xref: /dpdk/drivers/net/bnxt/bnxt_ethdev.c (revision b752fb4d)
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_TRUFLOW	"host-based-truflow"
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 
101 static const char *const bnxt_dev_args[] = {
102 	BNXT_DEVARG_REPRESENTOR,
103 	BNXT_DEVARG_TRUFLOW,
104 	BNXT_DEVARG_FLOW_XSTAT,
105 	BNXT_DEVARG_MAX_NUM_KFLOWS,
106 	BNXT_DEVARG_REP_BASED_PF,
107 	BNXT_DEVARG_REP_IS_PF,
108 	BNXT_DEVARG_REP_Q_R2F,
109 	BNXT_DEVARG_REP_Q_F2R,
110 	BNXT_DEVARG_REP_FC_R2F,
111 	BNXT_DEVARG_REP_FC_F2R,
112 	NULL
113 };
114 
115 /*
116  * truflow == false to disable the feature
117  * truflow == true to enable the feature
118  */
119 #define	BNXT_DEVARG_TRUFLOW_INVALID(truflow)	((truflow) > 1)
120 
121 /*
122  * flow_xstat == false to disable the feature
123  * flow_xstat == true to enable the feature
124  */
125 #define	BNXT_DEVARG_FLOW_XSTAT_INVALID(flow_xstat)	((flow_xstat) > 1)
126 
127 /*
128  * rep_is_pf == false to indicate VF representor
129  * rep_is_pf == true to indicate PF representor
130  */
131 #define	BNXT_DEVARG_REP_IS_PF_INVALID(rep_is_pf)	((rep_is_pf) > 1)
132 
133 /*
134  * rep_based_pf == Physical index of the PF
135  */
136 #define	BNXT_DEVARG_REP_BASED_PF_INVALID(rep_based_pf)	((rep_based_pf) > 15)
137 /*
138  * rep_q_r2f == Logical COS Queue index for the rep to endpoint direction
139  */
140 #define	BNXT_DEVARG_REP_Q_R2F_INVALID(rep_q_r2f)	((rep_q_r2f) > 3)
141 
142 /*
143  * rep_q_f2r == Logical COS Queue index for the endpoint to rep direction
144  */
145 #define	BNXT_DEVARG_REP_Q_F2R_INVALID(rep_q_f2r)	((rep_q_f2r) > 3)
146 
147 /*
148  * rep_fc_r2f == Flow control for the representor to endpoint direction
149  */
150 #define BNXT_DEVARG_REP_FC_R2F_INVALID(rep_fc_r2f)	((rep_fc_r2f) > 1)
151 
152 /*
153  * rep_fc_f2r == Flow control for the endpoint to representor direction
154  */
155 #define BNXT_DEVARG_REP_FC_F2R_INVALID(rep_fc_f2r)	((rep_fc_f2r) > 1)
156 
157 int bnxt_cfa_code_dynfield_offset = -1;
158 
159 /*
160  * max_num_kflows must be >= 32
161  * and must be a power-of-2 supported value
162  * return: 1 -> invalid
163  *         0 -> valid
164  */
165 static int bnxt_devarg_max_num_kflow_invalid(uint16_t max_num_kflows)
166 {
167 	if (max_num_kflows < 32 || !rte_is_power_of_2(max_num_kflows))
168 		return 1;
169 	return 0;
170 }
171 
172 static int bnxt_vlan_offload_set_op(struct rte_eth_dev *dev, int mask);
173 static int bnxt_dev_uninit(struct rte_eth_dev *eth_dev);
174 static int bnxt_init_resources(struct bnxt *bp, bool reconfig_dev);
175 static int bnxt_uninit_resources(struct bnxt *bp, bool reconfig_dev);
176 static void bnxt_cancel_fw_health_check(struct bnxt *bp);
177 static int bnxt_restore_vlan_filters(struct bnxt *bp);
178 static void bnxt_dev_recover(void *arg);
179 static void bnxt_free_error_recovery_info(struct bnxt *bp);
180 static void bnxt_free_rep_info(struct bnxt *bp);
181 
182 int is_bnxt_in_error(struct bnxt *bp)
183 {
184 	if (bp->flags & BNXT_FLAG_FATAL_ERROR)
185 		return -EIO;
186 	if (bp->flags & BNXT_FLAG_FW_RESET)
187 		return -EBUSY;
188 
189 	return 0;
190 }
191 
192 /***********************/
193 
194 /*
195  * High level utility functions
196  */
197 
198 static uint16_t bnxt_rss_ctxts(const struct bnxt *bp)
199 {
200 	unsigned int num_rss_rings = RTE_MIN(bp->rx_nr_rings,
201 					     BNXT_RSS_TBL_SIZE_P5);
202 
203 	if (!BNXT_CHIP_P5(bp))
204 		return 1;
205 
206 	return RTE_ALIGN_MUL_CEIL(num_rss_rings,
207 				  BNXT_RSS_ENTRIES_PER_CTX_P5) /
208 				  BNXT_RSS_ENTRIES_PER_CTX_P5;
209 }
210 
211 uint16_t bnxt_rss_hash_tbl_size(const struct bnxt *bp)
212 {
213 	if (!BNXT_CHIP_P5(bp))
214 		return HW_HASH_INDEX_SIZE;
215 
216 	return bnxt_rss_ctxts(bp) * BNXT_RSS_ENTRIES_PER_CTX_P5;
217 }
218 
219 static void bnxt_free_parent_info(struct bnxt *bp)
220 {
221 	rte_free(bp->parent);
222 	bp->parent = NULL;
223 }
224 
225 static void bnxt_free_pf_info(struct bnxt *bp)
226 {
227 	rte_free(bp->pf);
228 	bp->pf = NULL;
229 }
230 
231 static void bnxt_free_link_info(struct bnxt *bp)
232 {
233 	rte_free(bp->link_info);
234 	bp->link_info = NULL;
235 }
236 
237 static void bnxt_free_leds_info(struct bnxt *bp)
238 {
239 	if (BNXT_VF(bp))
240 		return;
241 
242 	rte_free(bp->leds);
243 	bp->leds = NULL;
244 }
245 
246 static void bnxt_free_flow_stats_info(struct bnxt *bp)
247 {
248 	rte_free(bp->flow_stat);
249 	bp->flow_stat = NULL;
250 }
251 
252 static void bnxt_free_cos_queues(struct bnxt *bp)
253 {
254 	rte_free(bp->rx_cos_queue);
255 	bp->rx_cos_queue = NULL;
256 	rte_free(bp->tx_cos_queue);
257 	bp->tx_cos_queue = NULL;
258 }
259 
260 static void bnxt_free_mem(struct bnxt *bp, bool reconfig)
261 {
262 	bnxt_free_filter_mem(bp);
263 	bnxt_free_vnic_attributes(bp);
264 	bnxt_free_vnic_mem(bp);
265 
266 	/* tx/rx rings are configured as part of *_queue_setup callbacks.
267 	 * If the number of rings change across fw update,
268 	 * we don't have much choice except to warn the user.
269 	 */
270 	if (!reconfig) {
271 		bnxt_free_stats(bp);
272 		bnxt_free_tx_rings(bp);
273 		bnxt_free_rx_rings(bp);
274 	}
275 	bnxt_free_async_cp_ring(bp);
276 	bnxt_free_rxtx_nq_ring(bp);
277 
278 	rte_free(bp->grp_info);
279 	bp->grp_info = NULL;
280 }
281 
282 static int bnxt_alloc_parent_info(struct bnxt *bp)
283 {
284 	bp->parent = rte_zmalloc("bnxt_parent_info",
285 				 sizeof(struct bnxt_parent_info), 0);
286 	if (bp->parent == NULL)
287 		return -ENOMEM;
288 
289 	return 0;
290 }
291 
292 static int bnxt_alloc_pf_info(struct bnxt *bp)
293 {
294 	bp->pf = rte_zmalloc("bnxt_pf_info", sizeof(struct bnxt_pf_info), 0);
295 	if (bp->pf == NULL)
296 		return -ENOMEM;
297 
298 	return 0;
299 }
300 
301 static int bnxt_alloc_link_info(struct bnxt *bp)
302 {
303 	bp->link_info =
304 		rte_zmalloc("bnxt_link_info", sizeof(struct bnxt_link_info), 0);
305 	if (bp->link_info == NULL)
306 		return -ENOMEM;
307 
308 	return 0;
309 }
310 
311 static int bnxt_alloc_leds_info(struct bnxt *bp)
312 {
313 	if (BNXT_VF(bp))
314 		return 0;
315 
316 	bp->leds = rte_zmalloc("bnxt_leds",
317 			       BNXT_MAX_LED * sizeof(struct bnxt_led_info),
318 			       0);
319 	if (bp->leds == NULL)
320 		return -ENOMEM;
321 
322 	return 0;
323 }
324 
325 static int bnxt_alloc_cos_queues(struct bnxt *bp)
326 {
327 	bp->rx_cos_queue =
328 		rte_zmalloc("bnxt_rx_cosq",
329 			    BNXT_COS_QUEUE_COUNT *
330 			    sizeof(struct bnxt_cos_queue_info),
331 			    0);
332 	if (bp->rx_cos_queue == NULL)
333 		return -ENOMEM;
334 
335 	bp->tx_cos_queue =
336 		rte_zmalloc("bnxt_tx_cosq",
337 			    BNXT_COS_QUEUE_COUNT *
338 			    sizeof(struct bnxt_cos_queue_info),
339 			    0);
340 	if (bp->tx_cos_queue == NULL)
341 		return -ENOMEM;
342 
343 	return 0;
344 }
345 
346 static int bnxt_alloc_flow_stats_info(struct bnxt *bp)
347 {
348 	bp->flow_stat = rte_zmalloc("bnxt_flow_xstat",
349 				    sizeof(struct bnxt_flow_stat_info), 0);
350 	if (bp->flow_stat == NULL)
351 		return -ENOMEM;
352 
353 	return 0;
354 }
355 
356 static int bnxt_alloc_mem(struct bnxt *bp, bool reconfig)
357 {
358 	int rc;
359 
360 	rc = bnxt_alloc_ring_grps(bp);
361 	if (rc)
362 		goto alloc_mem_err;
363 
364 	rc = bnxt_alloc_async_ring_struct(bp);
365 	if (rc)
366 		goto alloc_mem_err;
367 
368 	rc = bnxt_alloc_vnic_mem(bp);
369 	if (rc)
370 		goto alloc_mem_err;
371 
372 	rc = bnxt_alloc_vnic_attributes(bp);
373 	if (rc)
374 		goto alloc_mem_err;
375 
376 	rc = bnxt_alloc_filter_mem(bp);
377 	if (rc)
378 		goto alloc_mem_err;
379 
380 	rc = bnxt_alloc_async_cp_ring(bp);
381 	if (rc)
382 		goto alloc_mem_err;
383 
384 	rc = bnxt_alloc_rxtx_nq_ring(bp);
385 	if (rc)
386 		goto alloc_mem_err;
387 
388 	if (BNXT_FLOW_XSTATS_EN(bp)) {
389 		rc = bnxt_alloc_flow_stats_info(bp);
390 		if (rc)
391 			goto alloc_mem_err;
392 	}
393 
394 	return 0;
395 
396 alloc_mem_err:
397 	bnxt_free_mem(bp, reconfig);
398 	return rc;
399 }
400 
401 static int bnxt_setup_one_vnic(struct bnxt *bp, uint16_t vnic_id)
402 {
403 	struct rte_eth_conf *dev_conf = &bp->eth_dev->data->dev_conf;
404 	struct bnxt_vnic_info *vnic = &bp->vnic_info[vnic_id];
405 	uint64_t rx_offloads = dev_conf->rxmode.offloads;
406 	struct bnxt_rx_queue *rxq;
407 	unsigned int j;
408 	int rc;
409 
410 	rc = bnxt_vnic_grp_alloc(bp, vnic);
411 	if (rc)
412 		goto err_out;
413 
414 	PMD_DRV_LOG(DEBUG, "vnic[%d] = %p vnic->fw_grp_ids = %p\n",
415 		    vnic_id, vnic, vnic->fw_grp_ids);
416 
417 	rc = bnxt_hwrm_vnic_alloc(bp, vnic);
418 	if (rc)
419 		goto err_out;
420 
421 	/* Alloc RSS context only if RSS mode is enabled */
422 	if (dev_conf->rxmode.mq_mode & ETH_MQ_RX_RSS) {
423 		int j, nr_ctxs = bnxt_rss_ctxts(bp);
424 
425 		/* RSS table size in Thor is 512.
426 		 * Cap max Rx rings to same value
427 		 */
428 		if (bp->rx_nr_rings > BNXT_RSS_TBL_SIZE_P5) {
429 			PMD_DRV_LOG(ERR, "RxQ cnt %d > reta_size %d\n",
430 				    bp->rx_nr_rings, BNXT_RSS_TBL_SIZE_P5);
431 			goto err_out;
432 		}
433 
434 		rc = 0;
435 		for (j = 0; j < nr_ctxs; j++) {
436 			rc = bnxt_hwrm_vnic_ctx_alloc(bp, vnic, j);
437 			if (rc)
438 				break;
439 		}
440 		if (rc) {
441 			PMD_DRV_LOG(ERR,
442 				    "HWRM vnic %d ctx %d alloc failure rc: %x\n",
443 				    vnic_id, j, rc);
444 			goto err_out;
445 		}
446 		vnic->num_lb_ctxts = nr_ctxs;
447 	}
448 
449 	/*
450 	 * Firmware sets pf pair in default vnic cfg. If the VLAN strip
451 	 * setting is not available at this time, it will not be
452 	 * configured correctly in the CFA.
453 	 */
454 	if (rx_offloads & DEV_RX_OFFLOAD_VLAN_STRIP)
455 		vnic->vlan_strip = true;
456 	else
457 		vnic->vlan_strip = false;
458 
459 	rc = bnxt_hwrm_vnic_cfg(bp, vnic);
460 	if (rc)
461 		goto err_out;
462 
463 	rc = bnxt_set_hwrm_vnic_filters(bp, vnic);
464 	if (rc)
465 		goto err_out;
466 
467 	for (j = 0; j < bp->rx_num_qs_per_vnic; j++) {
468 		rxq = bp->eth_dev->data->rx_queues[j];
469 
470 		PMD_DRV_LOG(DEBUG,
471 			    "rxq[%d]->vnic=%p vnic->fw_grp_ids=%p\n",
472 			    j, rxq->vnic, rxq->vnic->fw_grp_ids);
473 
474 		if (BNXT_HAS_RING_GRPS(bp) && rxq->rx_deferred_start)
475 			rxq->vnic->fw_grp_ids[j] = INVALID_HW_RING_ID;
476 		else
477 			vnic->rx_queue_cnt++;
478 	}
479 
480 	PMD_DRV_LOG(DEBUG, "vnic->rx_queue_cnt = %d\n", vnic->rx_queue_cnt);
481 
482 	rc = bnxt_vnic_rss_configure(bp, vnic);
483 	if (rc)
484 		goto err_out;
485 
486 	bnxt_hwrm_vnic_plcmode_cfg(bp, vnic);
487 
488 	rc = bnxt_hwrm_vnic_tpa_cfg(bp, vnic,
489 				    (rx_offloads & DEV_RX_OFFLOAD_TCP_LRO) ?
490 				    true : false);
491 	if (rc)
492 		goto err_out;
493 
494 	return 0;
495 err_out:
496 	PMD_DRV_LOG(ERR, "HWRM vnic %d cfg failure rc: %x\n",
497 		    vnic_id, rc);
498 	return rc;
499 }
500 
501 static int bnxt_register_fc_ctx_mem(struct bnxt *bp)
502 {
503 	int rc = 0;
504 
505 	rc = bnxt_hwrm_ctx_rgtr(bp, bp->flow_stat->rx_fc_in_tbl.dma,
506 				&bp->flow_stat->rx_fc_in_tbl.ctx_id);
507 	if (rc)
508 		return rc;
509 
510 	PMD_DRV_LOG(DEBUG,
511 		    "rx_fc_in_tbl.va = %p rx_fc_in_tbl.dma = %p"
512 		    " rx_fc_in_tbl.ctx_id = %d\n",
513 		    bp->flow_stat->rx_fc_in_tbl.va,
514 		    (void *)((uintptr_t)bp->flow_stat->rx_fc_in_tbl.dma),
515 		    bp->flow_stat->rx_fc_in_tbl.ctx_id);
516 
517 	rc = bnxt_hwrm_ctx_rgtr(bp, bp->flow_stat->rx_fc_out_tbl.dma,
518 				&bp->flow_stat->rx_fc_out_tbl.ctx_id);
519 	if (rc)
520 		return rc;
521 
522 	PMD_DRV_LOG(DEBUG,
523 		    "rx_fc_out_tbl.va = %p rx_fc_out_tbl.dma = %p"
524 		    " rx_fc_out_tbl.ctx_id = %d\n",
525 		    bp->flow_stat->rx_fc_out_tbl.va,
526 		    (void *)((uintptr_t)bp->flow_stat->rx_fc_out_tbl.dma),
527 		    bp->flow_stat->rx_fc_out_tbl.ctx_id);
528 
529 	rc = bnxt_hwrm_ctx_rgtr(bp, bp->flow_stat->tx_fc_in_tbl.dma,
530 				&bp->flow_stat->tx_fc_in_tbl.ctx_id);
531 	if (rc)
532 		return rc;
533 
534 	PMD_DRV_LOG(DEBUG,
535 		    "tx_fc_in_tbl.va = %p tx_fc_in_tbl.dma = %p"
536 		    " tx_fc_in_tbl.ctx_id = %d\n",
537 		    bp->flow_stat->tx_fc_in_tbl.va,
538 		    (void *)((uintptr_t)bp->flow_stat->tx_fc_in_tbl.dma),
539 		    bp->flow_stat->tx_fc_in_tbl.ctx_id);
540 
541 	rc = bnxt_hwrm_ctx_rgtr(bp, bp->flow_stat->tx_fc_out_tbl.dma,
542 				&bp->flow_stat->tx_fc_out_tbl.ctx_id);
543 	if (rc)
544 		return rc;
545 
546 	PMD_DRV_LOG(DEBUG,
547 		    "tx_fc_out_tbl.va = %p tx_fc_out_tbl.dma = %p"
548 		    " tx_fc_out_tbl.ctx_id = %d\n",
549 		    bp->flow_stat->tx_fc_out_tbl.va,
550 		    (void *)((uintptr_t)bp->flow_stat->tx_fc_out_tbl.dma),
551 		    bp->flow_stat->tx_fc_out_tbl.ctx_id);
552 
553 	memset(bp->flow_stat->rx_fc_out_tbl.va,
554 	       0,
555 	       bp->flow_stat->rx_fc_out_tbl.size);
556 	rc = bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_RX,
557 				       CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
558 				       bp->flow_stat->rx_fc_out_tbl.ctx_id,
559 				       bp->flow_stat->max_fc,
560 				       true);
561 	if (rc)
562 		return rc;
563 
564 	memset(bp->flow_stat->tx_fc_out_tbl.va,
565 	       0,
566 	       bp->flow_stat->tx_fc_out_tbl.size);
567 	rc = bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_TX,
568 				       CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
569 				       bp->flow_stat->tx_fc_out_tbl.ctx_id,
570 				       bp->flow_stat->max_fc,
571 				       true);
572 
573 	return rc;
574 }
575 
576 static int bnxt_alloc_ctx_mem_buf(char *type, size_t size,
577 				  struct bnxt_ctx_mem_buf_info *ctx)
578 {
579 	if (!ctx)
580 		return -EINVAL;
581 
582 	ctx->va = rte_zmalloc(type, size, 0);
583 	if (ctx->va == NULL)
584 		return -ENOMEM;
585 	rte_mem_lock_page(ctx->va);
586 	ctx->size = size;
587 	ctx->dma = rte_mem_virt2iova(ctx->va);
588 	if (ctx->dma == RTE_BAD_IOVA)
589 		return -ENOMEM;
590 
591 	return 0;
592 }
593 
594 static int bnxt_init_fc_ctx_mem(struct bnxt *bp)
595 {
596 	struct rte_pci_device *pdev = bp->pdev;
597 	char type[RTE_MEMZONE_NAMESIZE];
598 	uint16_t max_fc;
599 	int rc = 0;
600 
601 	max_fc = bp->flow_stat->max_fc;
602 
603 	sprintf(type, "bnxt_rx_fc_in_" PCI_PRI_FMT, pdev->addr.domain,
604 		pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
605 	/* 4 bytes for each counter-id */
606 	rc = bnxt_alloc_ctx_mem_buf(type,
607 				    max_fc * 4,
608 				    &bp->flow_stat->rx_fc_in_tbl);
609 	if (rc)
610 		return rc;
611 
612 	sprintf(type, "bnxt_rx_fc_out_" PCI_PRI_FMT, pdev->addr.domain,
613 		pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
614 	/* 16 bytes for each counter - 8 bytes pkt_count, 8 bytes byte_count */
615 	rc = bnxt_alloc_ctx_mem_buf(type,
616 				    max_fc * 16,
617 				    &bp->flow_stat->rx_fc_out_tbl);
618 	if (rc)
619 		return rc;
620 
621 	sprintf(type, "bnxt_tx_fc_in_" PCI_PRI_FMT, pdev->addr.domain,
622 		pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
623 	/* 4 bytes for each counter-id */
624 	rc = bnxt_alloc_ctx_mem_buf(type,
625 				    max_fc * 4,
626 				    &bp->flow_stat->tx_fc_in_tbl);
627 	if (rc)
628 		return rc;
629 
630 	sprintf(type, "bnxt_tx_fc_out_" PCI_PRI_FMT, pdev->addr.domain,
631 		pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
632 	/* 16 bytes for each counter - 8 bytes pkt_count, 8 bytes byte_count */
633 	rc = bnxt_alloc_ctx_mem_buf(type,
634 				    max_fc * 16,
635 				    &bp->flow_stat->tx_fc_out_tbl);
636 	if (rc)
637 		return rc;
638 
639 	rc = bnxt_register_fc_ctx_mem(bp);
640 
641 	return rc;
642 }
643 
644 static int bnxt_init_ctx_mem(struct bnxt *bp)
645 {
646 	int rc = 0;
647 
648 	if (!(bp->fw_cap & BNXT_FW_CAP_ADV_FLOW_COUNTERS) ||
649 	    !(BNXT_PF(bp) || BNXT_VF_IS_TRUSTED(bp)) ||
650 	    !BNXT_FLOW_XSTATS_EN(bp))
651 		return 0;
652 
653 	rc = bnxt_hwrm_cfa_counter_qcaps(bp, &bp->flow_stat->max_fc);
654 	if (rc)
655 		return rc;
656 
657 	rc = bnxt_init_fc_ctx_mem(bp);
658 
659 	return rc;
660 }
661 
662 static int bnxt_update_phy_setting(struct bnxt *bp)
663 {
664 	struct rte_eth_link new;
665 	int rc;
666 
667 	rc = bnxt_get_hwrm_link_config(bp, &new);
668 	if (rc) {
669 		PMD_DRV_LOG(ERR, "Failed to get link settings\n");
670 		return rc;
671 	}
672 
673 	/*
674 	 * On BCM957508-N2100 adapters, FW will not allow any user other
675 	 * than BMC to shutdown the port. bnxt_get_hwrm_link_config() call
676 	 * always returns link up. Force phy update always in that case.
677 	 */
678 	if (!new.link_status || IS_BNXT_DEV_957508_N2100(bp)) {
679 		rc = bnxt_set_hwrm_link_config(bp, true);
680 		if (rc) {
681 			PMD_DRV_LOG(ERR, "Failed to update PHY settings\n");
682 			return rc;
683 		}
684 	}
685 
686 	return rc;
687 }
688 
689 static int bnxt_start_nic(struct bnxt *bp)
690 {
691 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(bp->eth_dev);
692 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
693 	uint32_t intr_vector = 0;
694 	uint32_t queue_id, base = BNXT_MISC_VEC_ID;
695 	uint32_t vec = BNXT_MISC_VEC_ID;
696 	unsigned int i, j;
697 	int rc;
698 
699 	if (bp->eth_dev->data->mtu > RTE_ETHER_MTU) {
700 		bp->eth_dev->data->dev_conf.rxmode.offloads |=
701 			DEV_RX_OFFLOAD_JUMBO_FRAME;
702 		bp->flags |= BNXT_FLAG_JUMBO;
703 	} else {
704 		bp->eth_dev->data->dev_conf.rxmode.offloads &=
705 			~DEV_RX_OFFLOAD_JUMBO_FRAME;
706 		bp->flags &= ~BNXT_FLAG_JUMBO;
707 	}
708 
709 	/* THOR does not support ring groups.
710 	 * But we will use the array to save RSS context IDs.
711 	 */
712 	if (BNXT_CHIP_P5(bp))
713 		bp->max_ring_grps = BNXT_MAX_RSS_CTXTS_P5;
714 
715 	rc = bnxt_alloc_all_hwrm_stat_ctxs(bp);
716 	if (rc) {
717 		PMD_DRV_LOG(ERR, "HWRM stat ctx alloc failure rc: %x\n", rc);
718 		goto err_out;
719 	}
720 
721 	rc = bnxt_alloc_hwrm_rings(bp);
722 	if (rc) {
723 		PMD_DRV_LOG(ERR, "HWRM ring alloc failure rc: %x\n", rc);
724 		goto err_out;
725 	}
726 
727 	rc = bnxt_alloc_all_hwrm_ring_grps(bp);
728 	if (rc) {
729 		PMD_DRV_LOG(ERR, "HWRM ring grp alloc failure: %x\n", rc);
730 		goto err_out;
731 	}
732 
733 	if (!(bp->vnic_cap_flags & BNXT_VNIC_CAP_COS_CLASSIFY))
734 		goto skip_cosq_cfg;
735 
736 	for (j = 0, i = 0; i < BNXT_COS_QUEUE_COUNT; i++) {
737 		if (bp->rx_cos_queue[i].id != 0xff) {
738 			struct bnxt_vnic_info *vnic = &bp->vnic_info[j++];
739 
740 			if (!vnic) {
741 				PMD_DRV_LOG(ERR,
742 					    "Num pools more than FW profile\n");
743 				rc = -EINVAL;
744 				goto err_out;
745 			}
746 			vnic->cos_queue_id = bp->rx_cos_queue[i].id;
747 			bp->rx_cosq_cnt++;
748 		}
749 	}
750 
751 skip_cosq_cfg:
752 	rc = bnxt_mq_rx_configure(bp);
753 	if (rc) {
754 		PMD_DRV_LOG(ERR, "MQ mode configure failure rc: %x\n", rc);
755 		goto err_out;
756 	}
757 
758 	/* default vnic 0 */
759 	rc = bnxt_setup_one_vnic(bp, 0);
760 	if (rc)
761 		goto err_out;
762 	/* VNIC configuration */
763 	if (BNXT_RFS_NEEDS_VNIC(bp)) {
764 		for (i = 1; i < bp->nr_vnics; i++) {
765 			rc = bnxt_setup_one_vnic(bp, i);
766 			if (rc)
767 				goto err_out;
768 		}
769 	}
770 
771 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, &bp->vnic_info[0], 0, NULL);
772 	if (rc) {
773 		PMD_DRV_LOG(ERR,
774 			"HWRM cfa l2 rx mask failure rc: %x\n", rc);
775 		goto err_out;
776 	}
777 
778 	/* check and configure queue intr-vector mapping */
779 	if ((rte_intr_cap_multiple(intr_handle) ||
780 	     !RTE_ETH_DEV_SRIOV(bp->eth_dev).active) &&
781 	    bp->eth_dev->data->dev_conf.intr_conf.rxq != 0) {
782 		intr_vector = bp->eth_dev->data->nb_rx_queues;
783 		PMD_DRV_LOG(DEBUG, "intr_vector = %d\n", intr_vector);
784 		if (intr_vector > bp->rx_cp_nr_rings) {
785 			PMD_DRV_LOG(ERR, "At most %d intr queues supported",
786 					bp->rx_cp_nr_rings);
787 			return -ENOTSUP;
788 		}
789 		rc = rte_intr_efd_enable(intr_handle, intr_vector);
790 		if (rc)
791 			return rc;
792 	}
793 
794 	if (rte_intr_dp_is_en(intr_handle) && !intr_handle->intr_vec) {
795 		intr_handle->intr_vec =
796 			rte_zmalloc("intr_vec",
797 				    bp->eth_dev->data->nb_rx_queues *
798 				    sizeof(int), 0);
799 		if (intr_handle->intr_vec == NULL) {
800 			PMD_DRV_LOG(ERR, "Failed to allocate %d rx_queues"
801 				" intr_vec", bp->eth_dev->data->nb_rx_queues);
802 			rc = -ENOMEM;
803 			goto err_out;
804 		}
805 		PMD_DRV_LOG(DEBUG, "intr_handle->intr_vec = %p "
806 			"intr_handle->nb_efd = %d intr_handle->max_intr = %d\n",
807 			 intr_handle->intr_vec, intr_handle->nb_efd,
808 			intr_handle->max_intr);
809 		for (queue_id = 0; queue_id < bp->eth_dev->data->nb_rx_queues;
810 		     queue_id++) {
811 			intr_handle->intr_vec[queue_id] =
812 							vec + BNXT_RX_VEC_START;
813 			if (vec < base + intr_handle->nb_efd - 1)
814 				vec++;
815 		}
816 	}
817 
818 	/* enable uio/vfio intr/eventfd mapping */
819 	rc = rte_intr_enable(intr_handle);
820 #ifndef RTE_EXEC_ENV_FREEBSD
821 	/* In FreeBSD OS, nic_uio driver does not support interrupts */
822 	if (rc)
823 		goto err_out;
824 #endif
825 
826 	rc = bnxt_update_phy_setting(bp);
827 	if (rc)
828 		goto err_out;
829 
830 	bp->mark_table = rte_zmalloc("bnxt_mark_table", BNXT_MARK_TABLE_SZ, 0);
831 	if (!bp->mark_table)
832 		PMD_DRV_LOG(ERR, "Allocation of mark table failed\n");
833 
834 	return 0;
835 
836 err_out:
837 	/* Some of the error status returned by FW may not be from errno.h */
838 	if (rc > 0)
839 		rc = -EIO;
840 
841 	return rc;
842 }
843 
844 static int bnxt_shutdown_nic(struct bnxt *bp)
845 {
846 	bnxt_free_all_hwrm_resources(bp);
847 	bnxt_free_all_filters(bp);
848 	bnxt_free_all_vnics(bp);
849 	return 0;
850 }
851 
852 /*
853  * Device configuration and status function
854  */
855 
856 uint32_t bnxt_get_speed_capabilities(struct bnxt *bp)
857 {
858 	uint32_t link_speed = 0;
859 	uint32_t speed_capa = 0;
860 
861 	if (bp->link_info == NULL)
862 		return 0;
863 
864 	link_speed = bp->link_info->support_speeds;
865 
866 	/* If PAM4 is configured, use PAM4 supported speed */
867 	if (link_speed == 0 && bp->link_info->support_pam4_speeds > 0)
868 		link_speed = bp->link_info->support_pam4_speeds;
869 
870 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_LINK_SPEED_100MB)
871 		speed_capa |= ETH_LINK_SPEED_100M;
872 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_100MBHD)
873 		speed_capa |= ETH_LINK_SPEED_100M_HD;
874 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_1GB)
875 		speed_capa |= ETH_LINK_SPEED_1G;
876 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_2_5GB)
877 		speed_capa |= ETH_LINK_SPEED_2_5G;
878 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_10GB)
879 		speed_capa |= ETH_LINK_SPEED_10G;
880 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_20GB)
881 		speed_capa |= ETH_LINK_SPEED_20G;
882 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_25GB)
883 		speed_capa |= ETH_LINK_SPEED_25G;
884 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_40GB)
885 		speed_capa |= ETH_LINK_SPEED_40G;
886 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_50GB)
887 		speed_capa |= ETH_LINK_SPEED_50G;
888 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_100GB)
889 		speed_capa |= ETH_LINK_SPEED_100G;
890 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_PAM4_SPEEDS_50G)
891 		speed_capa |= ETH_LINK_SPEED_50G;
892 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_PAM4_SPEEDS_100G)
893 		speed_capa |= ETH_LINK_SPEED_100G;
894 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_PAM4_SPEEDS_200G)
895 		speed_capa |= ETH_LINK_SPEED_200G;
896 
897 	if (bp->link_info->auto_mode ==
898 	    HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_MODE_NONE)
899 		speed_capa |= ETH_LINK_SPEED_FIXED;
900 
901 	return speed_capa;
902 }
903 
904 static int bnxt_dev_info_get_op(struct rte_eth_dev *eth_dev,
905 				struct rte_eth_dev_info *dev_info)
906 {
907 	struct rte_pci_device *pdev = RTE_DEV_TO_PCI(eth_dev->device);
908 	struct bnxt *bp = eth_dev->data->dev_private;
909 	uint16_t max_vnics, i, j, vpool, vrxq;
910 	unsigned int max_rx_rings;
911 	int rc;
912 
913 	rc = is_bnxt_in_error(bp);
914 	if (rc)
915 		return rc;
916 
917 	/* MAC Specifics */
918 	dev_info->max_mac_addrs = bp->max_l2_ctx;
919 	dev_info->max_hash_mac_addrs = 0;
920 
921 	/* PF/VF specifics */
922 	if (BNXT_PF(bp))
923 		dev_info->max_vfs = pdev->max_vfs;
924 
925 	max_rx_rings = bnxt_max_rings(bp);
926 	/* For the sake of symmetry, max_rx_queues = max_tx_queues */
927 	dev_info->max_rx_queues = max_rx_rings;
928 	dev_info->max_tx_queues = max_rx_rings;
929 	dev_info->reta_size = bnxt_rss_hash_tbl_size(bp);
930 	dev_info->hash_key_size = 40;
931 	max_vnics = bp->max_vnics;
932 
933 	/* MTU specifics */
934 	dev_info->min_mtu = RTE_ETHER_MIN_MTU;
935 	dev_info->max_mtu = BNXT_MAX_MTU;
936 
937 	/* Fast path specifics */
938 	dev_info->min_rx_bufsize = 1;
939 	dev_info->max_rx_pktlen = BNXT_MAX_PKT_LEN;
940 
941 	dev_info->rx_offload_capa = BNXT_DEV_RX_OFFLOAD_SUPPORT;
942 	if (bp->flags & BNXT_FLAG_PTP_SUPPORTED)
943 		dev_info->rx_offload_capa |= DEV_RX_OFFLOAD_TIMESTAMP;
944 	dev_info->tx_queue_offload_capa = DEV_TX_OFFLOAD_MBUF_FAST_FREE;
945 	dev_info->tx_offload_capa = BNXT_DEV_TX_OFFLOAD_SUPPORT |
946 				    dev_info->tx_queue_offload_capa;
947 	dev_info->flow_type_rss_offloads = BNXT_ETH_RSS_SUPPORT;
948 
949 	dev_info->speed_capa = bnxt_get_speed_capabilities(bp);
950 
951 	/* *INDENT-OFF* */
952 	dev_info->default_rxconf = (struct rte_eth_rxconf) {
953 		.rx_thresh = {
954 			.pthresh = 8,
955 			.hthresh = 8,
956 			.wthresh = 0,
957 		},
958 		.rx_free_thresh = 32,
959 		.rx_drop_en = BNXT_DEFAULT_RX_DROP_EN,
960 	};
961 
962 	dev_info->default_txconf = (struct rte_eth_txconf) {
963 		.tx_thresh = {
964 			.pthresh = 32,
965 			.hthresh = 0,
966 			.wthresh = 0,
967 		},
968 		.tx_free_thresh = 32,
969 		.tx_rs_thresh = 32,
970 	};
971 	eth_dev->data->dev_conf.intr_conf.lsc = 1;
972 
973 	eth_dev->data->dev_conf.intr_conf.rxq = 1;
974 	dev_info->rx_desc_lim.nb_min = BNXT_MIN_RING_DESC;
975 	dev_info->rx_desc_lim.nb_max = BNXT_MAX_RX_RING_DESC;
976 	dev_info->tx_desc_lim.nb_min = BNXT_MIN_RING_DESC;
977 	dev_info->tx_desc_lim.nb_max = BNXT_MAX_TX_RING_DESC;
978 
979 	if (BNXT_PF(bp) || BNXT_VF_IS_TRUSTED(bp)) {
980 		dev_info->switch_info.name = eth_dev->device->name;
981 		dev_info->switch_info.domain_id = bp->switch_domain_id;
982 		dev_info->switch_info.port_id =
983 				BNXT_PF(bp) ? BNXT_SWITCH_PORT_ID_PF :
984 				    BNXT_SWITCH_PORT_ID_TRUSTED_VF;
985 	}
986 
987 	/* *INDENT-ON* */
988 
989 	/*
990 	 * TODO: default_rxconf, default_txconf, rx_desc_lim, and tx_desc_lim
991 	 *       need further investigation.
992 	 */
993 
994 	/* VMDq resources */
995 	vpool = 64; /* ETH_64_POOLS */
996 	vrxq = 128; /* ETH_VMDQ_DCB_NUM_QUEUES */
997 	for (i = 0; i < 4; vpool >>= 1, i++) {
998 		if (max_vnics > vpool) {
999 			for (j = 0; j < 5; vrxq >>= 1, j++) {
1000 				if (dev_info->max_rx_queues > vrxq) {
1001 					if (vpool > vrxq)
1002 						vpool = vrxq;
1003 					goto found;
1004 				}
1005 			}
1006 			/* Not enough resources to support VMDq */
1007 			break;
1008 		}
1009 	}
1010 	/* Not enough resources to support VMDq */
1011 	vpool = 0;
1012 	vrxq = 0;
1013 found:
1014 	dev_info->max_vmdq_pools = vpool;
1015 	dev_info->vmdq_queue_num = vrxq;
1016 
1017 	dev_info->vmdq_pool_base = 0;
1018 	dev_info->vmdq_queue_base = 0;
1019 
1020 	return 0;
1021 }
1022 
1023 /* Configure the device based on the configuration provided */
1024 static int bnxt_dev_configure_op(struct rte_eth_dev *eth_dev)
1025 {
1026 	struct bnxt *bp = eth_dev->data->dev_private;
1027 	uint64_t rx_offloads = eth_dev->data->dev_conf.rxmode.offloads;
1028 	int rc;
1029 
1030 	bp->rx_queues = (void *)eth_dev->data->rx_queues;
1031 	bp->tx_queues = (void *)eth_dev->data->tx_queues;
1032 	bp->tx_nr_rings = eth_dev->data->nb_tx_queues;
1033 	bp->rx_nr_rings = eth_dev->data->nb_rx_queues;
1034 
1035 	rc = is_bnxt_in_error(bp);
1036 	if (rc)
1037 		return rc;
1038 
1039 	if (BNXT_VF(bp) && (bp->flags & BNXT_FLAG_NEW_RM)) {
1040 		rc = bnxt_hwrm_check_vf_rings(bp);
1041 		if (rc) {
1042 			PMD_DRV_LOG(ERR, "HWRM insufficient resources\n");
1043 			return -ENOSPC;
1044 		}
1045 
1046 		/* If a resource has already been allocated - in this case
1047 		 * it is the async completion ring, free it. Reallocate it after
1048 		 * resource reservation. This will ensure the resource counts
1049 		 * are calculated correctly.
1050 		 */
1051 
1052 		pthread_mutex_lock(&bp->def_cp_lock);
1053 
1054 		if (!BNXT_HAS_NQ(bp) && bp->async_cp_ring) {
1055 			bnxt_disable_int(bp);
1056 			bnxt_free_cp_ring(bp, bp->async_cp_ring);
1057 		}
1058 
1059 		rc = bnxt_hwrm_func_reserve_vf_resc(bp, false);
1060 		if (rc) {
1061 			PMD_DRV_LOG(ERR, "HWRM resource alloc fail:%x\n", rc);
1062 			pthread_mutex_unlock(&bp->def_cp_lock);
1063 			return -ENOSPC;
1064 		}
1065 
1066 		if (!BNXT_HAS_NQ(bp) && bp->async_cp_ring) {
1067 			rc = bnxt_alloc_async_cp_ring(bp);
1068 			if (rc) {
1069 				pthread_mutex_unlock(&bp->def_cp_lock);
1070 				return rc;
1071 			}
1072 			bnxt_enable_int(bp);
1073 		}
1074 
1075 		pthread_mutex_unlock(&bp->def_cp_lock);
1076 	}
1077 
1078 	/* Inherit new configurations */
1079 	if (eth_dev->data->nb_rx_queues > bp->max_rx_rings ||
1080 	    eth_dev->data->nb_tx_queues > bp->max_tx_rings ||
1081 	    eth_dev->data->nb_rx_queues + eth_dev->data->nb_tx_queues
1082 		+ BNXT_NUM_ASYNC_CPR(bp) > bp->max_cp_rings ||
1083 	    eth_dev->data->nb_rx_queues + eth_dev->data->nb_tx_queues >
1084 	    bp->max_stat_ctx)
1085 		goto resource_error;
1086 
1087 	if (BNXT_HAS_RING_GRPS(bp) &&
1088 	    (uint32_t)(eth_dev->data->nb_rx_queues) > bp->max_ring_grps)
1089 		goto resource_error;
1090 
1091 	if (!(eth_dev->data->dev_conf.rxmode.mq_mode & ETH_MQ_RX_RSS) &&
1092 	    bp->max_vnics < eth_dev->data->nb_rx_queues)
1093 		goto resource_error;
1094 
1095 	bp->rx_cp_nr_rings = bp->rx_nr_rings;
1096 	bp->tx_cp_nr_rings = bp->tx_nr_rings;
1097 
1098 	if (eth_dev->data->dev_conf.rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG)
1099 		rx_offloads |= DEV_RX_OFFLOAD_RSS_HASH;
1100 	eth_dev->data->dev_conf.rxmode.offloads = rx_offloads;
1101 
1102 	if (rx_offloads & DEV_RX_OFFLOAD_JUMBO_FRAME) {
1103 		eth_dev->data->mtu =
1104 			eth_dev->data->dev_conf.rxmode.max_rx_pkt_len -
1105 			RTE_ETHER_HDR_LEN - RTE_ETHER_CRC_LEN - VLAN_TAG_SIZE *
1106 			BNXT_NUM_VLANS;
1107 		bnxt_mtu_set_op(eth_dev, eth_dev->data->mtu);
1108 	}
1109 	return 0;
1110 
1111 resource_error:
1112 	PMD_DRV_LOG(ERR,
1113 		    "Insufficient resources to support requested config\n");
1114 	PMD_DRV_LOG(ERR,
1115 		    "Num Queues Requested: Tx %d, Rx %d\n",
1116 		    eth_dev->data->nb_tx_queues,
1117 		    eth_dev->data->nb_rx_queues);
1118 	PMD_DRV_LOG(ERR,
1119 		    "MAX: TxQ %d, RxQ %d, CQ %d Stat %d, Grp %d, Vnic %d\n",
1120 		    bp->max_tx_rings, bp->max_rx_rings, bp->max_cp_rings,
1121 		    bp->max_stat_ctx, bp->max_ring_grps, bp->max_vnics);
1122 	return -ENOSPC;
1123 }
1124 
1125 void bnxt_print_link_info(struct rte_eth_dev *eth_dev)
1126 {
1127 	struct rte_eth_link *link = &eth_dev->data->dev_link;
1128 
1129 	if (link->link_status)
1130 		PMD_DRV_LOG(INFO, "Port %d Link Up - speed %u Mbps - %s\n",
1131 			eth_dev->data->port_id,
1132 			(uint32_t)link->link_speed,
1133 			(link->link_duplex == ETH_LINK_FULL_DUPLEX) ?
1134 			("full-duplex") : ("half-duplex\n"));
1135 	else
1136 		PMD_DRV_LOG(INFO, "Port %d Link Down\n",
1137 			eth_dev->data->port_id);
1138 }
1139 
1140 /*
1141  * Determine whether the current configuration requires support for scattered
1142  * receive; return 1 if scattered receive is required and 0 if not.
1143  */
1144 static int bnxt_scattered_rx(struct rte_eth_dev *eth_dev)
1145 {
1146 	uint16_t buf_size;
1147 	int i;
1148 
1149 	if (eth_dev->data->dev_conf.rxmode.offloads & DEV_RX_OFFLOAD_SCATTER)
1150 		return 1;
1151 
1152 	if (eth_dev->data->dev_conf.rxmode.offloads & DEV_RX_OFFLOAD_TCP_LRO)
1153 		return 1;
1154 
1155 	for (i = 0; i < eth_dev->data->nb_rx_queues; i++) {
1156 		struct bnxt_rx_queue *rxq = eth_dev->data->rx_queues[i];
1157 
1158 		buf_size = (uint16_t)(rte_pktmbuf_data_room_size(rxq->mb_pool) -
1159 				      RTE_PKTMBUF_HEADROOM);
1160 		if (eth_dev->data->dev_conf.rxmode.max_rx_pkt_len > buf_size)
1161 			return 1;
1162 	}
1163 	return 0;
1164 }
1165 
1166 static eth_rx_burst_t
1167 bnxt_receive_function(struct rte_eth_dev *eth_dev)
1168 {
1169 	struct bnxt *bp = eth_dev->data->dev_private;
1170 
1171 	/* Disable vector mode RX for Stingray2 for now */
1172 	if (BNXT_CHIP_SR2(bp)) {
1173 		bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
1174 		return bnxt_recv_pkts;
1175 	}
1176 
1177 #if defined(RTE_ARCH_X86) || defined(RTE_ARCH_ARM64)
1178 #ifndef RTE_LIBRTE_IEEE1588
1179 	/*
1180 	 * Vector mode receive can be enabled only if scatter rx is not
1181 	 * in use and rx offloads are limited to VLAN stripping and
1182 	 * CRC stripping.
1183 	 */
1184 	if (!eth_dev->data->scattered_rx &&
1185 	    !(eth_dev->data->dev_conf.rxmode.offloads &
1186 	      ~(DEV_RX_OFFLOAD_VLAN_STRIP |
1187 		DEV_RX_OFFLOAD_KEEP_CRC |
1188 		DEV_RX_OFFLOAD_JUMBO_FRAME |
1189 		DEV_RX_OFFLOAD_IPV4_CKSUM |
1190 		DEV_RX_OFFLOAD_UDP_CKSUM |
1191 		DEV_RX_OFFLOAD_TCP_CKSUM |
1192 		DEV_RX_OFFLOAD_OUTER_IPV4_CKSUM |
1193 		DEV_RX_OFFLOAD_OUTER_UDP_CKSUM |
1194 		DEV_RX_OFFLOAD_RSS_HASH |
1195 		DEV_RX_OFFLOAD_VLAN_FILTER)) &&
1196 	    !BNXT_TRUFLOW_EN(bp) && BNXT_NUM_ASYNC_CPR(bp) &&
1197 	    rte_vect_get_max_simd_bitwidth() >= RTE_VECT_SIMD_128) {
1198 		PMD_DRV_LOG(INFO, "Using vector mode receive for port %d\n",
1199 			    eth_dev->data->port_id);
1200 		bp->flags |= BNXT_FLAG_RX_VECTOR_PKT_MODE;
1201 		return bnxt_recv_pkts_vec;
1202 	}
1203 	PMD_DRV_LOG(INFO, "Vector mode receive disabled for port %d\n",
1204 		    eth_dev->data->port_id);
1205 	PMD_DRV_LOG(INFO,
1206 		    "Port %d scatter: %d rx offload: %" PRIX64 "\n",
1207 		    eth_dev->data->port_id,
1208 		    eth_dev->data->scattered_rx,
1209 		    eth_dev->data->dev_conf.rxmode.offloads);
1210 #endif
1211 #endif
1212 	bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
1213 	return bnxt_recv_pkts;
1214 }
1215 
1216 static eth_tx_burst_t
1217 bnxt_transmit_function(struct rte_eth_dev *eth_dev)
1218 {
1219 	struct bnxt *bp = eth_dev->data->dev_private;
1220 
1221 	/* Disable vector mode TX for Stingray2 for now */
1222 	if (BNXT_CHIP_SR2(bp))
1223 		return bnxt_xmit_pkts;
1224 
1225 #if defined(RTE_ARCH_X86) || defined(RTE_ARCH_ARM64)
1226 #ifndef RTE_LIBRTE_IEEE1588
1227 	uint64_t offloads = eth_dev->data->dev_conf.txmode.offloads;
1228 
1229 	/*
1230 	 * Vector mode transmit can be enabled only if not using scatter rx
1231 	 * or tx offloads.
1232 	 */
1233 	if (!eth_dev->data->scattered_rx &&
1234 	    !(offloads & ~DEV_TX_OFFLOAD_MBUF_FAST_FREE) &&
1235 	    !BNXT_TRUFLOW_EN(bp) &&
1236 	    rte_vect_get_max_simd_bitwidth() >= RTE_VECT_SIMD_128) {
1237 		PMD_DRV_LOG(INFO, "Using vector mode transmit for port %d\n",
1238 			    eth_dev->data->port_id);
1239 		return bnxt_xmit_pkts_vec;
1240 	}
1241 	PMD_DRV_LOG(INFO, "Vector mode transmit disabled for port %d\n",
1242 		    eth_dev->data->port_id);
1243 	PMD_DRV_LOG(INFO,
1244 		    "Port %d scatter: %d tx offload: %" PRIX64 "\n",
1245 		    eth_dev->data->port_id,
1246 		    eth_dev->data->scattered_rx,
1247 		    offloads);
1248 #endif
1249 #endif
1250 	return bnxt_xmit_pkts;
1251 }
1252 
1253 static int bnxt_handle_if_change_status(struct bnxt *bp)
1254 {
1255 	int rc;
1256 
1257 	/* Since fw has undergone a reset and lost all contexts,
1258 	 * set fatal flag to not issue hwrm during cleanup
1259 	 */
1260 	bp->flags |= BNXT_FLAG_FATAL_ERROR;
1261 	bnxt_uninit_resources(bp, true);
1262 
1263 	/* clear fatal flag so that re-init happens */
1264 	bp->flags &= ~BNXT_FLAG_FATAL_ERROR;
1265 	rc = bnxt_init_resources(bp, true);
1266 
1267 	bp->flags &= ~BNXT_FLAG_IF_CHANGE_HOT_FW_RESET_DONE;
1268 
1269 	return rc;
1270 }
1271 
1272 static int bnxt_dev_set_link_up_op(struct rte_eth_dev *eth_dev)
1273 {
1274 	struct bnxt *bp = eth_dev->data->dev_private;
1275 	int rc = 0;
1276 
1277 	if (!BNXT_SINGLE_PF(bp))
1278 		return -ENOTSUP;
1279 
1280 	if (!bp->link_info->link_up)
1281 		rc = bnxt_set_hwrm_link_config(bp, true);
1282 	if (!rc)
1283 		eth_dev->data->dev_link.link_status = 1;
1284 
1285 	bnxt_print_link_info(eth_dev);
1286 	return rc;
1287 }
1288 
1289 static int bnxt_dev_set_link_down_op(struct rte_eth_dev *eth_dev)
1290 {
1291 	struct bnxt *bp = eth_dev->data->dev_private;
1292 
1293 	if (!BNXT_SINGLE_PF(bp))
1294 		return -ENOTSUP;
1295 
1296 	eth_dev->data->dev_link.link_status = 0;
1297 	bnxt_set_hwrm_link_config(bp, false);
1298 	bp->link_info->link_up = 0;
1299 
1300 	return 0;
1301 }
1302 
1303 static void bnxt_free_switch_domain(struct bnxt *bp)
1304 {
1305 	int rc = 0;
1306 
1307 	if (!(BNXT_PF(bp) || BNXT_VF_IS_TRUSTED(bp)))
1308 		return;
1309 
1310 	rc = rte_eth_switch_domain_free(bp->switch_domain_id);
1311 	if (rc)
1312 		PMD_DRV_LOG(ERR, "free switch domain:%d fail: %d\n",
1313 			    bp->switch_domain_id, rc);
1314 }
1315 
1316 static void bnxt_ptp_get_current_time(void *arg)
1317 {
1318 	struct bnxt *bp = arg;
1319 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
1320 	int rc;
1321 
1322 	rc = is_bnxt_in_error(bp);
1323 	if (rc)
1324 		return;
1325 
1326 	if (!ptp)
1327 		return;
1328 
1329 	bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_CURRENT_TIME,
1330 				&ptp->current_time);
1331 
1332 	rc = rte_eal_alarm_set(US_PER_S, bnxt_ptp_get_current_time, (void *)bp);
1333 	if (rc != 0) {
1334 		PMD_DRV_LOG(ERR, "Failed to re-schedule PTP alarm\n");
1335 		bp->flags2 &= ~BNXT_FLAGS2_PTP_ALARM_SCHEDULED;
1336 	}
1337 }
1338 
1339 static int bnxt_schedule_ptp_alarm(struct bnxt *bp)
1340 {
1341 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
1342 	int rc;
1343 
1344 	if (bp->flags2 & BNXT_FLAGS2_PTP_ALARM_SCHEDULED)
1345 		return 0;
1346 
1347 	bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_CURRENT_TIME,
1348 				&ptp->current_time);
1349 
1350 	rc = rte_eal_alarm_set(US_PER_S, bnxt_ptp_get_current_time, (void *)bp);
1351 	return rc;
1352 }
1353 
1354 static void bnxt_cancel_ptp_alarm(struct bnxt *bp)
1355 {
1356 	if (bp->flags2 & BNXT_FLAGS2_PTP_ALARM_SCHEDULED) {
1357 		rte_eal_alarm_cancel(bnxt_ptp_get_current_time, (void *)bp);
1358 		bp->flags2 &= ~BNXT_FLAGS2_PTP_ALARM_SCHEDULED;
1359 	}
1360 }
1361 
1362 static void bnxt_ptp_stop(struct bnxt *bp)
1363 {
1364 	bnxt_cancel_ptp_alarm(bp);
1365 	bp->flags2 &= ~BNXT_FLAGS2_PTP_TIMESYNC_ENABLED;
1366 }
1367 
1368 static int bnxt_ptp_start(struct bnxt *bp)
1369 {
1370 	int rc;
1371 
1372 	rc = bnxt_schedule_ptp_alarm(bp);
1373 	if (rc != 0) {
1374 		PMD_DRV_LOG(ERR, "Failed to schedule PTP alarm\n");
1375 	} else {
1376 		bp->flags2 |= BNXT_FLAGS2_PTP_TIMESYNC_ENABLED;
1377 		bp->flags2 |= BNXT_FLAGS2_PTP_ALARM_SCHEDULED;
1378 	}
1379 
1380 	return rc;
1381 }
1382 
1383 static int bnxt_dev_stop(struct rte_eth_dev *eth_dev)
1384 {
1385 	struct bnxt *bp = eth_dev->data->dev_private;
1386 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
1387 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
1388 	struct rte_eth_link link;
1389 	int ret;
1390 
1391 	eth_dev->data->dev_started = 0;
1392 	eth_dev->data->scattered_rx = 0;
1393 
1394 	/* Prevent crashes when queues are still in use */
1395 	eth_dev->rx_pkt_burst = &bnxt_dummy_recv_pkts;
1396 	eth_dev->tx_pkt_burst = &bnxt_dummy_xmit_pkts;
1397 
1398 	bnxt_disable_int(bp);
1399 
1400 	/* disable uio/vfio intr/eventfd mapping */
1401 	rte_intr_disable(intr_handle);
1402 
1403 	/* Stop the child representors for this device */
1404 	ret = bnxt_rep_stop_all(bp);
1405 	if (ret != 0)
1406 		return ret;
1407 
1408 	/* delete the bnxt ULP port details */
1409 	bnxt_ulp_port_deinit(bp);
1410 
1411 	bnxt_cancel_fw_health_check(bp);
1412 
1413 	if (BNXT_P5_PTP_TIMESYNC_ENABLED(bp))
1414 		bnxt_cancel_ptp_alarm(bp);
1415 
1416 	/* Do not bring link down during reset recovery */
1417 	if (!is_bnxt_in_error(bp)) {
1418 		bnxt_dev_set_link_down_op(eth_dev);
1419 		/* Wait for link to be reset */
1420 		if (BNXT_SINGLE_PF(bp))
1421 			rte_delay_ms(500);
1422 		/* clear the recorded link status */
1423 		memset(&link, 0, sizeof(link));
1424 		rte_eth_linkstatus_set(eth_dev, &link);
1425 	}
1426 
1427 	/* Clean queue intr-vector mapping */
1428 	rte_intr_efd_disable(intr_handle);
1429 	if (intr_handle->intr_vec != NULL) {
1430 		rte_free(intr_handle->intr_vec);
1431 		intr_handle->intr_vec = NULL;
1432 	}
1433 
1434 	bnxt_hwrm_port_clr_stats(bp);
1435 	bnxt_free_tx_mbufs(bp);
1436 	bnxt_free_rx_mbufs(bp);
1437 	/* Process any remaining notifications in default completion queue */
1438 	bnxt_int_handler(eth_dev);
1439 	bnxt_shutdown_nic(bp);
1440 	bnxt_hwrm_if_change(bp, false);
1441 
1442 	rte_free(bp->mark_table);
1443 	bp->mark_table = NULL;
1444 
1445 	bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
1446 	bp->rx_cosq_cnt = 0;
1447 	/* All filters are deleted on a port stop. */
1448 	if (BNXT_FLOW_XSTATS_EN(bp))
1449 		bp->flow_stat->flow_count = 0;
1450 
1451 	return 0;
1452 }
1453 
1454 /* Unload the driver, release resources */
1455 static int bnxt_dev_stop_op(struct rte_eth_dev *eth_dev)
1456 {
1457 	struct bnxt *bp = eth_dev->data->dev_private;
1458 
1459 	pthread_mutex_lock(&bp->err_recovery_lock);
1460 	if (bp->flags & BNXT_FLAG_FW_RESET) {
1461 		PMD_DRV_LOG(ERR,
1462 			    "Adapter recovering from error..Please retry\n");
1463 		pthread_mutex_unlock(&bp->err_recovery_lock);
1464 		return -EAGAIN;
1465 	}
1466 	pthread_mutex_unlock(&bp->err_recovery_lock);
1467 
1468 	return bnxt_dev_stop(eth_dev);
1469 }
1470 
1471 static int bnxt_dev_start_op(struct rte_eth_dev *eth_dev)
1472 {
1473 	struct bnxt *bp = eth_dev->data->dev_private;
1474 	uint64_t rx_offloads = eth_dev->data->dev_conf.rxmode.offloads;
1475 	int vlan_mask = 0;
1476 	int rc, retry_cnt = BNXT_IF_CHANGE_RETRY_COUNT;
1477 
1478 	if (!eth_dev->data->nb_tx_queues || !eth_dev->data->nb_rx_queues) {
1479 		PMD_DRV_LOG(ERR, "Queues are not configured yet!\n");
1480 		return -EINVAL;
1481 	}
1482 
1483 	if (bp->rx_cp_nr_rings > RTE_ETHDEV_QUEUE_STAT_CNTRS)
1484 		PMD_DRV_LOG(ERR,
1485 			    "RxQ cnt %d > RTE_ETHDEV_QUEUE_STAT_CNTRS %d\n",
1486 			    bp->rx_cp_nr_rings, RTE_ETHDEV_QUEUE_STAT_CNTRS);
1487 
1488 	do {
1489 		rc = bnxt_hwrm_if_change(bp, true);
1490 		if (rc == 0 || rc != -EAGAIN)
1491 			break;
1492 
1493 		rte_delay_ms(BNXT_IF_CHANGE_RETRY_INTERVAL);
1494 	} while (retry_cnt--);
1495 
1496 	if (rc)
1497 		return rc;
1498 
1499 	if (bp->flags & BNXT_FLAG_IF_CHANGE_HOT_FW_RESET_DONE) {
1500 		rc = bnxt_handle_if_change_status(bp);
1501 		if (rc)
1502 			return rc;
1503 	}
1504 
1505 	bnxt_enable_int(bp);
1506 
1507 	eth_dev->data->scattered_rx = bnxt_scattered_rx(eth_dev);
1508 
1509 	rc = bnxt_start_nic(bp);
1510 	if (rc)
1511 		goto error;
1512 
1513 	eth_dev->data->dev_started = 1;
1514 
1515 	bnxt_link_update_op(eth_dev, 1);
1516 
1517 	if (rx_offloads & DEV_RX_OFFLOAD_VLAN_FILTER)
1518 		vlan_mask |= ETH_VLAN_FILTER_MASK;
1519 	if (rx_offloads & DEV_RX_OFFLOAD_VLAN_STRIP)
1520 		vlan_mask |= ETH_VLAN_STRIP_MASK;
1521 	rc = bnxt_vlan_offload_set_op(eth_dev, vlan_mask);
1522 	if (rc)
1523 		goto error;
1524 
1525 	/* Initialize bnxt ULP port details */
1526 	rc = bnxt_ulp_port_init(bp);
1527 	if (rc)
1528 		goto error;
1529 
1530 	eth_dev->rx_pkt_burst = bnxt_receive_function(eth_dev);
1531 	eth_dev->tx_pkt_burst = bnxt_transmit_function(eth_dev);
1532 
1533 	bnxt_schedule_fw_health_check(bp);
1534 
1535 	if (BNXT_P5_PTP_TIMESYNC_ENABLED(bp))
1536 		bnxt_schedule_ptp_alarm(bp);
1537 
1538 	return 0;
1539 
1540 error:
1541 	bnxt_dev_stop(eth_dev);
1542 	return rc;
1543 }
1544 
1545 static void
1546 bnxt_uninit_locks(struct bnxt *bp)
1547 {
1548 	pthread_mutex_destroy(&bp->flow_lock);
1549 	pthread_mutex_destroy(&bp->def_cp_lock);
1550 	pthread_mutex_destroy(&bp->health_check_lock);
1551 	pthread_mutex_destroy(&bp->err_recovery_lock);
1552 	if (bp->rep_info) {
1553 		pthread_mutex_destroy(&bp->rep_info->vfr_lock);
1554 		pthread_mutex_destroy(&bp->rep_info->vfr_start_lock);
1555 	}
1556 }
1557 
1558 static void bnxt_drv_uninit(struct bnxt *bp)
1559 {
1560 	bnxt_free_leds_info(bp);
1561 	bnxt_free_cos_queues(bp);
1562 	bnxt_free_link_info(bp);
1563 	bnxt_free_parent_info(bp);
1564 	bnxt_uninit_locks(bp);
1565 
1566 	rte_memzone_free((const struct rte_memzone *)bp->tx_mem_zone);
1567 	bp->tx_mem_zone = NULL;
1568 	rte_memzone_free((const struct rte_memzone *)bp->rx_mem_zone);
1569 	bp->rx_mem_zone = NULL;
1570 
1571 	bnxt_free_vf_info(bp);
1572 	bnxt_free_pf_info(bp);
1573 
1574 	rte_free(bp->grp_info);
1575 	bp->grp_info = NULL;
1576 }
1577 
1578 static int bnxt_dev_close_op(struct rte_eth_dev *eth_dev)
1579 {
1580 	struct bnxt *bp = eth_dev->data->dev_private;
1581 	int ret = 0;
1582 
1583 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
1584 		return 0;
1585 
1586 	pthread_mutex_lock(&bp->err_recovery_lock);
1587 	if (bp->flags & BNXT_FLAG_FW_RESET) {
1588 		PMD_DRV_LOG(ERR,
1589 			    "Adapter recovering from error...Please retry\n");
1590 		pthread_mutex_unlock(&bp->err_recovery_lock);
1591 		return -EAGAIN;
1592 	}
1593 	pthread_mutex_unlock(&bp->err_recovery_lock);
1594 
1595 	/* cancel the recovery handler before remove dev */
1596 	rte_eal_alarm_cancel(bnxt_dev_reset_and_resume, (void *)bp);
1597 	rte_eal_alarm_cancel(bnxt_dev_recover, (void *)bp);
1598 	bnxt_cancel_fc_thread(bp);
1599 
1600 	if (eth_dev->data->dev_started)
1601 		ret = bnxt_dev_stop(eth_dev);
1602 
1603 	bnxt_uninit_resources(bp, false);
1604 
1605 	bnxt_drv_uninit(bp);
1606 
1607 	return ret;
1608 }
1609 
1610 static void bnxt_mac_addr_remove_op(struct rte_eth_dev *eth_dev,
1611 				    uint32_t index)
1612 {
1613 	struct bnxt *bp = eth_dev->data->dev_private;
1614 	uint64_t pool_mask = eth_dev->data->mac_pool_sel[index];
1615 	struct bnxt_vnic_info *vnic;
1616 	struct bnxt_filter_info *filter, *temp_filter;
1617 	uint32_t i;
1618 
1619 	if (is_bnxt_in_error(bp))
1620 		return;
1621 
1622 	/*
1623 	 * Loop through all VNICs from the specified filter flow pools to
1624 	 * remove the corresponding MAC addr filter
1625 	 */
1626 	for (i = 0; i < bp->nr_vnics; i++) {
1627 		if (!(pool_mask & (1ULL << i)))
1628 			continue;
1629 
1630 		vnic = &bp->vnic_info[i];
1631 		filter = STAILQ_FIRST(&vnic->filter);
1632 		while (filter) {
1633 			temp_filter = STAILQ_NEXT(filter, next);
1634 			if (filter->mac_index == index) {
1635 				STAILQ_REMOVE(&vnic->filter, filter,
1636 						bnxt_filter_info, next);
1637 				bnxt_hwrm_clear_l2_filter(bp, filter);
1638 				bnxt_free_filter(bp, filter);
1639 			}
1640 			filter = temp_filter;
1641 		}
1642 	}
1643 }
1644 
1645 static int bnxt_add_mac_filter(struct bnxt *bp, struct bnxt_vnic_info *vnic,
1646 			       struct rte_ether_addr *mac_addr, uint32_t index,
1647 			       uint32_t pool)
1648 {
1649 	struct bnxt_filter_info *filter;
1650 	int rc = 0;
1651 
1652 	/* Attach requested MAC address to the new l2_filter */
1653 	STAILQ_FOREACH(filter, &vnic->filter, next) {
1654 		if (filter->mac_index == index) {
1655 			PMD_DRV_LOG(DEBUG,
1656 				    "MAC addr already existed for pool %d\n",
1657 				    pool);
1658 			return 0;
1659 		}
1660 	}
1661 
1662 	filter = bnxt_alloc_filter(bp);
1663 	if (!filter) {
1664 		PMD_DRV_LOG(ERR, "L2 filter alloc failed\n");
1665 		return -ENODEV;
1666 	}
1667 
1668 	/* bnxt_alloc_filter copies default MAC to filter->l2_addr. So,
1669 	 * if the MAC that's been programmed now is a different one, then,
1670 	 * copy that addr to filter->l2_addr
1671 	 */
1672 	if (mac_addr)
1673 		memcpy(filter->l2_addr, mac_addr, RTE_ETHER_ADDR_LEN);
1674 	filter->flags |= HWRM_CFA_L2_FILTER_ALLOC_INPUT_FLAGS_OUTERMOST;
1675 
1676 	rc = bnxt_hwrm_set_l2_filter(bp, vnic->fw_vnic_id, filter);
1677 	if (!rc) {
1678 		filter->mac_index = index;
1679 		if (filter->mac_index == 0)
1680 			STAILQ_INSERT_HEAD(&vnic->filter, filter, next);
1681 		else
1682 			STAILQ_INSERT_TAIL(&vnic->filter, filter, next);
1683 	} else {
1684 		bnxt_free_filter(bp, filter);
1685 	}
1686 
1687 	return rc;
1688 }
1689 
1690 static int bnxt_mac_addr_add_op(struct rte_eth_dev *eth_dev,
1691 				struct rte_ether_addr *mac_addr,
1692 				uint32_t index, uint32_t pool)
1693 {
1694 	struct bnxt *bp = eth_dev->data->dev_private;
1695 	struct bnxt_vnic_info *vnic = &bp->vnic_info[pool];
1696 	int rc = 0;
1697 
1698 	rc = is_bnxt_in_error(bp);
1699 	if (rc)
1700 		return rc;
1701 
1702 	if (BNXT_VF(bp) && !BNXT_VF_IS_TRUSTED(bp)) {
1703 		PMD_DRV_LOG(ERR, "Cannot add MAC address to a VF interface\n");
1704 		return -ENOTSUP;
1705 	}
1706 
1707 	if (!vnic) {
1708 		PMD_DRV_LOG(ERR, "VNIC not found for pool %d!\n", pool);
1709 		return -EINVAL;
1710 	}
1711 
1712 	/* Filter settings will get applied when port is started */
1713 	if (!eth_dev->data->dev_started)
1714 		return 0;
1715 
1716 	rc = bnxt_add_mac_filter(bp, vnic, mac_addr, index, pool);
1717 
1718 	return rc;
1719 }
1720 
1721 int bnxt_link_update_op(struct rte_eth_dev *eth_dev, int wait_to_complete)
1722 {
1723 	int rc = 0;
1724 	struct bnxt *bp = eth_dev->data->dev_private;
1725 	struct rte_eth_link new;
1726 	int cnt = wait_to_complete ? BNXT_MAX_LINK_WAIT_CNT :
1727 			BNXT_MIN_LINK_WAIT_CNT;
1728 
1729 	rc = is_bnxt_in_error(bp);
1730 	if (rc)
1731 		return rc;
1732 
1733 	memset(&new, 0, sizeof(new));
1734 
1735 	if (bp->link_info == NULL)
1736 		goto out;
1737 
1738 	do {
1739 		/* Retrieve link info from hardware */
1740 		rc = bnxt_get_hwrm_link_config(bp, &new);
1741 		if (rc) {
1742 			new.link_speed = ETH_LINK_SPEED_100M;
1743 			new.link_duplex = ETH_LINK_FULL_DUPLEX;
1744 			PMD_DRV_LOG(ERR,
1745 				"Failed to retrieve link rc = 0x%x!\n", rc);
1746 			goto out;
1747 		}
1748 
1749 		if (!wait_to_complete || new.link_status)
1750 			break;
1751 
1752 		rte_delay_ms(BNXT_LINK_WAIT_INTERVAL);
1753 	} while (cnt--);
1754 
1755 	/* Only single function PF can bring phy down.
1756 	 * When port is stopped, report link down for VF/MH/NPAR functions.
1757 	 */
1758 	if (!BNXT_SINGLE_PF(bp) && !eth_dev->data->dev_started)
1759 		memset(&new, 0, sizeof(new));
1760 
1761 out:
1762 	/* Timed out or success */
1763 	if (new.link_status != eth_dev->data->dev_link.link_status ||
1764 	    new.link_speed != eth_dev->data->dev_link.link_speed) {
1765 		rte_eth_linkstatus_set(eth_dev, &new);
1766 
1767 		rte_eth_dev_callback_process(eth_dev,
1768 					     RTE_ETH_EVENT_INTR_LSC,
1769 					     NULL);
1770 
1771 		bnxt_print_link_info(eth_dev);
1772 	}
1773 
1774 	return rc;
1775 }
1776 
1777 static int bnxt_promiscuous_enable_op(struct rte_eth_dev *eth_dev)
1778 {
1779 	struct bnxt *bp = eth_dev->data->dev_private;
1780 	struct bnxt_vnic_info *vnic;
1781 	uint32_t old_flags;
1782 	int rc;
1783 
1784 	rc = is_bnxt_in_error(bp);
1785 	if (rc)
1786 		return rc;
1787 
1788 	/* Filter settings will get applied when port is started */
1789 	if (!eth_dev->data->dev_started)
1790 		return 0;
1791 
1792 	if (bp->vnic_info == NULL)
1793 		return 0;
1794 
1795 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
1796 
1797 	old_flags = vnic->flags;
1798 	vnic->flags |= BNXT_VNIC_INFO_PROMISC;
1799 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1800 	if (rc != 0)
1801 		vnic->flags = old_flags;
1802 
1803 	return rc;
1804 }
1805 
1806 static int bnxt_promiscuous_disable_op(struct rte_eth_dev *eth_dev)
1807 {
1808 	struct bnxt *bp = eth_dev->data->dev_private;
1809 	struct bnxt_vnic_info *vnic;
1810 	uint32_t old_flags;
1811 	int rc;
1812 
1813 	rc = is_bnxt_in_error(bp);
1814 	if (rc)
1815 		return rc;
1816 
1817 	/* Filter settings will get applied when port is started */
1818 	if (!eth_dev->data->dev_started)
1819 		return 0;
1820 
1821 	if (bp->vnic_info == NULL)
1822 		return 0;
1823 
1824 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
1825 
1826 	old_flags = vnic->flags;
1827 	vnic->flags &= ~BNXT_VNIC_INFO_PROMISC;
1828 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1829 	if (rc != 0)
1830 		vnic->flags = old_flags;
1831 
1832 	return rc;
1833 }
1834 
1835 static int bnxt_allmulticast_enable_op(struct rte_eth_dev *eth_dev)
1836 {
1837 	struct bnxt *bp = eth_dev->data->dev_private;
1838 	struct bnxt_vnic_info *vnic;
1839 	uint32_t old_flags;
1840 	int rc;
1841 
1842 	rc = is_bnxt_in_error(bp);
1843 	if (rc)
1844 		return rc;
1845 
1846 	/* Filter settings will get applied when port is started */
1847 	if (!eth_dev->data->dev_started)
1848 		return 0;
1849 
1850 	if (bp->vnic_info == NULL)
1851 		return 0;
1852 
1853 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
1854 
1855 	old_flags = vnic->flags;
1856 	vnic->flags |= BNXT_VNIC_INFO_ALLMULTI;
1857 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1858 	if (rc != 0)
1859 		vnic->flags = old_flags;
1860 
1861 	return rc;
1862 }
1863 
1864 static int bnxt_allmulticast_disable_op(struct rte_eth_dev *eth_dev)
1865 {
1866 	struct bnxt *bp = eth_dev->data->dev_private;
1867 	struct bnxt_vnic_info *vnic;
1868 	uint32_t old_flags;
1869 	int rc;
1870 
1871 	rc = is_bnxt_in_error(bp);
1872 	if (rc)
1873 		return rc;
1874 
1875 	/* Filter settings will get applied when port is started */
1876 	if (!eth_dev->data->dev_started)
1877 		return 0;
1878 
1879 	if (bp->vnic_info == NULL)
1880 		return 0;
1881 
1882 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
1883 
1884 	old_flags = vnic->flags;
1885 	vnic->flags &= ~BNXT_VNIC_INFO_ALLMULTI;
1886 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1887 	if (rc != 0)
1888 		vnic->flags = old_flags;
1889 
1890 	return rc;
1891 }
1892 
1893 /* Return bnxt_rx_queue pointer corresponding to a given rxq. */
1894 static struct bnxt_rx_queue *bnxt_qid_to_rxq(struct bnxt *bp, uint16_t qid)
1895 {
1896 	if (qid >= bp->rx_nr_rings)
1897 		return NULL;
1898 
1899 	return bp->eth_dev->data->rx_queues[qid];
1900 }
1901 
1902 /* Return rxq corresponding to a given rss table ring/group ID. */
1903 static uint16_t bnxt_rss_to_qid(struct bnxt *bp, uint16_t fwr)
1904 {
1905 	struct bnxt_rx_queue *rxq;
1906 	unsigned int i;
1907 
1908 	if (!BNXT_HAS_RING_GRPS(bp)) {
1909 		for (i = 0; i < bp->rx_nr_rings; i++) {
1910 			rxq = bp->eth_dev->data->rx_queues[i];
1911 			if (rxq->rx_ring->rx_ring_struct->fw_ring_id == fwr)
1912 				return rxq->index;
1913 		}
1914 	} else {
1915 		for (i = 0; i < bp->rx_nr_rings; i++) {
1916 			if (bp->grp_info[i].fw_grp_id == fwr)
1917 				return i;
1918 		}
1919 	}
1920 
1921 	return INVALID_HW_RING_ID;
1922 }
1923 
1924 static int bnxt_reta_update_op(struct rte_eth_dev *eth_dev,
1925 			    struct rte_eth_rss_reta_entry64 *reta_conf,
1926 			    uint16_t reta_size)
1927 {
1928 	struct bnxt *bp = eth_dev->data->dev_private;
1929 	struct rte_eth_conf *dev_conf = &bp->eth_dev->data->dev_conf;
1930 	struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
1931 	uint16_t tbl_size = bnxt_rss_hash_tbl_size(bp);
1932 	uint16_t idx, sft;
1933 	int i, rc;
1934 
1935 	rc = is_bnxt_in_error(bp);
1936 	if (rc)
1937 		return rc;
1938 
1939 	if (!vnic->rss_table)
1940 		return -EINVAL;
1941 
1942 	if (!(dev_conf->rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG))
1943 		return -EINVAL;
1944 
1945 	if (reta_size != tbl_size) {
1946 		PMD_DRV_LOG(ERR, "The configured hash table lookup size "
1947 			"(%d) must equal the size supported by the hardware "
1948 			"(%d)\n", reta_size, tbl_size);
1949 		return -EINVAL;
1950 	}
1951 
1952 	for (i = 0; i < reta_size; i++) {
1953 		struct bnxt_rx_queue *rxq;
1954 
1955 		idx = i / RTE_RETA_GROUP_SIZE;
1956 		sft = i % RTE_RETA_GROUP_SIZE;
1957 
1958 		if (!(reta_conf[idx].mask & (1ULL << sft)))
1959 			continue;
1960 
1961 		rxq = bnxt_qid_to_rxq(bp, reta_conf[idx].reta[sft]);
1962 		if (!rxq) {
1963 			PMD_DRV_LOG(ERR, "Invalid ring in reta_conf.\n");
1964 			return -EINVAL;
1965 		}
1966 
1967 		if (BNXT_CHIP_P5(bp)) {
1968 			vnic->rss_table[i * 2] =
1969 				rxq->rx_ring->rx_ring_struct->fw_ring_id;
1970 			vnic->rss_table[i * 2 + 1] =
1971 				rxq->cp_ring->cp_ring_struct->fw_ring_id;
1972 		} else {
1973 			vnic->rss_table[i] =
1974 			    vnic->fw_grp_ids[reta_conf[idx].reta[sft]];
1975 		}
1976 	}
1977 
1978 	rc = bnxt_hwrm_vnic_rss_cfg(bp, vnic);
1979 	return rc;
1980 }
1981 
1982 static int bnxt_reta_query_op(struct rte_eth_dev *eth_dev,
1983 			      struct rte_eth_rss_reta_entry64 *reta_conf,
1984 			      uint16_t reta_size)
1985 {
1986 	struct bnxt *bp = eth_dev->data->dev_private;
1987 	struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
1988 	uint16_t tbl_size = bnxt_rss_hash_tbl_size(bp);
1989 	uint16_t idx, sft, i;
1990 	int rc;
1991 
1992 	rc = is_bnxt_in_error(bp);
1993 	if (rc)
1994 		return rc;
1995 
1996 	/* Retrieve from the default VNIC */
1997 	if (!vnic)
1998 		return -EINVAL;
1999 	if (!vnic->rss_table)
2000 		return -EINVAL;
2001 
2002 	if (reta_size != tbl_size) {
2003 		PMD_DRV_LOG(ERR, "The configured hash table lookup size "
2004 			"(%d) must equal the size supported by the hardware "
2005 			"(%d)\n", reta_size, tbl_size);
2006 		return -EINVAL;
2007 	}
2008 
2009 	for (idx = 0, i = 0; i < reta_size; i++) {
2010 		idx = i / RTE_RETA_GROUP_SIZE;
2011 		sft = i % RTE_RETA_GROUP_SIZE;
2012 
2013 		if (reta_conf[idx].mask & (1ULL << sft)) {
2014 			uint16_t qid;
2015 
2016 			if (BNXT_CHIP_P5(bp))
2017 				qid = bnxt_rss_to_qid(bp,
2018 						      vnic->rss_table[i * 2]);
2019 			else
2020 				qid = bnxt_rss_to_qid(bp, vnic->rss_table[i]);
2021 
2022 			if (qid == INVALID_HW_RING_ID) {
2023 				PMD_DRV_LOG(ERR, "Inv. entry in rss table.\n");
2024 				return -EINVAL;
2025 			}
2026 			reta_conf[idx].reta[sft] = qid;
2027 		}
2028 	}
2029 
2030 	return 0;
2031 }
2032 
2033 static int bnxt_rss_hash_update_op(struct rte_eth_dev *eth_dev,
2034 				   struct rte_eth_rss_conf *rss_conf)
2035 {
2036 	struct bnxt *bp = eth_dev->data->dev_private;
2037 	struct rte_eth_conf *dev_conf = &bp->eth_dev->data->dev_conf;
2038 	struct bnxt_vnic_info *vnic;
2039 	int rc;
2040 
2041 	rc = is_bnxt_in_error(bp);
2042 	if (rc)
2043 		return rc;
2044 
2045 	/*
2046 	 * If RSS enablement were different than dev_configure,
2047 	 * then return -EINVAL
2048 	 */
2049 	if (dev_conf->rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG) {
2050 		if (!rss_conf->rss_hf)
2051 			PMD_DRV_LOG(ERR, "Hash type NONE\n");
2052 	} else {
2053 		if (rss_conf->rss_hf & BNXT_ETH_RSS_SUPPORT)
2054 			return -EINVAL;
2055 	}
2056 
2057 	bp->flags |= BNXT_FLAG_UPDATE_HASH;
2058 	memcpy(&eth_dev->data->dev_conf.rx_adv_conf.rss_conf,
2059 	       rss_conf,
2060 	       sizeof(*rss_conf));
2061 
2062 	/* Update the default RSS VNIC(s) */
2063 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
2064 	vnic->hash_type = bnxt_rte_to_hwrm_hash_types(rss_conf->rss_hf);
2065 	vnic->hash_mode =
2066 		bnxt_rte_to_hwrm_hash_level(bp, rss_conf->rss_hf,
2067 					    ETH_RSS_LEVEL(rss_conf->rss_hf));
2068 
2069 	/*
2070 	 * If hashkey is not specified, use the previously configured
2071 	 * hashkey
2072 	 */
2073 	if (!rss_conf->rss_key)
2074 		goto rss_config;
2075 
2076 	if (rss_conf->rss_key_len != HW_HASH_KEY_SIZE) {
2077 		PMD_DRV_LOG(ERR,
2078 			    "Invalid hashkey length, should be 16 bytes\n");
2079 		return -EINVAL;
2080 	}
2081 	memcpy(vnic->rss_hash_key, rss_conf->rss_key, rss_conf->rss_key_len);
2082 
2083 rss_config:
2084 	rc = bnxt_hwrm_vnic_rss_cfg(bp, vnic);
2085 	return rc;
2086 }
2087 
2088 static int bnxt_rss_hash_conf_get_op(struct rte_eth_dev *eth_dev,
2089 				     struct rte_eth_rss_conf *rss_conf)
2090 {
2091 	struct bnxt *bp = eth_dev->data->dev_private;
2092 	struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
2093 	int len, rc;
2094 	uint32_t hash_types;
2095 
2096 	rc = is_bnxt_in_error(bp);
2097 	if (rc)
2098 		return rc;
2099 
2100 	/* RSS configuration is the same for all VNICs */
2101 	if (vnic && vnic->rss_hash_key) {
2102 		if (rss_conf->rss_key) {
2103 			len = rss_conf->rss_key_len <= HW_HASH_KEY_SIZE ?
2104 			      rss_conf->rss_key_len : HW_HASH_KEY_SIZE;
2105 			memcpy(rss_conf->rss_key, vnic->rss_hash_key, len);
2106 		}
2107 
2108 		hash_types = vnic->hash_type;
2109 		rss_conf->rss_hf = 0;
2110 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV4) {
2111 			rss_conf->rss_hf |= ETH_RSS_IPV4;
2112 			hash_types &= ~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV4;
2113 		}
2114 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV4) {
2115 			rss_conf->rss_hf |= ETH_RSS_NONFRAG_IPV4_TCP;
2116 			hash_types &=
2117 				~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV4;
2118 		}
2119 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV4) {
2120 			rss_conf->rss_hf |= ETH_RSS_NONFRAG_IPV4_UDP;
2121 			hash_types &=
2122 				~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV4;
2123 		}
2124 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV6) {
2125 			rss_conf->rss_hf |= ETH_RSS_IPV6;
2126 			hash_types &= ~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV6;
2127 		}
2128 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV6) {
2129 			rss_conf->rss_hf |= ETH_RSS_NONFRAG_IPV6_TCP;
2130 			hash_types &=
2131 				~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV6;
2132 		}
2133 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV6) {
2134 			rss_conf->rss_hf |= ETH_RSS_NONFRAG_IPV6_UDP;
2135 			hash_types &=
2136 				~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV6;
2137 		}
2138 
2139 		rss_conf->rss_hf |=
2140 			bnxt_hwrm_to_rte_rss_level(bp, vnic->hash_mode);
2141 
2142 		if (hash_types) {
2143 			PMD_DRV_LOG(ERR,
2144 				"Unknown RSS config from firmware (%08x), RSS disabled",
2145 				vnic->hash_type);
2146 			return -ENOTSUP;
2147 		}
2148 	} else {
2149 		rss_conf->rss_hf = 0;
2150 	}
2151 	return 0;
2152 }
2153 
2154 static int bnxt_flow_ctrl_get_op(struct rte_eth_dev *dev,
2155 			       struct rte_eth_fc_conf *fc_conf)
2156 {
2157 	struct bnxt *bp = dev->data->dev_private;
2158 	struct rte_eth_link link_info;
2159 	int rc;
2160 
2161 	rc = is_bnxt_in_error(bp);
2162 	if (rc)
2163 		return rc;
2164 
2165 	rc = bnxt_get_hwrm_link_config(bp, &link_info);
2166 	if (rc)
2167 		return rc;
2168 
2169 	memset(fc_conf, 0, sizeof(*fc_conf));
2170 	if (bp->link_info->auto_pause)
2171 		fc_conf->autoneg = 1;
2172 	switch (bp->link_info->pause) {
2173 	case 0:
2174 		fc_conf->mode = RTE_FC_NONE;
2175 		break;
2176 	case HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_TX:
2177 		fc_conf->mode = RTE_FC_TX_PAUSE;
2178 		break;
2179 	case HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_RX:
2180 		fc_conf->mode = RTE_FC_RX_PAUSE;
2181 		break;
2182 	case (HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_TX |
2183 			HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_RX):
2184 		fc_conf->mode = RTE_FC_FULL;
2185 		break;
2186 	}
2187 	return 0;
2188 }
2189 
2190 static int bnxt_flow_ctrl_set_op(struct rte_eth_dev *dev,
2191 			       struct rte_eth_fc_conf *fc_conf)
2192 {
2193 	struct bnxt *bp = dev->data->dev_private;
2194 	int rc;
2195 
2196 	rc = is_bnxt_in_error(bp);
2197 	if (rc)
2198 		return rc;
2199 
2200 	if (!BNXT_SINGLE_PF(bp) || BNXT_VF(bp)) {
2201 		PMD_DRV_LOG(ERR, "Flow Control Settings cannot be modified\n");
2202 		return -ENOTSUP;
2203 	}
2204 
2205 	switch (fc_conf->mode) {
2206 	case RTE_FC_NONE:
2207 		bp->link_info->auto_pause = 0;
2208 		bp->link_info->force_pause = 0;
2209 		break;
2210 	case RTE_FC_RX_PAUSE:
2211 		if (fc_conf->autoneg) {
2212 			bp->link_info->auto_pause =
2213 					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_RX;
2214 			bp->link_info->force_pause = 0;
2215 		} else {
2216 			bp->link_info->auto_pause = 0;
2217 			bp->link_info->force_pause =
2218 					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_RX;
2219 		}
2220 		break;
2221 	case RTE_FC_TX_PAUSE:
2222 		if (fc_conf->autoneg) {
2223 			bp->link_info->auto_pause =
2224 					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_TX;
2225 			bp->link_info->force_pause = 0;
2226 		} else {
2227 			bp->link_info->auto_pause = 0;
2228 			bp->link_info->force_pause =
2229 					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_TX;
2230 		}
2231 		break;
2232 	case RTE_FC_FULL:
2233 		if (fc_conf->autoneg) {
2234 			bp->link_info->auto_pause =
2235 					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_TX |
2236 					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_RX;
2237 			bp->link_info->force_pause = 0;
2238 		} else {
2239 			bp->link_info->auto_pause = 0;
2240 			bp->link_info->force_pause =
2241 					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_TX |
2242 					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_RX;
2243 		}
2244 		break;
2245 	}
2246 	return bnxt_set_hwrm_link_config(bp, true);
2247 }
2248 
2249 /* Add UDP tunneling port */
2250 static int
2251 bnxt_udp_tunnel_port_add_op(struct rte_eth_dev *eth_dev,
2252 			 struct rte_eth_udp_tunnel *udp_tunnel)
2253 {
2254 	struct bnxt *bp = eth_dev->data->dev_private;
2255 	uint16_t tunnel_type = 0;
2256 	int rc = 0;
2257 
2258 	rc = is_bnxt_in_error(bp);
2259 	if (rc)
2260 		return rc;
2261 
2262 	switch (udp_tunnel->prot_type) {
2263 	case RTE_TUNNEL_TYPE_VXLAN:
2264 		if (bp->vxlan_port_cnt) {
2265 			PMD_DRV_LOG(ERR, "Tunnel Port %d already programmed\n",
2266 				udp_tunnel->udp_port);
2267 			if (bp->vxlan_port != udp_tunnel->udp_port) {
2268 				PMD_DRV_LOG(ERR, "Only one port allowed\n");
2269 				return -ENOSPC;
2270 			}
2271 			bp->vxlan_port_cnt++;
2272 			return 0;
2273 		}
2274 		tunnel_type =
2275 			HWRM_TUNNEL_DST_PORT_ALLOC_INPUT_TUNNEL_TYPE_VXLAN;
2276 		bp->vxlan_port_cnt++;
2277 		break;
2278 	case RTE_TUNNEL_TYPE_GENEVE:
2279 		if (bp->geneve_port_cnt) {
2280 			PMD_DRV_LOG(ERR, "Tunnel Port %d already programmed\n",
2281 				udp_tunnel->udp_port);
2282 			if (bp->geneve_port != udp_tunnel->udp_port) {
2283 				PMD_DRV_LOG(ERR, "Only one port allowed\n");
2284 				return -ENOSPC;
2285 			}
2286 			bp->geneve_port_cnt++;
2287 			return 0;
2288 		}
2289 		tunnel_type =
2290 			HWRM_TUNNEL_DST_PORT_ALLOC_INPUT_TUNNEL_TYPE_GENEVE;
2291 		bp->geneve_port_cnt++;
2292 		break;
2293 	default:
2294 		PMD_DRV_LOG(ERR, "Tunnel type is not supported\n");
2295 		return -ENOTSUP;
2296 	}
2297 	rc = bnxt_hwrm_tunnel_dst_port_alloc(bp, udp_tunnel->udp_port,
2298 					     tunnel_type);
2299 	return rc;
2300 }
2301 
2302 static int
2303 bnxt_udp_tunnel_port_del_op(struct rte_eth_dev *eth_dev,
2304 			 struct rte_eth_udp_tunnel *udp_tunnel)
2305 {
2306 	struct bnxt *bp = eth_dev->data->dev_private;
2307 	uint16_t tunnel_type = 0;
2308 	uint16_t port = 0;
2309 	int rc = 0;
2310 
2311 	rc = is_bnxt_in_error(bp);
2312 	if (rc)
2313 		return rc;
2314 
2315 	switch (udp_tunnel->prot_type) {
2316 	case RTE_TUNNEL_TYPE_VXLAN:
2317 		if (!bp->vxlan_port_cnt) {
2318 			PMD_DRV_LOG(ERR, "No Tunnel port configured yet\n");
2319 			return -EINVAL;
2320 		}
2321 		if (bp->vxlan_port != udp_tunnel->udp_port) {
2322 			PMD_DRV_LOG(ERR, "Req Port: %d. Configured port: %d\n",
2323 				udp_tunnel->udp_port, bp->vxlan_port);
2324 			return -EINVAL;
2325 		}
2326 		if (--bp->vxlan_port_cnt)
2327 			return 0;
2328 
2329 		tunnel_type =
2330 			HWRM_TUNNEL_DST_PORT_FREE_INPUT_TUNNEL_TYPE_VXLAN;
2331 		port = bp->vxlan_fw_dst_port_id;
2332 		break;
2333 	case RTE_TUNNEL_TYPE_GENEVE:
2334 		if (!bp->geneve_port_cnt) {
2335 			PMD_DRV_LOG(ERR, "No Tunnel port configured yet\n");
2336 			return -EINVAL;
2337 		}
2338 		if (bp->geneve_port != udp_tunnel->udp_port) {
2339 			PMD_DRV_LOG(ERR, "Req Port: %d. Configured port: %d\n",
2340 				udp_tunnel->udp_port, bp->geneve_port);
2341 			return -EINVAL;
2342 		}
2343 		if (--bp->geneve_port_cnt)
2344 			return 0;
2345 
2346 		tunnel_type =
2347 			HWRM_TUNNEL_DST_PORT_FREE_INPUT_TUNNEL_TYPE_GENEVE;
2348 		port = bp->geneve_fw_dst_port_id;
2349 		break;
2350 	default:
2351 		PMD_DRV_LOG(ERR, "Tunnel type is not supported\n");
2352 		return -ENOTSUP;
2353 	}
2354 
2355 	rc = bnxt_hwrm_tunnel_dst_port_free(bp, port, tunnel_type);
2356 	return rc;
2357 }
2358 
2359 static int bnxt_del_vlan_filter(struct bnxt *bp, uint16_t vlan_id)
2360 {
2361 	struct bnxt_filter_info *filter;
2362 	struct bnxt_vnic_info *vnic;
2363 	int rc = 0;
2364 	uint32_t chk = HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN;
2365 
2366 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
2367 	filter = STAILQ_FIRST(&vnic->filter);
2368 	while (filter) {
2369 		/* Search for this matching MAC+VLAN filter */
2370 		if (bnxt_vlan_filter_exists(bp, filter, chk, vlan_id)) {
2371 			/* Delete the filter */
2372 			rc = bnxt_hwrm_clear_l2_filter(bp, filter);
2373 			if (rc)
2374 				return rc;
2375 			STAILQ_REMOVE(&vnic->filter, filter,
2376 				      bnxt_filter_info, next);
2377 			bnxt_free_filter(bp, filter);
2378 			PMD_DRV_LOG(INFO,
2379 				    "Deleted vlan filter for %d\n",
2380 				    vlan_id);
2381 			return 0;
2382 		}
2383 		filter = STAILQ_NEXT(filter, next);
2384 	}
2385 	return -ENOENT;
2386 }
2387 
2388 static int bnxt_add_vlan_filter(struct bnxt *bp, uint16_t vlan_id)
2389 {
2390 	struct bnxt_filter_info *filter;
2391 	struct bnxt_vnic_info *vnic;
2392 	int rc = 0;
2393 	uint32_t en = HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN |
2394 		HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN_MASK;
2395 	uint32_t chk = HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN;
2396 
2397 	/* Implementation notes on the use of VNIC in this command:
2398 	 *
2399 	 * By default, these filters belong to default vnic for the function.
2400 	 * Once these filters are set up, only destination VNIC can be modified.
2401 	 * If the destination VNIC is not specified in this command,
2402 	 * then the HWRM shall only create an l2 context id.
2403 	 */
2404 
2405 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
2406 	filter = STAILQ_FIRST(&vnic->filter);
2407 	/* Check if the VLAN has already been added */
2408 	while (filter) {
2409 		if (bnxt_vlan_filter_exists(bp, filter, chk, vlan_id))
2410 			return -EEXIST;
2411 
2412 		filter = STAILQ_NEXT(filter, next);
2413 	}
2414 
2415 	/* No match found. Alloc a fresh filter and issue the L2_FILTER_ALLOC
2416 	 * command to create MAC+VLAN filter with the right flags, enables set.
2417 	 */
2418 	filter = bnxt_alloc_filter(bp);
2419 	if (!filter) {
2420 		PMD_DRV_LOG(ERR,
2421 			    "MAC/VLAN filter alloc failed\n");
2422 		return -ENOMEM;
2423 	}
2424 	/* MAC + VLAN ID filter */
2425 	/* If l2_ivlan == 0 and l2_ivlan_mask != 0, only
2426 	 * untagged packets are received
2427 	 *
2428 	 * If l2_ivlan != 0 and l2_ivlan_mask != 0, untagged
2429 	 * packets and only the programmed vlan's packets are received
2430 	 */
2431 	filter->l2_ivlan = vlan_id;
2432 	filter->l2_ivlan_mask = 0x0FFF;
2433 	filter->enables |= en;
2434 	filter->flags |= HWRM_CFA_L2_FILTER_ALLOC_INPUT_FLAGS_OUTERMOST;
2435 
2436 	rc = bnxt_hwrm_set_l2_filter(bp, vnic->fw_vnic_id, filter);
2437 	if (rc) {
2438 		/* Free the newly allocated filter as we were
2439 		 * not able to create the filter in hardware.
2440 		 */
2441 		bnxt_free_filter(bp, filter);
2442 		return rc;
2443 	}
2444 
2445 	filter->mac_index = 0;
2446 	/* Add this new filter to the list */
2447 	if (vlan_id == 0)
2448 		STAILQ_INSERT_HEAD(&vnic->filter, filter, next);
2449 	else
2450 		STAILQ_INSERT_TAIL(&vnic->filter, filter, next);
2451 
2452 	PMD_DRV_LOG(INFO,
2453 		    "Added Vlan filter for %d\n", vlan_id);
2454 	return rc;
2455 }
2456 
2457 static int bnxt_vlan_filter_set_op(struct rte_eth_dev *eth_dev,
2458 		uint16_t vlan_id, int on)
2459 {
2460 	struct bnxt *bp = eth_dev->data->dev_private;
2461 	int rc;
2462 
2463 	rc = is_bnxt_in_error(bp);
2464 	if (rc)
2465 		return rc;
2466 
2467 	if (!eth_dev->data->dev_started) {
2468 		PMD_DRV_LOG(ERR, "port must be started before setting vlan\n");
2469 		return -EINVAL;
2470 	}
2471 
2472 	/* These operations apply to ALL existing MAC/VLAN filters */
2473 	if (on)
2474 		return bnxt_add_vlan_filter(bp, vlan_id);
2475 	else
2476 		return bnxt_del_vlan_filter(bp, vlan_id);
2477 }
2478 
2479 static int bnxt_del_dflt_mac_filter(struct bnxt *bp,
2480 				    struct bnxt_vnic_info *vnic)
2481 {
2482 	struct bnxt_filter_info *filter;
2483 	int rc;
2484 
2485 	filter = STAILQ_FIRST(&vnic->filter);
2486 	while (filter) {
2487 		if (filter->mac_index == 0 &&
2488 		    !memcmp(filter->l2_addr, bp->mac_addr,
2489 			    RTE_ETHER_ADDR_LEN)) {
2490 			rc = bnxt_hwrm_clear_l2_filter(bp, filter);
2491 			if (!rc) {
2492 				STAILQ_REMOVE(&vnic->filter, filter,
2493 					      bnxt_filter_info, next);
2494 				bnxt_free_filter(bp, filter);
2495 			}
2496 			return rc;
2497 		}
2498 		filter = STAILQ_NEXT(filter, next);
2499 	}
2500 	return 0;
2501 }
2502 
2503 static int
2504 bnxt_config_vlan_hw_filter(struct bnxt *bp, uint64_t rx_offloads)
2505 {
2506 	struct bnxt_vnic_info *vnic;
2507 	unsigned int i;
2508 	int rc;
2509 
2510 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
2511 	if (!(rx_offloads & DEV_RX_OFFLOAD_VLAN_FILTER)) {
2512 		/* Remove any VLAN filters programmed */
2513 		for (i = 0; i < RTE_ETHER_MAX_VLAN_ID; i++)
2514 			bnxt_del_vlan_filter(bp, i);
2515 
2516 		rc = bnxt_add_mac_filter(bp, vnic, NULL, 0, 0);
2517 		if (rc)
2518 			return rc;
2519 	} else {
2520 		/* Default filter will allow packets that match the
2521 		 * dest mac. So, it has to be deleted, otherwise, we
2522 		 * will endup receiving vlan packets for which the
2523 		 * filter is not programmed, when hw-vlan-filter
2524 		 * configuration is ON
2525 		 */
2526 		bnxt_del_dflt_mac_filter(bp, vnic);
2527 		/* This filter will allow only untagged packets */
2528 		bnxt_add_vlan_filter(bp, 0);
2529 	}
2530 	PMD_DRV_LOG(DEBUG, "VLAN Filtering: %d\n",
2531 		    !!(rx_offloads & DEV_RX_OFFLOAD_VLAN_FILTER));
2532 
2533 	return 0;
2534 }
2535 
2536 static int bnxt_free_one_vnic(struct bnxt *bp, uint16_t vnic_id)
2537 {
2538 	struct bnxt_vnic_info *vnic = &bp->vnic_info[vnic_id];
2539 	unsigned int i;
2540 	int rc;
2541 
2542 	/* Destroy vnic filters and vnic */
2543 	if (bp->eth_dev->data->dev_conf.rxmode.offloads &
2544 	    DEV_RX_OFFLOAD_VLAN_FILTER) {
2545 		for (i = 0; i < RTE_ETHER_MAX_VLAN_ID; i++)
2546 			bnxt_del_vlan_filter(bp, i);
2547 	}
2548 	bnxt_del_dflt_mac_filter(bp, vnic);
2549 
2550 	rc = bnxt_hwrm_vnic_ctx_free(bp, vnic);
2551 	if (rc)
2552 		return rc;
2553 
2554 	rc = bnxt_hwrm_vnic_free(bp, vnic);
2555 	if (rc)
2556 		return rc;
2557 
2558 	rte_free(vnic->fw_grp_ids);
2559 	vnic->fw_grp_ids = NULL;
2560 
2561 	vnic->rx_queue_cnt = 0;
2562 
2563 	return 0;
2564 }
2565 
2566 static int
2567 bnxt_config_vlan_hw_stripping(struct bnxt *bp, uint64_t rx_offloads)
2568 {
2569 	struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
2570 	int rc;
2571 
2572 	/* Destroy, recreate and reconfigure the default vnic */
2573 	rc = bnxt_free_one_vnic(bp, 0);
2574 	if (rc)
2575 		return rc;
2576 
2577 	/* default vnic 0 */
2578 	rc = bnxt_setup_one_vnic(bp, 0);
2579 	if (rc)
2580 		return rc;
2581 
2582 	if (bp->eth_dev->data->dev_conf.rxmode.offloads &
2583 	    DEV_RX_OFFLOAD_VLAN_FILTER) {
2584 		rc = bnxt_add_vlan_filter(bp, 0);
2585 		if (rc)
2586 			return rc;
2587 		rc = bnxt_restore_vlan_filters(bp);
2588 		if (rc)
2589 			return rc;
2590 	} else {
2591 		rc = bnxt_add_mac_filter(bp, vnic, NULL, 0, 0);
2592 		if (rc)
2593 			return rc;
2594 	}
2595 
2596 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
2597 	if (rc)
2598 		return rc;
2599 
2600 	PMD_DRV_LOG(DEBUG, "VLAN Strip Offload: %d\n",
2601 		    !!(rx_offloads & DEV_RX_OFFLOAD_VLAN_STRIP));
2602 
2603 	return rc;
2604 }
2605 
2606 static int
2607 bnxt_vlan_offload_set_op(struct rte_eth_dev *dev, int mask)
2608 {
2609 	uint64_t rx_offloads = dev->data->dev_conf.rxmode.offloads;
2610 	struct bnxt *bp = dev->data->dev_private;
2611 	int rc;
2612 
2613 	rc = is_bnxt_in_error(bp);
2614 	if (rc)
2615 		return rc;
2616 
2617 	/* Filter settings will get applied when port is started */
2618 	if (!dev->data->dev_started)
2619 		return 0;
2620 
2621 	if (mask & ETH_VLAN_FILTER_MASK) {
2622 		/* Enable or disable VLAN filtering */
2623 		rc = bnxt_config_vlan_hw_filter(bp, rx_offloads);
2624 		if (rc)
2625 			return rc;
2626 	}
2627 
2628 	if (mask & ETH_VLAN_STRIP_MASK) {
2629 		/* Enable or disable VLAN stripping */
2630 		rc = bnxt_config_vlan_hw_stripping(bp, rx_offloads);
2631 		if (rc)
2632 			return rc;
2633 	}
2634 
2635 	if (mask & ETH_VLAN_EXTEND_MASK) {
2636 		if (rx_offloads & DEV_RX_OFFLOAD_VLAN_EXTEND)
2637 			PMD_DRV_LOG(DEBUG, "Extend VLAN supported\n");
2638 		else
2639 			PMD_DRV_LOG(INFO, "Extend VLAN unsupported\n");
2640 	}
2641 
2642 	return 0;
2643 }
2644 
2645 static int
2646 bnxt_vlan_tpid_set_op(struct rte_eth_dev *dev, enum rte_vlan_type vlan_type,
2647 		      uint16_t tpid)
2648 {
2649 	struct bnxt *bp = dev->data->dev_private;
2650 	int qinq = dev->data->dev_conf.rxmode.offloads &
2651 		   DEV_RX_OFFLOAD_VLAN_EXTEND;
2652 
2653 	if (vlan_type != ETH_VLAN_TYPE_INNER &&
2654 	    vlan_type != ETH_VLAN_TYPE_OUTER) {
2655 		PMD_DRV_LOG(ERR,
2656 			    "Unsupported vlan type.");
2657 		return -EINVAL;
2658 	}
2659 	if (!qinq) {
2660 		PMD_DRV_LOG(ERR,
2661 			    "QinQ not enabled. Needs to be ON as we can "
2662 			    "accelerate only outer vlan\n");
2663 		return -EINVAL;
2664 	}
2665 
2666 	if (vlan_type == ETH_VLAN_TYPE_OUTER) {
2667 		switch (tpid) {
2668 		case RTE_ETHER_TYPE_QINQ:
2669 			bp->outer_tpid_bd =
2670 				TX_BD_LONG_CFA_META_VLAN_TPID_TPID88A8;
2671 				break;
2672 		case RTE_ETHER_TYPE_VLAN:
2673 			bp->outer_tpid_bd =
2674 				TX_BD_LONG_CFA_META_VLAN_TPID_TPID8100;
2675 				break;
2676 		case RTE_ETHER_TYPE_QINQ1:
2677 			bp->outer_tpid_bd =
2678 				TX_BD_LONG_CFA_META_VLAN_TPID_TPID9100;
2679 				break;
2680 		case RTE_ETHER_TYPE_QINQ2:
2681 			bp->outer_tpid_bd =
2682 				TX_BD_LONG_CFA_META_VLAN_TPID_TPID9200;
2683 				break;
2684 		case RTE_ETHER_TYPE_QINQ3:
2685 			bp->outer_tpid_bd =
2686 				 TX_BD_LONG_CFA_META_VLAN_TPID_TPID9300;
2687 				break;
2688 		default:
2689 			PMD_DRV_LOG(ERR, "Invalid TPID: %x\n", tpid);
2690 			return -EINVAL;
2691 		}
2692 		bp->outer_tpid_bd |= tpid;
2693 		PMD_DRV_LOG(INFO, "outer_tpid_bd = %x\n", bp->outer_tpid_bd);
2694 	} else if (vlan_type == ETH_VLAN_TYPE_INNER) {
2695 		PMD_DRV_LOG(ERR,
2696 			    "Can accelerate only outer vlan in QinQ\n");
2697 		return -EINVAL;
2698 	}
2699 
2700 	return 0;
2701 }
2702 
2703 static int
2704 bnxt_set_default_mac_addr_op(struct rte_eth_dev *dev,
2705 			     struct rte_ether_addr *addr)
2706 {
2707 	struct bnxt *bp = dev->data->dev_private;
2708 	/* Default Filter is tied to VNIC 0 */
2709 	struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
2710 	int rc;
2711 
2712 	rc = is_bnxt_in_error(bp);
2713 	if (rc)
2714 		return rc;
2715 
2716 	if (BNXT_VF(bp) && !BNXT_VF_IS_TRUSTED(bp))
2717 		return -EPERM;
2718 
2719 	if (rte_is_zero_ether_addr(addr))
2720 		return -EINVAL;
2721 
2722 	/* Filter settings will get applied when port is started */
2723 	if (!dev->data->dev_started)
2724 		return 0;
2725 
2726 	/* Check if the requested MAC is already added */
2727 	if (memcmp(addr, bp->mac_addr, RTE_ETHER_ADDR_LEN) == 0)
2728 		return 0;
2729 
2730 	/* Destroy filter and re-create it */
2731 	bnxt_del_dflt_mac_filter(bp, vnic);
2732 
2733 	memcpy(bp->mac_addr, addr, RTE_ETHER_ADDR_LEN);
2734 	if (dev->data->dev_conf.rxmode.offloads & DEV_RX_OFFLOAD_VLAN_FILTER) {
2735 		/* This filter will allow only untagged packets */
2736 		rc = bnxt_add_vlan_filter(bp, 0);
2737 	} else {
2738 		rc = bnxt_add_mac_filter(bp, vnic, addr, 0, 0);
2739 	}
2740 
2741 	PMD_DRV_LOG(DEBUG, "Set MAC addr\n");
2742 	return rc;
2743 }
2744 
2745 static int
2746 bnxt_dev_set_mc_addr_list_op(struct rte_eth_dev *eth_dev,
2747 			  struct rte_ether_addr *mc_addr_set,
2748 			  uint32_t nb_mc_addr)
2749 {
2750 	struct bnxt *bp = eth_dev->data->dev_private;
2751 	char *mc_addr_list = (char *)mc_addr_set;
2752 	struct bnxt_vnic_info *vnic;
2753 	uint32_t off = 0, i = 0;
2754 	int rc;
2755 
2756 	rc = is_bnxt_in_error(bp);
2757 	if (rc)
2758 		return rc;
2759 
2760 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
2761 
2762 	if (nb_mc_addr > BNXT_MAX_MC_ADDRS) {
2763 		vnic->flags |= BNXT_VNIC_INFO_ALLMULTI;
2764 		goto allmulti;
2765 	}
2766 
2767 	/* TODO Check for Duplicate mcast addresses */
2768 	vnic->flags &= ~BNXT_VNIC_INFO_ALLMULTI;
2769 	for (i = 0; i < nb_mc_addr; i++) {
2770 		memcpy(vnic->mc_list + off, &mc_addr_list[i],
2771 			RTE_ETHER_ADDR_LEN);
2772 		off += RTE_ETHER_ADDR_LEN;
2773 	}
2774 
2775 	vnic->mc_addr_cnt = i;
2776 	if (vnic->mc_addr_cnt)
2777 		vnic->flags |= BNXT_VNIC_INFO_MCAST;
2778 	else
2779 		vnic->flags &= ~BNXT_VNIC_INFO_MCAST;
2780 
2781 allmulti:
2782 	return bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
2783 }
2784 
2785 static int
2786 bnxt_fw_version_get(struct rte_eth_dev *dev, char *fw_version, size_t fw_size)
2787 {
2788 	struct bnxt *bp = dev->data->dev_private;
2789 	uint8_t fw_major = (bp->fw_ver >> 24) & 0xff;
2790 	uint8_t fw_minor = (bp->fw_ver >> 16) & 0xff;
2791 	uint8_t fw_updt = (bp->fw_ver >> 8) & 0xff;
2792 	uint8_t fw_rsvd = bp->fw_ver & 0xff;
2793 	int ret;
2794 
2795 	ret = snprintf(fw_version, fw_size, "%d.%d.%d.%d",
2796 			fw_major, fw_minor, fw_updt, fw_rsvd);
2797 	if (ret < 0)
2798 		return -EINVAL;
2799 
2800 	ret += 1; /* add the size of '\0' */
2801 	if (fw_size < (size_t)ret)
2802 		return ret;
2803 	else
2804 		return 0;
2805 }
2806 
2807 static void
2808 bnxt_rxq_info_get_op(struct rte_eth_dev *dev, uint16_t queue_id,
2809 	struct rte_eth_rxq_info *qinfo)
2810 {
2811 	struct bnxt *bp = dev->data->dev_private;
2812 	struct bnxt_rx_queue *rxq;
2813 
2814 	if (is_bnxt_in_error(bp))
2815 		return;
2816 
2817 	rxq = dev->data->rx_queues[queue_id];
2818 
2819 	qinfo->mp = rxq->mb_pool;
2820 	qinfo->scattered_rx = dev->data->scattered_rx;
2821 	qinfo->nb_desc = rxq->nb_rx_desc;
2822 
2823 	qinfo->conf.rx_free_thresh = rxq->rx_free_thresh;
2824 	qinfo->conf.rx_drop_en = rxq->drop_en;
2825 	qinfo->conf.rx_deferred_start = rxq->rx_deferred_start;
2826 	qinfo->conf.offloads = dev->data->dev_conf.rxmode.offloads;
2827 }
2828 
2829 static void
2830 bnxt_txq_info_get_op(struct rte_eth_dev *dev, uint16_t queue_id,
2831 	struct rte_eth_txq_info *qinfo)
2832 {
2833 	struct bnxt *bp = dev->data->dev_private;
2834 	struct bnxt_tx_queue *txq;
2835 
2836 	if (is_bnxt_in_error(bp))
2837 		return;
2838 
2839 	txq = dev->data->tx_queues[queue_id];
2840 
2841 	qinfo->nb_desc = txq->nb_tx_desc;
2842 
2843 	qinfo->conf.tx_thresh.pthresh = txq->pthresh;
2844 	qinfo->conf.tx_thresh.hthresh = txq->hthresh;
2845 	qinfo->conf.tx_thresh.wthresh = txq->wthresh;
2846 
2847 	qinfo->conf.tx_free_thresh = txq->tx_free_thresh;
2848 	qinfo->conf.tx_rs_thresh = 0;
2849 	qinfo->conf.tx_deferred_start = txq->tx_deferred_start;
2850 	qinfo->conf.offloads = txq->offloads;
2851 }
2852 
2853 static const struct {
2854 	eth_rx_burst_t pkt_burst;
2855 	const char *info;
2856 } bnxt_rx_burst_info[] = {
2857 	{bnxt_recv_pkts,	"Scalar"},
2858 #if defined(RTE_ARCH_X86)
2859 	{bnxt_recv_pkts_vec,	"Vector SSE"},
2860 #elif defined(RTE_ARCH_ARM64)
2861 	{bnxt_recv_pkts_vec,	"Vector Neon"},
2862 #endif
2863 };
2864 
2865 static int
2866 bnxt_rx_burst_mode_get(struct rte_eth_dev *dev, __rte_unused uint16_t queue_id,
2867 		       struct rte_eth_burst_mode *mode)
2868 {
2869 	eth_rx_burst_t pkt_burst = dev->rx_pkt_burst;
2870 	size_t i;
2871 
2872 	for (i = 0; i < RTE_DIM(bnxt_rx_burst_info); i++) {
2873 		if (pkt_burst == bnxt_rx_burst_info[i].pkt_burst) {
2874 			snprintf(mode->info, sizeof(mode->info), "%s",
2875 				 bnxt_rx_burst_info[i].info);
2876 			return 0;
2877 		}
2878 	}
2879 
2880 	return -EINVAL;
2881 }
2882 
2883 static const struct {
2884 	eth_tx_burst_t pkt_burst;
2885 	const char *info;
2886 } bnxt_tx_burst_info[] = {
2887 	{bnxt_xmit_pkts,	"Scalar"},
2888 #if defined(RTE_ARCH_X86)
2889 	{bnxt_xmit_pkts_vec,	"Vector SSE"},
2890 #elif defined(RTE_ARCH_ARM64)
2891 	{bnxt_xmit_pkts_vec,	"Vector Neon"},
2892 #endif
2893 };
2894 
2895 static int
2896 bnxt_tx_burst_mode_get(struct rte_eth_dev *dev, __rte_unused uint16_t queue_id,
2897 		       struct rte_eth_burst_mode *mode)
2898 {
2899 	eth_tx_burst_t pkt_burst = dev->tx_pkt_burst;
2900 	size_t i;
2901 
2902 	for (i = 0; i < RTE_DIM(bnxt_tx_burst_info); i++) {
2903 		if (pkt_burst == bnxt_tx_burst_info[i].pkt_burst) {
2904 			snprintf(mode->info, sizeof(mode->info), "%s",
2905 				 bnxt_tx_burst_info[i].info);
2906 			return 0;
2907 		}
2908 	}
2909 
2910 	return -EINVAL;
2911 }
2912 
2913 int bnxt_mtu_set_op(struct rte_eth_dev *eth_dev, uint16_t new_mtu)
2914 {
2915 	struct bnxt *bp = eth_dev->data->dev_private;
2916 	uint32_t new_pkt_size;
2917 	uint32_t rc = 0;
2918 	uint32_t i;
2919 
2920 	rc = is_bnxt_in_error(bp);
2921 	if (rc)
2922 		return rc;
2923 
2924 	/* Exit if receive queues are not configured yet */
2925 	if (!eth_dev->data->nb_rx_queues)
2926 		return rc;
2927 
2928 	new_pkt_size = new_mtu + RTE_ETHER_HDR_LEN + RTE_ETHER_CRC_LEN +
2929 		       VLAN_TAG_SIZE * BNXT_NUM_VLANS;
2930 
2931 	/*
2932 	 * Disallow any MTU change that would require scattered receive support
2933 	 * if it is not already enabled.
2934 	 */
2935 	if (eth_dev->data->dev_started &&
2936 	    !eth_dev->data->scattered_rx &&
2937 	    (new_pkt_size >
2938 	     eth_dev->data->min_rx_buf_size - RTE_PKTMBUF_HEADROOM)) {
2939 		PMD_DRV_LOG(ERR,
2940 			    "MTU change would require scattered rx support. ");
2941 		PMD_DRV_LOG(ERR, "Stop port before changing MTU.\n");
2942 		return -EINVAL;
2943 	}
2944 
2945 	if (new_mtu > RTE_ETHER_MTU) {
2946 		bp->flags |= BNXT_FLAG_JUMBO;
2947 		bp->eth_dev->data->dev_conf.rxmode.offloads |=
2948 			DEV_RX_OFFLOAD_JUMBO_FRAME;
2949 	} else {
2950 		bp->eth_dev->data->dev_conf.rxmode.offloads &=
2951 			~DEV_RX_OFFLOAD_JUMBO_FRAME;
2952 		bp->flags &= ~BNXT_FLAG_JUMBO;
2953 	}
2954 
2955 	/* Is there a change in mtu setting? */
2956 	if (eth_dev->data->dev_conf.rxmode.max_rx_pkt_len == new_pkt_size)
2957 		return rc;
2958 
2959 	for (i = 0; i < bp->nr_vnics; i++) {
2960 		struct bnxt_vnic_info *vnic = &bp->vnic_info[i];
2961 		uint16_t size = 0;
2962 
2963 		vnic->mru = BNXT_VNIC_MRU(new_mtu);
2964 		rc = bnxt_hwrm_vnic_cfg(bp, vnic);
2965 		if (rc)
2966 			break;
2967 
2968 		size = rte_pktmbuf_data_room_size(bp->rx_queues[0]->mb_pool);
2969 		size -= RTE_PKTMBUF_HEADROOM;
2970 
2971 		if (size < new_mtu) {
2972 			rc = bnxt_hwrm_vnic_plcmode_cfg(bp, vnic);
2973 			if (rc)
2974 				return rc;
2975 		}
2976 	}
2977 
2978 	if (!rc)
2979 		eth_dev->data->dev_conf.rxmode.max_rx_pkt_len = new_pkt_size;
2980 
2981 	PMD_DRV_LOG(INFO, "New MTU is %d\n", new_mtu);
2982 
2983 	return rc;
2984 }
2985 
2986 static int
2987 bnxt_vlan_pvid_set_op(struct rte_eth_dev *dev, uint16_t pvid, int on)
2988 {
2989 	struct bnxt *bp = dev->data->dev_private;
2990 	uint16_t vlan = bp->vlan;
2991 	int rc;
2992 
2993 	rc = is_bnxt_in_error(bp);
2994 	if (rc)
2995 		return rc;
2996 
2997 	if (!BNXT_SINGLE_PF(bp) || BNXT_VF(bp)) {
2998 		PMD_DRV_LOG(ERR,
2999 			"PVID cannot be modified for this function\n");
3000 		return -ENOTSUP;
3001 	}
3002 	bp->vlan = on ? pvid : 0;
3003 
3004 	rc = bnxt_hwrm_set_default_vlan(bp, 0, 0);
3005 	if (rc)
3006 		bp->vlan = vlan;
3007 	return rc;
3008 }
3009 
3010 static int
3011 bnxt_dev_led_on_op(struct rte_eth_dev *dev)
3012 {
3013 	struct bnxt *bp = dev->data->dev_private;
3014 	int rc;
3015 
3016 	rc = is_bnxt_in_error(bp);
3017 	if (rc)
3018 		return rc;
3019 
3020 	return bnxt_hwrm_port_led_cfg(bp, true);
3021 }
3022 
3023 static int
3024 bnxt_dev_led_off_op(struct rte_eth_dev *dev)
3025 {
3026 	struct bnxt *bp = dev->data->dev_private;
3027 	int rc;
3028 
3029 	rc = is_bnxt_in_error(bp);
3030 	if (rc)
3031 		return rc;
3032 
3033 	return bnxt_hwrm_port_led_cfg(bp, false);
3034 }
3035 
3036 static uint32_t
3037 bnxt_rx_queue_count_op(struct rte_eth_dev *dev, uint16_t rx_queue_id)
3038 {
3039 	struct bnxt *bp = (struct bnxt *)dev->data->dev_private;
3040 	struct bnxt_cp_ring_info *cpr;
3041 	uint32_t desc = 0, raw_cons;
3042 	struct bnxt_rx_queue *rxq;
3043 	struct rx_pkt_cmpl *rxcmp;
3044 	int rc;
3045 
3046 	rc = is_bnxt_in_error(bp);
3047 	if (rc)
3048 		return rc;
3049 
3050 	rxq = dev->data->rx_queues[rx_queue_id];
3051 	cpr = rxq->cp_ring;
3052 	raw_cons = cpr->cp_raw_cons;
3053 
3054 	while (1) {
3055 		uint32_t agg_cnt, cons, cmpl_type;
3056 
3057 		cons = RING_CMP(cpr->cp_ring_struct, raw_cons);
3058 		rxcmp = (struct rx_pkt_cmpl *)&cpr->cp_desc_ring[cons];
3059 
3060 		if (!CMP_VALID(rxcmp, raw_cons, cpr->cp_ring_struct))
3061 			break;
3062 
3063 		cmpl_type = CMP_TYPE(rxcmp);
3064 
3065 		switch (cmpl_type) {
3066 		case CMPL_BASE_TYPE_RX_L2:
3067 		case CMPL_BASE_TYPE_RX_L2_V2:
3068 			agg_cnt = BNXT_RX_L2_AGG_BUFS(rxcmp);
3069 			raw_cons = raw_cons + CMP_LEN(cmpl_type) + agg_cnt;
3070 			desc++;
3071 			break;
3072 
3073 		case CMPL_BASE_TYPE_RX_TPA_END:
3074 			if (BNXT_CHIP_P5(rxq->bp)) {
3075 				struct rx_tpa_v2_end_cmpl_hi *p5_tpa_end;
3076 
3077 				p5_tpa_end = (void *)rxcmp;
3078 				agg_cnt = BNXT_TPA_END_AGG_BUFS_TH(p5_tpa_end);
3079 			} else {
3080 				struct rx_tpa_end_cmpl *tpa_end;
3081 
3082 				tpa_end = (void *)rxcmp;
3083 				agg_cnt = BNXT_TPA_END_AGG_BUFS(tpa_end);
3084 			}
3085 
3086 			raw_cons = raw_cons + CMP_LEN(cmpl_type) + agg_cnt;
3087 			desc++;
3088 			break;
3089 
3090 		default:
3091 			raw_cons += CMP_LEN(cmpl_type);
3092 		}
3093 	}
3094 
3095 	return desc;
3096 }
3097 
3098 static int
3099 bnxt_rx_descriptor_status_op(void *rx_queue, uint16_t offset)
3100 {
3101 	struct bnxt_rx_queue *rxq = rx_queue;
3102 	struct bnxt_cp_ring_info *cpr;
3103 	struct bnxt_rx_ring_info *rxr;
3104 	uint32_t desc, raw_cons;
3105 	struct bnxt *bp = rxq->bp;
3106 	struct rx_pkt_cmpl *rxcmp;
3107 	int rc;
3108 
3109 	rc = is_bnxt_in_error(bp);
3110 	if (rc)
3111 		return rc;
3112 
3113 	if (offset >= rxq->nb_rx_desc)
3114 		return -EINVAL;
3115 
3116 	rxr = rxq->rx_ring;
3117 	cpr = rxq->cp_ring;
3118 
3119 	/*
3120 	 * For the vector receive case, the completion at the requested
3121 	 * offset can be indexed directly.
3122 	 */
3123 #if defined(RTE_ARCH_X86) || defined(RTE_ARCH_ARM64)
3124 	if (bp->flags & BNXT_FLAG_RX_VECTOR_PKT_MODE) {
3125 		struct rx_pkt_cmpl *rxcmp;
3126 		uint32_t cons;
3127 
3128 		/* Check status of completion descriptor. */
3129 		raw_cons = cpr->cp_raw_cons +
3130 			   offset * CMP_LEN(CMPL_BASE_TYPE_RX_L2);
3131 		cons = RING_CMP(cpr->cp_ring_struct, raw_cons);
3132 		rxcmp = (struct rx_pkt_cmpl *)&cpr->cp_desc_ring[cons];
3133 
3134 		if (CMP_VALID(rxcmp, raw_cons, cpr->cp_ring_struct))
3135 			return RTE_ETH_RX_DESC_DONE;
3136 
3137 		/* Check whether rx desc has an mbuf attached. */
3138 		cons = RING_CMP(rxr->rx_ring_struct, raw_cons / 2);
3139 		if (cons >= rxq->rxrearm_start &&
3140 		    cons < rxq->rxrearm_start + rxq->rxrearm_nb) {
3141 			return RTE_ETH_RX_DESC_UNAVAIL;
3142 		}
3143 
3144 		return RTE_ETH_RX_DESC_AVAIL;
3145 	}
3146 #endif
3147 
3148 	/*
3149 	 * For the non-vector receive case, scan the completion ring to
3150 	 * locate the completion descriptor for the requested offset.
3151 	 */
3152 	raw_cons = cpr->cp_raw_cons;
3153 	desc = 0;
3154 	while (1) {
3155 		uint32_t agg_cnt, cons, cmpl_type;
3156 
3157 		cons = RING_CMP(cpr->cp_ring_struct, raw_cons);
3158 		rxcmp = (struct rx_pkt_cmpl *)&cpr->cp_desc_ring[cons];
3159 
3160 		if (!CMP_VALID(rxcmp, raw_cons, cpr->cp_ring_struct))
3161 			break;
3162 
3163 		cmpl_type = CMP_TYPE(rxcmp);
3164 
3165 		switch (cmpl_type) {
3166 		case CMPL_BASE_TYPE_RX_L2:
3167 		case CMPL_BASE_TYPE_RX_L2_V2:
3168 			if (desc == offset) {
3169 				cons = rxcmp->opaque;
3170 				if (rxr->rx_buf_ring[cons])
3171 					return RTE_ETH_RX_DESC_DONE;
3172 				else
3173 					return RTE_ETH_RX_DESC_UNAVAIL;
3174 			}
3175 			agg_cnt = BNXT_RX_L2_AGG_BUFS(rxcmp);
3176 			raw_cons = raw_cons + CMP_LEN(cmpl_type) + agg_cnt;
3177 			desc++;
3178 			break;
3179 
3180 		case CMPL_BASE_TYPE_RX_TPA_END:
3181 			if (desc == offset)
3182 				return RTE_ETH_RX_DESC_DONE;
3183 
3184 			if (BNXT_CHIP_P5(rxq->bp)) {
3185 				struct rx_tpa_v2_end_cmpl_hi *p5_tpa_end;
3186 
3187 				p5_tpa_end = (void *)rxcmp;
3188 				agg_cnt = BNXT_TPA_END_AGG_BUFS_TH(p5_tpa_end);
3189 			} else {
3190 				struct rx_tpa_end_cmpl *tpa_end;
3191 
3192 				tpa_end = (void *)rxcmp;
3193 				agg_cnt = BNXT_TPA_END_AGG_BUFS(tpa_end);
3194 			}
3195 
3196 			raw_cons = raw_cons + CMP_LEN(cmpl_type) + agg_cnt;
3197 			desc++;
3198 			break;
3199 
3200 		default:
3201 			raw_cons += CMP_LEN(cmpl_type);
3202 		}
3203 	}
3204 
3205 	return RTE_ETH_RX_DESC_AVAIL;
3206 }
3207 
3208 static int
3209 bnxt_tx_descriptor_status_op(void *tx_queue, uint16_t offset)
3210 {
3211 	struct bnxt_tx_queue *txq = (struct bnxt_tx_queue *)tx_queue;
3212 	struct bnxt_tx_ring_info *txr;
3213 	struct bnxt_cp_ring_info *cpr;
3214 	struct rte_mbuf **tx_buf;
3215 	struct tx_pkt_cmpl *txcmp;
3216 	uint32_t cons, cp_cons;
3217 	int rc;
3218 
3219 	if (!txq)
3220 		return -EINVAL;
3221 
3222 	rc = is_bnxt_in_error(txq->bp);
3223 	if (rc)
3224 		return rc;
3225 
3226 	cpr = txq->cp_ring;
3227 	txr = txq->tx_ring;
3228 
3229 	if (offset >= txq->nb_tx_desc)
3230 		return -EINVAL;
3231 
3232 	cons = RING_CMP(cpr->cp_ring_struct, offset);
3233 	txcmp = (struct tx_pkt_cmpl *)&cpr->cp_desc_ring[cons];
3234 	cp_cons = cpr->cp_raw_cons;
3235 
3236 	if (cons > cp_cons) {
3237 		if (CMPL_VALID(txcmp, cpr->valid))
3238 			return RTE_ETH_TX_DESC_UNAVAIL;
3239 	} else {
3240 		if (CMPL_VALID(txcmp, !cpr->valid))
3241 			return RTE_ETH_TX_DESC_UNAVAIL;
3242 	}
3243 	tx_buf = &txr->tx_buf_ring[cons];
3244 	if (*tx_buf == NULL)
3245 		return RTE_ETH_TX_DESC_DONE;
3246 
3247 	return RTE_ETH_TX_DESC_FULL;
3248 }
3249 
3250 int
3251 bnxt_flow_ops_get_op(struct rte_eth_dev *dev,
3252 		     const struct rte_flow_ops **ops)
3253 {
3254 	struct bnxt *bp = dev->data->dev_private;
3255 	int ret = 0;
3256 
3257 	if (!bp)
3258 		return -EIO;
3259 
3260 	if (BNXT_ETH_DEV_IS_REPRESENTOR(dev)) {
3261 		struct bnxt_representor *vfr = dev->data->dev_private;
3262 		bp = vfr->parent_dev->data->dev_private;
3263 		/* parent is deleted while children are still valid */
3264 		if (!bp) {
3265 			PMD_DRV_LOG(DEBUG, "BNXT Port:%d VFR Error\n",
3266 				    dev->data->port_id);
3267 			return -EIO;
3268 		}
3269 	}
3270 
3271 	ret = is_bnxt_in_error(bp);
3272 	if (ret)
3273 		return ret;
3274 
3275 	/* PMD supports thread-safe flow operations.  rte_flow API
3276 	 * functions can avoid mutex for multi-thread safety.
3277 	 */
3278 	dev->data->dev_flags |= RTE_ETH_DEV_FLOW_OPS_THREAD_SAFE;
3279 
3280 	if (BNXT_TRUFLOW_EN(bp))
3281 		*ops = &bnxt_ulp_rte_flow_ops;
3282 	else
3283 		*ops = &bnxt_flow_ops;
3284 
3285 	return ret;
3286 }
3287 
3288 static const uint32_t *
3289 bnxt_dev_supported_ptypes_get_op(struct rte_eth_dev *dev)
3290 {
3291 	static const uint32_t ptypes[] = {
3292 		RTE_PTYPE_L2_ETHER_VLAN,
3293 		RTE_PTYPE_L3_IPV4_EXT_UNKNOWN,
3294 		RTE_PTYPE_L3_IPV6_EXT_UNKNOWN,
3295 		RTE_PTYPE_L4_ICMP,
3296 		RTE_PTYPE_L4_TCP,
3297 		RTE_PTYPE_L4_UDP,
3298 		RTE_PTYPE_INNER_L3_IPV4_EXT_UNKNOWN,
3299 		RTE_PTYPE_INNER_L3_IPV6_EXT_UNKNOWN,
3300 		RTE_PTYPE_INNER_L4_ICMP,
3301 		RTE_PTYPE_INNER_L4_TCP,
3302 		RTE_PTYPE_INNER_L4_UDP,
3303 		RTE_PTYPE_UNKNOWN
3304 	};
3305 
3306 	if (!dev->rx_pkt_burst)
3307 		return NULL;
3308 
3309 	return ptypes;
3310 }
3311 
3312 static int bnxt_map_regs(struct bnxt *bp, uint32_t *reg_arr, int count,
3313 			 int reg_win)
3314 {
3315 	uint32_t reg_base = *reg_arr & 0xfffff000;
3316 	uint32_t win_off;
3317 	int i;
3318 
3319 	for (i = 0; i < count; i++) {
3320 		if ((reg_arr[i] & 0xfffff000) != reg_base)
3321 			return -ERANGE;
3322 	}
3323 	win_off = BNXT_GRCPF_REG_WINDOW_BASE_OUT + (reg_win - 1) * 4;
3324 	rte_write32(reg_base, (uint8_t *)bp->bar0 + win_off);
3325 	return 0;
3326 }
3327 
3328 static int bnxt_map_ptp_regs(struct bnxt *bp)
3329 {
3330 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3331 	uint32_t *reg_arr;
3332 	int rc, i;
3333 
3334 	reg_arr = ptp->rx_regs;
3335 	rc = bnxt_map_regs(bp, reg_arr, BNXT_PTP_RX_REGS, 5);
3336 	if (rc)
3337 		return rc;
3338 
3339 	reg_arr = ptp->tx_regs;
3340 	rc = bnxt_map_regs(bp, reg_arr, BNXT_PTP_TX_REGS, 6);
3341 	if (rc)
3342 		return rc;
3343 
3344 	for (i = 0; i < BNXT_PTP_RX_REGS; i++)
3345 		ptp->rx_mapped_regs[i] = 0x5000 + (ptp->rx_regs[i] & 0xfff);
3346 
3347 	for (i = 0; i < BNXT_PTP_TX_REGS; i++)
3348 		ptp->tx_mapped_regs[i] = 0x6000 + (ptp->tx_regs[i] & 0xfff);
3349 
3350 	return 0;
3351 }
3352 
3353 static void bnxt_unmap_ptp_regs(struct bnxt *bp)
3354 {
3355 	rte_write32(0, (uint8_t *)bp->bar0 +
3356 			 BNXT_GRCPF_REG_WINDOW_BASE_OUT + 16);
3357 	rte_write32(0, (uint8_t *)bp->bar0 +
3358 			 BNXT_GRCPF_REG_WINDOW_BASE_OUT + 20);
3359 }
3360 
3361 static uint64_t bnxt_cc_read(struct bnxt *bp)
3362 {
3363 	uint64_t ns;
3364 
3365 	ns = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3366 			      BNXT_GRCPF_REG_SYNC_TIME));
3367 	ns |= (uint64_t)(rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3368 					  BNXT_GRCPF_REG_SYNC_TIME + 4))) << 32;
3369 	return ns;
3370 }
3371 
3372 static int bnxt_get_tx_ts(struct bnxt *bp, uint64_t *ts)
3373 {
3374 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3375 	uint32_t fifo;
3376 
3377 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3378 				ptp->tx_mapped_regs[BNXT_PTP_TX_FIFO]));
3379 	if (fifo & BNXT_PTP_TX_FIFO_EMPTY)
3380 		return -EAGAIN;
3381 
3382 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3383 				ptp->tx_mapped_regs[BNXT_PTP_TX_FIFO]));
3384 	*ts = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3385 				ptp->tx_mapped_regs[BNXT_PTP_TX_TS_L]));
3386 	*ts |= (uint64_t)rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3387 				ptp->tx_mapped_regs[BNXT_PTP_TX_TS_H])) << 32;
3388 	rte_read32((uint8_t *)bp->bar0 + ptp->tx_mapped_regs[BNXT_PTP_TX_SEQ]);
3389 
3390 	return 0;
3391 }
3392 
3393 static int bnxt_get_rx_ts(struct bnxt *bp, uint64_t *ts)
3394 {
3395 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3396 	struct bnxt_pf_info *pf = bp->pf;
3397 	uint16_t port_id;
3398 	uint32_t fifo;
3399 
3400 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3401 				ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO]));
3402 	if (!(fifo & BNXT_PTP_RX_FIFO_PENDING))
3403 		return -EAGAIN;
3404 
3405 	port_id = pf->port_id;
3406 	rte_write32(1 << port_id, (uint8_t *)bp->bar0 +
3407 	       ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO_ADV]);
3408 
3409 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3410 				   ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO]));
3411 	if (fifo & BNXT_PTP_RX_FIFO_PENDING) {
3412 /*		bnxt_clr_rx_ts(bp);	  TBD  */
3413 		return -EBUSY;
3414 	}
3415 
3416 	*ts = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3417 				ptp->rx_mapped_regs[BNXT_PTP_RX_TS_L]));
3418 	*ts |= (uint64_t)rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3419 				ptp->rx_mapped_regs[BNXT_PTP_RX_TS_H])) << 32;
3420 
3421 	return 0;
3422 }
3423 
3424 static int
3425 bnxt_timesync_write_time(struct rte_eth_dev *dev, const struct timespec *ts)
3426 {
3427 	uint64_t ns;
3428 	struct bnxt *bp = dev->data->dev_private;
3429 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3430 
3431 	if (!ptp)
3432 		return -ENOTSUP;
3433 
3434 	ns = rte_timespec_to_ns(ts);
3435 	/* Set the timecounters to a new value. */
3436 	ptp->tc.nsec = ns;
3437 	ptp->tx_tstamp_tc.nsec = ns;
3438 	ptp->rx_tstamp_tc.nsec = ns;
3439 
3440 	return 0;
3441 }
3442 
3443 static int
3444 bnxt_timesync_read_time(struct rte_eth_dev *dev, struct timespec *ts)
3445 {
3446 	struct bnxt *bp = dev->data->dev_private;
3447 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3448 	uint64_t ns, systime_cycles = 0;
3449 	int rc = 0;
3450 
3451 	if (!ptp)
3452 		return -ENOTSUP;
3453 
3454 	if (BNXT_CHIP_P5(bp))
3455 		rc = bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_CURRENT_TIME,
3456 					     &systime_cycles);
3457 	else
3458 		systime_cycles = bnxt_cc_read(bp);
3459 
3460 	ns = rte_timecounter_update(&ptp->tc, systime_cycles);
3461 	*ts = rte_ns_to_timespec(ns);
3462 
3463 	return rc;
3464 }
3465 static int
3466 bnxt_timesync_enable(struct rte_eth_dev *dev)
3467 {
3468 	struct bnxt *bp = dev->data->dev_private;
3469 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3470 	uint32_t shift = 0;
3471 	int rc;
3472 
3473 	if (!ptp)
3474 		return -ENOTSUP;
3475 
3476 	ptp->rx_filter = 1;
3477 	ptp->tx_tstamp_en = 1;
3478 	ptp->rxctl = BNXT_PTP_MSG_EVENTS;
3479 
3480 	rc = bnxt_hwrm_ptp_cfg(bp);
3481 	if (rc)
3482 		return rc;
3483 
3484 	memset(&ptp->tc, 0, sizeof(struct rte_timecounter));
3485 	memset(&ptp->rx_tstamp_tc, 0, sizeof(struct rte_timecounter));
3486 	memset(&ptp->tx_tstamp_tc, 0, sizeof(struct rte_timecounter));
3487 
3488 	ptp->tc.cc_mask = BNXT_CYCLECOUNTER_MASK;
3489 	ptp->tc.cc_shift = shift;
3490 	ptp->tc.nsec_mask = (1ULL << shift) - 1;
3491 
3492 	ptp->rx_tstamp_tc.cc_mask = BNXT_CYCLECOUNTER_MASK;
3493 	ptp->rx_tstamp_tc.cc_shift = shift;
3494 	ptp->rx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
3495 
3496 	ptp->tx_tstamp_tc.cc_mask = BNXT_CYCLECOUNTER_MASK;
3497 	ptp->tx_tstamp_tc.cc_shift = shift;
3498 	ptp->tx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
3499 
3500 	if (!BNXT_CHIP_P5(bp))
3501 		bnxt_map_ptp_regs(bp);
3502 	else
3503 		rc = bnxt_ptp_start(bp);
3504 
3505 	return rc;
3506 }
3507 
3508 static int
3509 bnxt_timesync_disable(struct rte_eth_dev *dev)
3510 {
3511 	struct bnxt *bp = dev->data->dev_private;
3512 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3513 
3514 	if (!ptp)
3515 		return -ENOTSUP;
3516 
3517 	ptp->rx_filter = 0;
3518 	ptp->tx_tstamp_en = 0;
3519 	ptp->rxctl = 0;
3520 
3521 	bnxt_hwrm_ptp_cfg(bp);
3522 
3523 	if (!BNXT_CHIP_P5(bp))
3524 		bnxt_unmap_ptp_regs(bp);
3525 	else
3526 		bnxt_ptp_stop(bp);
3527 
3528 	return 0;
3529 }
3530 
3531 static int
3532 bnxt_timesync_read_rx_timestamp(struct rte_eth_dev *dev,
3533 				 struct timespec *timestamp,
3534 				 uint32_t flags __rte_unused)
3535 {
3536 	struct bnxt *bp = dev->data->dev_private;
3537 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3538 	uint64_t rx_tstamp_cycles = 0;
3539 	uint64_t ns;
3540 
3541 	if (!ptp)
3542 		return -ENOTSUP;
3543 
3544 	if (BNXT_CHIP_P5(bp))
3545 		rx_tstamp_cycles = ptp->rx_timestamp;
3546 	else
3547 		bnxt_get_rx_ts(bp, &rx_tstamp_cycles);
3548 
3549 	ns = rte_timecounter_update(&ptp->rx_tstamp_tc, rx_tstamp_cycles);
3550 	*timestamp = rte_ns_to_timespec(ns);
3551 	return  0;
3552 }
3553 
3554 static int
3555 bnxt_timesync_read_tx_timestamp(struct rte_eth_dev *dev,
3556 				 struct timespec *timestamp)
3557 {
3558 	struct bnxt *bp = dev->data->dev_private;
3559 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3560 	uint64_t tx_tstamp_cycles = 0;
3561 	uint64_t ns;
3562 	int rc = 0;
3563 
3564 	if (!ptp)
3565 		return -ENOTSUP;
3566 
3567 	if (BNXT_CHIP_P5(bp))
3568 		rc = bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_PATH_TX,
3569 					     &tx_tstamp_cycles);
3570 	else
3571 		rc = bnxt_get_tx_ts(bp, &tx_tstamp_cycles);
3572 
3573 	ns = rte_timecounter_update(&ptp->tx_tstamp_tc, tx_tstamp_cycles);
3574 	*timestamp = rte_ns_to_timespec(ns);
3575 
3576 	return rc;
3577 }
3578 
3579 static int
3580 bnxt_timesync_adjust_time(struct rte_eth_dev *dev, int64_t delta)
3581 {
3582 	struct bnxt *bp = dev->data->dev_private;
3583 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3584 
3585 	if (!ptp)
3586 		return -ENOTSUP;
3587 
3588 	ptp->tc.nsec += delta;
3589 	ptp->tx_tstamp_tc.nsec += delta;
3590 	ptp->rx_tstamp_tc.nsec += delta;
3591 
3592 	return 0;
3593 }
3594 
3595 static int
3596 bnxt_get_eeprom_length_op(struct rte_eth_dev *dev)
3597 {
3598 	struct bnxt *bp = dev->data->dev_private;
3599 	int rc;
3600 	uint32_t dir_entries;
3601 	uint32_t entry_length;
3602 
3603 	rc = is_bnxt_in_error(bp);
3604 	if (rc)
3605 		return rc;
3606 
3607 	PMD_DRV_LOG(INFO, PCI_PRI_FMT "\n",
3608 		    bp->pdev->addr.domain, bp->pdev->addr.bus,
3609 		    bp->pdev->addr.devid, bp->pdev->addr.function);
3610 
3611 	rc = bnxt_hwrm_nvm_get_dir_info(bp, &dir_entries, &entry_length);
3612 	if (rc != 0)
3613 		return rc;
3614 
3615 	return dir_entries * entry_length;
3616 }
3617 
3618 static int
3619 bnxt_get_eeprom_op(struct rte_eth_dev *dev,
3620 		struct rte_dev_eeprom_info *in_eeprom)
3621 {
3622 	struct bnxt *bp = dev->data->dev_private;
3623 	uint32_t index;
3624 	uint32_t offset;
3625 	int rc;
3626 
3627 	rc = is_bnxt_in_error(bp);
3628 	if (rc)
3629 		return rc;
3630 
3631 	PMD_DRV_LOG(INFO, PCI_PRI_FMT " in_eeprom->offset = %d len = %d\n",
3632 		    bp->pdev->addr.domain, bp->pdev->addr.bus,
3633 		    bp->pdev->addr.devid, bp->pdev->addr.function,
3634 		    in_eeprom->offset, in_eeprom->length);
3635 
3636 	if (in_eeprom->offset == 0) /* special offset value to get directory */
3637 		return bnxt_get_nvram_directory(bp, in_eeprom->length,
3638 						in_eeprom->data);
3639 
3640 	index = in_eeprom->offset >> 24;
3641 	offset = in_eeprom->offset & 0xffffff;
3642 
3643 	if (index != 0)
3644 		return bnxt_hwrm_get_nvram_item(bp, index - 1, offset,
3645 					   in_eeprom->length, in_eeprom->data);
3646 
3647 	return 0;
3648 }
3649 
3650 static bool bnxt_dir_type_is_ape_bin_format(uint16_t dir_type)
3651 {
3652 	switch (dir_type) {
3653 	case BNX_DIR_TYPE_CHIMP_PATCH:
3654 	case BNX_DIR_TYPE_BOOTCODE:
3655 	case BNX_DIR_TYPE_BOOTCODE_2:
3656 	case BNX_DIR_TYPE_APE_FW:
3657 	case BNX_DIR_TYPE_APE_PATCH:
3658 	case BNX_DIR_TYPE_KONG_FW:
3659 	case BNX_DIR_TYPE_KONG_PATCH:
3660 	case BNX_DIR_TYPE_BONO_FW:
3661 	case BNX_DIR_TYPE_BONO_PATCH:
3662 		/* FALLTHROUGH */
3663 		return true;
3664 	}
3665 
3666 	return false;
3667 }
3668 
3669 static bool bnxt_dir_type_is_other_exec_format(uint16_t dir_type)
3670 {
3671 	switch (dir_type) {
3672 	case BNX_DIR_TYPE_AVS:
3673 	case BNX_DIR_TYPE_EXP_ROM_MBA:
3674 	case BNX_DIR_TYPE_PCIE:
3675 	case BNX_DIR_TYPE_TSCF_UCODE:
3676 	case BNX_DIR_TYPE_EXT_PHY:
3677 	case BNX_DIR_TYPE_CCM:
3678 	case BNX_DIR_TYPE_ISCSI_BOOT:
3679 	case BNX_DIR_TYPE_ISCSI_BOOT_IPV6:
3680 	case BNX_DIR_TYPE_ISCSI_BOOT_IPV4N6:
3681 		/* FALLTHROUGH */
3682 		return true;
3683 	}
3684 
3685 	return false;
3686 }
3687 
3688 static bool bnxt_dir_type_is_executable(uint16_t dir_type)
3689 {
3690 	return bnxt_dir_type_is_ape_bin_format(dir_type) ||
3691 		bnxt_dir_type_is_other_exec_format(dir_type);
3692 }
3693 
3694 static int
3695 bnxt_set_eeprom_op(struct rte_eth_dev *dev,
3696 		struct rte_dev_eeprom_info *in_eeprom)
3697 {
3698 	struct bnxt *bp = dev->data->dev_private;
3699 	uint8_t index, dir_op;
3700 	uint16_t type, ext, ordinal, attr;
3701 	int rc;
3702 
3703 	rc = is_bnxt_in_error(bp);
3704 	if (rc)
3705 		return rc;
3706 
3707 	PMD_DRV_LOG(INFO, PCI_PRI_FMT " in_eeprom->offset = %d len = %d\n",
3708 		    bp->pdev->addr.domain, bp->pdev->addr.bus,
3709 		    bp->pdev->addr.devid, bp->pdev->addr.function,
3710 		    in_eeprom->offset, in_eeprom->length);
3711 
3712 	if (!BNXT_PF(bp)) {
3713 		PMD_DRV_LOG(ERR, "NVM write not supported from a VF\n");
3714 		return -EINVAL;
3715 	}
3716 
3717 	type = in_eeprom->magic >> 16;
3718 
3719 	if (type == 0xffff) { /* special value for directory operations */
3720 		index = in_eeprom->magic & 0xff;
3721 		dir_op = in_eeprom->magic >> 8;
3722 		if (index == 0)
3723 			return -EINVAL;
3724 		switch (dir_op) {
3725 		case 0x0e: /* erase */
3726 			if (in_eeprom->offset != ~in_eeprom->magic)
3727 				return -EINVAL;
3728 			return bnxt_hwrm_erase_nvram_directory(bp, index - 1);
3729 		default:
3730 			return -EINVAL;
3731 		}
3732 	}
3733 
3734 	/* Create or re-write an NVM item: */
3735 	if (bnxt_dir_type_is_executable(type) == true)
3736 		return -EOPNOTSUPP;
3737 	ext = in_eeprom->magic & 0xffff;
3738 	ordinal = in_eeprom->offset >> 16;
3739 	attr = in_eeprom->offset & 0xffff;
3740 
3741 	return bnxt_hwrm_flash_nvram(bp, type, ordinal, ext, attr,
3742 				     in_eeprom->data, in_eeprom->length);
3743 }
3744 
3745 /*
3746  * Initialization
3747  */
3748 
3749 static const struct eth_dev_ops bnxt_dev_ops = {
3750 	.dev_infos_get = bnxt_dev_info_get_op,
3751 	.dev_close = bnxt_dev_close_op,
3752 	.dev_configure = bnxt_dev_configure_op,
3753 	.dev_start = bnxt_dev_start_op,
3754 	.dev_stop = bnxt_dev_stop_op,
3755 	.dev_set_link_up = bnxt_dev_set_link_up_op,
3756 	.dev_set_link_down = bnxt_dev_set_link_down_op,
3757 	.stats_get = bnxt_stats_get_op,
3758 	.stats_reset = bnxt_stats_reset_op,
3759 	.rx_queue_setup = bnxt_rx_queue_setup_op,
3760 	.rx_queue_release = bnxt_rx_queue_release_op,
3761 	.tx_queue_setup = bnxt_tx_queue_setup_op,
3762 	.tx_queue_release = bnxt_tx_queue_release_op,
3763 	.rx_queue_intr_enable = bnxt_rx_queue_intr_enable_op,
3764 	.rx_queue_intr_disable = bnxt_rx_queue_intr_disable_op,
3765 	.reta_update = bnxt_reta_update_op,
3766 	.reta_query = bnxt_reta_query_op,
3767 	.rss_hash_update = bnxt_rss_hash_update_op,
3768 	.rss_hash_conf_get = bnxt_rss_hash_conf_get_op,
3769 	.link_update = bnxt_link_update_op,
3770 	.promiscuous_enable = bnxt_promiscuous_enable_op,
3771 	.promiscuous_disable = bnxt_promiscuous_disable_op,
3772 	.allmulticast_enable = bnxt_allmulticast_enable_op,
3773 	.allmulticast_disable = bnxt_allmulticast_disable_op,
3774 	.mac_addr_add = bnxt_mac_addr_add_op,
3775 	.mac_addr_remove = bnxt_mac_addr_remove_op,
3776 	.flow_ctrl_get = bnxt_flow_ctrl_get_op,
3777 	.flow_ctrl_set = bnxt_flow_ctrl_set_op,
3778 	.udp_tunnel_port_add  = bnxt_udp_tunnel_port_add_op,
3779 	.udp_tunnel_port_del  = bnxt_udp_tunnel_port_del_op,
3780 	.vlan_filter_set = bnxt_vlan_filter_set_op,
3781 	.vlan_offload_set = bnxt_vlan_offload_set_op,
3782 	.vlan_tpid_set = bnxt_vlan_tpid_set_op,
3783 	.vlan_pvid_set = bnxt_vlan_pvid_set_op,
3784 	.mtu_set = bnxt_mtu_set_op,
3785 	.mac_addr_set = bnxt_set_default_mac_addr_op,
3786 	.xstats_get = bnxt_dev_xstats_get_op,
3787 	.xstats_get_names = bnxt_dev_xstats_get_names_op,
3788 	.xstats_reset = bnxt_dev_xstats_reset_op,
3789 	.fw_version_get = bnxt_fw_version_get,
3790 	.set_mc_addr_list = bnxt_dev_set_mc_addr_list_op,
3791 	.rxq_info_get = bnxt_rxq_info_get_op,
3792 	.txq_info_get = bnxt_txq_info_get_op,
3793 	.rx_burst_mode_get = bnxt_rx_burst_mode_get,
3794 	.tx_burst_mode_get = bnxt_tx_burst_mode_get,
3795 	.dev_led_on = bnxt_dev_led_on_op,
3796 	.dev_led_off = bnxt_dev_led_off_op,
3797 	.rx_queue_start = bnxt_rx_queue_start,
3798 	.rx_queue_stop = bnxt_rx_queue_stop,
3799 	.tx_queue_start = bnxt_tx_queue_start,
3800 	.tx_queue_stop = bnxt_tx_queue_stop,
3801 	.flow_ops_get = bnxt_flow_ops_get_op,
3802 	.dev_supported_ptypes_get = bnxt_dev_supported_ptypes_get_op,
3803 	.get_eeprom_length    = bnxt_get_eeprom_length_op,
3804 	.get_eeprom           = bnxt_get_eeprom_op,
3805 	.set_eeprom           = bnxt_set_eeprom_op,
3806 	.timesync_enable      = bnxt_timesync_enable,
3807 	.timesync_disable     = bnxt_timesync_disable,
3808 	.timesync_read_time   = bnxt_timesync_read_time,
3809 	.timesync_write_time   = bnxt_timesync_write_time,
3810 	.timesync_adjust_time = bnxt_timesync_adjust_time,
3811 	.timesync_read_rx_timestamp = bnxt_timesync_read_rx_timestamp,
3812 	.timesync_read_tx_timestamp = bnxt_timesync_read_tx_timestamp,
3813 };
3814 
3815 static uint32_t bnxt_map_reset_regs(struct bnxt *bp, uint32_t reg)
3816 {
3817 	uint32_t offset;
3818 
3819 	/* Only pre-map the reset GRC registers using window 3 */
3820 	rte_write32(reg & 0xfffff000, (uint8_t *)bp->bar0 +
3821 		    BNXT_GRCPF_REG_WINDOW_BASE_OUT + 8);
3822 
3823 	offset = BNXT_GRCP_WINDOW_3_BASE + (reg & 0xffc);
3824 
3825 	return offset;
3826 }
3827 
3828 int bnxt_map_fw_health_status_regs(struct bnxt *bp)
3829 {
3830 	struct bnxt_error_recovery_info *info = bp->recovery_info;
3831 	uint32_t reg_base = 0xffffffff;
3832 	int i;
3833 
3834 	/* Only pre-map the monitoring GRC registers using window 2 */
3835 	for (i = 0; i < BNXT_FW_STATUS_REG_CNT; i++) {
3836 		uint32_t reg = info->status_regs[i];
3837 
3838 		if (BNXT_FW_STATUS_REG_TYPE(reg) != BNXT_FW_STATUS_REG_TYPE_GRC)
3839 			continue;
3840 
3841 		if (reg_base == 0xffffffff)
3842 			reg_base = reg & 0xfffff000;
3843 		if ((reg & 0xfffff000) != reg_base)
3844 			return -ERANGE;
3845 
3846 		/* Use mask 0xffc as the Lower 2 bits indicates
3847 		 * address space location
3848 		 */
3849 		info->mapped_status_regs[i] = BNXT_GRCP_WINDOW_2_BASE +
3850 						(reg & 0xffc);
3851 	}
3852 
3853 	if (reg_base == 0xffffffff)
3854 		return 0;
3855 
3856 	rte_write32(reg_base, (uint8_t *)bp->bar0 +
3857 		    BNXT_GRCPF_REG_WINDOW_BASE_OUT + 4);
3858 
3859 	return 0;
3860 }
3861 
3862 static void bnxt_write_fw_reset_reg(struct bnxt *bp, uint32_t index)
3863 {
3864 	struct bnxt_error_recovery_info *info = bp->recovery_info;
3865 	uint32_t delay = info->delay_after_reset[index];
3866 	uint32_t val = info->reset_reg_val[index];
3867 	uint32_t reg = info->reset_reg[index];
3868 	uint32_t type, offset;
3869 	int ret;
3870 
3871 	type = BNXT_FW_STATUS_REG_TYPE(reg);
3872 	offset = BNXT_FW_STATUS_REG_OFF(reg);
3873 
3874 	switch (type) {
3875 	case BNXT_FW_STATUS_REG_TYPE_CFG:
3876 		ret = rte_pci_write_config(bp->pdev, &val, sizeof(val), offset);
3877 		if (ret < 0) {
3878 			PMD_DRV_LOG(ERR, "Failed to write %#x at PCI offset %#x",
3879 				    val, offset);
3880 			return;
3881 		}
3882 		break;
3883 	case BNXT_FW_STATUS_REG_TYPE_GRC:
3884 		offset = bnxt_map_reset_regs(bp, offset);
3885 		rte_write32(val, (uint8_t *)bp->bar0 + offset);
3886 		break;
3887 	case BNXT_FW_STATUS_REG_TYPE_BAR0:
3888 		rte_write32(val, (uint8_t *)bp->bar0 + offset);
3889 		break;
3890 	}
3891 	/* wait on a specific interval of time until core reset is complete */
3892 	if (delay)
3893 		rte_delay_ms(delay);
3894 }
3895 
3896 static void bnxt_dev_cleanup(struct bnxt *bp)
3897 {
3898 	bp->eth_dev->data->dev_link.link_status = 0;
3899 	bp->link_info->link_up = 0;
3900 	if (bp->eth_dev->data->dev_started)
3901 		bnxt_dev_stop(bp->eth_dev);
3902 
3903 	bnxt_uninit_resources(bp, true);
3904 }
3905 
3906 static int
3907 bnxt_check_fw_reset_done(struct bnxt *bp)
3908 {
3909 	int timeout = bp->fw_reset_max_msecs;
3910 	uint16_t val = 0;
3911 	int rc;
3912 
3913 	do {
3914 		rc = rte_pci_read_config(bp->pdev, &val, sizeof(val), PCI_SUBSYSTEM_ID_OFFSET);
3915 		if (rc < 0) {
3916 			PMD_DRV_LOG(ERR, "Failed to read PCI offset 0x%x", PCI_SUBSYSTEM_ID_OFFSET);
3917 			return rc;
3918 		}
3919 		if (val != 0xffff)
3920 			break;
3921 		rte_delay_ms(1);
3922 	} while (timeout--);
3923 
3924 	if (val == 0xffff) {
3925 		PMD_DRV_LOG(ERR, "Firmware reset aborted, PCI config space invalid\n");
3926 		return -1;
3927 	}
3928 
3929 	return 0;
3930 }
3931 
3932 static int bnxt_restore_vlan_filters(struct bnxt *bp)
3933 {
3934 	struct rte_eth_dev *dev = bp->eth_dev;
3935 	struct rte_vlan_filter_conf *vfc;
3936 	int vidx, vbit, rc;
3937 	uint16_t vlan_id;
3938 
3939 	for (vlan_id = 1; vlan_id <= RTE_ETHER_MAX_VLAN_ID; vlan_id++) {
3940 		vfc = &dev->data->vlan_filter_conf;
3941 		vidx = vlan_id / 64;
3942 		vbit = vlan_id % 64;
3943 
3944 		/* Each bit corresponds to a VLAN id */
3945 		if (vfc->ids[vidx] & (UINT64_C(1) << vbit)) {
3946 			rc = bnxt_add_vlan_filter(bp, vlan_id);
3947 			if (rc)
3948 				return rc;
3949 		}
3950 	}
3951 
3952 	return 0;
3953 }
3954 
3955 static int bnxt_restore_mac_filters(struct bnxt *bp)
3956 {
3957 	struct rte_eth_dev *dev = bp->eth_dev;
3958 	struct rte_eth_dev_info dev_info;
3959 	struct rte_ether_addr *addr;
3960 	uint64_t pool_mask;
3961 	uint32_t pool = 0;
3962 	uint16_t i;
3963 	int rc;
3964 
3965 	if (BNXT_VF(bp) && !BNXT_VF_IS_TRUSTED(bp))
3966 		return 0;
3967 
3968 	rc = bnxt_dev_info_get_op(dev, &dev_info);
3969 	if (rc)
3970 		return rc;
3971 
3972 	/* replay MAC address configuration */
3973 	for (i = 1; i < dev_info.max_mac_addrs; i++) {
3974 		addr = &dev->data->mac_addrs[i];
3975 
3976 		/* skip zero address */
3977 		if (rte_is_zero_ether_addr(addr))
3978 			continue;
3979 
3980 		pool = 0;
3981 		pool_mask = dev->data->mac_pool_sel[i];
3982 
3983 		do {
3984 			if (pool_mask & 1ULL) {
3985 				rc = bnxt_mac_addr_add_op(dev, addr, i, pool);
3986 				if (rc)
3987 					return rc;
3988 			}
3989 			pool_mask >>= 1;
3990 			pool++;
3991 		} while (pool_mask);
3992 	}
3993 
3994 	return 0;
3995 }
3996 
3997 static int bnxt_restore_filters(struct bnxt *bp)
3998 {
3999 	struct rte_eth_dev *dev = bp->eth_dev;
4000 	int ret = 0;
4001 
4002 	if (dev->data->all_multicast) {
4003 		ret = bnxt_allmulticast_enable_op(dev);
4004 		if (ret)
4005 			return ret;
4006 	}
4007 	if (dev->data->promiscuous) {
4008 		ret = bnxt_promiscuous_enable_op(dev);
4009 		if (ret)
4010 			return ret;
4011 	}
4012 
4013 	ret = bnxt_restore_mac_filters(bp);
4014 	if (ret)
4015 		return ret;
4016 
4017 	ret = bnxt_restore_vlan_filters(bp);
4018 	/* TODO restore other filters as well */
4019 	return ret;
4020 }
4021 
4022 static int bnxt_check_fw_ready(struct bnxt *bp)
4023 {
4024 	int timeout = bp->fw_reset_max_msecs;
4025 	int rc = 0;
4026 
4027 	do {
4028 		rc = bnxt_hwrm_poll_ver_get(bp);
4029 		if (rc == 0)
4030 			break;
4031 		rte_delay_ms(BNXT_FW_READY_WAIT_INTERVAL);
4032 		timeout -= BNXT_FW_READY_WAIT_INTERVAL;
4033 	} while (rc && timeout > 0);
4034 
4035 	if (rc)
4036 		PMD_DRV_LOG(ERR, "FW is not Ready after reset\n");
4037 
4038 	return rc;
4039 }
4040 
4041 static void bnxt_dev_recover(void *arg)
4042 {
4043 	struct bnxt *bp = arg;
4044 	int rc = 0;
4045 
4046 	pthread_mutex_lock(&bp->err_recovery_lock);
4047 
4048 	if (!bp->fw_reset_min_msecs) {
4049 		rc = bnxt_check_fw_reset_done(bp);
4050 		if (rc)
4051 			goto err;
4052 	}
4053 
4054 	/* Clear Error flag so that device re-init should happen */
4055 	bp->flags &= ~BNXT_FLAG_FATAL_ERROR;
4056 
4057 	rc = bnxt_check_fw_ready(bp);
4058 	if (rc)
4059 		goto err;
4060 
4061 	rc = bnxt_init_resources(bp, true);
4062 	if (rc) {
4063 		PMD_DRV_LOG(ERR,
4064 			    "Failed to initialize resources after reset\n");
4065 		goto err;
4066 	}
4067 	/* clear reset flag as the device is initialized now */
4068 	bp->flags &= ~BNXT_FLAG_FW_RESET;
4069 
4070 	rc = bnxt_dev_start_op(bp->eth_dev);
4071 	if (rc) {
4072 		PMD_DRV_LOG(ERR, "Failed to start port after reset\n");
4073 		goto err_start;
4074 	}
4075 
4076 	rc = bnxt_restore_filters(bp);
4077 	if (rc)
4078 		goto err_start;
4079 
4080 	PMD_DRV_LOG(INFO, "Recovered from FW reset\n");
4081 	pthread_mutex_unlock(&bp->err_recovery_lock);
4082 
4083 	return;
4084 err_start:
4085 	bnxt_dev_stop(bp->eth_dev);
4086 err:
4087 	bp->flags |= BNXT_FLAG_FATAL_ERROR;
4088 	bnxt_uninit_resources(bp, false);
4089 	pthread_mutex_unlock(&bp->err_recovery_lock);
4090 	PMD_DRV_LOG(ERR, "Failed to recover from FW reset\n");
4091 }
4092 
4093 void bnxt_dev_reset_and_resume(void *arg)
4094 {
4095 	struct bnxt *bp = arg;
4096 	uint32_t us = US_PER_MS * bp->fw_reset_min_msecs;
4097 	uint16_t val = 0;
4098 	int rc;
4099 
4100 	bnxt_dev_cleanup(bp);
4101 
4102 	bnxt_wait_for_device_shutdown(bp);
4103 
4104 	/* During some fatal firmware error conditions, the PCI config space
4105 	 * register 0x2e which normally contains the subsystem ID will become
4106 	 * 0xffff. This register will revert back to the normal value after
4107 	 * the chip has completed core reset. If we detect this condition,
4108 	 * we can poll this config register immediately for the value to revert.
4109 	 */
4110 	if (bp->flags & BNXT_FLAG_FATAL_ERROR) {
4111 		rc = rte_pci_read_config(bp->pdev, &val, sizeof(val), PCI_SUBSYSTEM_ID_OFFSET);
4112 		if (rc < 0) {
4113 			PMD_DRV_LOG(ERR, "Failed to read PCI offset 0x%x", PCI_SUBSYSTEM_ID_OFFSET);
4114 			return;
4115 		}
4116 		if (val == 0xffff) {
4117 			bp->fw_reset_min_msecs = 0;
4118 			us = 1;
4119 		}
4120 	}
4121 
4122 	rc = rte_eal_alarm_set(us, bnxt_dev_recover, (void *)bp);
4123 	if (rc)
4124 		PMD_DRV_LOG(ERR, "Error setting recovery alarm");
4125 }
4126 
4127 uint32_t bnxt_read_fw_status_reg(struct bnxt *bp, uint32_t index)
4128 {
4129 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4130 	uint32_t reg = info->status_regs[index];
4131 	uint32_t type, offset, val = 0;
4132 
4133 	type = BNXT_FW_STATUS_REG_TYPE(reg);
4134 	offset = BNXT_FW_STATUS_REG_OFF(reg);
4135 
4136 	switch (type) {
4137 	case BNXT_FW_STATUS_REG_TYPE_CFG:
4138 		rte_pci_read_config(bp->pdev, &val, sizeof(val), offset);
4139 		break;
4140 	case BNXT_FW_STATUS_REG_TYPE_GRC:
4141 		offset = info->mapped_status_regs[index];
4142 		/* FALLTHROUGH */
4143 	case BNXT_FW_STATUS_REG_TYPE_BAR0:
4144 		val = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
4145 				       offset));
4146 		break;
4147 	}
4148 
4149 	return val;
4150 }
4151 
4152 static int bnxt_fw_reset_all(struct bnxt *bp)
4153 {
4154 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4155 	uint32_t i;
4156 	int rc = 0;
4157 
4158 	if (info->flags & BNXT_FLAG_ERROR_RECOVERY_HOST) {
4159 		/* Reset through master function driver */
4160 		for (i = 0; i < info->reg_array_cnt; i++)
4161 			bnxt_write_fw_reset_reg(bp, i);
4162 		/* Wait for time specified by FW after triggering reset */
4163 		rte_delay_ms(info->master_func_wait_period_after_reset);
4164 	} else if (info->flags & BNXT_FLAG_ERROR_RECOVERY_CO_CPU) {
4165 		/* Reset with the help of Kong processor */
4166 		rc = bnxt_hwrm_fw_reset(bp);
4167 		if (rc)
4168 			PMD_DRV_LOG(ERR, "Failed to reset FW\n");
4169 	}
4170 
4171 	return rc;
4172 }
4173 
4174 static void bnxt_fw_reset_cb(void *arg)
4175 {
4176 	struct bnxt *bp = arg;
4177 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4178 	int rc = 0;
4179 
4180 	/* Only Master function can do FW reset */
4181 	if (bnxt_is_master_func(bp) &&
4182 	    bnxt_is_recovery_enabled(bp)) {
4183 		rc = bnxt_fw_reset_all(bp);
4184 		if (rc) {
4185 			PMD_DRV_LOG(ERR, "Adapter recovery failed\n");
4186 			return;
4187 		}
4188 	}
4189 
4190 	/* if recovery method is ERROR_RECOVERY_CO_CPU, KONG will send
4191 	 * EXCEPTION_FATAL_ASYNC event to all the functions
4192 	 * (including MASTER FUNC). After receiving this Async, all the active
4193 	 * drivers should treat this case as FW initiated recovery
4194 	 */
4195 	if (info->flags & BNXT_FLAG_ERROR_RECOVERY_HOST) {
4196 		bp->fw_reset_min_msecs = BNXT_MIN_FW_READY_TIMEOUT;
4197 		bp->fw_reset_max_msecs = BNXT_MAX_FW_RESET_TIMEOUT;
4198 
4199 		/* To recover from error */
4200 		rte_eal_alarm_set(US_PER_MS, bnxt_dev_reset_and_resume,
4201 				  (void *)bp);
4202 	}
4203 }
4204 
4205 /* Driver should poll FW heartbeat, reset_counter with the frequency
4206  * advertised by FW in HWRM_ERROR_RECOVERY_QCFG.
4207  * When the driver detects heartbeat stop or change in reset_counter,
4208  * it has to trigger a reset to recover from the error condition.
4209  * A “master PF” is the function who will have the privilege to
4210  * initiate the chimp reset. The master PF will be elected by the
4211  * firmware and will be notified through async message.
4212  */
4213 static void bnxt_check_fw_health(void *arg)
4214 {
4215 	struct bnxt *bp = arg;
4216 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4217 	uint32_t val = 0, wait_msec;
4218 
4219 	if (!info || !bnxt_is_recovery_enabled(bp) ||
4220 	    is_bnxt_in_error(bp))
4221 		return;
4222 
4223 	val = bnxt_read_fw_status_reg(bp, BNXT_FW_HEARTBEAT_CNT_REG);
4224 	if (val == info->last_heart_beat)
4225 		goto reset;
4226 
4227 	info->last_heart_beat = val;
4228 
4229 	val = bnxt_read_fw_status_reg(bp, BNXT_FW_RECOVERY_CNT_REG);
4230 	if (val != info->last_reset_counter)
4231 		goto reset;
4232 
4233 	info->last_reset_counter = val;
4234 
4235 	rte_eal_alarm_set(US_PER_MS * info->driver_polling_freq,
4236 			  bnxt_check_fw_health, (void *)bp);
4237 
4238 	return;
4239 reset:
4240 	/* Stop DMA to/from device */
4241 	bp->flags |= BNXT_FLAG_FATAL_ERROR;
4242 	bp->flags |= BNXT_FLAG_FW_RESET;
4243 
4244 	PMD_DRV_LOG(ERR, "Detected FW dead condition\n");
4245 
4246 	if (bnxt_is_master_func(bp))
4247 		wait_msec = info->master_func_wait_period;
4248 	else
4249 		wait_msec = info->normal_func_wait_period;
4250 
4251 	rte_eal_alarm_set(US_PER_MS * wait_msec,
4252 			  bnxt_fw_reset_cb, (void *)bp);
4253 }
4254 
4255 void bnxt_schedule_fw_health_check(struct bnxt *bp)
4256 {
4257 	uint32_t polling_freq;
4258 
4259 	pthread_mutex_lock(&bp->health_check_lock);
4260 
4261 	if (!bnxt_is_recovery_enabled(bp))
4262 		goto done;
4263 
4264 	if (bp->flags & BNXT_FLAG_FW_HEALTH_CHECK_SCHEDULED)
4265 		goto done;
4266 
4267 	polling_freq = bp->recovery_info->driver_polling_freq;
4268 
4269 	rte_eal_alarm_set(US_PER_MS * polling_freq,
4270 			  bnxt_check_fw_health, (void *)bp);
4271 	bp->flags |= BNXT_FLAG_FW_HEALTH_CHECK_SCHEDULED;
4272 
4273 done:
4274 	pthread_mutex_unlock(&bp->health_check_lock);
4275 }
4276 
4277 static void bnxt_cancel_fw_health_check(struct bnxt *bp)
4278 {
4279 	rte_eal_alarm_cancel(bnxt_check_fw_health, (void *)bp);
4280 	bp->flags &= ~BNXT_FLAG_FW_HEALTH_CHECK_SCHEDULED;
4281 }
4282 
4283 static bool bnxt_vf_pciid(uint16_t device_id)
4284 {
4285 	switch (device_id) {
4286 	case BROADCOM_DEV_ID_57304_VF:
4287 	case BROADCOM_DEV_ID_57406_VF:
4288 	case BROADCOM_DEV_ID_5731X_VF:
4289 	case BROADCOM_DEV_ID_5741X_VF:
4290 	case BROADCOM_DEV_ID_57414_VF:
4291 	case BROADCOM_DEV_ID_STRATUS_NIC_VF1:
4292 	case BROADCOM_DEV_ID_STRATUS_NIC_VF2:
4293 	case BROADCOM_DEV_ID_58802_VF:
4294 	case BROADCOM_DEV_ID_57500_VF1:
4295 	case BROADCOM_DEV_ID_57500_VF2:
4296 	case BROADCOM_DEV_ID_58818_VF:
4297 		/* FALLTHROUGH */
4298 		return true;
4299 	default:
4300 		return false;
4301 	}
4302 }
4303 
4304 /* Phase 5 device */
4305 static bool bnxt_p5_device(uint16_t device_id)
4306 {
4307 	switch (device_id) {
4308 	case BROADCOM_DEV_ID_57508:
4309 	case BROADCOM_DEV_ID_57504:
4310 	case BROADCOM_DEV_ID_57502:
4311 	case BROADCOM_DEV_ID_57508_MF1:
4312 	case BROADCOM_DEV_ID_57504_MF1:
4313 	case BROADCOM_DEV_ID_57502_MF1:
4314 	case BROADCOM_DEV_ID_57508_MF2:
4315 	case BROADCOM_DEV_ID_57504_MF2:
4316 	case BROADCOM_DEV_ID_57502_MF2:
4317 	case BROADCOM_DEV_ID_57500_VF1:
4318 	case BROADCOM_DEV_ID_57500_VF2:
4319 	case BROADCOM_DEV_ID_58812:
4320 	case BROADCOM_DEV_ID_58814:
4321 	case BROADCOM_DEV_ID_58818:
4322 	case BROADCOM_DEV_ID_58818_VF:
4323 		/* FALLTHROUGH */
4324 		return true;
4325 	default:
4326 		return false;
4327 	}
4328 }
4329 
4330 bool bnxt_stratus_device(struct bnxt *bp)
4331 {
4332 	uint16_t device_id = bp->pdev->id.device_id;
4333 
4334 	switch (device_id) {
4335 	case BROADCOM_DEV_ID_STRATUS_NIC:
4336 	case BROADCOM_DEV_ID_STRATUS_NIC_VF1:
4337 	case BROADCOM_DEV_ID_STRATUS_NIC_VF2:
4338 		/* FALLTHROUGH */
4339 		return true;
4340 	default:
4341 		return false;
4342 	}
4343 }
4344 
4345 static int bnxt_map_pci_bars(struct rte_eth_dev *eth_dev)
4346 {
4347 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
4348 	struct bnxt *bp = eth_dev->data->dev_private;
4349 
4350 	/* enable device (incl. PCI PM wakeup), and bus-mastering */
4351 	bp->bar0 = (void *)pci_dev->mem_resource[0].addr;
4352 	bp->doorbell_base = (void *)pci_dev->mem_resource[2].addr;
4353 	if (!bp->bar0 || !bp->doorbell_base) {
4354 		PMD_DRV_LOG(ERR, "Unable to access Hardware\n");
4355 		return -ENODEV;
4356 	}
4357 
4358 	bp->eth_dev = eth_dev;
4359 	bp->pdev = pci_dev;
4360 
4361 	return 0;
4362 }
4363 
4364 static int bnxt_alloc_ctx_mem_blk(struct bnxt *bp,
4365 				  struct bnxt_ctx_pg_info *ctx_pg,
4366 				  uint32_t mem_size,
4367 				  const char *suffix,
4368 				  uint16_t idx)
4369 {
4370 	struct bnxt_ring_mem_info *rmem = &ctx_pg->ring_mem;
4371 	const struct rte_memzone *mz = NULL;
4372 	char mz_name[RTE_MEMZONE_NAMESIZE];
4373 	rte_iova_t mz_phys_addr;
4374 	uint64_t valid_bits = 0;
4375 	uint32_t sz;
4376 	int i;
4377 
4378 	if (!mem_size)
4379 		return 0;
4380 
4381 	rmem->nr_pages = RTE_ALIGN_MUL_CEIL(mem_size, BNXT_PAGE_SIZE) /
4382 			 BNXT_PAGE_SIZE;
4383 	rmem->page_size = BNXT_PAGE_SIZE;
4384 	rmem->pg_arr = ctx_pg->ctx_pg_arr;
4385 	rmem->dma_arr = ctx_pg->ctx_dma_arr;
4386 	rmem->flags = BNXT_RMEM_VALID_PTE_FLAG;
4387 
4388 	valid_bits = PTU_PTE_VALID;
4389 
4390 	if (rmem->nr_pages > 1) {
4391 		snprintf(mz_name, RTE_MEMZONE_NAMESIZE,
4392 			 "bnxt_ctx_pg_tbl%s_%x_%d",
4393 			 suffix, idx, bp->eth_dev->data->port_id);
4394 		mz_name[RTE_MEMZONE_NAMESIZE - 1] = 0;
4395 		mz = rte_memzone_lookup(mz_name);
4396 		if (!mz) {
4397 			mz = rte_memzone_reserve_aligned(mz_name,
4398 						rmem->nr_pages * 8,
4399 						SOCKET_ID_ANY,
4400 						RTE_MEMZONE_2MB |
4401 						RTE_MEMZONE_SIZE_HINT_ONLY |
4402 						RTE_MEMZONE_IOVA_CONTIG,
4403 						BNXT_PAGE_SIZE);
4404 			if (mz == NULL)
4405 				return -ENOMEM;
4406 		}
4407 
4408 		memset(mz->addr, 0, mz->len);
4409 		mz_phys_addr = mz->iova;
4410 
4411 		rmem->pg_tbl = mz->addr;
4412 		rmem->pg_tbl_map = mz_phys_addr;
4413 		rmem->pg_tbl_mz = mz;
4414 	}
4415 
4416 	snprintf(mz_name, RTE_MEMZONE_NAMESIZE, "bnxt_ctx_%s_%x_%d",
4417 		 suffix, idx, bp->eth_dev->data->port_id);
4418 	mz = rte_memzone_lookup(mz_name);
4419 	if (!mz) {
4420 		mz = rte_memzone_reserve_aligned(mz_name,
4421 						 mem_size,
4422 						 SOCKET_ID_ANY,
4423 						 RTE_MEMZONE_1GB |
4424 						 RTE_MEMZONE_SIZE_HINT_ONLY |
4425 						 RTE_MEMZONE_IOVA_CONTIG,
4426 						 BNXT_PAGE_SIZE);
4427 		if (mz == NULL)
4428 			return -ENOMEM;
4429 	}
4430 
4431 	memset(mz->addr, 0, mz->len);
4432 	mz_phys_addr = mz->iova;
4433 
4434 	for (sz = 0, i = 0; sz < mem_size; sz += BNXT_PAGE_SIZE, i++) {
4435 		rmem->pg_arr[i] = ((char *)mz->addr) + sz;
4436 		rmem->dma_arr[i] = mz_phys_addr + sz;
4437 
4438 		if (rmem->nr_pages > 1) {
4439 			if (i == rmem->nr_pages - 2 &&
4440 			    (rmem->flags & BNXT_RMEM_RING_PTE_FLAG))
4441 				valid_bits |= PTU_PTE_NEXT_TO_LAST;
4442 			else if (i == rmem->nr_pages - 1 &&
4443 				 (rmem->flags & BNXT_RMEM_RING_PTE_FLAG))
4444 				valid_bits |= PTU_PTE_LAST;
4445 
4446 			rmem->pg_tbl[i] = rte_cpu_to_le_64(rmem->dma_arr[i] |
4447 							   valid_bits);
4448 		}
4449 	}
4450 
4451 	rmem->mz = mz;
4452 	if (rmem->vmem_size)
4453 		rmem->vmem = (void **)mz->addr;
4454 	rmem->dma_arr[0] = mz_phys_addr;
4455 	return 0;
4456 }
4457 
4458 static void bnxt_free_ctx_mem(struct bnxt *bp)
4459 {
4460 	int i;
4461 
4462 	if (!bp->ctx || !(bp->ctx->flags & BNXT_CTX_FLAG_INITED))
4463 		return;
4464 
4465 	bp->ctx->flags &= ~BNXT_CTX_FLAG_INITED;
4466 	rte_memzone_free(bp->ctx->qp_mem.ring_mem.mz);
4467 	rte_memzone_free(bp->ctx->srq_mem.ring_mem.mz);
4468 	rte_memzone_free(bp->ctx->cq_mem.ring_mem.mz);
4469 	rte_memzone_free(bp->ctx->vnic_mem.ring_mem.mz);
4470 	rte_memzone_free(bp->ctx->stat_mem.ring_mem.mz);
4471 	rte_memzone_free(bp->ctx->qp_mem.ring_mem.pg_tbl_mz);
4472 	rte_memzone_free(bp->ctx->srq_mem.ring_mem.pg_tbl_mz);
4473 	rte_memzone_free(bp->ctx->cq_mem.ring_mem.pg_tbl_mz);
4474 	rte_memzone_free(bp->ctx->vnic_mem.ring_mem.pg_tbl_mz);
4475 	rte_memzone_free(bp->ctx->stat_mem.ring_mem.pg_tbl_mz);
4476 
4477 	for (i = 0; i < bp->ctx->tqm_fp_rings_count + 1; i++) {
4478 		if (bp->ctx->tqm_mem[i])
4479 			rte_memzone_free(bp->ctx->tqm_mem[i]->ring_mem.mz);
4480 	}
4481 
4482 	rte_free(bp->ctx);
4483 	bp->ctx = NULL;
4484 }
4485 
4486 #define bnxt_roundup(x, y)   ((((x) + ((y) - 1)) / (y)) * (y))
4487 
4488 #define min_t(type, x, y) ({                    \
4489 	type __min1 = (x);                      \
4490 	type __min2 = (y);                      \
4491 	__min1 < __min2 ? __min1 : __min2; })
4492 
4493 #define max_t(type, x, y) ({                    \
4494 	type __max1 = (x);                      \
4495 	type __max2 = (y);                      \
4496 	__max1 > __max2 ? __max1 : __max2; })
4497 
4498 #define clamp_t(type, _x, min, max)     min_t(type, max_t(type, _x, min), max)
4499 
4500 int bnxt_alloc_ctx_mem(struct bnxt *bp)
4501 {
4502 	struct bnxt_ctx_pg_info *ctx_pg;
4503 	struct bnxt_ctx_mem_info *ctx;
4504 	uint32_t mem_size, ena, entries;
4505 	uint32_t entries_sp, min;
4506 	int i, rc;
4507 
4508 	rc = bnxt_hwrm_func_backing_store_qcaps(bp);
4509 	if (rc) {
4510 		PMD_DRV_LOG(ERR, "Query context mem capability failed\n");
4511 		return rc;
4512 	}
4513 	ctx = bp->ctx;
4514 	if (!ctx || (ctx->flags & BNXT_CTX_FLAG_INITED))
4515 		return 0;
4516 
4517 	ctx_pg = &ctx->qp_mem;
4518 	ctx_pg->entries = ctx->qp_min_qp1_entries + ctx->qp_max_l2_entries;
4519 	if (ctx->qp_entry_size) {
4520 		mem_size = ctx->qp_entry_size * ctx_pg->entries;
4521 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "qp_mem", 0);
4522 		if (rc)
4523 			return rc;
4524 	}
4525 
4526 	ctx_pg = &ctx->srq_mem;
4527 	ctx_pg->entries = ctx->srq_max_l2_entries;
4528 	if (ctx->srq_entry_size) {
4529 		mem_size = ctx->srq_entry_size * ctx_pg->entries;
4530 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "srq_mem", 0);
4531 		if (rc)
4532 			return rc;
4533 	}
4534 
4535 	ctx_pg = &ctx->cq_mem;
4536 	ctx_pg->entries = ctx->cq_max_l2_entries;
4537 	if (ctx->cq_entry_size) {
4538 		mem_size = ctx->cq_entry_size * ctx_pg->entries;
4539 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "cq_mem", 0);
4540 		if (rc)
4541 			return rc;
4542 	}
4543 
4544 	ctx_pg = &ctx->vnic_mem;
4545 	ctx_pg->entries = ctx->vnic_max_vnic_entries +
4546 		ctx->vnic_max_ring_table_entries;
4547 	if (ctx->vnic_entry_size) {
4548 		mem_size = ctx->vnic_entry_size * ctx_pg->entries;
4549 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "vnic_mem", 0);
4550 		if (rc)
4551 			return rc;
4552 	}
4553 
4554 	ctx_pg = &ctx->stat_mem;
4555 	ctx_pg->entries = ctx->stat_max_entries;
4556 	if (ctx->stat_entry_size) {
4557 		mem_size = ctx->stat_entry_size * ctx_pg->entries;
4558 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "stat_mem", 0);
4559 		if (rc)
4560 			return rc;
4561 	}
4562 
4563 	min = ctx->tqm_min_entries_per_ring;
4564 
4565 	entries_sp = ctx->qp_max_l2_entries +
4566 		     ctx->vnic_max_vnic_entries +
4567 		     2 * ctx->qp_min_qp1_entries + min;
4568 	entries_sp = bnxt_roundup(entries_sp, ctx->tqm_entries_multiple);
4569 
4570 	entries = ctx->qp_max_l2_entries + ctx->qp_min_qp1_entries;
4571 	entries = bnxt_roundup(entries, ctx->tqm_entries_multiple);
4572 	entries = clamp_t(uint32_t, entries, min,
4573 			  ctx->tqm_max_entries_per_ring);
4574 	for (i = 0, ena = 0; i < ctx->tqm_fp_rings_count + 1; i++) {
4575 		/* i=0 is for TQM_SP. i=1 to i=8 applies to RING0 to RING7.
4576 		 * i > 8 is other ext rings.
4577 		 */
4578 		ctx_pg = ctx->tqm_mem[i];
4579 		ctx_pg->entries = i ? entries : entries_sp;
4580 		if (ctx->tqm_entry_size) {
4581 			mem_size = ctx->tqm_entry_size * ctx_pg->entries;
4582 			rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size,
4583 						    "tqm_mem", i);
4584 			if (rc)
4585 				return rc;
4586 		}
4587 		if (i < BNXT_MAX_TQM_LEGACY_RINGS)
4588 			ena |= HWRM_FUNC_BACKING_STORE_CFG_INPUT_ENABLES_TQM_SP << i;
4589 		else
4590 			ena |= HWRM_FUNC_BACKING_STORE_CFG_INPUT_ENABLES_TQM_RING8;
4591 	}
4592 
4593 	ena |= FUNC_BACKING_STORE_CFG_INPUT_DFLT_ENABLES;
4594 	rc = bnxt_hwrm_func_backing_store_cfg(bp, ena);
4595 	if (rc)
4596 		PMD_DRV_LOG(ERR,
4597 			    "Failed to configure context mem: rc = %d\n", rc);
4598 	else
4599 		ctx->flags |= BNXT_CTX_FLAG_INITED;
4600 
4601 	return rc;
4602 }
4603 
4604 static int bnxt_alloc_stats_mem(struct bnxt *bp)
4605 {
4606 	struct rte_pci_device *pci_dev = bp->pdev;
4607 	char mz_name[RTE_MEMZONE_NAMESIZE];
4608 	const struct rte_memzone *mz = NULL;
4609 	uint32_t total_alloc_len;
4610 	rte_iova_t mz_phys_addr;
4611 
4612 	if (pci_dev->id.device_id == BROADCOM_DEV_ID_NS2)
4613 		return 0;
4614 
4615 	snprintf(mz_name, RTE_MEMZONE_NAMESIZE,
4616 		 "bnxt_" PCI_PRI_FMT "-%s", pci_dev->addr.domain,
4617 		 pci_dev->addr.bus, pci_dev->addr.devid,
4618 		 pci_dev->addr.function, "rx_port_stats");
4619 	mz_name[RTE_MEMZONE_NAMESIZE - 1] = 0;
4620 	mz = rte_memzone_lookup(mz_name);
4621 	total_alloc_len =
4622 		RTE_CACHE_LINE_ROUNDUP(sizeof(struct rx_port_stats) +
4623 				       sizeof(struct rx_port_stats_ext) + 512);
4624 	if (!mz) {
4625 		mz = rte_memzone_reserve(mz_name, total_alloc_len,
4626 					 SOCKET_ID_ANY,
4627 					 RTE_MEMZONE_2MB |
4628 					 RTE_MEMZONE_SIZE_HINT_ONLY |
4629 					 RTE_MEMZONE_IOVA_CONTIG);
4630 		if (mz == NULL)
4631 			return -ENOMEM;
4632 	}
4633 	memset(mz->addr, 0, mz->len);
4634 	mz_phys_addr = mz->iova;
4635 
4636 	bp->rx_mem_zone = (const void *)mz;
4637 	bp->hw_rx_port_stats = mz->addr;
4638 	bp->hw_rx_port_stats_map = mz_phys_addr;
4639 
4640 	snprintf(mz_name, RTE_MEMZONE_NAMESIZE,
4641 		 "bnxt_" PCI_PRI_FMT "-%s", pci_dev->addr.domain,
4642 		 pci_dev->addr.bus, pci_dev->addr.devid,
4643 		 pci_dev->addr.function, "tx_port_stats");
4644 	mz_name[RTE_MEMZONE_NAMESIZE - 1] = 0;
4645 	mz = rte_memzone_lookup(mz_name);
4646 	total_alloc_len =
4647 		RTE_CACHE_LINE_ROUNDUP(sizeof(struct tx_port_stats) +
4648 				       sizeof(struct tx_port_stats_ext) + 512);
4649 	if (!mz) {
4650 		mz = rte_memzone_reserve(mz_name,
4651 					 total_alloc_len,
4652 					 SOCKET_ID_ANY,
4653 					 RTE_MEMZONE_2MB |
4654 					 RTE_MEMZONE_SIZE_HINT_ONLY |
4655 					 RTE_MEMZONE_IOVA_CONTIG);
4656 		if (mz == NULL)
4657 			return -ENOMEM;
4658 	}
4659 	memset(mz->addr, 0, mz->len);
4660 	mz_phys_addr = mz->iova;
4661 
4662 	bp->tx_mem_zone = (const void *)mz;
4663 	bp->hw_tx_port_stats = mz->addr;
4664 	bp->hw_tx_port_stats_map = mz_phys_addr;
4665 	bp->flags |= BNXT_FLAG_PORT_STATS;
4666 
4667 	/* Display extended statistics if FW supports it */
4668 	if (bp->hwrm_spec_code < HWRM_SPEC_CODE_1_8_4 ||
4669 	    bp->hwrm_spec_code == HWRM_SPEC_CODE_1_9_0 ||
4670 	    !(bp->flags & BNXT_FLAG_EXT_STATS_SUPPORTED))
4671 		return 0;
4672 
4673 	bp->hw_rx_port_stats_ext = (void *)
4674 		((uint8_t *)bp->hw_rx_port_stats +
4675 		 sizeof(struct rx_port_stats));
4676 	bp->hw_rx_port_stats_ext_map = bp->hw_rx_port_stats_map +
4677 		sizeof(struct rx_port_stats);
4678 	bp->flags |= BNXT_FLAG_EXT_RX_PORT_STATS;
4679 
4680 	if (bp->hwrm_spec_code < HWRM_SPEC_CODE_1_9_2 ||
4681 	    bp->flags & BNXT_FLAG_EXT_STATS_SUPPORTED) {
4682 		bp->hw_tx_port_stats_ext = (void *)
4683 			((uint8_t *)bp->hw_tx_port_stats +
4684 			 sizeof(struct tx_port_stats));
4685 		bp->hw_tx_port_stats_ext_map =
4686 			bp->hw_tx_port_stats_map +
4687 			sizeof(struct tx_port_stats);
4688 		bp->flags |= BNXT_FLAG_EXT_TX_PORT_STATS;
4689 	}
4690 
4691 	return 0;
4692 }
4693 
4694 static int bnxt_setup_mac_addr(struct rte_eth_dev *eth_dev)
4695 {
4696 	struct bnxt *bp = eth_dev->data->dev_private;
4697 	int rc = 0;
4698 
4699 	eth_dev->data->mac_addrs = rte_zmalloc("bnxt_mac_addr_tbl",
4700 					       RTE_ETHER_ADDR_LEN *
4701 					       bp->max_l2_ctx,
4702 					       0);
4703 	if (eth_dev->data->mac_addrs == NULL) {
4704 		PMD_DRV_LOG(ERR, "Failed to alloc MAC addr tbl\n");
4705 		return -ENOMEM;
4706 	}
4707 
4708 	if (!BNXT_HAS_DFLT_MAC_SET(bp)) {
4709 		if (BNXT_PF(bp))
4710 			return -EINVAL;
4711 
4712 		/* Generate a random MAC address, if none was assigned by PF */
4713 		PMD_DRV_LOG(INFO, "VF MAC address not assigned by Host PF\n");
4714 		bnxt_eth_hw_addr_random(bp->mac_addr);
4715 		PMD_DRV_LOG(INFO,
4716 			    "Assign random MAC:%02X:%02X:%02X:%02X:%02X:%02X\n",
4717 			    bp->mac_addr[0], bp->mac_addr[1], bp->mac_addr[2],
4718 			    bp->mac_addr[3], bp->mac_addr[4], bp->mac_addr[5]);
4719 
4720 		rc = bnxt_hwrm_set_mac(bp);
4721 		if (rc)
4722 			return rc;
4723 	}
4724 
4725 	/* Copy the permanent MAC from the FUNC_QCAPS response */
4726 	memcpy(&eth_dev->data->mac_addrs[0], bp->mac_addr, RTE_ETHER_ADDR_LEN);
4727 
4728 	return rc;
4729 }
4730 
4731 static int bnxt_restore_dflt_mac(struct bnxt *bp)
4732 {
4733 	int rc = 0;
4734 
4735 	/* MAC is already configured in FW */
4736 	if (BNXT_HAS_DFLT_MAC_SET(bp))
4737 		return 0;
4738 
4739 	/* Restore the old MAC configured */
4740 	rc = bnxt_hwrm_set_mac(bp);
4741 	if (rc)
4742 		PMD_DRV_LOG(ERR, "Failed to restore MAC address\n");
4743 
4744 	return rc;
4745 }
4746 
4747 static void bnxt_config_vf_req_fwd(struct bnxt *bp)
4748 {
4749 	if (!BNXT_PF(bp))
4750 		return;
4751 
4752 	memset(bp->pf->vf_req_fwd, 0, sizeof(bp->pf->vf_req_fwd));
4753 
4754 	if (!(bp->fw_cap & BNXT_FW_CAP_LINK_ADMIN))
4755 		BNXT_HWRM_CMD_TO_FORWARD(HWRM_PORT_PHY_QCFG);
4756 	BNXT_HWRM_CMD_TO_FORWARD(HWRM_FUNC_CFG);
4757 	BNXT_HWRM_CMD_TO_FORWARD(HWRM_FUNC_VF_CFG);
4758 	BNXT_HWRM_CMD_TO_FORWARD(HWRM_CFA_L2_FILTER_ALLOC);
4759 	BNXT_HWRM_CMD_TO_FORWARD(HWRM_OEM_CMD);
4760 }
4761 
4762 uint16_t
4763 bnxt_get_svif(uint16_t port_id, bool func_svif,
4764 	      enum bnxt_ulp_intf_type type)
4765 {
4766 	struct rte_eth_dev *eth_dev;
4767 	struct bnxt *bp;
4768 
4769 	eth_dev = &rte_eth_devices[port_id];
4770 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
4771 		struct bnxt_representor *vfr = eth_dev->data->dev_private;
4772 		if (!vfr)
4773 			return 0;
4774 
4775 		if (type == BNXT_ULP_INTF_TYPE_VF_REP)
4776 			return vfr->svif;
4777 
4778 		eth_dev = vfr->parent_dev;
4779 	}
4780 
4781 	bp = eth_dev->data->dev_private;
4782 
4783 	return func_svif ? bp->func_svif : bp->port_svif;
4784 }
4785 
4786 uint16_t
4787 bnxt_get_vnic_id(uint16_t port, enum bnxt_ulp_intf_type type)
4788 {
4789 	struct rte_eth_dev *eth_dev;
4790 	struct bnxt_vnic_info *vnic;
4791 	struct bnxt *bp;
4792 
4793 	eth_dev = &rte_eth_devices[port];
4794 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
4795 		struct bnxt_representor *vfr = eth_dev->data->dev_private;
4796 		if (!vfr)
4797 			return 0;
4798 
4799 		if (type == BNXT_ULP_INTF_TYPE_VF_REP)
4800 			return vfr->dflt_vnic_id;
4801 
4802 		eth_dev = vfr->parent_dev;
4803 	}
4804 
4805 	bp = eth_dev->data->dev_private;
4806 
4807 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
4808 
4809 	return vnic->fw_vnic_id;
4810 }
4811 
4812 uint16_t
4813 bnxt_get_fw_func_id(uint16_t port, enum bnxt_ulp_intf_type type)
4814 {
4815 	struct rte_eth_dev *eth_dev;
4816 	struct bnxt *bp;
4817 
4818 	eth_dev = &rte_eth_devices[port];
4819 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
4820 		struct bnxt_representor *vfr = eth_dev->data->dev_private;
4821 		if (!vfr)
4822 			return 0;
4823 
4824 		if (type == BNXT_ULP_INTF_TYPE_VF_REP)
4825 			return vfr->fw_fid;
4826 
4827 		eth_dev = vfr->parent_dev;
4828 	}
4829 
4830 	bp = eth_dev->data->dev_private;
4831 
4832 	return bp->fw_fid;
4833 }
4834 
4835 enum bnxt_ulp_intf_type
4836 bnxt_get_interface_type(uint16_t port)
4837 {
4838 	struct rte_eth_dev *eth_dev;
4839 	struct bnxt *bp;
4840 
4841 	eth_dev = &rte_eth_devices[port];
4842 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev))
4843 		return BNXT_ULP_INTF_TYPE_VF_REP;
4844 
4845 	bp = eth_dev->data->dev_private;
4846 	if (BNXT_PF(bp))
4847 		return BNXT_ULP_INTF_TYPE_PF;
4848 	else if (BNXT_VF_IS_TRUSTED(bp))
4849 		return BNXT_ULP_INTF_TYPE_TRUSTED_VF;
4850 	else if (BNXT_VF(bp))
4851 		return BNXT_ULP_INTF_TYPE_VF;
4852 
4853 	return BNXT_ULP_INTF_TYPE_INVALID;
4854 }
4855 
4856 uint16_t
4857 bnxt_get_phy_port_id(uint16_t port_id)
4858 {
4859 	struct bnxt_representor *vfr;
4860 	struct rte_eth_dev *eth_dev;
4861 	struct bnxt *bp;
4862 
4863 	eth_dev = &rte_eth_devices[port_id];
4864 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
4865 		vfr = eth_dev->data->dev_private;
4866 		if (!vfr)
4867 			return 0;
4868 
4869 		eth_dev = vfr->parent_dev;
4870 	}
4871 
4872 	bp = eth_dev->data->dev_private;
4873 
4874 	return BNXT_PF(bp) ? bp->pf->port_id : bp->parent->port_id;
4875 }
4876 
4877 uint16_t
4878 bnxt_get_parif(uint16_t port_id, enum bnxt_ulp_intf_type type)
4879 {
4880 	struct rte_eth_dev *eth_dev;
4881 	struct bnxt *bp;
4882 
4883 	eth_dev = &rte_eth_devices[port_id];
4884 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
4885 		struct bnxt_representor *vfr = eth_dev->data->dev_private;
4886 		if (!vfr)
4887 			return 0;
4888 
4889 		if (type == BNXT_ULP_INTF_TYPE_VF_REP)
4890 			return vfr->fw_fid - 1;
4891 
4892 		eth_dev = vfr->parent_dev;
4893 	}
4894 
4895 	bp = eth_dev->data->dev_private;
4896 
4897 	return BNXT_PF(bp) ? bp->fw_fid - 1 : bp->parent->fid - 1;
4898 }
4899 
4900 uint16_t
4901 bnxt_get_vport(uint16_t port_id)
4902 {
4903 	return (1 << bnxt_get_phy_port_id(port_id));
4904 }
4905 
4906 static void bnxt_alloc_error_recovery_info(struct bnxt *bp)
4907 {
4908 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4909 
4910 	if (info) {
4911 		if (!(bp->fw_cap & BNXT_FW_CAP_HCOMM_FW_STATUS))
4912 			memset(info, 0, sizeof(*info));
4913 		return;
4914 	}
4915 
4916 	if (!(bp->fw_cap & BNXT_FW_CAP_ERROR_RECOVERY))
4917 		return;
4918 
4919 	info = rte_zmalloc("bnxt_hwrm_error_recovery_qcfg",
4920 			   sizeof(*info), 0);
4921 	if (!info)
4922 		bp->fw_cap &= ~BNXT_FW_CAP_ERROR_RECOVERY;
4923 
4924 	bp->recovery_info = info;
4925 }
4926 
4927 static void bnxt_check_fw_status(struct bnxt *bp)
4928 {
4929 	uint32_t fw_status;
4930 
4931 	if (!(bp->recovery_info &&
4932 	      (bp->fw_cap & BNXT_FW_CAP_HCOMM_FW_STATUS)))
4933 		return;
4934 
4935 	fw_status = bnxt_read_fw_status_reg(bp, BNXT_FW_STATUS_REG);
4936 	if (fw_status != BNXT_FW_STATUS_HEALTHY)
4937 		PMD_DRV_LOG(ERR, "Firmware not responding, status: %#x\n",
4938 			    fw_status);
4939 }
4940 
4941 static int bnxt_map_hcomm_fw_status_reg(struct bnxt *bp)
4942 {
4943 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4944 	uint32_t status_loc;
4945 	uint32_t sig_ver;
4946 
4947 	rte_write32(HCOMM_STATUS_STRUCT_LOC, (uint8_t *)bp->bar0 +
4948 		    BNXT_GRCPF_REG_WINDOW_BASE_OUT + 4);
4949 	sig_ver = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
4950 				   BNXT_GRCP_WINDOW_2_BASE +
4951 				   offsetof(struct hcomm_status,
4952 					    sig_ver)));
4953 	/* If the signature is absent, then FW does not support this feature */
4954 	if ((sig_ver & HCOMM_STATUS_SIGNATURE_MASK) !=
4955 	    HCOMM_STATUS_SIGNATURE_VAL)
4956 		return 0;
4957 
4958 	if (!info) {
4959 		info = rte_zmalloc("bnxt_hwrm_error_recovery_qcfg",
4960 				   sizeof(*info), 0);
4961 		if (!info)
4962 			return -ENOMEM;
4963 		bp->recovery_info = info;
4964 	} else {
4965 		memset(info, 0, sizeof(*info));
4966 	}
4967 
4968 	status_loc = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
4969 				      BNXT_GRCP_WINDOW_2_BASE +
4970 				      offsetof(struct hcomm_status,
4971 					       fw_status_loc)));
4972 
4973 	/* Only pre-map the FW health status GRC register */
4974 	if (BNXT_FW_STATUS_REG_TYPE(status_loc) != BNXT_FW_STATUS_REG_TYPE_GRC)
4975 		return 0;
4976 
4977 	info->status_regs[BNXT_FW_STATUS_REG] = status_loc;
4978 	info->mapped_status_regs[BNXT_FW_STATUS_REG] =
4979 		BNXT_GRCP_WINDOW_2_BASE + (status_loc & BNXT_GRCP_OFFSET_MASK);
4980 
4981 	rte_write32((status_loc & BNXT_GRCP_BASE_MASK), (uint8_t *)bp->bar0 +
4982 		    BNXT_GRCPF_REG_WINDOW_BASE_OUT + 4);
4983 
4984 	bp->fw_cap |= BNXT_FW_CAP_HCOMM_FW_STATUS;
4985 
4986 	return 0;
4987 }
4988 
4989 /* This function gets the FW version along with the
4990  * capabilities(MAX and current) of the function, vnic,
4991  * error recovery, phy and other chip related info
4992  */
4993 static int bnxt_get_config(struct bnxt *bp)
4994 {
4995 	uint16_t mtu;
4996 	int rc = 0;
4997 
4998 	bp->fw_cap = 0;
4999 
5000 	rc = bnxt_map_hcomm_fw_status_reg(bp);
5001 	if (rc)
5002 		return rc;
5003 
5004 	rc = bnxt_hwrm_ver_get(bp, DFLT_HWRM_CMD_TIMEOUT);
5005 	if (rc) {
5006 		bnxt_check_fw_status(bp);
5007 		return rc;
5008 	}
5009 
5010 	rc = bnxt_hwrm_func_reset(bp);
5011 	if (rc)
5012 		return -EIO;
5013 
5014 	rc = bnxt_hwrm_vnic_qcaps(bp);
5015 	if (rc)
5016 		return rc;
5017 
5018 	rc = bnxt_hwrm_queue_qportcfg(bp);
5019 	if (rc)
5020 		return rc;
5021 
5022 	/* Get the MAX capabilities for this function.
5023 	 * This function also allocates context memory for TQM rings and
5024 	 * informs the firmware about this allocated backing store memory.
5025 	 */
5026 	rc = bnxt_hwrm_func_qcaps(bp);
5027 	if (rc)
5028 		return rc;
5029 
5030 	rc = bnxt_hwrm_func_qcfg(bp, &mtu);
5031 	if (rc)
5032 		return rc;
5033 
5034 	rc = bnxt_hwrm_cfa_adv_flow_mgmt_qcaps(bp);
5035 	if (rc)
5036 		return rc;
5037 
5038 	bnxt_hwrm_port_mac_qcfg(bp);
5039 
5040 	bnxt_hwrm_parent_pf_qcfg(bp);
5041 
5042 	bnxt_hwrm_port_phy_qcaps(bp);
5043 
5044 	bnxt_alloc_error_recovery_info(bp);
5045 	/* Get the adapter error recovery support info */
5046 	rc = bnxt_hwrm_error_recovery_qcfg(bp);
5047 	if (rc)
5048 		bp->fw_cap &= ~BNXT_FW_CAP_ERROR_RECOVERY;
5049 
5050 	bnxt_hwrm_port_led_qcaps(bp);
5051 
5052 	return 0;
5053 }
5054 
5055 static int
5056 bnxt_init_locks(struct bnxt *bp)
5057 {
5058 	int err;
5059 
5060 	err = pthread_mutex_init(&bp->flow_lock, NULL);
5061 	if (err) {
5062 		PMD_DRV_LOG(ERR, "Unable to initialize flow_lock\n");
5063 		return err;
5064 	}
5065 
5066 	err = pthread_mutex_init(&bp->def_cp_lock, NULL);
5067 	if (err) {
5068 		PMD_DRV_LOG(ERR, "Unable to initialize def_cp_lock\n");
5069 		return err;
5070 	}
5071 
5072 	err = pthread_mutex_init(&bp->health_check_lock, NULL);
5073 	if (err) {
5074 		PMD_DRV_LOG(ERR, "Unable to initialize health_check_lock\n");
5075 		return err;
5076 	}
5077 
5078 	err = pthread_mutex_init(&bp->err_recovery_lock, NULL);
5079 	if (err)
5080 		PMD_DRV_LOG(ERR, "Unable to initialize err_recovery_lock\n");
5081 
5082 	return err;
5083 }
5084 
5085 static int bnxt_init_resources(struct bnxt *bp, bool reconfig_dev)
5086 {
5087 	int rc = 0;
5088 
5089 	rc = bnxt_get_config(bp);
5090 	if (rc)
5091 		return rc;
5092 
5093 	if (!reconfig_dev) {
5094 		rc = bnxt_setup_mac_addr(bp->eth_dev);
5095 		if (rc)
5096 			return rc;
5097 	} else {
5098 		rc = bnxt_restore_dflt_mac(bp);
5099 		if (rc)
5100 			return rc;
5101 	}
5102 
5103 	bnxt_config_vf_req_fwd(bp);
5104 
5105 	rc = bnxt_hwrm_func_driver_register(bp);
5106 	if (rc) {
5107 		PMD_DRV_LOG(ERR, "Failed to register driver");
5108 		return -EBUSY;
5109 	}
5110 
5111 	if (BNXT_PF(bp)) {
5112 		if (bp->pdev->max_vfs) {
5113 			rc = bnxt_hwrm_allocate_vfs(bp, bp->pdev->max_vfs);
5114 			if (rc) {
5115 				PMD_DRV_LOG(ERR, "Failed to allocate VFs\n");
5116 				return rc;
5117 			}
5118 		} else {
5119 			rc = bnxt_hwrm_allocate_pf_only(bp);
5120 			if (rc) {
5121 				PMD_DRV_LOG(ERR,
5122 					    "Failed to allocate PF resources");
5123 				return rc;
5124 			}
5125 		}
5126 	}
5127 
5128 	rc = bnxt_alloc_mem(bp, reconfig_dev);
5129 	if (rc)
5130 		return rc;
5131 
5132 	rc = bnxt_setup_int(bp);
5133 	if (rc)
5134 		return rc;
5135 
5136 	rc = bnxt_request_int(bp);
5137 	if (rc)
5138 		return rc;
5139 
5140 	rc = bnxt_init_ctx_mem(bp);
5141 	if (rc) {
5142 		PMD_DRV_LOG(ERR, "Failed to init adv_flow_counters\n");
5143 		return rc;
5144 	}
5145 
5146 	return 0;
5147 }
5148 
5149 static int
5150 bnxt_parse_devarg_truflow(__rte_unused const char *key,
5151 			  const char *value, void *opaque_arg)
5152 {
5153 	struct bnxt *bp = opaque_arg;
5154 	unsigned long truflow;
5155 	char *end = NULL;
5156 
5157 	if (!value || !opaque_arg) {
5158 		PMD_DRV_LOG(ERR,
5159 			    "Invalid parameter passed to truflow devargs.\n");
5160 		return -EINVAL;
5161 	}
5162 
5163 	truflow = strtoul(value, &end, 10);
5164 	if (end == NULL || *end != '\0' ||
5165 	    (truflow == ULONG_MAX && errno == ERANGE)) {
5166 		PMD_DRV_LOG(ERR,
5167 			    "Invalid parameter passed to truflow devargs.\n");
5168 		return -EINVAL;
5169 	}
5170 
5171 	if (BNXT_DEVARG_TRUFLOW_INVALID(truflow)) {
5172 		PMD_DRV_LOG(ERR,
5173 			    "Invalid value passed to truflow devargs.\n");
5174 		return -EINVAL;
5175 	}
5176 
5177 	if (truflow) {
5178 		bp->flags |= BNXT_FLAG_TRUFLOW_EN;
5179 		PMD_DRV_LOG(INFO, "Host-based truflow feature enabled.\n");
5180 	} else {
5181 		bp->flags &= ~BNXT_FLAG_TRUFLOW_EN;
5182 		PMD_DRV_LOG(INFO, "Host-based truflow feature disabled.\n");
5183 	}
5184 
5185 	return 0;
5186 }
5187 
5188 static int
5189 bnxt_parse_devarg_flow_xstat(__rte_unused const char *key,
5190 			     const char *value, void *opaque_arg)
5191 {
5192 	struct bnxt *bp = opaque_arg;
5193 	unsigned long flow_xstat;
5194 	char *end = NULL;
5195 
5196 	if (!value || !opaque_arg) {
5197 		PMD_DRV_LOG(ERR,
5198 			    "Invalid parameter passed to flow_xstat devarg.\n");
5199 		return -EINVAL;
5200 	}
5201 
5202 	flow_xstat = strtoul(value, &end, 10);
5203 	if (end == NULL || *end != '\0' ||
5204 	    (flow_xstat == ULONG_MAX && errno == ERANGE)) {
5205 		PMD_DRV_LOG(ERR,
5206 			    "Invalid parameter passed to flow_xstat devarg.\n");
5207 		return -EINVAL;
5208 	}
5209 
5210 	if (BNXT_DEVARG_FLOW_XSTAT_INVALID(flow_xstat)) {
5211 		PMD_DRV_LOG(ERR,
5212 			    "Invalid value passed to flow_xstat devarg.\n");
5213 		return -EINVAL;
5214 	}
5215 
5216 	bp->flags |= BNXT_FLAG_FLOW_XSTATS_EN;
5217 	if (BNXT_FLOW_XSTATS_EN(bp))
5218 		PMD_DRV_LOG(INFO, "flow_xstat feature enabled.\n");
5219 
5220 	return 0;
5221 }
5222 
5223 static int
5224 bnxt_parse_devarg_max_num_kflows(__rte_unused const char *key,
5225 					const char *value, void *opaque_arg)
5226 {
5227 	struct bnxt *bp = opaque_arg;
5228 	unsigned long max_num_kflows;
5229 	char *end = NULL;
5230 
5231 	if (!value || !opaque_arg) {
5232 		PMD_DRV_LOG(ERR,
5233 			"Invalid parameter passed to max_num_kflows devarg.\n");
5234 		return -EINVAL;
5235 	}
5236 
5237 	max_num_kflows = strtoul(value, &end, 10);
5238 	if (end == NULL || *end != '\0' ||
5239 		(max_num_kflows == ULONG_MAX && errno == ERANGE)) {
5240 		PMD_DRV_LOG(ERR,
5241 			"Invalid parameter passed to max_num_kflows devarg.\n");
5242 		return -EINVAL;
5243 	}
5244 
5245 	if (bnxt_devarg_max_num_kflow_invalid(max_num_kflows)) {
5246 		PMD_DRV_LOG(ERR,
5247 			"Invalid value passed to max_num_kflows devarg.\n");
5248 		return -EINVAL;
5249 	}
5250 
5251 	bp->max_num_kflows = max_num_kflows;
5252 	if (bp->max_num_kflows)
5253 		PMD_DRV_LOG(INFO, "max_num_kflows set as %ldK.\n",
5254 				max_num_kflows);
5255 
5256 	return 0;
5257 }
5258 
5259 static int
5260 bnxt_parse_devarg_rep_is_pf(__rte_unused const char *key,
5261 			    const char *value, void *opaque_arg)
5262 {
5263 	struct bnxt_representor *vfr_bp = opaque_arg;
5264 	unsigned long rep_is_pf;
5265 	char *end = NULL;
5266 
5267 	if (!value || !opaque_arg) {
5268 		PMD_DRV_LOG(ERR,
5269 			    "Invalid parameter passed to rep_is_pf devargs.\n");
5270 		return -EINVAL;
5271 	}
5272 
5273 	rep_is_pf = strtoul(value, &end, 10);
5274 	if (end == NULL || *end != '\0' ||
5275 	    (rep_is_pf == ULONG_MAX && errno == ERANGE)) {
5276 		PMD_DRV_LOG(ERR,
5277 			    "Invalid parameter passed to rep_is_pf devargs.\n");
5278 		return -EINVAL;
5279 	}
5280 
5281 	if (BNXT_DEVARG_REP_IS_PF_INVALID(rep_is_pf)) {
5282 		PMD_DRV_LOG(ERR,
5283 			    "Invalid value passed to rep_is_pf devargs.\n");
5284 		return -EINVAL;
5285 	}
5286 
5287 	vfr_bp->flags |= rep_is_pf;
5288 	if (BNXT_REP_PF(vfr_bp))
5289 		PMD_DRV_LOG(INFO, "PF representor\n");
5290 	else
5291 		PMD_DRV_LOG(INFO, "VF representor\n");
5292 
5293 	return 0;
5294 }
5295 
5296 static int
5297 bnxt_parse_devarg_rep_based_pf(__rte_unused const char *key,
5298 			       const char *value, void *opaque_arg)
5299 {
5300 	struct bnxt_representor *vfr_bp = opaque_arg;
5301 	unsigned long rep_based_pf;
5302 	char *end = NULL;
5303 
5304 	if (!value || !opaque_arg) {
5305 		PMD_DRV_LOG(ERR,
5306 			    "Invalid parameter passed to rep_based_pf "
5307 			    "devargs.\n");
5308 		return -EINVAL;
5309 	}
5310 
5311 	rep_based_pf = strtoul(value, &end, 10);
5312 	if (end == NULL || *end != '\0' ||
5313 	    (rep_based_pf == ULONG_MAX && errno == ERANGE)) {
5314 		PMD_DRV_LOG(ERR,
5315 			    "Invalid parameter passed to rep_based_pf "
5316 			    "devargs.\n");
5317 		return -EINVAL;
5318 	}
5319 
5320 	if (BNXT_DEVARG_REP_BASED_PF_INVALID(rep_based_pf)) {
5321 		PMD_DRV_LOG(ERR,
5322 			    "Invalid value passed to rep_based_pf devargs.\n");
5323 		return -EINVAL;
5324 	}
5325 
5326 	vfr_bp->rep_based_pf = rep_based_pf;
5327 	vfr_bp->flags |= BNXT_REP_BASED_PF_VALID;
5328 
5329 	PMD_DRV_LOG(INFO, "rep-based-pf = %d\n", vfr_bp->rep_based_pf);
5330 
5331 	return 0;
5332 }
5333 
5334 static int
5335 bnxt_parse_devarg_rep_q_r2f(__rte_unused const char *key,
5336 			    const char *value, void *opaque_arg)
5337 {
5338 	struct bnxt_representor *vfr_bp = opaque_arg;
5339 	unsigned long rep_q_r2f;
5340 	char *end = NULL;
5341 
5342 	if (!value || !opaque_arg) {
5343 		PMD_DRV_LOG(ERR,
5344 			    "Invalid parameter passed to rep_q_r2f "
5345 			    "devargs.\n");
5346 		return -EINVAL;
5347 	}
5348 
5349 	rep_q_r2f = strtoul(value, &end, 10);
5350 	if (end == NULL || *end != '\0' ||
5351 	    (rep_q_r2f == ULONG_MAX && errno == ERANGE)) {
5352 		PMD_DRV_LOG(ERR,
5353 			    "Invalid parameter passed to rep_q_r2f "
5354 			    "devargs.\n");
5355 		return -EINVAL;
5356 	}
5357 
5358 	if (BNXT_DEVARG_REP_Q_R2F_INVALID(rep_q_r2f)) {
5359 		PMD_DRV_LOG(ERR,
5360 			    "Invalid value passed to rep_q_r2f devargs.\n");
5361 		return -EINVAL;
5362 	}
5363 
5364 	vfr_bp->rep_q_r2f = rep_q_r2f;
5365 	vfr_bp->flags |= BNXT_REP_Q_R2F_VALID;
5366 	PMD_DRV_LOG(INFO, "rep-q-r2f = %d\n", vfr_bp->rep_q_r2f);
5367 
5368 	return 0;
5369 }
5370 
5371 static int
5372 bnxt_parse_devarg_rep_q_f2r(__rte_unused const char *key,
5373 			    const char *value, void *opaque_arg)
5374 {
5375 	struct bnxt_representor *vfr_bp = opaque_arg;
5376 	unsigned long rep_q_f2r;
5377 	char *end = NULL;
5378 
5379 	if (!value || !opaque_arg) {
5380 		PMD_DRV_LOG(ERR,
5381 			    "Invalid parameter passed to rep_q_f2r "
5382 			    "devargs.\n");
5383 		return -EINVAL;
5384 	}
5385 
5386 	rep_q_f2r = strtoul(value, &end, 10);
5387 	if (end == NULL || *end != '\0' ||
5388 	    (rep_q_f2r == ULONG_MAX && errno == ERANGE)) {
5389 		PMD_DRV_LOG(ERR,
5390 			    "Invalid parameter passed to rep_q_f2r "
5391 			    "devargs.\n");
5392 		return -EINVAL;
5393 	}
5394 
5395 	if (BNXT_DEVARG_REP_Q_F2R_INVALID(rep_q_f2r)) {
5396 		PMD_DRV_LOG(ERR,
5397 			    "Invalid value passed to rep_q_f2r devargs.\n");
5398 		return -EINVAL;
5399 	}
5400 
5401 	vfr_bp->rep_q_f2r = rep_q_f2r;
5402 	vfr_bp->flags |= BNXT_REP_Q_F2R_VALID;
5403 	PMD_DRV_LOG(INFO, "rep-q-f2r = %d\n", vfr_bp->rep_q_f2r);
5404 
5405 	return 0;
5406 }
5407 
5408 static int
5409 bnxt_parse_devarg_rep_fc_r2f(__rte_unused const char *key,
5410 			     const char *value, void *opaque_arg)
5411 {
5412 	struct bnxt_representor *vfr_bp = opaque_arg;
5413 	unsigned long rep_fc_r2f;
5414 	char *end = NULL;
5415 
5416 	if (!value || !opaque_arg) {
5417 		PMD_DRV_LOG(ERR,
5418 			    "Invalid parameter passed to rep_fc_r2f "
5419 			    "devargs.\n");
5420 		return -EINVAL;
5421 	}
5422 
5423 	rep_fc_r2f = strtoul(value, &end, 10);
5424 	if (end == NULL || *end != '\0' ||
5425 	    (rep_fc_r2f == ULONG_MAX && errno == ERANGE)) {
5426 		PMD_DRV_LOG(ERR,
5427 			    "Invalid parameter passed to rep_fc_r2f "
5428 			    "devargs.\n");
5429 		return -EINVAL;
5430 	}
5431 
5432 	if (BNXT_DEVARG_REP_FC_R2F_INVALID(rep_fc_r2f)) {
5433 		PMD_DRV_LOG(ERR,
5434 			    "Invalid value passed to rep_fc_r2f devargs.\n");
5435 		return -EINVAL;
5436 	}
5437 
5438 	vfr_bp->flags |= BNXT_REP_FC_R2F_VALID;
5439 	vfr_bp->rep_fc_r2f = rep_fc_r2f;
5440 	PMD_DRV_LOG(INFO, "rep-fc-r2f = %lu\n", rep_fc_r2f);
5441 
5442 	return 0;
5443 }
5444 
5445 static int
5446 bnxt_parse_devarg_rep_fc_f2r(__rte_unused const char *key,
5447 			     const char *value, void *opaque_arg)
5448 {
5449 	struct bnxt_representor *vfr_bp = opaque_arg;
5450 	unsigned long rep_fc_f2r;
5451 	char *end = NULL;
5452 
5453 	if (!value || !opaque_arg) {
5454 		PMD_DRV_LOG(ERR,
5455 			    "Invalid parameter passed to rep_fc_f2r "
5456 			    "devargs.\n");
5457 		return -EINVAL;
5458 	}
5459 
5460 	rep_fc_f2r = strtoul(value, &end, 10);
5461 	if (end == NULL || *end != '\0' ||
5462 	    (rep_fc_f2r == ULONG_MAX && errno == ERANGE)) {
5463 		PMD_DRV_LOG(ERR,
5464 			    "Invalid parameter passed to rep_fc_f2r "
5465 			    "devargs.\n");
5466 		return -EINVAL;
5467 	}
5468 
5469 	if (BNXT_DEVARG_REP_FC_F2R_INVALID(rep_fc_f2r)) {
5470 		PMD_DRV_LOG(ERR,
5471 			    "Invalid value passed to rep_fc_f2r devargs.\n");
5472 		return -EINVAL;
5473 	}
5474 
5475 	vfr_bp->flags |= BNXT_REP_FC_F2R_VALID;
5476 	vfr_bp->rep_fc_f2r = rep_fc_f2r;
5477 	PMD_DRV_LOG(INFO, "rep-fc-f2r = %lu\n", rep_fc_f2r);
5478 
5479 	return 0;
5480 }
5481 
5482 static int
5483 bnxt_parse_dev_args(struct bnxt *bp, struct rte_devargs *devargs)
5484 {
5485 	struct rte_kvargs *kvlist;
5486 	int ret;
5487 
5488 	if (devargs == NULL)
5489 		return 0;
5490 
5491 	kvlist = rte_kvargs_parse(devargs->args, bnxt_dev_args);
5492 	if (kvlist == NULL)
5493 		return -EINVAL;
5494 
5495 	/*
5496 	 * Handler for "truflow" devarg.
5497 	 * Invoked as for ex: "-a 0000:00:0d.0,host-based-truflow=1"
5498 	 */
5499 	ret = rte_kvargs_process(kvlist, BNXT_DEVARG_TRUFLOW,
5500 				 bnxt_parse_devarg_truflow, bp);
5501 	if (ret)
5502 		goto err;
5503 
5504 	/*
5505 	 * Handler for "flow_xstat" devarg.
5506 	 * Invoked as for ex: "-a 0000:00:0d.0,flow_xstat=1"
5507 	 */
5508 	ret = rte_kvargs_process(kvlist, BNXT_DEVARG_FLOW_XSTAT,
5509 				 bnxt_parse_devarg_flow_xstat, bp);
5510 	if (ret)
5511 		goto err;
5512 
5513 	/*
5514 	 * Handler for "max_num_kflows" devarg.
5515 	 * Invoked as for ex: "-a 000:00:0d.0,max_num_kflows=32"
5516 	 */
5517 	ret = rte_kvargs_process(kvlist, BNXT_DEVARG_MAX_NUM_KFLOWS,
5518 				 bnxt_parse_devarg_max_num_kflows, bp);
5519 	if (ret)
5520 		goto err;
5521 
5522 err:
5523 	rte_kvargs_free(kvlist);
5524 	return ret;
5525 }
5526 
5527 static int bnxt_alloc_switch_domain(struct bnxt *bp)
5528 {
5529 	int rc = 0;
5530 
5531 	if (BNXT_PF(bp) || BNXT_VF_IS_TRUSTED(bp)) {
5532 		rc = rte_eth_switch_domain_alloc(&bp->switch_domain_id);
5533 		if (rc)
5534 			PMD_DRV_LOG(ERR,
5535 				    "Failed to alloc switch domain: %d\n", rc);
5536 		else
5537 			PMD_DRV_LOG(INFO,
5538 				    "Switch domain allocated %d\n",
5539 				    bp->switch_domain_id);
5540 	}
5541 
5542 	return rc;
5543 }
5544 
5545 /* Allocate and initialize various fields in bnxt struct that
5546  * need to be allocated/destroyed only once in the lifetime of the driver
5547  */
5548 static int bnxt_drv_init(struct rte_eth_dev *eth_dev)
5549 {
5550 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
5551 	struct bnxt *bp = eth_dev->data->dev_private;
5552 	int rc = 0;
5553 
5554 	bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
5555 
5556 	if (bnxt_vf_pciid(pci_dev->id.device_id))
5557 		bp->flags |= BNXT_FLAG_VF;
5558 
5559 	if (bnxt_p5_device(pci_dev->id.device_id))
5560 		bp->flags |= BNXT_FLAG_CHIP_P5;
5561 
5562 	if (pci_dev->id.device_id == BROADCOM_DEV_ID_58802 ||
5563 	    pci_dev->id.device_id == BROADCOM_DEV_ID_58804 ||
5564 	    pci_dev->id.device_id == BROADCOM_DEV_ID_58808 ||
5565 	    pci_dev->id.device_id == BROADCOM_DEV_ID_58802_VF)
5566 		bp->flags |= BNXT_FLAG_STINGRAY;
5567 
5568 	if (BNXT_TRUFLOW_EN(bp)) {
5569 		/* extra mbuf field is required to store CFA code from mark */
5570 		static const struct rte_mbuf_dynfield bnxt_cfa_code_dynfield_desc = {
5571 			.name = RTE_PMD_BNXT_CFA_CODE_DYNFIELD_NAME,
5572 			.size = sizeof(bnxt_cfa_code_dynfield_t),
5573 			.align = __alignof__(bnxt_cfa_code_dynfield_t),
5574 		};
5575 		bnxt_cfa_code_dynfield_offset =
5576 			rte_mbuf_dynfield_register(&bnxt_cfa_code_dynfield_desc);
5577 		if (bnxt_cfa_code_dynfield_offset < 0) {
5578 			PMD_DRV_LOG(ERR,
5579 			    "Failed to register mbuf field for TruFlow mark\n");
5580 			return -rte_errno;
5581 		}
5582 	}
5583 
5584 	rc = bnxt_map_pci_bars(eth_dev);
5585 	if (rc) {
5586 		PMD_DRV_LOG(ERR,
5587 			    "Failed to initialize board rc: %x\n", rc);
5588 		return rc;
5589 	}
5590 
5591 	rc = bnxt_alloc_pf_info(bp);
5592 	if (rc)
5593 		return rc;
5594 
5595 	rc = bnxt_alloc_link_info(bp);
5596 	if (rc)
5597 		return rc;
5598 
5599 	rc = bnxt_alloc_parent_info(bp);
5600 	if (rc)
5601 		return rc;
5602 
5603 	rc = bnxt_alloc_hwrm_resources(bp);
5604 	if (rc) {
5605 		PMD_DRV_LOG(ERR,
5606 			    "Failed to allocate response buffer rc: %x\n", rc);
5607 		return rc;
5608 	}
5609 	rc = bnxt_alloc_leds_info(bp);
5610 	if (rc)
5611 		return rc;
5612 
5613 	rc = bnxt_alloc_cos_queues(bp);
5614 	if (rc)
5615 		return rc;
5616 
5617 	rc = bnxt_init_locks(bp);
5618 	if (rc)
5619 		return rc;
5620 
5621 	rc = bnxt_alloc_switch_domain(bp);
5622 	if (rc)
5623 		return rc;
5624 
5625 	return rc;
5626 }
5627 
5628 static int
5629 bnxt_dev_init(struct rte_eth_dev *eth_dev, void *params __rte_unused)
5630 {
5631 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
5632 	static int version_printed;
5633 	struct bnxt *bp;
5634 	int rc;
5635 
5636 	if (version_printed++ == 0)
5637 		PMD_DRV_LOG(INFO, "%s\n", bnxt_version);
5638 
5639 	eth_dev->dev_ops = &bnxt_dev_ops;
5640 	eth_dev->rx_queue_count = bnxt_rx_queue_count_op;
5641 	eth_dev->rx_descriptor_status = bnxt_rx_descriptor_status_op;
5642 	eth_dev->tx_descriptor_status = bnxt_tx_descriptor_status_op;
5643 	eth_dev->rx_pkt_burst = &bnxt_recv_pkts;
5644 	eth_dev->tx_pkt_burst = &bnxt_xmit_pkts;
5645 
5646 	/*
5647 	 * For secondary processes, we don't initialise any further
5648 	 * as primary has already done this work.
5649 	 */
5650 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
5651 		return 0;
5652 
5653 	rte_eth_copy_pci_info(eth_dev, pci_dev);
5654 	eth_dev->data->dev_flags |= RTE_ETH_DEV_AUTOFILL_QUEUE_XSTATS;
5655 
5656 	bp = eth_dev->data->dev_private;
5657 
5658 	/* Parse dev arguments passed on when starting the DPDK application. */
5659 	rc = bnxt_parse_dev_args(bp, pci_dev->device.devargs);
5660 	if (rc)
5661 		goto error_free;
5662 
5663 	rc = bnxt_drv_init(eth_dev);
5664 	if (rc)
5665 		goto error_free;
5666 
5667 	rc = bnxt_init_resources(bp, false);
5668 	if (rc)
5669 		goto error_free;
5670 
5671 	rc = bnxt_alloc_stats_mem(bp);
5672 	if (rc)
5673 		goto error_free;
5674 
5675 	PMD_DRV_LOG(INFO,
5676 		    DRV_MODULE_NAME "found at mem %" PRIX64 ", node addr %pM\n",
5677 		    pci_dev->mem_resource[0].phys_addr,
5678 		    pci_dev->mem_resource[0].addr);
5679 
5680 	return 0;
5681 
5682 error_free:
5683 	bnxt_dev_uninit(eth_dev);
5684 	return rc;
5685 }
5686 
5687 
5688 static void bnxt_free_ctx_mem_buf(struct bnxt_ctx_mem_buf_info *ctx)
5689 {
5690 	if (!ctx)
5691 		return;
5692 
5693 	if (ctx->va)
5694 		rte_free(ctx->va);
5695 
5696 	ctx->va = NULL;
5697 	ctx->dma = RTE_BAD_IOVA;
5698 	ctx->ctx_id = BNXT_CTX_VAL_INVAL;
5699 }
5700 
5701 static void bnxt_unregister_fc_ctx_mem(struct bnxt *bp)
5702 {
5703 	bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_RX,
5704 				  CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
5705 				  bp->flow_stat->rx_fc_out_tbl.ctx_id,
5706 				  bp->flow_stat->max_fc,
5707 				  false);
5708 
5709 	bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_TX,
5710 				  CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
5711 				  bp->flow_stat->tx_fc_out_tbl.ctx_id,
5712 				  bp->flow_stat->max_fc,
5713 				  false);
5714 
5715 	if (bp->flow_stat->rx_fc_in_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
5716 		bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->rx_fc_in_tbl.ctx_id);
5717 	bp->flow_stat->rx_fc_in_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
5718 
5719 	if (bp->flow_stat->rx_fc_out_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
5720 		bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->rx_fc_out_tbl.ctx_id);
5721 	bp->flow_stat->rx_fc_out_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
5722 
5723 	if (bp->flow_stat->tx_fc_in_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
5724 		bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->tx_fc_in_tbl.ctx_id);
5725 	bp->flow_stat->tx_fc_in_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
5726 
5727 	if (bp->flow_stat->tx_fc_out_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
5728 		bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->tx_fc_out_tbl.ctx_id);
5729 	bp->flow_stat->tx_fc_out_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
5730 }
5731 
5732 static void bnxt_uninit_fc_ctx_mem(struct bnxt *bp)
5733 {
5734 	bnxt_unregister_fc_ctx_mem(bp);
5735 
5736 	bnxt_free_ctx_mem_buf(&bp->flow_stat->rx_fc_in_tbl);
5737 	bnxt_free_ctx_mem_buf(&bp->flow_stat->rx_fc_out_tbl);
5738 	bnxt_free_ctx_mem_buf(&bp->flow_stat->tx_fc_in_tbl);
5739 	bnxt_free_ctx_mem_buf(&bp->flow_stat->tx_fc_out_tbl);
5740 }
5741 
5742 static void bnxt_uninit_ctx_mem(struct bnxt *bp)
5743 {
5744 	if (BNXT_FLOW_XSTATS_EN(bp))
5745 		bnxt_uninit_fc_ctx_mem(bp);
5746 }
5747 
5748 static void
5749 bnxt_free_error_recovery_info(struct bnxt *bp)
5750 {
5751 	rte_free(bp->recovery_info);
5752 	bp->recovery_info = NULL;
5753 	bp->fw_cap &= ~BNXT_FW_CAP_ERROR_RECOVERY;
5754 }
5755 
5756 static int
5757 bnxt_uninit_resources(struct bnxt *bp, bool reconfig_dev)
5758 {
5759 	int rc;
5760 
5761 	bnxt_free_int(bp);
5762 	bnxt_free_mem(bp, reconfig_dev);
5763 
5764 	bnxt_hwrm_func_buf_unrgtr(bp);
5765 	if (bp->pf != NULL) {
5766 		rte_free(bp->pf->vf_req_buf);
5767 		bp->pf->vf_req_buf = NULL;
5768 	}
5769 
5770 	rc = bnxt_hwrm_func_driver_unregister(bp, 0);
5771 	bp->flags &= ~BNXT_FLAG_REGISTERED;
5772 	bnxt_free_ctx_mem(bp);
5773 	if (!reconfig_dev) {
5774 		bnxt_free_hwrm_resources(bp);
5775 		bnxt_free_error_recovery_info(bp);
5776 	}
5777 
5778 	bnxt_uninit_ctx_mem(bp);
5779 
5780 	bnxt_free_flow_stats_info(bp);
5781 	if (bp->rep_info != NULL)
5782 		bnxt_free_switch_domain(bp);
5783 	bnxt_free_rep_info(bp);
5784 	rte_free(bp->ptp_cfg);
5785 	bp->ptp_cfg = NULL;
5786 	return rc;
5787 }
5788 
5789 static int
5790 bnxt_dev_uninit(struct rte_eth_dev *eth_dev)
5791 {
5792 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
5793 		return -EPERM;
5794 
5795 	PMD_DRV_LOG(DEBUG, "Calling Device uninit\n");
5796 
5797 	if (eth_dev->state != RTE_ETH_DEV_UNUSED)
5798 		bnxt_dev_close_op(eth_dev);
5799 
5800 	return 0;
5801 }
5802 
5803 static int bnxt_pci_remove_dev_with_reps(struct rte_eth_dev *eth_dev)
5804 {
5805 	struct bnxt *bp = eth_dev->data->dev_private;
5806 	struct rte_eth_dev *vf_rep_eth_dev;
5807 	int ret = 0, i;
5808 
5809 	if (!bp)
5810 		return -EINVAL;
5811 
5812 	for (i = 0; i < bp->num_reps; i++) {
5813 		vf_rep_eth_dev = bp->rep_info[i].vfr_eth_dev;
5814 		if (!vf_rep_eth_dev)
5815 			continue;
5816 		PMD_DRV_LOG(DEBUG, "BNXT Port:%d VFR pci remove\n",
5817 			    vf_rep_eth_dev->data->port_id);
5818 		rte_eth_dev_destroy(vf_rep_eth_dev, bnxt_representor_uninit);
5819 	}
5820 	PMD_DRV_LOG(DEBUG, "BNXT Port:%d pci remove\n",
5821 		    eth_dev->data->port_id);
5822 	ret = rte_eth_dev_destroy(eth_dev, bnxt_dev_uninit);
5823 
5824 	return ret;
5825 }
5826 
5827 static void bnxt_free_rep_info(struct bnxt *bp)
5828 {
5829 	rte_free(bp->rep_info);
5830 	bp->rep_info = NULL;
5831 	rte_free(bp->cfa_code_map);
5832 	bp->cfa_code_map = NULL;
5833 }
5834 
5835 static int bnxt_init_rep_info(struct bnxt *bp)
5836 {
5837 	int i = 0, rc;
5838 
5839 	if (bp->rep_info)
5840 		return 0;
5841 
5842 	bp->rep_info = rte_zmalloc("bnxt_rep_info",
5843 				   sizeof(bp->rep_info[0]) * BNXT_MAX_VF_REPS,
5844 				   0);
5845 	if (!bp->rep_info) {
5846 		PMD_DRV_LOG(ERR, "Failed to alloc memory for rep info\n");
5847 		return -ENOMEM;
5848 	}
5849 	bp->cfa_code_map = rte_zmalloc("bnxt_cfa_code_map",
5850 				       sizeof(*bp->cfa_code_map) *
5851 				       BNXT_MAX_CFA_CODE, 0);
5852 	if (!bp->cfa_code_map) {
5853 		PMD_DRV_LOG(ERR, "Failed to alloc memory for cfa_code_map\n");
5854 		bnxt_free_rep_info(bp);
5855 		return -ENOMEM;
5856 	}
5857 
5858 	for (i = 0; i < BNXT_MAX_CFA_CODE; i++)
5859 		bp->cfa_code_map[i] = BNXT_VF_IDX_INVALID;
5860 
5861 	rc = pthread_mutex_init(&bp->rep_info->vfr_lock, NULL);
5862 	if (rc) {
5863 		PMD_DRV_LOG(ERR, "Unable to initialize vfr_lock\n");
5864 		bnxt_free_rep_info(bp);
5865 		return rc;
5866 	}
5867 
5868 	rc = pthread_mutex_init(&bp->rep_info->vfr_start_lock, NULL);
5869 	if (rc) {
5870 		PMD_DRV_LOG(ERR, "Unable to initialize vfr_start_lock\n");
5871 		bnxt_free_rep_info(bp);
5872 		return rc;
5873 	}
5874 
5875 	return rc;
5876 }
5877 
5878 static int bnxt_rep_port_probe(struct rte_pci_device *pci_dev,
5879 			       struct rte_eth_devargs *eth_da,
5880 			       struct rte_eth_dev *backing_eth_dev,
5881 			       const char *dev_args)
5882 {
5883 	struct rte_eth_dev *vf_rep_eth_dev;
5884 	char name[RTE_ETH_NAME_MAX_LEN];
5885 	struct bnxt *backing_bp;
5886 	uint16_t num_rep;
5887 	int i, ret = 0;
5888 	struct rte_kvargs *kvlist = NULL;
5889 
5890 	if (eth_da->type == RTE_ETH_REPRESENTOR_NONE)
5891 		return 0;
5892 	if (eth_da->type != RTE_ETH_REPRESENTOR_VF) {
5893 		PMD_DRV_LOG(ERR, "unsupported representor type %d\n",
5894 			    eth_da->type);
5895 		return -ENOTSUP;
5896 	}
5897 	num_rep = eth_da->nb_representor_ports;
5898 	if (num_rep > BNXT_MAX_VF_REPS) {
5899 		PMD_DRV_LOG(ERR, "nb_representor_ports = %d > %d MAX VF REPS\n",
5900 			    num_rep, BNXT_MAX_VF_REPS);
5901 		return -EINVAL;
5902 	}
5903 
5904 	if (num_rep >= RTE_MAX_ETHPORTS) {
5905 		PMD_DRV_LOG(ERR,
5906 			    "nb_representor_ports = %d > %d MAX ETHPORTS\n",
5907 			    num_rep, RTE_MAX_ETHPORTS);
5908 		return -EINVAL;
5909 	}
5910 
5911 	backing_bp = backing_eth_dev->data->dev_private;
5912 
5913 	if (!(BNXT_PF(backing_bp) || BNXT_VF_IS_TRUSTED(backing_bp))) {
5914 		PMD_DRV_LOG(ERR,
5915 			    "Not a PF or trusted VF. No Representor support\n");
5916 		/* Returning an error is not an option.
5917 		 * Applications are not handling this correctly
5918 		 */
5919 		return 0;
5920 	}
5921 
5922 	if (bnxt_init_rep_info(backing_bp))
5923 		return 0;
5924 
5925 	for (i = 0; i < num_rep; i++) {
5926 		struct bnxt_representor representor = {
5927 			.vf_id = eth_da->representor_ports[i],
5928 			.switch_domain_id = backing_bp->switch_domain_id,
5929 			.parent_dev = backing_eth_dev
5930 		};
5931 
5932 		if (representor.vf_id >= BNXT_MAX_VF_REPS) {
5933 			PMD_DRV_LOG(ERR, "VF-Rep id %d >= %d MAX VF ID\n",
5934 				    representor.vf_id, BNXT_MAX_VF_REPS);
5935 			continue;
5936 		}
5937 
5938 		/* representor port net_bdf_port */
5939 		snprintf(name, sizeof(name), "net_%s_representor_%d",
5940 			 pci_dev->device.name, eth_da->representor_ports[i]);
5941 
5942 		kvlist = rte_kvargs_parse(dev_args, bnxt_dev_args);
5943 		if (kvlist) {
5944 			/*
5945 			 * Handler for "rep_is_pf" devarg.
5946 			 * Invoked as for ex: "-a 000:00:0d.0,
5947 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
5948 			 */
5949 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_IS_PF,
5950 						 bnxt_parse_devarg_rep_is_pf,
5951 						 (void *)&representor);
5952 			if (ret) {
5953 				ret = -EINVAL;
5954 				goto err;
5955 			}
5956 			/*
5957 			 * Handler for "rep_based_pf" devarg.
5958 			 * Invoked as for ex: "-a 000:00:0d.0,
5959 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
5960 			 */
5961 			ret = rte_kvargs_process(kvlist,
5962 						 BNXT_DEVARG_REP_BASED_PF,
5963 						 bnxt_parse_devarg_rep_based_pf,
5964 						 (void *)&representor);
5965 			if (ret) {
5966 				ret = -EINVAL;
5967 				goto err;
5968 			}
5969 			/*
5970 			 * Handler for "rep_based_pf" devarg.
5971 			 * Invoked as for ex: "-a 000:00:0d.0,
5972 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
5973 			 */
5974 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_Q_R2F,
5975 						 bnxt_parse_devarg_rep_q_r2f,
5976 						 (void *)&representor);
5977 			if (ret) {
5978 				ret = -EINVAL;
5979 				goto err;
5980 			}
5981 			/*
5982 			 * Handler for "rep_based_pf" devarg.
5983 			 * Invoked as for ex: "-a 000:00:0d.0,
5984 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
5985 			 */
5986 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_Q_F2R,
5987 						 bnxt_parse_devarg_rep_q_f2r,
5988 						 (void *)&representor);
5989 			if (ret) {
5990 				ret = -EINVAL;
5991 				goto err;
5992 			}
5993 			/*
5994 			 * Handler for "rep_based_pf" devarg.
5995 			 * Invoked as for ex: "-a 000:00:0d.0,
5996 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
5997 			 */
5998 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_FC_R2F,
5999 						 bnxt_parse_devarg_rep_fc_r2f,
6000 						 (void *)&representor);
6001 			if (ret) {
6002 				ret = -EINVAL;
6003 				goto err;
6004 			}
6005 			/*
6006 			 * Handler for "rep_based_pf" devarg.
6007 			 * Invoked as for ex: "-a 000:00:0d.0,
6008 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6009 			 */
6010 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_FC_F2R,
6011 						 bnxt_parse_devarg_rep_fc_f2r,
6012 						 (void *)&representor);
6013 			if (ret) {
6014 				ret = -EINVAL;
6015 				goto err;
6016 			}
6017 		}
6018 
6019 		ret = rte_eth_dev_create(&pci_dev->device, name,
6020 					 sizeof(struct bnxt_representor),
6021 					 NULL, NULL,
6022 					 bnxt_representor_init,
6023 					 &representor);
6024 		if (ret) {
6025 			PMD_DRV_LOG(ERR, "failed to create bnxt vf "
6026 				    "representor %s.", name);
6027 			goto err;
6028 		}
6029 
6030 		vf_rep_eth_dev = rte_eth_dev_allocated(name);
6031 		if (!vf_rep_eth_dev) {
6032 			PMD_DRV_LOG(ERR, "Failed to find the eth_dev"
6033 				    " for VF-Rep: %s.", name);
6034 			ret = -ENODEV;
6035 			goto err;
6036 		}
6037 
6038 		PMD_DRV_LOG(DEBUG, "BNXT Port:%d VFR pci probe\n",
6039 			    backing_eth_dev->data->port_id);
6040 		backing_bp->rep_info[representor.vf_id].vfr_eth_dev =
6041 							 vf_rep_eth_dev;
6042 		backing_bp->num_reps++;
6043 
6044 	}
6045 
6046 	rte_kvargs_free(kvlist);
6047 	return 0;
6048 
6049 err:
6050 	/* If num_rep > 1, then rollback already created
6051 	 * ports, since we'll be failing the probe anyway
6052 	 */
6053 	if (num_rep > 1)
6054 		bnxt_pci_remove_dev_with_reps(backing_eth_dev);
6055 	rte_errno = -ret;
6056 	rte_kvargs_free(kvlist);
6057 
6058 	return ret;
6059 }
6060 
6061 static int bnxt_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
6062 			  struct rte_pci_device *pci_dev)
6063 {
6064 	struct rte_eth_devargs eth_da = { .nb_representor_ports = 0 };
6065 	struct rte_eth_dev *backing_eth_dev;
6066 	uint16_t num_rep;
6067 	int ret = 0;
6068 
6069 	if (pci_dev->device.devargs) {
6070 		ret = rte_eth_devargs_parse(pci_dev->device.devargs->args,
6071 					    &eth_da);
6072 		if (ret)
6073 			return ret;
6074 	}
6075 
6076 	num_rep = eth_da.nb_representor_ports;
6077 	PMD_DRV_LOG(DEBUG, "nb_representor_ports = %d\n",
6078 		    num_rep);
6079 
6080 	/* We could come here after first level of probe is already invoked
6081 	 * as part of an application bringup(OVS-DPDK vswitchd), so first check
6082 	 * for already allocated eth_dev for the backing device (PF/Trusted VF)
6083 	 */
6084 	backing_eth_dev = rte_eth_dev_allocated(pci_dev->device.name);
6085 	if (backing_eth_dev == NULL) {
6086 		ret = rte_eth_dev_create(&pci_dev->device, pci_dev->device.name,
6087 					 sizeof(struct bnxt),
6088 					 eth_dev_pci_specific_init, pci_dev,
6089 					 bnxt_dev_init, NULL);
6090 
6091 		if (ret || !num_rep)
6092 			return ret;
6093 
6094 		backing_eth_dev = rte_eth_dev_allocated(pci_dev->device.name);
6095 	}
6096 	PMD_DRV_LOG(DEBUG, "BNXT Port:%d pci probe\n",
6097 		    backing_eth_dev->data->port_id);
6098 
6099 	if (!num_rep)
6100 		return ret;
6101 
6102 	/* probe representor ports now */
6103 	ret = bnxt_rep_port_probe(pci_dev, &eth_da, backing_eth_dev,
6104 				  pci_dev->device.devargs->args);
6105 
6106 	return ret;
6107 }
6108 
6109 static int bnxt_pci_remove(struct rte_pci_device *pci_dev)
6110 {
6111 	struct rte_eth_dev *eth_dev;
6112 
6113 	eth_dev = rte_eth_dev_allocated(pci_dev->device.name);
6114 	if (!eth_dev)
6115 		return 0; /* Invoked typically only by OVS-DPDK, by the
6116 			   * time it comes here the eth_dev is already
6117 			   * deleted by rte_eth_dev_close(), so returning
6118 			   * +ve value will at least help in proper cleanup
6119 			   */
6120 
6121 	PMD_DRV_LOG(DEBUG, "BNXT Port:%d pci remove\n", eth_dev->data->port_id);
6122 	if (rte_eal_process_type() == RTE_PROC_PRIMARY) {
6123 		if (eth_dev->data->dev_flags & RTE_ETH_DEV_REPRESENTOR)
6124 			return rte_eth_dev_destroy(eth_dev,
6125 						   bnxt_representor_uninit);
6126 		else
6127 			return rte_eth_dev_destroy(eth_dev,
6128 						   bnxt_dev_uninit);
6129 	} else {
6130 		return rte_eth_dev_pci_generic_remove(pci_dev, NULL);
6131 	}
6132 }
6133 
6134 static struct rte_pci_driver bnxt_rte_pmd = {
6135 	.id_table = bnxt_pci_id_map,
6136 	.drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC |
6137 			RTE_PCI_DRV_PROBE_AGAIN, /* Needed in case of VF-REPs
6138 						  * and OVS-DPDK
6139 						  */
6140 	.probe = bnxt_pci_probe,
6141 	.remove = bnxt_pci_remove,
6142 };
6143 
6144 static bool
6145 is_device_supported(struct rte_eth_dev *dev, struct rte_pci_driver *drv)
6146 {
6147 	if (strcmp(dev->device->driver->name, drv->driver.name))
6148 		return false;
6149 
6150 	return true;
6151 }
6152 
6153 bool is_bnxt_supported(struct rte_eth_dev *dev)
6154 {
6155 	return is_device_supported(dev, &bnxt_rte_pmd);
6156 }
6157 
6158 RTE_LOG_REGISTER(bnxt_logtype_driver, pmd.net.bnxt.driver, NOTICE);
6159 RTE_PMD_REGISTER_PCI(net_bnxt, bnxt_rte_pmd);
6160 RTE_PMD_REGISTER_PCI_TABLE(net_bnxt, bnxt_pci_id_map);
6161 RTE_PMD_REGISTER_KMOD_DEP(net_bnxt, "* igb_uio | uio_pci_generic | vfio-pci");
6162