xref: /dpdk/drivers/net/bnxt/bnxt_ethdev.c (revision b19f366c)
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)) {
2201 		PMD_DRV_LOG(ERR,
2202 			    "Flow Control Settings cannot be modified on VF or on shared PF\n");
2203 		return -ENOTSUP;
2204 	}
2205 
2206 	switch (fc_conf->mode) {
2207 	case RTE_FC_NONE:
2208 		bp->link_info->auto_pause = 0;
2209 		bp->link_info->force_pause = 0;
2210 		break;
2211 	case RTE_FC_RX_PAUSE:
2212 		if (fc_conf->autoneg) {
2213 			bp->link_info->auto_pause =
2214 					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_RX;
2215 			bp->link_info->force_pause = 0;
2216 		} else {
2217 			bp->link_info->auto_pause = 0;
2218 			bp->link_info->force_pause =
2219 					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_RX;
2220 		}
2221 		break;
2222 	case RTE_FC_TX_PAUSE:
2223 		if (fc_conf->autoneg) {
2224 			bp->link_info->auto_pause =
2225 					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_TX;
2226 			bp->link_info->force_pause = 0;
2227 		} else {
2228 			bp->link_info->auto_pause = 0;
2229 			bp->link_info->force_pause =
2230 					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_TX;
2231 		}
2232 		break;
2233 	case RTE_FC_FULL:
2234 		if (fc_conf->autoneg) {
2235 			bp->link_info->auto_pause =
2236 					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_TX |
2237 					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_RX;
2238 			bp->link_info->force_pause = 0;
2239 		} else {
2240 			bp->link_info->auto_pause = 0;
2241 			bp->link_info->force_pause =
2242 					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_TX |
2243 					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_RX;
2244 		}
2245 		break;
2246 	}
2247 	return bnxt_set_hwrm_link_config(bp, true);
2248 }
2249 
2250 /* Add UDP tunneling port */
2251 static int
2252 bnxt_udp_tunnel_port_add_op(struct rte_eth_dev *eth_dev,
2253 			 struct rte_eth_udp_tunnel *udp_tunnel)
2254 {
2255 	struct bnxt *bp = eth_dev->data->dev_private;
2256 	uint16_t tunnel_type = 0;
2257 	int rc = 0;
2258 
2259 	rc = is_bnxt_in_error(bp);
2260 	if (rc)
2261 		return rc;
2262 
2263 	switch (udp_tunnel->prot_type) {
2264 	case RTE_TUNNEL_TYPE_VXLAN:
2265 		if (bp->vxlan_port_cnt) {
2266 			PMD_DRV_LOG(ERR, "Tunnel Port %d already programmed\n",
2267 				udp_tunnel->udp_port);
2268 			if (bp->vxlan_port != udp_tunnel->udp_port) {
2269 				PMD_DRV_LOG(ERR, "Only one port allowed\n");
2270 				return -ENOSPC;
2271 			}
2272 			bp->vxlan_port_cnt++;
2273 			return 0;
2274 		}
2275 		tunnel_type =
2276 			HWRM_TUNNEL_DST_PORT_ALLOC_INPUT_TUNNEL_TYPE_VXLAN;
2277 		bp->vxlan_port_cnt++;
2278 		break;
2279 	case RTE_TUNNEL_TYPE_GENEVE:
2280 		if (bp->geneve_port_cnt) {
2281 			PMD_DRV_LOG(ERR, "Tunnel Port %d already programmed\n",
2282 				udp_tunnel->udp_port);
2283 			if (bp->geneve_port != udp_tunnel->udp_port) {
2284 				PMD_DRV_LOG(ERR, "Only one port allowed\n");
2285 				return -ENOSPC;
2286 			}
2287 			bp->geneve_port_cnt++;
2288 			return 0;
2289 		}
2290 		tunnel_type =
2291 			HWRM_TUNNEL_DST_PORT_ALLOC_INPUT_TUNNEL_TYPE_GENEVE;
2292 		bp->geneve_port_cnt++;
2293 		break;
2294 	default:
2295 		PMD_DRV_LOG(ERR, "Tunnel type is not supported\n");
2296 		return -ENOTSUP;
2297 	}
2298 	rc = bnxt_hwrm_tunnel_dst_port_alloc(bp, udp_tunnel->udp_port,
2299 					     tunnel_type);
2300 	return rc;
2301 }
2302 
2303 static int
2304 bnxt_udp_tunnel_port_del_op(struct rte_eth_dev *eth_dev,
2305 			 struct rte_eth_udp_tunnel *udp_tunnel)
2306 {
2307 	struct bnxt *bp = eth_dev->data->dev_private;
2308 	uint16_t tunnel_type = 0;
2309 	uint16_t port = 0;
2310 	int rc = 0;
2311 
2312 	rc = is_bnxt_in_error(bp);
2313 	if (rc)
2314 		return rc;
2315 
2316 	switch (udp_tunnel->prot_type) {
2317 	case RTE_TUNNEL_TYPE_VXLAN:
2318 		if (!bp->vxlan_port_cnt) {
2319 			PMD_DRV_LOG(ERR, "No Tunnel port configured yet\n");
2320 			return -EINVAL;
2321 		}
2322 		if (bp->vxlan_port != udp_tunnel->udp_port) {
2323 			PMD_DRV_LOG(ERR, "Req Port: %d. Configured port: %d\n",
2324 				udp_tunnel->udp_port, bp->vxlan_port);
2325 			return -EINVAL;
2326 		}
2327 		if (--bp->vxlan_port_cnt)
2328 			return 0;
2329 
2330 		tunnel_type =
2331 			HWRM_TUNNEL_DST_PORT_FREE_INPUT_TUNNEL_TYPE_VXLAN;
2332 		port = bp->vxlan_fw_dst_port_id;
2333 		break;
2334 	case RTE_TUNNEL_TYPE_GENEVE:
2335 		if (!bp->geneve_port_cnt) {
2336 			PMD_DRV_LOG(ERR, "No Tunnel port configured yet\n");
2337 			return -EINVAL;
2338 		}
2339 		if (bp->geneve_port != udp_tunnel->udp_port) {
2340 			PMD_DRV_LOG(ERR, "Req Port: %d. Configured port: %d\n",
2341 				udp_tunnel->udp_port, bp->geneve_port);
2342 			return -EINVAL;
2343 		}
2344 		if (--bp->geneve_port_cnt)
2345 			return 0;
2346 
2347 		tunnel_type =
2348 			HWRM_TUNNEL_DST_PORT_FREE_INPUT_TUNNEL_TYPE_GENEVE;
2349 		port = bp->geneve_fw_dst_port_id;
2350 		break;
2351 	default:
2352 		PMD_DRV_LOG(ERR, "Tunnel type is not supported\n");
2353 		return -ENOTSUP;
2354 	}
2355 
2356 	rc = bnxt_hwrm_tunnel_dst_port_free(bp, port, tunnel_type);
2357 	return rc;
2358 }
2359 
2360 static int bnxt_del_vlan_filter(struct bnxt *bp, uint16_t vlan_id)
2361 {
2362 	struct bnxt_filter_info *filter;
2363 	struct bnxt_vnic_info *vnic;
2364 	int rc = 0;
2365 	uint32_t chk = HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN;
2366 
2367 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
2368 	filter = STAILQ_FIRST(&vnic->filter);
2369 	while (filter) {
2370 		/* Search for this matching MAC+VLAN filter */
2371 		if (bnxt_vlan_filter_exists(bp, filter, chk, vlan_id)) {
2372 			/* Delete the filter */
2373 			rc = bnxt_hwrm_clear_l2_filter(bp, filter);
2374 			if (rc)
2375 				return rc;
2376 			STAILQ_REMOVE(&vnic->filter, filter,
2377 				      bnxt_filter_info, next);
2378 			bnxt_free_filter(bp, filter);
2379 			PMD_DRV_LOG(INFO,
2380 				    "Deleted vlan filter for %d\n",
2381 				    vlan_id);
2382 			return 0;
2383 		}
2384 		filter = STAILQ_NEXT(filter, next);
2385 	}
2386 	return -ENOENT;
2387 }
2388 
2389 static int bnxt_add_vlan_filter(struct bnxt *bp, uint16_t vlan_id)
2390 {
2391 	struct bnxt_filter_info *filter;
2392 	struct bnxt_vnic_info *vnic;
2393 	int rc = 0;
2394 	uint32_t en = HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN |
2395 		HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN_MASK;
2396 	uint32_t chk = HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN;
2397 
2398 	/* Implementation notes on the use of VNIC in this command:
2399 	 *
2400 	 * By default, these filters belong to default vnic for the function.
2401 	 * Once these filters are set up, only destination VNIC can be modified.
2402 	 * If the destination VNIC is not specified in this command,
2403 	 * then the HWRM shall only create an l2 context id.
2404 	 */
2405 
2406 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
2407 	filter = STAILQ_FIRST(&vnic->filter);
2408 	/* Check if the VLAN has already been added */
2409 	while (filter) {
2410 		if (bnxt_vlan_filter_exists(bp, filter, chk, vlan_id))
2411 			return -EEXIST;
2412 
2413 		filter = STAILQ_NEXT(filter, next);
2414 	}
2415 
2416 	/* No match found. Alloc a fresh filter and issue the L2_FILTER_ALLOC
2417 	 * command to create MAC+VLAN filter with the right flags, enables set.
2418 	 */
2419 	filter = bnxt_alloc_filter(bp);
2420 	if (!filter) {
2421 		PMD_DRV_LOG(ERR,
2422 			    "MAC/VLAN filter alloc failed\n");
2423 		return -ENOMEM;
2424 	}
2425 	/* MAC + VLAN ID filter */
2426 	/* If l2_ivlan == 0 and l2_ivlan_mask != 0, only
2427 	 * untagged packets are received
2428 	 *
2429 	 * If l2_ivlan != 0 and l2_ivlan_mask != 0, untagged
2430 	 * packets and only the programmed vlan's packets are received
2431 	 */
2432 	filter->l2_ivlan = vlan_id;
2433 	filter->l2_ivlan_mask = 0x0FFF;
2434 	filter->enables |= en;
2435 	filter->flags |= HWRM_CFA_L2_FILTER_ALLOC_INPUT_FLAGS_OUTERMOST;
2436 
2437 	rc = bnxt_hwrm_set_l2_filter(bp, vnic->fw_vnic_id, filter);
2438 	if (rc) {
2439 		/* Free the newly allocated filter as we were
2440 		 * not able to create the filter in hardware.
2441 		 */
2442 		bnxt_free_filter(bp, filter);
2443 		return rc;
2444 	}
2445 
2446 	filter->mac_index = 0;
2447 	/* Add this new filter to the list */
2448 	if (vlan_id == 0)
2449 		STAILQ_INSERT_HEAD(&vnic->filter, filter, next);
2450 	else
2451 		STAILQ_INSERT_TAIL(&vnic->filter, filter, next);
2452 
2453 	PMD_DRV_LOG(INFO,
2454 		    "Added Vlan filter for %d\n", vlan_id);
2455 	return rc;
2456 }
2457 
2458 static int bnxt_vlan_filter_set_op(struct rte_eth_dev *eth_dev,
2459 		uint16_t vlan_id, int on)
2460 {
2461 	struct bnxt *bp = eth_dev->data->dev_private;
2462 	int rc;
2463 
2464 	rc = is_bnxt_in_error(bp);
2465 	if (rc)
2466 		return rc;
2467 
2468 	if (!eth_dev->data->dev_started) {
2469 		PMD_DRV_LOG(ERR, "port must be started before setting vlan\n");
2470 		return -EINVAL;
2471 	}
2472 
2473 	/* These operations apply to ALL existing MAC/VLAN filters */
2474 	if (on)
2475 		return bnxt_add_vlan_filter(bp, vlan_id);
2476 	else
2477 		return bnxt_del_vlan_filter(bp, vlan_id);
2478 }
2479 
2480 static int bnxt_del_dflt_mac_filter(struct bnxt *bp,
2481 				    struct bnxt_vnic_info *vnic)
2482 {
2483 	struct bnxt_filter_info *filter;
2484 	int rc;
2485 
2486 	filter = STAILQ_FIRST(&vnic->filter);
2487 	while (filter) {
2488 		if (filter->mac_index == 0 &&
2489 		    !memcmp(filter->l2_addr, bp->mac_addr,
2490 			    RTE_ETHER_ADDR_LEN)) {
2491 			rc = bnxt_hwrm_clear_l2_filter(bp, filter);
2492 			if (!rc) {
2493 				STAILQ_REMOVE(&vnic->filter, filter,
2494 					      bnxt_filter_info, next);
2495 				bnxt_free_filter(bp, filter);
2496 			}
2497 			return rc;
2498 		}
2499 		filter = STAILQ_NEXT(filter, next);
2500 	}
2501 	return 0;
2502 }
2503 
2504 static int
2505 bnxt_config_vlan_hw_filter(struct bnxt *bp, uint64_t rx_offloads)
2506 {
2507 	struct bnxt_vnic_info *vnic;
2508 	unsigned int i;
2509 	int rc;
2510 
2511 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
2512 	if (!(rx_offloads & DEV_RX_OFFLOAD_VLAN_FILTER)) {
2513 		/* Remove any VLAN filters programmed */
2514 		for (i = 0; i < RTE_ETHER_MAX_VLAN_ID; i++)
2515 			bnxt_del_vlan_filter(bp, i);
2516 
2517 		rc = bnxt_add_mac_filter(bp, vnic, NULL, 0, 0);
2518 		if (rc)
2519 			return rc;
2520 	} else {
2521 		/* Default filter will allow packets that match the
2522 		 * dest mac. So, it has to be deleted, otherwise, we
2523 		 * will endup receiving vlan packets for which the
2524 		 * filter is not programmed, when hw-vlan-filter
2525 		 * configuration is ON
2526 		 */
2527 		bnxt_del_dflt_mac_filter(bp, vnic);
2528 		/* This filter will allow only untagged packets */
2529 		bnxt_add_vlan_filter(bp, 0);
2530 	}
2531 	PMD_DRV_LOG(DEBUG, "VLAN Filtering: %d\n",
2532 		    !!(rx_offloads & DEV_RX_OFFLOAD_VLAN_FILTER));
2533 
2534 	return 0;
2535 }
2536 
2537 static int bnxt_free_one_vnic(struct bnxt *bp, uint16_t vnic_id)
2538 {
2539 	struct bnxt_vnic_info *vnic = &bp->vnic_info[vnic_id];
2540 	unsigned int i;
2541 	int rc;
2542 
2543 	/* Destroy vnic filters and vnic */
2544 	if (bp->eth_dev->data->dev_conf.rxmode.offloads &
2545 	    DEV_RX_OFFLOAD_VLAN_FILTER) {
2546 		for (i = 0; i < RTE_ETHER_MAX_VLAN_ID; i++)
2547 			bnxt_del_vlan_filter(bp, i);
2548 	}
2549 	bnxt_del_dflt_mac_filter(bp, vnic);
2550 
2551 	rc = bnxt_hwrm_vnic_ctx_free(bp, vnic);
2552 	if (rc)
2553 		return rc;
2554 
2555 	rc = bnxt_hwrm_vnic_free(bp, vnic);
2556 	if (rc)
2557 		return rc;
2558 
2559 	rte_free(vnic->fw_grp_ids);
2560 	vnic->fw_grp_ids = NULL;
2561 
2562 	vnic->rx_queue_cnt = 0;
2563 
2564 	return 0;
2565 }
2566 
2567 static int
2568 bnxt_config_vlan_hw_stripping(struct bnxt *bp, uint64_t rx_offloads)
2569 {
2570 	struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
2571 	int rc;
2572 
2573 	/* Destroy, recreate and reconfigure the default vnic */
2574 	rc = bnxt_free_one_vnic(bp, 0);
2575 	if (rc)
2576 		return rc;
2577 
2578 	/* default vnic 0 */
2579 	rc = bnxt_setup_one_vnic(bp, 0);
2580 	if (rc)
2581 		return rc;
2582 
2583 	if (bp->eth_dev->data->dev_conf.rxmode.offloads &
2584 	    DEV_RX_OFFLOAD_VLAN_FILTER) {
2585 		rc = bnxt_add_vlan_filter(bp, 0);
2586 		if (rc)
2587 			return rc;
2588 		rc = bnxt_restore_vlan_filters(bp);
2589 		if (rc)
2590 			return rc;
2591 	} else {
2592 		rc = bnxt_add_mac_filter(bp, vnic, NULL, 0, 0);
2593 		if (rc)
2594 			return rc;
2595 	}
2596 
2597 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
2598 	if (rc)
2599 		return rc;
2600 
2601 	PMD_DRV_LOG(DEBUG, "VLAN Strip Offload: %d\n",
2602 		    !!(rx_offloads & DEV_RX_OFFLOAD_VLAN_STRIP));
2603 
2604 	return rc;
2605 }
2606 
2607 static int
2608 bnxt_vlan_offload_set_op(struct rte_eth_dev *dev, int mask)
2609 {
2610 	uint64_t rx_offloads = dev->data->dev_conf.rxmode.offloads;
2611 	struct bnxt *bp = dev->data->dev_private;
2612 	int rc;
2613 
2614 	rc = is_bnxt_in_error(bp);
2615 	if (rc)
2616 		return rc;
2617 
2618 	/* Filter settings will get applied when port is started */
2619 	if (!dev->data->dev_started)
2620 		return 0;
2621 
2622 	if (mask & ETH_VLAN_FILTER_MASK) {
2623 		/* Enable or disable VLAN filtering */
2624 		rc = bnxt_config_vlan_hw_filter(bp, rx_offloads);
2625 		if (rc)
2626 			return rc;
2627 	}
2628 
2629 	if (mask & ETH_VLAN_STRIP_MASK) {
2630 		/* Enable or disable VLAN stripping */
2631 		rc = bnxt_config_vlan_hw_stripping(bp, rx_offloads);
2632 		if (rc)
2633 			return rc;
2634 	}
2635 
2636 	if (mask & ETH_VLAN_EXTEND_MASK) {
2637 		if (rx_offloads & DEV_RX_OFFLOAD_VLAN_EXTEND)
2638 			PMD_DRV_LOG(DEBUG, "Extend VLAN supported\n");
2639 		else
2640 			PMD_DRV_LOG(INFO, "Extend VLAN unsupported\n");
2641 	}
2642 
2643 	return 0;
2644 }
2645 
2646 static int
2647 bnxt_vlan_tpid_set_op(struct rte_eth_dev *dev, enum rte_vlan_type vlan_type,
2648 		      uint16_t tpid)
2649 {
2650 	struct bnxt *bp = dev->data->dev_private;
2651 	int qinq = dev->data->dev_conf.rxmode.offloads &
2652 		   DEV_RX_OFFLOAD_VLAN_EXTEND;
2653 
2654 	if (vlan_type != ETH_VLAN_TYPE_INNER &&
2655 	    vlan_type != ETH_VLAN_TYPE_OUTER) {
2656 		PMD_DRV_LOG(ERR,
2657 			    "Unsupported vlan type.");
2658 		return -EINVAL;
2659 	}
2660 	if (!qinq) {
2661 		PMD_DRV_LOG(ERR,
2662 			    "QinQ not enabled. Needs to be ON as we can "
2663 			    "accelerate only outer vlan\n");
2664 		return -EINVAL;
2665 	}
2666 
2667 	if (vlan_type == ETH_VLAN_TYPE_OUTER) {
2668 		switch (tpid) {
2669 		case RTE_ETHER_TYPE_QINQ:
2670 			bp->outer_tpid_bd =
2671 				TX_BD_LONG_CFA_META_VLAN_TPID_TPID88A8;
2672 				break;
2673 		case RTE_ETHER_TYPE_VLAN:
2674 			bp->outer_tpid_bd =
2675 				TX_BD_LONG_CFA_META_VLAN_TPID_TPID8100;
2676 				break;
2677 		case RTE_ETHER_TYPE_QINQ1:
2678 			bp->outer_tpid_bd =
2679 				TX_BD_LONG_CFA_META_VLAN_TPID_TPID9100;
2680 				break;
2681 		case RTE_ETHER_TYPE_QINQ2:
2682 			bp->outer_tpid_bd =
2683 				TX_BD_LONG_CFA_META_VLAN_TPID_TPID9200;
2684 				break;
2685 		case RTE_ETHER_TYPE_QINQ3:
2686 			bp->outer_tpid_bd =
2687 				 TX_BD_LONG_CFA_META_VLAN_TPID_TPID9300;
2688 				break;
2689 		default:
2690 			PMD_DRV_LOG(ERR, "Invalid TPID: %x\n", tpid);
2691 			return -EINVAL;
2692 		}
2693 		bp->outer_tpid_bd |= tpid;
2694 		PMD_DRV_LOG(INFO, "outer_tpid_bd = %x\n", bp->outer_tpid_bd);
2695 	} else if (vlan_type == ETH_VLAN_TYPE_INNER) {
2696 		PMD_DRV_LOG(ERR,
2697 			    "Can accelerate only outer vlan in QinQ\n");
2698 		return -EINVAL;
2699 	}
2700 
2701 	return 0;
2702 }
2703 
2704 static int
2705 bnxt_set_default_mac_addr_op(struct rte_eth_dev *dev,
2706 			     struct rte_ether_addr *addr)
2707 {
2708 	struct bnxt *bp = dev->data->dev_private;
2709 	/* Default Filter is tied to VNIC 0 */
2710 	struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
2711 	int rc;
2712 
2713 	rc = is_bnxt_in_error(bp);
2714 	if (rc)
2715 		return rc;
2716 
2717 	if (BNXT_VF(bp) && !BNXT_VF_IS_TRUSTED(bp))
2718 		return -EPERM;
2719 
2720 	if (rte_is_zero_ether_addr(addr))
2721 		return -EINVAL;
2722 
2723 	/* Filter settings will get applied when port is started */
2724 	if (!dev->data->dev_started)
2725 		return 0;
2726 
2727 	/* Check if the requested MAC is already added */
2728 	if (memcmp(addr, bp->mac_addr, RTE_ETHER_ADDR_LEN) == 0)
2729 		return 0;
2730 
2731 	/* Destroy filter and re-create it */
2732 	bnxt_del_dflt_mac_filter(bp, vnic);
2733 
2734 	memcpy(bp->mac_addr, addr, RTE_ETHER_ADDR_LEN);
2735 	if (dev->data->dev_conf.rxmode.offloads & DEV_RX_OFFLOAD_VLAN_FILTER) {
2736 		/* This filter will allow only untagged packets */
2737 		rc = bnxt_add_vlan_filter(bp, 0);
2738 	} else {
2739 		rc = bnxt_add_mac_filter(bp, vnic, addr, 0, 0);
2740 	}
2741 
2742 	PMD_DRV_LOG(DEBUG, "Set MAC addr\n");
2743 	return rc;
2744 }
2745 
2746 static int
2747 bnxt_dev_set_mc_addr_list_op(struct rte_eth_dev *eth_dev,
2748 			  struct rte_ether_addr *mc_addr_set,
2749 			  uint32_t nb_mc_addr)
2750 {
2751 	struct bnxt *bp = eth_dev->data->dev_private;
2752 	char *mc_addr_list = (char *)mc_addr_set;
2753 	struct bnxt_vnic_info *vnic;
2754 	uint32_t off = 0, i = 0;
2755 	int rc;
2756 
2757 	rc = is_bnxt_in_error(bp);
2758 	if (rc)
2759 		return rc;
2760 
2761 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
2762 
2763 	if (nb_mc_addr > BNXT_MAX_MC_ADDRS) {
2764 		vnic->flags |= BNXT_VNIC_INFO_ALLMULTI;
2765 		goto allmulti;
2766 	}
2767 
2768 	/* TODO Check for Duplicate mcast addresses */
2769 	vnic->flags &= ~BNXT_VNIC_INFO_ALLMULTI;
2770 	for (i = 0; i < nb_mc_addr; i++) {
2771 		memcpy(vnic->mc_list + off, &mc_addr_list[i],
2772 			RTE_ETHER_ADDR_LEN);
2773 		off += RTE_ETHER_ADDR_LEN;
2774 	}
2775 
2776 	vnic->mc_addr_cnt = i;
2777 	if (vnic->mc_addr_cnt)
2778 		vnic->flags |= BNXT_VNIC_INFO_MCAST;
2779 	else
2780 		vnic->flags &= ~BNXT_VNIC_INFO_MCAST;
2781 
2782 allmulti:
2783 	return bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
2784 }
2785 
2786 static int
2787 bnxt_fw_version_get(struct rte_eth_dev *dev, char *fw_version, size_t fw_size)
2788 {
2789 	struct bnxt *bp = dev->data->dev_private;
2790 	uint8_t fw_major = (bp->fw_ver >> 24) & 0xff;
2791 	uint8_t fw_minor = (bp->fw_ver >> 16) & 0xff;
2792 	uint8_t fw_updt = (bp->fw_ver >> 8) & 0xff;
2793 	uint8_t fw_rsvd = bp->fw_ver & 0xff;
2794 	int ret;
2795 
2796 	ret = snprintf(fw_version, fw_size, "%d.%d.%d.%d",
2797 			fw_major, fw_minor, fw_updt, fw_rsvd);
2798 	if (ret < 0)
2799 		return -EINVAL;
2800 
2801 	ret += 1; /* add the size of '\0' */
2802 	if (fw_size < (size_t)ret)
2803 		return ret;
2804 	else
2805 		return 0;
2806 }
2807 
2808 static void
2809 bnxt_rxq_info_get_op(struct rte_eth_dev *dev, uint16_t queue_id,
2810 	struct rte_eth_rxq_info *qinfo)
2811 {
2812 	struct bnxt *bp = dev->data->dev_private;
2813 	struct bnxt_rx_queue *rxq;
2814 
2815 	if (is_bnxt_in_error(bp))
2816 		return;
2817 
2818 	rxq = dev->data->rx_queues[queue_id];
2819 
2820 	qinfo->mp = rxq->mb_pool;
2821 	qinfo->scattered_rx = dev->data->scattered_rx;
2822 	qinfo->nb_desc = rxq->nb_rx_desc;
2823 
2824 	qinfo->conf.rx_free_thresh = rxq->rx_free_thresh;
2825 	qinfo->conf.rx_drop_en = rxq->drop_en;
2826 	qinfo->conf.rx_deferred_start = rxq->rx_deferred_start;
2827 	qinfo->conf.offloads = dev->data->dev_conf.rxmode.offloads;
2828 }
2829 
2830 static void
2831 bnxt_txq_info_get_op(struct rte_eth_dev *dev, uint16_t queue_id,
2832 	struct rte_eth_txq_info *qinfo)
2833 {
2834 	struct bnxt *bp = dev->data->dev_private;
2835 	struct bnxt_tx_queue *txq;
2836 
2837 	if (is_bnxt_in_error(bp))
2838 		return;
2839 
2840 	txq = dev->data->tx_queues[queue_id];
2841 
2842 	qinfo->nb_desc = txq->nb_tx_desc;
2843 
2844 	qinfo->conf.tx_thresh.pthresh = txq->pthresh;
2845 	qinfo->conf.tx_thresh.hthresh = txq->hthresh;
2846 	qinfo->conf.tx_thresh.wthresh = txq->wthresh;
2847 
2848 	qinfo->conf.tx_free_thresh = txq->tx_free_thresh;
2849 	qinfo->conf.tx_rs_thresh = 0;
2850 	qinfo->conf.tx_deferred_start = txq->tx_deferred_start;
2851 	qinfo->conf.offloads = txq->offloads;
2852 }
2853 
2854 static const struct {
2855 	eth_rx_burst_t pkt_burst;
2856 	const char *info;
2857 } bnxt_rx_burst_info[] = {
2858 	{bnxt_recv_pkts,	"Scalar"},
2859 #if defined(RTE_ARCH_X86)
2860 	{bnxt_recv_pkts_vec,	"Vector SSE"},
2861 #elif defined(RTE_ARCH_ARM64)
2862 	{bnxt_recv_pkts_vec,	"Vector Neon"},
2863 #endif
2864 };
2865 
2866 static int
2867 bnxt_rx_burst_mode_get(struct rte_eth_dev *dev, __rte_unused uint16_t queue_id,
2868 		       struct rte_eth_burst_mode *mode)
2869 {
2870 	eth_rx_burst_t pkt_burst = dev->rx_pkt_burst;
2871 	size_t i;
2872 
2873 	for (i = 0; i < RTE_DIM(bnxt_rx_burst_info); i++) {
2874 		if (pkt_burst == bnxt_rx_burst_info[i].pkt_burst) {
2875 			snprintf(mode->info, sizeof(mode->info), "%s",
2876 				 bnxt_rx_burst_info[i].info);
2877 			return 0;
2878 		}
2879 	}
2880 
2881 	return -EINVAL;
2882 }
2883 
2884 static const struct {
2885 	eth_tx_burst_t pkt_burst;
2886 	const char *info;
2887 } bnxt_tx_burst_info[] = {
2888 	{bnxt_xmit_pkts,	"Scalar"},
2889 #if defined(RTE_ARCH_X86)
2890 	{bnxt_xmit_pkts_vec,	"Vector SSE"},
2891 #elif defined(RTE_ARCH_ARM64)
2892 	{bnxt_xmit_pkts_vec,	"Vector Neon"},
2893 #endif
2894 };
2895 
2896 static int
2897 bnxt_tx_burst_mode_get(struct rte_eth_dev *dev, __rte_unused uint16_t queue_id,
2898 		       struct rte_eth_burst_mode *mode)
2899 {
2900 	eth_tx_burst_t pkt_burst = dev->tx_pkt_burst;
2901 	size_t i;
2902 
2903 	for (i = 0; i < RTE_DIM(bnxt_tx_burst_info); i++) {
2904 		if (pkt_burst == bnxt_tx_burst_info[i].pkt_burst) {
2905 			snprintf(mode->info, sizeof(mode->info), "%s",
2906 				 bnxt_tx_burst_info[i].info);
2907 			return 0;
2908 		}
2909 	}
2910 
2911 	return -EINVAL;
2912 }
2913 
2914 int bnxt_mtu_set_op(struct rte_eth_dev *eth_dev, uint16_t new_mtu)
2915 {
2916 	struct bnxt *bp = eth_dev->data->dev_private;
2917 	uint32_t new_pkt_size;
2918 	uint32_t rc = 0;
2919 	uint32_t i;
2920 
2921 	rc = is_bnxt_in_error(bp);
2922 	if (rc)
2923 		return rc;
2924 
2925 	/* Exit if receive queues are not configured yet */
2926 	if (!eth_dev->data->nb_rx_queues)
2927 		return rc;
2928 
2929 	new_pkt_size = new_mtu + RTE_ETHER_HDR_LEN + RTE_ETHER_CRC_LEN +
2930 		       VLAN_TAG_SIZE * BNXT_NUM_VLANS;
2931 
2932 	/*
2933 	 * Disallow any MTU change that would require scattered receive support
2934 	 * if it is not already enabled.
2935 	 */
2936 	if (eth_dev->data->dev_started &&
2937 	    !eth_dev->data->scattered_rx &&
2938 	    (new_pkt_size >
2939 	     eth_dev->data->min_rx_buf_size - RTE_PKTMBUF_HEADROOM)) {
2940 		PMD_DRV_LOG(ERR,
2941 			    "MTU change would require scattered rx support. ");
2942 		PMD_DRV_LOG(ERR, "Stop port before changing MTU.\n");
2943 		return -EINVAL;
2944 	}
2945 
2946 	if (new_mtu > RTE_ETHER_MTU) {
2947 		bp->flags |= BNXT_FLAG_JUMBO;
2948 		bp->eth_dev->data->dev_conf.rxmode.offloads |=
2949 			DEV_RX_OFFLOAD_JUMBO_FRAME;
2950 	} else {
2951 		bp->eth_dev->data->dev_conf.rxmode.offloads &=
2952 			~DEV_RX_OFFLOAD_JUMBO_FRAME;
2953 		bp->flags &= ~BNXT_FLAG_JUMBO;
2954 	}
2955 
2956 	/* Is there a change in mtu setting? */
2957 	if (eth_dev->data->dev_conf.rxmode.max_rx_pkt_len == new_pkt_size)
2958 		return rc;
2959 
2960 	for (i = 0; i < bp->nr_vnics; i++) {
2961 		struct bnxt_vnic_info *vnic = &bp->vnic_info[i];
2962 		uint16_t size = 0;
2963 
2964 		vnic->mru = BNXT_VNIC_MRU(new_mtu);
2965 		rc = bnxt_hwrm_vnic_cfg(bp, vnic);
2966 		if (rc)
2967 			break;
2968 
2969 		size = rte_pktmbuf_data_room_size(bp->rx_queues[0]->mb_pool);
2970 		size -= RTE_PKTMBUF_HEADROOM;
2971 
2972 		if (size < new_mtu) {
2973 			rc = bnxt_hwrm_vnic_plcmode_cfg(bp, vnic);
2974 			if (rc)
2975 				return rc;
2976 		}
2977 	}
2978 
2979 	if (!rc)
2980 		eth_dev->data->dev_conf.rxmode.max_rx_pkt_len = new_pkt_size;
2981 
2982 	PMD_DRV_LOG(INFO, "New MTU is %d\n", new_mtu);
2983 
2984 	return rc;
2985 }
2986 
2987 static int
2988 bnxt_vlan_pvid_set_op(struct rte_eth_dev *dev, uint16_t pvid, int on)
2989 {
2990 	struct bnxt *bp = dev->data->dev_private;
2991 	uint16_t vlan = bp->vlan;
2992 	int rc;
2993 
2994 	rc = is_bnxt_in_error(bp);
2995 	if (rc)
2996 		return rc;
2997 
2998 	if (!BNXT_SINGLE_PF(bp)) {
2999 		PMD_DRV_LOG(ERR, "PVID cannot be modified on VF or on shared PF\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_clr_rx_ts(struct bnxt *bp, uint64_t *last_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 	int i = 0;
3399 	uint32_t fifo;
3400 
3401 	if (!ptp || (bp->flags & BNXT_FLAG_CHIP_P5))
3402 		return -EINVAL;
3403 
3404 	port_id = pf->port_id;
3405 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3406 				ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO]));
3407 	while ((fifo & BNXT_PTP_RX_FIFO_PENDING) && (i < BNXT_PTP_RX_PND_CNT)) {
3408 		rte_write32(1 << port_id, (uint8_t *)bp->bar0 +
3409 			    ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO_ADV]);
3410 		fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3411 					ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO]));
3412 		*last_ts = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3413 					ptp->rx_mapped_regs[BNXT_PTP_RX_TS_L]));
3414 		*last_ts |= (uint64_t)rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3415 					ptp->rx_mapped_regs[BNXT_PTP_RX_TS_H])) << 32;
3416 		i++;
3417 	}
3418 
3419 	if (i >= BNXT_PTP_RX_PND_CNT)
3420 		return -EBUSY;
3421 
3422 	return 0;
3423 }
3424 
3425 static int bnxt_get_rx_ts(struct bnxt *bp, uint64_t *ts)
3426 {
3427 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3428 	struct bnxt_pf_info *pf = bp->pf;
3429 	uint16_t port_id;
3430 	uint32_t fifo;
3431 
3432 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3433 				ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO]));
3434 	if (!(fifo & BNXT_PTP_RX_FIFO_PENDING))
3435 		return -EAGAIN;
3436 
3437 	port_id = pf->port_id;
3438 	rte_write32(1 << port_id, (uint8_t *)bp->bar0 +
3439 	       ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO_ADV]);
3440 
3441 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3442 				   ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO]));
3443 	if (fifo & BNXT_PTP_RX_FIFO_PENDING)
3444 		return bnxt_clr_rx_ts(bp, ts);
3445 
3446 	*ts = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3447 				ptp->rx_mapped_regs[BNXT_PTP_RX_TS_L]));
3448 	*ts |= (uint64_t)rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3449 				ptp->rx_mapped_regs[BNXT_PTP_RX_TS_H])) << 32;
3450 
3451 	return 0;
3452 }
3453 
3454 static int
3455 bnxt_timesync_write_time(struct rte_eth_dev *dev, const struct timespec *ts)
3456 {
3457 	uint64_t ns;
3458 	struct bnxt *bp = dev->data->dev_private;
3459 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3460 
3461 	if (!ptp)
3462 		return -ENOTSUP;
3463 
3464 	ns = rte_timespec_to_ns(ts);
3465 	/* Set the timecounters to a new value. */
3466 	ptp->tc.nsec = ns;
3467 	ptp->tx_tstamp_tc.nsec = ns;
3468 	ptp->rx_tstamp_tc.nsec = ns;
3469 
3470 	return 0;
3471 }
3472 
3473 static int
3474 bnxt_timesync_read_time(struct rte_eth_dev *dev, struct timespec *ts)
3475 {
3476 	struct bnxt *bp = dev->data->dev_private;
3477 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3478 	uint64_t ns, systime_cycles = 0;
3479 	int rc = 0;
3480 
3481 	if (!ptp)
3482 		return -ENOTSUP;
3483 
3484 	if (BNXT_CHIP_P5(bp))
3485 		rc = bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_CURRENT_TIME,
3486 					     &systime_cycles);
3487 	else
3488 		systime_cycles = bnxt_cc_read(bp);
3489 
3490 	ns = rte_timecounter_update(&ptp->tc, systime_cycles);
3491 	*ts = rte_ns_to_timespec(ns);
3492 
3493 	return rc;
3494 }
3495 static int
3496 bnxt_timesync_enable(struct rte_eth_dev *dev)
3497 {
3498 	struct bnxt *bp = dev->data->dev_private;
3499 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3500 	uint32_t shift = 0;
3501 	int rc;
3502 
3503 	if (!ptp)
3504 		return -ENOTSUP;
3505 
3506 	ptp->rx_filter = 1;
3507 	ptp->tx_tstamp_en = 1;
3508 	ptp->rxctl = BNXT_PTP_MSG_EVENTS;
3509 
3510 	rc = bnxt_hwrm_ptp_cfg(bp);
3511 	if (rc)
3512 		return rc;
3513 
3514 	memset(&ptp->tc, 0, sizeof(struct rte_timecounter));
3515 	memset(&ptp->rx_tstamp_tc, 0, sizeof(struct rte_timecounter));
3516 	memset(&ptp->tx_tstamp_tc, 0, sizeof(struct rte_timecounter));
3517 
3518 	ptp->tc.cc_mask = BNXT_CYCLECOUNTER_MASK;
3519 	ptp->tc.cc_shift = shift;
3520 	ptp->tc.nsec_mask = (1ULL << shift) - 1;
3521 
3522 	ptp->rx_tstamp_tc.cc_mask = BNXT_CYCLECOUNTER_MASK;
3523 	ptp->rx_tstamp_tc.cc_shift = shift;
3524 	ptp->rx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
3525 
3526 	ptp->tx_tstamp_tc.cc_mask = BNXT_CYCLECOUNTER_MASK;
3527 	ptp->tx_tstamp_tc.cc_shift = shift;
3528 	ptp->tx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
3529 
3530 	if (!BNXT_CHIP_P5(bp))
3531 		bnxt_map_ptp_regs(bp);
3532 	else
3533 		rc = bnxt_ptp_start(bp);
3534 
3535 	return rc;
3536 }
3537 
3538 static int
3539 bnxt_timesync_disable(struct rte_eth_dev *dev)
3540 {
3541 	struct bnxt *bp = dev->data->dev_private;
3542 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3543 
3544 	if (!ptp)
3545 		return -ENOTSUP;
3546 
3547 	ptp->rx_filter = 0;
3548 	ptp->tx_tstamp_en = 0;
3549 	ptp->rxctl = 0;
3550 
3551 	bnxt_hwrm_ptp_cfg(bp);
3552 
3553 	if (!BNXT_CHIP_P5(bp))
3554 		bnxt_unmap_ptp_regs(bp);
3555 	else
3556 		bnxt_ptp_stop(bp);
3557 
3558 	return 0;
3559 }
3560 
3561 static int
3562 bnxt_timesync_read_rx_timestamp(struct rte_eth_dev *dev,
3563 				 struct timespec *timestamp,
3564 				 uint32_t flags __rte_unused)
3565 {
3566 	struct bnxt *bp = dev->data->dev_private;
3567 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3568 	uint64_t rx_tstamp_cycles = 0;
3569 	uint64_t ns;
3570 
3571 	if (!ptp)
3572 		return -ENOTSUP;
3573 
3574 	if (BNXT_CHIP_P5(bp))
3575 		rx_tstamp_cycles = ptp->rx_timestamp;
3576 	else
3577 		bnxt_get_rx_ts(bp, &rx_tstamp_cycles);
3578 
3579 	ns = rte_timecounter_update(&ptp->rx_tstamp_tc, rx_tstamp_cycles);
3580 	*timestamp = rte_ns_to_timespec(ns);
3581 	return  0;
3582 }
3583 
3584 static int
3585 bnxt_timesync_read_tx_timestamp(struct rte_eth_dev *dev,
3586 				 struct timespec *timestamp)
3587 {
3588 	struct bnxt *bp = dev->data->dev_private;
3589 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3590 	uint64_t tx_tstamp_cycles = 0;
3591 	uint64_t ns;
3592 	int rc = 0;
3593 
3594 	if (!ptp)
3595 		return -ENOTSUP;
3596 
3597 	if (BNXT_CHIP_P5(bp))
3598 		rc = bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_PATH_TX,
3599 					     &tx_tstamp_cycles);
3600 	else
3601 		rc = bnxt_get_tx_ts(bp, &tx_tstamp_cycles);
3602 
3603 	ns = rte_timecounter_update(&ptp->tx_tstamp_tc, tx_tstamp_cycles);
3604 	*timestamp = rte_ns_to_timespec(ns);
3605 
3606 	return rc;
3607 }
3608 
3609 static int
3610 bnxt_timesync_adjust_time(struct rte_eth_dev *dev, int64_t delta)
3611 {
3612 	struct bnxt *bp = dev->data->dev_private;
3613 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3614 
3615 	if (!ptp)
3616 		return -ENOTSUP;
3617 
3618 	ptp->tc.nsec += delta;
3619 	ptp->tx_tstamp_tc.nsec += delta;
3620 	ptp->rx_tstamp_tc.nsec += delta;
3621 
3622 	return 0;
3623 }
3624 
3625 static int
3626 bnxt_get_eeprom_length_op(struct rte_eth_dev *dev)
3627 {
3628 	struct bnxt *bp = dev->data->dev_private;
3629 	int rc;
3630 	uint32_t dir_entries;
3631 	uint32_t entry_length;
3632 
3633 	rc = is_bnxt_in_error(bp);
3634 	if (rc)
3635 		return rc;
3636 
3637 	PMD_DRV_LOG(INFO, PCI_PRI_FMT "\n",
3638 		    bp->pdev->addr.domain, bp->pdev->addr.bus,
3639 		    bp->pdev->addr.devid, bp->pdev->addr.function);
3640 
3641 	rc = bnxt_hwrm_nvm_get_dir_info(bp, &dir_entries, &entry_length);
3642 	if (rc != 0)
3643 		return rc;
3644 
3645 	return dir_entries * entry_length;
3646 }
3647 
3648 static int
3649 bnxt_get_eeprom_op(struct rte_eth_dev *dev,
3650 		struct rte_dev_eeprom_info *in_eeprom)
3651 {
3652 	struct bnxt *bp = dev->data->dev_private;
3653 	uint32_t index;
3654 	uint32_t offset;
3655 	int rc;
3656 
3657 	rc = is_bnxt_in_error(bp);
3658 	if (rc)
3659 		return rc;
3660 
3661 	PMD_DRV_LOG(INFO, PCI_PRI_FMT " in_eeprom->offset = %d len = %d\n",
3662 		    bp->pdev->addr.domain, bp->pdev->addr.bus,
3663 		    bp->pdev->addr.devid, bp->pdev->addr.function,
3664 		    in_eeprom->offset, in_eeprom->length);
3665 
3666 	if (in_eeprom->offset == 0) /* special offset value to get directory */
3667 		return bnxt_get_nvram_directory(bp, in_eeprom->length,
3668 						in_eeprom->data);
3669 
3670 	index = in_eeprom->offset >> 24;
3671 	offset = in_eeprom->offset & 0xffffff;
3672 
3673 	if (index != 0)
3674 		return bnxt_hwrm_get_nvram_item(bp, index - 1, offset,
3675 					   in_eeprom->length, in_eeprom->data);
3676 
3677 	return 0;
3678 }
3679 
3680 static bool bnxt_dir_type_is_ape_bin_format(uint16_t dir_type)
3681 {
3682 	switch (dir_type) {
3683 	case BNX_DIR_TYPE_CHIMP_PATCH:
3684 	case BNX_DIR_TYPE_BOOTCODE:
3685 	case BNX_DIR_TYPE_BOOTCODE_2:
3686 	case BNX_DIR_TYPE_APE_FW:
3687 	case BNX_DIR_TYPE_APE_PATCH:
3688 	case BNX_DIR_TYPE_KONG_FW:
3689 	case BNX_DIR_TYPE_KONG_PATCH:
3690 	case BNX_DIR_TYPE_BONO_FW:
3691 	case BNX_DIR_TYPE_BONO_PATCH:
3692 		/* FALLTHROUGH */
3693 		return true;
3694 	}
3695 
3696 	return false;
3697 }
3698 
3699 static bool bnxt_dir_type_is_other_exec_format(uint16_t dir_type)
3700 {
3701 	switch (dir_type) {
3702 	case BNX_DIR_TYPE_AVS:
3703 	case BNX_DIR_TYPE_EXP_ROM_MBA:
3704 	case BNX_DIR_TYPE_PCIE:
3705 	case BNX_DIR_TYPE_TSCF_UCODE:
3706 	case BNX_DIR_TYPE_EXT_PHY:
3707 	case BNX_DIR_TYPE_CCM:
3708 	case BNX_DIR_TYPE_ISCSI_BOOT:
3709 	case BNX_DIR_TYPE_ISCSI_BOOT_IPV6:
3710 	case BNX_DIR_TYPE_ISCSI_BOOT_IPV4N6:
3711 		/* FALLTHROUGH */
3712 		return true;
3713 	}
3714 
3715 	return false;
3716 }
3717 
3718 static bool bnxt_dir_type_is_executable(uint16_t dir_type)
3719 {
3720 	return bnxt_dir_type_is_ape_bin_format(dir_type) ||
3721 		bnxt_dir_type_is_other_exec_format(dir_type);
3722 }
3723 
3724 static int
3725 bnxt_set_eeprom_op(struct rte_eth_dev *dev,
3726 		struct rte_dev_eeprom_info *in_eeprom)
3727 {
3728 	struct bnxt *bp = dev->data->dev_private;
3729 	uint8_t index, dir_op;
3730 	uint16_t type, ext, ordinal, attr;
3731 	int rc;
3732 
3733 	rc = is_bnxt_in_error(bp);
3734 	if (rc)
3735 		return rc;
3736 
3737 	PMD_DRV_LOG(INFO, PCI_PRI_FMT " in_eeprom->offset = %d len = %d\n",
3738 		    bp->pdev->addr.domain, bp->pdev->addr.bus,
3739 		    bp->pdev->addr.devid, bp->pdev->addr.function,
3740 		    in_eeprom->offset, in_eeprom->length);
3741 
3742 	if (!BNXT_PF(bp)) {
3743 		PMD_DRV_LOG(ERR, "NVM write not supported from a VF\n");
3744 		return -EINVAL;
3745 	}
3746 
3747 	type = in_eeprom->magic >> 16;
3748 
3749 	if (type == 0xffff) { /* special value for directory operations */
3750 		index = in_eeprom->magic & 0xff;
3751 		dir_op = in_eeprom->magic >> 8;
3752 		if (index == 0)
3753 			return -EINVAL;
3754 		switch (dir_op) {
3755 		case 0x0e: /* erase */
3756 			if (in_eeprom->offset != ~in_eeprom->magic)
3757 				return -EINVAL;
3758 			return bnxt_hwrm_erase_nvram_directory(bp, index - 1);
3759 		default:
3760 			return -EINVAL;
3761 		}
3762 	}
3763 
3764 	/* Create or re-write an NVM item: */
3765 	if (bnxt_dir_type_is_executable(type) == true)
3766 		return -EOPNOTSUPP;
3767 	ext = in_eeprom->magic & 0xffff;
3768 	ordinal = in_eeprom->offset >> 16;
3769 	attr = in_eeprom->offset & 0xffff;
3770 
3771 	return bnxt_hwrm_flash_nvram(bp, type, ordinal, ext, attr,
3772 				     in_eeprom->data, in_eeprom->length);
3773 }
3774 
3775 /*
3776  * Initialization
3777  */
3778 
3779 static const struct eth_dev_ops bnxt_dev_ops = {
3780 	.dev_infos_get = bnxt_dev_info_get_op,
3781 	.dev_close = bnxt_dev_close_op,
3782 	.dev_configure = bnxt_dev_configure_op,
3783 	.dev_start = bnxt_dev_start_op,
3784 	.dev_stop = bnxt_dev_stop_op,
3785 	.dev_set_link_up = bnxt_dev_set_link_up_op,
3786 	.dev_set_link_down = bnxt_dev_set_link_down_op,
3787 	.stats_get = bnxt_stats_get_op,
3788 	.stats_reset = bnxt_stats_reset_op,
3789 	.rx_queue_setup = bnxt_rx_queue_setup_op,
3790 	.rx_queue_release = bnxt_rx_queue_release_op,
3791 	.tx_queue_setup = bnxt_tx_queue_setup_op,
3792 	.tx_queue_release = bnxt_tx_queue_release_op,
3793 	.rx_queue_intr_enable = bnxt_rx_queue_intr_enable_op,
3794 	.rx_queue_intr_disable = bnxt_rx_queue_intr_disable_op,
3795 	.reta_update = bnxt_reta_update_op,
3796 	.reta_query = bnxt_reta_query_op,
3797 	.rss_hash_update = bnxt_rss_hash_update_op,
3798 	.rss_hash_conf_get = bnxt_rss_hash_conf_get_op,
3799 	.link_update = bnxt_link_update_op,
3800 	.promiscuous_enable = bnxt_promiscuous_enable_op,
3801 	.promiscuous_disable = bnxt_promiscuous_disable_op,
3802 	.allmulticast_enable = bnxt_allmulticast_enable_op,
3803 	.allmulticast_disable = bnxt_allmulticast_disable_op,
3804 	.mac_addr_add = bnxt_mac_addr_add_op,
3805 	.mac_addr_remove = bnxt_mac_addr_remove_op,
3806 	.flow_ctrl_get = bnxt_flow_ctrl_get_op,
3807 	.flow_ctrl_set = bnxt_flow_ctrl_set_op,
3808 	.udp_tunnel_port_add  = bnxt_udp_tunnel_port_add_op,
3809 	.udp_tunnel_port_del  = bnxt_udp_tunnel_port_del_op,
3810 	.vlan_filter_set = bnxt_vlan_filter_set_op,
3811 	.vlan_offload_set = bnxt_vlan_offload_set_op,
3812 	.vlan_tpid_set = bnxt_vlan_tpid_set_op,
3813 	.vlan_pvid_set = bnxt_vlan_pvid_set_op,
3814 	.mtu_set = bnxt_mtu_set_op,
3815 	.mac_addr_set = bnxt_set_default_mac_addr_op,
3816 	.xstats_get = bnxt_dev_xstats_get_op,
3817 	.xstats_get_names = bnxt_dev_xstats_get_names_op,
3818 	.xstats_reset = bnxt_dev_xstats_reset_op,
3819 	.fw_version_get = bnxt_fw_version_get,
3820 	.set_mc_addr_list = bnxt_dev_set_mc_addr_list_op,
3821 	.rxq_info_get = bnxt_rxq_info_get_op,
3822 	.txq_info_get = bnxt_txq_info_get_op,
3823 	.rx_burst_mode_get = bnxt_rx_burst_mode_get,
3824 	.tx_burst_mode_get = bnxt_tx_burst_mode_get,
3825 	.dev_led_on = bnxt_dev_led_on_op,
3826 	.dev_led_off = bnxt_dev_led_off_op,
3827 	.rx_queue_start = bnxt_rx_queue_start,
3828 	.rx_queue_stop = bnxt_rx_queue_stop,
3829 	.tx_queue_start = bnxt_tx_queue_start,
3830 	.tx_queue_stop = bnxt_tx_queue_stop,
3831 	.flow_ops_get = bnxt_flow_ops_get_op,
3832 	.dev_supported_ptypes_get = bnxt_dev_supported_ptypes_get_op,
3833 	.get_eeprom_length    = bnxt_get_eeprom_length_op,
3834 	.get_eeprom           = bnxt_get_eeprom_op,
3835 	.set_eeprom           = bnxt_set_eeprom_op,
3836 	.timesync_enable      = bnxt_timesync_enable,
3837 	.timesync_disable     = bnxt_timesync_disable,
3838 	.timesync_read_time   = bnxt_timesync_read_time,
3839 	.timesync_write_time   = bnxt_timesync_write_time,
3840 	.timesync_adjust_time = bnxt_timesync_adjust_time,
3841 	.timesync_read_rx_timestamp = bnxt_timesync_read_rx_timestamp,
3842 	.timesync_read_tx_timestamp = bnxt_timesync_read_tx_timestamp,
3843 };
3844 
3845 static uint32_t bnxt_map_reset_regs(struct bnxt *bp, uint32_t reg)
3846 {
3847 	uint32_t offset;
3848 
3849 	/* Only pre-map the reset GRC registers using window 3 */
3850 	rte_write32(reg & 0xfffff000, (uint8_t *)bp->bar0 +
3851 		    BNXT_GRCPF_REG_WINDOW_BASE_OUT + 8);
3852 
3853 	offset = BNXT_GRCP_WINDOW_3_BASE + (reg & 0xffc);
3854 
3855 	return offset;
3856 }
3857 
3858 int bnxt_map_fw_health_status_regs(struct bnxt *bp)
3859 {
3860 	struct bnxt_error_recovery_info *info = bp->recovery_info;
3861 	uint32_t reg_base = 0xffffffff;
3862 	int i;
3863 
3864 	/* Only pre-map the monitoring GRC registers using window 2 */
3865 	for (i = 0; i < BNXT_FW_STATUS_REG_CNT; i++) {
3866 		uint32_t reg = info->status_regs[i];
3867 
3868 		if (BNXT_FW_STATUS_REG_TYPE(reg) != BNXT_FW_STATUS_REG_TYPE_GRC)
3869 			continue;
3870 
3871 		if (reg_base == 0xffffffff)
3872 			reg_base = reg & 0xfffff000;
3873 		if ((reg & 0xfffff000) != reg_base)
3874 			return -ERANGE;
3875 
3876 		/* Use mask 0xffc as the Lower 2 bits indicates
3877 		 * address space location
3878 		 */
3879 		info->mapped_status_regs[i] = BNXT_GRCP_WINDOW_2_BASE +
3880 						(reg & 0xffc);
3881 	}
3882 
3883 	if (reg_base == 0xffffffff)
3884 		return 0;
3885 
3886 	rte_write32(reg_base, (uint8_t *)bp->bar0 +
3887 		    BNXT_GRCPF_REG_WINDOW_BASE_OUT + 4);
3888 
3889 	return 0;
3890 }
3891 
3892 static void bnxt_write_fw_reset_reg(struct bnxt *bp, uint32_t index)
3893 {
3894 	struct bnxt_error_recovery_info *info = bp->recovery_info;
3895 	uint32_t delay = info->delay_after_reset[index];
3896 	uint32_t val = info->reset_reg_val[index];
3897 	uint32_t reg = info->reset_reg[index];
3898 	uint32_t type, offset;
3899 	int ret;
3900 
3901 	type = BNXT_FW_STATUS_REG_TYPE(reg);
3902 	offset = BNXT_FW_STATUS_REG_OFF(reg);
3903 
3904 	switch (type) {
3905 	case BNXT_FW_STATUS_REG_TYPE_CFG:
3906 		ret = rte_pci_write_config(bp->pdev, &val, sizeof(val), offset);
3907 		if (ret < 0) {
3908 			PMD_DRV_LOG(ERR, "Failed to write %#x at PCI offset %#x",
3909 				    val, offset);
3910 			return;
3911 		}
3912 		break;
3913 	case BNXT_FW_STATUS_REG_TYPE_GRC:
3914 		offset = bnxt_map_reset_regs(bp, offset);
3915 		rte_write32(val, (uint8_t *)bp->bar0 + offset);
3916 		break;
3917 	case BNXT_FW_STATUS_REG_TYPE_BAR0:
3918 		rte_write32(val, (uint8_t *)bp->bar0 + offset);
3919 		break;
3920 	}
3921 	/* wait on a specific interval of time until core reset is complete */
3922 	if (delay)
3923 		rte_delay_ms(delay);
3924 }
3925 
3926 static void bnxt_dev_cleanup(struct bnxt *bp)
3927 {
3928 	bp->eth_dev->data->dev_link.link_status = 0;
3929 	bp->link_info->link_up = 0;
3930 	if (bp->eth_dev->data->dev_started)
3931 		bnxt_dev_stop(bp->eth_dev);
3932 
3933 	bnxt_uninit_resources(bp, true);
3934 }
3935 
3936 static int
3937 bnxt_check_fw_reset_done(struct bnxt *bp)
3938 {
3939 	int timeout = bp->fw_reset_max_msecs;
3940 	uint16_t val = 0;
3941 	int rc;
3942 
3943 	do {
3944 		rc = rte_pci_read_config(bp->pdev, &val, sizeof(val), PCI_SUBSYSTEM_ID_OFFSET);
3945 		if (rc < 0) {
3946 			PMD_DRV_LOG(ERR, "Failed to read PCI offset 0x%x", PCI_SUBSYSTEM_ID_OFFSET);
3947 			return rc;
3948 		}
3949 		if (val != 0xffff)
3950 			break;
3951 		rte_delay_ms(1);
3952 	} while (timeout--);
3953 
3954 	if (val == 0xffff) {
3955 		PMD_DRV_LOG(ERR, "Firmware reset aborted, PCI config space invalid\n");
3956 		return -1;
3957 	}
3958 
3959 	return 0;
3960 }
3961 
3962 static int bnxt_restore_vlan_filters(struct bnxt *bp)
3963 {
3964 	struct rte_eth_dev *dev = bp->eth_dev;
3965 	struct rte_vlan_filter_conf *vfc;
3966 	int vidx, vbit, rc;
3967 	uint16_t vlan_id;
3968 
3969 	for (vlan_id = 1; vlan_id <= RTE_ETHER_MAX_VLAN_ID; vlan_id++) {
3970 		vfc = &dev->data->vlan_filter_conf;
3971 		vidx = vlan_id / 64;
3972 		vbit = vlan_id % 64;
3973 
3974 		/* Each bit corresponds to a VLAN id */
3975 		if (vfc->ids[vidx] & (UINT64_C(1) << vbit)) {
3976 			rc = bnxt_add_vlan_filter(bp, vlan_id);
3977 			if (rc)
3978 				return rc;
3979 		}
3980 	}
3981 
3982 	return 0;
3983 }
3984 
3985 static int bnxt_restore_mac_filters(struct bnxt *bp)
3986 {
3987 	struct rte_eth_dev *dev = bp->eth_dev;
3988 	struct rte_eth_dev_info dev_info;
3989 	struct rte_ether_addr *addr;
3990 	uint64_t pool_mask;
3991 	uint32_t pool = 0;
3992 	uint32_t i;
3993 	int rc;
3994 
3995 	if (BNXT_VF(bp) && !BNXT_VF_IS_TRUSTED(bp))
3996 		return 0;
3997 
3998 	rc = bnxt_dev_info_get_op(dev, &dev_info);
3999 	if (rc)
4000 		return rc;
4001 
4002 	/* replay MAC address configuration */
4003 	for (i = 1; i < dev_info.max_mac_addrs; i++) {
4004 		addr = &dev->data->mac_addrs[i];
4005 
4006 		/* skip zero address */
4007 		if (rte_is_zero_ether_addr(addr))
4008 			continue;
4009 
4010 		pool = 0;
4011 		pool_mask = dev->data->mac_pool_sel[i];
4012 
4013 		do {
4014 			if (pool_mask & 1ULL) {
4015 				rc = bnxt_mac_addr_add_op(dev, addr, i, pool);
4016 				if (rc)
4017 					return rc;
4018 			}
4019 			pool_mask >>= 1;
4020 			pool++;
4021 		} while (pool_mask);
4022 	}
4023 
4024 	return 0;
4025 }
4026 
4027 static int bnxt_restore_filters(struct bnxt *bp)
4028 {
4029 	struct rte_eth_dev *dev = bp->eth_dev;
4030 	int ret = 0;
4031 
4032 	if (dev->data->all_multicast) {
4033 		ret = bnxt_allmulticast_enable_op(dev);
4034 		if (ret)
4035 			return ret;
4036 	}
4037 	if (dev->data->promiscuous) {
4038 		ret = bnxt_promiscuous_enable_op(dev);
4039 		if (ret)
4040 			return ret;
4041 	}
4042 
4043 	ret = bnxt_restore_mac_filters(bp);
4044 	if (ret)
4045 		return ret;
4046 
4047 	ret = bnxt_restore_vlan_filters(bp);
4048 	/* TODO restore other filters as well */
4049 	return ret;
4050 }
4051 
4052 static int bnxt_check_fw_ready(struct bnxt *bp)
4053 {
4054 	int timeout = bp->fw_reset_max_msecs;
4055 	int rc = 0;
4056 
4057 	do {
4058 		rc = bnxt_hwrm_poll_ver_get(bp);
4059 		if (rc == 0)
4060 			break;
4061 		rte_delay_ms(BNXT_FW_READY_WAIT_INTERVAL);
4062 		timeout -= BNXT_FW_READY_WAIT_INTERVAL;
4063 	} while (rc && timeout > 0);
4064 
4065 	if (rc)
4066 		PMD_DRV_LOG(ERR, "FW is not Ready after reset\n");
4067 
4068 	return rc;
4069 }
4070 
4071 static void bnxt_dev_recover(void *arg)
4072 {
4073 	struct bnxt *bp = arg;
4074 	int rc = 0;
4075 
4076 	pthread_mutex_lock(&bp->err_recovery_lock);
4077 
4078 	if (!bp->fw_reset_min_msecs) {
4079 		rc = bnxt_check_fw_reset_done(bp);
4080 		if (rc)
4081 			goto err;
4082 	}
4083 
4084 	/* Clear Error flag so that device re-init should happen */
4085 	bp->flags &= ~BNXT_FLAG_FATAL_ERROR;
4086 
4087 	rc = bnxt_check_fw_ready(bp);
4088 	if (rc)
4089 		goto err;
4090 
4091 	rc = bnxt_init_resources(bp, true);
4092 	if (rc) {
4093 		PMD_DRV_LOG(ERR,
4094 			    "Failed to initialize resources after reset\n");
4095 		goto err;
4096 	}
4097 	/* clear reset flag as the device is initialized now */
4098 	bp->flags &= ~BNXT_FLAG_FW_RESET;
4099 
4100 	rc = bnxt_dev_start_op(bp->eth_dev);
4101 	if (rc) {
4102 		PMD_DRV_LOG(ERR, "Failed to start port after reset\n");
4103 		goto err_start;
4104 	}
4105 
4106 	rc = bnxt_restore_filters(bp);
4107 	if (rc)
4108 		goto err_start;
4109 
4110 	PMD_DRV_LOG(INFO, "Recovered from FW reset\n");
4111 	pthread_mutex_unlock(&bp->err_recovery_lock);
4112 
4113 	return;
4114 err_start:
4115 	bnxt_dev_stop(bp->eth_dev);
4116 err:
4117 	bp->flags |= BNXT_FLAG_FATAL_ERROR;
4118 	bnxt_uninit_resources(bp, false);
4119 	pthread_mutex_unlock(&bp->err_recovery_lock);
4120 	PMD_DRV_LOG(ERR, "Failed to recover from FW reset\n");
4121 }
4122 
4123 void bnxt_dev_reset_and_resume(void *arg)
4124 {
4125 	struct bnxt *bp = arg;
4126 	uint32_t us = US_PER_MS * bp->fw_reset_min_msecs;
4127 	uint16_t val = 0;
4128 	int rc;
4129 
4130 	bnxt_dev_cleanup(bp);
4131 
4132 	bnxt_wait_for_device_shutdown(bp);
4133 
4134 	/* During some fatal firmware error conditions, the PCI config space
4135 	 * register 0x2e which normally contains the subsystem ID will become
4136 	 * 0xffff. This register will revert back to the normal value after
4137 	 * the chip has completed core reset. If we detect this condition,
4138 	 * we can poll this config register immediately for the value to revert.
4139 	 */
4140 	if (bp->flags & BNXT_FLAG_FATAL_ERROR) {
4141 		rc = rte_pci_read_config(bp->pdev, &val, sizeof(val), PCI_SUBSYSTEM_ID_OFFSET);
4142 		if (rc < 0) {
4143 			PMD_DRV_LOG(ERR, "Failed to read PCI offset 0x%x", PCI_SUBSYSTEM_ID_OFFSET);
4144 			return;
4145 		}
4146 		if (val == 0xffff) {
4147 			bp->fw_reset_min_msecs = 0;
4148 			us = 1;
4149 		}
4150 	}
4151 
4152 	rc = rte_eal_alarm_set(us, bnxt_dev_recover, (void *)bp);
4153 	if (rc)
4154 		PMD_DRV_LOG(ERR, "Error setting recovery alarm");
4155 }
4156 
4157 uint32_t bnxt_read_fw_status_reg(struct bnxt *bp, uint32_t index)
4158 {
4159 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4160 	uint32_t reg = info->status_regs[index];
4161 	uint32_t type, offset, val = 0;
4162 	int ret = 0;
4163 
4164 	type = BNXT_FW_STATUS_REG_TYPE(reg);
4165 	offset = BNXT_FW_STATUS_REG_OFF(reg);
4166 
4167 	switch (type) {
4168 	case BNXT_FW_STATUS_REG_TYPE_CFG:
4169 		ret = rte_pci_read_config(bp->pdev, &val, sizeof(val), offset);
4170 		if (ret < 0)
4171 			PMD_DRV_LOG(ERR, "Failed to read PCI offset %#x",
4172 				    offset);
4173 		break;
4174 	case BNXT_FW_STATUS_REG_TYPE_GRC:
4175 		offset = info->mapped_status_regs[index];
4176 		/* FALLTHROUGH */
4177 	case BNXT_FW_STATUS_REG_TYPE_BAR0:
4178 		val = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
4179 				       offset));
4180 		break;
4181 	}
4182 
4183 	return val;
4184 }
4185 
4186 static int bnxt_fw_reset_all(struct bnxt *bp)
4187 {
4188 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4189 	uint32_t i;
4190 	int rc = 0;
4191 
4192 	if (info->flags & BNXT_FLAG_ERROR_RECOVERY_HOST) {
4193 		/* Reset through master function driver */
4194 		for (i = 0; i < info->reg_array_cnt; i++)
4195 			bnxt_write_fw_reset_reg(bp, i);
4196 		/* Wait for time specified by FW after triggering reset */
4197 		rte_delay_ms(info->master_func_wait_period_after_reset);
4198 	} else if (info->flags & BNXT_FLAG_ERROR_RECOVERY_CO_CPU) {
4199 		/* Reset with the help of Kong processor */
4200 		rc = bnxt_hwrm_fw_reset(bp);
4201 		if (rc)
4202 			PMD_DRV_LOG(ERR, "Failed to reset FW\n");
4203 	}
4204 
4205 	return rc;
4206 }
4207 
4208 static void bnxt_fw_reset_cb(void *arg)
4209 {
4210 	struct bnxt *bp = arg;
4211 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4212 	int rc = 0;
4213 
4214 	/* Only Master function can do FW reset */
4215 	if (bnxt_is_master_func(bp) &&
4216 	    bnxt_is_recovery_enabled(bp)) {
4217 		rc = bnxt_fw_reset_all(bp);
4218 		if (rc) {
4219 			PMD_DRV_LOG(ERR, "Adapter recovery failed\n");
4220 			return;
4221 		}
4222 	}
4223 
4224 	/* if recovery method is ERROR_RECOVERY_CO_CPU, KONG will send
4225 	 * EXCEPTION_FATAL_ASYNC event to all the functions
4226 	 * (including MASTER FUNC). After receiving this Async, all the active
4227 	 * drivers should treat this case as FW initiated recovery
4228 	 */
4229 	if (info->flags & BNXT_FLAG_ERROR_RECOVERY_HOST) {
4230 		bp->fw_reset_min_msecs = BNXT_MIN_FW_READY_TIMEOUT;
4231 		bp->fw_reset_max_msecs = BNXT_MAX_FW_RESET_TIMEOUT;
4232 
4233 		/* To recover from error */
4234 		rte_eal_alarm_set(US_PER_MS, bnxt_dev_reset_and_resume,
4235 				  (void *)bp);
4236 	}
4237 }
4238 
4239 /* Driver should poll FW heartbeat, reset_counter with the frequency
4240  * advertised by FW in HWRM_ERROR_RECOVERY_QCFG.
4241  * When the driver detects heartbeat stop or change in reset_counter,
4242  * it has to trigger a reset to recover from the error condition.
4243  * A “master PF” is the function who will have the privilege to
4244  * initiate the chimp reset. The master PF will be elected by the
4245  * firmware and will be notified through async message.
4246  */
4247 static void bnxt_check_fw_health(void *arg)
4248 {
4249 	struct bnxt *bp = arg;
4250 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4251 	uint32_t val = 0, wait_msec;
4252 
4253 	if (!info || !bnxt_is_recovery_enabled(bp) ||
4254 	    is_bnxt_in_error(bp))
4255 		return;
4256 
4257 	val = bnxt_read_fw_status_reg(bp, BNXT_FW_HEARTBEAT_CNT_REG);
4258 	if (val == info->last_heart_beat)
4259 		goto reset;
4260 
4261 	info->last_heart_beat = val;
4262 
4263 	val = bnxt_read_fw_status_reg(bp, BNXT_FW_RECOVERY_CNT_REG);
4264 	if (val != info->last_reset_counter)
4265 		goto reset;
4266 
4267 	info->last_reset_counter = val;
4268 
4269 	rte_eal_alarm_set(US_PER_MS * info->driver_polling_freq,
4270 			  bnxt_check_fw_health, (void *)bp);
4271 
4272 	return;
4273 reset:
4274 	/* Stop DMA to/from device */
4275 	bp->flags |= BNXT_FLAG_FATAL_ERROR;
4276 	bp->flags |= BNXT_FLAG_FW_RESET;
4277 
4278 	bnxt_stop_rxtx(bp);
4279 
4280 	PMD_DRV_LOG(ERR, "Detected FW dead condition\n");
4281 
4282 	if (bnxt_is_master_func(bp))
4283 		wait_msec = info->master_func_wait_period;
4284 	else
4285 		wait_msec = info->normal_func_wait_period;
4286 
4287 	rte_eal_alarm_set(US_PER_MS * wait_msec,
4288 			  bnxt_fw_reset_cb, (void *)bp);
4289 }
4290 
4291 void bnxt_schedule_fw_health_check(struct bnxt *bp)
4292 {
4293 	uint32_t polling_freq;
4294 
4295 	pthread_mutex_lock(&bp->health_check_lock);
4296 
4297 	if (!bnxt_is_recovery_enabled(bp))
4298 		goto done;
4299 
4300 	if (bp->flags & BNXT_FLAG_FW_HEALTH_CHECK_SCHEDULED)
4301 		goto done;
4302 
4303 	polling_freq = bp->recovery_info->driver_polling_freq;
4304 
4305 	rte_eal_alarm_set(US_PER_MS * polling_freq,
4306 			  bnxt_check_fw_health, (void *)bp);
4307 	bp->flags |= BNXT_FLAG_FW_HEALTH_CHECK_SCHEDULED;
4308 
4309 done:
4310 	pthread_mutex_unlock(&bp->health_check_lock);
4311 }
4312 
4313 static void bnxt_cancel_fw_health_check(struct bnxt *bp)
4314 {
4315 	rte_eal_alarm_cancel(bnxt_check_fw_health, (void *)bp);
4316 	bp->flags &= ~BNXT_FLAG_FW_HEALTH_CHECK_SCHEDULED;
4317 }
4318 
4319 static bool bnxt_vf_pciid(uint16_t device_id)
4320 {
4321 	switch (device_id) {
4322 	case BROADCOM_DEV_ID_57304_VF:
4323 	case BROADCOM_DEV_ID_57406_VF:
4324 	case BROADCOM_DEV_ID_5731X_VF:
4325 	case BROADCOM_DEV_ID_5741X_VF:
4326 	case BROADCOM_DEV_ID_57414_VF:
4327 	case BROADCOM_DEV_ID_STRATUS_NIC_VF1:
4328 	case BROADCOM_DEV_ID_STRATUS_NIC_VF2:
4329 	case BROADCOM_DEV_ID_58802_VF:
4330 	case BROADCOM_DEV_ID_57500_VF1:
4331 	case BROADCOM_DEV_ID_57500_VF2:
4332 	case BROADCOM_DEV_ID_58818_VF:
4333 		/* FALLTHROUGH */
4334 		return true;
4335 	default:
4336 		return false;
4337 	}
4338 }
4339 
4340 /* Phase 5 device */
4341 static bool bnxt_p5_device(uint16_t device_id)
4342 {
4343 	switch (device_id) {
4344 	case BROADCOM_DEV_ID_57508:
4345 	case BROADCOM_DEV_ID_57504:
4346 	case BROADCOM_DEV_ID_57502:
4347 	case BROADCOM_DEV_ID_57508_MF1:
4348 	case BROADCOM_DEV_ID_57504_MF1:
4349 	case BROADCOM_DEV_ID_57502_MF1:
4350 	case BROADCOM_DEV_ID_57508_MF2:
4351 	case BROADCOM_DEV_ID_57504_MF2:
4352 	case BROADCOM_DEV_ID_57502_MF2:
4353 	case BROADCOM_DEV_ID_57500_VF1:
4354 	case BROADCOM_DEV_ID_57500_VF2:
4355 	case BROADCOM_DEV_ID_58812:
4356 	case BROADCOM_DEV_ID_58814:
4357 	case BROADCOM_DEV_ID_58818:
4358 	case BROADCOM_DEV_ID_58818_VF:
4359 		/* FALLTHROUGH */
4360 		return true;
4361 	default:
4362 		return false;
4363 	}
4364 }
4365 
4366 bool bnxt_stratus_device(struct bnxt *bp)
4367 {
4368 	uint16_t device_id = bp->pdev->id.device_id;
4369 
4370 	switch (device_id) {
4371 	case BROADCOM_DEV_ID_STRATUS_NIC:
4372 	case BROADCOM_DEV_ID_STRATUS_NIC_VF1:
4373 	case BROADCOM_DEV_ID_STRATUS_NIC_VF2:
4374 		/* FALLTHROUGH */
4375 		return true;
4376 	default:
4377 		return false;
4378 	}
4379 }
4380 
4381 static int bnxt_map_pci_bars(struct rte_eth_dev *eth_dev)
4382 {
4383 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
4384 	struct bnxt *bp = eth_dev->data->dev_private;
4385 
4386 	/* enable device (incl. PCI PM wakeup), and bus-mastering */
4387 	bp->bar0 = (void *)pci_dev->mem_resource[0].addr;
4388 	bp->doorbell_base = (void *)pci_dev->mem_resource[2].addr;
4389 	if (!bp->bar0 || !bp->doorbell_base) {
4390 		PMD_DRV_LOG(ERR, "Unable to access Hardware\n");
4391 		return -ENODEV;
4392 	}
4393 
4394 	bp->eth_dev = eth_dev;
4395 	bp->pdev = pci_dev;
4396 
4397 	return 0;
4398 }
4399 
4400 static int bnxt_alloc_ctx_mem_blk(struct bnxt *bp,
4401 				  struct bnxt_ctx_pg_info *ctx_pg,
4402 				  uint32_t mem_size,
4403 				  const char *suffix,
4404 				  uint16_t idx)
4405 {
4406 	struct bnxt_ring_mem_info *rmem = &ctx_pg->ring_mem;
4407 	const struct rte_memzone *mz = NULL;
4408 	char mz_name[RTE_MEMZONE_NAMESIZE];
4409 	rte_iova_t mz_phys_addr;
4410 	uint64_t valid_bits = 0;
4411 	uint32_t sz;
4412 	int i;
4413 
4414 	if (!mem_size)
4415 		return 0;
4416 
4417 	rmem->nr_pages = RTE_ALIGN_MUL_CEIL(mem_size, BNXT_PAGE_SIZE) /
4418 			 BNXT_PAGE_SIZE;
4419 	rmem->page_size = BNXT_PAGE_SIZE;
4420 	rmem->pg_arr = ctx_pg->ctx_pg_arr;
4421 	rmem->dma_arr = ctx_pg->ctx_dma_arr;
4422 	rmem->flags = BNXT_RMEM_VALID_PTE_FLAG;
4423 
4424 	valid_bits = PTU_PTE_VALID;
4425 
4426 	if (rmem->nr_pages > 1) {
4427 		snprintf(mz_name, RTE_MEMZONE_NAMESIZE,
4428 			 "bnxt_ctx_pg_tbl%s_%x_%d",
4429 			 suffix, idx, bp->eth_dev->data->port_id);
4430 		mz_name[RTE_MEMZONE_NAMESIZE - 1] = 0;
4431 		mz = rte_memzone_lookup(mz_name);
4432 		if (!mz) {
4433 			mz = rte_memzone_reserve_aligned(mz_name,
4434 						rmem->nr_pages * 8,
4435 						SOCKET_ID_ANY,
4436 						RTE_MEMZONE_2MB |
4437 						RTE_MEMZONE_SIZE_HINT_ONLY |
4438 						RTE_MEMZONE_IOVA_CONTIG,
4439 						BNXT_PAGE_SIZE);
4440 			if (mz == NULL)
4441 				return -ENOMEM;
4442 		}
4443 
4444 		memset(mz->addr, 0, mz->len);
4445 		mz_phys_addr = mz->iova;
4446 
4447 		rmem->pg_tbl = mz->addr;
4448 		rmem->pg_tbl_map = mz_phys_addr;
4449 		rmem->pg_tbl_mz = mz;
4450 	}
4451 
4452 	snprintf(mz_name, RTE_MEMZONE_NAMESIZE, "bnxt_ctx_%s_%x_%d",
4453 		 suffix, idx, bp->eth_dev->data->port_id);
4454 	mz = rte_memzone_lookup(mz_name);
4455 	if (!mz) {
4456 		mz = rte_memzone_reserve_aligned(mz_name,
4457 						 mem_size,
4458 						 SOCKET_ID_ANY,
4459 						 RTE_MEMZONE_1GB |
4460 						 RTE_MEMZONE_SIZE_HINT_ONLY |
4461 						 RTE_MEMZONE_IOVA_CONTIG,
4462 						 BNXT_PAGE_SIZE);
4463 		if (mz == NULL)
4464 			return -ENOMEM;
4465 	}
4466 
4467 	memset(mz->addr, 0, mz->len);
4468 	mz_phys_addr = mz->iova;
4469 
4470 	for (sz = 0, i = 0; sz < mem_size; sz += BNXT_PAGE_SIZE, i++) {
4471 		rmem->pg_arr[i] = ((char *)mz->addr) + sz;
4472 		rmem->dma_arr[i] = mz_phys_addr + sz;
4473 
4474 		if (rmem->nr_pages > 1) {
4475 			if (i == rmem->nr_pages - 2 &&
4476 			    (rmem->flags & BNXT_RMEM_RING_PTE_FLAG))
4477 				valid_bits |= PTU_PTE_NEXT_TO_LAST;
4478 			else if (i == rmem->nr_pages - 1 &&
4479 				 (rmem->flags & BNXT_RMEM_RING_PTE_FLAG))
4480 				valid_bits |= PTU_PTE_LAST;
4481 
4482 			rmem->pg_tbl[i] = rte_cpu_to_le_64(rmem->dma_arr[i] |
4483 							   valid_bits);
4484 		}
4485 	}
4486 
4487 	rmem->mz = mz;
4488 	if (rmem->vmem_size)
4489 		rmem->vmem = (void **)mz->addr;
4490 	rmem->dma_arr[0] = mz_phys_addr;
4491 	return 0;
4492 }
4493 
4494 static void bnxt_free_ctx_mem(struct bnxt *bp)
4495 {
4496 	int i;
4497 
4498 	if (!bp->ctx || !(bp->ctx->flags & BNXT_CTX_FLAG_INITED))
4499 		return;
4500 
4501 	bp->ctx->flags &= ~BNXT_CTX_FLAG_INITED;
4502 	rte_memzone_free(bp->ctx->qp_mem.ring_mem.mz);
4503 	rte_memzone_free(bp->ctx->srq_mem.ring_mem.mz);
4504 	rte_memzone_free(bp->ctx->cq_mem.ring_mem.mz);
4505 	rte_memzone_free(bp->ctx->vnic_mem.ring_mem.mz);
4506 	rte_memzone_free(bp->ctx->stat_mem.ring_mem.mz);
4507 	rte_memzone_free(bp->ctx->qp_mem.ring_mem.pg_tbl_mz);
4508 	rte_memzone_free(bp->ctx->srq_mem.ring_mem.pg_tbl_mz);
4509 	rte_memzone_free(bp->ctx->cq_mem.ring_mem.pg_tbl_mz);
4510 	rte_memzone_free(bp->ctx->vnic_mem.ring_mem.pg_tbl_mz);
4511 	rte_memzone_free(bp->ctx->stat_mem.ring_mem.pg_tbl_mz);
4512 
4513 	for (i = 0; i < bp->ctx->tqm_fp_rings_count + 1; i++) {
4514 		if (bp->ctx->tqm_mem[i])
4515 			rte_memzone_free(bp->ctx->tqm_mem[i]->ring_mem.mz);
4516 	}
4517 
4518 	rte_free(bp->ctx);
4519 	bp->ctx = NULL;
4520 }
4521 
4522 #define bnxt_roundup(x, y)   ((((x) + ((y) - 1)) / (y)) * (y))
4523 
4524 #define min_t(type, x, y) ({                    \
4525 	type __min1 = (x);                      \
4526 	type __min2 = (y);                      \
4527 	__min1 < __min2 ? __min1 : __min2; })
4528 
4529 #define max_t(type, x, y) ({                    \
4530 	type __max1 = (x);                      \
4531 	type __max2 = (y);                      \
4532 	__max1 > __max2 ? __max1 : __max2; })
4533 
4534 #define clamp_t(type, _x, min, max)     min_t(type, max_t(type, _x, min), max)
4535 
4536 int bnxt_alloc_ctx_mem(struct bnxt *bp)
4537 {
4538 	struct bnxt_ctx_pg_info *ctx_pg;
4539 	struct bnxt_ctx_mem_info *ctx;
4540 	uint32_t mem_size, ena, entries;
4541 	uint32_t entries_sp, min;
4542 	int i, rc;
4543 
4544 	rc = bnxt_hwrm_func_backing_store_qcaps(bp);
4545 	if (rc) {
4546 		PMD_DRV_LOG(ERR, "Query context mem capability failed\n");
4547 		return rc;
4548 	}
4549 	ctx = bp->ctx;
4550 	if (!ctx || (ctx->flags & BNXT_CTX_FLAG_INITED))
4551 		return 0;
4552 
4553 	ctx_pg = &ctx->qp_mem;
4554 	ctx_pg->entries = ctx->qp_min_qp1_entries + ctx->qp_max_l2_entries;
4555 	if (ctx->qp_entry_size) {
4556 		mem_size = ctx->qp_entry_size * ctx_pg->entries;
4557 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "qp_mem", 0);
4558 		if (rc)
4559 			return rc;
4560 	}
4561 
4562 	ctx_pg = &ctx->srq_mem;
4563 	ctx_pg->entries = ctx->srq_max_l2_entries;
4564 	if (ctx->srq_entry_size) {
4565 		mem_size = ctx->srq_entry_size * ctx_pg->entries;
4566 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "srq_mem", 0);
4567 		if (rc)
4568 			return rc;
4569 	}
4570 
4571 	ctx_pg = &ctx->cq_mem;
4572 	ctx_pg->entries = ctx->cq_max_l2_entries;
4573 	if (ctx->cq_entry_size) {
4574 		mem_size = ctx->cq_entry_size * ctx_pg->entries;
4575 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "cq_mem", 0);
4576 		if (rc)
4577 			return rc;
4578 	}
4579 
4580 	ctx_pg = &ctx->vnic_mem;
4581 	ctx_pg->entries = ctx->vnic_max_vnic_entries +
4582 		ctx->vnic_max_ring_table_entries;
4583 	if (ctx->vnic_entry_size) {
4584 		mem_size = ctx->vnic_entry_size * ctx_pg->entries;
4585 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "vnic_mem", 0);
4586 		if (rc)
4587 			return rc;
4588 	}
4589 
4590 	ctx_pg = &ctx->stat_mem;
4591 	ctx_pg->entries = ctx->stat_max_entries;
4592 	if (ctx->stat_entry_size) {
4593 		mem_size = ctx->stat_entry_size * ctx_pg->entries;
4594 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "stat_mem", 0);
4595 		if (rc)
4596 			return rc;
4597 	}
4598 
4599 	min = ctx->tqm_min_entries_per_ring;
4600 
4601 	entries_sp = ctx->qp_max_l2_entries +
4602 		     ctx->vnic_max_vnic_entries +
4603 		     2 * ctx->qp_min_qp1_entries + min;
4604 	entries_sp = bnxt_roundup(entries_sp, ctx->tqm_entries_multiple);
4605 
4606 	entries = ctx->qp_max_l2_entries + ctx->qp_min_qp1_entries;
4607 	entries = bnxt_roundup(entries, ctx->tqm_entries_multiple);
4608 	entries = clamp_t(uint32_t, entries, min,
4609 			  ctx->tqm_max_entries_per_ring);
4610 	for (i = 0, ena = 0; i < ctx->tqm_fp_rings_count + 1; i++) {
4611 		/* i=0 is for TQM_SP. i=1 to i=8 applies to RING0 to RING7.
4612 		 * i > 8 is other ext rings.
4613 		 */
4614 		ctx_pg = ctx->tqm_mem[i];
4615 		ctx_pg->entries = i ? entries : entries_sp;
4616 		if (ctx->tqm_entry_size) {
4617 			mem_size = ctx->tqm_entry_size * ctx_pg->entries;
4618 			rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size,
4619 						    "tqm_mem", i);
4620 			if (rc)
4621 				return rc;
4622 		}
4623 		if (i < BNXT_MAX_TQM_LEGACY_RINGS)
4624 			ena |= HWRM_FUNC_BACKING_STORE_CFG_INPUT_ENABLES_TQM_SP << i;
4625 		else
4626 			ena |= HWRM_FUNC_BACKING_STORE_CFG_INPUT_ENABLES_TQM_RING8;
4627 	}
4628 
4629 	ena |= FUNC_BACKING_STORE_CFG_INPUT_DFLT_ENABLES;
4630 	rc = bnxt_hwrm_func_backing_store_cfg(bp, ena);
4631 	if (rc)
4632 		PMD_DRV_LOG(ERR,
4633 			    "Failed to configure context mem: rc = %d\n", rc);
4634 	else
4635 		ctx->flags |= BNXT_CTX_FLAG_INITED;
4636 
4637 	return rc;
4638 }
4639 
4640 static int bnxt_alloc_stats_mem(struct bnxt *bp)
4641 {
4642 	struct rte_pci_device *pci_dev = bp->pdev;
4643 	char mz_name[RTE_MEMZONE_NAMESIZE];
4644 	const struct rte_memzone *mz = NULL;
4645 	uint32_t total_alloc_len;
4646 	rte_iova_t mz_phys_addr;
4647 
4648 	if (pci_dev->id.device_id == BROADCOM_DEV_ID_NS2)
4649 		return 0;
4650 
4651 	snprintf(mz_name, RTE_MEMZONE_NAMESIZE,
4652 		 "bnxt_" PCI_PRI_FMT "-%s", pci_dev->addr.domain,
4653 		 pci_dev->addr.bus, pci_dev->addr.devid,
4654 		 pci_dev->addr.function, "rx_port_stats");
4655 	mz_name[RTE_MEMZONE_NAMESIZE - 1] = 0;
4656 	mz = rte_memzone_lookup(mz_name);
4657 	total_alloc_len =
4658 		RTE_CACHE_LINE_ROUNDUP(sizeof(struct rx_port_stats) +
4659 				       sizeof(struct rx_port_stats_ext) + 512);
4660 	if (!mz) {
4661 		mz = rte_memzone_reserve(mz_name, total_alloc_len,
4662 					 SOCKET_ID_ANY,
4663 					 RTE_MEMZONE_2MB |
4664 					 RTE_MEMZONE_SIZE_HINT_ONLY |
4665 					 RTE_MEMZONE_IOVA_CONTIG);
4666 		if (mz == NULL)
4667 			return -ENOMEM;
4668 	}
4669 	memset(mz->addr, 0, mz->len);
4670 	mz_phys_addr = mz->iova;
4671 
4672 	bp->rx_mem_zone = (const void *)mz;
4673 	bp->hw_rx_port_stats = mz->addr;
4674 	bp->hw_rx_port_stats_map = mz_phys_addr;
4675 
4676 	snprintf(mz_name, RTE_MEMZONE_NAMESIZE,
4677 		 "bnxt_" PCI_PRI_FMT "-%s", pci_dev->addr.domain,
4678 		 pci_dev->addr.bus, pci_dev->addr.devid,
4679 		 pci_dev->addr.function, "tx_port_stats");
4680 	mz_name[RTE_MEMZONE_NAMESIZE - 1] = 0;
4681 	mz = rte_memzone_lookup(mz_name);
4682 	total_alloc_len =
4683 		RTE_CACHE_LINE_ROUNDUP(sizeof(struct tx_port_stats) +
4684 				       sizeof(struct tx_port_stats_ext) + 512);
4685 	if (!mz) {
4686 		mz = rte_memzone_reserve(mz_name,
4687 					 total_alloc_len,
4688 					 SOCKET_ID_ANY,
4689 					 RTE_MEMZONE_2MB |
4690 					 RTE_MEMZONE_SIZE_HINT_ONLY |
4691 					 RTE_MEMZONE_IOVA_CONTIG);
4692 		if (mz == NULL)
4693 			return -ENOMEM;
4694 	}
4695 	memset(mz->addr, 0, mz->len);
4696 	mz_phys_addr = mz->iova;
4697 
4698 	bp->tx_mem_zone = (const void *)mz;
4699 	bp->hw_tx_port_stats = mz->addr;
4700 	bp->hw_tx_port_stats_map = mz_phys_addr;
4701 	bp->flags |= BNXT_FLAG_PORT_STATS;
4702 
4703 	/* Display extended statistics if FW supports it */
4704 	if (bp->hwrm_spec_code < HWRM_SPEC_CODE_1_8_4 ||
4705 	    bp->hwrm_spec_code == HWRM_SPEC_CODE_1_9_0 ||
4706 	    !(bp->flags & BNXT_FLAG_EXT_STATS_SUPPORTED))
4707 		return 0;
4708 
4709 	bp->hw_rx_port_stats_ext = (void *)
4710 		((uint8_t *)bp->hw_rx_port_stats +
4711 		 sizeof(struct rx_port_stats));
4712 	bp->hw_rx_port_stats_ext_map = bp->hw_rx_port_stats_map +
4713 		sizeof(struct rx_port_stats);
4714 	bp->flags |= BNXT_FLAG_EXT_RX_PORT_STATS;
4715 
4716 	if (bp->hwrm_spec_code < HWRM_SPEC_CODE_1_9_2 ||
4717 	    bp->flags & BNXT_FLAG_EXT_STATS_SUPPORTED) {
4718 		bp->hw_tx_port_stats_ext = (void *)
4719 			((uint8_t *)bp->hw_tx_port_stats +
4720 			 sizeof(struct tx_port_stats));
4721 		bp->hw_tx_port_stats_ext_map =
4722 			bp->hw_tx_port_stats_map +
4723 			sizeof(struct tx_port_stats);
4724 		bp->flags |= BNXT_FLAG_EXT_TX_PORT_STATS;
4725 	}
4726 
4727 	return 0;
4728 }
4729 
4730 static int bnxt_setup_mac_addr(struct rte_eth_dev *eth_dev)
4731 {
4732 	struct bnxt *bp = eth_dev->data->dev_private;
4733 	int rc = 0;
4734 
4735 	eth_dev->data->mac_addrs = rte_zmalloc("bnxt_mac_addr_tbl",
4736 					       RTE_ETHER_ADDR_LEN *
4737 					       bp->max_l2_ctx,
4738 					       0);
4739 	if (eth_dev->data->mac_addrs == NULL) {
4740 		PMD_DRV_LOG(ERR, "Failed to alloc MAC addr tbl\n");
4741 		return -ENOMEM;
4742 	}
4743 
4744 	if (!BNXT_HAS_DFLT_MAC_SET(bp)) {
4745 		if (BNXT_PF(bp))
4746 			return -EINVAL;
4747 
4748 		/* Generate a random MAC address, if none was assigned by PF */
4749 		PMD_DRV_LOG(INFO, "VF MAC address not assigned by Host PF\n");
4750 		bnxt_eth_hw_addr_random(bp->mac_addr);
4751 		PMD_DRV_LOG(INFO,
4752 			    "Assign random MAC:%02X:%02X:%02X:%02X:%02X:%02X\n",
4753 			    bp->mac_addr[0], bp->mac_addr[1], bp->mac_addr[2],
4754 			    bp->mac_addr[3], bp->mac_addr[4], bp->mac_addr[5]);
4755 
4756 		rc = bnxt_hwrm_set_mac(bp);
4757 		if (rc)
4758 			return rc;
4759 	}
4760 
4761 	/* Copy the permanent MAC from the FUNC_QCAPS response */
4762 	memcpy(&eth_dev->data->mac_addrs[0], bp->mac_addr, RTE_ETHER_ADDR_LEN);
4763 
4764 	return rc;
4765 }
4766 
4767 static int bnxt_restore_dflt_mac(struct bnxt *bp)
4768 {
4769 	int rc = 0;
4770 
4771 	/* MAC is already configured in FW */
4772 	if (BNXT_HAS_DFLT_MAC_SET(bp))
4773 		return 0;
4774 
4775 	/* Restore the old MAC configured */
4776 	rc = bnxt_hwrm_set_mac(bp);
4777 	if (rc)
4778 		PMD_DRV_LOG(ERR, "Failed to restore MAC address\n");
4779 
4780 	return rc;
4781 }
4782 
4783 static void bnxt_config_vf_req_fwd(struct bnxt *bp)
4784 {
4785 	if (!BNXT_PF(bp))
4786 		return;
4787 
4788 	memset(bp->pf->vf_req_fwd, 0, sizeof(bp->pf->vf_req_fwd));
4789 
4790 	if (!(bp->fw_cap & BNXT_FW_CAP_LINK_ADMIN))
4791 		BNXT_HWRM_CMD_TO_FORWARD(HWRM_PORT_PHY_QCFG);
4792 	BNXT_HWRM_CMD_TO_FORWARD(HWRM_FUNC_CFG);
4793 	BNXT_HWRM_CMD_TO_FORWARD(HWRM_FUNC_VF_CFG);
4794 	BNXT_HWRM_CMD_TO_FORWARD(HWRM_CFA_L2_FILTER_ALLOC);
4795 	BNXT_HWRM_CMD_TO_FORWARD(HWRM_OEM_CMD);
4796 }
4797 
4798 uint16_t
4799 bnxt_get_svif(uint16_t port_id, bool func_svif,
4800 	      enum bnxt_ulp_intf_type type)
4801 {
4802 	struct rte_eth_dev *eth_dev;
4803 	struct bnxt *bp;
4804 
4805 	eth_dev = &rte_eth_devices[port_id];
4806 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
4807 		struct bnxt_representor *vfr = eth_dev->data->dev_private;
4808 		if (!vfr)
4809 			return 0;
4810 
4811 		if (type == BNXT_ULP_INTF_TYPE_VF_REP)
4812 			return vfr->svif;
4813 
4814 		eth_dev = vfr->parent_dev;
4815 	}
4816 
4817 	bp = eth_dev->data->dev_private;
4818 
4819 	return func_svif ? bp->func_svif : bp->port_svif;
4820 }
4821 
4822 uint16_t
4823 bnxt_get_vnic_id(uint16_t port, enum bnxt_ulp_intf_type type)
4824 {
4825 	struct rte_eth_dev *eth_dev;
4826 	struct bnxt_vnic_info *vnic;
4827 	struct bnxt *bp;
4828 
4829 	eth_dev = &rte_eth_devices[port];
4830 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
4831 		struct bnxt_representor *vfr = eth_dev->data->dev_private;
4832 		if (!vfr)
4833 			return 0;
4834 
4835 		if (type == BNXT_ULP_INTF_TYPE_VF_REP)
4836 			return vfr->dflt_vnic_id;
4837 
4838 		eth_dev = vfr->parent_dev;
4839 	}
4840 
4841 	bp = eth_dev->data->dev_private;
4842 
4843 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
4844 
4845 	return vnic->fw_vnic_id;
4846 }
4847 
4848 uint16_t
4849 bnxt_get_fw_func_id(uint16_t port, enum bnxt_ulp_intf_type type)
4850 {
4851 	struct rte_eth_dev *eth_dev;
4852 	struct bnxt *bp;
4853 
4854 	eth_dev = &rte_eth_devices[port];
4855 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
4856 		struct bnxt_representor *vfr = eth_dev->data->dev_private;
4857 		if (!vfr)
4858 			return 0;
4859 
4860 		if (type == BNXT_ULP_INTF_TYPE_VF_REP)
4861 			return vfr->fw_fid;
4862 
4863 		eth_dev = vfr->parent_dev;
4864 	}
4865 
4866 	bp = eth_dev->data->dev_private;
4867 
4868 	return bp->fw_fid;
4869 }
4870 
4871 enum bnxt_ulp_intf_type
4872 bnxt_get_interface_type(uint16_t port)
4873 {
4874 	struct rte_eth_dev *eth_dev;
4875 	struct bnxt *bp;
4876 
4877 	eth_dev = &rte_eth_devices[port];
4878 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev))
4879 		return BNXT_ULP_INTF_TYPE_VF_REP;
4880 
4881 	bp = eth_dev->data->dev_private;
4882 	if (BNXT_PF(bp))
4883 		return BNXT_ULP_INTF_TYPE_PF;
4884 	else if (BNXT_VF_IS_TRUSTED(bp))
4885 		return BNXT_ULP_INTF_TYPE_TRUSTED_VF;
4886 	else if (BNXT_VF(bp))
4887 		return BNXT_ULP_INTF_TYPE_VF;
4888 
4889 	return BNXT_ULP_INTF_TYPE_INVALID;
4890 }
4891 
4892 uint16_t
4893 bnxt_get_phy_port_id(uint16_t port_id)
4894 {
4895 	struct bnxt_representor *vfr;
4896 	struct rte_eth_dev *eth_dev;
4897 	struct bnxt *bp;
4898 
4899 	eth_dev = &rte_eth_devices[port_id];
4900 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
4901 		vfr = eth_dev->data->dev_private;
4902 		if (!vfr)
4903 			return 0;
4904 
4905 		eth_dev = vfr->parent_dev;
4906 	}
4907 
4908 	bp = eth_dev->data->dev_private;
4909 
4910 	return BNXT_PF(bp) ? bp->pf->port_id : bp->parent->port_id;
4911 }
4912 
4913 uint16_t
4914 bnxt_get_parif(uint16_t port_id, enum bnxt_ulp_intf_type type)
4915 {
4916 	struct rte_eth_dev *eth_dev;
4917 	struct bnxt *bp;
4918 
4919 	eth_dev = &rte_eth_devices[port_id];
4920 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
4921 		struct bnxt_representor *vfr = eth_dev->data->dev_private;
4922 		if (!vfr)
4923 			return 0;
4924 
4925 		if (type == BNXT_ULP_INTF_TYPE_VF_REP)
4926 			return vfr->fw_fid - 1;
4927 
4928 		eth_dev = vfr->parent_dev;
4929 	}
4930 
4931 	bp = eth_dev->data->dev_private;
4932 
4933 	return BNXT_PF(bp) ? bp->fw_fid - 1 : bp->parent->fid - 1;
4934 }
4935 
4936 uint16_t
4937 bnxt_get_vport(uint16_t port_id)
4938 {
4939 	return (1 << bnxt_get_phy_port_id(port_id));
4940 }
4941 
4942 static void bnxt_alloc_error_recovery_info(struct bnxt *bp)
4943 {
4944 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4945 
4946 	if (info) {
4947 		if (!(bp->fw_cap & BNXT_FW_CAP_HCOMM_FW_STATUS))
4948 			memset(info, 0, sizeof(*info));
4949 		return;
4950 	}
4951 
4952 	if (!(bp->fw_cap & BNXT_FW_CAP_ERROR_RECOVERY))
4953 		return;
4954 
4955 	info = rte_zmalloc("bnxt_hwrm_error_recovery_qcfg",
4956 			   sizeof(*info), 0);
4957 	if (!info)
4958 		bp->fw_cap &= ~BNXT_FW_CAP_ERROR_RECOVERY;
4959 
4960 	bp->recovery_info = info;
4961 }
4962 
4963 static void bnxt_check_fw_status(struct bnxt *bp)
4964 {
4965 	uint32_t fw_status;
4966 
4967 	if (!(bp->recovery_info &&
4968 	      (bp->fw_cap & BNXT_FW_CAP_HCOMM_FW_STATUS)))
4969 		return;
4970 
4971 	fw_status = bnxt_read_fw_status_reg(bp, BNXT_FW_STATUS_REG);
4972 	if (fw_status != BNXT_FW_STATUS_HEALTHY)
4973 		PMD_DRV_LOG(ERR, "Firmware not responding, status: %#x\n",
4974 			    fw_status);
4975 }
4976 
4977 static int bnxt_map_hcomm_fw_status_reg(struct bnxt *bp)
4978 {
4979 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4980 	uint32_t status_loc;
4981 	uint32_t sig_ver;
4982 
4983 	rte_write32(HCOMM_STATUS_STRUCT_LOC, (uint8_t *)bp->bar0 +
4984 		    BNXT_GRCPF_REG_WINDOW_BASE_OUT + 4);
4985 	sig_ver = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
4986 				   BNXT_GRCP_WINDOW_2_BASE +
4987 				   offsetof(struct hcomm_status,
4988 					    sig_ver)));
4989 	/* If the signature is absent, then FW does not support this feature */
4990 	if ((sig_ver & HCOMM_STATUS_SIGNATURE_MASK) !=
4991 	    HCOMM_STATUS_SIGNATURE_VAL)
4992 		return 0;
4993 
4994 	if (!info) {
4995 		info = rte_zmalloc("bnxt_hwrm_error_recovery_qcfg",
4996 				   sizeof(*info), 0);
4997 		if (!info)
4998 			return -ENOMEM;
4999 		bp->recovery_info = info;
5000 	} else {
5001 		memset(info, 0, sizeof(*info));
5002 	}
5003 
5004 	status_loc = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
5005 				      BNXT_GRCP_WINDOW_2_BASE +
5006 				      offsetof(struct hcomm_status,
5007 					       fw_status_loc)));
5008 
5009 	/* Only pre-map the FW health status GRC register */
5010 	if (BNXT_FW_STATUS_REG_TYPE(status_loc) != BNXT_FW_STATUS_REG_TYPE_GRC)
5011 		return 0;
5012 
5013 	info->status_regs[BNXT_FW_STATUS_REG] = status_loc;
5014 	info->mapped_status_regs[BNXT_FW_STATUS_REG] =
5015 		BNXT_GRCP_WINDOW_2_BASE + (status_loc & BNXT_GRCP_OFFSET_MASK);
5016 
5017 	rte_write32((status_loc & BNXT_GRCP_BASE_MASK), (uint8_t *)bp->bar0 +
5018 		    BNXT_GRCPF_REG_WINDOW_BASE_OUT + 4);
5019 
5020 	bp->fw_cap |= BNXT_FW_CAP_HCOMM_FW_STATUS;
5021 
5022 	return 0;
5023 }
5024 
5025 /* This function gets the FW version along with the
5026  * capabilities(MAX and current) of the function, vnic,
5027  * error recovery, phy and other chip related info
5028  */
5029 static int bnxt_get_config(struct bnxt *bp)
5030 {
5031 	uint16_t mtu;
5032 	int rc = 0;
5033 
5034 	bp->fw_cap = 0;
5035 
5036 	rc = bnxt_map_hcomm_fw_status_reg(bp);
5037 	if (rc)
5038 		return rc;
5039 
5040 	rc = bnxt_hwrm_ver_get(bp, DFLT_HWRM_CMD_TIMEOUT);
5041 	if (rc) {
5042 		bnxt_check_fw_status(bp);
5043 		return rc;
5044 	}
5045 
5046 	rc = bnxt_hwrm_func_reset(bp);
5047 	if (rc)
5048 		return -EIO;
5049 
5050 	rc = bnxt_hwrm_vnic_qcaps(bp);
5051 	if (rc)
5052 		return rc;
5053 
5054 	rc = bnxt_hwrm_queue_qportcfg(bp);
5055 	if (rc)
5056 		return rc;
5057 
5058 	/* Get the MAX capabilities for this function.
5059 	 * This function also allocates context memory for TQM rings and
5060 	 * informs the firmware about this allocated backing store memory.
5061 	 */
5062 	rc = bnxt_hwrm_func_qcaps(bp);
5063 	if (rc)
5064 		return rc;
5065 
5066 	rc = bnxt_hwrm_func_qcfg(bp, &mtu);
5067 	if (rc)
5068 		return rc;
5069 
5070 	rc = bnxt_hwrm_cfa_adv_flow_mgmt_qcaps(bp);
5071 	if (rc)
5072 		return rc;
5073 
5074 	bnxt_hwrm_port_mac_qcfg(bp);
5075 
5076 	bnxt_hwrm_parent_pf_qcfg(bp);
5077 
5078 	bnxt_hwrm_port_phy_qcaps(bp);
5079 
5080 	bnxt_alloc_error_recovery_info(bp);
5081 	/* Get the adapter error recovery support info */
5082 	rc = bnxt_hwrm_error_recovery_qcfg(bp);
5083 	if (rc)
5084 		bp->fw_cap &= ~BNXT_FW_CAP_ERROR_RECOVERY;
5085 
5086 	bnxt_hwrm_port_led_qcaps(bp);
5087 
5088 	return 0;
5089 }
5090 
5091 static int
5092 bnxt_init_locks(struct bnxt *bp)
5093 {
5094 	int err;
5095 
5096 	err = pthread_mutex_init(&bp->flow_lock, NULL);
5097 	if (err) {
5098 		PMD_DRV_LOG(ERR, "Unable to initialize flow_lock\n");
5099 		return err;
5100 	}
5101 
5102 	err = pthread_mutex_init(&bp->def_cp_lock, NULL);
5103 	if (err) {
5104 		PMD_DRV_LOG(ERR, "Unable to initialize def_cp_lock\n");
5105 		return err;
5106 	}
5107 
5108 	err = pthread_mutex_init(&bp->health_check_lock, NULL);
5109 	if (err) {
5110 		PMD_DRV_LOG(ERR, "Unable to initialize health_check_lock\n");
5111 		return err;
5112 	}
5113 
5114 	err = pthread_mutex_init(&bp->err_recovery_lock, NULL);
5115 	if (err)
5116 		PMD_DRV_LOG(ERR, "Unable to initialize err_recovery_lock\n");
5117 
5118 	return err;
5119 }
5120 
5121 static int bnxt_init_resources(struct bnxt *bp, bool reconfig_dev)
5122 {
5123 	int rc = 0;
5124 
5125 	rc = bnxt_get_config(bp);
5126 	if (rc)
5127 		return rc;
5128 
5129 	if (!reconfig_dev) {
5130 		rc = bnxt_setup_mac_addr(bp->eth_dev);
5131 		if (rc)
5132 			return rc;
5133 	} else {
5134 		rc = bnxt_restore_dflt_mac(bp);
5135 		if (rc)
5136 			return rc;
5137 	}
5138 
5139 	bnxt_config_vf_req_fwd(bp);
5140 
5141 	rc = bnxt_hwrm_func_driver_register(bp);
5142 	if (rc) {
5143 		PMD_DRV_LOG(ERR, "Failed to register driver");
5144 		return -EBUSY;
5145 	}
5146 
5147 	if (BNXT_PF(bp)) {
5148 		if (bp->pdev->max_vfs) {
5149 			rc = bnxt_hwrm_allocate_vfs(bp, bp->pdev->max_vfs);
5150 			if (rc) {
5151 				PMD_DRV_LOG(ERR, "Failed to allocate VFs\n");
5152 				return rc;
5153 			}
5154 		} else {
5155 			rc = bnxt_hwrm_allocate_pf_only(bp);
5156 			if (rc) {
5157 				PMD_DRV_LOG(ERR,
5158 					    "Failed to allocate PF resources");
5159 				return rc;
5160 			}
5161 		}
5162 	}
5163 
5164 	rc = bnxt_alloc_mem(bp, reconfig_dev);
5165 	if (rc)
5166 		return rc;
5167 
5168 	rc = bnxt_setup_int(bp);
5169 	if (rc)
5170 		return rc;
5171 
5172 	rc = bnxt_request_int(bp);
5173 	if (rc)
5174 		return rc;
5175 
5176 	rc = bnxt_init_ctx_mem(bp);
5177 	if (rc) {
5178 		PMD_DRV_LOG(ERR, "Failed to init adv_flow_counters\n");
5179 		return rc;
5180 	}
5181 
5182 	return 0;
5183 }
5184 
5185 static int
5186 bnxt_parse_devarg_truflow(__rte_unused const char *key,
5187 			  const char *value, void *opaque_arg)
5188 {
5189 	struct bnxt *bp = opaque_arg;
5190 	unsigned long truflow;
5191 	char *end = NULL;
5192 
5193 	if (!value || !opaque_arg) {
5194 		PMD_DRV_LOG(ERR,
5195 			    "Invalid parameter passed to truflow devargs.\n");
5196 		return -EINVAL;
5197 	}
5198 
5199 	truflow = strtoul(value, &end, 10);
5200 	if (end == NULL || *end != '\0' ||
5201 	    (truflow == ULONG_MAX && errno == ERANGE)) {
5202 		PMD_DRV_LOG(ERR,
5203 			    "Invalid parameter passed to truflow devargs.\n");
5204 		return -EINVAL;
5205 	}
5206 
5207 	if (BNXT_DEVARG_TRUFLOW_INVALID(truflow)) {
5208 		PMD_DRV_LOG(ERR,
5209 			    "Invalid value passed to truflow devargs.\n");
5210 		return -EINVAL;
5211 	}
5212 
5213 	if (truflow) {
5214 		bp->flags |= BNXT_FLAG_TRUFLOW_EN;
5215 		PMD_DRV_LOG(INFO, "Host-based truflow feature enabled.\n");
5216 	} else {
5217 		bp->flags &= ~BNXT_FLAG_TRUFLOW_EN;
5218 		PMD_DRV_LOG(INFO, "Host-based truflow feature disabled.\n");
5219 	}
5220 
5221 	return 0;
5222 }
5223 
5224 static int
5225 bnxt_parse_devarg_flow_xstat(__rte_unused const char *key,
5226 			     const char *value, void *opaque_arg)
5227 {
5228 	struct bnxt *bp = opaque_arg;
5229 	unsigned long flow_xstat;
5230 	char *end = NULL;
5231 
5232 	if (!value || !opaque_arg) {
5233 		PMD_DRV_LOG(ERR,
5234 			    "Invalid parameter passed to flow_xstat devarg.\n");
5235 		return -EINVAL;
5236 	}
5237 
5238 	flow_xstat = strtoul(value, &end, 10);
5239 	if (end == NULL || *end != '\0' ||
5240 	    (flow_xstat == ULONG_MAX && errno == ERANGE)) {
5241 		PMD_DRV_LOG(ERR,
5242 			    "Invalid parameter passed to flow_xstat devarg.\n");
5243 		return -EINVAL;
5244 	}
5245 
5246 	if (BNXT_DEVARG_FLOW_XSTAT_INVALID(flow_xstat)) {
5247 		PMD_DRV_LOG(ERR,
5248 			    "Invalid value passed to flow_xstat devarg.\n");
5249 		return -EINVAL;
5250 	}
5251 
5252 	bp->flags |= BNXT_FLAG_FLOW_XSTATS_EN;
5253 	if (BNXT_FLOW_XSTATS_EN(bp))
5254 		PMD_DRV_LOG(INFO, "flow_xstat feature enabled.\n");
5255 
5256 	return 0;
5257 }
5258 
5259 static int
5260 bnxt_parse_devarg_max_num_kflows(__rte_unused const char *key,
5261 					const char *value, void *opaque_arg)
5262 {
5263 	struct bnxt *bp = opaque_arg;
5264 	unsigned long max_num_kflows;
5265 	char *end = NULL;
5266 
5267 	if (!value || !opaque_arg) {
5268 		PMD_DRV_LOG(ERR,
5269 			"Invalid parameter passed to max_num_kflows devarg.\n");
5270 		return -EINVAL;
5271 	}
5272 
5273 	max_num_kflows = strtoul(value, &end, 10);
5274 	if (end == NULL || *end != '\0' ||
5275 		(max_num_kflows == ULONG_MAX && errno == ERANGE)) {
5276 		PMD_DRV_LOG(ERR,
5277 			"Invalid parameter passed to max_num_kflows devarg.\n");
5278 		return -EINVAL;
5279 	}
5280 
5281 	if (bnxt_devarg_max_num_kflow_invalid(max_num_kflows)) {
5282 		PMD_DRV_LOG(ERR,
5283 			"Invalid value passed to max_num_kflows devarg.\n");
5284 		return -EINVAL;
5285 	}
5286 
5287 	bp->max_num_kflows = max_num_kflows;
5288 	if (bp->max_num_kflows)
5289 		PMD_DRV_LOG(INFO, "max_num_kflows set as %ldK.\n",
5290 				max_num_kflows);
5291 
5292 	return 0;
5293 }
5294 
5295 static int
5296 bnxt_parse_devarg_rep_is_pf(__rte_unused const char *key,
5297 			    const char *value, void *opaque_arg)
5298 {
5299 	struct bnxt_representor *vfr_bp = opaque_arg;
5300 	unsigned long rep_is_pf;
5301 	char *end = NULL;
5302 
5303 	if (!value || !opaque_arg) {
5304 		PMD_DRV_LOG(ERR,
5305 			    "Invalid parameter passed to rep_is_pf devargs.\n");
5306 		return -EINVAL;
5307 	}
5308 
5309 	rep_is_pf = strtoul(value, &end, 10);
5310 	if (end == NULL || *end != '\0' ||
5311 	    (rep_is_pf == ULONG_MAX && errno == ERANGE)) {
5312 		PMD_DRV_LOG(ERR,
5313 			    "Invalid parameter passed to rep_is_pf devargs.\n");
5314 		return -EINVAL;
5315 	}
5316 
5317 	if (BNXT_DEVARG_REP_IS_PF_INVALID(rep_is_pf)) {
5318 		PMD_DRV_LOG(ERR,
5319 			    "Invalid value passed to rep_is_pf devargs.\n");
5320 		return -EINVAL;
5321 	}
5322 
5323 	vfr_bp->flags |= rep_is_pf;
5324 	if (BNXT_REP_PF(vfr_bp))
5325 		PMD_DRV_LOG(INFO, "PF representor\n");
5326 	else
5327 		PMD_DRV_LOG(INFO, "VF representor\n");
5328 
5329 	return 0;
5330 }
5331 
5332 static int
5333 bnxt_parse_devarg_rep_based_pf(__rte_unused const char *key,
5334 			       const char *value, void *opaque_arg)
5335 {
5336 	struct bnxt_representor *vfr_bp = opaque_arg;
5337 	unsigned long rep_based_pf;
5338 	char *end = NULL;
5339 
5340 	if (!value || !opaque_arg) {
5341 		PMD_DRV_LOG(ERR,
5342 			    "Invalid parameter passed to rep_based_pf "
5343 			    "devargs.\n");
5344 		return -EINVAL;
5345 	}
5346 
5347 	rep_based_pf = strtoul(value, &end, 10);
5348 	if (end == NULL || *end != '\0' ||
5349 	    (rep_based_pf == ULONG_MAX && errno == ERANGE)) {
5350 		PMD_DRV_LOG(ERR,
5351 			    "Invalid parameter passed to rep_based_pf "
5352 			    "devargs.\n");
5353 		return -EINVAL;
5354 	}
5355 
5356 	if (BNXT_DEVARG_REP_BASED_PF_INVALID(rep_based_pf)) {
5357 		PMD_DRV_LOG(ERR,
5358 			    "Invalid value passed to rep_based_pf devargs.\n");
5359 		return -EINVAL;
5360 	}
5361 
5362 	vfr_bp->rep_based_pf = rep_based_pf;
5363 	vfr_bp->flags |= BNXT_REP_BASED_PF_VALID;
5364 
5365 	PMD_DRV_LOG(INFO, "rep-based-pf = %d\n", vfr_bp->rep_based_pf);
5366 
5367 	return 0;
5368 }
5369 
5370 static int
5371 bnxt_parse_devarg_rep_q_r2f(__rte_unused const char *key,
5372 			    const char *value, void *opaque_arg)
5373 {
5374 	struct bnxt_representor *vfr_bp = opaque_arg;
5375 	unsigned long rep_q_r2f;
5376 	char *end = NULL;
5377 
5378 	if (!value || !opaque_arg) {
5379 		PMD_DRV_LOG(ERR,
5380 			    "Invalid parameter passed to rep_q_r2f "
5381 			    "devargs.\n");
5382 		return -EINVAL;
5383 	}
5384 
5385 	rep_q_r2f = strtoul(value, &end, 10);
5386 	if (end == NULL || *end != '\0' ||
5387 	    (rep_q_r2f == ULONG_MAX && errno == ERANGE)) {
5388 		PMD_DRV_LOG(ERR,
5389 			    "Invalid parameter passed to rep_q_r2f "
5390 			    "devargs.\n");
5391 		return -EINVAL;
5392 	}
5393 
5394 	if (BNXT_DEVARG_REP_Q_R2F_INVALID(rep_q_r2f)) {
5395 		PMD_DRV_LOG(ERR,
5396 			    "Invalid value passed to rep_q_r2f devargs.\n");
5397 		return -EINVAL;
5398 	}
5399 
5400 	vfr_bp->rep_q_r2f = rep_q_r2f;
5401 	vfr_bp->flags |= BNXT_REP_Q_R2F_VALID;
5402 	PMD_DRV_LOG(INFO, "rep-q-r2f = %d\n", vfr_bp->rep_q_r2f);
5403 
5404 	return 0;
5405 }
5406 
5407 static int
5408 bnxt_parse_devarg_rep_q_f2r(__rte_unused const char *key,
5409 			    const char *value, void *opaque_arg)
5410 {
5411 	struct bnxt_representor *vfr_bp = opaque_arg;
5412 	unsigned long rep_q_f2r;
5413 	char *end = NULL;
5414 
5415 	if (!value || !opaque_arg) {
5416 		PMD_DRV_LOG(ERR,
5417 			    "Invalid parameter passed to rep_q_f2r "
5418 			    "devargs.\n");
5419 		return -EINVAL;
5420 	}
5421 
5422 	rep_q_f2r = strtoul(value, &end, 10);
5423 	if (end == NULL || *end != '\0' ||
5424 	    (rep_q_f2r == ULONG_MAX && errno == ERANGE)) {
5425 		PMD_DRV_LOG(ERR,
5426 			    "Invalid parameter passed to rep_q_f2r "
5427 			    "devargs.\n");
5428 		return -EINVAL;
5429 	}
5430 
5431 	if (BNXT_DEVARG_REP_Q_F2R_INVALID(rep_q_f2r)) {
5432 		PMD_DRV_LOG(ERR,
5433 			    "Invalid value passed to rep_q_f2r devargs.\n");
5434 		return -EINVAL;
5435 	}
5436 
5437 	vfr_bp->rep_q_f2r = rep_q_f2r;
5438 	vfr_bp->flags |= BNXT_REP_Q_F2R_VALID;
5439 	PMD_DRV_LOG(INFO, "rep-q-f2r = %d\n", vfr_bp->rep_q_f2r);
5440 
5441 	return 0;
5442 }
5443 
5444 static int
5445 bnxt_parse_devarg_rep_fc_r2f(__rte_unused const char *key,
5446 			     const char *value, void *opaque_arg)
5447 {
5448 	struct bnxt_representor *vfr_bp = opaque_arg;
5449 	unsigned long rep_fc_r2f;
5450 	char *end = NULL;
5451 
5452 	if (!value || !opaque_arg) {
5453 		PMD_DRV_LOG(ERR,
5454 			    "Invalid parameter passed to rep_fc_r2f "
5455 			    "devargs.\n");
5456 		return -EINVAL;
5457 	}
5458 
5459 	rep_fc_r2f = strtoul(value, &end, 10);
5460 	if (end == NULL || *end != '\0' ||
5461 	    (rep_fc_r2f == ULONG_MAX && errno == ERANGE)) {
5462 		PMD_DRV_LOG(ERR,
5463 			    "Invalid parameter passed to rep_fc_r2f "
5464 			    "devargs.\n");
5465 		return -EINVAL;
5466 	}
5467 
5468 	if (BNXT_DEVARG_REP_FC_R2F_INVALID(rep_fc_r2f)) {
5469 		PMD_DRV_LOG(ERR,
5470 			    "Invalid value passed to rep_fc_r2f devargs.\n");
5471 		return -EINVAL;
5472 	}
5473 
5474 	vfr_bp->flags |= BNXT_REP_FC_R2F_VALID;
5475 	vfr_bp->rep_fc_r2f = rep_fc_r2f;
5476 	PMD_DRV_LOG(INFO, "rep-fc-r2f = %lu\n", rep_fc_r2f);
5477 
5478 	return 0;
5479 }
5480 
5481 static int
5482 bnxt_parse_devarg_rep_fc_f2r(__rte_unused const char *key,
5483 			     const char *value, void *opaque_arg)
5484 {
5485 	struct bnxt_representor *vfr_bp = opaque_arg;
5486 	unsigned long rep_fc_f2r;
5487 	char *end = NULL;
5488 
5489 	if (!value || !opaque_arg) {
5490 		PMD_DRV_LOG(ERR,
5491 			    "Invalid parameter passed to rep_fc_f2r "
5492 			    "devargs.\n");
5493 		return -EINVAL;
5494 	}
5495 
5496 	rep_fc_f2r = strtoul(value, &end, 10);
5497 	if (end == NULL || *end != '\0' ||
5498 	    (rep_fc_f2r == ULONG_MAX && errno == ERANGE)) {
5499 		PMD_DRV_LOG(ERR,
5500 			    "Invalid parameter passed to rep_fc_f2r "
5501 			    "devargs.\n");
5502 		return -EINVAL;
5503 	}
5504 
5505 	if (BNXT_DEVARG_REP_FC_F2R_INVALID(rep_fc_f2r)) {
5506 		PMD_DRV_LOG(ERR,
5507 			    "Invalid value passed to rep_fc_f2r devargs.\n");
5508 		return -EINVAL;
5509 	}
5510 
5511 	vfr_bp->flags |= BNXT_REP_FC_F2R_VALID;
5512 	vfr_bp->rep_fc_f2r = rep_fc_f2r;
5513 	PMD_DRV_LOG(INFO, "rep-fc-f2r = %lu\n", rep_fc_f2r);
5514 
5515 	return 0;
5516 }
5517 
5518 static int
5519 bnxt_parse_dev_args(struct bnxt *bp, struct rte_devargs *devargs)
5520 {
5521 	struct rte_kvargs *kvlist;
5522 	int ret;
5523 
5524 	if (devargs == NULL)
5525 		return 0;
5526 
5527 	kvlist = rte_kvargs_parse(devargs->args, bnxt_dev_args);
5528 	if (kvlist == NULL)
5529 		return -EINVAL;
5530 
5531 	/*
5532 	 * Handler for "truflow" devarg.
5533 	 * Invoked as for ex: "-a 0000:00:0d.0,host-based-truflow=1"
5534 	 */
5535 	ret = rte_kvargs_process(kvlist, BNXT_DEVARG_TRUFLOW,
5536 				 bnxt_parse_devarg_truflow, bp);
5537 	if (ret)
5538 		goto err;
5539 
5540 	/*
5541 	 * Handler for "flow_xstat" devarg.
5542 	 * Invoked as for ex: "-a 0000:00:0d.0,flow_xstat=1"
5543 	 */
5544 	ret = rte_kvargs_process(kvlist, BNXT_DEVARG_FLOW_XSTAT,
5545 				 bnxt_parse_devarg_flow_xstat, bp);
5546 	if (ret)
5547 		goto err;
5548 
5549 	/*
5550 	 * Handler for "max_num_kflows" devarg.
5551 	 * Invoked as for ex: "-a 000:00:0d.0,max_num_kflows=32"
5552 	 */
5553 	ret = rte_kvargs_process(kvlist, BNXT_DEVARG_MAX_NUM_KFLOWS,
5554 				 bnxt_parse_devarg_max_num_kflows, bp);
5555 	if (ret)
5556 		goto err;
5557 
5558 err:
5559 	rte_kvargs_free(kvlist);
5560 	return ret;
5561 }
5562 
5563 static int bnxt_alloc_switch_domain(struct bnxt *bp)
5564 {
5565 	int rc = 0;
5566 
5567 	if (BNXT_PF(bp) || BNXT_VF_IS_TRUSTED(bp)) {
5568 		rc = rte_eth_switch_domain_alloc(&bp->switch_domain_id);
5569 		if (rc)
5570 			PMD_DRV_LOG(ERR,
5571 				    "Failed to alloc switch domain: %d\n", rc);
5572 		else
5573 			PMD_DRV_LOG(INFO,
5574 				    "Switch domain allocated %d\n",
5575 				    bp->switch_domain_id);
5576 	}
5577 
5578 	return rc;
5579 }
5580 
5581 /* Allocate and initialize various fields in bnxt struct that
5582  * need to be allocated/destroyed only once in the lifetime of the driver
5583  */
5584 static int bnxt_drv_init(struct rte_eth_dev *eth_dev)
5585 {
5586 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
5587 	struct bnxt *bp = eth_dev->data->dev_private;
5588 	int rc = 0;
5589 
5590 	bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
5591 
5592 	if (bnxt_vf_pciid(pci_dev->id.device_id))
5593 		bp->flags |= BNXT_FLAG_VF;
5594 
5595 	if (bnxt_p5_device(pci_dev->id.device_id))
5596 		bp->flags |= BNXT_FLAG_CHIP_P5;
5597 
5598 	if (pci_dev->id.device_id == BROADCOM_DEV_ID_58802 ||
5599 	    pci_dev->id.device_id == BROADCOM_DEV_ID_58804 ||
5600 	    pci_dev->id.device_id == BROADCOM_DEV_ID_58808 ||
5601 	    pci_dev->id.device_id == BROADCOM_DEV_ID_58802_VF)
5602 		bp->flags |= BNXT_FLAG_STINGRAY;
5603 
5604 	if (BNXT_TRUFLOW_EN(bp)) {
5605 		/* extra mbuf field is required to store CFA code from mark */
5606 		static const struct rte_mbuf_dynfield bnxt_cfa_code_dynfield_desc = {
5607 			.name = RTE_PMD_BNXT_CFA_CODE_DYNFIELD_NAME,
5608 			.size = sizeof(bnxt_cfa_code_dynfield_t),
5609 			.align = __alignof__(bnxt_cfa_code_dynfield_t),
5610 		};
5611 		bnxt_cfa_code_dynfield_offset =
5612 			rte_mbuf_dynfield_register(&bnxt_cfa_code_dynfield_desc);
5613 		if (bnxt_cfa_code_dynfield_offset < 0) {
5614 			PMD_DRV_LOG(ERR,
5615 			    "Failed to register mbuf field for TruFlow mark\n");
5616 			return -rte_errno;
5617 		}
5618 	}
5619 
5620 	rc = bnxt_map_pci_bars(eth_dev);
5621 	if (rc) {
5622 		PMD_DRV_LOG(ERR,
5623 			    "Failed to initialize board rc: %x\n", rc);
5624 		return rc;
5625 	}
5626 
5627 	rc = bnxt_alloc_pf_info(bp);
5628 	if (rc)
5629 		return rc;
5630 
5631 	rc = bnxt_alloc_link_info(bp);
5632 	if (rc)
5633 		return rc;
5634 
5635 	rc = bnxt_alloc_parent_info(bp);
5636 	if (rc)
5637 		return rc;
5638 
5639 	rc = bnxt_alloc_hwrm_resources(bp);
5640 	if (rc) {
5641 		PMD_DRV_LOG(ERR,
5642 			    "Failed to allocate response buffer rc: %x\n", rc);
5643 		return rc;
5644 	}
5645 	rc = bnxt_alloc_leds_info(bp);
5646 	if (rc)
5647 		return rc;
5648 
5649 	rc = bnxt_alloc_cos_queues(bp);
5650 	if (rc)
5651 		return rc;
5652 
5653 	rc = bnxt_init_locks(bp);
5654 	if (rc)
5655 		return rc;
5656 
5657 	rc = bnxt_alloc_switch_domain(bp);
5658 	if (rc)
5659 		return rc;
5660 
5661 	return rc;
5662 }
5663 
5664 static int
5665 bnxt_dev_init(struct rte_eth_dev *eth_dev, void *params __rte_unused)
5666 {
5667 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
5668 	static int version_printed;
5669 	struct bnxt *bp;
5670 	int rc;
5671 
5672 	if (version_printed++ == 0)
5673 		PMD_DRV_LOG(INFO, "%s\n", bnxt_version);
5674 
5675 	eth_dev->dev_ops = &bnxt_dev_ops;
5676 	eth_dev->rx_queue_count = bnxt_rx_queue_count_op;
5677 	eth_dev->rx_descriptor_status = bnxt_rx_descriptor_status_op;
5678 	eth_dev->tx_descriptor_status = bnxt_tx_descriptor_status_op;
5679 	eth_dev->rx_pkt_burst = &bnxt_recv_pkts;
5680 	eth_dev->tx_pkt_burst = &bnxt_xmit_pkts;
5681 
5682 	/*
5683 	 * For secondary processes, we don't initialise any further
5684 	 * as primary has already done this work.
5685 	 */
5686 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
5687 		return 0;
5688 
5689 	rte_eth_copy_pci_info(eth_dev, pci_dev);
5690 	eth_dev->data->dev_flags |= RTE_ETH_DEV_AUTOFILL_QUEUE_XSTATS;
5691 
5692 	bp = eth_dev->data->dev_private;
5693 
5694 	/* Parse dev arguments passed on when starting the DPDK application. */
5695 	rc = bnxt_parse_dev_args(bp, pci_dev->device.devargs);
5696 	if (rc)
5697 		goto error_free;
5698 
5699 	rc = bnxt_drv_init(eth_dev);
5700 	if (rc)
5701 		goto error_free;
5702 
5703 	rc = bnxt_init_resources(bp, false);
5704 	if (rc)
5705 		goto error_free;
5706 
5707 	rc = bnxt_alloc_stats_mem(bp);
5708 	if (rc)
5709 		goto error_free;
5710 
5711 	PMD_DRV_LOG(INFO,
5712 		    DRV_MODULE_NAME "found at mem %" PRIX64 ", node addr %pM\n",
5713 		    pci_dev->mem_resource[0].phys_addr,
5714 		    pci_dev->mem_resource[0].addr);
5715 
5716 	return 0;
5717 
5718 error_free:
5719 	bnxt_dev_uninit(eth_dev);
5720 	return rc;
5721 }
5722 
5723 
5724 static void bnxt_free_ctx_mem_buf(struct bnxt_ctx_mem_buf_info *ctx)
5725 {
5726 	if (!ctx)
5727 		return;
5728 
5729 	if (ctx->va)
5730 		rte_free(ctx->va);
5731 
5732 	ctx->va = NULL;
5733 	ctx->dma = RTE_BAD_IOVA;
5734 	ctx->ctx_id = BNXT_CTX_VAL_INVAL;
5735 }
5736 
5737 static void bnxt_unregister_fc_ctx_mem(struct bnxt *bp)
5738 {
5739 	bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_RX,
5740 				  CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
5741 				  bp->flow_stat->rx_fc_out_tbl.ctx_id,
5742 				  bp->flow_stat->max_fc,
5743 				  false);
5744 
5745 	bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_TX,
5746 				  CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
5747 				  bp->flow_stat->tx_fc_out_tbl.ctx_id,
5748 				  bp->flow_stat->max_fc,
5749 				  false);
5750 
5751 	if (bp->flow_stat->rx_fc_in_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
5752 		bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->rx_fc_in_tbl.ctx_id);
5753 	bp->flow_stat->rx_fc_in_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
5754 
5755 	if (bp->flow_stat->rx_fc_out_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
5756 		bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->rx_fc_out_tbl.ctx_id);
5757 	bp->flow_stat->rx_fc_out_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
5758 
5759 	if (bp->flow_stat->tx_fc_in_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
5760 		bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->tx_fc_in_tbl.ctx_id);
5761 	bp->flow_stat->tx_fc_in_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
5762 
5763 	if (bp->flow_stat->tx_fc_out_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
5764 		bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->tx_fc_out_tbl.ctx_id);
5765 	bp->flow_stat->tx_fc_out_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
5766 }
5767 
5768 static void bnxt_uninit_fc_ctx_mem(struct bnxt *bp)
5769 {
5770 	bnxt_unregister_fc_ctx_mem(bp);
5771 
5772 	bnxt_free_ctx_mem_buf(&bp->flow_stat->rx_fc_in_tbl);
5773 	bnxt_free_ctx_mem_buf(&bp->flow_stat->rx_fc_out_tbl);
5774 	bnxt_free_ctx_mem_buf(&bp->flow_stat->tx_fc_in_tbl);
5775 	bnxt_free_ctx_mem_buf(&bp->flow_stat->tx_fc_out_tbl);
5776 }
5777 
5778 static void bnxt_uninit_ctx_mem(struct bnxt *bp)
5779 {
5780 	if (BNXT_FLOW_XSTATS_EN(bp))
5781 		bnxt_uninit_fc_ctx_mem(bp);
5782 }
5783 
5784 static void
5785 bnxt_free_error_recovery_info(struct bnxt *bp)
5786 {
5787 	rte_free(bp->recovery_info);
5788 	bp->recovery_info = NULL;
5789 	bp->fw_cap &= ~BNXT_FW_CAP_ERROR_RECOVERY;
5790 }
5791 
5792 static int
5793 bnxt_uninit_resources(struct bnxt *bp, bool reconfig_dev)
5794 {
5795 	int rc;
5796 
5797 	bnxt_free_int(bp);
5798 	bnxt_free_mem(bp, reconfig_dev);
5799 
5800 	bnxt_hwrm_func_buf_unrgtr(bp);
5801 	if (bp->pf != NULL) {
5802 		rte_free(bp->pf->vf_req_buf);
5803 		bp->pf->vf_req_buf = NULL;
5804 	}
5805 
5806 	rc = bnxt_hwrm_func_driver_unregister(bp, 0);
5807 	bp->flags &= ~BNXT_FLAG_REGISTERED;
5808 	bnxt_free_ctx_mem(bp);
5809 	if (!reconfig_dev) {
5810 		bnxt_free_hwrm_resources(bp);
5811 		bnxt_free_error_recovery_info(bp);
5812 	}
5813 
5814 	bnxt_uninit_ctx_mem(bp);
5815 
5816 	bnxt_free_flow_stats_info(bp);
5817 	if (bp->rep_info != NULL)
5818 		bnxt_free_switch_domain(bp);
5819 	bnxt_free_rep_info(bp);
5820 	rte_free(bp->ptp_cfg);
5821 	bp->ptp_cfg = NULL;
5822 	return rc;
5823 }
5824 
5825 static int
5826 bnxt_dev_uninit(struct rte_eth_dev *eth_dev)
5827 {
5828 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
5829 		return -EPERM;
5830 
5831 	PMD_DRV_LOG(DEBUG, "Calling Device uninit\n");
5832 
5833 	if (eth_dev->state != RTE_ETH_DEV_UNUSED)
5834 		bnxt_dev_close_op(eth_dev);
5835 
5836 	return 0;
5837 }
5838 
5839 static int bnxt_pci_remove_dev_with_reps(struct rte_eth_dev *eth_dev)
5840 {
5841 	struct bnxt *bp = eth_dev->data->dev_private;
5842 	struct rte_eth_dev *vf_rep_eth_dev;
5843 	int ret = 0, i;
5844 
5845 	if (!bp)
5846 		return -EINVAL;
5847 
5848 	for (i = 0; i < bp->num_reps; i++) {
5849 		vf_rep_eth_dev = bp->rep_info[i].vfr_eth_dev;
5850 		if (!vf_rep_eth_dev)
5851 			continue;
5852 		PMD_DRV_LOG(DEBUG, "BNXT Port:%d VFR pci remove\n",
5853 			    vf_rep_eth_dev->data->port_id);
5854 		rte_eth_dev_destroy(vf_rep_eth_dev, bnxt_representor_uninit);
5855 	}
5856 	PMD_DRV_LOG(DEBUG, "BNXT Port:%d pci remove\n",
5857 		    eth_dev->data->port_id);
5858 	ret = rte_eth_dev_destroy(eth_dev, bnxt_dev_uninit);
5859 
5860 	return ret;
5861 }
5862 
5863 static void bnxt_free_rep_info(struct bnxt *bp)
5864 {
5865 	rte_free(bp->rep_info);
5866 	bp->rep_info = NULL;
5867 	rte_free(bp->cfa_code_map);
5868 	bp->cfa_code_map = NULL;
5869 }
5870 
5871 static int bnxt_init_rep_info(struct bnxt *bp)
5872 {
5873 	int i = 0, rc;
5874 
5875 	if (bp->rep_info)
5876 		return 0;
5877 
5878 	bp->rep_info = rte_zmalloc("bnxt_rep_info",
5879 				   sizeof(bp->rep_info[0]) * BNXT_MAX_VF_REPS,
5880 				   0);
5881 	if (!bp->rep_info) {
5882 		PMD_DRV_LOG(ERR, "Failed to alloc memory for rep info\n");
5883 		return -ENOMEM;
5884 	}
5885 	bp->cfa_code_map = rte_zmalloc("bnxt_cfa_code_map",
5886 				       sizeof(*bp->cfa_code_map) *
5887 				       BNXT_MAX_CFA_CODE, 0);
5888 	if (!bp->cfa_code_map) {
5889 		PMD_DRV_LOG(ERR, "Failed to alloc memory for cfa_code_map\n");
5890 		bnxt_free_rep_info(bp);
5891 		return -ENOMEM;
5892 	}
5893 
5894 	for (i = 0; i < BNXT_MAX_CFA_CODE; i++)
5895 		bp->cfa_code_map[i] = BNXT_VF_IDX_INVALID;
5896 
5897 	rc = pthread_mutex_init(&bp->rep_info->vfr_lock, NULL);
5898 	if (rc) {
5899 		PMD_DRV_LOG(ERR, "Unable to initialize vfr_lock\n");
5900 		bnxt_free_rep_info(bp);
5901 		return rc;
5902 	}
5903 
5904 	rc = pthread_mutex_init(&bp->rep_info->vfr_start_lock, NULL);
5905 	if (rc) {
5906 		PMD_DRV_LOG(ERR, "Unable to initialize vfr_start_lock\n");
5907 		bnxt_free_rep_info(bp);
5908 		return rc;
5909 	}
5910 
5911 	return rc;
5912 }
5913 
5914 static int bnxt_rep_port_probe(struct rte_pci_device *pci_dev,
5915 			       struct rte_eth_devargs *eth_da,
5916 			       struct rte_eth_dev *backing_eth_dev,
5917 			       const char *dev_args)
5918 {
5919 	struct rte_eth_dev *vf_rep_eth_dev;
5920 	char name[RTE_ETH_NAME_MAX_LEN];
5921 	struct bnxt *backing_bp;
5922 	uint16_t num_rep;
5923 	int i, ret = 0;
5924 	struct rte_kvargs *kvlist = NULL;
5925 
5926 	if (eth_da->type == RTE_ETH_REPRESENTOR_NONE)
5927 		return 0;
5928 	if (eth_da->type != RTE_ETH_REPRESENTOR_VF) {
5929 		PMD_DRV_LOG(ERR, "unsupported representor type %d\n",
5930 			    eth_da->type);
5931 		return -ENOTSUP;
5932 	}
5933 	num_rep = eth_da->nb_representor_ports;
5934 	if (num_rep > BNXT_MAX_VF_REPS) {
5935 		PMD_DRV_LOG(ERR, "nb_representor_ports = %d > %d MAX VF REPS\n",
5936 			    num_rep, BNXT_MAX_VF_REPS);
5937 		return -EINVAL;
5938 	}
5939 
5940 	if (num_rep >= RTE_MAX_ETHPORTS) {
5941 		PMD_DRV_LOG(ERR,
5942 			    "nb_representor_ports = %d > %d MAX ETHPORTS\n",
5943 			    num_rep, RTE_MAX_ETHPORTS);
5944 		return -EINVAL;
5945 	}
5946 
5947 	backing_bp = backing_eth_dev->data->dev_private;
5948 
5949 	if (!(BNXT_PF(backing_bp) || BNXT_VF_IS_TRUSTED(backing_bp))) {
5950 		PMD_DRV_LOG(ERR,
5951 			    "Not a PF or trusted VF. No Representor support\n");
5952 		/* Returning an error is not an option.
5953 		 * Applications are not handling this correctly
5954 		 */
5955 		return 0;
5956 	}
5957 
5958 	if (bnxt_init_rep_info(backing_bp))
5959 		return 0;
5960 
5961 	for (i = 0; i < num_rep; i++) {
5962 		struct bnxt_representor representor = {
5963 			.vf_id = eth_da->representor_ports[i],
5964 			.switch_domain_id = backing_bp->switch_domain_id,
5965 			.parent_dev = backing_eth_dev
5966 		};
5967 
5968 		if (representor.vf_id >= BNXT_MAX_VF_REPS) {
5969 			PMD_DRV_LOG(ERR, "VF-Rep id %d >= %d MAX VF ID\n",
5970 				    representor.vf_id, BNXT_MAX_VF_REPS);
5971 			continue;
5972 		}
5973 
5974 		/* representor port net_bdf_port */
5975 		snprintf(name, sizeof(name), "net_%s_representor_%d",
5976 			 pci_dev->device.name, eth_da->representor_ports[i]);
5977 
5978 		kvlist = rte_kvargs_parse(dev_args, bnxt_dev_args);
5979 		if (kvlist) {
5980 			/*
5981 			 * Handler for "rep_is_pf" devarg.
5982 			 * Invoked as for ex: "-a 000:00:0d.0,
5983 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
5984 			 */
5985 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_IS_PF,
5986 						 bnxt_parse_devarg_rep_is_pf,
5987 						 (void *)&representor);
5988 			if (ret) {
5989 				ret = -EINVAL;
5990 				goto err;
5991 			}
5992 			/*
5993 			 * Handler for "rep_based_pf" devarg.
5994 			 * Invoked as for ex: "-a 000:00:0d.0,
5995 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
5996 			 */
5997 			ret = rte_kvargs_process(kvlist,
5998 						 BNXT_DEVARG_REP_BASED_PF,
5999 						 bnxt_parse_devarg_rep_based_pf,
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_Q_R2F,
6011 						 bnxt_parse_devarg_rep_q_r2f,
6012 						 (void *)&representor);
6013 			if (ret) {
6014 				ret = -EINVAL;
6015 				goto err;
6016 			}
6017 			/*
6018 			 * Handler for "rep_based_pf" devarg.
6019 			 * Invoked as for ex: "-a 000:00:0d.0,
6020 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6021 			 */
6022 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_Q_F2R,
6023 						 bnxt_parse_devarg_rep_q_f2r,
6024 						 (void *)&representor);
6025 			if (ret) {
6026 				ret = -EINVAL;
6027 				goto err;
6028 			}
6029 			/*
6030 			 * Handler for "rep_based_pf" devarg.
6031 			 * Invoked as for ex: "-a 000:00:0d.0,
6032 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6033 			 */
6034 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_FC_R2F,
6035 						 bnxt_parse_devarg_rep_fc_r2f,
6036 						 (void *)&representor);
6037 			if (ret) {
6038 				ret = -EINVAL;
6039 				goto err;
6040 			}
6041 			/*
6042 			 * Handler for "rep_based_pf" devarg.
6043 			 * Invoked as for ex: "-a 000:00:0d.0,
6044 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6045 			 */
6046 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_FC_F2R,
6047 						 bnxt_parse_devarg_rep_fc_f2r,
6048 						 (void *)&representor);
6049 			if (ret) {
6050 				ret = -EINVAL;
6051 				goto err;
6052 			}
6053 		}
6054 
6055 		ret = rte_eth_dev_create(&pci_dev->device, name,
6056 					 sizeof(struct bnxt_representor),
6057 					 NULL, NULL,
6058 					 bnxt_representor_init,
6059 					 &representor);
6060 		if (ret) {
6061 			PMD_DRV_LOG(ERR, "failed to create bnxt vf "
6062 				    "representor %s.", name);
6063 			goto err;
6064 		}
6065 
6066 		vf_rep_eth_dev = rte_eth_dev_allocated(name);
6067 		if (!vf_rep_eth_dev) {
6068 			PMD_DRV_LOG(ERR, "Failed to find the eth_dev"
6069 				    " for VF-Rep: %s.", name);
6070 			ret = -ENODEV;
6071 			goto err;
6072 		}
6073 
6074 		PMD_DRV_LOG(DEBUG, "BNXT Port:%d VFR pci probe\n",
6075 			    backing_eth_dev->data->port_id);
6076 		backing_bp->rep_info[representor.vf_id].vfr_eth_dev =
6077 							 vf_rep_eth_dev;
6078 		backing_bp->num_reps++;
6079 
6080 	}
6081 
6082 	rte_kvargs_free(kvlist);
6083 	return 0;
6084 
6085 err:
6086 	/* If num_rep > 1, then rollback already created
6087 	 * ports, since we'll be failing the probe anyway
6088 	 */
6089 	if (num_rep > 1)
6090 		bnxt_pci_remove_dev_with_reps(backing_eth_dev);
6091 	rte_errno = -ret;
6092 	rte_kvargs_free(kvlist);
6093 
6094 	return ret;
6095 }
6096 
6097 static int bnxt_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
6098 			  struct rte_pci_device *pci_dev)
6099 {
6100 	struct rte_eth_devargs eth_da = { .nb_representor_ports = 0 };
6101 	struct rte_eth_dev *backing_eth_dev;
6102 	uint16_t num_rep;
6103 	int ret = 0;
6104 
6105 	if (pci_dev->device.devargs) {
6106 		ret = rte_eth_devargs_parse(pci_dev->device.devargs->args,
6107 					    &eth_da);
6108 		if (ret)
6109 			return ret;
6110 	}
6111 
6112 	num_rep = eth_da.nb_representor_ports;
6113 	PMD_DRV_LOG(DEBUG, "nb_representor_ports = %d\n",
6114 		    num_rep);
6115 
6116 	/* We could come here after first level of probe is already invoked
6117 	 * as part of an application bringup(OVS-DPDK vswitchd), so first check
6118 	 * for already allocated eth_dev for the backing device (PF/Trusted VF)
6119 	 */
6120 	backing_eth_dev = rte_eth_dev_allocated(pci_dev->device.name);
6121 	if (backing_eth_dev == NULL) {
6122 		ret = rte_eth_dev_create(&pci_dev->device, pci_dev->device.name,
6123 					 sizeof(struct bnxt),
6124 					 eth_dev_pci_specific_init, pci_dev,
6125 					 bnxt_dev_init, NULL);
6126 
6127 		if (ret || !num_rep)
6128 			return ret;
6129 
6130 		backing_eth_dev = rte_eth_dev_allocated(pci_dev->device.name);
6131 	}
6132 	PMD_DRV_LOG(DEBUG, "BNXT Port:%d pci probe\n",
6133 		    backing_eth_dev->data->port_id);
6134 
6135 	if (!num_rep)
6136 		return ret;
6137 
6138 	/* probe representor ports now */
6139 	ret = bnxt_rep_port_probe(pci_dev, &eth_da, backing_eth_dev,
6140 				  pci_dev->device.devargs->args);
6141 
6142 	return ret;
6143 }
6144 
6145 static int bnxt_pci_remove(struct rte_pci_device *pci_dev)
6146 {
6147 	struct rte_eth_dev *eth_dev;
6148 
6149 	eth_dev = rte_eth_dev_allocated(pci_dev->device.name);
6150 	if (!eth_dev)
6151 		return 0; /* Invoked typically only by OVS-DPDK, by the
6152 			   * time it comes here the eth_dev is already
6153 			   * deleted by rte_eth_dev_close(), so returning
6154 			   * +ve value will at least help in proper cleanup
6155 			   */
6156 
6157 	PMD_DRV_LOG(DEBUG, "BNXT Port:%d pci remove\n", eth_dev->data->port_id);
6158 	if (rte_eal_process_type() == RTE_PROC_PRIMARY) {
6159 		if (eth_dev->data->dev_flags & RTE_ETH_DEV_REPRESENTOR)
6160 			return rte_eth_dev_destroy(eth_dev,
6161 						   bnxt_representor_uninit);
6162 		else
6163 			return rte_eth_dev_destroy(eth_dev,
6164 						   bnxt_dev_uninit);
6165 	} else {
6166 		return rte_eth_dev_pci_generic_remove(pci_dev, NULL);
6167 	}
6168 }
6169 
6170 static struct rte_pci_driver bnxt_rte_pmd = {
6171 	.id_table = bnxt_pci_id_map,
6172 	.drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC |
6173 			RTE_PCI_DRV_PROBE_AGAIN, /* Needed in case of VF-REPs
6174 						  * and OVS-DPDK
6175 						  */
6176 	.probe = bnxt_pci_probe,
6177 	.remove = bnxt_pci_remove,
6178 };
6179 
6180 static bool
6181 is_device_supported(struct rte_eth_dev *dev, struct rte_pci_driver *drv)
6182 {
6183 	if (strcmp(dev->device->driver->name, drv->driver.name))
6184 		return false;
6185 
6186 	return true;
6187 }
6188 
6189 bool is_bnxt_supported(struct rte_eth_dev *dev)
6190 {
6191 	return is_device_supported(dev, &bnxt_rte_pmd);
6192 }
6193 
6194 RTE_LOG_REGISTER_SUFFIX(bnxt_logtype_driver, driver, NOTICE);
6195 RTE_PMD_REGISTER_PCI(net_bnxt, bnxt_rte_pmd);
6196 RTE_PMD_REGISTER_PCI_TABLE(net_bnxt, bnxt_pci_id_map);
6197 RTE_PMD_REGISTER_KMOD_DEP(net_bnxt, "* igb_uio | uio_pci_generic | vfio-pci");
6198