xref: /dpdk/drivers/net/bnxt/bnxt_ethdev.c (revision 2f6ac042)
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 & RTE_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 & RTE_ETH_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 & RTE_ETH_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->flags |= BNXT_FLAG_JUMBO;
741 	else
742 		bp->flags &= ~BNXT_FLAG_JUMBO;
743 
744 	/* THOR does not support ring groups.
745 	 * But we will use the array to save RSS context IDs.
746 	 */
747 	if (BNXT_CHIP_P5(bp))
748 		bp->max_ring_grps = BNXT_MAX_RSS_CTXTS_P5;
749 
750 	rc = bnxt_alloc_hwrm_rings(bp);
751 	if (rc) {
752 		PMD_DRV_LOG(ERR, "HWRM ring alloc failure rc: %x\n", rc);
753 		goto err_out;
754 	}
755 
756 	rc = bnxt_alloc_all_hwrm_ring_grps(bp);
757 	if (rc) {
758 		PMD_DRV_LOG(ERR, "HWRM ring grp alloc failure: %x\n", rc);
759 		goto err_out;
760 	}
761 
762 	if (!(bp->vnic_cap_flags & BNXT_VNIC_CAP_COS_CLASSIFY))
763 		goto skip_cosq_cfg;
764 
765 	for (j = 0, i = 0; i < BNXT_COS_QUEUE_COUNT; i++) {
766 		if (bp->rx_cos_queue[i].id != 0xff) {
767 			struct bnxt_vnic_info *vnic = &bp->vnic_info[j++];
768 
769 			if (!vnic) {
770 				PMD_DRV_LOG(ERR,
771 					    "Num pools more than FW profile\n");
772 				rc = -EINVAL;
773 				goto err_out;
774 			}
775 			vnic->cos_queue_id = bp->rx_cos_queue[i].id;
776 			bp->rx_cosq_cnt++;
777 		}
778 	}
779 
780 skip_cosq_cfg:
781 	rc = bnxt_mq_rx_configure(bp);
782 	if (rc) {
783 		PMD_DRV_LOG(ERR, "MQ mode configure failure rc: %x\n", rc);
784 		goto err_out;
785 	}
786 
787 	for (j = 0; j < bp->rx_nr_rings; j++) {
788 		struct bnxt_rx_queue *rxq = bp->rx_queues[j];
789 
790 		if (!rxq->rx_deferred_start) {
791 			bp->eth_dev->data->rx_queue_state[j] =
792 				RTE_ETH_QUEUE_STATE_STARTED;
793 			rxq->rx_started = true;
794 		}
795 	}
796 
797 	/* default vnic 0 */
798 	rc = bnxt_setup_one_vnic(bp, 0);
799 	if (rc)
800 		goto err_out;
801 	/* VNIC configuration */
802 	if (BNXT_RFS_NEEDS_VNIC(bp)) {
803 		for (i = 1; i < bp->nr_vnics; i++) {
804 			rc = bnxt_setup_one_vnic(bp, i);
805 			if (rc)
806 				goto err_out;
807 		}
808 	}
809 
810 	for (j = 0; j < bp->tx_nr_rings; j++) {
811 		struct bnxt_tx_queue *txq = bp->tx_queues[j];
812 
813 		if (!txq->tx_deferred_start) {
814 			bp->eth_dev->data->tx_queue_state[j] =
815 				RTE_ETH_QUEUE_STATE_STARTED;
816 			txq->tx_started = true;
817 		}
818 	}
819 
820 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, &bp->vnic_info[0], 0, NULL);
821 	if (rc) {
822 		PMD_DRV_LOG(ERR,
823 			"HWRM cfa l2 rx mask failure rc: %x\n", rc);
824 		goto err_out;
825 	}
826 
827 	/* check and configure queue intr-vector mapping */
828 	if ((rte_intr_cap_multiple(intr_handle) ||
829 	     !RTE_ETH_DEV_SRIOV(bp->eth_dev).active) &&
830 	    bp->eth_dev->data->dev_conf.intr_conf.rxq != 0) {
831 		intr_vector = bp->eth_dev->data->nb_rx_queues;
832 		PMD_DRV_LOG(DEBUG, "intr_vector = %d\n", intr_vector);
833 		if (intr_vector > bp->rx_cp_nr_rings) {
834 			PMD_DRV_LOG(ERR, "At most %d intr queues supported",
835 					bp->rx_cp_nr_rings);
836 			return -ENOTSUP;
837 		}
838 		rc = rte_intr_efd_enable(intr_handle, intr_vector);
839 		if (rc)
840 			return rc;
841 	}
842 
843 	if (rte_intr_dp_is_en(intr_handle)) {
844 		if (rte_intr_vec_list_alloc(intr_handle, "intr_vec",
845 					bp->eth_dev->data->nb_rx_queues)) {
846 			PMD_DRV_LOG(ERR, "Failed to allocate %d rx_queues"
847 				" intr_vec", bp->eth_dev->data->nb_rx_queues);
848 			rc = -ENOMEM;
849 			goto err_out;
850 		}
851 		PMD_DRV_LOG(DEBUG, "intr_handle->nb_efd = %d "
852 			    "intr_handle->max_intr = %d\n",
853 			    rte_intr_nb_efd_get(intr_handle),
854 			    rte_intr_max_intr_get(intr_handle));
855 		for (queue_id = 0; queue_id < bp->eth_dev->data->nb_rx_queues;
856 		     queue_id++) {
857 			rte_intr_vec_list_index_set(intr_handle,
858 					queue_id, vec + BNXT_RX_VEC_START);
859 			if (vec < base + rte_intr_nb_efd_get(intr_handle)
860 			    - 1)
861 				vec++;
862 		}
863 	}
864 
865 	/* enable uio/vfio intr/eventfd mapping */
866 	rc = rte_intr_enable(intr_handle);
867 #ifndef RTE_EXEC_ENV_FREEBSD
868 	/* In FreeBSD OS, nic_uio driver does not support interrupts */
869 	if (rc)
870 		goto err_out;
871 #endif
872 
873 	rc = bnxt_update_phy_setting(bp);
874 	if (rc)
875 		goto err_out;
876 
877 	bp->mark_table = rte_zmalloc("bnxt_mark_table", BNXT_MARK_TABLE_SZ, 0);
878 	if (!bp->mark_table)
879 		PMD_DRV_LOG(ERR, "Allocation of mark table failed\n");
880 
881 	return 0;
882 
883 err_out:
884 	/* Some of the error status returned by FW may not be from errno.h */
885 	if (rc > 0)
886 		rc = -EIO;
887 
888 	return rc;
889 }
890 
891 static int bnxt_shutdown_nic(struct bnxt *bp)
892 {
893 	bnxt_free_all_hwrm_resources(bp);
894 	bnxt_free_all_filters(bp);
895 	bnxt_free_all_vnics(bp);
896 	return 0;
897 }
898 
899 /*
900  * Device configuration and status function
901  */
902 
903 uint32_t bnxt_get_speed_capabilities(struct bnxt *bp)
904 {
905 	uint32_t link_speed = 0;
906 	uint32_t speed_capa = 0;
907 
908 	if (bp->link_info == NULL)
909 		return 0;
910 
911 	link_speed = bp->link_info->support_speeds;
912 
913 	/* If PAM4 is configured, use PAM4 supported speed */
914 	if (link_speed == 0 && bp->link_info->support_pam4_speeds > 0)
915 		link_speed = bp->link_info->support_pam4_speeds;
916 
917 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_LINK_SPEED_100MB)
918 		speed_capa |= RTE_ETH_LINK_SPEED_100M;
919 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_100MBHD)
920 		speed_capa |= RTE_ETH_LINK_SPEED_100M_HD;
921 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_1GB)
922 		speed_capa |= RTE_ETH_LINK_SPEED_1G;
923 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_2_5GB)
924 		speed_capa |= RTE_ETH_LINK_SPEED_2_5G;
925 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_10GB)
926 		speed_capa |= RTE_ETH_LINK_SPEED_10G;
927 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_20GB)
928 		speed_capa |= RTE_ETH_LINK_SPEED_20G;
929 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_25GB)
930 		speed_capa |= RTE_ETH_LINK_SPEED_25G;
931 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_40GB)
932 		speed_capa |= RTE_ETH_LINK_SPEED_40G;
933 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_50GB)
934 		speed_capa |= RTE_ETH_LINK_SPEED_50G;
935 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_100GB)
936 		speed_capa |= RTE_ETH_LINK_SPEED_100G;
937 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_PAM4_SPEEDS_50G)
938 		speed_capa |= RTE_ETH_LINK_SPEED_50G;
939 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_PAM4_SPEEDS_100G)
940 		speed_capa |= RTE_ETH_LINK_SPEED_100G;
941 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_PAM4_SPEEDS_200G)
942 		speed_capa |= RTE_ETH_LINK_SPEED_200G;
943 
944 	if (bp->link_info->auto_mode ==
945 	    HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_MODE_NONE)
946 		speed_capa |= RTE_ETH_LINK_SPEED_FIXED;
947 
948 	return speed_capa;
949 }
950 
951 static int bnxt_dev_info_get_op(struct rte_eth_dev *eth_dev,
952 				struct rte_eth_dev_info *dev_info)
953 {
954 	struct rte_pci_device *pdev = RTE_DEV_TO_PCI(eth_dev->device);
955 	struct bnxt *bp = eth_dev->data->dev_private;
956 	uint16_t max_vnics, i, j, vpool, vrxq;
957 	unsigned int max_rx_rings;
958 	int rc;
959 
960 	rc = is_bnxt_in_error(bp);
961 	if (rc)
962 		return rc;
963 
964 	/* MAC Specifics */
965 	dev_info->max_mac_addrs = bp->max_l2_ctx;
966 	dev_info->max_hash_mac_addrs = 0;
967 
968 	/* PF/VF specifics */
969 	if (BNXT_PF(bp))
970 		dev_info->max_vfs = pdev->max_vfs;
971 
972 	max_rx_rings = bnxt_max_rings(bp);
973 	/* For the sake of symmetry, max_rx_queues = max_tx_queues */
974 	dev_info->max_rx_queues = max_rx_rings;
975 	dev_info->max_tx_queues = max_rx_rings;
976 	dev_info->reta_size = bnxt_rss_hash_tbl_size(bp);
977 	dev_info->hash_key_size = HW_HASH_KEY_SIZE;
978 	max_vnics = bp->max_vnics;
979 
980 	/* MTU specifics */
981 	dev_info->min_mtu = RTE_ETHER_MIN_MTU;
982 	dev_info->max_mtu = BNXT_MAX_MTU;
983 
984 	/* Fast path specifics */
985 	dev_info->min_rx_bufsize = 1;
986 	dev_info->max_rx_pktlen = BNXT_MAX_PKT_LEN;
987 
988 	dev_info->rx_offload_capa = BNXT_DEV_RX_OFFLOAD_SUPPORT;
989 	if (bp->flags & BNXT_FLAG_PTP_SUPPORTED)
990 		dev_info->rx_offload_capa |= RTE_ETH_RX_OFFLOAD_TIMESTAMP;
991 	if (bp->vnic_cap_flags & BNXT_VNIC_CAP_VLAN_RX_STRIP)
992 		dev_info->rx_offload_capa |= RTE_ETH_RX_OFFLOAD_VLAN_STRIP;
993 	dev_info->tx_queue_offload_capa = RTE_ETH_TX_OFFLOAD_MBUF_FAST_FREE;
994 	dev_info->tx_offload_capa = BNXT_DEV_TX_OFFLOAD_SUPPORT |
995 				    dev_info->tx_queue_offload_capa;
996 	if (bp->fw_cap & BNXT_FW_CAP_VLAN_TX_INSERT)
997 		dev_info->tx_offload_capa |= RTE_ETH_TX_OFFLOAD_VLAN_INSERT;
998 	dev_info->flow_type_rss_offloads = BNXT_ETH_RSS_SUPPORT;
999 
1000 	dev_info->speed_capa = bnxt_get_speed_capabilities(bp);
1001 	dev_info->dev_capa = RTE_ETH_DEV_CAPA_RUNTIME_RX_QUEUE_SETUP |
1002 			     RTE_ETH_DEV_CAPA_RUNTIME_TX_QUEUE_SETUP;
1003 	dev_info->dev_capa &= ~RTE_ETH_DEV_CAPA_FLOW_RULE_KEEP;
1004 
1005 	dev_info->default_rxconf = (struct rte_eth_rxconf) {
1006 		.rx_thresh = {
1007 			.pthresh = 8,
1008 			.hthresh = 8,
1009 			.wthresh = 0,
1010 		},
1011 		.rx_free_thresh = 32,
1012 		.rx_drop_en = BNXT_DEFAULT_RX_DROP_EN,
1013 	};
1014 
1015 	dev_info->default_txconf = (struct rte_eth_txconf) {
1016 		.tx_thresh = {
1017 			.pthresh = 32,
1018 			.hthresh = 0,
1019 			.wthresh = 0,
1020 		},
1021 		.tx_free_thresh = 32,
1022 		.tx_rs_thresh = 32,
1023 	};
1024 	eth_dev->data->dev_conf.intr_conf.lsc = 1;
1025 
1026 	dev_info->rx_desc_lim.nb_min = BNXT_MIN_RING_DESC;
1027 	dev_info->rx_desc_lim.nb_max = BNXT_MAX_RX_RING_DESC;
1028 	dev_info->tx_desc_lim.nb_min = BNXT_MIN_RING_DESC;
1029 	dev_info->tx_desc_lim.nb_max = BNXT_MAX_TX_RING_DESC;
1030 
1031 	if (BNXT_PF(bp) || BNXT_VF_IS_TRUSTED(bp)) {
1032 		dev_info->switch_info.name = eth_dev->device->name;
1033 		dev_info->switch_info.domain_id = bp->switch_domain_id;
1034 		dev_info->switch_info.port_id =
1035 				BNXT_PF(bp) ? BNXT_SWITCH_PORT_ID_PF :
1036 				    BNXT_SWITCH_PORT_ID_TRUSTED_VF;
1037 	}
1038 
1039 	/*
1040 	 * TODO: default_rxconf, default_txconf, rx_desc_lim, and tx_desc_lim
1041 	 *       need further investigation.
1042 	 */
1043 
1044 	/* VMDq resources */
1045 	vpool = 64; /* RTE_ETH_64_POOLS */
1046 	vrxq = 128; /* RTE_ETH_VMDQ_DCB_NUM_QUEUES */
1047 	for (i = 0; i < 4; vpool >>= 1, i++) {
1048 		if (max_vnics > vpool) {
1049 			for (j = 0; j < 5; vrxq >>= 1, j++) {
1050 				if (dev_info->max_rx_queues > vrxq) {
1051 					if (vpool > vrxq)
1052 						vpool = vrxq;
1053 					goto found;
1054 				}
1055 			}
1056 			/* Not enough resources to support VMDq */
1057 			break;
1058 		}
1059 	}
1060 	/* Not enough resources to support VMDq */
1061 	vpool = 0;
1062 	vrxq = 0;
1063 found:
1064 	dev_info->max_vmdq_pools = vpool;
1065 	dev_info->vmdq_queue_num = vrxq;
1066 
1067 	dev_info->vmdq_pool_base = 0;
1068 	dev_info->vmdq_queue_base = 0;
1069 
1070 	return 0;
1071 }
1072 
1073 /* Configure the device based on the configuration provided */
1074 static int bnxt_dev_configure_op(struct rte_eth_dev *eth_dev)
1075 {
1076 	struct bnxt *bp = eth_dev->data->dev_private;
1077 	uint64_t rx_offloads = eth_dev->data->dev_conf.rxmode.offloads;
1078 	int rc;
1079 
1080 	bp->rx_queues = (void *)eth_dev->data->rx_queues;
1081 	bp->tx_queues = (void *)eth_dev->data->tx_queues;
1082 	bp->tx_nr_rings = eth_dev->data->nb_tx_queues;
1083 	bp->rx_nr_rings = eth_dev->data->nb_rx_queues;
1084 
1085 	rc = is_bnxt_in_error(bp);
1086 	if (rc)
1087 		return rc;
1088 
1089 	if (BNXT_VF(bp) && (bp->flags & BNXT_FLAG_NEW_RM)) {
1090 		rc = bnxt_hwrm_check_vf_rings(bp);
1091 		if (rc) {
1092 			PMD_DRV_LOG(ERR, "HWRM insufficient resources\n");
1093 			return -ENOSPC;
1094 		}
1095 
1096 		/* If a resource has already been allocated - in this case
1097 		 * it is the async completion ring, free it. Reallocate it after
1098 		 * resource reservation. This will ensure the resource counts
1099 		 * are calculated correctly.
1100 		 */
1101 
1102 		pthread_mutex_lock(&bp->def_cp_lock);
1103 
1104 		if (!BNXT_HAS_NQ(bp) && bp->async_cp_ring) {
1105 			bnxt_disable_int(bp);
1106 			bnxt_free_cp_ring(bp, bp->async_cp_ring);
1107 		}
1108 
1109 		rc = bnxt_hwrm_func_reserve_vf_resc(bp, false);
1110 		if (rc) {
1111 			PMD_DRV_LOG(ERR, "HWRM resource alloc fail:%x\n", rc);
1112 			pthread_mutex_unlock(&bp->def_cp_lock);
1113 			return -ENOSPC;
1114 		}
1115 
1116 		if (!BNXT_HAS_NQ(bp) && bp->async_cp_ring) {
1117 			rc = bnxt_alloc_async_cp_ring(bp);
1118 			if (rc) {
1119 				pthread_mutex_unlock(&bp->def_cp_lock);
1120 				return rc;
1121 			}
1122 			bnxt_enable_int(bp);
1123 		}
1124 
1125 		pthread_mutex_unlock(&bp->def_cp_lock);
1126 	}
1127 
1128 	/* Inherit new configurations */
1129 	if (eth_dev->data->nb_rx_queues > bp->max_rx_rings ||
1130 	    eth_dev->data->nb_tx_queues > bp->max_tx_rings ||
1131 	    eth_dev->data->nb_rx_queues + eth_dev->data->nb_tx_queues
1132 		+ BNXT_NUM_ASYNC_CPR(bp) > bp->max_cp_rings ||
1133 	    eth_dev->data->nb_rx_queues + eth_dev->data->nb_tx_queues >
1134 	    bp->max_stat_ctx)
1135 		goto resource_error;
1136 
1137 	if (BNXT_HAS_RING_GRPS(bp) &&
1138 	    (uint32_t)(eth_dev->data->nb_rx_queues) > bp->max_ring_grps)
1139 		goto resource_error;
1140 
1141 	if (!(eth_dev->data->dev_conf.rxmode.mq_mode & RTE_ETH_MQ_RX_RSS) &&
1142 	    bp->max_vnics < eth_dev->data->nb_rx_queues)
1143 		goto resource_error;
1144 
1145 	bp->rx_cp_nr_rings = bp->rx_nr_rings;
1146 	bp->tx_cp_nr_rings = bp->tx_nr_rings;
1147 
1148 	if (eth_dev->data->dev_conf.rxmode.mq_mode & RTE_ETH_MQ_RX_RSS_FLAG)
1149 		rx_offloads |= RTE_ETH_RX_OFFLOAD_RSS_HASH;
1150 	eth_dev->data->dev_conf.rxmode.offloads = rx_offloads;
1151 
1152 	bnxt_mtu_set_op(eth_dev, eth_dev->data->mtu);
1153 
1154 	return 0;
1155 
1156 resource_error:
1157 	PMD_DRV_LOG(ERR,
1158 		    "Insufficient resources to support requested config\n");
1159 	PMD_DRV_LOG(ERR,
1160 		    "Num Queues Requested: Tx %d, Rx %d\n",
1161 		    eth_dev->data->nb_tx_queues,
1162 		    eth_dev->data->nb_rx_queues);
1163 	PMD_DRV_LOG(ERR,
1164 		    "MAX: TxQ %d, RxQ %d, CQ %d Stat %d, Grp %d, Vnic %d\n",
1165 		    bp->max_tx_rings, bp->max_rx_rings, bp->max_cp_rings,
1166 		    bp->max_stat_ctx, bp->max_ring_grps, bp->max_vnics);
1167 	return -ENOSPC;
1168 }
1169 
1170 void bnxt_print_link_info(struct rte_eth_dev *eth_dev)
1171 {
1172 	struct rte_eth_link *link = &eth_dev->data->dev_link;
1173 
1174 	if (link->link_status)
1175 		PMD_DRV_LOG(INFO, "Port %d Link Up - speed %u Mbps - %s\n",
1176 			eth_dev->data->port_id,
1177 			(uint32_t)link->link_speed,
1178 			(link->link_duplex == RTE_ETH_LINK_FULL_DUPLEX) ?
1179 			("full-duplex") : ("half-duplex\n"));
1180 	else
1181 		PMD_DRV_LOG(INFO, "Port %d Link Down\n",
1182 			eth_dev->data->port_id);
1183 }
1184 
1185 /*
1186  * Determine whether the current configuration requires support for scattered
1187  * receive; return 1 if scattered receive is required and 0 if not.
1188  */
1189 static int bnxt_scattered_rx(struct rte_eth_dev *eth_dev)
1190 {
1191 	uint32_t overhead = BNXT_MAX_PKT_LEN - BNXT_MAX_MTU;
1192 	uint16_t buf_size;
1193 	int i;
1194 
1195 	if (eth_dev->data->dev_conf.rxmode.offloads & RTE_ETH_RX_OFFLOAD_SCATTER)
1196 		return 1;
1197 
1198 	if (eth_dev->data->dev_conf.rxmode.offloads & RTE_ETH_RX_OFFLOAD_TCP_LRO)
1199 		return 1;
1200 
1201 	for (i = 0; i < eth_dev->data->nb_rx_queues; i++) {
1202 		struct bnxt_rx_queue *rxq = eth_dev->data->rx_queues[i];
1203 
1204 		buf_size = (uint16_t)(rte_pktmbuf_data_room_size(rxq->mb_pool) -
1205 				      RTE_PKTMBUF_HEADROOM);
1206 		if (eth_dev->data->mtu + overhead > buf_size)
1207 			return 1;
1208 	}
1209 	return 0;
1210 }
1211 
1212 static eth_rx_burst_t
1213 bnxt_receive_function(struct rte_eth_dev *eth_dev)
1214 {
1215 	struct bnxt *bp = eth_dev->data->dev_private;
1216 
1217 	/* Disable vector mode RX for Stingray2 for now */
1218 	if (BNXT_CHIP_SR2(bp)) {
1219 		bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
1220 		return bnxt_recv_pkts;
1221 	}
1222 
1223 #if (defined(RTE_ARCH_X86) || defined(RTE_ARCH_ARM64)) && \
1224 	!defined(RTE_LIBRTE_IEEE1588)
1225 
1226 	/* Vector mode receive cannot be enabled if scattered rx is in use. */
1227 	if (eth_dev->data->scattered_rx)
1228 		goto use_scalar_rx;
1229 
1230 	/*
1231 	 * Vector mode receive cannot be enabled if Truflow is enabled or if
1232 	 * asynchronous completions and receive completions can be placed in
1233 	 * the same completion ring.
1234 	 */
1235 	if (BNXT_TRUFLOW_EN(bp) || !BNXT_NUM_ASYNC_CPR(bp))
1236 		goto use_scalar_rx;
1237 
1238 	/*
1239 	 * Vector mode receive cannot be enabled if any receive offloads outside
1240 	 * a limited subset have been enabled.
1241 	 */
1242 	if (eth_dev->data->dev_conf.rxmode.offloads &
1243 		~(RTE_ETH_RX_OFFLOAD_VLAN_STRIP |
1244 		  RTE_ETH_RX_OFFLOAD_KEEP_CRC |
1245 		  RTE_ETH_RX_OFFLOAD_IPV4_CKSUM |
1246 		  RTE_ETH_RX_OFFLOAD_UDP_CKSUM |
1247 		  RTE_ETH_RX_OFFLOAD_TCP_CKSUM |
1248 		  RTE_ETH_RX_OFFLOAD_OUTER_IPV4_CKSUM |
1249 		  RTE_ETH_RX_OFFLOAD_OUTER_UDP_CKSUM |
1250 		  RTE_ETH_RX_OFFLOAD_RSS_HASH |
1251 		  RTE_ETH_RX_OFFLOAD_VLAN_FILTER))
1252 		goto use_scalar_rx;
1253 
1254 #if defined(RTE_ARCH_X86) && defined(CC_AVX2_SUPPORT)
1255 	if (rte_vect_get_max_simd_bitwidth() >= RTE_VECT_SIMD_256 &&
1256 	    rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX2) == 1) {
1257 		PMD_DRV_LOG(INFO,
1258 			    "Using AVX2 vector mode receive for port %d\n",
1259 			    eth_dev->data->port_id);
1260 		bp->flags |= BNXT_FLAG_RX_VECTOR_PKT_MODE;
1261 		return bnxt_recv_pkts_vec_avx2;
1262 	}
1263  #endif
1264 	if (rte_vect_get_max_simd_bitwidth() >= RTE_VECT_SIMD_128) {
1265 		PMD_DRV_LOG(INFO,
1266 			    "Using SSE vector mode receive for port %d\n",
1267 			    eth_dev->data->port_id);
1268 		bp->flags |= BNXT_FLAG_RX_VECTOR_PKT_MODE;
1269 		return bnxt_recv_pkts_vec;
1270 	}
1271 
1272 use_scalar_rx:
1273 	PMD_DRV_LOG(INFO, "Vector mode receive disabled for port %d\n",
1274 		    eth_dev->data->port_id);
1275 	PMD_DRV_LOG(INFO,
1276 		    "Port %d scatter: %d rx offload: %" PRIX64 "\n",
1277 		    eth_dev->data->port_id,
1278 		    eth_dev->data->scattered_rx,
1279 		    eth_dev->data->dev_conf.rxmode.offloads);
1280 #endif
1281 	bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
1282 	return bnxt_recv_pkts;
1283 }
1284 
1285 static eth_tx_burst_t
1286 bnxt_transmit_function(struct rte_eth_dev *eth_dev)
1287 {
1288 	struct bnxt *bp = eth_dev->data->dev_private;
1289 
1290 	/* Disable vector mode TX for Stingray2 for now */
1291 	if (BNXT_CHIP_SR2(bp))
1292 		return bnxt_xmit_pkts;
1293 
1294 #if defined(RTE_ARCH_X86) || defined(RTE_ARCH_ARM64) && \
1295 	!defined(RTE_LIBRTE_IEEE1588)
1296 	uint64_t offloads = eth_dev->data->dev_conf.txmode.offloads;
1297 
1298 	/*
1299 	 * Vector mode transmit can be enabled only if not using scatter rx
1300 	 * or tx offloads.
1301 	 */
1302 	if (eth_dev->data->scattered_rx ||
1303 	    (offloads & ~RTE_ETH_TX_OFFLOAD_MBUF_FAST_FREE) ||
1304 	    BNXT_TRUFLOW_EN(bp))
1305 		goto use_scalar_tx;
1306 
1307 #if defined(RTE_ARCH_X86) && defined(CC_AVX2_SUPPORT)
1308 	if (rte_vect_get_max_simd_bitwidth() >= RTE_VECT_SIMD_256 &&
1309 	    rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX2) == 1) {
1310 		PMD_DRV_LOG(INFO,
1311 			    "Using AVX2 vector mode transmit for port %d\n",
1312 			    eth_dev->data->port_id);
1313 		return bnxt_xmit_pkts_vec_avx2;
1314 	}
1315 #endif
1316 	if (rte_vect_get_max_simd_bitwidth() >= RTE_VECT_SIMD_128) {
1317 		PMD_DRV_LOG(INFO,
1318 			    "Using SSE vector mode transmit for port %d\n",
1319 			    eth_dev->data->port_id);
1320 		return bnxt_xmit_pkts_vec;
1321 	}
1322 
1323 use_scalar_tx:
1324 	PMD_DRV_LOG(INFO, "Vector mode transmit disabled for port %d\n",
1325 		    eth_dev->data->port_id);
1326 	PMD_DRV_LOG(INFO,
1327 		    "Port %d scatter: %d tx offload: %" PRIX64 "\n",
1328 		    eth_dev->data->port_id,
1329 		    eth_dev->data->scattered_rx,
1330 		    offloads);
1331 #endif
1332 	return bnxt_xmit_pkts;
1333 }
1334 
1335 static int bnxt_handle_if_change_status(struct bnxt *bp)
1336 {
1337 	int rc;
1338 
1339 	/* Since fw has undergone a reset and lost all contexts,
1340 	 * set fatal flag to not issue hwrm during cleanup
1341 	 */
1342 	bp->flags |= BNXT_FLAG_FATAL_ERROR;
1343 	bnxt_uninit_resources(bp, true);
1344 
1345 	/* clear fatal flag so that re-init happens */
1346 	bp->flags &= ~BNXT_FLAG_FATAL_ERROR;
1347 	rc = bnxt_init_resources(bp, true);
1348 
1349 	bp->flags &= ~BNXT_FLAG_IF_CHANGE_HOT_FW_RESET_DONE;
1350 
1351 	return rc;
1352 }
1353 
1354 static int bnxt_dev_set_link_up_op(struct rte_eth_dev *eth_dev)
1355 {
1356 	struct bnxt *bp = eth_dev->data->dev_private;
1357 	int rc = 0;
1358 
1359 	if (!BNXT_SINGLE_PF(bp))
1360 		return -ENOTSUP;
1361 
1362 	if (!bp->link_info->link_up)
1363 		rc = bnxt_set_hwrm_link_config(bp, true);
1364 	if (!rc)
1365 		eth_dev->data->dev_link.link_status = 1;
1366 
1367 	bnxt_print_link_info(eth_dev);
1368 	return rc;
1369 }
1370 
1371 static int bnxt_dev_set_link_down_op(struct rte_eth_dev *eth_dev)
1372 {
1373 	struct bnxt *bp = eth_dev->data->dev_private;
1374 
1375 	if (!BNXT_SINGLE_PF(bp))
1376 		return -ENOTSUP;
1377 
1378 	eth_dev->data->dev_link.link_status = 0;
1379 	bnxt_set_hwrm_link_config(bp, false);
1380 	bp->link_info->link_up = 0;
1381 
1382 	return 0;
1383 }
1384 
1385 static void bnxt_free_switch_domain(struct bnxt *bp)
1386 {
1387 	int rc = 0;
1388 
1389 	if (!(BNXT_PF(bp) || BNXT_VF_IS_TRUSTED(bp)))
1390 		return;
1391 
1392 	rc = rte_eth_switch_domain_free(bp->switch_domain_id);
1393 	if (rc)
1394 		PMD_DRV_LOG(ERR, "free switch domain:%d fail: %d\n",
1395 			    bp->switch_domain_id, rc);
1396 }
1397 
1398 static void bnxt_ptp_get_current_time(void *arg)
1399 {
1400 	struct bnxt *bp = arg;
1401 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
1402 	int rc;
1403 
1404 	rc = is_bnxt_in_error(bp);
1405 	if (rc)
1406 		return;
1407 
1408 	if (!ptp)
1409 		return;
1410 
1411 	bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_CURRENT_TIME,
1412 				&ptp->current_time);
1413 
1414 	rc = rte_eal_alarm_set(US_PER_S, bnxt_ptp_get_current_time, (void *)bp);
1415 	if (rc != 0) {
1416 		PMD_DRV_LOG(ERR, "Failed to re-schedule PTP alarm\n");
1417 		bp->flags2 &= ~BNXT_FLAGS2_PTP_ALARM_SCHEDULED;
1418 	}
1419 }
1420 
1421 static int bnxt_schedule_ptp_alarm(struct bnxt *bp)
1422 {
1423 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
1424 	int rc;
1425 
1426 	if (bp->flags2 & BNXT_FLAGS2_PTP_ALARM_SCHEDULED)
1427 		return 0;
1428 
1429 	bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_CURRENT_TIME,
1430 				&ptp->current_time);
1431 
1432 	rc = rte_eal_alarm_set(US_PER_S, bnxt_ptp_get_current_time, (void *)bp);
1433 	return rc;
1434 }
1435 
1436 static void bnxt_cancel_ptp_alarm(struct bnxt *bp)
1437 {
1438 	if (bp->flags2 & BNXT_FLAGS2_PTP_ALARM_SCHEDULED) {
1439 		rte_eal_alarm_cancel(bnxt_ptp_get_current_time, (void *)bp);
1440 		bp->flags2 &= ~BNXT_FLAGS2_PTP_ALARM_SCHEDULED;
1441 	}
1442 }
1443 
1444 static void bnxt_ptp_stop(struct bnxt *bp)
1445 {
1446 	bnxt_cancel_ptp_alarm(bp);
1447 	bp->flags2 &= ~BNXT_FLAGS2_PTP_TIMESYNC_ENABLED;
1448 }
1449 
1450 static int bnxt_ptp_start(struct bnxt *bp)
1451 {
1452 	int rc;
1453 
1454 	rc = bnxt_schedule_ptp_alarm(bp);
1455 	if (rc != 0) {
1456 		PMD_DRV_LOG(ERR, "Failed to schedule PTP alarm\n");
1457 	} else {
1458 		bp->flags2 |= BNXT_FLAGS2_PTP_TIMESYNC_ENABLED;
1459 		bp->flags2 |= BNXT_FLAGS2_PTP_ALARM_SCHEDULED;
1460 	}
1461 
1462 	return rc;
1463 }
1464 
1465 static int bnxt_dev_stop(struct rte_eth_dev *eth_dev)
1466 {
1467 	struct bnxt *bp = eth_dev->data->dev_private;
1468 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
1469 	struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
1470 	struct rte_eth_link link;
1471 	int ret;
1472 
1473 	eth_dev->data->dev_started = 0;
1474 
1475 	/* Prevent crashes when queues are still in use */
1476 	eth_dev->rx_pkt_burst = &bnxt_dummy_recv_pkts;
1477 	eth_dev->tx_pkt_burst = &bnxt_dummy_xmit_pkts;
1478 
1479 	bnxt_disable_int(bp);
1480 
1481 	/* disable uio/vfio intr/eventfd mapping */
1482 	rte_intr_disable(intr_handle);
1483 
1484 	/* Stop the child representors for this device */
1485 	ret = bnxt_rep_stop_all(bp);
1486 	if (ret != 0)
1487 		return ret;
1488 
1489 	/* delete the bnxt ULP port details */
1490 	bnxt_ulp_port_deinit(bp);
1491 
1492 	bnxt_cancel_fw_health_check(bp);
1493 
1494 	if (BNXT_P5_PTP_TIMESYNC_ENABLED(bp))
1495 		bnxt_cancel_ptp_alarm(bp);
1496 
1497 	/* Do not bring link down during reset recovery */
1498 	if (!is_bnxt_in_error(bp)) {
1499 		bnxt_dev_set_link_down_op(eth_dev);
1500 		/* Wait for link to be reset */
1501 		if (BNXT_SINGLE_PF(bp))
1502 			rte_delay_ms(500);
1503 		/* clear the recorded link status */
1504 		memset(&link, 0, sizeof(link));
1505 		rte_eth_linkstatus_set(eth_dev, &link);
1506 	}
1507 
1508 	/* Clean queue intr-vector mapping */
1509 	rte_intr_efd_disable(intr_handle);
1510 	rte_intr_vec_list_free(intr_handle);
1511 
1512 	bnxt_hwrm_port_clr_stats(bp);
1513 	bnxt_free_tx_mbufs(bp);
1514 	bnxt_free_rx_mbufs(bp);
1515 	/* Process any remaining notifications in default completion queue */
1516 	bnxt_int_handler(eth_dev);
1517 	bnxt_shutdown_nic(bp);
1518 	bnxt_hwrm_if_change(bp, false);
1519 
1520 	bnxt_free_prev_ring_stats(bp);
1521 	rte_free(bp->mark_table);
1522 	bp->mark_table = NULL;
1523 
1524 	bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
1525 	bp->rx_cosq_cnt = 0;
1526 	/* All filters are deleted on a port stop. */
1527 	if (BNXT_FLOW_XSTATS_EN(bp))
1528 		bp->flow_stat->flow_count = 0;
1529 
1530 	eth_dev->data->scattered_rx = 0;
1531 
1532 	return 0;
1533 }
1534 
1535 /* Unload the driver, release resources */
1536 static int bnxt_dev_stop_op(struct rte_eth_dev *eth_dev)
1537 {
1538 	struct bnxt *bp = eth_dev->data->dev_private;
1539 
1540 	pthread_mutex_lock(&bp->err_recovery_lock);
1541 	if (bp->flags & BNXT_FLAG_FW_RESET) {
1542 		PMD_DRV_LOG(ERR,
1543 			    "Adapter recovering from error..Please retry\n");
1544 		pthread_mutex_unlock(&bp->err_recovery_lock);
1545 		return -EAGAIN;
1546 	}
1547 	pthread_mutex_unlock(&bp->err_recovery_lock);
1548 
1549 	return bnxt_dev_stop(eth_dev);
1550 }
1551 
1552 static int bnxt_dev_start_op(struct rte_eth_dev *eth_dev)
1553 {
1554 	struct bnxt *bp = eth_dev->data->dev_private;
1555 	uint64_t rx_offloads = eth_dev->data->dev_conf.rxmode.offloads;
1556 	int vlan_mask = 0;
1557 	int rc, retry_cnt = BNXT_IF_CHANGE_RETRY_COUNT;
1558 
1559 	if (!eth_dev->data->nb_tx_queues || !eth_dev->data->nb_rx_queues) {
1560 		PMD_DRV_LOG(ERR, "Queues are not configured yet!\n");
1561 		return -EINVAL;
1562 	}
1563 
1564 	if (bp->rx_cp_nr_rings > RTE_ETHDEV_QUEUE_STAT_CNTRS)
1565 		PMD_DRV_LOG(ERR,
1566 			    "RxQ cnt %d > RTE_ETHDEV_QUEUE_STAT_CNTRS %d\n",
1567 			    bp->rx_cp_nr_rings, RTE_ETHDEV_QUEUE_STAT_CNTRS);
1568 
1569 	do {
1570 		rc = bnxt_hwrm_if_change(bp, true);
1571 		if (rc == 0 || rc != -EAGAIN)
1572 			break;
1573 
1574 		rte_delay_ms(BNXT_IF_CHANGE_RETRY_INTERVAL);
1575 	} while (retry_cnt--);
1576 
1577 	if (rc)
1578 		return rc;
1579 
1580 	if (bp->flags & BNXT_FLAG_IF_CHANGE_HOT_FW_RESET_DONE) {
1581 		rc = bnxt_handle_if_change_status(bp);
1582 		if (rc)
1583 			return rc;
1584 	}
1585 
1586 	bnxt_enable_int(bp);
1587 
1588 	eth_dev->data->scattered_rx = bnxt_scattered_rx(eth_dev);
1589 
1590 	rc = bnxt_start_nic(bp);
1591 	if (rc)
1592 		goto error;
1593 
1594 	rc = bnxt_alloc_prev_ring_stats(bp);
1595 	if (rc)
1596 		goto error;
1597 
1598 	eth_dev->data->dev_started = 1;
1599 
1600 	bnxt_link_update_op(eth_dev, 1);
1601 
1602 	if (rx_offloads & RTE_ETH_RX_OFFLOAD_VLAN_FILTER)
1603 		vlan_mask |= RTE_ETH_VLAN_FILTER_MASK;
1604 	if (rx_offloads & RTE_ETH_RX_OFFLOAD_VLAN_STRIP)
1605 		vlan_mask |= RTE_ETH_VLAN_STRIP_MASK;
1606 	rc = bnxt_vlan_offload_set_op(eth_dev, vlan_mask);
1607 	if (rc)
1608 		goto error;
1609 
1610 	/* Initialize bnxt ULP port details */
1611 	rc = bnxt_ulp_port_init(bp);
1612 	if (rc)
1613 		goto error;
1614 
1615 	eth_dev->rx_pkt_burst = bnxt_receive_function(eth_dev);
1616 	eth_dev->tx_pkt_burst = bnxt_transmit_function(eth_dev);
1617 
1618 	bnxt_schedule_fw_health_check(bp);
1619 
1620 	if (BNXT_P5_PTP_TIMESYNC_ENABLED(bp))
1621 		bnxt_schedule_ptp_alarm(bp);
1622 
1623 	return 0;
1624 
1625 error:
1626 	bnxt_dev_stop(eth_dev);
1627 	return rc;
1628 }
1629 
1630 static void
1631 bnxt_uninit_locks(struct bnxt *bp)
1632 {
1633 	pthread_mutex_destroy(&bp->flow_lock);
1634 	pthread_mutex_destroy(&bp->def_cp_lock);
1635 	pthread_mutex_destroy(&bp->health_check_lock);
1636 	pthread_mutex_destroy(&bp->err_recovery_lock);
1637 	if (bp->rep_info) {
1638 		pthread_mutex_destroy(&bp->rep_info->vfr_lock);
1639 		pthread_mutex_destroy(&bp->rep_info->vfr_start_lock);
1640 	}
1641 }
1642 
1643 static void bnxt_drv_uninit(struct bnxt *bp)
1644 {
1645 	bnxt_free_leds_info(bp);
1646 	bnxt_free_cos_queues(bp);
1647 	bnxt_free_link_info(bp);
1648 	bnxt_free_parent_info(bp);
1649 	bnxt_uninit_locks(bp);
1650 
1651 	rte_memzone_free((const struct rte_memzone *)bp->tx_mem_zone);
1652 	bp->tx_mem_zone = NULL;
1653 	rte_memzone_free((const struct rte_memzone *)bp->rx_mem_zone);
1654 	bp->rx_mem_zone = NULL;
1655 
1656 	bnxt_free_vf_info(bp);
1657 	bnxt_free_pf_info(bp);
1658 
1659 	rte_free(bp->grp_info);
1660 	bp->grp_info = NULL;
1661 }
1662 
1663 static int bnxt_dev_close_op(struct rte_eth_dev *eth_dev)
1664 {
1665 	struct bnxt *bp = eth_dev->data->dev_private;
1666 	int ret = 0;
1667 
1668 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
1669 		return 0;
1670 
1671 	pthread_mutex_lock(&bp->err_recovery_lock);
1672 	if (bp->flags & BNXT_FLAG_FW_RESET) {
1673 		PMD_DRV_LOG(ERR,
1674 			    "Adapter recovering from error...Please retry\n");
1675 		pthread_mutex_unlock(&bp->err_recovery_lock);
1676 		return -EAGAIN;
1677 	}
1678 	pthread_mutex_unlock(&bp->err_recovery_lock);
1679 
1680 	/* cancel the recovery handler before remove dev */
1681 	rte_eal_alarm_cancel(bnxt_dev_reset_and_resume, (void *)bp);
1682 	rte_eal_alarm_cancel(bnxt_dev_recover, (void *)bp);
1683 	bnxt_cancel_fc_thread(bp);
1684 
1685 	if (eth_dev->data->dev_started)
1686 		ret = bnxt_dev_stop(eth_dev);
1687 
1688 	bnxt_uninit_resources(bp, false);
1689 
1690 	bnxt_drv_uninit(bp);
1691 
1692 	return ret;
1693 }
1694 
1695 static void bnxt_mac_addr_remove_op(struct rte_eth_dev *eth_dev,
1696 				    uint32_t index)
1697 {
1698 	struct bnxt *bp = eth_dev->data->dev_private;
1699 	uint64_t pool_mask = eth_dev->data->mac_pool_sel[index];
1700 	struct bnxt_vnic_info *vnic;
1701 	struct bnxt_filter_info *filter, *temp_filter;
1702 	uint32_t i;
1703 
1704 	if (is_bnxt_in_error(bp))
1705 		return;
1706 
1707 	/*
1708 	 * Loop through all VNICs from the specified filter flow pools to
1709 	 * remove the corresponding MAC addr filter
1710 	 */
1711 	for (i = 0; i < bp->nr_vnics; i++) {
1712 		if (!(pool_mask & (1ULL << i)))
1713 			continue;
1714 
1715 		vnic = &bp->vnic_info[i];
1716 		filter = STAILQ_FIRST(&vnic->filter);
1717 		while (filter) {
1718 			temp_filter = STAILQ_NEXT(filter, next);
1719 			if (filter->mac_index == index) {
1720 				STAILQ_REMOVE(&vnic->filter, filter,
1721 						bnxt_filter_info, next);
1722 				bnxt_hwrm_clear_l2_filter(bp, filter);
1723 				bnxt_free_filter(bp, filter);
1724 			}
1725 			filter = temp_filter;
1726 		}
1727 	}
1728 }
1729 
1730 static int bnxt_add_mac_filter(struct bnxt *bp, struct bnxt_vnic_info *vnic,
1731 			       struct rte_ether_addr *mac_addr, uint32_t index,
1732 			       uint32_t pool)
1733 {
1734 	struct bnxt_filter_info *filter;
1735 	int rc = 0;
1736 
1737 	/* Attach requested MAC address to the new l2_filter */
1738 	STAILQ_FOREACH(filter, &vnic->filter, next) {
1739 		if (filter->mac_index == index) {
1740 			PMD_DRV_LOG(DEBUG,
1741 				    "MAC addr already existed for pool %d\n",
1742 				    pool);
1743 			return 0;
1744 		}
1745 	}
1746 
1747 	filter = bnxt_alloc_filter(bp);
1748 	if (!filter) {
1749 		PMD_DRV_LOG(ERR, "L2 filter alloc failed\n");
1750 		return -ENODEV;
1751 	}
1752 
1753 	/* bnxt_alloc_filter copies default MAC to filter->l2_addr. So,
1754 	 * if the MAC that's been programmed now is a different one, then,
1755 	 * copy that addr to filter->l2_addr
1756 	 */
1757 	if (mac_addr)
1758 		memcpy(filter->l2_addr, mac_addr, RTE_ETHER_ADDR_LEN);
1759 	filter->flags |= HWRM_CFA_L2_FILTER_ALLOC_INPUT_FLAGS_OUTERMOST;
1760 
1761 	rc = bnxt_hwrm_set_l2_filter(bp, vnic->fw_vnic_id, filter);
1762 	if (!rc) {
1763 		filter->mac_index = index;
1764 		if (filter->mac_index == 0)
1765 			STAILQ_INSERT_HEAD(&vnic->filter, filter, next);
1766 		else
1767 			STAILQ_INSERT_TAIL(&vnic->filter, filter, next);
1768 	} else {
1769 		bnxt_free_filter(bp, filter);
1770 	}
1771 
1772 	return rc;
1773 }
1774 
1775 static int bnxt_mac_addr_add_op(struct rte_eth_dev *eth_dev,
1776 				struct rte_ether_addr *mac_addr,
1777 				uint32_t index, uint32_t pool)
1778 {
1779 	struct bnxt *bp = eth_dev->data->dev_private;
1780 	struct bnxt_vnic_info *vnic = &bp->vnic_info[pool];
1781 	int rc = 0;
1782 
1783 	rc = is_bnxt_in_error(bp);
1784 	if (rc)
1785 		return rc;
1786 
1787 	if (BNXT_VF(bp) && !BNXT_VF_IS_TRUSTED(bp)) {
1788 		PMD_DRV_LOG(ERR, "Cannot add MAC address to a VF interface\n");
1789 		return -ENOTSUP;
1790 	}
1791 
1792 	if (!vnic) {
1793 		PMD_DRV_LOG(ERR, "VNIC not found for pool %d!\n", pool);
1794 		return -EINVAL;
1795 	}
1796 
1797 	/* Filter settings will get applied when port is started */
1798 	if (!eth_dev->data->dev_started)
1799 		return 0;
1800 
1801 	rc = bnxt_add_mac_filter(bp, vnic, mac_addr, index, pool);
1802 
1803 	return rc;
1804 }
1805 
1806 int bnxt_link_update_op(struct rte_eth_dev *eth_dev, int wait_to_complete)
1807 {
1808 	int rc = 0;
1809 	struct bnxt *bp = eth_dev->data->dev_private;
1810 	struct rte_eth_link new;
1811 	int cnt = wait_to_complete ? BNXT_MAX_LINK_WAIT_CNT :
1812 			BNXT_MIN_LINK_WAIT_CNT;
1813 
1814 	rc = is_bnxt_in_error(bp);
1815 	if (rc)
1816 		return rc;
1817 
1818 	memset(&new, 0, sizeof(new));
1819 
1820 	if (bp->link_info == NULL)
1821 		goto out;
1822 
1823 	do {
1824 		/* Retrieve link info from hardware */
1825 		rc = bnxt_get_hwrm_link_config(bp, &new);
1826 		if (rc) {
1827 			new.link_speed = RTE_ETH_LINK_SPEED_100M;
1828 			new.link_duplex = RTE_ETH_LINK_FULL_DUPLEX;
1829 			PMD_DRV_LOG(ERR,
1830 				"Failed to retrieve link rc = 0x%x!\n", rc);
1831 			goto out;
1832 		}
1833 
1834 		if (!wait_to_complete || new.link_status)
1835 			break;
1836 
1837 		rte_delay_ms(BNXT_LINK_WAIT_INTERVAL);
1838 	} while (cnt--);
1839 
1840 	/* Only single function PF can bring phy down.
1841 	 * When port is stopped, report link down for VF/MH/NPAR functions.
1842 	 */
1843 	if (!BNXT_SINGLE_PF(bp) && !eth_dev->data->dev_started)
1844 		memset(&new, 0, sizeof(new));
1845 
1846 out:
1847 	/* Timed out or success */
1848 	if (new.link_status != eth_dev->data->dev_link.link_status ||
1849 	    new.link_speed != eth_dev->data->dev_link.link_speed) {
1850 		rte_eth_linkstatus_set(eth_dev, &new);
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 & RTE_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_ETH_RETA_GROUP_SIZE;
2036 		sft = i % RTE_ETH_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_ETH_RETA_GROUP_SIZE;
2090 		sft = i % RTE_ETH_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 & RTE_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 					    RTE_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 |= RTE_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 |= RTE_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 |= RTE_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 |= RTE_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 |= RTE_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 |= RTE_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_ETH_FC_NONE;
2255 		break;
2256 	case HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_TX:
2257 		fc_conf->mode = RTE_ETH_FC_TX_PAUSE;
2258 		break;
2259 	case HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_RX:
2260 		fc_conf->mode = RTE_ETH_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_ETH_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_ETH_FC_NONE:
2288 		bp->link_info->auto_pause = 0;
2289 		bp->link_info->force_pause = 0;
2290 		break;
2291 	case RTE_ETH_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_ETH_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_ETH_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_ETH_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 		break;
2358 	case RTE_ETH_TUNNEL_TYPE_GENEVE:
2359 		if (bp->geneve_port_cnt) {
2360 			PMD_DRV_LOG(ERR, "Tunnel Port %d already programmed\n",
2361 				udp_tunnel->udp_port);
2362 			if (bp->geneve_port != udp_tunnel->udp_port) {
2363 				PMD_DRV_LOG(ERR, "Only one port allowed\n");
2364 				return -ENOSPC;
2365 			}
2366 			bp->geneve_port_cnt++;
2367 			return 0;
2368 		}
2369 		tunnel_type =
2370 			HWRM_TUNNEL_DST_PORT_ALLOC_INPUT_TUNNEL_TYPE_GENEVE;
2371 		break;
2372 	default:
2373 		PMD_DRV_LOG(ERR, "Tunnel type is not supported\n");
2374 		return -ENOTSUP;
2375 	}
2376 	rc = bnxt_hwrm_tunnel_dst_port_alloc(bp, udp_tunnel->udp_port,
2377 					     tunnel_type);
2378 
2379 	if (rc != 0)
2380 		return rc;
2381 
2382 	if (tunnel_type ==
2383 	    HWRM_TUNNEL_DST_PORT_ALLOC_INPUT_TUNNEL_TYPE_VXLAN)
2384 		bp->vxlan_port_cnt++;
2385 
2386 	if (tunnel_type ==
2387 	    HWRM_TUNNEL_DST_PORT_ALLOC_INPUT_TUNNEL_TYPE_GENEVE)
2388 		bp->geneve_port_cnt++;
2389 
2390 	return rc;
2391 }
2392 
2393 static int
2394 bnxt_udp_tunnel_port_del_op(struct rte_eth_dev *eth_dev,
2395 			 struct rte_eth_udp_tunnel *udp_tunnel)
2396 {
2397 	struct bnxt *bp = eth_dev->data->dev_private;
2398 	uint16_t tunnel_type = 0;
2399 	uint16_t port = 0;
2400 	int rc = 0;
2401 
2402 	rc = is_bnxt_in_error(bp);
2403 	if (rc)
2404 		return rc;
2405 
2406 	switch (udp_tunnel->prot_type) {
2407 	case RTE_ETH_TUNNEL_TYPE_VXLAN:
2408 		if (!bp->vxlan_port_cnt) {
2409 			PMD_DRV_LOG(ERR, "No Tunnel port configured yet\n");
2410 			return -EINVAL;
2411 		}
2412 		if (bp->vxlan_port != udp_tunnel->udp_port) {
2413 			PMD_DRV_LOG(ERR, "Req Port: %d. Configured port: %d\n",
2414 				udp_tunnel->udp_port, bp->vxlan_port);
2415 			return -EINVAL;
2416 		}
2417 		if (--bp->vxlan_port_cnt)
2418 			return 0;
2419 
2420 		tunnel_type =
2421 			HWRM_TUNNEL_DST_PORT_FREE_INPUT_TUNNEL_TYPE_VXLAN;
2422 		port = bp->vxlan_fw_dst_port_id;
2423 		break;
2424 	case RTE_ETH_TUNNEL_TYPE_GENEVE:
2425 		if (!bp->geneve_port_cnt) {
2426 			PMD_DRV_LOG(ERR, "No Tunnel port configured yet\n");
2427 			return -EINVAL;
2428 		}
2429 		if (bp->geneve_port != udp_tunnel->udp_port) {
2430 			PMD_DRV_LOG(ERR, "Req Port: %d. Configured port: %d\n",
2431 				udp_tunnel->udp_port, bp->geneve_port);
2432 			return -EINVAL;
2433 		}
2434 		if (--bp->geneve_port_cnt)
2435 			return 0;
2436 
2437 		tunnel_type =
2438 			HWRM_TUNNEL_DST_PORT_FREE_INPUT_TUNNEL_TYPE_GENEVE;
2439 		port = bp->geneve_fw_dst_port_id;
2440 		break;
2441 	default:
2442 		PMD_DRV_LOG(ERR, "Tunnel type is not supported\n");
2443 		return -ENOTSUP;
2444 	}
2445 
2446 	rc = bnxt_hwrm_tunnel_dst_port_free(bp, port, tunnel_type);
2447 	return rc;
2448 }
2449 
2450 static int bnxt_del_vlan_filter(struct bnxt *bp, uint16_t vlan_id)
2451 {
2452 	struct bnxt_filter_info *filter;
2453 	struct bnxt_vnic_info *vnic;
2454 	int rc = 0;
2455 	uint32_t chk = HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN;
2456 
2457 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
2458 	filter = STAILQ_FIRST(&vnic->filter);
2459 	while (filter) {
2460 		/* Search for this matching MAC+VLAN filter */
2461 		if (bnxt_vlan_filter_exists(bp, filter, chk, vlan_id)) {
2462 			/* Delete the filter */
2463 			rc = bnxt_hwrm_clear_l2_filter(bp, filter);
2464 			if (rc)
2465 				return rc;
2466 			STAILQ_REMOVE(&vnic->filter, filter,
2467 				      bnxt_filter_info, next);
2468 			bnxt_free_filter(bp, filter);
2469 			PMD_DRV_LOG(INFO,
2470 				    "Deleted vlan filter for %d\n",
2471 				    vlan_id);
2472 			return 0;
2473 		}
2474 		filter = STAILQ_NEXT(filter, next);
2475 	}
2476 	return -ENOENT;
2477 }
2478 
2479 static int bnxt_add_vlan_filter(struct bnxt *bp, uint16_t vlan_id)
2480 {
2481 	struct bnxt_filter_info *filter;
2482 	struct bnxt_vnic_info *vnic;
2483 	int rc = 0;
2484 	uint32_t en = HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN |
2485 		HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN_MASK;
2486 	uint32_t chk = HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN;
2487 
2488 	/* Implementation notes on the use of VNIC in this command:
2489 	 *
2490 	 * By default, these filters belong to default vnic for the function.
2491 	 * Once these filters are set up, only destination VNIC can be modified.
2492 	 * If the destination VNIC is not specified in this command,
2493 	 * then the HWRM shall only create an l2 context id.
2494 	 */
2495 
2496 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
2497 	filter = STAILQ_FIRST(&vnic->filter);
2498 	/* Check if the VLAN has already been added */
2499 	while (filter) {
2500 		if (bnxt_vlan_filter_exists(bp, filter, chk, vlan_id))
2501 			return -EEXIST;
2502 
2503 		filter = STAILQ_NEXT(filter, next);
2504 	}
2505 
2506 	/* No match found. Alloc a fresh filter and issue the L2_FILTER_ALLOC
2507 	 * command to create MAC+VLAN filter with the right flags, enables set.
2508 	 */
2509 	filter = bnxt_alloc_filter(bp);
2510 	if (!filter) {
2511 		PMD_DRV_LOG(ERR,
2512 			    "MAC/VLAN filter alloc failed\n");
2513 		return -ENOMEM;
2514 	}
2515 	/* MAC + VLAN ID filter */
2516 	/* If l2_ivlan == 0 and l2_ivlan_mask != 0, only
2517 	 * untagged packets are received
2518 	 *
2519 	 * If l2_ivlan != 0 and l2_ivlan_mask != 0, untagged
2520 	 * packets and only the programmed vlan's packets are received
2521 	 */
2522 	filter->l2_ivlan = vlan_id;
2523 	filter->l2_ivlan_mask = 0x0FFF;
2524 	filter->enables |= en;
2525 	filter->flags |= HWRM_CFA_L2_FILTER_ALLOC_INPUT_FLAGS_OUTERMOST;
2526 
2527 	rc = bnxt_hwrm_set_l2_filter(bp, vnic->fw_vnic_id, filter);
2528 	if (rc) {
2529 		/* Free the newly allocated filter as we were
2530 		 * not able to create the filter in hardware.
2531 		 */
2532 		bnxt_free_filter(bp, filter);
2533 		return rc;
2534 	}
2535 
2536 	filter->mac_index = 0;
2537 	/* Add this new filter to the list */
2538 	if (vlan_id == 0)
2539 		STAILQ_INSERT_HEAD(&vnic->filter, filter, next);
2540 	else
2541 		STAILQ_INSERT_TAIL(&vnic->filter, filter, next);
2542 
2543 	PMD_DRV_LOG(INFO,
2544 		    "Added Vlan filter for %d\n", vlan_id);
2545 	return rc;
2546 }
2547 
2548 static int bnxt_vlan_filter_set_op(struct rte_eth_dev *eth_dev,
2549 		uint16_t vlan_id, int on)
2550 {
2551 	struct bnxt *bp = eth_dev->data->dev_private;
2552 	int rc;
2553 
2554 	rc = is_bnxt_in_error(bp);
2555 	if (rc)
2556 		return rc;
2557 
2558 	if (!eth_dev->data->dev_started) {
2559 		PMD_DRV_LOG(ERR, "port must be started before setting vlan\n");
2560 		return -EINVAL;
2561 	}
2562 
2563 	/* These operations apply to ALL existing MAC/VLAN filters */
2564 	if (on)
2565 		return bnxt_add_vlan_filter(bp, vlan_id);
2566 	else
2567 		return bnxt_del_vlan_filter(bp, vlan_id);
2568 }
2569 
2570 static int bnxt_del_dflt_mac_filter(struct bnxt *bp,
2571 				    struct bnxt_vnic_info *vnic)
2572 {
2573 	struct bnxt_filter_info *filter;
2574 	int rc;
2575 
2576 	filter = STAILQ_FIRST(&vnic->filter);
2577 	while (filter) {
2578 		if (filter->mac_index == 0 &&
2579 		    !memcmp(filter->l2_addr, bp->mac_addr,
2580 			    RTE_ETHER_ADDR_LEN)) {
2581 			rc = bnxt_hwrm_clear_l2_filter(bp, filter);
2582 			if (!rc) {
2583 				STAILQ_REMOVE(&vnic->filter, filter,
2584 					      bnxt_filter_info, next);
2585 				bnxt_free_filter(bp, filter);
2586 			}
2587 			return rc;
2588 		}
2589 		filter = STAILQ_NEXT(filter, next);
2590 	}
2591 	return 0;
2592 }
2593 
2594 static int
2595 bnxt_config_vlan_hw_filter(struct bnxt *bp, uint64_t rx_offloads)
2596 {
2597 	struct bnxt_vnic_info *vnic;
2598 	unsigned int i;
2599 	int rc;
2600 
2601 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
2602 	if (!(rx_offloads & RTE_ETH_RX_OFFLOAD_VLAN_FILTER)) {
2603 		/* Remove any VLAN filters programmed */
2604 		for (i = 0; i < RTE_ETHER_MAX_VLAN_ID; i++)
2605 			bnxt_del_vlan_filter(bp, i);
2606 
2607 		rc = bnxt_add_mac_filter(bp, vnic, NULL, 0, 0);
2608 		if (rc)
2609 			return rc;
2610 	} else {
2611 		/* Default filter will allow packets that match the
2612 		 * dest mac. So, it has to be deleted, otherwise, we
2613 		 * will endup receiving vlan packets for which the
2614 		 * filter is not programmed, when hw-vlan-filter
2615 		 * configuration is ON
2616 		 */
2617 		bnxt_del_dflt_mac_filter(bp, vnic);
2618 		/* This filter will allow only untagged packets */
2619 		bnxt_add_vlan_filter(bp, 0);
2620 	}
2621 	PMD_DRV_LOG(DEBUG, "VLAN Filtering: %d\n",
2622 		    !!(rx_offloads & RTE_ETH_RX_OFFLOAD_VLAN_FILTER));
2623 
2624 	return 0;
2625 }
2626 
2627 static int bnxt_free_one_vnic(struct bnxt *bp, uint16_t vnic_id)
2628 {
2629 	struct bnxt_vnic_info *vnic = &bp->vnic_info[vnic_id];
2630 	unsigned int i;
2631 	int rc;
2632 
2633 	/* Destroy vnic filters and vnic */
2634 	if (bp->eth_dev->data->dev_conf.rxmode.offloads &
2635 	    RTE_ETH_RX_OFFLOAD_VLAN_FILTER) {
2636 		for (i = 0; i < RTE_ETHER_MAX_VLAN_ID; i++)
2637 			bnxt_del_vlan_filter(bp, i);
2638 	}
2639 	bnxt_del_dflt_mac_filter(bp, vnic);
2640 
2641 	rc = bnxt_hwrm_vnic_ctx_free(bp, vnic);
2642 	if (rc)
2643 		return rc;
2644 
2645 	rc = bnxt_hwrm_vnic_free(bp, vnic);
2646 	if (rc)
2647 		return rc;
2648 
2649 	rte_free(vnic->fw_grp_ids);
2650 	vnic->fw_grp_ids = NULL;
2651 
2652 	vnic->rx_queue_cnt = 0;
2653 
2654 	return 0;
2655 }
2656 
2657 static int
2658 bnxt_config_vlan_hw_stripping(struct bnxt *bp, uint64_t rx_offloads)
2659 {
2660 	struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
2661 	int rc;
2662 
2663 	/* Destroy, recreate and reconfigure the default vnic */
2664 	rc = bnxt_free_one_vnic(bp, 0);
2665 	if (rc)
2666 		return rc;
2667 
2668 	/* default vnic 0 */
2669 	rc = bnxt_setup_one_vnic(bp, 0);
2670 	if (rc)
2671 		return rc;
2672 
2673 	if (bp->eth_dev->data->dev_conf.rxmode.offloads &
2674 	    RTE_ETH_RX_OFFLOAD_VLAN_FILTER) {
2675 		rc = bnxt_add_vlan_filter(bp, 0);
2676 		if (rc)
2677 			return rc;
2678 		rc = bnxt_restore_vlan_filters(bp);
2679 		if (rc)
2680 			return rc;
2681 	} else {
2682 		rc = bnxt_add_mac_filter(bp, vnic, NULL, 0, 0);
2683 		if (rc)
2684 			return rc;
2685 	}
2686 
2687 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
2688 	if (rc)
2689 		return rc;
2690 
2691 	PMD_DRV_LOG(DEBUG, "VLAN Strip Offload: %d\n",
2692 		    !!(rx_offloads & RTE_ETH_RX_OFFLOAD_VLAN_STRIP));
2693 
2694 	return rc;
2695 }
2696 
2697 static int
2698 bnxt_vlan_offload_set_op(struct rte_eth_dev *dev, int mask)
2699 {
2700 	uint64_t rx_offloads = dev->data->dev_conf.rxmode.offloads;
2701 	struct bnxt *bp = dev->data->dev_private;
2702 	int rc;
2703 
2704 	rc = is_bnxt_in_error(bp);
2705 	if (rc)
2706 		return rc;
2707 
2708 	/* Filter settings will get applied when port is started */
2709 	if (!dev->data->dev_started)
2710 		return 0;
2711 
2712 	if (mask & RTE_ETH_VLAN_FILTER_MASK) {
2713 		/* Enable or disable VLAN filtering */
2714 		rc = bnxt_config_vlan_hw_filter(bp, rx_offloads);
2715 		if (rc)
2716 			return rc;
2717 	}
2718 
2719 	if (mask & RTE_ETH_VLAN_STRIP_MASK) {
2720 		/* Enable or disable VLAN stripping */
2721 		rc = bnxt_config_vlan_hw_stripping(bp, rx_offloads);
2722 		if (rc)
2723 			return rc;
2724 	}
2725 
2726 	if (mask & RTE_ETH_VLAN_EXTEND_MASK) {
2727 		if (rx_offloads & RTE_ETH_RX_OFFLOAD_VLAN_EXTEND)
2728 			PMD_DRV_LOG(DEBUG, "Extend VLAN supported\n");
2729 		else
2730 			PMD_DRV_LOG(INFO, "Extend VLAN unsupported\n");
2731 	}
2732 
2733 	return 0;
2734 }
2735 
2736 static int
2737 bnxt_vlan_tpid_set_op(struct rte_eth_dev *dev, enum rte_vlan_type vlan_type,
2738 		      uint16_t tpid)
2739 {
2740 	struct bnxt *bp = dev->data->dev_private;
2741 	int qinq = dev->data->dev_conf.rxmode.offloads &
2742 		   RTE_ETH_RX_OFFLOAD_VLAN_EXTEND;
2743 
2744 	if (vlan_type != RTE_ETH_VLAN_TYPE_INNER &&
2745 	    vlan_type != RTE_ETH_VLAN_TYPE_OUTER) {
2746 		PMD_DRV_LOG(ERR,
2747 			    "Unsupported vlan type.");
2748 		return -EINVAL;
2749 	}
2750 	if (!qinq) {
2751 		PMD_DRV_LOG(ERR,
2752 			    "QinQ not enabled. Needs to be ON as we can "
2753 			    "accelerate only outer vlan\n");
2754 		return -EINVAL;
2755 	}
2756 
2757 	if (vlan_type == RTE_ETH_VLAN_TYPE_OUTER) {
2758 		switch (tpid) {
2759 		case RTE_ETHER_TYPE_QINQ:
2760 			bp->outer_tpid_bd =
2761 				TX_BD_LONG_CFA_META_VLAN_TPID_TPID88A8;
2762 				break;
2763 		case RTE_ETHER_TYPE_VLAN:
2764 			bp->outer_tpid_bd =
2765 				TX_BD_LONG_CFA_META_VLAN_TPID_TPID8100;
2766 				break;
2767 		case RTE_ETHER_TYPE_QINQ1:
2768 			bp->outer_tpid_bd =
2769 				TX_BD_LONG_CFA_META_VLAN_TPID_TPID9100;
2770 				break;
2771 		case RTE_ETHER_TYPE_QINQ2:
2772 			bp->outer_tpid_bd =
2773 				TX_BD_LONG_CFA_META_VLAN_TPID_TPID9200;
2774 				break;
2775 		case RTE_ETHER_TYPE_QINQ3:
2776 			bp->outer_tpid_bd =
2777 				 TX_BD_LONG_CFA_META_VLAN_TPID_TPID9300;
2778 				break;
2779 		default:
2780 			PMD_DRV_LOG(ERR, "Invalid TPID: %x\n", tpid);
2781 			return -EINVAL;
2782 		}
2783 		bp->outer_tpid_bd |= tpid;
2784 		PMD_DRV_LOG(INFO, "outer_tpid_bd = %x\n", bp->outer_tpid_bd);
2785 	} else if (vlan_type == RTE_ETH_VLAN_TYPE_INNER) {
2786 		PMD_DRV_LOG(ERR,
2787 			    "Can accelerate only outer vlan in QinQ\n");
2788 		return -EINVAL;
2789 	}
2790 
2791 	return 0;
2792 }
2793 
2794 static int
2795 bnxt_set_default_mac_addr_op(struct rte_eth_dev *dev,
2796 			     struct rte_ether_addr *addr)
2797 {
2798 	struct bnxt *bp = dev->data->dev_private;
2799 	/* Default Filter is tied to VNIC 0 */
2800 	struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
2801 	int rc;
2802 
2803 	rc = is_bnxt_in_error(bp);
2804 	if (rc)
2805 		return rc;
2806 
2807 	if (BNXT_VF(bp) && !BNXT_VF_IS_TRUSTED(bp))
2808 		return -EPERM;
2809 
2810 	if (rte_is_zero_ether_addr(addr))
2811 		return -EINVAL;
2812 
2813 	/* Filter settings will get applied when port is started */
2814 	if (!dev->data->dev_started)
2815 		return 0;
2816 
2817 	/* Check if the requested MAC is already added */
2818 	if (memcmp(addr, bp->mac_addr, RTE_ETHER_ADDR_LEN) == 0)
2819 		return 0;
2820 
2821 	/* Destroy filter and re-create it */
2822 	bnxt_del_dflt_mac_filter(bp, vnic);
2823 
2824 	memcpy(bp->mac_addr, addr, RTE_ETHER_ADDR_LEN);
2825 	if (dev->data->dev_conf.rxmode.offloads & RTE_ETH_RX_OFFLOAD_VLAN_FILTER) {
2826 		/* This filter will allow only untagged packets */
2827 		rc = bnxt_add_vlan_filter(bp, 0);
2828 	} else {
2829 		rc = bnxt_add_mac_filter(bp, vnic, addr, 0, 0);
2830 	}
2831 
2832 	PMD_DRV_LOG(DEBUG, "Set MAC addr\n");
2833 	return rc;
2834 }
2835 
2836 static int
2837 bnxt_dev_set_mc_addr_list_op(struct rte_eth_dev *eth_dev,
2838 			  struct rte_ether_addr *mc_addr_set,
2839 			  uint32_t nb_mc_addr)
2840 {
2841 	struct bnxt *bp = eth_dev->data->dev_private;
2842 	char *mc_addr_list = (char *)mc_addr_set;
2843 	struct bnxt_vnic_info *vnic;
2844 	uint32_t off = 0, i = 0;
2845 	int rc;
2846 
2847 	rc = is_bnxt_in_error(bp);
2848 	if (rc)
2849 		return rc;
2850 
2851 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
2852 
2853 	if (nb_mc_addr > BNXT_MAX_MC_ADDRS) {
2854 		vnic->flags |= BNXT_VNIC_INFO_ALLMULTI;
2855 		goto allmulti;
2856 	}
2857 
2858 	/* TODO Check for Duplicate mcast addresses */
2859 	vnic->flags &= ~BNXT_VNIC_INFO_ALLMULTI;
2860 	for (i = 0; i < nb_mc_addr; i++) {
2861 		memcpy(vnic->mc_list + off, &mc_addr_list[i],
2862 			RTE_ETHER_ADDR_LEN);
2863 		off += RTE_ETHER_ADDR_LEN;
2864 	}
2865 
2866 	vnic->mc_addr_cnt = i;
2867 	if (vnic->mc_addr_cnt)
2868 		vnic->flags |= BNXT_VNIC_INFO_MCAST;
2869 	else
2870 		vnic->flags &= ~BNXT_VNIC_INFO_MCAST;
2871 
2872 allmulti:
2873 	return bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
2874 }
2875 
2876 static int
2877 bnxt_fw_version_get(struct rte_eth_dev *dev, char *fw_version, size_t fw_size)
2878 {
2879 	struct bnxt *bp = dev->data->dev_private;
2880 	uint8_t fw_major = (bp->fw_ver >> 24) & 0xff;
2881 	uint8_t fw_minor = (bp->fw_ver >> 16) & 0xff;
2882 	uint8_t fw_updt = (bp->fw_ver >> 8) & 0xff;
2883 	uint8_t fw_rsvd = bp->fw_ver & 0xff;
2884 	int ret;
2885 
2886 	ret = snprintf(fw_version, fw_size, "%d.%d.%d.%d",
2887 			fw_major, fw_minor, fw_updt, fw_rsvd);
2888 	if (ret < 0)
2889 		return -EINVAL;
2890 
2891 	ret += 1; /* add the size of '\0' */
2892 	if (fw_size < (size_t)ret)
2893 		return ret;
2894 	else
2895 		return 0;
2896 }
2897 
2898 static void
2899 bnxt_rxq_info_get_op(struct rte_eth_dev *dev, uint16_t queue_id,
2900 	struct rte_eth_rxq_info *qinfo)
2901 {
2902 	struct bnxt *bp = dev->data->dev_private;
2903 	struct bnxt_rx_queue *rxq;
2904 
2905 	if (is_bnxt_in_error(bp))
2906 		return;
2907 
2908 	rxq = dev->data->rx_queues[queue_id];
2909 
2910 	qinfo->mp = rxq->mb_pool;
2911 	qinfo->scattered_rx = dev->data->scattered_rx;
2912 	qinfo->nb_desc = rxq->nb_rx_desc;
2913 
2914 	qinfo->conf.rx_free_thresh = rxq->rx_free_thresh;
2915 	qinfo->conf.rx_drop_en = rxq->drop_en;
2916 	qinfo->conf.rx_deferred_start = rxq->rx_deferred_start;
2917 	qinfo->conf.offloads = dev->data->dev_conf.rxmode.offloads;
2918 }
2919 
2920 static void
2921 bnxt_txq_info_get_op(struct rte_eth_dev *dev, uint16_t queue_id,
2922 	struct rte_eth_txq_info *qinfo)
2923 {
2924 	struct bnxt *bp = dev->data->dev_private;
2925 	struct bnxt_tx_queue *txq;
2926 
2927 	if (is_bnxt_in_error(bp))
2928 		return;
2929 
2930 	txq = dev->data->tx_queues[queue_id];
2931 
2932 	qinfo->nb_desc = txq->nb_tx_desc;
2933 
2934 	qinfo->conf.tx_thresh.pthresh = txq->pthresh;
2935 	qinfo->conf.tx_thresh.hthresh = txq->hthresh;
2936 	qinfo->conf.tx_thresh.wthresh = txq->wthresh;
2937 
2938 	qinfo->conf.tx_free_thresh = txq->tx_free_thresh;
2939 	qinfo->conf.tx_rs_thresh = 0;
2940 	qinfo->conf.tx_deferred_start = txq->tx_deferred_start;
2941 	qinfo->conf.offloads = txq->offloads;
2942 }
2943 
2944 static const struct {
2945 	eth_rx_burst_t pkt_burst;
2946 	const char *info;
2947 } bnxt_rx_burst_info[] = {
2948 	{bnxt_recv_pkts,		"Scalar"},
2949 #if defined(RTE_ARCH_X86)
2950 	{bnxt_recv_pkts_vec,		"Vector SSE"},
2951 #endif
2952 #if defined(RTE_ARCH_X86) && defined(CC_AVX2_SUPPORT)
2953 	{bnxt_recv_pkts_vec_avx2,	"Vector AVX2"},
2954 #endif
2955 #if defined(RTE_ARCH_ARM64)
2956 	{bnxt_recv_pkts_vec,		"Vector Neon"},
2957 #endif
2958 };
2959 
2960 static int
2961 bnxt_rx_burst_mode_get(struct rte_eth_dev *dev, __rte_unused uint16_t queue_id,
2962 		       struct rte_eth_burst_mode *mode)
2963 {
2964 	eth_rx_burst_t pkt_burst = dev->rx_pkt_burst;
2965 	size_t i;
2966 
2967 	for (i = 0; i < RTE_DIM(bnxt_rx_burst_info); i++) {
2968 		if (pkt_burst == bnxt_rx_burst_info[i].pkt_burst) {
2969 			snprintf(mode->info, sizeof(mode->info), "%s",
2970 				 bnxt_rx_burst_info[i].info);
2971 			return 0;
2972 		}
2973 	}
2974 
2975 	return -EINVAL;
2976 }
2977 
2978 static const struct {
2979 	eth_tx_burst_t pkt_burst;
2980 	const char *info;
2981 } bnxt_tx_burst_info[] = {
2982 	{bnxt_xmit_pkts,		"Scalar"},
2983 #if defined(RTE_ARCH_X86)
2984 	{bnxt_xmit_pkts_vec,		"Vector SSE"},
2985 #endif
2986 #if defined(RTE_ARCH_X86) && defined(CC_AVX2_SUPPORT)
2987 	{bnxt_xmit_pkts_vec_avx2,	"Vector AVX2"},
2988 #endif
2989 #if defined(RTE_ARCH_ARM64)
2990 	{bnxt_xmit_pkts_vec,		"Vector Neon"},
2991 #endif
2992 };
2993 
2994 static int
2995 bnxt_tx_burst_mode_get(struct rte_eth_dev *dev, __rte_unused uint16_t queue_id,
2996 		       struct rte_eth_burst_mode *mode)
2997 {
2998 	eth_tx_burst_t pkt_burst = dev->tx_pkt_burst;
2999 	size_t i;
3000 
3001 	for (i = 0; i < RTE_DIM(bnxt_tx_burst_info); i++) {
3002 		if (pkt_burst == bnxt_tx_burst_info[i].pkt_burst) {
3003 			snprintf(mode->info, sizeof(mode->info), "%s",
3004 				 bnxt_tx_burst_info[i].info);
3005 			return 0;
3006 		}
3007 	}
3008 
3009 	return -EINVAL;
3010 }
3011 
3012 int bnxt_mtu_set_op(struct rte_eth_dev *eth_dev, uint16_t new_mtu)
3013 {
3014 	uint32_t overhead = BNXT_MAX_PKT_LEN - BNXT_MAX_MTU;
3015 	struct bnxt *bp = eth_dev->data->dev_private;
3016 	uint32_t new_pkt_size;
3017 	uint32_t rc;
3018 	uint32_t i;
3019 
3020 	rc = is_bnxt_in_error(bp);
3021 	if (rc)
3022 		return rc;
3023 
3024 	/* Exit if receive queues are not configured yet */
3025 	if (!eth_dev->data->nb_rx_queues)
3026 		return rc;
3027 
3028 	new_pkt_size = new_mtu + overhead;
3029 
3030 	/*
3031 	 * Disallow any MTU change that would require scattered receive support
3032 	 * if it is not already enabled.
3033 	 */
3034 	if (eth_dev->data->dev_started &&
3035 	    !eth_dev->data->scattered_rx &&
3036 	    (new_pkt_size >
3037 	     eth_dev->data->min_rx_buf_size - RTE_PKTMBUF_HEADROOM)) {
3038 		PMD_DRV_LOG(ERR,
3039 			    "MTU change would require scattered rx support. ");
3040 		PMD_DRV_LOG(ERR, "Stop port before changing MTU.\n");
3041 		return -EINVAL;
3042 	}
3043 
3044 	if (new_mtu > RTE_ETHER_MTU)
3045 		bp->flags |= BNXT_FLAG_JUMBO;
3046 	else
3047 		bp->flags &= ~BNXT_FLAG_JUMBO;
3048 
3049 	/* Is there a change in mtu setting? */
3050 	if (eth_dev->data->mtu == new_mtu)
3051 		return rc;
3052 
3053 	for (i = 0; i < bp->nr_vnics; i++) {
3054 		struct bnxt_vnic_info *vnic = &bp->vnic_info[i];
3055 		uint16_t size = 0;
3056 
3057 		vnic->mru = BNXT_VNIC_MRU(new_mtu);
3058 		rc = bnxt_hwrm_vnic_cfg(bp, vnic);
3059 		if (rc)
3060 			break;
3061 
3062 		size = rte_pktmbuf_data_room_size(bp->rx_queues[0]->mb_pool);
3063 		size -= RTE_PKTMBUF_HEADROOM;
3064 
3065 		if (size < new_mtu) {
3066 			rc = bnxt_hwrm_vnic_plcmode_cfg(bp, vnic);
3067 			if (rc)
3068 				return rc;
3069 		}
3070 	}
3071 
3072 	if (bnxt_hwrm_config_host_mtu(bp))
3073 		PMD_DRV_LOG(WARNING, "Failed to configure host MTU\n");
3074 
3075 	PMD_DRV_LOG(INFO, "New MTU is %d\n", new_mtu);
3076 
3077 	return rc;
3078 }
3079 
3080 static int
3081 bnxt_vlan_pvid_set_op(struct rte_eth_dev *dev, uint16_t pvid, int on)
3082 {
3083 	struct bnxt *bp = dev->data->dev_private;
3084 	uint16_t vlan = bp->vlan;
3085 	int rc;
3086 
3087 	rc = is_bnxt_in_error(bp);
3088 	if (rc)
3089 		return rc;
3090 
3091 	if (!BNXT_SINGLE_PF(bp)) {
3092 		PMD_DRV_LOG(ERR, "PVID cannot be modified on VF or on shared PF\n");
3093 		return -ENOTSUP;
3094 	}
3095 	bp->vlan = on ? pvid : 0;
3096 
3097 	rc = bnxt_hwrm_set_default_vlan(bp, 0, 0);
3098 	if (rc)
3099 		bp->vlan = vlan;
3100 	return rc;
3101 }
3102 
3103 static int
3104 bnxt_dev_led_on_op(struct rte_eth_dev *dev)
3105 {
3106 	struct bnxt *bp = dev->data->dev_private;
3107 	int rc;
3108 
3109 	rc = is_bnxt_in_error(bp);
3110 	if (rc)
3111 		return rc;
3112 
3113 	return bnxt_hwrm_port_led_cfg(bp, true);
3114 }
3115 
3116 static int
3117 bnxt_dev_led_off_op(struct rte_eth_dev *dev)
3118 {
3119 	struct bnxt *bp = dev->data->dev_private;
3120 	int rc;
3121 
3122 	rc = is_bnxt_in_error(bp);
3123 	if (rc)
3124 		return rc;
3125 
3126 	return bnxt_hwrm_port_led_cfg(bp, false);
3127 }
3128 
3129 static uint32_t
3130 bnxt_rx_queue_count_op(void *rx_queue)
3131 {
3132 	struct bnxt *bp;
3133 	struct bnxt_cp_ring_info *cpr;
3134 	uint32_t desc = 0, raw_cons, cp_ring_size;
3135 	struct bnxt_rx_queue *rxq;
3136 	struct rx_pkt_cmpl *rxcmp;
3137 	int rc;
3138 
3139 	rxq = rx_queue;
3140 	bp = rxq->bp;
3141 
3142 	rc = is_bnxt_in_error(bp);
3143 	if (rc)
3144 		return rc;
3145 
3146 	cpr = rxq->cp_ring;
3147 	raw_cons = cpr->cp_raw_cons;
3148 	cp_ring_size = cpr->cp_ring_struct->ring_size;
3149 
3150 	while (1) {
3151 		uint32_t agg_cnt, cons, cmpl_type;
3152 
3153 		cons = RING_CMP(cpr->cp_ring_struct, raw_cons);
3154 		rxcmp = (struct rx_pkt_cmpl *)&cpr->cp_desc_ring[cons];
3155 
3156 		if (!bnxt_cpr_cmp_valid(rxcmp, raw_cons, cp_ring_size))
3157 			break;
3158 
3159 		cmpl_type = CMP_TYPE(rxcmp);
3160 
3161 		switch (cmpl_type) {
3162 		case CMPL_BASE_TYPE_RX_L2:
3163 		case CMPL_BASE_TYPE_RX_L2_V2:
3164 			agg_cnt = BNXT_RX_L2_AGG_BUFS(rxcmp);
3165 			raw_cons = raw_cons + CMP_LEN(cmpl_type) + agg_cnt;
3166 			desc++;
3167 			break;
3168 
3169 		case CMPL_BASE_TYPE_RX_TPA_END:
3170 			if (BNXT_CHIP_P5(rxq->bp)) {
3171 				struct rx_tpa_v2_end_cmpl_hi *p5_tpa_end;
3172 
3173 				p5_tpa_end = (void *)rxcmp;
3174 				agg_cnt = BNXT_TPA_END_AGG_BUFS_TH(p5_tpa_end);
3175 			} else {
3176 				struct rx_tpa_end_cmpl *tpa_end;
3177 
3178 				tpa_end = (void *)rxcmp;
3179 				agg_cnt = BNXT_TPA_END_AGG_BUFS(tpa_end);
3180 			}
3181 
3182 			raw_cons = raw_cons + CMP_LEN(cmpl_type) + agg_cnt;
3183 			desc++;
3184 			break;
3185 
3186 		default:
3187 			raw_cons += CMP_LEN(cmpl_type);
3188 		}
3189 	}
3190 
3191 	return desc;
3192 }
3193 
3194 static int
3195 bnxt_rx_descriptor_status_op(void *rx_queue, uint16_t offset)
3196 {
3197 	struct bnxt_rx_queue *rxq = rx_queue;
3198 	struct bnxt_cp_ring_info *cpr;
3199 	struct bnxt_rx_ring_info *rxr;
3200 	uint32_t desc, raw_cons, cp_ring_size;
3201 	struct bnxt *bp = rxq->bp;
3202 	struct rx_pkt_cmpl *rxcmp;
3203 	int rc;
3204 
3205 	rc = is_bnxt_in_error(bp);
3206 	if (rc)
3207 		return rc;
3208 
3209 	if (offset >= rxq->nb_rx_desc)
3210 		return -EINVAL;
3211 
3212 	rxr = rxq->rx_ring;
3213 	cpr = rxq->cp_ring;
3214 	cp_ring_size = cpr->cp_ring_struct->ring_size;
3215 
3216 	/*
3217 	 * For the vector receive case, the completion at the requested
3218 	 * offset can be indexed directly.
3219 	 */
3220 #if defined(RTE_ARCH_X86) || defined(RTE_ARCH_ARM64)
3221 	if (bp->flags & BNXT_FLAG_RX_VECTOR_PKT_MODE) {
3222 		struct rx_pkt_cmpl *rxcmp;
3223 		uint32_t cons;
3224 
3225 		/* Check status of completion descriptor. */
3226 		raw_cons = cpr->cp_raw_cons +
3227 			   offset * CMP_LEN(CMPL_BASE_TYPE_RX_L2);
3228 		cons = RING_CMP(cpr->cp_ring_struct, raw_cons);
3229 		rxcmp = (struct rx_pkt_cmpl *)&cpr->cp_desc_ring[cons];
3230 
3231 		if (bnxt_cpr_cmp_valid(rxcmp, raw_cons, cp_ring_size))
3232 			return RTE_ETH_RX_DESC_DONE;
3233 
3234 		/* Check whether rx desc has an mbuf attached. */
3235 		cons = RING_CMP(rxr->rx_ring_struct, raw_cons / 2);
3236 		if (cons >= rxq->rxrearm_start &&
3237 		    cons < rxq->rxrearm_start + rxq->rxrearm_nb) {
3238 			return RTE_ETH_RX_DESC_UNAVAIL;
3239 		}
3240 
3241 		return RTE_ETH_RX_DESC_AVAIL;
3242 	}
3243 #endif
3244 
3245 	/*
3246 	 * For the non-vector receive case, scan the completion ring to
3247 	 * locate the completion descriptor for the requested offset.
3248 	 */
3249 	raw_cons = cpr->cp_raw_cons;
3250 	desc = 0;
3251 	while (1) {
3252 		uint32_t agg_cnt, cons, cmpl_type;
3253 
3254 		cons = RING_CMP(cpr->cp_ring_struct, raw_cons);
3255 		rxcmp = (struct rx_pkt_cmpl *)&cpr->cp_desc_ring[cons];
3256 
3257 		if (!bnxt_cpr_cmp_valid(rxcmp, raw_cons, cp_ring_size))
3258 			break;
3259 
3260 		cmpl_type = CMP_TYPE(rxcmp);
3261 
3262 		switch (cmpl_type) {
3263 		case CMPL_BASE_TYPE_RX_L2:
3264 		case CMPL_BASE_TYPE_RX_L2_V2:
3265 			if (desc == offset) {
3266 				cons = rxcmp->opaque;
3267 				if (rxr->rx_buf_ring[cons])
3268 					return RTE_ETH_RX_DESC_DONE;
3269 				else
3270 					return RTE_ETH_RX_DESC_UNAVAIL;
3271 			}
3272 			agg_cnt = BNXT_RX_L2_AGG_BUFS(rxcmp);
3273 			raw_cons = raw_cons + CMP_LEN(cmpl_type) + agg_cnt;
3274 			desc++;
3275 			break;
3276 
3277 		case CMPL_BASE_TYPE_RX_TPA_END:
3278 			if (desc == offset)
3279 				return RTE_ETH_RX_DESC_DONE;
3280 
3281 			if (BNXT_CHIP_P5(rxq->bp)) {
3282 				struct rx_tpa_v2_end_cmpl_hi *p5_tpa_end;
3283 
3284 				p5_tpa_end = (void *)rxcmp;
3285 				agg_cnt = BNXT_TPA_END_AGG_BUFS_TH(p5_tpa_end);
3286 			} else {
3287 				struct rx_tpa_end_cmpl *tpa_end;
3288 
3289 				tpa_end = (void *)rxcmp;
3290 				agg_cnt = BNXT_TPA_END_AGG_BUFS(tpa_end);
3291 			}
3292 
3293 			raw_cons = raw_cons + CMP_LEN(cmpl_type) + agg_cnt;
3294 			desc++;
3295 			break;
3296 
3297 		default:
3298 			raw_cons += CMP_LEN(cmpl_type);
3299 		}
3300 	}
3301 
3302 	return RTE_ETH_RX_DESC_AVAIL;
3303 }
3304 
3305 static int
3306 bnxt_tx_descriptor_status_op(void *tx_queue, uint16_t offset)
3307 {
3308 	struct bnxt_tx_queue *txq = (struct bnxt_tx_queue *)tx_queue;
3309 	struct bnxt_cp_ring_info *cpr = txq->cp_ring;
3310 	uint32_t ring_mask, raw_cons, nb_tx_pkts = 0;
3311 	struct cmpl_base *cp_desc_ring;
3312 	int rc;
3313 
3314 	rc = is_bnxt_in_error(txq->bp);
3315 	if (rc)
3316 		return rc;
3317 
3318 	if (offset >= txq->nb_tx_desc)
3319 		return -EINVAL;
3320 
3321 	/* Return "desc done" if descriptor is available for use. */
3322 	if (bnxt_tx_bds_in_hw(txq) <= offset)
3323 		return RTE_ETH_TX_DESC_DONE;
3324 
3325 	raw_cons = cpr->cp_raw_cons;
3326 	cp_desc_ring = cpr->cp_desc_ring;
3327 	ring_mask = cpr->cp_ring_struct->ring_mask;
3328 
3329 	/* Check to see if hw has posted a completion for the descriptor. */
3330 	while (1) {
3331 		struct tx_cmpl *txcmp;
3332 		uint32_t cons;
3333 
3334 		cons = RING_CMPL(ring_mask, raw_cons);
3335 		txcmp = (struct tx_cmpl *)&cp_desc_ring[cons];
3336 
3337 		if (!bnxt_cpr_cmp_valid(txcmp, raw_cons, ring_mask + 1))
3338 			break;
3339 
3340 		if (CMP_TYPE(txcmp) == TX_CMPL_TYPE_TX_L2)
3341 			nb_tx_pkts += rte_le_to_cpu_32(txcmp->opaque);
3342 
3343 		if (nb_tx_pkts > offset)
3344 			return RTE_ETH_TX_DESC_DONE;
3345 
3346 		raw_cons = NEXT_RAW_CMP(raw_cons);
3347 	}
3348 
3349 	/* Descriptor is pending transmit, not yet completed by hardware. */
3350 	return RTE_ETH_TX_DESC_FULL;
3351 }
3352 
3353 int
3354 bnxt_flow_ops_get_op(struct rte_eth_dev *dev,
3355 		     const struct rte_flow_ops **ops)
3356 {
3357 	struct bnxt *bp = dev->data->dev_private;
3358 	int ret = 0;
3359 
3360 	if (!bp)
3361 		return -EIO;
3362 
3363 	if (BNXT_ETH_DEV_IS_REPRESENTOR(dev)) {
3364 		struct bnxt_representor *vfr = dev->data->dev_private;
3365 		bp = vfr->parent_dev->data->dev_private;
3366 		/* parent is deleted while children are still valid */
3367 		if (!bp) {
3368 			PMD_DRV_LOG(DEBUG, "BNXT Port:%d VFR Error\n",
3369 				    dev->data->port_id);
3370 			return -EIO;
3371 		}
3372 	}
3373 
3374 	ret = is_bnxt_in_error(bp);
3375 	if (ret)
3376 		return ret;
3377 
3378 	/* PMD supports thread-safe flow operations.  rte_flow API
3379 	 * functions can avoid mutex for multi-thread safety.
3380 	 */
3381 	dev->data->dev_flags |= RTE_ETH_DEV_FLOW_OPS_THREAD_SAFE;
3382 
3383 	if (BNXT_TRUFLOW_EN(bp))
3384 		*ops = &bnxt_ulp_rte_flow_ops;
3385 	else
3386 		*ops = &bnxt_flow_ops;
3387 
3388 	return ret;
3389 }
3390 
3391 static const uint32_t *
3392 bnxt_dev_supported_ptypes_get_op(struct rte_eth_dev *dev)
3393 {
3394 	static const uint32_t ptypes[] = {
3395 		RTE_PTYPE_L2_ETHER_VLAN,
3396 		RTE_PTYPE_L3_IPV4_EXT_UNKNOWN,
3397 		RTE_PTYPE_L3_IPV6_EXT_UNKNOWN,
3398 		RTE_PTYPE_L4_ICMP,
3399 		RTE_PTYPE_L4_TCP,
3400 		RTE_PTYPE_L4_UDP,
3401 		RTE_PTYPE_INNER_L3_IPV4_EXT_UNKNOWN,
3402 		RTE_PTYPE_INNER_L3_IPV6_EXT_UNKNOWN,
3403 		RTE_PTYPE_INNER_L4_ICMP,
3404 		RTE_PTYPE_INNER_L4_TCP,
3405 		RTE_PTYPE_INNER_L4_UDP,
3406 		RTE_PTYPE_UNKNOWN
3407 	};
3408 
3409 	if (!dev->rx_pkt_burst)
3410 		return NULL;
3411 
3412 	return ptypes;
3413 }
3414 
3415 static int bnxt_map_regs(struct bnxt *bp, uint32_t *reg_arr, int count,
3416 			 int reg_win)
3417 {
3418 	uint32_t reg_base = *reg_arr & 0xfffff000;
3419 	uint32_t win_off;
3420 	int i;
3421 
3422 	for (i = 0; i < count; i++) {
3423 		if ((reg_arr[i] & 0xfffff000) != reg_base)
3424 			return -ERANGE;
3425 	}
3426 	win_off = BNXT_GRCPF_REG_WINDOW_BASE_OUT + (reg_win - 1) * 4;
3427 	rte_write32(reg_base, (uint8_t *)bp->bar0 + win_off);
3428 	return 0;
3429 }
3430 
3431 static int bnxt_map_ptp_regs(struct bnxt *bp)
3432 {
3433 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3434 	uint32_t *reg_arr;
3435 	int rc, i;
3436 
3437 	reg_arr = ptp->rx_regs;
3438 	rc = bnxt_map_regs(bp, reg_arr, BNXT_PTP_RX_REGS, 5);
3439 	if (rc)
3440 		return rc;
3441 
3442 	reg_arr = ptp->tx_regs;
3443 	rc = bnxt_map_regs(bp, reg_arr, BNXT_PTP_TX_REGS, 6);
3444 	if (rc)
3445 		return rc;
3446 
3447 	for (i = 0; i < BNXT_PTP_RX_REGS; i++)
3448 		ptp->rx_mapped_regs[i] = 0x5000 + (ptp->rx_regs[i] & 0xfff);
3449 
3450 	for (i = 0; i < BNXT_PTP_TX_REGS; i++)
3451 		ptp->tx_mapped_regs[i] = 0x6000 + (ptp->tx_regs[i] & 0xfff);
3452 
3453 	return 0;
3454 }
3455 
3456 static void bnxt_unmap_ptp_regs(struct bnxt *bp)
3457 {
3458 	rte_write32(0, (uint8_t *)bp->bar0 +
3459 			 BNXT_GRCPF_REG_WINDOW_BASE_OUT + 16);
3460 	rte_write32(0, (uint8_t *)bp->bar0 +
3461 			 BNXT_GRCPF_REG_WINDOW_BASE_OUT + 20);
3462 }
3463 
3464 static uint64_t bnxt_cc_read(struct bnxt *bp)
3465 {
3466 	uint64_t ns;
3467 
3468 	ns = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3469 			      BNXT_GRCPF_REG_SYNC_TIME));
3470 	ns |= (uint64_t)(rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3471 					  BNXT_GRCPF_REG_SYNC_TIME + 4))) << 32;
3472 	return ns;
3473 }
3474 
3475 static int bnxt_get_tx_ts(struct bnxt *bp, uint64_t *ts)
3476 {
3477 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3478 	uint32_t fifo;
3479 
3480 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3481 				ptp->tx_mapped_regs[BNXT_PTP_TX_FIFO]));
3482 	if (fifo & BNXT_PTP_TX_FIFO_EMPTY)
3483 		return -EAGAIN;
3484 
3485 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3486 				ptp->tx_mapped_regs[BNXT_PTP_TX_FIFO]));
3487 	*ts = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3488 				ptp->tx_mapped_regs[BNXT_PTP_TX_TS_L]));
3489 	*ts |= (uint64_t)rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3490 				ptp->tx_mapped_regs[BNXT_PTP_TX_TS_H])) << 32;
3491 	rte_read32((uint8_t *)bp->bar0 + ptp->tx_mapped_regs[BNXT_PTP_TX_SEQ]);
3492 
3493 	return 0;
3494 }
3495 
3496 static int bnxt_clr_rx_ts(struct bnxt *bp, uint64_t *last_ts)
3497 {
3498 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3499 	struct bnxt_pf_info *pf = bp->pf;
3500 	uint16_t port_id;
3501 	int i = 0;
3502 	uint32_t fifo;
3503 
3504 	if (!ptp || (bp->flags & BNXT_FLAG_CHIP_P5))
3505 		return -EINVAL;
3506 
3507 	port_id = pf->port_id;
3508 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3509 				ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO]));
3510 	while ((fifo & BNXT_PTP_RX_FIFO_PENDING) && (i < BNXT_PTP_RX_PND_CNT)) {
3511 		rte_write32(1 << port_id, (uint8_t *)bp->bar0 +
3512 			    ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO_ADV]);
3513 		fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3514 					ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO]));
3515 		*last_ts = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3516 					ptp->rx_mapped_regs[BNXT_PTP_RX_TS_L]));
3517 		*last_ts |= (uint64_t)rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3518 					ptp->rx_mapped_regs[BNXT_PTP_RX_TS_H])) << 32;
3519 		i++;
3520 	}
3521 
3522 	if (i >= BNXT_PTP_RX_PND_CNT)
3523 		return -EBUSY;
3524 
3525 	return 0;
3526 }
3527 
3528 static int bnxt_get_rx_ts(struct bnxt *bp, uint64_t *ts)
3529 {
3530 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3531 	struct bnxt_pf_info *pf = bp->pf;
3532 	uint16_t port_id;
3533 	uint32_t fifo;
3534 
3535 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3536 				ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO]));
3537 	if (!(fifo & BNXT_PTP_RX_FIFO_PENDING))
3538 		return -EAGAIN;
3539 
3540 	port_id = pf->port_id;
3541 	rte_write32(1 << port_id, (uint8_t *)bp->bar0 +
3542 	       ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO_ADV]);
3543 
3544 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3545 				   ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO]));
3546 	if (fifo & BNXT_PTP_RX_FIFO_PENDING)
3547 		return bnxt_clr_rx_ts(bp, ts);
3548 
3549 	*ts = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3550 				ptp->rx_mapped_regs[BNXT_PTP_RX_TS_L]));
3551 	*ts |= (uint64_t)rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3552 				ptp->rx_mapped_regs[BNXT_PTP_RX_TS_H])) << 32;
3553 
3554 	return 0;
3555 }
3556 
3557 static int
3558 bnxt_timesync_write_time(struct rte_eth_dev *dev, const struct timespec *ts)
3559 {
3560 	uint64_t ns;
3561 	struct bnxt *bp = dev->data->dev_private;
3562 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3563 
3564 	if (!ptp)
3565 		return -ENOTSUP;
3566 
3567 	ns = rte_timespec_to_ns(ts);
3568 	/* Set the timecounters to a new value. */
3569 	ptp->tc.nsec = ns;
3570 	ptp->tx_tstamp_tc.nsec = ns;
3571 	ptp->rx_tstamp_tc.nsec = ns;
3572 
3573 	return 0;
3574 }
3575 
3576 static int
3577 bnxt_timesync_read_time(struct rte_eth_dev *dev, struct timespec *ts)
3578 {
3579 	struct bnxt *bp = dev->data->dev_private;
3580 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3581 	uint64_t ns, systime_cycles = 0;
3582 	int rc = 0;
3583 
3584 	if (!ptp)
3585 		return -ENOTSUP;
3586 
3587 	if (BNXT_CHIP_P5(bp))
3588 		rc = bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_CURRENT_TIME,
3589 					     &systime_cycles);
3590 	else
3591 		systime_cycles = bnxt_cc_read(bp);
3592 
3593 	ns = rte_timecounter_update(&ptp->tc, systime_cycles);
3594 	*ts = rte_ns_to_timespec(ns);
3595 
3596 	return rc;
3597 }
3598 static int
3599 bnxt_timesync_enable(struct rte_eth_dev *dev)
3600 {
3601 	struct bnxt *bp = dev->data->dev_private;
3602 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3603 	uint32_t shift = 0;
3604 	int rc;
3605 
3606 	if (!ptp)
3607 		return -ENOTSUP;
3608 
3609 	ptp->rx_filter = 1;
3610 	ptp->tx_tstamp_en = 1;
3611 	ptp->rxctl = BNXT_PTP_MSG_EVENTS;
3612 
3613 	rc = bnxt_hwrm_ptp_cfg(bp);
3614 	if (rc)
3615 		return rc;
3616 
3617 	memset(&ptp->tc, 0, sizeof(struct rte_timecounter));
3618 	memset(&ptp->rx_tstamp_tc, 0, sizeof(struct rte_timecounter));
3619 	memset(&ptp->tx_tstamp_tc, 0, sizeof(struct rte_timecounter));
3620 
3621 	ptp->tc.cc_mask = BNXT_CYCLECOUNTER_MASK;
3622 	ptp->tc.cc_shift = shift;
3623 	ptp->tc.nsec_mask = (1ULL << shift) - 1;
3624 
3625 	ptp->rx_tstamp_tc.cc_mask = BNXT_CYCLECOUNTER_MASK;
3626 	ptp->rx_tstamp_tc.cc_shift = shift;
3627 	ptp->rx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
3628 
3629 	ptp->tx_tstamp_tc.cc_mask = BNXT_CYCLECOUNTER_MASK;
3630 	ptp->tx_tstamp_tc.cc_shift = shift;
3631 	ptp->tx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
3632 
3633 	if (!BNXT_CHIP_P5(bp))
3634 		bnxt_map_ptp_regs(bp);
3635 	else
3636 		rc = bnxt_ptp_start(bp);
3637 
3638 	return rc;
3639 }
3640 
3641 static int
3642 bnxt_timesync_disable(struct rte_eth_dev *dev)
3643 {
3644 	struct bnxt *bp = dev->data->dev_private;
3645 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3646 
3647 	if (!ptp)
3648 		return -ENOTSUP;
3649 
3650 	ptp->rx_filter = 0;
3651 	ptp->tx_tstamp_en = 0;
3652 	ptp->rxctl = 0;
3653 
3654 	bnxt_hwrm_ptp_cfg(bp);
3655 
3656 	if (!BNXT_CHIP_P5(bp))
3657 		bnxt_unmap_ptp_regs(bp);
3658 	else
3659 		bnxt_ptp_stop(bp);
3660 
3661 	return 0;
3662 }
3663 
3664 static int
3665 bnxt_timesync_read_rx_timestamp(struct rte_eth_dev *dev,
3666 				 struct timespec *timestamp,
3667 				 uint32_t flags __rte_unused)
3668 {
3669 	struct bnxt *bp = dev->data->dev_private;
3670 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3671 	uint64_t rx_tstamp_cycles = 0;
3672 	uint64_t ns;
3673 
3674 	if (!ptp)
3675 		return -ENOTSUP;
3676 
3677 	if (BNXT_CHIP_P5(bp))
3678 		rx_tstamp_cycles = ptp->rx_timestamp;
3679 	else
3680 		bnxt_get_rx_ts(bp, &rx_tstamp_cycles);
3681 
3682 	ns = rte_timecounter_update(&ptp->rx_tstamp_tc, rx_tstamp_cycles);
3683 	*timestamp = rte_ns_to_timespec(ns);
3684 	return  0;
3685 }
3686 
3687 static int
3688 bnxt_timesync_read_tx_timestamp(struct rte_eth_dev *dev,
3689 				 struct timespec *timestamp)
3690 {
3691 	struct bnxt *bp = dev->data->dev_private;
3692 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3693 	uint64_t tx_tstamp_cycles = 0;
3694 	uint64_t ns;
3695 	int rc = 0;
3696 
3697 	if (!ptp)
3698 		return -ENOTSUP;
3699 
3700 	if (BNXT_CHIP_P5(bp))
3701 		rc = bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_PATH_TX,
3702 					     &tx_tstamp_cycles);
3703 	else
3704 		rc = bnxt_get_tx_ts(bp, &tx_tstamp_cycles);
3705 
3706 	ns = rte_timecounter_update(&ptp->tx_tstamp_tc, tx_tstamp_cycles);
3707 	*timestamp = rte_ns_to_timespec(ns);
3708 
3709 	return rc;
3710 }
3711 
3712 static int
3713 bnxt_timesync_adjust_time(struct rte_eth_dev *dev, int64_t delta)
3714 {
3715 	struct bnxt *bp = dev->data->dev_private;
3716 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3717 
3718 	if (!ptp)
3719 		return -ENOTSUP;
3720 
3721 	ptp->tc.nsec += delta;
3722 	ptp->tx_tstamp_tc.nsec += delta;
3723 	ptp->rx_tstamp_tc.nsec += delta;
3724 
3725 	return 0;
3726 }
3727 
3728 static int
3729 bnxt_get_eeprom_length_op(struct rte_eth_dev *dev)
3730 {
3731 	struct bnxt *bp = dev->data->dev_private;
3732 	int rc;
3733 	uint32_t dir_entries;
3734 	uint32_t entry_length;
3735 
3736 	rc = is_bnxt_in_error(bp);
3737 	if (rc)
3738 		return rc;
3739 
3740 	PMD_DRV_LOG(INFO, PCI_PRI_FMT "\n",
3741 		    bp->pdev->addr.domain, bp->pdev->addr.bus,
3742 		    bp->pdev->addr.devid, bp->pdev->addr.function);
3743 
3744 	rc = bnxt_hwrm_nvm_get_dir_info(bp, &dir_entries, &entry_length);
3745 	if (rc != 0)
3746 		return rc;
3747 
3748 	return dir_entries * entry_length;
3749 }
3750 
3751 static int
3752 bnxt_get_eeprom_op(struct rte_eth_dev *dev,
3753 		struct rte_dev_eeprom_info *in_eeprom)
3754 {
3755 	struct bnxt *bp = dev->data->dev_private;
3756 	uint32_t index;
3757 	uint32_t offset;
3758 	int rc;
3759 
3760 	rc = is_bnxt_in_error(bp);
3761 	if (rc)
3762 		return rc;
3763 
3764 	PMD_DRV_LOG(INFO, PCI_PRI_FMT " in_eeprom->offset = %d len = %d\n",
3765 		    bp->pdev->addr.domain, bp->pdev->addr.bus,
3766 		    bp->pdev->addr.devid, bp->pdev->addr.function,
3767 		    in_eeprom->offset, in_eeprom->length);
3768 
3769 	if (in_eeprom->offset == 0) /* special offset value to get directory */
3770 		return bnxt_get_nvram_directory(bp, in_eeprom->length,
3771 						in_eeprom->data);
3772 
3773 	index = in_eeprom->offset >> 24;
3774 	offset = in_eeprom->offset & 0xffffff;
3775 
3776 	if (index != 0)
3777 		return bnxt_hwrm_get_nvram_item(bp, index - 1, offset,
3778 					   in_eeprom->length, in_eeprom->data);
3779 
3780 	return 0;
3781 }
3782 
3783 static bool bnxt_dir_type_is_ape_bin_format(uint16_t dir_type)
3784 {
3785 	switch (dir_type) {
3786 	case BNX_DIR_TYPE_CHIMP_PATCH:
3787 	case BNX_DIR_TYPE_BOOTCODE:
3788 	case BNX_DIR_TYPE_BOOTCODE_2:
3789 	case BNX_DIR_TYPE_APE_FW:
3790 	case BNX_DIR_TYPE_APE_PATCH:
3791 	case BNX_DIR_TYPE_KONG_FW:
3792 	case BNX_DIR_TYPE_KONG_PATCH:
3793 	case BNX_DIR_TYPE_BONO_FW:
3794 	case BNX_DIR_TYPE_BONO_PATCH:
3795 		/* FALLTHROUGH */
3796 		return true;
3797 	}
3798 
3799 	return false;
3800 }
3801 
3802 static bool bnxt_dir_type_is_other_exec_format(uint16_t dir_type)
3803 {
3804 	switch (dir_type) {
3805 	case BNX_DIR_TYPE_AVS:
3806 	case BNX_DIR_TYPE_EXP_ROM_MBA:
3807 	case BNX_DIR_TYPE_PCIE:
3808 	case BNX_DIR_TYPE_TSCF_UCODE:
3809 	case BNX_DIR_TYPE_EXT_PHY:
3810 	case BNX_DIR_TYPE_CCM:
3811 	case BNX_DIR_TYPE_ISCSI_BOOT:
3812 	case BNX_DIR_TYPE_ISCSI_BOOT_IPV6:
3813 	case BNX_DIR_TYPE_ISCSI_BOOT_IPV4N6:
3814 		/* FALLTHROUGH */
3815 		return true;
3816 	}
3817 
3818 	return false;
3819 }
3820 
3821 static bool bnxt_dir_type_is_executable(uint16_t dir_type)
3822 {
3823 	return bnxt_dir_type_is_ape_bin_format(dir_type) ||
3824 		bnxt_dir_type_is_other_exec_format(dir_type);
3825 }
3826 
3827 static int
3828 bnxt_set_eeprom_op(struct rte_eth_dev *dev,
3829 		struct rte_dev_eeprom_info *in_eeprom)
3830 {
3831 	struct bnxt *bp = dev->data->dev_private;
3832 	uint8_t index, dir_op;
3833 	uint16_t type, ext, ordinal, attr;
3834 	int rc;
3835 
3836 	rc = is_bnxt_in_error(bp);
3837 	if (rc)
3838 		return rc;
3839 
3840 	PMD_DRV_LOG(INFO, PCI_PRI_FMT " in_eeprom->offset = %d len = %d\n",
3841 		    bp->pdev->addr.domain, bp->pdev->addr.bus,
3842 		    bp->pdev->addr.devid, bp->pdev->addr.function,
3843 		    in_eeprom->offset, in_eeprom->length);
3844 
3845 	if (!BNXT_PF(bp)) {
3846 		PMD_DRV_LOG(ERR, "NVM write not supported from a VF\n");
3847 		return -EINVAL;
3848 	}
3849 
3850 	type = in_eeprom->magic >> 16;
3851 
3852 	if (type == 0xffff) { /* special value for directory operations */
3853 		index = in_eeprom->magic & 0xff;
3854 		dir_op = in_eeprom->magic >> 8;
3855 		if (index == 0)
3856 			return -EINVAL;
3857 		switch (dir_op) {
3858 		case 0x0e: /* erase */
3859 			if (in_eeprom->offset != ~in_eeprom->magic)
3860 				return -EINVAL;
3861 			return bnxt_hwrm_erase_nvram_directory(bp, index - 1);
3862 		default:
3863 			return -EINVAL;
3864 		}
3865 	}
3866 
3867 	/* Create or re-write an NVM item: */
3868 	if (bnxt_dir_type_is_executable(type) == true)
3869 		return -EOPNOTSUPP;
3870 	ext = in_eeprom->magic & 0xffff;
3871 	ordinal = in_eeprom->offset >> 16;
3872 	attr = in_eeprom->offset & 0xffff;
3873 
3874 	return bnxt_hwrm_flash_nvram(bp, type, ordinal, ext, attr,
3875 				     in_eeprom->data, in_eeprom->length);
3876 }
3877 
3878 static int bnxt_get_module_info(struct rte_eth_dev *dev,
3879 				struct rte_eth_dev_module_info *modinfo)
3880 {
3881 	uint8_t module_info[SFF_DIAG_SUPPORT_OFFSET + 1];
3882 	struct bnxt *bp = dev->data->dev_private;
3883 	int rc;
3884 
3885 	/* No point in going further if phy status indicates
3886 	 * module is not inserted or if it is powered down or
3887 	 * if it is of type 10GBase-T
3888 	 */
3889 	if (bp->link_info->module_status >
3890 	    HWRM_PORT_PHY_QCFG_OUTPUT_MODULE_STATUS_WARNINGMSG) {
3891 		PMD_DRV_LOG(NOTICE, "Port %u : Module is not inserted or is powered down\n",
3892 			    dev->data->port_id);
3893 		return -ENOTSUP;
3894 	}
3895 
3896 	/* This feature is not supported in older firmware versions */
3897 	if (bp->hwrm_spec_code < 0x10202) {
3898 		PMD_DRV_LOG(NOTICE, "Port %u : Feature is not supported in older firmware\n",
3899 			    dev->data->port_id);
3900 		return -ENOTSUP;
3901 	}
3902 
3903 	rc = bnxt_hwrm_read_sfp_module_eeprom_info(bp, I2C_DEV_ADDR_A0, 0, 0,
3904 						   SFF_DIAG_SUPPORT_OFFSET + 1,
3905 						   module_info);
3906 
3907 	if (rc)
3908 		return rc;
3909 
3910 	switch (module_info[0]) {
3911 	case SFF_MODULE_ID_SFP:
3912 		modinfo->type = RTE_ETH_MODULE_SFF_8472;
3913 		modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8472_LEN;
3914 		if (module_info[SFF_DIAG_SUPPORT_OFFSET] == 0)
3915 			modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8436_LEN;
3916 		break;
3917 	case SFF_MODULE_ID_QSFP:
3918 	case SFF_MODULE_ID_QSFP_PLUS:
3919 		modinfo->type = RTE_ETH_MODULE_SFF_8436;
3920 		modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8436_LEN;
3921 		break;
3922 	case SFF_MODULE_ID_QSFP28:
3923 		modinfo->type = RTE_ETH_MODULE_SFF_8636;
3924 		modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8636_MAX_LEN;
3925 		if (module_info[SFF8636_FLATMEM_OFFSET] & SFF8636_FLATMEM_MASK)
3926 			modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8636_LEN;
3927 		break;
3928 	default:
3929 		PMD_DRV_LOG(NOTICE, "Port %u : Unsupported module\n", dev->data->port_id);
3930 		return -ENOTSUP;
3931 	}
3932 
3933 	PMD_DRV_LOG(INFO, "Port %u : modinfo->type = %d modinfo->eeprom_len = %d\n",
3934 		    dev->data->port_id, modinfo->type, modinfo->eeprom_len);
3935 
3936 	return 0;
3937 }
3938 
3939 static int bnxt_get_module_eeprom(struct rte_eth_dev *dev,
3940 				  struct rte_dev_eeprom_info *info)
3941 {
3942 	uint8_t pg_addr[5] = { I2C_DEV_ADDR_A0, I2C_DEV_ADDR_A0 };
3943 	uint32_t offset = info->offset, length = info->length;
3944 	uint8_t module_info[SFF_DIAG_SUPPORT_OFFSET + 1];
3945 	struct bnxt *bp = dev->data->dev_private;
3946 	uint8_t *data = info->data;
3947 	uint8_t page = offset >> 7;
3948 	uint8_t max_pages = 2;
3949 	uint8_t opt_pages;
3950 	int rc;
3951 
3952 	rc = bnxt_hwrm_read_sfp_module_eeprom_info(bp, I2C_DEV_ADDR_A0, 0, 0,
3953 						   SFF_DIAG_SUPPORT_OFFSET + 1,
3954 						   module_info);
3955 	if (rc)
3956 		return rc;
3957 
3958 	switch (module_info[0]) {
3959 	case SFF_MODULE_ID_SFP:
3960 		module_info[SFF_DIAG_SUPPORT_OFFSET] = 0;
3961 		if (module_info[SFF_DIAG_SUPPORT_OFFSET]) {
3962 			pg_addr[2] = I2C_DEV_ADDR_A2;
3963 			pg_addr[3] = I2C_DEV_ADDR_A2;
3964 			max_pages = 4;
3965 		}
3966 		break;
3967 	case SFF_MODULE_ID_QSFP28:
3968 		rc = bnxt_hwrm_read_sfp_module_eeprom_info(bp, I2C_DEV_ADDR_A0, 0,
3969 							   SFF8636_OPT_PAGES_OFFSET,
3970 							   1, &opt_pages);
3971 		if (rc)
3972 			return rc;
3973 
3974 		if (opt_pages & SFF8636_PAGE1_MASK) {
3975 			pg_addr[2] = I2C_DEV_ADDR_A0;
3976 			max_pages = 3;
3977 		}
3978 		if (opt_pages & SFF8636_PAGE2_MASK) {
3979 			pg_addr[3] = I2C_DEV_ADDR_A0;
3980 			max_pages = 4;
3981 		}
3982 		if (~module_info[SFF8636_FLATMEM_OFFSET] & SFF8636_FLATMEM_MASK) {
3983 			pg_addr[4] = I2C_DEV_ADDR_A0;
3984 			max_pages = 5;
3985 		}
3986 		break;
3987 	default:
3988 		break;
3989 	}
3990 
3991 	memset(data, 0, length);
3992 
3993 	offset &= 0xff;
3994 	while (length && page < max_pages) {
3995 		uint8_t raw_page = page ? page - 1 : 0;
3996 		uint16_t chunk;
3997 
3998 		if (pg_addr[page] == I2C_DEV_ADDR_A2)
3999 			raw_page = 0;
4000 		else if (page)
4001 			offset |= 0x80;
4002 		chunk = RTE_MIN(length, 256 - offset);
4003 
4004 		if (pg_addr[page]) {
4005 			rc = bnxt_hwrm_read_sfp_module_eeprom_info(bp, pg_addr[page],
4006 								   raw_page, offset,
4007 								   chunk, data);
4008 			if (rc)
4009 				return rc;
4010 		}
4011 
4012 		data += chunk;
4013 		length -= chunk;
4014 		offset = 0;
4015 		page += 1 + (chunk > 128);
4016 	}
4017 
4018 	return length ? -EINVAL : 0;
4019 }
4020 
4021 /*
4022  * Initialization
4023  */
4024 
4025 static const struct eth_dev_ops bnxt_dev_ops = {
4026 	.dev_infos_get = bnxt_dev_info_get_op,
4027 	.dev_close = bnxt_dev_close_op,
4028 	.dev_configure = bnxt_dev_configure_op,
4029 	.dev_start = bnxt_dev_start_op,
4030 	.dev_stop = bnxt_dev_stop_op,
4031 	.dev_set_link_up = bnxt_dev_set_link_up_op,
4032 	.dev_set_link_down = bnxt_dev_set_link_down_op,
4033 	.stats_get = bnxt_stats_get_op,
4034 	.stats_reset = bnxt_stats_reset_op,
4035 	.rx_queue_setup = bnxt_rx_queue_setup_op,
4036 	.rx_queue_release = bnxt_rx_queue_release_op,
4037 	.tx_queue_setup = bnxt_tx_queue_setup_op,
4038 	.tx_queue_release = bnxt_tx_queue_release_op,
4039 	.rx_queue_intr_enable = bnxt_rx_queue_intr_enable_op,
4040 	.rx_queue_intr_disable = bnxt_rx_queue_intr_disable_op,
4041 	.reta_update = bnxt_reta_update_op,
4042 	.reta_query = bnxt_reta_query_op,
4043 	.rss_hash_update = bnxt_rss_hash_update_op,
4044 	.rss_hash_conf_get = bnxt_rss_hash_conf_get_op,
4045 	.link_update = bnxt_link_update_op,
4046 	.promiscuous_enable = bnxt_promiscuous_enable_op,
4047 	.promiscuous_disable = bnxt_promiscuous_disable_op,
4048 	.allmulticast_enable = bnxt_allmulticast_enable_op,
4049 	.allmulticast_disable = bnxt_allmulticast_disable_op,
4050 	.mac_addr_add = bnxt_mac_addr_add_op,
4051 	.mac_addr_remove = bnxt_mac_addr_remove_op,
4052 	.flow_ctrl_get = bnxt_flow_ctrl_get_op,
4053 	.flow_ctrl_set = bnxt_flow_ctrl_set_op,
4054 	.udp_tunnel_port_add  = bnxt_udp_tunnel_port_add_op,
4055 	.udp_tunnel_port_del  = bnxt_udp_tunnel_port_del_op,
4056 	.vlan_filter_set = bnxt_vlan_filter_set_op,
4057 	.vlan_offload_set = bnxt_vlan_offload_set_op,
4058 	.vlan_tpid_set = bnxt_vlan_tpid_set_op,
4059 	.vlan_pvid_set = bnxt_vlan_pvid_set_op,
4060 	.mtu_set = bnxt_mtu_set_op,
4061 	.mac_addr_set = bnxt_set_default_mac_addr_op,
4062 	.xstats_get = bnxt_dev_xstats_get_op,
4063 	.xstats_get_names = bnxt_dev_xstats_get_names_op,
4064 	.xstats_reset = bnxt_dev_xstats_reset_op,
4065 	.fw_version_get = bnxt_fw_version_get,
4066 	.set_mc_addr_list = bnxt_dev_set_mc_addr_list_op,
4067 	.rxq_info_get = bnxt_rxq_info_get_op,
4068 	.txq_info_get = bnxt_txq_info_get_op,
4069 	.rx_burst_mode_get = bnxt_rx_burst_mode_get,
4070 	.tx_burst_mode_get = bnxt_tx_burst_mode_get,
4071 	.dev_led_on = bnxt_dev_led_on_op,
4072 	.dev_led_off = bnxt_dev_led_off_op,
4073 	.rx_queue_start = bnxt_rx_queue_start,
4074 	.rx_queue_stop = bnxt_rx_queue_stop,
4075 	.tx_queue_start = bnxt_tx_queue_start,
4076 	.tx_queue_stop = bnxt_tx_queue_stop,
4077 	.flow_ops_get = bnxt_flow_ops_get_op,
4078 	.dev_supported_ptypes_get = bnxt_dev_supported_ptypes_get_op,
4079 	.get_eeprom_length    = bnxt_get_eeprom_length_op,
4080 	.get_eeprom           = bnxt_get_eeprom_op,
4081 	.set_eeprom           = bnxt_set_eeprom_op,
4082 	.get_module_info = bnxt_get_module_info,
4083 	.get_module_eeprom = bnxt_get_module_eeprom,
4084 	.timesync_enable      = bnxt_timesync_enable,
4085 	.timesync_disable     = bnxt_timesync_disable,
4086 	.timesync_read_time   = bnxt_timesync_read_time,
4087 	.timesync_write_time   = bnxt_timesync_write_time,
4088 	.timesync_adjust_time = bnxt_timesync_adjust_time,
4089 	.timesync_read_rx_timestamp = bnxt_timesync_read_rx_timestamp,
4090 	.timesync_read_tx_timestamp = bnxt_timesync_read_tx_timestamp,
4091 };
4092 
4093 static uint32_t bnxt_map_reset_regs(struct bnxt *bp, uint32_t reg)
4094 {
4095 	uint32_t offset;
4096 
4097 	/* Only pre-map the reset GRC registers using window 3 */
4098 	rte_write32(reg & 0xfffff000, (uint8_t *)bp->bar0 +
4099 		    BNXT_GRCPF_REG_WINDOW_BASE_OUT + 8);
4100 
4101 	offset = BNXT_GRCP_WINDOW_3_BASE + (reg & 0xffc);
4102 
4103 	return offset;
4104 }
4105 
4106 int bnxt_map_fw_health_status_regs(struct bnxt *bp)
4107 {
4108 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4109 	uint32_t reg_base = 0xffffffff;
4110 	int i;
4111 
4112 	/* Only pre-map the monitoring GRC registers using window 2 */
4113 	for (i = 0; i < BNXT_FW_STATUS_REG_CNT; i++) {
4114 		uint32_t reg = info->status_regs[i];
4115 
4116 		if (BNXT_FW_STATUS_REG_TYPE(reg) != BNXT_FW_STATUS_REG_TYPE_GRC)
4117 			continue;
4118 
4119 		if (reg_base == 0xffffffff)
4120 			reg_base = reg & 0xfffff000;
4121 		if ((reg & 0xfffff000) != reg_base)
4122 			return -ERANGE;
4123 
4124 		/* Use mask 0xffc as the Lower 2 bits indicates
4125 		 * address space location
4126 		 */
4127 		info->mapped_status_regs[i] = BNXT_GRCP_WINDOW_2_BASE +
4128 						(reg & 0xffc);
4129 	}
4130 
4131 	if (reg_base == 0xffffffff)
4132 		return 0;
4133 
4134 	rte_write32(reg_base, (uint8_t *)bp->bar0 +
4135 		    BNXT_GRCPF_REG_WINDOW_BASE_OUT + 4);
4136 
4137 	return 0;
4138 }
4139 
4140 static void bnxt_write_fw_reset_reg(struct bnxt *bp, uint32_t index)
4141 {
4142 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4143 	uint32_t delay = info->delay_after_reset[index];
4144 	uint32_t val = info->reset_reg_val[index];
4145 	uint32_t reg = info->reset_reg[index];
4146 	uint32_t type, offset;
4147 	int ret;
4148 
4149 	type = BNXT_FW_STATUS_REG_TYPE(reg);
4150 	offset = BNXT_FW_STATUS_REG_OFF(reg);
4151 
4152 	switch (type) {
4153 	case BNXT_FW_STATUS_REG_TYPE_CFG:
4154 		ret = rte_pci_write_config(bp->pdev, &val, sizeof(val), offset);
4155 		if (ret < 0) {
4156 			PMD_DRV_LOG(ERR, "Failed to write %#x at PCI offset %#x",
4157 				    val, offset);
4158 			return;
4159 		}
4160 		break;
4161 	case BNXT_FW_STATUS_REG_TYPE_GRC:
4162 		offset = bnxt_map_reset_regs(bp, offset);
4163 		rte_write32(val, (uint8_t *)bp->bar0 + offset);
4164 		break;
4165 	case BNXT_FW_STATUS_REG_TYPE_BAR0:
4166 		rte_write32(val, (uint8_t *)bp->bar0 + offset);
4167 		break;
4168 	}
4169 	/* wait on a specific interval of time until core reset is complete */
4170 	if (delay)
4171 		rte_delay_ms(delay);
4172 }
4173 
4174 static void bnxt_dev_cleanup(struct bnxt *bp)
4175 {
4176 	bp->eth_dev->data->dev_link.link_status = 0;
4177 	bp->link_info->link_up = 0;
4178 	if (bp->eth_dev->data->dev_started)
4179 		bnxt_dev_stop(bp->eth_dev);
4180 
4181 	bnxt_uninit_resources(bp, true);
4182 }
4183 
4184 static int
4185 bnxt_check_fw_reset_done(struct bnxt *bp)
4186 {
4187 	int timeout = bp->fw_reset_max_msecs;
4188 	uint16_t val = 0;
4189 	int rc;
4190 
4191 	do {
4192 		rc = rte_pci_read_config(bp->pdev, &val, sizeof(val), PCI_SUBSYSTEM_ID_OFFSET);
4193 		if (rc < 0) {
4194 			PMD_DRV_LOG(ERR, "Failed to read PCI offset 0x%x", PCI_SUBSYSTEM_ID_OFFSET);
4195 			return rc;
4196 		}
4197 		if (val != 0xffff)
4198 			break;
4199 		rte_delay_ms(1);
4200 	} while (timeout--);
4201 
4202 	if (val == 0xffff) {
4203 		PMD_DRV_LOG(ERR, "Firmware reset aborted, PCI config space invalid\n");
4204 		return -1;
4205 	}
4206 
4207 	return 0;
4208 }
4209 
4210 static int bnxt_restore_vlan_filters(struct bnxt *bp)
4211 {
4212 	struct rte_eth_dev *dev = bp->eth_dev;
4213 	struct rte_vlan_filter_conf *vfc;
4214 	int vidx, vbit, rc;
4215 	uint16_t vlan_id;
4216 
4217 	for (vlan_id = 1; vlan_id <= RTE_ETHER_MAX_VLAN_ID; vlan_id++) {
4218 		vfc = &dev->data->vlan_filter_conf;
4219 		vidx = vlan_id / 64;
4220 		vbit = vlan_id % 64;
4221 
4222 		/* Each bit corresponds to a VLAN id */
4223 		if (vfc->ids[vidx] & (UINT64_C(1) << vbit)) {
4224 			rc = bnxt_add_vlan_filter(bp, vlan_id);
4225 			if (rc)
4226 				return rc;
4227 		}
4228 	}
4229 
4230 	return 0;
4231 }
4232 
4233 static int bnxt_restore_mac_filters(struct bnxt *bp)
4234 {
4235 	struct rte_eth_dev *dev = bp->eth_dev;
4236 	struct rte_eth_dev_info dev_info;
4237 	struct rte_ether_addr *addr;
4238 	uint64_t pool_mask;
4239 	uint32_t pool = 0;
4240 	uint32_t i;
4241 	int rc;
4242 
4243 	if (BNXT_VF(bp) && !BNXT_VF_IS_TRUSTED(bp))
4244 		return 0;
4245 
4246 	rc = bnxt_dev_info_get_op(dev, &dev_info);
4247 	if (rc)
4248 		return rc;
4249 
4250 	/* replay MAC address configuration */
4251 	for (i = 1; i < dev_info.max_mac_addrs; i++) {
4252 		addr = &dev->data->mac_addrs[i];
4253 
4254 		/* skip zero address */
4255 		if (rte_is_zero_ether_addr(addr))
4256 			continue;
4257 
4258 		pool = 0;
4259 		pool_mask = dev->data->mac_pool_sel[i];
4260 
4261 		do {
4262 			if (pool_mask & 1ULL) {
4263 				rc = bnxt_mac_addr_add_op(dev, addr, i, pool);
4264 				if (rc)
4265 					return rc;
4266 			}
4267 			pool_mask >>= 1;
4268 			pool++;
4269 		} while (pool_mask);
4270 	}
4271 
4272 	return 0;
4273 }
4274 
4275 static int bnxt_restore_filters(struct bnxt *bp)
4276 {
4277 	struct rte_eth_dev *dev = bp->eth_dev;
4278 	int ret = 0;
4279 
4280 	if (dev->data->all_multicast) {
4281 		ret = bnxt_allmulticast_enable_op(dev);
4282 		if (ret)
4283 			return ret;
4284 	}
4285 	if (dev->data->promiscuous) {
4286 		ret = bnxt_promiscuous_enable_op(dev);
4287 		if (ret)
4288 			return ret;
4289 	}
4290 
4291 	ret = bnxt_restore_mac_filters(bp);
4292 	if (ret)
4293 		return ret;
4294 
4295 	ret = bnxt_restore_vlan_filters(bp);
4296 	/* TODO restore other filters as well */
4297 	return ret;
4298 }
4299 
4300 static int bnxt_check_fw_ready(struct bnxt *bp)
4301 {
4302 	int timeout = bp->fw_reset_max_msecs;
4303 	int rc = 0;
4304 
4305 	do {
4306 		rc = bnxt_hwrm_poll_ver_get(bp);
4307 		if (rc == 0)
4308 			break;
4309 		rte_delay_ms(BNXT_FW_READY_WAIT_INTERVAL);
4310 		timeout -= BNXT_FW_READY_WAIT_INTERVAL;
4311 	} while (rc && timeout > 0);
4312 
4313 	if (rc)
4314 		PMD_DRV_LOG(ERR, "FW is not Ready after reset\n");
4315 
4316 	return rc;
4317 }
4318 
4319 static void bnxt_dev_recover(void *arg)
4320 {
4321 	struct bnxt *bp = arg;
4322 	int rc = 0;
4323 
4324 	pthread_mutex_lock(&bp->err_recovery_lock);
4325 
4326 	if (!bp->fw_reset_min_msecs) {
4327 		rc = bnxt_check_fw_reset_done(bp);
4328 		if (rc)
4329 			goto err;
4330 	}
4331 
4332 	/* Clear Error flag so that device re-init should happen */
4333 	bp->flags &= ~BNXT_FLAG_FATAL_ERROR;
4334 
4335 	rc = bnxt_check_fw_ready(bp);
4336 	if (rc)
4337 		goto err;
4338 
4339 	rc = bnxt_init_resources(bp, true);
4340 	if (rc) {
4341 		PMD_DRV_LOG(ERR,
4342 			    "Failed to initialize resources after reset\n");
4343 		goto err;
4344 	}
4345 	/* clear reset flag as the device is initialized now */
4346 	bp->flags &= ~BNXT_FLAG_FW_RESET;
4347 
4348 	rc = bnxt_dev_start_op(bp->eth_dev);
4349 	if (rc) {
4350 		PMD_DRV_LOG(ERR, "Failed to start port after reset\n");
4351 		goto err_start;
4352 	}
4353 
4354 	rc = bnxt_restore_filters(bp);
4355 	if (rc)
4356 		goto err_start;
4357 
4358 	PMD_DRV_LOG(INFO, "Recovered from FW reset\n");
4359 	pthread_mutex_unlock(&bp->err_recovery_lock);
4360 
4361 	return;
4362 err_start:
4363 	bnxt_dev_stop(bp->eth_dev);
4364 err:
4365 	bp->flags |= BNXT_FLAG_FATAL_ERROR;
4366 	bnxt_uninit_resources(bp, false);
4367 	if (bp->eth_dev->data->dev_conf.intr_conf.rmv)
4368 		rte_eth_dev_callback_process(bp->eth_dev,
4369 					     RTE_ETH_EVENT_INTR_RMV,
4370 					     NULL);
4371 	pthread_mutex_unlock(&bp->err_recovery_lock);
4372 	PMD_DRV_LOG(ERR, "Failed to recover from FW reset\n");
4373 }
4374 
4375 void bnxt_dev_reset_and_resume(void *arg)
4376 {
4377 	struct bnxt *bp = arg;
4378 	uint32_t us = US_PER_MS * bp->fw_reset_min_msecs;
4379 	uint16_t val = 0;
4380 	int rc;
4381 
4382 	bnxt_dev_cleanup(bp);
4383 
4384 	bnxt_wait_for_device_shutdown(bp);
4385 
4386 	/* During some fatal firmware error conditions, the PCI config space
4387 	 * register 0x2e which normally contains the subsystem ID will become
4388 	 * 0xffff. This register will revert back to the normal value after
4389 	 * the chip has completed core reset. If we detect this condition,
4390 	 * we can poll this config register immediately for the value to revert.
4391 	 */
4392 	if (bp->flags & BNXT_FLAG_FATAL_ERROR) {
4393 		rc = rte_pci_read_config(bp->pdev, &val, sizeof(val), PCI_SUBSYSTEM_ID_OFFSET);
4394 		if (rc < 0) {
4395 			PMD_DRV_LOG(ERR, "Failed to read PCI offset 0x%x", PCI_SUBSYSTEM_ID_OFFSET);
4396 			return;
4397 		}
4398 		if (val == 0xffff) {
4399 			bp->fw_reset_min_msecs = 0;
4400 			us = 1;
4401 		}
4402 	}
4403 
4404 	rc = rte_eal_alarm_set(us, bnxt_dev_recover, (void *)bp);
4405 	if (rc)
4406 		PMD_DRV_LOG(ERR, "Error setting recovery alarm");
4407 }
4408 
4409 uint32_t bnxt_read_fw_status_reg(struct bnxt *bp, uint32_t index)
4410 {
4411 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4412 	uint32_t reg = info->status_regs[index];
4413 	uint32_t type, offset, val = 0;
4414 	int ret = 0;
4415 
4416 	type = BNXT_FW_STATUS_REG_TYPE(reg);
4417 	offset = BNXT_FW_STATUS_REG_OFF(reg);
4418 
4419 	switch (type) {
4420 	case BNXT_FW_STATUS_REG_TYPE_CFG:
4421 		ret = rte_pci_read_config(bp->pdev, &val, sizeof(val), offset);
4422 		if (ret < 0)
4423 			PMD_DRV_LOG(ERR, "Failed to read PCI offset %#x",
4424 				    offset);
4425 		break;
4426 	case BNXT_FW_STATUS_REG_TYPE_GRC:
4427 		offset = info->mapped_status_regs[index];
4428 		/* FALLTHROUGH */
4429 	case BNXT_FW_STATUS_REG_TYPE_BAR0:
4430 		val = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
4431 				       offset));
4432 		break;
4433 	}
4434 
4435 	return val;
4436 }
4437 
4438 static int bnxt_fw_reset_all(struct bnxt *bp)
4439 {
4440 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4441 	uint32_t i;
4442 	int rc = 0;
4443 
4444 	if (info->flags & BNXT_FLAG_ERROR_RECOVERY_HOST) {
4445 		/* Reset through primary function driver */
4446 		for (i = 0; i < info->reg_array_cnt; i++)
4447 			bnxt_write_fw_reset_reg(bp, i);
4448 		/* Wait for time specified by FW after triggering reset */
4449 		rte_delay_ms(info->primary_func_wait_period_after_reset);
4450 	} else if (info->flags & BNXT_FLAG_ERROR_RECOVERY_CO_CPU) {
4451 		/* Reset with the help of Kong processor */
4452 		rc = bnxt_hwrm_fw_reset(bp);
4453 		if (rc)
4454 			PMD_DRV_LOG(ERR, "Failed to reset FW\n");
4455 	}
4456 
4457 	return rc;
4458 }
4459 
4460 static void bnxt_fw_reset_cb(void *arg)
4461 {
4462 	struct bnxt *bp = arg;
4463 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4464 	int rc = 0;
4465 
4466 	/* Only Primary function can do FW reset */
4467 	if (bnxt_is_primary_func(bp) &&
4468 	    bnxt_is_recovery_enabled(bp)) {
4469 		rc = bnxt_fw_reset_all(bp);
4470 		if (rc) {
4471 			PMD_DRV_LOG(ERR, "Adapter recovery failed\n");
4472 			return;
4473 		}
4474 	}
4475 
4476 	/* if recovery method is ERROR_RECOVERY_CO_CPU, KONG will send
4477 	 * EXCEPTION_FATAL_ASYNC event to all the functions
4478 	 * (including MASTER FUNC). After receiving this Async, all the active
4479 	 * drivers should treat this case as FW initiated recovery
4480 	 */
4481 	if (info->flags & BNXT_FLAG_ERROR_RECOVERY_HOST) {
4482 		bp->fw_reset_min_msecs = BNXT_MIN_FW_READY_TIMEOUT;
4483 		bp->fw_reset_max_msecs = BNXT_MAX_FW_RESET_TIMEOUT;
4484 
4485 		/* To recover from error */
4486 		rte_eal_alarm_set(US_PER_MS, bnxt_dev_reset_and_resume,
4487 				  (void *)bp);
4488 	}
4489 }
4490 
4491 /* Driver should poll FW heartbeat, reset_counter with the frequency
4492  * advertised by FW in HWRM_ERROR_RECOVERY_QCFG.
4493  * When the driver detects heartbeat stop or change in reset_counter,
4494  * it has to trigger a reset to recover from the error condition.
4495  * A “primary function” is the function who will have the privilege to
4496  * initiate the chimp reset. The primary function will be elected by the
4497  * firmware and will be notified through async message.
4498  */
4499 static void bnxt_check_fw_health(void *arg)
4500 {
4501 	struct bnxt *bp = arg;
4502 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4503 	uint32_t val = 0, wait_msec;
4504 
4505 	if (!info || !bnxt_is_recovery_enabled(bp) ||
4506 	    is_bnxt_in_error(bp))
4507 		return;
4508 
4509 	val = bnxt_read_fw_status_reg(bp, BNXT_FW_HEARTBEAT_CNT_REG);
4510 	if (val == info->last_heart_beat)
4511 		goto reset;
4512 
4513 	info->last_heart_beat = val;
4514 
4515 	val = bnxt_read_fw_status_reg(bp, BNXT_FW_RECOVERY_CNT_REG);
4516 	if (val != info->last_reset_counter)
4517 		goto reset;
4518 
4519 	info->last_reset_counter = val;
4520 
4521 	rte_eal_alarm_set(US_PER_MS * info->driver_polling_freq,
4522 			  bnxt_check_fw_health, (void *)bp);
4523 
4524 	return;
4525 reset:
4526 	/* Stop DMA to/from device */
4527 	bp->flags |= BNXT_FLAG_FATAL_ERROR;
4528 	bp->flags |= BNXT_FLAG_FW_RESET;
4529 
4530 	bnxt_stop_rxtx(bp);
4531 
4532 	PMD_DRV_LOG(ERR, "Detected FW dead condition\n");
4533 
4534 	if (bnxt_is_primary_func(bp))
4535 		wait_msec = info->primary_func_wait_period;
4536 	else
4537 		wait_msec = info->normal_func_wait_period;
4538 
4539 	rte_eal_alarm_set(US_PER_MS * wait_msec,
4540 			  bnxt_fw_reset_cb, (void *)bp);
4541 }
4542 
4543 void bnxt_schedule_fw_health_check(struct bnxt *bp)
4544 {
4545 	uint32_t polling_freq;
4546 
4547 	pthread_mutex_lock(&bp->health_check_lock);
4548 
4549 	if (!bnxt_is_recovery_enabled(bp))
4550 		goto done;
4551 
4552 	if (bp->flags & BNXT_FLAG_FW_HEALTH_CHECK_SCHEDULED)
4553 		goto done;
4554 
4555 	polling_freq = bp->recovery_info->driver_polling_freq;
4556 
4557 	rte_eal_alarm_set(US_PER_MS * polling_freq,
4558 			  bnxt_check_fw_health, (void *)bp);
4559 	bp->flags |= BNXT_FLAG_FW_HEALTH_CHECK_SCHEDULED;
4560 
4561 done:
4562 	pthread_mutex_unlock(&bp->health_check_lock);
4563 }
4564 
4565 static void bnxt_cancel_fw_health_check(struct bnxt *bp)
4566 {
4567 	rte_eal_alarm_cancel(bnxt_check_fw_health, (void *)bp);
4568 	bp->flags &= ~BNXT_FLAG_FW_HEALTH_CHECK_SCHEDULED;
4569 }
4570 
4571 static bool bnxt_vf_pciid(uint16_t device_id)
4572 {
4573 	switch (device_id) {
4574 	case BROADCOM_DEV_ID_57304_VF:
4575 	case BROADCOM_DEV_ID_57406_VF:
4576 	case BROADCOM_DEV_ID_5731X_VF:
4577 	case BROADCOM_DEV_ID_5741X_VF:
4578 	case BROADCOM_DEV_ID_57414_VF:
4579 	case BROADCOM_DEV_ID_STRATUS_NIC_VF1:
4580 	case BROADCOM_DEV_ID_STRATUS_NIC_VF2:
4581 	case BROADCOM_DEV_ID_58802_VF:
4582 	case BROADCOM_DEV_ID_57500_VF1:
4583 	case BROADCOM_DEV_ID_57500_VF2:
4584 	case BROADCOM_DEV_ID_58818_VF:
4585 		/* FALLTHROUGH */
4586 		return true;
4587 	default:
4588 		return false;
4589 	}
4590 }
4591 
4592 /* Phase 5 device */
4593 static bool bnxt_p5_device(uint16_t device_id)
4594 {
4595 	switch (device_id) {
4596 	case BROADCOM_DEV_ID_57508:
4597 	case BROADCOM_DEV_ID_57504:
4598 	case BROADCOM_DEV_ID_57502:
4599 	case BROADCOM_DEV_ID_57508_MF1:
4600 	case BROADCOM_DEV_ID_57504_MF1:
4601 	case BROADCOM_DEV_ID_57502_MF1:
4602 	case BROADCOM_DEV_ID_57508_MF2:
4603 	case BROADCOM_DEV_ID_57504_MF2:
4604 	case BROADCOM_DEV_ID_57502_MF2:
4605 	case BROADCOM_DEV_ID_57500_VF1:
4606 	case BROADCOM_DEV_ID_57500_VF2:
4607 	case BROADCOM_DEV_ID_58812:
4608 	case BROADCOM_DEV_ID_58814:
4609 	case BROADCOM_DEV_ID_58818:
4610 	case BROADCOM_DEV_ID_58818_VF:
4611 		/* FALLTHROUGH */
4612 		return true;
4613 	default:
4614 		return false;
4615 	}
4616 }
4617 
4618 bool bnxt_stratus_device(struct bnxt *bp)
4619 {
4620 	uint16_t device_id = bp->pdev->id.device_id;
4621 
4622 	switch (device_id) {
4623 	case BROADCOM_DEV_ID_STRATUS_NIC:
4624 	case BROADCOM_DEV_ID_STRATUS_NIC_VF1:
4625 	case BROADCOM_DEV_ID_STRATUS_NIC_VF2:
4626 		/* FALLTHROUGH */
4627 		return true;
4628 	default:
4629 		return false;
4630 	}
4631 }
4632 
4633 static int bnxt_map_pci_bars(struct rte_eth_dev *eth_dev)
4634 {
4635 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
4636 	struct bnxt *bp = eth_dev->data->dev_private;
4637 
4638 	/* enable device (incl. PCI PM wakeup), and bus-mastering */
4639 	bp->bar0 = (void *)pci_dev->mem_resource[0].addr;
4640 	bp->doorbell_base = (void *)pci_dev->mem_resource[2].addr;
4641 	if (!bp->bar0 || !bp->doorbell_base) {
4642 		PMD_DRV_LOG(ERR, "Unable to access Hardware\n");
4643 		return -ENODEV;
4644 	}
4645 
4646 	bp->eth_dev = eth_dev;
4647 	bp->pdev = pci_dev;
4648 
4649 	return 0;
4650 }
4651 
4652 static int bnxt_alloc_ctx_mem_blk(struct bnxt *bp,
4653 				  struct bnxt_ctx_pg_info *ctx_pg,
4654 				  uint32_t mem_size,
4655 				  const char *suffix,
4656 				  uint16_t idx)
4657 {
4658 	struct bnxt_ring_mem_info *rmem = &ctx_pg->ring_mem;
4659 	const struct rte_memzone *mz = NULL;
4660 	char mz_name[RTE_MEMZONE_NAMESIZE];
4661 	rte_iova_t mz_phys_addr;
4662 	uint64_t valid_bits = 0;
4663 	uint32_t sz;
4664 	int i;
4665 
4666 	if (!mem_size)
4667 		return 0;
4668 
4669 	rmem->nr_pages = RTE_ALIGN_MUL_CEIL(mem_size, BNXT_PAGE_SIZE) /
4670 			 BNXT_PAGE_SIZE;
4671 	rmem->page_size = BNXT_PAGE_SIZE;
4672 	rmem->pg_arr = ctx_pg->ctx_pg_arr;
4673 	rmem->dma_arr = ctx_pg->ctx_dma_arr;
4674 	rmem->flags = BNXT_RMEM_VALID_PTE_FLAG;
4675 
4676 	valid_bits = PTU_PTE_VALID;
4677 
4678 	if (rmem->nr_pages > 1) {
4679 		snprintf(mz_name, RTE_MEMZONE_NAMESIZE,
4680 			 "bnxt_ctx_pg_tbl%s_%x_%d",
4681 			 suffix, idx, bp->eth_dev->data->port_id);
4682 		mz_name[RTE_MEMZONE_NAMESIZE - 1] = 0;
4683 		mz = rte_memzone_lookup(mz_name);
4684 		if (!mz) {
4685 			mz = rte_memzone_reserve_aligned(mz_name,
4686 						rmem->nr_pages * 8,
4687 						bp->eth_dev->device->numa_node,
4688 						RTE_MEMZONE_2MB |
4689 						RTE_MEMZONE_SIZE_HINT_ONLY |
4690 						RTE_MEMZONE_IOVA_CONTIG,
4691 						BNXT_PAGE_SIZE);
4692 			if (mz == NULL)
4693 				return -ENOMEM;
4694 		}
4695 
4696 		memset(mz->addr, 0, mz->len);
4697 		mz_phys_addr = mz->iova;
4698 
4699 		rmem->pg_tbl = mz->addr;
4700 		rmem->pg_tbl_map = mz_phys_addr;
4701 		rmem->pg_tbl_mz = mz;
4702 	}
4703 
4704 	snprintf(mz_name, RTE_MEMZONE_NAMESIZE, "bnxt_ctx_%s_%x_%d",
4705 		 suffix, idx, bp->eth_dev->data->port_id);
4706 	mz = rte_memzone_lookup(mz_name);
4707 	if (!mz) {
4708 		mz = rte_memzone_reserve_aligned(mz_name,
4709 						 mem_size,
4710 						 bp->eth_dev->device->numa_node,
4711 						 RTE_MEMZONE_1GB |
4712 						 RTE_MEMZONE_SIZE_HINT_ONLY |
4713 						 RTE_MEMZONE_IOVA_CONTIG,
4714 						 BNXT_PAGE_SIZE);
4715 		if (mz == NULL)
4716 			return -ENOMEM;
4717 	}
4718 
4719 	memset(mz->addr, 0, mz->len);
4720 	mz_phys_addr = mz->iova;
4721 
4722 	for (sz = 0, i = 0; sz < mem_size; sz += BNXT_PAGE_SIZE, i++) {
4723 		rmem->pg_arr[i] = ((char *)mz->addr) + sz;
4724 		rmem->dma_arr[i] = mz_phys_addr + sz;
4725 
4726 		if (rmem->nr_pages > 1) {
4727 			if (i == rmem->nr_pages - 2 &&
4728 			    (rmem->flags & BNXT_RMEM_RING_PTE_FLAG))
4729 				valid_bits |= PTU_PTE_NEXT_TO_LAST;
4730 			else if (i == rmem->nr_pages - 1 &&
4731 				 (rmem->flags & BNXT_RMEM_RING_PTE_FLAG))
4732 				valid_bits |= PTU_PTE_LAST;
4733 
4734 			rmem->pg_tbl[i] = rte_cpu_to_le_64(rmem->dma_arr[i] |
4735 							   valid_bits);
4736 		}
4737 	}
4738 
4739 	rmem->mz = mz;
4740 	if (rmem->vmem_size)
4741 		rmem->vmem = (void **)mz->addr;
4742 	rmem->dma_arr[0] = mz_phys_addr;
4743 	return 0;
4744 }
4745 
4746 static void bnxt_free_ctx_mem(struct bnxt *bp)
4747 {
4748 	int i;
4749 
4750 	if (!bp->ctx || !(bp->ctx->flags & BNXT_CTX_FLAG_INITED))
4751 		return;
4752 
4753 	bp->ctx->flags &= ~BNXT_CTX_FLAG_INITED;
4754 	rte_memzone_free(bp->ctx->qp_mem.ring_mem.mz);
4755 	rte_memzone_free(bp->ctx->srq_mem.ring_mem.mz);
4756 	rte_memzone_free(bp->ctx->cq_mem.ring_mem.mz);
4757 	rte_memzone_free(bp->ctx->vnic_mem.ring_mem.mz);
4758 	rte_memzone_free(bp->ctx->stat_mem.ring_mem.mz);
4759 	rte_memzone_free(bp->ctx->qp_mem.ring_mem.pg_tbl_mz);
4760 	rte_memzone_free(bp->ctx->srq_mem.ring_mem.pg_tbl_mz);
4761 	rte_memzone_free(bp->ctx->cq_mem.ring_mem.pg_tbl_mz);
4762 	rte_memzone_free(bp->ctx->vnic_mem.ring_mem.pg_tbl_mz);
4763 	rte_memzone_free(bp->ctx->stat_mem.ring_mem.pg_tbl_mz);
4764 
4765 	for (i = 0; i < bp->ctx->tqm_fp_rings_count + 1; i++) {
4766 		if (bp->ctx->tqm_mem[i])
4767 			rte_memzone_free(bp->ctx->tqm_mem[i]->ring_mem.mz);
4768 	}
4769 
4770 	rte_free(bp->ctx);
4771 	bp->ctx = NULL;
4772 }
4773 
4774 #define bnxt_roundup(x, y)   ((((x) + ((y) - 1)) / (y)) * (y))
4775 
4776 #define min_t(type, x, y) ({                    \
4777 	type __min1 = (x);                      \
4778 	type __min2 = (y);                      \
4779 	__min1 < __min2 ? __min1 : __min2; })
4780 
4781 #define max_t(type, x, y) ({                    \
4782 	type __max1 = (x);                      \
4783 	type __max2 = (y);                      \
4784 	__max1 > __max2 ? __max1 : __max2; })
4785 
4786 #define clamp_t(type, _x, min, max)     min_t(type, max_t(type, _x, min), max)
4787 
4788 int bnxt_alloc_ctx_mem(struct bnxt *bp)
4789 {
4790 	struct bnxt_ctx_pg_info *ctx_pg;
4791 	struct bnxt_ctx_mem_info *ctx;
4792 	uint32_t mem_size, ena, entries;
4793 	uint32_t entries_sp, min;
4794 	int i, rc;
4795 
4796 	rc = bnxt_hwrm_func_backing_store_qcaps(bp);
4797 	if (rc) {
4798 		PMD_DRV_LOG(ERR, "Query context mem capability failed\n");
4799 		return rc;
4800 	}
4801 	ctx = bp->ctx;
4802 	if (!ctx || (ctx->flags & BNXT_CTX_FLAG_INITED))
4803 		return 0;
4804 
4805 	ctx_pg = &ctx->qp_mem;
4806 	ctx_pg->entries = ctx->qp_min_qp1_entries + ctx->qp_max_l2_entries;
4807 	if (ctx->qp_entry_size) {
4808 		mem_size = ctx->qp_entry_size * ctx_pg->entries;
4809 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "qp_mem", 0);
4810 		if (rc)
4811 			return rc;
4812 	}
4813 
4814 	ctx_pg = &ctx->srq_mem;
4815 	ctx_pg->entries = ctx->srq_max_l2_entries;
4816 	if (ctx->srq_entry_size) {
4817 		mem_size = ctx->srq_entry_size * ctx_pg->entries;
4818 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "srq_mem", 0);
4819 		if (rc)
4820 			return rc;
4821 	}
4822 
4823 	ctx_pg = &ctx->cq_mem;
4824 	ctx_pg->entries = ctx->cq_max_l2_entries;
4825 	if (ctx->cq_entry_size) {
4826 		mem_size = ctx->cq_entry_size * ctx_pg->entries;
4827 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "cq_mem", 0);
4828 		if (rc)
4829 			return rc;
4830 	}
4831 
4832 	ctx_pg = &ctx->vnic_mem;
4833 	ctx_pg->entries = ctx->vnic_max_vnic_entries +
4834 		ctx->vnic_max_ring_table_entries;
4835 	if (ctx->vnic_entry_size) {
4836 		mem_size = ctx->vnic_entry_size * ctx_pg->entries;
4837 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "vnic_mem", 0);
4838 		if (rc)
4839 			return rc;
4840 	}
4841 
4842 	ctx_pg = &ctx->stat_mem;
4843 	ctx_pg->entries = ctx->stat_max_entries;
4844 	if (ctx->stat_entry_size) {
4845 		mem_size = ctx->stat_entry_size * ctx_pg->entries;
4846 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "stat_mem", 0);
4847 		if (rc)
4848 			return rc;
4849 	}
4850 
4851 	min = ctx->tqm_min_entries_per_ring;
4852 
4853 	entries_sp = ctx->qp_max_l2_entries +
4854 		     ctx->vnic_max_vnic_entries +
4855 		     2 * ctx->qp_min_qp1_entries + min;
4856 	entries_sp = bnxt_roundup(entries_sp, ctx->tqm_entries_multiple);
4857 
4858 	entries = ctx->qp_max_l2_entries + ctx->qp_min_qp1_entries;
4859 	entries = bnxt_roundup(entries, ctx->tqm_entries_multiple);
4860 	entries = clamp_t(uint32_t, entries, min,
4861 			  ctx->tqm_max_entries_per_ring);
4862 	for (i = 0, ena = 0; i < ctx->tqm_fp_rings_count + 1; i++) {
4863 		/* i=0 is for TQM_SP. i=1 to i=8 applies to RING0 to RING7.
4864 		 * i > 8 is other ext rings.
4865 		 */
4866 		ctx_pg = ctx->tqm_mem[i];
4867 		ctx_pg->entries = i ? entries : entries_sp;
4868 		if (ctx->tqm_entry_size) {
4869 			mem_size = ctx->tqm_entry_size * ctx_pg->entries;
4870 			rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size,
4871 						    "tqm_mem", i);
4872 			if (rc)
4873 				return rc;
4874 		}
4875 		if (i < BNXT_MAX_TQM_LEGACY_RINGS)
4876 			ena |= HWRM_FUNC_BACKING_STORE_CFG_INPUT_ENABLES_TQM_SP << i;
4877 		else
4878 			ena |= HWRM_FUNC_BACKING_STORE_CFG_INPUT_ENABLES_TQM_RING8;
4879 	}
4880 
4881 	ena |= FUNC_BACKING_STORE_CFG_INPUT_DFLT_ENABLES;
4882 	rc = bnxt_hwrm_func_backing_store_cfg(bp, ena);
4883 	if (rc)
4884 		PMD_DRV_LOG(ERR,
4885 			    "Failed to configure context mem: rc = %d\n", rc);
4886 	else
4887 		ctx->flags |= BNXT_CTX_FLAG_INITED;
4888 
4889 	return rc;
4890 }
4891 
4892 static int bnxt_alloc_stats_mem(struct bnxt *bp)
4893 {
4894 	struct rte_pci_device *pci_dev = bp->pdev;
4895 	char mz_name[RTE_MEMZONE_NAMESIZE];
4896 	const struct rte_memzone *mz = NULL;
4897 	uint32_t total_alloc_len;
4898 	rte_iova_t mz_phys_addr;
4899 
4900 	if (pci_dev->id.device_id == BROADCOM_DEV_ID_NS2)
4901 		return 0;
4902 
4903 	snprintf(mz_name, RTE_MEMZONE_NAMESIZE,
4904 		 "bnxt_" PCI_PRI_FMT "-%s", pci_dev->addr.domain,
4905 		 pci_dev->addr.bus, pci_dev->addr.devid,
4906 		 pci_dev->addr.function, "rx_port_stats");
4907 	mz_name[RTE_MEMZONE_NAMESIZE - 1] = 0;
4908 	mz = rte_memzone_lookup(mz_name);
4909 	total_alloc_len =
4910 		RTE_CACHE_LINE_ROUNDUP(sizeof(struct rx_port_stats) +
4911 				       sizeof(struct rx_port_stats_ext) + 512);
4912 	if (!mz) {
4913 		mz = rte_memzone_reserve(mz_name, total_alloc_len,
4914 					 SOCKET_ID_ANY,
4915 					 RTE_MEMZONE_2MB |
4916 					 RTE_MEMZONE_SIZE_HINT_ONLY |
4917 					 RTE_MEMZONE_IOVA_CONTIG);
4918 		if (mz == NULL)
4919 			return -ENOMEM;
4920 	}
4921 	memset(mz->addr, 0, mz->len);
4922 	mz_phys_addr = mz->iova;
4923 
4924 	bp->rx_mem_zone = (const void *)mz;
4925 	bp->hw_rx_port_stats = mz->addr;
4926 	bp->hw_rx_port_stats_map = mz_phys_addr;
4927 
4928 	snprintf(mz_name, RTE_MEMZONE_NAMESIZE,
4929 		 "bnxt_" PCI_PRI_FMT "-%s", pci_dev->addr.domain,
4930 		 pci_dev->addr.bus, pci_dev->addr.devid,
4931 		 pci_dev->addr.function, "tx_port_stats");
4932 	mz_name[RTE_MEMZONE_NAMESIZE - 1] = 0;
4933 	mz = rte_memzone_lookup(mz_name);
4934 	total_alloc_len =
4935 		RTE_CACHE_LINE_ROUNDUP(sizeof(struct tx_port_stats) +
4936 				       sizeof(struct tx_port_stats_ext) + 512);
4937 	if (!mz) {
4938 		mz = rte_memzone_reserve(mz_name,
4939 					 total_alloc_len,
4940 					 SOCKET_ID_ANY,
4941 					 RTE_MEMZONE_2MB |
4942 					 RTE_MEMZONE_SIZE_HINT_ONLY |
4943 					 RTE_MEMZONE_IOVA_CONTIG);
4944 		if (mz == NULL)
4945 			return -ENOMEM;
4946 	}
4947 	memset(mz->addr, 0, mz->len);
4948 	mz_phys_addr = mz->iova;
4949 
4950 	bp->tx_mem_zone = (const void *)mz;
4951 	bp->hw_tx_port_stats = mz->addr;
4952 	bp->hw_tx_port_stats_map = mz_phys_addr;
4953 	bp->flags |= BNXT_FLAG_PORT_STATS;
4954 
4955 	/* Display extended statistics if FW supports it */
4956 	if (bp->hwrm_spec_code < HWRM_SPEC_CODE_1_8_4 ||
4957 	    bp->hwrm_spec_code == HWRM_SPEC_CODE_1_9_0 ||
4958 	    !(bp->flags & BNXT_FLAG_EXT_STATS_SUPPORTED))
4959 		return 0;
4960 
4961 	bp->hw_rx_port_stats_ext = (void *)
4962 		((uint8_t *)bp->hw_rx_port_stats +
4963 		 sizeof(struct rx_port_stats));
4964 	bp->hw_rx_port_stats_ext_map = bp->hw_rx_port_stats_map +
4965 		sizeof(struct rx_port_stats);
4966 	bp->flags |= BNXT_FLAG_EXT_RX_PORT_STATS;
4967 
4968 	if (bp->hwrm_spec_code < HWRM_SPEC_CODE_1_9_2 ||
4969 	    bp->flags & BNXT_FLAG_EXT_STATS_SUPPORTED) {
4970 		bp->hw_tx_port_stats_ext = (void *)
4971 			((uint8_t *)bp->hw_tx_port_stats +
4972 			 sizeof(struct tx_port_stats));
4973 		bp->hw_tx_port_stats_ext_map =
4974 			bp->hw_tx_port_stats_map +
4975 			sizeof(struct tx_port_stats);
4976 		bp->flags |= BNXT_FLAG_EXT_TX_PORT_STATS;
4977 	}
4978 
4979 	return 0;
4980 }
4981 
4982 static int bnxt_setup_mac_addr(struct rte_eth_dev *eth_dev)
4983 {
4984 	struct bnxt *bp = eth_dev->data->dev_private;
4985 	int rc = 0;
4986 
4987 	eth_dev->data->mac_addrs = rte_zmalloc("bnxt_mac_addr_tbl",
4988 					       RTE_ETHER_ADDR_LEN *
4989 					       bp->max_l2_ctx,
4990 					       0);
4991 	if (eth_dev->data->mac_addrs == NULL) {
4992 		PMD_DRV_LOG(ERR, "Failed to alloc MAC addr tbl\n");
4993 		return -ENOMEM;
4994 	}
4995 
4996 	if (!BNXT_HAS_DFLT_MAC_SET(bp)) {
4997 		if (BNXT_PF(bp))
4998 			return -EINVAL;
4999 
5000 		/* Generate a random MAC address, if none was assigned by PF */
5001 		PMD_DRV_LOG(INFO, "VF MAC address not assigned by Host PF\n");
5002 		bnxt_eth_hw_addr_random(bp->mac_addr);
5003 		PMD_DRV_LOG(INFO,
5004 			    "Assign random MAC:" RTE_ETHER_ADDR_PRT_FMT "\n",
5005 			    bp->mac_addr[0], bp->mac_addr[1], bp->mac_addr[2],
5006 			    bp->mac_addr[3], bp->mac_addr[4], bp->mac_addr[5]);
5007 
5008 		rc = bnxt_hwrm_set_mac(bp);
5009 		if (rc)
5010 			return rc;
5011 	}
5012 
5013 	/* Copy the permanent MAC from the FUNC_QCAPS response */
5014 	memcpy(&eth_dev->data->mac_addrs[0], bp->mac_addr, RTE_ETHER_ADDR_LEN);
5015 
5016 	return rc;
5017 }
5018 
5019 static int bnxt_restore_dflt_mac(struct bnxt *bp)
5020 {
5021 	int rc = 0;
5022 
5023 	/* MAC is already configured in FW */
5024 	if (BNXT_HAS_DFLT_MAC_SET(bp))
5025 		return 0;
5026 
5027 	/* Restore the old MAC configured */
5028 	rc = bnxt_hwrm_set_mac(bp);
5029 	if (rc)
5030 		PMD_DRV_LOG(ERR, "Failed to restore MAC address\n");
5031 
5032 	return rc;
5033 }
5034 
5035 static void bnxt_config_vf_req_fwd(struct bnxt *bp)
5036 {
5037 	if (!BNXT_PF(bp))
5038 		return;
5039 
5040 	memset(bp->pf->vf_req_fwd, 0, sizeof(bp->pf->vf_req_fwd));
5041 
5042 	if (!(bp->fw_cap & BNXT_FW_CAP_LINK_ADMIN))
5043 		BNXT_HWRM_CMD_TO_FORWARD(HWRM_PORT_PHY_QCFG);
5044 	BNXT_HWRM_CMD_TO_FORWARD(HWRM_FUNC_CFG);
5045 	BNXT_HWRM_CMD_TO_FORWARD(HWRM_FUNC_VF_CFG);
5046 	BNXT_HWRM_CMD_TO_FORWARD(HWRM_CFA_L2_FILTER_ALLOC);
5047 	BNXT_HWRM_CMD_TO_FORWARD(HWRM_OEM_CMD);
5048 }
5049 
5050 struct bnxt *
5051 bnxt_get_bp(uint16_t port)
5052 {
5053 	struct bnxt *bp;
5054 	struct rte_eth_dev *dev;
5055 
5056 	if (!rte_eth_dev_is_valid_port(port)) {
5057 		PMD_DRV_LOG(ERR, "Invalid port %d\n", port);
5058 		return NULL;
5059 	}
5060 
5061 	dev = &rte_eth_devices[port];
5062 	if (!is_bnxt_supported(dev)) {
5063 		PMD_DRV_LOG(ERR, "Device %d not supported\n", port);
5064 		return NULL;
5065 	}
5066 
5067 	bp = (struct bnxt *)dev->data->dev_private;
5068 	if (!BNXT_TRUFLOW_EN(bp)) {
5069 		PMD_DRV_LOG(ERR, "TRUFLOW not enabled\n");
5070 		return NULL;
5071 	}
5072 
5073 	return bp;
5074 }
5075 
5076 uint16_t
5077 bnxt_get_svif(uint16_t port_id, bool func_svif,
5078 	      enum bnxt_ulp_intf_type type)
5079 {
5080 	struct rte_eth_dev *eth_dev;
5081 	struct bnxt *bp;
5082 
5083 	eth_dev = &rte_eth_devices[port_id];
5084 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
5085 		struct bnxt_representor *vfr = eth_dev->data->dev_private;
5086 		if (!vfr)
5087 			return 0;
5088 
5089 		if (type == BNXT_ULP_INTF_TYPE_VF_REP)
5090 			return vfr->svif;
5091 
5092 		eth_dev = vfr->parent_dev;
5093 	}
5094 
5095 	bp = eth_dev->data->dev_private;
5096 
5097 	return func_svif ? bp->func_svif : bp->port_svif;
5098 }
5099 
5100 void
5101 bnxt_get_iface_mac(uint16_t port, enum bnxt_ulp_intf_type type,
5102 		   uint8_t *mac, uint8_t *parent_mac)
5103 {
5104 	struct rte_eth_dev *eth_dev;
5105 	struct bnxt *bp;
5106 
5107 	if (type != BNXT_ULP_INTF_TYPE_TRUSTED_VF &&
5108 	    type != BNXT_ULP_INTF_TYPE_PF)
5109 		return;
5110 
5111 	eth_dev = &rte_eth_devices[port];
5112 	bp = eth_dev->data->dev_private;
5113 	memcpy(mac, bp->mac_addr, RTE_ETHER_ADDR_LEN);
5114 
5115 	if (type == BNXT_ULP_INTF_TYPE_TRUSTED_VF)
5116 		memcpy(parent_mac, bp->parent->mac_addr, RTE_ETHER_ADDR_LEN);
5117 }
5118 
5119 uint16_t
5120 bnxt_get_parent_vnic_id(uint16_t port, enum bnxt_ulp_intf_type type)
5121 {
5122 	struct rte_eth_dev *eth_dev;
5123 	struct bnxt *bp;
5124 
5125 	if (type != BNXT_ULP_INTF_TYPE_TRUSTED_VF)
5126 		return 0;
5127 
5128 	eth_dev = &rte_eth_devices[port];
5129 	bp = eth_dev->data->dev_private;
5130 
5131 	return bp->parent->vnic;
5132 }
5133 uint16_t
5134 bnxt_get_vnic_id(uint16_t port, enum bnxt_ulp_intf_type type)
5135 {
5136 	struct rte_eth_dev *eth_dev;
5137 	struct bnxt_vnic_info *vnic;
5138 	struct bnxt *bp;
5139 
5140 	eth_dev = &rte_eth_devices[port];
5141 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
5142 		struct bnxt_representor *vfr = eth_dev->data->dev_private;
5143 		if (!vfr)
5144 			return 0;
5145 
5146 		if (type == BNXT_ULP_INTF_TYPE_VF_REP)
5147 			return vfr->dflt_vnic_id;
5148 
5149 		eth_dev = vfr->parent_dev;
5150 	}
5151 
5152 	bp = eth_dev->data->dev_private;
5153 
5154 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
5155 
5156 	return vnic->fw_vnic_id;
5157 }
5158 
5159 uint16_t
5160 bnxt_get_fw_func_id(uint16_t port, enum bnxt_ulp_intf_type type)
5161 {
5162 	struct rte_eth_dev *eth_dev;
5163 	struct bnxt *bp;
5164 
5165 	eth_dev = &rte_eth_devices[port];
5166 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
5167 		struct bnxt_representor *vfr = eth_dev->data->dev_private;
5168 		if (!vfr)
5169 			return 0;
5170 
5171 		if (type == BNXT_ULP_INTF_TYPE_VF_REP)
5172 			return vfr->fw_fid;
5173 
5174 		eth_dev = vfr->parent_dev;
5175 	}
5176 
5177 	bp = eth_dev->data->dev_private;
5178 
5179 	return bp->fw_fid;
5180 }
5181 
5182 enum bnxt_ulp_intf_type
5183 bnxt_get_interface_type(uint16_t port)
5184 {
5185 	struct rte_eth_dev *eth_dev;
5186 	struct bnxt *bp;
5187 
5188 	eth_dev = &rte_eth_devices[port];
5189 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev))
5190 		return BNXT_ULP_INTF_TYPE_VF_REP;
5191 
5192 	bp = eth_dev->data->dev_private;
5193 	if (BNXT_PF(bp))
5194 		return BNXT_ULP_INTF_TYPE_PF;
5195 	else if (BNXT_VF_IS_TRUSTED(bp))
5196 		return BNXT_ULP_INTF_TYPE_TRUSTED_VF;
5197 	else if (BNXT_VF(bp))
5198 		return BNXT_ULP_INTF_TYPE_VF;
5199 
5200 	return BNXT_ULP_INTF_TYPE_INVALID;
5201 }
5202 
5203 uint16_t
5204 bnxt_get_phy_port_id(uint16_t port_id)
5205 {
5206 	struct bnxt_representor *vfr;
5207 	struct rte_eth_dev *eth_dev;
5208 	struct bnxt *bp;
5209 
5210 	eth_dev = &rte_eth_devices[port_id];
5211 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
5212 		vfr = eth_dev->data->dev_private;
5213 		if (!vfr)
5214 			return 0;
5215 
5216 		eth_dev = vfr->parent_dev;
5217 	}
5218 
5219 	bp = eth_dev->data->dev_private;
5220 
5221 	return BNXT_PF(bp) ? bp->pf->port_id : bp->parent->port_id;
5222 }
5223 
5224 uint16_t
5225 bnxt_get_parif(uint16_t port_id, enum bnxt_ulp_intf_type type)
5226 {
5227 	struct rte_eth_dev *eth_dev;
5228 	struct bnxt *bp;
5229 
5230 	eth_dev = &rte_eth_devices[port_id];
5231 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
5232 		struct bnxt_representor *vfr = eth_dev->data->dev_private;
5233 		if (!vfr)
5234 			return 0;
5235 
5236 		if (type == BNXT_ULP_INTF_TYPE_VF_REP)
5237 			return vfr->fw_fid - 1;
5238 
5239 		eth_dev = vfr->parent_dev;
5240 	}
5241 
5242 	bp = eth_dev->data->dev_private;
5243 
5244 	return BNXT_PF(bp) ? bp->fw_fid - 1 : bp->parent->fid - 1;
5245 }
5246 
5247 uint16_t
5248 bnxt_get_vport(uint16_t port_id)
5249 {
5250 	return (1 << bnxt_get_phy_port_id(port_id));
5251 }
5252 
5253 static void bnxt_alloc_error_recovery_info(struct bnxt *bp)
5254 {
5255 	struct bnxt_error_recovery_info *info = bp->recovery_info;
5256 
5257 	if (info) {
5258 		if (!(bp->fw_cap & BNXT_FW_CAP_HCOMM_FW_STATUS))
5259 			memset(info, 0, sizeof(*info));
5260 		return;
5261 	}
5262 
5263 	if (!(bp->fw_cap & BNXT_FW_CAP_ERROR_RECOVERY))
5264 		return;
5265 
5266 	info = rte_zmalloc("bnxt_hwrm_error_recovery_qcfg",
5267 			   sizeof(*info), 0);
5268 	if (!info)
5269 		bp->fw_cap &= ~BNXT_FW_CAP_ERROR_RECOVERY;
5270 
5271 	bp->recovery_info = info;
5272 }
5273 
5274 static void bnxt_check_fw_status(struct bnxt *bp)
5275 {
5276 	uint32_t fw_status;
5277 
5278 	if (!(bp->recovery_info &&
5279 	      (bp->fw_cap & BNXT_FW_CAP_HCOMM_FW_STATUS)))
5280 		return;
5281 
5282 	fw_status = bnxt_read_fw_status_reg(bp, BNXT_FW_STATUS_REG);
5283 	if (fw_status != BNXT_FW_STATUS_HEALTHY)
5284 		PMD_DRV_LOG(ERR, "Firmware not responding, status: %#x\n",
5285 			    fw_status);
5286 }
5287 
5288 static int bnxt_map_hcomm_fw_status_reg(struct bnxt *bp)
5289 {
5290 	struct bnxt_error_recovery_info *info = bp->recovery_info;
5291 	uint32_t status_loc;
5292 	uint32_t sig_ver;
5293 
5294 	rte_write32(HCOMM_STATUS_STRUCT_LOC, (uint8_t *)bp->bar0 +
5295 		    BNXT_GRCPF_REG_WINDOW_BASE_OUT + 4);
5296 	sig_ver = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
5297 				   BNXT_GRCP_WINDOW_2_BASE +
5298 				   offsetof(struct hcomm_status,
5299 					    sig_ver)));
5300 	/* If the signature is absent, then FW does not support this feature */
5301 	if ((sig_ver & HCOMM_STATUS_SIGNATURE_MASK) !=
5302 	    HCOMM_STATUS_SIGNATURE_VAL)
5303 		return 0;
5304 
5305 	if (!info) {
5306 		info = rte_zmalloc("bnxt_hwrm_error_recovery_qcfg",
5307 				   sizeof(*info), 0);
5308 		if (!info)
5309 			return -ENOMEM;
5310 		bp->recovery_info = info;
5311 	} else {
5312 		memset(info, 0, sizeof(*info));
5313 	}
5314 
5315 	status_loc = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
5316 				      BNXT_GRCP_WINDOW_2_BASE +
5317 				      offsetof(struct hcomm_status,
5318 					       fw_status_loc)));
5319 
5320 	/* Only pre-map the FW health status GRC register */
5321 	if (BNXT_FW_STATUS_REG_TYPE(status_loc) != BNXT_FW_STATUS_REG_TYPE_GRC)
5322 		return 0;
5323 
5324 	info->status_regs[BNXT_FW_STATUS_REG] = status_loc;
5325 	info->mapped_status_regs[BNXT_FW_STATUS_REG] =
5326 		BNXT_GRCP_WINDOW_2_BASE + (status_loc & BNXT_GRCP_OFFSET_MASK);
5327 
5328 	rte_write32((status_loc & BNXT_GRCP_BASE_MASK), (uint8_t *)bp->bar0 +
5329 		    BNXT_GRCPF_REG_WINDOW_BASE_OUT + 4);
5330 
5331 	bp->fw_cap |= BNXT_FW_CAP_HCOMM_FW_STATUS;
5332 
5333 	return 0;
5334 }
5335 
5336 /* This function gets the FW version along with the
5337  * capabilities(MAX and current) of the function, vnic,
5338  * error recovery, phy and other chip related info
5339  */
5340 static int bnxt_get_config(struct bnxt *bp)
5341 {
5342 	uint16_t mtu;
5343 	int rc = 0;
5344 
5345 	bp->fw_cap = 0;
5346 
5347 	rc = bnxt_map_hcomm_fw_status_reg(bp);
5348 	if (rc)
5349 		return rc;
5350 
5351 	rc = bnxt_hwrm_ver_get(bp, DFLT_HWRM_CMD_TIMEOUT);
5352 	if (rc) {
5353 		bnxt_check_fw_status(bp);
5354 		return rc;
5355 	}
5356 
5357 	rc = bnxt_hwrm_func_reset(bp);
5358 	if (rc)
5359 		return -EIO;
5360 
5361 	rc = bnxt_hwrm_vnic_qcaps(bp);
5362 	if (rc)
5363 		return rc;
5364 
5365 	rc = bnxt_hwrm_queue_qportcfg(bp);
5366 	if (rc)
5367 		return rc;
5368 
5369 	/* Get the MAX capabilities for this function.
5370 	 * This function also allocates context memory for TQM rings and
5371 	 * informs the firmware about this allocated backing store memory.
5372 	 */
5373 	rc = bnxt_hwrm_func_qcaps(bp);
5374 	if (rc)
5375 		return rc;
5376 
5377 	rc = bnxt_hwrm_func_qcfg(bp, &mtu);
5378 	if (rc)
5379 		return rc;
5380 
5381 	rc = bnxt_hwrm_cfa_adv_flow_mgmt_qcaps(bp);
5382 	if (rc)
5383 		return rc;
5384 
5385 	bnxt_hwrm_port_mac_qcfg(bp);
5386 
5387 	bnxt_hwrm_parent_pf_qcfg(bp);
5388 
5389 	bnxt_hwrm_port_phy_qcaps(bp);
5390 
5391 	bnxt_alloc_error_recovery_info(bp);
5392 	/* Get the adapter error recovery support info */
5393 	rc = bnxt_hwrm_error_recovery_qcfg(bp);
5394 	if (rc)
5395 		bp->fw_cap &= ~BNXT_FW_CAP_ERROR_RECOVERY;
5396 
5397 	bnxt_hwrm_port_led_qcaps(bp);
5398 
5399 	return 0;
5400 }
5401 
5402 static int
5403 bnxt_init_locks(struct bnxt *bp)
5404 {
5405 	int err;
5406 
5407 	err = pthread_mutex_init(&bp->flow_lock, NULL);
5408 	if (err) {
5409 		PMD_DRV_LOG(ERR, "Unable to initialize flow_lock\n");
5410 		return err;
5411 	}
5412 
5413 	err = pthread_mutex_init(&bp->def_cp_lock, NULL);
5414 	if (err) {
5415 		PMD_DRV_LOG(ERR, "Unable to initialize def_cp_lock\n");
5416 		return err;
5417 	}
5418 
5419 	err = pthread_mutex_init(&bp->health_check_lock, NULL);
5420 	if (err) {
5421 		PMD_DRV_LOG(ERR, "Unable to initialize health_check_lock\n");
5422 		return err;
5423 	}
5424 
5425 	err = pthread_mutex_init(&bp->err_recovery_lock, NULL);
5426 	if (err)
5427 		PMD_DRV_LOG(ERR, "Unable to initialize err_recovery_lock\n");
5428 
5429 	return err;
5430 }
5431 
5432 static int bnxt_init_resources(struct bnxt *bp, bool reconfig_dev)
5433 {
5434 	int rc = 0;
5435 
5436 	rc = bnxt_get_config(bp);
5437 	if (rc)
5438 		return rc;
5439 
5440 	if (!reconfig_dev) {
5441 		rc = bnxt_setup_mac_addr(bp->eth_dev);
5442 		if (rc)
5443 			return rc;
5444 	} else {
5445 		rc = bnxt_restore_dflt_mac(bp);
5446 		if (rc)
5447 			return rc;
5448 	}
5449 
5450 	bnxt_config_vf_req_fwd(bp);
5451 
5452 	rc = bnxt_hwrm_func_driver_register(bp);
5453 	if (rc) {
5454 		PMD_DRV_LOG(ERR, "Failed to register driver");
5455 		return -EBUSY;
5456 	}
5457 
5458 	if (BNXT_PF(bp)) {
5459 		if (bp->pdev->max_vfs) {
5460 			rc = bnxt_hwrm_allocate_vfs(bp, bp->pdev->max_vfs);
5461 			if (rc) {
5462 				PMD_DRV_LOG(ERR, "Failed to allocate VFs\n");
5463 				return rc;
5464 			}
5465 		} else {
5466 			rc = bnxt_hwrm_allocate_pf_only(bp);
5467 			if (rc) {
5468 				PMD_DRV_LOG(ERR,
5469 					    "Failed to allocate PF resources");
5470 				return rc;
5471 			}
5472 		}
5473 	}
5474 
5475 	rc = bnxt_alloc_mem(bp, reconfig_dev);
5476 	if (rc)
5477 		return rc;
5478 
5479 	rc = bnxt_setup_int(bp);
5480 	if (rc)
5481 		return rc;
5482 
5483 	rc = bnxt_request_int(bp);
5484 	if (rc)
5485 		return rc;
5486 
5487 	rc = bnxt_init_ctx_mem(bp);
5488 	if (rc) {
5489 		PMD_DRV_LOG(ERR, "Failed to init adv_flow_counters\n");
5490 		return rc;
5491 	}
5492 
5493 	return 0;
5494 }
5495 
5496 static int
5497 bnxt_parse_devarg_accum_stats(__rte_unused const char *key,
5498 			      const char *value, void *opaque_arg)
5499 {
5500 	struct bnxt *bp = opaque_arg;
5501 	unsigned long accum_stats;
5502 	char *end = NULL;
5503 
5504 	if (!value || !opaque_arg) {
5505 		PMD_DRV_LOG(ERR,
5506 			    "Invalid parameter passed to accum-stats devargs.\n");
5507 		return -EINVAL;
5508 	}
5509 
5510 	accum_stats = strtoul(value, &end, 10);
5511 	if (end == NULL || *end != '\0' ||
5512 	    (accum_stats == ULONG_MAX && errno == ERANGE)) {
5513 		PMD_DRV_LOG(ERR,
5514 			    "Invalid parameter passed to accum-stats devargs.\n");
5515 		return -EINVAL;
5516 	}
5517 
5518 	if (BNXT_DEVARG_ACCUM_STATS_INVALID(accum_stats)) {
5519 		PMD_DRV_LOG(ERR,
5520 			    "Invalid value passed to accum-stats devargs.\n");
5521 		return -EINVAL;
5522 	}
5523 
5524 	if (accum_stats) {
5525 		bp->flags2 |= BNXT_FLAGS2_ACCUM_STATS_EN;
5526 		PMD_DRV_LOG(INFO, "Host-based accum-stats feature enabled.\n");
5527 	} else {
5528 		bp->flags2 &= ~BNXT_FLAGS2_ACCUM_STATS_EN;
5529 		PMD_DRV_LOG(INFO, "Host-based accum-stats feature disabled.\n");
5530 	}
5531 
5532 	return 0;
5533 }
5534 
5535 static int
5536 bnxt_parse_devarg_flow_xstat(__rte_unused const char *key,
5537 			     const char *value, void *opaque_arg)
5538 {
5539 	struct bnxt *bp = opaque_arg;
5540 	unsigned long flow_xstat;
5541 	char *end = NULL;
5542 
5543 	if (!value || !opaque_arg) {
5544 		PMD_DRV_LOG(ERR,
5545 			    "Invalid parameter passed to flow_xstat devarg.\n");
5546 		return -EINVAL;
5547 	}
5548 
5549 	flow_xstat = strtoul(value, &end, 10);
5550 	if (end == NULL || *end != '\0' ||
5551 	    (flow_xstat == ULONG_MAX && errno == ERANGE)) {
5552 		PMD_DRV_LOG(ERR,
5553 			    "Invalid parameter passed to flow_xstat devarg.\n");
5554 		return -EINVAL;
5555 	}
5556 
5557 	if (BNXT_DEVARG_FLOW_XSTAT_INVALID(flow_xstat)) {
5558 		PMD_DRV_LOG(ERR,
5559 			    "Invalid value passed to flow_xstat devarg.\n");
5560 		return -EINVAL;
5561 	}
5562 
5563 	bp->flags |= BNXT_FLAG_FLOW_XSTATS_EN;
5564 	if (BNXT_FLOW_XSTATS_EN(bp))
5565 		PMD_DRV_LOG(INFO, "flow_xstat feature enabled.\n");
5566 
5567 	return 0;
5568 }
5569 
5570 static int
5571 bnxt_parse_devarg_max_num_kflows(__rte_unused const char *key,
5572 					const char *value, void *opaque_arg)
5573 {
5574 	struct bnxt *bp = opaque_arg;
5575 	unsigned long max_num_kflows;
5576 	char *end = NULL;
5577 
5578 	if (!value || !opaque_arg) {
5579 		PMD_DRV_LOG(ERR,
5580 			"Invalid parameter passed to max_num_kflows devarg.\n");
5581 		return -EINVAL;
5582 	}
5583 
5584 	max_num_kflows = strtoul(value, &end, 10);
5585 	if (end == NULL || *end != '\0' ||
5586 		(max_num_kflows == ULONG_MAX && errno == ERANGE)) {
5587 		PMD_DRV_LOG(ERR,
5588 			"Invalid parameter passed to max_num_kflows devarg.\n");
5589 		return -EINVAL;
5590 	}
5591 
5592 	if (bnxt_devarg_max_num_kflow_invalid(max_num_kflows)) {
5593 		PMD_DRV_LOG(ERR,
5594 			"Invalid value passed to max_num_kflows devarg.\n");
5595 		return -EINVAL;
5596 	}
5597 
5598 	bp->max_num_kflows = max_num_kflows;
5599 	if (bp->max_num_kflows)
5600 		PMD_DRV_LOG(INFO, "max_num_kflows set as %ldK.\n",
5601 				max_num_kflows);
5602 
5603 	return 0;
5604 }
5605 
5606 static int
5607 bnxt_parse_devarg_app_id(__rte_unused const char *key,
5608 				 const char *value, void *opaque_arg)
5609 {
5610 	struct bnxt *bp = opaque_arg;
5611 	unsigned long app_id;
5612 	char *end = NULL;
5613 
5614 	if (!value || !opaque_arg) {
5615 		PMD_DRV_LOG(ERR,
5616 			    "Invalid parameter passed to app-id "
5617 			    "devargs.\n");
5618 		return -EINVAL;
5619 	}
5620 
5621 	app_id = strtoul(value, &end, 10);
5622 	if (end == NULL || *end != '\0' ||
5623 	    (app_id == ULONG_MAX && errno == ERANGE)) {
5624 		PMD_DRV_LOG(ERR,
5625 			    "Invalid parameter passed to app_id "
5626 			    "devargs.\n");
5627 		return -EINVAL;
5628 	}
5629 
5630 	if (BNXT_DEVARG_APP_ID_INVALID(app_id)) {
5631 		PMD_DRV_LOG(ERR, "Invalid app-id(%d) devargs.\n",
5632 			    (uint16_t)app_id);
5633 		return -EINVAL;
5634 	}
5635 
5636 	bp->app_id = app_id;
5637 	PMD_DRV_LOG(INFO, "app-id=%d feature enabled.\n", (uint16_t)app_id);
5638 
5639 	return 0;
5640 }
5641 
5642 static int
5643 bnxt_parse_devarg_rep_is_pf(__rte_unused const char *key,
5644 			    const char *value, void *opaque_arg)
5645 {
5646 	struct bnxt_representor *vfr_bp = opaque_arg;
5647 	unsigned long rep_is_pf;
5648 	char *end = NULL;
5649 
5650 	if (!value || !opaque_arg) {
5651 		PMD_DRV_LOG(ERR,
5652 			    "Invalid parameter passed to rep_is_pf devargs.\n");
5653 		return -EINVAL;
5654 	}
5655 
5656 	rep_is_pf = strtoul(value, &end, 10);
5657 	if (end == NULL || *end != '\0' ||
5658 	    (rep_is_pf == ULONG_MAX && errno == ERANGE)) {
5659 		PMD_DRV_LOG(ERR,
5660 			    "Invalid parameter passed to rep_is_pf devargs.\n");
5661 		return -EINVAL;
5662 	}
5663 
5664 	if (BNXT_DEVARG_REP_IS_PF_INVALID(rep_is_pf)) {
5665 		PMD_DRV_LOG(ERR,
5666 			    "Invalid value passed to rep_is_pf devargs.\n");
5667 		return -EINVAL;
5668 	}
5669 
5670 	vfr_bp->flags |= rep_is_pf;
5671 	if (BNXT_REP_PF(vfr_bp))
5672 		PMD_DRV_LOG(INFO, "PF representor\n");
5673 	else
5674 		PMD_DRV_LOG(INFO, "VF representor\n");
5675 
5676 	return 0;
5677 }
5678 
5679 static int
5680 bnxt_parse_devarg_rep_based_pf(__rte_unused const char *key,
5681 			       const char *value, void *opaque_arg)
5682 {
5683 	struct bnxt_representor *vfr_bp = opaque_arg;
5684 	unsigned long rep_based_pf;
5685 	char *end = NULL;
5686 
5687 	if (!value || !opaque_arg) {
5688 		PMD_DRV_LOG(ERR,
5689 			    "Invalid parameter passed to rep_based_pf "
5690 			    "devargs.\n");
5691 		return -EINVAL;
5692 	}
5693 
5694 	rep_based_pf = strtoul(value, &end, 10);
5695 	if (end == NULL || *end != '\0' ||
5696 	    (rep_based_pf == ULONG_MAX && errno == ERANGE)) {
5697 		PMD_DRV_LOG(ERR,
5698 			    "Invalid parameter passed to rep_based_pf "
5699 			    "devargs.\n");
5700 		return -EINVAL;
5701 	}
5702 
5703 	if (BNXT_DEVARG_REP_BASED_PF_INVALID(rep_based_pf)) {
5704 		PMD_DRV_LOG(ERR,
5705 			    "Invalid value passed to rep_based_pf devargs.\n");
5706 		return -EINVAL;
5707 	}
5708 
5709 	vfr_bp->rep_based_pf = rep_based_pf;
5710 	vfr_bp->flags |= BNXT_REP_BASED_PF_VALID;
5711 
5712 	PMD_DRV_LOG(INFO, "rep-based-pf = %d\n", vfr_bp->rep_based_pf);
5713 
5714 	return 0;
5715 }
5716 
5717 static int
5718 bnxt_parse_devarg_rep_q_r2f(__rte_unused const char *key,
5719 			    const char *value, void *opaque_arg)
5720 {
5721 	struct bnxt_representor *vfr_bp = opaque_arg;
5722 	unsigned long rep_q_r2f;
5723 	char *end = NULL;
5724 
5725 	if (!value || !opaque_arg) {
5726 		PMD_DRV_LOG(ERR,
5727 			    "Invalid parameter passed to rep_q_r2f "
5728 			    "devargs.\n");
5729 		return -EINVAL;
5730 	}
5731 
5732 	rep_q_r2f = strtoul(value, &end, 10);
5733 	if (end == NULL || *end != '\0' ||
5734 	    (rep_q_r2f == ULONG_MAX && errno == ERANGE)) {
5735 		PMD_DRV_LOG(ERR,
5736 			    "Invalid parameter passed to rep_q_r2f "
5737 			    "devargs.\n");
5738 		return -EINVAL;
5739 	}
5740 
5741 	if (BNXT_DEVARG_REP_Q_R2F_INVALID(rep_q_r2f)) {
5742 		PMD_DRV_LOG(ERR,
5743 			    "Invalid value passed to rep_q_r2f devargs.\n");
5744 		return -EINVAL;
5745 	}
5746 
5747 	vfr_bp->rep_q_r2f = rep_q_r2f;
5748 	vfr_bp->flags |= BNXT_REP_Q_R2F_VALID;
5749 	PMD_DRV_LOG(INFO, "rep-q-r2f = %d\n", vfr_bp->rep_q_r2f);
5750 
5751 	return 0;
5752 }
5753 
5754 static int
5755 bnxt_parse_devarg_rep_q_f2r(__rte_unused const char *key,
5756 			    const char *value, void *opaque_arg)
5757 {
5758 	struct bnxt_representor *vfr_bp = opaque_arg;
5759 	unsigned long rep_q_f2r;
5760 	char *end = NULL;
5761 
5762 	if (!value || !opaque_arg) {
5763 		PMD_DRV_LOG(ERR,
5764 			    "Invalid parameter passed to rep_q_f2r "
5765 			    "devargs.\n");
5766 		return -EINVAL;
5767 	}
5768 
5769 	rep_q_f2r = strtoul(value, &end, 10);
5770 	if (end == NULL || *end != '\0' ||
5771 	    (rep_q_f2r == ULONG_MAX && errno == ERANGE)) {
5772 		PMD_DRV_LOG(ERR,
5773 			    "Invalid parameter passed to rep_q_f2r "
5774 			    "devargs.\n");
5775 		return -EINVAL;
5776 	}
5777 
5778 	if (BNXT_DEVARG_REP_Q_F2R_INVALID(rep_q_f2r)) {
5779 		PMD_DRV_LOG(ERR,
5780 			    "Invalid value passed to rep_q_f2r devargs.\n");
5781 		return -EINVAL;
5782 	}
5783 
5784 	vfr_bp->rep_q_f2r = rep_q_f2r;
5785 	vfr_bp->flags |= BNXT_REP_Q_F2R_VALID;
5786 	PMD_DRV_LOG(INFO, "rep-q-f2r = %d\n", vfr_bp->rep_q_f2r);
5787 
5788 	return 0;
5789 }
5790 
5791 static int
5792 bnxt_parse_devarg_rep_fc_r2f(__rte_unused const char *key,
5793 			     const char *value, void *opaque_arg)
5794 {
5795 	struct bnxt_representor *vfr_bp = opaque_arg;
5796 	unsigned long rep_fc_r2f;
5797 	char *end = NULL;
5798 
5799 	if (!value || !opaque_arg) {
5800 		PMD_DRV_LOG(ERR,
5801 			    "Invalid parameter passed to rep_fc_r2f "
5802 			    "devargs.\n");
5803 		return -EINVAL;
5804 	}
5805 
5806 	rep_fc_r2f = strtoul(value, &end, 10);
5807 	if (end == NULL || *end != '\0' ||
5808 	    (rep_fc_r2f == ULONG_MAX && errno == ERANGE)) {
5809 		PMD_DRV_LOG(ERR,
5810 			    "Invalid parameter passed to rep_fc_r2f "
5811 			    "devargs.\n");
5812 		return -EINVAL;
5813 	}
5814 
5815 	if (BNXT_DEVARG_REP_FC_R2F_INVALID(rep_fc_r2f)) {
5816 		PMD_DRV_LOG(ERR,
5817 			    "Invalid value passed to rep_fc_r2f devargs.\n");
5818 		return -EINVAL;
5819 	}
5820 
5821 	vfr_bp->flags |= BNXT_REP_FC_R2F_VALID;
5822 	vfr_bp->rep_fc_r2f = rep_fc_r2f;
5823 	PMD_DRV_LOG(INFO, "rep-fc-r2f = %lu\n", rep_fc_r2f);
5824 
5825 	return 0;
5826 }
5827 
5828 static int
5829 bnxt_parse_devarg_rep_fc_f2r(__rte_unused const char *key,
5830 			     const char *value, void *opaque_arg)
5831 {
5832 	struct bnxt_representor *vfr_bp = opaque_arg;
5833 	unsigned long rep_fc_f2r;
5834 	char *end = NULL;
5835 
5836 	if (!value || !opaque_arg) {
5837 		PMD_DRV_LOG(ERR,
5838 			    "Invalid parameter passed to rep_fc_f2r "
5839 			    "devargs.\n");
5840 		return -EINVAL;
5841 	}
5842 
5843 	rep_fc_f2r = strtoul(value, &end, 10);
5844 	if (end == NULL || *end != '\0' ||
5845 	    (rep_fc_f2r == ULONG_MAX && errno == ERANGE)) {
5846 		PMD_DRV_LOG(ERR,
5847 			    "Invalid parameter passed to rep_fc_f2r "
5848 			    "devargs.\n");
5849 		return -EINVAL;
5850 	}
5851 
5852 	if (BNXT_DEVARG_REP_FC_F2R_INVALID(rep_fc_f2r)) {
5853 		PMD_DRV_LOG(ERR,
5854 			    "Invalid value passed to rep_fc_f2r devargs.\n");
5855 		return -EINVAL;
5856 	}
5857 
5858 	vfr_bp->flags |= BNXT_REP_FC_F2R_VALID;
5859 	vfr_bp->rep_fc_f2r = rep_fc_f2r;
5860 	PMD_DRV_LOG(INFO, "rep-fc-f2r = %lu\n", rep_fc_f2r);
5861 
5862 	return 0;
5863 }
5864 
5865 static int
5866 bnxt_parse_dev_args(struct bnxt *bp, struct rte_devargs *devargs)
5867 {
5868 	struct rte_kvargs *kvlist;
5869 	int ret;
5870 
5871 	if (devargs == NULL)
5872 		return 0;
5873 
5874 	kvlist = rte_kvargs_parse(devargs->args, bnxt_dev_args);
5875 	if (kvlist == NULL)
5876 		return -EINVAL;
5877 
5878 	/*
5879 	 * Handler for "flow_xstat" devarg.
5880 	 * Invoked as for ex: "-a 0000:00:0d.0,flow_xstat=1"
5881 	 */
5882 	ret = rte_kvargs_process(kvlist, BNXT_DEVARG_FLOW_XSTAT,
5883 				 bnxt_parse_devarg_flow_xstat, bp);
5884 	if (ret)
5885 		goto err;
5886 
5887 	/*
5888 	 * Handler for "accum-stats" devarg.
5889 	 * Invoked as for ex: "-a 0000:00:0d.0,accum-stats=1"
5890 	 */
5891 	rte_kvargs_process(kvlist, BNXT_DEVARG_ACCUM_STATS,
5892 			   bnxt_parse_devarg_accum_stats, bp);
5893 	/*
5894 	 * Handler for "max_num_kflows" devarg.
5895 	 * Invoked as for ex: "-a 000:00:0d.0,max_num_kflows=32"
5896 	 */
5897 	ret = rte_kvargs_process(kvlist, BNXT_DEVARG_MAX_NUM_KFLOWS,
5898 				 bnxt_parse_devarg_max_num_kflows, bp);
5899 	if (ret)
5900 		goto err;
5901 
5902 err:
5903 	/*
5904 	 * Handler for "app-id" devarg.
5905 	 * Invoked as for ex: "-a 000:00:0d.0,app-id=1"
5906 	 */
5907 	rte_kvargs_process(kvlist, BNXT_DEVARG_APP_ID,
5908 			   bnxt_parse_devarg_app_id, bp);
5909 
5910 	rte_kvargs_free(kvlist);
5911 	return ret;
5912 }
5913 
5914 static int bnxt_alloc_switch_domain(struct bnxt *bp)
5915 {
5916 	int rc = 0;
5917 
5918 	if (BNXT_PF(bp) || BNXT_VF_IS_TRUSTED(bp)) {
5919 		rc = rte_eth_switch_domain_alloc(&bp->switch_domain_id);
5920 		if (rc)
5921 			PMD_DRV_LOG(ERR,
5922 				    "Failed to alloc switch domain: %d\n", rc);
5923 		else
5924 			PMD_DRV_LOG(INFO,
5925 				    "Switch domain allocated %d\n",
5926 				    bp->switch_domain_id);
5927 	}
5928 
5929 	return rc;
5930 }
5931 
5932 /* Allocate and initialize various fields in bnxt struct that
5933  * need to be allocated/destroyed only once in the lifetime of the driver
5934  */
5935 static int bnxt_drv_init(struct rte_eth_dev *eth_dev)
5936 {
5937 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
5938 	struct bnxt *bp = eth_dev->data->dev_private;
5939 	int rc = 0;
5940 
5941 	bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
5942 
5943 	if (bnxt_vf_pciid(pci_dev->id.device_id))
5944 		bp->flags |= BNXT_FLAG_VF;
5945 
5946 	if (bnxt_p5_device(pci_dev->id.device_id))
5947 		bp->flags |= BNXT_FLAG_CHIP_P5;
5948 
5949 	if (pci_dev->id.device_id == BROADCOM_DEV_ID_58802 ||
5950 	    pci_dev->id.device_id == BROADCOM_DEV_ID_58804 ||
5951 	    pci_dev->id.device_id == BROADCOM_DEV_ID_58808 ||
5952 	    pci_dev->id.device_id == BROADCOM_DEV_ID_58802_VF)
5953 		bp->flags |= BNXT_FLAG_STINGRAY;
5954 
5955 	if (BNXT_TRUFLOW_EN(bp)) {
5956 		/* extra mbuf field is required to store CFA code from mark */
5957 		static const struct rte_mbuf_dynfield bnxt_cfa_code_dynfield_desc = {
5958 			.name = RTE_PMD_BNXT_CFA_CODE_DYNFIELD_NAME,
5959 			.size = sizeof(bnxt_cfa_code_dynfield_t),
5960 			.align = __alignof__(bnxt_cfa_code_dynfield_t),
5961 		};
5962 		bnxt_cfa_code_dynfield_offset =
5963 			rte_mbuf_dynfield_register(&bnxt_cfa_code_dynfield_desc);
5964 		if (bnxt_cfa_code_dynfield_offset < 0) {
5965 			PMD_DRV_LOG(ERR,
5966 			    "Failed to register mbuf field for TruFlow mark\n");
5967 			return -rte_errno;
5968 		}
5969 	}
5970 
5971 	rc = bnxt_map_pci_bars(eth_dev);
5972 	if (rc) {
5973 		PMD_DRV_LOG(ERR,
5974 			    "Failed to initialize board rc: %x\n", rc);
5975 		return rc;
5976 	}
5977 
5978 	rc = bnxt_alloc_pf_info(bp);
5979 	if (rc)
5980 		return rc;
5981 
5982 	rc = bnxt_alloc_link_info(bp);
5983 	if (rc)
5984 		return rc;
5985 
5986 	rc = bnxt_alloc_parent_info(bp);
5987 	if (rc)
5988 		return rc;
5989 
5990 	rc = bnxt_alloc_hwrm_resources(bp);
5991 	if (rc) {
5992 		PMD_DRV_LOG(ERR,
5993 			    "Failed to allocate response buffer rc: %x\n", rc);
5994 		return rc;
5995 	}
5996 	rc = bnxt_alloc_leds_info(bp);
5997 	if (rc)
5998 		return rc;
5999 
6000 	rc = bnxt_alloc_cos_queues(bp);
6001 	if (rc)
6002 		return rc;
6003 
6004 	rc = bnxt_init_locks(bp);
6005 	if (rc)
6006 		return rc;
6007 
6008 	rc = bnxt_alloc_switch_domain(bp);
6009 	if (rc)
6010 		return rc;
6011 
6012 	return rc;
6013 }
6014 
6015 static int
6016 bnxt_dev_init(struct rte_eth_dev *eth_dev, void *params __rte_unused)
6017 {
6018 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
6019 	static int version_printed;
6020 	struct bnxt *bp;
6021 	int rc;
6022 
6023 	if (version_printed++ == 0)
6024 		PMD_DRV_LOG(INFO, "%s\n", bnxt_version);
6025 
6026 	eth_dev->dev_ops = &bnxt_dev_ops;
6027 	eth_dev->rx_queue_count = bnxt_rx_queue_count_op;
6028 	eth_dev->rx_descriptor_status = bnxt_rx_descriptor_status_op;
6029 	eth_dev->tx_descriptor_status = bnxt_tx_descriptor_status_op;
6030 	eth_dev->rx_pkt_burst = &bnxt_recv_pkts;
6031 	eth_dev->tx_pkt_burst = &bnxt_xmit_pkts;
6032 
6033 	/*
6034 	 * For secondary processes, we don't initialise any further
6035 	 * as primary has already done this work.
6036 	 */
6037 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
6038 		return 0;
6039 
6040 	rte_eth_copy_pci_info(eth_dev, pci_dev);
6041 	eth_dev->data->dev_flags |= RTE_ETH_DEV_AUTOFILL_QUEUE_XSTATS;
6042 
6043 	bp = eth_dev->data->dev_private;
6044 
6045 	/* Parse dev arguments passed on when starting the DPDK application. */
6046 	rc = bnxt_parse_dev_args(bp, pci_dev->device.devargs);
6047 	if (rc)
6048 		goto error_free;
6049 
6050 	rc = bnxt_drv_init(eth_dev);
6051 	if (rc)
6052 		goto error_free;
6053 
6054 	rc = bnxt_init_resources(bp, false);
6055 	if (rc)
6056 		goto error_free;
6057 
6058 	rc = bnxt_alloc_stats_mem(bp);
6059 	if (rc)
6060 		goto error_free;
6061 
6062 	PMD_DRV_LOG(INFO,
6063 		    "Found %s device at mem %" PRIX64 ", node addr %pM\n",
6064 		    DRV_MODULE_NAME,
6065 		    pci_dev->mem_resource[0].phys_addr,
6066 		    pci_dev->mem_resource[0].addr);
6067 
6068 	return 0;
6069 
6070 error_free:
6071 	bnxt_dev_uninit(eth_dev);
6072 	return rc;
6073 }
6074 
6075 
6076 static void bnxt_free_ctx_mem_buf(struct bnxt_ctx_mem_buf_info *ctx)
6077 {
6078 	if (!ctx)
6079 		return;
6080 
6081 	if (ctx->va)
6082 		rte_free(ctx->va);
6083 
6084 	ctx->va = NULL;
6085 	ctx->dma = RTE_BAD_IOVA;
6086 	ctx->ctx_id = BNXT_CTX_VAL_INVAL;
6087 }
6088 
6089 static void bnxt_unregister_fc_ctx_mem(struct bnxt *bp)
6090 {
6091 	bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_RX,
6092 				  CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
6093 				  bp->flow_stat->rx_fc_out_tbl.ctx_id,
6094 				  bp->flow_stat->max_fc,
6095 				  false);
6096 
6097 	bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_TX,
6098 				  CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
6099 				  bp->flow_stat->tx_fc_out_tbl.ctx_id,
6100 				  bp->flow_stat->max_fc,
6101 				  false);
6102 
6103 	if (bp->flow_stat->rx_fc_in_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
6104 		bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->rx_fc_in_tbl.ctx_id);
6105 	bp->flow_stat->rx_fc_in_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
6106 
6107 	if (bp->flow_stat->rx_fc_out_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
6108 		bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->rx_fc_out_tbl.ctx_id);
6109 	bp->flow_stat->rx_fc_out_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
6110 
6111 	if (bp->flow_stat->tx_fc_in_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
6112 		bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->tx_fc_in_tbl.ctx_id);
6113 	bp->flow_stat->tx_fc_in_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
6114 
6115 	if (bp->flow_stat->tx_fc_out_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
6116 		bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->tx_fc_out_tbl.ctx_id);
6117 	bp->flow_stat->tx_fc_out_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
6118 }
6119 
6120 static void bnxt_uninit_fc_ctx_mem(struct bnxt *bp)
6121 {
6122 	bnxt_unregister_fc_ctx_mem(bp);
6123 
6124 	bnxt_free_ctx_mem_buf(&bp->flow_stat->rx_fc_in_tbl);
6125 	bnxt_free_ctx_mem_buf(&bp->flow_stat->rx_fc_out_tbl);
6126 	bnxt_free_ctx_mem_buf(&bp->flow_stat->tx_fc_in_tbl);
6127 	bnxt_free_ctx_mem_buf(&bp->flow_stat->tx_fc_out_tbl);
6128 }
6129 
6130 static void bnxt_uninit_ctx_mem(struct bnxt *bp)
6131 {
6132 	if (BNXT_FLOW_XSTATS_EN(bp))
6133 		bnxt_uninit_fc_ctx_mem(bp);
6134 }
6135 
6136 static void
6137 bnxt_free_error_recovery_info(struct bnxt *bp)
6138 {
6139 	rte_free(bp->recovery_info);
6140 	bp->recovery_info = NULL;
6141 	bp->fw_cap &= ~BNXT_FW_CAP_ERROR_RECOVERY;
6142 }
6143 
6144 static int
6145 bnxt_uninit_resources(struct bnxt *bp, bool reconfig_dev)
6146 {
6147 	int rc;
6148 
6149 	bnxt_free_int(bp);
6150 	bnxt_free_mem(bp, reconfig_dev);
6151 
6152 	bnxt_hwrm_func_buf_unrgtr(bp);
6153 	if (bp->pf != NULL) {
6154 		rte_free(bp->pf->vf_req_buf);
6155 		bp->pf->vf_req_buf = NULL;
6156 	}
6157 
6158 	rc = bnxt_hwrm_func_driver_unregister(bp);
6159 	bp->flags &= ~BNXT_FLAG_REGISTERED;
6160 	bnxt_free_ctx_mem(bp);
6161 	if (!reconfig_dev) {
6162 		bnxt_free_hwrm_resources(bp);
6163 		bnxt_free_error_recovery_info(bp);
6164 	}
6165 
6166 	bnxt_uninit_ctx_mem(bp);
6167 
6168 	bnxt_free_flow_stats_info(bp);
6169 	if (bp->rep_info != NULL)
6170 		bnxt_free_switch_domain(bp);
6171 	bnxt_free_rep_info(bp);
6172 	rte_free(bp->ptp_cfg);
6173 	bp->ptp_cfg = NULL;
6174 	return rc;
6175 }
6176 
6177 static int
6178 bnxt_dev_uninit(struct rte_eth_dev *eth_dev)
6179 {
6180 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
6181 		return -EPERM;
6182 
6183 	PMD_DRV_LOG(DEBUG, "Calling Device uninit\n");
6184 
6185 	if (eth_dev->state != RTE_ETH_DEV_UNUSED)
6186 		bnxt_dev_close_op(eth_dev);
6187 
6188 	return 0;
6189 }
6190 
6191 static int bnxt_pci_remove_dev_with_reps(struct rte_eth_dev *eth_dev)
6192 {
6193 	struct bnxt *bp = eth_dev->data->dev_private;
6194 	struct rte_eth_dev *vf_rep_eth_dev;
6195 	int ret = 0, i;
6196 
6197 	if (!bp)
6198 		return -EINVAL;
6199 
6200 	for (i = 0; i < bp->num_reps; i++) {
6201 		vf_rep_eth_dev = bp->rep_info[i].vfr_eth_dev;
6202 		if (!vf_rep_eth_dev)
6203 			continue;
6204 		PMD_DRV_LOG(DEBUG, "BNXT Port:%d VFR pci remove\n",
6205 			    vf_rep_eth_dev->data->port_id);
6206 		rte_eth_dev_destroy(vf_rep_eth_dev, bnxt_representor_uninit);
6207 	}
6208 	PMD_DRV_LOG(DEBUG, "BNXT Port:%d pci remove\n",
6209 		    eth_dev->data->port_id);
6210 	ret = rte_eth_dev_destroy(eth_dev, bnxt_dev_uninit);
6211 
6212 	return ret;
6213 }
6214 
6215 static void bnxt_free_rep_info(struct bnxt *bp)
6216 {
6217 	rte_free(bp->rep_info);
6218 	bp->rep_info = NULL;
6219 	rte_free(bp->cfa_code_map);
6220 	bp->cfa_code_map = NULL;
6221 }
6222 
6223 static int bnxt_init_rep_info(struct bnxt *bp)
6224 {
6225 	int i = 0, rc;
6226 
6227 	if (bp->rep_info)
6228 		return 0;
6229 
6230 	bp->rep_info = rte_zmalloc("bnxt_rep_info",
6231 				   sizeof(bp->rep_info[0]) * BNXT_MAX_VF_REPS,
6232 				   0);
6233 	if (!bp->rep_info) {
6234 		PMD_DRV_LOG(ERR, "Failed to alloc memory for rep info\n");
6235 		return -ENOMEM;
6236 	}
6237 	bp->cfa_code_map = rte_zmalloc("bnxt_cfa_code_map",
6238 				       sizeof(*bp->cfa_code_map) *
6239 				       BNXT_MAX_CFA_CODE, 0);
6240 	if (!bp->cfa_code_map) {
6241 		PMD_DRV_LOG(ERR, "Failed to alloc memory for cfa_code_map\n");
6242 		bnxt_free_rep_info(bp);
6243 		return -ENOMEM;
6244 	}
6245 
6246 	for (i = 0; i < BNXT_MAX_CFA_CODE; i++)
6247 		bp->cfa_code_map[i] = BNXT_VF_IDX_INVALID;
6248 
6249 	rc = pthread_mutex_init(&bp->rep_info->vfr_lock, NULL);
6250 	if (rc) {
6251 		PMD_DRV_LOG(ERR, "Unable to initialize vfr_lock\n");
6252 		bnxt_free_rep_info(bp);
6253 		return rc;
6254 	}
6255 
6256 	rc = pthread_mutex_init(&bp->rep_info->vfr_start_lock, NULL);
6257 	if (rc) {
6258 		PMD_DRV_LOG(ERR, "Unable to initialize vfr_start_lock\n");
6259 		bnxt_free_rep_info(bp);
6260 		return rc;
6261 	}
6262 
6263 	return rc;
6264 }
6265 
6266 static int bnxt_rep_port_probe(struct rte_pci_device *pci_dev,
6267 			       struct rte_eth_devargs *eth_da,
6268 			       struct rte_eth_dev *backing_eth_dev,
6269 			       const char *dev_args)
6270 {
6271 	struct rte_eth_dev *vf_rep_eth_dev;
6272 	char name[RTE_ETH_NAME_MAX_LEN];
6273 	struct bnxt *backing_bp;
6274 	uint16_t num_rep;
6275 	int i, ret = 0;
6276 	struct rte_kvargs *kvlist = NULL;
6277 
6278 	if (eth_da->type == RTE_ETH_REPRESENTOR_NONE)
6279 		return 0;
6280 	if (eth_da->type != RTE_ETH_REPRESENTOR_VF) {
6281 		PMD_DRV_LOG(ERR, "unsupported representor type %d\n",
6282 			    eth_da->type);
6283 		return -ENOTSUP;
6284 	}
6285 	num_rep = eth_da->nb_representor_ports;
6286 	if (num_rep > BNXT_MAX_VF_REPS) {
6287 		PMD_DRV_LOG(ERR, "nb_representor_ports = %d > %d MAX VF REPS\n",
6288 			    num_rep, BNXT_MAX_VF_REPS);
6289 		return -EINVAL;
6290 	}
6291 
6292 	if (num_rep >= RTE_MAX_ETHPORTS) {
6293 		PMD_DRV_LOG(ERR,
6294 			    "nb_representor_ports = %d > %d MAX ETHPORTS\n",
6295 			    num_rep, RTE_MAX_ETHPORTS);
6296 		return -EINVAL;
6297 	}
6298 
6299 	backing_bp = backing_eth_dev->data->dev_private;
6300 
6301 	if (!(BNXT_PF(backing_bp) || BNXT_VF_IS_TRUSTED(backing_bp))) {
6302 		PMD_DRV_LOG(ERR,
6303 			    "Not a PF or trusted VF. No Representor support\n");
6304 		/* Returning an error is not an option.
6305 		 * Applications are not handling this correctly
6306 		 */
6307 		return 0;
6308 	}
6309 
6310 	if (bnxt_init_rep_info(backing_bp))
6311 		return 0;
6312 
6313 	for (i = 0; i < num_rep; i++) {
6314 		struct bnxt_representor representor = {
6315 			.vf_id = eth_da->representor_ports[i],
6316 			.switch_domain_id = backing_bp->switch_domain_id,
6317 			.parent_dev = backing_eth_dev
6318 		};
6319 
6320 		if (representor.vf_id >= BNXT_MAX_VF_REPS) {
6321 			PMD_DRV_LOG(ERR, "VF-Rep id %d >= %d MAX VF ID\n",
6322 				    representor.vf_id, BNXT_MAX_VF_REPS);
6323 			continue;
6324 		}
6325 
6326 		/* representor port net_bdf_port */
6327 		snprintf(name, sizeof(name), "net_%s_representor_%d",
6328 			 pci_dev->device.name, eth_da->representor_ports[i]);
6329 
6330 		kvlist = rte_kvargs_parse(dev_args, bnxt_dev_args);
6331 		if (kvlist) {
6332 			/*
6333 			 * Handler for "rep_is_pf" devarg.
6334 			 * Invoked as for ex: "-a 000:00:0d.0,
6335 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6336 			 */
6337 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_IS_PF,
6338 						 bnxt_parse_devarg_rep_is_pf,
6339 						 (void *)&representor);
6340 			if (ret) {
6341 				ret = -EINVAL;
6342 				goto err;
6343 			}
6344 			/*
6345 			 * Handler for "rep_based_pf" devarg.
6346 			 * Invoked as for ex: "-a 000:00:0d.0,
6347 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6348 			 */
6349 			ret = rte_kvargs_process(kvlist,
6350 						 BNXT_DEVARG_REP_BASED_PF,
6351 						 bnxt_parse_devarg_rep_based_pf,
6352 						 (void *)&representor);
6353 			if (ret) {
6354 				ret = -EINVAL;
6355 				goto err;
6356 			}
6357 			/*
6358 			 * Handler for "rep_based_pf" devarg.
6359 			 * Invoked as for ex: "-a 000:00:0d.0,
6360 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6361 			 */
6362 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_Q_R2F,
6363 						 bnxt_parse_devarg_rep_q_r2f,
6364 						 (void *)&representor);
6365 			if (ret) {
6366 				ret = -EINVAL;
6367 				goto err;
6368 			}
6369 			/*
6370 			 * Handler for "rep_based_pf" devarg.
6371 			 * Invoked as for ex: "-a 000:00:0d.0,
6372 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6373 			 */
6374 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_Q_F2R,
6375 						 bnxt_parse_devarg_rep_q_f2r,
6376 						 (void *)&representor);
6377 			if (ret) {
6378 				ret = -EINVAL;
6379 				goto err;
6380 			}
6381 			/*
6382 			 * Handler for "rep_based_pf" devarg.
6383 			 * Invoked as for ex: "-a 000:00:0d.0,
6384 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6385 			 */
6386 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_FC_R2F,
6387 						 bnxt_parse_devarg_rep_fc_r2f,
6388 						 (void *)&representor);
6389 			if (ret) {
6390 				ret = -EINVAL;
6391 				goto err;
6392 			}
6393 			/*
6394 			 * Handler for "rep_based_pf" devarg.
6395 			 * Invoked as for ex: "-a 000:00:0d.0,
6396 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6397 			 */
6398 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_FC_F2R,
6399 						 bnxt_parse_devarg_rep_fc_f2r,
6400 						 (void *)&representor);
6401 			if (ret) {
6402 				ret = -EINVAL;
6403 				goto err;
6404 			}
6405 		}
6406 
6407 		ret = rte_eth_dev_create(&pci_dev->device, name,
6408 					 sizeof(struct bnxt_representor),
6409 					 NULL, NULL,
6410 					 bnxt_representor_init,
6411 					 &representor);
6412 		if (ret) {
6413 			PMD_DRV_LOG(ERR, "failed to create bnxt vf "
6414 				    "representor %s.", name);
6415 			goto err;
6416 		}
6417 
6418 		vf_rep_eth_dev = rte_eth_dev_allocated(name);
6419 		if (!vf_rep_eth_dev) {
6420 			PMD_DRV_LOG(ERR, "Failed to find the eth_dev"
6421 				    " for VF-Rep: %s.", name);
6422 			ret = -ENODEV;
6423 			goto err;
6424 		}
6425 
6426 		PMD_DRV_LOG(DEBUG, "BNXT Port:%d VFR pci probe\n",
6427 			    backing_eth_dev->data->port_id);
6428 		backing_bp->rep_info[representor.vf_id].vfr_eth_dev =
6429 							 vf_rep_eth_dev;
6430 		backing_bp->num_reps++;
6431 
6432 	}
6433 
6434 	rte_kvargs_free(kvlist);
6435 	return 0;
6436 
6437 err:
6438 	/* If num_rep > 1, then rollback already created
6439 	 * ports, since we'll be failing the probe anyway
6440 	 */
6441 	if (num_rep > 1)
6442 		bnxt_pci_remove_dev_with_reps(backing_eth_dev);
6443 	rte_errno = -ret;
6444 	rte_kvargs_free(kvlist);
6445 
6446 	return ret;
6447 }
6448 
6449 static int bnxt_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
6450 			  struct rte_pci_device *pci_dev)
6451 {
6452 	struct rte_eth_devargs eth_da = { .nb_representor_ports = 0 };
6453 	struct rte_eth_dev *backing_eth_dev;
6454 	uint16_t num_rep;
6455 	int ret = 0;
6456 
6457 	if (pci_dev->device.devargs) {
6458 		ret = rte_eth_devargs_parse(pci_dev->device.devargs->args,
6459 					    &eth_da);
6460 		if (ret)
6461 			return ret;
6462 	}
6463 
6464 	num_rep = eth_da.nb_representor_ports;
6465 	PMD_DRV_LOG(DEBUG, "nb_representor_ports = %d\n",
6466 		    num_rep);
6467 
6468 	/* We could come here after first level of probe is already invoked
6469 	 * as part of an application bringup(OVS-DPDK vswitchd), so first check
6470 	 * for already allocated eth_dev for the backing device (PF/Trusted VF)
6471 	 */
6472 	backing_eth_dev = rte_eth_dev_allocated(pci_dev->device.name);
6473 	if (backing_eth_dev == NULL) {
6474 		ret = rte_eth_dev_create(&pci_dev->device, pci_dev->device.name,
6475 					 sizeof(struct bnxt),
6476 					 eth_dev_pci_specific_init, pci_dev,
6477 					 bnxt_dev_init, NULL);
6478 
6479 		if (ret || !num_rep)
6480 			return ret;
6481 
6482 		backing_eth_dev = rte_eth_dev_allocated(pci_dev->device.name);
6483 	}
6484 	PMD_DRV_LOG(DEBUG, "BNXT Port:%d pci probe\n",
6485 		    backing_eth_dev->data->port_id);
6486 
6487 	if (!num_rep)
6488 		return ret;
6489 
6490 	/* probe representor ports now */
6491 	ret = bnxt_rep_port_probe(pci_dev, &eth_da, backing_eth_dev,
6492 				  pci_dev->device.devargs->args);
6493 
6494 	return ret;
6495 }
6496 
6497 static int bnxt_pci_remove(struct rte_pci_device *pci_dev)
6498 {
6499 	struct rte_eth_dev *eth_dev;
6500 
6501 	eth_dev = rte_eth_dev_allocated(pci_dev->device.name);
6502 	if (!eth_dev)
6503 		return 0; /* Invoked typically only by OVS-DPDK, by the
6504 			   * time it comes here the eth_dev is already
6505 			   * deleted by rte_eth_dev_close(), so returning
6506 			   * +ve value will at least help in proper cleanup
6507 			   */
6508 
6509 	PMD_DRV_LOG(DEBUG, "BNXT Port:%d pci remove\n", eth_dev->data->port_id);
6510 	if (rte_eal_process_type() == RTE_PROC_PRIMARY) {
6511 		if (eth_dev->data->dev_flags & RTE_ETH_DEV_REPRESENTOR)
6512 			return rte_eth_dev_destroy(eth_dev,
6513 						   bnxt_representor_uninit);
6514 		else
6515 			return rte_eth_dev_destroy(eth_dev,
6516 						   bnxt_dev_uninit);
6517 	} else {
6518 		return rte_eth_dev_pci_generic_remove(pci_dev, NULL);
6519 	}
6520 }
6521 
6522 static struct rte_pci_driver bnxt_rte_pmd = {
6523 	.id_table = bnxt_pci_id_map,
6524 	.drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC |
6525 			RTE_PCI_DRV_INTR_RMV |
6526 			RTE_PCI_DRV_PROBE_AGAIN, /* Needed in case of VF-REPs
6527 						  * and OVS-DPDK
6528 						  */
6529 	.probe = bnxt_pci_probe,
6530 	.remove = bnxt_pci_remove,
6531 };
6532 
6533 static bool
6534 is_device_supported(struct rte_eth_dev *dev, struct rte_pci_driver *drv)
6535 {
6536 	if (strcmp(dev->device->driver->name, drv->driver.name))
6537 		return false;
6538 
6539 	return true;
6540 }
6541 
6542 bool is_bnxt_supported(struct rte_eth_dev *dev)
6543 {
6544 	return is_device_supported(dev, &bnxt_rte_pmd);
6545 }
6546 
6547 RTE_LOG_REGISTER_SUFFIX(bnxt_logtype_driver, driver, NOTICE);
6548 RTE_PMD_REGISTER_PCI(net_bnxt, bnxt_rte_pmd);
6549 RTE_PMD_REGISTER_PCI_TABLE(net_bnxt, bnxt_pci_id_map);
6550 
6551