xref: /dpdk/drivers/net/bnxt/bnxt_ethdev.c (revision 41a9d881cb86162e8b4e5795f4fcc86281ebaa05)
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 
16 #include "bnxt.h"
17 #include "bnxt_filter.h"
18 #include "bnxt_hwrm.h"
19 #include "bnxt_irq.h"
20 #include "bnxt_ring.h"
21 #include "bnxt_rxq.h"
22 #include "bnxt_rxr.h"
23 #include "bnxt_stats.h"
24 #include "bnxt_txq.h"
25 #include "bnxt_txr.h"
26 #include "bnxt_vnic.h"
27 #include "hsi_struct_def_dpdk.h"
28 #include "bnxt_nvm_defs.h"
29 
30 #define DRV_MODULE_NAME		"bnxt"
31 static const char bnxt_version[] =
32 	"Broadcom NetXtreme driver " DRV_MODULE_NAME;
33 int bnxt_logtype_driver;
34 
35 /*
36  * The set of PCI devices this driver supports
37  */
38 static const struct rte_pci_id bnxt_pci_id_map[] = {
39 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM,
40 			 BROADCOM_DEV_ID_STRATUS_NIC_VF1) },
41 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM,
42 			 BROADCOM_DEV_ID_STRATUS_NIC_VF2) },
43 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_STRATUS_NIC) },
44 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57414_VF) },
45 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57301) },
46 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57302) },
47 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57304_PF) },
48 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57304_VF) },
49 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_NS2) },
50 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57402) },
51 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57404) },
52 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57406_PF) },
53 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57406_VF) },
54 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57402_MF) },
55 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57407_RJ45) },
56 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57404_MF) },
57 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57406_MF) },
58 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57407_SFP) },
59 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57407_MF) },
60 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_5741X_VF) },
61 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_5731X_VF) },
62 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57314) },
63 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57417_MF) },
64 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57311) },
65 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57312) },
66 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57412) },
67 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57414) },
68 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57416_RJ45) },
69 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57417_RJ45) },
70 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57412_MF) },
71 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57317_RJ45) },
72 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57417_SFP) },
73 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57416_SFP) },
74 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57317_SFP) },
75 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57414_MF) },
76 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57416_MF) },
77 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58802) },
78 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58804) },
79 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58808) },
80 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58802_VF) },
81 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57508) },
82 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57504) },
83 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57502) },
84 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57500_VF1) },
85 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57500_VF2) },
86 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57508_MF1) },
87 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57504_MF1) },
88 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57502_MF1) },
89 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57508_MF2) },
90 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57504_MF2) },
91 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57502_MF2) },
92 	{ .vendor_id = 0, /* sentinel */ },
93 };
94 
95 #define BNXT_ETH_RSS_SUPPORT (	\
96 	ETH_RSS_IPV4 |		\
97 	ETH_RSS_NONFRAG_IPV4_TCP |	\
98 	ETH_RSS_NONFRAG_IPV4_UDP |	\
99 	ETH_RSS_IPV6 |		\
100 	ETH_RSS_NONFRAG_IPV6_TCP |	\
101 	ETH_RSS_NONFRAG_IPV6_UDP)
102 
103 #define BNXT_DEV_TX_OFFLOAD_SUPPORT (DEV_TX_OFFLOAD_VLAN_INSERT | \
104 				     DEV_TX_OFFLOAD_IPV4_CKSUM | \
105 				     DEV_TX_OFFLOAD_TCP_CKSUM | \
106 				     DEV_TX_OFFLOAD_UDP_CKSUM | \
107 				     DEV_TX_OFFLOAD_TCP_TSO | \
108 				     DEV_TX_OFFLOAD_OUTER_IPV4_CKSUM | \
109 				     DEV_TX_OFFLOAD_VXLAN_TNL_TSO | \
110 				     DEV_TX_OFFLOAD_GRE_TNL_TSO | \
111 				     DEV_TX_OFFLOAD_IPIP_TNL_TSO | \
112 				     DEV_TX_OFFLOAD_GENEVE_TNL_TSO | \
113 				     DEV_TX_OFFLOAD_QINQ_INSERT | \
114 				     DEV_TX_OFFLOAD_MULTI_SEGS)
115 
116 #define BNXT_DEV_RX_OFFLOAD_SUPPORT (DEV_RX_OFFLOAD_VLAN_FILTER | \
117 				     DEV_RX_OFFLOAD_VLAN_STRIP | \
118 				     DEV_RX_OFFLOAD_IPV4_CKSUM | \
119 				     DEV_RX_OFFLOAD_UDP_CKSUM | \
120 				     DEV_RX_OFFLOAD_TCP_CKSUM | \
121 				     DEV_RX_OFFLOAD_OUTER_IPV4_CKSUM | \
122 				     DEV_RX_OFFLOAD_JUMBO_FRAME | \
123 				     DEV_RX_OFFLOAD_KEEP_CRC | \
124 				     DEV_RX_OFFLOAD_VLAN_EXTEND | \
125 				     DEV_RX_OFFLOAD_TCP_LRO | \
126 				     DEV_RX_OFFLOAD_SCATTER | \
127 				     DEV_RX_OFFLOAD_RSS_HASH)
128 
129 static int bnxt_vlan_offload_set_op(struct rte_eth_dev *dev, int mask);
130 static void bnxt_print_link_info(struct rte_eth_dev *eth_dev);
131 static int bnxt_dev_uninit(struct rte_eth_dev *eth_dev);
132 static int bnxt_init_resources(struct bnxt *bp, bool reconfig_dev);
133 static int bnxt_uninit_resources(struct bnxt *bp, bool reconfig_dev);
134 static void bnxt_cancel_fw_health_check(struct bnxt *bp);
135 
136 int is_bnxt_in_error(struct bnxt *bp)
137 {
138 	if (bp->flags & BNXT_FLAG_FATAL_ERROR)
139 		return -EIO;
140 	if (bp->flags & BNXT_FLAG_FW_RESET)
141 		return -EBUSY;
142 
143 	return 0;
144 }
145 
146 /***********************/
147 
148 /*
149  * High level utility functions
150  */
151 
152 uint16_t bnxt_rss_ctxts(const struct bnxt *bp)
153 {
154 	if (!BNXT_CHIP_THOR(bp))
155 		return 1;
156 
157 	return RTE_ALIGN_MUL_CEIL(bp->rx_nr_rings,
158 				  BNXT_RSS_ENTRIES_PER_CTX_THOR) /
159 				    BNXT_RSS_ENTRIES_PER_CTX_THOR;
160 }
161 
162 static uint16_t  bnxt_rss_hash_tbl_size(const struct bnxt *bp)
163 {
164 	if (!BNXT_CHIP_THOR(bp))
165 		return HW_HASH_INDEX_SIZE;
166 
167 	return bnxt_rss_ctxts(bp) * BNXT_RSS_ENTRIES_PER_CTX_THOR;
168 }
169 
170 static void bnxt_free_mem(struct bnxt *bp, bool reconfig)
171 {
172 	bnxt_free_filter_mem(bp);
173 	bnxt_free_vnic_attributes(bp);
174 	bnxt_free_vnic_mem(bp);
175 
176 	/* tx/rx rings are configured as part of *_queue_setup callbacks.
177 	 * If the number of rings change across fw update,
178 	 * we don't have much choice except to warn the user.
179 	 */
180 	if (!reconfig) {
181 		bnxt_free_stats(bp);
182 		bnxt_free_tx_rings(bp);
183 		bnxt_free_rx_rings(bp);
184 	}
185 	bnxt_free_async_cp_ring(bp);
186 	bnxt_free_rxtx_nq_ring(bp);
187 
188 	rte_free(bp->grp_info);
189 	bp->grp_info = NULL;
190 }
191 
192 static int bnxt_alloc_mem(struct bnxt *bp, bool reconfig)
193 {
194 	int rc;
195 
196 	rc = bnxt_alloc_ring_grps(bp);
197 	if (rc)
198 		goto alloc_mem_err;
199 
200 	rc = bnxt_alloc_async_ring_struct(bp);
201 	if (rc)
202 		goto alloc_mem_err;
203 
204 	rc = bnxt_alloc_vnic_mem(bp);
205 	if (rc)
206 		goto alloc_mem_err;
207 
208 	rc = bnxt_alloc_vnic_attributes(bp);
209 	if (rc)
210 		goto alloc_mem_err;
211 
212 	rc = bnxt_alloc_filter_mem(bp);
213 	if (rc)
214 		goto alloc_mem_err;
215 
216 	rc = bnxt_alloc_async_cp_ring(bp);
217 	if (rc)
218 		goto alloc_mem_err;
219 
220 	rc = bnxt_alloc_rxtx_nq_ring(bp);
221 	if (rc)
222 		goto alloc_mem_err;
223 
224 	return 0;
225 
226 alloc_mem_err:
227 	bnxt_free_mem(bp, reconfig);
228 	return rc;
229 }
230 
231 static int bnxt_init_chip(struct bnxt *bp)
232 {
233 	struct bnxt_rx_queue *rxq;
234 	struct rte_eth_link new;
235 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(bp->eth_dev);
236 	struct rte_eth_conf *dev_conf = &bp->eth_dev->data->dev_conf;
237 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
238 	uint64_t rx_offloads = dev_conf->rxmode.offloads;
239 	uint32_t intr_vector = 0;
240 	uint32_t queue_id, base = BNXT_MISC_VEC_ID;
241 	uint32_t vec = BNXT_MISC_VEC_ID;
242 	unsigned int i, j;
243 	int rc;
244 
245 	if (bp->eth_dev->data->mtu > RTE_ETHER_MTU) {
246 		bp->eth_dev->data->dev_conf.rxmode.offloads |=
247 			DEV_RX_OFFLOAD_JUMBO_FRAME;
248 		bp->flags |= BNXT_FLAG_JUMBO;
249 	} else {
250 		bp->eth_dev->data->dev_conf.rxmode.offloads &=
251 			~DEV_RX_OFFLOAD_JUMBO_FRAME;
252 		bp->flags &= ~BNXT_FLAG_JUMBO;
253 	}
254 
255 	/* THOR does not support ring groups.
256 	 * But we will use the array to save RSS context IDs.
257 	 */
258 	if (BNXT_CHIP_THOR(bp))
259 		bp->max_ring_grps = BNXT_MAX_RSS_CTXTS_THOR;
260 
261 	rc = bnxt_alloc_all_hwrm_stat_ctxs(bp);
262 	if (rc) {
263 		PMD_DRV_LOG(ERR, "HWRM stat ctx alloc failure rc: %x\n", rc);
264 		goto err_out;
265 	}
266 
267 	rc = bnxt_alloc_hwrm_rings(bp);
268 	if (rc) {
269 		PMD_DRV_LOG(ERR, "HWRM ring alloc failure rc: %x\n", rc);
270 		goto err_out;
271 	}
272 
273 	rc = bnxt_alloc_all_hwrm_ring_grps(bp);
274 	if (rc) {
275 		PMD_DRV_LOG(ERR, "HWRM ring grp alloc failure: %x\n", rc);
276 		goto err_out;
277 	}
278 
279 	if (!(bp->vnic_cap_flags & BNXT_VNIC_CAP_COS_CLASSIFY))
280 		goto skip_cosq_cfg;
281 
282 	for (j = 0, i = 0; i < BNXT_COS_QUEUE_COUNT; i++) {
283 		if (bp->rx_cos_queue[i].id != 0xff) {
284 			struct bnxt_vnic_info *vnic = &bp->vnic_info[j++];
285 
286 			if (!vnic) {
287 				PMD_DRV_LOG(ERR,
288 					    "Num pools more than FW profile\n");
289 				rc = -EINVAL;
290 				goto err_out;
291 			}
292 			vnic->cos_queue_id = bp->rx_cos_queue[i].id;
293 			bp->rx_cosq_cnt++;
294 		}
295 	}
296 
297 skip_cosq_cfg:
298 	rc = bnxt_mq_rx_configure(bp);
299 	if (rc) {
300 		PMD_DRV_LOG(ERR, "MQ mode configure failure rc: %x\n", rc);
301 		goto err_out;
302 	}
303 
304 	/* VNIC configuration */
305 	for (i = 0; i < bp->nr_vnics; i++) {
306 		struct rte_eth_conf *dev_conf = &bp->eth_dev->data->dev_conf;
307 		struct bnxt_vnic_info *vnic = &bp->vnic_info[i];
308 
309 		rc = bnxt_vnic_grp_alloc(bp, vnic);
310 		if (rc)
311 			goto err_out;
312 
313 		PMD_DRV_LOG(DEBUG, "vnic[%d] = %p vnic->fw_grp_ids = %p\n",
314 			    i, vnic, vnic->fw_grp_ids);
315 
316 		rc = bnxt_hwrm_vnic_alloc(bp, vnic);
317 		if (rc) {
318 			PMD_DRV_LOG(ERR, "HWRM vnic %d alloc failure rc: %x\n",
319 				i, rc);
320 			goto err_out;
321 		}
322 
323 		/* Alloc RSS context only if RSS mode is enabled */
324 		if (dev_conf->rxmode.mq_mode & ETH_MQ_RX_RSS) {
325 			int j, nr_ctxs = bnxt_rss_ctxts(bp);
326 
327 			rc = 0;
328 			for (j = 0; j < nr_ctxs; j++) {
329 				rc = bnxt_hwrm_vnic_ctx_alloc(bp, vnic, j);
330 				if (rc)
331 					break;
332 			}
333 			if (rc) {
334 				PMD_DRV_LOG(ERR,
335 				  "HWRM vnic %d ctx %d alloc failure rc: %x\n",
336 				  i, j, rc);
337 				goto err_out;
338 			}
339 			vnic->num_lb_ctxts = nr_ctxs;
340 		}
341 
342 		/*
343 		 * Firmware sets pf pair in default vnic cfg. If the VLAN strip
344 		 * setting is not available at this time, it will not be
345 		 * configured correctly in the CFA.
346 		 */
347 		if (rx_offloads & DEV_RX_OFFLOAD_VLAN_STRIP)
348 			vnic->vlan_strip = true;
349 		else
350 			vnic->vlan_strip = false;
351 
352 		rc = bnxt_hwrm_vnic_cfg(bp, vnic);
353 		if (rc) {
354 			PMD_DRV_LOG(ERR, "HWRM vnic %d cfg failure rc: %x\n",
355 				i, rc);
356 			goto err_out;
357 		}
358 
359 		rc = bnxt_set_hwrm_vnic_filters(bp, vnic);
360 		if (rc) {
361 			PMD_DRV_LOG(ERR,
362 				"HWRM vnic %d filter failure rc: %x\n",
363 				i, rc);
364 			goto err_out;
365 		}
366 
367 		for (j = 0; j < bp->rx_num_qs_per_vnic; j++) {
368 			rxq = bp->eth_dev->data->rx_queues[j];
369 
370 			PMD_DRV_LOG(DEBUG,
371 				    "rxq[%d]->vnic=%p vnic->fw_grp_ids=%p\n",
372 				    j, rxq->vnic, rxq->vnic->fw_grp_ids);
373 
374 			if (BNXT_HAS_RING_GRPS(bp) && rxq->rx_deferred_start)
375 				rxq->vnic->fw_grp_ids[j] = INVALID_HW_RING_ID;
376 		}
377 
378 		rc = bnxt_vnic_rss_configure(bp, vnic);
379 		if (rc) {
380 			PMD_DRV_LOG(ERR,
381 				    "HWRM vnic set RSS failure rc: %x\n", rc);
382 			goto err_out;
383 		}
384 
385 		bnxt_hwrm_vnic_plcmode_cfg(bp, vnic);
386 
387 		if (bp->eth_dev->data->dev_conf.rxmode.offloads &
388 		    DEV_RX_OFFLOAD_TCP_LRO)
389 			bnxt_hwrm_vnic_tpa_cfg(bp, vnic, 1);
390 		else
391 			bnxt_hwrm_vnic_tpa_cfg(bp, vnic, 0);
392 	}
393 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, &bp->vnic_info[0], 0, NULL);
394 	if (rc) {
395 		PMD_DRV_LOG(ERR,
396 			"HWRM cfa l2 rx mask failure rc: %x\n", rc);
397 		goto err_out;
398 	}
399 
400 	/* check and configure queue intr-vector mapping */
401 	if ((rte_intr_cap_multiple(intr_handle) ||
402 	     !RTE_ETH_DEV_SRIOV(bp->eth_dev).active) &&
403 	    bp->eth_dev->data->dev_conf.intr_conf.rxq != 0) {
404 		intr_vector = bp->eth_dev->data->nb_rx_queues;
405 		PMD_DRV_LOG(DEBUG, "intr_vector = %d\n", intr_vector);
406 		if (intr_vector > bp->rx_cp_nr_rings) {
407 			PMD_DRV_LOG(ERR, "At most %d intr queues supported",
408 					bp->rx_cp_nr_rings);
409 			return -ENOTSUP;
410 		}
411 		rc = rte_intr_efd_enable(intr_handle, intr_vector);
412 		if (rc)
413 			return rc;
414 	}
415 
416 	if (rte_intr_dp_is_en(intr_handle) && !intr_handle->intr_vec) {
417 		intr_handle->intr_vec =
418 			rte_zmalloc("intr_vec",
419 				    bp->eth_dev->data->nb_rx_queues *
420 				    sizeof(int), 0);
421 		if (intr_handle->intr_vec == NULL) {
422 			PMD_DRV_LOG(ERR, "Failed to allocate %d rx_queues"
423 				" intr_vec", bp->eth_dev->data->nb_rx_queues);
424 			rc = -ENOMEM;
425 			goto err_disable;
426 		}
427 		PMD_DRV_LOG(DEBUG, "intr_handle->intr_vec = %p "
428 			"intr_handle->nb_efd = %d intr_handle->max_intr = %d\n",
429 			 intr_handle->intr_vec, intr_handle->nb_efd,
430 			intr_handle->max_intr);
431 		for (queue_id = 0; queue_id < bp->eth_dev->data->nb_rx_queues;
432 		     queue_id++) {
433 			intr_handle->intr_vec[queue_id] =
434 							vec + BNXT_RX_VEC_START;
435 			if (vec < base + intr_handle->nb_efd - 1)
436 				vec++;
437 		}
438 	}
439 
440 	/* enable uio/vfio intr/eventfd mapping */
441 	rc = rte_intr_enable(intr_handle);
442 	if (rc)
443 		goto err_free;
444 
445 	rc = bnxt_get_hwrm_link_config(bp, &new);
446 	if (rc) {
447 		PMD_DRV_LOG(ERR, "HWRM Get link config failure rc: %x\n", rc);
448 		goto err_free;
449 	}
450 
451 	if (!bp->link_info.link_up) {
452 		rc = bnxt_set_hwrm_link_config(bp, true);
453 		if (rc) {
454 			PMD_DRV_LOG(ERR,
455 				"HWRM link config failure rc: %x\n", rc);
456 			goto err_free;
457 		}
458 	}
459 	bnxt_print_link_info(bp->eth_dev);
460 
461 	return 0;
462 
463 err_free:
464 	rte_free(intr_handle->intr_vec);
465 err_disable:
466 	rte_intr_efd_disable(intr_handle);
467 err_out:
468 	/* Some of the error status returned by FW may not be from errno.h */
469 	if (rc > 0)
470 		rc = -EIO;
471 
472 	return rc;
473 }
474 
475 static int bnxt_shutdown_nic(struct bnxt *bp)
476 {
477 	bnxt_free_all_hwrm_resources(bp);
478 	bnxt_free_all_filters(bp);
479 	bnxt_free_all_vnics(bp);
480 	return 0;
481 }
482 
483 /*
484  * Device configuration and status function
485  */
486 
487 static int bnxt_dev_info_get_op(struct rte_eth_dev *eth_dev,
488 				struct rte_eth_dev_info *dev_info)
489 {
490 	struct rte_pci_device *pdev = RTE_DEV_TO_PCI(eth_dev->device);
491 	struct bnxt *bp = eth_dev->data->dev_private;
492 	uint16_t max_vnics, i, j, vpool, vrxq;
493 	unsigned int max_rx_rings;
494 	int rc;
495 
496 	rc = is_bnxt_in_error(bp);
497 	if (rc)
498 		return rc;
499 
500 	/* MAC Specifics */
501 	dev_info->max_mac_addrs = bp->max_l2_ctx;
502 	dev_info->max_hash_mac_addrs = 0;
503 
504 	/* PF/VF specifics */
505 	if (BNXT_PF(bp))
506 		dev_info->max_vfs = pdev->max_vfs;
507 
508 	max_rx_rings = BNXT_MAX_RINGS(bp);
509 	/* For the sake of symmetry, max_rx_queues = max_tx_queues */
510 	dev_info->max_rx_queues = max_rx_rings;
511 	dev_info->max_tx_queues = max_rx_rings;
512 	dev_info->reta_size = bnxt_rss_hash_tbl_size(bp);
513 	dev_info->hash_key_size = 40;
514 	max_vnics = bp->max_vnics;
515 
516 	/* MTU specifics */
517 	dev_info->min_mtu = RTE_ETHER_MIN_MTU;
518 	dev_info->max_mtu = BNXT_MAX_MTU;
519 
520 	/* Fast path specifics */
521 	dev_info->min_rx_bufsize = 1;
522 	dev_info->max_rx_pktlen = BNXT_MAX_PKT_LEN;
523 
524 	dev_info->rx_offload_capa = BNXT_DEV_RX_OFFLOAD_SUPPORT;
525 	if (bp->flags & BNXT_FLAG_PTP_SUPPORTED)
526 		dev_info->rx_offload_capa |= DEV_RX_OFFLOAD_TIMESTAMP;
527 	dev_info->tx_offload_capa = BNXT_DEV_TX_OFFLOAD_SUPPORT;
528 	dev_info->flow_type_rss_offloads = BNXT_ETH_RSS_SUPPORT;
529 
530 	/* *INDENT-OFF* */
531 	dev_info->default_rxconf = (struct rte_eth_rxconf) {
532 		.rx_thresh = {
533 			.pthresh = 8,
534 			.hthresh = 8,
535 			.wthresh = 0,
536 		},
537 		.rx_free_thresh = 32,
538 		/* If no descriptors available, pkts are dropped by default */
539 		.rx_drop_en = 1,
540 	};
541 
542 	dev_info->default_txconf = (struct rte_eth_txconf) {
543 		.tx_thresh = {
544 			.pthresh = 32,
545 			.hthresh = 0,
546 			.wthresh = 0,
547 		},
548 		.tx_free_thresh = 32,
549 		.tx_rs_thresh = 32,
550 	};
551 	eth_dev->data->dev_conf.intr_conf.lsc = 1;
552 
553 	eth_dev->data->dev_conf.intr_conf.rxq = 1;
554 	dev_info->rx_desc_lim.nb_min = BNXT_MIN_RING_DESC;
555 	dev_info->rx_desc_lim.nb_max = BNXT_MAX_RX_RING_DESC;
556 	dev_info->tx_desc_lim.nb_min = BNXT_MIN_RING_DESC;
557 	dev_info->tx_desc_lim.nb_max = BNXT_MAX_TX_RING_DESC;
558 
559 	/* *INDENT-ON* */
560 
561 	/*
562 	 * TODO: default_rxconf, default_txconf, rx_desc_lim, and tx_desc_lim
563 	 *       need further investigation.
564 	 */
565 
566 	/* VMDq resources */
567 	vpool = 64; /* ETH_64_POOLS */
568 	vrxq = 128; /* ETH_VMDQ_DCB_NUM_QUEUES */
569 	for (i = 0; i < 4; vpool >>= 1, i++) {
570 		if (max_vnics > vpool) {
571 			for (j = 0; j < 5; vrxq >>= 1, j++) {
572 				if (dev_info->max_rx_queues > vrxq) {
573 					if (vpool > vrxq)
574 						vpool = vrxq;
575 					goto found;
576 				}
577 			}
578 			/* Not enough resources to support VMDq */
579 			break;
580 		}
581 	}
582 	/* Not enough resources to support VMDq */
583 	vpool = 0;
584 	vrxq = 0;
585 found:
586 	dev_info->max_vmdq_pools = vpool;
587 	dev_info->vmdq_queue_num = vrxq;
588 
589 	dev_info->vmdq_pool_base = 0;
590 	dev_info->vmdq_queue_base = 0;
591 
592 	return 0;
593 }
594 
595 /* Configure the device based on the configuration provided */
596 static int bnxt_dev_configure_op(struct rte_eth_dev *eth_dev)
597 {
598 	struct bnxt *bp = eth_dev->data->dev_private;
599 	uint64_t rx_offloads = eth_dev->data->dev_conf.rxmode.offloads;
600 	int rc;
601 
602 	bp->rx_queues = (void *)eth_dev->data->rx_queues;
603 	bp->tx_queues = (void *)eth_dev->data->tx_queues;
604 	bp->tx_nr_rings = eth_dev->data->nb_tx_queues;
605 	bp->rx_nr_rings = eth_dev->data->nb_rx_queues;
606 
607 	rc = is_bnxt_in_error(bp);
608 	if (rc)
609 		return rc;
610 
611 	if (BNXT_VF(bp) && (bp->flags & BNXT_FLAG_NEW_RM)) {
612 		rc = bnxt_hwrm_check_vf_rings(bp);
613 		if (rc) {
614 			PMD_DRV_LOG(ERR, "HWRM insufficient resources\n");
615 			return -ENOSPC;
616 		}
617 
618 		/* If a resource has already been allocated - in this case
619 		 * it is the async completion ring, free it. Reallocate it after
620 		 * resource reservation. This will ensure the resource counts
621 		 * are calculated correctly.
622 		 */
623 
624 		pthread_mutex_lock(&bp->def_cp_lock);
625 
626 		if (!BNXT_HAS_NQ(bp) && bp->async_cp_ring) {
627 			bnxt_disable_int(bp);
628 			bnxt_free_cp_ring(bp, bp->async_cp_ring);
629 		}
630 
631 		rc = bnxt_hwrm_func_reserve_vf_resc(bp, false);
632 		if (rc) {
633 			PMD_DRV_LOG(ERR, "HWRM resource alloc fail:%x\n", rc);
634 			pthread_mutex_unlock(&bp->def_cp_lock);
635 			return -ENOSPC;
636 		}
637 
638 		if (!BNXT_HAS_NQ(bp) && bp->async_cp_ring) {
639 			rc = bnxt_alloc_async_cp_ring(bp);
640 			if (rc) {
641 				pthread_mutex_unlock(&bp->def_cp_lock);
642 				return rc;
643 			}
644 			bnxt_enable_int(bp);
645 		}
646 
647 		pthread_mutex_unlock(&bp->def_cp_lock);
648 	} else {
649 		/* legacy driver needs to get updated values */
650 		rc = bnxt_hwrm_func_qcaps(bp);
651 		if (rc) {
652 			PMD_DRV_LOG(ERR, "hwrm func qcaps fail:%d\n", rc);
653 			return rc;
654 		}
655 	}
656 
657 	/* Inherit new configurations */
658 	if (eth_dev->data->nb_rx_queues > bp->max_rx_rings ||
659 	    eth_dev->data->nb_tx_queues > bp->max_tx_rings ||
660 	    eth_dev->data->nb_rx_queues + eth_dev->data->nb_tx_queues
661 		+ BNXT_NUM_ASYNC_CPR(bp) > bp->max_cp_rings ||
662 	    eth_dev->data->nb_rx_queues + eth_dev->data->nb_tx_queues >
663 	    bp->max_stat_ctx)
664 		goto resource_error;
665 
666 	if (BNXT_HAS_RING_GRPS(bp) &&
667 	    (uint32_t)(eth_dev->data->nb_rx_queues) > bp->max_ring_grps)
668 		goto resource_error;
669 
670 	if (!(eth_dev->data->dev_conf.rxmode.mq_mode & ETH_MQ_RX_RSS) &&
671 	    bp->max_vnics < eth_dev->data->nb_rx_queues)
672 		goto resource_error;
673 
674 	bp->rx_cp_nr_rings = bp->rx_nr_rings;
675 	bp->tx_cp_nr_rings = bp->tx_nr_rings;
676 
677 	if (eth_dev->data->dev_conf.rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG)
678 		rx_offloads |= DEV_RX_OFFLOAD_RSS_HASH;
679 	eth_dev->data->dev_conf.rxmode.offloads = rx_offloads;
680 
681 	if (rx_offloads & DEV_RX_OFFLOAD_JUMBO_FRAME) {
682 		eth_dev->data->mtu =
683 			eth_dev->data->dev_conf.rxmode.max_rx_pkt_len -
684 			RTE_ETHER_HDR_LEN - RTE_ETHER_CRC_LEN - VLAN_TAG_SIZE *
685 			BNXT_NUM_VLANS;
686 		bnxt_mtu_set_op(eth_dev, eth_dev->data->mtu);
687 	}
688 	return 0;
689 
690 resource_error:
691 	PMD_DRV_LOG(ERR,
692 		    "Insufficient resources to support requested config\n");
693 	PMD_DRV_LOG(ERR,
694 		    "Num Queues Requested: Tx %d, Rx %d\n",
695 		    eth_dev->data->nb_tx_queues,
696 		    eth_dev->data->nb_rx_queues);
697 	PMD_DRV_LOG(ERR,
698 		    "MAX: TxQ %d, RxQ %d, CQ %d Stat %d, Grp %d, Vnic %d\n",
699 		    bp->max_tx_rings, bp->max_rx_rings, bp->max_cp_rings,
700 		    bp->max_stat_ctx, bp->max_ring_grps, bp->max_vnics);
701 	return -ENOSPC;
702 }
703 
704 static void bnxt_print_link_info(struct rte_eth_dev *eth_dev)
705 {
706 	struct rte_eth_link *link = &eth_dev->data->dev_link;
707 
708 	if (link->link_status)
709 		PMD_DRV_LOG(INFO, "Port %d Link Up - speed %u Mbps - %s\n",
710 			eth_dev->data->port_id,
711 			(uint32_t)link->link_speed,
712 			(link->link_duplex == ETH_LINK_FULL_DUPLEX) ?
713 			("full-duplex") : ("half-duplex\n"));
714 	else
715 		PMD_DRV_LOG(INFO, "Port %d Link Down\n",
716 			eth_dev->data->port_id);
717 }
718 
719 /*
720  * Determine whether the current configuration requires support for scattered
721  * receive; return 1 if scattered receive is required and 0 if not.
722  */
723 static int bnxt_scattered_rx(struct rte_eth_dev *eth_dev)
724 {
725 	uint16_t buf_size;
726 	int i;
727 
728 	if (eth_dev->data->dev_conf.rxmode.offloads & DEV_RX_OFFLOAD_SCATTER)
729 		return 1;
730 
731 	for (i = 0; i < eth_dev->data->nb_rx_queues; i++) {
732 		struct bnxt_rx_queue *rxq = eth_dev->data->rx_queues[i];
733 
734 		buf_size = (uint16_t)(rte_pktmbuf_data_room_size(rxq->mb_pool) -
735 				      RTE_PKTMBUF_HEADROOM);
736 		if (eth_dev->data->dev_conf.rxmode.max_rx_pkt_len > buf_size)
737 			return 1;
738 	}
739 	return 0;
740 }
741 
742 static eth_rx_burst_t
743 bnxt_receive_function(__rte_unused struct rte_eth_dev *eth_dev)
744 {
745 #ifdef RTE_ARCH_X86
746 #ifndef RTE_LIBRTE_IEEE1588
747 	/*
748 	 * Vector mode receive can be enabled only if scatter rx is not
749 	 * in use and rx offloads are limited to VLAN stripping and
750 	 * CRC stripping.
751 	 */
752 	if (!eth_dev->data->scattered_rx &&
753 	    !(eth_dev->data->dev_conf.rxmode.offloads &
754 	      ~(DEV_RX_OFFLOAD_VLAN_STRIP |
755 		DEV_RX_OFFLOAD_KEEP_CRC |
756 		DEV_RX_OFFLOAD_JUMBO_FRAME |
757 		DEV_RX_OFFLOAD_IPV4_CKSUM |
758 		DEV_RX_OFFLOAD_UDP_CKSUM |
759 		DEV_RX_OFFLOAD_TCP_CKSUM |
760 		DEV_RX_OFFLOAD_OUTER_IPV4_CKSUM |
761 		DEV_RX_OFFLOAD_VLAN_FILTER))) {
762 		PMD_DRV_LOG(INFO, "Using vector mode receive for port %d\n",
763 			    eth_dev->data->port_id);
764 		return bnxt_recv_pkts_vec;
765 	}
766 	PMD_DRV_LOG(INFO, "Vector mode receive disabled for port %d\n",
767 		    eth_dev->data->port_id);
768 	PMD_DRV_LOG(INFO,
769 		    "Port %d scatter: %d rx offload: %" PRIX64 "\n",
770 		    eth_dev->data->port_id,
771 		    eth_dev->data->scattered_rx,
772 		    eth_dev->data->dev_conf.rxmode.offloads);
773 #endif
774 #endif
775 	return bnxt_recv_pkts;
776 }
777 
778 static eth_tx_burst_t
779 bnxt_transmit_function(__rte_unused struct rte_eth_dev *eth_dev)
780 {
781 #ifdef RTE_ARCH_X86
782 #ifndef RTE_LIBRTE_IEEE1588
783 	/*
784 	 * Vector mode transmit can be enabled only if not using scatter rx
785 	 * or tx offloads.
786 	 */
787 	if (!eth_dev->data->scattered_rx &&
788 	    !eth_dev->data->dev_conf.txmode.offloads) {
789 		PMD_DRV_LOG(INFO, "Using vector mode transmit for port %d\n",
790 			    eth_dev->data->port_id);
791 		return bnxt_xmit_pkts_vec;
792 	}
793 	PMD_DRV_LOG(INFO, "Vector mode transmit disabled for port %d\n",
794 		    eth_dev->data->port_id);
795 	PMD_DRV_LOG(INFO,
796 		    "Port %d scatter: %d tx offload: %" PRIX64 "\n",
797 		    eth_dev->data->port_id,
798 		    eth_dev->data->scattered_rx,
799 		    eth_dev->data->dev_conf.txmode.offloads);
800 #endif
801 #endif
802 	return bnxt_xmit_pkts;
803 }
804 
805 static int bnxt_handle_if_change_status(struct bnxt *bp)
806 {
807 	int rc;
808 
809 	/* Since fw has undergone a reset and lost all contexts,
810 	 * set fatal flag to not issue hwrm during cleanup
811 	 */
812 	bp->flags |= BNXT_FLAG_FATAL_ERROR;
813 	bnxt_uninit_resources(bp, true);
814 
815 	/* clear fatal flag so that re-init happens */
816 	bp->flags &= ~BNXT_FLAG_FATAL_ERROR;
817 	rc = bnxt_init_resources(bp, true);
818 
819 	bp->flags &= ~BNXT_FLAG_IF_CHANGE_HOT_FW_RESET_DONE;
820 
821 	return rc;
822 }
823 
824 static int bnxt_dev_start_op(struct rte_eth_dev *eth_dev)
825 {
826 	struct bnxt *bp = eth_dev->data->dev_private;
827 	uint64_t rx_offloads = eth_dev->data->dev_conf.rxmode.offloads;
828 	int vlan_mask = 0;
829 	int rc;
830 
831 	if (!eth_dev->data->nb_tx_queues || !eth_dev->data->nb_rx_queues) {
832 		PMD_DRV_LOG(ERR, "Queues are not configured yet!\n");
833 		return -EINVAL;
834 	}
835 
836 	if (bp->rx_cp_nr_rings > RTE_ETHDEV_QUEUE_STAT_CNTRS) {
837 		PMD_DRV_LOG(ERR,
838 			"RxQ cnt %d > CONFIG_RTE_ETHDEV_QUEUE_STAT_CNTRS %d\n",
839 			bp->rx_cp_nr_rings, RTE_ETHDEV_QUEUE_STAT_CNTRS);
840 	}
841 
842 	rc = bnxt_hwrm_if_change(bp, 1);
843 	if (!rc) {
844 		if (bp->flags & BNXT_FLAG_IF_CHANGE_HOT_FW_RESET_DONE) {
845 			rc = bnxt_handle_if_change_status(bp);
846 			if (rc)
847 				return rc;
848 		}
849 	}
850 	bnxt_enable_int(bp);
851 
852 	rc = bnxt_init_chip(bp);
853 	if (rc)
854 		goto error;
855 
856 	eth_dev->data->scattered_rx = bnxt_scattered_rx(eth_dev);
857 
858 	bnxt_link_update_op(eth_dev, 1);
859 
860 	if (rx_offloads & DEV_RX_OFFLOAD_VLAN_FILTER)
861 		vlan_mask |= ETH_VLAN_FILTER_MASK;
862 	if (rx_offloads & DEV_RX_OFFLOAD_VLAN_STRIP)
863 		vlan_mask |= ETH_VLAN_STRIP_MASK;
864 	rc = bnxt_vlan_offload_set_op(eth_dev, vlan_mask);
865 	if (rc)
866 		goto error;
867 
868 	eth_dev->rx_pkt_burst = bnxt_receive_function(eth_dev);
869 	eth_dev->tx_pkt_burst = bnxt_transmit_function(eth_dev);
870 
871 	bp->flags |= BNXT_FLAG_INIT_DONE;
872 	eth_dev->data->dev_started = 1;
873 	bp->dev_stopped = 0;
874 	pthread_mutex_lock(&bp->def_cp_lock);
875 	bnxt_schedule_fw_health_check(bp);
876 	pthread_mutex_unlock(&bp->def_cp_lock);
877 	return 0;
878 
879 error:
880 	bnxt_hwrm_if_change(bp, 0);
881 	bnxt_shutdown_nic(bp);
882 	bnxt_free_tx_mbufs(bp);
883 	bnxt_free_rx_mbufs(bp);
884 	return rc;
885 }
886 
887 static int bnxt_dev_set_link_up_op(struct rte_eth_dev *eth_dev)
888 {
889 	struct bnxt *bp = eth_dev->data->dev_private;
890 	int rc = 0;
891 
892 	if (!bp->link_info.link_up)
893 		rc = bnxt_set_hwrm_link_config(bp, true);
894 	if (!rc)
895 		eth_dev->data->dev_link.link_status = 1;
896 
897 	bnxt_print_link_info(eth_dev);
898 	return rc;
899 }
900 
901 static int bnxt_dev_set_link_down_op(struct rte_eth_dev *eth_dev)
902 {
903 	struct bnxt *bp = eth_dev->data->dev_private;
904 
905 	eth_dev->data->dev_link.link_status = 0;
906 	bnxt_set_hwrm_link_config(bp, false);
907 	bp->link_info.link_up = 0;
908 
909 	return 0;
910 }
911 
912 /* Unload the driver, release resources */
913 static void bnxt_dev_stop_op(struct rte_eth_dev *eth_dev)
914 {
915 	struct bnxt *bp = eth_dev->data->dev_private;
916 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
917 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
918 
919 	eth_dev->data->dev_started = 0;
920 	/* Prevent crashes when queues are still in use */
921 	eth_dev->rx_pkt_burst = &bnxt_dummy_recv_pkts;
922 	eth_dev->tx_pkt_burst = &bnxt_dummy_xmit_pkts;
923 
924 	bnxt_disable_int(bp);
925 
926 	/* disable uio/vfio intr/eventfd mapping */
927 	rte_intr_disable(intr_handle);
928 
929 	bnxt_cancel_fw_health_check(bp);
930 
931 	bp->flags &= ~BNXT_FLAG_INIT_DONE;
932 	if (bp->eth_dev->data->dev_started) {
933 		/* TBD: STOP HW queues DMA */
934 		eth_dev->data->dev_link.link_status = 0;
935 	}
936 	bnxt_dev_set_link_down_op(eth_dev);
937 
938 	/* Wait for link to be reset and the async notification to process.
939 	 * During reset recovery, there is no need to wait
940 	 */
941 	if (!is_bnxt_in_error(bp))
942 		rte_delay_ms(BNXT_LINK_WAIT_INTERVAL * 2);
943 
944 	/* Clean queue intr-vector mapping */
945 	rte_intr_efd_disable(intr_handle);
946 	if (intr_handle->intr_vec != NULL) {
947 		rte_free(intr_handle->intr_vec);
948 		intr_handle->intr_vec = NULL;
949 	}
950 
951 	bnxt_hwrm_port_clr_stats(bp);
952 	bnxt_free_tx_mbufs(bp);
953 	bnxt_free_rx_mbufs(bp);
954 	/* Process any remaining notifications in default completion queue */
955 	bnxt_int_handler(eth_dev);
956 	bnxt_shutdown_nic(bp);
957 	bnxt_hwrm_if_change(bp, 0);
958 	bp->dev_stopped = 1;
959 	bp->rx_cosq_cnt = 0;
960 }
961 
962 static void bnxt_dev_close_op(struct rte_eth_dev *eth_dev)
963 {
964 	struct bnxt *bp = eth_dev->data->dev_private;
965 
966 	if (bp->dev_stopped == 0)
967 		bnxt_dev_stop_op(eth_dev);
968 
969 	if (eth_dev->data->mac_addrs != NULL) {
970 		rte_free(eth_dev->data->mac_addrs);
971 		eth_dev->data->mac_addrs = NULL;
972 	}
973 	if (bp->grp_info != NULL) {
974 		rte_free(bp->grp_info);
975 		bp->grp_info = NULL;
976 	}
977 
978 	bnxt_dev_uninit(eth_dev);
979 }
980 
981 static void bnxt_mac_addr_remove_op(struct rte_eth_dev *eth_dev,
982 				    uint32_t index)
983 {
984 	struct bnxt *bp = eth_dev->data->dev_private;
985 	uint64_t pool_mask = eth_dev->data->mac_pool_sel[index];
986 	struct bnxt_vnic_info *vnic;
987 	struct bnxt_filter_info *filter, *temp_filter;
988 	uint32_t i;
989 
990 	if (is_bnxt_in_error(bp))
991 		return;
992 
993 	/*
994 	 * Loop through all VNICs from the specified filter flow pools to
995 	 * remove the corresponding MAC addr filter
996 	 */
997 	for (i = 0; i < bp->nr_vnics; i++) {
998 		if (!(pool_mask & (1ULL << i)))
999 			continue;
1000 
1001 		vnic = &bp->vnic_info[i];
1002 		filter = STAILQ_FIRST(&vnic->filter);
1003 		while (filter) {
1004 			temp_filter = STAILQ_NEXT(filter, next);
1005 			if (filter->mac_index == index) {
1006 				STAILQ_REMOVE(&vnic->filter, filter,
1007 						bnxt_filter_info, next);
1008 				bnxt_hwrm_clear_l2_filter(bp, filter);
1009 				filter->mac_index = INVALID_MAC_INDEX;
1010 				memset(&filter->l2_addr, 0, RTE_ETHER_ADDR_LEN);
1011 				bnxt_free_filter(bp, filter);
1012 			}
1013 			filter = temp_filter;
1014 		}
1015 	}
1016 }
1017 
1018 static int bnxt_add_mac_filter(struct bnxt *bp, struct bnxt_vnic_info *vnic,
1019 			       struct rte_ether_addr *mac_addr, uint32_t index,
1020 			       uint32_t pool)
1021 {
1022 	struct bnxt_filter_info *filter;
1023 	int rc = 0;
1024 
1025 	/* Attach requested MAC address to the new l2_filter */
1026 	STAILQ_FOREACH(filter, &vnic->filter, next) {
1027 		if (filter->mac_index == index) {
1028 			PMD_DRV_LOG(DEBUG,
1029 				    "MAC addr already existed for pool %d\n",
1030 				    pool);
1031 			return 0;
1032 		}
1033 	}
1034 
1035 	filter = bnxt_alloc_filter(bp);
1036 	if (!filter) {
1037 		PMD_DRV_LOG(ERR, "L2 filter alloc failed\n");
1038 		return -ENODEV;
1039 	}
1040 
1041 	/* bnxt_alloc_filter copies default MAC to filter->l2_addr. So,
1042 	 * if the MAC that's been programmed now is a different one, then,
1043 	 * copy that addr to filter->l2_addr
1044 	 */
1045 	if (mac_addr)
1046 		memcpy(filter->l2_addr, mac_addr, RTE_ETHER_ADDR_LEN);
1047 	filter->flags |= HWRM_CFA_L2_FILTER_ALLOC_INPUT_FLAGS_OUTERMOST;
1048 
1049 	rc = bnxt_hwrm_set_l2_filter(bp, vnic->fw_vnic_id, filter);
1050 	if (!rc) {
1051 		filter->mac_index = index;
1052 		if (filter->mac_index == 0)
1053 			STAILQ_INSERT_HEAD(&vnic->filter, filter, next);
1054 		else
1055 			STAILQ_INSERT_TAIL(&vnic->filter, filter, next);
1056 	} else {
1057 		memset(&filter->l2_addr, 0, RTE_ETHER_ADDR_LEN);
1058 		bnxt_free_filter(bp, filter);
1059 	}
1060 
1061 	return rc;
1062 }
1063 
1064 static int bnxt_mac_addr_add_op(struct rte_eth_dev *eth_dev,
1065 				struct rte_ether_addr *mac_addr,
1066 				uint32_t index, uint32_t pool)
1067 {
1068 	struct bnxt *bp = eth_dev->data->dev_private;
1069 	struct bnxt_vnic_info *vnic = &bp->vnic_info[pool];
1070 	int rc = 0;
1071 
1072 	rc = is_bnxt_in_error(bp);
1073 	if (rc)
1074 		return rc;
1075 
1076 	if (BNXT_VF(bp) & !BNXT_VF_IS_TRUSTED(bp)) {
1077 		PMD_DRV_LOG(ERR, "Cannot add MAC address to a VF interface\n");
1078 		return -ENOTSUP;
1079 	}
1080 
1081 	if (!vnic) {
1082 		PMD_DRV_LOG(ERR, "VNIC not found for pool %d!\n", pool);
1083 		return -EINVAL;
1084 	}
1085 
1086 	rc = bnxt_add_mac_filter(bp, vnic, mac_addr, index, pool);
1087 
1088 	return rc;
1089 }
1090 
1091 int bnxt_link_update_op(struct rte_eth_dev *eth_dev, int wait_to_complete)
1092 {
1093 	int rc = 0;
1094 	struct bnxt *bp = eth_dev->data->dev_private;
1095 	struct rte_eth_link new;
1096 	unsigned int cnt = BNXT_LINK_WAIT_CNT;
1097 
1098 	rc = is_bnxt_in_error(bp);
1099 	if (rc)
1100 		return rc;
1101 
1102 	memset(&new, 0, sizeof(new));
1103 	do {
1104 		/* Retrieve link info from hardware */
1105 		rc = bnxt_get_hwrm_link_config(bp, &new);
1106 		if (rc) {
1107 			new.link_speed = ETH_LINK_SPEED_100M;
1108 			new.link_duplex = ETH_LINK_FULL_DUPLEX;
1109 			PMD_DRV_LOG(ERR,
1110 				"Failed to retrieve link rc = 0x%x!\n", rc);
1111 			goto out;
1112 		}
1113 
1114 		if (!wait_to_complete || new.link_status)
1115 			break;
1116 
1117 		rte_delay_ms(BNXT_LINK_WAIT_INTERVAL);
1118 	} while (cnt--);
1119 
1120 out:
1121 	/* Timed out or success */
1122 	if (new.link_status != eth_dev->data->dev_link.link_status ||
1123 	new.link_speed != eth_dev->data->dev_link.link_speed) {
1124 		rte_eth_linkstatus_set(eth_dev, &new);
1125 
1126 		_rte_eth_dev_callback_process(eth_dev,
1127 					      RTE_ETH_EVENT_INTR_LSC,
1128 					      NULL);
1129 
1130 		bnxt_print_link_info(eth_dev);
1131 	}
1132 
1133 	return rc;
1134 }
1135 
1136 static int bnxt_promiscuous_enable_op(struct rte_eth_dev *eth_dev)
1137 {
1138 	struct bnxt *bp = eth_dev->data->dev_private;
1139 	struct bnxt_vnic_info *vnic;
1140 	uint32_t old_flags;
1141 	int rc;
1142 
1143 	rc = is_bnxt_in_error(bp);
1144 	if (rc)
1145 		return rc;
1146 
1147 	if (bp->vnic_info == NULL)
1148 		return 0;
1149 
1150 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
1151 
1152 	old_flags = vnic->flags;
1153 	vnic->flags |= BNXT_VNIC_INFO_PROMISC;
1154 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1155 	if (rc != 0)
1156 		vnic->flags = old_flags;
1157 
1158 	return rc;
1159 }
1160 
1161 static int bnxt_promiscuous_disable_op(struct rte_eth_dev *eth_dev)
1162 {
1163 	struct bnxt *bp = eth_dev->data->dev_private;
1164 	struct bnxt_vnic_info *vnic;
1165 	uint32_t old_flags;
1166 	int rc;
1167 
1168 	rc = is_bnxt_in_error(bp);
1169 	if (rc)
1170 		return rc;
1171 
1172 	if (bp->vnic_info == NULL)
1173 		return 0;
1174 
1175 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
1176 
1177 	old_flags = vnic->flags;
1178 	vnic->flags &= ~BNXT_VNIC_INFO_PROMISC;
1179 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1180 	if (rc != 0)
1181 		vnic->flags = old_flags;
1182 
1183 	return rc;
1184 }
1185 
1186 static int bnxt_allmulticast_enable_op(struct rte_eth_dev *eth_dev)
1187 {
1188 	struct bnxt *bp = eth_dev->data->dev_private;
1189 	struct bnxt_vnic_info *vnic;
1190 	uint32_t old_flags;
1191 	int rc;
1192 
1193 	rc = is_bnxt_in_error(bp);
1194 	if (rc)
1195 		return rc;
1196 
1197 	if (bp->vnic_info == NULL)
1198 		return 0;
1199 
1200 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
1201 
1202 	old_flags = vnic->flags;
1203 	vnic->flags |= BNXT_VNIC_INFO_ALLMULTI;
1204 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1205 	if (rc != 0)
1206 		vnic->flags = old_flags;
1207 
1208 	return rc;
1209 }
1210 
1211 static int bnxt_allmulticast_disable_op(struct rte_eth_dev *eth_dev)
1212 {
1213 	struct bnxt *bp = eth_dev->data->dev_private;
1214 	struct bnxt_vnic_info *vnic;
1215 	uint32_t old_flags;
1216 	int rc;
1217 
1218 	rc = is_bnxt_in_error(bp);
1219 	if (rc)
1220 		return rc;
1221 
1222 	if (bp->vnic_info == NULL)
1223 		return 0;
1224 
1225 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
1226 
1227 	old_flags = vnic->flags;
1228 	vnic->flags &= ~BNXT_VNIC_INFO_ALLMULTI;
1229 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1230 	if (rc != 0)
1231 		vnic->flags = old_flags;
1232 
1233 	return rc;
1234 }
1235 
1236 /* Return bnxt_rx_queue pointer corresponding to a given rxq. */
1237 static struct bnxt_rx_queue *bnxt_qid_to_rxq(struct bnxt *bp, uint16_t qid)
1238 {
1239 	if (qid >= bp->rx_nr_rings)
1240 		return NULL;
1241 
1242 	return bp->eth_dev->data->rx_queues[qid];
1243 }
1244 
1245 /* Return rxq corresponding to a given rss table ring/group ID. */
1246 static uint16_t bnxt_rss_to_qid(struct bnxt *bp, uint16_t fwr)
1247 {
1248 	struct bnxt_rx_queue *rxq;
1249 	unsigned int i;
1250 
1251 	if (!BNXT_HAS_RING_GRPS(bp)) {
1252 		for (i = 0; i < bp->rx_nr_rings; i++) {
1253 			rxq = bp->eth_dev->data->rx_queues[i];
1254 			if (rxq->rx_ring->rx_ring_struct->fw_ring_id == fwr)
1255 				return rxq->index;
1256 		}
1257 	} else {
1258 		for (i = 0; i < bp->rx_nr_rings; i++) {
1259 			if (bp->grp_info[i].fw_grp_id == fwr)
1260 				return i;
1261 		}
1262 	}
1263 
1264 	return INVALID_HW_RING_ID;
1265 }
1266 
1267 static int bnxt_reta_update_op(struct rte_eth_dev *eth_dev,
1268 			    struct rte_eth_rss_reta_entry64 *reta_conf,
1269 			    uint16_t reta_size)
1270 {
1271 	struct bnxt *bp = eth_dev->data->dev_private;
1272 	struct rte_eth_conf *dev_conf = &bp->eth_dev->data->dev_conf;
1273 	struct bnxt_vnic_info *vnic = &bp->vnic_info[0];
1274 	uint16_t tbl_size = bnxt_rss_hash_tbl_size(bp);
1275 	uint16_t idx, sft;
1276 	int i, rc;
1277 
1278 	rc = is_bnxt_in_error(bp);
1279 	if (rc)
1280 		return rc;
1281 
1282 	if (!vnic->rss_table)
1283 		return -EINVAL;
1284 
1285 	if (!(dev_conf->rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG))
1286 		return -EINVAL;
1287 
1288 	if (reta_size != tbl_size) {
1289 		PMD_DRV_LOG(ERR, "The configured hash table lookup size "
1290 			"(%d) must equal the size supported by the hardware "
1291 			"(%d)\n", reta_size, tbl_size);
1292 		return -EINVAL;
1293 	}
1294 
1295 	for (i = 0; i < reta_size; i++) {
1296 		struct bnxt_rx_queue *rxq;
1297 
1298 		idx = i / RTE_RETA_GROUP_SIZE;
1299 		sft = i % RTE_RETA_GROUP_SIZE;
1300 
1301 		if (!(reta_conf[idx].mask & (1ULL << sft)))
1302 			continue;
1303 
1304 		rxq = bnxt_qid_to_rxq(bp, reta_conf[idx].reta[sft]);
1305 		if (!rxq) {
1306 			PMD_DRV_LOG(ERR, "Invalid ring in reta_conf.\n");
1307 			return -EINVAL;
1308 		}
1309 
1310 		if (BNXT_CHIP_THOR(bp)) {
1311 			vnic->rss_table[i * 2] =
1312 				rxq->rx_ring->rx_ring_struct->fw_ring_id;
1313 			vnic->rss_table[i * 2 + 1] =
1314 				rxq->cp_ring->cp_ring_struct->fw_ring_id;
1315 		} else {
1316 			vnic->rss_table[i] =
1317 			    vnic->fw_grp_ids[reta_conf[idx].reta[sft]];
1318 		}
1319 	}
1320 
1321 	bnxt_hwrm_vnic_rss_cfg(bp, vnic);
1322 	return 0;
1323 }
1324 
1325 static int bnxt_reta_query_op(struct rte_eth_dev *eth_dev,
1326 			      struct rte_eth_rss_reta_entry64 *reta_conf,
1327 			      uint16_t reta_size)
1328 {
1329 	struct bnxt *bp = eth_dev->data->dev_private;
1330 	struct bnxt_vnic_info *vnic = &bp->vnic_info[0];
1331 	uint16_t tbl_size = bnxt_rss_hash_tbl_size(bp);
1332 	uint16_t idx, sft, i;
1333 	int rc;
1334 
1335 	rc = is_bnxt_in_error(bp);
1336 	if (rc)
1337 		return rc;
1338 
1339 	/* Retrieve from the default VNIC */
1340 	if (!vnic)
1341 		return -EINVAL;
1342 	if (!vnic->rss_table)
1343 		return -EINVAL;
1344 
1345 	if (reta_size != tbl_size) {
1346 		PMD_DRV_LOG(ERR, "The configured hash table lookup size "
1347 			"(%d) must equal the size supported by the hardware "
1348 			"(%d)\n", reta_size, tbl_size);
1349 		return -EINVAL;
1350 	}
1351 
1352 	for (idx = 0, i = 0; i < reta_size; i++) {
1353 		idx = i / RTE_RETA_GROUP_SIZE;
1354 		sft = i % RTE_RETA_GROUP_SIZE;
1355 
1356 		if (reta_conf[idx].mask & (1ULL << sft)) {
1357 			uint16_t qid;
1358 
1359 			if (BNXT_CHIP_THOR(bp))
1360 				qid = bnxt_rss_to_qid(bp,
1361 						      vnic->rss_table[i * 2]);
1362 			else
1363 				qid = bnxt_rss_to_qid(bp, vnic->rss_table[i]);
1364 
1365 			if (qid == INVALID_HW_RING_ID) {
1366 				PMD_DRV_LOG(ERR, "Inv. entry in rss table.\n");
1367 				return -EINVAL;
1368 			}
1369 			reta_conf[idx].reta[sft] = qid;
1370 		}
1371 	}
1372 
1373 	return 0;
1374 }
1375 
1376 static int bnxt_rss_hash_update_op(struct rte_eth_dev *eth_dev,
1377 				   struct rte_eth_rss_conf *rss_conf)
1378 {
1379 	struct bnxt *bp = eth_dev->data->dev_private;
1380 	struct rte_eth_conf *dev_conf = &bp->eth_dev->data->dev_conf;
1381 	struct bnxt_vnic_info *vnic;
1382 	int rc;
1383 
1384 	rc = is_bnxt_in_error(bp);
1385 	if (rc)
1386 		return rc;
1387 
1388 	/*
1389 	 * If RSS enablement were different than dev_configure,
1390 	 * then return -EINVAL
1391 	 */
1392 	if (dev_conf->rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG) {
1393 		if (!rss_conf->rss_hf)
1394 			PMD_DRV_LOG(ERR, "Hash type NONE\n");
1395 	} else {
1396 		if (rss_conf->rss_hf & BNXT_ETH_RSS_SUPPORT)
1397 			return -EINVAL;
1398 	}
1399 
1400 	bp->flags |= BNXT_FLAG_UPDATE_HASH;
1401 	memcpy(&bp->rss_conf, rss_conf, sizeof(*rss_conf));
1402 
1403 	/* Update the default RSS VNIC(s) */
1404 	vnic = &bp->vnic_info[0];
1405 	vnic->hash_type = bnxt_rte_to_hwrm_hash_types(rss_conf->rss_hf);
1406 
1407 	/*
1408 	 * If hashkey is not specified, use the previously configured
1409 	 * hashkey
1410 	 */
1411 	if (!rss_conf->rss_key)
1412 		goto rss_config;
1413 
1414 	if (rss_conf->rss_key_len != HW_HASH_KEY_SIZE) {
1415 		PMD_DRV_LOG(ERR,
1416 			    "Invalid hashkey length, should be 16 bytes\n");
1417 		return -EINVAL;
1418 	}
1419 	memcpy(vnic->rss_hash_key, rss_conf->rss_key, rss_conf->rss_key_len);
1420 
1421 rss_config:
1422 	bnxt_hwrm_vnic_rss_cfg(bp, vnic);
1423 	return 0;
1424 }
1425 
1426 static int bnxt_rss_hash_conf_get_op(struct rte_eth_dev *eth_dev,
1427 				     struct rte_eth_rss_conf *rss_conf)
1428 {
1429 	struct bnxt *bp = eth_dev->data->dev_private;
1430 	struct bnxt_vnic_info *vnic = &bp->vnic_info[0];
1431 	int len, rc;
1432 	uint32_t hash_types;
1433 
1434 	rc = is_bnxt_in_error(bp);
1435 	if (rc)
1436 		return rc;
1437 
1438 	/* RSS configuration is the same for all VNICs */
1439 	if (vnic && vnic->rss_hash_key) {
1440 		if (rss_conf->rss_key) {
1441 			len = rss_conf->rss_key_len <= HW_HASH_KEY_SIZE ?
1442 			      rss_conf->rss_key_len : HW_HASH_KEY_SIZE;
1443 			memcpy(rss_conf->rss_key, vnic->rss_hash_key, len);
1444 		}
1445 
1446 		hash_types = vnic->hash_type;
1447 		rss_conf->rss_hf = 0;
1448 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV4) {
1449 			rss_conf->rss_hf |= ETH_RSS_IPV4;
1450 			hash_types &= ~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV4;
1451 		}
1452 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV4) {
1453 			rss_conf->rss_hf |= ETH_RSS_NONFRAG_IPV4_TCP;
1454 			hash_types &=
1455 				~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV4;
1456 		}
1457 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV4) {
1458 			rss_conf->rss_hf |= ETH_RSS_NONFRAG_IPV4_UDP;
1459 			hash_types &=
1460 				~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV4;
1461 		}
1462 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV6) {
1463 			rss_conf->rss_hf |= ETH_RSS_IPV6;
1464 			hash_types &= ~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV6;
1465 		}
1466 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV6) {
1467 			rss_conf->rss_hf |= ETH_RSS_NONFRAG_IPV6_TCP;
1468 			hash_types &=
1469 				~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV6;
1470 		}
1471 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV6) {
1472 			rss_conf->rss_hf |= ETH_RSS_NONFRAG_IPV6_UDP;
1473 			hash_types &=
1474 				~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV6;
1475 		}
1476 		if (hash_types) {
1477 			PMD_DRV_LOG(ERR,
1478 				"Unknwon RSS config from firmware (%08x), RSS disabled",
1479 				vnic->hash_type);
1480 			return -ENOTSUP;
1481 		}
1482 	} else {
1483 		rss_conf->rss_hf = 0;
1484 	}
1485 	return 0;
1486 }
1487 
1488 static int bnxt_flow_ctrl_get_op(struct rte_eth_dev *dev,
1489 			       struct rte_eth_fc_conf *fc_conf)
1490 {
1491 	struct bnxt *bp = dev->data->dev_private;
1492 	struct rte_eth_link link_info;
1493 	int rc;
1494 
1495 	rc = is_bnxt_in_error(bp);
1496 	if (rc)
1497 		return rc;
1498 
1499 	rc = bnxt_get_hwrm_link_config(bp, &link_info);
1500 	if (rc)
1501 		return rc;
1502 
1503 	memset(fc_conf, 0, sizeof(*fc_conf));
1504 	if (bp->link_info.auto_pause)
1505 		fc_conf->autoneg = 1;
1506 	switch (bp->link_info.pause) {
1507 	case 0:
1508 		fc_conf->mode = RTE_FC_NONE;
1509 		break;
1510 	case HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_TX:
1511 		fc_conf->mode = RTE_FC_TX_PAUSE;
1512 		break;
1513 	case HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_RX:
1514 		fc_conf->mode = RTE_FC_RX_PAUSE;
1515 		break;
1516 	case (HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_TX |
1517 			HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_RX):
1518 		fc_conf->mode = RTE_FC_FULL;
1519 		break;
1520 	}
1521 	return 0;
1522 }
1523 
1524 static int bnxt_flow_ctrl_set_op(struct rte_eth_dev *dev,
1525 			       struct rte_eth_fc_conf *fc_conf)
1526 {
1527 	struct bnxt *bp = dev->data->dev_private;
1528 	int rc;
1529 
1530 	rc = is_bnxt_in_error(bp);
1531 	if (rc)
1532 		return rc;
1533 
1534 	if (!BNXT_SINGLE_PF(bp) || BNXT_VF(bp)) {
1535 		PMD_DRV_LOG(ERR, "Flow Control Settings cannot be modified\n");
1536 		return -ENOTSUP;
1537 	}
1538 
1539 	switch (fc_conf->mode) {
1540 	case RTE_FC_NONE:
1541 		bp->link_info.auto_pause = 0;
1542 		bp->link_info.force_pause = 0;
1543 		break;
1544 	case RTE_FC_RX_PAUSE:
1545 		if (fc_conf->autoneg) {
1546 			bp->link_info.auto_pause =
1547 					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_RX;
1548 			bp->link_info.force_pause = 0;
1549 		} else {
1550 			bp->link_info.auto_pause = 0;
1551 			bp->link_info.force_pause =
1552 					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_RX;
1553 		}
1554 		break;
1555 	case RTE_FC_TX_PAUSE:
1556 		if (fc_conf->autoneg) {
1557 			bp->link_info.auto_pause =
1558 					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_TX;
1559 			bp->link_info.force_pause = 0;
1560 		} else {
1561 			bp->link_info.auto_pause = 0;
1562 			bp->link_info.force_pause =
1563 					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_TX;
1564 		}
1565 		break;
1566 	case RTE_FC_FULL:
1567 		if (fc_conf->autoneg) {
1568 			bp->link_info.auto_pause =
1569 					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_TX |
1570 					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_RX;
1571 			bp->link_info.force_pause = 0;
1572 		} else {
1573 			bp->link_info.auto_pause = 0;
1574 			bp->link_info.force_pause =
1575 					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_TX |
1576 					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_RX;
1577 		}
1578 		break;
1579 	}
1580 	return bnxt_set_hwrm_link_config(bp, true);
1581 }
1582 
1583 /* Add UDP tunneling port */
1584 static int
1585 bnxt_udp_tunnel_port_add_op(struct rte_eth_dev *eth_dev,
1586 			 struct rte_eth_udp_tunnel *udp_tunnel)
1587 {
1588 	struct bnxt *bp = eth_dev->data->dev_private;
1589 	uint16_t tunnel_type = 0;
1590 	int rc = 0;
1591 
1592 	rc = is_bnxt_in_error(bp);
1593 	if (rc)
1594 		return rc;
1595 
1596 	switch (udp_tunnel->prot_type) {
1597 	case RTE_TUNNEL_TYPE_VXLAN:
1598 		if (bp->vxlan_port_cnt) {
1599 			PMD_DRV_LOG(ERR, "Tunnel Port %d already programmed\n",
1600 				udp_tunnel->udp_port);
1601 			if (bp->vxlan_port != udp_tunnel->udp_port) {
1602 				PMD_DRV_LOG(ERR, "Only one port allowed\n");
1603 				return -ENOSPC;
1604 			}
1605 			bp->vxlan_port_cnt++;
1606 			return 0;
1607 		}
1608 		tunnel_type =
1609 			HWRM_TUNNEL_DST_PORT_ALLOC_INPUT_TUNNEL_TYPE_VXLAN;
1610 		bp->vxlan_port_cnt++;
1611 		break;
1612 	case RTE_TUNNEL_TYPE_GENEVE:
1613 		if (bp->geneve_port_cnt) {
1614 			PMD_DRV_LOG(ERR, "Tunnel Port %d already programmed\n",
1615 				udp_tunnel->udp_port);
1616 			if (bp->geneve_port != udp_tunnel->udp_port) {
1617 				PMD_DRV_LOG(ERR, "Only one port allowed\n");
1618 				return -ENOSPC;
1619 			}
1620 			bp->geneve_port_cnt++;
1621 			return 0;
1622 		}
1623 		tunnel_type =
1624 			HWRM_TUNNEL_DST_PORT_ALLOC_INPUT_TUNNEL_TYPE_GENEVE;
1625 		bp->geneve_port_cnt++;
1626 		break;
1627 	default:
1628 		PMD_DRV_LOG(ERR, "Tunnel type is not supported\n");
1629 		return -ENOTSUP;
1630 	}
1631 	rc = bnxt_hwrm_tunnel_dst_port_alloc(bp, udp_tunnel->udp_port,
1632 					     tunnel_type);
1633 	return rc;
1634 }
1635 
1636 static int
1637 bnxt_udp_tunnel_port_del_op(struct rte_eth_dev *eth_dev,
1638 			 struct rte_eth_udp_tunnel *udp_tunnel)
1639 {
1640 	struct bnxt *bp = eth_dev->data->dev_private;
1641 	uint16_t tunnel_type = 0;
1642 	uint16_t port = 0;
1643 	int rc = 0;
1644 
1645 	rc = is_bnxt_in_error(bp);
1646 	if (rc)
1647 		return rc;
1648 
1649 	switch (udp_tunnel->prot_type) {
1650 	case RTE_TUNNEL_TYPE_VXLAN:
1651 		if (!bp->vxlan_port_cnt) {
1652 			PMD_DRV_LOG(ERR, "No Tunnel port configured yet\n");
1653 			return -EINVAL;
1654 		}
1655 		if (bp->vxlan_port != udp_tunnel->udp_port) {
1656 			PMD_DRV_LOG(ERR, "Req Port: %d. Configured port: %d\n",
1657 				udp_tunnel->udp_port, bp->vxlan_port);
1658 			return -EINVAL;
1659 		}
1660 		if (--bp->vxlan_port_cnt)
1661 			return 0;
1662 
1663 		tunnel_type =
1664 			HWRM_TUNNEL_DST_PORT_FREE_INPUT_TUNNEL_TYPE_VXLAN;
1665 		port = bp->vxlan_fw_dst_port_id;
1666 		break;
1667 	case RTE_TUNNEL_TYPE_GENEVE:
1668 		if (!bp->geneve_port_cnt) {
1669 			PMD_DRV_LOG(ERR, "No Tunnel port configured yet\n");
1670 			return -EINVAL;
1671 		}
1672 		if (bp->geneve_port != udp_tunnel->udp_port) {
1673 			PMD_DRV_LOG(ERR, "Req Port: %d. Configured port: %d\n",
1674 				udp_tunnel->udp_port, bp->geneve_port);
1675 			return -EINVAL;
1676 		}
1677 		if (--bp->geneve_port_cnt)
1678 			return 0;
1679 
1680 		tunnel_type =
1681 			HWRM_TUNNEL_DST_PORT_FREE_INPUT_TUNNEL_TYPE_GENEVE;
1682 		port = bp->geneve_fw_dst_port_id;
1683 		break;
1684 	default:
1685 		PMD_DRV_LOG(ERR, "Tunnel type is not supported\n");
1686 		return -ENOTSUP;
1687 	}
1688 
1689 	rc = bnxt_hwrm_tunnel_dst_port_free(bp, port, tunnel_type);
1690 	if (!rc) {
1691 		if (tunnel_type ==
1692 		    HWRM_TUNNEL_DST_PORT_FREE_INPUT_TUNNEL_TYPE_VXLAN)
1693 			bp->vxlan_port = 0;
1694 		if (tunnel_type ==
1695 		    HWRM_TUNNEL_DST_PORT_FREE_INPUT_TUNNEL_TYPE_GENEVE)
1696 			bp->geneve_port = 0;
1697 	}
1698 	return rc;
1699 }
1700 
1701 static int bnxt_del_vlan_filter(struct bnxt *bp, uint16_t vlan_id)
1702 {
1703 	struct bnxt_filter_info *filter;
1704 	struct bnxt_vnic_info *vnic;
1705 	int rc = 0;
1706 	uint32_t chk = HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN;
1707 
1708 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
1709 	filter = STAILQ_FIRST(&vnic->filter);
1710 	while (filter) {
1711 		/* Search for this matching MAC+VLAN filter */
1712 		if (bnxt_vlan_filter_exists(bp, filter, chk, vlan_id)) {
1713 			/* Delete the filter */
1714 			rc = bnxt_hwrm_clear_l2_filter(bp, filter);
1715 			if (rc)
1716 				return rc;
1717 			STAILQ_REMOVE(&vnic->filter, filter,
1718 				      bnxt_filter_info, next);
1719 			bnxt_free_filter(bp, filter);
1720 			PMD_DRV_LOG(INFO,
1721 				    "Deleted vlan filter for %d\n",
1722 				    vlan_id);
1723 			return 0;
1724 		}
1725 		filter = STAILQ_NEXT(filter, next);
1726 	}
1727 	return -ENOENT;
1728 }
1729 
1730 static int bnxt_add_vlan_filter(struct bnxt *bp, uint16_t vlan_id)
1731 {
1732 	struct bnxt_filter_info *filter;
1733 	struct bnxt_vnic_info *vnic;
1734 	int rc = 0;
1735 	uint32_t en = HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN |
1736 		HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN_MASK;
1737 	uint32_t chk = HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN;
1738 
1739 	/* Implementation notes on the use of VNIC in this command:
1740 	 *
1741 	 * By default, these filters belong to default vnic for the function.
1742 	 * Once these filters are set up, only destination VNIC can be modified.
1743 	 * If the destination VNIC is not specified in this command,
1744 	 * then the HWRM shall only create an l2 context id.
1745 	 */
1746 
1747 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
1748 	filter = STAILQ_FIRST(&vnic->filter);
1749 	/* Check if the VLAN has already been added */
1750 	while (filter) {
1751 		if (bnxt_vlan_filter_exists(bp, filter, chk, vlan_id))
1752 			return -EEXIST;
1753 
1754 		filter = STAILQ_NEXT(filter, next);
1755 	}
1756 
1757 	/* No match found. Alloc a fresh filter and issue the L2_FILTER_ALLOC
1758 	 * command to create MAC+VLAN filter with the right flags, enables set.
1759 	 */
1760 	filter = bnxt_alloc_filter(bp);
1761 	if (!filter) {
1762 		PMD_DRV_LOG(ERR,
1763 			    "MAC/VLAN filter alloc failed\n");
1764 		return -ENOMEM;
1765 	}
1766 	/* MAC + VLAN ID filter */
1767 	/* If l2_ivlan == 0 and l2_ivlan_mask != 0, only
1768 	 * untagged packets are received
1769 	 *
1770 	 * If l2_ivlan != 0 and l2_ivlan_mask != 0, untagged
1771 	 * packets and only the programmed vlan's packets are received
1772 	 */
1773 	filter->l2_ivlan = vlan_id;
1774 	filter->l2_ivlan_mask = 0x0FFF;
1775 	filter->enables |= en;
1776 	filter->flags |= HWRM_CFA_L2_FILTER_ALLOC_INPUT_FLAGS_OUTERMOST;
1777 
1778 	rc = bnxt_hwrm_set_l2_filter(bp, vnic->fw_vnic_id, filter);
1779 	if (rc) {
1780 		/* Free the newly allocated filter as we were
1781 		 * not able to create the filter in hardware.
1782 		 */
1783 		filter->fw_l2_filter_id = UINT64_MAX;
1784 		bnxt_free_filter(bp, filter);
1785 		return rc;
1786 	}
1787 
1788 	filter->mac_index = 0;
1789 	/* Add this new filter to the list */
1790 	if (vlan_id == 0)
1791 		STAILQ_INSERT_HEAD(&vnic->filter, filter, next);
1792 	else
1793 		STAILQ_INSERT_TAIL(&vnic->filter, filter, next);
1794 
1795 	PMD_DRV_LOG(INFO,
1796 		    "Added Vlan filter for %d\n", vlan_id);
1797 	return rc;
1798 }
1799 
1800 static int bnxt_vlan_filter_set_op(struct rte_eth_dev *eth_dev,
1801 		uint16_t vlan_id, int on)
1802 {
1803 	struct bnxt *bp = eth_dev->data->dev_private;
1804 	int rc;
1805 
1806 	rc = is_bnxt_in_error(bp);
1807 	if (rc)
1808 		return rc;
1809 
1810 	/* These operations apply to ALL existing MAC/VLAN filters */
1811 	if (on)
1812 		return bnxt_add_vlan_filter(bp, vlan_id);
1813 	else
1814 		return bnxt_del_vlan_filter(bp, vlan_id);
1815 }
1816 
1817 static int bnxt_del_dflt_mac_filter(struct bnxt *bp,
1818 				    struct bnxt_vnic_info *vnic)
1819 {
1820 	struct bnxt_filter_info *filter;
1821 	int rc;
1822 
1823 	filter = STAILQ_FIRST(&vnic->filter);
1824 	while (filter) {
1825 		if (filter->mac_index == 0 &&
1826 		    !memcmp(filter->l2_addr, bp->mac_addr,
1827 			    RTE_ETHER_ADDR_LEN)) {
1828 			rc = bnxt_hwrm_clear_l2_filter(bp, filter);
1829 			if (!rc) {
1830 				STAILQ_REMOVE(&vnic->filter, filter,
1831 					      bnxt_filter_info, next);
1832 				bnxt_free_filter(bp, filter);
1833 				filter->fw_l2_filter_id = UINT64_MAX;
1834 			}
1835 			return rc;
1836 		}
1837 		filter = STAILQ_NEXT(filter, next);
1838 	}
1839 	return 0;
1840 }
1841 
1842 static int
1843 bnxt_vlan_offload_set_op(struct rte_eth_dev *dev, int mask)
1844 {
1845 	struct bnxt *bp = dev->data->dev_private;
1846 	uint64_t rx_offloads = dev->data->dev_conf.rxmode.offloads;
1847 	struct bnxt_vnic_info *vnic;
1848 	unsigned int i;
1849 	int rc;
1850 
1851 	rc = is_bnxt_in_error(bp);
1852 	if (rc)
1853 		return rc;
1854 
1855 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
1856 	if (!(rx_offloads & DEV_RX_OFFLOAD_VLAN_FILTER)) {
1857 		/* Remove any VLAN filters programmed */
1858 		for (i = 0; i < RTE_ETHER_MAX_VLAN_ID; i++)
1859 			bnxt_del_vlan_filter(bp, i);
1860 
1861 		rc = bnxt_add_mac_filter(bp, vnic, NULL, 0, 0);
1862 		if (rc)
1863 			return rc;
1864 	} else {
1865 		/* Default filter will allow packets that match the
1866 		 * dest mac. So, it has to be deleted, otherwise, we
1867 		 * will endup receiving vlan packets for which the
1868 		 * filter is not programmed, when hw-vlan-filter
1869 		 * configuration is ON
1870 		 */
1871 		bnxt_del_dflt_mac_filter(bp, vnic);
1872 		/* This filter will allow only untagged packets */
1873 		bnxt_add_vlan_filter(bp, 0);
1874 	}
1875 	PMD_DRV_LOG(DEBUG, "VLAN Filtering: %d\n",
1876 		    !!(rx_offloads & DEV_RX_OFFLOAD_VLAN_FILTER));
1877 
1878 	if (mask & ETH_VLAN_STRIP_MASK) {
1879 		/* Enable or disable VLAN stripping */
1880 		for (i = 0; i < bp->nr_vnics; i++) {
1881 			struct bnxt_vnic_info *vnic = &bp->vnic_info[i];
1882 			if (rx_offloads & DEV_RX_OFFLOAD_VLAN_STRIP)
1883 				vnic->vlan_strip = true;
1884 			else
1885 				vnic->vlan_strip = false;
1886 			bnxt_hwrm_vnic_cfg(bp, vnic);
1887 		}
1888 		PMD_DRV_LOG(DEBUG, "VLAN Strip Offload: %d\n",
1889 			!!(rx_offloads & DEV_RX_OFFLOAD_VLAN_STRIP));
1890 	}
1891 
1892 	if (mask & ETH_VLAN_EXTEND_MASK) {
1893 		if (rx_offloads & DEV_RX_OFFLOAD_VLAN_EXTEND)
1894 			PMD_DRV_LOG(DEBUG, "Extend VLAN supported\n");
1895 		else
1896 			PMD_DRV_LOG(INFO, "Extend VLAN unsupported\n");
1897 	}
1898 
1899 	return 0;
1900 }
1901 
1902 static int
1903 bnxt_vlan_tpid_set_op(struct rte_eth_dev *dev, enum rte_vlan_type vlan_type,
1904 		      uint16_t tpid)
1905 {
1906 	struct bnxt *bp = dev->data->dev_private;
1907 	int qinq = dev->data->dev_conf.rxmode.offloads &
1908 		   DEV_RX_OFFLOAD_VLAN_EXTEND;
1909 
1910 	if (vlan_type != ETH_VLAN_TYPE_INNER &&
1911 	    vlan_type != ETH_VLAN_TYPE_OUTER) {
1912 		PMD_DRV_LOG(ERR,
1913 			    "Unsupported vlan type.");
1914 		return -EINVAL;
1915 	}
1916 	if (!qinq) {
1917 		PMD_DRV_LOG(ERR,
1918 			    "QinQ not enabled. Needs to be ON as we can "
1919 			    "accelerate only outer vlan\n");
1920 		return -EINVAL;
1921 	}
1922 
1923 	if (vlan_type == ETH_VLAN_TYPE_OUTER) {
1924 		switch (tpid) {
1925 		case RTE_ETHER_TYPE_QINQ:
1926 			bp->outer_tpid_bd =
1927 				TX_BD_LONG_CFA_META_VLAN_TPID_TPID88A8;
1928 				break;
1929 		case RTE_ETHER_TYPE_VLAN:
1930 			bp->outer_tpid_bd =
1931 				TX_BD_LONG_CFA_META_VLAN_TPID_TPID8100;
1932 				break;
1933 		case 0x9100:
1934 			bp->outer_tpid_bd =
1935 				TX_BD_LONG_CFA_META_VLAN_TPID_TPID9100;
1936 				break;
1937 		case 0x9200:
1938 			bp->outer_tpid_bd =
1939 				TX_BD_LONG_CFA_META_VLAN_TPID_TPID9200;
1940 				break;
1941 		case 0x9300:
1942 			bp->outer_tpid_bd =
1943 				 TX_BD_LONG_CFA_META_VLAN_TPID_TPID9300;
1944 				break;
1945 		default:
1946 			PMD_DRV_LOG(ERR, "Invalid TPID: %x\n", tpid);
1947 			return -EINVAL;
1948 		}
1949 		bp->outer_tpid_bd |= tpid;
1950 		PMD_DRV_LOG(INFO, "outer_tpid_bd = %x\n", bp->outer_tpid_bd);
1951 	} else if (vlan_type == ETH_VLAN_TYPE_INNER) {
1952 		PMD_DRV_LOG(ERR,
1953 			    "Can accelerate only outer vlan in QinQ\n");
1954 		return -EINVAL;
1955 	}
1956 
1957 	return 0;
1958 }
1959 
1960 static int
1961 bnxt_set_default_mac_addr_op(struct rte_eth_dev *dev,
1962 			     struct rte_ether_addr *addr)
1963 {
1964 	struct bnxt *bp = dev->data->dev_private;
1965 	/* Default Filter is tied to VNIC 0 */
1966 	struct bnxt_vnic_info *vnic = &bp->vnic_info[0];
1967 	struct bnxt_filter_info *filter;
1968 	int rc;
1969 
1970 	rc = is_bnxt_in_error(bp);
1971 	if (rc)
1972 		return rc;
1973 
1974 	if (BNXT_VF(bp) && !BNXT_VF_IS_TRUSTED(bp))
1975 		return -EPERM;
1976 
1977 	if (rte_is_zero_ether_addr(addr))
1978 		return -EINVAL;
1979 
1980 	STAILQ_FOREACH(filter, &vnic->filter, next) {
1981 		/* Default Filter is at Index 0 */
1982 		if (filter->mac_index != 0)
1983 			continue;
1984 
1985 		memcpy(filter->l2_addr, addr, RTE_ETHER_ADDR_LEN);
1986 		memset(filter->l2_addr_mask, 0xff, RTE_ETHER_ADDR_LEN);
1987 		filter->flags |= HWRM_CFA_L2_FILTER_ALLOC_INPUT_FLAGS_PATH_RX |
1988 			HWRM_CFA_L2_FILTER_ALLOC_INPUT_FLAGS_OUTERMOST;
1989 		filter->enables |=
1990 			HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_ADDR |
1991 			HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_ADDR_MASK;
1992 
1993 		rc = bnxt_hwrm_set_l2_filter(bp, vnic->fw_vnic_id, filter);
1994 		if (rc) {
1995 			memcpy(filter->l2_addr, bp->mac_addr,
1996 			       RTE_ETHER_ADDR_LEN);
1997 			return rc;
1998 		}
1999 
2000 		memcpy(bp->mac_addr, addr, RTE_ETHER_ADDR_LEN);
2001 		PMD_DRV_LOG(DEBUG, "Set MAC addr\n");
2002 		return 0;
2003 	}
2004 
2005 	return 0;
2006 }
2007 
2008 static int
2009 bnxt_dev_set_mc_addr_list_op(struct rte_eth_dev *eth_dev,
2010 			  struct rte_ether_addr *mc_addr_set,
2011 			  uint32_t nb_mc_addr)
2012 {
2013 	struct bnxt *bp = eth_dev->data->dev_private;
2014 	char *mc_addr_list = (char *)mc_addr_set;
2015 	struct bnxt_vnic_info *vnic;
2016 	uint32_t off = 0, i = 0;
2017 	int rc;
2018 
2019 	rc = is_bnxt_in_error(bp);
2020 	if (rc)
2021 		return rc;
2022 
2023 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
2024 
2025 	if (nb_mc_addr > BNXT_MAX_MC_ADDRS) {
2026 		vnic->flags |= BNXT_VNIC_INFO_ALLMULTI;
2027 		goto allmulti;
2028 	}
2029 
2030 	/* TODO Check for Duplicate mcast addresses */
2031 	vnic->flags &= ~BNXT_VNIC_INFO_ALLMULTI;
2032 	for (i = 0; i < nb_mc_addr; i++) {
2033 		memcpy(vnic->mc_list + off, &mc_addr_list[i],
2034 			RTE_ETHER_ADDR_LEN);
2035 		off += RTE_ETHER_ADDR_LEN;
2036 	}
2037 
2038 	vnic->mc_addr_cnt = i;
2039 	if (vnic->mc_addr_cnt)
2040 		vnic->flags |= BNXT_VNIC_INFO_MCAST;
2041 	else
2042 		vnic->flags &= ~BNXT_VNIC_INFO_MCAST;
2043 
2044 allmulti:
2045 	return bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
2046 }
2047 
2048 static int
2049 bnxt_fw_version_get(struct rte_eth_dev *dev, char *fw_version, size_t fw_size)
2050 {
2051 	struct bnxt *bp = dev->data->dev_private;
2052 	uint8_t fw_major = (bp->fw_ver >> 24) & 0xff;
2053 	uint8_t fw_minor = (bp->fw_ver >> 16) & 0xff;
2054 	uint8_t fw_updt = (bp->fw_ver >> 8) & 0xff;
2055 	int ret;
2056 
2057 	ret = snprintf(fw_version, fw_size, "%d.%d.%d",
2058 			fw_major, fw_minor, fw_updt);
2059 
2060 	ret += 1; /* add the size of '\0' */
2061 	if (fw_size < (uint32_t)ret)
2062 		return ret;
2063 	else
2064 		return 0;
2065 }
2066 
2067 static void
2068 bnxt_rxq_info_get_op(struct rte_eth_dev *dev, uint16_t queue_id,
2069 	struct rte_eth_rxq_info *qinfo)
2070 {
2071 	struct bnxt *bp = dev->data->dev_private;
2072 	struct bnxt_rx_queue *rxq;
2073 
2074 	if (is_bnxt_in_error(bp))
2075 		return;
2076 
2077 	rxq = dev->data->rx_queues[queue_id];
2078 
2079 	qinfo->mp = rxq->mb_pool;
2080 	qinfo->scattered_rx = dev->data->scattered_rx;
2081 	qinfo->nb_desc = rxq->nb_rx_desc;
2082 
2083 	qinfo->conf.rx_free_thresh = rxq->rx_free_thresh;
2084 	qinfo->conf.rx_drop_en = 0;
2085 	qinfo->conf.rx_deferred_start = rxq->rx_deferred_start;
2086 }
2087 
2088 static void
2089 bnxt_txq_info_get_op(struct rte_eth_dev *dev, uint16_t queue_id,
2090 	struct rte_eth_txq_info *qinfo)
2091 {
2092 	struct bnxt *bp = dev->data->dev_private;
2093 	struct bnxt_tx_queue *txq;
2094 
2095 	if (is_bnxt_in_error(bp))
2096 		return;
2097 
2098 	txq = dev->data->tx_queues[queue_id];
2099 
2100 	qinfo->nb_desc = txq->nb_tx_desc;
2101 
2102 	qinfo->conf.tx_thresh.pthresh = txq->pthresh;
2103 	qinfo->conf.tx_thresh.hthresh = txq->hthresh;
2104 	qinfo->conf.tx_thresh.wthresh = txq->wthresh;
2105 
2106 	qinfo->conf.tx_free_thresh = txq->tx_free_thresh;
2107 	qinfo->conf.tx_rs_thresh = 0;
2108 	qinfo->conf.tx_deferred_start = txq->tx_deferred_start;
2109 }
2110 
2111 int bnxt_mtu_set_op(struct rte_eth_dev *eth_dev, uint16_t new_mtu)
2112 {
2113 	struct bnxt *bp = eth_dev->data->dev_private;
2114 	uint32_t new_pkt_size;
2115 	uint32_t rc = 0;
2116 	uint32_t i;
2117 
2118 	rc = is_bnxt_in_error(bp);
2119 	if (rc)
2120 		return rc;
2121 
2122 	/* Exit if receive queues are not configured yet */
2123 	if (!eth_dev->data->nb_rx_queues)
2124 		return rc;
2125 
2126 	new_pkt_size = new_mtu + RTE_ETHER_HDR_LEN + RTE_ETHER_CRC_LEN +
2127 		       VLAN_TAG_SIZE * BNXT_NUM_VLANS;
2128 
2129 #ifdef RTE_ARCH_X86
2130 	/*
2131 	 * If vector-mode tx/rx is active, disallow any MTU change that would
2132 	 * require scattered receive support.
2133 	 */
2134 	if (eth_dev->data->dev_started &&
2135 	    (eth_dev->rx_pkt_burst == bnxt_recv_pkts_vec ||
2136 	     eth_dev->tx_pkt_burst == bnxt_xmit_pkts_vec) &&
2137 	    (new_pkt_size >
2138 	     eth_dev->data->min_rx_buf_size - RTE_PKTMBUF_HEADROOM)) {
2139 		PMD_DRV_LOG(ERR,
2140 			    "MTU change would require scattered rx support. ");
2141 		PMD_DRV_LOG(ERR, "Stop port before changing MTU.\n");
2142 		return -EINVAL;
2143 	}
2144 #endif
2145 
2146 	if (new_mtu > RTE_ETHER_MTU) {
2147 		bp->flags |= BNXT_FLAG_JUMBO;
2148 		bp->eth_dev->data->dev_conf.rxmode.offloads |=
2149 			DEV_RX_OFFLOAD_JUMBO_FRAME;
2150 	} else {
2151 		bp->eth_dev->data->dev_conf.rxmode.offloads &=
2152 			~DEV_RX_OFFLOAD_JUMBO_FRAME;
2153 		bp->flags &= ~BNXT_FLAG_JUMBO;
2154 	}
2155 
2156 	/* Is there a change in mtu setting? */
2157 	if (eth_dev->data->dev_conf.rxmode.max_rx_pkt_len == new_pkt_size)
2158 		return rc;
2159 
2160 	for (i = 0; i < bp->nr_vnics; i++) {
2161 		struct bnxt_vnic_info *vnic = &bp->vnic_info[i];
2162 		uint16_t size = 0;
2163 
2164 		vnic->mru = BNXT_VNIC_MRU(new_mtu);
2165 		rc = bnxt_hwrm_vnic_cfg(bp, vnic);
2166 		if (rc)
2167 			break;
2168 
2169 		size = rte_pktmbuf_data_room_size(bp->rx_queues[0]->mb_pool);
2170 		size -= RTE_PKTMBUF_HEADROOM;
2171 
2172 		if (size < new_mtu) {
2173 			rc = bnxt_hwrm_vnic_plcmode_cfg(bp, vnic);
2174 			if (rc)
2175 				return rc;
2176 		}
2177 	}
2178 
2179 	if (!rc)
2180 		eth_dev->data->dev_conf.rxmode.max_rx_pkt_len = new_pkt_size;
2181 
2182 	PMD_DRV_LOG(INFO, "New MTU is %d\n", new_mtu);
2183 
2184 	return rc;
2185 }
2186 
2187 static int
2188 bnxt_vlan_pvid_set_op(struct rte_eth_dev *dev, uint16_t pvid, int on)
2189 {
2190 	struct bnxt *bp = dev->data->dev_private;
2191 	uint16_t vlan = bp->vlan;
2192 	int rc;
2193 
2194 	rc = is_bnxt_in_error(bp);
2195 	if (rc)
2196 		return rc;
2197 
2198 	if (!BNXT_SINGLE_PF(bp) || BNXT_VF(bp)) {
2199 		PMD_DRV_LOG(ERR,
2200 			"PVID cannot be modified for this function\n");
2201 		return -ENOTSUP;
2202 	}
2203 	bp->vlan = on ? pvid : 0;
2204 
2205 	rc = bnxt_hwrm_set_default_vlan(bp, 0, 0);
2206 	if (rc)
2207 		bp->vlan = vlan;
2208 	return rc;
2209 }
2210 
2211 static int
2212 bnxt_dev_led_on_op(struct rte_eth_dev *dev)
2213 {
2214 	struct bnxt *bp = dev->data->dev_private;
2215 	int rc;
2216 
2217 	rc = is_bnxt_in_error(bp);
2218 	if (rc)
2219 		return rc;
2220 
2221 	return bnxt_hwrm_port_led_cfg(bp, true);
2222 }
2223 
2224 static int
2225 bnxt_dev_led_off_op(struct rte_eth_dev *dev)
2226 {
2227 	struct bnxt *bp = dev->data->dev_private;
2228 	int rc;
2229 
2230 	rc = is_bnxt_in_error(bp);
2231 	if (rc)
2232 		return rc;
2233 
2234 	return bnxt_hwrm_port_led_cfg(bp, false);
2235 }
2236 
2237 static uint32_t
2238 bnxt_rx_queue_count_op(struct rte_eth_dev *dev, uint16_t rx_queue_id)
2239 {
2240 	struct bnxt *bp = (struct bnxt *)dev->data->dev_private;
2241 	uint32_t desc = 0, raw_cons = 0, cons;
2242 	struct bnxt_cp_ring_info *cpr;
2243 	struct bnxt_rx_queue *rxq;
2244 	struct rx_pkt_cmpl *rxcmp;
2245 	int rc;
2246 
2247 	rc = is_bnxt_in_error(bp);
2248 	if (rc)
2249 		return rc;
2250 
2251 	rxq = dev->data->rx_queues[rx_queue_id];
2252 	cpr = rxq->cp_ring;
2253 	raw_cons = cpr->cp_raw_cons;
2254 
2255 	while (1) {
2256 		cons = RING_CMP(cpr->cp_ring_struct, raw_cons);
2257 		rte_prefetch0(&cpr->cp_desc_ring[cons]);
2258 		rxcmp = (struct rx_pkt_cmpl *)&cpr->cp_desc_ring[cons];
2259 
2260 		if (!CMP_VALID(rxcmp, raw_cons, cpr->cp_ring_struct)) {
2261 			break;
2262 		} else {
2263 			raw_cons++;
2264 			desc++;
2265 		}
2266 	}
2267 
2268 	return desc;
2269 }
2270 
2271 static int
2272 bnxt_rx_descriptor_status_op(void *rx_queue, uint16_t offset)
2273 {
2274 	struct bnxt_rx_queue *rxq = (struct bnxt_rx_queue *)rx_queue;
2275 	struct bnxt_rx_ring_info *rxr;
2276 	struct bnxt_cp_ring_info *cpr;
2277 	struct bnxt_sw_rx_bd *rx_buf;
2278 	struct rx_pkt_cmpl *rxcmp;
2279 	uint32_t cons, cp_cons;
2280 	int rc;
2281 
2282 	if (!rxq)
2283 		return -EINVAL;
2284 
2285 	rc = is_bnxt_in_error(rxq->bp);
2286 	if (rc)
2287 		return rc;
2288 
2289 	cpr = rxq->cp_ring;
2290 	rxr = rxq->rx_ring;
2291 
2292 	if (offset >= rxq->nb_rx_desc)
2293 		return -EINVAL;
2294 
2295 	cons = RING_CMP(cpr->cp_ring_struct, offset);
2296 	cp_cons = cpr->cp_raw_cons;
2297 	rxcmp = (struct rx_pkt_cmpl *)&cpr->cp_desc_ring[cons];
2298 
2299 	if (cons > cp_cons) {
2300 		if (CMPL_VALID(rxcmp, cpr->valid))
2301 			return RTE_ETH_RX_DESC_DONE;
2302 	} else {
2303 		if (CMPL_VALID(rxcmp, !cpr->valid))
2304 			return RTE_ETH_RX_DESC_DONE;
2305 	}
2306 	rx_buf = &rxr->rx_buf_ring[cons];
2307 	if (rx_buf->mbuf == NULL)
2308 		return RTE_ETH_RX_DESC_UNAVAIL;
2309 
2310 
2311 	return RTE_ETH_RX_DESC_AVAIL;
2312 }
2313 
2314 static int
2315 bnxt_tx_descriptor_status_op(void *tx_queue, uint16_t offset)
2316 {
2317 	struct bnxt_tx_queue *txq = (struct bnxt_tx_queue *)tx_queue;
2318 	struct bnxt_tx_ring_info *txr;
2319 	struct bnxt_cp_ring_info *cpr;
2320 	struct bnxt_sw_tx_bd *tx_buf;
2321 	struct tx_pkt_cmpl *txcmp;
2322 	uint32_t cons, cp_cons;
2323 	int rc;
2324 
2325 	if (!txq)
2326 		return -EINVAL;
2327 
2328 	rc = is_bnxt_in_error(txq->bp);
2329 	if (rc)
2330 		return rc;
2331 
2332 	cpr = txq->cp_ring;
2333 	txr = txq->tx_ring;
2334 
2335 	if (offset >= txq->nb_tx_desc)
2336 		return -EINVAL;
2337 
2338 	cons = RING_CMP(cpr->cp_ring_struct, offset);
2339 	txcmp = (struct tx_pkt_cmpl *)&cpr->cp_desc_ring[cons];
2340 	cp_cons = cpr->cp_raw_cons;
2341 
2342 	if (cons > cp_cons) {
2343 		if (CMPL_VALID(txcmp, cpr->valid))
2344 			return RTE_ETH_TX_DESC_UNAVAIL;
2345 	} else {
2346 		if (CMPL_VALID(txcmp, !cpr->valid))
2347 			return RTE_ETH_TX_DESC_UNAVAIL;
2348 	}
2349 	tx_buf = &txr->tx_buf_ring[cons];
2350 	if (tx_buf->mbuf == NULL)
2351 		return RTE_ETH_TX_DESC_DONE;
2352 
2353 	return RTE_ETH_TX_DESC_FULL;
2354 }
2355 
2356 static struct bnxt_filter_info *
2357 bnxt_match_and_validate_ether_filter(struct bnxt *bp,
2358 				struct rte_eth_ethertype_filter *efilter,
2359 				struct bnxt_vnic_info *vnic0,
2360 				struct bnxt_vnic_info *vnic,
2361 				int *ret)
2362 {
2363 	struct bnxt_filter_info *mfilter = NULL;
2364 	int match = 0;
2365 	*ret = 0;
2366 
2367 	if (efilter->ether_type == RTE_ETHER_TYPE_IPV4 ||
2368 		efilter->ether_type == RTE_ETHER_TYPE_IPV6) {
2369 		PMD_DRV_LOG(ERR, "invalid ether_type(0x%04x) in"
2370 			" ethertype filter.", efilter->ether_type);
2371 		*ret = -EINVAL;
2372 		goto exit;
2373 	}
2374 	if (efilter->queue >= bp->rx_nr_rings) {
2375 		PMD_DRV_LOG(ERR, "Invalid queue %d\n", efilter->queue);
2376 		*ret = -EINVAL;
2377 		goto exit;
2378 	}
2379 
2380 	vnic0 = &bp->vnic_info[0];
2381 	vnic = &bp->vnic_info[efilter->queue];
2382 	if (vnic == NULL) {
2383 		PMD_DRV_LOG(ERR, "Invalid queue %d\n", efilter->queue);
2384 		*ret = -EINVAL;
2385 		goto exit;
2386 	}
2387 
2388 	if (efilter->flags & RTE_ETHTYPE_FLAGS_DROP) {
2389 		STAILQ_FOREACH(mfilter, &vnic0->filter, next) {
2390 			if ((!memcmp(efilter->mac_addr.addr_bytes,
2391 				     mfilter->l2_addr, RTE_ETHER_ADDR_LEN) &&
2392 			     mfilter->flags ==
2393 			     HWRM_CFA_NTUPLE_FILTER_ALLOC_INPUT_FLAGS_DROP &&
2394 			     mfilter->ethertype == efilter->ether_type)) {
2395 				match = 1;
2396 				break;
2397 			}
2398 		}
2399 	} else {
2400 		STAILQ_FOREACH(mfilter, &vnic->filter, next)
2401 			if ((!memcmp(efilter->mac_addr.addr_bytes,
2402 				     mfilter->l2_addr, RTE_ETHER_ADDR_LEN) &&
2403 			     mfilter->ethertype == efilter->ether_type &&
2404 			     mfilter->flags ==
2405 			     HWRM_CFA_L2_FILTER_CFG_INPUT_FLAGS_PATH_RX)) {
2406 				match = 1;
2407 				break;
2408 			}
2409 	}
2410 
2411 	if (match)
2412 		*ret = -EEXIST;
2413 
2414 exit:
2415 	return mfilter;
2416 }
2417 
2418 static int
2419 bnxt_ethertype_filter(struct rte_eth_dev *dev,
2420 			enum rte_filter_op filter_op,
2421 			void *arg)
2422 {
2423 	struct bnxt *bp = dev->data->dev_private;
2424 	struct rte_eth_ethertype_filter *efilter =
2425 			(struct rte_eth_ethertype_filter *)arg;
2426 	struct bnxt_filter_info *bfilter, *filter1;
2427 	struct bnxt_vnic_info *vnic, *vnic0;
2428 	int ret;
2429 
2430 	if (filter_op == RTE_ETH_FILTER_NOP)
2431 		return 0;
2432 
2433 	if (arg == NULL) {
2434 		PMD_DRV_LOG(ERR, "arg shouldn't be NULL for operation %u.",
2435 			    filter_op);
2436 		return -EINVAL;
2437 	}
2438 
2439 	vnic0 = &bp->vnic_info[0];
2440 	vnic = &bp->vnic_info[efilter->queue];
2441 
2442 	switch (filter_op) {
2443 	case RTE_ETH_FILTER_ADD:
2444 		bnxt_match_and_validate_ether_filter(bp, efilter,
2445 							vnic0, vnic, &ret);
2446 		if (ret < 0)
2447 			return ret;
2448 
2449 		bfilter = bnxt_get_unused_filter(bp);
2450 		if (bfilter == NULL) {
2451 			PMD_DRV_LOG(ERR,
2452 				"Not enough resources for a new filter.\n");
2453 			return -ENOMEM;
2454 		}
2455 		bfilter->filter_type = HWRM_CFA_NTUPLE_FILTER;
2456 		memcpy(bfilter->l2_addr, efilter->mac_addr.addr_bytes,
2457 		       RTE_ETHER_ADDR_LEN);
2458 		memcpy(bfilter->dst_macaddr, efilter->mac_addr.addr_bytes,
2459 		       RTE_ETHER_ADDR_LEN);
2460 		bfilter->enables |= NTUPLE_FLTR_ALLOC_INPUT_EN_DST_MACADDR;
2461 		bfilter->ethertype = efilter->ether_type;
2462 		bfilter->enables |= NTUPLE_FLTR_ALLOC_INPUT_EN_ETHERTYPE;
2463 
2464 		filter1 = bnxt_get_l2_filter(bp, bfilter, vnic0);
2465 		if (filter1 == NULL) {
2466 			ret = -EINVAL;
2467 			goto cleanup;
2468 		}
2469 		bfilter->enables |=
2470 			HWRM_CFA_NTUPLE_FILTER_ALLOC_INPUT_ENABLES_L2_FILTER_ID;
2471 		bfilter->fw_l2_filter_id = filter1->fw_l2_filter_id;
2472 
2473 		bfilter->dst_id = vnic->fw_vnic_id;
2474 
2475 		if (efilter->flags & RTE_ETHTYPE_FLAGS_DROP) {
2476 			bfilter->flags =
2477 				HWRM_CFA_NTUPLE_FILTER_ALLOC_INPUT_FLAGS_DROP;
2478 		}
2479 
2480 		ret = bnxt_hwrm_set_ntuple_filter(bp, bfilter->dst_id, bfilter);
2481 		if (ret)
2482 			goto cleanup;
2483 		STAILQ_INSERT_TAIL(&vnic->filter, bfilter, next);
2484 		break;
2485 	case RTE_ETH_FILTER_DELETE:
2486 		filter1 = bnxt_match_and_validate_ether_filter(bp, efilter,
2487 							vnic0, vnic, &ret);
2488 		if (ret == -EEXIST) {
2489 			ret = bnxt_hwrm_clear_ntuple_filter(bp, filter1);
2490 
2491 			STAILQ_REMOVE(&vnic->filter, filter1, bnxt_filter_info,
2492 				      next);
2493 			bnxt_free_filter(bp, filter1);
2494 		} else if (ret == 0) {
2495 			PMD_DRV_LOG(ERR, "No matching filter found\n");
2496 		}
2497 		break;
2498 	default:
2499 		PMD_DRV_LOG(ERR, "unsupported operation %u.", filter_op);
2500 		ret = -EINVAL;
2501 		goto error;
2502 	}
2503 	return ret;
2504 cleanup:
2505 	bnxt_free_filter(bp, bfilter);
2506 error:
2507 	return ret;
2508 }
2509 
2510 static inline int
2511 parse_ntuple_filter(struct bnxt *bp,
2512 		    struct rte_eth_ntuple_filter *nfilter,
2513 		    struct bnxt_filter_info *bfilter)
2514 {
2515 	uint32_t en = 0;
2516 
2517 	if (nfilter->queue >= bp->rx_nr_rings) {
2518 		PMD_DRV_LOG(ERR, "Invalid queue %d\n", nfilter->queue);
2519 		return -EINVAL;
2520 	}
2521 
2522 	switch (nfilter->dst_port_mask) {
2523 	case UINT16_MAX:
2524 		bfilter->dst_port_mask = -1;
2525 		bfilter->dst_port = nfilter->dst_port;
2526 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_DST_PORT |
2527 			NTUPLE_FLTR_ALLOC_INPUT_EN_DST_PORT_MASK;
2528 		break;
2529 	default:
2530 		PMD_DRV_LOG(ERR, "invalid dst_port mask.");
2531 		return -EINVAL;
2532 	}
2533 
2534 	bfilter->ip_addr_type = NTUPLE_FLTR_ALLOC_INPUT_IP_ADDR_TYPE_IPV4;
2535 	en |= NTUPLE_FLTR_ALLOC_IN_EN_IP_PROTO;
2536 
2537 	switch (nfilter->proto_mask) {
2538 	case UINT8_MAX:
2539 		if (nfilter->proto == 17) /* IPPROTO_UDP */
2540 			bfilter->ip_protocol = 17;
2541 		else if (nfilter->proto == 6) /* IPPROTO_TCP */
2542 			bfilter->ip_protocol = 6;
2543 		else
2544 			return -EINVAL;
2545 		en |= NTUPLE_FLTR_ALLOC_IN_EN_IP_PROTO;
2546 		break;
2547 	default:
2548 		PMD_DRV_LOG(ERR, "invalid protocol mask.");
2549 		return -EINVAL;
2550 	}
2551 
2552 	switch (nfilter->dst_ip_mask) {
2553 	case UINT32_MAX:
2554 		bfilter->dst_ipaddr_mask[0] = -1;
2555 		bfilter->dst_ipaddr[0] = nfilter->dst_ip;
2556 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_DST_IPADDR |
2557 			NTUPLE_FLTR_ALLOC_INPUT_EN_DST_IPADDR_MASK;
2558 		break;
2559 	default:
2560 		PMD_DRV_LOG(ERR, "invalid dst_ip mask.");
2561 		return -EINVAL;
2562 	}
2563 
2564 	switch (nfilter->src_ip_mask) {
2565 	case UINT32_MAX:
2566 		bfilter->src_ipaddr_mask[0] = -1;
2567 		bfilter->src_ipaddr[0] = nfilter->src_ip;
2568 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_IPADDR |
2569 			NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_IPADDR_MASK;
2570 		break;
2571 	default:
2572 		PMD_DRV_LOG(ERR, "invalid src_ip mask.");
2573 		return -EINVAL;
2574 	}
2575 
2576 	switch (nfilter->src_port_mask) {
2577 	case UINT16_MAX:
2578 		bfilter->src_port_mask = -1;
2579 		bfilter->src_port = nfilter->src_port;
2580 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_PORT |
2581 			NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_PORT_MASK;
2582 		break;
2583 	default:
2584 		PMD_DRV_LOG(ERR, "invalid src_port mask.");
2585 		return -EINVAL;
2586 	}
2587 
2588 	bfilter->enables = en;
2589 	return 0;
2590 }
2591 
2592 static struct bnxt_filter_info*
2593 bnxt_match_ntuple_filter(struct bnxt *bp,
2594 			 struct bnxt_filter_info *bfilter,
2595 			 struct bnxt_vnic_info **mvnic)
2596 {
2597 	struct bnxt_filter_info *mfilter = NULL;
2598 	int i;
2599 
2600 	for (i = bp->nr_vnics - 1; i >= 0; i--) {
2601 		struct bnxt_vnic_info *vnic = &bp->vnic_info[i];
2602 		STAILQ_FOREACH(mfilter, &vnic->filter, next) {
2603 			if (bfilter->src_ipaddr[0] == mfilter->src_ipaddr[0] &&
2604 			    bfilter->src_ipaddr_mask[0] ==
2605 			    mfilter->src_ipaddr_mask[0] &&
2606 			    bfilter->src_port == mfilter->src_port &&
2607 			    bfilter->src_port_mask == mfilter->src_port_mask &&
2608 			    bfilter->dst_ipaddr[0] == mfilter->dst_ipaddr[0] &&
2609 			    bfilter->dst_ipaddr_mask[0] ==
2610 			    mfilter->dst_ipaddr_mask[0] &&
2611 			    bfilter->dst_port == mfilter->dst_port &&
2612 			    bfilter->dst_port_mask == mfilter->dst_port_mask &&
2613 			    bfilter->flags == mfilter->flags &&
2614 			    bfilter->enables == mfilter->enables) {
2615 				if (mvnic)
2616 					*mvnic = vnic;
2617 				return mfilter;
2618 			}
2619 		}
2620 	}
2621 	return NULL;
2622 }
2623 
2624 static int
2625 bnxt_cfg_ntuple_filter(struct bnxt *bp,
2626 		       struct rte_eth_ntuple_filter *nfilter,
2627 		       enum rte_filter_op filter_op)
2628 {
2629 	struct bnxt_filter_info *bfilter, *mfilter, *filter1;
2630 	struct bnxt_vnic_info *vnic, *vnic0, *mvnic;
2631 	int ret;
2632 
2633 	if (nfilter->flags != RTE_5TUPLE_FLAGS) {
2634 		PMD_DRV_LOG(ERR, "only 5tuple is supported.");
2635 		return -EINVAL;
2636 	}
2637 
2638 	if (nfilter->flags & RTE_NTUPLE_FLAGS_TCP_FLAG) {
2639 		PMD_DRV_LOG(ERR, "Ntuple filter: TCP flags not supported\n");
2640 		return -EINVAL;
2641 	}
2642 
2643 	bfilter = bnxt_get_unused_filter(bp);
2644 	if (bfilter == NULL) {
2645 		PMD_DRV_LOG(ERR,
2646 			"Not enough resources for a new filter.\n");
2647 		return -ENOMEM;
2648 	}
2649 	ret = parse_ntuple_filter(bp, nfilter, bfilter);
2650 	if (ret < 0)
2651 		goto free_filter;
2652 
2653 	vnic = &bp->vnic_info[nfilter->queue];
2654 	vnic0 = &bp->vnic_info[0];
2655 	filter1 = STAILQ_FIRST(&vnic0->filter);
2656 	if (filter1 == NULL) {
2657 		ret = -EINVAL;
2658 		goto free_filter;
2659 	}
2660 
2661 	bfilter->dst_id = vnic->fw_vnic_id;
2662 	bfilter->fw_l2_filter_id = filter1->fw_l2_filter_id;
2663 	bfilter->enables |=
2664 		HWRM_CFA_NTUPLE_FILTER_ALLOC_INPUT_ENABLES_L2_FILTER_ID;
2665 	bfilter->ethertype = 0x800;
2666 	bfilter->enables |= NTUPLE_FLTR_ALLOC_INPUT_EN_ETHERTYPE;
2667 
2668 	mfilter = bnxt_match_ntuple_filter(bp, bfilter, &mvnic);
2669 
2670 	if (mfilter != NULL && filter_op == RTE_ETH_FILTER_ADD &&
2671 	    bfilter->dst_id == mfilter->dst_id) {
2672 		PMD_DRV_LOG(ERR, "filter exists.\n");
2673 		ret = -EEXIST;
2674 		goto free_filter;
2675 	} else if (mfilter != NULL && filter_op == RTE_ETH_FILTER_ADD &&
2676 		   bfilter->dst_id != mfilter->dst_id) {
2677 		mfilter->dst_id = vnic->fw_vnic_id;
2678 		ret = bnxt_hwrm_set_ntuple_filter(bp, mfilter->dst_id, mfilter);
2679 		STAILQ_REMOVE(&mvnic->filter, mfilter, bnxt_filter_info, next);
2680 		STAILQ_INSERT_TAIL(&vnic->filter, mfilter, next);
2681 		PMD_DRV_LOG(ERR, "filter with matching pattern exists.\n");
2682 		PMD_DRV_LOG(ERR, " Updated it to the new destination queue\n");
2683 		goto free_filter;
2684 	}
2685 	if (mfilter == NULL && filter_op == RTE_ETH_FILTER_DELETE) {
2686 		PMD_DRV_LOG(ERR, "filter doesn't exist.");
2687 		ret = -ENOENT;
2688 		goto free_filter;
2689 	}
2690 
2691 	if (filter_op == RTE_ETH_FILTER_ADD) {
2692 		bfilter->filter_type = HWRM_CFA_NTUPLE_FILTER;
2693 		ret = bnxt_hwrm_set_ntuple_filter(bp, bfilter->dst_id, bfilter);
2694 		if (ret)
2695 			goto free_filter;
2696 		STAILQ_INSERT_TAIL(&vnic->filter, bfilter, next);
2697 	} else {
2698 		if (mfilter == NULL) {
2699 			/* This should not happen. But for Coverity! */
2700 			ret = -ENOENT;
2701 			goto free_filter;
2702 		}
2703 		ret = bnxt_hwrm_clear_ntuple_filter(bp, mfilter);
2704 
2705 		STAILQ_REMOVE(&vnic->filter, mfilter, bnxt_filter_info, next);
2706 		bnxt_free_filter(bp, mfilter);
2707 		mfilter->fw_l2_filter_id = -1;
2708 		bnxt_free_filter(bp, bfilter);
2709 		bfilter->fw_l2_filter_id = -1;
2710 	}
2711 
2712 	return 0;
2713 free_filter:
2714 	bfilter->fw_l2_filter_id = -1;
2715 	bnxt_free_filter(bp, bfilter);
2716 	return ret;
2717 }
2718 
2719 static int
2720 bnxt_ntuple_filter(struct rte_eth_dev *dev,
2721 			enum rte_filter_op filter_op,
2722 			void *arg)
2723 {
2724 	struct bnxt *bp = dev->data->dev_private;
2725 	int ret;
2726 
2727 	if (filter_op == RTE_ETH_FILTER_NOP)
2728 		return 0;
2729 
2730 	if (arg == NULL) {
2731 		PMD_DRV_LOG(ERR, "arg shouldn't be NULL for operation %u.",
2732 			    filter_op);
2733 		return -EINVAL;
2734 	}
2735 
2736 	switch (filter_op) {
2737 	case RTE_ETH_FILTER_ADD:
2738 		ret = bnxt_cfg_ntuple_filter(bp,
2739 			(struct rte_eth_ntuple_filter *)arg,
2740 			filter_op);
2741 		break;
2742 	case RTE_ETH_FILTER_DELETE:
2743 		ret = bnxt_cfg_ntuple_filter(bp,
2744 			(struct rte_eth_ntuple_filter *)arg,
2745 			filter_op);
2746 		break;
2747 	default:
2748 		PMD_DRV_LOG(ERR, "unsupported operation %u.", filter_op);
2749 		ret = -EINVAL;
2750 		break;
2751 	}
2752 	return ret;
2753 }
2754 
2755 static int
2756 bnxt_parse_fdir_filter(struct bnxt *bp,
2757 		       struct rte_eth_fdir_filter *fdir,
2758 		       struct bnxt_filter_info *filter)
2759 {
2760 	enum rte_fdir_mode fdir_mode =
2761 		bp->eth_dev->data->dev_conf.fdir_conf.mode;
2762 	struct bnxt_vnic_info *vnic0, *vnic;
2763 	struct bnxt_filter_info *filter1;
2764 	uint32_t en = 0;
2765 	int i;
2766 
2767 	if (fdir_mode == RTE_FDIR_MODE_PERFECT_TUNNEL)
2768 		return -EINVAL;
2769 
2770 	filter->l2_ovlan = fdir->input.flow_ext.vlan_tci;
2771 	en |= EM_FLOW_ALLOC_INPUT_EN_OVLAN_VID;
2772 
2773 	switch (fdir->input.flow_type) {
2774 	case RTE_ETH_FLOW_IPV4:
2775 	case RTE_ETH_FLOW_NONFRAG_IPV4_OTHER:
2776 		/* FALLTHROUGH */
2777 		filter->src_ipaddr[0] = fdir->input.flow.ip4_flow.src_ip;
2778 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_IPADDR;
2779 		filter->dst_ipaddr[0] = fdir->input.flow.ip4_flow.dst_ip;
2780 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_DST_IPADDR;
2781 		filter->ip_protocol = fdir->input.flow.ip4_flow.proto;
2782 		en |= NTUPLE_FLTR_ALLOC_IN_EN_IP_PROTO;
2783 		filter->ip_addr_type =
2784 			NTUPLE_FLTR_ALLOC_INPUT_IP_ADDR_TYPE_IPV4;
2785 		filter->src_ipaddr_mask[0] = 0xffffffff;
2786 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_IPADDR_MASK;
2787 		filter->dst_ipaddr_mask[0] = 0xffffffff;
2788 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_DST_IPADDR_MASK;
2789 		filter->ethertype = 0x800;
2790 		filter->enables |= NTUPLE_FLTR_ALLOC_INPUT_EN_ETHERTYPE;
2791 		break;
2792 	case RTE_ETH_FLOW_NONFRAG_IPV4_TCP:
2793 		filter->src_port = fdir->input.flow.tcp4_flow.src_port;
2794 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_PORT;
2795 		filter->dst_port = fdir->input.flow.tcp4_flow.dst_port;
2796 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_DST_PORT;
2797 		filter->dst_port_mask = 0xffff;
2798 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_DST_PORT_MASK;
2799 		filter->src_port_mask = 0xffff;
2800 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_PORT_MASK;
2801 		filter->src_ipaddr[0] = fdir->input.flow.tcp4_flow.ip.src_ip;
2802 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_IPADDR;
2803 		filter->dst_ipaddr[0] = fdir->input.flow.tcp4_flow.ip.dst_ip;
2804 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_DST_IPADDR;
2805 		filter->ip_protocol = 6;
2806 		en |= NTUPLE_FLTR_ALLOC_IN_EN_IP_PROTO;
2807 		filter->ip_addr_type =
2808 			NTUPLE_FLTR_ALLOC_INPUT_IP_ADDR_TYPE_IPV4;
2809 		filter->src_ipaddr_mask[0] = 0xffffffff;
2810 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_IPADDR_MASK;
2811 		filter->dst_ipaddr_mask[0] = 0xffffffff;
2812 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_DST_IPADDR_MASK;
2813 		filter->ethertype = 0x800;
2814 		filter->enables |= NTUPLE_FLTR_ALLOC_INPUT_EN_ETHERTYPE;
2815 		break;
2816 	case RTE_ETH_FLOW_NONFRAG_IPV4_UDP:
2817 		filter->src_port = fdir->input.flow.udp4_flow.src_port;
2818 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_PORT;
2819 		filter->dst_port = fdir->input.flow.udp4_flow.dst_port;
2820 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_DST_PORT;
2821 		filter->dst_port_mask = 0xffff;
2822 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_DST_PORT_MASK;
2823 		filter->src_port_mask = 0xffff;
2824 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_PORT_MASK;
2825 		filter->src_ipaddr[0] = fdir->input.flow.udp4_flow.ip.src_ip;
2826 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_IPADDR;
2827 		filter->dst_ipaddr[0] = fdir->input.flow.udp4_flow.ip.dst_ip;
2828 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_DST_IPADDR;
2829 		filter->ip_protocol = 17;
2830 		en |= NTUPLE_FLTR_ALLOC_IN_EN_IP_PROTO;
2831 		filter->ip_addr_type =
2832 			NTUPLE_FLTR_ALLOC_INPUT_IP_ADDR_TYPE_IPV4;
2833 		filter->src_ipaddr_mask[0] = 0xffffffff;
2834 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_IPADDR_MASK;
2835 		filter->dst_ipaddr_mask[0] = 0xffffffff;
2836 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_DST_IPADDR_MASK;
2837 		filter->ethertype = 0x800;
2838 		filter->enables |= NTUPLE_FLTR_ALLOC_INPUT_EN_ETHERTYPE;
2839 		break;
2840 	case RTE_ETH_FLOW_IPV6:
2841 	case RTE_ETH_FLOW_NONFRAG_IPV6_OTHER:
2842 		/* FALLTHROUGH */
2843 		filter->ip_addr_type =
2844 			NTUPLE_FLTR_ALLOC_INPUT_IP_ADDR_TYPE_IPV6;
2845 		filter->ip_protocol = fdir->input.flow.ipv6_flow.proto;
2846 		en |= NTUPLE_FLTR_ALLOC_IN_EN_IP_PROTO;
2847 		rte_memcpy(filter->src_ipaddr,
2848 			   fdir->input.flow.ipv6_flow.src_ip, 16);
2849 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_IPADDR;
2850 		rte_memcpy(filter->dst_ipaddr,
2851 			   fdir->input.flow.ipv6_flow.dst_ip, 16);
2852 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_DST_IPADDR;
2853 		memset(filter->dst_ipaddr_mask, 0xff, 16);
2854 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_DST_IPADDR_MASK;
2855 		memset(filter->src_ipaddr_mask, 0xff, 16);
2856 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_IPADDR_MASK;
2857 		filter->ethertype = 0x86dd;
2858 		filter->enables |= NTUPLE_FLTR_ALLOC_INPUT_EN_ETHERTYPE;
2859 		break;
2860 	case RTE_ETH_FLOW_NONFRAG_IPV6_TCP:
2861 		filter->src_port = fdir->input.flow.tcp6_flow.src_port;
2862 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_PORT;
2863 		filter->dst_port = fdir->input.flow.tcp6_flow.dst_port;
2864 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_DST_PORT;
2865 		filter->dst_port_mask = 0xffff;
2866 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_DST_PORT_MASK;
2867 		filter->src_port_mask = 0xffff;
2868 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_PORT_MASK;
2869 		filter->ip_addr_type =
2870 			NTUPLE_FLTR_ALLOC_INPUT_IP_ADDR_TYPE_IPV6;
2871 		filter->ip_protocol = fdir->input.flow.tcp6_flow.ip.proto;
2872 		en |= NTUPLE_FLTR_ALLOC_IN_EN_IP_PROTO;
2873 		rte_memcpy(filter->src_ipaddr,
2874 			   fdir->input.flow.tcp6_flow.ip.src_ip, 16);
2875 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_IPADDR;
2876 		rte_memcpy(filter->dst_ipaddr,
2877 			   fdir->input.flow.tcp6_flow.ip.dst_ip, 16);
2878 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_DST_IPADDR;
2879 		memset(filter->dst_ipaddr_mask, 0xff, 16);
2880 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_DST_IPADDR_MASK;
2881 		memset(filter->src_ipaddr_mask, 0xff, 16);
2882 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_IPADDR_MASK;
2883 		filter->ethertype = 0x86dd;
2884 		filter->enables |= NTUPLE_FLTR_ALLOC_INPUT_EN_ETHERTYPE;
2885 		break;
2886 	case RTE_ETH_FLOW_NONFRAG_IPV6_UDP:
2887 		filter->src_port = fdir->input.flow.udp6_flow.src_port;
2888 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_PORT;
2889 		filter->dst_port = fdir->input.flow.udp6_flow.dst_port;
2890 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_DST_PORT;
2891 		filter->dst_port_mask = 0xffff;
2892 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_DST_PORT_MASK;
2893 		filter->src_port_mask = 0xffff;
2894 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_PORT_MASK;
2895 		filter->ip_addr_type =
2896 			NTUPLE_FLTR_ALLOC_INPUT_IP_ADDR_TYPE_IPV6;
2897 		filter->ip_protocol = fdir->input.flow.udp6_flow.ip.proto;
2898 		en |= NTUPLE_FLTR_ALLOC_IN_EN_IP_PROTO;
2899 		rte_memcpy(filter->src_ipaddr,
2900 			   fdir->input.flow.udp6_flow.ip.src_ip, 16);
2901 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_IPADDR;
2902 		rte_memcpy(filter->dst_ipaddr,
2903 			   fdir->input.flow.udp6_flow.ip.dst_ip, 16);
2904 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_DST_IPADDR;
2905 		memset(filter->dst_ipaddr_mask, 0xff, 16);
2906 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_DST_IPADDR_MASK;
2907 		memset(filter->src_ipaddr_mask, 0xff, 16);
2908 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_IPADDR_MASK;
2909 		filter->ethertype = 0x86dd;
2910 		filter->enables |= NTUPLE_FLTR_ALLOC_INPUT_EN_ETHERTYPE;
2911 		break;
2912 	case RTE_ETH_FLOW_L2_PAYLOAD:
2913 		filter->ethertype = fdir->input.flow.l2_flow.ether_type;
2914 		en |= NTUPLE_FLTR_ALLOC_INPUT_EN_ETHERTYPE;
2915 		break;
2916 	case RTE_ETH_FLOW_VXLAN:
2917 		if (fdir->action.behavior == RTE_ETH_FDIR_REJECT)
2918 			return -EINVAL;
2919 		filter->vni = fdir->input.flow.tunnel_flow.tunnel_id;
2920 		filter->tunnel_type =
2921 			CFA_NTUPLE_FILTER_ALLOC_REQ_TUNNEL_TYPE_VXLAN;
2922 		en |= HWRM_CFA_NTUPLE_FILTER_ALLOC_INPUT_ENABLES_TUNNEL_TYPE;
2923 		break;
2924 	case RTE_ETH_FLOW_NVGRE:
2925 		if (fdir->action.behavior == RTE_ETH_FDIR_REJECT)
2926 			return -EINVAL;
2927 		filter->vni = fdir->input.flow.tunnel_flow.tunnel_id;
2928 		filter->tunnel_type =
2929 			CFA_NTUPLE_FILTER_ALLOC_REQ_TUNNEL_TYPE_NVGRE;
2930 		en |= HWRM_CFA_NTUPLE_FILTER_ALLOC_INPUT_ENABLES_TUNNEL_TYPE;
2931 		break;
2932 	case RTE_ETH_FLOW_UNKNOWN:
2933 	case RTE_ETH_FLOW_RAW:
2934 	case RTE_ETH_FLOW_FRAG_IPV4:
2935 	case RTE_ETH_FLOW_NONFRAG_IPV4_SCTP:
2936 	case RTE_ETH_FLOW_FRAG_IPV6:
2937 	case RTE_ETH_FLOW_NONFRAG_IPV6_SCTP:
2938 	case RTE_ETH_FLOW_IPV6_EX:
2939 	case RTE_ETH_FLOW_IPV6_TCP_EX:
2940 	case RTE_ETH_FLOW_IPV6_UDP_EX:
2941 	case RTE_ETH_FLOW_GENEVE:
2942 		/* FALLTHROUGH */
2943 	default:
2944 		return -EINVAL;
2945 	}
2946 
2947 	vnic0 = &bp->vnic_info[0];
2948 	vnic = &bp->vnic_info[fdir->action.rx_queue];
2949 	if (vnic == NULL) {
2950 		PMD_DRV_LOG(ERR, "Invalid queue %d\n", fdir->action.rx_queue);
2951 		return -EINVAL;
2952 	}
2953 
2954 	if (fdir_mode == RTE_FDIR_MODE_PERFECT_MAC_VLAN) {
2955 		rte_memcpy(filter->dst_macaddr,
2956 			fdir->input.flow.mac_vlan_flow.mac_addr.addr_bytes, 6);
2957 			en |= NTUPLE_FLTR_ALLOC_INPUT_EN_DST_MACADDR;
2958 	}
2959 
2960 	if (fdir->action.behavior == RTE_ETH_FDIR_REJECT) {
2961 		filter->flags = HWRM_CFA_NTUPLE_FILTER_ALLOC_INPUT_FLAGS_DROP;
2962 		filter1 = STAILQ_FIRST(&vnic0->filter);
2963 		//filter1 = bnxt_get_l2_filter(bp, filter, vnic0);
2964 	} else {
2965 		filter->dst_id = vnic->fw_vnic_id;
2966 		for (i = 0; i < RTE_ETHER_ADDR_LEN; i++)
2967 			if (filter->dst_macaddr[i] == 0x00)
2968 				filter1 = STAILQ_FIRST(&vnic0->filter);
2969 			else
2970 				filter1 = bnxt_get_l2_filter(bp, filter, vnic);
2971 	}
2972 
2973 	if (filter1 == NULL)
2974 		return -EINVAL;
2975 
2976 	en |= HWRM_CFA_NTUPLE_FILTER_ALLOC_INPUT_ENABLES_L2_FILTER_ID;
2977 	filter->fw_l2_filter_id = filter1->fw_l2_filter_id;
2978 
2979 	filter->enables = en;
2980 
2981 	return 0;
2982 }
2983 
2984 static struct bnxt_filter_info *
2985 bnxt_match_fdir(struct bnxt *bp, struct bnxt_filter_info *nf,
2986 		struct bnxt_vnic_info **mvnic)
2987 {
2988 	struct bnxt_filter_info *mf = NULL;
2989 	int i;
2990 
2991 	for (i = bp->nr_vnics - 1; i >= 0; i--) {
2992 		struct bnxt_vnic_info *vnic = &bp->vnic_info[i];
2993 
2994 		STAILQ_FOREACH(mf, &vnic->filter, next) {
2995 			if (mf->filter_type == nf->filter_type &&
2996 			    mf->flags == nf->flags &&
2997 			    mf->src_port == nf->src_port &&
2998 			    mf->src_port_mask == nf->src_port_mask &&
2999 			    mf->dst_port == nf->dst_port &&
3000 			    mf->dst_port_mask == nf->dst_port_mask &&
3001 			    mf->ip_protocol == nf->ip_protocol &&
3002 			    mf->ip_addr_type == nf->ip_addr_type &&
3003 			    mf->ethertype == nf->ethertype &&
3004 			    mf->vni == nf->vni &&
3005 			    mf->tunnel_type == nf->tunnel_type &&
3006 			    mf->l2_ovlan == nf->l2_ovlan &&
3007 			    mf->l2_ovlan_mask == nf->l2_ovlan_mask &&
3008 			    mf->l2_ivlan == nf->l2_ivlan &&
3009 			    mf->l2_ivlan_mask == nf->l2_ivlan_mask &&
3010 			    !memcmp(mf->l2_addr, nf->l2_addr,
3011 				    RTE_ETHER_ADDR_LEN) &&
3012 			    !memcmp(mf->l2_addr_mask, nf->l2_addr_mask,
3013 				    RTE_ETHER_ADDR_LEN) &&
3014 			    !memcmp(mf->src_macaddr, nf->src_macaddr,
3015 				    RTE_ETHER_ADDR_LEN) &&
3016 			    !memcmp(mf->dst_macaddr, nf->dst_macaddr,
3017 				    RTE_ETHER_ADDR_LEN) &&
3018 			    !memcmp(mf->src_ipaddr, nf->src_ipaddr,
3019 				    sizeof(nf->src_ipaddr)) &&
3020 			    !memcmp(mf->src_ipaddr_mask, nf->src_ipaddr_mask,
3021 				    sizeof(nf->src_ipaddr_mask)) &&
3022 			    !memcmp(mf->dst_ipaddr, nf->dst_ipaddr,
3023 				    sizeof(nf->dst_ipaddr)) &&
3024 			    !memcmp(mf->dst_ipaddr_mask, nf->dst_ipaddr_mask,
3025 				    sizeof(nf->dst_ipaddr_mask))) {
3026 				if (mvnic)
3027 					*mvnic = vnic;
3028 				return mf;
3029 			}
3030 		}
3031 	}
3032 	return NULL;
3033 }
3034 
3035 static int
3036 bnxt_fdir_filter(struct rte_eth_dev *dev,
3037 		 enum rte_filter_op filter_op,
3038 		 void *arg)
3039 {
3040 	struct bnxt *bp = dev->data->dev_private;
3041 	struct rte_eth_fdir_filter *fdir  = (struct rte_eth_fdir_filter *)arg;
3042 	struct bnxt_filter_info *filter, *match;
3043 	struct bnxt_vnic_info *vnic, *mvnic;
3044 	int ret = 0, i;
3045 
3046 	if (filter_op == RTE_ETH_FILTER_NOP)
3047 		return 0;
3048 
3049 	if (arg == NULL && filter_op != RTE_ETH_FILTER_FLUSH)
3050 		return -EINVAL;
3051 
3052 	switch (filter_op) {
3053 	case RTE_ETH_FILTER_ADD:
3054 	case RTE_ETH_FILTER_DELETE:
3055 		/* FALLTHROUGH */
3056 		filter = bnxt_get_unused_filter(bp);
3057 		if (filter == NULL) {
3058 			PMD_DRV_LOG(ERR,
3059 				"Not enough resources for a new flow.\n");
3060 			return -ENOMEM;
3061 		}
3062 
3063 		ret = bnxt_parse_fdir_filter(bp, fdir, filter);
3064 		if (ret != 0)
3065 			goto free_filter;
3066 		filter->filter_type = HWRM_CFA_NTUPLE_FILTER;
3067 
3068 		if (fdir->action.behavior == RTE_ETH_FDIR_REJECT)
3069 			vnic = &bp->vnic_info[0];
3070 		else
3071 			vnic = &bp->vnic_info[fdir->action.rx_queue];
3072 
3073 		match = bnxt_match_fdir(bp, filter, &mvnic);
3074 		if (match != NULL && filter_op == RTE_ETH_FILTER_ADD) {
3075 			if (match->dst_id == vnic->fw_vnic_id) {
3076 				PMD_DRV_LOG(ERR, "Flow already exists.\n");
3077 				ret = -EEXIST;
3078 				goto free_filter;
3079 			} else {
3080 				match->dst_id = vnic->fw_vnic_id;
3081 				ret = bnxt_hwrm_set_ntuple_filter(bp,
3082 								  match->dst_id,
3083 								  match);
3084 				STAILQ_REMOVE(&mvnic->filter, match,
3085 					      bnxt_filter_info, next);
3086 				STAILQ_INSERT_TAIL(&vnic->filter, match, next);
3087 				PMD_DRV_LOG(ERR,
3088 					"Filter with matching pattern exist\n");
3089 				PMD_DRV_LOG(ERR,
3090 					"Updated it to new destination q\n");
3091 				goto free_filter;
3092 			}
3093 		}
3094 		if (match == NULL && filter_op == RTE_ETH_FILTER_DELETE) {
3095 			PMD_DRV_LOG(ERR, "Flow does not exist.\n");
3096 			ret = -ENOENT;
3097 			goto free_filter;
3098 		}
3099 
3100 		if (filter_op == RTE_ETH_FILTER_ADD) {
3101 			ret = bnxt_hwrm_set_ntuple_filter(bp,
3102 							  filter->dst_id,
3103 							  filter);
3104 			if (ret)
3105 				goto free_filter;
3106 			STAILQ_INSERT_TAIL(&vnic->filter, filter, next);
3107 		} else {
3108 			ret = bnxt_hwrm_clear_ntuple_filter(bp, match);
3109 			STAILQ_REMOVE(&vnic->filter, match,
3110 				      bnxt_filter_info, next);
3111 			bnxt_free_filter(bp, match);
3112 			filter->fw_l2_filter_id = -1;
3113 			bnxt_free_filter(bp, filter);
3114 		}
3115 		break;
3116 	case RTE_ETH_FILTER_FLUSH:
3117 		for (i = bp->nr_vnics - 1; i >= 0; i--) {
3118 			struct bnxt_vnic_info *vnic = &bp->vnic_info[i];
3119 
3120 			STAILQ_FOREACH(filter, &vnic->filter, next) {
3121 				if (filter->filter_type ==
3122 				    HWRM_CFA_NTUPLE_FILTER) {
3123 					ret =
3124 					bnxt_hwrm_clear_ntuple_filter(bp,
3125 								      filter);
3126 					STAILQ_REMOVE(&vnic->filter, filter,
3127 						      bnxt_filter_info, next);
3128 				}
3129 			}
3130 		}
3131 		return ret;
3132 	case RTE_ETH_FILTER_UPDATE:
3133 	case RTE_ETH_FILTER_STATS:
3134 	case RTE_ETH_FILTER_INFO:
3135 		PMD_DRV_LOG(ERR, "operation %u not implemented", filter_op);
3136 		break;
3137 	default:
3138 		PMD_DRV_LOG(ERR, "unknown operation %u", filter_op);
3139 		ret = -EINVAL;
3140 		break;
3141 	}
3142 	return ret;
3143 
3144 free_filter:
3145 	filter->fw_l2_filter_id = -1;
3146 	bnxt_free_filter(bp, filter);
3147 	return ret;
3148 }
3149 
3150 static int
3151 bnxt_filter_ctrl_op(struct rte_eth_dev *dev __rte_unused,
3152 		    enum rte_filter_type filter_type,
3153 		    enum rte_filter_op filter_op, void *arg)
3154 {
3155 	int ret = 0;
3156 
3157 	ret = is_bnxt_in_error(dev->data->dev_private);
3158 	if (ret)
3159 		return ret;
3160 
3161 	switch (filter_type) {
3162 	case RTE_ETH_FILTER_TUNNEL:
3163 		PMD_DRV_LOG(ERR,
3164 			"filter type: %d: To be implemented\n", filter_type);
3165 		break;
3166 	case RTE_ETH_FILTER_FDIR:
3167 		ret = bnxt_fdir_filter(dev, filter_op, arg);
3168 		break;
3169 	case RTE_ETH_FILTER_NTUPLE:
3170 		ret = bnxt_ntuple_filter(dev, filter_op, arg);
3171 		break;
3172 	case RTE_ETH_FILTER_ETHERTYPE:
3173 		ret = bnxt_ethertype_filter(dev, filter_op, arg);
3174 		break;
3175 	case RTE_ETH_FILTER_GENERIC:
3176 		if (filter_op != RTE_ETH_FILTER_GET)
3177 			return -EINVAL;
3178 		*(const void **)arg = &bnxt_flow_ops;
3179 		break;
3180 	default:
3181 		PMD_DRV_LOG(ERR,
3182 			"Filter type (%d) not supported", filter_type);
3183 		ret = -EINVAL;
3184 		break;
3185 	}
3186 	return ret;
3187 }
3188 
3189 static const uint32_t *
3190 bnxt_dev_supported_ptypes_get_op(struct rte_eth_dev *dev)
3191 {
3192 	static const uint32_t ptypes[] = {
3193 		RTE_PTYPE_L2_ETHER_VLAN,
3194 		RTE_PTYPE_L3_IPV4_EXT_UNKNOWN,
3195 		RTE_PTYPE_L3_IPV6_EXT_UNKNOWN,
3196 		RTE_PTYPE_L4_ICMP,
3197 		RTE_PTYPE_L4_TCP,
3198 		RTE_PTYPE_L4_UDP,
3199 		RTE_PTYPE_INNER_L3_IPV4_EXT_UNKNOWN,
3200 		RTE_PTYPE_INNER_L3_IPV6_EXT_UNKNOWN,
3201 		RTE_PTYPE_INNER_L4_ICMP,
3202 		RTE_PTYPE_INNER_L4_TCP,
3203 		RTE_PTYPE_INNER_L4_UDP,
3204 		RTE_PTYPE_UNKNOWN
3205 	};
3206 
3207 	if (!dev->rx_pkt_burst)
3208 		return NULL;
3209 
3210 	return ptypes;
3211 }
3212 
3213 static int bnxt_map_regs(struct bnxt *bp, uint32_t *reg_arr, int count,
3214 			 int reg_win)
3215 {
3216 	uint32_t reg_base = *reg_arr & 0xfffff000;
3217 	uint32_t win_off;
3218 	int i;
3219 
3220 	for (i = 0; i < count; i++) {
3221 		if ((reg_arr[i] & 0xfffff000) != reg_base)
3222 			return -ERANGE;
3223 	}
3224 	win_off = BNXT_GRCPF_REG_WINDOW_BASE_OUT + (reg_win - 1) * 4;
3225 	rte_write32(reg_base, (uint8_t *)bp->bar0 + win_off);
3226 	return 0;
3227 }
3228 
3229 static int bnxt_map_ptp_regs(struct bnxt *bp)
3230 {
3231 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3232 	uint32_t *reg_arr;
3233 	int rc, i;
3234 
3235 	reg_arr = ptp->rx_regs;
3236 	rc = bnxt_map_regs(bp, reg_arr, BNXT_PTP_RX_REGS, 5);
3237 	if (rc)
3238 		return rc;
3239 
3240 	reg_arr = ptp->tx_regs;
3241 	rc = bnxt_map_regs(bp, reg_arr, BNXT_PTP_TX_REGS, 6);
3242 	if (rc)
3243 		return rc;
3244 
3245 	for (i = 0; i < BNXT_PTP_RX_REGS; i++)
3246 		ptp->rx_mapped_regs[i] = 0x5000 + (ptp->rx_regs[i] & 0xfff);
3247 
3248 	for (i = 0; i < BNXT_PTP_TX_REGS; i++)
3249 		ptp->tx_mapped_regs[i] = 0x6000 + (ptp->tx_regs[i] & 0xfff);
3250 
3251 	return 0;
3252 }
3253 
3254 static void bnxt_unmap_ptp_regs(struct bnxt *bp)
3255 {
3256 	rte_write32(0, (uint8_t *)bp->bar0 +
3257 			 BNXT_GRCPF_REG_WINDOW_BASE_OUT + 16);
3258 	rte_write32(0, (uint8_t *)bp->bar0 +
3259 			 BNXT_GRCPF_REG_WINDOW_BASE_OUT + 20);
3260 }
3261 
3262 static uint64_t bnxt_cc_read(struct bnxt *bp)
3263 {
3264 	uint64_t ns;
3265 
3266 	ns = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3267 			      BNXT_GRCPF_REG_SYNC_TIME));
3268 	ns |= (uint64_t)(rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3269 					  BNXT_GRCPF_REG_SYNC_TIME + 4))) << 32;
3270 	return ns;
3271 }
3272 
3273 static int bnxt_get_tx_ts(struct bnxt *bp, uint64_t *ts)
3274 {
3275 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3276 	uint32_t fifo;
3277 
3278 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3279 				ptp->tx_mapped_regs[BNXT_PTP_TX_FIFO]));
3280 	if (fifo & BNXT_PTP_TX_FIFO_EMPTY)
3281 		return -EAGAIN;
3282 
3283 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3284 				ptp->tx_mapped_regs[BNXT_PTP_TX_FIFO]));
3285 	*ts = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3286 				ptp->tx_mapped_regs[BNXT_PTP_TX_TS_L]));
3287 	*ts |= (uint64_t)rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3288 				ptp->tx_mapped_regs[BNXT_PTP_TX_TS_H])) << 32;
3289 
3290 	return 0;
3291 }
3292 
3293 static int bnxt_get_rx_ts(struct bnxt *bp, uint64_t *ts)
3294 {
3295 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3296 	struct bnxt_pf_info *pf = &bp->pf;
3297 	uint16_t port_id;
3298 	uint32_t fifo;
3299 
3300 	if (!ptp)
3301 		return -ENODEV;
3302 
3303 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3304 				ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO]));
3305 	if (!(fifo & BNXT_PTP_RX_FIFO_PENDING))
3306 		return -EAGAIN;
3307 
3308 	port_id = pf->port_id;
3309 	rte_write32(1 << port_id, (uint8_t *)bp->bar0 +
3310 	       ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO_ADV]);
3311 
3312 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3313 				   ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO]));
3314 	if (fifo & BNXT_PTP_RX_FIFO_PENDING) {
3315 /*		bnxt_clr_rx_ts(bp);	  TBD  */
3316 		return -EBUSY;
3317 	}
3318 
3319 	*ts = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3320 				ptp->rx_mapped_regs[BNXT_PTP_RX_TS_L]));
3321 	*ts |= (uint64_t)rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3322 				ptp->rx_mapped_regs[BNXT_PTP_RX_TS_H])) << 32;
3323 
3324 	return 0;
3325 }
3326 
3327 static int
3328 bnxt_timesync_write_time(struct rte_eth_dev *dev, const struct timespec *ts)
3329 {
3330 	uint64_t ns;
3331 	struct bnxt *bp = dev->data->dev_private;
3332 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3333 
3334 	if (!ptp)
3335 		return 0;
3336 
3337 	ns = rte_timespec_to_ns(ts);
3338 	/* Set the timecounters to a new value. */
3339 	ptp->tc.nsec = ns;
3340 
3341 	return 0;
3342 }
3343 
3344 static int
3345 bnxt_timesync_read_time(struct rte_eth_dev *dev, struct timespec *ts)
3346 {
3347 	struct bnxt *bp = dev->data->dev_private;
3348 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3349 	uint64_t ns, systime_cycles = 0;
3350 	int rc = 0;
3351 
3352 	if (!ptp)
3353 		return 0;
3354 
3355 	if (BNXT_CHIP_THOR(bp))
3356 		rc = bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_CURRENT_TIME,
3357 					     &systime_cycles);
3358 	else
3359 		systime_cycles = bnxt_cc_read(bp);
3360 
3361 	ns = rte_timecounter_update(&ptp->tc, systime_cycles);
3362 	*ts = rte_ns_to_timespec(ns);
3363 
3364 	return rc;
3365 }
3366 static int
3367 bnxt_timesync_enable(struct rte_eth_dev *dev)
3368 {
3369 	struct bnxt *bp = dev->data->dev_private;
3370 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3371 	uint32_t shift = 0;
3372 	int rc;
3373 
3374 	if (!ptp)
3375 		return 0;
3376 
3377 	ptp->rx_filter = 1;
3378 	ptp->tx_tstamp_en = 1;
3379 	ptp->rxctl = BNXT_PTP_MSG_EVENTS;
3380 
3381 	rc = bnxt_hwrm_ptp_cfg(bp);
3382 	if (rc)
3383 		return rc;
3384 
3385 	memset(&ptp->tc, 0, sizeof(struct rte_timecounter));
3386 	memset(&ptp->rx_tstamp_tc, 0, sizeof(struct rte_timecounter));
3387 	memset(&ptp->tx_tstamp_tc, 0, sizeof(struct rte_timecounter));
3388 
3389 	ptp->tc.cc_mask = BNXT_CYCLECOUNTER_MASK;
3390 	ptp->tc.cc_shift = shift;
3391 	ptp->tc.nsec_mask = (1ULL << shift) - 1;
3392 
3393 	ptp->rx_tstamp_tc.cc_mask = BNXT_CYCLECOUNTER_MASK;
3394 	ptp->rx_tstamp_tc.cc_shift = shift;
3395 	ptp->rx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
3396 
3397 	ptp->tx_tstamp_tc.cc_mask = BNXT_CYCLECOUNTER_MASK;
3398 	ptp->tx_tstamp_tc.cc_shift = shift;
3399 	ptp->tx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
3400 
3401 	if (!BNXT_CHIP_THOR(bp))
3402 		bnxt_map_ptp_regs(bp);
3403 
3404 	return 0;
3405 }
3406 
3407 static int
3408 bnxt_timesync_disable(struct rte_eth_dev *dev)
3409 {
3410 	struct bnxt *bp = dev->data->dev_private;
3411 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3412 
3413 	if (!ptp)
3414 		return 0;
3415 
3416 	ptp->rx_filter = 0;
3417 	ptp->tx_tstamp_en = 0;
3418 	ptp->rxctl = 0;
3419 
3420 	bnxt_hwrm_ptp_cfg(bp);
3421 
3422 	if (!BNXT_CHIP_THOR(bp))
3423 		bnxt_unmap_ptp_regs(bp);
3424 
3425 	return 0;
3426 }
3427 
3428 static int
3429 bnxt_timesync_read_rx_timestamp(struct rte_eth_dev *dev,
3430 				 struct timespec *timestamp,
3431 				 uint32_t flags __rte_unused)
3432 {
3433 	struct bnxt *bp = dev->data->dev_private;
3434 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3435 	uint64_t rx_tstamp_cycles = 0;
3436 	uint64_t ns;
3437 
3438 	if (!ptp)
3439 		return 0;
3440 
3441 	if (BNXT_CHIP_THOR(bp))
3442 		rx_tstamp_cycles = ptp->rx_timestamp;
3443 	else
3444 		bnxt_get_rx_ts(bp, &rx_tstamp_cycles);
3445 
3446 	ns = rte_timecounter_update(&ptp->rx_tstamp_tc, rx_tstamp_cycles);
3447 	*timestamp = rte_ns_to_timespec(ns);
3448 	return  0;
3449 }
3450 
3451 static int
3452 bnxt_timesync_read_tx_timestamp(struct rte_eth_dev *dev,
3453 				 struct timespec *timestamp)
3454 {
3455 	struct bnxt *bp = dev->data->dev_private;
3456 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3457 	uint64_t tx_tstamp_cycles = 0;
3458 	uint64_t ns;
3459 	int rc = 0;
3460 
3461 	if (!ptp)
3462 		return 0;
3463 
3464 	if (BNXT_CHIP_THOR(bp))
3465 		rc = bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_PATH_TX,
3466 					     &tx_tstamp_cycles);
3467 	else
3468 		rc = bnxt_get_tx_ts(bp, &tx_tstamp_cycles);
3469 
3470 	ns = rte_timecounter_update(&ptp->tx_tstamp_tc, tx_tstamp_cycles);
3471 	*timestamp = rte_ns_to_timespec(ns);
3472 
3473 	return rc;
3474 }
3475 
3476 static int
3477 bnxt_timesync_adjust_time(struct rte_eth_dev *dev, int64_t delta)
3478 {
3479 	struct bnxt *bp = dev->data->dev_private;
3480 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3481 
3482 	if (!ptp)
3483 		return 0;
3484 
3485 	ptp->tc.nsec += delta;
3486 
3487 	return 0;
3488 }
3489 
3490 static int
3491 bnxt_get_eeprom_length_op(struct rte_eth_dev *dev)
3492 {
3493 	struct bnxt *bp = dev->data->dev_private;
3494 	int rc;
3495 	uint32_t dir_entries;
3496 	uint32_t entry_length;
3497 
3498 	rc = is_bnxt_in_error(bp);
3499 	if (rc)
3500 		return rc;
3501 
3502 	PMD_DRV_LOG(INFO, "%04x:%02x:%02x:%02x\n",
3503 		bp->pdev->addr.domain, bp->pdev->addr.bus,
3504 		bp->pdev->addr.devid, bp->pdev->addr.function);
3505 
3506 	rc = bnxt_hwrm_nvm_get_dir_info(bp, &dir_entries, &entry_length);
3507 	if (rc != 0)
3508 		return rc;
3509 
3510 	return dir_entries * entry_length;
3511 }
3512 
3513 static int
3514 bnxt_get_eeprom_op(struct rte_eth_dev *dev,
3515 		struct rte_dev_eeprom_info *in_eeprom)
3516 {
3517 	struct bnxt *bp = dev->data->dev_private;
3518 	uint32_t index;
3519 	uint32_t offset;
3520 	int rc;
3521 
3522 	rc = is_bnxt_in_error(bp);
3523 	if (rc)
3524 		return rc;
3525 
3526 	PMD_DRV_LOG(INFO, "%04x:%02x:%02x:%02x in_eeprom->offset = %d "
3527 		"len = %d\n", bp->pdev->addr.domain,
3528 		bp->pdev->addr.bus, bp->pdev->addr.devid,
3529 		bp->pdev->addr.function, in_eeprom->offset, in_eeprom->length);
3530 
3531 	if (in_eeprom->offset == 0) /* special offset value to get directory */
3532 		return bnxt_get_nvram_directory(bp, in_eeprom->length,
3533 						in_eeprom->data);
3534 
3535 	index = in_eeprom->offset >> 24;
3536 	offset = in_eeprom->offset & 0xffffff;
3537 
3538 	if (index != 0)
3539 		return bnxt_hwrm_get_nvram_item(bp, index - 1, offset,
3540 					   in_eeprom->length, in_eeprom->data);
3541 
3542 	return 0;
3543 }
3544 
3545 static bool bnxt_dir_type_is_ape_bin_format(uint16_t dir_type)
3546 {
3547 	switch (dir_type) {
3548 	case BNX_DIR_TYPE_CHIMP_PATCH:
3549 	case BNX_DIR_TYPE_BOOTCODE:
3550 	case BNX_DIR_TYPE_BOOTCODE_2:
3551 	case BNX_DIR_TYPE_APE_FW:
3552 	case BNX_DIR_TYPE_APE_PATCH:
3553 	case BNX_DIR_TYPE_KONG_FW:
3554 	case BNX_DIR_TYPE_KONG_PATCH:
3555 	case BNX_DIR_TYPE_BONO_FW:
3556 	case BNX_DIR_TYPE_BONO_PATCH:
3557 		/* FALLTHROUGH */
3558 		return true;
3559 	}
3560 
3561 	return false;
3562 }
3563 
3564 static bool bnxt_dir_type_is_other_exec_format(uint16_t dir_type)
3565 {
3566 	switch (dir_type) {
3567 	case BNX_DIR_TYPE_AVS:
3568 	case BNX_DIR_TYPE_EXP_ROM_MBA:
3569 	case BNX_DIR_TYPE_PCIE:
3570 	case BNX_DIR_TYPE_TSCF_UCODE:
3571 	case BNX_DIR_TYPE_EXT_PHY:
3572 	case BNX_DIR_TYPE_CCM:
3573 	case BNX_DIR_TYPE_ISCSI_BOOT:
3574 	case BNX_DIR_TYPE_ISCSI_BOOT_IPV6:
3575 	case BNX_DIR_TYPE_ISCSI_BOOT_IPV4N6:
3576 		/* FALLTHROUGH */
3577 		return true;
3578 	}
3579 
3580 	return false;
3581 }
3582 
3583 static bool bnxt_dir_type_is_executable(uint16_t dir_type)
3584 {
3585 	return bnxt_dir_type_is_ape_bin_format(dir_type) ||
3586 		bnxt_dir_type_is_other_exec_format(dir_type);
3587 }
3588 
3589 static int
3590 bnxt_set_eeprom_op(struct rte_eth_dev *dev,
3591 		struct rte_dev_eeprom_info *in_eeprom)
3592 {
3593 	struct bnxt *bp = dev->data->dev_private;
3594 	uint8_t index, dir_op;
3595 	uint16_t type, ext, ordinal, attr;
3596 	int rc;
3597 
3598 	rc = is_bnxt_in_error(bp);
3599 	if (rc)
3600 		return rc;
3601 
3602 	PMD_DRV_LOG(INFO, "%04x:%02x:%02x:%02x in_eeprom->offset = %d "
3603 		"len = %d\n", bp->pdev->addr.domain,
3604 		bp->pdev->addr.bus, bp->pdev->addr.devid,
3605 		bp->pdev->addr.function, in_eeprom->offset, in_eeprom->length);
3606 
3607 	if (!BNXT_PF(bp)) {
3608 		PMD_DRV_LOG(ERR, "NVM write not supported from a VF\n");
3609 		return -EINVAL;
3610 	}
3611 
3612 	type = in_eeprom->magic >> 16;
3613 
3614 	if (type == 0xffff) { /* special value for directory operations */
3615 		index = in_eeprom->magic & 0xff;
3616 		dir_op = in_eeprom->magic >> 8;
3617 		if (index == 0)
3618 			return -EINVAL;
3619 		switch (dir_op) {
3620 		case 0x0e: /* erase */
3621 			if (in_eeprom->offset != ~in_eeprom->magic)
3622 				return -EINVAL;
3623 			return bnxt_hwrm_erase_nvram_directory(bp, index - 1);
3624 		default:
3625 			return -EINVAL;
3626 		}
3627 	}
3628 
3629 	/* Create or re-write an NVM item: */
3630 	if (bnxt_dir_type_is_executable(type) == true)
3631 		return -EOPNOTSUPP;
3632 	ext = in_eeprom->magic & 0xffff;
3633 	ordinal = in_eeprom->offset >> 16;
3634 	attr = in_eeprom->offset & 0xffff;
3635 
3636 	return bnxt_hwrm_flash_nvram(bp, type, ordinal, ext, attr,
3637 				     in_eeprom->data, in_eeprom->length);
3638 }
3639 
3640 /*
3641  * Initialization
3642  */
3643 
3644 static const struct eth_dev_ops bnxt_dev_ops = {
3645 	.dev_infos_get = bnxt_dev_info_get_op,
3646 	.dev_close = bnxt_dev_close_op,
3647 	.dev_configure = bnxt_dev_configure_op,
3648 	.dev_start = bnxt_dev_start_op,
3649 	.dev_stop = bnxt_dev_stop_op,
3650 	.dev_set_link_up = bnxt_dev_set_link_up_op,
3651 	.dev_set_link_down = bnxt_dev_set_link_down_op,
3652 	.stats_get = bnxt_stats_get_op,
3653 	.stats_reset = bnxt_stats_reset_op,
3654 	.rx_queue_setup = bnxt_rx_queue_setup_op,
3655 	.rx_queue_release = bnxt_rx_queue_release_op,
3656 	.tx_queue_setup = bnxt_tx_queue_setup_op,
3657 	.tx_queue_release = bnxt_tx_queue_release_op,
3658 	.rx_queue_intr_enable = bnxt_rx_queue_intr_enable_op,
3659 	.rx_queue_intr_disable = bnxt_rx_queue_intr_disable_op,
3660 	.reta_update = bnxt_reta_update_op,
3661 	.reta_query = bnxt_reta_query_op,
3662 	.rss_hash_update = bnxt_rss_hash_update_op,
3663 	.rss_hash_conf_get = bnxt_rss_hash_conf_get_op,
3664 	.link_update = bnxt_link_update_op,
3665 	.promiscuous_enable = bnxt_promiscuous_enable_op,
3666 	.promiscuous_disable = bnxt_promiscuous_disable_op,
3667 	.allmulticast_enable = bnxt_allmulticast_enable_op,
3668 	.allmulticast_disable = bnxt_allmulticast_disable_op,
3669 	.mac_addr_add = bnxt_mac_addr_add_op,
3670 	.mac_addr_remove = bnxt_mac_addr_remove_op,
3671 	.flow_ctrl_get = bnxt_flow_ctrl_get_op,
3672 	.flow_ctrl_set = bnxt_flow_ctrl_set_op,
3673 	.udp_tunnel_port_add  = bnxt_udp_tunnel_port_add_op,
3674 	.udp_tunnel_port_del  = bnxt_udp_tunnel_port_del_op,
3675 	.vlan_filter_set = bnxt_vlan_filter_set_op,
3676 	.vlan_offload_set = bnxt_vlan_offload_set_op,
3677 	.vlan_tpid_set = bnxt_vlan_tpid_set_op,
3678 	.vlan_pvid_set = bnxt_vlan_pvid_set_op,
3679 	.mtu_set = bnxt_mtu_set_op,
3680 	.mac_addr_set = bnxt_set_default_mac_addr_op,
3681 	.xstats_get = bnxt_dev_xstats_get_op,
3682 	.xstats_get_names = bnxt_dev_xstats_get_names_op,
3683 	.xstats_reset = bnxt_dev_xstats_reset_op,
3684 	.fw_version_get = bnxt_fw_version_get,
3685 	.set_mc_addr_list = bnxt_dev_set_mc_addr_list_op,
3686 	.rxq_info_get = bnxt_rxq_info_get_op,
3687 	.txq_info_get = bnxt_txq_info_get_op,
3688 	.dev_led_on = bnxt_dev_led_on_op,
3689 	.dev_led_off = bnxt_dev_led_off_op,
3690 	.xstats_get_by_id = bnxt_dev_xstats_get_by_id_op,
3691 	.xstats_get_names_by_id = bnxt_dev_xstats_get_names_by_id_op,
3692 	.rx_queue_count = bnxt_rx_queue_count_op,
3693 	.rx_descriptor_status = bnxt_rx_descriptor_status_op,
3694 	.tx_descriptor_status = bnxt_tx_descriptor_status_op,
3695 	.rx_queue_start = bnxt_rx_queue_start,
3696 	.rx_queue_stop = bnxt_rx_queue_stop,
3697 	.tx_queue_start = bnxt_tx_queue_start,
3698 	.tx_queue_stop = bnxt_tx_queue_stop,
3699 	.filter_ctrl = bnxt_filter_ctrl_op,
3700 	.dev_supported_ptypes_get = bnxt_dev_supported_ptypes_get_op,
3701 	.get_eeprom_length    = bnxt_get_eeprom_length_op,
3702 	.get_eeprom           = bnxt_get_eeprom_op,
3703 	.set_eeprom           = bnxt_set_eeprom_op,
3704 	.timesync_enable      = bnxt_timesync_enable,
3705 	.timesync_disable     = bnxt_timesync_disable,
3706 	.timesync_read_time   = bnxt_timesync_read_time,
3707 	.timesync_write_time   = bnxt_timesync_write_time,
3708 	.timesync_adjust_time = bnxt_timesync_adjust_time,
3709 	.timesync_read_rx_timestamp = bnxt_timesync_read_rx_timestamp,
3710 	.timesync_read_tx_timestamp = bnxt_timesync_read_tx_timestamp,
3711 };
3712 
3713 static uint32_t bnxt_map_reset_regs(struct bnxt *bp, uint32_t reg)
3714 {
3715 	uint32_t offset;
3716 
3717 	/* Only pre-map the reset GRC registers using window 3 */
3718 	rte_write32(reg & 0xfffff000, (uint8_t *)bp->bar0 +
3719 		    BNXT_GRCPF_REG_WINDOW_BASE_OUT + 8);
3720 
3721 	offset = BNXT_GRCP_WINDOW_3_BASE + (reg & 0xffc);
3722 
3723 	return offset;
3724 }
3725 
3726 int bnxt_map_fw_health_status_regs(struct bnxt *bp)
3727 {
3728 	struct bnxt_error_recovery_info *info = bp->recovery_info;
3729 	uint32_t reg_base = 0xffffffff;
3730 	int i;
3731 
3732 	/* Only pre-map the monitoring GRC registers using window 2 */
3733 	for (i = 0; i < BNXT_FW_STATUS_REG_CNT; i++) {
3734 		uint32_t reg = info->status_regs[i];
3735 
3736 		if (BNXT_FW_STATUS_REG_TYPE(reg) != BNXT_FW_STATUS_REG_TYPE_GRC)
3737 			continue;
3738 
3739 		if (reg_base == 0xffffffff)
3740 			reg_base = reg & 0xfffff000;
3741 		if ((reg & 0xfffff000) != reg_base)
3742 			return -ERANGE;
3743 
3744 		/* Use mask 0xffc as the Lower 2 bits indicates
3745 		 * address space location
3746 		 */
3747 		info->mapped_status_regs[i] = BNXT_GRCP_WINDOW_2_BASE +
3748 						(reg & 0xffc);
3749 	}
3750 
3751 	if (reg_base == 0xffffffff)
3752 		return 0;
3753 
3754 	rte_write32(reg_base, (uint8_t *)bp->bar0 +
3755 		    BNXT_GRCPF_REG_WINDOW_BASE_OUT + 4);
3756 
3757 	return 0;
3758 }
3759 
3760 static void bnxt_write_fw_reset_reg(struct bnxt *bp, uint32_t index)
3761 {
3762 	struct bnxt_error_recovery_info *info = bp->recovery_info;
3763 	uint32_t delay = info->delay_after_reset[index];
3764 	uint32_t val = info->reset_reg_val[index];
3765 	uint32_t reg = info->reset_reg[index];
3766 	uint32_t type, offset;
3767 
3768 	type = BNXT_FW_STATUS_REG_TYPE(reg);
3769 	offset = BNXT_FW_STATUS_REG_OFF(reg);
3770 
3771 	switch (type) {
3772 	case BNXT_FW_STATUS_REG_TYPE_CFG:
3773 		rte_pci_write_config(bp->pdev, &val, sizeof(val), offset);
3774 		break;
3775 	case BNXT_FW_STATUS_REG_TYPE_GRC:
3776 		offset = bnxt_map_reset_regs(bp, offset);
3777 		rte_write32(val, (uint8_t *)bp->bar0 + offset);
3778 		break;
3779 	case BNXT_FW_STATUS_REG_TYPE_BAR0:
3780 		rte_write32(val, (uint8_t *)bp->bar0 + offset);
3781 		break;
3782 	}
3783 	/* wait on a specific interval of time until core reset is complete */
3784 	if (delay)
3785 		rte_delay_ms(delay);
3786 }
3787 
3788 static void bnxt_dev_cleanup(struct bnxt *bp)
3789 {
3790 	bnxt_set_hwrm_link_config(bp, false);
3791 	bp->link_info.link_up = 0;
3792 	if (bp->dev_stopped == 0)
3793 		bnxt_dev_stop_op(bp->eth_dev);
3794 
3795 	bnxt_uninit_resources(bp, true);
3796 }
3797 
3798 static int bnxt_restore_filters(struct bnxt *bp)
3799 {
3800 	struct rte_eth_dev *dev = bp->eth_dev;
3801 	int ret = 0;
3802 
3803 	if (dev->data->all_multicast)
3804 		ret = bnxt_allmulticast_enable_op(dev);
3805 	if (dev->data->promiscuous)
3806 		ret = bnxt_promiscuous_enable_op(dev);
3807 
3808 	/* TODO restore other filters as well */
3809 	return ret;
3810 }
3811 
3812 static void bnxt_dev_recover(void *arg)
3813 {
3814 	struct bnxt *bp = arg;
3815 	int timeout = bp->fw_reset_max_msecs;
3816 	int rc = 0;
3817 
3818 	/* Clear Error flag so that device re-init should happen */
3819 	bp->flags &= ~BNXT_FLAG_FATAL_ERROR;
3820 
3821 	do {
3822 		rc = bnxt_hwrm_ver_get(bp);
3823 		if (rc == 0)
3824 			break;
3825 		rte_delay_ms(BNXT_FW_READY_WAIT_INTERVAL);
3826 		timeout -= BNXT_FW_READY_WAIT_INTERVAL;
3827 	} while (rc && timeout);
3828 
3829 	if (rc) {
3830 		PMD_DRV_LOG(ERR, "FW is not Ready after reset\n");
3831 		goto err;
3832 	}
3833 
3834 	rc = bnxt_init_resources(bp, true);
3835 	if (rc) {
3836 		PMD_DRV_LOG(ERR,
3837 			    "Failed to initialize resources after reset\n");
3838 		goto err;
3839 	}
3840 	/* clear reset flag as the device is initialized now */
3841 	bp->flags &= ~BNXT_FLAG_FW_RESET;
3842 
3843 	rc = bnxt_dev_start_op(bp->eth_dev);
3844 	if (rc) {
3845 		PMD_DRV_LOG(ERR, "Failed to start port after reset\n");
3846 		goto err;
3847 	}
3848 
3849 	rc = bnxt_restore_filters(bp);
3850 	if (rc)
3851 		goto err;
3852 
3853 	PMD_DRV_LOG(INFO, "Recovered from FW reset\n");
3854 	return;
3855 err:
3856 	bp->flags |= BNXT_FLAG_FATAL_ERROR;
3857 	bnxt_uninit_resources(bp, false);
3858 	PMD_DRV_LOG(ERR, "Failed to recover from FW reset\n");
3859 }
3860 
3861 void bnxt_dev_reset_and_resume(void *arg)
3862 {
3863 	struct bnxt *bp = arg;
3864 	int rc;
3865 
3866 	bnxt_dev_cleanup(bp);
3867 
3868 	bnxt_wait_for_device_shutdown(bp);
3869 
3870 	rc = rte_eal_alarm_set(US_PER_MS * bp->fw_reset_min_msecs,
3871 			       bnxt_dev_recover, (void *)bp);
3872 	if (rc)
3873 		PMD_DRV_LOG(ERR, "Error setting recovery alarm");
3874 }
3875 
3876 uint32_t bnxt_read_fw_status_reg(struct bnxt *bp, uint32_t index)
3877 {
3878 	struct bnxt_error_recovery_info *info = bp->recovery_info;
3879 	uint32_t reg = info->status_regs[index];
3880 	uint32_t type, offset, val = 0;
3881 
3882 	type = BNXT_FW_STATUS_REG_TYPE(reg);
3883 	offset = BNXT_FW_STATUS_REG_OFF(reg);
3884 
3885 	switch (type) {
3886 	case BNXT_FW_STATUS_REG_TYPE_CFG:
3887 		rte_pci_read_config(bp->pdev, &val, sizeof(val), offset);
3888 		break;
3889 	case BNXT_FW_STATUS_REG_TYPE_GRC:
3890 		offset = info->mapped_status_regs[index];
3891 		/* FALLTHROUGH */
3892 	case BNXT_FW_STATUS_REG_TYPE_BAR0:
3893 		val = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3894 				       offset));
3895 		break;
3896 	}
3897 
3898 	return val;
3899 }
3900 
3901 static int bnxt_fw_reset_all(struct bnxt *bp)
3902 {
3903 	struct bnxt_error_recovery_info *info = bp->recovery_info;
3904 	uint32_t i;
3905 	int rc = 0;
3906 
3907 	if (info->flags & BNXT_FLAG_ERROR_RECOVERY_HOST) {
3908 		/* Reset through master function driver */
3909 		for (i = 0; i < info->reg_array_cnt; i++)
3910 			bnxt_write_fw_reset_reg(bp, i);
3911 		/* Wait for time specified by FW after triggering reset */
3912 		rte_delay_ms(info->master_func_wait_period_after_reset);
3913 	} else if (info->flags & BNXT_FLAG_ERROR_RECOVERY_CO_CPU) {
3914 		/* Reset with the help of Kong processor */
3915 		rc = bnxt_hwrm_fw_reset(bp);
3916 		if (rc)
3917 			PMD_DRV_LOG(ERR, "Failed to reset FW\n");
3918 	}
3919 
3920 	return rc;
3921 }
3922 
3923 static void bnxt_fw_reset_cb(void *arg)
3924 {
3925 	struct bnxt *bp = arg;
3926 	struct bnxt_error_recovery_info *info = bp->recovery_info;
3927 	int rc = 0;
3928 
3929 	/* Only Master function can do FW reset */
3930 	if (bnxt_is_master_func(bp) &&
3931 	    bnxt_is_recovery_enabled(bp)) {
3932 		rc = bnxt_fw_reset_all(bp);
3933 		if (rc) {
3934 			PMD_DRV_LOG(ERR, "Adapter recovery failed\n");
3935 			return;
3936 		}
3937 	}
3938 
3939 	/* if recovery method is ERROR_RECOVERY_CO_CPU, KONG will send
3940 	 * EXCEPTION_FATAL_ASYNC event to all the functions
3941 	 * (including MASTER FUNC). After receiving this Async, all the active
3942 	 * drivers should treat this case as FW initiated recovery
3943 	 */
3944 	if (info->flags & BNXT_FLAG_ERROR_RECOVERY_HOST) {
3945 		bp->fw_reset_min_msecs = BNXT_MIN_FW_READY_TIMEOUT;
3946 		bp->fw_reset_max_msecs = BNXT_MAX_FW_RESET_TIMEOUT;
3947 
3948 		/* To recover from error */
3949 		rte_eal_alarm_set(US_PER_MS, bnxt_dev_reset_and_resume,
3950 				  (void *)bp);
3951 	}
3952 }
3953 
3954 /* Driver should poll FW heartbeat, reset_counter with the frequency
3955  * advertised by FW in HWRM_ERROR_RECOVERY_QCFG.
3956  * When the driver detects heartbeat stop or change in reset_counter,
3957  * it has to trigger a reset to recover from the error condition.
3958  * A “master PF” is the function who will have the privilege to
3959  * initiate the chimp reset. The master PF will be elected by the
3960  * firmware and will be notified through async message.
3961  */
3962 static void bnxt_check_fw_health(void *arg)
3963 {
3964 	struct bnxt *bp = arg;
3965 	struct bnxt_error_recovery_info *info = bp->recovery_info;
3966 	uint32_t val = 0, wait_msec;
3967 
3968 	if (!info || !bnxt_is_recovery_enabled(bp) ||
3969 	    is_bnxt_in_error(bp))
3970 		return;
3971 
3972 	val = bnxt_read_fw_status_reg(bp, BNXT_FW_HEARTBEAT_CNT_REG);
3973 	if (val == info->last_heart_beat)
3974 		goto reset;
3975 
3976 	info->last_heart_beat = val;
3977 
3978 	val = bnxt_read_fw_status_reg(bp, BNXT_FW_RECOVERY_CNT_REG);
3979 	if (val != info->last_reset_counter)
3980 		goto reset;
3981 
3982 	info->last_reset_counter = val;
3983 
3984 	rte_eal_alarm_set(US_PER_MS * info->driver_polling_freq,
3985 			  bnxt_check_fw_health, (void *)bp);
3986 
3987 	return;
3988 reset:
3989 	/* Stop DMA to/from device */
3990 	bp->flags |= BNXT_FLAG_FATAL_ERROR;
3991 	bp->flags |= BNXT_FLAG_FW_RESET;
3992 
3993 	PMD_DRV_LOG(ERR, "Detected FW dead condition\n");
3994 
3995 	if (bnxt_is_master_func(bp))
3996 		wait_msec = info->master_func_wait_period;
3997 	else
3998 		wait_msec = info->normal_func_wait_period;
3999 
4000 	rte_eal_alarm_set(US_PER_MS * wait_msec,
4001 			  bnxt_fw_reset_cb, (void *)bp);
4002 }
4003 
4004 void bnxt_schedule_fw_health_check(struct bnxt *bp)
4005 {
4006 	uint32_t polling_freq;
4007 
4008 	if (!bnxt_is_recovery_enabled(bp))
4009 		return;
4010 
4011 	if (bp->flags & BNXT_FLAG_FW_HEALTH_CHECK_SCHEDULED)
4012 		return;
4013 
4014 	polling_freq = bp->recovery_info->driver_polling_freq;
4015 
4016 	rte_eal_alarm_set(US_PER_MS * polling_freq,
4017 			  bnxt_check_fw_health, (void *)bp);
4018 	bp->flags |= BNXT_FLAG_FW_HEALTH_CHECK_SCHEDULED;
4019 }
4020 
4021 static void bnxt_cancel_fw_health_check(struct bnxt *bp)
4022 {
4023 	if (!bnxt_is_recovery_enabled(bp))
4024 		return;
4025 
4026 	rte_eal_alarm_cancel(bnxt_check_fw_health, (void *)bp);
4027 	bp->flags &= ~BNXT_FLAG_FW_HEALTH_CHECK_SCHEDULED;
4028 }
4029 
4030 static bool bnxt_vf_pciid(uint16_t id)
4031 {
4032 	if (id == BROADCOM_DEV_ID_57304_VF ||
4033 	    id == BROADCOM_DEV_ID_57406_VF ||
4034 	    id == BROADCOM_DEV_ID_5731X_VF ||
4035 	    id == BROADCOM_DEV_ID_5741X_VF ||
4036 	    id == BROADCOM_DEV_ID_57414_VF ||
4037 	    id == BROADCOM_DEV_ID_STRATUS_NIC_VF1 ||
4038 	    id == BROADCOM_DEV_ID_STRATUS_NIC_VF2 ||
4039 	    id == BROADCOM_DEV_ID_58802_VF ||
4040 	    id == BROADCOM_DEV_ID_57500_VF1 ||
4041 	    id == BROADCOM_DEV_ID_57500_VF2)
4042 		return true;
4043 	return false;
4044 }
4045 
4046 static bool bnxt_thor_device(uint16_t id)
4047 {
4048 	if (id == BROADCOM_DEV_ID_57508 ||
4049 	    id == BROADCOM_DEV_ID_57504 ||
4050 	    id == BROADCOM_DEV_ID_57502 ||
4051 	    id == BROADCOM_DEV_ID_57508_MF1 ||
4052 	    id == BROADCOM_DEV_ID_57504_MF1 ||
4053 	    id == BROADCOM_DEV_ID_57502_MF1 ||
4054 	    id == BROADCOM_DEV_ID_57508_MF2 ||
4055 	    id == BROADCOM_DEV_ID_57504_MF2 ||
4056 	    id == BROADCOM_DEV_ID_57502_MF2 ||
4057 	    id == BROADCOM_DEV_ID_57500_VF1 ||
4058 	    id == BROADCOM_DEV_ID_57500_VF2)
4059 		return true;
4060 
4061 	return false;
4062 }
4063 
4064 bool bnxt_stratus_device(struct bnxt *bp)
4065 {
4066 	uint16_t id = bp->pdev->id.device_id;
4067 
4068 	if (id == BROADCOM_DEV_ID_STRATUS_NIC ||
4069 	    id == BROADCOM_DEV_ID_STRATUS_NIC_VF1 ||
4070 	    id == BROADCOM_DEV_ID_STRATUS_NIC_VF2)
4071 		return true;
4072 	return false;
4073 }
4074 
4075 static int bnxt_init_board(struct rte_eth_dev *eth_dev)
4076 {
4077 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
4078 	struct bnxt *bp = eth_dev->data->dev_private;
4079 
4080 	/* enable device (incl. PCI PM wakeup), and bus-mastering */
4081 	bp->bar0 = (void *)pci_dev->mem_resource[0].addr;
4082 	bp->doorbell_base = (void *)pci_dev->mem_resource[2].addr;
4083 	if (!bp->bar0 || !bp->doorbell_base) {
4084 		PMD_DRV_LOG(ERR, "Unable to access Hardware\n");
4085 		return -ENODEV;
4086 	}
4087 
4088 	bp->eth_dev = eth_dev;
4089 	bp->pdev = pci_dev;
4090 
4091 	return 0;
4092 }
4093 
4094 static int bnxt_alloc_ctx_mem_blk(__rte_unused struct bnxt *bp,
4095 				  struct bnxt_ctx_pg_info *ctx_pg,
4096 				  uint32_t mem_size,
4097 				  const char *suffix,
4098 				  uint16_t idx)
4099 {
4100 	struct bnxt_ring_mem_info *rmem = &ctx_pg->ring_mem;
4101 	const struct rte_memzone *mz = NULL;
4102 	char mz_name[RTE_MEMZONE_NAMESIZE];
4103 	rte_iova_t mz_phys_addr;
4104 	uint64_t valid_bits = 0;
4105 	uint32_t sz;
4106 	int i;
4107 
4108 	if (!mem_size)
4109 		return 0;
4110 
4111 	rmem->nr_pages = RTE_ALIGN_MUL_CEIL(mem_size, BNXT_PAGE_SIZE) /
4112 			 BNXT_PAGE_SIZE;
4113 	rmem->page_size = BNXT_PAGE_SIZE;
4114 	rmem->pg_arr = ctx_pg->ctx_pg_arr;
4115 	rmem->dma_arr = ctx_pg->ctx_dma_arr;
4116 	rmem->flags = BNXT_RMEM_VALID_PTE_FLAG;
4117 
4118 	valid_bits = PTU_PTE_VALID;
4119 
4120 	if (rmem->nr_pages > 1) {
4121 		snprintf(mz_name, RTE_MEMZONE_NAMESIZE,
4122 			 "bnxt_ctx_pg_tbl%s_%x_%d",
4123 			 suffix, idx, bp->eth_dev->data->port_id);
4124 		mz_name[RTE_MEMZONE_NAMESIZE - 1] = 0;
4125 		mz = rte_memzone_lookup(mz_name);
4126 		if (!mz) {
4127 			mz = rte_memzone_reserve_aligned(mz_name,
4128 						rmem->nr_pages * 8,
4129 						SOCKET_ID_ANY,
4130 						RTE_MEMZONE_2MB |
4131 						RTE_MEMZONE_SIZE_HINT_ONLY |
4132 						RTE_MEMZONE_IOVA_CONTIG,
4133 						BNXT_PAGE_SIZE);
4134 			if (mz == NULL)
4135 				return -ENOMEM;
4136 		}
4137 
4138 		memset(mz->addr, 0, mz->len);
4139 		mz_phys_addr = mz->iova;
4140 		if ((unsigned long)mz->addr == mz_phys_addr) {
4141 			PMD_DRV_LOG(DEBUG,
4142 				    "physical address same as virtual\n");
4143 			PMD_DRV_LOG(DEBUG, "Using rte_mem_virt2iova()\n");
4144 			mz_phys_addr = rte_mem_virt2iova(mz->addr);
4145 			if (mz_phys_addr == RTE_BAD_IOVA) {
4146 				PMD_DRV_LOG(ERR,
4147 					"unable to map addr to phys memory\n");
4148 				return -ENOMEM;
4149 			}
4150 		}
4151 		rte_mem_lock_page(((char *)mz->addr));
4152 
4153 		rmem->pg_tbl = mz->addr;
4154 		rmem->pg_tbl_map = mz_phys_addr;
4155 		rmem->pg_tbl_mz = mz;
4156 	}
4157 
4158 	snprintf(mz_name, RTE_MEMZONE_NAMESIZE, "bnxt_ctx_%s_%x_%d",
4159 		 suffix, idx, bp->eth_dev->data->port_id);
4160 	mz = rte_memzone_lookup(mz_name);
4161 	if (!mz) {
4162 		mz = rte_memzone_reserve_aligned(mz_name,
4163 						 mem_size,
4164 						 SOCKET_ID_ANY,
4165 						 RTE_MEMZONE_1GB |
4166 						 RTE_MEMZONE_SIZE_HINT_ONLY |
4167 						 RTE_MEMZONE_IOVA_CONTIG,
4168 						 BNXT_PAGE_SIZE);
4169 		if (mz == NULL)
4170 			return -ENOMEM;
4171 	}
4172 
4173 	memset(mz->addr, 0, mz->len);
4174 	mz_phys_addr = mz->iova;
4175 	if ((unsigned long)mz->addr == mz_phys_addr) {
4176 		PMD_DRV_LOG(DEBUG,
4177 			    "Memzone physical address same as virtual.\n");
4178 		PMD_DRV_LOG(DEBUG, "Using rte_mem_virt2iova()\n");
4179 		for (sz = 0; sz < mem_size; sz += BNXT_PAGE_SIZE)
4180 			rte_mem_lock_page(((char *)mz->addr) + sz);
4181 		mz_phys_addr = rte_mem_virt2iova(mz->addr);
4182 		if (mz_phys_addr == RTE_BAD_IOVA) {
4183 			PMD_DRV_LOG(ERR,
4184 				    "unable to map addr to phys memory\n");
4185 			return -ENOMEM;
4186 		}
4187 	}
4188 
4189 	for (sz = 0, i = 0; sz < mem_size; sz += BNXT_PAGE_SIZE, i++) {
4190 		rte_mem_lock_page(((char *)mz->addr) + sz);
4191 		rmem->pg_arr[i] = ((char *)mz->addr) + sz;
4192 		rmem->dma_arr[i] = mz_phys_addr + sz;
4193 
4194 		if (rmem->nr_pages > 1) {
4195 			if (i == rmem->nr_pages - 2 &&
4196 			    (rmem->flags & BNXT_RMEM_RING_PTE_FLAG))
4197 				valid_bits |= PTU_PTE_NEXT_TO_LAST;
4198 			else if (i == rmem->nr_pages - 1 &&
4199 				 (rmem->flags & BNXT_RMEM_RING_PTE_FLAG))
4200 				valid_bits |= PTU_PTE_LAST;
4201 
4202 			rmem->pg_tbl[i] = rte_cpu_to_le_64(rmem->dma_arr[i] |
4203 							   valid_bits);
4204 		}
4205 	}
4206 
4207 	rmem->mz = mz;
4208 	if (rmem->vmem_size)
4209 		rmem->vmem = (void **)mz->addr;
4210 	rmem->dma_arr[0] = mz_phys_addr;
4211 	return 0;
4212 }
4213 
4214 static void bnxt_free_ctx_mem(struct bnxt *bp)
4215 {
4216 	int i;
4217 
4218 	if (!bp->ctx || !(bp->ctx->flags & BNXT_CTX_FLAG_INITED))
4219 		return;
4220 
4221 	bp->ctx->flags &= ~BNXT_CTX_FLAG_INITED;
4222 	rte_memzone_free(bp->ctx->qp_mem.ring_mem.mz);
4223 	rte_memzone_free(bp->ctx->srq_mem.ring_mem.mz);
4224 	rte_memzone_free(bp->ctx->cq_mem.ring_mem.mz);
4225 	rte_memzone_free(bp->ctx->vnic_mem.ring_mem.mz);
4226 	rte_memzone_free(bp->ctx->stat_mem.ring_mem.mz);
4227 	rte_memzone_free(bp->ctx->qp_mem.ring_mem.pg_tbl_mz);
4228 	rte_memzone_free(bp->ctx->srq_mem.ring_mem.pg_tbl_mz);
4229 	rte_memzone_free(bp->ctx->cq_mem.ring_mem.pg_tbl_mz);
4230 	rte_memzone_free(bp->ctx->vnic_mem.ring_mem.pg_tbl_mz);
4231 	rte_memzone_free(bp->ctx->stat_mem.ring_mem.pg_tbl_mz);
4232 
4233 	for (i = 0; i < BNXT_MAX_Q; i++) {
4234 		if (bp->ctx->tqm_mem[i])
4235 			rte_memzone_free(bp->ctx->tqm_mem[i]->ring_mem.mz);
4236 	}
4237 
4238 	rte_free(bp->ctx);
4239 	bp->ctx = NULL;
4240 }
4241 
4242 #define bnxt_roundup(x, y)   ((((x) + ((y) - 1)) / (y)) * (y))
4243 
4244 #define min_t(type, x, y) ({                    \
4245 	type __min1 = (x);                      \
4246 	type __min2 = (y);                      \
4247 	__min1 < __min2 ? __min1 : __min2; })
4248 
4249 #define max_t(type, x, y) ({                    \
4250 	type __max1 = (x);                      \
4251 	type __max2 = (y);                      \
4252 	__max1 > __max2 ? __max1 : __max2; })
4253 
4254 #define clamp_t(type, _x, min, max)     min_t(type, max_t(type, _x, min), max)
4255 
4256 int bnxt_alloc_ctx_mem(struct bnxt *bp)
4257 {
4258 	struct bnxt_ctx_pg_info *ctx_pg;
4259 	struct bnxt_ctx_mem_info *ctx;
4260 	uint32_t mem_size, ena, entries;
4261 	int i, rc;
4262 
4263 	rc = bnxt_hwrm_func_backing_store_qcaps(bp);
4264 	if (rc) {
4265 		PMD_DRV_LOG(ERR, "Query context mem capability failed\n");
4266 		return rc;
4267 	}
4268 	ctx = bp->ctx;
4269 	if (!ctx || (ctx->flags & BNXT_CTX_FLAG_INITED))
4270 		return 0;
4271 
4272 	ctx_pg = &ctx->qp_mem;
4273 	ctx_pg->entries = ctx->qp_min_qp1_entries + ctx->qp_max_l2_entries;
4274 	mem_size = ctx->qp_entry_size * ctx_pg->entries;
4275 	rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "qp_mem", 0);
4276 	if (rc)
4277 		return rc;
4278 
4279 	ctx_pg = &ctx->srq_mem;
4280 	ctx_pg->entries = ctx->srq_max_l2_entries;
4281 	mem_size = ctx->srq_entry_size * ctx_pg->entries;
4282 	rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "srq_mem", 0);
4283 	if (rc)
4284 		return rc;
4285 
4286 	ctx_pg = &ctx->cq_mem;
4287 	ctx_pg->entries = ctx->cq_max_l2_entries;
4288 	mem_size = ctx->cq_entry_size * ctx_pg->entries;
4289 	rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "cq_mem", 0);
4290 	if (rc)
4291 		return rc;
4292 
4293 	ctx_pg = &ctx->vnic_mem;
4294 	ctx_pg->entries = ctx->vnic_max_vnic_entries +
4295 		ctx->vnic_max_ring_table_entries;
4296 	mem_size = ctx->vnic_entry_size * ctx_pg->entries;
4297 	rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "vnic_mem", 0);
4298 	if (rc)
4299 		return rc;
4300 
4301 	ctx_pg = &ctx->stat_mem;
4302 	ctx_pg->entries = ctx->stat_max_entries;
4303 	mem_size = ctx->stat_entry_size * ctx_pg->entries;
4304 	rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "stat_mem", 0);
4305 	if (rc)
4306 		return rc;
4307 
4308 	entries = ctx->qp_max_l2_entries +
4309 		  ctx->vnic_max_vnic_entries +
4310 		  ctx->tqm_min_entries_per_ring;
4311 	entries = bnxt_roundup(entries, ctx->tqm_entries_multiple);
4312 	entries = clamp_t(uint32_t, entries, ctx->tqm_min_entries_per_ring,
4313 			  ctx->tqm_max_entries_per_ring);
4314 	for (i = 0, ena = 0; i < BNXT_MAX_Q; i++) {
4315 		ctx_pg = ctx->tqm_mem[i];
4316 		/* use min tqm entries for now. */
4317 		ctx_pg->entries = entries;
4318 		mem_size = ctx->tqm_entry_size * ctx_pg->entries;
4319 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "tqm_mem", i);
4320 		if (rc)
4321 			return rc;
4322 		ena |= HWRM_FUNC_BACKING_STORE_CFG_INPUT_ENABLES_TQM_SP << i;
4323 	}
4324 
4325 	ena |= FUNC_BACKING_STORE_CFG_INPUT_DFLT_ENABLES;
4326 	rc = bnxt_hwrm_func_backing_store_cfg(bp, ena);
4327 	if (rc)
4328 		PMD_DRV_LOG(ERR,
4329 			    "Failed to configure context mem: rc = %d\n", rc);
4330 	else
4331 		ctx->flags |= BNXT_CTX_FLAG_INITED;
4332 
4333 	return rc;
4334 }
4335 
4336 static int bnxt_alloc_stats_mem(struct bnxt *bp)
4337 {
4338 	struct rte_pci_device *pci_dev = bp->pdev;
4339 	char mz_name[RTE_MEMZONE_NAMESIZE];
4340 	const struct rte_memzone *mz = NULL;
4341 	uint32_t total_alloc_len;
4342 	rte_iova_t mz_phys_addr;
4343 
4344 	if (pci_dev->id.device_id == BROADCOM_DEV_ID_NS2)
4345 		return 0;
4346 
4347 	snprintf(mz_name, RTE_MEMZONE_NAMESIZE,
4348 		 "bnxt_" PCI_PRI_FMT "-%s", pci_dev->addr.domain,
4349 		 pci_dev->addr.bus, pci_dev->addr.devid,
4350 		 pci_dev->addr.function, "rx_port_stats");
4351 	mz_name[RTE_MEMZONE_NAMESIZE - 1] = 0;
4352 	mz = rte_memzone_lookup(mz_name);
4353 	total_alloc_len =
4354 		RTE_CACHE_LINE_ROUNDUP(sizeof(struct rx_port_stats) +
4355 				       sizeof(struct rx_port_stats_ext) + 512);
4356 	if (!mz) {
4357 		mz = rte_memzone_reserve(mz_name, total_alloc_len,
4358 					 SOCKET_ID_ANY,
4359 					 RTE_MEMZONE_2MB |
4360 					 RTE_MEMZONE_SIZE_HINT_ONLY |
4361 					 RTE_MEMZONE_IOVA_CONTIG);
4362 		if (mz == NULL)
4363 			return -ENOMEM;
4364 	}
4365 	memset(mz->addr, 0, mz->len);
4366 	mz_phys_addr = mz->iova;
4367 	if ((unsigned long)mz->addr == mz_phys_addr) {
4368 		PMD_DRV_LOG(DEBUG,
4369 			    "Memzone physical address same as virtual.\n");
4370 		PMD_DRV_LOG(DEBUG,
4371 			    "Using rte_mem_virt2iova()\n");
4372 		mz_phys_addr = rte_mem_virt2iova(mz->addr);
4373 		if (mz_phys_addr == RTE_BAD_IOVA) {
4374 			PMD_DRV_LOG(ERR,
4375 				    "Can't map address to physical memory\n");
4376 			return -ENOMEM;
4377 		}
4378 	}
4379 
4380 	bp->rx_mem_zone = (const void *)mz;
4381 	bp->hw_rx_port_stats = mz->addr;
4382 	bp->hw_rx_port_stats_map = mz_phys_addr;
4383 
4384 	snprintf(mz_name, RTE_MEMZONE_NAMESIZE,
4385 		 "bnxt_" PCI_PRI_FMT "-%s", pci_dev->addr.domain,
4386 		 pci_dev->addr.bus, pci_dev->addr.devid,
4387 		 pci_dev->addr.function, "tx_port_stats");
4388 	mz_name[RTE_MEMZONE_NAMESIZE - 1] = 0;
4389 	mz = rte_memzone_lookup(mz_name);
4390 	total_alloc_len =
4391 		RTE_CACHE_LINE_ROUNDUP(sizeof(struct tx_port_stats) +
4392 				       sizeof(struct tx_port_stats_ext) + 512);
4393 	if (!mz) {
4394 		mz = rte_memzone_reserve(mz_name,
4395 					 total_alloc_len,
4396 					 SOCKET_ID_ANY,
4397 					 RTE_MEMZONE_2MB |
4398 					 RTE_MEMZONE_SIZE_HINT_ONLY |
4399 					 RTE_MEMZONE_IOVA_CONTIG);
4400 		if (mz == NULL)
4401 			return -ENOMEM;
4402 	}
4403 	memset(mz->addr, 0, mz->len);
4404 	mz_phys_addr = mz->iova;
4405 	if ((unsigned long)mz->addr == mz_phys_addr) {
4406 		PMD_DRV_LOG(DEBUG,
4407 			    "Memzone physical address same as virtual\n");
4408 		PMD_DRV_LOG(DEBUG, "Using rte_mem_virt2iova()\n");
4409 		mz_phys_addr = rte_mem_virt2iova(mz->addr);
4410 		if (mz_phys_addr == RTE_BAD_IOVA) {
4411 			PMD_DRV_LOG(ERR,
4412 				    "Can't map address to physical memory\n");
4413 			return -ENOMEM;
4414 		}
4415 	}
4416 
4417 	bp->tx_mem_zone = (const void *)mz;
4418 	bp->hw_tx_port_stats = mz->addr;
4419 	bp->hw_tx_port_stats_map = mz_phys_addr;
4420 	bp->flags |= BNXT_FLAG_PORT_STATS;
4421 
4422 	/* Display extended statistics if FW supports it */
4423 	if (bp->hwrm_spec_code < HWRM_SPEC_CODE_1_8_4 ||
4424 	    bp->hwrm_spec_code == HWRM_SPEC_CODE_1_9_0 ||
4425 	    !(bp->flags & BNXT_FLAG_EXT_STATS_SUPPORTED))
4426 		return 0;
4427 
4428 	bp->hw_rx_port_stats_ext = (void *)
4429 		((uint8_t *)bp->hw_rx_port_stats +
4430 		 sizeof(struct rx_port_stats));
4431 	bp->hw_rx_port_stats_ext_map = bp->hw_rx_port_stats_map +
4432 		sizeof(struct rx_port_stats);
4433 	bp->flags |= BNXT_FLAG_EXT_RX_PORT_STATS;
4434 
4435 	if (bp->hwrm_spec_code < HWRM_SPEC_CODE_1_9_2 ||
4436 	    bp->flags & BNXT_FLAG_EXT_STATS_SUPPORTED) {
4437 		bp->hw_tx_port_stats_ext = (void *)
4438 			((uint8_t *)bp->hw_tx_port_stats +
4439 			 sizeof(struct tx_port_stats));
4440 		bp->hw_tx_port_stats_ext_map =
4441 			bp->hw_tx_port_stats_map +
4442 			sizeof(struct tx_port_stats);
4443 		bp->flags |= BNXT_FLAG_EXT_TX_PORT_STATS;
4444 	}
4445 
4446 	return 0;
4447 }
4448 
4449 static int bnxt_setup_mac_addr(struct rte_eth_dev *eth_dev)
4450 {
4451 	struct bnxt *bp = eth_dev->data->dev_private;
4452 	int rc = 0;
4453 
4454 	eth_dev->data->mac_addrs = rte_zmalloc("bnxt_mac_addr_tbl",
4455 					       RTE_ETHER_ADDR_LEN *
4456 					       bp->max_l2_ctx,
4457 					       0);
4458 	if (eth_dev->data->mac_addrs == NULL) {
4459 		PMD_DRV_LOG(ERR, "Failed to alloc MAC addr tbl\n");
4460 		return -ENOMEM;
4461 	}
4462 
4463 	if (bnxt_check_zero_bytes(bp->dflt_mac_addr, RTE_ETHER_ADDR_LEN)) {
4464 		if (BNXT_PF(bp))
4465 			return -EINVAL;
4466 
4467 		/* Generate a random MAC address, if none was assigned by PF */
4468 		PMD_DRV_LOG(INFO, "VF MAC address not assigned by Host PF\n");
4469 		bnxt_eth_hw_addr_random(bp->mac_addr);
4470 		PMD_DRV_LOG(INFO,
4471 			    "Assign random MAC:%02X:%02X:%02X:%02X:%02X:%02X\n",
4472 			    bp->mac_addr[0], bp->mac_addr[1], bp->mac_addr[2],
4473 			    bp->mac_addr[3], bp->mac_addr[4], bp->mac_addr[5]);
4474 
4475 		rc = bnxt_hwrm_set_mac(bp);
4476 		if (!rc)
4477 			memcpy(&bp->eth_dev->data->mac_addrs[0], bp->mac_addr,
4478 			       RTE_ETHER_ADDR_LEN);
4479 		return rc;
4480 	}
4481 
4482 	/* Copy the permanent MAC from the FUNC_QCAPS response */
4483 	memcpy(bp->mac_addr, bp->dflt_mac_addr, RTE_ETHER_ADDR_LEN);
4484 	memcpy(&eth_dev->data->mac_addrs[0], bp->mac_addr, RTE_ETHER_ADDR_LEN);
4485 
4486 	return rc;
4487 }
4488 
4489 static int bnxt_restore_dflt_mac(struct bnxt *bp)
4490 {
4491 	int rc = 0;
4492 
4493 	/* MAC is already configured in FW */
4494 	if (!bnxt_check_zero_bytes(bp->dflt_mac_addr, RTE_ETHER_ADDR_LEN))
4495 		return 0;
4496 
4497 	/* Restore the old MAC configured */
4498 	rc = bnxt_hwrm_set_mac(bp);
4499 	if (rc)
4500 		PMD_DRV_LOG(ERR, "Failed to restore MAC address\n");
4501 
4502 	return rc;
4503 }
4504 
4505 static void bnxt_config_vf_req_fwd(struct bnxt *bp)
4506 {
4507 	if (!BNXT_PF(bp))
4508 		return;
4509 
4510 #define ALLOW_FUNC(x)	\
4511 	{ \
4512 		uint32_t arg = (x); \
4513 		bp->pf.vf_req_fwd[((arg) >> 5)] &= \
4514 		~rte_cpu_to_le_32(1 << ((arg) & 0x1f)); \
4515 	}
4516 
4517 	/* Forward all requests if firmware is new enough */
4518 	if (((bp->fw_ver >= ((20 << 24) | (6 << 16) | (100 << 8))) &&
4519 	     (bp->fw_ver < ((20 << 24) | (7 << 16)))) ||
4520 	    ((bp->fw_ver >= ((20 << 24) | (8 << 16))))) {
4521 		memset(bp->pf.vf_req_fwd, 0xff, sizeof(bp->pf.vf_req_fwd));
4522 	} else {
4523 		PMD_DRV_LOG(WARNING,
4524 			    "Firmware too old for VF mailbox functionality\n");
4525 		memset(bp->pf.vf_req_fwd, 0, sizeof(bp->pf.vf_req_fwd));
4526 	}
4527 
4528 	/*
4529 	 * The following are used for driver cleanup. If we disallow these,
4530 	 * VF drivers can't clean up cleanly.
4531 	 */
4532 	ALLOW_FUNC(HWRM_FUNC_DRV_UNRGTR);
4533 	ALLOW_FUNC(HWRM_VNIC_FREE);
4534 	ALLOW_FUNC(HWRM_RING_FREE);
4535 	ALLOW_FUNC(HWRM_RING_GRP_FREE);
4536 	ALLOW_FUNC(HWRM_VNIC_RSS_COS_LB_CTX_FREE);
4537 	ALLOW_FUNC(HWRM_CFA_L2_FILTER_FREE);
4538 	ALLOW_FUNC(HWRM_STAT_CTX_FREE);
4539 	ALLOW_FUNC(HWRM_PORT_PHY_QCFG);
4540 	ALLOW_FUNC(HWRM_VNIC_TPA_CFG);
4541 }
4542 
4543 static int bnxt_init_fw(struct bnxt *bp)
4544 {
4545 	uint16_t mtu;
4546 	int rc = 0;
4547 
4548 	rc = bnxt_hwrm_ver_get(bp);
4549 	if (rc)
4550 		return rc;
4551 
4552 	rc = bnxt_hwrm_func_reset(bp);
4553 	if (rc)
4554 		return -EIO;
4555 
4556 	rc = bnxt_hwrm_vnic_qcaps(bp);
4557 	if (rc)
4558 		return rc;
4559 
4560 	rc = bnxt_hwrm_queue_qportcfg(bp);
4561 	if (rc)
4562 		return rc;
4563 
4564 	/* Get the MAX capabilities for this function.
4565 	 * This function also allocates context memory for TQM rings and
4566 	 * informs the firmware about this allocated backing store memory.
4567 	 */
4568 	rc = bnxt_hwrm_func_qcaps(bp);
4569 	if (rc)
4570 		return rc;
4571 
4572 	rc = bnxt_hwrm_func_qcfg(bp, &mtu);
4573 	if (rc)
4574 		return rc;
4575 
4576 	rc = bnxt_hwrm_cfa_adv_flow_mgmt_qcaps(bp);
4577 	if (rc)
4578 		return rc;
4579 
4580 	/* Get the adapter error recovery support info */
4581 	rc = bnxt_hwrm_error_recovery_qcfg(bp);
4582 	if (rc)
4583 		bp->flags &= ~BNXT_FLAG_FW_CAP_ERROR_RECOVERY;
4584 
4585 	bnxt_hwrm_port_led_qcaps(bp);
4586 
4587 	return 0;
4588 }
4589 
4590 static int
4591 bnxt_init_locks(struct bnxt *bp)
4592 {
4593 	int err;
4594 
4595 	err = pthread_mutex_init(&bp->flow_lock, NULL);
4596 	if (err) {
4597 		PMD_DRV_LOG(ERR, "Unable to initialize flow_lock\n");
4598 		return err;
4599 	}
4600 
4601 	err = pthread_mutex_init(&bp->def_cp_lock, NULL);
4602 	if (err)
4603 		PMD_DRV_LOG(ERR, "Unable to initialize def_cp_lock\n");
4604 	return err;
4605 }
4606 
4607 static int bnxt_init_resources(struct bnxt *bp, bool reconfig_dev)
4608 {
4609 	int rc;
4610 
4611 	rc = bnxt_init_fw(bp);
4612 	if (rc)
4613 		return rc;
4614 
4615 	if (!reconfig_dev) {
4616 		rc = bnxt_setup_mac_addr(bp->eth_dev);
4617 		if (rc)
4618 			return rc;
4619 	} else {
4620 		rc = bnxt_restore_dflt_mac(bp);
4621 		if (rc)
4622 			return rc;
4623 	}
4624 
4625 	bnxt_config_vf_req_fwd(bp);
4626 
4627 	rc = bnxt_hwrm_func_driver_register(bp);
4628 	if (rc) {
4629 		PMD_DRV_LOG(ERR, "Failed to register driver");
4630 		return -EBUSY;
4631 	}
4632 
4633 	if (BNXT_PF(bp)) {
4634 		if (bp->pdev->max_vfs) {
4635 			rc = bnxt_hwrm_allocate_vfs(bp, bp->pdev->max_vfs);
4636 			if (rc) {
4637 				PMD_DRV_LOG(ERR, "Failed to allocate VFs\n");
4638 				return rc;
4639 			}
4640 		} else {
4641 			rc = bnxt_hwrm_allocate_pf_only(bp);
4642 			if (rc) {
4643 				PMD_DRV_LOG(ERR,
4644 					    "Failed to allocate PF resources");
4645 				return rc;
4646 			}
4647 		}
4648 	}
4649 
4650 	rc = bnxt_alloc_mem(bp, reconfig_dev);
4651 	if (rc)
4652 		return rc;
4653 
4654 	rc = bnxt_setup_int(bp);
4655 	if (rc)
4656 		return rc;
4657 
4658 	rc = bnxt_request_int(bp);
4659 	if (rc)
4660 		return rc;
4661 
4662 	rc = bnxt_init_locks(bp);
4663 	if (rc)
4664 		return rc;
4665 
4666 	return 0;
4667 }
4668 
4669 static int
4670 bnxt_dev_init(struct rte_eth_dev *eth_dev)
4671 {
4672 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
4673 	static int version_printed;
4674 	struct bnxt *bp;
4675 	int rc;
4676 
4677 	if (version_printed++ == 0)
4678 		PMD_DRV_LOG(INFO, "%s\n", bnxt_version);
4679 
4680 	eth_dev->dev_ops = &bnxt_dev_ops;
4681 	eth_dev->rx_pkt_burst = &bnxt_recv_pkts;
4682 	eth_dev->tx_pkt_burst = &bnxt_xmit_pkts;
4683 
4684 	/*
4685 	 * For secondary processes, we don't initialise any further
4686 	 * as primary has already done this work.
4687 	 */
4688 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
4689 		return 0;
4690 
4691 	rte_eth_copy_pci_info(eth_dev, pci_dev);
4692 
4693 	bp = eth_dev->data->dev_private;
4694 
4695 	bp->dev_stopped = 1;
4696 
4697 	if (bnxt_vf_pciid(pci_dev->id.device_id))
4698 		bp->flags |= BNXT_FLAG_VF;
4699 
4700 	if (bnxt_thor_device(pci_dev->id.device_id))
4701 		bp->flags |= BNXT_FLAG_THOR_CHIP;
4702 
4703 	if (pci_dev->id.device_id == BROADCOM_DEV_ID_58802 ||
4704 	    pci_dev->id.device_id == BROADCOM_DEV_ID_58804 ||
4705 	    pci_dev->id.device_id == BROADCOM_DEV_ID_58808 ||
4706 	    pci_dev->id.device_id == BROADCOM_DEV_ID_58802_VF)
4707 		bp->flags |= BNXT_FLAG_STINGRAY;
4708 
4709 	rc = bnxt_init_board(eth_dev);
4710 	if (rc) {
4711 		PMD_DRV_LOG(ERR,
4712 			    "Failed to initialize board rc: %x\n", rc);
4713 		return rc;
4714 	}
4715 
4716 	rc = bnxt_alloc_hwrm_resources(bp);
4717 	if (rc) {
4718 		PMD_DRV_LOG(ERR,
4719 			    "Failed to allocate hwrm resource rc: %x\n", rc);
4720 		goto error_free;
4721 	}
4722 	rc = bnxt_init_resources(bp, false);
4723 	if (rc)
4724 		goto error_free;
4725 
4726 	rc = bnxt_alloc_stats_mem(bp);
4727 	if (rc)
4728 		goto error_free;
4729 
4730 	PMD_DRV_LOG(INFO,
4731 		    DRV_MODULE_NAME "found at mem %" PRIX64 ", node addr %pM\n",
4732 		    pci_dev->mem_resource[0].phys_addr,
4733 		    pci_dev->mem_resource[0].addr);
4734 
4735 	return 0;
4736 
4737 error_free:
4738 	bnxt_dev_uninit(eth_dev);
4739 	return rc;
4740 }
4741 
4742 static void
4743 bnxt_uninit_locks(struct bnxt *bp)
4744 {
4745 	pthread_mutex_destroy(&bp->flow_lock);
4746 	pthread_mutex_destroy(&bp->def_cp_lock);
4747 }
4748 
4749 static int
4750 bnxt_uninit_resources(struct bnxt *bp, bool reconfig_dev)
4751 {
4752 	int rc;
4753 
4754 	bnxt_free_int(bp);
4755 	bnxt_free_mem(bp, reconfig_dev);
4756 	bnxt_hwrm_func_buf_unrgtr(bp);
4757 	rc = bnxt_hwrm_func_driver_unregister(bp, 0);
4758 	bp->flags &= ~BNXT_FLAG_REGISTERED;
4759 	bnxt_free_ctx_mem(bp);
4760 	if (!reconfig_dev) {
4761 		bnxt_free_hwrm_resources(bp);
4762 
4763 		if (bp->recovery_info != NULL) {
4764 			rte_free(bp->recovery_info);
4765 			bp->recovery_info = NULL;
4766 		}
4767 	}
4768 
4769 	bnxt_uninit_locks(bp);
4770 	rte_free(bp->ptp_cfg);
4771 	bp->ptp_cfg = NULL;
4772 	return rc;
4773 }
4774 
4775 static int
4776 bnxt_dev_uninit(struct rte_eth_dev *eth_dev)
4777 {
4778 	struct bnxt *bp = eth_dev->data->dev_private;
4779 	int rc;
4780 
4781 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
4782 		return -EPERM;
4783 
4784 	PMD_DRV_LOG(DEBUG, "Calling Device uninit\n");
4785 
4786 	rc = bnxt_uninit_resources(bp, false);
4787 
4788 	if (bp->tx_mem_zone) {
4789 		rte_memzone_free((const struct rte_memzone *)bp->tx_mem_zone);
4790 		bp->tx_mem_zone = NULL;
4791 	}
4792 
4793 	if (bp->rx_mem_zone) {
4794 		rte_memzone_free((const struct rte_memzone *)bp->rx_mem_zone);
4795 		bp->rx_mem_zone = NULL;
4796 	}
4797 
4798 	if (bp->dev_stopped == 0)
4799 		bnxt_dev_close_op(eth_dev);
4800 	if (bp->pf.vf_info)
4801 		rte_free(bp->pf.vf_info);
4802 	eth_dev->dev_ops = NULL;
4803 	eth_dev->rx_pkt_burst = NULL;
4804 	eth_dev->tx_pkt_burst = NULL;
4805 
4806 	return rc;
4807 }
4808 
4809 static int bnxt_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
4810 	struct rte_pci_device *pci_dev)
4811 {
4812 	return rte_eth_dev_pci_generic_probe(pci_dev, sizeof(struct bnxt),
4813 		bnxt_dev_init);
4814 }
4815 
4816 static int bnxt_pci_remove(struct rte_pci_device *pci_dev)
4817 {
4818 	if (rte_eal_process_type() == RTE_PROC_PRIMARY)
4819 		return rte_eth_dev_pci_generic_remove(pci_dev,
4820 				bnxt_dev_uninit);
4821 	else
4822 		return rte_eth_dev_pci_generic_remove(pci_dev, NULL);
4823 }
4824 
4825 static struct rte_pci_driver bnxt_rte_pmd = {
4826 	.id_table = bnxt_pci_id_map,
4827 	.drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC,
4828 	.probe = bnxt_pci_probe,
4829 	.remove = bnxt_pci_remove,
4830 };
4831 
4832 static bool
4833 is_device_supported(struct rte_eth_dev *dev, struct rte_pci_driver *drv)
4834 {
4835 	if (strcmp(dev->device->driver->name, drv->driver.name))
4836 		return false;
4837 
4838 	return true;
4839 }
4840 
4841 bool is_bnxt_supported(struct rte_eth_dev *dev)
4842 {
4843 	return is_device_supported(dev, &bnxt_rte_pmd);
4844 }
4845 
4846 RTE_INIT(bnxt_init_log)
4847 {
4848 	bnxt_logtype_driver = rte_log_register("pmd.net.bnxt.driver");
4849 	if (bnxt_logtype_driver >= 0)
4850 		rte_log_set_level(bnxt_logtype_driver, RTE_LOG_NOTICE);
4851 }
4852 
4853 RTE_PMD_REGISTER_PCI(net_bnxt, bnxt_rte_pmd);
4854 RTE_PMD_REGISTER_PCI_TABLE(net_bnxt, bnxt_pci_id_map);
4855 RTE_PMD_REGISTER_KMOD_DEP(net_bnxt, "* igb_uio | uio_pci_generic | vfio-pci");
4856