct rte_eth_burst_mode *mode)
2953 {
2954 	eth_rx_burst_t pkt_burst = dev->rx_pkt_burst;
2955 	size_t i;
2956 
2957 	for (i = 0; i < RTE_DIM(bnxt_rx_burst_info); i++) {
2958 		if (pkt_burst == bnxt_rx_burst_info[i].pkt_burst) {
2959 			snprintf(mode->info, sizeof(mode->info), "%s",
2960 				 bnxt_rx_burst_info[i].info);
2961 			return 0;
2962 		}
2963 	}
2964 
2965 	return -EINVAL;
2966 }
2967 
2968 static const struct {
2969 	eth_tx_burst_t pkt_burst;
2970 	const char *info;
2971 } bnxt_tx_burst_info[] = {
2972 	{bnxt_xmit_pkts,		"Scalar"},
2973 #if defined(RTE_ARCH_X86)
2974 	{bnxt_xmit_pkts_vec,		"Vector SSE"},
2975 #endif
2976 #if defined(RTE_ARCH_X86) && defined(CC_AVX2_SUPPORT)
2977 	{bnxt_xmit_pkts_vec_avx2,	"Vector AVX2"},
2978 #endif
2979 #if defined(RTE_ARCH_ARM64)
2980 	{bnxt_xmit_pkts_vec,		"Vector Neon"},
2981 #endif
2982 };
2983 
2984 static int
2985 bnxt_tx_burst_mode_get(struct rte_eth_dev *dev, __rte_unused uint16_t queue_id,
2986 		       struct rte_eth_burst_mode *mode)
2987 {
2988 	eth_tx_burst_t pkt_burst = dev->tx_pkt_burst;
2989 	size_t i;
2990 
2991 	for (i = 0; i < RTE_DIM(bnxt_tx_burst_info); i++) {
2992 		if (pkt_burst == bnxt_tx_burst_info[i].pkt_burst) {
2993 			snprintf(mode->info, sizeof(mode->info), "%s",
2994 				 bnxt_tx_burst_info[i].info);
2995 			return 0;
2996 		}
2997 	}
2998 
2999 	return -EINVAL;
3000 }
3001 
3002 int bnxt_mtu_set_op(struct rte_eth_dev *eth_dev, uint16_t new_mtu)
3003 {
3004 	struct bnxt *bp = eth_dev->data->dev_private;
3005 	uint32_t new_pkt_size;
3006 	uint32_t rc = 0;
3007 	uint32_t i;
3008 
3009 	rc = is_bnxt_in_error(bp);
3010 	if (rc)
3011 		return rc;
3012 
3013 	/* Exit if receive queues are not configured yet */
3014 	if (!eth_dev->data->nb_rx_queues)
3015 		return rc;
3016 
3017 	new_pkt_size = new_mtu + RTE_ETHER_HDR_LEN + RTE_ETHER_CRC_LEN +
3018 		       VLAN_TAG_SIZE * BNXT_NUM_VLANS;
3019 
3020 	/*
3021 	 * Disallow any MTU change that would require scattered receive support
3022 	 * if it is not already enabled.
3023 	 */
3024 	if (eth_dev->data->dev_started &&
3025 	    !eth_dev->data->scattered_rx &&
3026 	    (new_pkt_size >
3027 	     eth_dev->data->min_rx_buf_size - RTE_PKTMBUF_HEADROOM)) {
3028 		PMD_DRV_LOG(ERR,
3029 			    "MTU change would require scattered rx support. ");
3030 		PMD_DRV_LOG(ERR, "Stop port before changing MTU.\n");
3031 		return -EINVAL;
3032 	}
3033 
3034 	if (new_mtu > RTE_ETHER_MTU) {
3035 		bp->flags |= BNXT_FLAG_JUMBO;
3036 		bp->eth_dev->data->dev_conf.rxmode.offloads |=
3037 			DEV_RX_OFFLOAD_JUMBO_FRAME;
3038 	} else {
3039 		bp->eth_dev->data->dev_conf.rxmode.offloads &=
3040 			~DEV_RX_OFFLOAD_JUMBO_FRAME;
3041 		bp->flags &= ~BNXT_FLAG_JUMBO;
3042 	}
3043 
3044 	/* Is there a change in mtu setting? */
3045 	if (eth_dev->data->dev_conf.rxmode.max_rx_pkt_len == new_pkt_size)
3046 		return rc;
3047 
3048 	for (i = 0; i < bp->nr_vnics; i++) {
3049 		struct bnxt_vnic_info *vnic = &bp->vnic_info[i];
3050 		uint16_t size = 0;
3051 
3052 		vnic->mru = BNXT_VNIC_MRU(new_mtu);
3053 		rc = bnxt_hwrm_vnic_cfg(bp, vnic);
3054 		if (rc)
3055 			break;
3056 
3057 		size = rte_pktmbuf_data_room_size(bp->rx_queues[0]->mb_pool);
3058 		size -= RTE_PKTMBUF_HEADROOM;
3059 
3060 		if (size < new_mtu) {
3061 			rc = bnxt_hwrm_vnic_plcmode_cfg(bp, vnic);
3062 			if (rc)
3063 				return rc;
3064 		}
3065 	}
3066 
3067 	if (!rc)
3068 		eth_dev->data->dev_conf.rxmode.max_rx_pkt_len = new_pkt_size;
3069 
3070 	PMD_DRV_LOG(INFO, "New MTU is %d\n", new_mtu);
3071 
3072 	return rc;
3073 }
3074 
3075 static int
3076 bnxt_vlan_pvid_set_op(struct rte_eth_dev *dev, uint16_t pvid, int on)
3077 {
3078 	struct bnxt *bp = dev->data->dev_private;
3079 	uint16_t vlan = bp->vlan;
3080 	int rc;
3081 
3082 	rc = is_bnxt_in_error(bp);
3083 	if (rc)
3084 		return rc;
3085 
3086 	if (!BNXT_SINGLE_PF(bp)) {
3087 		PMD_DRV_LOG(ERR, "PVID cannot be modified on VF or on shared PF\n");
3088 		return -ENOTSUP;
3089 	}
3090 	bp->vlan = on ? pvid : 0;
3091 
3092 	rc = bnxt_hwrm_set_default_vlan(bp, 0, 0);
3093 	if (rc)
3094 		bp->vlan = vlan;
3095 	return rc;
3096 }
3097 
3098 static int
3099 bnxt_dev_led_on_op(struct rte_eth_dev *dev)
3100 {
3101 	struct bnxt *bp = dev->data->dev_private;
3102 	int rc;
3103 
3104 	rc = is_bnxt_in_error(bp);
3105 	if (rc)
3106 		return rc;
3107 
3108 	return bnxt_hwrm_port_led_cfg(bp, true);
3109 }
3110 
3111 static int
3112 bnxt_dev_led_off_op(struct rte_eth_dev *dev)
3113 {
3114 	struct bnxt *bp = dev->data->dev_private;
3115 	int rc;
3116 
3117 	rc = is_bnxt_in_error(bp);
3118 	if (rc)
3119 		return rc;
3120 
3121 	return bnxt_hwrm_port_led_cfg(bp, false);
3122 }
3123 
3124 static uint32_t
3125 bnxt_rx_queue_count_op(struct rte_eth_dev *dev, uint16_t rx_queue_id)
3126 {
3127 	struct bnxt *bp = (struct bnxt *)dev->data->dev_private;
3128 	struct bnxt_cp_ring_info *cpr;
3129 	uint32_t desc = 0, raw_cons;
3130 	struct bnxt_rx_queue *rxq;
3131 	struct rx_pkt_cmpl *rxcmp;
3132 	int rc;
3133 
3134 	rc = is_bnxt_in_error(bp);
3135 	if (rc)
3136 		return rc;
3137 
3138 	rxq = dev->data->rx_queues[rx_queue_id];
3139 	cpr = rxq->cp_ring;
3140 	raw_cons = cpr->cp_raw_cons;
3141 
3142 	while (1) {
3143 		uint32_t agg_cnt, cons, cmpl_type;
3144 
3145 		cons = RING_CMP(cpr->cp_ring_struct, raw_cons);
3146 		rxcmp = (struct rx_pkt_cmpl *)&cpr->cp_desc_ring[cons];
3147 
3148 		if (!CMP_VALID(rxcmp, raw_cons, cpr->cp_ring_struct))
3149 			break;
3150 
3151 		cmpl_type = CMP_TYPE(rxcmp);
3152 
3153 		switch (cmpl_type) {
3154 		case CMPL_BASE_TYPE_RX_L2:
3155 		case CMPL_BASE_TYPE_RX_L2_V2:
3156 			agg_cnt = BNXT_RX_L2_AGG_BUFS(rxcmp);
3157 			raw_cons = raw_cons + CMP_LEN(cmpl_type) + agg_cnt;
3158 			desc++;
3159 			break;
3160 
3161 		case CMPL_BASE_TYPE_RX_TPA_END:
3162 			if (BNXT_CHIP_P5(rxq->bp)) {
3163 				struct rx_tpa_v2_end_cmpl_hi *p5_tpa_end;
3164 
3165 				p5_tpa_end = (void *)rxcmp;
3166 				agg_cnt = BNXT_TPA_END_AGG_BUFS_TH(p5_tpa_end);
3167 			} else {
3168 				struct rx_tpa_end_cmpl *tpa_end;
3169 
3170 				tpa_end = (void *)rxcmp;
3171 				agg_cnt = BNXT_TPA_END_AGG_BUFS(tpa_end);
3172 			}
3173 
3174 			raw_cons = raw_cons + CMP_LEN(cmpl_type) + agg_cnt;
3175 			desc++;
3176 			break;
3177 
3178 		default:
3179 			raw_cons += CMP_LEN(cmpl_type);
3180 		}
3181 	}
3182 
3183 	return desc;
3184 }
3185 
3186 static int
3187 bnxt_rx_descriptor_status_op(void *rx_queue, uint16_t offset)
3188 {
3189 	struct bnxt_rx_queue *rxq = rx_queue;
3190 	struct bnxt_cp_ring_info *cpr;
3191 	struct bnxt_rx_ring_info *rxr;
3192 	uint32_t desc, raw_cons;
3193 	struct bnxt *bp = rxq->bp;
3194 	struct rx_pkt_cmpl *rxcmp;
3195 	int rc;
3196 
3197 	rc = is_bnxt_in_error(bp);
3198 	if (rc)
3199 		return rc;
3200 
3201 	if (offset >= rxq->nb_rx_desc)
3202 		return -EINVAL;
3203 
3204 	rxr = rxq->rx_ring;
3205 	cpr = rxq->cp_ring;
3206 
3207 	/*
3208 	 * For the vector receive case, the completion at the requested
3209 	 * offset can be indexed directly.
3210 	 */
3211 #if defined(RTE_ARCH_X86) || defined(RTE_ARCH_ARM64)
3212 	if (bp->flags & BNXT_FLAG_RX_VECTOR_PKT_MODE) {
3213 		struct rx_pkt_cmpl *rxcmp;
3214 		uint32_t cons;
3215 
3216 		/* Check status of completion descriptor. */
3217 		raw_cons = cpr->cp_raw_cons +
3218 			   offset * CMP_LEN(CMPL_BASE_TYPE_RX_L2);
3219 		cons = RING_CMP(cpr->cp_ring_struct, raw_cons);
3220 		rxcmp = (struct rx_pkt_cmpl *)&cpr->cp_desc_ring[cons];
3221 
3222 		if (CMP_VALID(rxcmp, raw_cons, cpr->cp_ring_struct))
3223 			return RTE_ETH_RX_DESC_DONE;
3224 
3225 		/* Check whether rx desc has an mbuf attached. */
3226 		cons = RING_CMP(rxr->rx_ring_struct, raw_cons / 2);
3227 		if (cons >= rxq->rxrearm_start &&
3228 		    cons < rxq->rxrearm_start + rxq->rxrearm_nb) {
3229 			return RTE_ETH_RX_DESC_UNAVAIL;
3230 		}
3231 
3232 		return RTE_ETH_RX_DESC_AVAIL;
3233 	}
3234 #endif
3235 
3236 	/*
3237 	 * For the non-vector receive case, scan the completion ring to
3238 	 * locate the completion descriptor for the requested offset.
3239 	 */
3240 	raw_cons = cpr->cp_raw_cons;
3241 	desc = 0;
3242 	while (1) {
3243 		uint32_t agg_cnt, cons, cmpl_type;
3244 
3245 		cons = RING_CMP(cpr->cp_ring_struct, raw_cons);
3246 		rxcmp = (struct rx_pkt_cmpl *)&cpr->cp_desc_ring[cons];
3247 
3248 		if (!CMP_VALID(rxcmp, raw_cons, cpr->cp_ring_struct))
3249 			break;
3250 
3251 		cmpl_type = CMP_TYPE(rxcmp);
3252 
3253 		switch (cmpl_type) {
3254 		case CMPL_BASE_TYPE_RX_L2:
3255 		case CMPL_BASE_TYPE_RX_L2_V2:
3256 			if (desc == offset) {
3257 				cons = rxcmp->opaque;
3258 				if (rxr->rx_buf_ring[cons])
3259 					return RTE_ETH_RX_DESC_DONE;
3260 				else
3261 					return RTE_ETH_RX_DESC_UNAVAIL;
3262 			}
3263 			agg_cnt = BNXT_RX_L2_AGG_BUFS(rxcmp);
3264 			raw_cons = raw_cons + CMP_LEN(cmpl_type) + agg_cnt;
3265 			desc++;
3266 			break;
3267 
3268 		case CMPL_BASE_TYPE_RX_TPA_END:
3269 			if (desc == offset)
3270 				return RTE_ETH_RX_DESC_DONE;
3271 
3272 			if (BNXT_CHIP_P5(rxq->bp)) {
3273 				struct rx_tpa_v2_end_cmpl_hi *p5_tpa_end;
3274 
3275 				p5_tpa_end = (void *)rxcmp;
3276 				agg_cnt = BNXT_TPA_END_AGG_BUFS_TH(p5_tpa_end);
3277 			} else {
3278 				struct rx_tpa_end_cmpl *tpa_end;
3279 
3280 				tpa_end = (void *)rxcmp;
3281 				agg_cnt = BNXT_TPA_END_AGG_BUFS(tpa_end);
3282 			}
3283 
3284 			raw_cons = raw_cons + CMP_LEN(cmpl_type) + agg_cnt;
3285 			desc++;
3286 			break;
3287 
3288 		default:
3289 			raw_cons += CMP_LEN(cmpl_type);
3290 		}
3291 	}
3292 
3293 	return RTE_ETH_RX_DESC_AVAIL;
3294 }
3295 
3296 static int
3297 bnxt_tx_descriptor_status_op(void *tx_queue, uint16_t offset)
3298 {
3299 	struct bnxt_tx_queue *txq = (struct bnxt_tx_queue *)tx_queue;
3300 	struct bnxt_cp_ring_info *cpr = txq->cp_ring;
3301 	uint32_t ring_mask, raw_cons, nb_tx_pkts = 0;
3302 	struct bnxt_ring *cp_ring_struct;
3303 	struct cmpl_base *cp_desc_ring;
3304 	int rc;
3305 
3306 	rc = is_bnxt_in_error(txq->bp);
3307 	if (rc)
3308 		return rc;
3309 
3310 	if (offset >= txq->nb_tx_desc)
3311 		return -EINVAL;
3312 
3313 	/* Return "desc done" if descriptor is available for use. */
3314 	if (bnxt_tx_bds_in_hw(txq) <= offset)
3315 		return RTE_ETH_TX_DESC_DONE;
3316 
3317 	raw_cons = cpr->cp_raw_cons;
3318 	cp_desc_ring = cpr->cp_desc_ring;
3319 	cp_ring_struct = cpr->cp_ring_struct;
3320 	ring_mask = cpr->cp_ring_struct->ring_mask;
3321 
3322 	/* Check to see if hw has posted a completion for the descriptor. */
3323 	while (1) {
3324 		struct tx_cmpl *txcmp;
3325 		uint32_t cons;
3326 
3327 		cons = RING_CMPL(ring_mask, raw_cons);
3328 		txcmp = (struct tx_cmpl *)&cp_desc_ring[cons];
3329 
3330 		if (!CMP_VALID(txcmp, raw_cons, cp_ring_struct))
3331 			break;
3332 
3333 		if (CMP_TYPE(txcmp) == TX_CMPL_TYPE_TX_L2)
3334 			nb_tx_pkts += rte_le_to_cpu_32(txcmp->opaque);
3335 
3336 		if (nb_tx_pkts > offset)
3337 			return RTE_ETH_TX_DESC_DONE;
3338 
3339 		raw_cons = NEXT_RAW_CMP(raw_cons);
3340 	}
3341 
3342 	/* Descriptor is pending transmit, not yet completed by hardware. */
3343 	return RTE_ETH_TX_DESC_FULL;
3344 }
3345 
3346 int
3347 bnxt_flow_ops_get_op(struct rte_eth_dev *dev,
3348 		     const struct rte_flow_ops **ops)
3349 {
3350 	struct bnxt *bp = dev->data->dev_private;
3351 	int ret = 0;
3352 
3353 	if (!bp)
3354 		return -EIO;
3355 
3356 	if (BNXT_ETH_DEV_IS_REPRESENTOR(dev)) {
3357 		struct bnxt_representor *vfr = dev->data->dev_private;
3358 		bp = vfr->parent_dev->data->dev_private;
3359 		/* parent is deleted while children are still valid */
3360 		if (!bp) {
3361 			PMD_DRV_LOG(DEBUG, "BNXT Port:%d VFR Error\n",
3362 				    dev->data->port_id);
3363 			return -EIO;
3364 		}
3365 	}
3366 
3367 	ret = is_bnxt_in_error(bp);
3368 	if (ret)
3369 		return ret;
3370 
3371 	/* PMD supports thread-safe flow operations.  rte_flow API
3372 	 * functions can avoid mutex for multi-thread safety.
3373 	 */
3374 	dev->data->dev_flags |= RTE_ETH_DEV_FLOW_OPS_THREAD_SAFE;
3375 
3376 	if (BNXT_TRUFLOW_EN(bp))
3377 		*ops = &bnxt_ulp_rte_flow_ops;
3378 	else
3379 		*ops = &bnxt_flow_ops;
3380 
3381 	return ret;
3382 }
3383 
3384 static const uint32_t *
3385 bnxt_dev_supported_ptypes_get_op(struct rte_eth_dev *dev)
3386 {
3387 	static const uint32_t ptypes[] = {
3388 		RTE_PTYPE_L2_ETHER_VLAN,
3389 		RTE_PTYPE_L3_IPV4_EXT_UNKNOWN,
3390 		RTE_PTYPE_L3_IPV6_EXT_UNKNOWN,
3391 		RTE_PTYPE_L4_ICMP,
3392 		RTE_PTYPE_L4_TCP,
3393 		RTE_PTYPE_L4_UDP,
3394 		RTE_PTYPE_INNER_L3_IPV4_EXT_UNKNOWN,
3395 		RTE_PTYPE_INNER_L3_IPV6_EXT_UNKNOWN,
3396 		RTE_PTYPE_INNER_L4_ICMP,
3397 		RTE_PTYPE_INNER_L4_TCP,
3398 		RTE_PTYPE_INNER_L4_UDP,
3399 		RTE_PTYPE_UNKNOWN
3400 	};
3401 
3402 	if (!dev->rx_pkt_burst)
3403 		return NULL;
3404 
3405 	return ptypes;
3406 }
3407 
3408 static int bnxt_map_regs(struct bnxt *bp, uint32_t *reg_arr, int count,
3409 			 int reg_win)
3410 {
3411 	uint32_t reg_base = *reg_arr & 0xfffff000;
3412 	uint32_t win_off;
3413 	int i;
3414 
3415 	for (i = 0; i < count; i++) {
3416 		if ((reg_arr[i] & 0xfffff000) != reg_base)
3417 			return -ERANGE;
3418 	}
3419 	win_off = BNXT_GRCPF_REG_WINDOW_BASE_OUT + (reg_win - 1) * 4;
3420 	rte_write32(reg_base, (uint8_t *)bp->bar0 + win_off);
3421 	return 0;
3422 }
3423 
3424 static int bnxt_map_ptp_regs(struct bnxt *bp)
3425 {
3426 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3427 	uint32_t *reg_arr;
3428 	int rc, i;
3429 
3430 	reg_arr = ptp->rx_regs;
3431 	rc = bnxt_map_regs(bp, reg_arr, BNXT_PTP_RX_REGS, 5);
3432 	if (rc)
3433 		return rc;
3434 
3435 	reg_arr = ptp->tx_regs;
3436 	rc = bnxt_map_regs(bp, reg_arr, BNXT_PTP_TX_REGS, 6);
3437 	if (rc)
3438 		return rc;
3439 
3440 	for (i = 0; i < BNXT_PTP_RX_REGS; i++)
3441 		ptp->rx_mapped_regs[i] = 0x5000 + (ptp->rx_regs[i] & 0xfff);
3442 
3443 	for (i = 0; i < BNXT_PTP_TX_REGS; i++)
3444 		ptp->tx_mapped_regs[i] = 0x6000 + (ptp->tx_regs[i] & 0xfff);
3445 
3446 	return 0;
3447 }
3448 
3449 static void bnxt_unmap_ptp_regs(struct bnxt *bp)
3450 {
3451 	rte_write32(0, (uint8_t *)bp->bar0 +
3452 			 BNXT_GRCPF_REG_WINDOW_BASE_OUT + 16);
3453 	rte_write32(0, (uint8_t *)bp->bar0 +
3454 			 BNXT_GRCPF_REG_WINDOW_BASE_OUT + 20);
3455 }
3456 
3457 static uint64_t bnxt_cc_read(struct bnxt *bp)
3458 {
3459 	uint64_t ns;
3460 
3461 	ns = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3462 			      BNXT_GRCPF_REG_SYNC_TIME));
3463 	ns |= (uint64_t)(rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3464 					  BNXT_GRCPF_REG_SYNC_TIME + 4))) << 32;
3465 	return ns;
3466 }
3467 
3468 static int bnxt_get_tx_ts(struct bnxt *bp, uint64_t *ts)
3469 {
3470 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3471 	uint32_t fifo;
3472 
3473 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3474 				ptp->tx_mapped_regs[BNXT_PTP_TX_FIFO]));
3475 	if (fifo & BNXT_PTP_TX_FIFO_EMPTY)
3476 		return -EAGAIN;
3477 
3478 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3479 				ptp->tx_mapped_regs[BNXT_PTP_TX_FIFO]));
3480 	*ts = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3481 				ptp->tx_mapped_regs[BNXT_PTP_TX_TS_L]));
3482 	*ts |= (uint64_t)rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3483 				ptp->tx_mapped_regs[BNXT_PTP_TX_TS_H])) << 32;
3484 	rte_read32((uint8_t *)bp->bar0 + ptp->tx_mapped_regs[BNXT_PTP_TX_SEQ]);
3485 
3486 	return 0;
3487 }
3488 
3489 static int bnxt_clr_rx_ts(struct bnxt *bp, uint64_t *last_ts)
3490 {
3491 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3492 	struct bnxt_pf_info *pf = bp->pf;
3493 	uint16_t port_id;
3494 	int i = 0;
3495 	uint32_t fifo;
3496 
3497 	if (!ptp || (bp->flags & BNXT_FLAG_CHIP_P5))
3498 		return -EINVAL;
3499 
3500 	port_id = pf->port_id;
3501 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3502 				ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO]));
3503 	while ((fifo & BNXT_PTP_RX_FIFO_PENDING) && (i < BNXT_PTP_RX_PND_CNT)) {
3504 		rte_write32(1 << port_id, (uint8_t *)bp->bar0 +
3505 			    ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO_ADV]);
3506 		fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3507 					ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO]));
3508 		*last_ts = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3509 					ptp->rx_mapped_regs[BNXT_PTP_RX_TS_L]));
3510 		*last_ts |= (uint64_t)rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3511 					ptp->rx_mapped_regs[BNXT_PTP_RX_TS_H])) << 32;
3512 		i++;
3513 	}
3514 
3515 	if (i >= BNXT_PTP_RX_PND_CNT)
3516 		return -EBUSY;
3517 
3518 	return 0;
3519 }
3520 
3521 static int bnxt_get_rx_ts(struct bnxt *bp, uint64_t *ts)
3522 {
3523 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3524 	struct bnxt_pf_info *pf = bp->pf;
3525 	uint16_t port_id;
3526 	uint32_t fifo;
3527 
3528 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3529 				ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO]));
3530 	if (!(fifo & BNXT_PTP_RX_FIFO_PENDING))
3531 		return -EAGAIN;
3532 
3533 	port_id = pf->port_id;
3534 	rte_write32(1 << port_id, (uint8_t *)bp->bar0 +
3535 	       ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO_ADV]);
3536 
3537 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3538 				   ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO]));
3539 	if (fifo & BNXT_PTP_RX_FIFO_PENDING)
3540 		return bnxt_clr_rx_ts(bp, ts);
3541 
3542 	*ts = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3543 				ptp->rx_mapped_regs[BNXT_PTP_RX_TS_L]));
3544 	*ts |= (uint64_t)rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3545 				ptp->rx_mapped_regs[BNXT_PTP_RX_TS_H])) << 32;
3546 
3547 	return 0;
3548 }
3549 
3550 static int
3551 bnxt_timesync_write_time(struct rte_eth_dev *dev, const struct timespec *ts)
3552 {
3553 	uint64_t ns;
3554 	struct bnxt *bp = dev->data->dev_private;
3555 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3556 
3557 	if (!ptp)
3558 		return -ENOTSUP;
3559 
3560 	ns = rte_timespec_to_ns(ts);
3561 	/* Set the timecounters to a new value. */
3562 	ptp->tc.nsec = ns;
3563 	ptp->tx_tstamp_tc.nsec = ns;
3564 	ptp->rx_tstamp_tc.nsec = ns;
3565 
3566 	return 0;
3567 }
3568 
3569 static int
3570 bnxt_timesync_read_time(struct rte_eth_dev *dev, struct timespec *ts)
3571 {
3572 	struct bnxt *bp = dev->data->dev_private;
3573 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3574 	uint64_t ns, systime_cycles = 0;
3575 	int rc = 0;
3576 
3577 	if (!ptp)
3578 		return -ENOTSUP;
3579 
3580 	if (BNXT_CHIP_P5(bp))
3581 		rc = bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_CURRENT_TIME,
3582 					     &systime_cycles);
3583 	else
3584 		systime_cycles = bnxt_cc_read(bp);
3585 
3586 	ns = rte_timecounter_update(&ptp->tc, systime_cycles);
3587 	*ts = rte_ns_to_timespec(ns);
3588 
3589 	return rc;
3590 }
3591 static int
3592 bnxt_timesync_enable(struct rte_eth_dev *dev)
3593 {
3594 	struct bnxt *bp = dev->data->dev_private;
3595 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3596 	uint32_t shift = 0;
3597 	int rc;
3598 
3599 	if (!ptp)
3600 		return -ENOTSUP;
3601 
3602 	ptp->rx_filter = 1;
3603 	ptp->tx_tstamp_en = 1;
3604 	ptp->rxctl = BNXT_PTP_MSG_EVENTS;
3605 
3606 	rc = bnxt_hwrm_ptp_cfg(bp);
3607 	if (rc)
3608 		return rc;
3609 
3610 	memset(&ptp->tc, 0, sizeof(struct rte_timecounter));
3611 	memset(&ptp->rx_tstamp_tc, 0, sizeof(struct rte_timecounter));
3612 	memset(&ptp->tx_tstamp_tc, 0, sizeof(struct rte_timecounter));
3613 
3614 	ptp->tc.cc_mask = BNXT_CYCLECOUNTER_MASK;
3615 	ptp->tc.cc_shift = shift;
3616 	ptp->tc.nsec_mask = (1ULL << shift) - 1;
3617 
3618 	ptp->rx_tstamp_tc.cc_mask = BNXT_CYCLECOUNTER_MASK;
3619 	ptp->rx_tstamp_tc.cc_shift = shift;
3620 	ptp->rx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
3621 
3622 	ptp->tx_tstamp_tc.cc_mask = BNXT_CYCLECOUNTER_MASK;
3623 	ptp->tx_tstamp_tc.cc_shift = shift;
3624 	ptp->tx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
3625 
3626 	if (!BNXT_CHIP_P5(bp))
3627 		bnxt_map_ptp_regs(bp);
3628 	else
3629 		rc = bnxt_ptp_start(bp);
3630 
3631 	return rc;
3632 }
3633 
3634 static int
3635 bnxt_timesync_disable(struct rte_eth_dev *dev)
3636 {
3637 	struct bnxt *bp = dev->data->dev_private;
3638 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3639 
3640 	if (!ptp)
3641 		return -ENOTSUP;
3642 
3643 	ptp->rx_filter = 0;
3644 	ptp->tx_tstamp_en = 0;
3645 	ptp->rxctl = 0;
3646 
3647 	bnxt_hwrm_ptp_cfg(bp);
3648 
3649 	if (!BNXT_CHIP_P5(bp))
3650 		bnxt_unmap_ptp_regs(bp);
3651 	else
3652 		bnxt_ptp_stop(bp);
3653 
3654 	return 0;
3655 }
3656 
3657 static int
3658 bnxt_timesync_read_rx_timestamp(struct rte_eth_dev *dev,
3659 				 struct timespec *timestamp,
3660 				 uint32_t flags __rte_unused)
3661 {
3662 	struct bnxt *bp = dev->data->dev_private;
3663 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3664 	uint64_t rx_tstamp_cycles = 0;
3665 	uint64_t ns;
3666 
3667 	if (!ptp)
3668 		return -ENOTSUP;
3669 
3670 	if (BNXT_CHIP_P5(bp))
3671 		rx_tstamp_cycles = ptp->rx_timestamp;
3672 	else
3673 		bnxt_get_rx_ts(bp, &rx_tstamp_cycles);
3674 
3675 	ns = rte_timecounter_update(&ptp->rx_tstamp_tc, rx_tstamp_cycles);
3676 	*timestamp = rte_ns_to_timespec(ns);
3677 	return  0;
3678 }
3679 
3680 static int
3681 bnxt_timesync_read_tx_timestamp(struct rte_eth_dev *dev,
3682 				 struct timespec *timestamp)
3683 {
3684 	struct bnxt *bp = dev->data->dev_private;
3685 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3686 	uint64_t tx_tstamp_cycles = 0;
3687 	uint64_t ns;
3688 	int rc = 0;
3689 
3690 	if (!ptp)
3691 		return -ENOTSUP;
3692 
3693 	if (BNXT_CHIP_P5(bp))
3694 		rc = bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_PATH_TX,
3695 					     &tx_tstamp_cycles);
3696 	else
3697 		rc = bnxt_get_tx_ts(bp, &tx_tstamp_cycles);
3698 
3699 	ns = rte_timecounter_update(&ptp->tx_tstamp_tc, tx_tstamp_cycles);
3700 	*timestamp = rte_ns_to_timespec(ns);
3701 
3702 	return rc;
3703 }
3704 
3705 static int
3706 bnxt_timesync_adjust_time(struct rte_eth_dev *dev, int64_t delta)
3707 {
3708 	struct bnxt *bp = dev->data->dev_private;
3709 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3710 
3711 	if (!ptp)
3712 		return -ENOTSUP;
3713 
3714 	ptp->tc.nsec += delta;
3715 	ptp->tx_tstamp_tc.nsec += delta;
3716 	ptp->rx_tstamp_tc.nsec += delta;
3717 
3718 	return 0;
3719 }
3720 
3721 static int
3722 bnxt_get_eeprom_length_op(struct rte_eth_dev *dev)
3723 {
3724 	struct bnxt *bp = dev->data->dev_private;
3725 	int rc;
3726 	uint32_t dir_entries;
3727 	uint32_t entry_length;
3728 
3729 	rc = is_bnxt_in_error(bp);
3730 	if (rc)
3731 		return rc;
3732 
3733 	PMD_DRV_LOG(INFO, PCI_PRI_FMT "\n",
3734 		    bp->pdev->addr.domain, bp->pdev->addr.bus,
3735 		    bp->pdev->addr.devid, bp->pdev->addr.function);
3736 
3737 	rc = bnxt_hwrm_nvm_get_dir_info(bp, &dir_entries, &entry_length);
3738 	if (rc != 0)
3739 		return rc;
3740 
3741 	return dir_entries * entry_length;
3742 }
3743 
3744 static int
3745 bnxt_get_eeprom_op(struct rte_eth_dev *dev,
3746 		struct rte_dev_eeprom_info *in_eeprom)
3747 {
3748 	struct bnxt *bp = dev->data->dev_private;
3749 	uint32_t index;
3750 	uint32_t offset;
3751 	int rc;
3752 
3753 	rc = is_bnxt_in_error(bp);
3754 	if (rc)
3755 		return rc;
3756 
3757 	PMD_DRV_LOG(INFO, PCI_PRI_FMT " in_eeprom->offset = %d len = %d\n",
3758 		    bp->pdev->addr.domain, bp->pdev->addr.bus,
3759 		    bp->pdev->addr.devid, bp->pdev->addr.function,
3760 		    in_eeprom->offset, in_eeprom->length);
3761 
3762 	if (in_eeprom->offset == 0) /* special offset value to get directory */
3763 		return bnxt_get_nvram_directory(bp, in_eeprom->length,
3764 						in_eeprom->data);
3765 
3766 	index = in_eeprom->offset >> 24;
3767 	offset = in_eeprom->offset & 0xffffff;
3768 
3769 	if (index != 0)
3770 		return bnxt_hwrm_get_nvram_item(bp, index - 1, offset,
3771 					   in_eeprom->length, in_eeprom->data);
3772 
3773 	return 0;
3774 }
3775 
3776 static bool bnxt_dir_type_is_ape_bin_format(uint16_t dir_type)
3777 {
3778 	switch (dir_type) {
3779 	case BNX_DIR_TYPE_CHIMP_PATCH:
3780 	case BNX_DIR_TYPE_BOOTCODE:
3781 	case BNX_DIR_TYPE_BOOTCODE_2:
3782 	case BNX_DIR_TYPE_APE_FW:
3783 	case BNX_DIR_TYPE_APE_PATCH:
3784 	case BNX_DIR_TYPE_KONG_FW:
3785 	case BNX_DIR_TYPE_KONG_PATCH:
3786 	case BNX_DIR_TYPE_BONO_FW:
3787 	case BNX_DIR_TYPE_BONO_PATCH:
3788 		/* FALLTHROUGH */
3789 		return true;
3790 	}
3791 
3792 	return false;
3793 }
3794 
3795 static bool bnxt_dir_type_is_other_exec_format(uint16_t dir_type)
3796 {
3797 	switch (dir_type) {
3798 	case BNX_DIR_TYPE_AVS:
3799 	case BNX_DIR_TYPE_EXP_ROM_MBA:
3800 	case BNX_DIR_TYPE_PCIE:
3801 	case BNX_DIR_TYPE_TSCF_UCODE:
3802 	case BNX_DIR_TYPE_EXT_PHY:
3803 	case BNX_DIR_TYPE_CCM:
3804 	case BNX_DIR_TYPE_ISCSI_BOOT:
3805 	case BNX_DIR_TYPE_ISCSI_BOOT_IPV6:
3806 	case BNX_DIR_TYPE_ISCSI_BOOT_IPV4N6:
3807 		/* FALLTHROUGH */
3808 		return true;
3809 	}
3810 
3811 	return false;
3812 }
3813 
3814 static bool bnxt_dir_type_is_executable(uint16_t dir_type)
3815 {
3816 	return bnxt_dir_type_is_ape_bin_format(dir_type) ||
3817 		bnxt_dir_type_is_other_exec_format(dir_type);
3818 }
3819 
3820 static int
3821 bnxt_set_eeprom_op(struct rte_eth_dev *dev,
3822 		struct rte_dev_eeprom_info *in_eeprom)
3823 {
3824 	struct bnxt *bp = dev->data->dev_private;
3825 	uint8_t index, dir_op;
3826 	uint16_t type, ext, ordinal, attr;
3827 	int rc;
3828 
3829 	rc = is_bnxt_in_error(bp);
3830 	if (rc)
3831 		return rc;
3832 
3833 	PMD_DRV_LOG(INFO, PCI_PRI_FMT " in_eeprom->offset = %d len = %d\n",
3834 		    bp->pdev->addr.domain, bp->pdev->addr.bus,
3835 		    bp->pdev->addr.devid, bp->pdev->addr.function,
3836 		    in_eeprom->offset, in_eeprom->length);
3837 
3838 	if (!BNXT_PF(bp)) {
3839 		PMD_DRV_LOG(ERR, "NVM write not supported from a VF\n");
3840 		return -EINVAL;
3841 	}
3842 
3843 	type = in_eeprom->magic >> 16;
3844 
3845 	if (type == 0xffff) { /* special value for directory operations */
3846 		index = in_eeprom->magic & 0xff;
3847 		dir_op = in_eeprom->magic >> 8;
3848 		if (index == 0)
3849 			return -EINVAL;
3850 		switch (dir_op) {
3851 		case 0x0e: /* erase */
3852 			if (in_eeprom->offset != ~in_eeprom->magic)
3853 				return -EINVAL;
3854 			return bnxt_hwrm_erase_nvram_directory(bp, index - 1);
3855 		default:
3856 			return -EINVAL;
3857 		}
3858 	}
3859 
3860 	/* Create or re-write an NVM item: */
3861 	if (bnxt_dir_type_is_executable(type) == true)
3862 		return -EOPNOTSUPP;
3863 	ext = in_eeprom->magic & 0xffff;
3864 	ordinal = in_eeprom->offset >> 16;
3865 	attr = in_eeprom->offset & 0xffff;
3866 
3867 	return bnxt_hwrm_flash_nvram(bp, type, ordinal, ext, attr,
3868 				     in_eeprom->data, in_eeprom->length);
3869 }
3870 
3871 static int bnxt_get_module_info(struct rte_eth_dev *dev,
3872 				struct rte_eth_dev_module_info *modinfo)
3873 {
3874 	uint8_t module_info[SFF_DIAG_SUPPORT_OFFSET + 1];
3875 	struct bnxt *bp = dev->data->dev_private;
3876 	int rc;
3877 
3878 	/* No point in going further if phy status indicates
3879 	 * module is not inserted or if it is powered down or
3880 	 * if it is of type 10GBase-T
3881 	 */
3882 	if (bp->link_info->module_status >
3883 	    HWRM_PORT_PHY_QCFG_OUTPUT_MODULE_STATUS_WARNINGMSG) {
3884 		PMD_DRV_LOG(NOTICE, "Port %u : Module is not inserted or is powered down\n",
3885 			    dev->data->port_id);
3886 		return -ENOTSUP;
3887 	}
3888 
3889 	/* This feature is not supported in older firmware versions */
3890 	if (bp->hwrm_spec_code < 0x10202) {
3891 		PMD_DRV_LOG(NOTICE, "Port %u : Feature is not supported in older firmware\n",
3892 			    dev->data->port_id);
3893 		return -ENOTSUP;
3894 	}
3895 
3896 	rc = bnxt_hwrm_read_sfp_module_eeprom_info(bp, I2C_DEV_ADDR_A0, 0, 0,
3897 						   SFF_DIAG_SUPPORT_OFFSET + 1,
3898 						   module_info);
3899 
3900 	if (rc)
3901 		return rc;
3902 
3903 	switch (module_info[0]) {
3904 	case SFF_MODULE_ID_SFP:
3905 		modinfo->type = RTE_ETH_MODULE_SFF_8472;
3906 		modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8472_LEN;
3907 		if (module_info[SFF_DIAG_SUPPORT_OFFSET] == 0)
3908 			modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8436_LEN;
3909 		break;
3910 	case SFF_MODULE_ID_QSFP:
3911 	case SFF_MODULE_ID_QSFP_PLUS:
3912 		modinfo->type = RTE_ETH_MODULE_SFF_8436;
3913 		modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8436_LEN;
3914 		break;
3915 	case SFF_MODULE_ID_QSFP28:
3916 		modinfo->type = RTE_ETH_MODULE_SFF_8636;
3917 		modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8636_MAX_LEN;
3918 		if (module_info[SFF8636_FLATMEM_OFFSET] & SFF8636_FLATMEM_MASK)
3919 			modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8636_LEN;
3920 		break;
3921 	default:
3922 		PMD_DRV_LOG(NOTICE, "Port %u : Unsupported module\n", dev->data->port_id);
3923 		return -ENOTSUP;
3924 	}
3925 
3926 	PMD_DRV_LOG(INFO, "Port %u : modinfo->type = %d modinfo->eeprom_len = %d\n",
3927 		    dev->data->port_id, modinfo->type, modinfo->eeprom_len);
3928 
3929 	return 0;
3930 }
3931 
3932 static int bnxt_get_module_eeprom(struct rte_eth_dev *dev,
3933 				  struct rte_dev_eeprom_info *info)
3934 {
3935 	uint8_t pg_addr[5] = { I2C_DEV_ADDR_A0, I2C_DEV_ADDR_A0 };
3936 	uint32_t offset = info->offset, length = info->length;
3937 	uint8_t module_info[SFF_DIAG_SUPPORT_OFFSET + 1];
3938 	struct bnxt *bp = dev->data->dev_private;
3939 	uint8_t *data = info->data;
3940 	uint8_t page = offset >> 7;
3941 	uint8_t max_pages = 2;
3942 	uint8_t opt_pages;
3943 	int rc;
3944 
3945 	rc = bnxt_hwrm_read_sfp_module_eeprom_info(bp, I2C_DEV_ADDR_A0, 0, 0,
3946 						   SFF_DIAG_SUPPORT_OFFSET + 1,
3947 						   module_info);
3948 	if (rc)
3949 		return rc;
3950 
3951 	switch (module_info[0]) {
3952 	case SFF_MODULE_ID_SFP:
3953 		module_info[SFF_DIAG_SUPPORT_OFFSET] = 0;
3954 		if (module_info[SFF_DIAG_SUPPORT_OFFSET]) {
3955 			pg_addr[2] = I2C_DEV_ADDR_A2;
3956 			pg_addr[3] = I2C_DEV_ADDR_A2;
3957 			max_pages = 4;
3958 		}
3959 		break;
3960 	case SFF_MODULE_ID_QSFP28:
3961 		rc = bnxt_hwrm_read_sfp_module_eeprom_info(bp, I2C_DEV_ADDR_A0, 0,
3962 							   SFF8636_OPT_PAGES_OFFSET,
3963 							   1, &opt_pages);
3964 		if (rc)
3965 			return rc;
3966 
3967 		if (opt_pages & SFF8636_PAGE1_MASK) {
3968 			pg_addr[2] = I2C_DEV_ADDR_A0;
3969 			max_pages = 3;
3970 		}
3971 		if (opt_pages & SFF8636_PAGE2_MASK) {
3972 			pg_addr[3] = I2C_DEV_ADDR_A0;
3973 			max_pages = 4;
3974 		}
3975 		if (~module_info[SFF8636_FLATMEM_OFFSET] & SFF8636_FLATMEM_MASK) {
3976 			pg_addr[4] = I2C_DEV_ADDR_A0;
3977 			max_pages = 5;
3978 		}
3979 		break;
3980 	default:
3981 		break;
3982 	}
3983 
3984 	memset(data, 0, length);
3985 
3986 	offset &= 0xff;
3987 	while (length && page < max_pages) {
3988 		uint8_t raw_page = page ? page - 1 : 0;
3989 		uint16_t chunk;
3990 
3991 		if (pg_addr[page] == I2C_DEV_ADDR_A2)
3992 			raw_page = 0;
3993 		else if (page)
3994 			offset |= 0x80;
3995 		chunk = RTE_MIN(length, 256 - offset);
3996 
3997 		if (pg_addr[page]) {
3998 			rc = bnxt_hwrm_read_sfp_module_eeprom_info(bp, pg_addr[page],
3999 								   raw_page, offset,
4000 								   chunk, data);
4001 			if (rc)
4002 				return rc;
4003 		}
4004 
4005 		data += chunk;
4006 		length -= chunk;
4007 		offset = 0;
4008 		page += 1 + (chunk > 128);
4009 	}
4010 
4011 	return length ? -EINVAL : 0;
4012 }
4013 
4014 /*
4015  * Initialization
4016  */
4017 
4018 static const struct eth_dev_ops bnxt_dev_ops = {
4019 	.dev_infos_get = bnxt_dev_info_get_op,
4020 	.dev_close = bnxt_dev_close_op,
4021 	.dev_configure = bnxt_dev_configure_op,
4022 	.dev_start = bnxt_dev_start_op,
4023 	.dev_stop = bnxt_dev_stop_op,
4024 	.dev_set_link_up = bnxt_dev_set_link_up_op,
4025 	.dev_set_link_down = bnxt_dev_set_link_down_op,
4026 	.stats_get = bnxt_stats_get_op,
4027 	.stats_reset = bnxt_stats_reset_op,
4028 	.rx_queue_setup = bnxt_rx_queue_setup_op,
4029 	.rx_queue_release = bnxt_rx_queue_release_op,
4030 	.tx_queue_setup = bnxt_tx_queue_setup_op,
4031 	.tx_queue_release = bnxt_tx_queue_release_op,
4032 	.rx_queue_intr_enable = bnxt_rx_queue_intr_enable_op,
4033 	.rx_queue_intr_disable = bnxt_rx_queue_intr_disable_op,
4034 	.reta_update = bnxt_reta_update_op,
4035 	.reta_query = bnxt_reta_query_op,
4036 	.rss_hash_update = bnxt_rss_hash_update_op,
4037 	.rss_hash_conf_get = bnxt_rss_hash_conf_get_op,
4038 	.link_update = bnxt_link_update_op,
4039 	.promiscuous_enable = bnxt_promiscuous_enable_op,
4040 	.promiscuous_disable = bnxt_promiscuous_disable_op,
4041 	.allmulticast_enable = bnxt_allmulticast_enable_op,
4042 	.allmulticast_disable = bnxt_allmulticast_disable_op,
4043 	.mac_addr_add = bnxt_mac_addr_add_op,
4044 	.mac_addr_remove = bnxt_mac_addr_remove_op,
4045 	.flow_ctrl_get = bnxt_flow_ctrl_get_op,
4046 	.flow_ctrl_set = bnxt_flow_ctrl_set_op,
4047 	.udp_tunnel_port_add  = bnxt_udp_tunnel_port_add_op,
4048 	.udp_tunnel_port_del  = bnxt_udp_tunnel_port_del_op,
4049 	.vlan_filter_set = bnxt_vlan_filter_set_op,
4050 	.vlan_offload_set = bnxt_vlan_offload_set_op,
4051 	.vlan_tpid_set = bnxt_vlan_tpid_set_op,
4052 	.vlan_pvid_set = bnxt_vlan_pvid_set_op,
4053 	.mtu_set = bnxt_mtu_set_op,
4054 	.mac_addr_set = bnxt_set_default_mac_addr_op,
4055 	.xstats_get = bnxt_dev_xstats_get_op,
4056 	.xstats_get_names = bnxt_dev_xstats_get_names_op,
4057 	.xstats_reset = bnxt_dev_xstats_reset_op,
4058 	.fw_version_get = bnxt_fw_version_get,
4059 	.set_mc_addr_list = bnxt_dev_set_mc_addr_list_op,
4060 	.rxq_info_get = bnxt_rxq_info_get_op,
4061 	.txq_info_get = bnxt_txq_info_get_op,
4062 	.rx_burst_mode_get = bnxt_rx_burst_mode_get,
4063 	.tx_burst_mode_get = bnxt_tx_burst_mode_get,
4064 	.dev_led_on = bnxt_dev_led_on_op,
4065 	.dev_led_off = bnxt_dev_led_off_op,
4066 	.rx_queue_start = bnxt_rx_queue_start,
4067 	.rx_queue_stop = bnxt_rx_queue_stop,
4068 	.tx_queue_start = bnxt_tx_queue_start,
4069 	.tx_queue_stop = bnxt_tx_queue_stop,
4070 	.flow_ops_get = bnxt_flow_ops_get_op,
4071 	.dev_supported_ptypes_get = bnxt_dev_supported_ptypes_get_op,
4072 	.get_eeprom_length    = bnxt_get_eeprom_length_op,
4073 	.get_eeprom           = bnxt_get_eeprom_op,
4074 	.set_eeprom           = bnxt_set_eeprom_op,
4075 	.get_module_info = bnxt_get_module_info,
4076 	.get_module_eeprom = bnxt_get_module_eeprom,
4077 	.timesync_enable      = bnxt_timesync_enable,
4078 	.timesync_disable     = bnxt_timesync_disable,
4079 	.timesync_read_time   = bnxt_timesync_read_time,
4080 	.timesync_write_time   = bnxt_timesync_write_time,
4081 	.timesync_adjust_time = bnxt_timesync_adjust_time,
4082 	.timesync_read_rx_timestamp = bnxt_timesync_read_rx_timestamp,
4083 	.timesync_read_tx_timestamp = bnxt_timesync_read_tx_timestamp,
4084 };
4085 
4086 static uint32_t bnxt_map_reset_regs(struct bnxt *bp, uint32_t reg)
4087 {
4088 	uint32_t offset;
4089 
4090 	/* Only pre-map the reset GRC registers using window 3 */
4091 	rte_write32(reg & 0xfffff000, (uint8_t *)bp->bar0 +
4092 		    BNXT_GRCPF_REG_WINDOW_BASE_OUT + 8);
4093 
4094 	offset = BNXT_GRCP_WINDOW_3_BASE + (reg & 0xffc);
4095 
4096 	return offset;
4097 }
4098 
4099 int bnxt_map_fw_health_status_regs(struct bnxt *bp)
4100 {
4101 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4102 	uint32_t reg_base = 0xffffffff;
4103 	int i;
4104 
4105 	/* Only pre-map the monitoring GRC registers using window 2 */
4106 	for (i = 0; i < BNXT_FW_STATUS_REG_CNT; i++) {
4107 		uint32_t reg = info->status_regs[i];
4108 
4109 		if (BNXT_FW_STATUS_REG_TYPE(reg) != BNXT_FW_STATUS_REG_TYPE_GRC)
4110 			continue;
4111 
4112 		if (reg_base == 0xffffffff)
4113 			reg_base = reg & 0xfffff000;
4114 		if ((reg & 0xfffff000) != reg_base)
4115 			return -ERANGE;
4116 
4117 		/* Use mask 0xffc as the Lower 2 bits indicates
4118 		 * address space location
4119 		 */
4120 		info->mapped_status_regs[i] = BNXT_GRCP_WINDOW_2_BASE +
4121 						(reg & 0xffc);
4122 	}
4123 
4124 	if (reg_base == 0xffffffff)
4125 		return 0;
4126 
4127 	rte_write32(reg_base, (uint8_t *)bp->bar0 +
4128 		    BNXT_GRCPF_REG_WINDOW_BASE_OUT + 4);
4129 
4130 	return 0;
4131 }
4132 
4133 static void bnxt_write_fw_reset_reg(struct bnxt *bp, uint32_t index)
4134 {
4135 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4136 	uint32_t delay = info->delay_after_reset[index];
4137 	uint32_t val = info->reset_reg_val[index];
4138 	uint32_t reg = info->reset_reg[index];
4139 	uint32_t type, offset;
4140 	int ret;
4141 
4142 	type = BNXT_FW_STATUS_REG_TYPE(reg);
4143 	offset = BNXT_FW_STATUS_REG_OFF(reg);
4144 
4145 	switch (type) {
4146 	case BNXT_FW_STATUS_REG_TYPE_CFG:
4147 		ret = rte_pci_write_config(bp->pdev, &val, sizeof(val), offset);
4148 		if (ret < 0) {
4149 			PMD_DRV_LOG(ERR, "Failed to write %#x at PCI offset %#x",
4150 				    val, offset);
4151 			return;
4152 		}
4153 		break;
4154 	case BNXT_FW_STATUS_REG_TYPE_GRC:
4155 		offset = bnxt_map_reset_regs(bp, offset);
4156 		rte_write32(val, (uint8_t *)bp->bar0 + offset);
4157 		break;
4158 	case BNXT_FW_STATUS_REG_TYPE_BAR0:
4159 		rte_write32(val, (uint8_t *)bp->bar0 + offset);
4160 		break;
4161 	}
4162 	/* wait on a specific interval of time until core reset is complete */
4163 	if (delay)
4164 		rte_delay_ms(delay);
4165 }
4166 
4167 static void bnxt_dev_cleanup(struct bnxt *bp)
4168 {
4169 	bp->eth_dev->data->dev_link.link_status = 0;
4170 	bp->link_info->link_up = 0;
4171 	if (bp->eth_dev->data->dev_started)
4172 		bnxt_dev_stop(bp->eth_dev);
4173 
4174 	bnxt_uninit_resources(bp, true);
4175 }
4176 
4177 static int
4178 bnxt_check_fw_reset_done(struct bnxt *bp)
4179 {
4180 	int timeout = bp->fw_reset_max_msecs;
4181 	uint16_t val = 0;
4182 	int rc;
4183 
4184 	do {
4185 		rc = rte_pci_read_config(bp->pdev, &val, sizeof(val), PCI_SUBSYSTEM_ID_OFFSET);
4186 		if (rc < 0) {
4187 			PMD_DRV_LOG(ERR, "Failed to read PCI offset 0x%x", PCI_SUBSYSTEM_ID_OFFSET);
4188 			return rc;
4189 		}
4190 		if (val != 0xffff)
4191 			break;
4192 		rte_delay_ms(1);
4193 	} while (timeout--);
4194 
4195 	if (val == 0xffff) {
4196 		PMD_DRV_LOG(ERR, "Firmware reset aborted, PCI config space invalid\n");
4197 		return -1;
4198 	}
4199 
4200 	return 0;
4201 }
4202 
4203 static int bnxt_restore_vlan_filters(struct bnxt *bp)
4204 {
4205 	struct rte_eth_dev *dev = bp->eth_dev;
4206 	struct rte_vlan_filter_conf *vfc;
4207 	int vidx, vbit, rc;
4208 	uint16_t vlan_id;
4209 
4210 	for (vlan_id = 1; vlan_id <= RTE_ETHER_MAX_VLAN_ID; vlan_id++) {
4211 		vfc = &dev->data->vlan_filter_conf;
4212 		vidx = vlan_id / 64;
4213 		vbit = vlan_id % 64;
4214 
4215 		/* Each bit corresponds to a VLAN id */
4216 		if (vfc->ids[vidx] & (UINT64_C(1) << vbit)) {
4217 			rc = bnxt_add_vlan_filter(bp, vlan_id);
4218 			if (rc)
4219 				return rc;
4220 		}
4221 	}
4222 
4223 	return 0;
4224 }
4225 
4226 static int bnxt_restore_mac_filters(struct bnxt *bp)
4227 {
4228 	struct rte_eth_dev *dev = bp->eth_dev;
4229 	struct rte_eth_dev_info dev_info;
4230 	struct rte_ether_addr *addr;
4231 	uint64_t pool_mask;
4232 	uint32_t pool = 0;
4233 	uint32_t i;
4234 	int rc;
4235 
4236 	if (BNXT_VF(bp) && !BNXT_VF_IS_TRUSTED(bp))
4237 		return 0;
4238 
4239 	rc = bnxt_dev_info_get_op(dev, &dev_info);
4240 	if (rc)
4241 		return rc;
4242 
4243 	/* replay MAC address configuration */
4244 	for (i = 1; i < dev_info.max_mac_addrs; i++) {
4245 		addr = &dev->data->mac_addrs[i];
4246 
4247 		/* skip zero address */
4248 		if (rte_is_zero_ether_addr(addr))
4249 			continue;
4250 
4251 		pool = 0;
4252 		pool_mask = dev->data->mac_pool_sel[i];
4253 
4254 		do {
4255 			if (pool_mask & 1ULL) {
4256 				rc = bnxt_mac_addr_add_op(dev, addr, i, pool);
4257 				if (rc)
4258 					return rc;
4259 			}
4260 			pool_mask >>= 1;
4261 			pool++;
4262 		} while (pool_mask);
4263 	}
4264 
4265 	return 0;
4266 }
4267 
4268 static int bnxt_restore_filters(struct bnxt *bp)
4269 {
4270 	struct rte_eth_dev *dev = bp->eth_dev;
4271 	int ret = 0;
4272 
4273 	if (dev->data->all_multicast) {
4274 		ret = bnxt_allmulticast_enable_op(dev);
4275 		if (ret)
4276 			return ret;
4277 	}
4278 	if (dev->data->promiscuous) {
4279 		ret = bnxt_promiscuous_enable_op(dev);
4280 		if (ret)
4281 			return ret;
4282 	}
4283 
4284 	ret = bnxt_restore_mac_filters(bp);
4285 	if (ret)
4286 		return ret;
4287 
4288 	ret = bnxt_restore_vlan_filters(bp);
4289 	/* TODO restore other filters as well */
4290 	return ret;
4291 }
4292 
4293 static int bnxt_check_fw_ready(struct bnxt *bp)
4294 {
4295 	int timeout = bp->fw_reset_max_msecs;
4296 	int rc = 0;
4297 
4298 	do {
4299 		rc = bnxt_hwrm_poll_ver_get(bp);
4300 		if (rc == 0)
4301 			break;
4302 		rte_delay_ms(BNXT_FW_READY_WAIT_INTERVAL);
4303 		timeout -= BNXT_FW_READY_WAIT_INTERVAL;
4304 	} while (rc && timeout > 0);
4305 
4306 	if (rc)
4307 		PMD_DRV_LOG(ERR, "FW is not Ready after reset\n");
4308 
4309 	return rc;
4310 }
4311 
4312 static void bnxt_dev_recover(void *arg)
4313 {
4314 	struct bnxt *bp = arg;
4315 	int rc = 0;
4316 
4317 	pthread_mutex_lock(&bp->err_recovery_lock);
4318 
4319 	if (!bp->fw_reset_min_msecs) {
4320 		rc = bnxt_check_fw_reset_done(bp);
4321 		if (rc)
4322 			goto err;
4323 	}
4324 
4325 	/* Clear Error flag so that device re-init should happen */
4326 	bp->flags &= ~BNXT_FLAG_FATAL_ERROR;
4327 
4328 	rc = bnxt_check_fw_ready(bp);
4329 	if (rc)
4330 		goto err;
4331 
4332 	rc = bnxt_init_resources(bp, true);
4333 	if (rc) {
4334 		PMD_DRV_LOG(ERR,
4335 			    "Failed to initialize resources after reset\n");
4336 		goto err;
4337 	}
4338 	/* clear reset flag as the device is initialized now */
4339 	bp->flags &= ~BNXT_FLAG_FW_RESET;
4340 
4341 	rc = bnxt_dev_start_op(bp->eth_dev);
4342 	if (rc) {
4343 		PMD_DRV_LOG(ERR, "Failed to start port after reset\n");
4344 		goto err_start;
4345 	}
4346 
4347 	rc = bnxt_restore_filters(bp);
4348 	if (rc)
4349 		goto err_start;
4350 
4351 	PMD_DRV_LOG(INFO, "Recovered from FW reset\n");
4352 	pthread_mutex_unlock(&bp->err_recovery_lock);
4353 
4354 	return;
4355 err_start:
4356 	bnxt_dev_stop(bp->eth_dev);
4357 err:
4358 	bp->flags |= BNXT_FLAG_FATAL_ERROR;
4359 	bnxt_uninit_resources(bp, false);
4360 	if (bp->eth_dev->data->dev_conf.intr_conf.rmv)
4361 		rte_eth_dev_callback_process(bp->eth_dev,
4362 					     RTE_ETH_EVENT_INTR_RMV,
4363 					     NULL);
4364 	pthread_mutex_unlock(&bp->err_recovery_lock);
4365 	PMD_DRV_LOG(ERR, "Failed to recover from FW reset\n");
4366 }
4367 
4368 void bnxt_dev_reset_and_resume(void *arg)
4369 {
4370 	struct bnxt *bp = arg;
4371 	uint32_t us = US_PER_MS * bp->fw_reset_min_msecs;
4372 	uint16_t val = 0;
4373 	int rc;
4374 
4375 	bnxt_dev_cleanup(bp);
4376 
4377 	bnxt_wait_for_device_shutdown(bp);
4378 
4379 	/* During some fatal firmware error conditions, the PCI config space
4380 	 * register 0x2e which normally contains the subsystem ID will become
4381 	 * 0xffff. This register will revert back to the normal value after
4382 	 * the chip has completed core reset. If we detect this condition,
4383 	 * we can poll this config register immediately for the value to revert.
4384 	 */
4385 	if (bp->flags & BNXT_FLAG_FATAL_ERROR) {
4386 		rc = rte_pci_read_config(bp->pdev, &val, sizeof(val), PCI_SUBSYSTEM_ID_OFFSET);
4387 		if (rc < 0) {
4388 			PMD_DRV_LOG(ERR, "Failed to read PCI offset 0x%x", PCI_SUBSYSTEM_ID_OFFSET);
4389 			return;
4390 		}
4391 		if (val == 0xffff) {
4392 			bp->fw_reset_min_msecs = 0;
4393 			us = 1;
4394 		}
4395 	}
4396 
4397 	rc = rte_eal_alarm_set(us, bnxt_dev_recover, (void *)bp);
4398 	if (rc)
4399 		PMD_DRV_LOG(ERR, "Error setting recovery alarm");
4400 }
4401 
4402 uint32_t bnxt_read_fw_status_reg(struct bnxt *bp, uint32_t index)
4403 {
4404 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4405 	uint32_t reg = info->status_regs[index];
4406 	uint32_t type, offset, val = 0;
4407 	int ret = 0;
4408 
4409 	type = BNXT_FW_STATUS_REG_TYPE(reg);
4410 	offset = BNXT_FW_STATUS_REG_OFF(reg);
4411 
4412 	switch (type) {
4413 	case BNXT_FW_STATUS_REG_TYPE_CFG:
4414 		ret = rte_pci_read_config(bp->pdev, &val, sizeof(val), offset);
4415 		if (ret < 0)
4416 			PMD_DRV_LOG(ERR, "Failed to read PCI offset %#x",
4417 				    offset);
4418 		break;
4419 	case BNXT_FW_STATUS_REG_TYPE_GRC:
4420 		offset = info->mapped_status_regs[index];
4421 		/* FALLTHROUGH */
4422 	case BNXT_FW_STATUS_REG_TYPE_BAR0:
4423 		val = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
4424 				       offset));
4425 		break;
4426 	}
4427 
4428 	return val;
4429 }
4430 
4431 static int bnxt_fw_reset_all(struct bnxt *bp)
4432 {
4433 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4434 	uint32_t i;
4435 	int rc = 0;
4436 
4437 	if (info->flags & BNXT_FLAG_ERROR_RECOVERY_HOST) {
4438 		/* Reset through master function driver */
4439 		for (i = 0; i < info->reg_array_cnt; i++)
4440 			bnxt_write_fw_reset_reg(bp, i);
4441 		/* Wait for time specified by FW after triggering reset */
4442 		rte_delay_ms(info->master_func_wait_period_after_reset);
4443 	} else if (info->flags & BNXT_FLAG_ERROR_RECOVERY_CO_CPU) {
4444 		/* Reset with the help of Kong processor */
4445 		rc = bnxt_hwrm_fw_reset(bp);
4446 		if (rc)
4447 			PMD_DRV_LOG(ERR, "Failed to reset FW\n");
4448 	}
4449 
4450 	return rc;
4451 }
4452 
4453 static void bnxt_fw_reset_cb(void *arg)
4454 {
4455 	struct bnxt *bp = arg;
4456 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4457 	int rc = 0;
4458 
4459 	/* Only Master function can do FW reset */
4460 	if (bnxt_is_master_func(bp) &&
4461 	    bnxt_is_recovery_enabled(bp)) {
4462 		rc = bnxt_fw_reset_all(bp);
4463 		if (rc) {
4464 			PMD_DRV_LOG(ERR, "Adapter recovery failed\n");
4465 			return;
4466 		}
4467 	}
4468 
4469 	/* if recovery method is ERROR_RECOVERY_CO_CPU, KONG will send
4470 	 * EXCEPTION_FATAL_ASYNC event to all the functions
4471 	 * (including MASTER FUNC). After receiving this Async, all the active
4472 	 * drivers should treat this case as FW initiated recovery
4473 	 */
4474 	if (info->flags & BNXT_FLAG_ERROR_RECOVERY_HOST) {
4475 		bp->fw_reset_min_msecs = BNXT_MIN_FW_READY_TIMEOUT;
4476 		bp->fw_reset_max_msecs = BNXT_MAX_FW_RESET_TIMEOUT;
4477 
4478 		/* To recover from error */
4479 		rte_eal_alarm_set(US_PER_MS, bnxt_dev_reset_and_resume,
4480 				  (void *)bp);
4481 	}
4482 }
4483 
4484 /* Driver should poll FW heartbeat, reset_counter with the frequency
4485  * advertised by FW in HWRM_ERROR_RECOVERY_QCFG.
4486  * When the driver detects heartbeat stop or change in reset_counter,
4487  * it has to trigger a reset to recover from the error condition.
4488  * A “master PF” is the function who will have the privilege to
4489  * initiate the chimp reset. The master PF will be elected by the
4490  * firmware and will be notified through async message.
4491  */
4492 static void bnxt_check_fw_health(void *arg)
4493 {
4494 	struct bnxt *bp = arg;
4495 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4496 	uint32_t val = 0, wait_msec;
4497 
4498 	if (!info || !bnxt_is_recovery_enabled(bp) ||
4499 	    is_bnxt_in_error(bp))
4500 		return;
4501 
4502 	val = bnxt_read_fw_status_reg(bp, BNXT_FW_HEARTBEAT_CNT_REG);
4503 	if (val == info->last_heart_beat)
4504 		goto reset;
4505 
4506 	info->last_heart_beat = val;
4507 
4508 	val = bnxt_read_fw_status_reg(bp, BNXT_FW_RECOVERY_CNT_REG);
4509 	if (val != info->last_reset_counter)
4510 		goto reset;
4511 
4512 	info->last_reset_counter = val;
4513 
4514 	rte_eal_alarm_set(US_PER_MS * info->driver_polling_freq,
4515 			  bnxt_check_fw_health, (void *)bp);
4516 
4517 	return;
4518 reset:
4519 	/* Stop DMA to/from device */
4520 	bp->flags |= BNXT_FLAG_FATAL_ERROR;
4521 	bp->flags |= BNXT_FLAG_FW_RESET;
4522 
4523 	bnxt_stop_rxtx(bp);
4524 
4525 	PMD_DRV_LOG(ERR, "Detected FW dead condition\n");
4526 
4527 	if (bnxt_is_master_func(bp))
4528 		wait_msec = info->master_func_wait_period;
4529 	else
4530 		wait_msec = info->normal_func_wait_period;
4531 
4532 	rte_eal_alarm_set(US_PER_MS * wait_msec,
4533 			  bnxt_fw_reset_cb, (void *)bp);
4534 }
4535 
4536 void bnxt_schedule_fw_health_check(struct bnxt *bp)
4537 {
4538 	uint32_t polling_freq;
4539 
4540 	pthread_mutex_lock(&bp->health_check_lock);
4541 
4542 	if (!bnxt_is_recovery_enabled(bp))
4543 		goto done;
4544 
4545 	if (bp->flags & BNXT_FLAG_FW_HEALTH_CHECK_SCHEDULED)
4546 		goto done;
4547 
4548 	polling_freq = bp->recovery_info->driver_polling_freq;
4549 
4550 	rte_eal_alarm_set(US_PER_MS * polling_freq,
4551 			  bnxt_check_fw_health, (void *)bp);
4552 	bp->flags |= BNXT_FLAG_FW_HEALTH_CHECK_SCHEDULED;
4553 
4554 done:
4555 	pthread_mutex_unlock(&bp->health_check_lock);
4556 }
4557 
4558 static void bnxt_cancel_fw_health_check(struct bnxt *bp)
4559 {
4560 	rte_eal_alarm_cancel(bnxt_check_fw_health, (void *)bp);
4561 	bp->flags &= ~BNXT_FLAG_FW_HEALTH_CHECK_SCHEDULED;
4562 }
4563 
4564 static bool bnxt_vf_pciid(uint16_t device_id)
4565 {
4566 	switch (device_id) {
4567 	case BROADCOM_DEV_ID_57304_VF:
4568 	case BROADCOM_DEV_ID_57406_VF:
4569 	case BROADCOM_DEV_ID_5731X_VF:
4570 	case BROADCOM_DEV_ID_5741X_VF:
4571 	case BROADCOM_DEV_ID_57414_VF:
4572 	case BROADCOM_DEV_ID_STRATUS_NIC_VF1:
4573 	case BROADCOM_DEV_ID_STRATUS_NIC_VF2:
4574 	case BROADCOM_DEV_ID_58802_VF:
4575 	case BROADCOM_DEV_ID_57500_VF1:
4576 	case BROADCOM_DEV_ID_57500_VF2:
4577 	case BROADCOM_DEV_ID_58818_VF:
4578 		/* FALLTHROUGH */
4579 		return true;
4580 	default:
4581 		return false;
4582 	}
4583 }
4584 
4585 /* Phase 5 device */
4586 static bool bnxt_p5_device(uint16_t device_id)
4587 {
4588 	switch (device_id) {
4589 	case BROADCOM_DEV_ID_57508:
4590 	case BROADCOM_DEV_ID_57504:
4591 	case BROADCOM_DEV_ID_57502:
4592 	case BROADCOM_DEV_ID_57508_MF1:
4593 	case BROADCOM_DEV_ID_57504_MF1:
4594 	case BROADCOM_DEV_ID_57502_MF1:
4595 	case BROADCOM_DEV_ID_57508_MF2:
4596 	case BROADCOM_DEV_ID_57504_MF2:
4597 	case BROADCOM_DEV_ID_57502_MF2:
4598 	case BROADCOM_DEV_ID_57500_VF1:
4599 	case BROADCOM_DEV_ID_57500_VF2:
4600 	case BROADCOM_DEV_ID_58812:
4601 	case BROADCOM_DEV_ID_58814:
4602 	case BROADCOM_DEV_ID_58818:
4603 	case BROADCOM_DEV_ID_58818_VF:
4604 		/* FALLTHROUGH */
4605 		return true;
4606 	default:
4607 		return false;
4608 	}
4609 }
4610 
4611 bool bnxt_stratus_device(struct bnxt *bp)
4612 {
4613 	uint16_t device_id = bp->pdev->id.device_id;
4614 
4615 	switch (device_id) {
4616 	case BROADCOM_DEV_ID_STRATUS_NIC:
4617 	case BROADCOM_DEV_ID_STRATUS_NIC_VF1:
4618 	case BROADCOM_DEV_ID_STRATUS_NIC_VF2:
4619 		/* FALLTHROUGH */
4620 		return true;
4621 	default:
4622 		return false;
4623 	}
4624 }
4625 
4626 static int bnxt_map_pci_bars(struct rte_eth_dev *eth_dev)
4627 {
4628 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
4629 	struct bnxt *bp = eth_dev->data->dev_private;
4630 
4631 	/* enable device (incl. PCI PM wakeup), and bus-mastering */
4632 	bp->bar0 = (void *)pci_dev->mem_resource[0].addr;
4633 	bp->doorbell_base = (void *)pci_dev->mem_resource[2].addr;
4634 	if (!bp->bar0 || !bp->doorbell_base) {
4635 		PMD_DRV_LOG(ERR, "Unable to access Hardware\n");
4636 		return -ENODEV;
4637 	}
4638 
4639 	bp->eth_dev = eth_dev;
4640 	bp->pdev = pci_dev;
4641 
4642 	return 0;
4643 }
4644 
4645 static int bnxt_alloc_ctx_mem_blk(struct bnxt *bp,
4646 				  struct bnxt_ctx_pg_info *ctx_pg,
4647 				  uint32_t mem_size,
4648 				  const char *suffix,
4649 				  uint16_t idx)
4650 {
4651 	struct bnxt_ring_mem_info *rmem = &ctx_pg->ring_mem;
4652 	const struct rte_memzone *mz = NULL;
4653 	char mz_name[RTE_MEMZONE_NAMESIZE];
4654 	rte_iova_t mz_phys_addr;
4655 	uint64_t valid_bits = 0;
4656 	uint32_t sz;
4657 	int i;
4658 
4659 	if (!mem_size)
4660 		return 0;
4661 
4662 	rmem->nr_pages = RTE_ALIGN_MUL_CEIL(mem_size, BNXT_PAGE_SIZE) /
4663 			 BNXT_PAGE_SIZE;
4664 	rmem->page_size = BNXT_PAGE_SIZE;
4665 	rmem->pg_arr = ctx_pg->ctx_pg_arr;
4666 	rmem->dma_arr = ctx_pg->ctx_dma_arr;
4667 	rmem->flags = BNXT_RMEM_VALID_PTE_FLAG;
4668 
4669 	valid_bits = PTU_PTE_VALID;
4670 
4671 	if (rmem->nr_pages > 1) {
4672 		snprintf(mz_name, RTE_MEMZONE_NAMESIZE,
4673 			 "bnxt_ctx_pg_tbl%s_%x_%d",
4674 			 suffix, idx, bp->eth_dev->data->port_id);
4675 		mz_name[RTE_MEMZONE_NAMESIZE - 1] = 0;
4676 		mz = rte_memzone_lookup(mz_name);
4677 		if (!mz) {
4678 			mz = rte_memzone_reserve_aligned(mz_name,
4679 						rmem->nr_pages * 8,
4680 						bp->eth_dev->device->numa_node,
4681 						RTE_MEMZONE_2MB |
4682 						RTE_MEMZONE_SIZE_HINT_ONLY |
4683 						RTE_MEMZONE_IOVA_CONTIG,
4684 						BNXT_PAGE_SIZE);
4685 			if (mz == NULL)
4686 				return -ENOMEM;
4687 		}
4688 
4689 		memset(mz->addr, 0, mz->len);
4690 		mz_phys_addr = mz->iova;
4691 
4692 		rmem->pg_tbl = mz->addr;
4693 		rmem->pg_tbl_map = mz_phys_addr;
4694 		rmem->pg_tbl_mz = mz;
4695 	}
4696 
4697 	snprintf(mz_name, RTE_MEMZONE_NAMESIZE, "bnxt_ctx_%s_%x_%d",
4698 		 suffix, idx, bp->eth_dev->data->port_id);
4699 	mz = rte_memzone_lookup(mz_name);
4700 	if (!mz) {
4701 		mz = rte_memzone_reserve_aligned(mz_name,
4702 						 mem_size,
4703 						 bp->eth_dev->device->numa_node,
4704 						 RTE_MEMZONE_1GB |
4705 						 RTE_MEMZONE_SIZE_HINT_ONLY |
4706 						 RTE_MEMZONE_IOVA_CONTIG,
4707 						 BNXT_PAGE_SIZE);
4708 		if (mz == NULL)
4709 			return -ENOMEM;
4710 	}
4711 
4712 	memset(mz->addr, 0, mz->len);
4713 	mz_phys_addr = mz->iova;
4714 
4715 	for (sz = 0, i = 0; sz < mem_size; sz += BNXT_PAGE_SIZE, i++) {
4716 		rmem->pg_arr[i] = ((char *)mz->addr) + sz;
4717 		rmem->dma_arr[i] = mz_phys_addr + sz;
4718 
4719 		if (rmem->nr_pages > 1) {
4720 			if (i == rmem->nr_pages - 2 &&
4721 			    (rmem->flags & BNXT_RMEM_RING_PTE_FLAG))
4722 				valid_bits |= PTU_PTE_NEXT_TO_LAST;
4723 			else if (i == rmem->nr_pages - 1 &&
4724 				 (rmem->flags & BNXT_RMEM_RING_PTE_FLAG))
4725 				valid_bits |= PTU_PTE_LAST;
4726 
4727 			rmem->pg_tbl[i] = rte_cpu_to_le_64(rmem->dma_arr[i] |
4728 							   valid_bits);
4729 		}
4730 	}
4731 
4732 	rmem->mz = mz;
4733 	if (rmem->vmem_size)
4734 		rmem->vmem = (void **)mz->addr;
4735 	rmem->dma_arr[0] = mz_phys_addr;
4736 	return 0;
4737 }
4738 
4739 static void bnxt_free_ctx_mem(struct bnxt *bp)
4740 {
4741 	int i;
4742 
4743 	if (!bp->ctx || !(bp->ctx->flags & BNXT_CTX_FLAG_INITED))
4744 		return;
4745 
4746 	bp->ctx->flags &= ~BNXT_CTX_FLAG_INITED;
4747 	rte_memzone_free(bp->ctx->qp_mem.ring_mem.mz);
4748 	rte_memzone_free(bp->ctx->srq_mem.ring_mem.mz);
4749 	rte_memzone_free(bp->ctx->cq_mem.ring_mem.mz);
4750 	rte_memzone_free(bp->ctx->vnic_mem.ring_mem.mz);
4751 	rte_memzone_free(bp->ctx->stat_mem.ring_mem.mz);
4752 	rte_memzone_free(bp->ctx->qp_mem.ring_mem.pg_tbl_mz);
4753 	rte_memzone_free(bp->ctx->srq_mem.ring_mem.pg_tbl_mz);
4754 	rte_memzone_free(bp->ctx->cq_mem.ring_mem.pg_tbl_mz);
4755 	rte_memzone_free(bp->ctx->vnic_mem.ring_mem.pg_tbl_mz);
4756 	rte_memzone_free(bp->ctx->stat_mem.ring_mem.pg_tbl_mz);
4757 
4758 	for (i = 0; i < bp->ctx->tqm_fp_rings_count + 1; i++) {
4759 		if (bp->ctx->tqm_mem[i])
4760 			rte_memzone_free(bp->ctx->tqm_mem[i]->ring_mem.mz);
4761 	}
4762 
4763 	rte_free(bp->ctx);
4764 	bp->ctx = NULL;
4765 }
4766 
4767 #define bnxt_roundup(x, y)   ((((x) + ((y) - 1)) / (y)) * (y))
4768 
4769 #define min_t(type, x, y) ({                    \
4770 	type __min1 = (x);                      \
4771 	type __min2 = (y);                      \
4772 	__min1 < __min2 ? __min1 : __min2; })
4773 
4774 #define max_t(type, x, y) ({                    \
4775 	type __max1 = (x);                      \
4776 	type __max2 = (y);                      \
4777 	__max1 > __max2 ? __max1 : __max2; })
4778 
4779 #define clamp_t(type, _x, min, max)     min_t(type, max_t(type, _x, min), max)
4780 
4781 int bnxt_alloc_ctx_mem(struct bnxt *bp)
4782 {
4783 	struct bnxt_ctx_pg_info *ctx_pg;
4784 	struct bnxt_ctx_mem_info *ctx;
4785 	uint32_t mem_size, ena, entries;
4786 	uint32_t entries_sp, min;
4787 	int i, rc;
4788 
4789 	rc = bnxt_hwrm_func_backing_store_qcaps(bp);
4790 	if (rc) {
4791 		PMD_DRV_LOG(ERR, "Query context mem capability failed\n");
4792 		return rc;
4793 	}
4794 	ctx = bp->ctx;
4795 	if (!ctx || (ctx->flags & BNXT_CTX_FLAG_INITED))
4796 		return 0;
4797 
4798 	ctx_pg = &ctx->qp_mem;
4799 	ctx_pg->entries = ctx->qp_min_qp1_entries + ctx->qp_max_l2_entries;
4800 	if (ctx->qp_entry_size) {
4801 		mem_size = ctx->qp_entry_size * ctx_pg->entries;
4802 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "qp_mem", 0);
4803 		if (rc)
4804 			return rc;
4805 	}
4806 
4807 	ctx_pg = &ctx->srq_mem;
4808 	ctx_pg->entries = ctx->srq_max_l2_entries;
4809 	if (ctx->srq_entry_size) {
4810 		mem_size = ctx->srq_entry_size * ctx_pg->entries;
4811 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "srq_mem", 0);
4812 		if (rc)
4813 			return rc;
4814 	}
4815 
4816 	ctx_pg = &ctx->cq_mem;
4817 	ctx_pg->entries = ctx->cq_max_l2_entries;
4818 	if (ctx->cq_entry_size) {
4819 		mem_size = ctx->cq_entry_size * ctx_pg->entries;
4820 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "cq_mem", 0);
4821 		if (rc)
4822 			return rc;
4823 	}
4824 
4825 	ctx_pg = &ctx->vnic_mem;
4826 	ctx_pg->entries = ctx->vnic_max_vnic_entries +
4827 		ctx->vnic_max_ring_table_entries;
4828 	if (ctx->vnic_entry_size) {
4829 		mem_size = ctx->vnic_entry_size * ctx_pg->entries;
4830 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "vnic_mem", 0);
4831 		if (rc)
4832 			return rc;
4833 	}
4834 
4835 	ctx_pg = &ctx->stat_mem;
4836 	ctx_pg->entries = ctx->stat_max_entries;
4837 	if (ctx->stat_entry_size) {
4838 		mem_size = ctx->stat_entry_size * ctx_pg->entries;
4839 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "stat_mem", 0);
4840 		if (rc)
4841 			return rc;
4842 	}
4843 
4844 	min = ctx->tqm_min_entries_per_ring;
4845 
4846 	entries_sp = ctx->qp_max_l2_entries +
4847 		     ctx->vnic_max_vnic_entries +
4848 		     2 * ctx->qp_min_qp1_entries + min;
4849 	entries_sp = bnxt_roundup(entries_sp, ctx->tqm_entries_multiple);
4850 
4851 	entries = ctx->qp_max_l2_entries + ctx->qp_min_qp1_entries;
4852 	entries = bnxt_roundup(entries, ctx->tqm_entries_multiple);
4853 	entries = clamp_t(uint32_t, entries, min,
4854 			  ctx->tqm_max_entries_per_ring);
4855 	for (i = 0, ena = 0; i < ctx->tqm_fp_rings_count + 1; i++) {
4856 		/* i=0 is for TQM_SP. i=1 to i=8 applies to RING0 to RING7.
4857 		 * i > 8 is other ext rings.
4858 		 */
4859 		ctx_pg = ctx->tqm_mem[i];
4860 		ctx_pg->entries = i ? entries : entries_sp;
4861 		if (ctx->tqm_entry_size) {
4862 			mem_size = ctx->tqm_entry_size * ctx_pg->entries;
4863 			rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size,
4864 						    "tqm_mem", i);
4865 			if (rc)
4866 				return rc;
4867 		}
4868 		if (i < BNXT_MAX_TQM_LEGACY_RINGS)
4869 			ena |= HWRM_FUNC_BACKING_STORE_CFG_INPUT_ENABLES_TQM_SP << i;
4870 		else
4871 			ena |= HWRM_FUNC_BACKING_STORE_CFG_INPUT_ENABLES_TQM_RING8;
4872 	}
4873 
4874 	ena |= FUNC_BACKING_STORE_CFG_INPUT_DFLT_ENABLES;
4875 	rc = bnxt_hwrm_func_backing_store_cfg(bp, ena);
4876 	if (rc)
4877 		PMD_DRV_LOG(ERR,
4878 			    "Failed to configure context mem: rc = %d\n", rc);
4879 	else
4880 		ctx->flags |= BNXT_CTX_FLAG_INITED;
4881 
4882 	return rc;
4883 }
4884 
4885 static int bnxt_alloc_stats_mem(struct bnxt *bp)
4886 {
4887 	struct rte_pci_device *pci_dev = bp->pdev;
4888 	char mz_name[RTE_MEMZONE_NAMESIZE];
4889 	const struct rte_memzone *mz = NULL;
4890 	uint32_t total_alloc_len;
4891 	rte_iova_t mz_phys_addr;
4892 
4893 	if (pci_dev->id.device_id == BROADCOM_DEV_ID_NS2)
4894 		return 0;
4895 
4896 	snprintf(mz_name, RTE_MEMZONE_NAMESIZE,
4897 		 "bnxt_" PCI_PRI_FMT "-%s", pci_dev->addr.domain,
4898 		 pci_dev->addr.bus, pci_dev->addr.devid,
4899 		 pci_dev->addr.function, "rx_port_stats");
4900 	mz_name[RTE_MEMZONE_NAMESIZE - 1] = 0;
4901 	mz = rte_memzone_lookup(mz_name);
4902 	total_alloc_len =
4903 		RTE_CACHE_LINE_ROUNDUP(sizeof(struct rx_port_stats) +
4904 				       sizeof(struct rx_port_stats_ext) + 512);
4905 	if (!mz) {
4906 		mz = rte_memzone_reserve(mz_name, total_alloc_len,
4907 					 SOCKET_ID_ANY,
4908 					 RTE_MEMZONE_2MB |
4909 					 RTE_MEMZONE_SIZE_HINT_ONLY |
4910 					 RTE_MEMZONE_IOVA_CONTIG);
4911 		if (mz == NULL)
4912 			return -ENOMEM;
4913 	}
4914 	memset(mz->addr, 0, mz->len);
4915 	mz_phys_addr = mz->iova;
4916 
4917 	bp->rx_mem_zone = (const void *)mz;
4918 	bp->hw_rx_port_stats = mz->addr;
4919 	bp->hw_rx_port_stats_map = mz_phys_addr;
4920 
4921 	snprintf(mz_name, RTE_MEMZONE_NAMESIZE,
4922 		 "bnxt_" PCI_PRI_FMT "-%s", pci_dev->addr.domain,
4923 		 pci_dev->addr.bus, pci_dev->addr.devid,
4924 		 pci_dev->addr.function, "tx_port_stats");
4925 	mz_name[RTE_MEMZONE_NAMESIZE - 1] = 0;
4926 	mz = rte_memzone_lookup(mz_name);
4927 	total_alloc_len =
4928 		RTE_CACHE_LINE_ROUNDUP(sizeof(struct tx_port_stats) +
4929 				       sizeof(struct tx_port_stats_ext) + 512);
4930 	if (!mz) {
4931 		mz = rte_memzone_reserve(mz_name,
4932 					 total_alloc_len,
4933 					 SOCKET_ID_ANY,
4934 					 RTE_MEMZONE_2MB |
4935 					 RTE_MEMZONE_SIZE_HINT_ONLY |
4936 					 RTE_MEMZONE_IOVA_CONTIG);
4937 		if (mz == NULL)
4938 			return -ENOMEM;
4939 	}
4940 	memset(mz->addr, 0, mz->len);
4941 	mz_phys_addr = mz->iova;
4942 
4943 	bp->tx_mem_zone = (const void *)mz;
4944 	bp->hw_tx_port_stats = mz->addr;
4945 	bp->hw_tx_port_stats_map = mz_phys_addr;
4946 	bp->flags |= BNXT_FLAG_PORT_STATS;
4947 
4948 	/* Display extended statistics if FW supports it */
4949 	if (bp->hwrm_spec_code < HWRM_SPEC_CODE_1_8_4 ||
4950 	    bp->hwrm_spec_code == HWRM_SPEC_CODE_1_9_0 ||
4951 	    !(bp->flags & BNXT_FLAG_EXT_STATS_SUPPORTED))
4952 		return 0;
4953 
4954 	bp->hw_rx_port_stats_ext = (void *)
4955 		((uint8_t *)bp->hw_rx_port_stats +
4956 		 sizeof(struct rx_port_stats));
4957 	bp->hw_rx_port_stats_ext_map = bp->hw_rx_port_stats_map +
4958 		sizeof(struct rx_port_stats);
4959 	bp->flags |= BNXT_FLAG_EXT_RX_PORT_STATS;
4960 
4961 	if (bp->hwrm_spec_code < HWRM_SPEC_CODE_1_9_2 ||
4962 	    bp->flags & BNXT_FLAG_EXT_STATS_SUPPORTED) {
4963 		bp->hw_tx_port_stats_ext = (void *)
4964 			((uint8_t *)bp->hw_tx_port_stats +
4965 			 sizeof(struct tx_port_stats));
4966 		bp->hw_tx_port_stats_ext_map =
4967 			bp->hw_tx_port_stats_map +
4968 			sizeof(struct tx_port_stats);
4969 		bp->flags |= BNXT_FLAG_EXT_TX_PORT_STATS;
4970 	}
4971 
4972 	return 0;
4973 }
4974 
4975 static int bnxt_setup_mac_addr(struct rte_eth_dev *eth_dev)
4976 {
4977 	struct bnxt *bp = eth_dev->data->dev_private;
4978 	int rc = 0;
4979 
4980 	eth_dev->data->mac_addrs = rte_zmalloc("bnxt_mac_addr_tbl",
4981 					       RTE_ETHER_ADDR_LEN *
4982 					       bp->max_l2_ctx,
4983 					       0);
4984 	if (eth_dev->data->mac_addrs == NULL) {
4985 		PMD_DRV_LOG(ERR, "Failed to alloc MAC addr tbl\n");
4986 		return -ENOMEM;
4987 	}
4988 
4989 	if (!BNXT_HAS_DFLT_MAC_SET(bp)) {
4990 		if (BNXT_PF(bp))
4991 			return -EINVAL;
4992 
4993 		/* Generate a random MAC address, if none was assigned by PF */
4994 		PMD_DRV_LOG(INFO, "VF MAC address not assigned by Host PF\n");
4995 		bnxt_eth_hw_addr_random(bp->mac_addr);
4996 		PMD_DRV_LOG(INFO,
4997 			    "Assign random MAC:%02X:%02X:%02X:%02X:%02X:%02X\n",
4998 			    bp->mac_addr[0], bp->mac_addr[1], bp->mac_addr[2],
4999 			    bp->mac_addr[3], bp->mac_addr[4], bp->mac_addr[5]);
5000 
5001 		rc = bnxt_hwrm_set_mac(bp);
5002 		if (rc)
5003 			return rc;
5004 	}
5005 
5006 	/* Copy the permanent MAC from the FUNC_QCAPS response */
5007 	memcpy(&eth_dev->data->mac_addrs[0], bp->mac_addr, RTE_ETHER_ADDR_LEN);
5008 
5009 	return rc;
5010 }
5011 
5012 static int bnxt_restore_dflt_mac(struct bnxt *bp)
5013 {
5014 	int rc = 0;
5015 
5016 	/* MAC is already configured in FW */
5017 	if (BNXT_HAS_DFLT_MAC_SET(bp))
5018 		return 0;
5019 
5020 	/* Restore the old MAC configured */
5021 	rc = bnxt_hwrm_set_mac(bp);
5022 	if (rc)
5023 		PMD_DRV_LOG(ERR, "Failed to restore MAC address\n");
5024 
5025 	return rc;
5026 }
5027 
5028 static void bnxt_config_vf_req_fwd(struct bnxt *bp)
5029 {
5030 	if (!BNXT_PF(bp))
5031 		return;
5032 
5033 	memset(bp->pf->vf_req_fwd, 0, sizeof(bp->pf->vf_req_fwd));
5034 
5035 	if (!(bp->fw_cap & BNXT_FW_CAP_LINK_ADMIN))
5036 		BNXT_HWRM_CMD_TO_FORWARD(HWRM_PORT_PHY_QCFG);
5037 	BNXT_HWRM_CMD_TO_FORWARD(HWRM_FUNC_CFG);
5038 	BNXT_HWRM_CMD_TO_FORWARD(HWRM_FUNC_VF_CFG);
5039 	BNXT_HWRM_CMD_TO_FORWARD(HWRM_CFA_L2_FILTER_ALLOC);
5040 	BNXT_HWRM_CMD_TO_FORWARD(HWRM_OEM_CMD);
5041 }
5042 
5043 struct bnxt *
5044 bnxt_get_bp(uint16_t port)
5045 {
5046 	struct bnxt *bp;
5047 	struct rte_eth_dev *dev;
5048 
5049 	if (!rte_eth_dev_is_valid_port(port)) {
5050 		PMD_DRV_LOG(ERR, "Invalid port %d\n", port);
5051 		return NULL;
5052 	}
5053 
5054 	dev = &rte_eth_devices[port];
5055 	if (!is_bnxt_supported(dev)) {
5056 		PMD_DRV_LOG(ERR, "Device %d not supported\n", port);
5057 		return NULL;
5058 	}
5059 
5060 	bp = (struct bnxt *)dev->data->dev_private;
5061 	if (!BNXT_TRUFLOW_EN(bp)) {
5062 		PMD_DRV_LOG(ERR, "TRUFLOW not enabled\n");
5063 		return NULL;
5064 	}
5065 
5066 	return bp;
5067 }
5068 
5069 uint16_t
5070 bnxt_get_svif(uint16_t port_id, bool func_svif,
5071 	      enum bnxt_ulp_intf_type type)
5072 {
5073 	struct rte_eth_dev *eth_dev;
5074 	struct bnxt *bp;
5075 
5076 	eth_dev = &rte_eth_devices[port_id];
5077 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
5078 		struct bnxt_representor *vfr = eth_dev->data->dev_private;
5079 		if (!vfr)
5080 			return 0;
5081 
5082 		if (type == BNXT_ULP_INTF_TYPE_VF_REP)
5083 			return vfr->svif;
5084 
5085 		eth_dev = vfr->parent_dev;
5086 	}
5087 
5088 	bp = eth_dev->data->dev_private;
5089 
5090 	return func_svif ? bp->func_svif : bp->port_svif;
5091 }
5092 
5093 void
5094 bnxt_get_iface_mac(uint16_t port, enum bnxt_ulp_intf_type type,
5095 		   uint8_t *mac, uint8_t *parent_mac)
5096 {
5097 	struct rte_eth_dev *eth_dev;
5098 	struct bnxt *bp;
5099 
5100 	if (type != BNXT_ULP_INTF_TYPE_TRUSTED_VF &&
5101 	    type != BNXT_ULP_INTF_TYPE_PF)
5102 		return;
5103 
5104 	eth_dev = &rte_eth_devices[port];
5105 	bp = eth_dev->data->dev_private;
5106 	memcpy(mac, bp->mac_addr, RTE_ETHER_ADDR_LEN);
5107 
5108 	if (type == BNXT_ULP_INTF_TYPE_TRUSTED_VF)
5109 		memcpy(parent_mac, bp->parent->mac_addr, RTE_ETHER_ADDR_LEN);
5110 }
5111 
5112 uint16_t
5113 bnxt_get_parent_vnic_id(uint16_t port, enum bnxt_ulp_intf_type type)
5114 {
5115 	struct rte_eth_dev *eth_dev;
5116 	struct bnxt *bp;
5117 
5118 	if (type != BNXT_ULP_INTF_TYPE_TRUSTED_VF)
5119 		return 0;
5120 
5121 	eth_dev = &rte_eth_devices[port];
5122 	bp = eth_dev->data->dev_private;
5123 
5124 	return bp->parent->vnic;
5125 }
5126 uint16_t
5127 bnxt_get_vnic_id(uint16_t port, enum bnxt_ulp_intf_type type)
5128 {
5129 	struct rte_eth_dev *eth_dev;
5130 	struct bnxt_vnic_info *vnic;
5131 	struct bnxt *bp;
5132 
5133 	eth_dev = &rte_eth_devices[port];
5134 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
5135 		struct bnxt_representor *vfr = eth_dev->data->dev_private;
5136 		if (!vfr)
5137 			return 0;
5138 
5139 		if (type == BNXT_ULP_INTF_TYPE_VF_REP)
5140 			return vfr->dflt_vnic_id;
5141 
5142 		eth_dev = vfr->parent_dev;
5143 	}
5144 
5145 	bp = eth_dev->data->dev_private;
5146 
5147 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
5148 
5149 	return vnic->fw_vnic_id;
5150 }
5151 
5152 uint16_t
5153 bnxt_get_fw_func_id(uint16_t port, enum bnxt_ulp_intf_type type)
5154 {
5155 	struct rte_eth_dev *eth_dev;
5156 	struct bnxt *bp;
5157 
5158 	eth_dev = &rte_eth_devices[port];
5159 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
5160 		struct bnxt_representor *vfr = eth_dev->data->dev_private;
5161 		if (!vfr)
5162 			return 0;
5163 
5164 		if (type == BNXT_ULP_INTF_TYPE_VF_REP)
5165 			return vfr->fw_fid;
5166 
5167 		eth_dev = vfr->parent_dev;
5168 	}
5169 
5170 	bp = eth_dev->data->dev_private;
5171 
5172 	return bp->fw_fid;
5173 }
5174 
5175 enum bnxt_ulp_intf_type
5176 bnxt_get_interface_type(uint16_t port)
5177 {
5178 	struct rte_eth_dev *eth_dev;
5179 	struct bnxt *bp;
5180 
5181 	eth_dev = &rte_eth_devices[port];
5182 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev))
5183 		return BNXT_ULP_INTF_TYPE_VF_REP;
5184 
5185 	bp = eth_dev->data->dev_private;
5186 	if (BNXT_PF(bp))
5187 		return BNXT_ULP_INTF_TYPE_PF;
5188 	else if (BNXT_VF_IS_TRUSTED(bp))
5189 		return BNXT_ULP_INTF_TYPE_TRUSTED_VF;
5190 	else if (BNXT_VF(bp))
5191 		return BNXT_ULP_INTF_TYPE_VF;
5192 
5193 	return BNXT_ULP_INTF_TYPE_INVALID;
5194 }
5195 
5196 uint16_t
5197 bnxt_get_phy_port_id(uint16_t port_id)
5198 {
5199 	struct bnxt_representor *vfr;
5200 	struct rte_eth_dev *eth_dev;
5201 	struct bnxt *bp;
5202 
5203 	eth_dev = &rte_eth_devices[port_id];
5204 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
5205 		vfr = eth_dev->data->dev_private;
5206 		if (!vfr)
5207 			return 0;
5208 
5209 		eth_dev = vfr->parent_dev;
5210 	}
5211 
5212 	bp = eth_dev->data->dev_private;
5213 
5214 	return BNXT_PF(bp) ? bp->pf->port_id : bp->parent->port_id;
5215 }
5216 
5217 uint16_t
5218 bnxt_get_parif(uint16_t port_id, enum bnxt_ulp_intf_type type)
5219 {
5220 	struct rte_eth_dev *eth_dev;
5221 	struct bnxt *bp;
5222 
5223 	eth_dev = &rte_eth_devices[port_id];
5224 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
5225 		struct bnxt_representor *vfr = eth_dev->data->dev_private;
5226 		if (!vfr)
5227 			return 0;
5228 
5229 		if (type == BNXT_ULP_INTF_TYPE_VF_REP)
5230 			return vfr->fw_fid - 1;
5231 
5232 		eth_dev = vfr->parent_dev;
5233 	}
5234 
5235 	bp = eth_dev->data->dev_private;
5236 
5237 	return BNXT_PF(bp) ? bp->fw_fid - 1 : bp->parent->fid - 1;
5238 }
5239 
5240 uint16_t
5241 bnxt_get_vport(uint16_t port_id)
5242 {
5243 	return (1 << bnxt_get_phy_port_id(port_id));
5244 }
5245 
5246 static void bnxt_alloc_error_recovery_info(struct bnxt *bp)
5247 {
5248 	struct bnxt_error_recovery_info *info = bp->recovery_info;
5249 
5250 	if (info) {
5251 		if (!(bp->fw_cap & BNXT_FW_CAP_HCOMM_FW_STATUS))
5252 			memset(info, 0, sizeof(*info));
5253 		return;
5254 	}
5255 
5256 	if (!(bp->fw_cap & BNXT_FW_CAP_ERROR_RECOVERY))
5257 		return;
5258 
5259 	info = rte_zmalloc("bnxt_hwrm_error_recovery_qcfg",
5260 			   sizeof(*info), 0);
5261 	if (!info)
5262 		bp->fw_cap &= ~BNXT_FW_CAP_ERROR_RECOVERY;
5263 
5264 	bp->recovery_info = info;
5265 }
5266 
5267 static void bnxt_check_fw_status(struct bnxt *bp)
5268 {
5269 	uint32_t fw_status;
5270 
5271 	if (!(bp->recovery_info &&
5272 	      (bp->fw_cap & BNXT_FW_CAP_HCOMM_FW_STATUS)))
5273 		return;
5274 
5275 	fw_status = bnxt_read_fw_status_reg(bp, BNXT_FW_STATUS_REG);
5276 	if (fw_status != BNXT_FW_STATUS_HEALTHY)
5277 		PMD_DRV_LOG(ERR, "Firmware not responding, status: %#x\n",
5278 			    fw_status);
5279 }
5280 
5281 static int bnxt_map_hcomm_fw_status_reg(struct bnxt *bp)
5282 {
5283 	struct bnxt_error_recovery_info *info = bp->recovery_info;
5284 	uint32_t status_loc;
5285 	uint32_t sig_ver;
5286 
5287 	rte_write32(HCOMM_STATUS_STRUCT_LOC, (uint8_t *)bp->bar0 +
5288 		    BNXT_GRCPF_REG_WINDOW_BASE_OUT + 4);
5289 	sig_ver = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
5290 				   BNXT_GRCP_WINDOW_2_BASE +
5291 				   offsetof(struct hcomm_status,
5292 					    sig_ver)));
5293 	/* If the signature is absent, then FW does not support this feature */
5294 	if ((sig_ver & HCOMM_STATUS_SIGNATURE_MASK) !=
5295 	    HCOMM_STATUS_SIGNATURE_VAL)
5296 		return 0;
5297 
5298 	if (!info) {
5299 		info = rte_zmalloc("bnxt_hwrm_error_recovery_qcfg",
5300 				   sizeof(*info), 0);
5301 		if (!info)
5302 			return -ENOMEM;
5303 		bp->recovery_info = info;
5304 	} else {
5305 		memset(info, 0, sizeof(*info));
5306 	}
5307 
5308 	status_loc = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
5309 				      BNXT_GRCP_WINDOW_2_BASE +
5310 				      offsetof(struct hcomm_status,
5311 					       fw_status_loc)));
5312 
5313 	/* Only pre-map the FW health status GRC register */
5314 	if (BNXT_FW_STATUS_REG_TYPE(status_loc) != BNXT_FW_STATUS_REG_TYPE_GRC)
5315 		return 0;
5316 
5317 	info->status_regs[BNXT_FW_STATUS_REG] = status_loc;
5318 	info->mapped_status_regs[BNXT_FW_STATUS_REG] =
5319 		BNXT_GRCP_WINDOW_2_BASE + (status_loc & BNXT_GRCP_OFFSET_MASK);
5320 
5321 	rte_write32((status_loc & BNXT_GRCP_BASE_MASK), (uint8_t *)bp->bar0 +
5322 		    BNXT_GRCPF_REG_WINDOW_BASE_OUT + 4);
5323 
5324 	bp->fw_cap |= BNXT_FW_CAP_HCOMM_FW_STATUS;
5325 
5326 	return 0;
5327 }
5328 
5329 /* This function gets the FW version along with the
5330  * capabilities(MAX and current) of the function, vnic,
5331  * error recovery, phy and other chip related info
5332  */
5333 static int bnxt_get_config(struct bnxt *bp)
5334 {
5335 	uint16_t mtu;
5336 	int rc = 0;
5337 
5338 	bp->fw_cap = 0;
5339 
5340 	rc = bnxt_map_hcomm_fw_status_reg(bp);
5341 	if (rc)
5342 		return rc;
5343 
5344 	rc = bnxt_hwrm_ver_get(bp, DFLT_HWRM_CMD_TIMEOUT);
5345 	if (rc) {
5346 		bnxt_check_fw_status(bp);
5347 		return rc;
5348 	}
5349 
5350 	rc = bnxt_hwrm_func_reset(bp);
5351 	if (rc)
5352 		return -EIO;
5353 
5354 	rc = bnxt_hwrm_vnic_qcaps(bp);
5355 	if (rc)
5356 		return rc;
5357 
5358 	rc = bnxt_hwrm_queue_qportcfg(bp);
5359 	if (rc)
5360 		return rc;
5361 
5362 	/* Get the MAX capabilities for this function.
5363 	 * This function also allocates context memory for TQM rings and
5364 	 * informs the firmware about this allocated backing store memory.
5365 	 */
5366 	rc = bnxt_hwrm_func_qcaps(bp);
5367 	if (rc)
5368 		return rc;
5369 
5370 	rc = bnxt_hwrm_func_qcfg(bp, &mtu);
5371 	if (rc)
5372 		return rc;
5373 
5374 	rc = bnxt_hwrm_cfa_adv_flow_mgmt_qcaps(bp);
5375 	if (rc)
5376 		return rc;
5377 
5378 	bnxt_hwrm_port_mac_qcfg(bp);
5379 
5380 	bnxt_hwrm_parent_pf_qcfg(bp);
5381 
5382 	bnxt_hwrm_port_phy_qcaps(bp);
5383 
5384 	bnxt_alloc_error_recovery_info(bp);
5385 	/* Get the adapter error recovery support info */
5386 	rc = bnxt_hwrm_error_recovery_qcfg(bp);
5387 	if (rc)
5388 		bp->fw_cap &= ~BNXT_FW_CAP_ERROR_RECOVERY;
5389 
5390 	bnxt_hwrm_port_led_qcaps(bp);
5391 
5392 	return 0;
5393 }
5394 
5395 static int
5396 bnxt_init_locks(struct bnxt *bp)
5397 {
5398 	int err;
5399 
5400 	err = pthread_mutex_init(&bp->flow_lock, NULL);
5401 	if (err) {
5402 		PMD_DRV_LOG(ERR, "Unable to initialize flow_lock\n");
5403 		return err;
5404 	}
5405 
5406 	err = pthread_mutex_init(&bp->def_cp_lock, NULL);
5407 	if (err) {
5408 		PMD_DRV_LOG(ERR, "Unable to initialize def_cp_lock\n");
5409 		return err;
5410 	}
5411 
5412 	err = pthread_mutex_init(&bp->health_check_lock, NULL);
5413 	if (err) {
5414 		PMD_DRV_LOG(ERR, "Unable to initialize health_check_lock\n");
5415 		return err;
5416 	}
5417 
5418 	err = pthread_mutex_init(&bp->err_recovery_lock, NULL);
5419 	if (err)
5420 		PMD_DRV_LOG(ERR, "Unable to initialize err_recovery_lock\n");
5421 
5422 	return err;
5423 }
5424 
5425 static int bnxt_init_resources(struct bnxt *bp, bool reconfig_dev)
5426 {
5427 	int rc = 0;
5428 
5429 	rc = bnxt_get_config(bp);
5430 	if (rc)
5431 		return rc;
5432 
5433 	if (!reconfig_dev) {
5434 		rc = bnxt_setup_mac_addr(bp->eth_dev);
5435 		if (rc)
5436 			return rc;
5437 	} else {
5438 		rc = bnxt_restore_dflt_mac(bp);
5439 		if (rc)
5440 			return rc;
5441 	}
5442 
5443 	bnxt_config_vf_req_fwd(bp);
5444 
5445 	rc = bnxt_hwrm_func_driver_register(bp);
5446 	if (rc) {
5447 		PMD_DRV_LOG(ERR, "Failed to register driver");
5448 		return -EBUSY;
5449 	}
5450 
5451 	if (BNXT_PF(bp)) {
5452 		if (bp->pdev->max_vfs) {
5453 			rc = bnxt_hwrm_allocate_vfs(bp, bp->pdev->max_vfs);
5454 			if (rc) {
5455 				PMD_DRV_LOG(ERR, "Failed to allocate VFs\n");
5456 				return rc;
5457 			}
5458 		} else {
5459 			rc = bnxt_hwrm_allocate_pf_only(bp);
5460 			if (rc) {
5461 				PMD_DRV_LOG(ERR,
5462 					    "Failed to allocate PF resources");
5463 				return rc;
5464 			}
5465 		}
5466 	}
5467 
5468 	rc = bnxt_alloc_mem(bp, reconfig_dev);
5469 	if (rc)
5470 		return rc;
5471 
5472 	rc = bnxt_setup_int(bp);
5473 	if (rc)
5474 		return rc;
5475 
5476 	rc = bnxt_request_int(bp);
5477 	if (rc)
5478 		return rc;
5479 
5480 	rc = bnxt_init_ctx_mem(bp);
5481 	if (rc) {
5482 		PMD_DRV_LOG(ERR, "Failed to init adv_flow_counters\n");
5483 		return rc;
5484 	}
5485 
5486 	return 0;
5487 }
5488 
5489 static int
5490 bnxt_parse_devarg_accum_stats(__rte_unused const char *key,
5491 			      const char *value, void *opaque_arg)
5492 {
5493 	struct bnxt *bp = opaque_arg;
5494 	unsigned long accum_stats;
5495 	char *end = NULL;
5496 
5497 	if (!value || !opaque_arg) {
5498 		PMD_DRV_LOG(ERR,
5499 			    "Invalid parameter passed to accum-stats devargs.\n");
5500 		return -EINVAL;
5501 	}
5502 
5503 	accum_stats = strtoul(value, &end, 10);
5504 	if (end == NULL || *end != '\0' ||
5505 	    (accum_stats == ULONG_MAX && errno == ERANGE)) {
5506 		PMD_DRV_LOG(ERR,
5507 			    "Invalid parameter passed to accum-stats devargs.\n");
5508 		return -EINVAL;
5509 	}
5510 
5511 	if (BNXT_DEVARG_ACCUM_STATS_INVALID(accum_stats)) {
5512 		PMD_DRV_LOG(ERR,
5513 			    "Invalid value passed to accum-stats devargs.\n");
5514 		return -EINVAL;
5515 	}
5516 
5517 	if (accum_stats) {
5518 		bp->flags2 |= BNXT_FLAGS2_ACCUM_STATS_EN;
5519 		PMD_DRV_LOG(INFO, "Host-based accum-stats feature enabled.\n");
5520 	} else {
5521 		bp->flags2 &= ~BNXT_FLAGS2_ACCUM_STATS_EN;
5522 		PMD_DRV_LOG(INFO, "Host-based accum-stats feature disabled.\n");
5523 	}
5524 
5525 	return 0;
5526 }
5527 
5528 static int
5529 bnxt_parse_devarg_flow_xstat(__rte_unused const char *key,
5530 			     const char *value, void *opaque_arg)
5531 {
5532 	struct bnxt *bp = opaque_arg;
5533 	unsigned long flow_xstat;
5534 	char *end = NULL;
5535 
5536 	if (!value || !opaque_arg) {
5537 		PMD_DRV_LOG(ERR,
5538 			    "Invalid parameter passed to flow_xstat devarg.\n");
5539 		return -EINVAL;
5540 	}
5541 
5542 	flow_xstat = strtoul(value, &end, 10);
5543 	if (end == NULL || *end != '\0' ||
5544 	    (flow_xstat == ULONG_MAX && errno == ERANGE)) {
5545 		PMD_DRV_LOG(ERR,
5546 			    "Invalid parameter passed to flow_xstat devarg.\n");
5547 		return -EINVAL;
5548 	}
5549 
5550 	if (BNXT_DEVARG_FLOW_XSTAT_INVALID(flow_xstat)) {
5551 		PMD_DRV_LOG(ERR,
5552 			    "Invalid value passed to flow_xstat devarg.\n");
5553 		return -EINVAL;
5554 	}
5555 
5556 	bp->flags |= BNXT_FLAG_FLOW_XSTATS_EN;
5557 	if (BNXT_FLOW_XSTATS_EN(bp))
5558 		PMD_DRV_LOG(INFO, "flow_xstat feature enabled.\n");
5559 
5560 	return 0;
5561 }
5562 
5563 static int
5564 bnxt_parse_devarg_max_num_kflows(__rte_unused const char *key,
5565 					const char *value, void *opaque_arg)
5566 {
5567 	struct bnxt *bp = opaque_arg;
5568 	unsigned long max_num_kflows;
5569 	char *end = NULL;
5570 
5571 	if (!value || !opaque_arg) {
5572 		PMD_DRV_LOG(ERR,
5573 			"Invalid parameter passed to max_num_kflows devarg.\n");
5574 		return -EINVAL;
5575 	}
5576 
5577 	max_num_kflows = strtoul(value, &end, 10);
5578 	if (end == NULL || *end != '\0' ||
5579 		(max_num_kflows == ULONG_MAX && errno == ERANGE)) {
5580 		PMD_DRV_LOG(ERR,
5581 			"Invalid parameter passed to max_num_kflows devarg.\n");
5582 		return -EINVAL;
5583 	}
5584 
5585 	if (bnxt_devarg_max_num_kflow_invalid(max_num_kflows)) {
5586 		PMD_DRV_LOG(ERR,
5587 			"Invalid value passed to max_num_kflows devarg.\n");
5588 		return -EINVAL;
5589 	}
5590 
5591 	bp->max_num_kflows = max_num_kflows;
5592 	if (bp->max_num_kflows)
5593 		PMD_DRV_LOG(INFO, "max_num_kflows set as %ldK.\n",
5594 				max_num_kflows);
5595 
5596 	return 0;
5597 }
5598 
5599 static int
5600 bnxt_parse_devarg_app_id(__rte_unused const char *key,
5601 				 const char *value, void *opaque_arg)
5602 {
5603 	struct bnxt *bp = opaque_arg;
5604 	unsigned long app_id;
5605 	char *end = NULL;
5606 
5607 	if (!value || !opaque_arg) {
5608 		PMD_DRV_LOG(ERR,
5609 			    "Invalid parameter passed to app-id "
5610 			    "devargs.\n");
5611 		return -EINVAL;
5612 	}
5613 
5614 	app_id = strtoul(value, &end, 10);
5615 	if (end == NULL || *end != '\0' ||
5616 	    (app_id == ULONG_MAX && errno == ERANGE)) {
5617 		PMD_DRV_LOG(ERR,
5618 			    "Invalid parameter passed to app_id "
5619 			    "devargs.\n");
5620 		return -EINVAL;
5621 	}
5622 
5623 	if (BNXT_DEVARG_APP_ID_INVALID(app_id)) {
5624 		PMD_DRV_LOG(ERR, "Invalid app-id(%d) devargs.\n",
5625 			    (uint16_t)app_id);
5626 		return -EINVAL;
5627 	}
5628 
5629 	bp->app_id = app_id;
5630 	PMD_DRV_LOG(INFO, "app-id=%d feature enabled.\n", (uint16_t)app_id);
5631 
5632 	return 0;
5633 }
5634 
5635 static int
5636 bnxt_parse_devarg_rep_is_pf(__rte_unused const char *key,
5637 			    const char *value, void *opaque_arg)
5638 {
5639 	struct bnxt_representor *vfr_bp = opaque_arg;
5640 	unsigned long rep_is_pf;
5641 	char *end = NULL;
5642 
5643 	if (!value || !opaque_arg) {
5644 		PMD_DRV_LOG(ERR,
5645 			    "Invalid parameter passed to rep_is_pf devargs.\n");
5646 		return -EINVAL;
5647 	}
5648 
5649 	rep_is_pf = strtoul(value, &end, 10);
5650 	if (end == NULL || *end != '\0' ||
5651 	    (rep_is_pf == ULONG_MAX && errno == ERANGE)) {
5652 		PMD_DRV_LOG(ERR,
5653 			    "Invalid parameter passed to rep_is_pf devargs.\n");
5654 		return -EINVAL;
5655 	}
5656 
5657 	if (BNXT_DEVARG_REP_IS_PF_INVALID(rep_is_pf)) {
5658 		PMD_DRV_LOG(ERR,
5659 			    "Invalid value passed to rep_is_pf devargs.\n");
5660 		return -EINVAL;
5661 	}
5662 
5663 	vfr_bp->flags |= rep_is_pf;
5664 	if (BNXT_REP_PF(vfr_bp))
5665 		PMD_DRV_LOG(INFO, "PF representor\n");
5666 	else
5667 		PMD_DRV_LOG(INFO, "VF representor\n");
5668 
5669 	return 0;
5670 }
5671 
5672 static int
5673 bnxt_parse_devarg_rep_based_pf(__rte_unused const char *key,
5674 			       const char *value, void *opaque_arg)
5675 {
5676 	struct bnxt_representor *vfr_bp = opaque_arg;
5677 	unsigned long rep_based_pf;
5678 	char *end = NULL;
5679 
5680 	if (!value || !opaque_arg) {
5681 		PMD_DRV_LOG(ERR,
5682 			    "Invalid parameter passed to rep_based_pf "
5683 			    "devargs.\n");
5684 		return -EINVAL;
5685 	}
5686 
5687 	rep_based_pf = strtoul(value, &end, 10);
5688 	if (end == NULL || *end != '\0' ||
5689 	    (rep_based_pf == ULONG_MAX && errno == ERANGE)) {
5690 		PMD_DRV_LOG(ERR,
5691 			    "Invalid parameter passed to rep_based_pf "
5692 			    "devargs.\n");
5693 		return -EINVAL;
5694 	}
5695 
5696 	if (BNXT_DEVARG_REP_BASED_PF_INVALID(rep_based_pf)) {
5697 		PMD_DRV_LOG(ERR,
5698 			    "Invalid value passed to rep_based_pf devargs.\n");
5699 		return -EINVAL;
5700 	}
5701 
5702 	vfr_bp->rep_based_pf = rep_based_pf;
5703 	vfr_bp->flags |= BNXT_REP_BASED_PF_VALID;
5704 
5705 	PMD_DRV_LOG(INFO, "rep-based-pf = %d\n", vfr_bp->rep_based_pf);
5706 
5707 	return 0;
5708 }
5709 
5710 static int
5711 bnxt_parse_devarg_rep_q_r2f(__rte_unused const char *key,
5712 			    const char *value, void *opaque_arg)
5713 {
5714 	struct bnxt_representor *vfr_bp = opaque_arg;
5715 	unsigned long rep_q_r2f;
5716 	char *end = NULL;
5717 
5718 	if (!value || !opaque_arg) {
5719 		PMD_DRV_LOG(ERR,
5720 			    "Invalid parameter passed to rep_q_r2f "
5721 			    "devargs.\n");
5722 		return -EINVAL;
5723 	}
5724 
5725 	rep_q_r2f = strtoul(value, &end, 10);
5726 	if (end == NULL || *end != '\0' ||
5727 	    (rep_q_r2f == ULONG_MAX && errno == ERANGE)) {
5728 		PMD_DRV_LOG(ERR,
5729 			    "Invalid parameter passed to rep_q_r2f "
5730 			    "devargs.\n");
5731 		return -EINVAL;
5732 	}
5733 
5734 	if (BNXT_DEVARG_REP_Q_R2F_INVALID(rep_q_r2f)) {
5735 		PMD_DRV_LOG(ERR,
5736 			    "Invalid value passed to rep_q_r2f devargs.\n");
5737 		return -EINVAL;
5738 	}
5739 
5740 	vfr_bp->rep_q_r2f = rep_q_r2f;
5741 	vfr_bp->flags |= BNXT_REP_Q_R2F_VALID;
5742 	PMD_DRV_LOG(INFO, "rep-q-r2f = %d\n", vfr_bp->rep_q_r2f);
5743 
5744 	return 0;
5745 }
5746 
5747 static int
5748 bnxt_parse_devarg_rep_q_f2r(__rte_unused const char *key,
5749 			    const char *value, void *opaque_arg)
5750 {
5751 	struct bnxt_representor *vfr_bp = opaque_arg;
5752 	unsigned long rep_q_f2r;
5753 	char *end = NULL;
5754 
5755 	if (!value || !opaque_arg) {
5756 		PMD_DRV_LOG(ERR,
5757 			    "Invalid parameter passed to rep_q_f2r "
5758 			    "devargs.\n");
5759 		return -EINVAL;
5760 	}
5761 
5762 	rep_q_f2r = strtoul(value, &end, 10);
5763 	if (end == NULL || *end != '\0' ||
5764 	    (rep_q_f2r == ULONG_MAX && errno == ERANGE)) {
5765 		PMD_DRV_LOG(ERR,
5766 			    "Invalid parameter passed to rep_q_f2r "
5767 			    "devargs.\n");
5768 		return -EINVAL;
5769 	}
5770 
5771 	if (BNXT_DEVARG_REP_Q_F2R_INVALID(rep_q_f2r)) {
5772 		PMD_DRV_LOG(ERR,
5773 			    "Invalid value passed to rep_q_f2r devargs.\n");
5774 		return -EINVAL;
5775 	}
5776 
5777 	vfr_bp->rep_q_f2r = rep_q_f2r;
5778 	vfr_bp->flags |= BNXT_REP_Q_F2R_VALID;
5779 	PMD_DRV_LOG(INFO, "rep-q-f2r = %d\n", vfr_bp->rep_q_f2r);
5780 
5781 	return 0;
5782 }
5783 
5784 static int
5785 bnxt_parse_devarg_rep_fc_r2f(__rte_unused const char *key,
5786 			     const char *value, void *opaque_arg)
5787 {
5788 	struct bnxt_representor *vfr_bp = opaque_arg;
5789 	unsigned long rep_fc_r2f;
5790 	char *end = NULL;
5791 
5792 	if (!value || !opaque_arg) {
5793 		PMD_DRV_LOG(ERR,
5794 			    "Invalid parameter passed to rep_fc_r2f "
5795 			    "devargs.\n");
5796 		return -EINVAL;
5797 	}
5798 
5799 	rep_fc_r2f = strtoul(value, &end, 10);
5800 	if (end == NULL || *end != '\0' ||
5801 	    (rep_fc_r2f == ULONG_MAX && errno == ERANGE)) {
5802 		PMD_DRV_LOG(ERR,
5803 			    "Invalid parameter passed to rep_fc_r2f "
5804 			    "devargs.\n");
5805 		return -EINVAL;
5806 	}
5807 
5808 	if (BNXT_DEVARG_REP_FC_R2F_INVALID(rep_fc_r2f)) {
5809 		PMD_DRV_LOG(ERR,
5810 			    "Invalid value passed to rep_fc_r2f devargs.\n");
5811 		return -EINVAL;
5812 	}
5813 
5814 	vfr_bp->flags |= BNXT_REP_FC_R2F_VALID;
5815 	vfr_bp->rep_fc_r2f = rep_fc_r2f;
5816 	PMD_DRV_LOG(INFO, "rep-fc-r2f = %lu\n", rep_fc_r2f);
5817 
5818 	return 0;
5819 }
5820 
5821 static int
5822 bnxt_parse_devarg_rep_fc_f2r(__rte_unused const char *key,
5823 			     const char *value, void *opaque_arg)
5824 {
5825 	struct bnxt_representor *vfr_bp = opaque_arg;
5826 	unsigned long rep_fc_f2r;
5827 	char *end = NULL;
5828 
5829 	if (!value || !opaque_arg) {
5830 		PMD_DRV_LOG(ERR,
5831 			    "Invalid parameter passed to rep_fc_f2r "
5832 			    "devargs.\n");
5833 		return -EINVAL;
5834 	}
5835 
5836 	rep_fc_f2r = strtoul(value, &end, 10);
5837 	if (end == NULL || *end != '\0' ||
5838 	    (rep_fc_f2r == ULONG_MAX && errno == ERANGE)) {
5839 		PMD_DRV_LOG(ERR,
5840 			    "Invalid parameter passed to rep_fc_f2r "
5841 			    "devargs.\n");
5842 		return -EINVAL;
5843 	}
5844 
5845 	if (BNXT_DEVARG_REP_FC_F2R_INVALID(rep_fc_f2r)) {
5846 		PMD_DRV_LOG(ERR,
5847 			    "Invalid value passed to rep_fc_f2r devargs.\n");
5848 		return -EINVAL;
5849 	}
5850 
5851 	vfr_bp->flags |= BNXT_REP_FC_F2R_VALID;
5852 	vfr_bp->rep_fc_f2r = rep_fc_f2r;
5853 	PMD_DRV_LOG(INFO, "rep-fc-f2r = %lu\n", rep_fc_f2r);
5854 
5855 	return 0;
5856 }
5857 
5858 static int
5859 bnxt_parse_dev_args(struct bnxt *bp, struct rte_devargs *devargs)
5860 {
5861 	struct rte_kvargs *kvlist;
5862 	int ret;
5863 
5864 	if (devargs == NULL)
5865 		return 0;
5866 
5867 	kvlist = rte_kvargs_parse(devargs->args, bnxt_dev_args);
5868 	if (kvlist == NULL)
5869 		return -EINVAL;
5870 
5871 	/*
5872 	 * Handler for "flow_xstat" devarg.
5873 	 * Invoked as for ex: "-a 0000:00:0d.0,flow_xstat=1"
5874 	 */
5875 	ret = rte_kvargs_process(kvlist, BNXT_DEVARG_FLOW_XSTAT,
5876 				 bnxt_parse_devarg_flow_xstat, bp);
5877 	if (ret)
5878 		goto err;
5879 
5880 	/*
5881 	 * Handler for "accum-stats" devarg.
5882 	 * Invoked as for ex: "-a 0000:00:0d.0,accum-stats=1"
5883 	 */
5884 	rte_kvargs_process(kvlist, BNXT_DEVARG_ACCUM_STATS,
5885 			   bnxt_parse_devarg_accum_stats, bp);
5886 	/*
5887 	 * Handler for "max_num_kflows" devarg.
5888 	 * Invoked as for ex: "-a 000:00:0d.0,max_num_kflows=32"
5889 	 */
5890 	ret = rte_kvargs_process(kvlist, BNXT_DEVARG_MAX_NUM_KFLOWS,
5891 				 bnxt_parse_devarg_max_num_kflows, bp);
5892 	if (ret)
5893 		goto err;
5894 
5895 err:
5896 	/*
5897 	 * Handler for "app-id" devarg.
5898 	 * Invoked as for ex: "-a 000:00:0d.0,app-id=1"
5899 	 */
5900 	rte_kvargs_process(kvlist, BNXT_DEVARG_APP_ID,
5901 			   bnxt_parse_devarg_app_id, bp);
5902 
5903 	rte_kvargs_free(kvlist);
5904 	return ret;
5905 }
5906 
5907 static int bnxt_alloc_switch_domain(struct bnxt *bp)
5908 {
5909 	int rc = 0;
5910 
5911 	if (BNXT_PF(bp) || BNXT_VF_IS_TRUSTED(bp)) {
5912 		rc = rte_eth_switch_domain_alloc(&bp->switch_domain_id);
5913 		if (rc)
5914 			PMD_DRV_LOG(ERR,
5915 				    "Failed to alloc switch domain: %d\n", rc);
5916 		else
5917 			PMD_DRV_LOG(INFO,
5918 				    "Switch domain allocated %d\n",
5919 				    bp->switch_domain_id);
5920 	}
5921 
5922 	return rc;
5923 }
5924 
5925 /* Allocate and initialize various fields in bnxt struct that
5926  * need to be allocated/destroyed only once in the lifetime of the driver
5927  */
5928 static int bnxt_drv_init(struct rte_eth_dev *eth_dev)
5929 {
5930 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
5931 	struct bnxt *bp = eth_dev->data->dev_private;
5932 	int rc = 0;
5933 
5934 	bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
5935 
5936 	if (bnxt_vf_pciid(pci_dev->id.device_id))
5937 		bp->flags |= BNXT_FLAG_VF;
5938 
5939 	if (bnxt_p5_device(pci_dev->id.device_id))
5940 		bp->flags |= BNXT_FLAG_CHIP_P5;
5941 
5942 	if (pci_dev->id.device_id == BROADCOM_DEV_ID_58802 ||
5943 	    pci_dev->id.device_id == BROADCOM_DEV_ID_58804 ||
5944 	    pci_dev->id.device_id == BROADCOM_DEV_ID_58808 ||
5945 	    pci_dev->id.device_id == BROADCOM_DEV_ID_58802_VF)
5946 		bp->flags |= BNXT_FLAG_STINGRAY;
5947 
5948 	if (BNXT_TRUFLOW_EN(bp)) {
5949 		/* extra mbuf field is required to store CFA code from mark */
5950 		static const struct rte_mbuf_dynfield bnxt_cfa_code_dynfield_desc = {
5951 			.name = RTE_PMD_BNXT_CFA_CODE_DYNFIELD_NAME,
5952 			.size = sizeof(bnxt_cfa_code_dynfield_t),
5953 			.align = __alignof__(bnxt_cfa_code_dynfield_t),
5954 		};
5955 		bnxt_cfa_code_dynfield_offset =
5956 			rte_mbuf_dynfield_register(&bnxt_cfa_code_dynfield_desc);
5957 		if (bnxt_cfa_code_dynfield_offset < 0) {
5958 			PMD_DRV_LOG(ERR,
5959 			    "Failed to register mbuf field for TruFlow mark\n");
5960 			return -rte_errno;
5961 		}
5962 	}
5963 
5964 	rc = bnxt_map_pci_bars(eth_dev);
5965 	if (rc) {
5966 		PMD_DRV_LOG(ERR,
5967 			    "Failed to initialize board rc: %x\n", rc);
5968 		return rc;
5969 	}
5970 
5971 	rc = bnxt_alloc_pf_info(bp);
5972 	if (rc)
5973 		return rc;
5974 
5975 	rc = bnxt_alloc_link_info(bp);
5976 	if (rc)
5977 		return rc;
5978 
5979 	rc = bnxt_alloc_parent_info(bp);
5980 	if (rc)
5981 		return rc;
5982 
5983 	rc = bnxt_alloc_hwrm_resources(bp);
5984 	if (rc) {
5985 		PMD_DRV_LOG(ERR,
5986 			    "Failed to allocate response buffer rc: %x\n", rc);
5987 		return rc;
5988 	}
5989 	rc = bnxt_alloc_leds_info(bp);
5990 	if (rc)
5991 		return rc;
5992 
5993 	rc = bnxt_alloc_cos_queues(bp);
5994 	if (rc)
5995 		return rc;
5996 
5997 	rc = bnxt_init_locks(bp);
5998 	if (rc)
5999 		return rc;
6000 
6001 	rc = bnxt_alloc_switch_domain(bp);
6002 	if (rc)
6003 		return rc;
6004 
6005 	return rc;
6006 }
6007 
6008 static int
6009 bnxt_dev_init(struct rte_eth_dev *eth_dev, void *params __rte_unused)
6010 {
6011 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
6012 	static int version_printed;
6013 	struct bnxt *bp;
6014 	int rc;
6015 
6016 	if (version_printed++ == 0)
6017 		PMD_DRV_LOG(INFO, "%s\n", bnxt_version);
6018 
6019 	eth_dev->dev_ops = &bnxt_dev_ops;
6020 	eth_dev->rx_queue_count = bnxt_rx_queue_count_op;
6021 	eth_dev->rx_descriptor_status = bnxt_rx_descriptor_status_op;
6022 	eth_dev->tx_descriptor_status = bnxt_tx_descriptor_status_op;
6023 	eth_dev->rx_pkt_burst = &bnxt_recv_pkts;
6024 	eth_dev->tx_pkt_burst = &bnxt_xmit_pkts;
6025 
6026 	/*
6027 	 * For secondary processes, we don't initialise any further
6028 	 * as primary has already done this work.
6029 	 */
6030 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
6031 		return 0;
6032 
6033 	rte_eth_copy_pci_info(eth_dev, pci_dev);
6034 	eth_dev->data->dev_flags |= RTE_ETH_DEV_AUTOFILL_QUEUE_XSTATS;
6035 
6036 	bp = eth_dev->data->dev_private;
6037 
6038 	/* Parse dev arguments passed on when starting the DPDK application. */
6039 	rc = bnxt_parse_dev_args(bp, pci_dev->device.devargs);
6040 	if (rc)
6041 		goto error_free;
6042 
6043 	rc = bnxt_drv_init(eth_dev);
6044 	if (rc)
6045 		goto error_free;
6046 
6047 	rc = bnxt_init_resources(bp, false);
6048 	if (rc)
6049 		goto error_free;
6050 
6051 	rc = bnxt_alloc_stats_mem(bp);
6052 	if (rc)
6053 		goto error_free;
6054 
6055 	PMD_DRV_LOG(INFO,
6056 		    "Found %s device at mem %" PRIX64 ", node addr %pM\n",
6057 		    DRV_MODULE_NAME,
6058 		    pci_dev->mem_resource[0].phys_addr,
6059 		    pci_dev->mem_resource[0].addr);
6060 
6061 	return 0;
6062 
6063 error_free:
6064 	bnxt_dev_uninit(eth_dev);
6065 	return rc;
6066 }
6067 
6068 
6069 static void bnxt_free_ctx_mem_buf(struct bnxt_ctx_mem_buf_info *ctx)
6070 {
6071 	if (!ctx)
6072 		return;
6073 
6074 	if (ctx->va)
6075 		rte_free(ctx->va);
6076 
6077 	ctx->va = NULL;
6078 	ctx->dma = RTE_BAD_IOVA;
6079 	ctx->ctx_id = BNXT_CTX_VAL_INVAL;
6080 }
6081 
6082 static void bnxt_unregister_fc_ctx_mem(struct bnxt *bp)
6083 {
6084 	bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_RX,
6085 				  CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
6086 				  bp->flow_stat->rx_fc_out_tbl.ctx_id,
6087 				  bp->flow_stat->max_fc,
6088 				  false);
6089 
6090 	bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_TX,
6091 				  CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
6092 				  bp->flow_stat->tx_fc_out_tbl.ctx_id,
6093 				  bp->flow_stat->max_fc,
6094 				  false);
6095 
6096 	if (bp->flow_stat->rx_fc_in_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
6097 		bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->rx_fc_in_tbl.ctx_id);
6098 	bp->flow_stat->rx_fc_in_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
6099 
6100 	if (bp->flow_stat->rx_fc_out_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
6101 		bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->rx_fc_out_tbl.ctx_id);
6102 	bp->flow_stat->rx_fc_out_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
6103 
6104 	if (bp->flow_stat->tx_fc_in_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
6105 		bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->tx_fc_in_tbl.ctx_id);
6106 	bp->flow_stat->tx_fc_in_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
6107 
6108 	if (bp->flow_stat->tx_fc_out_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
6109 		bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->tx_fc_out_tbl.ctx_id);
6110 	bp->flow_stat->tx_fc_out_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
6111 }
6112 
6113 static void bnxt_uninit_fc_ctx_mem(struct bnxt *bp)
6114 {
6115 	bnxt_unregister_fc_ctx_mem(bp);
6116 
6117 	bnxt_free_ctx_mem_buf(&bp->flow_stat->rx_fc_in_tbl);
6118 	bnxt_free_ctx_mem_buf(&bp->flow_stat->rx_fc_out_tbl);
6119 	bnxt_free_ctx_mem_buf(&bp->flow_stat->tx_fc_in_tbl);
6120 	bnxt_free_ctx_mem_buf(&bp->flow_stat->tx_fc_out_tbl);
6121 }
6122 
6123 static void bnxt_uninit_ctx_mem(struct bnxt *bp)
6124 {
6125 	if (BNXT_FLOW_XSTATS_EN(bp))
6126 		bnxt_uninit_fc_ctx_mem(bp);
6127 }
6128 
6129 static void
6130 bnxt_free_error_recovery_info(struct bnxt *bp)
6131 {
6132 	rte_free(bp->recovery_info);
6133 	bp->recovery_info = NULL;
6134 	bp->fw_cap &= ~BNXT_FW_CAP_ERROR_RECOVERY;
6135 }
6136 
6137 static int
6138 bnxt_uninit_resources(struct bnxt *bp, bool reconfig_dev)
6139 {
6140 	int rc;
6141 
6142 	bnxt_free_int(bp);
6143 	bnxt_free_mem(bp, reconfig_dev);
6144 
6145 	bnxt_hwrm_func_buf_unrgtr(bp);
6146 	if (bp->pf != NULL) {
6147 		rte_free(bp->pf->vf_req_buf);
6148 		bp->pf->vf_req_buf = NULL;
6149 	}
6150 
6151 	rc = bnxt_hwrm_func_driver_unregister(bp, 0);
6152 	bp->flags &= ~BNXT_FLAG_REGISTERED;
6153 	bnxt_free_ctx_mem(bp);
6154 	if (!reconfig_dev) {
6155 		bnxt_free_hwrm_resources(bp);
6156 		bnxt_free_error_recovery_info(bp);
6157 	}
6158 
6159 	bnxt_uninit_ctx_mem(bp);
6160 
6161 	bnxt_free_flow_stats_info(bp);
6162 	if (bp->rep_info != NULL)
6163 		bnxt_free_switch_domain(bp);
6164 	bnxt_free_rep_info(bp);
6165 	rte_free(bp->ptp_cfg);
6166 	bp->ptp_cfg = NULL;
6167 	return rc;
6168 }
6169 
6170 static int
6171 bnxt_dev_uninit(struct rte_eth_dev *eth_dev)
6172 {
6173 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
6174 		return -EPERM;
6175 
6176 	PMD_DRV_LOG(DEBUG, "Calling Device uninit\n");
6177 
6178 	if (eth_dev->state != RTE_ETH_DEV_UNUSED)
6179 		bnxt_dev_close_op(eth_dev);
6180 
6181 	return 0;
6182 }
6183 
6184 static int bnxt_pci_remove_dev_with_reps(struct rte_eth_dev *eth_dev)
6185 {
6186 	struct bnxt *bp = eth_dev->data->dev_private;
6187 	struct rte_eth_dev *vf_rep_eth_dev;
6188 	int ret = 0, i;
6189 
6190 	if (!bp)
6191 		return -EINVAL;
6192 
6193 	for (i = 0; i < bp->num_reps; i++) {
6194 		vf_rep_eth_dev = bp->rep_info[i].vfr_eth_dev;
6195 		if (!vf_rep_eth_dev)
6196 			continue;
6197 		PMD_DRV_LOG(DEBUG, "BNXT Port:%d VFR pci remove\n",
6198 			    vf_rep_eth_dev->data->port_id);
6199 		rte_eth_dev_destroy(vf_rep_eth_dev, bnxt_representor_uninit);
6200 	}
6201 	PMD_DRV_LOG(DEBUG, "BNXT Port:%d pci remove\n",
6202 		    eth_dev->data->port_id);
6203 	ret = rte_eth_dev_destroy(eth_dev, bnxt_dev_uninit);
6204 
6205 	return ret;
6206 }
6207 
6208 static void bnxt_free_rep_info(struct bnxt *bp)
6209 {
6210 	rte_free(bp->rep_info);
6211 	bp->rep_info = NULL;
6212 	rte_free(bp->cfa_code_map);
6213 	bp->cfa_code_map = NULL;
6214 }
6215 
6216 static int bnxt_init_rep_info(struct bnxt *bp)
6217 {
6218 	int i = 0, rc;
6219 
6220 	if (bp->rep_info)
6221 		return 0;
6222 
6223 	bp->rep_info = rte_zmalloc("bnxt_rep_info",
6224 				   sizeof(bp->rep_info[0]) * BNXT_MAX_VF_REPS,
6225 				   0);
6226 	if (!bp->rep_info) {
6227 		PMD_DRV_LOG(ERR, "Failed to alloc memory for rep info\n");
6228 		return -ENOMEM;
6229 	}
6230 	bp->cfa_code_map = rte_zmalloc("bnxt_cfa_code_map",
6231 				       sizeof(*bp->cfa_code_map) *
6232 				       BNXT_MAX_CFA_CODE, 0);
6233 	if (!bp->cfa_code_map) {
6234 		PMD_DRV_LOG(ERR, "Failed to alloc memory for cfa_code_map\n");
6235 		bnxt_free_rep_info(bp);
6236 		return -ENOMEM;
6237 	}
6238 
6239 	for (i = 0; i < BNXT_MAX_CFA_CODE; i++)
6240 		bp->cfa_code_map[i] = BNXT_VF_IDX_INVALID;
6241 
6242 	rc = pthread_mutex_init(&bp->rep_info->vfr_lock, NULL);
6243 	if (rc) {
6244 		PMD_DRV_LOG(ERR, "Unable to initialize vfr_lock\n");
6245 		bnxt_free_rep_info(bp);
6246 		return rc;
6247 	}
6248 
6249 	rc = pthread_mutex_init(&bp->rep_info->vfr_start_lock, NULL);
6250 	if (rc) {
6251 		PMD_DRV_LOG(ERR, "Unable to initialize vfr_start_lock\n");
6252 		bnxt_free_rep_info(bp);
6253 		return rc;
6254 	}
6255 
6256 	return rc;
6257 }
6258 
6259 static int bnxt_rep_port_probe(struct rte_pci_device *pci_dev,
6260 			       struct rte_eth_devargs *eth_da,
6261 			       struct rte_eth_dev *backing_eth_dev,
6262 			       const char *dev_args)
6263 {
6264 	struct rte_eth_dev *vf_rep_eth_dev;
6265 	char name[RTE_ETH_NAME_MAX_LEN];
6266 	struct bnxt *backing_bp;
6267 	uint16_t num_rep;
6268 	int i, ret = 0;
6269 	struct rte_kvargs *kvlist = NULL;
6270 
6271 	if (eth_da->type == RTE_ETH_REPRESENTOR_NONE)
6272 		return 0;
6273 	if (eth_da->type != RTE_ETH_REPRESENTOR_VF) {
6274 		PMD_DRV_LOG(ERR, "unsupported representor type %d\n",
6275 			    eth_da->type);
6276 		return -ENOTSUP;
6277 	}
6278 	num_rep = eth_da->nb_representor_ports;
6279 	if (num_rep > BNXT_MAX_VF_REPS) {
6280 		PMD_DRV_LOG(ERR, "nb_representor_ports = %d > %d MAX VF REPS\n",
6281 			    num_rep, BNXT_MAX_VF_REPS);
6282 		return -EINVAL;
6283 	}
6284 
6285 	if (num_rep >= RTE_MAX_ETHPORTS) {
6286 		PMD_DRV_LOG(ERR,
6287 			    "nb_representor_ports = %d > %d MAX ETHPORTS\n",
6288 			    num_rep, RTE_MAX_ETHPORTS);
6289 		return -EINVAL;
6290 	}
6291 
6292 	backing_bp = backing_eth_dev->data->dev_private;
6293 
6294 	if (!(BNXT_PF(backing_bp) || BNXT_VF_IS_TRUSTED(backing_bp))) {
6295 		PMD_DRV_LOG(ERR,
6296 			    "Not a PF or trusted VF. No Representor support\n");
6297 		/* Returning an error is not an option.
6298 		 * Applications are not handling this correctly
6299 		 */
6300 		return 0;
6301 	}
6302 
6303 	if (bnxt_init_rep_info(backing_bp))
6304 		return 0;
6305 
6306 	for (i = 0; i < num_rep; i++) {
6307 		struct bnxt_representor representor = {
6308 			.vf_id = eth_da->representor_ports[i],
6309 			.switch_domain_id = backing_bp->switch_domain_id,
6310 			.parent_dev = backing_eth_dev
6311 		};
6312 
6313 		if (representor.vf_id >= BNXT_MAX_VF_REPS) {
6314 			PMD_DRV_LOG(ERR, "VF-Rep id %d >= %d MAX VF ID\n",
6315 				    representor.vf_id, BNXT_MAX_VF_REPS);
6316 			continue;
6317 		}
6318 
6319 		/* representor port net_bdf_port */
6320 		snprintf(name, sizeof(name), "net_%s_representor_%d",
6321 			 pci_dev->device.name, eth_da->representor_ports[i]);
6322 
6323 		kvlist = rte_kvargs_parse(dev_args, bnxt_dev_args);
6324 		if (kvlist) {
6325 			/*
6326 			 * Handler for "rep_is_pf" devarg.
6327 			 * Invoked as for ex: "-a 000:00:0d.0,
6328 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6329 			 */
6330 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_IS_PF,
6331 						 bnxt_parse_devarg_rep_is_pf,
6332 						 (void *)&representor);
6333 			if (ret) {
6334 				ret = -EINVAL;
6335 				goto err;
6336 			}
6337 			/*
6338 			 * Handler for "rep_based_pf" devarg.
6339 			 * Invoked as for ex: "-a 000:00:0d.0,
6340 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6341 			 */
6342 			ret = rte_kvargs_process(kvlist,
6343 						 BNXT_DEVARG_REP_BASED_PF,
6344 						 bnxt_parse_devarg_rep_based_pf,
6345 						 (void *)&representor);
6346 			if (ret) {
6347 				ret = -EINVAL;
6348 				goto err;
6349 			}
6350 			/*
6351 			 * Handler for "rep_based_pf" devarg.
6352 			 * Invoked as for ex: "-a 000:00:0d.0,
6353 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6354 			 */
6355 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_Q_R2F,
6356 						 bnxt_parse_devarg_rep_q_r2f,
6357 						 (void *)&representor);
6358 			if (ret) {
6359 				ret = -EINVAL;
6360 				goto err;
6361 			}
6362 			/*
6363 			 * Handler for "rep_based_pf" devarg.
6364 			 * Invoked as for ex: "-a 000:00:0d.0,
6365 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6366 			 */
6367 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_Q_F2R,
6368 						 bnxt_parse_devarg_rep_q_f2r,
6369 						 (void *)&representor);
6370 			if (ret) {
6371 				ret = -EINVAL;
6372 				goto err;
6373 			}
6374 			/*
6375 			 * Handler for "rep_based_pf" devarg.
6376 			 * Invoked as for ex: "-a 000:00:0d.0,
6377 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6378 			 */
6379 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_FC_R2F,
6380 						 bnxt_parse_devarg_rep_fc_r2f,
6381 						 (void *)&representor);
6382 			if (ret) {
6383 				ret = -EINVAL;
6384 				goto err;
6385 			}
6386 			/*
6387 			 * Handler for "rep_based_pf" devarg.
6388 			 * Invoked as for ex: "-a 000:00:0d.0,
6389 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6390 			 */
6391 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_FC_F2R,
6392 						 bnxt_parse_devarg_rep_fc_f2r,
6393 						 (void *)&representor);
6394 			if (ret) {
6395 				ret = -EINVAL;
6396 				goto err;
6397 			}
6398 		}
6399 
6400 		ret = rte_eth_dev_create(&pci_dev->device, name,
6401 					 sizeof(struct bnxt_representor),
6402 					 NULL, NULL,
6403 					 bnxt_representor_init,
6404 					 &representor);
6405 		if (ret) {
6406 			PMD_DRV_LOG(ERR, "failed to create bnxt vf "
6407 				    "representor %s.", name);
6408 			goto err;
6409 		}
6410 
6411 		vf_rep_eth_dev = rte_eth_dev_allocated(name);
6412 		if (!vf_rep_eth_dev) {
6413 			PMD_DRV_LOG(ERR, "Failed to find the eth_dev"
6414 				    " for VF-Rep: %s.", name);
6415 			ret = -ENODEV;
6416 			goto err;
6417 		}
6418 
6419 		PMD_DRV_LOG(DEBUG, "BNXT Port:%d VFR pci probe\n",
6420 			    backing_eth_dev->data->port_id);
6421 		backing_bp->rep_info[representor.vf_id].vfr_eth_dev =
6422 							 vf_rep_eth_dev;
6423 		backing_bp->num_reps++;
6424 
6425 	}
6426 
6427 	rte_kvargs_free(kvlist);
6428 	return 0;
6429 
6430 err:
6431 	/* If num_rep > 1, then rollback already created
6432 	 * ports, since we'll be failing the probe anyway
6433 	 */
6434 	if (num_rep > 1)
6435 		bnxt_pci_remove_dev_with_reps(backing_eth_dev);
6436 	rte_errno = -ret;
6437 	rte_kvargs_free(kvlist);
6438 
6439 	return ret;
6440 }
6441 
6442 static int bnxt_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
6443 			  struct rte_pci_device *pci_dev)
6444 {
6445 	struct rte_eth_devargs eth_da = { .nb_representor_ports = 0 };
6446 	struct rte_eth_dev *backing_eth_dev;
6447 	uint16_t num_rep;
6448 	int ret = 0;
6449 
6450 	if (pci_dev->device.devargs) {
6451 		ret = rte_eth_devargs_parse(pci_dev->device.devargs->args,
6452 					    &eth_da);
6453 		if (ret)
6454 			return ret;
6455 	}
6456 
6457 	num_rep = eth_da.nb_representor_ports;
6458 	PMD_DRV_LOG(DEBUG, "nb_representor_ports = %d\n",
6459 		    num_rep);
6460 
6461 	/* We could come here after first level of probe is already invoked
6462 	 * as part of an application bringup(OVS-DPDK vswitchd), so first check
6463 	 * for already allocated eth_dev for the backing device (PF/Trusted VF)
6464 	 */
6465 	backing_eth_dev = rte_eth_dev_allocated(pci_dev->device.name);
6466 	if (backing_eth_dev == NULL) {
6467 		ret = rte_eth_dev_create(&pci_dev->device, pci_dev->device.name,
6468 					 sizeof(struct bnxt),
6469 					 eth_dev_pci_specific_init, pci_dev,
6470 					 bnxt_dev_init, NULL);
6471 
6472 		if (ret || !num_rep)
6473 			return ret;
6474 
6475 		backing_eth_dev = rte_eth_dev_allocated(pci_dev->device.name);
6476 	}
6477 	PMD_DRV_LOG(DEBUG, "BNXT Port:%d pci probe\n",
6478 		    backing_eth_dev->data->port_id);
6479 
6480 	if (!num_rep)
6481 		return ret;
6482 
6483 	/* probe representor ports now */
6484 	ret = bnxt_rep_port_probe(pci_dev, &eth_da, backing_eth_dev,
6485 				  pci_dev->device.devargs->args);
6486 
6487 	return ret;
6488 }
6489 
6490 static int bnxt_pci_remove(struct rte_pci_device *pci_dev)
6491 {
6492 	struct rte_eth_dev *eth_dev;
6493 
6494 	eth_dev = rte_eth_dev_allocated(pci_dev->device.name);
6495 	if (!eth_dev)
6496 		return 0; /* Invoked typically only by OVS-DPDK, by the
6497 			   * time it comes here the eth_dev is already
6498 			   * deleted by rte_eth_dev_close(), so returning
6499 			   * +ve value will at least help in proper cleanup
6500 			   */
6501 
6502 	PMD_DRV_LOG(DEBUG, "BNXT Port:%d pci remove\n", eth_dev->data->port_id);
6503 	if (rte_eal_process_type() == RTE_PROC_PRIMARY) {
6504 		if (eth_dev->data->dev_flags & RTE_ETH_DEV_REPRESENTOR)
6505 			return rte_eth_dev_destroy(eth_dev,
6506 						   bnxt_representor_uninit);
6507 		else
6508 			return rte_eth_dev_destroy(eth_dev,
6509 						   bnxt_dev_uninit);
6510 	} else {
6511 		return rte_eth_dev_pci_generic_remove(pci_dev, NULL);
6512 	}
6513 }
6514 
6515 static struct rte_pci_driver bnxt_rte_pmd = {
6516 	.id_table = bnxt_pci_id_map,
6517 	.drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC |
6518 			RTE_PCI_DRV_INTR_RMV |
6519 			RTE_PCI_DRV_PROBE_AGAIN, /* Needed in case of VF-REPs
6520 						  * and OVS-DPDK
6521 						  */
6522 	.probe = bnxt_pci_probe,
6523 	.remove = bnxt_pci_remove,
6524 };
6525 
6526 static bool
6527 is_device_supported(struct rte_eth_dev *dev, struct rte_pci_driver *drv)
6528 {
6529 	if (strcmp(dev->device->driver->name, drv->driver.name))
6530 		return false;
6531 
6532 	return true;
6533 }
6534 
6535 bool is_bnxt_supported(struct rte_eth_dev *dev)
6536 {
6537 	return is_device_supported(dev, &bnxt_rte_pmd);
6538 }
6539 
6540 RTE_LOG_REGISTER_SUFFIX(bnxt_logtype_driver, driver, NOTICE);
6541 RTE_PMD_REGISTER_PCI(net_bnxt, bnxt_rte_pmd);
6542 RTE_PMD_REGISTER_PCI_TABLE(net_bnxt, bnxt_pci_id_map);
6543 RTE_PMD_REGISTER_KMOD_DEP(net_bnxt, "* igb_uio | uio_pci_generic | vfio-pci");
6544