xref: /dpdk/drivers/net/bnxt/bnxt_ethdev.c (revision ee2cf75e)
1 /* SPDX-License-Identifier: BSD-3-Clause
2  * Copyright(c) 2014-2018 Broadcom
3  * All rights reserved.
4  */
5 
6 #include <inttypes.h>
7 #include <stdbool.h>
8 
9 #include <rte_dev.h>
10 #include <rte_ethdev_driver.h>
11 #include <rte_ethdev_pci.h>
12 #include <rte_malloc.h>
13 #include <rte_cycles.h>
14 #include <rte_alarm.h>
15 #include <rte_kvargs.h>
16 
17 #include "bnxt.h"
18 #include "bnxt_filter.h"
19 #include "bnxt_hwrm.h"
20 #include "bnxt_irq.h"
21 #include "bnxt_reps.h"
22 #include "bnxt_ring.h"
23 #include "bnxt_rxq.h"
24 #include "bnxt_rxr.h"
25 #include "bnxt_stats.h"
26 #include "bnxt_txq.h"
27 #include "bnxt_txr.h"
28 #include "bnxt_vnic.h"
29 #include "hsi_struct_def_dpdk.h"
30 #include "bnxt_nvm_defs.h"
31 #include "bnxt_tf_common.h"
32 #include "ulp_flow_db.h"
33 
34 #define DRV_MODULE_NAME		"bnxt"
35 static const char bnxt_version[] =
36 	"Broadcom NetXtreme driver " DRV_MODULE_NAME;
37 
38 /*
39  * The set of PCI devices this driver supports
40  */
41 static const struct rte_pci_id bnxt_pci_id_map[] = {
42 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM,
43 			 BROADCOM_DEV_ID_STRATUS_NIC_VF1) },
44 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM,
45 			 BROADCOM_DEV_ID_STRATUS_NIC_VF2) },
46 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_STRATUS_NIC) },
47 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57414_VF) },
48 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57301) },
49 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57302) },
50 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57304_PF) },
51 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57304_VF) },
52 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_NS2) },
53 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57402) },
54 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57404) },
55 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57406_PF) },
56 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57406_VF) },
57 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57402_MF) },
58 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57407_RJ45) },
59 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57404_MF) },
60 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57406_MF) },
61 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57407_SFP) },
62 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57407_MF) },
63 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_5741X_VF) },
64 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_5731X_VF) },
65 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57314) },
66 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57417_MF) },
67 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57311) },
68 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57312) },
69 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57412) },
70 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57414) },
71 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57416_RJ45) },
72 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57417_RJ45) },
73 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57412_MF) },
74 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57317_RJ45) },
75 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57417_SFP) },
76 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57416_SFP) },
77 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57317_SFP) },
78 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57414_MF) },
79 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57416_MF) },
80 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58802) },
81 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58804) },
82 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58808) },
83 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58802_VF) },
84 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57508) },
85 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57504) },
86 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57502) },
87 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57500_VF1) },
88 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57500_VF2) },
89 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57508_MF1) },
90 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57504_MF1) },
91 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57502_MF1) },
92 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57508_MF2) },
93 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57504_MF2) },
94 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57502_MF2) },
95 	{ .vendor_id = 0, /* sentinel */ },
96 };
97 
98 #define BNXT_DEVARG_TRUFLOW	"host-based-truflow"
99 #define BNXT_DEVARG_FLOW_XSTAT	"flow-xstat"
100 #define BNXT_DEVARG_MAX_NUM_KFLOWS  "max-num-kflows"
101 #define BNXT_DEVARG_REPRESENTOR	"representor"
102 
103 static const char *const bnxt_dev_args[] = {
104 	BNXT_DEVARG_REPRESENTOR,
105 	BNXT_DEVARG_TRUFLOW,
106 	BNXT_DEVARG_FLOW_XSTAT,
107 	BNXT_DEVARG_MAX_NUM_KFLOWS,
108 	NULL
109 };
110 
111 /*
112  * truflow == false to disable the feature
113  * truflow == true to enable the feature
114  */
115 #define	BNXT_DEVARG_TRUFLOW_INVALID(truflow)	((truflow) > 1)
116 
117 /*
118  * flow_xstat == false to disable the feature
119  * flow_xstat == true to enable the feature
120  */
121 #define	BNXT_DEVARG_FLOW_XSTAT_INVALID(flow_xstat)	((flow_xstat) > 1)
122 
123 /*
124  * max_num_kflows must be >= 32
125  * and must be a power-of-2 supported value
126  * return: 1 -> invalid
127  *         0 -> valid
128  */
129 static int bnxt_devarg_max_num_kflow_invalid(uint16_t max_num_kflows)
130 {
131 	if (max_num_kflows < 32 || !rte_is_power_of_2(max_num_kflows))
132 		return 1;
133 	return 0;
134 }
135 
136 static int bnxt_vlan_offload_set_op(struct rte_eth_dev *dev, int mask);
137 static int bnxt_dev_uninit(struct rte_eth_dev *eth_dev);
138 static int bnxt_init_resources(struct bnxt *bp, bool reconfig_dev);
139 static int bnxt_uninit_resources(struct bnxt *bp, bool reconfig_dev);
140 static void bnxt_cancel_fw_health_check(struct bnxt *bp);
141 static int bnxt_restore_vlan_filters(struct bnxt *bp);
142 static void bnxt_dev_recover(void *arg);
143 static void bnxt_free_error_recovery_info(struct bnxt *bp);
144 static void bnxt_free_rep_info(struct bnxt *bp);
145 
146 int is_bnxt_in_error(struct bnxt *bp)
147 {
148 	if (bp->flags & BNXT_FLAG_FATAL_ERROR)
149 		return -EIO;
150 	if (bp->flags & BNXT_FLAG_FW_RESET)
151 		return -EBUSY;
152 
153 	return 0;
154 }
155 
156 /***********************/
157 
158 /*
159  * High level utility functions
160  */
161 
162 static uint16_t bnxt_rss_ctxts(const struct bnxt *bp)
163 {
164 	if (!BNXT_CHIP_THOR(bp))
165 		return 1;
166 
167 	return RTE_ALIGN_MUL_CEIL(bp->rx_nr_rings,
168 				  BNXT_RSS_ENTRIES_PER_CTX_THOR) /
169 				    BNXT_RSS_ENTRIES_PER_CTX_THOR;
170 }
171 
172 uint16_t bnxt_rss_hash_tbl_size(const struct bnxt *bp)
173 {
174 	if (!BNXT_CHIP_THOR(bp))
175 		return HW_HASH_INDEX_SIZE;
176 
177 	return bnxt_rss_ctxts(bp) * BNXT_RSS_ENTRIES_PER_CTX_THOR;
178 }
179 
180 static void bnxt_free_parent_info(struct bnxt *bp)
181 {
182 	rte_free(bp->parent);
183 }
184 
185 static void bnxt_free_pf_info(struct bnxt *bp)
186 {
187 	rte_free(bp->pf);
188 }
189 
190 static void bnxt_free_link_info(struct bnxt *bp)
191 {
192 	rte_free(bp->link_info);
193 }
194 
195 static void bnxt_free_leds_info(struct bnxt *bp)
196 {
197 	if (BNXT_VF(bp))
198 		return;
199 
200 	rte_free(bp->leds);
201 	bp->leds = NULL;
202 }
203 
204 static void bnxt_free_flow_stats_info(struct bnxt *bp)
205 {
206 	rte_free(bp->flow_stat);
207 	bp->flow_stat = NULL;
208 }
209 
210 static void bnxt_free_cos_queues(struct bnxt *bp)
211 {
212 	rte_free(bp->rx_cos_queue);
213 	rte_free(bp->tx_cos_queue);
214 }
215 
216 static void bnxt_free_mem(struct bnxt *bp, bool reconfig)
217 {
218 	bnxt_free_filter_mem(bp);
219 	bnxt_free_vnic_attributes(bp);
220 	bnxt_free_vnic_mem(bp);
221 
222 	/* tx/rx rings are configured as part of *_queue_setup callbacks.
223 	 * If the number of rings change across fw update,
224 	 * we don't have much choice except to warn the user.
225 	 */
226 	if (!reconfig) {
227 		bnxt_free_stats(bp);
228 		bnxt_free_tx_rings(bp);
229 		bnxt_free_rx_rings(bp);
230 	}
231 	bnxt_free_async_cp_ring(bp);
232 	bnxt_free_rxtx_nq_ring(bp);
233 
234 	rte_free(bp->grp_info);
235 	bp->grp_info = NULL;
236 }
237 
238 static int bnxt_alloc_parent_info(struct bnxt *bp)
239 {
240 	bp->parent = rte_zmalloc("bnxt_parent_info",
241 				 sizeof(struct bnxt_parent_info), 0);
242 	if (bp->parent == NULL)
243 		return -ENOMEM;
244 
245 	return 0;
246 }
247 
248 static int bnxt_alloc_pf_info(struct bnxt *bp)
249 {
250 	bp->pf = rte_zmalloc("bnxt_pf_info", sizeof(struct bnxt_pf_info), 0);
251 	if (bp->pf == NULL)
252 		return -ENOMEM;
253 
254 	return 0;
255 }
256 
257 static int bnxt_alloc_link_info(struct bnxt *bp)
258 {
259 	bp->link_info =
260 		rte_zmalloc("bnxt_link_info", sizeof(struct bnxt_link_info), 0);
261 	if (bp->link_info == NULL)
262 		return -ENOMEM;
263 
264 	return 0;
265 }
266 
267 static int bnxt_alloc_leds_info(struct bnxt *bp)
268 {
269 	if (BNXT_VF(bp))
270 		return 0;
271 
272 	bp->leds = rte_zmalloc("bnxt_leds",
273 			       BNXT_MAX_LED * sizeof(struct bnxt_led_info),
274 			       0);
275 	if (bp->leds == NULL)
276 		return -ENOMEM;
277 
278 	return 0;
279 }
280 
281 static int bnxt_alloc_cos_queues(struct bnxt *bp)
282 {
283 	bp->rx_cos_queue =
284 		rte_zmalloc("bnxt_rx_cosq",
285 			    BNXT_COS_QUEUE_COUNT *
286 			    sizeof(struct bnxt_cos_queue_info),
287 			    0);
288 	if (bp->rx_cos_queue == NULL)
289 		return -ENOMEM;
290 
291 	bp->tx_cos_queue =
292 		rte_zmalloc("bnxt_tx_cosq",
293 			    BNXT_COS_QUEUE_COUNT *
294 			    sizeof(struct bnxt_cos_queue_info),
295 			    0);
296 	if (bp->tx_cos_queue == NULL)
297 		return -ENOMEM;
298 
299 	return 0;
300 }
301 
302 static int bnxt_alloc_flow_stats_info(struct bnxt *bp)
303 {
304 	bp->flow_stat = rte_zmalloc("bnxt_flow_xstat",
305 				    sizeof(struct bnxt_flow_stat_info), 0);
306 	if (bp->flow_stat == NULL)
307 		return -ENOMEM;
308 
309 	return 0;
310 }
311 
312 static int bnxt_alloc_mem(struct bnxt *bp, bool reconfig)
313 {
314 	int rc;
315 
316 	rc = bnxt_alloc_ring_grps(bp);
317 	if (rc)
318 		goto alloc_mem_err;
319 
320 	rc = bnxt_alloc_async_ring_struct(bp);
321 	if (rc)
322 		goto alloc_mem_err;
323 
324 	rc = bnxt_alloc_vnic_mem(bp);
325 	if (rc)
326 		goto alloc_mem_err;
327 
328 	rc = bnxt_alloc_vnic_attributes(bp);
329 	if (rc)
330 		goto alloc_mem_err;
331 
332 	rc = bnxt_alloc_filter_mem(bp);
333 	if (rc)
334 		goto alloc_mem_err;
335 
336 	rc = bnxt_alloc_async_cp_ring(bp);
337 	if (rc)
338 		goto alloc_mem_err;
339 
340 	rc = bnxt_alloc_rxtx_nq_ring(bp);
341 	if (rc)
342 		goto alloc_mem_err;
343 
344 	if (BNXT_FLOW_XSTATS_EN(bp)) {
345 		rc = bnxt_alloc_flow_stats_info(bp);
346 		if (rc)
347 			goto alloc_mem_err;
348 	}
349 
350 	return 0;
351 
352 alloc_mem_err:
353 	bnxt_free_mem(bp, reconfig);
354 	return rc;
355 }
356 
357 static int bnxt_setup_one_vnic(struct bnxt *bp, uint16_t vnic_id)
358 {
359 	struct rte_eth_conf *dev_conf = &bp->eth_dev->data->dev_conf;
360 	struct bnxt_vnic_info *vnic = &bp->vnic_info[vnic_id];
361 	uint64_t rx_offloads = dev_conf->rxmode.offloads;
362 	struct bnxt_rx_queue *rxq;
363 	unsigned int j;
364 	int rc;
365 
366 	rc = bnxt_vnic_grp_alloc(bp, vnic);
367 	if (rc)
368 		goto err_out;
369 
370 	PMD_DRV_LOG(DEBUG, "vnic[%d] = %p vnic->fw_grp_ids = %p\n",
371 		    vnic_id, vnic, vnic->fw_grp_ids);
372 
373 	rc = bnxt_hwrm_vnic_alloc(bp, vnic);
374 	if (rc)
375 		goto err_out;
376 
377 	/* Alloc RSS context only if RSS mode is enabled */
378 	if (dev_conf->rxmode.mq_mode & ETH_MQ_RX_RSS) {
379 		int j, nr_ctxs = bnxt_rss_ctxts(bp);
380 
381 		rc = 0;
382 		for (j = 0; j < nr_ctxs; j++) {
383 			rc = bnxt_hwrm_vnic_ctx_alloc(bp, vnic, j);
384 			if (rc)
385 				break;
386 		}
387 		if (rc) {
388 			PMD_DRV_LOG(ERR,
389 				    "HWRM vnic %d ctx %d alloc failure rc: %x\n",
390 				    vnic_id, j, rc);
391 			goto err_out;
392 		}
393 		vnic->num_lb_ctxts = nr_ctxs;
394 	}
395 
396 	/*
397 	 * Firmware sets pf pair in default vnic cfg. If the VLAN strip
398 	 * setting is not available at this time, it will not be
399 	 * configured correctly in the CFA.
400 	 */
401 	if (rx_offloads & DEV_RX_OFFLOAD_VLAN_STRIP)
402 		vnic->vlan_strip = true;
403 	else
404 		vnic->vlan_strip = false;
405 
406 	rc = bnxt_hwrm_vnic_cfg(bp, vnic);
407 	if (rc)
408 		goto err_out;
409 
410 	rc = bnxt_set_hwrm_vnic_filters(bp, vnic);
411 	if (rc)
412 		goto err_out;
413 
414 	for (j = 0; j < bp->rx_num_qs_per_vnic; j++) {
415 		rxq = bp->eth_dev->data->rx_queues[j];
416 
417 		PMD_DRV_LOG(DEBUG,
418 			    "rxq[%d]->vnic=%p vnic->fw_grp_ids=%p\n",
419 			    j, rxq->vnic, rxq->vnic->fw_grp_ids);
420 
421 		if (BNXT_HAS_RING_GRPS(bp) && rxq->rx_deferred_start)
422 			rxq->vnic->fw_grp_ids[j] = INVALID_HW_RING_ID;
423 		else
424 			vnic->rx_queue_cnt++;
425 	}
426 
427 	PMD_DRV_LOG(DEBUG, "vnic->rx_queue_cnt = %d\n", vnic->rx_queue_cnt);
428 
429 	rc = bnxt_vnic_rss_configure(bp, vnic);
430 	if (rc)
431 		goto err_out;
432 
433 	bnxt_hwrm_vnic_plcmode_cfg(bp, vnic);
434 
435 	if (rx_offloads & DEV_RX_OFFLOAD_TCP_LRO)
436 		bnxt_hwrm_vnic_tpa_cfg(bp, vnic, 1);
437 	else
438 		bnxt_hwrm_vnic_tpa_cfg(bp, vnic, 0);
439 
440 	return 0;
441 err_out:
442 	PMD_DRV_LOG(ERR, "HWRM vnic %d cfg failure rc: %x\n",
443 		    vnic_id, rc);
444 	return rc;
445 }
446 
447 static int bnxt_register_fc_ctx_mem(struct bnxt *bp)
448 {
449 	int rc = 0;
450 
451 	rc = bnxt_hwrm_ctx_rgtr(bp, bp->flow_stat->rx_fc_in_tbl.dma,
452 				&bp->flow_stat->rx_fc_in_tbl.ctx_id);
453 	if (rc)
454 		return rc;
455 
456 	PMD_DRV_LOG(DEBUG,
457 		    "rx_fc_in_tbl.va = %p rx_fc_in_tbl.dma = %p"
458 		    " rx_fc_in_tbl.ctx_id = %d\n",
459 		    bp->flow_stat->rx_fc_in_tbl.va,
460 		    (void *)((uintptr_t)bp->flow_stat->rx_fc_in_tbl.dma),
461 		    bp->flow_stat->rx_fc_in_tbl.ctx_id);
462 
463 	rc = bnxt_hwrm_ctx_rgtr(bp, bp->flow_stat->rx_fc_out_tbl.dma,
464 				&bp->flow_stat->rx_fc_out_tbl.ctx_id);
465 	if (rc)
466 		return rc;
467 
468 	PMD_DRV_LOG(DEBUG,
469 		    "rx_fc_out_tbl.va = %p rx_fc_out_tbl.dma = %p"
470 		    " rx_fc_out_tbl.ctx_id = %d\n",
471 		    bp->flow_stat->rx_fc_out_tbl.va,
472 		    (void *)((uintptr_t)bp->flow_stat->rx_fc_out_tbl.dma),
473 		    bp->flow_stat->rx_fc_out_tbl.ctx_id);
474 
475 	rc = bnxt_hwrm_ctx_rgtr(bp, bp->flow_stat->tx_fc_in_tbl.dma,
476 				&bp->flow_stat->tx_fc_in_tbl.ctx_id);
477 	if (rc)
478 		return rc;
479 
480 	PMD_DRV_LOG(DEBUG,
481 		    "tx_fc_in_tbl.va = %p tx_fc_in_tbl.dma = %p"
482 		    " tx_fc_in_tbl.ctx_id = %d\n",
483 		    bp->flow_stat->tx_fc_in_tbl.va,
484 		    (void *)((uintptr_t)bp->flow_stat->tx_fc_in_tbl.dma),
485 		    bp->flow_stat->tx_fc_in_tbl.ctx_id);
486 
487 	rc = bnxt_hwrm_ctx_rgtr(bp, bp->flow_stat->tx_fc_out_tbl.dma,
488 				&bp->flow_stat->tx_fc_out_tbl.ctx_id);
489 	if (rc)
490 		return rc;
491 
492 	PMD_DRV_LOG(DEBUG,
493 		    "tx_fc_out_tbl.va = %p tx_fc_out_tbl.dma = %p"
494 		    " tx_fc_out_tbl.ctx_id = %d\n",
495 		    bp->flow_stat->tx_fc_out_tbl.va,
496 		    (void *)((uintptr_t)bp->flow_stat->tx_fc_out_tbl.dma),
497 		    bp->flow_stat->tx_fc_out_tbl.ctx_id);
498 
499 	memset(bp->flow_stat->rx_fc_out_tbl.va,
500 	       0,
501 	       bp->flow_stat->rx_fc_out_tbl.size);
502 	rc = bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_RX,
503 				       CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
504 				       bp->flow_stat->rx_fc_out_tbl.ctx_id,
505 				       bp->flow_stat->max_fc,
506 				       true);
507 	if (rc)
508 		return rc;
509 
510 	memset(bp->flow_stat->tx_fc_out_tbl.va,
511 	       0,
512 	       bp->flow_stat->tx_fc_out_tbl.size);
513 	rc = bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_TX,
514 				       CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
515 				       bp->flow_stat->tx_fc_out_tbl.ctx_id,
516 				       bp->flow_stat->max_fc,
517 				       true);
518 
519 	return rc;
520 }
521 
522 static int bnxt_alloc_ctx_mem_buf(char *type, size_t size,
523 				  struct bnxt_ctx_mem_buf_info *ctx)
524 {
525 	if (!ctx)
526 		return -EINVAL;
527 
528 	ctx->va = rte_zmalloc(type, size, 0);
529 	if (ctx->va == NULL)
530 		return -ENOMEM;
531 	rte_mem_lock_page(ctx->va);
532 	ctx->size = size;
533 	ctx->dma = rte_mem_virt2iova(ctx->va);
534 	if (ctx->dma == RTE_BAD_IOVA)
535 		return -ENOMEM;
536 
537 	return 0;
538 }
539 
540 static int bnxt_init_fc_ctx_mem(struct bnxt *bp)
541 {
542 	struct rte_pci_device *pdev = bp->pdev;
543 	char type[RTE_MEMZONE_NAMESIZE];
544 	uint16_t max_fc;
545 	int rc = 0;
546 
547 	max_fc = bp->flow_stat->max_fc;
548 
549 	sprintf(type, "bnxt_rx_fc_in_" PCI_PRI_FMT, pdev->addr.domain,
550 		pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
551 	/* 4 bytes for each counter-id */
552 	rc = bnxt_alloc_ctx_mem_buf(type,
553 				    max_fc * 4,
554 				    &bp->flow_stat->rx_fc_in_tbl);
555 	if (rc)
556 		return rc;
557 
558 	sprintf(type, "bnxt_rx_fc_out_" PCI_PRI_FMT, pdev->addr.domain,
559 		pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
560 	/* 16 bytes for each counter - 8 bytes pkt_count, 8 bytes byte_count */
561 	rc = bnxt_alloc_ctx_mem_buf(type,
562 				    max_fc * 16,
563 				    &bp->flow_stat->rx_fc_out_tbl);
564 	if (rc)
565 		return rc;
566 
567 	sprintf(type, "bnxt_tx_fc_in_" PCI_PRI_FMT, pdev->addr.domain,
568 		pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
569 	/* 4 bytes for each counter-id */
570 	rc = bnxt_alloc_ctx_mem_buf(type,
571 				    max_fc * 4,
572 				    &bp->flow_stat->tx_fc_in_tbl);
573 	if (rc)
574 		return rc;
575 
576 	sprintf(type, "bnxt_tx_fc_out_" PCI_PRI_FMT, pdev->addr.domain,
577 		pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
578 	/* 16 bytes for each counter - 8 bytes pkt_count, 8 bytes byte_count */
579 	rc = bnxt_alloc_ctx_mem_buf(type,
580 				    max_fc * 16,
581 				    &bp->flow_stat->tx_fc_out_tbl);
582 	if (rc)
583 		return rc;
584 
585 	rc = bnxt_register_fc_ctx_mem(bp);
586 
587 	return rc;
588 }
589 
590 static int bnxt_init_ctx_mem(struct bnxt *bp)
591 {
592 	int rc = 0;
593 
594 	if (!(bp->fw_cap & BNXT_FW_CAP_ADV_FLOW_COUNTERS) ||
595 	    !(BNXT_PF(bp) || BNXT_VF_IS_TRUSTED(bp)) ||
596 	    !BNXT_FLOW_XSTATS_EN(bp))
597 		return 0;
598 
599 	rc = bnxt_hwrm_cfa_counter_qcaps(bp, &bp->flow_stat->max_fc);
600 	if (rc)
601 		return rc;
602 
603 	rc = bnxt_init_fc_ctx_mem(bp);
604 
605 	return rc;
606 }
607 
608 static int bnxt_init_chip(struct bnxt *bp)
609 {
610 	struct rte_eth_link new;
611 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(bp->eth_dev);
612 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
613 	uint32_t intr_vector = 0;
614 	uint32_t queue_id, base = BNXT_MISC_VEC_ID;
615 	uint32_t vec = BNXT_MISC_VEC_ID;
616 	unsigned int i, j;
617 	int rc;
618 
619 	if (bp->eth_dev->data->mtu > RTE_ETHER_MTU) {
620 		bp->eth_dev->data->dev_conf.rxmode.offloads |=
621 			DEV_RX_OFFLOAD_JUMBO_FRAME;
622 		bp->flags |= BNXT_FLAG_JUMBO;
623 	} else {
624 		bp->eth_dev->data->dev_conf.rxmode.offloads &=
625 			~DEV_RX_OFFLOAD_JUMBO_FRAME;
626 		bp->flags &= ~BNXT_FLAG_JUMBO;
627 	}
628 
629 	/* THOR does not support ring groups.
630 	 * But we will use the array to save RSS context IDs.
631 	 */
632 	if (BNXT_CHIP_THOR(bp))
633 		bp->max_ring_grps = BNXT_MAX_RSS_CTXTS_THOR;
634 
635 	rc = bnxt_alloc_all_hwrm_stat_ctxs(bp);
636 	if (rc) {
637 		PMD_DRV_LOG(ERR, "HWRM stat ctx alloc failure rc: %x\n", rc);
638 		goto err_out;
639 	}
640 
641 	rc = bnxt_alloc_hwrm_rings(bp);
642 	if (rc) {
643 		PMD_DRV_LOG(ERR, "HWRM ring alloc failure rc: %x\n", rc);
644 		goto err_out;
645 	}
646 
647 	rc = bnxt_alloc_all_hwrm_ring_grps(bp);
648 	if (rc) {
649 		PMD_DRV_LOG(ERR, "HWRM ring grp alloc failure: %x\n", rc);
650 		goto err_out;
651 	}
652 
653 	if (!(bp->vnic_cap_flags & BNXT_VNIC_CAP_COS_CLASSIFY))
654 		goto skip_cosq_cfg;
655 
656 	for (j = 0, i = 0; i < BNXT_COS_QUEUE_COUNT; i++) {
657 		if (bp->rx_cos_queue[i].id != 0xff) {
658 			struct bnxt_vnic_info *vnic = &bp->vnic_info[j++];
659 
660 			if (!vnic) {
661 				PMD_DRV_LOG(ERR,
662 					    "Num pools more than FW profile\n");
663 				rc = -EINVAL;
664 				goto err_out;
665 			}
666 			vnic->cos_queue_id = bp->rx_cos_queue[i].id;
667 			bp->rx_cosq_cnt++;
668 		}
669 	}
670 
671 skip_cosq_cfg:
672 	rc = bnxt_mq_rx_configure(bp);
673 	if (rc) {
674 		PMD_DRV_LOG(ERR, "MQ mode configure failure rc: %x\n", rc);
675 		goto err_out;
676 	}
677 
678 	/* VNIC configuration */
679 	for (i = 0; i < bp->nr_vnics; i++) {
680 		rc = bnxt_setup_one_vnic(bp, i);
681 		if (rc)
682 			goto err_out;
683 	}
684 
685 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, &bp->vnic_info[0], 0, NULL);
686 	if (rc) {
687 		PMD_DRV_LOG(ERR,
688 			"HWRM cfa l2 rx mask failure rc: %x\n", rc);
689 		goto err_out;
690 	}
691 
692 	/* check and configure queue intr-vector mapping */
693 	if ((rte_intr_cap_multiple(intr_handle) ||
694 	     !RTE_ETH_DEV_SRIOV(bp->eth_dev).active) &&
695 	    bp->eth_dev->data->dev_conf.intr_conf.rxq != 0) {
696 		intr_vector = bp->eth_dev->data->nb_rx_queues;
697 		PMD_DRV_LOG(DEBUG, "intr_vector = %d\n", intr_vector);
698 		if (intr_vector > bp->rx_cp_nr_rings) {
699 			PMD_DRV_LOG(ERR, "At most %d intr queues supported",
700 					bp->rx_cp_nr_rings);
701 			return -ENOTSUP;
702 		}
703 		rc = rte_intr_efd_enable(intr_handle, intr_vector);
704 		if (rc)
705 			return rc;
706 	}
707 
708 	if (rte_intr_dp_is_en(intr_handle) && !intr_handle->intr_vec) {
709 		intr_handle->intr_vec =
710 			rte_zmalloc("intr_vec",
711 				    bp->eth_dev->data->nb_rx_queues *
712 				    sizeof(int), 0);
713 		if (intr_handle->intr_vec == NULL) {
714 			PMD_DRV_LOG(ERR, "Failed to allocate %d rx_queues"
715 				" intr_vec", bp->eth_dev->data->nb_rx_queues);
716 			rc = -ENOMEM;
717 			goto err_disable;
718 		}
719 		PMD_DRV_LOG(DEBUG, "intr_handle->intr_vec = %p "
720 			"intr_handle->nb_efd = %d intr_handle->max_intr = %d\n",
721 			 intr_handle->intr_vec, intr_handle->nb_efd,
722 			intr_handle->max_intr);
723 		for (queue_id = 0; queue_id < bp->eth_dev->data->nb_rx_queues;
724 		     queue_id++) {
725 			intr_handle->intr_vec[queue_id] =
726 							vec + BNXT_RX_VEC_START;
727 			if (vec < base + intr_handle->nb_efd - 1)
728 				vec++;
729 		}
730 	}
731 
732 	/* enable uio/vfio intr/eventfd mapping */
733 	rc = rte_intr_enable(intr_handle);
734 #ifndef RTE_EXEC_ENV_FREEBSD
735 	/* In FreeBSD OS, nic_uio driver does not support interrupts */
736 	if (rc)
737 		goto err_free;
738 #endif
739 
740 	rc = bnxt_get_hwrm_link_config(bp, &new);
741 	if (rc) {
742 		PMD_DRV_LOG(ERR, "HWRM Get link config failure rc: %x\n", rc);
743 		goto err_free;
744 	}
745 
746 	if (!bp->link_info->link_up) {
747 		rc = bnxt_set_hwrm_link_config(bp, true);
748 		if (rc) {
749 			PMD_DRV_LOG(ERR,
750 				"HWRM link config failure rc: %x\n", rc);
751 			goto err_free;
752 		}
753 	}
754 	bnxt_print_link_info(bp->eth_dev);
755 
756 	bp->mark_table = rte_zmalloc("bnxt_mark_table", BNXT_MARK_TABLE_SZ, 0);
757 	if (!bp->mark_table)
758 		PMD_DRV_LOG(ERR, "Allocation of mark table failed\n");
759 
760 	return 0;
761 
762 err_free:
763 	rte_free(intr_handle->intr_vec);
764 err_disable:
765 	rte_intr_efd_disable(intr_handle);
766 err_out:
767 	/* Some of the error status returned by FW may not be from errno.h */
768 	if (rc > 0)
769 		rc = -EIO;
770 
771 	return rc;
772 }
773 
774 static int bnxt_shutdown_nic(struct bnxt *bp)
775 {
776 	bnxt_free_all_hwrm_resources(bp);
777 	bnxt_free_all_filters(bp);
778 	bnxt_free_all_vnics(bp);
779 	return 0;
780 }
781 
782 /*
783  * Device configuration and status function
784  */
785 
786 uint32_t bnxt_get_speed_capabilities(struct bnxt *bp)
787 {
788 	uint32_t link_speed = bp->link_info->support_speeds;
789 	uint32_t speed_capa = 0;
790 
791 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_LINK_SPEED_100MB)
792 		speed_capa |= ETH_LINK_SPEED_100M;
793 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_100MBHD)
794 		speed_capa |= ETH_LINK_SPEED_100M_HD;
795 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_1GB)
796 		speed_capa |= ETH_LINK_SPEED_1G;
797 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_2_5GB)
798 		speed_capa |= ETH_LINK_SPEED_2_5G;
799 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_10GB)
800 		speed_capa |= ETH_LINK_SPEED_10G;
801 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_20GB)
802 		speed_capa |= ETH_LINK_SPEED_20G;
803 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_25GB)
804 		speed_capa |= ETH_LINK_SPEED_25G;
805 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_40GB)
806 		speed_capa |= ETH_LINK_SPEED_40G;
807 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_50GB)
808 		speed_capa |= ETH_LINK_SPEED_50G;
809 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_100GB)
810 		speed_capa |= ETH_LINK_SPEED_100G;
811 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_200GB)
812 		speed_capa |= ETH_LINK_SPEED_200G;
813 
814 	if (bp->link_info->auto_mode ==
815 	    HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_MODE_NONE)
816 		speed_capa |= ETH_LINK_SPEED_FIXED;
817 	else
818 		speed_capa |= ETH_LINK_SPEED_AUTONEG;
819 
820 	return speed_capa;
821 }
822 
823 static int bnxt_dev_info_get_op(struct rte_eth_dev *eth_dev,
824 				struct rte_eth_dev_info *dev_info)
825 {
826 	struct rte_pci_device *pdev = RTE_DEV_TO_PCI(eth_dev->device);
827 	struct bnxt *bp = eth_dev->data->dev_private;
828 	uint16_t max_vnics, i, j, vpool, vrxq;
829 	unsigned int max_rx_rings;
830 	int rc;
831 
832 	rc = is_bnxt_in_error(bp);
833 	if (rc)
834 		return rc;
835 
836 	/* MAC Specifics */
837 	dev_info->max_mac_addrs = bp->max_l2_ctx;
838 	dev_info->max_hash_mac_addrs = 0;
839 
840 	/* PF/VF specifics */
841 	if (BNXT_PF(bp))
842 		dev_info->max_vfs = pdev->max_vfs;
843 
844 	max_rx_rings = BNXT_MAX_RINGS(bp);
845 	/* For the sake of symmetry, max_rx_queues = max_tx_queues */
846 	dev_info->max_rx_queues = max_rx_rings;
847 	dev_info->max_tx_queues = max_rx_rings;
848 	dev_info->reta_size = bnxt_rss_hash_tbl_size(bp);
849 	dev_info->hash_key_size = 40;
850 	max_vnics = bp->max_vnics;
851 
852 	/* MTU specifics */
853 	dev_info->min_mtu = RTE_ETHER_MIN_MTU;
854 	dev_info->max_mtu = BNXT_MAX_MTU;
855 
856 	/* Fast path specifics */
857 	dev_info->min_rx_bufsize = 1;
858 	dev_info->max_rx_pktlen = BNXT_MAX_PKT_LEN;
859 
860 	dev_info->rx_offload_capa = BNXT_DEV_RX_OFFLOAD_SUPPORT;
861 	if (bp->flags & BNXT_FLAG_PTP_SUPPORTED)
862 		dev_info->rx_offload_capa |= DEV_RX_OFFLOAD_TIMESTAMP;
863 	dev_info->tx_offload_capa = BNXT_DEV_TX_OFFLOAD_SUPPORT;
864 	dev_info->flow_type_rss_offloads = BNXT_ETH_RSS_SUPPORT;
865 
866 	dev_info->speed_capa = bnxt_get_speed_capabilities(bp);
867 
868 	/* *INDENT-OFF* */
869 	dev_info->default_rxconf = (struct rte_eth_rxconf) {
870 		.rx_thresh = {
871 			.pthresh = 8,
872 			.hthresh = 8,
873 			.wthresh = 0,
874 		},
875 		.rx_free_thresh = 32,
876 		/* If no descriptors available, pkts are dropped by default */
877 		.rx_drop_en = 1,
878 	};
879 
880 	dev_info->default_txconf = (struct rte_eth_txconf) {
881 		.tx_thresh = {
882 			.pthresh = 32,
883 			.hthresh = 0,
884 			.wthresh = 0,
885 		},
886 		.tx_free_thresh = 32,
887 		.tx_rs_thresh = 32,
888 	};
889 	eth_dev->data->dev_conf.intr_conf.lsc = 1;
890 
891 	eth_dev->data->dev_conf.intr_conf.rxq = 1;
892 	dev_info->rx_desc_lim.nb_min = BNXT_MIN_RING_DESC;
893 	dev_info->rx_desc_lim.nb_max = BNXT_MAX_RX_RING_DESC;
894 	dev_info->tx_desc_lim.nb_min = BNXT_MIN_RING_DESC;
895 	dev_info->tx_desc_lim.nb_max = BNXT_MAX_TX_RING_DESC;
896 
897 	/* *INDENT-ON* */
898 
899 	/*
900 	 * TODO: default_rxconf, default_txconf, rx_desc_lim, and tx_desc_lim
901 	 *       need further investigation.
902 	 */
903 
904 	/* VMDq resources */
905 	vpool = 64; /* ETH_64_POOLS */
906 	vrxq = 128; /* ETH_VMDQ_DCB_NUM_QUEUES */
907 	for (i = 0; i < 4; vpool >>= 1, i++) {
908 		if (max_vnics > vpool) {
909 			for (j = 0; j < 5; vrxq >>= 1, j++) {
910 				if (dev_info->max_rx_queues > vrxq) {
911 					if (vpool > vrxq)
912 						vpool = vrxq;
913 					goto found;
914 				}
915 			}
916 			/* Not enough resources to support VMDq */
917 			break;
918 		}
919 	}
920 	/* Not enough resources to support VMDq */
921 	vpool = 0;
922 	vrxq = 0;
923 found:
924 	dev_info->max_vmdq_pools = vpool;
925 	dev_info->vmdq_queue_num = vrxq;
926 
927 	dev_info->vmdq_pool_base = 0;
928 	dev_info->vmdq_queue_base = 0;
929 
930 	return 0;
931 }
932 
933 /* Configure the device based on the configuration provided */
934 static int bnxt_dev_configure_op(struct rte_eth_dev *eth_dev)
935 {
936 	struct bnxt *bp = eth_dev->data->dev_private;
937 	uint64_t rx_offloads = eth_dev->data->dev_conf.rxmode.offloads;
938 	int rc;
939 
940 	bp->rx_queues = (void *)eth_dev->data->rx_queues;
941 	bp->tx_queues = (void *)eth_dev->data->tx_queues;
942 	bp->tx_nr_rings = eth_dev->data->nb_tx_queues;
943 	bp->rx_nr_rings = eth_dev->data->nb_rx_queues;
944 
945 	rc = is_bnxt_in_error(bp);
946 	if (rc)
947 		return rc;
948 
949 	if (BNXT_VF(bp) && (bp->flags & BNXT_FLAG_NEW_RM)) {
950 		rc = bnxt_hwrm_check_vf_rings(bp);
951 		if (rc) {
952 			PMD_DRV_LOG(ERR, "HWRM insufficient resources\n");
953 			return -ENOSPC;
954 		}
955 
956 		/* If a resource has already been allocated - in this case
957 		 * it is the async completion ring, free it. Reallocate it after
958 		 * resource reservation. This will ensure the resource counts
959 		 * are calculated correctly.
960 		 */
961 
962 		pthread_mutex_lock(&bp->def_cp_lock);
963 
964 		if (!BNXT_HAS_NQ(bp) && bp->async_cp_ring) {
965 			bnxt_disable_int(bp);
966 			bnxt_free_cp_ring(bp, bp->async_cp_ring);
967 		}
968 
969 		rc = bnxt_hwrm_func_reserve_vf_resc(bp, false);
970 		if (rc) {
971 			PMD_DRV_LOG(ERR, "HWRM resource alloc fail:%x\n", rc);
972 			pthread_mutex_unlock(&bp->def_cp_lock);
973 			return -ENOSPC;
974 		}
975 
976 		if (!BNXT_HAS_NQ(bp) && bp->async_cp_ring) {
977 			rc = bnxt_alloc_async_cp_ring(bp);
978 			if (rc) {
979 				pthread_mutex_unlock(&bp->def_cp_lock);
980 				return rc;
981 			}
982 			bnxt_enable_int(bp);
983 		}
984 
985 		pthread_mutex_unlock(&bp->def_cp_lock);
986 	} else {
987 		/* legacy driver needs to get updated values */
988 		rc = bnxt_hwrm_func_qcaps(bp);
989 		if (rc) {
990 			PMD_DRV_LOG(ERR, "hwrm func qcaps fail:%d\n", rc);
991 			return rc;
992 		}
993 	}
994 
995 	/* Inherit new configurations */
996 	if (eth_dev->data->nb_rx_queues > bp->max_rx_rings ||
997 	    eth_dev->data->nb_tx_queues > bp->max_tx_rings ||
998 	    eth_dev->data->nb_rx_queues + eth_dev->data->nb_tx_queues
999 		+ BNXT_NUM_ASYNC_CPR(bp) > bp->max_cp_rings ||
1000 	    eth_dev->data->nb_rx_queues + eth_dev->data->nb_tx_queues >
1001 	    bp->max_stat_ctx)
1002 		goto resource_error;
1003 
1004 	if (BNXT_HAS_RING_GRPS(bp) &&
1005 	    (uint32_t)(eth_dev->data->nb_rx_queues) > bp->max_ring_grps)
1006 		goto resource_error;
1007 
1008 	if (!(eth_dev->data->dev_conf.rxmode.mq_mode & ETH_MQ_RX_RSS) &&
1009 	    bp->max_vnics < eth_dev->data->nb_rx_queues)
1010 		goto resource_error;
1011 
1012 	bp->rx_cp_nr_rings = bp->rx_nr_rings;
1013 	bp->tx_cp_nr_rings = bp->tx_nr_rings;
1014 
1015 	if (eth_dev->data->dev_conf.rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG)
1016 		rx_offloads |= DEV_RX_OFFLOAD_RSS_HASH;
1017 	eth_dev->data->dev_conf.rxmode.offloads = rx_offloads;
1018 
1019 	if (rx_offloads & DEV_RX_OFFLOAD_JUMBO_FRAME) {
1020 		eth_dev->data->mtu =
1021 			eth_dev->data->dev_conf.rxmode.max_rx_pkt_len -
1022 			RTE_ETHER_HDR_LEN - RTE_ETHER_CRC_LEN - VLAN_TAG_SIZE *
1023 			BNXT_NUM_VLANS;
1024 		bnxt_mtu_set_op(eth_dev, eth_dev->data->mtu);
1025 	}
1026 	return 0;
1027 
1028 resource_error:
1029 	PMD_DRV_LOG(ERR,
1030 		    "Insufficient resources to support requested config\n");
1031 	PMD_DRV_LOG(ERR,
1032 		    "Num Queues Requested: Tx %d, Rx %d\n",
1033 		    eth_dev->data->nb_tx_queues,
1034 		    eth_dev->data->nb_rx_queues);
1035 	PMD_DRV_LOG(ERR,
1036 		    "MAX: TxQ %d, RxQ %d, CQ %d Stat %d, Grp %d, Vnic %d\n",
1037 		    bp->max_tx_rings, bp->max_rx_rings, bp->max_cp_rings,
1038 		    bp->max_stat_ctx, bp->max_ring_grps, bp->max_vnics);
1039 	return -ENOSPC;
1040 }
1041 
1042 void bnxt_print_link_info(struct rte_eth_dev *eth_dev)
1043 {
1044 	struct rte_eth_link *link = &eth_dev->data->dev_link;
1045 
1046 	if (link->link_status)
1047 		PMD_DRV_LOG(INFO, "Port %d Link Up - speed %u Mbps - %s\n",
1048 			eth_dev->data->port_id,
1049 			(uint32_t)link->link_speed,
1050 			(link->link_duplex == ETH_LINK_FULL_DUPLEX) ?
1051 			("full-duplex") : ("half-duplex\n"));
1052 	else
1053 		PMD_DRV_LOG(INFO, "Port %d Link Down\n",
1054 			eth_dev->data->port_id);
1055 }
1056 
1057 /*
1058  * Determine whether the current configuration requires support for scattered
1059  * receive; return 1 if scattered receive is required and 0 if not.
1060  */
1061 static int bnxt_scattered_rx(struct rte_eth_dev *eth_dev)
1062 {
1063 	uint16_t buf_size;
1064 	int i;
1065 
1066 	if (eth_dev->data->dev_conf.rxmode.offloads & DEV_RX_OFFLOAD_SCATTER)
1067 		return 1;
1068 
1069 	for (i = 0; i < eth_dev->data->nb_rx_queues; i++) {
1070 		struct bnxt_rx_queue *rxq = eth_dev->data->rx_queues[i];
1071 
1072 		buf_size = (uint16_t)(rte_pktmbuf_data_room_size(rxq->mb_pool) -
1073 				      RTE_PKTMBUF_HEADROOM);
1074 		if (eth_dev->data->dev_conf.rxmode.max_rx_pkt_len > buf_size)
1075 			return 1;
1076 	}
1077 	return 0;
1078 }
1079 
1080 static eth_rx_burst_t
1081 bnxt_receive_function(struct rte_eth_dev *eth_dev)
1082 {
1083 	struct bnxt *bp = eth_dev->data->dev_private;
1084 
1085 #if defined(RTE_ARCH_X86) || defined(RTE_ARCH_ARM64)
1086 #ifndef RTE_LIBRTE_IEEE1588
1087 	/*
1088 	 * Vector mode receive can be enabled only if scatter rx is not
1089 	 * in use and rx offloads are limited to VLAN stripping and
1090 	 * CRC stripping.
1091 	 */
1092 	if (!eth_dev->data->scattered_rx &&
1093 	    !(eth_dev->data->dev_conf.rxmode.offloads &
1094 	      ~(DEV_RX_OFFLOAD_VLAN_STRIP |
1095 		DEV_RX_OFFLOAD_KEEP_CRC |
1096 		DEV_RX_OFFLOAD_JUMBO_FRAME |
1097 		DEV_RX_OFFLOAD_IPV4_CKSUM |
1098 		DEV_RX_OFFLOAD_UDP_CKSUM |
1099 		DEV_RX_OFFLOAD_TCP_CKSUM |
1100 		DEV_RX_OFFLOAD_OUTER_IPV4_CKSUM |
1101 		DEV_RX_OFFLOAD_RSS_HASH |
1102 		DEV_RX_OFFLOAD_VLAN_FILTER)) &&
1103 	    !BNXT_TRUFLOW_EN(bp)) {
1104 		PMD_DRV_LOG(INFO, "Using vector mode receive for port %d\n",
1105 			    eth_dev->data->port_id);
1106 		bp->flags |= BNXT_FLAG_RX_VECTOR_PKT_MODE;
1107 		return bnxt_recv_pkts_vec;
1108 	}
1109 	PMD_DRV_LOG(INFO, "Vector mode receive disabled for port %d\n",
1110 		    eth_dev->data->port_id);
1111 	PMD_DRV_LOG(INFO,
1112 		    "Port %d scatter: %d rx offload: %" PRIX64 "\n",
1113 		    eth_dev->data->port_id,
1114 		    eth_dev->data->scattered_rx,
1115 		    eth_dev->data->dev_conf.rxmode.offloads);
1116 #endif
1117 #endif
1118 	bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
1119 	return bnxt_recv_pkts;
1120 }
1121 
1122 static eth_tx_burst_t
1123 bnxt_transmit_function(__rte_unused struct rte_eth_dev *eth_dev)
1124 {
1125 #if defined(RTE_ARCH_X86) || defined(RTE_ARCH_ARM64)
1126 #ifndef RTE_LIBRTE_IEEE1588
1127 	struct bnxt *bp = eth_dev->data->dev_private;
1128 
1129 	/*
1130 	 * Vector mode transmit can be enabled only if not using scatter rx
1131 	 * or tx offloads.
1132 	 */
1133 	if (!eth_dev->data->scattered_rx &&
1134 	    !eth_dev->data->dev_conf.txmode.offloads &&
1135 	    !BNXT_TRUFLOW_EN(bp)) {
1136 		PMD_DRV_LOG(INFO, "Using vector mode transmit for port %d\n",
1137 			    eth_dev->data->port_id);
1138 		return bnxt_xmit_pkts_vec;
1139 	}
1140 	PMD_DRV_LOG(INFO, "Vector mode transmit disabled for port %d\n",
1141 		    eth_dev->data->port_id);
1142 	PMD_DRV_LOG(INFO,
1143 		    "Port %d scatter: %d tx offload: %" PRIX64 "\n",
1144 		    eth_dev->data->port_id,
1145 		    eth_dev->data->scattered_rx,
1146 		    eth_dev->data->dev_conf.txmode.offloads);
1147 #endif
1148 #endif
1149 	return bnxt_xmit_pkts;
1150 }
1151 
1152 static int bnxt_handle_if_change_status(struct bnxt *bp)
1153 {
1154 	int rc;
1155 
1156 	/* Since fw has undergone a reset and lost all contexts,
1157 	 * set fatal flag to not issue hwrm during cleanup
1158 	 */
1159 	bp->flags |= BNXT_FLAG_FATAL_ERROR;
1160 	bnxt_uninit_resources(bp, true);
1161 
1162 	/* clear fatal flag so that re-init happens */
1163 	bp->flags &= ~BNXT_FLAG_FATAL_ERROR;
1164 	rc = bnxt_init_resources(bp, true);
1165 
1166 	bp->flags &= ~BNXT_FLAG_IF_CHANGE_HOT_FW_RESET_DONE;
1167 
1168 	return rc;
1169 }
1170 
1171 static int32_t
1172 bnxt_create_port_app_df_rule(struct bnxt *bp, uint8_t flow_type,
1173 			     uint32_t *flow_id)
1174 {
1175 	uint16_t port_id = bp->eth_dev->data->port_id;
1176 	struct ulp_tlv_param param_list[] = {
1177 		{
1178 			.type = BNXT_ULP_DF_PARAM_TYPE_DEV_PORT_ID,
1179 			.length = 2,
1180 			.value = {(port_id >> 8) & 0xff, port_id & 0xff}
1181 		},
1182 		{
1183 			.type = BNXT_ULP_DF_PARAM_TYPE_LAST,
1184 			.length = 0,
1185 			.value = {0}
1186 		}
1187 	};
1188 
1189 	return ulp_default_flow_create(bp->eth_dev, param_list, flow_type,
1190 				       flow_id);
1191 }
1192 
1193 static int32_t
1194 bnxt_create_df_rules(struct bnxt *bp)
1195 {
1196 	struct bnxt_ulp_data *cfg_data;
1197 	int rc;
1198 
1199 	cfg_data = bp->ulp_ctx->cfg_data;
1200 	rc = bnxt_create_port_app_df_rule(bp, BNXT_ULP_DF_TPL_PORT_TO_VS,
1201 					  &cfg_data->port_to_app_flow_id);
1202 	if (rc) {
1203 		PMD_DRV_LOG(ERR,
1204 			    "Failed to create port to app default rule\n");
1205 		return rc;
1206 	}
1207 
1208 	BNXT_TF_DBG(DEBUG, "***** created port to app default rule ******\n");
1209 	rc = bnxt_create_port_app_df_rule(bp, BNXT_ULP_DF_TPL_VS_TO_PORT,
1210 					  &cfg_data->app_to_port_flow_id);
1211 	if (!rc) {
1212 		rc = ulp_default_flow_db_cfa_action_get(bp->ulp_ctx,
1213 							cfg_data->app_to_port_flow_id,
1214 							&cfg_data->tx_cfa_action);
1215 		if (rc)
1216 			goto err;
1217 
1218 		BNXT_TF_DBG(DEBUG,
1219 			    "***** created app to port default rule *****\n");
1220 		return 0;
1221 	}
1222 
1223 err:
1224 	BNXT_TF_DBG(DEBUG, "Failed to create app to port default rule\n");
1225 	return rc;
1226 }
1227 
1228 static void
1229 bnxt_destroy_df_rules(struct bnxt *bp)
1230 {
1231 	struct bnxt_ulp_data *cfg_data;
1232 
1233 	cfg_data = bp->ulp_ctx->cfg_data;
1234 	ulp_default_flow_destroy(bp->eth_dev, cfg_data->port_to_app_flow_id);
1235 	ulp_default_flow_destroy(bp->eth_dev, cfg_data->app_to_port_flow_id);
1236 }
1237 
1238 static int bnxt_dev_start_op(struct rte_eth_dev *eth_dev)
1239 {
1240 	struct bnxt *bp = eth_dev->data->dev_private;
1241 	uint64_t rx_offloads = eth_dev->data->dev_conf.rxmode.offloads;
1242 	int vlan_mask = 0;
1243 	int rc, retry_cnt = BNXT_IF_CHANGE_RETRY_COUNT;
1244 
1245 	if (!eth_dev->data->nb_tx_queues || !eth_dev->data->nb_rx_queues) {
1246 		PMD_DRV_LOG(ERR, "Queues are not configured yet!\n");
1247 		return -EINVAL;
1248 	}
1249 
1250 	if (bp->rx_cp_nr_rings > RTE_ETHDEV_QUEUE_STAT_CNTRS) {
1251 		PMD_DRV_LOG(ERR,
1252 			"RxQ cnt %d > CONFIG_RTE_ETHDEV_QUEUE_STAT_CNTRS %d\n",
1253 			bp->rx_cp_nr_rings, RTE_ETHDEV_QUEUE_STAT_CNTRS);
1254 	}
1255 
1256 	do {
1257 		rc = bnxt_hwrm_if_change(bp, true);
1258 		if (rc == 0 || rc != -EAGAIN)
1259 			break;
1260 
1261 		rte_delay_ms(BNXT_IF_CHANGE_RETRY_INTERVAL);
1262 	} while (retry_cnt--);
1263 
1264 	if (rc)
1265 		return rc;
1266 
1267 	if (bp->flags & BNXT_FLAG_IF_CHANGE_HOT_FW_RESET_DONE) {
1268 		rc = bnxt_handle_if_change_status(bp);
1269 		if (rc)
1270 			return rc;
1271 	}
1272 
1273 	bnxt_enable_int(bp);
1274 
1275 	rc = bnxt_init_chip(bp);
1276 	if (rc)
1277 		goto error;
1278 
1279 	eth_dev->data->scattered_rx = bnxt_scattered_rx(eth_dev);
1280 	eth_dev->data->dev_started = 1;
1281 
1282 	bnxt_link_update(eth_dev, 1, ETH_LINK_UP);
1283 
1284 	if (rx_offloads & DEV_RX_OFFLOAD_VLAN_FILTER)
1285 		vlan_mask |= ETH_VLAN_FILTER_MASK;
1286 	if (rx_offloads & DEV_RX_OFFLOAD_VLAN_STRIP)
1287 		vlan_mask |= ETH_VLAN_STRIP_MASK;
1288 	rc = bnxt_vlan_offload_set_op(eth_dev, vlan_mask);
1289 	if (rc)
1290 		goto error;
1291 
1292 	eth_dev->rx_pkt_burst = bnxt_receive_function(eth_dev);
1293 	eth_dev->tx_pkt_burst = bnxt_transmit_function(eth_dev);
1294 
1295 	pthread_mutex_lock(&bp->def_cp_lock);
1296 	bnxt_schedule_fw_health_check(bp);
1297 	pthread_mutex_unlock(&bp->def_cp_lock);
1298 
1299 	if (BNXT_TRUFLOW_EN(bp))
1300 		bnxt_ulp_init(bp);
1301 
1302 	return 0;
1303 
1304 error:
1305 	bnxt_shutdown_nic(bp);
1306 	bnxt_free_tx_mbufs(bp);
1307 	bnxt_free_rx_mbufs(bp);
1308 	bnxt_hwrm_if_change(bp, false);
1309 	eth_dev->data->dev_started = 0;
1310 	return rc;
1311 }
1312 
1313 static int bnxt_dev_set_link_up_op(struct rte_eth_dev *eth_dev)
1314 {
1315 	struct bnxt *bp = eth_dev->data->dev_private;
1316 	int rc = 0;
1317 
1318 	if (!bp->link_info->link_up)
1319 		rc = bnxt_set_hwrm_link_config(bp, true);
1320 	if (!rc)
1321 		eth_dev->data->dev_link.link_status = 1;
1322 
1323 	bnxt_print_link_info(eth_dev);
1324 	return rc;
1325 }
1326 
1327 static int bnxt_dev_set_link_down_op(struct rte_eth_dev *eth_dev)
1328 {
1329 	struct bnxt *bp = eth_dev->data->dev_private;
1330 
1331 	eth_dev->data->dev_link.link_status = 0;
1332 	bnxt_set_hwrm_link_config(bp, false);
1333 	bp->link_info->link_up = 0;
1334 
1335 	return 0;
1336 }
1337 
1338 static void bnxt_free_switch_domain(struct bnxt *bp)
1339 {
1340 	if (bp->switch_domain_id)
1341 		rte_eth_switch_domain_free(bp->switch_domain_id);
1342 }
1343 
1344 /* Unload the driver, release resources */
1345 static void bnxt_dev_stop_op(struct rte_eth_dev *eth_dev)
1346 {
1347 	struct bnxt *bp = eth_dev->data->dev_private;
1348 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
1349 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
1350 
1351 	eth_dev->data->dev_started = 0;
1352 	/* Prevent crashes when queues are still in use */
1353 	eth_dev->rx_pkt_burst = &bnxt_dummy_recv_pkts;
1354 	eth_dev->tx_pkt_burst = &bnxt_dummy_xmit_pkts;
1355 
1356 	bnxt_disable_int(bp);
1357 
1358 	/* disable uio/vfio intr/eventfd mapping */
1359 	rte_intr_disable(intr_handle);
1360 
1361 	bnxt_cancel_fw_health_check(bp);
1362 
1363 	bnxt_dev_set_link_down_op(eth_dev);
1364 
1365 	/* Wait for link to be reset and the async notification to process.
1366 	 * During reset recovery, there is no need to wait and
1367 	 * VF/NPAR functions do not have privilege to change PHY config.
1368 	 */
1369 	if (!is_bnxt_in_error(bp) && BNXT_SINGLE_PF(bp))
1370 		bnxt_link_update(eth_dev, 1, ETH_LINK_DOWN);
1371 
1372 	/* Clean queue intr-vector mapping */
1373 	rte_intr_efd_disable(intr_handle);
1374 	if (intr_handle->intr_vec != NULL) {
1375 		rte_free(intr_handle->intr_vec);
1376 		intr_handle->intr_vec = NULL;
1377 	}
1378 
1379 	bnxt_hwrm_port_clr_stats(bp);
1380 	bnxt_free_tx_mbufs(bp);
1381 	bnxt_free_rx_mbufs(bp);
1382 	/* Process any remaining notifications in default completion queue */
1383 	bnxt_int_handler(eth_dev);
1384 	bnxt_shutdown_nic(bp);
1385 	bnxt_hwrm_if_change(bp, false);
1386 
1387 	rte_free(bp->mark_table);
1388 	bp->mark_table = NULL;
1389 
1390 	bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
1391 	bp->rx_cosq_cnt = 0;
1392 	/* All filters are deleted on a port stop. */
1393 	if (BNXT_FLOW_XSTATS_EN(bp))
1394 		bp->flow_stat->flow_count = 0;
1395 }
1396 
1397 static void bnxt_dev_close_op(struct rte_eth_dev *eth_dev)
1398 {
1399 	struct bnxt *bp = eth_dev->data->dev_private;
1400 
1401 	/* cancel the recovery handler before remove dev */
1402 	rte_eal_alarm_cancel(bnxt_dev_reset_and_resume, (void *)bp);
1403 	rte_eal_alarm_cancel(bnxt_dev_recover, (void *)bp);
1404 	bnxt_cancel_fc_thread(bp);
1405 
1406 	if (BNXT_TRUFLOW_EN(bp)) {
1407 		bnxt_destroy_df_rules(bp);
1408 		bnxt_ulp_deinit(bp);
1409 	}
1410 
1411 	if (eth_dev->data->dev_started)
1412 		bnxt_dev_stop_op(eth_dev);
1413 
1414 	bnxt_free_switch_domain(bp);
1415 
1416 	bnxt_uninit_resources(bp, false);
1417 
1418 	bnxt_free_leds_info(bp);
1419 	bnxt_free_cos_queues(bp);
1420 	bnxt_free_link_info(bp);
1421 	bnxt_free_pf_info(bp);
1422 	bnxt_free_parent_info(bp);
1423 
1424 	eth_dev->dev_ops = NULL;
1425 	eth_dev->rx_pkt_burst = NULL;
1426 	eth_dev->tx_pkt_burst = NULL;
1427 
1428 	rte_memzone_free((const struct rte_memzone *)bp->tx_mem_zone);
1429 	bp->tx_mem_zone = NULL;
1430 	rte_memzone_free((const struct rte_memzone *)bp->rx_mem_zone);
1431 	bp->rx_mem_zone = NULL;
1432 
1433 	rte_free(bp->pf->vf_info);
1434 	bp->pf->vf_info = NULL;
1435 
1436 	rte_free(bp->grp_info);
1437 	bp->grp_info = NULL;
1438 }
1439 
1440 static void bnxt_mac_addr_remove_op(struct rte_eth_dev *eth_dev,
1441 				    uint32_t index)
1442 {
1443 	struct bnxt *bp = eth_dev->data->dev_private;
1444 	uint64_t pool_mask = eth_dev->data->mac_pool_sel[index];
1445 	struct bnxt_vnic_info *vnic;
1446 	struct bnxt_filter_info *filter, *temp_filter;
1447 	uint32_t i;
1448 
1449 	if (is_bnxt_in_error(bp))
1450 		return;
1451 
1452 	/*
1453 	 * Loop through all VNICs from the specified filter flow pools to
1454 	 * remove the corresponding MAC addr filter
1455 	 */
1456 	for (i = 0; i < bp->nr_vnics; i++) {
1457 		if (!(pool_mask & (1ULL << i)))
1458 			continue;
1459 
1460 		vnic = &bp->vnic_info[i];
1461 		filter = STAILQ_FIRST(&vnic->filter);
1462 		while (filter) {
1463 			temp_filter = STAILQ_NEXT(filter, next);
1464 			if (filter->mac_index == index) {
1465 				STAILQ_REMOVE(&vnic->filter, filter,
1466 						bnxt_filter_info, next);
1467 				bnxt_hwrm_clear_l2_filter(bp, filter);
1468 				bnxt_free_filter(bp, filter);
1469 			}
1470 			filter = temp_filter;
1471 		}
1472 	}
1473 }
1474 
1475 static int bnxt_add_mac_filter(struct bnxt *bp, struct bnxt_vnic_info *vnic,
1476 			       struct rte_ether_addr *mac_addr, uint32_t index,
1477 			       uint32_t pool)
1478 {
1479 	struct bnxt_filter_info *filter;
1480 	int rc = 0;
1481 
1482 	/* Attach requested MAC address to the new l2_filter */
1483 	STAILQ_FOREACH(filter, &vnic->filter, next) {
1484 		if (filter->mac_index == index) {
1485 			PMD_DRV_LOG(DEBUG,
1486 				    "MAC addr already existed for pool %d\n",
1487 				    pool);
1488 			return 0;
1489 		}
1490 	}
1491 
1492 	filter = bnxt_alloc_filter(bp);
1493 	if (!filter) {
1494 		PMD_DRV_LOG(ERR, "L2 filter alloc failed\n");
1495 		return -ENODEV;
1496 	}
1497 
1498 	/* bnxt_alloc_filter copies default MAC to filter->l2_addr. So,
1499 	 * if the MAC that's been programmed now is a different one, then,
1500 	 * copy that addr to filter->l2_addr
1501 	 */
1502 	if (mac_addr)
1503 		memcpy(filter->l2_addr, mac_addr, RTE_ETHER_ADDR_LEN);
1504 	filter->flags |= HWRM_CFA_L2_FILTER_ALLOC_INPUT_FLAGS_OUTERMOST;
1505 
1506 	rc = bnxt_hwrm_set_l2_filter(bp, vnic->fw_vnic_id, filter);
1507 	if (!rc) {
1508 		filter->mac_index = index;
1509 		if (filter->mac_index == 0)
1510 			STAILQ_INSERT_HEAD(&vnic->filter, filter, next);
1511 		else
1512 			STAILQ_INSERT_TAIL(&vnic->filter, filter, next);
1513 	} else {
1514 		bnxt_free_filter(bp, filter);
1515 	}
1516 
1517 	return rc;
1518 }
1519 
1520 static int bnxt_mac_addr_add_op(struct rte_eth_dev *eth_dev,
1521 				struct rte_ether_addr *mac_addr,
1522 				uint32_t index, uint32_t pool)
1523 {
1524 	struct bnxt *bp = eth_dev->data->dev_private;
1525 	struct bnxt_vnic_info *vnic = &bp->vnic_info[pool];
1526 	int rc = 0;
1527 
1528 	rc = is_bnxt_in_error(bp);
1529 	if (rc)
1530 		return rc;
1531 
1532 	if (BNXT_VF(bp) & !BNXT_VF_IS_TRUSTED(bp)) {
1533 		PMD_DRV_LOG(ERR, "Cannot add MAC address to a VF interface\n");
1534 		return -ENOTSUP;
1535 	}
1536 
1537 	if (!vnic) {
1538 		PMD_DRV_LOG(ERR, "VNIC not found for pool %d!\n", pool);
1539 		return -EINVAL;
1540 	}
1541 
1542 	/* Filter settings will get applied when port is started */
1543 	if (!eth_dev->data->dev_started)
1544 		return 0;
1545 
1546 	rc = bnxt_add_mac_filter(bp, vnic, mac_addr, index, pool);
1547 
1548 	return rc;
1549 }
1550 
1551 int bnxt_link_update(struct rte_eth_dev *eth_dev, int wait_to_complete,
1552 		     bool exp_link_status)
1553 {
1554 	int rc = 0;
1555 	struct bnxt *bp = eth_dev->data->dev_private;
1556 	struct rte_eth_link new;
1557 	int cnt = exp_link_status ? BNXT_LINK_UP_WAIT_CNT :
1558 		  BNXT_LINK_DOWN_WAIT_CNT;
1559 
1560 	rc = is_bnxt_in_error(bp);
1561 	if (rc)
1562 		return rc;
1563 
1564 	memset(&new, 0, sizeof(new));
1565 	do {
1566 		/* Retrieve link info from hardware */
1567 		rc = bnxt_get_hwrm_link_config(bp, &new);
1568 		if (rc) {
1569 			new.link_speed = ETH_LINK_SPEED_100M;
1570 			new.link_duplex = ETH_LINK_FULL_DUPLEX;
1571 			PMD_DRV_LOG(ERR,
1572 				"Failed to retrieve link rc = 0x%x!\n", rc);
1573 			goto out;
1574 		}
1575 
1576 		if (!wait_to_complete || new.link_status == exp_link_status)
1577 			break;
1578 
1579 		rte_delay_ms(BNXT_LINK_WAIT_INTERVAL);
1580 	} while (cnt--);
1581 
1582 out:
1583 	/* Timed out or success */
1584 	if (new.link_status != eth_dev->data->dev_link.link_status ||
1585 	new.link_speed != eth_dev->data->dev_link.link_speed) {
1586 		rte_eth_linkstatus_set(eth_dev, &new);
1587 
1588 		_rte_eth_dev_callback_process(eth_dev,
1589 					      RTE_ETH_EVENT_INTR_LSC,
1590 					      NULL);
1591 
1592 		bnxt_print_link_info(eth_dev);
1593 	}
1594 
1595 	return rc;
1596 }
1597 
1598 int bnxt_link_update_op(struct rte_eth_dev *eth_dev,
1599 			int wait_to_complete)
1600 {
1601 	return bnxt_link_update(eth_dev, wait_to_complete, ETH_LINK_UP);
1602 }
1603 
1604 static int bnxt_promiscuous_enable_op(struct rte_eth_dev *eth_dev)
1605 {
1606 	struct bnxt *bp = eth_dev->data->dev_private;
1607 	struct bnxt_vnic_info *vnic;
1608 	uint32_t old_flags;
1609 	int rc;
1610 
1611 	rc = is_bnxt_in_error(bp);
1612 	if (rc)
1613 		return rc;
1614 
1615 	/* Filter settings will get applied when port is started */
1616 	if (!eth_dev->data->dev_started)
1617 		return 0;
1618 
1619 	if (bp->vnic_info == NULL)
1620 		return 0;
1621 
1622 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
1623 
1624 	old_flags = vnic->flags;
1625 	vnic->flags |= BNXT_VNIC_INFO_PROMISC;
1626 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1627 	if (rc != 0)
1628 		vnic->flags = old_flags;
1629 
1630 	return rc;
1631 }
1632 
1633 static int bnxt_promiscuous_disable_op(struct rte_eth_dev *eth_dev)
1634 {
1635 	struct bnxt *bp = eth_dev->data->dev_private;
1636 	struct bnxt_vnic_info *vnic;
1637 	uint32_t old_flags;
1638 	int rc;
1639 
1640 	rc = is_bnxt_in_error(bp);
1641 	if (rc)
1642 		return rc;
1643 
1644 	/* Filter settings will get applied when port is started */
1645 	if (!eth_dev->data->dev_started)
1646 		return 0;
1647 
1648 	if (bp->vnic_info == NULL)
1649 		return 0;
1650 
1651 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
1652 
1653 	old_flags = vnic->flags;
1654 	vnic->flags &= ~BNXT_VNIC_INFO_PROMISC;
1655 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1656 	if (rc != 0)
1657 		vnic->flags = old_flags;
1658 
1659 	if (BNXT_TRUFLOW_EN(bp))
1660 		bnxt_create_df_rules(bp);
1661 
1662 	return rc;
1663 }
1664 
1665 static int bnxt_allmulticast_enable_op(struct rte_eth_dev *eth_dev)
1666 {
1667 	struct bnxt *bp = eth_dev->data->dev_private;
1668 	struct bnxt_vnic_info *vnic;
1669 	uint32_t old_flags;
1670 	int rc;
1671 
1672 	rc = is_bnxt_in_error(bp);
1673 	if (rc)
1674 		return rc;
1675 
1676 	/* Filter settings will get applied when port is started */
1677 	if (!eth_dev->data->dev_started)
1678 		return 0;
1679 
1680 	if (bp->vnic_info == NULL)
1681 		return 0;
1682 
1683 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
1684 
1685 	old_flags = vnic->flags;
1686 	vnic->flags |= BNXT_VNIC_INFO_ALLMULTI;
1687 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1688 	if (rc != 0)
1689 		vnic->flags = old_flags;
1690 
1691 	return rc;
1692 }
1693 
1694 static int bnxt_allmulticast_disable_op(struct rte_eth_dev *eth_dev)
1695 {
1696 	struct bnxt *bp = eth_dev->data->dev_private;
1697 	struct bnxt_vnic_info *vnic;
1698 	uint32_t old_flags;
1699 	int rc;
1700 
1701 	rc = is_bnxt_in_error(bp);
1702 	if (rc)
1703 		return rc;
1704 
1705 	/* Filter settings will get applied when port is started */
1706 	if (!eth_dev->data->dev_started)
1707 		return 0;
1708 
1709 	if (bp->vnic_info == NULL)
1710 		return 0;
1711 
1712 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
1713 
1714 	old_flags = vnic->flags;
1715 	vnic->flags &= ~BNXT_VNIC_INFO_ALLMULTI;
1716 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1717 	if (rc != 0)
1718 		vnic->flags = old_flags;
1719 
1720 	return rc;
1721 }
1722 
1723 /* Return bnxt_rx_queue pointer corresponding to a given rxq. */
1724 static struct bnxt_rx_queue *bnxt_qid_to_rxq(struct bnxt *bp, uint16_t qid)
1725 {
1726 	if (qid >= bp->rx_nr_rings)
1727 		return NULL;
1728 
1729 	return bp->eth_dev->data->rx_queues[qid];
1730 }
1731 
1732 /* Return rxq corresponding to a given rss table ring/group ID. */
1733 static uint16_t bnxt_rss_to_qid(struct bnxt *bp, uint16_t fwr)
1734 {
1735 	struct bnxt_rx_queue *rxq;
1736 	unsigned int i;
1737 
1738 	if (!BNXT_HAS_RING_GRPS(bp)) {
1739 		for (i = 0; i < bp->rx_nr_rings; i++) {
1740 			rxq = bp->eth_dev->data->rx_queues[i];
1741 			if (rxq->rx_ring->rx_ring_struct->fw_ring_id == fwr)
1742 				return rxq->index;
1743 		}
1744 	} else {
1745 		for (i = 0; i < bp->rx_nr_rings; i++) {
1746 			if (bp->grp_info[i].fw_grp_id == fwr)
1747 				return i;
1748 		}
1749 	}
1750 
1751 	return INVALID_HW_RING_ID;
1752 }
1753 
1754 static int bnxt_reta_update_op(struct rte_eth_dev *eth_dev,
1755 			    struct rte_eth_rss_reta_entry64 *reta_conf,
1756 			    uint16_t reta_size)
1757 {
1758 	struct bnxt *bp = eth_dev->data->dev_private;
1759 	struct rte_eth_conf *dev_conf = &bp->eth_dev->data->dev_conf;
1760 	struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
1761 	uint16_t tbl_size = bnxt_rss_hash_tbl_size(bp);
1762 	uint16_t idx, sft;
1763 	int i, rc;
1764 
1765 	rc = is_bnxt_in_error(bp);
1766 	if (rc)
1767 		return rc;
1768 
1769 	if (!vnic->rss_table)
1770 		return -EINVAL;
1771 
1772 	if (!(dev_conf->rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG))
1773 		return -EINVAL;
1774 
1775 	if (reta_size != tbl_size) {
1776 		PMD_DRV_LOG(ERR, "The configured hash table lookup size "
1777 			"(%d) must equal the size supported by the hardware "
1778 			"(%d)\n", reta_size, tbl_size);
1779 		return -EINVAL;
1780 	}
1781 
1782 	for (i = 0; i < reta_size; i++) {
1783 		struct bnxt_rx_queue *rxq;
1784 
1785 		idx = i / RTE_RETA_GROUP_SIZE;
1786 		sft = i % RTE_RETA_GROUP_SIZE;
1787 
1788 		if (!(reta_conf[idx].mask & (1ULL << sft)))
1789 			continue;
1790 
1791 		rxq = bnxt_qid_to_rxq(bp, reta_conf[idx].reta[sft]);
1792 		if (!rxq) {
1793 			PMD_DRV_LOG(ERR, "Invalid ring in reta_conf.\n");
1794 			return -EINVAL;
1795 		}
1796 
1797 		if (BNXT_CHIP_THOR(bp)) {
1798 			vnic->rss_table[i * 2] =
1799 				rxq->rx_ring->rx_ring_struct->fw_ring_id;
1800 			vnic->rss_table[i * 2 + 1] =
1801 				rxq->cp_ring->cp_ring_struct->fw_ring_id;
1802 		} else {
1803 			vnic->rss_table[i] =
1804 			    vnic->fw_grp_ids[reta_conf[idx].reta[sft]];
1805 		}
1806 	}
1807 
1808 	bnxt_hwrm_vnic_rss_cfg(bp, vnic);
1809 	return 0;
1810 }
1811 
1812 static int bnxt_reta_query_op(struct rte_eth_dev *eth_dev,
1813 			      struct rte_eth_rss_reta_entry64 *reta_conf,
1814 			      uint16_t reta_size)
1815 {
1816 	struct bnxt *bp = eth_dev->data->dev_private;
1817 	struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
1818 	uint16_t tbl_size = bnxt_rss_hash_tbl_size(bp);
1819 	uint16_t idx, sft, i;
1820 	int rc;
1821 
1822 	rc = is_bnxt_in_error(bp);
1823 	if (rc)
1824 		return rc;
1825 
1826 	/* Retrieve from the default VNIC */
1827 	if (!vnic)
1828 		return -EINVAL;
1829 	if (!vnic->rss_table)
1830 		return -EINVAL;
1831 
1832 	if (reta_size != tbl_size) {
1833 		PMD_DRV_LOG(ERR, "The configured hash table lookup size "
1834 			"(%d) must equal the size supported by the hardware "
1835 			"(%d)\n", reta_size, tbl_size);
1836 		return -EINVAL;
1837 	}
1838 
1839 	for (idx = 0, i = 0; i < reta_size; i++) {
1840 		idx = i / RTE_RETA_GROUP_SIZE;
1841 		sft = i % RTE_RETA_GROUP_SIZE;
1842 
1843 		if (reta_conf[idx].mask & (1ULL << sft)) {
1844 			uint16_t qid;
1845 
1846 			if (BNXT_CHIP_THOR(bp))
1847 				qid = bnxt_rss_to_qid(bp,
1848 						      vnic->rss_table[i * 2]);
1849 			else
1850 				qid = bnxt_rss_to_qid(bp, vnic->rss_table[i]);
1851 
1852 			if (qid == INVALID_HW_RING_ID) {
1853 				PMD_DRV_LOG(ERR, "Inv. entry in rss table.\n");
1854 				return -EINVAL;
1855 			}
1856 			reta_conf[idx].reta[sft] = qid;
1857 		}
1858 	}
1859 
1860 	return 0;
1861 }
1862 
1863 static int bnxt_rss_hash_update_op(struct rte_eth_dev *eth_dev,
1864 				   struct rte_eth_rss_conf *rss_conf)
1865 {
1866 	struct bnxt *bp = eth_dev->data->dev_private;
1867 	struct rte_eth_conf *dev_conf = &bp->eth_dev->data->dev_conf;
1868 	struct bnxt_vnic_info *vnic;
1869 	int rc;
1870 
1871 	rc = is_bnxt_in_error(bp);
1872 	if (rc)
1873 		return rc;
1874 
1875 	/*
1876 	 * If RSS enablement were different than dev_configure,
1877 	 * then return -EINVAL
1878 	 */
1879 	if (dev_conf->rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG) {
1880 		if (!rss_conf->rss_hf)
1881 			PMD_DRV_LOG(ERR, "Hash type NONE\n");
1882 	} else {
1883 		if (rss_conf->rss_hf & BNXT_ETH_RSS_SUPPORT)
1884 			return -EINVAL;
1885 	}
1886 
1887 	bp->flags |= BNXT_FLAG_UPDATE_HASH;
1888 	memcpy(&eth_dev->data->dev_conf.rx_adv_conf.rss_conf,
1889 	       rss_conf,
1890 	       sizeof(*rss_conf));
1891 
1892 	/* Update the default RSS VNIC(s) */
1893 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
1894 	vnic->hash_type = bnxt_rte_to_hwrm_hash_types(rss_conf->rss_hf);
1895 
1896 	/*
1897 	 * If hashkey is not specified, use the previously configured
1898 	 * hashkey
1899 	 */
1900 	if (!rss_conf->rss_key)
1901 		goto rss_config;
1902 
1903 	if (rss_conf->rss_key_len != HW_HASH_KEY_SIZE) {
1904 		PMD_DRV_LOG(ERR,
1905 			    "Invalid hashkey length, should be 16 bytes\n");
1906 		return -EINVAL;
1907 	}
1908 	memcpy(vnic->rss_hash_key, rss_conf->rss_key, rss_conf->rss_key_len);
1909 
1910 rss_config:
1911 	bnxt_hwrm_vnic_rss_cfg(bp, vnic);
1912 	return 0;
1913 }
1914 
1915 static int bnxt_rss_hash_conf_get_op(struct rte_eth_dev *eth_dev,
1916 				     struct rte_eth_rss_conf *rss_conf)
1917 {
1918 	struct bnxt *bp = eth_dev->data->dev_private;
1919 	struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
1920 	int len, rc;
1921 	uint32_t hash_types;
1922 
1923 	rc = is_bnxt_in_error(bp);
1924 	if (rc)
1925 		return rc;
1926 
1927 	/* RSS configuration is the same for all VNICs */
1928 	if (vnic && vnic->rss_hash_key) {
1929 		if (rss_conf->rss_key) {
1930 			len = rss_conf->rss_key_len <= HW_HASH_KEY_SIZE ?
1931 			      rss_conf->rss_key_len : HW_HASH_KEY_SIZE;
1932 			memcpy(rss_conf->rss_key, vnic->rss_hash_key, len);
1933 		}
1934 
1935 		hash_types = vnic->hash_type;
1936 		rss_conf->rss_hf = 0;
1937 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV4) {
1938 			rss_conf->rss_hf |= ETH_RSS_IPV4;
1939 			hash_types &= ~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV4;
1940 		}
1941 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV4) {
1942 			rss_conf->rss_hf |= ETH_RSS_NONFRAG_IPV4_TCP;
1943 			hash_types &=
1944 				~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV4;
1945 		}
1946 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV4) {
1947 			rss_conf->rss_hf |= ETH_RSS_NONFRAG_IPV4_UDP;
1948 			hash_types &=
1949 				~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV4;
1950 		}
1951 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV6) {
1952 			rss_conf->rss_hf |= ETH_RSS_IPV6;
1953 			hash_types &= ~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV6;
1954 		}
1955 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV6) {
1956 			rss_conf->rss_hf |= ETH_RSS_NONFRAG_IPV6_TCP;
1957 			hash_types &=
1958 				~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV6;
1959 		}
1960 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV6) {
1961 			rss_conf->rss_hf |= ETH_RSS_NONFRAG_IPV6_UDP;
1962 			hash_types &=
1963 				~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV6;
1964 		}
1965 		if (hash_types) {
1966 			PMD_DRV_LOG(ERR,
1967 				"Unknown RSS config from firmware (%08x), RSS disabled",
1968 				vnic->hash_type);
1969 			return -ENOTSUP;
1970 		}
1971 	} else {
1972 		rss_conf->rss_hf = 0;
1973 	}
1974 	return 0;
1975 }
1976 
1977 static int bnxt_flow_ctrl_get_op(struct rte_eth_dev *dev,
1978 			       struct rte_eth_fc_conf *fc_conf)
1979 {
1980 	struct bnxt *bp = dev->data->dev_private;
1981 	struct rte_eth_link link_info;
1982 	int rc;
1983 
1984 	rc = is_bnxt_in_error(bp);
1985 	if (rc)
1986 		return rc;
1987 
1988 	rc = bnxt_get_hwrm_link_config(bp, &link_info);
1989 	if (rc)
1990 		return rc;
1991 
1992 	memset(fc_conf, 0, sizeof(*fc_conf));
1993 	if (bp->link_info->auto_pause)
1994 		fc_conf->autoneg = 1;
1995 	switch (bp->link_info->pause) {
1996 	case 0:
1997 		fc_conf->mode = RTE_FC_NONE;
1998 		break;
1999 	case HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_TX:
2000 		fc_conf->mode = RTE_FC_TX_PAUSE;
2001 		break;
2002 	case HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_RX:
2003 		fc_conf->mode = RTE_FC_RX_PAUSE;
2004 		break;
2005 	case (HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_TX |
2006 			HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_RX):
2007 		fc_conf->mode = RTE_FC_FULL;
2008 		break;
2009 	}
2010 	return 0;
2011 }
2012 
2013 static int bnxt_flow_ctrl_set_op(struct rte_eth_dev *dev,
2014 			       struct rte_eth_fc_conf *fc_conf)
2015 {
2016 	struct bnxt *bp = dev->data->dev_private;
2017 	int rc;
2018 
2019 	rc = is_bnxt_in_error(bp);
2020 	if (rc)
2021 		return rc;
2022 
2023 	if (!BNXT_SINGLE_PF(bp) || BNXT_VF(bp)) {
2024 		PMD_DRV_LOG(ERR, "Flow Control Settings cannot be modified\n");
2025 		return -ENOTSUP;
2026 	}
2027 
2028 	switch (fc_conf->mode) {
2029 	case RTE_FC_NONE:
2030 		bp->link_info->auto_pause = 0;
2031 		bp->link_info->force_pause = 0;
2032 		break;
2033 	case RTE_FC_RX_PAUSE:
2034 		if (fc_conf->autoneg) {
2035 			bp->link_info->auto_pause =
2036 					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_RX;
2037 			bp->link_info->force_pause = 0;
2038 		} else {
2039 			bp->link_info->auto_pause = 0;
2040 			bp->link_info->force_pause =
2041 					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_RX;
2042 		}
2043 		break;
2044 	case RTE_FC_TX_PAUSE:
2045 		if (fc_conf->autoneg) {
2046 			bp->link_info->auto_pause =
2047 					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_TX;
2048 			bp->link_info->force_pause = 0;
2049 		} else {
2050 			bp->link_info->auto_pause = 0;
2051 			bp->link_info->force_pause =
2052 					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_TX;
2053 		}
2054 		break;
2055 	case RTE_FC_FULL:
2056 		if (fc_conf->autoneg) {
2057 			bp->link_info->auto_pause =
2058 					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_TX |
2059 					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_RX;
2060 			bp->link_info->force_pause = 0;
2061 		} else {
2062 			bp->link_info->auto_pause = 0;
2063 			bp->link_info->force_pause =
2064 					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_TX |
2065 					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_RX;
2066 		}
2067 		break;
2068 	}
2069 	return bnxt_set_hwrm_link_config(bp, true);
2070 }
2071 
2072 /* Add UDP tunneling port */
2073 static int
2074 bnxt_udp_tunnel_port_add_op(struct rte_eth_dev *eth_dev,
2075 			 struct rte_eth_udp_tunnel *udp_tunnel)
2076 {
2077 	struct bnxt *bp = eth_dev->data->dev_private;
2078 	uint16_t tunnel_type = 0;
2079 	int rc = 0;
2080 
2081 	rc = is_bnxt_in_error(bp);
2082 	if (rc)
2083 		return rc;
2084 
2085 	switch (udp_tunnel->prot_type) {
2086 	case RTE_TUNNEL_TYPE_VXLAN:
2087 		if (bp->vxlan_port_cnt) {
2088 			PMD_DRV_LOG(ERR, "Tunnel Port %d already programmed\n",
2089 				udp_tunnel->udp_port);
2090 			if (bp->vxlan_port != udp_tunnel->udp_port) {
2091 				PMD_DRV_LOG(ERR, "Only one port allowed\n");
2092 				return -ENOSPC;
2093 			}
2094 			bp->vxlan_port_cnt++;
2095 			return 0;
2096 		}
2097 		tunnel_type =
2098 			HWRM_TUNNEL_DST_PORT_ALLOC_INPUT_TUNNEL_TYPE_VXLAN;
2099 		bp->vxlan_port_cnt++;
2100 		break;
2101 	case RTE_TUNNEL_TYPE_GENEVE:
2102 		if (bp->geneve_port_cnt) {
2103 			PMD_DRV_LOG(ERR, "Tunnel Port %d already programmed\n",
2104 				udp_tunnel->udp_port);
2105 			if (bp->geneve_port != udp_tunnel->udp_port) {
2106 				PMD_DRV_LOG(ERR, "Only one port allowed\n");
2107 				return -ENOSPC;
2108 			}
2109 			bp->geneve_port_cnt++;
2110 			return 0;
2111 		}
2112 		tunnel_type =
2113 			HWRM_TUNNEL_DST_PORT_ALLOC_INPUT_TUNNEL_TYPE_GENEVE;
2114 		bp->geneve_port_cnt++;
2115 		break;
2116 	default:
2117 		PMD_DRV_LOG(ERR, "Tunnel type is not supported\n");
2118 		return -ENOTSUP;
2119 	}
2120 	rc = bnxt_hwrm_tunnel_dst_port_alloc(bp, udp_tunnel->udp_port,
2121 					     tunnel_type);
2122 	return rc;
2123 }
2124 
2125 static int
2126 bnxt_udp_tunnel_port_del_op(struct rte_eth_dev *eth_dev,
2127 			 struct rte_eth_udp_tunnel *udp_tunnel)
2128 {
2129 	struct bnxt *bp = eth_dev->data->dev_private;
2130 	uint16_t tunnel_type = 0;
2131 	uint16_t port = 0;
2132 	int rc = 0;
2133 
2134 	rc = is_bnxt_in_error(bp);
2135 	if (rc)
2136 		return rc;
2137 
2138 	switch (udp_tunnel->prot_type) {
2139 	case RTE_TUNNEL_TYPE_VXLAN:
2140 		if (!bp->vxlan_port_cnt) {
2141 			PMD_DRV_LOG(ERR, "No Tunnel port configured yet\n");
2142 			return -EINVAL;
2143 		}
2144 		if (bp->vxlan_port != udp_tunnel->udp_port) {
2145 			PMD_DRV_LOG(ERR, "Req Port: %d. Configured port: %d\n",
2146 				udp_tunnel->udp_port, bp->vxlan_port);
2147 			return -EINVAL;
2148 		}
2149 		if (--bp->vxlan_port_cnt)
2150 			return 0;
2151 
2152 		tunnel_type =
2153 			HWRM_TUNNEL_DST_PORT_FREE_INPUT_TUNNEL_TYPE_VXLAN;
2154 		port = bp->vxlan_fw_dst_port_id;
2155 		break;
2156 	case RTE_TUNNEL_TYPE_GENEVE:
2157 		if (!bp->geneve_port_cnt) {
2158 			PMD_DRV_LOG(ERR, "No Tunnel port configured yet\n");
2159 			return -EINVAL;
2160 		}
2161 		if (bp->geneve_port != udp_tunnel->udp_port) {
2162 			PMD_DRV_LOG(ERR, "Req Port: %d. Configured port: %d\n",
2163 				udp_tunnel->udp_port, bp->geneve_port);
2164 			return -EINVAL;
2165 		}
2166 		if (--bp->geneve_port_cnt)
2167 			return 0;
2168 
2169 		tunnel_type =
2170 			HWRM_TUNNEL_DST_PORT_FREE_INPUT_TUNNEL_TYPE_GENEVE;
2171 		port = bp->geneve_fw_dst_port_id;
2172 		break;
2173 	default:
2174 		PMD_DRV_LOG(ERR, "Tunnel type is not supported\n");
2175 		return -ENOTSUP;
2176 	}
2177 
2178 	rc = bnxt_hwrm_tunnel_dst_port_free(bp, port, tunnel_type);
2179 	if (!rc) {
2180 		if (tunnel_type ==
2181 		    HWRM_TUNNEL_DST_PORT_FREE_INPUT_TUNNEL_TYPE_VXLAN)
2182 			bp->vxlan_port = 0;
2183 		if (tunnel_type ==
2184 		    HWRM_TUNNEL_DST_PORT_FREE_INPUT_TUNNEL_TYPE_GENEVE)
2185 			bp->geneve_port = 0;
2186 	}
2187 	return rc;
2188 }
2189 
2190 static int bnxt_del_vlan_filter(struct bnxt *bp, uint16_t vlan_id)
2191 {
2192 	struct bnxt_filter_info *filter;
2193 	struct bnxt_vnic_info *vnic;
2194 	int rc = 0;
2195 	uint32_t chk = HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN;
2196 
2197 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
2198 	filter = STAILQ_FIRST(&vnic->filter);
2199 	while (filter) {
2200 		/* Search for this matching MAC+VLAN filter */
2201 		if (bnxt_vlan_filter_exists(bp, filter, chk, vlan_id)) {
2202 			/* Delete the filter */
2203 			rc = bnxt_hwrm_clear_l2_filter(bp, filter);
2204 			if (rc)
2205 				return rc;
2206 			STAILQ_REMOVE(&vnic->filter, filter,
2207 				      bnxt_filter_info, next);
2208 			bnxt_free_filter(bp, filter);
2209 			PMD_DRV_LOG(INFO,
2210 				    "Deleted vlan filter for %d\n",
2211 				    vlan_id);
2212 			return 0;
2213 		}
2214 		filter = STAILQ_NEXT(filter, next);
2215 	}
2216 	return -ENOENT;
2217 }
2218 
2219 static int bnxt_add_vlan_filter(struct bnxt *bp, uint16_t vlan_id)
2220 {
2221 	struct bnxt_filter_info *filter;
2222 	struct bnxt_vnic_info *vnic;
2223 	int rc = 0;
2224 	uint32_t en = HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN |
2225 		HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN_MASK;
2226 	uint32_t chk = HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN;
2227 
2228 	/* Implementation notes on the use of VNIC in this command:
2229 	 *
2230 	 * By default, these filters belong to default vnic for the function.
2231 	 * Once these filters are set up, only destination VNIC can be modified.
2232 	 * If the destination VNIC is not specified in this command,
2233 	 * then the HWRM shall only create an l2 context id.
2234 	 */
2235 
2236 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
2237 	filter = STAILQ_FIRST(&vnic->filter);
2238 	/* Check if the VLAN has already been added */
2239 	while (filter) {
2240 		if (bnxt_vlan_filter_exists(bp, filter, chk, vlan_id))
2241 			return -EEXIST;
2242 
2243 		filter = STAILQ_NEXT(filter, next);
2244 	}
2245 
2246 	/* No match found. Alloc a fresh filter and issue the L2_FILTER_ALLOC
2247 	 * command to create MAC+VLAN filter with the right flags, enables set.
2248 	 */
2249 	filter = bnxt_alloc_filter(bp);
2250 	if (!filter) {
2251 		PMD_DRV_LOG(ERR,
2252 			    "MAC/VLAN filter alloc failed\n");
2253 		return -ENOMEM;
2254 	}
2255 	/* MAC + VLAN ID filter */
2256 	/* If l2_ivlan == 0 and l2_ivlan_mask != 0, only
2257 	 * untagged packets are received
2258 	 *
2259 	 * If l2_ivlan != 0 and l2_ivlan_mask != 0, untagged
2260 	 * packets and only the programmed vlan's packets are received
2261 	 */
2262 	filter->l2_ivlan = vlan_id;
2263 	filter->l2_ivlan_mask = 0x0FFF;
2264 	filter->enables |= en;
2265 	filter->flags |= HWRM_CFA_L2_FILTER_ALLOC_INPUT_FLAGS_OUTERMOST;
2266 
2267 	rc = bnxt_hwrm_set_l2_filter(bp, vnic->fw_vnic_id, filter);
2268 	if (rc) {
2269 		/* Free the newly allocated filter as we were
2270 		 * not able to create the filter in hardware.
2271 		 */
2272 		bnxt_free_filter(bp, filter);
2273 		return rc;
2274 	}
2275 
2276 	filter->mac_index = 0;
2277 	/* Add this new filter to the list */
2278 	if (vlan_id == 0)
2279 		STAILQ_INSERT_HEAD(&vnic->filter, filter, next);
2280 	else
2281 		STAILQ_INSERT_TAIL(&vnic->filter, filter, next);
2282 
2283 	PMD_DRV_LOG(INFO,
2284 		    "Added Vlan filter for %d\n", vlan_id);
2285 	return rc;
2286 }
2287 
2288 static int bnxt_vlan_filter_set_op(struct rte_eth_dev *eth_dev,
2289 		uint16_t vlan_id, int on)
2290 {
2291 	struct bnxt *bp = eth_dev->data->dev_private;
2292 	int rc;
2293 
2294 	rc = is_bnxt_in_error(bp);
2295 	if (rc)
2296 		return rc;
2297 
2298 	if (!eth_dev->data->dev_started) {
2299 		PMD_DRV_LOG(ERR, "port must be started before setting vlan\n");
2300 		return -EINVAL;
2301 	}
2302 
2303 	/* These operations apply to ALL existing MAC/VLAN filters */
2304 	if (on)
2305 		return bnxt_add_vlan_filter(bp, vlan_id);
2306 	else
2307 		return bnxt_del_vlan_filter(bp, vlan_id);
2308 }
2309 
2310 static int bnxt_del_dflt_mac_filter(struct bnxt *bp,
2311 				    struct bnxt_vnic_info *vnic)
2312 {
2313 	struct bnxt_filter_info *filter;
2314 	int rc;
2315 
2316 	filter = STAILQ_FIRST(&vnic->filter);
2317 	while (filter) {
2318 		if (filter->mac_index == 0 &&
2319 		    !memcmp(filter->l2_addr, bp->mac_addr,
2320 			    RTE_ETHER_ADDR_LEN)) {
2321 			rc = bnxt_hwrm_clear_l2_filter(bp, filter);
2322 			if (!rc) {
2323 				STAILQ_REMOVE(&vnic->filter, filter,
2324 					      bnxt_filter_info, next);
2325 				bnxt_free_filter(bp, filter);
2326 			}
2327 			return rc;
2328 		}
2329 		filter = STAILQ_NEXT(filter, next);
2330 	}
2331 	return 0;
2332 }
2333 
2334 static int
2335 bnxt_config_vlan_hw_filter(struct bnxt *bp, uint64_t rx_offloads)
2336 {
2337 	struct bnxt_vnic_info *vnic;
2338 	unsigned int i;
2339 	int rc;
2340 
2341 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
2342 	if (!(rx_offloads & DEV_RX_OFFLOAD_VLAN_FILTER)) {
2343 		/* Remove any VLAN filters programmed */
2344 		for (i = 0; i < RTE_ETHER_MAX_VLAN_ID; i++)
2345 			bnxt_del_vlan_filter(bp, i);
2346 
2347 		rc = bnxt_add_mac_filter(bp, vnic, NULL, 0, 0);
2348 		if (rc)
2349 			return rc;
2350 	} else {
2351 		/* Default filter will allow packets that match the
2352 		 * dest mac. So, it has to be deleted, otherwise, we
2353 		 * will endup receiving vlan packets for which the
2354 		 * filter is not programmed, when hw-vlan-filter
2355 		 * configuration is ON
2356 		 */
2357 		bnxt_del_dflt_mac_filter(bp, vnic);
2358 		/* This filter will allow only untagged packets */
2359 		bnxt_add_vlan_filter(bp, 0);
2360 	}
2361 	PMD_DRV_LOG(DEBUG, "VLAN Filtering: %d\n",
2362 		    !!(rx_offloads & DEV_RX_OFFLOAD_VLAN_FILTER));
2363 
2364 	return 0;
2365 }
2366 
2367 static int bnxt_free_one_vnic(struct bnxt *bp, uint16_t vnic_id)
2368 {
2369 	struct bnxt_vnic_info *vnic = &bp->vnic_info[vnic_id];
2370 	unsigned int i;
2371 	int rc;
2372 
2373 	/* Destroy vnic filters and vnic */
2374 	if (bp->eth_dev->data->dev_conf.rxmode.offloads &
2375 	    DEV_RX_OFFLOAD_VLAN_FILTER) {
2376 		for (i = 0; i < RTE_ETHER_MAX_VLAN_ID; i++)
2377 			bnxt_del_vlan_filter(bp, i);
2378 	}
2379 	bnxt_del_dflt_mac_filter(bp, vnic);
2380 
2381 	rc = bnxt_hwrm_vnic_free(bp, vnic);
2382 	if (rc)
2383 		return rc;
2384 
2385 	rte_free(vnic->fw_grp_ids);
2386 	vnic->fw_grp_ids = NULL;
2387 
2388 	vnic->rx_queue_cnt = 0;
2389 
2390 	return 0;
2391 }
2392 
2393 static int
2394 bnxt_config_vlan_hw_stripping(struct bnxt *bp, uint64_t rx_offloads)
2395 {
2396 	struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
2397 	int rc;
2398 
2399 	/* Destroy, recreate and reconfigure the default vnic */
2400 	rc = bnxt_free_one_vnic(bp, 0);
2401 	if (rc)
2402 		return rc;
2403 
2404 	/* default vnic 0 */
2405 	rc = bnxt_setup_one_vnic(bp, 0);
2406 	if (rc)
2407 		return rc;
2408 
2409 	if (bp->eth_dev->data->dev_conf.rxmode.offloads &
2410 	    DEV_RX_OFFLOAD_VLAN_FILTER) {
2411 		rc = bnxt_add_vlan_filter(bp, 0);
2412 		if (rc)
2413 			return rc;
2414 		rc = bnxt_restore_vlan_filters(bp);
2415 		if (rc)
2416 			return rc;
2417 	} else {
2418 		rc = bnxt_add_mac_filter(bp, vnic, NULL, 0, 0);
2419 		if (rc)
2420 			return rc;
2421 	}
2422 
2423 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
2424 	if (rc)
2425 		return rc;
2426 
2427 	PMD_DRV_LOG(DEBUG, "VLAN Strip Offload: %d\n",
2428 		    !!(rx_offloads & DEV_RX_OFFLOAD_VLAN_STRIP));
2429 
2430 	return rc;
2431 }
2432 
2433 static int
2434 bnxt_vlan_offload_set_op(struct rte_eth_dev *dev, int mask)
2435 {
2436 	uint64_t rx_offloads = dev->data->dev_conf.rxmode.offloads;
2437 	struct bnxt *bp = dev->data->dev_private;
2438 	int rc;
2439 
2440 	rc = is_bnxt_in_error(bp);
2441 	if (rc)
2442 		return rc;
2443 
2444 	/* Filter settings will get applied when port is started */
2445 	if (!dev->data->dev_started)
2446 		return 0;
2447 
2448 	if (mask & ETH_VLAN_FILTER_MASK) {
2449 		/* Enable or disable VLAN filtering */
2450 		rc = bnxt_config_vlan_hw_filter(bp, rx_offloads);
2451 		if (rc)
2452 			return rc;
2453 	}
2454 
2455 	if (mask & ETH_VLAN_STRIP_MASK) {
2456 		/* Enable or disable VLAN stripping */
2457 		rc = bnxt_config_vlan_hw_stripping(bp, rx_offloads);
2458 		if (rc)
2459 			return rc;
2460 	}
2461 
2462 	if (mask & ETH_VLAN_EXTEND_MASK) {
2463 		if (rx_offloads & DEV_RX_OFFLOAD_VLAN_EXTEND)
2464 			PMD_DRV_LOG(DEBUG, "Extend VLAN supported\n");
2465 		else
2466 			PMD_DRV_LOG(INFO, "Extend VLAN unsupported\n");
2467 	}
2468 
2469 	return 0;
2470 }
2471 
2472 static int
2473 bnxt_vlan_tpid_set_op(struct rte_eth_dev *dev, enum rte_vlan_type vlan_type,
2474 		      uint16_t tpid)
2475 {
2476 	struct bnxt *bp = dev->data->dev_private;
2477 	int qinq = dev->data->dev_conf.rxmode.offloads &
2478 		   DEV_RX_OFFLOAD_VLAN_EXTEND;
2479 
2480 	if (vlan_type != ETH_VLAN_TYPE_INNER &&
2481 	    vlan_type != ETH_VLAN_TYPE_OUTER) {
2482 		PMD_DRV_LOG(ERR,
2483 			    "Unsupported vlan type.");
2484 		return -EINVAL;
2485 	}
2486 	if (!qinq) {
2487 		PMD_DRV_LOG(ERR,
2488 			    "QinQ not enabled. Needs to be ON as we can "
2489 			    "accelerate only outer vlan\n");
2490 		return -EINVAL;
2491 	}
2492 
2493 	if (vlan_type == ETH_VLAN_TYPE_OUTER) {
2494 		switch (tpid) {
2495 		case RTE_ETHER_TYPE_QINQ:
2496 			bp->outer_tpid_bd =
2497 				TX_BD_LONG_CFA_META_VLAN_TPID_TPID88A8;
2498 				break;
2499 		case RTE_ETHER_TYPE_VLAN:
2500 			bp->outer_tpid_bd =
2501 				TX_BD_LONG_CFA_META_VLAN_TPID_TPID8100;
2502 				break;
2503 		case RTE_ETHER_TYPE_QINQ1:
2504 			bp->outer_tpid_bd =
2505 				TX_BD_LONG_CFA_META_VLAN_TPID_TPID9100;
2506 				break;
2507 		case RTE_ETHER_TYPE_QINQ2:
2508 			bp->outer_tpid_bd =
2509 				TX_BD_LONG_CFA_META_VLAN_TPID_TPID9200;
2510 				break;
2511 		case RTE_ETHER_TYPE_QINQ3:
2512 			bp->outer_tpid_bd =
2513 				 TX_BD_LONG_CFA_META_VLAN_TPID_TPID9300;
2514 				break;
2515 		default:
2516 			PMD_DRV_LOG(ERR, "Invalid TPID: %x\n", tpid);
2517 			return -EINVAL;
2518 		}
2519 		bp->outer_tpid_bd |= tpid;
2520 		PMD_DRV_LOG(INFO, "outer_tpid_bd = %x\n", bp->outer_tpid_bd);
2521 	} else if (vlan_type == ETH_VLAN_TYPE_INNER) {
2522 		PMD_DRV_LOG(ERR,
2523 			    "Can accelerate only outer vlan in QinQ\n");
2524 		return -EINVAL;
2525 	}
2526 
2527 	return 0;
2528 }
2529 
2530 static int
2531 bnxt_set_default_mac_addr_op(struct rte_eth_dev *dev,
2532 			     struct rte_ether_addr *addr)
2533 {
2534 	struct bnxt *bp = dev->data->dev_private;
2535 	/* Default Filter is tied to VNIC 0 */
2536 	struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
2537 	int rc;
2538 
2539 	rc = is_bnxt_in_error(bp);
2540 	if (rc)
2541 		return rc;
2542 
2543 	if (BNXT_VF(bp) && !BNXT_VF_IS_TRUSTED(bp))
2544 		return -EPERM;
2545 
2546 	if (rte_is_zero_ether_addr(addr))
2547 		return -EINVAL;
2548 
2549 	/* Filter settings will get applied when port is started */
2550 	if (!dev->data->dev_started)
2551 		return 0;
2552 
2553 	/* Check if the requested MAC is already added */
2554 	if (memcmp(addr, bp->mac_addr, RTE_ETHER_ADDR_LEN) == 0)
2555 		return 0;
2556 
2557 	/* Destroy filter and re-create it */
2558 	bnxt_del_dflt_mac_filter(bp, vnic);
2559 
2560 	memcpy(bp->mac_addr, addr, RTE_ETHER_ADDR_LEN);
2561 	if (dev->data->dev_conf.rxmode.offloads & DEV_RX_OFFLOAD_VLAN_FILTER) {
2562 		/* This filter will allow only untagged packets */
2563 		rc = bnxt_add_vlan_filter(bp, 0);
2564 	} else {
2565 		rc = bnxt_add_mac_filter(bp, vnic, addr, 0, 0);
2566 	}
2567 
2568 	PMD_DRV_LOG(DEBUG, "Set MAC addr\n");
2569 	return rc;
2570 }
2571 
2572 static int
2573 bnxt_dev_set_mc_addr_list_op(struct rte_eth_dev *eth_dev,
2574 			  struct rte_ether_addr *mc_addr_set,
2575 			  uint32_t nb_mc_addr)
2576 {
2577 	struct bnxt *bp = eth_dev->data->dev_private;
2578 	char *mc_addr_list = (char *)mc_addr_set;
2579 	struct bnxt_vnic_info *vnic;
2580 	uint32_t off = 0, i = 0;
2581 	int rc;
2582 
2583 	rc = is_bnxt_in_error(bp);
2584 	if (rc)
2585 		return rc;
2586 
2587 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
2588 
2589 	if (nb_mc_addr > BNXT_MAX_MC_ADDRS) {
2590 		vnic->flags |= BNXT_VNIC_INFO_ALLMULTI;
2591 		goto allmulti;
2592 	}
2593 
2594 	/* TODO Check for Duplicate mcast addresses */
2595 	vnic->flags &= ~BNXT_VNIC_INFO_ALLMULTI;
2596 	for (i = 0; i < nb_mc_addr; i++) {
2597 		memcpy(vnic->mc_list + off, &mc_addr_list[i],
2598 			RTE_ETHER_ADDR_LEN);
2599 		off += RTE_ETHER_ADDR_LEN;
2600 	}
2601 
2602 	vnic->mc_addr_cnt = i;
2603 	if (vnic->mc_addr_cnt)
2604 		vnic->flags |= BNXT_VNIC_INFO_MCAST;
2605 	else
2606 		vnic->flags &= ~BNXT_VNIC_INFO_MCAST;
2607 
2608 allmulti:
2609 	return bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
2610 }
2611 
2612 static int
2613 bnxt_fw_version_get(struct rte_eth_dev *dev, char *fw_version, size_t fw_size)
2614 {
2615 	struct bnxt *bp = dev->data->dev_private;
2616 	uint8_t fw_major = (bp->fw_ver >> 24) & 0xff;
2617 	uint8_t fw_minor = (bp->fw_ver >> 16) & 0xff;
2618 	uint8_t fw_updt = (bp->fw_ver >> 8) & 0xff;
2619 	uint8_t fw_rsvd = bp->fw_ver & 0xff;
2620 	int ret;
2621 
2622 	ret = snprintf(fw_version, fw_size, "%d.%d.%d.%d",
2623 			fw_major, fw_minor, fw_updt, fw_rsvd);
2624 
2625 	ret += 1; /* add the size of '\0' */
2626 	if (fw_size < (uint32_t)ret)
2627 		return ret;
2628 	else
2629 		return 0;
2630 }
2631 
2632 static void
2633 bnxt_rxq_info_get_op(struct rte_eth_dev *dev, uint16_t queue_id,
2634 	struct rte_eth_rxq_info *qinfo)
2635 {
2636 	struct bnxt *bp = dev->data->dev_private;
2637 	struct bnxt_rx_queue *rxq;
2638 
2639 	if (is_bnxt_in_error(bp))
2640 		return;
2641 
2642 	rxq = dev->data->rx_queues[queue_id];
2643 
2644 	qinfo->mp = rxq->mb_pool;
2645 	qinfo->scattered_rx = dev->data->scattered_rx;
2646 	qinfo->nb_desc = rxq->nb_rx_desc;
2647 
2648 	qinfo->conf.rx_free_thresh = rxq->rx_free_thresh;
2649 	qinfo->conf.rx_drop_en = 0;
2650 	qinfo->conf.rx_deferred_start = rxq->rx_deferred_start;
2651 }
2652 
2653 static void
2654 bnxt_txq_info_get_op(struct rte_eth_dev *dev, uint16_t queue_id,
2655 	struct rte_eth_txq_info *qinfo)
2656 {
2657 	struct bnxt *bp = dev->data->dev_private;
2658 	struct bnxt_tx_queue *txq;
2659 
2660 	if (is_bnxt_in_error(bp))
2661 		return;
2662 
2663 	txq = dev->data->tx_queues[queue_id];
2664 
2665 	qinfo->nb_desc = txq->nb_tx_desc;
2666 
2667 	qinfo->conf.tx_thresh.pthresh = txq->pthresh;
2668 	qinfo->conf.tx_thresh.hthresh = txq->hthresh;
2669 	qinfo->conf.tx_thresh.wthresh = txq->wthresh;
2670 
2671 	qinfo->conf.tx_free_thresh = txq->tx_free_thresh;
2672 	qinfo->conf.tx_rs_thresh = 0;
2673 	qinfo->conf.tx_deferred_start = txq->tx_deferred_start;
2674 }
2675 
2676 static int
2677 bnxt_rx_burst_mode_get(struct rte_eth_dev *dev, __rte_unused uint16_t queue_id,
2678 		       struct rte_eth_burst_mode *mode)
2679 {
2680 	eth_rx_burst_t pkt_burst = dev->rx_pkt_burst;
2681 
2682 	if (pkt_burst == bnxt_recv_pkts) {
2683 		snprintf(mode->info, sizeof(mode->info), "%s",
2684 			 "Scalar");
2685 		return 0;
2686 	}
2687 #if defined(RTE_ARCH_X86) || defined(RTE_ARCH_ARM64)
2688 	if (pkt_burst == bnxt_recv_pkts_vec) {
2689 		snprintf(mode->info, sizeof(mode->info), "%s",
2690 			 "Vector SSE");
2691 		return 0;
2692 	}
2693 #endif
2694 
2695 	return -EINVAL;
2696 }
2697 
2698 static int
2699 bnxt_tx_burst_mode_get(struct rte_eth_dev *dev, __rte_unused uint16_t queue_id,
2700 		       struct rte_eth_burst_mode *mode)
2701 {
2702 	eth_tx_burst_t pkt_burst = dev->tx_pkt_burst;
2703 
2704 	if (pkt_burst == bnxt_xmit_pkts) {
2705 		snprintf(mode->info, sizeof(mode->info), "%s",
2706 			 "Scalar");
2707 		return 0;
2708 	}
2709 #if defined(RTE_ARCH_X86) || defined(RTE_ARCH_ARM64)
2710 	if (pkt_burst == bnxt_xmit_pkts_vec) {
2711 		snprintf(mode->info, sizeof(mode->info), "%s",
2712 			 "Vector SSE");
2713 		return 0;
2714 	}
2715 #endif
2716 
2717 	return -EINVAL;
2718 }
2719 
2720 int bnxt_mtu_set_op(struct rte_eth_dev *eth_dev, uint16_t new_mtu)
2721 {
2722 	struct bnxt *bp = eth_dev->data->dev_private;
2723 	uint32_t new_pkt_size;
2724 	uint32_t rc = 0;
2725 	uint32_t i;
2726 
2727 	rc = is_bnxt_in_error(bp);
2728 	if (rc)
2729 		return rc;
2730 
2731 	/* Exit if receive queues are not configured yet */
2732 	if (!eth_dev->data->nb_rx_queues)
2733 		return rc;
2734 
2735 	new_pkt_size = new_mtu + RTE_ETHER_HDR_LEN + RTE_ETHER_CRC_LEN +
2736 		       VLAN_TAG_SIZE * BNXT_NUM_VLANS;
2737 
2738 #if defined(RTE_ARCH_X86) || defined(RTE_ARCH_ARM64)
2739 	/*
2740 	 * If vector-mode tx/rx is active, disallow any MTU change that would
2741 	 * require scattered receive support.
2742 	 */
2743 	if (eth_dev->data->dev_started &&
2744 	    (eth_dev->rx_pkt_burst == bnxt_recv_pkts_vec ||
2745 	     eth_dev->tx_pkt_burst == bnxt_xmit_pkts_vec) &&
2746 	    (new_pkt_size >
2747 	     eth_dev->data->min_rx_buf_size - RTE_PKTMBUF_HEADROOM)) {
2748 		PMD_DRV_LOG(ERR,
2749 			    "MTU change would require scattered rx support. ");
2750 		PMD_DRV_LOG(ERR, "Stop port before changing MTU.\n");
2751 		return -EINVAL;
2752 	}
2753 #endif
2754 
2755 	if (new_mtu > RTE_ETHER_MTU) {
2756 		bp->flags |= BNXT_FLAG_JUMBO;
2757 		bp->eth_dev->data->dev_conf.rxmode.offloads |=
2758 			DEV_RX_OFFLOAD_JUMBO_FRAME;
2759 	} else {
2760 		bp->eth_dev->data->dev_conf.rxmode.offloads &=
2761 			~DEV_RX_OFFLOAD_JUMBO_FRAME;
2762 		bp->flags &= ~BNXT_FLAG_JUMBO;
2763 	}
2764 
2765 	/* Is there a change in mtu setting? */
2766 	if (eth_dev->data->dev_conf.rxmode.max_rx_pkt_len == new_pkt_size)
2767 		return rc;
2768 
2769 	for (i = 0; i < bp->nr_vnics; i++) {
2770 		struct bnxt_vnic_info *vnic = &bp->vnic_info[i];
2771 		uint16_t size = 0;
2772 
2773 		vnic->mru = BNXT_VNIC_MRU(new_mtu);
2774 		rc = bnxt_hwrm_vnic_cfg(bp, vnic);
2775 		if (rc)
2776 			break;
2777 
2778 		size = rte_pktmbuf_data_room_size(bp->rx_queues[0]->mb_pool);
2779 		size -= RTE_PKTMBUF_HEADROOM;
2780 
2781 		if (size < new_mtu) {
2782 			rc = bnxt_hwrm_vnic_plcmode_cfg(bp, vnic);
2783 			if (rc)
2784 				return rc;
2785 		}
2786 	}
2787 
2788 	if (!rc)
2789 		eth_dev->data->dev_conf.rxmode.max_rx_pkt_len = new_pkt_size;
2790 
2791 	PMD_DRV_LOG(INFO, "New MTU is %d\n", new_mtu);
2792 
2793 	return rc;
2794 }
2795 
2796 static int
2797 bnxt_vlan_pvid_set_op(struct rte_eth_dev *dev, uint16_t pvid, int on)
2798 {
2799 	struct bnxt *bp = dev->data->dev_private;
2800 	uint16_t vlan = bp->vlan;
2801 	int rc;
2802 
2803 	rc = is_bnxt_in_error(bp);
2804 	if (rc)
2805 		return rc;
2806 
2807 	if (!BNXT_SINGLE_PF(bp) || BNXT_VF(bp)) {
2808 		PMD_DRV_LOG(ERR,
2809 			"PVID cannot be modified for this function\n");
2810 		return -ENOTSUP;
2811 	}
2812 	bp->vlan = on ? pvid : 0;
2813 
2814 	rc = bnxt_hwrm_set_default_vlan(bp, 0, 0);
2815 	if (rc)
2816 		bp->vlan = vlan;
2817 	return rc;
2818 }
2819 
2820 static int
2821 bnxt_dev_led_on_op(struct rte_eth_dev *dev)
2822 {
2823 	struct bnxt *bp = dev->data->dev_private;
2824 	int rc;
2825 
2826 	rc = is_bnxt_in_error(bp);
2827 	if (rc)
2828 		return rc;
2829 
2830 	return bnxt_hwrm_port_led_cfg(bp, true);
2831 }
2832 
2833 static int
2834 bnxt_dev_led_off_op(struct rte_eth_dev *dev)
2835 {
2836 	struct bnxt *bp = dev->data->dev_private;
2837 	int rc;
2838 
2839 	rc = is_bnxt_in_error(bp);
2840 	if (rc)
2841 		return rc;
2842 
2843 	return bnxt_hwrm_port_led_cfg(bp, false);
2844 }
2845 
2846 static uint32_t
2847 bnxt_rx_queue_count_op(struct rte_eth_dev *dev, uint16_t rx_queue_id)
2848 {
2849 	struct bnxt *bp = (struct bnxt *)dev->data->dev_private;
2850 	uint32_t desc = 0, raw_cons = 0, cons;
2851 	struct bnxt_cp_ring_info *cpr;
2852 	struct bnxt_rx_queue *rxq;
2853 	struct rx_pkt_cmpl *rxcmp;
2854 	int rc;
2855 
2856 	rc = is_bnxt_in_error(bp);
2857 	if (rc)
2858 		return rc;
2859 
2860 	rxq = dev->data->rx_queues[rx_queue_id];
2861 	cpr = rxq->cp_ring;
2862 	raw_cons = cpr->cp_raw_cons;
2863 
2864 	while (1) {
2865 		cons = RING_CMP(cpr->cp_ring_struct, raw_cons);
2866 		rte_prefetch0(&cpr->cp_desc_ring[cons]);
2867 		rxcmp = (struct rx_pkt_cmpl *)&cpr->cp_desc_ring[cons];
2868 
2869 		if (!CMP_VALID(rxcmp, raw_cons, cpr->cp_ring_struct)) {
2870 			break;
2871 		} else {
2872 			raw_cons++;
2873 			desc++;
2874 		}
2875 	}
2876 
2877 	return desc;
2878 }
2879 
2880 static int
2881 bnxt_rx_descriptor_status_op(void *rx_queue, uint16_t offset)
2882 {
2883 	struct bnxt_rx_queue *rxq = (struct bnxt_rx_queue *)rx_queue;
2884 	struct bnxt_rx_ring_info *rxr;
2885 	struct bnxt_cp_ring_info *cpr;
2886 	struct bnxt_sw_rx_bd *rx_buf;
2887 	struct rx_pkt_cmpl *rxcmp;
2888 	uint32_t cons, cp_cons;
2889 	int rc;
2890 
2891 	if (!rxq)
2892 		return -EINVAL;
2893 
2894 	rc = is_bnxt_in_error(rxq->bp);
2895 	if (rc)
2896 		return rc;
2897 
2898 	cpr = rxq->cp_ring;
2899 	rxr = rxq->rx_ring;
2900 
2901 	if (offset >= rxq->nb_rx_desc)
2902 		return -EINVAL;
2903 
2904 	cons = RING_CMP(cpr->cp_ring_struct, offset);
2905 	cp_cons = cpr->cp_raw_cons;
2906 	rxcmp = (struct rx_pkt_cmpl *)&cpr->cp_desc_ring[cons];
2907 
2908 	if (cons > cp_cons) {
2909 		if (CMPL_VALID(rxcmp, cpr->valid))
2910 			return RTE_ETH_RX_DESC_DONE;
2911 	} else {
2912 		if (CMPL_VALID(rxcmp, !cpr->valid))
2913 			return RTE_ETH_RX_DESC_DONE;
2914 	}
2915 	rx_buf = &rxr->rx_buf_ring[cons];
2916 	if (rx_buf->mbuf == NULL)
2917 		return RTE_ETH_RX_DESC_UNAVAIL;
2918 
2919 
2920 	return RTE_ETH_RX_DESC_AVAIL;
2921 }
2922 
2923 static int
2924 bnxt_tx_descriptor_status_op(void *tx_queue, uint16_t offset)
2925 {
2926 	struct bnxt_tx_queue *txq = (struct bnxt_tx_queue *)tx_queue;
2927 	struct bnxt_tx_ring_info *txr;
2928 	struct bnxt_cp_ring_info *cpr;
2929 	struct bnxt_sw_tx_bd *tx_buf;
2930 	struct tx_pkt_cmpl *txcmp;
2931 	uint32_t cons, cp_cons;
2932 	int rc;
2933 
2934 	if (!txq)
2935 		return -EINVAL;
2936 
2937 	rc = is_bnxt_in_error(txq->bp);
2938 	if (rc)
2939 		return rc;
2940 
2941 	cpr = txq->cp_ring;
2942 	txr = txq->tx_ring;
2943 
2944 	if (offset >= txq->nb_tx_desc)
2945 		return -EINVAL;
2946 
2947 	cons = RING_CMP(cpr->cp_ring_struct, offset);
2948 	txcmp = (struct tx_pkt_cmpl *)&cpr->cp_desc_ring[cons];
2949 	cp_cons = cpr->cp_raw_cons;
2950 
2951 	if (cons > cp_cons) {
2952 		if (CMPL_VALID(txcmp, cpr->valid))
2953 			return RTE_ETH_TX_DESC_UNAVAIL;
2954 	} else {
2955 		if (CMPL_VALID(txcmp, !cpr->valid))
2956 			return RTE_ETH_TX_DESC_UNAVAIL;
2957 	}
2958 	tx_buf = &txr->tx_buf_ring[cons];
2959 	if (tx_buf->mbuf == NULL)
2960 		return RTE_ETH_TX_DESC_DONE;
2961 
2962 	return RTE_ETH_TX_DESC_FULL;
2963 }
2964 
2965 static struct bnxt_filter_info *
2966 bnxt_match_and_validate_ether_filter(struct bnxt *bp,
2967 				struct rte_eth_ethertype_filter *efilter,
2968 				struct bnxt_vnic_info *vnic0,
2969 				struct bnxt_vnic_info *vnic,
2970 				int *ret)
2971 {
2972 	struct bnxt_filter_info *mfilter = NULL;
2973 	int match = 0;
2974 	*ret = 0;
2975 
2976 	if (efilter->ether_type == RTE_ETHER_TYPE_IPV4 ||
2977 		efilter->ether_type == RTE_ETHER_TYPE_IPV6) {
2978 		PMD_DRV_LOG(ERR, "invalid ether_type(0x%04x) in"
2979 			" ethertype filter.", efilter->ether_type);
2980 		*ret = -EINVAL;
2981 		goto exit;
2982 	}
2983 	if (efilter->queue >= bp->rx_nr_rings) {
2984 		PMD_DRV_LOG(ERR, "Invalid queue %d\n", efilter->queue);
2985 		*ret = -EINVAL;
2986 		goto exit;
2987 	}
2988 
2989 	vnic0 = BNXT_GET_DEFAULT_VNIC(bp);
2990 	vnic = &bp->vnic_info[efilter->queue];
2991 	if (vnic == NULL) {
2992 		PMD_DRV_LOG(ERR, "Invalid queue %d\n", efilter->queue);
2993 		*ret = -EINVAL;
2994 		goto exit;
2995 	}
2996 
2997 	if (efilter->flags & RTE_ETHTYPE_FLAGS_DROP) {
2998 		STAILQ_FOREACH(mfilter, &vnic0->filter, next) {
2999 			if ((!memcmp(efilter->mac_addr.addr_bytes,
3000 				     mfilter->l2_addr, RTE_ETHER_ADDR_LEN) &&
3001 			     mfilter->flags ==
3002 			     HWRM_CFA_NTUPLE_FILTER_ALLOC_INPUT_FLAGS_DROP &&
3003 			     mfilter->ethertype == efilter->ether_type)) {
3004 				match = 1;
3005 				break;
3006 			}
3007 		}
3008 	} else {
3009 		STAILQ_FOREACH(mfilter, &vnic->filter, next)
3010 			if ((!memcmp(efilter->mac_addr.addr_bytes,
3011 				     mfilter->l2_addr, RTE_ETHER_ADDR_LEN) &&
3012 			     mfilter->ethertype == efilter->ether_type &&
3013 			     mfilter->flags ==
3014 			     HWRM_CFA_L2_FILTER_CFG_INPUT_FLAGS_PATH_RX)) {
3015 				match = 1;
3016 				break;
3017 			}
3018 	}
3019 
3020 	if (match)
3021 		*ret = -EEXIST;
3022 
3023 exit:
3024 	return mfilter;
3025 }
3026 
3027 static int
3028 bnxt_ethertype_filter(struct rte_eth_dev *dev,
3029 			enum rte_filter_op filter_op,
3030 			void *arg)
3031 {
3032 	struct bnxt *bp = dev->data->dev_private;
3033 	struct rte_eth_ethertype_filter *efilter =
3034 			(struct rte_eth_ethertype_filter *)arg;
3035 	struct bnxt_filter_info *bfilter, *filter1;
3036 	struct bnxt_vnic_info *vnic, *vnic0;
3037 	int ret;
3038 
3039 	if (filter_op == RTE_ETH_FILTER_NOP)
3040 		return 0;
3041 
3042 	if (arg == NULL) {
3043 		PMD_DRV_LOG(ERR, "arg shouldn't be NULL for operation %u.",
3044 			    filter_op);
3045 		return -EINVAL;
3046 	}
3047 
3048 	vnic0 = BNXT_GET_DEFAULT_VNIC(bp);
3049 	vnic = &bp->vnic_info[efilter->queue];
3050 
3051 	switch (filter_op) {
3052 	case RTE_ETH_FILTER_ADD:
3053 		bnxt_match_and_validate_ether_filter(bp, efilter,
3054 							vnic0, vnic, &ret);
3055 		if (ret < 0)
3056 			return ret;
3057 
3058 		bfilter = bnxt_get_unused_filter(bp);
3059 		if (bfilter == NULL) {
3060 			PMD_DRV_LOG(ERR,
3061 				"Not enough resources for a new filter.\n");
3062 			return -ENOMEM;
3063 		}
3064 		bfilter->filter_type = HWRM_CFA_NTUPLE_FILTER;
3065 		memcpy(bfilter->l2_addr, efilter->mac_addr.addr_bytes,
3066 		       RTE_ETHER_ADDR_LEN);
3067 		memcpy(bfilter->dst_macaddr, efilter->mac_addr.addr_bytes,
3068 		       RTE_ETHER_ADDR_LEN);
3069 		bfilter->enables |= NTUPLE_FLTR_ALLOC_INPUT_EN_DST_MACADDR;
3070 		bfilter->ethertype = efilter->ether_type;
3071 		bfilter->enables |= NTUPLE_FLTR_ALLOC_INPUT_EN_ETHERTYPE;
3072 
3073 		filter1 = bnxt_get_l2_filter(bp, bfilter, vnic0);
3074 		if (filter1 == NULL) {
3075 			ret = -EINVAL;
3076 			goto cleanup;
3077 		}
3078 		bfilter->enables |=
3079 			HWRM_CFA_NTUPLE_FILTER_ALLOC_INPUT_ENABLES_L2_FILTER_ID;
3080 		bfilter->fw_l2_filter_id = filter1->fw_l2_filter_id;
3081 
3082 		bfilter->dst_id = vnic->fw_vnic_id;
3083 
3084 		if (efilter->flags & RTE_ETHTYPE_FLAGS_DROP) {
3085 			bfilter->flags =
3086 				HWRM_CFA_NTUPLE_FILTER_ALLOC_INPUT_FLAGS_DROP;
3087 		}
3088 
3089 		ret = bnxt_hwrm_set_ntuple_filter(bp, bfilter->dst_id, bfilter);
3090 		if (ret)
3091 			goto cleanup;
3092 		STAILQ_INSERT_TAIL(&vnic->filter, bfilter, next);
3093 		break;
3094 	case RTE_ETH_FILTER_DELETE:
3095 		filter1 = bnxt_match_and_validate_ether_filter(bp, efilter,
3096 							vnic0, vnic, &ret);
3097 		if (ret == -EEXIST) {
3098 			ret = bnxt_hwrm_clear_ntuple_filter(bp, filter1);
3099 
3100 			STAILQ_REMOVE(&vnic->filter, filter1, bnxt_filter_info,
3101 				      next);
3102 			bnxt_free_filter(bp, filter1);
3103 		} else if (ret == 0) {
3104 			PMD_DRV_LOG(ERR, "No matching filter found\n");
3105 		}
3106 		break;
3107 	default:
3108 		PMD_DRV_LOG(ERR, "unsupported operation %u.", filter_op);
3109 		ret = -EINVAL;
3110 		goto error;
3111 	}
3112 	return ret;
3113 cleanup:
3114 	bnxt_free_filter(bp, bfilter);
3115 error:
3116 	return ret;
3117 }
3118 
3119 static inline int
3120 parse_ntuple_filter(struct bnxt *bp,
3121 		    struct rte_eth_ntuple_filter *nfilter,
3122 		    struct bnxt_filter_info *bfilter)
3123 {
3124 	uint32_t en = 0;
3125 
3126 	if (nfilter->queue >= bp->rx_nr_rings) {
3127 		PMD_DRV_LOG(ERR, "Invalid queue %d\n", nfilter->queue);
3128 		return -EINVAL;
3129 	}
3130 
3131 	switch (nfilter->dst_port_mask) {
3132 	case UINT16_MAX:
3133 		bfilter->dst_port_mask = -1;
3134 		bfilter->dst_port = nfilter->dst_port;
3135 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_DST_PORT |
3136 			NTUPLE_FLTR_ALLOC_INPUT_EN_DST_PORT_MASK;
3137 		break;
3138 	default:
3139 		PMD_DRV_LOG(ERR, "invalid dst_port mask.");
3140 		return -EINVAL;
3141 	}
3142 
3143 	bfilter->ip_addr_type = NTUPLE_FLTR_ALLOC_INPUT_IP_ADDR_TYPE_IPV4;
3144 	en |= NTUPLE_FLTR_ALLOC_IN_EN_IP_PROTO;
3145 
3146 	switch (nfilter->proto_mask) {
3147 	case UINT8_MAX:
3148 		if (nfilter->proto == 17) /* IPPROTO_UDP */
3149 			bfilter->ip_protocol = 17;
3150 		else if (nfilter->proto == 6) /* IPPROTO_TCP */
3151 			bfilter->ip_protocol = 6;
3152 		else
3153 			return -EINVAL;
3154 		en |= NTUPLE_FLTR_ALLOC_IN_EN_IP_PROTO;
3155 		break;
3156 	default:
3157 		PMD_DRV_LOG(ERR, "invalid protocol mask.");
3158 		return -EINVAL;
3159 	}
3160 
3161 	switch (nfilter->dst_ip_mask) {
3162 	case UINT32_MAX:
3163 		bfilter->dst_ipaddr_mask[0] = -1;
3164 		bfilter->dst_ipaddr[0] = nfilter->dst_ip;
3165 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_DST_IPADDR |
3166 			NTUPLE_FLTR_ALLOC_INPUT_EN_DST_IPADDR_MASK;
3167 		break;
3168 	default:
3169 		PMD_DRV_LOG(ERR, "invalid dst_ip mask.");
3170 		return -EINVAL;
3171 	}
3172 
3173 	switch (nfilter->src_ip_mask) {
3174 	case UINT32_MAX:
3175 		bfilter->src_ipaddr_mask[0] = -1;
3176 		bfilter->src_ipaddr[0] = nfilter->src_ip;
3177 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_IPADDR |
3178 			NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_IPADDR_MASK;
3179 		break;
3180 	default:
3181 		PMD_DRV_LOG(ERR, "invalid src_ip mask.");
3182 		return -EINVAL;
3183 	}
3184 
3185 	switch (nfilter->src_port_mask) {
3186 	case UINT16_MAX:
3187 		bfilter->src_port_mask = -1;
3188 		bfilter->src_port = nfilter->src_port;
3189 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_PORT |
3190 			NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_PORT_MASK;
3191 		break;
3192 	default:
3193 		PMD_DRV_LOG(ERR, "invalid src_port mask.");
3194 		return -EINVAL;
3195 	}
3196 
3197 	bfilter->enables = en;
3198 	return 0;
3199 }
3200 
3201 static struct bnxt_filter_info*
3202 bnxt_match_ntuple_filter(struct bnxt *bp,
3203 			 struct bnxt_filter_info *bfilter,
3204 			 struct bnxt_vnic_info **mvnic)
3205 {
3206 	struct bnxt_filter_info *mfilter = NULL;
3207 	int i;
3208 
3209 	for (i = bp->nr_vnics - 1; i >= 0; i--) {
3210 		struct bnxt_vnic_info *vnic = &bp->vnic_info[i];
3211 		STAILQ_FOREACH(mfilter, &vnic->filter, next) {
3212 			if (bfilter->src_ipaddr[0] == mfilter->src_ipaddr[0] &&
3213 			    bfilter->src_ipaddr_mask[0] ==
3214 			    mfilter->src_ipaddr_mask[0] &&
3215 			    bfilter->src_port == mfilter->src_port &&
3216 			    bfilter->src_port_mask == mfilter->src_port_mask &&
3217 			    bfilter->dst_ipaddr[0] == mfilter->dst_ipaddr[0] &&
3218 			    bfilter->dst_ipaddr_mask[0] ==
3219 			    mfilter->dst_ipaddr_mask[0] &&
3220 			    bfilter->dst_port == mfilter->dst_port &&
3221 			    bfilter->dst_port_mask == mfilter->dst_port_mask &&
3222 			    bfilter->flags == mfilter->flags &&
3223 			    bfilter->enables == mfilter->enables) {
3224 				if (mvnic)
3225 					*mvnic = vnic;
3226 				return mfilter;
3227 			}
3228 		}
3229 	}
3230 	return NULL;
3231 }
3232 
3233 static int
3234 bnxt_cfg_ntuple_filter(struct bnxt *bp,
3235 		       struct rte_eth_ntuple_filter *nfilter,
3236 		       enum rte_filter_op filter_op)
3237 {
3238 	struct bnxt_filter_info *bfilter, *mfilter, *filter1;
3239 	struct bnxt_vnic_info *vnic, *vnic0, *mvnic;
3240 	int ret;
3241 
3242 	if (nfilter->flags != RTE_5TUPLE_FLAGS) {
3243 		PMD_DRV_LOG(ERR, "only 5tuple is supported.");
3244 		return -EINVAL;
3245 	}
3246 
3247 	if (nfilter->flags & RTE_NTUPLE_FLAGS_TCP_FLAG) {
3248 		PMD_DRV_LOG(ERR, "Ntuple filter: TCP flags not supported\n");
3249 		return -EINVAL;
3250 	}
3251 
3252 	bfilter = bnxt_get_unused_filter(bp);
3253 	if (bfilter == NULL) {
3254 		PMD_DRV_LOG(ERR,
3255 			"Not enough resources for a new filter.\n");
3256 		return -ENOMEM;
3257 	}
3258 	ret = parse_ntuple_filter(bp, nfilter, bfilter);
3259 	if (ret < 0)
3260 		goto free_filter;
3261 
3262 	vnic = &bp->vnic_info[nfilter->queue];
3263 	vnic0 = BNXT_GET_DEFAULT_VNIC(bp);
3264 	filter1 = STAILQ_FIRST(&vnic0->filter);
3265 	if (filter1 == NULL) {
3266 		ret = -EINVAL;
3267 		goto free_filter;
3268 	}
3269 
3270 	bfilter->dst_id = vnic->fw_vnic_id;
3271 	bfilter->fw_l2_filter_id = filter1->fw_l2_filter_id;
3272 	bfilter->enables |=
3273 		HWRM_CFA_NTUPLE_FILTER_ALLOC_INPUT_ENABLES_L2_FILTER_ID;
3274 	bfilter->ethertype = 0x800;
3275 	bfilter->enables |= NTUPLE_FLTR_ALLOC_INPUT_EN_ETHERTYPE;
3276 
3277 	mfilter = bnxt_match_ntuple_filter(bp, bfilter, &mvnic);
3278 
3279 	if (mfilter != NULL && filter_op == RTE_ETH_FILTER_ADD &&
3280 	    bfilter->dst_id == mfilter->dst_id) {
3281 		PMD_DRV_LOG(ERR, "filter exists.\n");
3282 		ret = -EEXIST;
3283 		goto free_filter;
3284 	} else if (mfilter != NULL && filter_op == RTE_ETH_FILTER_ADD &&
3285 		   bfilter->dst_id != mfilter->dst_id) {
3286 		mfilter->dst_id = vnic->fw_vnic_id;
3287 		ret = bnxt_hwrm_set_ntuple_filter(bp, mfilter->dst_id, mfilter);
3288 		STAILQ_REMOVE(&mvnic->filter, mfilter, bnxt_filter_info, next);
3289 		STAILQ_INSERT_TAIL(&vnic->filter, mfilter, next);
3290 		PMD_DRV_LOG(ERR, "filter with matching pattern exists.\n");
3291 		PMD_DRV_LOG(ERR, " Updated it to the new destination queue\n");
3292 		goto free_filter;
3293 	}
3294 	if (mfilter == NULL && filter_op == RTE_ETH_FILTER_DELETE) {
3295 		PMD_DRV_LOG(ERR, "filter doesn't exist.");
3296 		ret = -ENOENT;
3297 		goto free_filter;
3298 	}
3299 
3300 	if (filter_op == RTE_ETH_FILTER_ADD) {
3301 		bfilter->filter_type = HWRM_CFA_NTUPLE_FILTER;
3302 		ret = bnxt_hwrm_set_ntuple_filter(bp, bfilter->dst_id, bfilter);
3303 		if (ret)
3304 			goto free_filter;
3305 		STAILQ_INSERT_TAIL(&vnic->filter, bfilter, next);
3306 	} else {
3307 		if (mfilter == NULL) {
3308 			/* This should not happen. But for Coverity! */
3309 			ret = -ENOENT;
3310 			goto free_filter;
3311 		}
3312 		ret = bnxt_hwrm_clear_ntuple_filter(bp, mfilter);
3313 
3314 		STAILQ_REMOVE(&vnic->filter, mfilter, bnxt_filter_info, next);
3315 		bnxt_free_filter(bp, mfilter);
3316 		bnxt_free_filter(bp, bfilter);
3317 	}
3318 
3319 	return 0;
3320 free_filter:
3321 	bnxt_free_filter(bp, bfilter);
3322 	return ret;
3323 }
3324 
3325 static int
3326 bnxt_ntuple_filter(struct rte_eth_dev *dev,
3327 			enum rte_filter_op filter_op,
3328 			void *arg)
3329 {
3330 	struct bnxt *bp = dev->data->dev_private;
3331 	int ret;
3332 
3333 	if (filter_op == RTE_ETH_FILTER_NOP)
3334 		return 0;
3335 
3336 	if (arg == NULL) {
3337 		PMD_DRV_LOG(ERR, "arg shouldn't be NULL for operation %u.",
3338 			    filter_op);
3339 		return -EINVAL;
3340 	}
3341 
3342 	switch (filter_op) {
3343 	case RTE_ETH_FILTER_ADD:
3344 		ret = bnxt_cfg_ntuple_filter(bp,
3345 			(struct rte_eth_ntuple_filter *)arg,
3346 			filter_op);
3347 		break;
3348 	case RTE_ETH_FILTER_DELETE:
3349 		ret = bnxt_cfg_ntuple_filter(bp,
3350 			(struct rte_eth_ntuple_filter *)arg,
3351 			filter_op);
3352 		break;
3353 	default:
3354 		PMD_DRV_LOG(ERR, "unsupported operation %u.", filter_op);
3355 		ret = -EINVAL;
3356 		break;
3357 	}
3358 	return ret;
3359 }
3360 
3361 static int
3362 bnxt_parse_fdir_filter(struct bnxt *bp,
3363 		       struct rte_eth_fdir_filter *fdir,
3364 		       struct bnxt_filter_info *filter)
3365 {
3366 	enum rte_fdir_mode fdir_mode =
3367 		bp->eth_dev->data->dev_conf.fdir_conf.mode;
3368 	struct bnxt_vnic_info *vnic0, *vnic;
3369 	struct bnxt_filter_info *filter1;
3370 	uint32_t en = 0;
3371 	int i;
3372 
3373 	if (fdir_mode == RTE_FDIR_MODE_PERFECT_TUNNEL)
3374 		return -EINVAL;
3375 
3376 	filter->l2_ovlan = fdir->input.flow_ext.vlan_tci;
3377 	en |= EM_FLOW_ALLOC_INPUT_EN_OVLAN_VID;
3378 
3379 	switch (fdir->input.flow_type) {
3380 	case RTE_ETH_FLOW_IPV4:
3381 	case RTE_ETH_FLOW_NONFRAG_IPV4_OTHER:
3382 		/* FALLTHROUGH */
3383 		filter->src_ipaddr[0] = fdir->input.flow.ip4_flow.src_ip;
3384 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_IPADDR;
3385 		filter->dst_ipaddr[0] = fdir->input.flow.ip4_flow.dst_ip;
3386 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_DST_IPADDR;
3387 		filter->ip_protocol = fdir->input.flow.ip4_flow.proto;
3388 		en |= NTUPLE_FLTR_ALLOC_IN_EN_IP_PROTO;
3389 		filter->ip_addr_type =
3390 			NTUPLE_FLTR_ALLOC_INPUT_IP_ADDR_TYPE_IPV4;
3391 		filter->src_ipaddr_mask[0] = 0xffffffff;
3392 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_IPADDR_MASK;
3393 		filter->dst_ipaddr_mask[0] = 0xffffffff;
3394 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_DST_IPADDR_MASK;
3395 		filter->ethertype = 0x800;
3396 		filter->enables |= NTUPLE_FLTR_ALLOC_INPUT_EN_ETHERTYPE;
3397 		break;
3398 	case RTE_ETH_FLOW_NONFRAG_IPV4_TCP:
3399 		filter->src_port = fdir->input.flow.tcp4_flow.src_port;
3400 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_PORT;
3401 		filter->dst_port = fdir->input.flow.tcp4_flow.dst_port;
3402 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_DST_PORT;
3403 		filter->dst_port_mask = 0xffff;
3404 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_DST_PORT_MASK;
3405 		filter->src_port_mask = 0xffff;
3406 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_PORT_MASK;
3407 		filter->src_ipaddr[0] = fdir->input.flow.tcp4_flow.ip.src_ip;
3408 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_IPADDR;
3409 		filter->dst_ipaddr[0] = fdir->input.flow.tcp4_flow.ip.dst_ip;
3410 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_DST_IPADDR;
3411 		filter->ip_protocol = 6;
3412 		en |= NTUPLE_FLTR_ALLOC_IN_EN_IP_PROTO;
3413 		filter->ip_addr_type =
3414 			NTUPLE_FLTR_ALLOC_INPUT_IP_ADDR_TYPE_IPV4;
3415 		filter->src_ipaddr_mask[0] = 0xffffffff;
3416 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_IPADDR_MASK;
3417 		filter->dst_ipaddr_mask[0] = 0xffffffff;
3418 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_DST_IPADDR_MASK;
3419 		filter->ethertype = 0x800;
3420 		filter->enables |= NTUPLE_FLTR_ALLOC_INPUT_EN_ETHERTYPE;
3421 		break;
3422 	case RTE_ETH_FLOW_NONFRAG_IPV4_UDP:
3423 		filter->src_port = fdir->input.flow.udp4_flow.src_port;
3424 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_PORT;
3425 		filter->dst_port = fdir->input.flow.udp4_flow.dst_port;
3426 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_DST_PORT;
3427 		filter->dst_port_mask = 0xffff;
3428 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_DST_PORT_MASK;
3429 		filter->src_port_mask = 0xffff;
3430 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_PORT_MASK;
3431 		filter->src_ipaddr[0] = fdir->input.flow.udp4_flow.ip.src_ip;
3432 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_IPADDR;
3433 		filter->dst_ipaddr[0] = fdir->input.flow.udp4_flow.ip.dst_ip;
3434 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_DST_IPADDR;
3435 		filter->ip_protocol = 17;
3436 		en |= NTUPLE_FLTR_ALLOC_IN_EN_IP_PROTO;
3437 		filter->ip_addr_type =
3438 			NTUPLE_FLTR_ALLOC_INPUT_IP_ADDR_TYPE_IPV4;
3439 		filter->src_ipaddr_mask[0] = 0xffffffff;
3440 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_IPADDR_MASK;
3441 		filter->dst_ipaddr_mask[0] = 0xffffffff;
3442 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_DST_IPADDR_MASK;
3443 		filter->ethertype = 0x800;
3444 		filter->enables |= NTUPLE_FLTR_ALLOC_INPUT_EN_ETHERTYPE;
3445 		break;
3446 	case RTE_ETH_FLOW_IPV6:
3447 	case RTE_ETH_FLOW_NONFRAG_IPV6_OTHER:
3448 		/* FALLTHROUGH */
3449 		filter->ip_addr_type =
3450 			NTUPLE_FLTR_ALLOC_INPUT_IP_ADDR_TYPE_IPV6;
3451 		filter->ip_protocol = fdir->input.flow.ipv6_flow.proto;
3452 		en |= NTUPLE_FLTR_ALLOC_IN_EN_IP_PROTO;
3453 		rte_memcpy(filter->src_ipaddr,
3454 			   fdir->input.flow.ipv6_flow.src_ip, 16);
3455 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_IPADDR;
3456 		rte_memcpy(filter->dst_ipaddr,
3457 			   fdir->input.flow.ipv6_flow.dst_ip, 16);
3458 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_DST_IPADDR;
3459 		memset(filter->dst_ipaddr_mask, 0xff, 16);
3460 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_DST_IPADDR_MASK;
3461 		memset(filter->src_ipaddr_mask, 0xff, 16);
3462 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_IPADDR_MASK;
3463 		filter->ethertype = 0x86dd;
3464 		filter->enables |= NTUPLE_FLTR_ALLOC_INPUT_EN_ETHERTYPE;
3465 		break;
3466 	case RTE_ETH_FLOW_NONFRAG_IPV6_TCP:
3467 		filter->src_port = fdir->input.flow.tcp6_flow.src_port;
3468 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_PORT;
3469 		filter->dst_port = fdir->input.flow.tcp6_flow.dst_port;
3470 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_DST_PORT;
3471 		filter->dst_port_mask = 0xffff;
3472 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_DST_PORT_MASK;
3473 		filter->src_port_mask = 0xffff;
3474 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_PORT_MASK;
3475 		filter->ip_addr_type =
3476 			NTUPLE_FLTR_ALLOC_INPUT_IP_ADDR_TYPE_IPV6;
3477 		filter->ip_protocol = fdir->input.flow.tcp6_flow.ip.proto;
3478 		en |= NTUPLE_FLTR_ALLOC_IN_EN_IP_PROTO;
3479 		rte_memcpy(filter->src_ipaddr,
3480 			   fdir->input.flow.tcp6_flow.ip.src_ip, 16);
3481 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_IPADDR;
3482 		rte_memcpy(filter->dst_ipaddr,
3483 			   fdir->input.flow.tcp6_flow.ip.dst_ip, 16);
3484 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_DST_IPADDR;
3485 		memset(filter->dst_ipaddr_mask, 0xff, 16);
3486 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_DST_IPADDR_MASK;
3487 		memset(filter->src_ipaddr_mask, 0xff, 16);
3488 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_IPADDR_MASK;
3489 		filter->ethertype = 0x86dd;
3490 		filter->enables |= NTUPLE_FLTR_ALLOC_INPUT_EN_ETHERTYPE;
3491 		break;
3492 	case RTE_ETH_FLOW_NONFRAG_IPV6_UDP:
3493 		filter->src_port = fdir->input.flow.udp6_flow.src_port;
3494 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_PORT;
3495 		filter->dst_port = fdir->input.flow.udp6_flow.dst_port;
3496 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_DST_PORT;
3497 		filter->dst_port_mask = 0xffff;
3498 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_DST_PORT_MASK;
3499 		filter->src_port_mask = 0xffff;
3500 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_PORT_MASK;
3501 		filter->ip_addr_type =
3502 			NTUPLE_FLTR_ALLOC_INPUT_IP_ADDR_TYPE_IPV6;
3503 		filter->ip_protocol = fdir->input.flow.udp6_flow.ip.proto;
3504 		en |= NTUPLE_FLTR_ALLOC_IN_EN_IP_PROTO;
3505 		rte_memcpy(filter->src_ipaddr,
3506 			   fdir->input.flow.udp6_flow.ip.src_ip, 16);
3507 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_IPADDR;
3508 		rte_memcpy(filter->dst_ipaddr,
3509 			   fdir->input.flow.udp6_flow.ip.dst_ip, 16);
3510 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_DST_IPADDR;
3511 		memset(filter->dst_ipaddr_mask, 0xff, 16);
3512 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_DST_IPADDR_MASK;
3513 		memset(filter->src_ipaddr_mask, 0xff, 16);
3514 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_IPADDR_MASK;
3515 		filter->ethertype = 0x86dd;
3516 		filter->enables |= NTUPLE_FLTR_ALLOC_INPUT_EN_ETHERTYPE;
3517 		break;
3518 	case RTE_ETH_FLOW_L2_PAYLOAD:
3519 		filter->ethertype = fdir->input.flow.l2_flow.ether_type;
3520 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_ETHERTYPE;
3521 		break;
3522 	case RTE_ETH_FLOW_VXLAN:
3523 		if (fdir->action.behavior == RTE_ETH_FDIR_REJECT)
3524 			return -EINVAL;
3525 		filter->vni = fdir->input.flow.tunnel_flow.tunnel_id;
3526 		filter->tunnel_type =
3527 			CFA_NTUPLE_FILTER_ALLOC_REQ_TUNNEL_TYPE_VXLAN;
3528 		en |= HWRM_CFA_NTUPLE_FILTER_ALLOC_INPUT_ENABLES_TUNNEL_TYPE;
3529 		break;
3530 	case RTE_ETH_FLOW_NVGRE:
3531 		if (fdir->action.behavior == RTE_ETH_FDIR_REJECT)
3532 			return -EINVAL;
3533 		filter->vni = fdir->input.flow.tunnel_flow.tunnel_id;
3534 		filter->tunnel_type =
3535 			CFA_NTUPLE_FILTER_ALLOC_REQ_TUNNEL_TYPE_NVGRE;
3536 		en |= HWRM_CFA_NTUPLE_FILTER_ALLOC_INPUT_ENABLES_TUNNEL_TYPE;
3537 		break;
3538 	case RTE_ETH_FLOW_UNKNOWN:
3539 	case RTE_ETH_FLOW_RAW:
3540 	case RTE_ETH_FLOW_FRAG_IPV4:
3541 	case RTE_ETH_FLOW_NONFRAG_IPV4_SCTP:
3542 	case RTE_ETH_FLOW_FRAG_IPV6:
3543 	case RTE_ETH_FLOW_NONFRAG_IPV6_SCTP:
3544 	case RTE_ETH_FLOW_IPV6_EX:
3545 	case RTE_ETH_FLOW_IPV6_TCP_EX:
3546 	case RTE_ETH_FLOW_IPV6_UDP_EX:
3547 	case RTE_ETH_FLOW_GENEVE:
3548 		/* FALLTHROUGH */
3549 	default:
3550 		return -EINVAL;
3551 	}
3552 
3553 	vnic0 = BNXT_GET_DEFAULT_VNIC(bp);
3554 	vnic = &bp->vnic_info[fdir->action.rx_queue];
3555 	if (vnic == NULL) {
3556 		PMD_DRV_LOG(ERR, "Invalid queue %d\n", fdir->action.rx_queue);
3557 		return -EINVAL;
3558 	}
3559 
3560 	if (fdir_mode == RTE_FDIR_MODE_PERFECT_MAC_VLAN) {
3561 		rte_memcpy(filter->dst_macaddr,
3562 			fdir->input.flow.mac_vlan_flow.mac_addr.addr_bytes, 6);
3563 			en |= NTUPLE_FLTR_ALLOC_INPUT_EN_DST_MACADDR;
3564 	}
3565 
3566 	if (fdir->action.behavior == RTE_ETH_FDIR_REJECT) {
3567 		filter->flags = HWRM_CFA_NTUPLE_FILTER_ALLOC_INPUT_FLAGS_DROP;
3568 		filter1 = STAILQ_FIRST(&vnic0->filter);
3569 		//filter1 = bnxt_get_l2_filter(bp, filter, vnic0);
3570 	} else {
3571 		filter->dst_id = vnic->fw_vnic_id;
3572 		for (i = 0; i < RTE_ETHER_ADDR_LEN; i++)
3573 			if (filter->dst_macaddr[i] == 0x00)
3574 				filter1 = STAILQ_FIRST(&vnic0->filter);
3575 			else
3576 				filter1 = bnxt_get_l2_filter(bp, filter, vnic);
3577 	}
3578 
3579 	if (filter1 == NULL)
3580 		return -EINVAL;
3581 
3582 	en |= HWRM_CFA_NTUPLE_FILTER_ALLOC_INPUT_ENABLES_L2_FILTER_ID;
3583 	filter->fw_l2_filter_id = filter1->fw_l2_filter_id;
3584 
3585 	filter->enables = en;
3586 
3587 	return 0;
3588 }
3589 
3590 static struct bnxt_filter_info *
3591 bnxt_match_fdir(struct bnxt *bp, struct bnxt_filter_info *nf,
3592 		struct bnxt_vnic_info **mvnic)
3593 {
3594 	struct bnxt_filter_info *mf = NULL;
3595 	int i;
3596 
3597 	for (i = bp->nr_vnics - 1; i >= 0; i--) {
3598 		struct bnxt_vnic_info *vnic = &bp->vnic_info[i];
3599 
3600 		STAILQ_FOREACH(mf, &vnic->filter, next) {
3601 			if (mf->filter_type == nf->filter_type &&
3602 			    mf->flags == nf->flags &&
3603 			    mf->src_port == nf->src_port &&
3604 			    mf->src_port_mask == nf->src_port_mask &&
3605 			    mf->dst_port == nf->dst_port &&
3606 			    mf->dst_port_mask == nf->dst_port_mask &&
3607 			    mf->ip_protocol == nf->ip_protocol &&
3608 			    mf->ip_addr_type == nf->ip_addr_type &&
3609 			    mf->ethertype == nf->ethertype &&
3610 			    mf->vni == nf->vni &&
3611 			    mf->tunnel_type == nf->tunnel_type &&
3612 			    mf->l2_ovlan == nf->l2_ovlan &&
3613 			    mf->l2_ovlan_mask == nf->l2_ovlan_mask &&
3614 			    mf->l2_ivlan == nf->l2_ivlan &&
3615 			    mf->l2_ivlan_mask == nf->l2_ivlan_mask &&
3616 			    !memcmp(mf->l2_addr, nf->l2_addr,
3617 				    RTE_ETHER_ADDR_LEN) &&
3618 			    !memcmp(mf->l2_addr_mask, nf->l2_addr_mask,
3619 				    RTE_ETHER_ADDR_LEN) &&
3620 			    !memcmp(mf->src_macaddr, nf->src_macaddr,
3621 				    RTE_ETHER_ADDR_LEN) &&
3622 			    !memcmp(mf->dst_macaddr, nf->dst_macaddr,
3623 				    RTE_ETHER_ADDR_LEN) &&
3624 			    !memcmp(mf->src_ipaddr, nf->src_ipaddr,
3625 				    sizeof(nf->src_ipaddr)) &&
3626 			    !memcmp(mf->src_ipaddr_mask, nf->src_ipaddr_mask,
3627 				    sizeof(nf->src_ipaddr_mask)) &&
3628 			    !memcmp(mf->dst_ipaddr, nf->dst_ipaddr,
3629 				    sizeof(nf->dst_ipaddr)) &&
3630 			    !memcmp(mf->dst_ipaddr_mask, nf->dst_ipaddr_mask,
3631 				    sizeof(nf->dst_ipaddr_mask))) {
3632 				if (mvnic)
3633 					*mvnic = vnic;
3634 				return mf;
3635 			}
3636 		}
3637 	}
3638 	return NULL;
3639 }
3640 
3641 static int
3642 bnxt_fdir_filter(struct rte_eth_dev *dev,
3643 		 enum rte_filter_op filter_op,
3644 		 void *arg)
3645 {
3646 	struct bnxt *bp = dev->data->dev_private;
3647 	struct rte_eth_fdir_filter *fdir  = (struct rte_eth_fdir_filter *)arg;
3648 	struct bnxt_filter_info *filter, *match;
3649 	struct bnxt_vnic_info *vnic, *mvnic;
3650 	int ret = 0, i;
3651 
3652 	if (filter_op == RTE_ETH_FILTER_NOP)
3653 		return 0;
3654 
3655 	if (arg == NULL && filter_op != RTE_ETH_FILTER_FLUSH)
3656 		return -EINVAL;
3657 
3658 	switch (filter_op) {
3659 	case RTE_ETH_FILTER_ADD:
3660 	case RTE_ETH_FILTER_DELETE:
3661 		/* FALLTHROUGH */
3662 		filter = bnxt_get_unused_filter(bp);
3663 		if (filter == NULL) {
3664 			PMD_DRV_LOG(ERR,
3665 				"Not enough resources for a new flow.\n");
3666 			return -ENOMEM;
3667 		}
3668 
3669 		ret = bnxt_parse_fdir_filter(bp, fdir, filter);
3670 		if (ret != 0)
3671 			goto free_filter;
3672 		filter->filter_type = HWRM_CFA_NTUPLE_FILTER;
3673 
3674 		if (fdir->action.behavior == RTE_ETH_FDIR_REJECT)
3675 			vnic = &bp->vnic_info[0];
3676 		else
3677 			vnic = &bp->vnic_info[fdir->action.rx_queue];
3678 
3679 		match = bnxt_match_fdir(bp, filter, &mvnic);
3680 		if (match != NULL && filter_op == RTE_ETH_FILTER_ADD) {
3681 			if (match->dst_id == vnic->fw_vnic_id) {
3682 				PMD_DRV_LOG(ERR, "Flow already exists.\n");
3683 				ret = -EEXIST;
3684 				goto free_filter;
3685 			} else {
3686 				match->dst_id = vnic->fw_vnic_id;
3687 				ret = bnxt_hwrm_set_ntuple_filter(bp,
3688 								  match->dst_id,
3689 								  match);
3690 				STAILQ_REMOVE(&mvnic->filter, match,
3691 					      bnxt_filter_info, next);
3692 				STAILQ_INSERT_TAIL(&vnic->filter, match, next);
3693 				PMD_DRV_LOG(ERR,
3694 					"Filter with matching pattern exist\n");
3695 				PMD_DRV_LOG(ERR,
3696 					"Updated it to new destination q\n");
3697 				goto free_filter;
3698 			}
3699 		}
3700 		if (match == NULL && filter_op == RTE_ETH_FILTER_DELETE) {
3701 			PMD_DRV_LOG(ERR, "Flow does not exist.\n");
3702 			ret = -ENOENT;
3703 			goto free_filter;
3704 		}
3705 
3706 		if (filter_op == RTE_ETH_FILTER_ADD) {
3707 			ret = bnxt_hwrm_set_ntuple_filter(bp,
3708 							  filter->dst_id,
3709 							  filter);
3710 			if (ret)
3711 				goto free_filter;
3712 			STAILQ_INSERT_TAIL(&vnic->filter, filter, next);
3713 		} else {
3714 			ret = bnxt_hwrm_clear_ntuple_filter(bp, match);
3715 			STAILQ_REMOVE(&vnic->filter, match,
3716 				      bnxt_filter_info, next);
3717 			bnxt_free_filter(bp, match);
3718 			bnxt_free_filter(bp, filter);
3719 		}
3720 		break;
3721 	case RTE_ETH_FILTER_FLUSH:
3722 		for (i = bp->nr_vnics - 1; i >= 0; i--) {
3723 			struct bnxt_vnic_info *vnic = &bp->vnic_info[i];
3724 
3725 			STAILQ_FOREACH(filter, &vnic->filter, next) {
3726 				if (filter->filter_type ==
3727 				    HWRM_CFA_NTUPLE_FILTER) {
3728 					ret =
3729 					bnxt_hwrm_clear_ntuple_filter(bp,
3730 								      filter);
3731 					STAILQ_REMOVE(&vnic->filter, filter,
3732 						      bnxt_filter_info, next);
3733 				}
3734 			}
3735 		}
3736 		return ret;
3737 	case RTE_ETH_FILTER_UPDATE:
3738 	case RTE_ETH_FILTER_STATS:
3739 	case RTE_ETH_FILTER_INFO:
3740 		PMD_DRV_LOG(ERR, "operation %u not implemented", filter_op);
3741 		break;
3742 	default:
3743 		PMD_DRV_LOG(ERR, "unknown operation %u", filter_op);
3744 		ret = -EINVAL;
3745 		break;
3746 	}
3747 	return ret;
3748 
3749 free_filter:
3750 	bnxt_free_filter(bp, filter);
3751 	return ret;
3752 }
3753 
3754 int
3755 bnxt_filter_ctrl_op(struct rte_eth_dev *dev,
3756 		    enum rte_filter_type filter_type,
3757 		    enum rte_filter_op filter_op, void *arg)
3758 {
3759 	struct bnxt *bp = dev->data->dev_private;
3760 	int ret = 0;
3761 
3762 	if (BNXT_ETH_DEV_IS_REPRESENTOR(dev)) {
3763 		struct bnxt_vf_representor *vfr = dev->data->dev_private;
3764 		bp = vfr->parent_dev->data->dev_private;
3765 	}
3766 
3767 	ret = is_bnxt_in_error(bp);
3768 	if (ret)
3769 		return ret;
3770 
3771 	switch (filter_type) {
3772 	case RTE_ETH_FILTER_TUNNEL:
3773 		PMD_DRV_LOG(ERR,
3774 			"filter type: %d: To be implemented\n", filter_type);
3775 		break;
3776 	case RTE_ETH_FILTER_FDIR:
3777 		ret = bnxt_fdir_filter(dev, filter_op, arg);
3778 		break;
3779 	case RTE_ETH_FILTER_NTUPLE:
3780 		ret = bnxt_ntuple_filter(dev, filter_op, arg);
3781 		break;
3782 	case RTE_ETH_FILTER_ETHERTYPE:
3783 		ret = bnxt_ethertype_filter(dev, filter_op, arg);
3784 		break;
3785 	case RTE_ETH_FILTER_GENERIC:
3786 		if (filter_op != RTE_ETH_FILTER_GET)
3787 			return -EINVAL;
3788 		if (BNXT_TRUFLOW_EN(bp))
3789 			*(const void **)arg = &bnxt_ulp_rte_flow_ops;
3790 		else
3791 			*(const void **)arg = &bnxt_flow_ops;
3792 		break;
3793 	default:
3794 		PMD_DRV_LOG(ERR,
3795 			"Filter type (%d) not supported", filter_type);
3796 		ret = -EINVAL;
3797 		break;
3798 	}
3799 	return ret;
3800 }
3801 
3802 static const uint32_t *
3803 bnxt_dev_supported_ptypes_get_op(struct rte_eth_dev *dev)
3804 {
3805 	static const uint32_t ptypes[] = {
3806 		RTE_PTYPE_L2_ETHER_VLAN,
3807 		RTE_PTYPE_L3_IPV4_EXT_UNKNOWN,
3808 		RTE_PTYPE_L3_IPV6_EXT_UNKNOWN,
3809 		RTE_PTYPE_L4_ICMP,
3810 		RTE_PTYPE_L4_TCP,
3811 		RTE_PTYPE_L4_UDP,
3812 		RTE_PTYPE_INNER_L3_IPV4_EXT_UNKNOWN,
3813 		RTE_PTYPE_INNER_L3_IPV6_EXT_UNKNOWN,
3814 		RTE_PTYPE_INNER_L4_ICMP,
3815 		RTE_PTYPE_INNER_L4_TCP,
3816 		RTE_PTYPE_INNER_L4_UDP,
3817 		RTE_PTYPE_UNKNOWN
3818 	};
3819 
3820 	if (!dev->rx_pkt_burst)
3821 		return NULL;
3822 
3823 	return ptypes;
3824 }
3825 
3826 static int bnxt_map_regs(struct bnxt *bp, uint32_t *reg_arr, int count,
3827 			 int reg_win)
3828 {
3829 	uint32_t reg_base = *reg_arr & 0xfffff000;
3830 	uint32_t win_off;
3831 	int i;
3832 
3833 	for (i = 0; i < count; i++) {
3834 		if ((reg_arr[i] & 0xfffff000) != reg_base)
3835 			return -ERANGE;
3836 	}
3837 	win_off = BNXT_GRCPF_REG_WINDOW_BASE_OUT + (reg_win - 1) * 4;
3838 	rte_write32(reg_base, (uint8_t *)bp->bar0 + win_off);
3839 	return 0;
3840 }
3841 
3842 static int bnxt_map_ptp_regs(struct bnxt *bp)
3843 {
3844 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3845 	uint32_t *reg_arr;
3846 	int rc, i;
3847 
3848 	reg_arr = ptp->rx_regs;
3849 	rc = bnxt_map_regs(bp, reg_arr, BNXT_PTP_RX_REGS, 5);
3850 	if (rc)
3851 		return rc;
3852 
3853 	reg_arr = ptp->tx_regs;
3854 	rc = bnxt_map_regs(bp, reg_arr, BNXT_PTP_TX_REGS, 6);
3855 	if (rc)
3856 		return rc;
3857 
3858 	for (i = 0; i < BNXT_PTP_RX_REGS; i++)
3859 		ptp->rx_mapped_regs[i] = 0x5000 + (ptp->rx_regs[i] & 0xfff);
3860 
3861 	for (i = 0; i < BNXT_PTP_TX_REGS; i++)
3862 		ptp->tx_mapped_regs[i] = 0x6000 + (ptp->tx_regs[i] & 0xfff);
3863 
3864 	return 0;
3865 }
3866 
3867 static void bnxt_unmap_ptp_regs(struct bnxt *bp)
3868 {
3869 	rte_write32(0, (uint8_t *)bp->bar0 +
3870 			 BNXT_GRCPF_REG_WINDOW_BASE_OUT + 16);
3871 	rte_write32(0, (uint8_t *)bp->bar0 +
3872 			 BNXT_GRCPF_REG_WINDOW_BASE_OUT + 20);
3873 }
3874 
3875 static uint64_t bnxt_cc_read(struct bnxt *bp)
3876 {
3877 	uint64_t ns;
3878 
3879 	ns = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3880 			      BNXT_GRCPF_REG_SYNC_TIME));
3881 	ns |= (uint64_t)(rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3882 					  BNXT_GRCPF_REG_SYNC_TIME + 4))) << 32;
3883 	return ns;
3884 }
3885 
3886 static int bnxt_get_tx_ts(struct bnxt *bp, uint64_t *ts)
3887 {
3888 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3889 	uint32_t fifo;
3890 
3891 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3892 				ptp->tx_mapped_regs[BNXT_PTP_TX_FIFO]));
3893 	if (fifo & BNXT_PTP_TX_FIFO_EMPTY)
3894 		return -EAGAIN;
3895 
3896 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3897 				ptp->tx_mapped_regs[BNXT_PTP_TX_FIFO]));
3898 	*ts = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3899 				ptp->tx_mapped_regs[BNXT_PTP_TX_TS_L]));
3900 	*ts |= (uint64_t)rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3901 				ptp->tx_mapped_regs[BNXT_PTP_TX_TS_H])) << 32;
3902 
3903 	return 0;
3904 }
3905 
3906 static int bnxt_get_rx_ts(struct bnxt *bp, uint64_t *ts)
3907 {
3908 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3909 	struct bnxt_pf_info *pf = bp->pf;
3910 	uint16_t port_id;
3911 	uint32_t fifo;
3912 
3913 	if (!ptp)
3914 		return -ENODEV;
3915 
3916 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3917 				ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO]));
3918 	if (!(fifo & BNXT_PTP_RX_FIFO_PENDING))
3919 		return -EAGAIN;
3920 
3921 	port_id = pf->port_id;
3922 	rte_write32(1 << port_id, (uint8_t *)bp->bar0 +
3923 	       ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO_ADV]);
3924 
3925 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3926 				   ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO]));
3927 	if (fifo & BNXT_PTP_RX_FIFO_PENDING) {
3928 /*		bnxt_clr_rx_ts(bp);	  TBD  */
3929 		return -EBUSY;
3930 	}
3931 
3932 	*ts = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3933 				ptp->rx_mapped_regs[BNXT_PTP_RX_TS_L]));
3934 	*ts |= (uint64_t)rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3935 				ptp->rx_mapped_regs[BNXT_PTP_RX_TS_H])) << 32;
3936 
3937 	return 0;
3938 }
3939 
3940 static int
3941 bnxt_timesync_write_time(struct rte_eth_dev *dev, const struct timespec *ts)
3942 {
3943 	uint64_t ns;
3944 	struct bnxt *bp = dev->data->dev_private;
3945 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3946 
3947 	if (!ptp)
3948 		return 0;
3949 
3950 	ns = rte_timespec_to_ns(ts);
3951 	/* Set the timecounters to a new value. */
3952 	ptp->tc.nsec = ns;
3953 
3954 	return 0;
3955 }
3956 
3957 static int
3958 bnxt_timesync_read_time(struct rte_eth_dev *dev, struct timespec *ts)
3959 {
3960 	struct bnxt *bp = dev->data->dev_private;
3961 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3962 	uint64_t ns, systime_cycles = 0;
3963 	int rc = 0;
3964 
3965 	if (!ptp)
3966 		return 0;
3967 
3968 	if (BNXT_CHIP_THOR(bp))
3969 		rc = bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_CURRENT_TIME,
3970 					     &systime_cycles);
3971 	else
3972 		systime_cycles = bnxt_cc_read(bp);
3973 
3974 	ns = rte_timecounter_update(&ptp->tc, systime_cycles);
3975 	*ts = rte_ns_to_timespec(ns);
3976 
3977 	return rc;
3978 }
3979 static int
3980 bnxt_timesync_enable(struct rte_eth_dev *dev)
3981 {
3982 	struct bnxt *bp = dev->data->dev_private;
3983 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3984 	uint32_t shift = 0;
3985 	int rc;
3986 
3987 	if (!ptp)
3988 		return 0;
3989 
3990 	ptp->rx_filter = 1;
3991 	ptp->tx_tstamp_en = 1;
3992 	ptp->rxctl = BNXT_PTP_MSG_EVENTS;
3993 
3994 	rc = bnxt_hwrm_ptp_cfg(bp);
3995 	if (rc)
3996 		return rc;
3997 
3998 	memset(&ptp->tc, 0, sizeof(struct rte_timecounter));
3999 	memset(&ptp->rx_tstamp_tc, 0, sizeof(struct rte_timecounter));
4000 	memset(&ptp->tx_tstamp_tc, 0, sizeof(struct rte_timecounter));
4001 
4002 	ptp->tc.cc_mask = BNXT_CYCLECOUNTER_MASK;
4003 	ptp->tc.cc_shift = shift;
4004 	ptp->tc.nsec_mask = (1ULL << shift) - 1;
4005 
4006 	ptp->rx_tstamp_tc.cc_mask = BNXT_CYCLECOUNTER_MASK;
4007 	ptp->rx_tstamp_tc.cc_shift = shift;
4008 	ptp->rx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
4009 
4010 	ptp->tx_tstamp_tc.cc_mask = BNXT_CYCLECOUNTER_MASK;
4011 	ptp->tx_tstamp_tc.cc_shift = shift;
4012 	ptp->tx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
4013 
4014 	if (!BNXT_CHIP_THOR(bp))
4015 		bnxt_map_ptp_regs(bp);
4016 
4017 	return 0;
4018 }
4019 
4020 static int
4021 bnxt_timesync_disable(struct rte_eth_dev *dev)
4022 {
4023 	struct bnxt *bp = dev->data->dev_private;
4024 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
4025 
4026 	if (!ptp)
4027 		return 0;
4028 
4029 	ptp->rx_filter = 0;
4030 	ptp->tx_tstamp_en = 0;
4031 	ptp->rxctl = 0;
4032 
4033 	bnxt_hwrm_ptp_cfg(bp);
4034 
4035 	if (!BNXT_CHIP_THOR(bp))
4036 		bnxt_unmap_ptp_regs(bp);
4037 
4038 	return 0;
4039 }
4040 
4041 static int
4042 bnxt_timesync_read_rx_timestamp(struct rte_eth_dev *dev,
4043 				 struct timespec *timestamp,
4044 				 uint32_t flags __rte_unused)
4045 {
4046 	struct bnxt *bp = dev->data->dev_private;
4047 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
4048 	uint64_t rx_tstamp_cycles = 0;
4049 	uint64_t ns;
4050 
4051 	if (!ptp)
4052 		return 0;
4053 
4054 	if (BNXT_CHIP_THOR(bp))
4055 		rx_tstamp_cycles = ptp->rx_timestamp;
4056 	else
4057 		bnxt_get_rx_ts(bp, &rx_tstamp_cycles);
4058 
4059 	ns = rte_timecounter_update(&ptp->rx_tstamp_tc, rx_tstamp_cycles);
4060 	*timestamp = rte_ns_to_timespec(ns);
4061 	return  0;
4062 }
4063 
4064 static int
4065 bnxt_timesync_read_tx_timestamp(struct rte_eth_dev *dev,
4066 				 struct timespec *timestamp)
4067 {
4068 	struct bnxt *bp = dev->data->dev_private;
4069 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
4070 	uint64_t tx_tstamp_cycles = 0;
4071 	uint64_t ns;
4072 	int rc = 0;
4073 
4074 	if (!ptp)
4075 		return 0;
4076 
4077 	if (BNXT_CHIP_THOR(bp))
4078 		rc = bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_PATH_TX,
4079 					     &tx_tstamp_cycles);
4080 	else
4081 		rc = bnxt_get_tx_ts(bp, &tx_tstamp_cycles);
4082 
4083 	ns = rte_timecounter_update(&ptp->tx_tstamp_tc, tx_tstamp_cycles);
4084 	*timestamp = rte_ns_to_timespec(ns);
4085 
4086 	return rc;
4087 }
4088 
4089 static int
4090 bnxt_timesync_adjust_time(struct rte_eth_dev *dev, int64_t delta)
4091 {
4092 	struct bnxt *bp = dev->data->dev_private;
4093 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
4094 
4095 	if (!ptp)
4096 		return 0;
4097 
4098 	ptp->tc.nsec += delta;
4099 
4100 	return 0;
4101 }
4102 
4103 static int
4104 bnxt_get_eeprom_length_op(struct rte_eth_dev *dev)
4105 {
4106 	struct bnxt *bp = dev->data->dev_private;
4107 	int rc;
4108 	uint32_t dir_entries;
4109 	uint32_t entry_length;
4110 
4111 	rc = is_bnxt_in_error(bp);
4112 	if (rc)
4113 		return rc;
4114 
4115 	PMD_DRV_LOG(INFO, PCI_PRI_FMT "\n",
4116 		    bp->pdev->addr.domain, bp->pdev->addr.bus,
4117 		    bp->pdev->addr.devid, bp->pdev->addr.function);
4118 
4119 	rc = bnxt_hwrm_nvm_get_dir_info(bp, &dir_entries, &entry_length);
4120 	if (rc != 0)
4121 		return rc;
4122 
4123 	return dir_entries * entry_length;
4124 }
4125 
4126 static int
4127 bnxt_get_eeprom_op(struct rte_eth_dev *dev,
4128 		struct rte_dev_eeprom_info *in_eeprom)
4129 {
4130 	struct bnxt *bp = dev->data->dev_private;
4131 	uint32_t index;
4132 	uint32_t offset;
4133 	int rc;
4134 
4135 	rc = is_bnxt_in_error(bp);
4136 	if (rc)
4137 		return rc;
4138 
4139 	PMD_DRV_LOG(INFO, PCI_PRI_FMT " in_eeprom->offset = %d len = %d\n",
4140 		    bp->pdev->addr.domain, bp->pdev->addr.bus,
4141 		    bp->pdev->addr.devid, bp->pdev->addr.function,
4142 		    in_eeprom->offset, in_eeprom->length);
4143 
4144 	if (in_eeprom->offset == 0) /* special offset value to get directory */
4145 		return bnxt_get_nvram_directory(bp, in_eeprom->length,
4146 						in_eeprom->data);
4147 
4148 	index = in_eeprom->offset >> 24;
4149 	offset = in_eeprom->offset & 0xffffff;
4150 
4151 	if (index != 0)
4152 		return bnxt_hwrm_get_nvram_item(bp, index - 1, offset,
4153 					   in_eeprom->length, in_eeprom->data);
4154 
4155 	return 0;
4156 }
4157 
4158 static bool bnxt_dir_type_is_ape_bin_format(uint16_t dir_type)
4159 {
4160 	switch (dir_type) {
4161 	case BNX_DIR_TYPE_CHIMP_PATCH:
4162 	case BNX_DIR_TYPE_BOOTCODE:
4163 	case BNX_DIR_TYPE_BOOTCODE_2:
4164 	case BNX_DIR_TYPE_APE_FW:
4165 	case BNX_DIR_TYPE_APE_PATCH:
4166 	case BNX_DIR_TYPE_KONG_FW:
4167 	case BNX_DIR_TYPE_KONG_PATCH:
4168 	case BNX_DIR_TYPE_BONO_FW:
4169 	case BNX_DIR_TYPE_BONO_PATCH:
4170 		/* FALLTHROUGH */
4171 		return true;
4172 	}
4173 
4174 	return false;
4175 }
4176 
4177 static bool bnxt_dir_type_is_other_exec_format(uint16_t dir_type)
4178 {
4179 	switch (dir_type) {
4180 	case BNX_DIR_TYPE_AVS:
4181 	case BNX_DIR_TYPE_EXP_ROM_MBA:
4182 	case BNX_DIR_TYPE_PCIE:
4183 	case BNX_DIR_TYPE_TSCF_UCODE:
4184 	case BNX_DIR_TYPE_EXT_PHY:
4185 	case BNX_DIR_TYPE_CCM:
4186 	case BNX_DIR_TYPE_ISCSI_BOOT:
4187 	case BNX_DIR_TYPE_ISCSI_BOOT_IPV6:
4188 	case BNX_DIR_TYPE_ISCSI_BOOT_IPV4N6:
4189 		/* FALLTHROUGH */
4190 		return true;
4191 	}
4192 
4193 	return false;
4194 }
4195 
4196 static bool bnxt_dir_type_is_executable(uint16_t dir_type)
4197 {
4198 	return bnxt_dir_type_is_ape_bin_format(dir_type) ||
4199 		bnxt_dir_type_is_other_exec_format(dir_type);
4200 }
4201 
4202 static int
4203 bnxt_set_eeprom_op(struct rte_eth_dev *dev,
4204 		struct rte_dev_eeprom_info *in_eeprom)
4205 {
4206 	struct bnxt *bp = dev->data->dev_private;
4207 	uint8_t index, dir_op;
4208 	uint16_t type, ext, ordinal, attr;
4209 	int rc;
4210 
4211 	rc = is_bnxt_in_error(bp);
4212 	if (rc)
4213 		return rc;
4214 
4215 	PMD_DRV_LOG(INFO, PCI_PRI_FMT " in_eeprom->offset = %d len = %d\n",
4216 		    bp->pdev->addr.domain, bp->pdev->addr.bus,
4217 		    bp->pdev->addr.devid, bp->pdev->addr.function,
4218 		    in_eeprom->offset, in_eeprom->length);
4219 
4220 	if (!BNXT_PF(bp)) {
4221 		PMD_DRV_LOG(ERR, "NVM write not supported from a VF\n");
4222 		return -EINVAL;
4223 	}
4224 
4225 	type = in_eeprom->magic >> 16;
4226 
4227 	if (type == 0xffff) { /* special value for directory operations */
4228 		index = in_eeprom->magic & 0xff;
4229 		dir_op = in_eeprom->magic >> 8;
4230 		if (index == 0)
4231 			return -EINVAL;
4232 		switch (dir_op) {
4233 		case 0x0e: /* erase */
4234 			if (in_eeprom->offset != ~in_eeprom->magic)
4235 				return -EINVAL;
4236 			return bnxt_hwrm_erase_nvram_directory(bp, index - 1);
4237 		default:
4238 			return -EINVAL;
4239 		}
4240 	}
4241 
4242 	/* Create or re-write an NVM item: */
4243 	if (bnxt_dir_type_is_executable(type) == true)
4244 		return -EOPNOTSUPP;
4245 	ext = in_eeprom->magic & 0xffff;
4246 	ordinal = in_eeprom->offset >> 16;
4247 	attr = in_eeprom->offset & 0xffff;
4248 
4249 	return bnxt_hwrm_flash_nvram(bp, type, ordinal, ext, attr,
4250 				     in_eeprom->data, in_eeprom->length);
4251 }
4252 
4253 /*
4254  * Initialization
4255  */
4256 
4257 static const struct eth_dev_ops bnxt_dev_ops = {
4258 	.dev_infos_get = bnxt_dev_info_get_op,
4259 	.dev_close = bnxt_dev_close_op,
4260 	.dev_configure = bnxt_dev_configure_op,
4261 	.dev_start = bnxt_dev_start_op,
4262 	.dev_stop = bnxt_dev_stop_op,
4263 	.dev_set_link_up = bnxt_dev_set_link_up_op,
4264 	.dev_set_link_down = bnxt_dev_set_link_down_op,
4265 	.stats_get = bnxt_stats_get_op,
4266 	.stats_reset = bnxt_stats_reset_op,
4267 	.rx_queue_setup = bnxt_rx_queue_setup_op,
4268 	.rx_queue_release = bnxt_rx_queue_release_op,
4269 	.tx_queue_setup = bnxt_tx_queue_setup_op,
4270 	.tx_queue_release = bnxt_tx_queue_release_op,
4271 	.rx_queue_intr_enable = bnxt_rx_queue_intr_enable_op,
4272 	.rx_queue_intr_disable = bnxt_rx_queue_intr_disable_op,
4273 	.reta_update = bnxt_reta_update_op,
4274 	.reta_query = bnxt_reta_query_op,
4275 	.rss_hash_update = bnxt_rss_hash_update_op,
4276 	.rss_hash_conf_get = bnxt_rss_hash_conf_get_op,
4277 	.link_update = bnxt_link_update_op,
4278 	.promiscuous_enable = bnxt_promiscuous_enable_op,
4279 	.promiscuous_disable = bnxt_promiscuous_disable_op,
4280 	.allmulticast_enable = bnxt_allmulticast_enable_op,
4281 	.allmulticast_disable = bnxt_allmulticast_disable_op,
4282 	.mac_addr_add = bnxt_mac_addr_add_op,
4283 	.mac_addr_remove = bnxt_mac_addr_remove_op,
4284 	.flow_ctrl_get = bnxt_flow_ctrl_get_op,
4285 	.flow_ctrl_set = bnxt_flow_ctrl_set_op,
4286 	.udp_tunnel_port_add  = bnxt_udp_tunnel_port_add_op,
4287 	.udp_tunnel_port_del  = bnxt_udp_tunnel_port_del_op,
4288 	.vlan_filter_set = bnxt_vlan_filter_set_op,
4289 	.vlan_offload_set = bnxt_vlan_offload_set_op,
4290 	.vlan_tpid_set = bnxt_vlan_tpid_set_op,
4291 	.vlan_pvid_set = bnxt_vlan_pvid_set_op,
4292 	.mtu_set = bnxt_mtu_set_op,
4293 	.mac_addr_set = bnxt_set_default_mac_addr_op,
4294 	.xstats_get = bnxt_dev_xstats_get_op,
4295 	.xstats_get_names = bnxt_dev_xstats_get_names_op,
4296 	.xstats_reset = bnxt_dev_xstats_reset_op,
4297 	.fw_version_get = bnxt_fw_version_get,
4298 	.set_mc_addr_list = bnxt_dev_set_mc_addr_list_op,
4299 	.rxq_info_get = bnxt_rxq_info_get_op,
4300 	.txq_info_get = bnxt_txq_info_get_op,
4301 	.rx_burst_mode_get = bnxt_rx_burst_mode_get,
4302 	.tx_burst_mode_get = bnxt_tx_burst_mode_get,
4303 	.dev_led_on = bnxt_dev_led_on_op,
4304 	.dev_led_off = bnxt_dev_led_off_op,
4305 	.xstats_get_by_id = bnxt_dev_xstats_get_by_id_op,
4306 	.xstats_get_names_by_id = bnxt_dev_xstats_get_names_by_id_op,
4307 	.rx_queue_count = bnxt_rx_queue_count_op,
4308 	.rx_descriptor_status = bnxt_rx_descriptor_status_op,
4309 	.tx_descriptor_status = bnxt_tx_descriptor_status_op,
4310 	.rx_queue_start = bnxt_rx_queue_start,
4311 	.rx_queue_stop = bnxt_rx_queue_stop,
4312 	.tx_queue_start = bnxt_tx_queue_start,
4313 	.tx_queue_stop = bnxt_tx_queue_stop,
4314 	.filter_ctrl = bnxt_filter_ctrl_op,
4315 	.dev_supported_ptypes_get = bnxt_dev_supported_ptypes_get_op,
4316 	.get_eeprom_length    = bnxt_get_eeprom_length_op,
4317 	.get_eeprom           = bnxt_get_eeprom_op,
4318 	.set_eeprom           = bnxt_set_eeprom_op,
4319 	.timesync_enable      = bnxt_timesync_enable,
4320 	.timesync_disable     = bnxt_timesync_disable,
4321 	.timesync_read_time   = bnxt_timesync_read_time,
4322 	.timesync_write_time   = bnxt_timesync_write_time,
4323 	.timesync_adjust_time = bnxt_timesync_adjust_time,
4324 	.timesync_read_rx_timestamp = bnxt_timesync_read_rx_timestamp,
4325 	.timesync_read_tx_timestamp = bnxt_timesync_read_tx_timestamp,
4326 };
4327 
4328 static uint32_t bnxt_map_reset_regs(struct bnxt *bp, uint32_t reg)
4329 {
4330 	uint32_t offset;
4331 
4332 	/* Only pre-map the reset GRC registers using window 3 */
4333 	rte_write32(reg & 0xfffff000, (uint8_t *)bp->bar0 +
4334 		    BNXT_GRCPF_REG_WINDOW_BASE_OUT + 8);
4335 
4336 	offset = BNXT_GRCP_WINDOW_3_BASE + (reg & 0xffc);
4337 
4338 	return offset;
4339 }
4340 
4341 int bnxt_map_fw_health_status_regs(struct bnxt *bp)
4342 {
4343 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4344 	uint32_t reg_base = 0xffffffff;
4345 	int i;
4346 
4347 	/* Only pre-map the monitoring GRC registers using window 2 */
4348 	for (i = 0; i < BNXT_FW_STATUS_REG_CNT; i++) {
4349 		uint32_t reg = info->status_regs[i];
4350 
4351 		if (BNXT_FW_STATUS_REG_TYPE(reg) != BNXT_FW_STATUS_REG_TYPE_GRC)
4352 			continue;
4353 
4354 		if (reg_base == 0xffffffff)
4355 			reg_base = reg & 0xfffff000;
4356 		if ((reg & 0xfffff000) != reg_base)
4357 			return -ERANGE;
4358 
4359 		/* Use mask 0xffc as the Lower 2 bits indicates
4360 		 * address space location
4361 		 */
4362 		info->mapped_status_regs[i] = BNXT_GRCP_WINDOW_2_BASE +
4363 						(reg & 0xffc);
4364 	}
4365 
4366 	if (reg_base == 0xffffffff)
4367 		return 0;
4368 
4369 	rte_write32(reg_base, (uint8_t *)bp->bar0 +
4370 		    BNXT_GRCPF_REG_WINDOW_BASE_OUT + 4);
4371 
4372 	return 0;
4373 }
4374 
4375 static void bnxt_write_fw_reset_reg(struct bnxt *bp, uint32_t index)
4376 {
4377 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4378 	uint32_t delay = info->delay_after_reset[index];
4379 	uint32_t val = info->reset_reg_val[index];
4380 	uint32_t reg = info->reset_reg[index];
4381 	uint32_t type, offset;
4382 
4383 	type = BNXT_FW_STATUS_REG_TYPE(reg);
4384 	offset = BNXT_FW_STATUS_REG_OFF(reg);
4385 
4386 	switch (type) {
4387 	case BNXT_FW_STATUS_REG_TYPE_CFG:
4388 		rte_pci_write_config(bp->pdev, &val, sizeof(val), offset);
4389 		break;
4390 	case BNXT_FW_STATUS_REG_TYPE_GRC:
4391 		offset = bnxt_map_reset_regs(bp, offset);
4392 		rte_write32(val, (uint8_t *)bp->bar0 + offset);
4393 		break;
4394 	case BNXT_FW_STATUS_REG_TYPE_BAR0:
4395 		rte_write32(val, (uint8_t *)bp->bar0 + offset);
4396 		break;
4397 	}
4398 	/* wait on a specific interval of time until core reset is complete */
4399 	if (delay)
4400 		rte_delay_ms(delay);
4401 }
4402 
4403 static void bnxt_dev_cleanup(struct bnxt *bp)
4404 {
4405 	bnxt_set_hwrm_link_config(bp, false);
4406 	bp->link_info->link_up = 0;
4407 	if (bp->eth_dev->data->dev_started)
4408 		bnxt_dev_stop_op(bp->eth_dev);
4409 
4410 	bnxt_uninit_resources(bp, true);
4411 }
4412 
4413 static int bnxt_restore_vlan_filters(struct bnxt *bp)
4414 {
4415 	struct rte_eth_dev *dev = bp->eth_dev;
4416 	struct rte_vlan_filter_conf *vfc;
4417 	int vidx, vbit, rc;
4418 	uint16_t vlan_id;
4419 
4420 	for (vlan_id = 1; vlan_id <= RTE_ETHER_MAX_VLAN_ID; vlan_id++) {
4421 		vfc = &dev->data->vlan_filter_conf;
4422 		vidx = vlan_id / 64;
4423 		vbit = vlan_id % 64;
4424 
4425 		/* Each bit corresponds to a VLAN id */
4426 		if (vfc->ids[vidx] & (UINT64_C(1) << vbit)) {
4427 			rc = bnxt_add_vlan_filter(bp, vlan_id);
4428 			if (rc)
4429 				return rc;
4430 		}
4431 	}
4432 
4433 	return 0;
4434 }
4435 
4436 static int bnxt_restore_mac_filters(struct bnxt *bp)
4437 {
4438 	struct rte_eth_dev *dev = bp->eth_dev;
4439 	struct rte_eth_dev_info dev_info;
4440 	struct rte_ether_addr *addr;
4441 	uint64_t pool_mask;
4442 	uint32_t pool = 0;
4443 	uint16_t i;
4444 	int rc;
4445 
4446 	if (BNXT_VF(bp) & !BNXT_VF_IS_TRUSTED(bp))
4447 		return 0;
4448 
4449 	rc = bnxt_dev_info_get_op(dev, &dev_info);
4450 	if (rc)
4451 		return rc;
4452 
4453 	/* replay MAC address configuration */
4454 	for (i = 1; i < dev_info.max_mac_addrs; i++) {
4455 		addr = &dev->data->mac_addrs[i];
4456 
4457 		/* skip zero address */
4458 		if (rte_is_zero_ether_addr(addr))
4459 			continue;
4460 
4461 		pool = 0;
4462 		pool_mask = dev->data->mac_pool_sel[i];
4463 
4464 		do {
4465 			if (pool_mask & 1ULL) {
4466 				rc = bnxt_mac_addr_add_op(dev, addr, i, pool);
4467 				if (rc)
4468 					return rc;
4469 			}
4470 			pool_mask >>= 1;
4471 			pool++;
4472 		} while (pool_mask);
4473 	}
4474 
4475 	return 0;
4476 }
4477 
4478 static int bnxt_restore_filters(struct bnxt *bp)
4479 {
4480 	struct rte_eth_dev *dev = bp->eth_dev;
4481 	int ret = 0;
4482 
4483 	if (dev->data->all_multicast) {
4484 		ret = bnxt_allmulticast_enable_op(dev);
4485 		if (ret)
4486 			return ret;
4487 	}
4488 	if (dev->data->promiscuous) {
4489 		ret = bnxt_promiscuous_enable_op(dev);
4490 		if (ret)
4491 			return ret;
4492 	}
4493 
4494 	ret = bnxt_restore_mac_filters(bp);
4495 	if (ret)
4496 		return ret;
4497 
4498 	ret = bnxt_restore_vlan_filters(bp);
4499 	/* TODO restore other filters as well */
4500 	return ret;
4501 }
4502 
4503 static void bnxt_dev_recover(void *arg)
4504 {
4505 	struct bnxt *bp = arg;
4506 	int timeout = bp->fw_reset_max_msecs;
4507 	int rc = 0;
4508 
4509 	/* Clear Error flag so that device re-init should happen */
4510 	bp->flags &= ~BNXT_FLAG_FATAL_ERROR;
4511 
4512 	do {
4513 		rc = bnxt_hwrm_ver_get(bp, SHORT_HWRM_CMD_TIMEOUT);
4514 		if (rc == 0)
4515 			break;
4516 		rte_delay_ms(BNXT_FW_READY_WAIT_INTERVAL);
4517 		timeout -= BNXT_FW_READY_WAIT_INTERVAL;
4518 	} while (rc && timeout);
4519 
4520 	if (rc) {
4521 		PMD_DRV_LOG(ERR, "FW is not Ready after reset\n");
4522 		goto err;
4523 	}
4524 
4525 	rc = bnxt_init_resources(bp, true);
4526 	if (rc) {
4527 		PMD_DRV_LOG(ERR,
4528 			    "Failed to initialize resources after reset\n");
4529 		goto err;
4530 	}
4531 	/* clear reset flag as the device is initialized now */
4532 	bp->flags &= ~BNXT_FLAG_FW_RESET;
4533 
4534 	rc = bnxt_dev_start_op(bp->eth_dev);
4535 	if (rc) {
4536 		PMD_DRV_LOG(ERR, "Failed to start port after reset\n");
4537 		goto err_start;
4538 	}
4539 
4540 	rc = bnxt_restore_filters(bp);
4541 	if (rc)
4542 		goto err_start;
4543 
4544 	PMD_DRV_LOG(INFO, "Recovered from FW reset\n");
4545 	return;
4546 err_start:
4547 	bnxt_dev_stop_op(bp->eth_dev);
4548 err:
4549 	bp->flags |= BNXT_FLAG_FATAL_ERROR;
4550 	bnxt_uninit_resources(bp, false);
4551 	PMD_DRV_LOG(ERR, "Failed to recover from FW reset\n");
4552 }
4553 
4554 void bnxt_dev_reset_and_resume(void *arg)
4555 {
4556 	struct bnxt *bp = arg;
4557 	int rc;
4558 
4559 	bnxt_dev_cleanup(bp);
4560 
4561 	bnxt_wait_for_device_shutdown(bp);
4562 
4563 	rc = rte_eal_alarm_set(US_PER_MS * bp->fw_reset_min_msecs,
4564 			       bnxt_dev_recover, (void *)bp);
4565 	if (rc)
4566 		PMD_DRV_LOG(ERR, "Error setting recovery alarm");
4567 }
4568 
4569 uint32_t bnxt_read_fw_status_reg(struct bnxt *bp, uint32_t index)
4570 {
4571 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4572 	uint32_t reg = info->status_regs[index];
4573 	uint32_t type, offset, val = 0;
4574 
4575 	type = BNXT_FW_STATUS_REG_TYPE(reg);
4576 	offset = BNXT_FW_STATUS_REG_OFF(reg);
4577 
4578 	switch (type) {
4579 	case BNXT_FW_STATUS_REG_TYPE_CFG:
4580 		rte_pci_read_config(bp->pdev, &val, sizeof(val), offset);
4581 		break;
4582 	case BNXT_FW_STATUS_REG_TYPE_GRC:
4583 		offset = info->mapped_status_regs[index];
4584 		/* FALLTHROUGH */
4585 	case BNXT_FW_STATUS_REG_TYPE_BAR0:
4586 		val = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
4587 				       offset));
4588 		break;
4589 	}
4590 
4591 	return val;
4592 }
4593 
4594 static int bnxt_fw_reset_all(struct bnxt *bp)
4595 {
4596 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4597 	uint32_t i;
4598 	int rc = 0;
4599 
4600 	if (info->flags & BNXT_FLAG_ERROR_RECOVERY_HOST) {
4601 		/* Reset through master function driver */
4602 		for (i = 0; i < info->reg_array_cnt; i++)
4603 			bnxt_write_fw_reset_reg(bp, i);
4604 		/* Wait for time specified by FW after triggering reset */
4605 		rte_delay_ms(info->master_func_wait_period_after_reset);
4606 	} else if (info->flags & BNXT_FLAG_ERROR_RECOVERY_CO_CPU) {
4607 		/* Reset with the help of Kong processor */
4608 		rc = bnxt_hwrm_fw_reset(bp);
4609 		if (rc)
4610 			PMD_DRV_LOG(ERR, "Failed to reset FW\n");
4611 	}
4612 
4613 	return rc;
4614 }
4615 
4616 static void bnxt_fw_reset_cb(void *arg)
4617 {
4618 	struct bnxt *bp = arg;
4619 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4620 	int rc = 0;
4621 
4622 	/* Only Master function can do FW reset */
4623 	if (bnxt_is_master_func(bp) &&
4624 	    bnxt_is_recovery_enabled(bp)) {
4625 		rc = bnxt_fw_reset_all(bp);
4626 		if (rc) {
4627 			PMD_DRV_LOG(ERR, "Adapter recovery failed\n");
4628 			return;
4629 		}
4630 	}
4631 
4632 	/* if recovery method is ERROR_RECOVERY_CO_CPU, KONG will send
4633 	 * EXCEPTION_FATAL_ASYNC event to all the functions
4634 	 * (including MASTER FUNC). After receiving this Async, all the active
4635 	 * drivers should treat this case as FW initiated recovery
4636 	 */
4637 	if (info->flags & BNXT_FLAG_ERROR_RECOVERY_HOST) {
4638 		bp->fw_reset_min_msecs = BNXT_MIN_FW_READY_TIMEOUT;
4639 		bp->fw_reset_max_msecs = BNXT_MAX_FW_RESET_TIMEOUT;
4640 
4641 		/* To recover from error */
4642 		rte_eal_alarm_set(US_PER_MS, bnxt_dev_reset_and_resume,
4643 				  (void *)bp);
4644 	}
4645 }
4646 
4647 /* Driver should poll FW heartbeat, reset_counter with the frequency
4648  * advertised by FW in HWRM_ERROR_RECOVERY_QCFG.
4649  * When the driver detects heartbeat stop or change in reset_counter,
4650  * it has to trigger a reset to recover from the error condition.
4651  * A “master PF” is the function who will have the privilege to
4652  * initiate the chimp reset. The master PF will be elected by the
4653  * firmware and will be notified through async message.
4654  */
4655 static void bnxt_check_fw_health(void *arg)
4656 {
4657 	struct bnxt *bp = arg;
4658 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4659 	uint32_t val = 0, wait_msec;
4660 
4661 	if (!info || !bnxt_is_recovery_enabled(bp) ||
4662 	    is_bnxt_in_error(bp))
4663 		return;
4664 
4665 	val = bnxt_read_fw_status_reg(bp, BNXT_FW_HEARTBEAT_CNT_REG);
4666 	if (val == info->last_heart_beat)
4667 		goto reset;
4668 
4669 	info->last_heart_beat = val;
4670 
4671 	val = bnxt_read_fw_status_reg(bp, BNXT_FW_RECOVERY_CNT_REG);
4672 	if (val != info->last_reset_counter)
4673 		goto reset;
4674 
4675 	info->last_reset_counter = val;
4676 
4677 	rte_eal_alarm_set(US_PER_MS * info->driver_polling_freq,
4678 			  bnxt_check_fw_health, (void *)bp);
4679 
4680 	return;
4681 reset:
4682 	/* Stop DMA to/from device */
4683 	bp->flags |= BNXT_FLAG_FATAL_ERROR;
4684 	bp->flags |= BNXT_FLAG_FW_RESET;
4685 
4686 	PMD_DRV_LOG(ERR, "Detected FW dead condition\n");
4687 
4688 	if (bnxt_is_master_func(bp))
4689 		wait_msec = info->master_func_wait_period;
4690 	else
4691 		wait_msec = info->normal_func_wait_period;
4692 
4693 	rte_eal_alarm_set(US_PER_MS * wait_msec,
4694 			  bnxt_fw_reset_cb, (void *)bp);
4695 }
4696 
4697 void bnxt_schedule_fw_health_check(struct bnxt *bp)
4698 {
4699 	uint32_t polling_freq;
4700 
4701 	if (!bnxt_is_recovery_enabled(bp))
4702 		return;
4703 
4704 	if (bp->flags & BNXT_FLAG_FW_HEALTH_CHECK_SCHEDULED)
4705 		return;
4706 
4707 	polling_freq = bp->recovery_info->driver_polling_freq;
4708 
4709 	rte_eal_alarm_set(US_PER_MS * polling_freq,
4710 			  bnxt_check_fw_health, (void *)bp);
4711 	bp->flags |= BNXT_FLAG_FW_HEALTH_CHECK_SCHEDULED;
4712 }
4713 
4714 static void bnxt_cancel_fw_health_check(struct bnxt *bp)
4715 {
4716 	if (!bnxt_is_recovery_enabled(bp))
4717 		return;
4718 
4719 	rte_eal_alarm_cancel(bnxt_check_fw_health, (void *)bp);
4720 	bp->flags &= ~BNXT_FLAG_FW_HEALTH_CHECK_SCHEDULED;
4721 }
4722 
4723 static bool bnxt_vf_pciid(uint16_t device_id)
4724 {
4725 	switch (device_id) {
4726 	case BROADCOM_DEV_ID_57304_VF:
4727 	case BROADCOM_DEV_ID_57406_VF:
4728 	case BROADCOM_DEV_ID_5731X_VF:
4729 	case BROADCOM_DEV_ID_5741X_VF:
4730 	case BROADCOM_DEV_ID_57414_VF:
4731 	case BROADCOM_DEV_ID_STRATUS_NIC_VF1:
4732 	case BROADCOM_DEV_ID_STRATUS_NIC_VF2:
4733 	case BROADCOM_DEV_ID_58802_VF:
4734 	case BROADCOM_DEV_ID_57500_VF1:
4735 	case BROADCOM_DEV_ID_57500_VF2:
4736 		/* FALLTHROUGH */
4737 		return true;
4738 	default:
4739 		return false;
4740 	}
4741 }
4742 
4743 static bool bnxt_thor_device(uint16_t device_id)
4744 {
4745 	switch (device_id) {
4746 	case BROADCOM_DEV_ID_57508:
4747 	case BROADCOM_DEV_ID_57504:
4748 	case BROADCOM_DEV_ID_57502:
4749 	case BROADCOM_DEV_ID_57508_MF1:
4750 	case BROADCOM_DEV_ID_57504_MF1:
4751 	case BROADCOM_DEV_ID_57502_MF1:
4752 	case BROADCOM_DEV_ID_57508_MF2:
4753 	case BROADCOM_DEV_ID_57504_MF2:
4754 	case BROADCOM_DEV_ID_57502_MF2:
4755 	case BROADCOM_DEV_ID_57500_VF1:
4756 	case BROADCOM_DEV_ID_57500_VF2:
4757 		/* FALLTHROUGH */
4758 		return true;
4759 	default:
4760 		return false;
4761 	}
4762 }
4763 
4764 bool bnxt_stratus_device(struct bnxt *bp)
4765 {
4766 	uint16_t device_id = bp->pdev->id.device_id;
4767 
4768 	switch (device_id) {
4769 	case BROADCOM_DEV_ID_STRATUS_NIC:
4770 	case BROADCOM_DEV_ID_STRATUS_NIC_VF1:
4771 	case BROADCOM_DEV_ID_STRATUS_NIC_VF2:
4772 		/* FALLTHROUGH */
4773 		return true;
4774 	default:
4775 		return false;
4776 	}
4777 }
4778 
4779 static int bnxt_init_board(struct rte_eth_dev *eth_dev)
4780 {
4781 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
4782 	struct bnxt *bp = eth_dev->data->dev_private;
4783 
4784 	/* enable device (incl. PCI PM wakeup), and bus-mastering */
4785 	bp->bar0 = (void *)pci_dev->mem_resource[0].addr;
4786 	bp->doorbell_base = (void *)pci_dev->mem_resource[2].addr;
4787 	if (!bp->bar0 || !bp->doorbell_base) {
4788 		PMD_DRV_LOG(ERR, "Unable to access Hardware\n");
4789 		return -ENODEV;
4790 	}
4791 
4792 	bp->eth_dev = eth_dev;
4793 	bp->pdev = pci_dev;
4794 
4795 	return 0;
4796 }
4797 
4798 static int bnxt_alloc_ctx_mem_blk(struct bnxt *bp,
4799 				  struct bnxt_ctx_pg_info *ctx_pg,
4800 				  uint32_t mem_size,
4801 				  const char *suffix,
4802 				  uint16_t idx)
4803 {
4804 	struct bnxt_ring_mem_info *rmem = &ctx_pg->ring_mem;
4805 	const struct rte_memzone *mz = NULL;
4806 	char mz_name[RTE_MEMZONE_NAMESIZE];
4807 	rte_iova_t mz_phys_addr;
4808 	uint64_t valid_bits = 0;
4809 	uint32_t sz;
4810 	int i;
4811 
4812 	if (!mem_size)
4813 		return 0;
4814 
4815 	rmem->nr_pages = RTE_ALIGN_MUL_CEIL(mem_size, BNXT_PAGE_SIZE) /
4816 			 BNXT_PAGE_SIZE;
4817 	rmem->page_size = BNXT_PAGE_SIZE;
4818 	rmem->pg_arr = ctx_pg->ctx_pg_arr;
4819 	rmem->dma_arr = ctx_pg->ctx_dma_arr;
4820 	rmem->flags = BNXT_RMEM_VALID_PTE_FLAG;
4821 
4822 	valid_bits = PTU_PTE_VALID;
4823 
4824 	if (rmem->nr_pages > 1) {
4825 		snprintf(mz_name, RTE_MEMZONE_NAMESIZE,
4826 			 "bnxt_ctx_pg_tbl%s_%x_%d",
4827 			 suffix, idx, bp->eth_dev->data->port_id);
4828 		mz_name[RTE_MEMZONE_NAMESIZE - 1] = 0;
4829 		mz = rte_memzone_lookup(mz_name);
4830 		if (!mz) {
4831 			mz = rte_memzone_reserve_aligned(mz_name,
4832 						rmem->nr_pages * 8,
4833 						SOCKET_ID_ANY,
4834 						RTE_MEMZONE_2MB |
4835 						RTE_MEMZONE_SIZE_HINT_ONLY |
4836 						RTE_MEMZONE_IOVA_CONTIG,
4837 						BNXT_PAGE_SIZE);
4838 			if (mz == NULL)
4839 				return -ENOMEM;
4840 		}
4841 
4842 		memset(mz->addr, 0, mz->len);
4843 		mz_phys_addr = mz->iova;
4844 
4845 		rmem->pg_tbl = mz->addr;
4846 		rmem->pg_tbl_map = mz_phys_addr;
4847 		rmem->pg_tbl_mz = mz;
4848 	}
4849 
4850 	snprintf(mz_name, RTE_MEMZONE_NAMESIZE, "bnxt_ctx_%s_%x_%d",
4851 		 suffix, idx, bp->eth_dev->data->port_id);
4852 	mz = rte_memzone_lookup(mz_name);
4853 	if (!mz) {
4854 		mz = rte_memzone_reserve_aligned(mz_name,
4855 						 mem_size,
4856 						 SOCKET_ID_ANY,
4857 						 RTE_MEMZONE_1GB |
4858 						 RTE_MEMZONE_SIZE_HINT_ONLY |
4859 						 RTE_MEMZONE_IOVA_CONTIG,
4860 						 BNXT_PAGE_SIZE);
4861 		if (mz == NULL)
4862 			return -ENOMEM;
4863 	}
4864 
4865 	memset(mz->addr, 0, mz->len);
4866 	mz_phys_addr = mz->iova;
4867 
4868 	for (sz = 0, i = 0; sz < mem_size; sz += BNXT_PAGE_SIZE, i++) {
4869 		rmem->pg_arr[i] = ((char *)mz->addr) + sz;
4870 		rmem->dma_arr[i] = mz_phys_addr + sz;
4871 
4872 		if (rmem->nr_pages > 1) {
4873 			if (i == rmem->nr_pages - 2 &&
4874 			    (rmem->flags & BNXT_RMEM_RING_PTE_FLAG))
4875 				valid_bits |= PTU_PTE_NEXT_TO_LAST;
4876 			else if (i == rmem->nr_pages - 1 &&
4877 				 (rmem->flags & BNXT_RMEM_RING_PTE_FLAG))
4878 				valid_bits |= PTU_PTE_LAST;
4879 
4880 			rmem->pg_tbl[i] = rte_cpu_to_le_64(rmem->dma_arr[i] |
4881 							   valid_bits);
4882 		}
4883 	}
4884 
4885 	rmem->mz = mz;
4886 	if (rmem->vmem_size)
4887 		rmem->vmem = (void **)mz->addr;
4888 	rmem->dma_arr[0] = mz_phys_addr;
4889 	return 0;
4890 }
4891 
4892 static void bnxt_free_ctx_mem(struct bnxt *bp)
4893 {
4894 	int i;
4895 
4896 	if (!bp->ctx || !(bp->ctx->flags & BNXT_CTX_FLAG_INITED))
4897 		return;
4898 
4899 	bp->ctx->flags &= ~BNXT_CTX_FLAG_INITED;
4900 	rte_memzone_free(bp->ctx->qp_mem.ring_mem.mz);
4901 	rte_memzone_free(bp->ctx->srq_mem.ring_mem.mz);
4902 	rte_memzone_free(bp->ctx->cq_mem.ring_mem.mz);
4903 	rte_memzone_free(bp->ctx->vnic_mem.ring_mem.mz);
4904 	rte_memzone_free(bp->ctx->stat_mem.ring_mem.mz);
4905 	rte_memzone_free(bp->ctx->qp_mem.ring_mem.pg_tbl_mz);
4906 	rte_memzone_free(bp->ctx->srq_mem.ring_mem.pg_tbl_mz);
4907 	rte_memzone_free(bp->ctx->cq_mem.ring_mem.pg_tbl_mz);
4908 	rte_memzone_free(bp->ctx->vnic_mem.ring_mem.pg_tbl_mz);
4909 	rte_memzone_free(bp->ctx->stat_mem.ring_mem.pg_tbl_mz);
4910 
4911 	for (i = 0; i < bp->ctx->tqm_fp_rings_count + 1; i++) {
4912 		if (bp->ctx->tqm_mem[i])
4913 			rte_memzone_free(bp->ctx->tqm_mem[i]->ring_mem.mz);
4914 	}
4915 
4916 	rte_free(bp->ctx);
4917 	bp->ctx = NULL;
4918 }
4919 
4920 #define bnxt_roundup(x, y)   ((((x) + ((y) - 1)) / (y)) * (y))
4921 
4922 #define min_t(type, x, y) ({                    \
4923 	type __min1 = (x);                      \
4924 	type __min2 = (y);                      \
4925 	__min1 < __min2 ? __min1 : __min2; })
4926 
4927 #define max_t(type, x, y) ({                    \
4928 	type __max1 = (x);                      \
4929 	type __max2 = (y);                      \
4930 	__max1 > __max2 ? __max1 : __max2; })
4931 
4932 #define clamp_t(type, _x, min, max)     min_t(type, max_t(type, _x, min), max)
4933 
4934 int bnxt_alloc_ctx_mem(struct bnxt *bp)
4935 {
4936 	struct bnxt_ctx_pg_info *ctx_pg;
4937 	struct bnxt_ctx_mem_info *ctx;
4938 	uint32_t mem_size, ena, entries;
4939 	uint32_t entries_sp, min;
4940 	int i, rc;
4941 
4942 	rc = bnxt_hwrm_func_backing_store_qcaps(bp);
4943 	if (rc) {
4944 		PMD_DRV_LOG(ERR, "Query context mem capability failed\n");
4945 		return rc;
4946 	}
4947 	ctx = bp->ctx;
4948 	if (!ctx || (ctx->flags & BNXT_CTX_FLAG_INITED))
4949 		return 0;
4950 
4951 	ctx_pg = &ctx->qp_mem;
4952 	ctx_pg->entries = ctx->qp_min_qp1_entries + ctx->qp_max_l2_entries;
4953 	mem_size = ctx->qp_entry_size * ctx_pg->entries;
4954 	rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "qp_mem", 0);
4955 	if (rc)
4956 		return rc;
4957 
4958 	ctx_pg = &ctx->srq_mem;
4959 	ctx_pg->entries = ctx->srq_max_l2_entries;
4960 	mem_size = ctx->srq_entry_size * ctx_pg->entries;
4961 	rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "srq_mem", 0);
4962 	if (rc)
4963 		return rc;
4964 
4965 	ctx_pg = &ctx->cq_mem;
4966 	ctx_pg->entries = ctx->cq_max_l2_entries;
4967 	mem_size = ctx->cq_entry_size * ctx_pg->entries;
4968 	rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "cq_mem", 0);
4969 	if (rc)
4970 		return rc;
4971 
4972 	ctx_pg = &ctx->vnic_mem;
4973 	ctx_pg->entries = ctx->vnic_max_vnic_entries +
4974 		ctx->vnic_max_ring_table_entries;
4975 	mem_size = ctx->vnic_entry_size * ctx_pg->entries;
4976 	rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "vnic_mem", 0);
4977 	if (rc)
4978 		return rc;
4979 
4980 	ctx_pg = &ctx->stat_mem;
4981 	ctx_pg->entries = ctx->stat_max_entries;
4982 	mem_size = ctx->stat_entry_size * ctx_pg->entries;
4983 	rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "stat_mem", 0);
4984 	if (rc)
4985 		return rc;
4986 
4987 	min = ctx->tqm_min_entries_per_ring;
4988 
4989 	entries_sp = ctx->qp_max_l2_entries +
4990 		     ctx->vnic_max_vnic_entries +
4991 		     2 * ctx->qp_min_qp1_entries + min;
4992 	entries_sp = bnxt_roundup(entries_sp, ctx->tqm_entries_multiple);
4993 
4994 	entries = ctx->qp_max_l2_entries + ctx->qp_min_qp1_entries;
4995 	entries = bnxt_roundup(entries, ctx->tqm_entries_multiple);
4996 	entries = clamp_t(uint32_t, entries, min,
4997 			  ctx->tqm_max_entries_per_ring);
4998 	for (i = 0, ena = 0; i < ctx->tqm_fp_rings_count + 1; i++) {
4999 		ctx_pg = ctx->tqm_mem[i];
5000 		ctx_pg->entries = i ? entries : entries_sp;
5001 		mem_size = ctx->tqm_entry_size * ctx_pg->entries;
5002 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "tqm_mem", i);
5003 		if (rc)
5004 			return rc;
5005 		ena |= HWRM_FUNC_BACKING_STORE_CFG_INPUT_ENABLES_TQM_SP << i;
5006 	}
5007 
5008 	ena |= FUNC_BACKING_STORE_CFG_INPUT_DFLT_ENABLES;
5009 	rc = bnxt_hwrm_func_backing_store_cfg(bp, ena);
5010 	if (rc)
5011 		PMD_DRV_LOG(ERR,
5012 			    "Failed to configure context mem: rc = %d\n", rc);
5013 	else
5014 		ctx->flags |= BNXT_CTX_FLAG_INITED;
5015 
5016 	return rc;
5017 }
5018 
5019 static int bnxt_alloc_stats_mem(struct bnxt *bp)
5020 {
5021 	struct rte_pci_device *pci_dev = bp->pdev;
5022 	char mz_name[RTE_MEMZONE_NAMESIZE];
5023 	const struct rte_memzone *mz = NULL;
5024 	uint32_t total_alloc_len;
5025 	rte_iova_t mz_phys_addr;
5026 
5027 	if (pci_dev->id.device_id == BROADCOM_DEV_ID_NS2)
5028 		return 0;
5029 
5030 	snprintf(mz_name, RTE_MEMZONE_NAMESIZE,
5031 		 "bnxt_" PCI_PRI_FMT "-%s", pci_dev->addr.domain,
5032 		 pci_dev->addr.bus, pci_dev->addr.devid,
5033 		 pci_dev->addr.function, "rx_port_stats");
5034 	mz_name[RTE_MEMZONE_NAMESIZE - 1] = 0;
5035 	mz = rte_memzone_lookup(mz_name);
5036 	total_alloc_len =
5037 		RTE_CACHE_LINE_ROUNDUP(sizeof(struct rx_port_stats) +
5038 				       sizeof(struct rx_port_stats_ext) + 512);
5039 	if (!mz) {
5040 		mz = rte_memzone_reserve(mz_name, total_alloc_len,
5041 					 SOCKET_ID_ANY,
5042 					 RTE_MEMZONE_2MB |
5043 					 RTE_MEMZONE_SIZE_HINT_ONLY |
5044 					 RTE_MEMZONE_IOVA_CONTIG);
5045 		if (mz == NULL)
5046 			return -ENOMEM;
5047 	}
5048 	memset(mz->addr, 0, mz->len);
5049 	mz_phys_addr = mz->iova;
5050 
5051 	bp->rx_mem_zone = (const void *)mz;
5052 	bp->hw_rx_port_stats = mz->addr;
5053 	bp->hw_rx_port_stats_map = mz_phys_addr;
5054 
5055 	snprintf(mz_name, RTE_MEMZONE_NAMESIZE,
5056 		 "bnxt_" PCI_PRI_FMT "-%s", pci_dev->addr.domain,
5057 		 pci_dev->addr.bus, pci_dev->addr.devid,
5058 		 pci_dev->addr.function, "tx_port_stats");
5059 	mz_name[RTE_MEMZONE_NAMESIZE - 1] = 0;
5060 	mz = rte_memzone_lookup(mz_name);
5061 	total_alloc_len =
5062 		RTE_CACHE_LINE_ROUNDUP(sizeof(struct tx_port_stats) +
5063 				       sizeof(struct tx_port_stats_ext) + 512);
5064 	if (!mz) {
5065 		mz = rte_memzone_reserve(mz_name,
5066 					 total_alloc_len,
5067 					 SOCKET_ID_ANY,
5068 					 RTE_MEMZONE_2MB |
5069 					 RTE_MEMZONE_SIZE_HINT_ONLY |
5070 					 RTE_MEMZONE_IOVA_CONTIG);
5071 		if (mz == NULL)
5072 			return -ENOMEM;
5073 	}
5074 	memset(mz->addr, 0, mz->len);
5075 	mz_phys_addr = mz->iova;
5076 
5077 	bp->tx_mem_zone = (const void *)mz;
5078 	bp->hw_tx_port_stats = mz->addr;
5079 	bp->hw_tx_port_stats_map = mz_phys_addr;
5080 	bp->flags |= BNXT_FLAG_PORT_STATS;
5081 
5082 	/* Display extended statistics if FW supports it */
5083 	if (bp->hwrm_spec_code < HWRM_SPEC_CODE_1_8_4 ||
5084 	    bp->hwrm_spec_code == HWRM_SPEC_CODE_1_9_0 ||
5085 	    !(bp->flags & BNXT_FLAG_EXT_STATS_SUPPORTED))
5086 		return 0;
5087 
5088 	bp->hw_rx_port_stats_ext = (void *)
5089 		((uint8_t *)bp->hw_rx_port_stats +
5090 		 sizeof(struct rx_port_stats));
5091 	bp->hw_rx_port_stats_ext_map = bp->hw_rx_port_stats_map +
5092 		sizeof(struct rx_port_stats);
5093 	bp->flags |= BNXT_FLAG_EXT_RX_PORT_STATS;
5094 
5095 	if (bp->hwrm_spec_code < HWRM_SPEC_CODE_1_9_2 ||
5096 	    bp->flags & BNXT_FLAG_EXT_STATS_SUPPORTED) {
5097 		bp->hw_tx_port_stats_ext = (void *)
5098 			((uint8_t *)bp->hw_tx_port_stats +
5099 			 sizeof(struct tx_port_stats));
5100 		bp->hw_tx_port_stats_ext_map =
5101 			bp->hw_tx_port_stats_map +
5102 			sizeof(struct tx_port_stats);
5103 		bp->flags |= BNXT_FLAG_EXT_TX_PORT_STATS;
5104 	}
5105 
5106 	return 0;
5107 }
5108 
5109 static int bnxt_setup_mac_addr(struct rte_eth_dev *eth_dev)
5110 {
5111 	struct bnxt *bp = eth_dev->data->dev_private;
5112 	int rc = 0;
5113 
5114 	eth_dev->data->mac_addrs = rte_zmalloc("bnxt_mac_addr_tbl",
5115 					       RTE_ETHER_ADDR_LEN *
5116 					       bp->max_l2_ctx,
5117 					       0);
5118 	if (eth_dev->data->mac_addrs == NULL) {
5119 		PMD_DRV_LOG(ERR, "Failed to alloc MAC addr tbl\n");
5120 		return -ENOMEM;
5121 	}
5122 
5123 	if (!BNXT_HAS_DFLT_MAC_SET(bp)) {
5124 		if (BNXT_PF(bp))
5125 			return -EINVAL;
5126 
5127 		/* Generate a random MAC address, if none was assigned by PF */
5128 		PMD_DRV_LOG(INFO, "VF MAC address not assigned by Host PF\n");
5129 		bnxt_eth_hw_addr_random(bp->mac_addr);
5130 		PMD_DRV_LOG(INFO,
5131 			    "Assign random MAC:%02X:%02X:%02X:%02X:%02X:%02X\n",
5132 			    bp->mac_addr[0], bp->mac_addr[1], bp->mac_addr[2],
5133 			    bp->mac_addr[3], bp->mac_addr[4], bp->mac_addr[5]);
5134 
5135 		rc = bnxt_hwrm_set_mac(bp);
5136 		if (rc)
5137 			return rc;
5138 	}
5139 
5140 	/* Copy the permanent MAC from the FUNC_QCAPS response */
5141 	memcpy(&eth_dev->data->mac_addrs[0], bp->mac_addr, RTE_ETHER_ADDR_LEN);
5142 
5143 	return rc;
5144 }
5145 
5146 static int bnxt_restore_dflt_mac(struct bnxt *bp)
5147 {
5148 	int rc = 0;
5149 
5150 	/* MAC is already configured in FW */
5151 	if (BNXT_HAS_DFLT_MAC_SET(bp))
5152 		return 0;
5153 
5154 	/* Restore the old MAC configured */
5155 	rc = bnxt_hwrm_set_mac(bp);
5156 	if (rc)
5157 		PMD_DRV_LOG(ERR, "Failed to restore MAC address\n");
5158 
5159 	return rc;
5160 }
5161 
5162 static void bnxt_config_vf_req_fwd(struct bnxt *bp)
5163 {
5164 	if (!BNXT_PF(bp))
5165 		return;
5166 
5167 #define ALLOW_FUNC(x)	\
5168 	{ \
5169 		uint32_t arg = (x); \
5170 		bp->pf->vf_req_fwd[((arg) >> 5)] &= \
5171 		~rte_cpu_to_le_32(1 << ((arg) & 0x1f)); \
5172 	}
5173 
5174 	/* Forward all requests if firmware is new enough */
5175 	if (((bp->fw_ver >= ((20 << 24) | (6 << 16) | (100 << 8))) &&
5176 	     (bp->fw_ver < ((20 << 24) | (7 << 16)))) ||
5177 	    ((bp->fw_ver >= ((20 << 24) | (8 << 16))))) {
5178 		memset(bp->pf->vf_req_fwd, 0xff, sizeof(bp->pf->vf_req_fwd));
5179 	} else {
5180 		PMD_DRV_LOG(WARNING,
5181 			    "Firmware too old for VF mailbox functionality\n");
5182 		memset(bp->pf->vf_req_fwd, 0, sizeof(bp->pf->vf_req_fwd));
5183 	}
5184 
5185 	/*
5186 	 * The following are used for driver cleanup. If we disallow these,
5187 	 * VF drivers can't clean up cleanly.
5188 	 */
5189 	ALLOW_FUNC(HWRM_FUNC_DRV_UNRGTR);
5190 	ALLOW_FUNC(HWRM_VNIC_FREE);
5191 	ALLOW_FUNC(HWRM_RING_FREE);
5192 	ALLOW_FUNC(HWRM_RING_GRP_FREE);
5193 	ALLOW_FUNC(HWRM_VNIC_RSS_COS_LB_CTX_FREE);
5194 	ALLOW_FUNC(HWRM_CFA_L2_FILTER_FREE);
5195 	ALLOW_FUNC(HWRM_STAT_CTX_FREE);
5196 	ALLOW_FUNC(HWRM_PORT_PHY_QCFG);
5197 	ALLOW_FUNC(HWRM_VNIC_TPA_CFG);
5198 }
5199 
5200 uint16_t
5201 bnxt_get_svif(uint16_t port_id, bool func_svif,
5202 	      enum bnxt_ulp_intf_type type)
5203 {
5204 	struct rte_eth_dev *eth_dev;
5205 	struct bnxt *bp;
5206 
5207 	eth_dev = &rte_eth_devices[port_id];
5208 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
5209 		struct bnxt_vf_representor *vfr = eth_dev->data->dev_private;
5210 		if (!vfr)
5211 			return 0;
5212 
5213 		if (type == BNXT_ULP_INTF_TYPE_VF_REP)
5214 			return vfr->svif;
5215 
5216 		eth_dev = vfr->parent_dev;
5217 	}
5218 
5219 	bp = eth_dev->data->dev_private;
5220 
5221 	return func_svif ? bp->func_svif : bp->port_svif;
5222 }
5223 
5224 uint16_t
5225 bnxt_get_vnic_id(uint16_t port, enum bnxt_ulp_intf_type type)
5226 {
5227 	struct rte_eth_dev *eth_dev;
5228 	struct bnxt_vnic_info *vnic;
5229 	struct bnxt *bp;
5230 
5231 	eth_dev = &rte_eth_devices[port];
5232 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
5233 		struct bnxt_vf_representor *vfr = eth_dev->data->dev_private;
5234 		if (!vfr)
5235 			return 0;
5236 
5237 		if (type == BNXT_ULP_INTF_TYPE_VF_REP)
5238 			return vfr->dflt_vnic_id;
5239 
5240 		eth_dev = vfr->parent_dev;
5241 	}
5242 
5243 	bp = eth_dev->data->dev_private;
5244 
5245 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
5246 
5247 	return vnic->fw_vnic_id;
5248 }
5249 
5250 uint16_t
5251 bnxt_get_fw_func_id(uint16_t port, enum bnxt_ulp_intf_type type)
5252 {
5253 	struct rte_eth_dev *eth_dev;
5254 	struct bnxt *bp;
5255 
5256 	eth_dev = &rte_eth_devices[port];
5257 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
5258 		struct bnxt_vf_representor *vfr = eth_dev->data->dev_private;
5259 		if (!vfr)
5260 			return 0;
5261 
5262 		if (type == BNXT_ULP_INTF_TYPE_VF_REP)
5263 			return vfr->fw_fid;
5264 
5265 		eth_dev = vfr->parent_dev;
5266 	}
5267 
5268 	bp = eth_dev->data->dev_private;
5269 
5270 	return bp->fw_fid;
5271 }
5272 
5273 enum bnxt_ulp_intf_type
5274 bnxt_get_interface_type(uint16_t port)
5275 {
5276 	struct rte_eth_dev *eth_dev;
5277 	struct bnxt *bp;
5278 
5279 	eth_dev = &rte_eth_devices[port];
5280 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev))
5281 		return BNXT_ULP_INTF_TYPE_VF_REP;
5282 
5283 	bp = eth_dev->data->dev_private;
5284 	if (BNXT_PF(bp))
5285 		return BNXT_ULP_INTF_TYPE_PF;
5286 	else if (BNXT_VF_IS_TRUSTED(bp))
5287 		return BNXT_ULP_INTF_TYPE_TRUSTED_VF;
5288 	else if (BNXT_VF(bp))
5289 		return BNXT_ULP_INTF_TYPE_VF;
5290 
5291 	return BNXT_ULP_INTF_TYPE_INVALID;
5292 }
5293 
5294 uint16_t
5295 bnxt_get_phy_port_id(uint16_t port_id)
5296 {
5297 	struct bnxt_vf_representor *vfr;
5298 	struct rte_eth_dev *eth_dev;
5299 	struct bnxt *bp;
5300 
5301 	eth_dev = &rte_eth_devices[port_id];
5302 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
5303 		vfr = eth_dev->data->dev_private;
5304 		if (!vfr)
5305 			return 0;
5306 
5307 		eth_dev = vfr->parent_dev;
5308 	}
5309 
5310 	bp = eth_dev->data->dev_private;
5311 
5312 	return BNXT_PF(bp) ? bp->pf->port_id : bp->parent->port_id;
5313 }
5314 
5315 uint16_t
5316 bnxt_get_parif(uint16_t port_id, enum bnxt_ulp_intf_type type)
5317 {
5318 	struct rte_eth_dev *eth_dev;
5319 	struct bnxt *bp;
5320 
5321 	eth_dev = &rte_eth_devices[port_id];
5322 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
5323 		struct bnxt_vf_representor *vfr = eth_dev->data->dev_private;
5324 		if (!vfr)
5325 			return 0;
5326 
5327 		if (type == BNXT_ULP_INTF_TYPE_VF_REP)
5328 			return vfr->fw_fid - 1;
5329 
5330 		eth_dev = vfr->parent_dev;
5331 	}
5332 
5333 	bp = eth_dev->data->dev_private;
5334 
5335 	return BNXT_PF(bp) ? bp->fw_fid - 1 : bp->parent->fid - 1;
5336 }
5337 
5338 uint16_t
5339 bnxt_get_vport(uint16_t port_id)
5340 {
5341 	return (1 << bnxt_get_phy_port_id(port_id));
5342 }
5343 
5344 static void bnxt_alloc_error_recovery_info(struct bnxt *bp)
5345 {
5346 	struct bnxt_error_recovery_info *info = bp->recovery_info;
5347 
5348 	if (info) {
5349 		if (!(bp->fw_cap & BNXT_FW_CAP_HCOMM_FW_STATUS))
5350 			memset(info, 0, sizeof(*info));
5351 		return;
5352 	}
5353 
5354 	if (!(bp->fw_cap & BNXT_FW_CAP_ERROR_RECOVERY))
5355 		return;
5356 
5357 	info = rte_zmalloc("bnxt_hwrm_error_recovery_qcfg",
5358 			   sizeof(*info), 0);
5359 	if (!info)
5360 		bp->fw_cap &= ~BNXT_FW_CAP_ERROR_RECOVERY;
5361 
5362 	bp->recovery_info = info;
5363 }
5364 
5365 static void bnxt_check_fw_status(struct bnxt *bp)
5366 {
5367 	uint32_t fw_status;
5368 
5369 	if (!(bp->recovery_info &&
5370 	      (bp->fw_cap & BNXT_FW_CAP_HCOMM_FW_STATUS)))
5371 		return;
5372 
5373 	fw_status = bnxt_read_fw_status_reg(bp, BNXT_FW_STATUS_REG);
5374 	if (fw_status != BNXT_FW_STATUS_HEALTHY)
5375 		PMD_DRV_LOG(ERR, "Firmware not responding, status: %#x\n",
5376 			    fw_status);
5377 }
5378 
5379 static int bnxt_map_hcomm_fw_status_reg(struct bnxt *bp)
5380 {
5381 	struct bnxt_error_recovery_info *info = bp->recovery_info;
5382 	uint32_t status_loc;
5383 	uint32_t sig_ver;
5384 
5385 	rte_write32(HCOMM_STATUS_STRUCT_LOC, (uint8_t *)bp->bar0 +
5386 		    BNXT_GRCPF_REG_WINDOW_BASE_OUT + 4);
5387 	sig_ver = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
5388 				   BNXT_GRCP_WINDOW_2_BASE +
5389 				   offsetof(struct hcomm_status,
5390 					    sig_ver)));
5391 	/* If the signature is absent, then FW does not support this feature */
5392 	if ((sig_ver & HCOMM_STATUS_SIGNATURE_MASK) !=
5393 	    HCOMM_STATUS_SIGNATURE_VAL)
5394 		return 0;
5395 
5396 	if (!info) {
5397 		info = rte_zmalloc("bnxt_hwrm_error_recovery_qcfg",
5398 				   sizeof(*info), 0);
5399 		if (!info)
5400 			return -ENOMEM;
5401 		bp->recovery_info = info;
5402 	} else {
5403 		memset(info, 0, sizeof(*info));
5404 	}
5405 
5406 	status_loc = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
5407 				      BNXT_GRCP_WINDOW_2_BASE +
5408 				      offsetof(struct hcomm_status,
5409 					       fw_status_loc)));
5410 
5411 	/* Only pre-map the FW health status GRC register */
5412 	if (BNXT_FW_STATUS_REG_TYPE(status_loc) != BNXT_FW_STATUS_REG_TYPE_GRC)
5413 		return 0;
5414 
5415 	info->status_regs[BNXT_FW_STATUS_REG] = status_loc;
5416 	info->mapped_status_regs[BNXT_FW_STATUS_REG] =
5417 		BNXT_GRCP_WINDOW_2_BASE + (status_loc & BNXT_GRCP_OFFSET_MASK);
5418 
5419 	rte_write32((status_loc & BNXT_GRCP_BASE_MASK), (uint8_t *)bp->bar0 +
5420 		    BNXT_GRCPF_REG_WINDOW_BASE_OUT + 4);
5421 
5422 	bp->fw_cap |= BNXT_FW_CAP_HCOMM_FW_STATUS;
5423 
5424 	return 0;
5425 }
5426 
5427 static int bnxt_init_fw(struct bnxt *bp)
5428 {
5429 	uint16_t mtu;
5430 	int rc = 0;
5431 
5432 	bp->fw_cap = 0;
5433 
5434 	rc = bnxt_map_hcomm_fw_status_reg(bp);
5435 	if (rc)
5436 		return rc;
5437 
5438 	rc = bnxt_hwrm_ver_get(bp, DFLT_HWRM_CMD_TIMEOUT);
5439 	if (rc) {
5440 		bnxt_check_fw_status(bp);
5441 		return rc;
5442 	}
5443 
5444 	rc = bnxt_hwrm_func_reset(bp);
5445 	if (rc)
5446 		return -EIO;
5447 
5448 	rc = bnxt_hwrm_vnic_qcaps(bp);
5449 	if (rc)
5450 		return rc;
5451 
5452 	rc = bnxt_hwrm_queue_qportcfg(bp);
5453 	if (rc)
5454 		return rc;
5455 
5456 	/* Get the MAX capabilities for this function.
5457 	 * This function also allocates context memory for TQM rings and
5458 	 * informs the firmware about this allocated backing store memory.
5459 	 */
5460 	rc = bnxt_hwrm_func_qcaps(bp);
5461 	if (rc)
5462 		return rc;
5463 
5464 	rc = bnxt_hwrm_func_qcfg(bp, &mtu);
5465 	if (rc)
5466 		return rc;
5467 
5468 	bnxt_hwrm_port_mac_qcfg(bp);
5469 
5470 	bnxt_hwrm_parent_pf_qcfg(bp);
5471 
5472 	bnxt_hwrm_port_phy_qcaps(bp);
5473 
5474 	bnxt_alloc_error_recovery_info(bp);
5475 	/* Get the adapter error recovery support info */
5476 	rc = bnxt_hwrm_error_recovery_qcfg(bp);
5477 	if (rc)
5478 		bp->fw_cap &= ~BNXT_FW_CAP_ERROR_RECOVERY;
5479 
5480 	bnxt_hwrm_port_led_qcaps(bp);
5481 
5482 	return 0;
5483 }
5484 
5485 static int
5486 bnxt_init_locks(struct bnxt *bp)
5487 {
5488 	int err;
5489 
5490 	err = pthread_mutex_init(&bp->flow_lock, NULL);
5491 	if (err) {
5492 		PMD_DRV_LOG(ERR, "Unable to initialize flow_lock\n");
5493 		return err;
5494 	}
5495 
5496 	err = pthread_mutex_init(&bp->def_cp_lock, NULL);
5497 	if (err)
5498 		PMD_DRV_LOG(ERR, "Unable to initialize def_cp_lock\n");
5499 	return err;
5500 }
5501 
5502 static int bnxt_init_resources(struct bnxt *bp, bool reconfig_dev)
5503 {
5504 	int rc = 0;
5505 
5506 	rc = bnxt_init_fw(bp);
5507 	if (rc)
5508 		return rc;
5509 
5510 	if (!reconfig_dev) {
5511 		rc = bnxt_setup_mac_addr(bp->eth_dev);
5512 		if (rc)
5513 			return rc;
5514 	} else {
5515 		rc = bnxt_restore_dflt_mac(bp);
5516 		if (rc)
5517 			return rc;
5518 	}
5519 
5520 	bnxt_config_vf_req_fwd(bp);
5521 
5522 	rc = bnxt_hwrm_func_driver_register(bp);
5523 	if (rc) {
5524 		PMD_DRV_LOG(ERR, "Failed to register driver");
5525 		return -EBUSY;
5526 	}
5527 
5528 	if (BNXT_PF(bp)) {
5529 		if (bp->pdev->max_vfs) {
5530 			rc = bnxt_hwrm_allocate_vfs(bp, bp->pdev->max_vfs);
5531 			if (rc) {
5532 				PMD_DRV_LOG(ERR, "Failed to allocate VFs\n");
5533 				return rc;
5534 			}
5535 		} else {
5536 			rc = bnxt_hwrm_allocate_pf_only(bp);
5537 			if (rc) {
5538 				PMD_DRV_LOG(ERR,
5539 					    "Failed to allocate PF resources");
5540 				return rc;
5541 			}
5542 		}
5543 	}
5544 
5545 	rc = bnxt_alloc_mem(bp, reconfig_dev);
5546 	if (rc)
5547 		return rc;
5548 
5549 	rc = bnxt_setup_int(bp);
5550 	if (rc)
5551 		return rc;
5552 
5553 	rc = bnxt_request_int(bp);
5554 	if (rc)
5555 		return rc;
5556 
5557 	rc = bnxt_init_ctx_mem(bp);
5558 	if (rc) {
5559 		PMD_DRV_LOG(ERR, "Failed to init adv_flow_counters\n");
5560 		return rc;
5561 	}
5562 
5563 	rc = bnxt_init_locks(bp);
5564 	if (rc)
5565 		return rc;
5566 
5567 	return 0;
5568 }
5569 
5570 static int
5571 bnxt_parse_devarg_truflow(__rte_unused const char *key,
5572 			  const char *value, void *opaque_arg)
5573 {
5574 	struct bnxt *bp = opaque_arg;
5575 	unsigned long truflow;
5576 	char *end = NULL;
5577 
5578 	if (!value || !opaque_arg) {
5579 		PMD_DRV_LOG(ERR,
5580 			    "Invalid parameter passed to truflow devargs.\n");
5581 		return -EINVAL;
5582 	}
5583 
5584 	truflow = strtoul(value, &end, 10);
5585 	if (end == NULL || *end != '\0' ||
5586 	    (truflow == ULONG_MAX && errno == ERANGE)) {
5587 		PMD_DRV_LOG(ERR,
5588 			    "Invalid parameter passed to truflow devargs.\n");
5589 		return -EINVAL;
5590 	}
5591 
5592 	if (BNXT_DEVARG_TRUFLOW_INVALID(truflow)) {
5593 		PMD_DRV_LOG(ERR,
5594 			    "Invalid value passed to truflow devargs.\n");
5595 		return -EINVAL;
5596 	}
5597 
5598 	bp->flags |= BNXT_FLAG_TRUFLOW_EN;
5599 	if (BNXT_TRUFLOW_EN(bp))
5600 		PMD_DRV_LOG(INFO, "Host-based truflow feature enabled.\n");
5601 
5602 	return 0;
5603 }
5604 
5605 static int
5606 bnxt_parse_devarg_flow_xstat(__rte_unused const char *key,
5607 			     const char *value, void *opaque_arg)
5608 {
5609 	struct bnxt *bp = opaque_arg;
5610 	unsigned long flow_xstat;
5611 	char *end = NULL;
5612 
5613 	if (!value || !opaque_arg) {
5614 		PMD_DRV_LOG(ERR,
5615 			    "Invalid parameter passed to flow_xstat devarg.\n");
5616 		return -EINVAL;
5617 	}
5618 
5619 	flow_xstat = strtoul(value, &end, 10);
5620 	if (end == NULL || *end != '\0' ||
5621 	    (flow_xstat == ULONG_MAX && errno == ERANGE)) {
5622 		PMD_DRV_LOG(ERR,
5623 			    "Invalid parameter passed to flow_xstat devarg.\n");
5624 		return -EINVAL;
5625 	}
5626 
5627 	if (BNXT_DEVARG_FLOW_XSTAT_INVALID(flow_xstat)) {
5628 		PMD_DRV_LOG(ERR,
5629 			    "Invalid value passed to flow_xstat devarg.\n");
5630 		return -EINVAL;
5631 	}
5632 
5633 	bp->flags |= BNXT_FLAG_FLOW_XSTATS_EN;
5634 	if (BNXT_FLOW_XSTATS_EN(bp))
5635 		PMD_DRV_LOG(INFO, "flow_xstat feature enabled.\n");
5636 
5637 	return 0;
5638 }
5639 
5640 static int
5641 bnxt_parse_devarg_max_num_kflows(__rte_unused const char *key,
5642 					const char *value, void *opaque_arg)
5643 {
5644 	struct bnxt *bp = opaque_arg;
5645 	unsigned long max_num_kflows;
5646 	char *end = NULL;
5647 
5648 	if (!value || !opaque_arg) {
5649 		PMD_DRV_LOG(ERR,
5650 			"Invalid parameter passed to max_num_kflows devarg.\n");
5651 		return -EINVAL;
5652 	}
5653 
5654 	max_num_kflows = strtoul(value, &end, 10);
5655 	if (end == NULL || *end != '\0' ||
5656 		(max_num_kflows == ULONG_MAX && errno == ERANGE)) {
5657 		PMD_DRV_LOG(ERR,
5658 			"Invalid parameter passed to max_num_kflows devarg.\n");
5659 		return -EINVAL;
5660 	}
5661 
5662 	if (bnxt_devarg_max_num_kflow_invalid(max_num_kflows)) {
5663 		PMD_DRV_LOG(ERR,
5664 			"Invalid value passed to max_num_kflows devarg.\n");
5665 		return -EINVAL;
5666 	}
5667 
5668 	bp->max_num_kflows = max_num_kflows;
5669 	if (bp->max_num_kflows)
5670 		PMD_DRV_LOG(INFO, "max_num_kflows set as %ldK.\n",
5671 				max_num_kflows);
5672 
5673 	return 0;
5674 }
5675 
5676 static void
5677 bnxt_parse_dev_args(struct bnxt *bp, struct rte_devargs *devargs)
5678 {
5679 	struct rte_kvargs *kvlist;
5680 
5681 	if (devargs == NULL)
5682 		return;
5683 
5684 	kvlist = rte_kvargs_parse(devargs->args, bnxt_dev_args);
5685 	if (kvlist == NULL)
5686 		return;
5687 
5688 	/*
5689 	 * Handler for "truflow" devarg.
5690 	 * Invoked as for ex: "-w 0000:00:0d.0,host-based-truflow=1"
5691 	 */
5692 	rte_kvargs_process(kvlist, BNXT_DEVARG_TRUFLOW,
5693 			   bnxt_parse_devarg_truflow, bp);
5694 
5695 	/*
5696 	 * Handler for "flow_xstat" devarg.
5697 	 * Invoked as for ex: "-w 0000:00:0d.0,flow_xstat=1"
5698 	 */
5699 	rte_kvargs_process(kvlist, BNXT_DEVARG_FLOW_XSTAT,
5700 			   bnxt_parse_devarg_flow_xstat, bp);
5701 
5702 	/*
5703 	 * Handler for "max_num_kflows" devarg.
5704 	 * Invoked as for ex: "-w 000:00:0d.0,max_num_kflows=32"
5705 	 */
5706 	rte_kvargs_process(kvlist, BNXT_DEVARG_MAX_NUM_KFLOWS,
5707 			   bnxt_parse_devarg_max_num_kflows, bp);
5708 
5709 	rte_kvargs_free(kvlist);
5710 }
5711 
5712 static int bnxt_alloc_switch_domain(struct bnxt *bp)
5713 {
5714 	int rc = 0;
5715 
5716 	if (BNXT_PF(bp) || BNXT_VF_IS_TRUSTED(bp)) {
5717 		rc = rte_eth_switch_domain_alloc(&bp->switch_domain_id);
5718 		if (rc)
5719 			PMD_DRV_LOG(ERR,
5720 				    "Failed to alloc switch domain: %d\n", rc);
5721 		else
5722 			PMD_DRV_LOG(INFO,
5723 				    "Switch domain allocated %d\n",
5724 				    bp->switch_domain_id);
5725 	}
5726 
5727 	return rc;
5728 }
5729 
5730 static int
5731 bnxt_dev_init(struct rte_eth_dev *eth_dev, void *params __rte_unused)
5732 {
5733 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
5734 	static int version_printed;
5735 	struct bnxt *bp;
5736 	int rc;
5737 
5738 	if (version_printed++ == 0)
5739 		PMD_DRV_LOG(INFO, "%s\n", bnxt_version);
5740 
5741 	eth_dev->dev_ops = &bnxt_dev_ops;
5742 	eth_dev->rx_pkt_burst = &bnxt_recv_pkts;
5743 	eth_dev->tx_pkt_burst = &bnxt_xmit_pkts;
5744 
5745 	/*
5746 	 * For secondary processes, we don't initialise any further
5747 	 * as primary has already done this work.
5748 	 */
5749 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
5750 		return 0;
5751 
5752 	rte_eth_copy_pci_info(eth_dev, pci_dev);
5753 
5754 	bp = eth_dev->data->dev_private;
5755 
5756 	/* Parse dev arguments passed on when starting the DPDK application. */
5757 	bnxt_parse_dev_args(bp, pci_dev->device.devargs);
5758 
5759 	bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
5760 
5761 	if (bnxt_vf_pciid(pci_dev->id.device_id))
5762 		bp->flags |= BNXT_FLAG_VF;
5763 
5764 	if (bnxt_thor_device(pci_dev->id.device_id))
5765 		bp->flags |= BNXT_FLAG_THOR_CHIP;
5766 
5767 	if (pci_dev->id.device_id == BROADCOM_DEV_ID_58802 ||
5768 	    pci_dev->id.device_id == BROADCOM_DEV_ID_58804 ||
5769 	    pci_dev->id.device_id == BROADCOM_DEV_ID_58808 ||
5770 	    pci_dev->id.device_id == BROADCOM_DEV_ID_58802_VF)
5771 		bp->flags |= BNXT_FLAG_STINGRAY;
5772 
5773 	rc = bnxt_init_board(eth_dev);
5774 	if (rc) {
5775 		PMD_DRV_LOG(ERR,
5776 			    "Failed to initialize board rc: %x\n", rc);
5777 		return rc;
5778 	}
5779 
5780 	rc = bnxt_alloc_pf_info(bp);
5781 	if (rc)
5782 		goto error_free;
5783 
5784 	rc = bnxt_alloc_link_info(bp);
5785 	if (rc)
5786 		goto error_free;
5787 
5788 	rc = bnxt_alloc_parent_info(bp);
5789 	if (rc)
5790 		goto error_free;
5791 
5792 	rc = bnxt_alloc_hwrm_resources(bp);
5793 	if (rc) {
5794 		PMD_DRV_LOG(ERR,
5795 			    "Failed to allocate hwrm resource rc: %x\n", rc);
5796 		goto error_free;
5797 	}
5798 	rc = bnxt_alloc_leds_info(bp);
5799 	if (rc)
5800 		goto error_free;
5801 
5802 	rc = bnxt_alloc_cos_queues(bp);
5803 	if (rc)
5804 		goto error_free;
5805 
5806 	rc = bnxt_init_resources(bp, false);
5807 	if (rc)
5808 		goto error_free;
5809 
5810 	rc = bnxt_alloc_stats_mem(bp);
5811 	if (rc)
5812 		goto error_free;
5813 
5814 	bnxt_alloc_switch_domain(bp);
5815 
5816 	/* Pass the information to the rte_eth_dev_close() that it should also
5817 	 * release the private port resources.
5818 	 */
5819 	eth_dev->data->dev_flags |= RTE_ETH_DEV_CLOSE_REMOVE;
5820 
5821 	PMD_DRV_LOG(INFO,
5822 		    DRV_MODULE_NAME "found at mem %" PRIX64 ", node addr %pM\n",
5823 		    pci_dev->mem_resource[0].phys_addr,
5824 		    pci_dev->mem_resource[0].addr);
5825 
5826 	return 0;
5827 
5828 error_free:
5829 	bnxt_dev_uninit(eth_dev);
5830 	return rc;
5831 }
5832 
5833 
5834 static void bnxt_free_ctx_mem_buf(struct bnxt_ctx_mem_buf_info *ctx)
5835 {
5836 	if (!ctx)
5837 		return;
5838 
5839 	if (ctx->va)
5840 		rte_free(ctx->va);
5841 
5842 	ctx->va = NULL;
5843 	ctx->dma = RTE_BAD_IOVA;
5844 	ctx->ctx_id = BNXT_CTX_VAL_INVAL;
5845 }
5846 
5847 static void bnxt_unregister_fc_ctx_mem(struct bnxt *bp)
5848 {
5849 	bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_RX,
5850 				  CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
5851 				  bp->flow_stat->rx_fc_out_tbl.ctx_id,
5852 				  bp->flow_stat->max_fc,
5853 				  false);
5854 
5855 	bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_TX,
5856 				  CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
5857 				  bp->flow_stat->tx_fc_out_tbl.ctx_id,
5858 				  bp->flow_stat->max_fc,
5859 				  false);
5860 
5861 	if (bp->flow_stat->rx_fc_in_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
5862 		bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->rx_fc_in_tbl.ctx_id);
5863 	bp->flow_stat->rx_fc_in_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
5864 
5865 	if (bp->flow_stat->rx_fc_out_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
5866 		bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->rx_fc_out_tbl.ctx_id);
5867 	bp->flow_stat->rx_fc_out_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
5868 
5869 	if (bp->flow_stat->tx_fc_in_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
5870 		bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->tx_fc_in_tbl.ctx_id);
5871 	bp->flow_stat->tx_fc_in_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
5872 
5873 	if (bp->flow_stat->tx_fc_out_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
5874 		bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->tx_fc_out_tbl.ctx_id);
5875 	bp->flow_stat->tx_fc_out_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
5876 }
5877 
5878 static void bnxt_uninit_fc_ctx_mem(struct bnxt *bp)
5879 {
5880 	bnxt_unregister_fc_ctx_mem(bp);
5881 
5882 	bnxt_free_ctx_mem_buf(&bp->flow_stat->rx_fc_in_tbl);
5883 	bnxt_free_ctx_mem_buf(&bp->flow_stat->rx_fc_out_tbl);
5884 	bnxt_free_ctx_mem_buf(&bp->flow_stat->tx_fc_in_tbl);
5885 	bnxt_free_ctx_mem_buf(&bp->flow_stat->tx_fc_out_tbl);
5886 }
5887 
5888 static void bnxt_uninit_ctx_mem(struct bnxt *bp)
5889 {
5890 	if (BNXT_FLOW_XSTATS_EN(bp))
5891 		bnxt_uninit_fc_ctx_mem(bp);
5892 }
5893 
5894 static void
5895 bnxt_free_error_recovery_info(struct bnxt *bp)
5896 {
5897 	rte_free(bp->recovery_info);
5898 	bp->recovery_info = NULL;
5899 	bp->fw_cap &= ~BNXT_FW_CAP_ERROR_RECOVERY;
5900 }
5901 
5902 static void
5903 bnxt_uninit_locks(struct bnxt *bp)
5904 {
5905 	pthread_mutex_destroy(&bp->flow_lock);
5906 	pthread_mutex_destroy(&bp->def_cp_lock);
5907 	if (bp->rep_info)
5908 		pthread_mutex_destroy(&bp->rep_info->vfr_lock);
5909 }
5910 
5911 static int
5912 bnxt_uninit_resources(struct bnxt *bp, bool reconfig_dev)
5913 {
5914 	int rc;
5915 
5916 	bnxt_free_int(bp);
5917 	bnxt_free_mem(bp, reconfig_dev);
5918 	bnxt_hwrm_func_buf_unrgtr(bp);
5919 	rc = bnxt_hwrm_func_driver_unregister(bp, 0);
5920 	bp->flags &= ~BNXT_FLAG_REGISTERED;
5921 	bnxt_free_ctx_mem(bp);
5922 	if (!reconfig_dev) {
5923 		bnxt_free_hwrm_resources(bp);
5924 		bnxt_free_error_recovery_info(bp);
5925 	}
5926 
5927 	bnxt_uninit_ctx_mem(bp);
5928 
5929 	bnxt_uninit_locks(bp);
5930 	bnxt_free_flow_stats_info(bp);
5931 	bnxt_free_rep_info(bp);
5932 	rte_free(bp->ptp_cfg);
5933 	bp->ptp_cfg = NULL;
5934 	return rc;
5935 }
5936 
5937 static int
5938 bnxt_dev_uninit(struct rte_eth_dev *eth_dev)
5939 {
5940 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
5941 		return -EPERM;
5942 
5943 	PMD_DRV_LOG(DEBUG, "Calling Device uninit\n");
5944 
5945 	if (eth_dev->state != RTE_ETH_DEV_UNUSED)
5946 		bnxt_dev_close_op(eth_dev);
5947 
5948 	return 0;
5949 }
5950 
5951 static int bnxt_pci_remove_dev_with_reps(struct rte_eth_dev *eth_dev)
5952 {
5953 	struct bnxt *bp = eth_dev->data->dev_private;
5954 	struct rte_eth_dev *vf_rep_eth_dev;
5955 	int ret = 0, i;
5956 
5957 	if (!bp)
5958 		return -EINVAL;
5959 
5960 	for (i = 0; i < bp->num_reps; i++) {
5961 		vf_rep_eth_dev = bp->rep_info[i].vfr_eth_dev;
5962 		if (!vf_rep_eth_dev)
5963 			continue;
5964 		rte_eth_dev_destroy(vf_rep_eth_dev, bnxt_vf_representor_uninit);
5965 	}
5966 	ret = rte_eth_dev_destroy(eth_dev, bnxt_dev_uninit);
5967 
5968 	return ret;
5969 }
5970 
5971 static void bnxt_free_rep_info(struct bnxt *bp)
5972 {
5973 	rte_free(bp->rep_info);
5974 	bp->rep_info = NULL;
5975 	rte_free(bp->cfa_code_map);
5976 	bp->cfa_code_map = NULL;
5977 }
5978 
5979 static int bnxt_init_rep_info(struct bnxt *bp)
5980 {
5981 	int i = 0, rc;
5982 
5983 	if (bp->rep_info)
5984 		return 0;
5985 
5986 	bp->rep_info = rte_zmalloc("bnxt_rep_info",
5987 				   sizeof(bp->rep_info[0]) * BNXT_MAX_VF_REPS,
5988 				   0);
5989 	if (!bp->rep_info) {
5990 		PMD_DRV_LOG(ERR, "Failed to alloc memory for rep info\n");
5991 		return -ENOMEM;
5992 	}
5993 	bp->cfa_code_map = rte_zmalloc("bnxt_cfa_code_map",
5994 				       sizeof(*bp->cfa_code_map) *
5995 				       BNXT_MAX_CFA_CODE, 0);
5996 	if (!bp->cfa_code_map) {
5997 		PMD_DRV_LOG(ERR, "Failed to alloc memory for cfa_code_map\n");
5998 		bnxt_free_rep_info(bp);
5999 		return -ENOMEM;
6000 	}
6001 
6002 	for (i = 0; i < BNXT_MAX_CFA_CODE; i++)
6003 		bp->cfa_code_map[i] = BNXT_VF_IDX_INVALID;
6004 
6005 	rc = pthread_mutex_init(&bp->rep_info->vfr_lock, NULL);
6006 	if (rc) {
6007 		PMD_DRV_LOG(ERR, "Unable to initialize vfr_lock\n");
6008 		bnxt_free_rep_info(bp);
6009 		return rc;
6010 	}
6011 	return rc;
6012 }
6013 
6014 static int bnxt_rep_port_probe(struct rte_pci_device *pci_dev,
6015 			       struct rte_eth_devargs eth_da,
6016 			       struct rte_eth_dev *backing_eth_dev)
6017 {
6018 	struct rte_eth_dev *vf_rep_eth_dev;
6019 	char name[RTE_ETH_NAME_MAX_LEN];
6020 	struct bnxt *backing_bp;
6021 	uint16_t num_rep;
6022 	int i, ret = 0;
6023 
6024 	num_rep = eth_da.nb_representor_ports;
6025 	if (num_rep > BNXT_MAX_VF_REPS) {
6026 		PMD_DRV_LOG(ERR, "nb_representor_ports = %d > %d MAX VF REPS\n",
6027 			    num_rep, BNXT_MAX_VF_REPS);
6028 		return -EINVAL;
6029 	}
6030 
6031 	if (num_rep > RTE_MAX_ETHPORTS) {
6032 		PMD_DRV_LOG(ERR,
6033 			    "nb_representor_ports = %d > %d MAX ETHPORTS\n",
6034 			    num_rep, RTE_MAX_ETHPORTS);
6035 		return -EINVAL;
6036 	}
6037 
6038 	backing_bp = backing_eth_dev->data->dev_private;
6039 
6040 	if (!(BNXT_PF(backing_bp) || BNXT_VF_IS_TRUSTED(backing_bp))) {
6041 		PMD_DRV_LOG(ERR,
6042 			    "Not a PF or trusted VF. No Representor support\n");
6043 		/* Returning an error is not an option.
6044 		 * Applications are not handling this correctly
6045 		 */
6046 		return 0;
6047 	}
6048 
6049 	if (bnxt_init_rep_info(backing_bp))
6050 		return 0;
6051 
6052 	for (i = 0; i < num_rep; i++) {
6053 		struct bnxt_vf_representor representor = {
6054 			.vf_id = eth_da.representor_ports[i],
6055 			.switch_domain_id = backing_bp->switch_domain_id,
6056 			.parent_dev = backing_eth_dev
6057 		};
6058 
6059 		if (representor.vf_id >= BNXT_MAX_VF_REPS) {
6060 			PMD_DRV_LOG(ERR, "VF-Rep id %d >= %d MAX VF ID\n",
6061 				    representor.vf_id, BNXT_MAX_VF_REPS);
6062 			continue;
6063 		}
6064 
6065 		/* representor port net_bdf_port */
6066 		snprintf(name, sizeof(name), "net_%s_representor_%d",
6067 			 pci_dev->device.name, eth_da.representor_ports[i]);
6068 
6069 		ret = rte_eth_dev_create(&pci_dev->device, name,
6070 					 sizeof(struct bnxt_vf_representor),
6071 					 NULL, NULL,
6072 					 bnxt_vf_representor_init,
6073 					 &representor);
6074 
6075 		if (!ret) {
6076 			vf_rep_eth_dev = rte_eth_dev_allocated(name);
6077 			if (!vf_rep_eth_dev) {
6078 				PMD_DRV_LOG(ERR, "Failed to find the eth_dev"
6079 					    " for VF-Rep: %s.", name);
6080 				bnxt_pci_remove_dev_with_reps(backing_eth_dev);
6081 				ret = -ENODEV;
6082 				return ret;
6083 			}
6084 			backing_bp->rep_info[representor.vf_id].vfr_eth_dev =
6085 				vf_rep_eth_dev;
6086 			backing_bp->num_reps++;
6087 		} else {
6088 			PMD_DRV_LOG(ERR, "failed to create bnxt vf "
6089 				    "representor %s.", name);
6090 			bnxt_pci_remove_dev_with_reps(backing_eth_dev);
6091 		}
6092 	}
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 
6133 	/* probe representor ports now */
6134 	ret = bnxt_rep_port_probe(pci_dev, eth_da, backing_eth_dev);
6135 
6136 	return ret;
6137 }
6138 
6139 static int bnxt_pci_remove(struct rte_pci_device *pci_dev)
6140 {
6141 	struct rte_eth_dev *eth_dev;
6142 
6143 	eth_dev = rte_eth_dev_allocated(pci_dev->device.name);
6144 	if (!eth_dev)
6145 		return 0; /* Invoked typically only by OVS-DPDK, by the
6146 			   * time it comes here the eth_dev is already
6147 			   * deleted by rte_eth_dev_close(), so returning
6148 			   * +ve value will at least help in proper cleanup
6149 			   */
6150 
6151 	if (rte_eal_process_type() == RTE_PROC_PRIMARY) {
6152 		if (eth_dev->data->dev_flags & RTE_ETH_DEV_REPRESENTOR)
6153 			return rte_eth_dev_destroy(eth_dev,
6154 						   bnxt_vf_representor_uninit);
6155 		else
6156 			return rte_eth_dev_destroy(eth_dev,
6157 						   bnxt_dev_uninit);
6158 	} else {
6159 		return rte_eth_dev_pci_generic_remove(pci_dev, NULL);
6160 	}
6161 }
6162 
6163 static struct rte_pci_driver bnxt_rte_pmd = {
6164 	.id_table = bnxt_pci_id_map,
6165 	.drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC |
6166 			RTE_PCI_DRV_PROBE_AGAIN, /* Needed in case of VF-REPs
6167 						  * and OVS-DPDK
6168 						  */
6169 	.probe = bnxt_pci_probe,
6170 	.remove = bnxt_pci_remove,
6171 };
6172 
6173 static bool
6174 is_device_supported(struct rte_eth_dev *dev, struct rte_pci_driver *drv)
6175 {
6176 	if (strcmp(dev->device->driver->name, drv->driver.name))
6177 		return false;
6178 
6179 	return true;
6180 }
6181 
6182 bool is_bnxt_supported(struct rte_eth_dev *dev)
6183 {
6184 	return is_device_supported(dev, &bnxt_rte_pmd);
6185 }
6186 
6187 RTE_LOG_REGISTER(bnxt_logtype_driver, pmd.net.bnxt.driver, NOTICE);
6188 RTE_PMD_REGISTER_PCI(net_bnxt, bnxt_rte_pmd);
6189 RTE_PMD_REGISTER_PCI_TABLE(net_bnxt, bnxt_pci_id_map);
6190 RTE_PMD_REGISTER_KMOD_DEP(net_bnxt, "* igb_uio | uio_pci_generic | vfio-pci");
6191