xref: /dpdk/drivers/net/bnxt/bnxt_ethdev.c (revision 3cc6ecfdfe85d2577fef30e1791bb7534e3d60b3)
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 int bnxt_dev_start_op(struct rte_eth_dev *eth_dev)
1172 {
1173 	struct bnxt *bp = eth_dev->data->dev_private;
1174 	uint64_t rx_offloads = eth_dev->data->dev_conf.rxmode.offloads;
1175 	int vlan_mask = 0;
1176 	int rc, retry_cnt = BNXT_IF_CHANGE_RETRY_COUNT;
1177 
1178 	if (!eth_dev->data->nb_tx_queues || !eth_dev->data->nb_rx_queues) {
1179 		PMD_DRV_LOG(ERR, "Queues are not configured yet!\n");
1180 		return -EINVAL;
1181 	}
1182 
1183 	if (bp->rx_cp_nr_rings > RTE_ETHDEV_QUEUE_STAT_CNTRS) {
1184 		PMD_DRV_LOG(ERR,
1185 			"RxQ cnt %d > CONFIG_RTE_ETHDEV_QUEUE_STAT_CNTRS %d\n",
1186 			bp->rx_cp_nr_rings, RTE_ETHDEV_QUEUE_STAT_CNTRS);
1187 	}
1188 
1189 	do {
1190 		rc = bnxt_hwrm_if_change(bp, true);
1191 		if (rc == 0 || rc != -EAGAIN)
1192 			break;
1193 
1194 		rte_delay_ms(BNXT_IF_CHANGE_RETRY_INTERVAL);
1195 	} while (retry_cnt--);
1196 
1197 	if (rc)
1198 		return rc;
1199 
1200 	if (bp->flags & BNXT_FLAG_IF_CHANGE_HOT_FW_RESET_DONE) {
1201 		rc = bnxt_handle_if_change_status(bp);
1202 		if (rc)
1203 			return rc;
1204 	}
1205 
1206 	bnxt_enable_int(bp);
1207 
1208 	rc = bnxt_init_chip(bp);
1209 	if (rc)
1210 		goto error;
1211 
1212 	eth_dev->data->scattered_rx = bnxt_scattered_rx(eth_dev);
1213 	eth_dev->data->dev_started = 1;
1214 
1215 	bnxt_link_update(eth_dev, 1, ETH_LINK_UP);
1216 
1217 	if (rx_offloads & DEV_RX_OFFLOAD_VLAN_FILTER)
1218 		vlan_mask |= ETH_VLAN_FILTER_MASK;
1219 	if (rx_offloads & DEV_RX_OFFLOAD_VLAN_STRIP)
1220 		vlan_mask |= ETH_VLAN_STRIP_MASK;
1221 	rc = bnxt_vlan_offload_set_op(eth_dev, vlan_mask);
1222 	if (rc)
1223 		goto error;
1224 
1225 	eth_dev->rx_pkt_burst = bnxt_receive_function(eth_dev);
1226 	eth_dev->tx_pkt_burst = bnxt_transmit_function(eth_dev);
1227 
1228 	pthread_mutex_lock(&bp->def_cp_lock);
1229 	bnxt_schedule_fw_health_check(bp);
1230 	pthread_mutex_unlock(&bp->def_cp_lock);
1231 
1232 	bnxt_ulp_init(bp);
1233 
1234 	return 0;
1235 
1236 error:
1237 	bnxt_shutdown_nic(bp);
1238 	bnxt_free_tx_mbufs(bp);
1239 	bnxt_free_rx_mbufs(bp);
1240 	bnxt_hwrm_if_change(bp, false);
1241 	eth_dev->data->dev_started = 0;
1242 	return rc;
1243 }
1244 
1245 static int bnxt_dev_set_link_up_op(struct rte_eth_dev *eth_dev)
1246 {
1247 	struct bnxt *bp = eth_dev->data->dev_private;
1248 	int rc = 0;
1249 
1250 	if (!bp->link_info->link_up)
1251 		rc = bnxt_set_hwrm_link_config(bp, true);
1252 	if (!rc)
1253 		eth_dev->data->dev_link.link_status = 1;
1254 
1255 	bnxt_print_link_info(eth_dev);
1256 	return rc;
1257 }
1258 
1259 static int bnxt_dev_set_link_down_op(struct rte_eth_dev *eth_dev)
1260 {
1261 	struct bnxt *bp = eth_dev->data->dev_private;
1262 
1263 	eth_dev->data->dev_link.link_status = 0;
1264 	bnxt_set_hwrm_link_config(bp, false);
1265 	bp->link_info->link_up = 0;
1266 
1267 	return 0;
1268 }
1269 
1270 static void bnxt_free_switch_domain(struct bnxt *bp)
1271 {
1272 	if (bp->switch_domain_id)
1273 		rte_eth_switch_domain_free(bp->switch_domain_id);
1274 }
1275 
1276 /* Unload the driver, release resources */
1277 static void bnxt_dev_stop_op(struct rte_eth_dev *eth_dev)
1278 {
1279 	struct bnxt *bp = eth_dev->data->dev_private;
1280 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
1281 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
1282 
1283 	eth_dev->data->dev_started = 0;
1284 	/* Prevent crashes when queues are still in use */
1285 	eth_dev->rx_pkt_burst = &bnxt_dummy_recv_pkts;
1286 	eth_dev->tx_pkt_burst = &bnxt_dummy_xmit_pkts;
1287 
1288 	bnxt_disable_int(bp);
1289 
1290 	/* disable uio/vfio intr/eventfd mapping */
1291 	rte_intr_disable(intr_handle);
1292 
1293 	bnxt_ulp_destroy_df_rules(bp, false);
1294 	bnxt_ulp_deinit(bp);
1295 
1296 	bnxt_cancel_fw_health_check(bp);
1297 
1298 	bnxt_dev_set_link_down_op(eth_dev);
1299 
1300 	/* Wait for link to be reset and the async notification to process.
1301 	 * During reset recovery, there is no need to wait and
1302 	 * VF/NPAR functions do not have privilege to change PHY config.
1303 	 */
1304 	if (!is_bnxt_in_error(bp) && BNXT_SINGLE_PF(bp))
1305 		bnxt_link_update(eth_dev, 1, ETH_LINK_DOWN);
1306 
1307 	/* Clean queue intr-vector mapping */
1308 	rte_intr_efd_disable(intr_handle);
1309 	if (intr_handle->intr_vec != NULL) {
1310 		rte_free(intr_handle->intr_vec);
1311 		intr_handle->intr_vec = NULL;
1312 	}
1313 
1314 	bnxt_hwrm_port_clr_stats(bp);
1315 	bnxt_free_tx_mbufs(bp);
1316 	bnxt_free_rx_mbufs(bp);
1317 	/* Process any remaining notifications in default completion queue */
1318 	bnxt_int_handler(eth_dev);
1319 	bnxt_shutdown_nic(bp);
1320 	bnxt_hwrm_if_change(bp, false);
1321 
1322 	rte_free(bp->mark_table);
1323 	bp->mark_table = NULL;
1324 
1325 	bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
1326 	bp->rx_cosq_cnt = 0;
1327 	/* All filters are deleted on a port stop. */
1328 	if (BNXT_FLOW_XSTATS_EN(bp))
1329 		bp->flow_stat->flow_count = 0;
1330 }
1331 
1332 static void bnxt_dev_close_op(struct rte_eth_dev *eth_dev)
1333 {
1334 	struct bnxt *bp = eth_dev->data->dev_private;
1335 
1336 	/* cancel the recovery handler before remove dev */
1337 	rte_eal_alarm_cancel(bnxt_dev_reset_and_resume, (void *)bp);
1338 	rte_eal_alarm_cancel(bnxt_dev_recover, (void *)bp);
1339 	bnxt_cancel_fc_thread(bp);
1340 
1341 	if (eth_dev->data->dev_started)
1342 		bnxt_dev_stop_op(eth_dev);
1343 
1344 	bnxt_free_switch_domain(bp);
1345 
1346 	bnxt_uninit_resources(bp, false);
1347 
1348 	bnxt_free_leds_info(bp);
1349 	bnxt_free_cos_queues(bp);
1350 	bnxt_free_link_info(bp);
1351 	bnxt_free_pf_info(bp);
1352 	bnxt_free_parent_info(bp);
1353 
1354 	eth_dev->dev_ops = NULL;
1355 	eth_dev->rx_pkt_burst = NULL;
1356 	eth_dev->tx_pkt_burst = NULL;
1357 
1358 	rte_memzone_free((const struct rte_memzone *)bp->tx_mem_zone);
1359 	bp->tx_mem_zone = NULL;
1360 	rte_memzone_free((const struct rte_memzone *)bp->rx_mem_zone);
1361 	bp->rx_mem_zone = NULL;
1362 
1363 	rte_free(bp->pf->vf_info);
1364 	bp->pf->vf_info = NULL;
1365 
1366 	rte_free(bp->grp_info);
1367 	bp->grp_info = NULL;
1368 }
1369 
1370 static void bnxt_mac_addr_remove_op(struct rte_eth_dev *eth_dev,
1371 				    uint32_t index)
1372 {
1373 	struct bnxt *bp = eth_dev->data->dev_private;
1374 	uint64_t pool_mask = eth_dev->data->mac_pool_sel[index];
1375 	struct bnxt_vnic_info *vnic;
1376 	struct bnxt_filter_info *filter, *temp_filter;
1377 	uint32_t i;
1378 
1379 	if (is_bnxt_in_error(bp))
1380 		return;
1381 
1382 	/*
1383 	 * Loop through all VNICs from the specified filter flow pools to
1384 	 * remove the corresponding MAC addr filter
1385 	 */
1386 	for (i = 0; i < bp->nr_vnics; i++) {
1387 		if (!(pool_mask & (1ULL << i)))
1388 			continue;
1389 
1390 		vnic = &bp->vnic_info[i];
1391 		filter = STAILQ_FIRST(&vnic->filter);
1392 		while (filter) {
1393 			temp_filter = STAILQ_NEXT(filter, next);
1394 			if (filter->mac_index == index) {
1395 				STAILQ_REMOVE(&vnic->filter, filter,
1396 						bnxt_filter_info, next);
1397 				bnxt_hwrm_clear_l2_filter(bp, filter);
1398 				bnxt_free_filter(bp, filter);
1399 			}
1400 			filter = temp_filter;
1401 		}
1402 	}
1403 }
1404 
1405 static int bnxt_add_mac_filter(struct bnxt *bp, struct bnxt_vnic_info *vnic,
1406 			       struct rte_ether_addr *mac_addr, uint32_t index,
1407 			       uint32_t pool)
1408 {
1409 	struct bnxt_filter_info *filter;
1410 	int rc = 0;
1411 
1412 	/* Attach requested MAC address to the new l2_filter */
1413 	STAILQ_FOREACH(filter, &vnic->filter, next) {
1414 		if (filter->mac_index == index) {
1415 			PMD_DRV_LOG(DEBUG,
1416 				    "MAC addr already existed for pool %d\n",
1417 				    pool);
1418 			return 0;
1419 		}
1420 	}
1421 
1422 	filter = bnxt_alloc_filter(bp);
1423 	if (!filter) {
1424 		PMD_DRV_LOG(ERR, "L2 filter alloc failed\n");
1425 		return -ENODEV;
1426 	}
1427 
1428 	/* bnxt_alloc_filter copies default MAC to filter->l2_addr. So,
1429 	 * if the MAC that's been programmed now is a different one, then,
1430 	 * copy that addr to filter->l2_addr
1431 	 */
1432 	if (mac_addr)
1433 		memcpy(filter->l2_addr, mac_addr, RTE_ETHER_ADDR_LEN);
1434 	filter->flags |= HWRM_CFA_L2_FILTER_ALLOC_INPUT_FLAGS_OUTERMOST;
1435 
1436 	rc = bnxt_hwrm_set_l2_filter(bp, vnic->fw_vnic_id, filter);
1437 	if (!rc) {
1438 		filter->mac_index = index;
1439 		if (filter->mac_index == 0)
1440 			STAILQ_INSERT_HEAD(&vnic->filter, filter, next);
1441 		else
1442 			STAILQ_INSERT_TAIL(&vnic->filter, filter, next);
1443 	} else {
1444 		bnxt_free_filter(bp, filter);
1445 	}
1446 
1447 	return rc;
1448 }
1449 
1450 static int bnxt_mac_addr_add_op(struct rte_eth_dev *eth_dev,
1451 				struct rte_ether_addr *mac_addr,
1452 				uint32_t index, uint32_t pool)
1453 {
1454 	struct bnxt *bp = eth_dev->data->dev_private;
1455 	struct bnxt_vnic_info *vnic = &bp->vnic_info[pool];
1456 	int rc = 0;
1457 
1458 	rc = is_bnxt_in_error(bp);
1459 	if (rc)
1460 		return rc;
1461 
1462 	if (BNXT_VF(bp) & !BNXT_VF_IS_TRUSTED(bp)) {
1463 		PMD_DRV_LOG(ERR, "Cannot add MAC address to a VF interface\n");
1464 		return -ENOTSUP;
1465 	}
1466 
1467 	if (!vnic) {
1468 		PMD_DRV_LOG(ERR, "VNIC not found for pool %d!\n", pool);
1469 		return -EINVAL;
1470 	}
1471 
1472 	/* Filter settings will get applied when port is started */
1473 	if (!eth_dev->data->dev_started)
1474 		return 0;
1475 
1476 	rc = bnxt_add_mac_filter(bp, vnic, mac_addr, index, pool);
1477 
1478 	return rc;
1479 }
1480 
1481 int bnxt_link_update(struct rte_eth_dev *eth_dev, int wait_to_complete,
1482 		     bool exp_link_status)
1483 {
1484 	int rc = 0;
1485 	struct bnxt *bp = eth_dev->data->dev_private;
1486 	struct rte_eth_link new;
1487 	int cnt = exp_link_status ? BNXT_LINK_UP_WAIT_CNT :
1488 		  BNXT_LINK_DOWN_WAIT_CNT;
1489 
1490 	rc = is_bnxt_in_error(bp);
1491 	if (rc)
1492 		return rc;
1493 
1494 	memset(&new, 0, sizeof(new));
1495 	do {
1496 		/* Retrieve link info from hardware */
1497 		rc = bnxt_get_hwrm_link_config(bp, &new);
1498 		if (rc) {
1499 			new.link_speed = ETH_LINK_SPEED_100M;
1500 			new.link_duplex = ETH_LINK_FULL_DUPLEX;
1501 			PMD_DRV_LOG(ERR,
1502 				"Failed to retrieve link rc = 0x%x!\n", rc);
1503 			goto out;
1504 		}
1505 
1506 		if (!wait_to_complete || new.link_status == exp_link_status)
1507 			break;
1508 
1509 		rte_delay_ms(BNXT_LINK_WAIT_INTERVAL);
1510 	} while (cnt--);
1511 
1512 out:
1513 	/* Timed out or success */
1514 	if (new.link_status != eth_dev->data->dev_link.link_status ||
1515 	new.link_speed != eth_dev->data->dev_link.link_speed) {
1516 		rte_eth_linkstatus_set(eth_dev, &new);
1517 
1518 		_rte_eth_dev_callback_process(eth_dev,
1519 					      RTE_ETH_EVENT_INTR_LSC,
1520 					      NULL);
1521 
1522 		bnxt_print_link_info(eth_dev);
1523 	}
1524 
1525 	return rc;
1526 }
1527 
1528 int bnxt_link_update_op(struct rte_eth_dev *eth_dev,
1529 			int wait_to_complete)
1530 {
1531 	return bnxt_link_update(eth_dev, wait_to_complete, ETH_LINK_UP);
1532 }
1533 
1534 static int bnxt_promiscuous_enable_op(struct rte_eth_dev *eth_dev)
1535 {
1536 	struct bnxt *bp = eth_dev->data->dev_private;
1537 	struct bnxt_vnic_info *vnic;
1538 	uint32_t old_flags;
1539 	int rc;
1540 
1541 	rc = is_bnxt_in_error(bp);
1542 	if (rc)
1543 		return rc;
1544 
1545 	/* Filter settings will get applied when port is started */
1546 	if (!eth_dev->data->dev_started)
1547 		return 0;
1548 
1549 	if (bp->vnic_info == NULL)
1550 		return 0;
1551 
1552 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
1553 
1554 	old_flags = vnic->flags;
1555 	vnic->flags |= BNXT_VNIC_INFO_PROMISC;
1556 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1557 	if (rc != 0)
1558 		vnic->flags = old_flags;
1559 
1560 	return rc;
1561 }
1562 
1563 static int bnxt_promiscuous_disable_op(struct rte_eth_dev *eth_dev)
1564 {
1565 	struct bnxt *bp = eth_dev->data->dev_private;
1566 	struct bnxt_vnic_info *vnic;
1567 	uint32_t old_flags;
1568 	int rc;
1569 
1570 	rc = is_bnxt_in_error(bp);
1571 	if (rc)
1572 		return rc;
1573 
1574 	/* Filter settings will get applied when port is started */
1575 	if (!eth_dev->data->dev_started)
1576 		return 0;
1577 
1578 	if (bp->vnic_info == NULL)
1579 		return 0;
1580 
1581 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
1582 
1583 	old_flags = vnic->flags;
1584 	vnic->flags &= ~BNXT_VNIC_INFO_PROMISC;
1585 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1586 	if (rc != 0)
1587 		vnic->flags = old_flags;
1588 
1589 	bnxt_ulp_create_df_rules(bp);
1590 
1591 	return rc;
1592 }
1593 
1594 static int bnxt_allmulticast_enable_op(struct rte_eth_dev *eth_dev)
1595 {
1596 	struct bnxt *bp = eth_dev->data->dev_private;
1597 	struct bnxt_vnic_info *vnic;
1598 	uint32_t old_flags;
1599 	int rc;
1600 
1601 	rc = is_bnxt_in_error(bp);
1602 	if (rc)
1603 		return rc;
1604 
1605 	/* Filter settings will get applied when port is started */
1606 	if (!eth_dev->data->dev_started)
1607 		return 0;
1608 
1609 	if (bp->vnic_info == NULL)
1610 		return 0;
1611 
1612 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
1613 
1614 	old_flags = vnic->flags;
1615 	vnic->flags |= BNXT_VNIC_INFO_ALLMULTI;
1616 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1617 	if (rc != 0)
1618 		vnic->flags = old_flags;
1619 
1620 	return rc;
1621 }
1622 
1623 static int bnxt_allmulticast_disable_op(struct rte_eth_dev *eth_dev)
1624 {
1625 	struct bnxt *bp = eth_dev->data->dev_private;
1626 	struct bnxt_vnic_info *vnic;
1627 	uint32_t old_flags;
1628 	int rc;
1629 
1630 	rc = is_bnxt_in_error(bp);
1631 	if (rc)
1632 		return rc;
1633 
1634 	/* Filter settings will get applied when port is started */
1635 	if (!eth_dev->data->dev_started)
1636 		return 0;
1637 
1638 	if (bp->vnic_info == NULL)
1639 		return 0;
1640 
1641 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
1642 
1643 	old_flags = vnic->flags;
1644 	vnic->flags &= ~BNXT_VNIC_INFO_ALLMULTI;
1645 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1646 	if (rc != 0)
1647 		vnic->flags = old_flags;
1648 
1649 	return rc;
1650 }
1651 
1652 /* Return bnxt_rx_queue pointer corresponding to a given rxq. */
1653 static struct bnxt_rx_queue *bnxt_qid_to_rxq(struct bnxt *bp, uint16_t qid)
1654 {
1655 	if (qid >= bp->rx_nr_rings)
1656 		return NULL;
1657 
1658 	return bp->eth_dev->data->rx_queues[qid];
1659 }
1660 
1661 /* Return rxq corresponding to a given rss table ring/group ID. */
1662 static uint16_t bnxt_rss_to_qid(struct bnxt *bp, uint16_t fwr)
1663 {
1664 	struct bnxt_rx_queue *rxq;
1665 	unsigned int i;
1666 
1667 	if (!BNXT_HAS_RING_GRPS(bp)) {
1668 		for (i = 0; i < bp->rx_nr_rings; i++) {
1669 			rxq = bp->eth_dev->data->rx_queues[i];
1670 			if (rxq->rx_ring->rx_ring_struct->fw_ring_id == fwr)
1671 				return rxq->index;
1672 		}
1673 	} else {
1674 		for (i = 0; i < bp->rx_nr_rings; i++) {
1675 			if (bp->grp_info[i].fw_grp_id == fwr)
1676 				return i;
1677 		}
1678 	}
1679 
1680 	return INVALID_HW_RING_ID;
1681 }
1682 
1683 static int bnxt_reta_update_op(struct rte_eth_dev *eth_dev,
1684 			    struct rte_eth_rss_reta_entry64 *reta_conf,
1685 			    uint16_t reta_size)
1686 {
1687 	struct bnxt *bp = eth_dev->data->dev_private;
1688 	struct rte_eth_conf *dev_conf = &bp->eth_dev->data->dev_conf;
1689 	struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
1690 	uint16_t tbl_size = bnxt_rss_hash_tbl_size(bp);
1691 	uint16_t idx, sft;
1692 	int i, rc;
1693 
1694 	rc = is_bnxt_in_error(bp);
1695 	if (rc)
1696 		return rc;
1697 
1698 	if (!vnic->rss_table)
1699 		return -EINVAL;
1700 
1701 	if (!(dev_conf->rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG))
1702 		return -EINVAL;
1703 
1704 	if (reta_size != tbl_size) {
1705 		PMD_DRV_LOG(ERR, "The configured hash table lookup size "
1706 			"(%d) must equal the size supported by the hardware "
1707 			"(%d)\n", reta_size, tbl_size);
1708 		return -EINVAL;
1709 	}
1710 
1711 	for (i = 0; i < reta_size; i++) {
1712 		struct bnxt_rx_queue *rxq;
1713 
1714 		idx = i / RTE_RETA_GROUP_SIZE;
1715 		sft = i % RTE_RETA_GROUP_SIZE;
1716 
1717 		if (!(reta_conf[idx].mask & (1ULL << sft)))
1718 			continue;
1719 
1720 		rxq = bnxt_qid_to_rxq(bp, reta_conf[idx].reta[sft]);
1721 		if (!rxq) {
1722 			PMD_DRV_LOG(ERR, "Invalid ring in reta_conf.\n");
1723 			return -EINVAL;
1724 		}
1725 
1726 		if (BNXT_CHIP_THOR(bp)) {
1727 			vnic->rss_table[i * 2] =
1728 				rxq->rx_ring->rx_ring_struct->fw_ring_id;
1729 			vnic->rss_table[i * 2 + 1] =
1730 				rxq->cp_ring->cp_ring_struct->fw_ring_id;
1731 		} else {
1732 			vnic->rss_table[i] =
1733 			    vnic->fw_grp_ids[reta_conf[idx].reta[sft]];
1734 		}
1735 	}
1736 
1737 	bnxt_hwrm_vnic_rss_cfg(bp, vnic);
1738 	return 0;
1739 }
1740 
1741 static int bnxt_reta_query_op(struct rte_eth_dev *eth_dev,
1742 			      struct rte_eth_rss_reta_entry64 *reta_conf,
1743 			      uint16_t reta_size)
1744 {
1745 	struct bnxt *bp = eth_dev->data->dev_private;
1746 	struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
1747 	uint16_t tbl_size = bnxt_rss_hash_tbl_size(bp);
1748 	uint16_t idx, sft, i;
1749 	int rc;
1750 
1751 	rc = is_bnxt_in_error(bp);
1752 	if (rc)
1753 		return rc;
1754 
1755 	/* Retrieve from the default VNIC */
1756 	if (!vnic)
1757 		return -EINVAL;
1758 	if (!vnic->rss_table)
1759 		return -EINVAL;
1760 
1761 	if (reta_size != tbl_size) {
1762 		PMD_DRV_LOG(ERR, "The configured hash table lookup size "
1763 			"(%d) must equal the size supported by the hardware "
1764 			"(%d)\n", reta_size, tbl_size);
1765 		return -EINVAL;
1766 	}
1767 
1768 	for (idx = 0, i = 0; i < reta_size; i++) {
1769 		idx = i / RTE_RETA_GROUP_SIZE;
1770 		sft = i % RTE_RETA_GROUP_SIZE;
1771 
1772 		if (reta_conf[idx].mask & (1ULL << sft)) {
1773 			uint16_t qid;
1774 
1775 			if (BNXT_CHIP_THOR(bp))
1776 				qid = bnxt_rss_to_qid(bp,
1777 						      vnic->rss_table[i * 2]);
1778 			else
1779 				qid = bnxt_rss_to_qid(bp, vnic->rss_table[i]);
1780 
1781 			if (qid == INVALID_HW_RING_ID) {
1782 				PMD_DRV_LOG(ERR, "Inv. entry in rss table.\n");
1783 				return -EINVAL;
1784 			}
1785 			reta_conf[idx].reta[sft] = qid;
1786 		}
1787 	}
1788 
1789 	return 0;
1790 }
1791 
1792 static int bnxt_rss_hash_update_op(struct rte_eth_dev *eth_dev,
1793 				   struct rte_eth_rss_conf *rss_conf)
1794 {
1795 	struct bnxt *bp = eth_dev->data->dev_private;
1796 	struct rte_eth_conf *dev_conf = &bp->eth_dev->data->dev_conf;
1797 	struct bnxt_vnic_info *vnic;
1798 	int rc;
1799 
1800 	rc = is_bnxt_in_error(bp);
1801 	if (rc)
1802 		return rc;
1803 
1804 	/*
1805 	 * If RSS enablement were different than dev_configure,
1806 	 * then return -EINVAL
1807 	 */
1808 	if (dev_conf->rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG) {
1809 		if (!rss_conf->rss_hf)
1810 			PMD_DRV_LOG(ERR, "Hash type NONE\n");
1811 	} else {
1812 		if (rss_conf->rss_hf & BNXT_ETH_RSS_SUPPORT)
1813 			return -EINVAL;
1814 	}
1815 
1816 	bp->flags |= BNXT_FLAG_UPDATE_HASH;
1817 	memcpy(&eth_dev->data->dev_conf.rx_adv_conf.rss_conf,
1818 	       rss_conf,
1819 	       sizeof(*rss_conf));
1820 
1821 	/* Update the default RSS VNIC(s) */
1822 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
1823 	vnic->hash_type = bnxt_rte_to_hwrm_hash_types(rss_conf->rss_hf);
1824 
1825 	/*
1826 	 * If hashkey is not specified, use the previously configured
1827 	 * hashkey
1828 	 */
1829 	if (!rss_conf->rss_key)
1830 		goto rss_config;
1831 
1832 	if (rss_conf->rss_key_len != HW_HASH_KEY_SIZE) {
1833 		PMD_DRV_LOG(ERR,
1834 			    "Invalid hashkey length, should be 16 bytes\n");
1835 		return -EINVAL;
1836 	}
1837 	memcpy(vnic->rss_hash_key, rss_conf->rss_key, rss_conf->rss_key_len);
1838 
1839 rss_config:
1840 	bnxt_hwrm_vnic_rss_cfg(bp, vnic);
1841 	return 0;
1842 }
1843 
1844 static int bnxt_rss_hash_conf_get_op(struct rte_eth_dev *eth_dev,
1845 				     struct rte_eth_rss_conf *rss_conf)
1846 {
1847 	struct bnxt *bp = eth_dev->data->dev_private;
1848 	struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
1849 	int len, rc;
1850 	uint32_t hash_types;
1851 
1852 	rc = is_bnxt_in_error(bp);
1853 	if (rc)
1854 		return rc;
1855 
1856 	/* RSS configuration is the same for all VNICs */
1857 	if (vnic && vnic->rss_hash_key) {
1858 		if (rss_conf->rss_key) {
1859 			len = rss_conf->rss_key_len <= HW_HASH_KEY_SIZE ?
1860 			      rss_conf->rss_key_len : HW_HASH_KEY_SIZE;
1861 			memcpy(rss_conf->rss_key, vnic->rss_hash_key, len);
1862 		}
1863 
1864 		hash_types = vnic->hash_type;
1865 		rss_conf->rss_hf = 0;
1866 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV4) {
1867 			rss_conf->rss_hf |= ETH_RSS_IPV4;
1868 			hash_types &= ~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV4;
1869 		}
1870 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV4) {
1871 			rss_conf->rss_hf |= ETH_RSS_NONFRAG_IPV4_TCP;
1872 			hash_types &=
1873 				~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV4;
1874 		}
1875 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV4) {
1876 			rss_conf->rss_hf |= ETH_RSS_NONFRAG_IPV4_UDP;
1877 			hash_types &=
1878 				~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV4;
1879 		}
1880 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV6) {
1881 			rss_conf->rss_hf |= ETH_RSS_IPV6;
1882 			hash_types &= ~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV6;
1883 		}
1884 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV6) {
1885 			rss_conf->rss_hf |= ETH_RSS_NONFRAG_IPV6_TCP;
1886 			hash_types &=
1887 				~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV6;
1888 		}
1889 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV6) {
1890 			rss_conf->rss_hf |= ETH_RSS_NONFRAG_IPV6_UDP;
1891 			hash_types &=
1892 				~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV6;
1893 		}
1894 		if (hash_types) {
1895 			PMD_DRV_LOG(ERR,
1896 				"Unknown RSS config from firmware (%08x), RSS disabled",
1897 				vnic->hash_type);
1898 			return -ENOTSUP;
1899 		}
1900 	} else {
1901 		rss_conf->rss_hf = 0;
1902 	}
1903 	return 0;
1904 }
1905 
1906 static int bnxt_flow_ctrl_get_op(struct rte_eth_dev *dev,
1907 			       struct rte_eth_fc_conf *fc_conf)
1908 {
1909 	struct bnxt *bp = dev->data->dev_private;
1910 	struct rte_eth_link link_info;
1911 	int rc;
1912 
1913 	rc = is_bnxt_in_error(bp);
1914 	if (rc)
1915 		return rc;
1916 
1917 	rc = bnxt_get_hwrm_link_config(bp, &link_info);
1918 	if (rc)
1919 		return rc;
1920 
1921 	memset(fc_conf, 0, sizeof(*fc_conf));
1922 	if (bp->link_info->auto_pause)
1923 		fc_conf->autoneg = 1;
1924 	switch (bp->link_info->pause) {
1925 	case 0:
1926 		fc_conf->mode = RTE_FC_NONE;
1927 		break;
1928 	case HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_TX:
1929 		fc_conf->mode = RTE_FC_TX_PAUSE;
1930 		break;
1931 	case HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_RX:
1932 		fc_conf->mode = RTE_FC_RX_PAUSE;
1933 		break;
1934 	case (HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_TX |
1935 			HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_RX):
1936 		fc_conf->mode = RTE_FC_FULL;
1937 		break;
1938 	}
1939 	return 0;
1940 }
1941 
1942 static int bnxt_flow_ctrl_set_op(struct rte_eth_dev *dev,
1943 			       struct rte_eth_fc_conf *fc_conf)
1944 {
1945 	struct bnxt *bp = dev->data->dev_private;
1946 	int rc;
1947 
1948 	rc = is_bnxt_in_error(bp);
1949 	if (rc)
1950 		return rc;
1951 
1952 	if (!BNXT_SINGLE_PF(bp) || BNXT_VF(bp)) {
1953 		PMD_DRV_LOG(ERR, "Flow Control Settings cannot be modified\n");
1954 		return -ENOTSUP;
1955 	}
1956 
1957 	switch (fc_conf->mode) {
1958 	case RTE_FC_NONE:
1959 		bp->link_info->auto_pause = 0;
1960 		bp->link_info->force_pause = 0;
1961 		break;
1962 	case RTE_FC_RX_PAUSE:
1963 		if (fc_conf->autoneg) {
1964 			bp->link_info->auto_pause =
1965 					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_RX;
1966 			bp->link_info->force_pause = 0;
1967 		} else {
1968 			bp->link_info->auto_pause = 0;
1969 			bp->link_info->force_pause =
1970 					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_RX;
1971 		}
1972 		break;
1973 	case RTE_FC_TX_PAUSE:
1974 		if (fc_conf->autoneg) {
1975 			bp->link_info->auto_pause =
1976 					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_TX;
1977 			bp->link_info->force_pause = 0;
1978 		} else {
1979 			bp->link_info->auto_pause = 0;
1980 			bp->link_info->force_pause =
1981 					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_TX;
1982 		}
1983 		break;
1984 	case RTE_FC_FULL:
1985 		if (fc_conf->autoneg) {
1986 			bp->link_info->auto_pause =
1987 					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_TX |
1988 					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_RX;
1989 			bp->link_info->force_pause = 0;
1990 		} else {
1991 			bp->link_info->auto_pause = 0;
1992 			bp->link_info->force_pause =
1993 					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_TX |
1994 					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_RX;
1995 		}
1996 		break;
1997 	}
1998 	return bnxt_set_hwrm_link_config(bp, true);
1999 }
2000 
2001 /* Add UDP tunneling port */
2002 static int
2003 bnxt_udp_tunnel_port_add_op(struct rte_eth_dev *eth_dev,
2004 			 struct rte_eth_udp_tunnel *udp_tunnel)
2005 {
2006 	struct bnxt *bp = eth_dev->data->dev_private;
2007 	uint16_t tunnel_type = 0;
2008 	int rc = 0;
2009 
2010 	rc = is_bnxt_in_error(bp);
2011 	if (rc)
2012 		return rc;
2013 
2014 	switch (udp_tunnel->prot_type) {
2015 	case RTE_TUNNEL_TYPE_VXLAN:
2016 		if (bp->vxlan_port_cnt) {
2017 			PMD_DRV_LOG(ERR, "Tunnel Port %d already programmed\n",
2018 				udp_tunnel->udp_port);
2019 			if (bp->vxlan_port != udp_tunnel->udp_port) {
2020 				PMD_DRV_LOG(ERR, "Only one port allowed\n");
2021 				return -ENOSPC;
2022 			}
2023 			bp->vxlan_port_cnt++;
2024 			return 0;
2025 		}
2026 		tunnel_type =
2027 			HWRM_TUNNEL_DST_PORT_ALLOC_INPUT_TUNNEL_TYPE_VXLAN;
2028 		bp->vxlan_port_cnt++;
2029 		break;
2030 	case RTE_TUNNEL_TYPE_GENEVE:
2031 		if (bp->geneve_port_cnt) {
2032 			PMD_DRV_LOG(ERR, "Tunnel Port %d already programmed\n",
2033 				udp_tunnel->udp_port);
2034 			if (bp->geneve_port != udp_tunnel->udp_port) {
2035 				PMD_DRV_LOG(ERR, "Only one port allowed\n");
2036 				return -ENOSPC;
2037 			}
2038 			bp->geneve_port_cnt++;
2039 			return 0;
2040 		}
2041 		tunnel_type =
2042 			HWRM_TUNNEL_DST_PORT_ALLOC_INPUT_TUNNEL_TYPE_GENEVE;
2043 		bp->geneve_port_cnt++;
2044 		break;
2045 	default:
2046 		PMD_DRV_LOG(ERR, "Tunnel type is not supported\n");
2047 		return -ENOTSUP;
2048 	}
2049 	rc = bnxt_hwrm_tunnel_dst_port_alloc(bp, udp_tunnel->udp_port,
2050 					     tunnel_type);
2051 	return rc;
2052 }
2053 
2054 static int
2055 bnxt_udp_tunnel_port_del_op(struct rte_eth_dev *eth_dev,
2056 			 struct rte_eth_udp_tunnel *udp_tunnel)
2057 {
2058 	struct bnxt *bp = eth_dev->data->dev_private;
2059 	uint16_t tunnel_type = 0;
2060 	uint16_t port = 0;
2061 	int rc = 0;
2062 
2063 	rc = is_bnxt_in_error(bp);
2064 	if (rc)
2065 		return rc;
2066 
2067 	switch (udp_tunnel->prot_type) {
2068 	case RTE_TUNNEL_TYPE_VXLAN:
2069 		if (!bp->vxlan_port_cnt) {
2070 			PMD_DRV_LOG(ERR, "No Tunnel port configured yet\n");
2071 			return -EINVAL;
2072 		}
2073 		if (bp->vxlan_port != udp_tunnel->udp_port) {
2074 			PMD_DRV_LOG(ERR, "Req Port: %d. Configured port: %d\n",
2075 				udp_tunnel->udp_port, bp->vxlan_port);
2076 			return -EINVAL;
2077 		}
2078 		if (--bp->vxlan_port_cnt)
2079 			return 0;
2080 
2081 		tunnel_type =
2082 			HWRM_TUNNEL_DST_PORT_FREE_INPUT_TUNNEL_TYPE_VXLAN;
2083 		port = bp->vxlan_fw_dst_port_id;
2084 		break;
2085 	case RTE_TUNNEL_TYPE_GENEVE:
2086 		if (!bp->geneve_port_cnt) {
2087 			PMD_DRV_LOG(ERR, "No Tunnel port configured yet\n");
2088 			return -EINVAL;
2089 		}
2090 		if (bp->geneve_port != udp_tunnel->udp_port) {
2091 			PMD_DRV_LOG(ERR, "Req Port: %d. Configured port: %d\n",
2092 				udp_tunnel->udp_port, bp->geneve_port);
2093 			return -EINVAL;
2094 		}
2095 		if (--bp->geneve_port_cnt)
2096 			return 0;
2097 
2098 		tunnel_type =
2099 			HWRM_TUNNEL_DST_PORT_FREE_INPUT_TUNNEL_TYPE_GENEVE;
2100 		port = bp->geneve_fw_dst_port_id;
2101 		break;
2102 	default:
2103 		PMD_DRV_LOG(ERR, "Tunnel type is not supported\n");
2104 		return -ENOTSUP;
2105 	}
2106 
2107 	rc = bnxt_hwrm_tunnel_dst_port_free(bp, port, tunnel_type);
2108 	if (!rc) {
2109 		if (tunnel_type ==
2110 		    HWRM_TUNNEL_DST_PORT_FREE_INPUT_TUNNEL_TYPE_VXLAN)
2111 			bp->vxlan_port = 0;
2112 		if (tunnel_type ==
2113 		    HWRM_TUNNEL_DST_PORT_FREE_INPUT_TUNNEL_TYPE_GENEVE)
2114 			bp->geneve_port = 0;
2115 	}
2116 	return rc;
2117 }
2118 
2119 static int bnxt_del_vlan_filter(struct bnxt *bp, uint16_t vlan_id)
2120 {
2121 	struct bnxt_filter_info *filter;
2122 	struct bnxt_vnic_info *vnic;
2123 	int rc = 0;
2124 	uint32_t chk = HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN;
2125 
2126 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
2127 	filter = STAILQ_FIRST(&vnic->filter);
2128 	while (filter) {
2129 		/* Search for this matching MAC+VLAN filter */
2130 		if (bnxt_vlan_filter_exists(bp, filter, chk, vlan_id)) {
2131 			/* Delete the filter */
2132 			rc = bnxt_hwrm_clear_l2_filter(bp, filter);
2133 			if (rc)
2134 				return rc;
2135 			STAILQ_REMOVE(&vnic->filter, filter,
2136 				      bnxt_filter_info, next);
2137 			bnxt_free_filter(bp, filter);
2138 			PMD_DRV_LOG(INFO,
2139 				    "Deleted vlan filter for %d\n",
2140 				    vlan_id);
2141 			return 0;
2142 		}
2143 		filter = STAILQ_NEXT(filter, next);
2144 	}
2145 	return -ENOENT;
2146 }
2147 
2148 static int bnxt_add_vlan_filter(struct bnxt *bp, uint16_t vlan_id)
2149 {
2150 	struct bnxt_filter_info *filter;
2151 	struct bnxt_vnic_info *vnic;
2152 	int rc = 0;
2153 	uint32_t en = HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN |
2154 		HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN_MASK;
2155 	uint32_t chk = HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN;
2156 
2157 	/* Implementation notes on the use of VNIC in this command:
2158 	 *
2159 	 * By default, these filters belong to default vnic for the function.
2160 	 * Once these filters are set up, only destination VNIC can be modified.
2161 	 * If the destination VNIC is not specified in this command,
2162 	 * then the HWRM shall only create an l2 context id.
2163 	 */
2164 
2165 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
2166 	filter = STAILQ_FIRST(&vnic->filter);
2167 	/* Check if the VLAN has already been added */
2168 	while (filter) {
2169 		if (bnxt_vlan_filter_exists(bp, filter, chk, vlan_id))
2170 			return -EEXIST;
2171 
2172 		filter = STAILQ_NEXT(filter, next);
2173 	}
2174 
2175 	/* No match found. Alloc a fresh filter and issue the L2_FILTER_ALLOC
2176 	 * command to create MAC+VLAN filter with the right flags, enables set.
2177 	 */
2178 	filter = bnxt_alloc_filter(bp);
2179 	if (!filter) {
2180 		PMD_DRV_LOG(ERR,
2181 			    "MAC/VLAN filter alloc failed\n");
2182 		return -ENOMEM;
2183 	}
2184 	/* MAC + VLAN ID filter */
2185 	/* If l2_ivlan == 0 and l2_ivlan_mask != 0, only
2186 	 * untagged packets are received
2187 	 *
2188 	 * If l2_ivlan != 0 and l2_ivlan_mask != 0, untagged
2189 	 * packets and only the programmed vlan's packets are received
2190 	 */
2191 	filter->l2_ivlan = vlan_id;
2192 	filter->l2_ivlan_mask = 0x0FFF;
2193 	filter->enables |= en;
2194 	filter->flags |= HWRM_CFA_L2_FILTER_ALLOC_INPUT_FLAGS_OUTERMOST;
2195 
2196 	rc = bnxt_hwrm_set_l2_filter(bp, vnic->fw_vnic_id, filter);
2197 	if (rc) {
2198 		/* Free the newly allocated filter as we were
2199 		 * not able to create the filter in hardware.
2200 		 */
2201 		bnxt_free_filter(bp, filter);
2202 		return rc;
2203 	}
2204 
2205 	filter->mac_index = 0;
2206 	/* Add this new filter to the list */
2207 	if (vlan_id == 0)
2208 		STAILQ_INSERT_HEAD(&vnic->filter, filter, next);
2209 	else
2210 		STAILQ_INSERT_TAIL(&vnic->filter, filter, next);
2211 
2212 	PMD_DRV_LOG(INFO,
2213 		    "Added Vlan filter for %d\n", vlan_id);
2214 	return rc;
2215 }
2216 
2217 static int bnxt_vlan_filter_set_op(struct rte_eth_dev *eth_dev,
2218 		uint16_t vlan_id, int on)
2219 {
2220 	struct bnxt *bp = eth_dev->data->dev_private;
2221 	int rc;
2222 
2223 	rc = is_bnxt_in_error(bp);
2224 	if (rc)
2225 		return rc;
2226 
2227 	if (!eth_dev->data->dev_started) {
2228 		PMD_DRV_LOG(ERR, "port must be started before setting vlan\n");
2229 		return -EINVAL;
2230 	}
2231 
2232 	/* These operations apply to ALL existing MAC/VLAN filters */
2233 	if (on)
2234 		return bnxt_add_vlan_filter(bp, vlan_id);
2235 	else
2236 		return bnxt_del_vlan_filter(bp, vlan_id);
2237 }
2238 
2239 static int bnxt_del_dflt_mac_filter(struct bnxt *bp,
2240 				    struct bnxt_vnic_info *vnic)
2241 {
2242 	struct bnxt_filter_info *filter;
2243 	int rc;
2244 
2245 	filter = STAILQ_FIRST(&vnic->filter);
2246 	while (filter) {
2247 		if (filter->mac_index == 0 &&
2248 		    !memcmp(filter->l2_addr, bp->mac_addr,
2249 			    RTE_ETHER_ADDR_LEN)) {
2250 			rc = bnxt_hwrm_clear_l2_filter(bp, filter);
2251 			if (!rc) {
2252 				STAILQ_REMOVE(&vnic->filter, filter,
2253 					      bnxt_filter_info, next);
2254 				bnxt_free_filter(bp, filter);
2255 			}
2256 			return rc;
2257 		}
2258 		filter = STAILQ_NEXT(filter, next);
2259 	}
2260 	return 0;
2261 }
2262 
2263 static int
2264 bnxt_config_vlan_hw_filter(struct bnxt *bp, uint64_t rx_offloads)
2265 {
2266 	struct bnxt_vnic_info *vnic;
2267 	unsigned int i;
2268 	int rc;
2269 
2270 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
2271 	if (!(rx_offloads & DEV_RX_OFFLOAD_VLAN_FILTER)) {
2272 		/* Remove any VLAN filters programmed */
2273 		for (i = 0; i < RTE_ETHER_MAX_VLAN_ID; i++)
2274 			bnxt_del_vlan_filter(bp, i);
2275 
2276 		rc = bnxt_add_mac_filter(bp, vnic, NULL, 0, 0);
2277 		if (rc)
2278 			return rc;
2279 	} else {
2280 		/* Default filter will allow packets that match the
2281 		 * dest mac. So, it has to be deleted, otherwise, we
2282 		 * will endup receiving vlan packets for which the
2283 		 * filter is not programmed, when hw-vlan-filter
2284 		 * configuration is ON
2285 		 */
2286 		bnxt_del_dflt_mac_filter(bp, vnic);
2287 		/* This filter will allow only untagged packets */
2288 		bnxt_add_vlan_filter(bp, 0);
2289 	}
2290 	PMD_DRV_LOG(DEBUG, "VLAN Filtering: %d\n",
2291 		    !!(rx_offloads & DEV_RX_OFFLOAD_VLAN_FILTER));
2292 
2293 	return 0;
2294 }
2295 
2296 static int bnxt_free_one_vnic(struct bnxt *bp, uint16_t vnic_id)
2297 {
2298 	struct bnxt_vnic_info *vnic = &bp->vnic_info[vnic_id];
2299 	unsigned int i;
2300 	int rc;
2301 
2302 	/* Destroy vnic filters and vnic */
2303 	if (bp->eth_dev->data->dev_conf.rxmode.offloads &
2304 	    DEV_RX_OFFLOAD_VLAN_FILTER) {
2305 		for (i = 0; i < RTE_ETHER_MAX_VLAN_ID; i++)
2306 			bnxt_del_vlan_filter(bp, i);
2307 	}
2308 	bnxt_del_dflt_mac_filter(bp, vnic);
2309 
2310 	rc = bnxt_hwrm_vnic_free(bp, vnic);
2311 	if (rc)
2312 		return rc;
2313 
2314 	rte_free(vnic->fw_grp_ids);
2315 	vnic->fw_grp_ids = NULL;
2316 
2317 	vnic->rx_queue_cnt = 0;
2318 
2319 	return 0;
2320 }
2321 
2322 static int
2323 bnxt_config_vlan_hw_stripping(struct bnxt *bp, uint64_t rx_offloads)
2324 {
2325 	struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
2326 	int rc;
2327 
2328 	/* Destroy, recreate and reconfigure the default vnic */
2329 	rc = bnxt_free_one_vnic(bp, 0);
2330 	if (rc)
2331 		return rc;
2332 
2333 	/* default vnic 0 */
2334 	rc = bnxt_setup_one_vnic(bp, 0);
2335 	if (rc)
2336 		return rc;
2337 
2338 	if (bp->eth_dev->data->dev_conf.rxmode.offloads &
2339 	    DEV_RX_OFFLOAD_VLAN_FILTER) {
2340 		rc = bnxt_add_vlan_filter(bp, 0);
2341 		if (rc)
2342 			return rc;
2343 		rc = bnxt_restore_vlan_filters(bp);
2344 		if (rc)
2345 			return rc;
2346 	} else {
2347 		rc = bnxt_add_mac_filter(bp, vnic, NULL, 0, 0);
2348 		if (rc)
2349 			return rc;
2350 	}
2351 
2352 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
2353 	if (rc)
2354 		return rc;
2355 
2356 	PMD_DRV_LOG(DEBUG, "VLAN Strip Offload: %d\n",
2357 		    !!(rx_offloads & DEV_RX_OFFLOAD_VLAN_STRIP));
2358 
2359 	return rc;
2360 }
2361 
2362 static int
2363 bnxt_vlan_offload_set_op(struct rte_eth_dev *dev, int mask)
2364 {
2365 	uint64_t rx_offloads = dev->data->dev_conf.rxmode.offloads;
2366 	struct bnxt *bp = dev->data->dev_private;
2367 	int rc;
2368 
2369 	rc = is_bnxt_in_error(bp);
2370 	if (rc)
2371 		return rc;
2372 
2373 	/* Filter settings will get applied when port is started */
2374 	if (!dev->data->dev_started)
2375 		return 0;
2376 
2377 	if (mask & ETH_VLAN_FILTER_MASK) {
2378 		/* Enable or disable VLAN filtering */
2379 		rc = bnxt_config_vlan_hw_filter(bp, rx_offloads);
2380 		if (rc)
2381 			return rc;
2382 	}
2383 
2384 	if (mask & ETH_VLAN_STRIP_MASK) {
2385 		/* Enable or disable VLAN stripping */
2386 		rc = bnxt_config_vlan_hw_stripping(bp, rx_offloads);
2387 		if (rc)
2388 			return rc;
2389 	}
2390 
2391 	if (mask & ETH_VLAN_EXTEND_MASK) {
2392 		if (rx_offloads & DEV_RX_OFFLOAD_VLAN_EXTEND)
2393 			PMD_DRV_LOG(DEBUG, "Extend VLAN supported\n");
2394 		else
2395 			PMD_DRV_LOG(INFO, "Extend VLAN unsupported\n");
2396 	}
2397 
2398 	return 0;
2399 }
2400 
2401 static int
2402 bnxt_vlan_tpid_set_op(struct rte_eth_dev *dev, enum rte_vlan_type vlan_type,
2403 		      uint16_t tpid)
2404 {
2405 	struct bnxt *bp = dev->data->dev_private;
2406 	int qinq = dev->data->dev_conf.rxmode.offloads &
2407 		   DEV_RX_OFFLOAD_VLAN_EXTEND;
2408 
2409 	if (vlan_type != ETH_VLAN_TYPE_INNER &&
2410 	    vlan_type != ETH_VLAN_TYPE_OUTER) {
2411 		PMD_DRV_LOG(ERR,
2412 			    "Unsupported vlan type.");
2413 		return -EINVAL;
2414 	}
2415 	if (!qinq) {
2416 		PMD_DRV_LOG(ERR,
2417 			    "QinQ not enabled. Needs to be ON as we can "
2418 			    "accelerate only outer vlan\n");
2419 		return -EINVAL;
2420 	}
2421 
2422 	if (vlan_type == ETH_VLAN_TYPE_OUTER) {
2423 		switch (tpid) {
2424 		case RTE_ETHER_TYPE_QINQ:
2425 			bp->outer_tpid_bd =
2426 				TX_BD_LONG_CFA_META_VLAN_TPID_TPID88A8;
2427 				break;
2428 		case RTE_ETHER_TYPE_VLAN:
2429 			bp->outer_tpid_bd =
2430 				TX_BD_LONG_CFA_META_VLAN_TPID_TPID8100;
2431 				break;
2432 		case RTE_ETHER_TYPE_QINQ1:
2433 			bp->outer_tpid_bd =
2434 				TX_BD_LONG_CFA_META_VLAN_TPID_TPID9100;
2435 				break;
2436 		case RTE_ETHER_TYPE_QINQ2:
2437 			bp->outer_tpid_bd =
2438 				TX_BD_LONG_CFA_META_VLAN_TPID_TPID9200;
2439 				break;
2440 		case RTE_ETHER_TYPE_QINQ3:
2441 			bp->outer_tpid_bd =
2442 				 TX_BD_LONG_CFA_META_VLAN_TPID_TPID9300;
2443 				break;
2444 		default:
2445 			PMD_DRV_LOG(ERR, "Invalid TPID: %x\n", tpid);
2446 			return -EINVAL;
2447 		}
2448 		bp->outer_tpid_bd |= tpid;
2449 		PMD_DRV_LOG(INFO, "outer_tpid_bd = %x\n", bp->outer_tpid_bd);
2450 	} else if (vlan_type == ETH_VLAN_TYPE_INNER) {
2451 		PMD_DRV_LOG(ERR,
2452 			    "Can accelerate only outer vlan in QinQ\n");
2453 		return -EINVAL;
2454 	}
2455 
2456 	return 0;
2457 }
2458 
2459 static int
2460 bnxt_set_default_mac_addr_op(struct rte_eth_dev *dev,
2461 			     struct rte_ether_addr *addr)
2462 {
2463 	struct bnxt *bp = dev->data->dev_private;
2464 	/* Default Filter is tied to VNIC 0 */
2465 	struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
2466 	int rc;
2467 
2468 	rc = is_bnxt_in_error(bp);
2469 	if (rc)
2470 		return rc;
2471 
2472 	if (BNXT_VF(bp) && !BNXT_VF_IS_TRUSTED(bp))
2473 		return -EPERM;
2474 
2475 	if (rte_is_zero_ether_addr(addr))
2476 		return -EINVAL;
2477 
2478 	/* Filter settings will get applied when port is started */
2479 	if (!dev->data->dev_started)
2480 		return 0;
2481 
2482 	/* Check if the requested MAC is already added */
2483 	if (memcmp(addr, bp->mac_addr, RTE_ETHER_ADDR_LEN) == 0)
2484 		return 0;
2485 
2486 	/* Destroy filter and re-create it */
2487 	bnxt_del_dflt_mac_filter(bp, vnic);
2488 
2489 	memcpy(bp->mac_addr, addr, RTE_ETHER_ADDR_LEN);
2490 	if (dev->data->dev_conf.rxmode.offloads & DEV_RX_OFFLOAD_VLAN_FILTER) {
2491 		/* This filter will allow only untagged packets */
2492 		rc = bnxt_add_vlan_filter(bp, 0);
2493 	} else {
2494 		rc = bnxt_add_mac_filter(bp, vnic, addr, 0, 0);
2495 	}
2496 
2497 	PMD_DRV_LOG(DEBUG, "Set MAC addr\n");
2498 	return rc;
2499 }
2500 
2501 static int
2502 bnxt_dev_set_mc_addr_list_op(struct rte_eth_dev *eth_dev,
2503 			  struct rte_ether_addr *mc_addr_set,
2504 			  uint32_t nb_mc_addr)
2505 {
2506 	struct bnxt *bp = eth_dev->data->dev_private;
2507 	char *mc_addr_list = (char *)mc_addr_set;
2508 	struct bnxt_vnic_info *vnic;
2509 	uint32_t off = 0, i = 0;
2510 	int rc;
2511 
2512 	rc = is_bnxt_in_error(bp);
2513 	if (rc)
2514 		return rc;
2515 
2516 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
2517 
2518 	if (nb_mc_addr > BNXT_MAX_MC_ADDRS) {
2519 		vnic->flags |= BNXT_VNIC_INFO_ALLMULTI;
2520 		goto allmulti;
2521 	}
2522 
2523 	/* TODO Check for Duplicate mcast addresses */
2524 	vnic->flags &= ~BNXT_VNIC_INFO_ALLMULTI;
2525 	for (i = 0; i < nb_mc_addr; i++) {
2526 		memcpy(vnic->mc_list + off, &mc_addr_list[i],
2527 			RTE_ETHER_ADDR_LEN);
2528 		off += RTE_ETHER_ADDR_LEN;
2529 	}
2530 
2531 	vnic->mc_addr_cnt = i;
2532 	if (vnic->mc_addr_cnt)
2533 		vnic->flags |= BNXT_VNIC_INFO_MCAST;
2534 	else
2535 		vnic->flags &= ~BNXT_VNIC_INFO_MCAST;
2536 
2537 allmulti:
2538 	return bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
2539 }
2540 
2541 static int
2542 bnxt_fw_version_get(struct rte_eth_dev *dev, char *fw_version, size_t fw_size)
2543 {
2544 	struct bnxt *bp = dev->data->dev_private;
2545 	uint8_t fw_major = (bp->fw_ver >> 24) & 0xff;
2546 	uint8_t fw_minor = (bp->fw_ver >> 16) & 0xff;
2547 	uint8_t fw_updt = (bp->fw_ver >> 8) & 0xff;
2548 	uint8_t fw_rsvd = bp->fw_ver & 0xff;
2549 	int ret;
2550 
2551 	ret = snprintf(fw_version, fw_size, "%d.%d.%d.%d",
2552 			fw_major, fw_minor, fw_updt, fw_rsvd);
2553 
2554 	ret += 1; /* add the size of '\0' */
2555 	if (fw_size < (uint32_t)ret)
2556 		return ret;
2557 	else
2558 		return 0;
2559 }
2560 
2561 static void
2562 bnxt_rxq_info_get_op(struct rte_eth_dev *dev, uint16_t queue_id,
2563 	struct rte_eth_rxq_info *qinfo)
2564 {
2565 	struct bnxt *bp = dev->data->dev_private;
2566 	struct bnxt_rx_queue *rxq;
2567 
2568 	if (is_bnxt_in_error(bp))
2569 		return;
2570 
2571 	rxq = dev->data->rx_queues[queue_id];
2572 
2573 	qinfo->mp = rxq->mb_pool;
2574 	qinfo->scattered_rx = dev->data->scattered_rx;
2575 	qinfo->nb_desc = rxq->nb_rx_desc;
2576 
2577 	qinfo->conf.rx_free_thresh = rxq->rx_free_thresh;
2578 	qinfo->conf.rx_drop_en = 0;
2579 	qinfo->conf.rx_deferred_start = rxq->rx_deferred_start;
2580 }
2581 
2582 static void
2583 bnxt_txq_info_get_op(struct rte_eth_dev *dev, uint16_t queue_id,
2584 	struct rte_eth_txq_info *qinfo)
2585 {
2586 	struct bnxt *bp = dev->data->dev_private;
2587 	struct bnxt_tx_queue *txq;
2588 
2589 	if (is_bnxt_in_error(bp))
2590 		return;
2591 
2592 	txq = dev->data->tx_queues[queue_id];
2593 
2594 	qinfo->nb_desc = txq->nb_tx_desc;
2595 
2596 	qinfo->conf.tx_thresh.pthresh = txq->pthresh;
2597 	qinfo->conf.tx_thresh.hthresh = txq->hthresh;
2598 	qinfo->conf.tx_thresh.wthresh = txq->wthresh;
2599 
2600 	qinfo->conf.tx_free_thresh = txq->tx_free_thresh;
2601 	qinfo->conf.tx_rs_thresh = 0;
2602 	qinfo->conf.tx_deferred_start = txq->tx_deferred_start;
2603 }
2604 
2605 static int
2606 bnxt_rx_burst_mode_get(struct rte_eth_dev *dev, __rte_unused uint16_t queue_id,
2607 		       struct rte_eth_burst_mode *mode)
2608 {
2609 	eth_rx_burst_t pkt_burst = dev->rx_pkt_burst;
2610 
2611 	if (pkt_burst == bnxt_recv_pkts) {
2612 		snprintf(mode->info, sizeof(mode->info), "%s",
2613 			 "Scalar");
2614 		return 0;
2615 	}
2616 #if defined(RTE_ARCH_X86) || defined(RTE_ARCH_ARM64)
2617 	if (pkt_burst == bnxt_recv_pkts_vec) {
2618 		snprintf(mode->info, sizeof(mode->info), "%s",
2619 			 "Vector SSE");
2620 		return 0;
2621 	}
2622 #endif
2623 
2624 	return -EINVAL;
2625 }
2626 
2627 static int
2628 bnxt_tx_burst_mode_get(struct rte_eth_dev *dev, __rte_unused uint16_t queue_id,
2629 		       struct rte_eth_burst_mode *mode)
2630 {
2631 	eth_tx_burst_t pkt_burst = dev->tx_pkt_burst;
2632 
2633 	if (pkt_burst == bnxt_xmit_pkts) {
2634 		snprintf(mode->info, sizeof(mode->info), "%s",
2635 			 "Scalar");
2636 		return 0;
2637 	}
2638 #if defined(RTE_ARCH_X86) || defined(RTE_ARCH_ARM64)
2639 	if (pkt_burst == bnxt_xmit_pkts_vec) {
2640 		snprintf(mode->info, sizeof(mode->info), "%s",
2641 			 "Vector SSE");
2642 		return 0;
2643 	}
2644 #endif
2645 
2646 	return -EINVAL;
2647 }
2648 
2649 int bnxt_mtu_set_op(struct rte_eth_dev *eth_dev, uint16_t new_mtu)
2650 {
2651 	struct bnxt *bp = eth_dev->data->dev_private;
2652 	uint32_t new_pkt_size;
2653 	uint32_t rc = 0;
2654 	uint32_t i;
2655 
2656 	rc = is_bnxt_in_error(bp);
2657 	if (rc)
2658 		return rc;
2659 
2660 	/* Exit if receive queues are not configured yet */
2661 	if (!eth_dev->data->nb_rx_queues)
2662 		return rc;
2663 
2664 	new_pkt_size = new_mtu + RTE_ETHER_HDR_LEN + RTE_ETHER_CRC_LEN +
2665 		       VLAN_TAG_SIZE * BNXT_NUM_VLANS;
2666 
2667 #if defined(RTE_ARCH_X86) || defined(RTE_ARCH_ARM64)
2668 	/*
2669 	 * If vector-mode tx/rx is active, disallow any MTU change that would
2670 	 * require scattered receive support.
2671 	 */
2672 	if (eth_dev->data->dev_started &&
2673 	    (eth_dev->rx_pkt_burst == bnxt_recv_pkts_vec ||
2674 	     eth_dev->tx_pkt_burst == bnxt_xmit_pkts_vec) &&
2675 	    (new_pkt_size >
2676 	     eth_dev->data->min_rx_buf_size - RTE_PKTMBUF_HEADROOM)) {
2677 		PMD_DRV_LOG(ERR,
2678 			    "MTU change would require scattered rx support. ");
2679 		PMD_DRV_LOG(ERR, "Stop port before changing MTU.\n");
2680 		return -EINVAL;
2681 	}
2682 #endif
2683 
2684 	if (new_mtu > RTE_ETHER_MTU) {
2685 		bp->flags |= BNXT_FLAG_JUMBO;
2686 		bp->eth_dev->data->dev_conf.rxmode.offloads |=
2687 			DEV_RX_OFFLOAD_JUMBO_FRAME;
2688 	} else {
2689 		bp->eth_dev->data->dev_conf.rxmode.offloads &=
2690 			~DEV_RX_OFFLOAD_JUMBO_FRAME;
2691 		bp->flags &= ~BNXT_FLAG_JUMBO;
2692 	}
2693 
2694 	/* Is there a change in mtu setting? */
2695 	if (eth_dev->data->dev_conf.rxmode.max_rx_pkt_len == new_pkt_size)
2696 		return rc;
2697 
2698 	for (i = 0; i < bp->nr_vnics; i++) {
2699 		struct bnxt_vnic_info *vnic = &bp->vnic_info[i];
2700 		uint16_t size = 0;
2701 
2702 		vnic->mru = BNXT_VNIC_MRU(new_mtu);
2703 		rc = bnxt_hwrm_vnic_cfg(bp, vnic);
2704 		if (rc)
2705 			break;
2706 
2707 		size = rte_pktmbuf_data_room_size(bp->rx_queues[0]->mb_pool);
2708 		size -= RTE_PKTMBUF_HEADROOM;
2709 
2710 		if (size < new_mtu) {
2711 			rc = bnxt_hwrm_vnic_plcmode_cfg(bp, vnic);
2712 			if (rc)
2713 				return rc;
2714 		}
2715 	}
2716 
2717 	if (!rc)
2718 		eth_dev->data->dev_conf.rxmode.max_rx_pkt_len = new_pkt_size;
2719 
2720 	PMD_DRV_LOG(INFO, "New MTU is %d\n", new_mtu);
2721 
2722 	return rc;
2723 }
2724 
2725 static int
2726 bnxt_vlan_pvid_set_op(struct rte_eth_dev *dev, uint16_t pvid, int on)
2727 {
2728 	struct bnxt *bp = dev->data->dev_private;
2729 	uint16_t vlan = bp->vlan;
2730 	int rc;
2731 
2732 	rc = is_bnxt_in_error(bp);
2733 	if (rc)
2734 		return rc;
2735 
2736 	if (!BNXT_SINGLE_PF(bp) || BNXT_VF(bp)) {
2737 		PMD_DRV_LOG(ERR,
2738 			"PVID cannot be modified for this function\n");
2739 		return -ENOTSUP;
2740 	}
2741 	bp->vlan = on ? pvid : 0;
2742 
2743 	rc = bnxt_hwrm_set_default_vlan(bp, 0, 0);
2744 	if (rc)
2745 		bp->vlan = vlan;
2746 	return rc;
2747 }
2748 
2749 static int
2750 bnxt_dev_led_on_op(struct rte_eth_dev *dev)
2751 {
2752 	struct bnxt *bp = dev->data->dev_private;
2753 	int rc;
2754 
2755 	rc = is_bnxt_in_error(bp);
2756 	if (rc)
2757 		return rc;
2758 
2759 	return bnxt_hwrm_port_led_cfg(bp, true);
2760 }
2761 
2762 static int
2763 bnxt_dev_led_off_op(struct rte_eth_dev *dev)
2764 {
2765 	struct bnxt *bp = dev->data->dev_private;
2766 	int rc;
2767 
2768 	rc = is_bnxt_in_error(bp);
2769 	if (rc)
2770 		return rc;
2771 
2772 	return bnxt_hwrm_port_led_cfg(bp, false);
2773 }
2774 
2775 static uint32_t
2776 bnxt_rx_queue_count_op(struct rte_eth_dev *dev, uint16_t rx_queue_id)
2777 {
2778 	struct bnxt *bp = (struct bnxt *)dev->data->dev_private;
2779 	uint32_t desc = 0, raw_cons = 0, cons;
2780 	struct bnxt_cp_ring_info *cpr;
2781 	struct bnxt_rx_queue *rxq;
2782 	struct rx_pkt_cmpl *rxcmp;
2783 	int rc;
2784 
2785 	rc = is_bnxt_in_error(bp);
2786 	if (rc)
2787 		return rc;
2788 
2789 	rxq = dev->data->rx_queues[rx_queue_id];
2790 	cpr = rxq->cp_ring;
2791 	raw_cons = cpr->cp_raw_cons;
2792 
2793 	while (1) {
2794 		cons = RING_CMP(cpr->cp_ring_struct, raw_cons);
2795 		rte_prefetch0(&cpr->cp_desc_ring[cons]);
2796 		rxcmp = (struct rx_pkt_cmpl *)&cpr->cp_desc_ring[cons];
2797 
2798 		if (!CMP_VALID(rxcmp, raw_cons, cpr->cp_ring_struct)) {
2799 			break;
2800 		} else {
2801 			raw_cons++;
2802 			desc++;
2803 		}
2804 	}
2805 
2806 	return desc;
2807 }
2808 
2809 static int
2810 bnxt_rx_descriptor_status_op(void *rx_queue, uint16_t offset)
2811 {
2812 	struct bnxt_rx_queue *rxq = (struct bnxt_rx_queue *)rx_queue;
2813 	struct bnxt_rx_ring_info *rxr;
2814 	struct bnxt_cp_ring_info *cpr;
2815 	struct bnxt_sw_rx_bd *rx_buf;
2816 	struct rx_pkt_cmpl *rxcmp;
2817 	uint32_t cons, cp_cons;
2818 	int rc;
2819 
2820 	if (!rxq)
2821 		return -EINVAL;
2822 
2823 	rc = is_bnxt_in_error(rxq->bp);
2824 	if (rc)
2825 		return rc;
2826 
2827 	cpr = rxq->cp_ring;
2828 	rxr = rxq->rx_ring;
2829 
2830 	if (offset >= rxq->nb_rx_desc)
2831 		return -EINVAL;
2832 
2833 	cons = RING_CMP(cpr->cp_ring_struct, offset);
2834 	cp_cons = cpr->cp_raw_cons;
2835 	rxcmp = (struct rx_pkt_cmpl *)&cpr->cp_desc_ring[cons];
2836 
2837 	if (cons > cp_cons) {
2838 		if (CMPL_VALID(rxcmp, cpr->valid))
2839 			return RTE_ETH_RX_DESC_DONE;
2840 	} else {
2841 		if (CMPL_VALID(rxcmp, !cpr->valid))
2842 			return RTE_ETH_RX_DESC_DONE;
2843 	}
2844 	rx_buf = &rxr->rx_buf_ring[cons];
2845 	if (rx_buf->mbuf == NULL)
2846 		return RTE_ETH_RX_DESC_UNAVAIL;
2847 
2848 
2849 	return RTE_ETH_RX_DESC_AVAIL;
2850 }
2851 
2852 static int
2853 bnxt_tx_descriptor_status_op(void *tx_queue, uint16_t offset)
2854 {
2855 	struct bnxt_tx_queue *txq = (struct bnxt_tx_queue *)tx_queue;
2856 	struct bnxt_tx_ring_info *txr;
2857 	struct bnxt_cp_ring_info *cpr;
2858 	struct bnxt_sw_tx_bd *tx_buf;
2859 	struct tx_pkt_cmpl *txcmp;
2860 	uint32_t cons, cp_cons;
2861 	int rc;
2862 
2863 	if (!txq)
2864 		return -EINVAL;
2865 
2866 	rc = is_bnxt_in_error(txq->bp);
2867 	if (rc)
2868 		return rc;
2869 
2870 	cpr = txq->cp_ring;
2871 	txr = txq->tx_ring;
2872 
2873 	if (offset >= txq->nb_tx_desc)
2874 		return -EINVAL;
2875 
2876 	cons = RING_CMP(cpr->cp_ring_struct, offset);
2877 	txcmp = (struct tx_pkt_cmpl *)&cpr->cp_desc_ring[cons];
2878 	cp_cons = cpr->cp_raw_cons;
2879 
2880 	if (cons > cp_cons) {
2881 		if (CMPL_VALID(txcmp, cpr->valid))
2882 			return RTE_ETH_TX_DESC_UNAVAIL;
2883 	} else {
2884 		if (CMPL_VALID(txcmp, !cpr->valid))
2885 			return RTE_ETH_TX_DESC_UNAVAIL;
2886 	}
2887 	tx_buf = &txr->tx_buf_ring[cons];
2888 	if (tx_buf->mbuf == NULL)
2889 		return RTE_ETH_TX_DESC_DONE;
2890 
2891 	return RTE_ETH_TX_DESC_FULL;
2892 }
2893 
2894 static struct bnxt_filter_info *
2895 bnxt_match_and_validate_ether_filter(struct bnxt *bp,
2896 				struct rte_eth_ethertype_filter *efilter,
2897 				struct bnxt_vnic_info *vnic0,
2898 				struct bnxt_vnic_info *vnic,
2899 				int *ret)
2900 {
2901 	struct bnxt_filter_info *mfilter = NULL;
2902 	int match = 0;
2903 	*ret = 0;
2904 
2905 	if (efilter->ether_type == RTE_ETHER_TYPE_IPV4 ||
2906 		efilter->ether_type == RTE_ETHER_TYPE_IPV6) {
2907 		PMD_DRV_LOG(ERR, "invalid ether_type(0x%04x) in"
2908 			" ethertype filter.", efilter->ether_type);
2909 		*ret = -EINVAL;
2910 		goto exit;
2911 	}
2912 	if (efilter->queue >= bp->rx_nr_rings) {
2913 		PMD_DRV_LOG(ERR, "Invalid queue %d\n", efilter->queue);
2914 		*ret = -EINVAL;
2915 		goto exit;
2916 	}
2917 
2918 	vnic0 = BNXT_GET_DEFAULT_VNIC(bp);
2919 	vnic = &bp->vnic_info[efilter->queue];
2920 	if (vnic == NULL) {
2921 		PMD_DRV_LOG(ERR, "Invalid queue %d\n", efilter->queue);
2922 		*ret = -EINVAL;
2923 		goto exit;
2924 	}
2925 
2926 	if (efilter->flags & RTE_ETHTYPE_FLAGS_DROP) {
2927 		STAILQ_FOREACH(mfilter, &vnic0->filter, next) {
2928 			if ((!memcmp(efilter->mac_addr.addr_bytes,
2929 				     mfilter->l2_addr, RTE_ETHER_ADDR_LEN) &&
2930 			     mfilter->flags ==
2931 			     HWRM_CFA_NTUPLE_FILTER_ALLOC_INPUT_FLAGS_DROP &&
2932 			     mfilter->ethertype == efilter->ether_type)) {
2933 				match = 1;
2934 				break;
2935 			}
2936 		}
2937 	} else {
2938 		STAILQ_FOREACH(mfilter, &vnic->filter, next)
2939 			if ((!memcmp(efilter->mac_addr.addr_bytes,
2940 				     mfilter->l2_addr, RTE_ETHER_ADDR_LEN) &&
2941 			     mfilter->ethertype == efilter->ether_type &&
2942 			     mfilter->flags ==
2943 			     HWRM_CFA_L2_FILTER_CFG_INPUT_FLAGS_PATH_RX)) {
2944 				match = 1;
2945 				break;
2946 			}
2947 	}
2948 
2949 	if (match)
2950 		*ret = -EEXIST;
2951 
2952 exit:
2953 	return mfilter;
2954 }
2955 
2956 static int
2957 bnxt_ethertype_filter(struct rte_eth_dev *dev,
2958 			enum rte_filter_op filter_op,
2959 			void *arg)
2960 {
2961 	struct bnxt *bp = dev->data->dev_private;
2962 	struct rte_eth_ethertype_filter *efilter =
2963 			(struct rte_eth_ethertype_filter *)arg;
2964 	struct bnxt_filter_info *bfilter, *filter1;
2965 	struct bnxt_vnic_info *vnic, *vnic0;
2966 	int ret;
2967 
2968 	if (filter_op == RTE_ETH_FILTER_NOP)
2969 		return 0;
2970 
2971 	if (arg == NULL) {
2972 		PMD_DRV_LOG(ERR, "arg shouldn't be NULL for operation %u.",
2973 			    filter_op);
2974 		return -EINVAL;
2975 	}
2976 
2977 	vnic0 = BNXT_GET_DEFAULT_VNIC(bp);
2978 	vnic = &bp->vnic_info[efilter->queue];
2979 
2980 	switch (filter_op) {
2981 	case RTE_ETH_FILTER_ADD:
2982 		bnxt_match_and_validate_ether_filter(bp, efilter,
2983 							vnic0, vnic, &ret);
2984 		if (ret < 0)
2985 			return ret;
2986 
2987 		bfilter = bnxt_get_unused_filter(bp);
2988 		if (bfilter == NULL) {
2989 			PMD_DRV_LOG(ERR,
2990 				"Not enough resources for a new filter.\n");
2991 			return -ENOMEM;
2992 		}
2993 		bfilter->filter_type = HWRM_CFA_NTUPLE_FILTER;
2994 		memcpy(bfilter->l2_addr, efilter->mac_addr.addr_bytes,
2995 		       RTE_ETHER_ADDR_LEN);
2996 		memcpy(bfilter->dst_macaddr, efilter->mac_addr.addr_bytes,
2997 		       RTE_ETHER_ADDR_LEN);
2998 		bfilter->enables |= NTUPLE_FLTR_ALLOC_INPUT_EN_DST_MACADDR;
2999 		bfilter->ethertype = efilter->ether_type;
3000 		bfilter->enables |= NTUPLE_FLTR_ALLOC_INPUT_EN_ETHERTYPE;
3001 
3002 		filter1 = bnxt_get_l2_filter(bp, bfilter, vnic0);
3003 		if (filter1 == NULL) {
3004 			ret = -EINVAL;
3005 			goto cleanup;
3006 		}
3007 		bfilter->enables |=
3008 			HWRM_CFA_NTUPLE_FILTER_ALLOC_INPUT_ENABLES_L2_FILTER_ID;
3009 		bfilter->fw_l2_filter_id = filter1->fw_l2_filter_id;
3010 
3011 		bfilter->dst_id = vnic->fw_vnic_id;
3012 
3013 		if (efilter->flags & RTE_ETHTYPE_FLAGS_DROP) {
3014 			bfilter->flags =
3015 				HWRM_CFA_NTUPLE_FILTER_ALLOC_INPUT_FLAGS_DROP;
3016 		}
3017 
3018 		ret = bnxt_hwrm_set_ntuple_filter(bp, bfilter->dst_id, bfilter);
3019 		if (ret)
3020 			goto cleanup;
3021 		STAILQ_INSERT_TAIL(&vnic->filter, bfilter, next);
3022 		break;
3023 	case RTE_ETH_FILTER_DELETE:
3024 		filter1 = bnxt_match_and_validate_ether_filter(bp, efilter,
3025 							vnic0, vnic, &ret);
3026 		if (ret == -EEXIST) {
3027 			ret = bnxt_hwrm_clear_ntuple_filter(bp, filter1);
3028 
3029 			STAILQ_REMOVE(&vnic->filter, filter1, bnxt_filter_info,
3030 				      next);
3031 			bnxt_free_filter(bp, filter1);
3032 		} else if (ret == 0) {
3033 			PMD_DRV_LOG(ERR, "No matching filter found\n");
3034 		}
3035 		break;
3036 	default:
3037 		PMD_DRV_LOG(ERR, "unsupported operation %u.", filter_op);
3038 		ret = -EINVAL;
3039 		goto error;
3040 	}
3041 	return ret;
3042 cleanup:
3043 	bnxt_free_filter(bp, bfilter);
3044 error:
3045 	return ret;
3046 }
3047 
3048 static inline int
3049 parse_ntuple_filter(struct bnxt *bp,
3050 		    struct rte_eth_ntuple_filter *nfilter,
3051 		    struct bnxt_filter_info *bfilter)
3052 {
3053 	uint32_t en = 0;
3054 
3055 	if (nfilter->queue >= bp->rx_nr_rings) {
3056 		PMD_DRV_LOG(ERR, "Invalid queue %d\n", nfilter->queue);
3057 		return -EINVAL;
3058 	}
3059 
3060 	switch (nfilter->dst_port_mask) {
3061 	case UINT16_MAX:
3062 		bfilter->dst_port_mask = -1;
3063 		bfilter->dst_port = nfilter->dst_port;
3064 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_DST_PORT |
3065 			NTUPLE_FLTR_ALLOC_INPUT_EN_DST_PORT_MASK;
3066 		break;
3067 	default:
3068 		PMD_DRV_LOG(ERR, "invalid dst_port mask.");
3069 		return -EINVAL;
3070 	}
3071 
3072 	bfilter->ip_addr_type = NTUPLE_FLTR_ALLOC_INPUT_IP_ADDR_TYPE_IPV4;
3073 	en |= NTUPLE_FLTR_ALLOC_IN_EN_IP_PROTO;
3074 
3075 	switch (nfilter->proto_mask) {
3076 	case UINT8_MAX:
3077 		if (nfilter->proto == 17) /* IPPROTO_UDP */
3078 			bfilter->ip_protocol = 17;
3079 		else if (nfilter->proto == 6) /* IPPROTO_TCP */
3080 			bfilter->ip_protocol = 6;
3081 		else
3082 			return -EINVAL;
3083 		en |= NTUPLE_FLTR_ALLOC_IN_EN_IP_PROTO;
3084 		break;
3085 	default:
3086 		PMD_DRV_LOG(ERR, "invalid protocol mask.");
3087 		return -EINVAL;
3088 	}
3089 
3090 	switch (nfilter->dst_ip_mask) {
3091 	case UINT32_MAX:
3092 		bfilter->dst_ipaddr_mask[0] = -1;
3093 		bfilter->dst_ipaddr[0] = nfilter->dst_ip;
3094 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_DST_IPADDR |
3095 			NTUPLE_FLTR_ALLOC_INPUT_EN_DST_IPADDR_MASK;
3096 		break;
3097 	default:
3098 		PMD_DRV_LOG(ERR, "invalid dst_ip mask.");
3099 		return -EINVAL;
3100 	}
3101 
3102 	switch (nfilter->src_ip_mask) {
3103 	case UINT32_MAX:
3104 		bfilter->src_ipaddr_mask[0] = -1;
3105 		bfilter->src_ipaddr[0] = nfilter->src_ip;
3106 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_IPADDR |
3107 			NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_IPADDR_MASK;
3108 		break;
3109 	default:
3110 		PMD_DRV_LOG(ERR, "invalid src_ip mask.");
3111 		return -EINVAL;
3112 	}
3113 
3114 	switch (nfilter->src_port_mask) {
3115 	case UINT16_MAX:
3116 		bfilter->src_port_mask = -1;
3117 		bfilter->src_port = nfilter->src_port;
3118 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_PORT |
3119 			NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_PORT_MASK;
3120 		break;
3121 	default:
3122 		PMD_DRV_LOG(ERR, "invalid src_port mask.");
3123 		return -EINVAL;
3124 	}
3125 
3126 	bfilter->enables = en;
3127 	return 0;
3128 }
3129 
3130 static struct bnxt_filter_info*
3131 bnxt_match_ntuple_filter(struct bnxt *bp,
3132 			 struct bnxt_filter_info *bfilter,
3133 			 struct bnxt_vnic_info **mvnic)
3134 {
3135 	struct bnxt_filter_info *mfilter = NULL;
3136 	int i;
3137 
3138 	for (i = bp->nr_vnics - 1; i >= 0; i--) {
3139 		struct bnxt_vnic_info *vnic = &bp->vnic_info[i];
3140 		STAILQ_FOREACH(mfilter, &vnic->filter, next) {
3141 			if (bfilter->src_ipaddr[0] == mfilter->src_ipaddr[0] &&
3142 			    bfilter->src_ipaddr_mask[0] ==
3143 			    mfilter->src_ipaddr_mask[0] &&
3144 			    bfilter->src_port == mfilter->src_port &&
3145 			    bfilter->src_port_mask == mfilter->src_port_mask &&
3146 			    bfilter->dst_ipaddr[0] == mfilter->dst_ipaddr[0] &&
3147 			    bfilter->dst_ipaddr_mask[0] ==
3148 			    mfilter->dst_ipaddr_mask[0] &&
3149 			    bfilter->dst_port == mfilter->dst_port &&
3150 			    bfilter->dst_port_mask == mfilter->dst_port_mask &&
3151 			    bfilter->flags == mfilter->flags &&
3152 			    bfilter->enables == mfilter->enables) {
3153 				if (mvnic)
3154 					*mvnic = vnic;
3155 				return mfilter;
3156 			}
3157 		}
3158 	}
3159 	return NULL;
3160 }
3161 
3162 static int
3163 bnxt_cfg_ntuple_filter(struct bnxt *bp,
3164 		       struct rte_eth_ntuple_filter *nfilter,
3165 		       enum rte_filter_op filter_op)
3166 {
3167 	struct bnxt_filter_info *bfilter, *mfilter, *filter1;
3168 	struct bnxt_vnic_info *vnic, *vnic0, *mvnic;
3169 	int ret;
3170 
3171 	if (nfilter->flags != RTE_5TUPLE_FLAGS) {
3172 		PMD_DRV_LOG(ERR, "only 5tuple is supported.");
3173 		return -EINVAL;
3174 	}
3175 
3176 	if (nfilter->flags & RTE_NTUPLE_FLAGS_TCP_FLAG) {
3177 		PMD_DRV_LOG(ERR, "Ntuple filter: TCP flags not supported\n");
3178 		return -EINVAL;
3179 	}
3180 
3181 	bfilter = bnxt_get_unused_filter(bp);
3182 	if (bfilter == NULL) {
3183 		PMD_DRV_LOG(ERR,
3184 			"Not enough resources for a new filter.\n");
3185 		return -ENOMEM;
3186 	}
3187 	ret = parse_ntuple_filter(bp, nfilter, bfilter);
3188 	if (ret < 0)
3189 		goto free_filter;
3190 
3191 	vnic = &bp->vnic_info[nfilter->queue];
3192 	vnic0 = BNXT_GET_DEFAULT_VNIC(bp);
3193 	filter1 = STAILQ_FIRST(&vnic0->filter);
3194 	if (filter1 == NULL) {
3195 		ret = -EINVAL;
3196 		goto free_filter;
3197 	}
3198 
3199 	bfilter->dst_id = vnic->fw_vnic_id;
3200 	bfilter->fw_l2_filter_id = filter1->fw_l2_filter_id;
3201 	bfilter->enables |=
3202 		HWRM_CFA_NTUPLE_FILTER_ALLOC_INPUT_ENABLES_L2_FILTER_ID;
3203 	bfilter->ethertype = 0x800;
3204 	bfilter->enables |= NTUPLE_FLTR_ALLOC_INPUT_EN_ETHERTYPE;
3205 
3206 	mfilter = bnxt_match_ntuple_filter(bp, bfilter, &mvnic);
3207 
3208 	if (mfilter != NULL && filter_op == RTE_ETH_FILTER_ADD &&
3209 	    bfilter->dst_id == mfilter->dst_id) {
3210 		PMD_DRV_LOG(ERR, "filter exists.\n");
3211 		ret = -EEXIST;
3212 		goto free_filter;
3213 	} else if (mfilter != NULL && filter_op == RTE_ETH_FILTER_ADD &&
3214 		   bfilter->dst_id != mfilter->dst_id) {
3215 		mfilter->dst_id = vnic->fw_vnic_id;
3216 		ret = bnxt_hwrm_set_ntuple_filter(bp, mfilter->dst_id, mfilter);
3217 		STAILQ_REMOVE(&mvnic->filter, mfilter, bnxt_filter_info, next);
3218 		STAILQ_INSERT_TAIL(&vnic->filter, mfilter, next);
3219 		PMD_DRV_LOG(ERR, "filter with matching pattern exists.\n");
3220 		PMD_DRV_LOG(ERR, " Updated it to the new destination queue\n");
3221 		goto free_filter;
3222 	}
3223 	if (mfilter == NULL && filter_op == RTE_ETH_FILTER_DELETE) {
3224 		PMD_DRV_LOG(ERR, "filter doesn't exist.");
3225 		ret = -ENOENT;
3226 		goto free_filter;
3227 	}
3228 
3229 	if (filter_op == RTE_ETH_FILTER_ADD) {
3230 		bfilter->filter_type = HWRM_CFA_NTUPLE_FILTER;
3231 		ret = bnxt_hwrm_set_ntuple_filter(bp, bfilter->dst_id, bfilter);
3232 		if (ret)
3233 			goto free_filter;
3234 		STAILQ_INSERT_TAIL(&vnic->filter, bfilter, next);
3235 	} else {
3236 		if (mfilter == NULL) {
3237 			/* This should not happen. But for Coverity! */
3238 			ret = -ENOENT;
3239 			goto free_filter;
3240 		}
3241 		ret = bnxt_hwrm_clear_ntuple_filter(bp, mfilter);
3242 
3243 		STAILQ_REMOVE(&vnic->filter, mfilter, bnxt_filter_info, next);
3244 		bnxt_free_filter(bp, mfilter);
3245 		bnxt_free_filter(bp, bfilter);
3246 	}
3247 
3248 	return 0;
3249 free_filter:
3250 	bnxt_free_filter(bp, bfilter);
3251 	return ret;
3252 }
3253 
3254 static int
3255 bnxt_ntuple_filter(struct rte_eth_dev *dev,
3256 			enum rte_filter_op filter_op,
3257 			void *arg)
3258 {
3259 	struct bnxt *bp = dev->data->dev_private;
3260 	int ret;
3261 
3262 	if (filter_op == RTE_ETH_FILTER_NOP)
3263 		return 0;
3264 
3265 	if (arg == NULL) {
3266 		PMD_DRV_LOG(ERR, "arg shouldn't be NULL for operation %u.",
3267 			    filter_op);
3268 		return -EINVAL;
3269 	}
3270 
3271 	switch (filter_op) {
3272 	case RTE_ETH_FILTER_ADD:
3273 		ret = bnxt_cfg_ntuple_filter(bp,
3274 			(struct rte_eth_ntuple_filter *)arg,
3275 			filter_op);
3276 		break;
3277 	case RTE_ETH_FILTER_DELETE:
3278 		ret = bnxt_cfg_ntuple_filter(bp,
3279 			(struct rte_eth_ntuple_filter *)arg,
3280 			filter_op);
3281 		break;
3282 	default:
3283 		PMD_DRV_LOG(ERR, "unsupported operation %u.", filter_op);
3284 		ret = -EINVAL;
3285 		break;
3286 	}
3287 	return ret;
3288 }
3289 
3290 static int
3291 bnxt_parse_fdir_filter(struct bnxt *bp,
3292 		       struct rte_eth_fdir_filter *fdir,
3293 		       struct bnxt_filter_info *filter)
3294 {
3295 	enum rte_fdir_mode fdir_mode =
3296 		bp->eth_dev->data->dev_conf.fdir_conf.mode;
3297 	struct bnxt_vnic_info *vnic0, *vnic;
3298 	struct bnxt_filter_info *filter1;
3299 	uint32_t en = 0;
3300 	int i;
3301 
3302 	if (fdir_mode == RTE_FDIR_MODE_PERFECT_TUNNEL)
3303 		return -EINVAL;
3304 
3305 	filter->l2_ovlan = fdir->input.flow_ext.vlan_tci;
3306 	en |= EM_FLOW_ALLOC_INPUT_EN_OVLAN_VID;
3307 
3308 	switch (fdir->input.flow_type) {
3309 	case RTE_ETH_FLOW_IPV4:
3310 	case RTE_ETH_FLOW_NONFRAG_IPV4_OTHER:
3311 		/* FALLTHROUGH */
3312 		filter->src_ipaddr[0] = fdir->input.flow.ip4_flow.src_ip;
3313 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_IPADDR;
3314 		filter->dst_ipaddr[0] = fdir->input.flow.ip4_flow.dst_ip;
3315 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_DST_IPADDR;
3316 		filter->ip_protocol = fdir->input.flow.ip4_flow.proto;
3317 		en |= NTUPLE_FLTR_ALLOC_IN_EN_IP_PROTO;
3318 		filter->ip_addr_type =
3319 			NTUPLE_FLTR_ALLOC_INPUT_IP_ADDR_TYPE_IPV4;
3320 		filter->src_ipaddr_mask[0] = 0xffffffff;
3321 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_IPADDR_MASK;
3322 		filter->dst_ipaddr_mask[0] = 0xffffffff;
3323 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_DST_IPADDR_MASK;
3324 		filter->ethertype = 0x800;
3325 		filter->enables |= NTUPLE_FLTR_ALLOC_INPUT_EN_ETHERTYPE;
3326 		break;
3327 	case RTE_ETH_FLOW_NONFRAG_IPV4_TCP:
3328 		filter->src_port = fdir->input.flow.tcp4_flow.src_port;
3329 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_PORT;
3330 		filter->dst_port = fdir->input.flow.tcp4_flow.dst_port;
3331 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_DST_PORT;
3332 		filter->dst_port_mask = 0xffff;
3333 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_DST_PORT_MASK;
3334 		filter->src_port_mask = 0xffff;
3335 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_PORT_MASK;
3336 		filter->src_ipaddr[0] = fdir->input.flow.tcp4_flow.ip.src_ip;
3337 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_IPADDR;
3338 		filter->dst_ipaddr[0] = fdir->input.flow.tcp4_flow.ip.dst_ip;
3339 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_DST_IPADDR;
3340 		filter->ip_protocol = 6;
3341 		en |= NTUPLE_FLTR_ALLOC_IN_EN_IP_PROTO;
3342 		filter->ip_addr_type =
3343 			NTUPLE_FLTR_ALLOC_INPUT_IP_ADDR_TYPE_IPV4;
3344 		filter->src_ipaddr_mask[0] = 0xffffffff;
3345 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_IPADDR_MASK;
3346 		filter->dst_ipaddr_mask[0] = 0xffffffff;
3347 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_DST_IPADDR_MASK;
3348 		filter->ethertype = 0x800;
3349 		filter->enables |= NTUPLE_FLTR_ALLOC_INPUT_EN_ETHERTYPE;
3350 		break;
3351 	case RTE_ETH_FLOW_NONFRAG_IPV4_UDP:
3352 		filter->src_port = fdir->input.flow.udp4_flow.src_port;
3353 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_PORT;
3354 		filter->dst_port = fdir->input.flow.udp4_flow.dst_port;
3355 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_DST_PORT;
3356 		filter->dst_port_mask = 0xffff;
3357 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_DST_PORT_MASK;
3358 		filter->src_port_mask = 0xffff;
3359 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_PORT_MASK;
3360 		filter->src_ipaddr[0] = fdir->input.flow.udp4_flow.ip.src_ip;
3361 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_IPADDR;
3362 		filter->dst_ipaddr[0] = fdir->input.flow.udp4_flow.ip.dst_ip;
3363 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_DST_IPADDR;
3364 		filter->ip_protocol = 17;
3365 		en |= NTUPLE_FLTR_ALLOC_IN_EN_IP_PROTO;
3366 		filter->ip_addr_type =
3367 			NTUPLE_FLTR_ALLOC_INPUT_IP_ADDR_TYPE_IPV4;
3368 		filter->src_ipaddr_mask[0] = 0xffffffff;
3369 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_IPADDR_MASK;
3370 		filter->dst_ipaddr_mask[0] = 0xffffffff;
3371 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_DST_IPADDR_MASK;
3372 		filter->ethertype = 0x800;
3373 		filter->enables |= NTUPLE_FLTR_ALLOC_INPUT_EN_ETHERTYPE;
3374 		break;
3375 	case RTE_ETH_FLOW_IPV6:
3376 	case RTE_ETH_FLOW_NONFRAG_IPV6_OTHER:
3377 		/* FALLTHROUGH */
3378 		filter->ip_addr_type =
3379 			NTUPLE_FLTR_ALLOC_INPUT_IP_ADDR_TYPE_IPV6;
3380 		filter->ip_protocol = fdir->input.flow.ipv6_flow.proto;
3381 		en |= NTUPLE_FLTR_ALLOC_IN_EN_IP_PROTO;
3382 		rte_memcpy(filter->src_ipaddr,
3383 			   fdir->input.flow.ipv6_flow.src_ip, 16);
3384 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_IPADDR;
3385 		rte_memcpy(filter->dst_ipaddr,
3386 			   fdir->input.flow.ipv6_flow.dst_ip, 16);
3387 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_DST_IPADDR;
3388 		memset(filter->dst_ipaddr_mask, 0xff, 16);
3389 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_DST_IPADDR_MASK;
3390 		memset(filter->src_ipaddr_mask, 0xff, 16);
3391 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_IPADDR_MASK;
3392 		filter->ethertype = 0x86dd;
3393 		filter->enables |= NTUPLE_FLTR_ALLOC_INPUT_EN_ETHERTYPE;
3394 		break;
3395 	case RTE_ETH_FLOW_NONFRAG_IPV6_TCP:
3396 		filter->src_port = fdir->input.flow.tcp6_flow.src_port;
3397 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_PORT;
3398 		filter->dst_port = fdir->input.flow.tcp6_flow.dst_port;
3399 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_DST_PORT;
3400 		filter->dst_port_mask = 0xffff;
3401 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_DST_PORT_MASK;
3402 		filter->src_port_mask = 0xffff;
3403 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_PORT_MASK;
3404 		filter->ip_addr_type =
3405 			NTUPLE_FLTR_ALLOC_INPUT_IP_ADDR_TYPE_IPV6;
3406 		filter->ip_protocol = fdir->input.flow.tcp6_flow.ip.proto;
3407 		en |= NTUPLE_FLTR_ALLOC_IN_EN_IP_PROTO;
3408 		rte_memcpy(filter->src_ipaddr,
3409 			   fdir->input.flow.tcp6_flow.ip.src_ip, 16);
3410 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_IPADDR;
3411 		rte_memcpy(filter->dst_ipaddr,
3412 			   fdir->input.flow.tcp6_flow.ip.dst_ip, 16);
3413 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_DST_IPADDR;
3414 		memset(filter->dst_ipaddr_mask, 0xff, 16);
3415 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_DST_IPADDR_MASK;
3416 		memset(filter->src_ipaddr_mask, 0xff, 16);
3417 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_IPADDR_MASK;
3418 		filter->ethertype = 0x86dd;
3419 		filter->enables |= NTUPLE_FLTR_ALLOC_INPUT_EN_ETHERTYPE;
3420 		break;
3421 	case RTE_ETH_FLOW_NONFRAG_IPV6_UDP:
3422 		filter->src_port = fdir->input.flow.udp6_flow.src_port;
3423 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_PORT;
3424 		filter->dst_port = fdir->input.flow.udp6_flow.dst_port;
3425 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_DST_PORT;
3426 		filter->dst_port_mask = 0xffff;
3427 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_DST_PORT_MASK;
3428 		filter->src_port_mask = 0xffff;
3429 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_PORT_MASK;
3430 		filter->ip_addr_type =
3431 			NTUPLE_FLTR_ALLOC_INPUT_IP_ADDR_TYPE_IPV6;
3432 		filter->ip_protocol = fdir->input.flow.udp6_flow.ip.proto;
3433 		en |= NTUPLE_FLTR_ALLOC_IN_EN_IP_PROTO;
3434 		rte_memcpy(filter->src_ipaddr,
3435 			   fdir->input.flow.udp6_flow.ip.src_ip, 16);
3436 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_IPADDR;
3437 		rte_memcpy(filter->dst_ipaddr,
3438 			   fdir->input.flow.udp6_flow.ip.dst_ip, 16);
3439 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_DST_IPADDR;
3440 		memset(filter->dst_ipaddr_mask, 0xff, 16);
3441 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_DST_IPADDR_MASK;
3442 		memset(filter->src_ipaddr_mask, 0xff, 16);
3443 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_IPADDR_MASK;
3444 		filter->ethertype = 0x86dd;
3445 		filter->enables |= NTUPLE_FLTR_ALLOC_INPUT_EN_ETHERTYPE;
3446 		break;
3447 	case RTE_ETH_FLOW_L2_PAYLOAD:
3448 		filter->ethertype = fdir->input.flow.l2_flow.ether_type;
3449 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_ETHERTYPE;
3450 		break;
3451 	case RTE_ETH_FLOW_VXLAN:
3452 		if (fdir->action.behavior == RTE_ETH_FDIR_REJECT)
3453 			return -EINVAL;
3454 		filter->vni = fdir->input.flow.tunnel_flow.tunnel_id;
3455 		filter->tunnel_type =
3456 			CFA_NTUPLE_FILTER_ALLOC_REQ_TUNNEL_TYPE_VXLAN;
3457 		en |= HWRM_CFA_NTUPLE_FILTER_ALLOC_INPUT_ENABLES_TUNNEL_TYPE;
3458 		break;
3459 	case RTE_ETH_FLOW_NVGRE:
3460 		if (fdir->action.behavior == RTE_ETH_FDIR_REJECT)
3461 			return -EINVAL;
3462 		filter->vni = fdir->input.flow.tunnel_flow.tunnel_id;
3463 		filter->tunnel_type =
3464 			CFA_NTUPLE_FILTER_ALLOC_REQ_TUNNEL_TYPE_NVGRE;
3465 		en |= HWRM_CFA_NTUPLE_FILTER_ALLOC_INPUT_ENABLES_TUNNEL_TYPE;
3466 		break;
3467 	case RTE_ETH_FLOW_UNKNOWN:
3468 	case RTE_ETH_FLOW_RAW:
3469 	case RTE_ETH_FLOW_FRAG_IPV4:
3470 	case RTE_ETH_FLOW_NONFRAG_IPV4_SCTP:
3471 	case RTE_ETH_FLOW_FRAG_IPV6:
3472 	case RTE_ETH_FLOW_NONFRAG_IPV6_SCTP:
3473 	case RTE_ETH_FLOW_IPV6_EX:
3474 	case RTE_ETH_FLOW_IPV6_TCP_EX:
3475 	case RTE_ETH_FLOW_IPV6_UDP_EX:
3476 	case RTE_ETH_FLOW_GENEVE:
3477 		/* FALLTHROUGH */
3478 	default:
3479 		return -EINVAL;
3480 	}
3481 
3482 	vnic0 = BNXT_GET_DEFAULT_VNIC(bp);
3483 	vnic = &bp->vnic_info[fdir->action.rx_queue];
3484 	if (vnic == NULL) {
3485 		PMD_DRV_LOG(ERR, "Invalid queue %d\n", fdir->action.rx_queue);
3486 		return -EINVAL;
3487 	}
3488 
3489 	if (fdir_mode == RTE_FDIR_MODE_PERFECT_MAC_VLAN) {
3490 		rte_memcpy(filter->dst_macaddr,
3491 			fdir->input.flow.mac_vlan_flow.mac_addr.addr_bytes, 6);
3492 			en |= NTUPLE_FLTR_ALLOC_INPUT_EN_DST_MACADDR;
3493 	}
3494 
3495 	if (fdir->action.behavior == RTE_ETH_FDIR_REJECT) {
3496 		filter->flags = HWRM_CFA_NTUPLE_FILTER_ALLOC_INPUT_FLAGS_DROP;
3497 		filter1 = STAILQ_FIRST(&vnic0->filter);
3498 		//filter1 = bnxt_get_l2_filter(bp, filter, vnic0);
3499 	} else {
3500 		filter->dst_id = vnic->fw_vnic_id;
3501 		for (i = 0; i < RTE_ETHER_ADDR_LEN; i++)
3502 			if (filter->dst_macaddr[i] == 0x00)
3503 				filter1 = STAILQ_FIRST(&vnic0->filter);
3504 			else
3505 				filter1 = bnxt_get_l2_filter(bp, filter, vnic);
3506 	}
3507 
3508 	if (filter1 == NULL)
3509 		return -EINVAL;
3510 
3511 	en |= HWRM_CFA_NTUPLE_FILTER_ALLOC_INPUT_ENABLES_L2_FILTER_ID;
3512 	filter->fw_l2_filter_id = filter1->fw_l2_filter_id;
3513 
3514 	filter->enables = en;
3515 
3516 	return 0;
3517 }
3518 
3519 static struct bnxt_filter_info *
3520 bnxt_match_fdir(struct bnxt *bp, struct bnxt_filter_info *nf,
3521 		struct bnxt_vnic_info **mvnic)
3522 {
3523 	struct bnxt_filter_info *mf = NULL;
3524 	int i;
3525 
3526 	for (i = bp->nr_vnics - 1; i >= 0; i--) {
3527 		struct bnxt_vnic_info *vnic = &bp->vnic_info[i];
3528 
3529 		STAILQ_FOREACH(mf, &vnic->filter, next) {
3530 			if (mf->filter_type == nf->filter_type &&
3531 			    mf->flags == nf->flags &&
3532 			    mf->src_port == nf->src_port &&
3533 			    mf->src_port_mask == nf->src_port_mask &&
3534 			    mf->dst_port == nf->dst_port &&
3535 			    mf->dst_port_mask == nf->dst_port_mask &&
3536 			    mf->ip_protocol == nf->ip_protocol &&
3537 			    mf->ip_addr_type == nf->ip_addr_type &&
3538 			    mf->ethertype == nf->ethertype &&
3539 			    mf->vni == nf->vni &&
3540 			    mf->tunnel_type == nf->tunnel_type &&
3541 			    mf->l2_ovlan == nf->l2_ovlan &&
3542 			    mf->l2_ovlan_mask == nf->l2_ovlan_mask &&
3543 			    mf->l2_ivlan == nf->l2_ivlan &&
3544 			    mf->l2_ivlan_mask == nf->l2_ivlan_mask &&
3545 			    !memcmp(mf->l2_addr, nf->l2_addr,
3546 				    RTE_ETHER_ADDR_LEN) &&
3547 			    !memcmp(mf->l2_addr_mask, nf->l2_addr_mask,
3548 				    RTE_ETHER_ADDR_LEN) &&
3549 			    !memcmp(mf->src_macaddr, nf->src_macaddr,
3550 				    RTE_ETHER_ADDR_LEN) &&
3551 			    !memcmp(mf->dst_macaddr, nf->dst_macaddr,
3552 				    RTE_ETHER_ADDR_LEN) &&
3553 			    !memcmp(mf->src_ipaddr, nf->src_ipaddr,
3554 				    sizeof(nf->src_ipaddr)) &&
3555 			    !memcmp(mf->src_ipaddr_mask, nf->src_ipaddr_mask,
3556 				    sizeof(nf->src_ipaddr_mask)) &&
3557 			    !memcmp(mf->dst_ipaddr, nf->dst_ipaddr,
3558 				    sizeof(nf->dst_ipaddr)) &&
3559 			    !memcmp(mf->dst_ipaddr_mask, nf->dst_ipaddr_mask,
3560 				    sizeof(nf->dst_ipaddr_mask))) {
3561 				if (mvnic)
3562 					*mvnic = vnic;
3563 				return mf;
3564 			}
3565 		}
3566 	}
3567 	return NULL;
3568 }
3569 
3570 static int
3571 bnxt_fdir_filter(struct rte_eth_dev *dev,
3572 		 enum rte_filter_op filter_op,
3573 		 void *arg)
3574 {
3575 	struct bnxt *bp = dev->data->dev_private;
3576 	struct rte_eth_fdir_filter *fdir  = (struct rte_eth_fdir_filter *)arg;
3577 	struct bnxt_filter_info *filter, *match;
3578 	struct bnxt_vnic_info *vnic, *mvnic;
3579 	int ret = 0, i;
3580 
3581 	if (filter_op == RTE_ETH_FILTER_NOP)
3582 		return 0;
3583 
3584 	if (arg == NULL && filter_op != RTE_ETH_FILTER_FLUSH)
3585 		return -EINVAL;
3586 
3587 	switch (filter_op) {
3588 	case RTE_ETH_FILTER_ADD:
3589 	case RTE_ETH_FILTER_DELETE:
3590 		/* FALLTHROUGH */
3591 		filter = bnxt_get_unused_filter(bp);
3592 		if (filter == NULL) {
3593 			PMD_DRV_LOG(ERR,
3594 				"Not enough resources for a new flow.\n");
3595 			return -ENOMEM;
3596 		}
3597 
3598 		ret = bnxt_parse_fdir_filter(bp, fdir, filter);
3599 		if (ret != 0)
3600 			goto free_filter;
3601 		filter->filter_type = HWRM_CFA_NTUPLE_FILTER;
3602 
3603 		if (fdir->action.behavior == RTE_ETH_FDIR_REJECT)
3604 			vnic = &bp->vnic_info[0];
3605 		else
3606 			vnic = &bp->vnic_info[fdir->action.rx_queue];
3607 
3608 		match = bnxt_match_fdir(bp, filter, &mvnic);
3609 		if (match != NULL && filter_op == RTE_ETH_FILTER_ADD) {
3610 			if (match->dst_id == vnic->fw_vnic_id) {
3611 				PMD_DRV_LOG(ERR, "Flow already exists.\n");
3612 				ret = -EEXIST;
3613 				goto free_filter;
3614 			} else {
3615 				match->dst_id = vnic->fw_vnic_id;
3616 				ret = bnxt_hwrm_set_ntuple_filter(bp,
3617 								  match->dst_id,
3618 								  match);
3619 				STAILQ_REMOVE(&mvnic->filter, match,
3620 					      bnxt_filter_info, next);
3621 				STAILQ_INSERT_TAIL(&vnic->filter, match, next);
3622 				PMD_DRV_LOG(ERR,
3623 					"Filter with matching pattern exist\n");
3624 				PMD_DRV_LOG(ERR,
3625 					"Updated it to new destination q\n");
3626 				goto free_filter;
3627 			}
3628 		}
3629 		if (match == NULL && filter_op == RTE_ETH_FILTER_DELETE) {
3630 			PMD_DRV_LOG(ERR, "Flow does not exist.\n");
3631 			ret = -ENOENT;
3632 			goto free_filter;
3633 		}
3634 
3635 		if (filter_op == RTE_ETH_FILTER_ADD) {
3636 			ret = bnxt_hwrm_set_ntuple_filter(bp,
3637 							  filter->dst_id,
3638 							  filter);
3639 			if (ret)
3640 				goto free_filter;
3641 			STAILQ_INSERT_TAIL(&vnic->filter, filter, next);
3642 		} else {
3643 			ret = bnxt_hwrm_clear_ntuple_filter(bp, match);
3644 			STAILQ_REMOVE(&vnic->filter, match,
3645 				      bnxt_filter_info, next);
3646 			bnxt_free_filter(bp, match);
3647 			bnxt_free_filter(bp, filter);
3648 		}
3649 		break;
3650 	case RTE_ETH_FILTER_FLUSH:
3651 		for (i = bp->nr_vnics - 1; i >= 0; i--) {
3652 			struct bnxt_vnic_info *vnic = &bp->vnic_info[i];
3653 
3654 			STAILQ_FOREACH(filter, &vnic->filter, next) {
3655 				if (filter->filter_type ==
3656 				    HWRM_CFA_NTUPLE_FILTER) {
3657 					ret =
3658 					bnxt_hwrm_clear_ntuple_filter(bp,
3659 								      filter);
3660 					STAILQ_REMOVE(&vnic->filter, filter,
3661 						      bnxt_filter_info, next);
3662 				}
3663 			}
3664 		}
3665 		return ret;
3666 	case RTE_ETH_FILTER_UPDATE:
3667 	case RTE_ETH_FILTER_STATS:
3668 	case RTE_ETH_FILTER_INFO:
3669 		PMD_DRV_LOG(ERR, "operation %u not implemented", filter_op);
3670 		break;
3671 	default:
3672 		PMD_DRV_LOG(ERR, "unknown operation %u", filter_op);
3673 		ret = -EINVAL;
3674 		break;
3675 	}
3676 	return ret;
3677 
3678 free_filter:
3679 	bnxt_free_filter(bp, filter);
3680 	return ret;
3681 }
3682 
3683 int
3684 bnxt_filter_ctrl_op(struct rte_eth_dev *dev,
3685 		    enum rte_filter_type filter_type,
3686 		    enum rte_filter_op filter_op, void *arg)
3687 {
3688 	struct bnxt *bp = dev->data->dev_private;
3689 	int ret = 0;
3690 
3691 	if (BNXT_ETH_DEV_IS_REPRESENTOR(dev)) {
3692 		struct bnxt_vf_representor *vfr = dev->data->dev_private;
3693 		bp = vfr->parent_dev->data->dev_private;
3694 	}
3695 
3696 	ret = is_bnxt_in_error(bp);
3697 	if (ret)
3698 		return ret;
3699 
3700 	switch (filter_type) {
3701 	case RTE_ETH_FILTER_TUNNEL:
3702 		PMD_DRV_LOG(ERR,
3703 			"filter type: %d: To be implemented\n", filter_type);
3704 		break;
3705 	case RTE_ETH_FILTER_FDIR:
3706 		ret = bnxt_fdir_filter(dev, filter_op, arg);
3707 		break;
3708 	case RTE_ETH_FILTER_NTUPLE:
3709 		ret = bnxt_ntuple_filter(dev, filter_op, arg);
3710 		break;
3711 	case RTE_ETH_FILTER_ETHERTYPE:
3712 		ret = bnxt_ethertype_filter(dev, filter_op, arg);
3713 		break;
3714 	case RTE_ETH_FILTER_GENERIC:
3715 		if (filter_op != RTE_ETH_FILTER_GET)
3716 			return -EINVAL;
3717 		if (BNXT_TRUFLOW_EN(bp))
3718 			*(const void **)arg = &bnxt_ulp_rte_flow_ops;
3719 		else
3720 			*(const void **)arg = &bnxt_flow_ops;
3721 		break;
3722 	default:
3723 		PMD_DRV_LOG(ERR,
3724 			"Filter type (%d) not supported", filter_type);
3725 		ret = -EINVAL;
3726 		break;
3727 	}
3728 	return ret;
3729 }
3730 
3731 static const uint32_t *
3732 bnxt_dev_supported_ptypes_get_op(struct rte_eth_dev *dev)
3733 {
3734 	static const uint32_t ptypes[] = {
3735 		RTE_PTYPE_L2_ETHER_VLAN,
3736 		RTE_PTYPE_L3_IPV4_EXT_UNKNOWN,
3737 		RTE_PTYPE_L3_IPV6_EXT_UNKNOWN,
3738 		RTE_PTYPE_L4_ICMP,
3739 		RTE_PTYPE_L4_TCP,
3740 		RTE_PTYPE_L4_UDP,
3741 		RTE_PTYPE_INNER_L3_IPV4_EXT_UNKNOWN,
3742 		RTE_PTYPE_INNER_L3_IPV6_EXT_UNKNOWN,
3743 		RTE_PTYPE_INNER_L4_ICMP,
3744 		RTE_PTYPE_INNER_L4_TCP,
3745 		RTE_PTYPE_INNER_L4_UDP,
3746 		RTE_PTYPE_UNKNOWN
3747 	};
3748 
3749 	if (!dev->rx_pkt_burst)
3750 		return NULL;
3751 
3752 	return ptypes;
3753 }
3754 
3755 static int bnxt_map_regs(struct bnxt *bp, uint32_t *reg_arr, int count,
3756 			 int reg_win)
3757 {
3758 	uint32_t reg_base = *reg_arr & 0xfffff000;
3759 	uint32_t win_off;
3760 	int i;
3761 
3762 	for (i = 0; i < count; i++) {
3763 		if ((reg_arr[i] & 0xfffff000) != reg_base)
3764 			return -ERANGE;
3765 	}
3766 	win_off = BNXT_GRCPF_REG_WINDOW_BASE_OUT + (reg_win - 1) * 4;
3767 	rte_write32(reg_base, (uint8_t *)bp->bar0 + win_off);
3768 	return 0;
3769 }
3770 
3771 static int bnxt_map_ptp_regs(struct bnxt *bp)
3772 {
3773 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3774 	uint32_t *reg_arr;
3775 	int rc, i;
3776 
3777 	reg_arr = ptp->rx_regs;
3778 	rc = bnxt_map_regs(bp, reg_arr, BNXT_PTP_RX_REGS, 5);
3779 	if (rc)
3780 		return rc;
3781 
3782 	reg_arr = ptp->tx_regs;
3783 	rc = bnxt_map_regs(bp, reg_arr, BNXT_PTP_TX_REGS, 6);
3784 	if (rc)
3785 		return rc;
3786 
3787 	for (i = 0; i < BNXT_PTP_RX_REGS; i++)
3788 		ptp->rx_mapped_regs[i] = 0x5000 + (ptp->rx_regs[i] & 0xfff);
3789 
3790 	for (i = 0; i < BNXT_PTP_TX_REGS; i++)
3791 		ptp->tx_mapped_regs[i] = 0x6000 + (ptp->tx_regs[i] & 0xfff);
3792 
3793 	return 0;
3794 }
3795 
3796 static void bnxt_unmap_ptp_regs(struct bnxt *bp)
3797 {
3798 	rte_write32(0, (uint8_t *)bp->bar0 +
3799 			 BNXT_GRCPF_REG_WINDOW_BASE_OUT + 16);
3800 	rte_write32(0, (uint8_t *)bp->bar0 +
3801 			 BNXT_GRCPF_REG_WINDOW_BASE_OUT + 20);
3802 }
3803 
3804 static uint64_t bnxt_cc_read(struct bnxt *bp)
3805 {
3806 	uint64_t ns;
3807 
3808 	ns = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3809 			      BNXT_GRCPF_REG_SYNC_TIME));
3810 	ns |= (uint64_t)(rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3811 					  BNXT_GRCPF_REG_SYNC_TIME + 4))) << 32;
3812 	return ns;
3813 }
3814 
3815 static int bnxt_get_tx_ts(struct bnxt *bp, uint64_t *ts)
3816 {
3817 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3818 	uint32_t fifo;
3819 
3820 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3821 				ptp->tx_mapped_regs[BNXT_PTP_TX_FIFO]));
3822 	if (fifo & BNXT_PTP_TX_FIFO_EMPTY)
3823 		return -EAGAIN;
3824 
3825 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3826 				ptp->tx_mapped_regs[BNXT_PTP_TX_FIFO]));
3827 	*ts = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3828 				ptp->tx_mapped_regs[BNXT_PTP_TX_TS_L]));
3829 	*ts |= (uint64_t)rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3830 				ptp->tx_mapped_regs[BNXT_PTP_TX_TS_H])) << 32;
3831 
3832 	return 0;
3833 }
3834 
3835 static int bnxt_get_rx_ts(struct bnxt *bp, uint64_t *ts)
3836 {
3837 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3838 	struct bnxt_pf_info *pf = bp->pf;
3839 	uint16_t port_id;
3840 	uint32_t fifo;
3841 
3842 	if (!ptp)
3843 		return -ENODEV;
3844 
3845 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3846 				ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO]));
3847 	if (!(fifo & BNXT_PTP_RX_FIFO_PENDING))
3848 		return -EAGAIN;
3849 
3850 	port_id = pf->port_id;
3851 	rte_write32(1 << port_id, (uint8_t *)bp->bar0 +
3852 	       ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO_ADV]);
3853 
3854 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3855 				   ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO]));
3856 	if (fifo & BNXT_PTP_RX_FIFO_PENDING) {
3857 /*		bnxt_clr_rx_ts(bp);	  TBD  */
3858 		return -EBUSY;
3859 	}
3860 
3861 	*ts = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3862 				ptp->rx_mapped_regs[BNXT_PTP_RX_TS_L]));
3863 	*ts |= (uint64_t)rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3864 				ptp->rx_mapped_regs[BNXT_PTP_RX_TS_H])) << 32;
3865 
3866 	return 0;
3867 }
3868 
3869 static int
3870 bnxt_timesync_write_time(struct rte_eth_dev *dev, const struct timespec *ts)
3871 {
3872 	uint64_t ns;
3873 	struct bnxt *bp = dev->data->dev_private;
3874 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3875 
3876 	if (!ptp)
3877 		return 0;
3878 
3879 	ns = rte_timespec_to_ns(ts);
3880 	/* Set the timecounters to a new value. */
3881 	ptp->tc.nsec = ns;
3882 
3883 	return 0;
3884 }
3885 
3886 static int
3887 bnxt_timesync_read_time(struct rte_eth_dev *dev, struct timespec *ts)
3888 {
3889 	struct bnxt *bp = dev->data->dev_private;
3890 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3891 	uint64_t ns, systime_cycles = 0;
3892 	int rc = 0;
3893 
3894 	if (!ptp)
3895 		return 0;
3896 
3897 	if (BNXT_CHIP_THOR(bp))
3898 		rc = bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_CURRENT_TIME,
3899 					     &systime_cycles);
3900 	else
3901 		systime_cycles = bnxt_cc_read(bp);
3902 
3903 	ns = rte_timecounter_update(&ptp->tc, systime_cycles);
3904 	*ts = rte_ns_to_timespec(ns);
3905 
3906 	return rc;
3907 }
3908 static int
3909 bnxt_timesync_enable(struct rte_eth_dev *dev)
3910 {
3911 	struct bnxt *bp = dev->data->dev_private;
3912 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3913 	uint32_t shift = 0;
3914 	int rc;
3915 
3916 	if (!ptp)
3917 		return 0;
3918 
3919 	ptp->rx_filter = 1;
3920 	ptp->tx_tstamp_en = 1;
3921 	ptp->rxctl = BNXT_PTP_MSG_EVENTS;
3922 
3923 	rc = bnxt_hwrm_ptp_cfg(bp);
3924 	if (rc)
3925 		return rc;
3926 
3927 	memset(&ptp->tc, 0, sizeof(struct rte_timecounter));
3928 	memset(&ptp->rx_tstamp_tc, 0, sizeof(struct rte_timecounter));
3929 	memset(&ptp->tx_tstamp_tc, 0, sizeof(struct rte_timecounter));
3930 
3931 	ptp->tc.cc_mask = BNXT_CYCLECOUNTER_MASK;
3932 	ptp->tc.cc_shift = shift;
3933 	ptp->tc.nsec_mask = (1ULL << shift) - 1;
3934 
3935 	ptp->rx_tstamp_tc.cc_mask = BNXT_CYCLECOUNTER_MASK;
3936 	ptp->rx_tstamp_tc.cc_shift = shift;
3937 	ptp->rx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
3938 
3939 	ptp->tx_tstamp_tc.cc_mask = BNXT_CYCLECOUNTER_MASK;
3940 	ptp->tx_tstamp_tc.cc_shift = shift;
3941 	ptp->tx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
3942 
3943 	if (!BNXT_CHIP_THOR(bp))
3944 		bnxt_map_ptp_regs(bp);
3945 
3946 	return 0;
3947 }
3948 
3949 static int
3950 bnxt_timesync_disable(struct rte_eth_dev *dev)
3951 {
3952 	struct bnxt *bp = dev->data->dev_private;
3953 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3954 
3955 	if (!ptp)
3956 		return 0;
3957 
3958 	ptp->rx_filter = 0;
3959 	ptp->tx_tstamp_en = 0;
3960 	ptp->rxctl = 0;
3961 
3962 	bnxt_hwrm_ptp_cfg(bp);
3963 
3964 	if (!BNXT_CHIP_THOR(bp))
3965 		bnxt_unmap_ptp_regs(bp);
3966 
3967 	return 0;
3968 }
3969 
3970 static int
3971 bnxt_timesync_read_rx_timestamp(struct rte_eth_dev *dev,
3972 				 struct timespec *timestamp,
3973 				 uint32_t flags __rte_unused)
3974 {
3975 	struct bnxt *bp = dev->data->dev_private;
3976 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3977 	uint64_t rx_tstamp_cycles = 0;
3978 	uint64_t ns;
3979 
3980 	if (!ptp)
3981 		return 0;
3982 
3983 	if (BNXT_CHIP_THOR(bp))
3984 		rx_tstamp_cycles = ptp->rx_timestamp;
3985 	else
3986 		bnxt_get_rx_ts(bp, &rx_tstamp_cycles);
3987 
3988 	ns = rte_timecounter_update(&ptp->rx_tstamp_tc, rx_tstamp_cycles);
3989 	*timestamp = rte_ns_to_timespec(ns);
3990 	return  0;
3991 }
3992 
3993 static int
3994 bnxt_timesync_read_tx_timestamp(struct rte_eth_dev *dev,
3995 				 struct timespec *timestamp)
3996 {
3997 	struct bnxt *bp = dev->data->dev_private;
3998 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3999 	uint64_t tx_tstamp_cycles = 0;
4000 	uint64_t ns;
4001 	int rc = 0;
4002 
4003 	if (!ptp)
4004 		return 0;
4005 
4006 	if (BNXT_CHIP_THOR(bp))
4007 		rc = bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_PATH_TX,
4008 					     &tx_tstamp_cycles);
4009 	else
4010 		rc = bnxt_get_tx_ts(bp, &tx_tstamp_cycles);
4011 
4012 	ns = rte_timecounter_update(&ptp->tx_tstamp_tc, tx_tstamp_cycles);
4013 	*timestamp = rte_ns_to_timespec(ns);
4014 
4015 	return rc;
4016 }
4017 
4018 static int
4019 bnxt_timesync_adjust_time(struct rte_eth_dev *dev, int64_t delta)
4020 {
4021 	struct bnxt *bp = dev->data->dev_private;
4022 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
4023 
4024 	if (!ptp)
4025 		return 0;
4026 
4027 	ptp->tc.nsec += delta;
4028 
4029 	return 0;
4030 }
4031 
4032 static int
4033 bnxt_get_eeprom_length_op(struct rte_eth_dev *dev)
4034 {
4035 	struct bnxt *bp = dev->data->dev_private;
4036 	int rc;
4037 	uint32_t dir_entries;
4038 	uint32_t entry_length;
4039 
4040 	rc = is_bnxt_in_error(bp);
4041 	if (rc)
4042 		return rc;
4043 
4044 	PMD_DRV_LOG(INFO, PCI_PRI_FMT "\n",
4045 		    bp->pdev->addr.domain, bp->pdev->addr.bus,
4046 		    bp->pdev->addr.devid, bp->pdev->addr.function);
4047 
4048 	rc = bnxt_hwrm_nvm_get_dir_info(bp, &dir_entries, &entry_length);
4049 	if (rc != 0)
4050 		return rc;
4051 
4052 	return dir_entries * entry_length;
4053 }
4054 
4055 static int
4056 bnxt_get_eeprom_op(struct rte_eth_dev *dev,
4057 		struct rte_dev_eeprom_info *in_eeprom)
4058 {
4059 	struct bnxt *bp = dev->data->dev_private;
4060 	uint32_t index;
4061 	uint32_t offset;
4062 	int rc;
4063 
4064 	rc = is_bnxt_in_error(bp);
4065 	if (rc)
4066 		return rc;
4067 
4068 	PMD_DRV_LOG(INFO, PCI_PRI_FMT " in_eeprom->offset = %d len = %d\n",
4069 		    bp->pdev->addr.domain, bp->pdev->addr.bus,
4070 		    bp->pdev->addr.devid, bp->pdev->addr.function,
4071 		    in_eeprom->offset, in_eeprom->length);
4072 
4073 	if (in_eeprom->offset == 0) /* special offset value to get directory */
4074 		return bnxt_get_nvram_directory(bp, in_eeprom->length,
4075 						in_eeprom->data);
4076 
4077 	index = in_eeprom->offset >> 24;
4078 	offset = in_eeprom->offset & 0xffffff;
4079 
4080 	if (index != 0)
4081 		return bnxt_hwrm_get_nvram_item(bp, index - 1, offset,
4082 					   in_eeprom->length, in_eeprom->data);
4083 
4084 	return 0;
4085 }
4086 
4087 static bool bnxt_dir_type_is_ape_bin_format(uint16_t dir_type)
4088 {
4089 	switch (dir_type) {
4090 	case BNX_DIR_TYPE_CHIMP_PATCH:
4091 	case BNX_DIR_TYPE_BOOTCODE:
4092 	case BNX_DIR_TYPE_BOOTCODE_2:
4093 	case BNX_DIR_TYPE_APE_FW:
4094 	case BNX_DIR_TYPE_APE_PATCH:
4095 	case BNX_DIR_TYPE_KONG_FW:
4096 	case BNX_DIR_TYPE_KONG_PATCH:
4097 	case BNX_DIR_TYPE_BONO_FW:
4098 	case BNX_DIR_TYPE_BONO_PATCH:
4099 		/* FALLTHROUGH */
4100 		return true;
4101 	}
4102 
4103 	return false;
4104 }
4105 
4106 static bool bnxt_dir_type_is_other_exec_format(uint16_t dir_type)
4107 {
4108 	switch (dir_type) {
4109 	case BNX_DIR_TYPE_AVS:
4110 	case BNX_DIR_TYPE_EXP_ROM_MBA:
4111 	case BNX_DIR_TYPE_PCIE:
4112 	case BNX_DIR_TYPE_TSCF_UCODE:
4113 	case BNX_DIR_TYPE_EXT_PHY:
4114 	case BNX_DIR_TYPE_CCM:
4115 	case BNX_DIR_TYPE_ISCSI_BOOT:
4116 	case BNX_DIR_TYPE_ISCSI_BOOT_IPV6:
4117 	case BNX_DIR_TYPE_ISCSI_BOOT_IPV4N6:
4118 		/* FALLTHROUGH */
4119 		return true;
4120 	}
4121 
4122 	return false;
4123 }
4124 
4125 static bool bnxt_dir_type_is_executable(uint16_t dir_type)
4126 {
4127 	return bnxt_dir_type_is_ape_bin_format(dir_type) ||
4128 		bnxt_dir_type_is_other_exec_format(dir_type);
4129 }
4130 
4131 static int
4132 bnxt_set_eeprom_op(struct rte_eth_dev *dev,
4133 		struct rte_dev_eeprom_info *in_eeprom)
4134 {
4135 	struct bnxt *bp = dev->data->dev_private;
4136 	uint8_t index, dir_op;
4137 	uint16_t type, ext, ordinal, attr;
4138 	int rc;
4139 
4140 	rc = is_bnxt_in_error(bp);
4141 	if (rc)
4142 		return rc;
4143 
4144 	PMD_DRV_LOG(INFO, PCI_PRI_FMT " in_eeprom->offset = %d len = %d\n",
4145 		    bp->pdev->addr.domain, bp->pdev->addr.bus,
4146 		    bp->pdev->addr.devid, bp->pdev->addr.function,
4147 		    in_eeprom->offset, in_eeprom->length);
4148 
4149 	if (!BNXT_PF(bp)) {
4150 		PMD_DRV_LOG(ERR, "NVM write not supported from a VF\n");
4151 		return -EINVAL;
4152 	}
4153 
4154 	type = in_eeprom->magic >> 16;
4155 
4156 	if (type == 0xffff) { /* special value for directory operations */
4157 		index = in_eeprom->magic & 0xff;
4158 		dir_op = in_eeprom->magic >> 8;
4159 		if (index == 0)
4160 			return -EINVAL;
4161 		switch (dir_op) {
4162 		case 0x0e: /* erase */
4163 			if (in_eeprom->offset != ~in_eeprom->magic)
4164 				return -EINVAL;
4165 			return bnxt_hwrm_erase_nvram_directory(bp, index - 1);
4166 		default:
4167 			return -EINVAL;
4168 		}
4169 	}
4170 
4171 	/* Create or re-write an NVM item: */
4172 	if (bnxt_dir_type_is_executable(type) == true)
4173 		return -EOPNOTSUPP;
4174 	ext = in_eeprom->magic & 0xffff;
4175 	ordinal = in_eeprom->offset >> 16;
4176 	attr = in_eeprom->offset & 0xffff;
4177 
4178 	return bnxt_hwrm_flash_nvram(bp, type, ordinal, ext, attr,
4179 				     in_eeprom->data, in_eeprom->length);
4180 }
4181 
4182 /*
4183  * Initialization
4184  */
4185 
4186 static const struct eth_dev_ops bnxt_dev_ops = {
4187 	.dev_infos_get = bnxt_dev_info_get_op,
4188 	.dev_close = bnxt_dev_close_op,
4189 	.dev_configure = bnxt_dev_configure_op,
4190 	.dev_start = bnxt_dev_start_op,
4191 	.dev_stop = bnxt_dev_stop_op,
4192 	.dev_set_link_up = bnxt_dev_set_link_up_op,
4193 	.dev_set_link_down = bnxt_dev_set_link_down_op,
4194 	.stats_get = bnxt_stats_get_op,
4195 	.stats_reset = bnxt_stats_reset_op,
4196 	.rx_queue_setup = bnxt_rx_queue_setup_op,
4197 	.rx_queue_release = bnxt_rx_queue_release_op,
4198 	.tx_queue_setup = bnxt_tx_queue_setup_op,
4199 	.tx_queue_release = bnxt_tx_queue_release_op,
4200 	.rx_queue_intr_enable = bnxt_rx_queue_intr_enable_op,
4201 	.rx_queue_intr_disable = bnxt_rx_queue_intr_disable_op,
4202 	.reta_update = bnxt_reta_update_op,
4203 	.reta_query = bnxt_reta_query_op,
4204 	.rss_hash_update = bnxt_rss_hash_update_op,
4205 	.rss_hash_conf_get = bnxt_rss_hash_conf_get_op,
4206 	.link_update = bnxt_link_update_op,
4207 	.promiscuous_enable = bnxt_promiscuous_enable_op,
4208 	.promiscuous_disable = bnxt_promiscuous_disable_op,
4209 	.allmulticast_enable = bnxt_allmulticast_enable_op,
4210 	.allmulticast_disable = bnxt_allmulticast_disable_op,
4211 	.mac_addr_add = bnxt_mac_addr_add_op,
4212 	.mac_addr_remove = bnxt_mac_addr_remove_op,
4213 	.flow_ctrl_get = bnxt_flow_ctrl_get_op,
4214 	.flow_ctrl_set = bnxt_flow_ctrl_set_op,
4215 	.udp_tunnel_port_add  = bnxt_udp_tunnel_port_add_op,
4216 	.udp_tunnel_port_del  = bnxt_udp_tunnel_port_del_op,
4217 	.vlan_filter_set = bnxt_vlan_filter_set_op,
4218 	.vlan_offload_set = bnxt_vlan_offload_set_op,
4219 	.vlan_tpid_set = bnxt_vlan_tpid_set_op,
4220 	.vlan_pvid_set = bnxt_vlan_pvid_set_op,
4221 	.mtu_set = bnxt_mtu_set_op,
4222 	.mac_addr_set = bnxt_set_default_mac_addr_op,
4223 	.xstats_get = bnxt_dev_xstats_get_op,
4224 	.xstats_get_names = bnxt_dev_xstats_get_names_op,
4225 	.xstats_reset = bnxt_dev_xstats_reset_op,
4226 	.fw_version_get = bnxt_fw_version_get,
4227 	.set_mc_addr_list = bnxt_dev_set_mc_addr_list_op,
4228 	.rxq_info_get = bnxt_rxq_info_get_op,
4229 	.txq_info_get = bnxt_txq_info_get_op,
4230 	.rx_burst_mode_get = bnxt_rx_burst_mode_get,
4231 	.tx_burst_mode_get = bnxt_tx_burst_mode_get,
4232 	.dev_led_on = bnxt_dev_led_on_op,
4233 	.dev_led_off = bnxt_dev_led_off_op,
4234 	.xstats_get_by_id = bnxt_dev_xstats_get_by_id_op,
4235 	.xstats_get_names_by_id = bnxt_dev_xstats_get_names_by_id_op,
4236 	.rx_queue_count = bnxt_rx_queue_count_op,
4237 	.rx_descriptor_status = bnxt_rx_descriptor_status_op,
4238 	.tx_descriptor_status = bnxt_tx_descriptor_status_op,
4239 	.rx_queue_start = bnxt_rx_queue_start,
4240 	.rx_queue_stop = bnxt_rx_queue_stop,
4241 	.tx_queue_start = bnxt_tx_queue_start,
4242 	.tx_queue_stop = bnxt_tx_queue_stop,
4243 	.filter_ctrl = bnxt_filter_ctrl_op,
4244 	.dev_supported_ptypes_get = bnxt_dev_supported_ptypes_get_op,
4245 	.get_eeprom_length    = bnxt_get_eeprom_length_op,
4246 	.get_eeprom           = bnxt_get_eeprom_op,
4247 	.set_eeprom           = bnxt_set_eeprom_op,
4248 	.timesync_enable      = bnxt_timesync_enable,
4249 	.timesync_disable     = bnxt_timesync_disable,
4250 	.timesync_read_time   = bnxt_timesync_read_time,
4251 	.timesync_write_time   = bnxt_timesync_write_time,
4252 	.timesync_adjust_time = bnxt_timesync_adjust_time,
4253 	.timesync_read_rx_timestamp = bnxt_timesync_read_rx_timestamp,
4254 	.timesync_read_tx_timestamp = bnxt_timesync_read_tx_timestamp,
4255 };
4256 
4257 static uint32_t bnxt_map_reset_regs(struct bnxt *bp, uint32_t reg)
4258 {
4259 	uint32_t offset;
4260 
4261 	/* Only pre-map the reset GRC registers using window 3 */
4262 	rte_write32(reg & 0xfffff000, (uint8_t *)bp->bar0 +
4263 		    BNXT_GRCPF_REG_WINDOW_BASE_OUT + 8);
4264 
4265 	offset = BNXT_GRCP_WINDOW_3_BASE + (reg & 0xffc);
4266 
4267 	return offset;
4268 }
4269 
4270 int bnxt_map_fw_health_status_regs(struct bnxt *bp)
4271 {
4272 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4273 	uint32_t reg_base = 0xffffffff;
4274 	int i;
4275 
4276 	/* Only pre-map the monitoring GRC registers using window 2 */
4277 	for (i = 0; i < BNXT_FW_STATUS_REG_CNT; i++) {
4278 		uint32_t reg = info->status_regs[i];
4279 
4280 		if (BNXT_FW_STATUS_REG_TYPE(reg) != BNXT_FW_STATUS_REG_TYPE_GRC)
4281 			continue;
4282 
4283 		if (reg_base == 0xffffffff)
4284 			reg_base = reg & 0xfffff000;
4285 		if ((reg & 0xfffff000) != reg_base)
4286 			return -ERANGE;
4287 
4288 		/* Use mask 0xffc as the Lower 2 bits indicates
4289 		 * address space location
4290 		 */
4291 		info->mapped_status_regs[i] = BNXT_GRCP_WINDOW_2_BASE +
4292 						(reg & 0xffc);
4293 	}
4294 
4295 	if (reg_base == 0xffffffff)
4296 		return 0;
4297 
4298 	rte_write32(reg_base, (uint8_t *)bp->bar0 +
4299 		    BNXT_GRCPF_REG_WINDOW_BASE_OUT + 4);
4300 
4301 	return 0;
4302 }
4303 
4304 static void bnxt_write_fw_reset_reg(struct bnxt *bp, uint32_t index)
4305 {
4306 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4307 	uint32_t delay = info->delay_after_reset[index];
4308 	uint32_t val = info->reset_reg_val[index];
4309 	uint32_t reg = info->reset_reg[index];
4310 	uint32_t type, offset;
4311 
4312 	type = BNXT_FW_STATUS_REG_TYPE(reg);
4313 	offset = BNXT_FW_STATUS_REG_OFF(reg);
4314 
4315 	switch (type) {
4316 	case BNXT_FW_STATUS_REG_TYPE_CFG:
4317 		rte_pci_write_config(bp->pdev, &val, sizeof(val), offset);
4318 		break;
4319 	case BNXT_FW_STATUS_REG_TYPE_GRC:
4320 		offset = bnxt_map_reset_regs(bp, offset);
4321 		rte_write32(val, (uint8_t *)bp->bar0 + offset);
4322 		break;
4323 	case BNXT_FW_STATUS_REG_TYPE_BAR0:
4324 		rte_write32(val, (uint8_t *)bp->bar0 + offset);
4325 		break;
4326 	}
4327 	/* wait on a specific interval of time until core reset is complete */
4328 	if (delay)
4329 		rte_delay_ms(delay);
4330 }
4331 
4332 static void bnxt_dev_cleanup(struct bnxt *bp)
4333 {
4334 	bnxt_set_hwrm_link_config(bp, false);
4335 	bp->link_info->link_up = 0;
4336 	if (bp->eth_dev->data->dev_started)
4337 		bnxt_dev_stop_op(bp->eth_dev);
4338 
4339 	bnxt_uninit_resources(bp, true);
4340 }
4341 
4342 static int bnxt_restore_vlan_filters(struct bnxt *bp)
4343 {
4344 	struct rte_eth_dev *dev = bp->eth_dev;
4345 	struct rte_vlan_filter_conf *vfc;
4346 	int vidx, vbit, rc;
4347 	uint16_t vlan_id;
4348 
4349 	for (vlan_id = 1; vlan_id <= RTE_ETHER_MAX_VLAN_ID; vlan_id++) {
4350 		vfc = &dev->data->vlan_filter_conf;
4351 		vidx = vlan_id / 64;
4352 		vbit = vlan_id % 64;
4353 
4354 		/* Each bit corresponds to a VLAN id */
4355 		if (vfc->ids[vidx] & (UINT64_C(1) << vbit)) {
4356 			rc = bnxt_add_vlan_filter(bp, vlan_id);
4357 			if (rc)
4358 				return rc;
4359 		}
4360 	}
4361 
4362 	return 0;
4363 }
4364 
4365 static int bnxt_restore_mac_filters(struct bnxt *bp)
4366 {
4367 	struct rte_eth_dev *dev = bp->eth_dev;
4368 	struct rte_eth_dev_info dev_info;
4369 	struct rte_ether_addr *addr;
4370 	uint64_t pool_mask;
4371 	uint32_t pool = 0;
4372 	uint16_t i;
4373 	int rc;
4374 
4375 	if (BNXT_VF(bp) && !BNXT_VF_IS_TRUSTED(bp))
4376 		return 0;
4377 
4378 	rc = bnxt_dev_info_get_op(dev, &dev_info);
4379 	if (rc)
4380 		return rc;
4381 
4382 	/* replay MAC address configuration */
4383 	for (i = 1; i < dev_info.max_mac_addrs; i++) {
4384 		addr = &dev->data->mac_addrs[i];
4385 
4386 		/* skip zero address */
4387 		if (rte_is_zero_ether_addr(addr))
4388 			continue;
4389 
4390 		pool = 0;
4391 		pool_mask = dev->data->mac_pool_sel[i];
4392 
4393 		do {
4394 			if (pool_mask & 1ULL) {
4395 				rc = bnxt_mac_addr_add_op(dev, addr, i, pool);
4396 				if (rc)
4397 					return rc;
4398 			}
4399 			pool_mask >>= 1;
4400 			pool++;
4401 		} while (pool_mask);
4402 	}
4403 
4404 	return 0;
4405 }
4406 
4407 static int bnxt_restore_filters(struct bnxt *bp)
4408 {
4409 	struct rte_eth_dev *dev = bp->eth_dev;
4410 	int ret = 0;
4411 
4412 	if (dev->data->all_multicast) {
4413 		ret = bnxt_allmulticast_enable_op(dev);
4414 		if (ret)
4415 			return ret;
4416 	}
4417 	if (dev->data->promiscuous) {
4418 		ret = bnxt_promiscuous_enable_op(dev);
4419 		if (ret)
4420 			return ret;
4421 	}
4422 
4423 	ret = bnxt_restore_mac_filters(bp);
4424 	if (ret)
4425 		return ret;
4426 
4427 	ret = bnxt_restore_vlan_filters(bp);
4428 	/* TODO restore other filters as well */
4429 	return ret;
4430 }
4431 
4432 static void bnxt_dev_recover(void *arg)
4433 {
4434 	struct bnxt *bp = arg;
4435 	int timeout = bp->fw_reset_max_msecs;
4436 	int rc = 0;
4437 
4438 	/* Clear Error flag so that device re-init should happen */
4439 	bp->flags &= ~BNXT_FLAG_FATAL_ERROR;
4440 
4441 	do {
4442 		rc = bnxt_hwrm_ver_get(bp, SHORT_HWRM_CMD_TIMEOUT);
4443 		if (rc == 0)
4444 			break;
4445 		rte_delay_ms(BNXT_FW_READY_WAIT_INTERVAL);
4446 		timeout -= BNXT_FW_READY_WAIT_INTERVAL;
4447 	} while (rc && timeout);
4448 
4449 	if (rc) {
4450 		PMD_DRV_LOG(ERR, "FW is not Ready after reset\n");
4451 		goto err;
4452 	}
4453 
4454 	rc = bnxt_init_resources(bp, true);
4455 	if (rc) {
4456 		PMD_DRV_LOG(ERR,
4457 			    "Failed to initialize resources after reset\n");
4458 		goto err;
4459 	}
4460 	/* clear reset flag as the device is initialized now */
4461 	bp->flags &= ~BNXT_FLAG_FW_RESET;
4462 
4463 	rc = bnxt_dev_start_op(bp->eth_dev);
4464 	if (rc) {
4465 		PMD_DRV_LOG(ERR, "Failed to start port after reset\n");
4466 		goto err_start;
4467 	}
4468 
4469 	rc = bnxt_restore_filters(bp);
4470 	if (rc)
4471 		goto err_start;
4472 
4473 	PMD_DRV_LOG(INFO, "Recovered from FW reset\n");
4474 	return;
4475 err_start:
4476 	bnxt_dev_stop_op(bp->eth_dev);
4477 err:
4478 	bp->flags |= BNXT_FLAG_FATAL_ERROR;
4479 	bnxt_uninit_resources(bp, false);
4480 	PMD_DRV_LOG(ERR, "Failed to recover from FW reset\n");
4481 }
4482 
4483 void bnxt_dev_reset_and_resume(void *arg)
4484 {
4485 	struct bnxt *bp = arg;
4486 	int rc;
4487 
4488 	bnxt_dev_cleanup(bp);
4489 
4490 	bnxt_wait_for_device_shutdown(bp);
4491 
4492 	rc = rte_eal_alarm_set(US_PER_MS * bp->fw_reset_min_msecs,
4493 			       bnxt_dev_recover, (void *)bp);
4494 	if (rc)
4495 		PMD_DRV_LOG(ERR, "Error setting recovery alarm");
4496 }
4497 
4498 uint32_t bnxt_read_fw_status_reg(struct bnxt *bp, uint32_t index)
4499 {
4500 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4501 	uint32_t reg = info->status_regs[index];
4502 	uint32_t type, offset, val = 0;
4503 
4504 	type = BNXT_FW_STATUS_REG_TYPE(reg);
4505 	offset = BNXT_FW_STATUS_REG_OFF(reg);
4506 
4507 	switch (type) {
4508 	case BNXT_FW_STATUS_REG_TYPE_CFG:
4509 		rte_pci_read_config(bp->pdev, &val, sizeof(val), offset);
4510 		break;
4511 	case BNXT_FW_STATUS_REG_TYPE_GRC:
4512 		offset = info->mapped_status_regs[index];
4513 		/* FALLTHROUGH */
4514 	case BNXT_FW_STATUS_REG_TYPE_BAR0:
4515 		val = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
4516 				       offset));
4517 		break;
4518 	}
4519 
4520 	return val;
4521 }
4522 
4523 static int bnxt_fw_reset_all(struct bnxt *bp)
4524 {
4525 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4526 	uint32_t i;
4527 	int rc = 0;
4528 
4529 	if (info->flags & BNXT_FLAG_ERROR_RECOVERY_HOST) {
4530 		/* Reset through master function driver */
4531 		for (i = 0; i < info->reg_array_cnt; i++)
4532 			bnxt_write_fw_reset_reg(bp, i);
4533 		/* Wait for time specified by FW after triggering reset */
4534 		rte_delay_ms(info->master_func_wait_period_after_reset);
4535 	} else if (info->flags & BNXT_FLAG_ERROR_RECOVERY_CO_CPU) {
4536 		/* Reset with the help of Kong processor */
4537 		rc = bnxt_hwrm_fw_reset(bp);
4538 		if (rc)
4539 			PMD_DRV_LOG(ERR, "Failed to reset FW\n");
4540 	}
4541 
4542 	return rc;
4543 }
4544 
4545 static void bnxt_fw_reset_cb(void *arg)
4546 {
4547 	struct bnxt *bp = arg;
4548 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4549 	int rc = 0;
4550 
4551 	/* Only Master function can do FW reset */
4552 	if (bnxt_is_master_func(bp) &&
4553 	    bnxt_is_recovery_enabled(bp)) {
4554 		rc = bnxt_fw_reset_all(bp);
4555 		if (rc) {
4556 			PMD_DRV_LOG(ERR, "Adapter recovery failed\n");
4557 			return;
4558 		}
4559 	}
4560 
4561 	/* if recovery method is ERROR_RECOVERY_CO_CPU, KONG will send
4562 	 * EXCEPTION_FATAL_ASYNC event to all the functions
4563 	 * (including MASTER FUNC). After receiving this Async, all the active
4564 	 * drivers should treat this case as FW initiated recovery
4565 	 */
4566 	if (info->flags & BNXT_FLAG_ERROR_RECOVERY_HOST) {
4567 		bp->fw_reset_min_msecs = BNXT_MIN_FW_READY_TIMEOUT;
4568 		bp->fw_reset_max_msecs = BNXT_MAX_FW_RESET_TIMEOUT;
4569 
4570 		/* To recover from error */
4571 		rte_eal_alarm_set(US_PER_MS, bnxt_dev_reset_and_resume,
4572 				  (void *)bp);
4573 	}
4574 }
4575 
4576 /* Driver should poll FW heartbeat, reset_counter with the frequency
4577  * advertised by FW in HWRM_ERROR_RECOVERY_QCFG.
4578  * When the driver detects heartbeat stop or change in reset_counter,
4579  * it has to trigger a reset to recover from the error condition.
4580  * A “master PF” is the function who will have the privilege to
4581  * initiate the chimp reset. The master PF will be elected by the
4582  * firmware and will be notified through async message.
4583  */
4584 static void bnxt_check_fw_health(void *arg)
4585 {
4586 	struct bnxt *bp = arg;
4587 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4588 	uint32_t val = 0, wait_msec;
4589 
4590 	if (!info || !bnxt_is_recovery_enabled(bp) ||
4591 	    is_bnxt_in_error(bp))
4592 		return;
4593 
4594 	val = bnxt_read_fw_status_reg(bp, BNXT_FW_HEARTBEAT_CNT_REG);
4595 	if (val == info->last_heart_beat)
4596 		goto reset;
4597 
4598 	info->last_heart_beat = val;
4599 
4600 	val = bnxt_read_fw_status_reg(bp, BNXT_FW_RECOVERY_CNT_REG);
4601 	if (val != info->last_reset_counter)
4602 		goto reset;
4603 
4604 	info->last_reset_counter = val;
4605 
4606 	rte_eal_alarm_set(US_PER_MS * info->driver_polling_freq,
4607 			  bnxt_check_fw_health, (void *)bp);
4608 
4609 	return;
4610 reset:
4611 	/* Stop DMA to/from device */
4612 	bp->flags |= BNXT_FLAG_FATAL_ERROR;
4613 	bp->flags |= BNXT_FLAG_FW_RESET;
4614 
4615 	PMD_DRV_LOG(ERR, "Detected FW dead condition\n");
4616 
4617 	if (bnxt_is_master_func(bp))
4618 		wait_msec = info->master_func_wait_period;
4619 	else
4620 		wait_msec = info->normal_func_wait_period;
4621 
4622 	rte_eal_alarm_set(US_PER_MS * wait_msec,
4623 			  bnxt_fw_reset_cb, (void *)bp);
4624 }
4625 
4626 void bnxt_schedule_fw_health_check(struct bnxt *bp)
4627 {
4628 	uint32_t polling_freq;
4629 
4630 	if (!bnxt_is_recovery_enabled(bp))
4631 		return;
4632 
4633 	if (bp->flags & BNXT_FLAG_FW_HEALTH_CHECK_SCHEDULED)
4634 		return;
4635 
4636 	polling_freq = bp->recovery_info->driver_polling_freq;
4637 
4638 	rte_eal_alarm_set(US_PER_MS * polling_freq,
4639 			  bnxt_check_fw_health, (void *)bp);
4640 	bp->flags |= BNXT_FLAG_FW_HEALTH_CHECK_SCHEDULED;
4641 }
4642 
4643 static void bnxt_cancel_fw_health_check(struct bnxt *bp)
4644 {
4645 	if (!bnxt_is_recovery_enabled(bp))
4646 		return;
4647 
4648 	rte_eal_alarm_cancel(bnxt_check_fw_health, (void *)bp);
4649 	bp->flags &= ~BNXT_FLAG_FW_HEALTH_CHECK_SCHEDULED;
4650 }
4651 
4652 static bool bnxt_vf_pciid(uint16_t device_id)
4653 {
4654 	switch (device_id) {
4655 	case BROADCOM_DEV_ID_57304_VF:
4656 	case BROADCOM_DEV_ID_57406_VF:
4657 	case BROADCOM_DEV_ID_5731X_VF:
4658 	case BROADCOM_DEV_ID_5741X_VF:
4659 	case BROADCOM_DEV_ID_57414_VF:
4660 	case BROADCOM_DEV_ID_STRATUS_NIC_VF1:
4661 	case BROADCOM_DEV_ID_STRATUS_NIC_VF2:
4662 	case BROADCOM_DEV_ID_58802_VF:
4663 	case BROADCOM_DEV_ID_57500_VF1:
4664 	case BROADCOM_DEV_ID_57500_VF2:
4665 		/* FALLTHROUGH */
4666 		return true;
4667 	default:
4668 		return false;
4669 	}
4670 }
4671 
4672 static bool bnxt_thor_device(uint16_t device_id)
4673 {
4674 	switch (device_id) {
4675 	case BROADCOM_DEV_ID_57508:
4676 	case BROADCOM_DEV_ID_57504:
4677 	case BROADCOM_DEV_ID_57502:
4678 	case BROADCOM_DEV_ID_57508_MF1:
4679 	case BROADCOM_DEV_ID_57504_MF1:
4680 	case BROADCOM_DEV_ID_57502_MF1:
4681 	case BROADCOM_DEV_ID_57508_MF2:
4682 	case BROADCOM_DEV_ID_57504_MF2:
4683 	case BROADCOM_DEV_ID_57502_MF2:
4684 	case BROADCOM_DEV_ID_57500_VF1:
4685 	case BROADCOM_DEV_ID_57500_VF2:
4686 		/* FALLTHROUGH */
4687 		return true;
4688 	default:
4689 		return false;
4690 	}
4691 }
4692 
4693 bool bnxt_stratus_device(struct bnxt *bp)
4694 {
4695 	uint16_t device_id = bp->pdev->id.device_id;
4696 
4697 	switch (device_id) {
4698 	case BROADCOM_DEV_ID_STRATUS_NIC:
4699 	case BROADCOM_DEV_ID_STRATUS_NIC_VF1:
4700 	case BROADCOM_DEV_ID_STRATUS_NIC_VF2:
4701 		/* FALLTHROUGH */
4702 		return true;
4703 	default:
4704 		return false;
4705 	}
4706 }
4707 
4708 static int bnxt_init_board(struct rte_eth_dev *eth_dev)
4709 {
4710 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
4711 	struct bnxt *bp = eth_dev->data->dev_private;
4712 
4713 	/* enable device (incl. PCI PM wakeup), and bus-mastering */
4714 	bp->bar0 = (void *)pci_dev->mem_resource[0].addr;
4715 	bp->doorbell_base = (void *)pci_dev->mem_resource[2].addr;
4716 	if (!bp->bar0 || !bp->doorbell_base) {
4717 		PMD_DRV_LOG(ERR, "Unable to access Hardware\n");
4718 		return -ENODEV;
4719 	}
4720 
4721 	bp->eth_dev = eth_dev;
4722 	bp->pdev = pci_dev;
4723 
4724 	return 0;
4725 }
4726 
4727 static int bnxt_alloc_ctx_mem_blk(struct bnxt *bp,
4728 				  struct bnxt_ctx_pg_info *ctx_pg,
4729 				  uint32_t mem_size,
4730 				  const char *suffix,
4731 				  uint16_t idx)
4732 {
4733 	struct bnxt_ring_mem_info *rmem = &ctx_pg->ring_mem;
4734 	const struct rte_memzone *mz = NULL;
4735 	char mz_name[RTE_MEMZONE_NAMESIZE];
4736 	rte_iova_t mz_phys_addr;
4737 	uint64_t valid_bits = 0;
4738 	uint32_t sz;
4739 	int i;
4740 
4741 	if (!mem_size)
4742 		return 0;
4743 
4744 	rmem->nr_pages = RTE_ALIGN_MUL_CEIL(mem_size, BNXT_PAGE_SIZE) /
4745 			 BNXT_PAGE_SIZE;
4746 	rmem->page_size = BNXT_PAGE_SIZE;
4747 	rmem->pg_arr = ctx_pg->ctx_pg_arr;
4748 	rmem->dma_arr = ctx_pg->ctx_dma_arr;
4749 	rmem->flags = BNXT_RMEM_VALID_PTE_FLAG;
4750 
4751 	valid_bits = PTU_PTE_VALID;
4752 
4753 	if (rmem->nr_pages > 1) {
4754 		snprintf(mz_name, RTE_MEMZONE_NAMESIZE,
4755 			 "bnxt_ctx_pg_tbl%s_%x_%d",
4756 			 suffix, idx, bp->eth_dev->data->port_id);
4757 		mz_name[RTE_MEMZONE_NAMESIZE - 1] = 0;
4758 		mz = rte_memzone_lookup(mz_name);
4759 		if (!mz) {
4760 			mz = rte_memzone_reserve_aligned(mz_name,
4761 						rmem->nr_pages * 8,
4762 						SOCKET_ID_ANY,
4763 						RTE_MEMZONE_2MB |
4764 						RTE_MEMZONE_SIZE_HINT_ONLY |
4765 						RTE_MEMZONE_IOVA_CONTIG,
4766 						BNXT_PAGE_SIZE);
4767 			if (mz == NULL)
4768 				return -ENOMEM;
4769 		}
4770 
4771 		memset(mz->addr, 0, mz->len);
4772 		mz_phys_addr = mz->iova;
4773 
4774 		rmem->pg_tbl = mz->addr;
4775 		rmem->pg_tbl_map = mz_phys_addr;
4776 		rmem->pg_tbl_mz = mz;
4777 	}
4778 
4779 	snprintf(mz_name, RTE_MEMZONE_NAMESIZE, "bnxt_ctx_%s_%x_%d",
4780 		 suffix, idx, bp->eth_dev->data->port_id);
4781 	mz = rte_memzone_lookup(mz_name);
4782 	if (!mz) {
4783 		mz = rte_memzone_reserve_aligned(mz_name,
4784 						 mem_size,
4785 						 SOCKET_ID_ANY,
4786 						 RTE_MEMZONE_1GB |
4787 						 RTE_MEMZONE_SIZE_HINT_ONLY |
4788 						 RTE_MEMZONE_IOVA_CONTIG,
4789 						 BNXT_PAGE_SIZE);
4790 		if (mz == NULL)
4791 			return -ENOMEM;
4792 	}
4793 
4794 	memset(mz->addr, 0, mz->len);
4795 	mz_phys_addr = mz->iova;
4796 
4797 	for (sz = 0, i = 0; sz < mem_size; sz += BNXT_PAGE_SIZE, i++) {
4798 		rmem->pg_arr[i] = ((char *)mz->addr) + sz;
4799 		rmem->dma_arr[i] = mz_phys_addr + sz;
4800 
4801 		if (rmem->nr_pages > 1) {
4802 			if (i == rmem->nr_pages - 2 &&
4803 			    (rmem->flags & BNXT_RMEM_RING_PTE_FLAG))
4804 				valid_bits |= PTU_PTE_NEXT_TO_LAST;
4805 			else if (i == rmem->nr_pages - 1 &&
4806 				 (rmem->flags & BNXT_RMEM_RING_PTE_FLAG))
4807 				valid_bits |= PTU_PTE_LAST;
4808 
4809 			rmem->pg_tbl[i] = rte_cpu_to_le_64(rmem->dma_arr[i] |
4810 							   valid_bits);
4811 		}
4812 	}
4813 
4814 	rmem->mz = mz;
4815 	if (rmem->vmem_size)
4816 		rmem->vmem = (void **)mz->addr;
4817 	rmem->dma_arr[0] = mz_phys_addr;
4818 	return 0;
4819 }
4820 
4821 static void bnxt_free_ctx_mem(struct bnxt *bp)
4822 {
4823 	int i;
4824 
4825 	if (!bp->ctx || !(bp->ctx->flags & BNXT_CTX_FLAG_INITED))
4826 		return;
4827 
4828 	bp->ctx->flags &= ~BNXT_CTX_FLAG_INITED;
4829 	rte_memzone_free(bp->ctx->qp_mem.ring_mem.mz);
4830 	rte_memzone_free(bp->ctx->srq_mem.ring_mem.mz);
4831 	rte_memzone_free(bp->ctx->cq_mem.ring_mem.mz);
4832 	rte_memzone_free(bp->ctx->vnic_mem.ring_mem.mz);
4833 	rte_memzone_free(bp->ctx->stat_mem.ring_mem.mz);
4834 	rte_memzone_free(bp->ctx->qp_mem.ring_mem.pg_tbl_mz);
4835 	rte_memzone_free(bp->ctx->srq_mem.ring_mem.pg_tbl_mz);
4836 	rte_memzone_free(bp->ctx->cq_mem.ring_mem.pg_tbl_mz);
4837 	rte_memzone_free(bp->ctx->vnic_mem.ring_mem.pg_tbl_mz);
4838 	rte_memzone_free(bp->ctx->stat_mem.ring_mem.pg_tbl_mz);
4839 
4840 	for (i = 0; i < bp->ctx->tqm_fp_rings_count + 1; i++) {
4841 		if (bp->ctx->tqm_mem[i])
4842 			rte_memzone_free(bp->ctx->tqm_mem[i]->ring_mem.mz);
4843 	}
4844 
4845 	rte_free(bp->ctx);
4846 	bp->ctx = NULL;
4847 }
4848 
4849 #define bnxt_roundup(x, y)   ((((x) + ((y) - 1)) / (y)) * (y))
4850 
4851 #define min_t(type, x, y) ({                    \
4852 	type __min1 = (x);                      \
4853 	type __min2 = (y);                      \
4854 	__min1 < __min2 ? __min1 : __min2; })
4855 
4856 #define max_t(type, x, y) ({                    \
4857 	type __max1 = (x);                      \
4858 	type __max2 = (y);                      \
4859 	__max1 > __max2 ? __max1 : __max2; })
4860 
4861 #define clamp_t(type, _x, min, max)     min_t(type, max_t(type, _x, min), max)
4862 
4863 int bnxt_alloc_ctx_mem(struct bnxt *bp)
4864 {
4865 	struct bnxt_ctx_pg_info *ctx_pg;
4866 	struct bnxt_ctx_mem_info *ctx;
4867 	uint32_t mem_size, ena, entries;
4868 	uint32_t entries_sp, min;
4869 	int i, rc;
4870 
4871 	rc = bnxt_hwrm_func_backing_store_qcaps(bp);
4872 	if (rc) {
4873 		PMD_DRV_LOG(ERR, "Query context mem capability failed\n");
4874 		return rc;
4875 	}
4876 	ctx = bp->ctx;
4877 	if (!ctx || (ctx->flags & BNXT_CTX_FLAG_INITED))
4878 		return 0;
4879 
4880 	ctx_pg = &ctx->qp_mem;
4881 	ctx_pg->entries = ctx->qp_min_qp1_entries + ctx->qp_max_l2_entries;
4882 	mem_size = ctx->qp_entry_size * ctx_pg->entries;
4883 	rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "qp_mem", 0);
4884 	if (rc)
4885 		return rc;
4886 
4887 	ctx_pg = &ctx->srq_mem;
4888 	ctx_pg->entries = ctx->srq_max_l2_entries;
4889 	mem_size = ctx->srq_entry_size * ctx_pg->entries;
4890 	rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "srq_mem", 0);
4891 	if (rc)
4892 		return rc;
4893 
4894 	ctx_pg = &ctx->cq_mem;
4895 	ctx_pg->entries = ctx->cq_max_l2_entries;
4896 	mem_size = ctx->cq_entry_size * ctx_pg->entries;
4897 	rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "cq_mem", 0);
4898 	if (rc)
4899 		return rc;
4900 
4901 	ctx_pg = &ctx->vnic_mem;
4902 	ctx_pg->entries = ctx->vnic_max_vnic_entries +
4903 		ctx->vnic_max_ring_table_entries;
4904 	mem_size = ctx->vnic_entry_size * ctx_pg->entries;
4905 	rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "vnic_mem", 0);
4906 	if (rc)
4907 		return rc;
4908 
4909 	ctx_pg = &ctx->stat_mem;
4910 	ctx_pg->entries = ctx->stat_max_entries;
4911 	mem_size = ctx->stat_entry_size * ctx_pg->entries;
4912 	rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "stat_mem", 0);
4913 	if (rc)
4914 		return rc;
4915 
4916 	min = ctx->tqm_min_entries_per_ring;
4917 
4918 	entries_sp = ctx->qp_max_l2_entries +
4919 		     ctx->vnic_max_vnic_entries +
4920 		     2 * ctx->qp_min_qp1_entries + min;
4921 	entries_sp = bnxt_roundup(entries_sp, ctx->tqm_entries_multiple);
4922 
4923 	entries = ctx->qp_max_l2_entries + ctx->qp_min_qp1_entries;
4924 	entries = bnxt_roundup(entries, ctx->tqm_entries_multiple);
4925 	entries = clamp_t(uint32_t, entries, min,
4926 			  ctx->tqm_max_entries_per_ring);
4927 	for (i = 0, ena = 0; i < ctx->tqm_fp_rings_count + 1; i++) {
4928 		ctx_pg = ctx->tqm_mem[i];
4929 		ctx_pg->entries = i ? entries : entries_sp;
4930 		mem_size = ctx->tqm_entry_size * ctx_pg->entries;
4931 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "tqm_mem", i);
4932 		if (rc)
4933 			return rc;
4934 		ena |= HWRM_FUNC_BACKING_STORE_CFG_INPUT_ENABLES_TQM_SP << i;
4935 	}
4936 
4937 	ena |= FUNC_BACKING_STORE_CFG_INPUT_DFLT_ENABLES;
4938 	rc = bnxt_hwrm_func_backing_store_cfg(bp, ena);
4939 	if (rc)
4940 		PMD_DRV_LOG(ERR,
4941 			    "Failed to configure context mem: rc = %d\n", rc);
4942 	else
4943 		ctx->flags |= BNXT_CTX_FLAG_INITED;
4944 
4945 	return rc;
4946 }
4947 
4948 static int bnxt_alloc_stats_mem(struct bnxt *bp)
4949 {
4950 	struct rte_pci_device *pci_dev = bp->pdev;
4951 	char mz_name[RTE_MEMZONE_NAMESIZE];
4952 	const struct rte_memzone *mz = NULL;
4953 	uint32_t total_alloc_len;
4954 	rte_iova_t mz_phys_addr;
4955 
4956 	if (pci_dev->id.device_id == BROADCOM_DEV_ID_NS2)
4957 		return 0;
4958 
4959 	snprintf(mz_name, RTE_MEMZONE_NAMESIZE,
4960 		 "bnxt_" PCI_PRI_FMT "-%s", pci_dev->addr.domain,
4961 		 pci_dev->addr.bus, pci_dev->addr.devid,
4962 		 pci_dev->addr.function, "rx_port_stats");
4963 	mz_name[RTE_MEMZONE_NAMESIZE - 1] = 0;
4964 	mz = rte_memzone_lookup(mz_name);
4965 	total_alloc_len =
4966 		RTE_CACHE_LINE_ROUNDUP(sizeof(struct rx_port_stats) +
4967 				       sizeof(struct rx_port_stats_ext) + 512);
4968 	if (!mz) {
4969 		mz = rte_memzone_reserve(mz_name, total_alloc_len,
4970 					 SOCKET_ID_ANY,
4971 					 RTE_MEMZONE_2MB |
4972 					 RTE_MEMZONE_SIZE_HINT_ONLY |
4973 					 RTE_MEMZONE_IOVA_CONTIG);
4974 		if (mz == NULL)
4975 			return -ENOMEM;
4976 	}
4977 	memset(mz->addr, 0, mz->len);
4978 	mz_phys_addr = mz->iova;
4979 
4980 	bp->rx_mem_zone = (const void *)mz;
4981 	bp->hw_rx_port_stats = mz->addr;
4982 	bp->hw_rx_port_stats_map = mz_phys_addr;
4983 
4984 	snprintf(mz_name, RTE_MEMZONE_NAMESIZE,
4985 		 "bnxt_" PCI_PRI_FMT "-%s", pci_dev->addr.domain,
4986 		 pci_dev->addr.bus, pci_dev->addr.devid,
4987 		 pci_dev->addr.function, "tx_port_stats");
4988 	mz_name[RTE_MEMZONE_NAMESIZE - 1] = 0;
4989 	mz = rte_memzone_lookup(mz_name);
4990 	total_alloc_len =
4991 		RTE_CACHE_LINE_ROUNDUP(sizeof(struct tx_port_stats) +
4992 				       sizeof(struct tx_port_stats_ext) + 512);
4993 	if (!mz) {
4994 		mz = rte_memzone_reserve(mz_name,
4995 					 total_alloc_len,
4996 					 SOCKET_ID_ANY,
4997 					 RTE_MEMZONE_2MB |
4998 					 RTE_MEMZONE_SIZE_HINT_ONLY |
4999 					 RTE_MEMZONE_IOVA_CONTIG);
5000 		if (mz == NULL)
5001 			return -ENOMEM;
5002 	}
5003 	memset(mz->addr, 0, mz->len);
5004 	mz_phys_addr = mz->iova;
5005 
5006 	bp->tx_mem_zone = (const void *)mz;
5007 	bp->hw_tx_port_stats = mz->addr;
5008 	bp->hw_tx_port_stats_map = mz_phys_addr;
5009 	bp->flags |= BNXT_FLAG_PORT_STATS;
5010 
5011 	/* Display extended statistics if FW supports it */
5012 	if (bp->hwrm_spec_code < HWRM_SPEC_CODE_1_8_4 ||
5013 	    bp->hwrm_spec_code == HWRM_SPEC_CODE_1_9_0 ||
5014 	    !(bp->flags & BNXT_FLAG_EXT_STATS_SUPPORTED))
5015 		return 0;
5016 
5017 	bp->hw_rx_port_stats_ext = (void *)
5018 		((uint8_t *)bp->hw_rx_port_stats +
5019 		 sizeof(struct rx_port_stats));
5020 	bp->hw_rx_port_stats_ext_map = bp->hw_rx_port_stats_map +
5021 		sizeof(struct rx_port_stats);
5022 	bp->flags |= BNXT_FLAG_EXT_RX_PORT_STATS;
5023 
5024 	if (bp->hwrm_spec_code < HWRM_SPEC_CODE_1_9_2 ||
5025 	    bp->flags & BNXT_FLAG_EXT_STATS_SUPPORTED) {
5026 		bp->hw_tx_port_stats_ext = (void *)
5027 			((uint8_t *)bp->hw_tx_port_stats +
5028 			 sizeof(struct tx_port_stats));
5029 		bp->hw_tx_port_stats_ext_map =
5030 			bp->hw_tx_port_stats_map +
5031 			sizeof(struct tx_port_stats);
5032 		bp->flags |= BNXT_FLAG_EXT_TX_PORT_STATS;
5033 	}
5034 
5035 	return 0;
5036 }
5037 
5038 static int bnxt_setup_mac_addr(struct rte_eth_dev *eth_dev)
5039 {
5040 	struct bnxt *bp = eth_dev->data->dev_private;
5041 	int rc = 0;
5042 
5043 	eth_dev->data->mac_addrs = rte_zmalloc("bnxt_mac_addr_tbl",
5044 					       RTE_ETHER_ADDR_LEN *
5045 					       bp->max_l2_ctx,
5046 					       0);
5047 	if (eth_dev->data->mac_addrs == NULL) {
5048 		PMD_DRV_LOG(ERR, "Failed to alloc MAC addr tbl\n");
5049 		return -ENOMEM;
5050 	}
5051 
5052 	if (!BNXT_HAS_DFLT_MAC_SET(bp)) {
5053 		if (BNXT_PF(bp))
5054 			return -EINVAL;
5055 
5056 		/* Generate a random MAC address, if none was assigned by PF */
5057 		PMD_DRV_LOG(INFO, "VF MAC address not assigned by Host PF\n");
5058 		bnxt_eth_hw_addr_random(bp->mac_addr);
5059 		PMD_DRV_LOG(INFO,
5060 			    "Assign random MAC:%02X:%02X:%02X:%02X:%02X:%02X\n",
5061 			    bp->mac_addr[0], bp->mac_addr[1], bp->mac_addr[2],
5062 			    bp->mac_addr[3], bp->mac_addr[4], bp->mac_addr[5]);
5063 
5064 		rc = bnxt_hwrm_set_mac(bp);
5065 		if (rc)
5066 			return rc;
5067 	}
5068 
5069 	/* Copy the permanent MAC from the FUNC_QCAPS response */
5070 	memcpy(&eth_dev->data->mac_addrs[0], bp->mac_addr, RTE_ETHER_ADDR_LEN);
5071 
5072 	return rc;
5073 }
5074 
5075 static int bnxt_restore_dflt_mac(struct bnxt *bp)
5076 {
5077 	int rc = 0;
5078 
5079 	/* MAC is already configured in FW */
5080 	if (BNXT_HAS_DFLT_MAC_SET(bp))
5081 		return 0;
5082 
5083 	/* Restore the old MAC configured */
5084 	rc = bnxt_hwrm_set_mac(bp);
5085 	if (rc)
5086 		PMD_DRV_LOG(ERR, "Failed to restore MAC address\n");
5087 
5088 	return rc;
5089 }
5090 
5091 static void bnxt_config_vf_req_fwd(struct bnxt *bp)
5092 {
5093 	if (!BNXT_PF(bp))
5094 		return;
5095 
5096 #define ALLOW_FUNC(x)	\
5097 	{ \
5098 		uint32_t arg = (x); \
5099 		bp->pf->vf_req_fwd[((arg) >> 5)] &= \
5100 		~rte_cpu_to_le_32(1 << ((arg) & 0x1f)); \
5101 	}
5102 
5103 	/* Forward all requests if firmware is new enough */
5104 	if (((bp->fw_ver >= ((20 << 24) | (6 << 16) | (100 << 8))) &&
5105 	     (bp->fw_ver < ((20 << 24) | (7 << 16)))) ||
5106 	    ((bp->fw_ver >= ((20 << 24) | (8 << 16))))) {
5107 		memset(bp->pf->vf_req_fwd, 0xff, sizeof(bp->pf->vf_req_fwd));
5108 	} else {
5109 		PMD_DRV_LOG(WARNING,
5110 			    "Firmware too old for VF mailbox functionality\n");
5111 		memset(bp->pf->vf_req_fwd, 0, sizeof(bp->pf->vf_req_fwd));
5112 	}
5113 
5114 	/*
5115 	 * The following are used for driver cleanup. If we disallow these,
5116 	 * VF drivers can't clean up cleanly.
5117 	 */
5118 	ALLOW_FUNC(HWRM_FUNC_DRV_UNRGTR);
5119 	ALLOW_FUNC(HWRM_VNIC_FREE);
5120 	ALLOW_FUNC(HWRM_RING_FREE);
5121 	ALLOW_FUNC(HWRM_RING_GRP_FREE);
5122 	ALLOW_FUNC(HWRM_VNIC_RSS_COS_LB_CTX_FREE);
5123 	ALLOW_FUNC(HWRM_CFA_L2_FILTER_FREE);
5124 	ALLOW_FUNC(HWRM_STAT_CTX_FREE);
5125 	ALLOW_FUNC(HWRM_PORT_PHY_QCFG);
5126 	ALLOW_FUNC(HWRM_VNIC_TPA_CFG);
5127 }
5128 
5129 uint16_t
5130 bnxt_get_svif(uint16_t port_id, bool func_svif,
5131 	      enum bnxt_ulp_intf_type type)
5132 {
5133 	struct rte_eth_dev *eth_dev;
5134 	struct bnxt *bp;
5135 
5136 	eth_dev = &rte_eth_devices[port_id];
5137 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
5138 		struct bnxt_vf_representor *vfr = eth_dev->data->dev_private;
5139 		if (!vfr)
5140 			return 0;
5141 
5142 		if (type == BNXT_ULP_INTF_TYPE_VF_REP)
5143 			return vfr->svif;
5144 
5145 		eth_dev = vfr->parent_dev;
5146 	}
5147 
5148 	bp = eth_dev->data->dev_private;
5149 
5150 	return func_svif ? bp->func_svif : bp->port_svif;
5151 }
5152 
5153 uint16_t
5154 bnxt_get_vnic_id(uint16_t port, enum bnxt_ulp_intf_type type)
5155 {
5156 	struct rte_eth_dev *eth_dev;
5157 	struct bnxt_vnic_info *vnic;
5158 	struct bnxt *bp;
5159 
5160 	eth_dev = &rte_eth_devices[port];
5161 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
5162 		struct bnxt_vf_representor *vfr = eth_dev->data->dev_private;
5163 		if (!vfr)
5164 			return 0;
5165 
5166 		if (type == BNXT_ULP_INTF_TYPE_VF_REP)
5167 			return vfr->dflt_vnic_id;
5168 
5169 		eth_dev = vfr->parent_dev;
5170 	}
5171 
5172 	bp = eth_dev->data->dev_private;
5173 
5174 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
5175 
5176 	return vnic->fw_vnic_id;
5177 }
5178 
5179 uint16_t
5180 bnxt_get_fw_func_id(uint16_t port, enum bnxt_ulp_intf_type type)
5181 {
5182 	struct rte_eth_dev *eth_dev;
5183 	struct bnxt *bp;
5184 
5185 	eth_dev = &rte_eth_devices[port];
5186 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
5187 		struct bnxt_vf_representor *vfr = eth_dev->data->dev_private;
5188 		if (!vfr)
5189 			return 0;
5190 
5191 		if (type == BNXT_ULP_INTF_TYPE_VF_REP)
5192 			return vfr->fw_fid;
5193 
5194 		eth_dev = vfr->parent_dev;
5195 	}
5196 
5197 	bp = eth_dev->data->dev_private;
5198 
5199 	return bp->fw_fid;
5200 }
5201 
5202 enum bnxt_ulp_intf_type
5203 bnxt_get_interface_type(uint16_t port)
5204 {
5205 	struct rte_eth_dev *eth_dev;
5206 	struct bnxt *bp;
5207 
5208 	eth_dev = &rte_eth_devices[port];
5209 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev))
5210 		return BNXT_ULP_INTF_TYPE_VF_REP;
5211 
5212 	bp = eth_dev->data->dev_private;
5213 	if (BNXT_PF(bp))
5214 		return BNXT_ULP_INTF_TYPE_PF;
5215 	else if (BNXT_VF_IS_TRUSTED(bp))
5216 		return BNXT_ULP_INTF_TYPE_TRUSTED_VF;
5217 	else if (BNXT_VF(bp))
5218 		return BNXT_ULP_INTF_TYPE_VF;
5219 
5220 	return BNXT_ULP_INTF_TYPE_INVALID;
5221 }
5222 
5223 uint16_t
5224 bnxt_get_phy_port_id(uint16_t port_id)
5225 {
5226 	struct bnxt_vf_representor *vfr;
5227 	struct rte_eth_dev *eth_dev;
5228 	struct bnxt *bp;
5229 
5230 	eth_dev = &rte_eth_devices[port_id];
5231 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
5232 		vfr = eth_dev->data->dev_private;
5233 		if (!vfr)
5234 			return 0;
5235 
5236 		eth_dev = vfr->parent_dev;
5237 	}
5238 
5239 	bp = eth_dev->data->dev_private;
5240 
5241 	return BNXT_PF(bp) ? bp->pf->port_id : bp->parent->port_id;
5242 }
5243 
5244 uint16_t
5245 bnxt_get_parif(uint16_t port_id, enum bnxt_ulp_intf_type type)
5246 {
5247 	struct rte_eth_dev *eth_dev;
5248 	struct bnxt *bp;
5249 
5250 	eth_dev = &rte_eth_devices[port_id];
5251 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
5252 		struct bnxt_vf_representor *vfr = eth_dev->data->dev_private;
5253 		if (!vfr)
5254 			return 0;
5255 
5256 		if (type == BNXT_ULP_INTF_TYPE_VF_REP)
5257 			return vfr->fw_fid - 1;
5258 
5259 		eth_dev = vfr->parent_dev;
5260 	}
5261 
5262 	bp = eth_dev->data->dev_private;
5263 
5264 	return BNXT_PF(bp) ? bp->fw_fid - 1 : bp->parent->fid - 1;
5265 }
5266 
5267 uint16_t
5268 bnxt_get_vport(uint16_t port_id)
5269 {
5270 	return (1 << bnxt_get_phy_port_id(port_id));
5271 }
5272 
5273 static void bnxt_alloc_error_recovery_info(struct bnxt *bp)
5274 {
5275 	struct bnxt_error_recovery_info *info = bp->recovery_info;
5276 
5277 	if (info) {
5278 		if (!(bp->fw_cap & BNXT_FW_CAP_HCOMM_FW_STATUS))
5279 			memset(info, 0, sizeof(*info));
5280 		return;
5281 	}
5282 
5283 	if (!(bp->fw_cap & BNXT_FW_CAP_ERROR_RECOVERY))
5284 		return;
5285 
5286 	info = rte_zmalloc("bnxt_hwrm_error_recovery_qcfg",
5287 			   sizeof(*info), 0);
5288 	if (!info)
5289 		bp->fw_cap &= ~BNXT_FW_CAP_ERROR_RECOVERY;
5290 
5291 	bp->recovery_info = info;
5292 }
5293 
5294 static void bnxt_check_fw_status(struct bnxt *bp)
5295 {
5296 	uint32_t fw_status;
5297 
5298 	if (!(bp->recovery_info &&
5299 	      (bp->fw_cap & BNXT_FW_CAP_HCOMM_FW_STATUS)))
5300 		return;
5301 
5302 	fw_status = bnxt_read_fw_status_reg(bp, BNXT_FW_STATUS_REG);
5303 	if (fw_status != BNXT_FW_STATUS_HEALTHY)
5304 		PMD_DRV_LOG(ERR, "Firmware not responding, status: %#x\n",
5305 			    fw_status);
5306 }
5307 
5308 static int bnxt_map_hcomm_fw_status_reg(struct bnxt *bp)
5309 {
5310 	struct bnxt_error_recovery_info *info = bp->recovery_info;
5311 	uint32_t status_loc;
5312 	uint32_t sig_ver;
5313 
5314 	rte_write32(HCOMM_STATUS_STRUCT_LOC, (uint8_t *)bp->bar0 +
5315 		    BNXT_GRCPF_REG_WINDOW_BASE_OUT + 4);
5316 	sig_ver = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
5317 				   BNXT_GRCP_WINDOW_2_BASE +
5318 				   offsetof(struct hcomm_status,
5319 					    sig_ver)));
5320 	/* If the signature is absent, then FW does not support this feature */
5321 	if ((sig_ver & HCOMM_STATUS_SIGNATURE_MASK) !=
5322 	    HCOMM_STATUS_SIGNATURE_VAL)
5323 		return 0;
5324 
5325 	if (!info) {
5326 		info = rte_zmalloc("bnxt_hwrm_error_recovery_qcfg",
5327 				   sizeof(*info), 0);
5328 		if (!info)
5329 			return -ENOMEM;
5330 		bp->recovery_info = info;
5331 	} else {
5332 		memset(info, 0, sizeof(*info));
5333 	}
5334 
5335 	status_loc = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
5336 				      BNXT_GRCP_WINDOW_2_BASE +
5337 				      offsetof(struct hcomm_status,
5338 					       fw_status_loc)));
5339 
5340 	/* Only pre-map the FW health status GRC register */
5341 	if (BNXT_FW_STATUS_REG_TYPE(status_loc) != BNXT_FW_STATUS_REG_TYPE_GRC)
5342 		return 0;
5343 
5344 	info->status_regs[BNXT_FW_STATUS_REG] = status_loc;
5345 	info->mapped_status_regs[BNXT_FW_STATUS_REG] =
5346 		BNXT_GRCP_WINDOW_2_BASE + (status_loc & BNXT_GRCP_OFFSET_MASK);
5347 
5348 	rte_write32((status_loc & BNXT_GRCP_BASE_MASK), (uint8_t *)bp->bar0 +
5349 		    BNXT_GRCPF_REG_WINDOW_BASE_OUT + 4);
5350 
5351 	bp->fw_cap |= BNXT_FW_CAP_HCOMM_FW_STATUS;
5352 
5353 	return 0;
5354 }
5355 
5356 static int bnxt_init_fw(struct bnxt *bp)
5357 {
5358 	uint16_t mtu;
5359 	int rc = 0;
5360 
5361 	bp->fw_cap = 0;
5362 
5363 	rc = bnxt_map_hcomm_fw_status_reg(bp);
5364 	if (rc)
5365 		return rc;
5366 
5367 	rc = bnxt_hwrm_ver_get(bp, DFLT_HWRM_CMD_TIMEOUT);
5368 	if (rc) {
5369 		bnxt_check_fw_status(bp);
5370 		return rc;
5371 	}
5372 
5373 	rc = bnxt_hwrm_func_reset(bp);
5374 	if (rc)
5375 		return -EIO;
5376 
5377 	rc = bnxt_hwrm_vnic_qcaps(bp);
5378 	if (rc)
5379 		return rc;
5380 
5381 	rc = bnxt_hwrm_queue_qportcfg(bp);
5382 	if (rc)
5383 		return rc;
5384 
5385 	/* Get the MAX capabilities for this function.
5386 	 * This function also allocates context memory for TQM rings and
5387 	 * informs the firmware about this allocated backing store memory.
5388 	 */
5389 	rc = bnxt_hwrm_func_qcaps(bp);
5390 	if (rc)
5391 		return rc;
5392 
5393 	rc = bnxt_hwrm_func_qcfg(bp, &mtu);
5394 	if (rc)
5395 		return rc;
5396 
5397 	bnxt_hwrm_port_mac_qcfg(bp);
5398 
5399 	bnxt_hwrm_parent_pf_qcfg(bp);
5400 
5401 	bnxt_hwrm_port_phy_qcaps(bp);
5402 
5403 	bnxt_alloc_error_recovery_info(bp);
5404 	/* Get the adapter error recovery support info */
5405 	rc = bnxt_hwrm_error_recovery_qcfg(bp);
5406 	if (rc)
5407 		bp->fw_cap &= ~BNXT_FW_CAP_ERROR_RECOVERY;
5408 
5409 	bnxt_hwrm_port_led_qcaps(bp);
5410 
5411 	return 0;
5412 }
5413 
5414 static int
5415 bnxt_init_locks(struct bnxt *bp)
5416 {
5417 	int err;
5418 
5419 	err = pthread_mutex_init(&bp->flow_lock, NULL);
5420 	if (err) {
5421 		PMD_DRV_LOG(ERR, "Unable to initialize flow_lock\n");
5422 		return err;
5423 	}
5424 
5425 	err = pthread_mutex_init(&bp->def_cp_lock, NULL);
5426 	if (err)
5427 		PMD_DRV_LOG(ERR, "Unable to initialize def_cp_lock\n");
5428 	return err;
5429 }
5430 
5431 static int bnxt_init_resources(struct bnxt *bp, bool reconfig_dev)
5432 {
5433 	int rc = 0;
5434 
5435 	rc = bnxt_init_fw(bp);
5436 	if (rc)
5437 		return rc;
5438 
5439 	if (!reconfig_dev) {
5440 		rc = bnxt_setup_mac_addr(bp->eth_dev);
5441 		if (rc)
5442 			return rc;
5443 	} else {
5444 		rc = bnxt_restore_dflt_mac(bp);
5445 		if (rc)
5446 			return rc;
5447 	}
5448 
5449 	bnxt_config_vf_req_fwd(bp);
5450 
5451 	rc = bnxt_hwrm_func_driver_register(bp);
5452 	if (rc) {
5453 		PMD_DRV_LOG(ERR, "Failed to register driver");
5454 		return -EBUSY;
5455 	}
5456 
5457 	if (BNXT_PF(bp)) {
5458 		if (bp->pdev->max_vfs) {
5459 			rc = bnxt_hwrm_allocate_vfs(bp, bp->pdev->max_vfs);
5460 			if (rc) {
5461 				PMD_DRV_LOG(ERR, "Failed to allocate VFs\n");
5462 				return rc;
5463 			}
5464 		} else {
5465 			rc = bnxt_hwrm_allocate_pf_only(bp);
5466 			if (rc) {
5467 				PMD_DRV_LOG(ERR,
5468 					    "Failed to allocate PF resources");
5469 				return rc;
5470 			}
5471 		}
5472 	}
5473 
5474 	rc = bnxt_alloc_mem(bp, reconfig_dev);
5475 	if (rc)
5476 		return rc;
5477 
5478 	rc = bnxt_setup_int(bp);
5479 	if (rc)
5480 		return rc;
5481 
5482 	rc = bnxt_request_int(bp);
5483 	if (rc)
5484 		return rc;
5485 
5486 	rc = bnxt_init_ctx_mem(bp);
5487 	if (rc) {
5488 		PMD_DRV_LOG(ERR, "Failed to init adv_flow_counters\n");
5489 		return rc;
5490 	}
5491 
5492 	rc = bnxt_init_locks(bp);
5493 	if (rc)
5494 		return rc;
5495 
5496 	return 0;
5497 }
5498 
5499 static int
5500 bnxt_parse_devarg_truflow(__rte_unused const char *key,
5501 			  const char *value, void *opaque_arg)
5502 {
5503 	struct bnxt *bp = opaque_arg;
5504 	unsigned long truflow;
5505 	char *end = NULL;
5506 
5507 	if (!value || !opaque_arg) {
5508 		PMD_DRV_LOG(ERR,
5509 			    "Invalid parameter passed to truflow devargs.\n");
5510 		return -EINVAL;
5511 	}
5512 
5513 	truflow = strtoul(value, &end, 10);
5514 	if (end == NULL || *end != '\0' ||
5515 	    (truflow == ULONG_MAX && errno == ERANGE)) {
5516 		PMD_DRV_LOG(ERR,
5517 			    "Invalid parameter passed to truflow devargs.\n");
5518 		return -EINVAL;
5519 	}
5520 
5521 	if (BNXT_DEVARG_TRUFLOW_INVALID(truflow)) {
5522 		PMD_DRV_LOG(ERR,
5523 			    "Invalid value passed to truflow devargs.\n");
5524 		return -EINVAL;
5525 	}
5526 
5527 	bp->flags |= BNXT_FLAG_TRUFLOW_EN;
5528 	if (BNXT_TRUFLOW_EN(bp))
5529 		PMD_DRV_LOG(INFO, "Host-based truflow feature enabled.\n");
5530 
5531 	return 0;
5532 }
5533 
5534 static int
5535 bnxt_parse_devarg_flow_xstat(__rte_unused const char *key,
5536 			     const char *value, void *opaque_arg)
5537 {
5538 	struct bnxt *bp = opaque_arg;
5539 	unsigned long flow_xstat;
5540 	char *end = NULL;
5541 
5542 	if (!value || !opaque_arg) {
5543 		PMD_DRV_LOG(ERR,
5544 			    "Invalid parameter passed to flow_xstat devarg.\n");
5545 		return -EINVAL;
5546 	}
5547 
5548 	flow_xstat = strtoul(value, &end, 10);
5549 	if (end == NULL || *end != '\0' ||
5550 	    (flow_xstat == ULONG_MAX && errno == ERANGE)) {
5551 		PMD_DRV_LOG(ERR,
5552 			    "Invalid parameter passed to flow_xstat devarg.\n");
5553 		return -EINVAL;
5554 	}
5555 
5556 	if (BNXT_DEVARG_FLOW_XSTAT_INVALID(flow_xstat)) {
5557 		PMD_DRV_LOG(ERR,
5558 			    "Invalid value passed to flow_xstat devarg.\n");
5559 		return -EINVAL;
5560 	}
5561 
5562 	bp->flags |= BNXT_FLAG_FLOW_XSTATS_EN;
5563 	if (BNXT_FLOW_XSTATS_EN(bp))
5564 		PMD_DRV_LOG(INFO, "flow_xstat feature enabled.\n");
5565 
5566 	return 0;
5567 }
5568 
5569 static int
5570 bnxt_parse_devarg_max_num_kflows(__rte_unused const char *key,
5571 					const char *value, void *opaque_arg)
5572 {
5573 	struct bnxt *bp = opaque_arg;
5574 	unsigned long max_num_kflows;
5575 	char *end = NULL;
5576 
5577 	if (!value || !opaque_arg) {
5578 		PMD_DRV_LOG(ERR,
5579 			"Invalid parameter passed to max_num_kflows devarg.\n");
5580 		return -EINVAL;
5581 	}
5582 
5583 	max_num_kflows = strtoul(value, &end, 10);
5584 	if (end == NULL || *end != '\0' ||
5585 		(max_num_kflows == ULONG_MAX && errno == ERANGE)) {
5586 		PMD_DRV_LOG(ERR,
5587 			"Invalid parameter passed to max_num_kflows devarg.\n");
5588 		return -EINVAL;
5589 	}
5590 
5591 	if (bnxt_devarg_max_num_kflow_invalid(max_num_kflows)) {
5592 		PMD_DRV_LOG(ERR,
5593 			"Invalid value passed to max_num_kflows devarg.\n");
5594 		return -EINVAL;
5595 	}
5596 
5597 	bp->max_num_kflows = max_num_kflows;
5598 	if (bp->max_num_kflows)
5599 		PMD_DRV_LOG(INFO, "max_num_kflows set as %ldK.\n",
5600 				max_num_kflows);
5601 
5602 	return 0;
5603 }
5604 
5605 static void
5606 bnxt_parse_dev_args(struct bnxt *bp, struct rte_devargs *devargs)
5607 {
5608 	struct rte_kvargs *kvlist;
5609 
5610 	if (devargs == NULL)
5611 		return;
5612 
5613 	kvlist = rte_kvargs_parse(devargs->args, bnxt_dev_args);
5614 	if (kvlist == NULL)
5615 		return;
5616 
5617 	/*
5618 	 * Handler for "truflow" devarg.
5619 	 * Invoked as for ex: "-w 0000:00:0d.0,host-based-truflow=1"
5620 	 */
5621 	rte_kvargs_process(kvlist, BNXT_DEVARG_TRUFLOW,
5622 			   bnxt_parse_devarg_truflow, bp);
5623 
5624 	/*
5625 	 * Handler for "flow_xstat" devarg.
5626 	 * Invoked as for ex: "-w 0000:00:0d.0,flow_xstat=1"
5627 	 */
5628 	rte_kvargs_process(kvlist, BNXT_DEVARG_FLOW_XSTAT,
5629 			   bnxt_parse_devarg_flow_xstat, bp);
5630 
5631 	/*
5632 	 * Handler for "max_num_kflows" devarg.
5633 	 * Invoked as for ex: "-w 000:00:0d.0,max_num_kflows=32"
5634 	 */
5635 	rte_kvargs_process(kvlist, BNXT_DEVARG_MAX_NUM_KFLOWS,
5636 			   bnxt_parse_devarg_max_num_kflows, bp);
5637 
5638 	rte_kvargs_free(kvlist);
5639 }
5640 
5641 static int bnxt_alloc_switch_domain(struct bnxt *bp)
5642 {
5643 	int rc = 0;
5644 
5645 	if (BNXT_PF(bp) || BNXT_VF_IS_TRUSTED(bp)) {
5646 		rc = rte_eth_switch_domain_alloc(&bp->switch_domain_id);
5647 		if (rc)
5648 			PMD_DRV_LOG(ERR,
5649 				    "Failed to alloc switch domain: %d\n", rc);
5650 		else
5651 			PMD_DRV_LOG(INFO,
5652 				    "Switch domain allocated %d\n",
5653 				    bp->switch_domain_id);
5654 	}
5655 
5656 	return rc;
5657 }
5658 
5659 static int
5660 bnxt_dev_init(struct rte_eth_dev *eth_dev, void *params __rte_unused)
5661 {
5662 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
5663 	static int version_printed;
5664 	struct bnxt *bp;
5665 	int rc;
5666 
5667 	if (version_printed++ == 0)
5668 		PMD_DRV_LOG(INFO, "%s\n", bnxt_version);
5669 
5670 	eth_dev->dev_ops = &bnxt_dev_ops;
5671 	eth_dev->rx_pkt_burst = &bnxt_recv_pkts;
5672 	eth_dev->tx_pkt_burst = &bnxt_xmit_pkts;
5673 
5674 	/*
5675 	 * For secondary processes, we don't initialise any further
5676 	 * as primary has already done this work.
5677 	 */
5678 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
5679 		return 0;
5680 
5681 	rte_eth_copy_pci_info(eth_dev, pci_dev);
5682 
5683 	bp = eth_dev->data->dev_private;
5684 
5685 	/* Parse dev arguments passed on when starting the DPDK application. */
5686 	bnxt_parse_dev_args(bp, pci_dev->device.devargs);
5687 
5688 	bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
5689 
5690 	if (bnxt_vf_pciid(pci_dev->id.device_id))
5691 		bp->flags |= BNXT_FLAG_VF;
5692 
5693 	if (bnxt_thor_device(pci_dev->id.device_id))
5694 		bp->flags |= BNXT_FLAG_THOR_CHIP;
5695 
5696 	if (pci_dev->id.device_id == BROADCOM_DEV_ID_58802 ||
5697 	    pci_dev->id.device_id == BROADCOM_DEV_ID_58804 ||
5698 	    pci_dev->id.device_id == BROADCOM_DEV_ID_58808 ||
5699 	    pci_dev->id.device_id == BROADCOM_DEV_ID_58802_VF)
5700 		bp->flags |= BNXT_FLAG_STINGRAY;
5701 
5702 	rc = bnxt_init_board(eth_dev);
5703 	if (rc) {
5704 		PMD_DRV_LOG(ERR,
5705 			    "Failed to initialize board rc: %x\n", rc);
5706 		return rc;
5707 	}
5708 
5709 	rc = bnxt_alloc_pf_info(bp);
5710 	if (rc)
5711 		goto error_free;
5712 
5713 	rc = bnxt_alloc_link_info(bp);
5714 	if (rc)
5715 		goto error_free;
5716 
5717 	rc = bnxt_alloc_parent_info(bp);
5718 	if (rc)
5719 		goto error_free;
5720 
5721 	rc = bnxt_alloc_hwrm_resources(bp);
5722 	if (rc) {
5723 		PMD_DRV_LOG(ERR,
5724 			    "Failed to allocate hwrm resource rc: %x\n", rc);
5725 		goto error_free;
5726 	}
5727 	rc = bnxt_alloc_leds_info(bp);
5728 	if (rc)
5729 		goto error_free;
5730 
5731 	rc = bnxt_alloc_cos_queues(bp);
5732 	if (rc)
5733 		goto error_free;
5734 
5735 	rc = bnxt_init_resources(bp, false);
5736 	if (rc)
5737 		goto error_free;
5738 
5739 	rc = bnxt_alloc_stats_mem(bp);
5740 	if (rc)
5741 		goto error_free;
5742 
5743 	bnxt_alloc_switch_domain(bp);
5744 
5745 	/* Pass the information to the rte_eth_dev_close() that it should also
5746 	 * release the private port resources.
5747 	 */
5748 	eth_dev->data->dev_flags |= RTE_ETH_DEV_CLOSE_REMOVE;
5749 
5750 	PMD_DRV_LOG(INFO,
5751 		    DRV_MODULE_NAME "found at mem %" PRIX64 ", node addr %pM\n",
5752 		    pci_dev->mem_resource[0].phys_addr,
5753 		    pci_dev->mem_resource[0].addr);
5754 
5755 	return 0;
5756 
5757 error_free:
5758 	bnxt_dev_uninit(eth_dev);
5759 	return rc;
5760 }
5761 
5762 
5763 static void bnxt_free_ctx_mem_buf(struct bnxt_ctx_mem_buf_info *ctx)
5764 {
5765 	if (!ctx)
5766 		return;
5767 
5768 	if (ctx->va)
5769 		rte_free(ctx->va);
5770 
5771 	ctx->va = NULL;
5772 	ctx->dma = RTE_BAD_IOVA;
5773 	ctx->ctx_id = BNXT_CTX_VAL_INVAL;
5774 }
5775 
5776 static void bnxt_unregister_fc_ctx_mem(struct bnxt *bp)
5777 {
5778 	bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_RX,
5779 				  CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
5780 				  bp->flow_stat->rx_fc_out_tbl.ctx_id,
5781 				  bp->flow_stat->max_fc,
5782 				  false);
5783 
5784 	bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_TX,
5785 				  CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
5786 				  bp->flow_stat->tx_fc_out_tbl.ctx_id,
5787 				  bp->flow_stat->max_fc,
5788 				  false);
5789 
5790 	if (bp->flow_stat->rx_fc_in_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
5791 		bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->rx_fc_in_tbl.ctx_id);
5792 	bp->flow_stat->rx_fc_in_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
5793 
5794 	if (bp->flow_stat->rx_fc_out_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
5795 		bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->rx_fc_out_tbl.ctx_id);
5796 	bp->flow_stat->rx_fc_out_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
5797 
5798 	if (bp->flow_stat->tx_fc_in_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
5799 		bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->tx_fc_in_tbl.ctx_id);
5800 	bp->flow_stat->tx_fc_in_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
5801 
5802 	if (bp->flow_stat->tx_fc_out_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
5803 		bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->tx_fc_out_tbl.ctx_id);
5804 	bp->flow_stat->tx_fc_out_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
5805 }
5806 
5807 static void bnxt_uninit_fc_ctx_mem(struct bnxt *bp)
5808 {
5809 	bnxt_unregister_fc_ctx_mem(bp);
5810 
5811 	bnxt_free_ctx_mem_buf(&bp->flow_stat->rx_fc_in_tbl);
5812 	bnxt_free_ctx_mem_buf(&bp->flow_stat->rx_fc_out_tbl);
5813 	bnxt_free_ctx_mem_buf(&bp->flow_stat->tx_fc_in_tbl);
5814 	bnxt_free_ctx_mem_buf(&bp->flow_stat->tx_fc_out_tbl);
5815 }
5816 
5817 static void bnxt_uninit_ctx_mem(struct bnxt *bp)
5818 {
5819 	if (BNXT_FLOW_XSTATS_EN(bp))
5820 		bnxt_uninit_fc_ctx_mem(bp);
5821 }
5822 
5823 static void
5824 bnxt_free_error_recovery_info(struct bnxt *bp)
5825 {
5826 	rte_free(bp->recovery_info);
5827 	bp->recovery_info = NULL;
5828 	bp->fw_cap &= ~BNXT_FW_CAP_ERROR_RECOVERY;
5829 }
5830 
5831 static void
5832 bnxt_uninit_locks(struct bnxt *bp)
5833 {
5834 	pthread_mutex_destroy(&bp->flow_lock);
5835 	pthread_mutex_destroy(&bp->def_cp_lock);
5836 	if (bp->rep_info)
5837 		pthread_mutex_destroy(&bp->rep_info->vfr_lock);
5838 }
5839 
5840 static int
5841 bnxt_uninit_resources(struct bnxt *bp, bool reconfig_dev)
5842 {
5843 	int rc;
5844 
5845 	bnxt_free_int(bp);
5846 	bnxt_free_mem(bp, reconfig_dev);
5847 	bnxt_hwrm_func_buf_unrgtr(bp);
5848 	rc = bnxt_hwrm_func_driver_unregister(bp, 0);
5849 	bp->flags &= ~BNXT_FLAG_REGISTERED;
5850 	bnxt_free_ctx_mem(bp);
5851 	if (!reconfig_dev) {
5852 		bnxt_free_hwrm_resources(bp);
5853 		bnxt_free_error_recovery_info(bp);
5854 	}
5855 
5856 	bnxt_uninit_ctx_mem(bp);
5857 
5858 	bnxt_uninit_locks(bp);
5859 	bnxt_free_flow_stats_info(bp);
5860 	bnxt_free_rep_info(bp);
5861 	rte_free(bp->ptp_cfg);
5862 	bp->ptp_cfg = NULL;
5863 	return rc;
5864 }
5865 
5866 static int
5867 bnxt_dev_uninit(struct rte_eth_dev *eth_dev)
5868 {
5869 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
5870 		return -EPERM;
5871 
5872 	PMD_DRV_LOG(DEBUG, "Calling Device uninit\n");
5873 
5874 	if (eth_dev->state != RTE_ETH_DEV_UNUSED)
5875 		bnxt_dev_close_op(eth_dev);
5876 
5877 	return 0;
5878 }
5879 
5880 static int bnxt_pci_remove_dev_with_reps(struct rte_eth_dev *eth_dev)
5881 {
5882 	struct bnxt *bp = eth_dev->data->dev_private;
5883 	struct rte_eth_dev *vf_rep_eth_dev;
5884 	int ret = 0, i;
5885 
5886 	if (!bp)
5887 		return -EINVAL;
5888 
5889 	for (i = 0; i < bp->num_reps; i++) {
5890 		vf_rep_eth_dev = bp->rep_info[i].vfr_eth_dev;
5891 		if (!vf_rep_eth_dev)
5892 			continue;
5893 		rte_eth_dev_destroy(vf_rep_eth_dev, bnxt_vf_representor_uninit);
5894 	}
5895 	ret = rte_eth_dev_destroy(eth_dev, bnxt_dev_uninit);
5896 
5897 	return ret;
5898 }
5899 
5900 static void bnxt_free_rep_info(struct bnxt *bp)
5901 {
5902 	rte_free(bp->rep_info);
5903 	bp->rep_info = NULL;
5904 	rte_free(bp->cfa_code_map);
5905 	bp->cfa_code_map = NULL;
5906 }
5907 
5908 static int bnxt_init_rep_info(struct bnxt *bp)
5909 {
5910 	int i = 0, rc;
5911 
5912 	if (bp->rep_info)
5913 		return 0;
5914 
5915 	bp->rep_info = rte_zmalloc("bnxt_rep_info",
5916 				   sizeof(bp->rep_info[0]) * BNXT_MAX_VF_REPS,
5917 				   0);
5918 	if (!bp->rep_info) {
5919 		PMD_DRV_LOG(ERR, "Failed to alloc memory for rep info\n");
5920 		return -ENOMEM;
5921 	}
5922 	bp->cfa_code_map = rte_zmalloc("bnxt_cfa_code_map",
5923 				       sizeof(*bp->cfa_code_map) *
5924 				       BNXT_MAX_CFA_CODE, 0);
5925 	if (!bp->cfa_code_map) {
5926 		PMD_DRV_LOG(ERR, "Failed to alloc memory for cfa_code_map\n");
5927 		bnxt_free_rep_info(bp);
5928 		return -ENOMEM;
5929 	}
5930 
5931 	for (i = 0; i < BNXT_MAX_CFA_CODE; i++)
5932 		bp->cfa_code_map[i] = BNXT_VF_IDX_INVALID;
5933 
5934 	rc = pthread_mutex_init(&bp->rep_info->vfr_lock, NULL);
5935 	if (rc) {
5936 		PMD_DRV_LOG(ERR, "Unable to initialize vfr_lock\n");
5937 		bnxt_free_rep_info(bp);
5938 		return rc;
5939 	}
5940 	return rc;
5941 }
5942 
5943 static int bnxt_rep_port_probe(struct rte_pci_device *pci_dev,
5944 			       struct rte_eth_devargs eth_da,
5945 			       struct rte_eth_dev *backing_eth_dev)
5946 {
5947 	struct rte_eth_dev *vf_rep_eth_dev;
5948 	char name[RTE_ETH_NAME_MAX_LEN];
5949 	struct bnxt *backing_bp;
5950 	uint16_t num_rep;
5951 	int i, ret = 0;
5952 
5953 	num_rep = eth_da.nb_representor_ports;
5954 	if (num_rep > BNXT_MAX_VF_REPS) {
5955 		PMD_DRV_LOG(ERR, "nb_representor_ports = %d > %d MAX VF REPS\n",
5956 			    num_rep, BNXT_MAX_VF_REPS);
5957 		return -EINVAL;
5958 	}
5959 
5960 	if (num_rep > RTE_MAX_ETHPORTS) {
5961 		PMD_DRV_LOG(ERR,
5962 			    "nb_representor_ports = %d > %d MAX ETHPORTS\n",
5963 			    num_rep, RTE_MAX_ETHPORTS);
5964 		return -EINVAL;
5965 	}
5966 
5967 	backing_bp = backing_eth_dev->data->dev_private;
5968 
5969 	if (!(BNXT_PF(backing_bp) || BNXT_VF_IS_TRUSTED(backing_bp))) {
5970 		PMD_DRV_LOG(ERR,
5971 			    "Not a PF or trusted VF. No Representor support\n");
5972 		/* Returning an error is not an option.
5973 		 * Applications are not handling this correctly
5974 		 */
5975 		return 0;
5976 	}
5977 
5978 	if (bnxt_init_rep_info(backing_bp))
5979 		return 0;
5980 
5981 	for (i = 0; i < num_rep; i++) {
5982 		struct bnxt_vf_representor representor = {
5983 			.vf_id = eth_da.representor_ports[i],
5984 			.switch_domain_id = backing_bp->switch_domain_id,
5985 			.parent_dev = backing_eth_dev
5986 		};
5987 
5988 		if (representor.vf_id >= BNXT_MAX_VF_REPS) {
5989 			PMD_DRV_LOG(ERR, "VF-Rep id %d >= %d MAX VF ID\n",
5990 				    representor.vf_id, BNXT_MAX_VF_REPS);
5991 			continue;
5992 		}
5993 
5994 		/* representor port net_bdf_port */
5995 		snprintf(name, sizeof(name), "net_%s_representor_%d",
5996 			 pci_dev->device.name, eth_da.representor_ports[i]);
5997 
5998 		ret = rte_eth_dev_create(&pci_dev->device, name,
5999 					 sizeof(struct bnxt_vf_representor),
6000 					 NULL, NULL,
6001 					 bnxt_vf_representor_init,
6002 					 &representor);
6003 
6004 		if (!ret) {
6005 			vf_rep_eth_dev = rte_eth_dev_allocated(name);
6006 			if (!vf_rep_eth_dev) {
6007 				PMD_DRV_LOG(ERR, "Failed to find the eth_dev"
6008 					    " for VF-Rep: %s.", name);
6009 				bnxt_pci_remove_dev_with_reps(backing_eth_dev);
6010 				ret = -ENODEV;
6011 				return ret;
6012 			}
6013 			backing_bp->rep_info[representor.vf_id].vfr_eth_dev =
6014 				vf_rep_eth_dev;
6015 			backing_bp->num_reps++;
6016 		} else {
6017 			PMD_DRV_LOG(ERR, "failed to create bnxt vf "
6018 				    "representor %s.", name);
6019 			bnxt_pci_remove_dev_with_reps(backing_eth_dev);
6020 		}
6021 	}
6022 
6023 	return ret;
6024 }
6025 
6026 static int bnxt_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
6027 			  struct rte_pci_device *pci_dev)
6028 {
6029 	struct rte_eth_devargs eth_da = { .nb_representor_ports = 0 };
6030 	struct rte_eth_dev *backing_eth_dev;
6031 	uint16_t num_rep;
6032 	int ret = 0;
6033 
6034 	if (pci_dev->device.devargs) {
6035 		ret = rte_eth_devargs_parse(pci_dev->device.devargs->args,
6036 					    &eth_da);
6037 		if (ret)
6038 			return ret;
6039 	}
6040 
6041 	num_rep = eth_da.nb_representor_ports;
6042 	PMD_DRV_LOG(DEBUG, "nb_representor_ports = %d\n",
6043 		    num_rep);
6044 
6045 	/* We could come here after first level of probe is already invoked
6046 	 * as part of an application bringup(OVS-DPDK vswitchd), so first check
6047 	 * for already allocated eth_dev for the backing device (PF/Trusted VF)
6048 	 */
6049 	backing_eth_dev = rte_eth_dev_allocated(pci_dev->device.name);
6050 	if (backing_eth_dev == NULL) {
6051 		ret = rte_eth_dev_create(&pci_dev->device, pci_dev->device.name,
6052 					 sizeof(struct bnxt),
6053 					 eth_dev_pci_specific_init, pci_dev,
6054 					 bnxt_dev_init, NULL);
6055 
6056 		if (ret || !num_rep)
6057 			return ret;
6058 
6059 		backing_eth_dev = rte_eth_dev_allocated(pci_dev->device.name);
6060 	}
6061 
6062 	/* probe representor ports now */
6063 	ret = bnxt_rep_port_probe(pci_dev, eth_da, backing_eth_dev);
6064 
6065 	return ret;
6066 }
6067 
6068 static int bnxt_pci_remove(struct rte_pci_device *pci_dev)
6069 {
6070 	struct rte_eth_dev *eth_dev;
6071 
6072 	eth_dev = rte_eth_dev_allocated(pci_dev->device.name);
6073 	if (!eth_dev)
6074 		return 0; /* Invoked typically only by OVS-DPDK, by the
6075 			   * time it comes here the eth_dev is already
6076 			   * deleted by rte_eth_dev_close(), so returning
6077 			   * +ve value will at least help in proper cleanup
6078 			   */
6079 
6080 	if (rte_eal_process_type() == RTE_PROC_PRIMARY) {
6081 		if (eth_dev->data->dev_flags & RTE_ETH_DEV_REPRESENTOR)
6082 			return rte_eth_dev_destroy(eth_dev,
6083 						   bnxt_vf_representor_uninit);
6084 		else
6085 			return rte_eth_dev_destroy(eth_dev,
6086 						   bnxt_dev_uninit);
6087 	} else {
6088 		return rte_eth_dev_pci_generic_remove(pci_dev, NULL);
6089 	}
6090 }
6091 
6092 static struct rte_pci_driver bnxt_rte_pmd = {
6093 	.id_table = bnxt_pci_id_map,
6094 	.drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC |
6095 			RTE_PCI_DRV_PROBE_AGAIN, /* Needed in case of VF-REPs
6096 						  * and OVS-DPDK
6097 						  */
6098 	.probe = bnxt_pci_probe,
6099 	.remove = bnxt_pci_remove,
6100 };
6101 
6102 static bool
6103 is_device_supported(struct rte_eth_dev *dev, struct rte_pci_driver *drv)
6104 {
6105 	if (strcmp(dev->device->driver->name, drv->driver.name))
6106 		return false;
6107 
6108 	return true;
6109 }
6110 
6111 bool is_bnxt_supported(struct rte_eth_dev *dev)
6112 {
6113 	return is_device_supported(dev, &bnxt_rte_pmd);
6114 }
6115 
6116 RTE_LOG_REGISTER(bnxt_logtype_driver, pmd.net.bnxt.driver, NOTICE);
6117 RTE_PMD_REGISTER_PCI(net_bnxt, bnxt_rte_pmd);
6118 RTE_PMD_REGISTER_PCI_TABLE(net_bnxt, bnxt_pci_id_map);
6119 RTE_PMD_REGISTER_KMOD_DEP(net_bnxt, "* igb_uio | uio_pci_generic | vfio-pci");
6120