xref: /dpdk/drivers/net/bnxt/bnxt_ethdev.c (revision daa02b5cddbb8e11b31d41e2bf7bb1ae64dcae2f)
1 /* SPDX-License-Identifier: BSD-3-Clause
2  * Copyright(c) 2014-2021 Broadcom
3  * All rights reserved.
4  */
5 
6 #include <inttypes.h>
7 #include <stdbool.h>
8 
9 #include <rte_dev.h>
10 #include <ethdev_driver.h>
11 #include <ethdev_pci.h>
12 #include <rte_malloc.h>
13 #include <rte_cycles.h>
14 #include <rte_alarm.h>
15 #include <rte_kvargs.h>
16 #include <rte_vect.h>
17 
18 #include "bnxt.h"
19 #include "bnxt_filter.h"
20 #include "bnxt_hwrm.h"
21 #include "bnxt_irq.h"
22 #include "bnxt_reps.h"
23 #include "bnxt_ring.h"
24 #include "bnxt_rxq.h"
25 #include "bnxt_rxr.h"
26 #include "bnxt_stats.h"
27 #include "bnxt_txq.h"
28 #include "bnxt_txr.h"
29 #include "bnxt_vnic.h"
30 #include "hsi_struct_def_dpdk.h"
31 #include "bnxt_nvm_defs.h"
32 #include "bnxt_tf_common.h"
33 #include "ulp_flow_db.h"
34 #include "rte_pmd_bnxt.h"
35 
36 #define DRV_MODULE_NAME		"bnxt"
37 static const char bnxt_version[] =
38 	"Broadcom NetXtreme driver " DRV_MODULE_NAME;
39 
40 /*
41  * The set of PCI devices this driver supports
42  */
43 static const struct rte_pci_id bnxt_pci_id_map[] = {
44 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM,
45 			 BROADCOM_DEV_ID_STRATUS_NIC_VF1) },
46 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM,
47 			 BROADCOM_DEV_ID_STRATUS_NIC_VF2) },
48 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_STRATUS_NIC) },
49 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57414_VF) },
50 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57304_VF) },
51 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_NS2) },
52 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57406_VF) },
53 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57407_MF) },
54 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_5741X_VF) },
55 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_5731X_VF) },
56 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57417_MF) },
57 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57412) },
58 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57414) },
59 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57416_RJ45) },
60 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57417_RJ45) },
61 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57412_MF) },
62 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57317_RJ45) },
63 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57417_SFP) },
64 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57416_SFP) },
65 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57317_SFP) },
66 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57414_MF) },
67 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57416_MF) },
68 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58802) },
69 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58804) },
70 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58808) },
71 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58802_VF) },
72 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57508) },
73 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57504) },
74 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57502) },
75 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57500_VF1) },
76 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57500_VF2) },
77 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57508_MF1) },
78 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57504_MF1) },
79 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57502_MF1) },
80 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57508_MF2) },
81 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57504_MF2) },
82 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57502_MF2) },
83 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58812) },
84 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58814) },
85 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58818) },
86 	{ RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58818_VF) },
87 	{ .vendor_id = 0, /* sentinel */ },
88 };
89 
90 #define	BNXT_DEVARG_ACCUM_STATS	"accum-stats"
91 #define BNXT_DEVARG_FLOW_XSTAT	"flow-xstat"
92 #define BNXT_DEVARG_MAX_NUM_KFLOWS  "max-num-kflows"
93 #define BNXT_DEVARG_REPRESENTOR	"representor"
94 #define BNXT_DEVARG_REP_BASED_PF  "rep-based-pf"
95 #define BNXT_DEVARG_REP_IS_PF  "rep-is-pf"
96 #define BNXT_DEVARG_REP_Q_R2F  "rep-q-r2f"
97 #define BNXT_DEVARG_REP_Q_F2R  "rep-q-f2r"
98 #define BNXT_DEVARG_REP_FC_R2F  "rep-fc-r2f"
99 #define BNXT_DEVARG_REP_FC_F2R  "rep-fc-f2r"
100 #define BNXT_DEVARG_APP_ID	"app-id"
101 
102 static const char *const bnxt_dev_args[] = {
103 	BNXT_DEVARG_REPRESENTOR,
104 	BNXT_DEVARG_ACCUM_STATS,
105 	BNXT_DEVARG_FLOW_XSTAT,
106 	BNXT_DEVARG_MAX_NUM_KFLOWS,
107 	BNXT_DEVARG_REP_BASED_PF,
108 	BNXT_DEVARG_REP_IS_PF,
109 	BNXT_DEVARG_REP_Q_R2F,
110 	BNXT_DEVARG_REP_Q_F2R,
111 	BNXT_DEVARG_REP_FC_R2F,
112 	BNXT_DEVARG_REP_FC_F2R,
113 	BNXT_DEVARG_APP_ID,
114 	NULL
115 };
116 
117 /*
118  * accum-stats == false to disable flow counter accumulation
119  * accum-stats == true to enable flow counter accumulation
120  */
121 #define	BNXT_DEVARG_ACCUM_STATS_INVALID(accum_stats)	((accum_stats) > 1)
122 
123 /*
124  * app-id = an non-negative 8-bit number
125  */
126 #define BNXT_DEVARG_APP_ID_INVALID(val)			((val) > 255)
127 
128 /*
129  * flow_xstat == false to disable the feature
130  * flow_xstat == true to enable the feature
131  */
132 #define	BNXT_DEVARG_FLOW_XSTAT_INVALID(flow_xstat)	((flow_xstat) > 1)
133 
134 /*
135  * rep_is_pf == false to indicate VF representor
136  * rep_is_pf == true to indicate PF representor
137  */
138 #define	BNXT_DEVARG_REP_IS_PF_INVALID(rep_is_pf)	((rep_is_pf) > 1)
139 
140 /*
141  * rep_based_pf == Physical index of the PF
142  */
143 #define	BNXT_DEVARG_REP_BASED_PF_INVALID(rep_based_pf)	((rep_based_pf) > 15)
144 /*
145  * rep_q_r2f == Logical COS Queue index for the rep to endpoint direction
146  */
147 #define	BNXT_DEVARG_REP_Q_R2F_INVALID(rep_q_r2f)	((rep_q_r2f) > 3)
148 
149 /*
150  * rep_q_f2r == Logical COS Queue index for the endpoint to rep direction
151  */
152 #define	BNXT_DEVARG_REP_Q_F2R_INVALID(rep_q_f2r)	((rep_q_f2r) > 3)
153 
154 /*
155  * rep_fc_r2f == Flow control for the representor to endpoint direction
156  */
157 #define BNXT_DEVARG_REP_FC_R2F_INVALID(rep_fc_r2f)	((rep_fc_r2f) > 1)
158 
159 /*
160  * rep_fc_f2r == Flow control for the endpoint to representor direction
161  */
162 #define BNXT_DEVARG_REP_FC_F2R_INVALID(rep_fc_f2r)	((rep_fc_f2r) > 1)
163 
164 int bnxt_cfa_code_dynfield_offset = -1;
165 
166 /*
167  * max_num_kflows must be >= 32
168  * and must be a power-of-2 supported value
169  * return: 1 -> invalid
170  *         0 -> valid
171  */
172 static int bnxt_devarg_max_num_kflow_invalid(uint16_t max_num_kflows)
173 {
174 	if (max_num_kflows < 32 || !rte_is_power_of_2(max_num_kflows))
175 		return 1;
176 	return 0;
177 }
178 
179 static int bnxt_vlan_offload_set_op(struct rte_eth_dev *dev, int mask);
180 static int bnxt_dev_uninit(struct rte_eth_dev *eth_dev);
181 static int bnxt_init_resources(struct bnxt *bp, bool reconfig_dev);
182 static int bnxt_uninit_resources(struct bnxt *bp, bool reconfig_dev);
183 static void bnxt_cancel_fw_health_check(struct bnxt *bp);
184 static int bnxt_restore_vlan_filters(struct bnxt *bp);
185 static void bnxt_dev_recover(void *arg);
186 static void bnxt_free_error_recovery_info(struct bnxt *bp);
187 static void bnxt_free_rep_info(struct bnxt *bp);
188 
189 int is_bnxt_in_error(struct bnxt *bp)
190 {
191 	if (bp->flags & BNXT_FLAG_FATAL_ERROR)
192 		return -EIO;
193 	if (bp->flags & BNXT_FLAG_FW_RESET)
194 		return -EBUSY;
195 
196 	return 0;
197 }
198 
199 /***********************/
200 
201 /*
202  * High level utility functions
203  */
204 
205 static uint16_t bnxt_rss_ctxts(const struct bnxt *bp)
206 {
207 	unsigned int num_rss_rings = RTE_MIN(bp->rx_nr_rings,
208 					     BNXT_RSS_TBL_SIZE_P5);
209 
210 	if (!BNXT_CHIP_P5(bp))
211 		return 1;
212 
213 	return RTE_ALIGN_MUL_CEIL(num_rss_rings,
214 				  BNXT_RSS_ENTRIES_PER_CTX_P5) /
215 				  BNXT_RSS_ENTRIES_PER_CTX_P5;
216 }
217 
218 uint16_t bnxt_rss_hash_tbl_size(const struct bnxt *bp)
219 {
220 	if (!BNXT_CHIP_P5(bp))
221 		return HW_HASH_INDEX_SIZE;
222 
223 	return bnxt_rss_ctxts(bp) * BNXT_RSS_ENTRIES_PER_CTX_P5;
224 }
225 
226 static void bnxt_free_parent_info(struct bnxt *bp)
227 {
228 	rte_free(bp->parent);
229 	bp->parent = NULL;
230 }
231 
232 static void bnxt_free_pf_info(struct bnxt *bp)
233 {
234 	rte_free(bp->pf);
235 	bp->pf = NULL;
236 }
237 
238 static void bnxt_free_link_info(struct bnxt *bp)
239 {
240 	rte_free(bp->link_info);
241 	bp->link_info = NULL;
242 }
243 
244 static void bnxt_free_leds_info(struct bnxt *bp)
245 {
246 	if (BNXT_VF(bp))
247 		return;
248 
249 	rte_free(bp->leds);
250 	bp->leds = NULL;
251 }
252 
253 static void bnxt_free_flow_stats_info(struct bnxt *bp)
254 {
255 	rte_free(bp->flow_stat);
256 	bp->flow_stat = NULL;
257 }
258 
259 static void bnxt_free_cos_queues(struct bnxt *bp)
260 {
261 	rte_free(bp->rx_cos_queue);
262 	bp->rx_cos_queue = NULL;
263 	rte_free(bp->tx_cos_queue);
264 	bp->tx_cos_queue = NULL;
265 }
266 
267 static void bnxt_free_mem(struct bnxt *bp, bool reconfig)
268 {
269 	bnxt_free_filter_mem(bp);
270 	bnxt_free_vnic_attributes(bp);
271 	bnxt_free_vnic_mem(bp);
272 
273 	/* tx/rx rings are configured as part of *_queue_setup callbacks.
274 	 * If the number of rings change across fw update,
275 	 * we don't have much choice except to warn the user.
276 	 */
277 	if (!reconfig) {
278 		bnxt_free_stats(bp);
279 		bnxt_free_tx_rings(bp);
280 		bnxt_free_rx_rings(bp);
281 	}
282 	bnxt_free_async_cp_ring(bp);
283 	bnxt_free_rxtx_nq_ring(bp);
284 
285 	rte_free(bp->grp_info);
286 	bp->grp_info = NULL;
287 }
288 
289 static int bnxt_alloc_parent_info(struct bnxt *bp)
290 {
291 	bp->parent = rte_zmalloc("bnxt_parent_info",
292 				 sizeof(struct bnxt_parent_info), 0);
293 	if (bp->parent == NULL)
294 		return -ENOMEM;
295 
296 	return 0;
297 }
298 
299 static int bnxt_alloc_pf_info(struct bnxt *bp)
300 {
301 	bp->pf = rte_zmalloc("bnxt_pf_info", sizeof(struct bnxt_pf_info), 0);
302 	if (bp->pf == NULL)
303 		return -ENOMEM;
304 
305 	return 0;
306 }
307 
308 static int bnxt_alloc_link_info(struct bnxt *bp)
309 {
310 	bp->link_info =
311 		rte_zmalloc("bnxt_link_info", sizeof(struct bnxt_link_info), 0);
312 	if (bp->link_info == NULL)
313 		return -ENOMEM;
314 
315 	return 0;
316 }
317 
318 static int bnxt_alloc_leds_info(struct bnxt *bp)
319 {
320 	if (BNXT_VF(bp))
321 		return 0;
322 
323 	bp->leds = rte_zmalloc("bnxt_leds",
324 			       BNXT_MAX_LED * sizeof(struct bnxt_led_info),
325 			       0);
326 	if (bp->leds == NULL)
327 		return -ENOMEM;
328 
329 	return 0;
330 }
331 
332 static int bnxt_alloc_cos_queues(struct bnxt *bp)
333 {
334 	bp->rx_cos_queue =
335 		rte_zmalloc("bnxt_rx_cosq",
336 			    BNXT_COS_QUEUE_COUNT *
337 			    sizeof(struct bnxt_cos_queue_info),
338 			    0);
339 	if (bp->rx_cos_queue == NULL)
340 		return -ENOMEM;
341 
342 	bp->tx_cos_queue =
343 		rte_zmalloc("bnxt_tx_cosq",
344 			    BNXT_COS_QUEUE_COUNT *
345 			    sizeof(struct bnxt_cos_queue_info),
346 			    0);
347 	if (bp->tx_cos_queue == NULL)
348 		return -ENOMEM;
349 
350 	return 0;
351 }
352 
353 static int bnxt_alloc_flow_stats_info(struct bnxt *bp)
354 {
355 	bp->flow_stat = rte_zmalloc("bnxt_flow_xstat",
356 				    sizeof(struct bnxt_flow_stat_info), 0);
357 	if (bp->flow_stat == NULL)
358 		return -ENOMEM;
359 
360 	return 0;
361 }
362 
363 static int bnxt_alloc_mem(struct bnxt *bp, bool reconfig)
364 {
365 	int rc;
366 
367 	rc = bnxt_alloc_ring_grps(bp);
368 	if (rc)
369 		goto alloc_mem_err;
370 
371 	rc = bnxt_alloc_async_ring_struct(bp);
372 	if (rc)
373 		goto alloc_mem_err;
374 
375 	rc = bnxt_alloc_vnic_mem(bp);
376 	if (rc)
377 		goto alloc_mem_err;
378 
379 	rc = bnxt_alloc_vnic_attributes(bp);
380 	if (rc)
381 		goto alloc_mem_err;
382 
383 	rc = bnxt_alloc_filter_mem(bp);
384 	if (rc)
385 		goto alloc_mem_err;
386 
387 	rc = bnxt_alloc_async_cp_ring(bp);
388 	if (rc)
389 		goto alloc_mem_err;
390 
391 	rc = bnxt_alloc_rxtx_nq_ring(bp);
392 	if (rc)
393 		goto alloc_mem_err;
394 
395 	if (BNXT_FLOW_XSTATS_EN(bp)) {
396 		rc = bnxt_alloc_flow_stats_info(bp);
397 		if (rc)
398 			goto alloc_mem_err;
399 	}
400 
401 	return 0;
402 
403 alloc_mem_err:
404 	bnxt_free_mem(bp, reconfig);
405 	return rc;
406 }
407 
408 static int bnxt_setup_one_vnic(struct bnxt *bp, uint16_t vnic_id)
409 {
410 	struct rte_eth_conf *dev_conf = &bp->eth_dev->data->dev_conf;
411 	struct bnxt_vnic_info *vnic = &bp->vnic_info[vnic_id];
412 	uint64_t rx_offloads = dev_conf->rxmode.offloads;
413 	struct bnxt_rx_queue *rxq;
414 	unsigned int j;
415 	int rc;
416 
417 	rc = bnxt_vnic_grp_alloc(bp, vnic);
418 	if (rc)
419 		goto err_out;
420 
421 	PMD_DRV_LOG(DEBUG, "vnic[%d] = %p vnic->fw_grp_ids = %p\n",
422 		    vnic_id, vnic, vnic->fw_grp_ids);
423 
424 	rc = bnxt_hwrm_vnic_alloc(bp, vnic);
425 	if (rc)
426 		goto err_out;
427 
428 	/* Alloc RSS context only if RSS mode is enabled */
429 	if (dev_conf->rxmode.mq_mode & RTE_ETH_MQ_RX_RSS) {
430 		int j, nr_ctxs = bnxt_rss_ctxts(bp);
431 
432 		/* RSS table size in Thor is 512.
433 		 * Cap max Rx rings to same value
434 		 */
435 		if (bp->rx_nr_rings > BNXT_RSS_TBL_SIZE_P5) {
436 			PMD_DRV_LOG(ERR, "RxQ cnt %d > reta_size %d\n",
437 				    bp->rx_nr_rings, BNXT_RSS_TBL_SIZE_P5);
438 			goto err_out;
439 		}
440 
441 		rc = 0;
442 		for (j = 0; j < nr_ctxs; j++) {
443 			rc = bnxt_hwrm_vnic_ctx_alloc(bp, vnic, j);
444 			if (rc)
445 				break;
446 		}
447 		if (rc) {
448 			PMD_DRV_LOG(ERR,
449 				    "HWRM vnic %d ctx %d alloc failure rc: %x\n",
450 				    vnic_id, j, rc);
451 			goto err_out;
452 		}
453 		vnic->num_lb_ctxts = nr_ctxs;
454 	}
455 
456 	/*
457 	 * Firmware sets pf pair in default vnic cfg. If the VLAN strip
458 	 * setting is not available at this time, it will not be
459 	 * configured correctly in the CFA.
460 	 */
461 	if (rx_offloads & RTE_ETH_RX_OFFLOAD_VLAN_STRIP)
462 		vnic->vlan_strip = true;
463 	else
464 		vnic->vlan_strip = false;
465 
466 	rc = bnxt_hwrm_vnic_cfg(bp, vnic);
467 	if (rc)
468 		goto err_out;
469 
470 	rc = bnxt_set_hwrm_vnic_filters(bp, vnic);
471 	if (rc)
472 		goto err_out;
473 
474 	for (j = 0; j < bp->rx_num_qs_per_vnic; j++) {
475 		rxq = bp->eth_dev->data->rx_queues[j];
476 
477 		PMD_DRV_LOG(DEBUG,
478 			    "rxq[%d]->vnic=%p vnic->fw_grp_ids=%p\n",
479 			    j, rxq->vnic, rxq->vnic->fw_grp_ids);
480 
481 		if (BNXT_HAS_RING_GRPS(bp) && rxq->rx_deferred_start)
482 			rxq->vnic->fw_grp_ids[j] = INVALID_HW_RING_ID;
483 		else
484 			vnic->rx_queue_cnt++;
485 	}
486 
487 	PMD_DRV_LOG(DEBUG, "vnic->rx_queue_cnt = %d\n", vnic->rx_queue_cnt);
488 
489 	rc = bnxt_vnic_rss_configure(bp, vnic);
490 	if (rc)
491 		goto err_out;
492 
493 	bnxt_hwrm_vnic_plcmode_cfg(bp, vnic);
494 
495 	rc = bnxt_hwrm_vnic_tpa_cfg(bp, vnic,
496 				    (rx_offloads & RTE_ETH_RX_OFFLOAD_TCP_LRO) ?
497 				    true : false);
498 	if (rc)
499 		goto err_out;
500 
501 	return 0;
502 err_out:
503 	PMD_DRV_LOG(ERR, "HWRM vnic %d cfg failure rc: %x\n",
504 		    vnic_id, rc);
505 	return rc;
506 }
507 
508 static int bnxt_register_fc_ctx_mem(struct bnxt *bp)
509 {
510 	int rc = 0;
511 
512 	rc = bnxt_hwrm_ctx_rgtr(bp, bp->flow_stat->rx_fc_in_tbl.dma,
513 				&bp->flow_stat->rx_fc_in_tbl.ctx_id);
514 	if (rc)
515 		return rc;
516 
517 	PMD_DRV_LOG(DEBUG,
518 		    "rx_fc_in_tbl.va = %p rx_fc_in_tbl.dma = %p"
519 		    " rx_fc_in_tbl.ctx_id = %d\n",
520 		    bp->flow_stat->rx_fc_in_tbl.va,
521 		    (void *)((uintptr_t)bp->flow_stat->rx_fc_in_tbl.dma),
522 		    bp->flow_stat->rx_fc_in_tbl.ctx_id);
523 
524 	rc = bnxt_hwrm_ctx_rgtr(bp, bp->flow_stat->rx_fc_out_tbl.dma,
525 				&bp->flow_stat->rx_fc_out_tbl.ctx_id);
526 	if (rc)
527 		return rc;
528 
529 	PMD_DRV_LOG(DEBUG,
530 		    "rx_fc_out_tbl.va = %p rx_fc_out_tbl.dma = %p"
531 		    " rx_fc_out_tbl.ctx_id = %d\n",
532 		    bp->flow_stat->rx_fc_out_tbl.va,
533 		    (void *)((uintptr_t)bp->flow_stat->rx_fc_out_tbl.dma),
534 		    bp->flow_stat->rx_fc_out_tbl.ctx_id);
535 
536 	rc = bnxt_hwrm_ctx_rgtr(bp, bp->flow_stat->tx_fc_in_tbl.dma,
537 				&bp->flow_stat->tx_fc_in_tbl.ctx_id);
538 	if (rc)
539 		return rc;
540 
541 	PMD_DRV_LOG(DEBUG,
542 		    "tx_fc_in_tbl.va = %p tx_fc_in_tbl.dma = %p"
543 		    " tx_fc_in_tbl.ctx_id = %d\n",
544 		    bp->flow_stat->tx_fc_in_tbl.va,
545 		    (void *)((uintptr_t)bp->flow_stat->tx_fc_in_tbl.dma),
546 		    bp->flow_stat->tx_fc_in_tbl.ctx_id);
547 
548 	rc = bnxt_hwrm_ctx_rgtr(bp, bp->flow_stat->tx_fc_out_tbl.dma,
549 				&bp->flow_stat->tx_fc_out_tbl.ctx_id);
550 	if (rc)
551 		return rc;
552 
553 	PMD_DRV_LOG(DEBUG,
554 		    "tx_fc_out_tbl.va = %p tx_fc_out_tbl.dma = %p"
555 		    " tx_fc_out_tbl.ctx_id = %d\n",
556 		    bp->flow_stat->tx_fc_out_tbl.va,
557 		    (void *)((uintptr_t)bp->flow_stat->tx_fc_out_tbl.dma),
558 		    bp->flow_stat->tx_fc_out_tbl.ctx_id);
559 
560 	memset(bp->flow_stat->rx_fc_out_tbl.va,
561 	       0,
562 	       bp->flow_stat->rx_fc_out_tbl.size);
563 	rc = bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_RX,
564 				       CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
565 				       bp->flow_stat->rx_fc_out_tbl.ctx_id,
566 				       bp->flow_stat->max_fc,
567 				       true);
568 	if (rc)
569 		return rc;
570 
571 	memset(bp->flow_stat->tx_fc_out_tbl.va,
572 	       0,
573 	       bp->flow_stat->tx_fc_out_tbl.size);
574 	rc = bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_TX,
575 				       CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
576 				       bp->flow_stat->tx_fc_out_tbl.ctx_id,
577 				       bp->flow_stat->max_fc,
578 				       true);
579 
580 	return rc;
581 }
582 
583 static int bnxt_alloc_ctx_mem_buf(struct bnxt *bp, char *type, size_t size,
584 				  struct bnxt_ctx_mem_buf_info *ctx)
585 {
586 	if (!ctx)
587 		return -EINVAL;
588 
589 	ctx->va = rte_zmalloc_socket(type, size, 0,
590 				     bp->eth_dev->device->numa_node);
591 	if (ctx->va == NULL)
592 		return -ENOMEM;
593 	rte_mem_lock_page(ctx->va);
594 	ctx->size = size;
595 	ctx->dma = rte_mem_virt2iova(ctx->va);
596 	if (ctx->dma == RTE_BAD_IOVA)
597 		return -ENOMEM;
598 
599 	return 0;
600 }
601 
602 static int bnxt_init_fc_ctx_mem(struct bnxt *bp)
603 {
604 	struct rte_pci_device *pdev = bp->pdev;
605 	char type[RTE_MEMZONE_NAMESIZE];
606 	uint16_t max_fc;
607 	int rc = 0;
608 
609 	max_fc = bp->flow_stat->max_fc;
610 
611 	sprintf(type, "bnxt_rx_fc_in_" PCI_PRI_FMT, pdev->addr.domain,
612 		pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
613 	/* 4 bytes for each counter-id */
614 	rc = bnxt_alloc_ctx_mem_buf(bp, type,
615 				    max_fc * 4,
616 				    &bp->flow_stat->rx_fc_in_tbl);
617 	if (rc)
618 		return rc;
619 
620 	sprintf(type, "bnxt_rx_fc_out_" PCI_PRI_FMT, pdev->addr.domain,
621 		pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
622 	/* 16 bytes for each counter - 8 bytes pkt_count, 8 bytes byte_count */
623 	rc = bnxt_alloc_ctx_mem_buf(bp, type,
624 				    max_fc * 16,
625 				    &bp->flow_stat->rx_fc_out_tbl);
626 	if (rc)
627 		return rc;
628 
629 	sprintf(type, "bnxt_tx_fc_in_" PCI_PRI_FMT, pdev->addr.domain,
630 		pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
631 	/* 4 bytes for each counter-id */
632 	rc = bnxt_alloc_ctx_mem_buf(bp, type,
633 				    max_fc * 4,
634 				    &bp->flow_stat->tx_fc_in_tbl);
635 	if (rc)
636 		return rc;
637 
638 	sprintf(type, "bnxt_tx_fc_out_" PCI_PRI_FMT, pdev->addr.domain,
639 		pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
640 	/* 16 bytes for each counter - 8 bytes pkt_count, 8 bytes byte_count */
641 	rc = bnxt_alloc_ctx_mem_buf(bp, type,
642 				    max_fc * 16,
643 				    &bp->flow_stat->tx_fc_out_tbl);
644 	if (rc)
645 		return rc;
646 
647 	rc = bnxt_register_fc_ctx_mem(bp);
648 
649 	return rc;
650 }
651 
652 static int bnxt_init_ctx_mem(struct bnxt *bp)
653 {
654 	int rc = 0;
655 
656 	if (!(bp->fw_cap & BNXT_FW_CAP_ADV_FLOW_COUNTERS) ||
657 	    !(BNXT_PF(bp) || BNXT_VF_IS_TRUSTED(bp)) ||
658 	    !BNXT_FLOW_XSTATS_EN(bp))
659 		return 0;
660 
661 	rc = bnxt_hwrm_cfa_counter_qcaps(bp, &bp->flow_stat->max_fc);
662 	if (rc)
663 		return rc;
664 
665 	rc = bnxt_init_fc_ctx_mem(bp);
666 
667 	return rc;
668 }
669 
670 static int bnxt_update_phy_setting(struct bnxt *bp)
671 {
672 	struct rte_eth_link new;
673 	int rc;
674 
675 	rc = bnxt_get_hwrm_link_config(bp, &new);
676 	if (rc) {
677 		PMD_DRV_LOG(ERR, "Failed to get link settings\n");
678 		return rc;
679 	}
680 
681 	/*
682 	 * On BCM957508-N2100 adapters, FW will not allow any user other
683 	 * than BMC to shutdown the port. bnxt_get_hwrm_link_config() call
684 	 * always returns link up. Force phy update always in that case.
685 	 */
686 	if (!new.link_status || IS_BNXT_DEV_957508_N2100(bp)) {
687 		rc = bnxt_set_hwrm_link_config(bp, true);
688 		if (rc) {
689 			PMD_DRV_LOG(ERR, "Failed to update PHY settings\n");
690 			return rc;
691 		}
692 	}
693 
694 	return rc;
695 }
696 
697 static void bnxt_free_prev_ring_stats(struct bnxt *bp)
698 {
699 	rte_free(bp->prev_rx_ring_stats);
700 	rte_free(bp->prev_tx_ring_stats);
701 
702 	bp->prev_rx_ring_stats = NULL;
703 	bp->prev_tx_ring_stats = NULL;
704 }
705 
706 static int bnxt_alloc_prev_ring_stats(struct bnxt *bp)
707 {
708 	bp->prev_rx_ring_stats =  rte_zmalloc("bnxt_prev_rx_ring_stats",
709 					      sizeof(struct bnxt_ring_stats) *
710 					      bp->rx_cp_nr_rings,
711 					      0);
712 	if (bp->prev_rx_ring_stats == NULL)
713 		return -ENOMEM;
714 
715 	bp->prev_tx_ring_stats = rte_zmalloc("bnxt_prev_tx_ring_stats",
716 					     sizeof(struct bnxt_ring_stats) *
717 					     bp->tx_cp_nr_rings,
718 					     0);
719 	if (bp->prev_tx_ring_stats == NULL)
720 		goto error;
721 
722 	return 0;
723 
724 error:
725 	bnxt_free_prev_ring_stats(bp);
726 	return -ENOMEM;
727 }
728 
729 static int bnxt_start_nic(struct bnxt *bp)
730 {
731 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(bp->eth_dev);
732 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
733 	uint32_t intr_vector = 0;
734 	uint32_t queue_id, base = BNXT_MISC_VEC_ID;
735 	uint32_t vec = BNXT_MISC_VEC_ID;
736 	unsigned int i, j;
737 	int rc;
738 
739 	if (bp->eth_dev->data->mtu > RTE_ETHER_MTU)
740 		bp->flags |= BNXT_FLAG_JUMBO;
741 	else
742 		bp->flags &= ~BNXT_FLAG_JUMBO;
743 
744 	/* THOR does not support ring groups.
745 	 * But we will use the array to save RSS context IDs.
746 	 */
747 	if (BNXT_CHIP_P5(bp))
748 		bp->max_ring_grps = BNXT_MAX_RSS_CTXTS_P5;
749 
750 	rc = bnxt_alloc_all_hwrm_stat_ctxs(bp);
751 	if (rc) {
752 		PMD_DRV_LOG(ERR, "HWRM stat ctx alloc failure rc: %x\n", rc);
753 		goto err_out;
754 	}
755 
756 	rc = bnxt_alloc_hwrm_rings(bp);
757 	if (rc) {
758 		PMD_DRV_LOG(ERR, "HWRM ring alloc failure rc: %x\n", rc);
759 		goto err_out;
760 	}
761 
762 	rc = bnxt_alloc_all_hwrm_ring_grps(bp);
763 	if (rc) {
764 		PMD_DRV_LOG(ERR, "HWRM ring grp alloc failure: %x\n", rc);
765 		goto err_out;
766 	}
767 
768 	if (!(bp->vnic_cap_flags & BNXT_VNIC_CAP_COS_CLASSIFY))
769 		goto skip_cosq_cfg;
770 
771 	for (j = 0, i = 0; i < BNXT_COS_QUEUE_COUNT; i++) {
772 		if (bp->rx_cos_queue[i].id != 0xff) {
773 			struct bnxt_vnic_info *vnic = &bp->vnic_info[j++];
774 
775 			if (!vnic) {
776 				PMD_DRV_LOG(ERR,
777 					    "Num pools more than FW profile\n");
778 				rc = -EINVAL;
779 				goto err_out;
780 			}
781 			vnic->cos_queue_id = bp->rx_cos_queue[i].id;
782 			bp->rx_cosq_cnt++;
783 		}
784 	}
785 
786 skip_cosq_cfg:
787 	rc = bnxt_mq_rx_configure(bp);
788 	if (rc) {
789 		PMD_DRV_LOG(ERR, "MQ mode configure failure rc: %x\n", rc);
790 		goto err_out;
791 	}
792 
793 	/* default vnic 0 */
794 	rc = bnxt_setup_one_vnic(bp, 0);
795 	if (rc)
796 		goto err_out;
797 	/* VNIC configuration */
798 	if (BNXT_RFS_NEEDS_VNIC(bp)) {
799 		for (i = 1; i < bp->nr_vnics; i++) {
800 			rc = bnxt_setup_one_vnic(bp, i);
801 			if (rc)
802 				goto err_out;
803 		}
804 	}
805 
806 	for (j = 0; j < bp->tx_nr_rings; j++) {
807 		struct bnxt_tx_queue *txq = bp->tx_queues[j];
808 
809 		if (!txq->tx_deferred_start) {
810 			bp->eth_dev->data->tx_queue_state[j] =
811 				RTE_ETH_QUEUE_STATE_STARTED;
812 			txq->tx_started = true;
813 		}
814 	}
815 
816 	for (j = 0; j < bp->rx_nr_rings; j++) {
817 		struct bnxt_rx_queue *rxq = bp->rx_queues[j];
818 
819 		if (!rxq->rx_deferred_start) {
820 			bp->eth_dev->data->rx_queue_state[j] =
821 				RTE_ETH_QUEUE_STATE_STARTED;
822 			rxq->rx_started = true;
823 		}
824 	}
825 
826 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, &bp->vnic_info[0], 0, NULL);
827 	if (rc) {
828 		PMD_DRV_LOG(ERR,
829 			"HWRM cfa l2 rx mask failure rc: %x\n", rc);
830 		goto err_out;
831 	}
832 
833 	/* check and configure queue intr-vector mapping */
834 	if ((rte_intr_cap_multiple(intr_handle) ||
835 	     !RTE_ETH_DEV_SRIOV(bp->eth_dev).active) &&
836 	    bp->eth_dev->data->dev_conf.intr_conf.rxq != 0) {
837 		intr_vector = bp->eth_dev->data->nb_rx_queues;
838 		PMD_DRV_LOG(DEBUG, "intr_vector = %d\n", intr_vector);
839 		if (intr_vector > bp->rx_cp_nr_rings) {
840 			PMD_DRV_LOG(ERR, "At most %d intr queues supported",
841 					bp->rx_cp_nr_rings);
842 			return -ENOTSUP;
843 		}
844 		rc = rte_intr_efd_enable(intr_handle, intr_vector);
845 		if (rc)
846 			return rc;
847 	}
848 
849 	if (rte_intr_dp_is_en(intr_handle) && !intr_handle->intr_vec) {
850 		intr_handle->intr_vec =
851 			rte_zmalloc("intr_vec",
852 				    bp->eth_dev->data->nb_rx_queues *
853 				    sizeof(int), 0);
854 		if (intr_handle->intr_vec == NULL) {
855 			PMD_DRV_LOG(ERR, "Failed to allocate %d rx_queues"
856 				" intr_vec", bp->eth_dev->data->nb_rx_queues);
857 			rc = -ENOMEM;
858 			goto err_out;
859 		}
860 		PMD_DRV_LOG(DEBUG, "intr_handle->intr_vec = %p "
861 			"intr_handle->nb_efd = %d intr_handle->max_intr = %d\n",
862 			 intr_handle->intr_vec, intr_handle->nb_efd,
863 			intr_handle->max_intr);
864 		for (queue_id = 0; queue_id < bp->eth_dev->data->nb_rx_queues;
865 		     queue_id++) {
866 			intr_handle->intr_vec[queue_id] =
867 							vec + BNXT_RX_VEC_START;
868 			if (vec < base + intr_handle->nb_efd - 1)
869 				vec++;
870 		}
871 	}
872 
873 	/* enable uio/vfio intr/eventfd mapping */
874 	rc = rte_intr_enable(intr_handle);
875 #ifndef RTE_EXEC_ENV_FREEBSD
876 	/* In FreeBSD OS, nic_uio driver does not support interrupts */
877 	if (rc)
878 		goto err_out;
879 #endif
880 
881 	rc = bnxt_update_phy_setting(bp);
882 	if (rc)
883 		goto err_out;
884 
885 	bp->mark_table = rte_zmalloc("bnxt_mark_table", BNXT_MARK_TABLE_SZ, 0);
886 	if (!bp->mark_table)
887 		PMD_DRV_LOG(ERR, "Allocation of mark table failed\n");
888 
889 	return 0;
890 
891 err_out:
892 	/* Some of the error status returned by FW may not be from errno.h */
893 	if (rc > 0)
894 		rc = -EIO;
895 
896 	return rc;
897 }
898 
899 static int bnxt_shutdown_nic(struct bnxt *bp)
900 {
901 	bnxt_free_all_hwrm_resources(bp);
902 	bnxt_free_all_filters(bp);
903 	bnxt_free_all_vnics(bp);
904 	return 0;
905 }
906 
907 /*
908  * Device configuration and status function
909  */
910 
911 uint32_t bnxt_get_speed_capabilities(struct bnxt *bp)
912 {
913 	uint32_t link_speed = 0;
914 	uint32_t speed_capa = 0;
915 
916 	if (bp->link_info == NULL)
917 		return 0;
918 
919 	link_speed = bp->link_info->support_speeds;
920 
921 	/* If PAM4 is configured, use PAM4 supported speed */
922 	if (link_speed == 0 && bp->link_info->support_pam4_speeds > 0)
923 		link_speed = bp->link_info->support_pam4_speeds;
924 
925 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_LINK_SPEED_100MB)
926 		speed_capa |= RTE_ETH_LINK_SPEED_100M;
927 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_100MBHD)
928 		speed_capa |= RTE_ETH_LINK_SPEED_100M_HD;
929 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_1GB)
930 		speed_capa |= RTE_ETH_LINK_SPEED_1G;
931 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_2_5GB)
932 		speed_capa |= RTE_ETH_LINK_SPEED_2_5G;
933 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_10GB)
934 		speed_capa |= RTE_ETH_LINK_SPEED_10G;
935 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_20GB)
936 		speed_capa |= RTE_ETH_LINK_SPEED_20G;
937 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_25GB)
938 		speed_capa |= RTE_ETH_LINK_SPEED_25G;
939 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_40GB)
940 		speed_capa |= RTE_ETH_LINK_SPEED_40G;
941 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_50GB)
942 		speed_capa |= RTE_ETH_LINK_SPEED_50G;
943 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_100GB)
944 		speed_capa |= RTE_ETH_LINK_SPEED_100G;
945 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_PAM4_SPEEDS_50G)
946 		speed_capa |= RTE_ETH_LINK_SPEED_50G;
947 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_PAM4_SPEEDS_100G)
948 		speed_capa |= RTE_ETH_LINK_SPEED_100G;
949 	if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_PAM4_SPEEDS_200G)
950 		speed_capa |= RTE_ETH_LINK_SPEED_200G;
951 
952 	if (bp->link_info->auto_mode ==
953 	    HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_MODE_NONE)
954 		speed_capa |= RTE_ETH_LINK_SPEED_FIXED;
955 
956 	return speed_capa;
957 }
958 
959 static int bnxt_dev_info_get_op(struct rte_eth_dev *eth_dev,
960 				struct rte_eth_dev_info *dev_info)
961 {
962 	struct rte_pci_device *pdev = RTE_DEV_TO_PCI(eth_dev->device);
963 	struct bnxt *bp = eth_dev->data->dev_private;
964 	uint16_t max_vnics, i, j, vpool, vrxq;
965 	unsigned int max_rx_rings;
966 	int rc;
967 
968 	rc = is_bnxt_in_error(bp);
969 	if (rc)
970 		return rc;
971 
972 	/* MAC Specifics */
973 	dev_info->max_mac_addrs = bp->max_l2_ctx;
974 	dev_info->max_hash_mac_addrs = 0;
975 
976 	/* PF/VF specifics */
977 	if (BNXT_PF(bp))
978 		dev_info->max_vfs = pdev->max_vfs;
979 
980 	max_rx_rings = bnxt_max_rings(bp);
981 	/* For the sake of symmetry, max_rx_queues = max_tx_queues */
982 	dev_info->max_rx_queues = max_rx_rings;
983 	dev_info->max_tx_queues = max_rx_rings;
984 	dev_info->reta_size = bnxt_rss_hash_tbl_size(bp);
985 	dev_info->hash_key_size = HW_HASH_KEY_SIZE;
986 	max_vnics = bp->max_vnics;
987 
988 	/* MTU specifics */
989 	dev_info->min_mtu = RTE_ETHER_MIN_MTU;
990 	dev_info->max_mtu = BNXT_MAX_MTU;
991 
992 	/* Fast path specifics */
993 	dev_info->min_rx_bufsize = 1;
994 	dev_info->max_rx_pktlen = BNXT_MAX_PKT_LEN;
995 
996 	dev_info->rx_offload_capa = BNXT_DEV_RX_OFFLOAD_SUPPORT;
997 	if (bp->flags & BNXT_FLAG_PTP_SUPPORTED)
998 		dev_info->rx_offload_capa |= RTE_ETH_RX_OFFLOAD_TIMESTAMP;
999 	if (bp->vnic_cap_flags & BNXT_VNIC_CAP_VLAN_RX_STRIP)
1000 		dev_info->rx_offload_capa |= RTE_ETH_RX_OFFLOAD_VLAN_STRIP;
1001 	dev_info->tx_queue_offload_capa = RTE_ETH_TX_OFFLOAD_MBUF_FAST_FREE;
1002 	dev_info->tx_offload_capa = BNXT_DEV_TX_OFFLOAD_SUPPORT |
1003 				    dev_info->tx_queue_offload_capa;
1004 	if (bp->fw_cap & BNXT_FW_CAP_VLAN_TX_INSERT)
1005 		dev_info->tx_offload_capa |= RTE_ETH_TX_OFFLOAD_VLAN_INSERT;
1006 	dev_info->flow_type_rss_offloads = BNXT_ETH_RSS_SUPPORT;
1007 
1008 	dev_info->speed_capa = bnxt_get_speed_capabilities(bp);
1009 	dev_info->dev_capa = RTE_ETH_DEV_CAPA_RUNTIME_RX_QUEUE_SETUP |
1010 			     RTE_ETH_DEV_CAPA_RUNTIME_TX_QUEUE_SETUP;
1011 
1012 	dev_info->default_rxconf = (struct rte_eth_rxconf) {
1013 		.rx_thresh = {
1014 			.pthresh = 8,
1015 			.hthresh = 8,
1016 			.wthresh = 0,
1017 		},
1018 		.rx_free_thresh = 32,
1019 		.rx_drop_en = BNXT_DEFAULT_RX_DROP_EN,
1020 	};
1021 
1022 	dev_info->default_txconf = (struct rte_eth_txconf) {
1023 		.tx_thresh = {
1024 			.pthresh = 32,
1025 			.hthresh = 0,
1026 			.wthresh = 0,
1027 		},
1028 		.tx_free_thresh = 32,
1029 		.tx_rs_thresh = 32,
1030 	};
1031 	eth_dev->data->dev_conf.intr_conf.lsc = 1;
1032 
1033 	dev_info->rx_desc_lim.nb_min = BNXT_MIN_RING_DESC;
1034 	dev_info->rx_desc_lim.nb_max = BNXT_MAX_RX_RING_DESC;
1035 	dev_info->tx_desc_lim.nb_min = BNXT_MIN_RING_DESC;
1036 	dev_info->tx_desc_lim.nb_max = BNXT_MAX_TX_RING_DESC;
1037 
1038 	if (BNXT_PF(bp) || BNXT_VF_IS_TRUSTED(bp)) {
1039 		dev_info->switch_info.name = eth_dev->device->name;
1040 		dev_info->switch_info.domain_id = bp->switch_domain_id;
1041 		dev_info->switch_info.port_id =
1042 				BNXT_PF(bp) ? BNXT_SWITCH_PORT_ID_PF :
1043 				    BNXT_SWITCH_PORT_ID_TRUSTED_VF;
1044 	}
1045 
1046 	/*
1047 	 * TODO: default_rxconf, default_txconf, rx_desc_lim, and tx_desc_lim
1048 	 *       need further investigation.
1049 	 */
1050 
1051 	/* VMDq resources */
1052 	vpool = 64; /* RTE_ETH_64_POOLS */
1053 	vrxq = 128; /* RTE_ETH_VMDQ_DCB_NUM_QUEUES */
1054 	for (i = 0; i < 4; vpool >>= 1, i++) {
1055 		if (max_vnics > vpool) {
1056 			for (j = 0; j < 5; vrxq >>= 1, j++) {
1057 				if (dev_info->max_rx_queues > vrxq) {
1058 					if (vpool > vrxq)
1059 						vpool = vrxq;
1060 					goto found;
1061 				}
1062 			}
1063 			/* Not enough resources to support VMDq */
1064 			break;
1065 		}
1066 	}
1067 	/* Not enough resources to support VMDq */
1068 	vpool = 0;
1069 	vrxq = 0;
1070 found:
1071 	dev_info->max_vmdq_pools = vpool;
1072 	dev_info->vmdq_queue_num = vrxq;
1073 
1074 	dev_info->vmdq_pool_base = 0;
1075 	dev_info->vmdq_queue_base = 0;
1076 
1077 	return 0;
1078 }
1079 
1080 /* Configure the device based on the configuration provided */
1081 static int bnxt_dev_configure_op(struct rte_eth_dev *eth_dev)
1082 {
1083 	struct bnxt *bp = eth_dev->data->dev_private;
1084 	uint64_t rx_offloads = eth_dev->data->dev_conf.rxmode.offloads;
1085 	int rc;
1086 
1087 	bp->rx_queues = (void *)eth_dev->data->rx_queues;
1088 	bp->tx_queues = (void *)eth_dev->data->tx_queues;
1089 	bp->tx_nr_rings = eth_dev->data->nb_tx_queues;
1090 	bp->rx_nr_rings = eth_dev->data->nb_rx_queues;
1091 
1092 	rc = is_bnxt_in_error(bp);
1093 	if (rc)
1094 		return rc;
1095 
1096 	if (BNXT_VF(bp) && (bp->flags & BNXT_FLAG_NEW_RM)) {
1097 		rc = bnxt_hwrm_check_vf_rings(bp);
1098 		if (rc) {
1099 			PMD_DRV_LOG(ERR, "HWRM insufficient resources\n");
1100 			return -ENOSPC;
1101 		}
1102 
1103 		/* If a resource has already been allocated - in this case
1104 		 * it is the async completion ring, free it. Reallocate it after
1105 		 * resource reservation. This will ensure the resource counts
1106 		 * are calculated correctly.
1107 		 */
1108 
1109 		pthread_mutex_lock(&bp->def_cp_lock);
1110 
1111 		if (!BNXT_HAS_NQ(bp) && bp->async_cp_ring) {
1112 			bnxt_disable_int(bp);
1113 			bnxt_free_cp_ring(bp, bp->async_cp_ring);
1114 		}
1115 
1116 		rc = bnxt_hwrm_func_reserve_vf_resc(bp, false);
1117 		if (rc) {
1118 			PMD_DRV_LOG(ERR, "HWRM resource alloc fail:%x\n", rc);
1119 			pthread_mutex_unlock(&bp->def_cp_lock);
1120 			return -ENOSPC;
1121 		}
1122 
1123 		if (!BNXT_HAS_NQ(bp) && bp->async_cp_ring) {
1124 			rc = bnxt_alloc_async_cp_ring(bp);
1125 			if (rc) {
1126 				pthread_mutex_unlock(&bp->def_cp_lock);
1127 				return rc;
1128 			}
1129 			bnxt_enable_int(bp);
1130 		}
1131 
1132 		pthread_mutex_unlock(&bp->def_cp_lock);
1133 	}
1134 
1135 	/* Inherit new configurations */
1136 	if (eth_dev->data->nb_rx_queues > bp->max_rx_rings ||
1137 	    eth_dev->data->nb_tx_queues > bp->max_tx_rings ||
1138 	    eth_dev->data->nb_rx_queues + eth_dev->data->nb_tx_queues
1139 		+ BNXT_NUM_ASYNC_CPR(bp) > bp->max_cp_rings ||
1140 	    eth_dev->data->nb_rx_queues + eth_dev->data->nb_tx_queues >
1141 	    bp->max_stat_ctx)
1142 		goto resource_error;
1143 
1144 	if (BNXT_HAS_RING_GRPS(bp) &&
1145 	    (uint32_t)(eth_dev->data->nb_rx_queues) > bp->max_ring_grps)
1146 		goto resource_error;
1147 
1148 	if (!(eth_dev->data->dev_conf.rxmode.mq_mode & RTE_ETH_MQ_RX_RSS) &&
1149 	    bp->max_vnics < eth_dev->data->nb_rx_queues)
1150 		goto resource_error;
1151 
1152 	bp->rx_cp_nr_rings = bp->rx_nr_rings;
1153 	bp->tx_cp_nr_rings = bp->tx_nr_rings;
1154 
1155 	if (eth_dev->data->dev_conf.rxmode.mq_mode & RTE_ETH_MQ_RX_RSS_FLAG)
1156 		rx_offloads |= RTE_ETH_RX_OFFLOAD_RSS_HASH;
1157 	eth_dev->data->dev_conf.rxmode.offloads = rx_offloads;
1158 
1159 	bnxt_mtu_set_op(eth_dev, eth_dev->data->mtu);
1160 
1161 	return 0;
1162 
1163 resource_error:
1164 	PMD_DRV_LOG(ERR,
1165 		    "Insufficient resources to support requested config\n");
1166 	PMD_DRV_LOG(ERR,
1167 		    "Num Queues Requested: Tx %d, Rx %d\n",
1168 		    eth_dev->data->nb_tx_queues,
1169 		    eth_dev->data->nb_rx_queues);
1170 	PMD_DRV_LOG(ERR,
1171 		    "MAX: TxQ %d, RxQ %d, CQ %d Stat %d, Grp %d, Vnic %d\n",
1172 		    bp->max_tx_rings, bp->max_rx_rings, bp->max_cp_rings,
1173 		    bp->max_stat_ctx, bp->max_ring_grps, bp->max_vnics);
1174 	return -ENOSPC;
1175 }
1176 
1177 void bnxt_print_link_info(struct rte_eth_dev *eth_dev)
1178 {
1179 	struct rte_eth_link *link = &eth_dev->data->dev_link;
1180 
1181 	if (link->link_status)
1182 		PMD_DRV_LOG(INFO, "Port %d Link Up - speed %u Mbps - %s\n",
1183 			eth_dev->data->port_id,
1184 			(uint32_t)link->link_speed,
1185 			(link->link_duplex == RTE_ETH_LINK_FULL_DUPLEX) ?
1186 			("full-duplex") : ("half-duplex\n"));
1187 	else
1188 		PMD_DRV_LOG(INFO, "Port %d Link Down\n",
1189 			eth_dev->data->port_id);
1190 }
1191 
1192 /*
1193  * Determine whether the current configuration requires support for scattered
1194  * receive; return 1 if scattered receive is required and 0 if not.
1195  */
1196 static int bnxt_scattered_rx(struct rte_eth_dev *eth_dev)
1197 {
1198 	uint32_t overhead = BNXT_MAX_PKT_LEN - BNXT_MAX_MTU;
1199 	uint16_t buf_size;
1200 	int i;
1201 
1202 	if (eth_dev->data->dev_conf.rxmode.offloads & RTE_ETH_RX_OFFLOAD_SCATTER)
1203 		return 1;
1204 
1205 	if (eth_dev->data->dev_conf.rxmode.offloads & RTE_ETH_RX_OFFLOAD_TCP_LRO)
1206 		return 1;
1207 
1208 	for (i = 0; i < eth_dev->data->nb_rx_queues; i++) {
1209 		struct bnxt_rx_queue *rxq = eth_dev->data->rx_queues[i];
1210 
1211 		buf_size = (uint16_t)(rte_pktmbuf_data_room_size(rxq->mb_pool) -
1212 				      RTE_PKTMBUF_HEADROOM);
1213 		if (eth_dev->data->mtu + overhead > buf_size)
1214 			return 1;
1215 	}
1216 	return 0;
1217 }
1218 
1219 static eth_rx_burst_t
1220 bnxt_receive_function(struct rte_eth_dev *eth_dev)
1221 {
1222 	struct bnxt *bp = eth_dev->data->dev_private;
1223 
1224 	/* Disable vector mode RX for Stingray2 for now */
1225 	if (BNXT_CHIP_SR2(bp)) {
1226 		bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
1227 		return bnxt_recv_pkts;
1228 	}
1229 
1230 #if (defined(RTE_ARCH_X86) || defined(RTE_ARCH_ARM64)) && \
1231 	!defined(RTE_LIBRTE_IEEE1588)
1232 
1233 	/* Vector mode receive cannot be enabled if scattered rx is in use. */
1234 	if (eth_dev->data->scattered_rx)
1235 		goto use_scalar_rx;
1236 
1237 	/*
1238 	 * Vector mode receive cannot be enabled if Truflow is enabled or if
1239 	 * asynchronous completions and receive completions can be placed in
1240 	 * the same completion ring.
1241 	 */
1242 	if (BNXT_TRUFLOW_EN(bp) || !BNXT_NUM_ASYNC_CPR(bp))
1243 		goto use_scalar_rx;
1244 
1245 	/*
1246 	 * Vector mode receive cannot be enabled if any receive offloads outside
1247 	 * a limited subset have been enabled.
1248 	 */
1249 	if (eth_dev->data->dev_conf.rxmode.offloads &
1250 		~(RTE_ETH_RX_OFFLOAD_VLAN_STRIP |
1251 		  RTE_ETH_RX_OFFLOAD_KEEP_CRC |
1252 		  RTE_ETH_RX_OFFLOAD_IPV4_CKSUM |
1253 		  RTE_ETH_RX_OFFLOAD_UDP_CKSUM |
1254 		  RTE_ETH_RX_OFFLOAD_TCP_CKSUM |
1255 		  RTE_ETH_RX_OFFLOAD_OUTER_IPV4_CKSUM |
1256 		  RTE_ETH_RX_OFFLOAD_OUTER_UDP_CKSUM |
1257 		  RTE_ETH_RX_OFFLOAD_RSS_HASH |
1258 		  RTE_ETH_RX_OFFLOAD_VLAN_FILTER))
1259 		goto use_scalar_rx;
1260 
1261 #if defined(RTE_ARCH_X86) && defined(CC_AVX2_SUPPORT)
1262 	if (rte_vect_get_max_simd_bitwidth() >= RTE_VECT_SIMD_256 &&
1263 	    rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX2) == 1) {
1264 		PMD_DRV_LOG(INFO,
1265 			    "Using AVX2 vector mode receive for port %d\n",
1266 			    eth_dev->data->port_id);
1267 		bp->flags |= BNXT_FLAG_RX_VECTOR_PKT_MODE;
1268 		return bnxt_recv_pkts_vec_avx2;
1269 	}
1270  #endif
1271 	if (rte_vect_get_max_simd_bitwidth() >= RTE_VECT_SIMD_128) {
1272 		PMD_DRV_LOG(INFO,
1273 			    "Using SSE vector mode receive for port %d\n",
1274 			    eth_dev->data->port_id);
1275 		bp->flags |= BNXT_FLAG_RX_VECTOR_PKT_MODE;
1276 		return bnxt_recv_pkts_vec;
1277 	}
1278 
1279 use_scalar_rx:
1280 	PMD_DRV_LOG(INFO, "Vector mode receive disabled for port %d\n",
1281 		    eth_dev->data->port_id);
1282 	PMD_DRV_LOG(INFO,
1283 		    "Port %d scatter: %d rx offload: %" PRIX64 "\n",
1284 		    eth_dev->data->port_id,
1285 		    eth_dev->data->scattered_rx,
1286 		    eth_dev->data->dev_conf.rxmode.offloads);
1287 #endif
1288 	bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
1289 	return bnxt_recv_pkts;
1290 }
1291 
1292 static eth_tx_burst_t
1293 bnxt_transmit_function(struct rte_eth_dev *eth_dev)
1294 {
1295 	struct bnxt *bp = eth_dev->data->dev_private;
1296 
1297 	/* Disable vector mode TX for Stingray2 for now */
1298 	if (BNXT_CHIP_SR2(bp))
1299 		return bnxt_xmit_pkts;
1300 
1301 #if defined(RTE_ARCH_X86) || defined(RTE_ARCH_ARM64) && \
1302 	!defined(RTE_LIBRTE_IEEE1588)
1303 	uint64_t offloads = eth_dev->data->dev_conf.txmode.offloads;
1304 
1305 	/*
1306 	 * Vector mode transmit can be enabled only if not using scatter rx
1307 	 * or tx offloads.
1308 	 */
1309 	if (eth_dev->data->scattered_rx ||
1310 	    (offloads & ~RTE_ETH_TX_OFFLOAD_MBUF_FAST_FREE) ||
1311 	    BNXT_TRUFLOW_EN(bp))
1312 		goto use_scalar_tx;
1313 
1314 #if defined(RTE_ARCH_X86) && defined(CC_AVX2_SUPPORT)
1315 	if (rte_vect_get_max_simd_bitwidth() >= RTE_VECT_SIMD_256 &&
1316 	    rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX2) == 1) {
1317 		PMD_DRV_LOG(INFO,
1318 			    "Using AVX2 vector mode transmit for port %d\n",
1319 			    eth_dev->data->port_id);
1320 		return bnxt_xmit_pkts_vec_avx2;
1321 	}
1322 #endif
1323 	if (rte_vect_get_max_simd_bitwidth() >= RTE_VECT_SIMD_128) {
1324 		PMD_DRV_LOG(INFO,
1325 			    "Using SSE vector mode transmit for port %d\n",
1326 			    eth_dev->data->port_id);
1327 		return bnxt_xmit_pkts_vec;
1328 	}
1329 
1330 use_scalar_tx:
1331 	PMD_DRV_LOG(INFO, "Vector mode transmit disabled for port %d\n",
1332 		    eth_dev->data->port_id);
1333 	PMD_DRV_LOG(INFO,
1334 		    "Port %d scatter: %d tx offload: %" PRIX64 "\n",
1335 		    eth_dev->data->port_id,
1336 		    eth_dev->data->scattered_rx,
1337 		    offloads);
1338 #endif
1339 	return bnxt_xmit_pkts;
1340 }
1341 
1342 static int bnxt_handle_if_change_status(struct bnxt *bp)
1343 {
1344 	int rc;
1345 
1346 	/* Since fw has undergone a reset and lost all contexts,
1347 	 * set fatal flag to not issue hwrm during cleanup
1348 	 */
1349 	bp->flags |= BNXT_FLAG_FATAL_ERROR;
1350 	bnxt_uninit_resources(bp, true);
1351 
1352 	/* clear fatal flag so that re-init happens */
1353 	bp->flags &= ~BNXT_FLAG_FATAL_ERROR;
1354 	rc = bnxt_init_resources(bp, true);
1355 
1356 	bp->flags &= ~BNXT_FLAG_IF_CHANGE_HOT_FW_RESET_DONE;
1357 
1358 	return rc;
1359 }
1360 
1361 static int bnxt_dev_set_link_up_op(struct rte_eth_dev *eth_dev)
1362 {
1363 	struct bnxt *bp = eth_dev->data->dev_private;
1364 	int rc = 0;
1365 
1366 	if (!BNXT_SINGLE_PF(bp))
1367 		return -ENOTSUP;
1368 
1369 	if (!bp->link_info->link_up)
1370 		rc = bnxt_set_hwrm_link_config(bp, true);
1371 	if (!rc)
1372 		eth_dev->data->dev_link.link_status = 1;
1373 
1374 	bnxt_print_link_info(eth_dev);
1375 	return rc;
1376 }
1377 
1378 static int bnxt_dev_set_link_down_op(struct rte_eth_dev *eth_dev)
1379 {
1380 	struct bnxt *bp = eth_dev->data->dev_private;
1381 
1382 	if (!BNXT_SINGLE_PF(bp))
1383 		return -ENOTSUP;
1384 
1385 	eth_dev->data->dev_link.link_status = 0;
1386 	bnxt_set_hwrm_link_config(bp, false);
1387 	bp->link_info->link_up = 0;
1388 
1389 	return 0;
1390 }
1391 
1392 static void bnxt_free_switch_domain(struct bnxt *bp)
1393 {
1394 	int rc = 0;
1395 
1396 	if (!(BNXT_PF(bp) || BNXT_VF_IS_TRUSTED(bp)))
1397 		return;
1398 
1399 	rc = rte_eth_switch_domain_free(bp->switch_domain_id);
1400 	if (rc)
1401 		PMD_DRV_LOG(ERR, "free switch domain:%d fail: %d\n",
1402 			    bp->switch_domain_id, rc);
1403 }
1404 
1405 static void bnxt_ptp_get_current_time(void *arg)
1406 {
1407 	struct bnxt *bp = arg;
1408 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
1409 	int rc;
1410 
1411 	rc = is_bnxt_in_error(bp);
1412 	if (rc)
1413 		return;
1414 
1415 	if (!ptp)
1416 		return;
1417 
1418 	bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_CURRENT_TIME,
1419 				&ptp->current_time);
1420 
1421 	rc = rte_eal_alarm_set(US_PER_S, bnxt_ptp_get_current_time, (void *)bp);
1422 	if (rc != 0) {
1423 		PMD_DRV_LOG(ERR, "Failed to re-schedule PTP alarm\n");
1424 		bp->flags2 &= ~BNXT_FLAGS2_PTP_ALARM_SCHEDULED;
1425 	}
1426 }
1427 
1428 static int bnxt_schedule_ptp_alarm(struct bnxt *bp)
1429 {
1430 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
1431 	int rc;
1432 
1433 	if (bp->flags2 & BNXT_FLAGS2_PTP_ALARM_SCHEDULED)
1434 		return 0;
1435 
1436 	bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_CURRENT_TIME,
1437 				&ptp->current_time);
1438 
1439 	rc = rte_eal_alarm_set(US_PER_S, bnxt_ptp_get_current_time, (void *)bp);
1440 	return rc;
1441 }
1442 
1443 static void bnxt_cancel_ptp_alarm(struct bnxt *bp)
1444 {
1445 	if (bp->flags2 & BNXT_FLAGS2_PTP_ALARM_SCHEDULED) {
1446 		rte_eal_alarm_cancel(bnxt_ptp_get_current_time, (void *)bp);
1447 		bp->flags2 &= ~BNXT_FLAGS2_PTP_ALARM_SCHEDULED;
1448 	}
1449 }
1450 
1451 static void bnxt_ptp_stop(struct bnxt *bp)
1452 {
1453 	bnxt_cancel_ptp_alarm(bp);
1454 	bp->flags2 &= ~BNXT_FLAGS2_PTP_TIMESYNC_ENABLED;
1455 }
1456 
1457 static int bnxt_ptp_start(struct bnxt *bp)
1458 {
1459 	int rc;
1460 
1461 	rc = bnxt_schedule_ptp_alarm(bp);
1462 	if (rc != 0) {
1463 		PMD_DRV_LOG(ERR, "Failed to schedule PTP alarm\n");
1464 	} else {
1465 		bp->flags2 |= BNXT_FLAGS2_PTP_TIMESYNC_ENABLED;
1466 		bp->flags2 |= BNXT_FLAGS2_PTP_ALARM_SCHEDULED;
1467 	}
1468 
1469 	return rc;
1470 }
1471 
1472 static int bnxt_dev_stop(struct rte_eth_dev *eth_dev)
1473 {
1474 	struct bnxt *bp = eth_dev->data->dev_private;
1475 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
1476 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
1477 	struct rte_eth_link link;
1478 	int ret;
1479 
1480 	eth_dev->data->dev_started = 0;
1481 	eth_dev->data->scattered_rx = 0;
1482 
1483 	/* Prevent crashes when queues are still in use */
1484 	eth_dev->rx_pkt_burst = &bnxt_dummy_recv_pkts;
1485 	eth_dev->tx_pkt_burst = &bnxt_dummy_xmit_pkts;
1486 
1487 	bnxt_disable_int(bp);
1488 
1489 	/* disable uio/vfio intr/eventfd mapping */
1490 	rte_intr_disable(intr_handle);
1491 
1492 	/* Stop the child representors for this device */
1493 	ret = bnxt_rep_stop_all(bp);
1494 	if (ret != 0)
1495 		return ret;
1496 
1497 	/* delete the bnxt ULP port details */
1498 	bnxt_ulp_port_deinit(bp);
1499 
1500 	bnxt_cancel_fw_health_check(bp);
1501 
1502 	if (BNXT_P5_PTP_TIMESYNC_ENABLED(bp))
1503 		bnxt_cancel_ptp_alarm(bp);
1504 
1505 	/* Do not bring link down during reset recovery */
1506 	if (!is_bnxt_in_error(bp)) {
1507 		bnxt_dev_set_link_down_op(eth_dev);
1508 		/* Wait for link to be reset */
1509 		if (BNXT_SINGLE_PF(bp))
1510 			rte_delay_ms(500);
1511 		/* clear the recorded link status */
1512 		memset(&link, 0, sizeof(link));
1513 		rte_eth_linkstatus_set(eth_dev, &link);
1514 	}
1515 
1516 	/* Clean queue intr-vector mapping */
1517 	rte_intr_efd_disable(intr_handle);
1518 	if (intr_handle->intr_vec != NULL) {
1519 		rte_free(intr_handle->intr_vec);
1520 		intr_handle->intr_vec = NULL;
1521 	}
1522 
1523 	bnxt_hwrm_port_clr_stats(bp);
1524 	bnxt_free_tx_mbufs(bp);
1525 	bnxt_free_rx_mbufs(bp);
1526 	/* Process any remaining notifications in default completion queue */
1527 	bnxt_int_handler(eth_dev);
1528 	bnxt_shutdown_nic(bp);
1529 	bnxt_hwrm_if_change(bp, false);
1530 
1531 	bnxt_free_prev_ring_stats(bp);
1532 	rte_free(bp->mark_table);
1533 	bp->mark_table = NULL;
1534 
1535 	bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
1536 	bp->rx_cosq_cnt = 0;
1537 	/* All filters are deleted on a port stop. */
1538 	if (BNXT_FLOW_XSTATS_EN(bp))
1539 		bp->flow_stat->flow_count = 0;
1540 
1541 	return 0;
1542 }
1543 
1544 /* Unload the driver, release resources */
1545 static int bnxt_dev_stop_op(struct rte_eth_dev *eth_dev)
1546 {
1547 	struct bnxt *bp = eth_dev->data->dev_private;
1548 
1549 	pthread_mutex_lock(&bp->err_recovery_lock);
1550 	if (bp->flags & BNXT_FLAG_FW_RESET) {
1551 		PMD_DRV_LOG(ERR,
1552 			    "Adapter recovering from error..Please retry\n");
1553 		pthread_mutex_unlock(&bp->err_recovery_lock);
1554 		return -EAGAIN;
1555 	}
1556 	pthread_mutex_unlock(&bp->err_recovery_lock);
1557 
1558 	return bnxt_dev_stop(eth_dev);
1559 }
1560 
1561 static int bnxt_dev_start_op(struct rte_eth_dev *eth_dev)
1562 {
1563 	struct bnxt *bp = eth_dev->data->dev_private;
1564 	uint64_t rx_offloads = eth_dev->data->dev_conf.rxmode.offloads;
1565 	int vlan_mask = 0;
1566 	int rc, retry_cnt = BNXT_IF_CHANGE_RETRY_COUNT;
1567 
1568 	if (!eth_dev->data->nb_tx_queues || !eth_dev->data->nb_rx_queues) {
1569 		PMD_DRV_LOG(ERR, "Queues are not configured yet!\n");
1570 		return -EINVAL;
1571 	}
1572 
1573 	if (bp->rx_cp_nr_rings > RTE_ETHDEV_QUEUE_STAT_CNTRS)
1574 		PMD_DRV_LOG(ERR,
1575 			    "RxQ cnt %d > RTE_ETHDEV_QUEUE_STAT_CNTRS %d\n",
1576 			    bp->rx_cp_nr_rings, RTE_ETHDEV_QUEUE_STAT_CNTRS);
1577 
1578 	do {
1579 		rc = bnxt_hwrm_if_change(bp, true);
1580 		if (rc == 0 || rc != -EAGAIN)
1581 			break;
1582 
1583 		rte_delay_ms(BNXT_IF_CHANGE_RETRY_INTERVAL);
1584 	} while (retry_cnt--);
1585 
1586 	if (rc)
1587 		return rc;
1588 
1589 	if (bp->flags & BNXT_FLAG_IF_CHANGE_HOT_FW_RESET_DONE) {
1590 		rc = bnxt_handle_if_change_status(bp);
1591 		if (rc)
1592 			return rc;
1593 	}
1594 
1595 	bnxt_enable_int(bp);
1596 
1597 	eth_dev->data->scattered_rx = bnxt_scattered_rx(eth_dev);
1598 
1599 	rc = bnxt_start_nic(bp);
1600 	if (rc)
1601 		goto error;
1602 
1603 	rc = bnxt_alloc_prev_ring_stats(bp);
1604 	if (rc)
1605 		goto error;
1606 
1607 	eth_dev->data->dev_started = 1;
1608 
1609 	bnxt_link_update_op(eth_dev, 1);
1610 
1611 	if (rx_offloads & RTE_ETH_RX_OFFLOAD_VLAN_FILTER)
1612 		vlan_mask |= RTE_ETH_VLAN_FILTER_MASK;
1613 	if (rx_offloads & RTE_ETH_RX_OFFLOAD_VLAN_STRIP)
1614 		vlan_mask |= RTE_ETH_VLAN_STRIP_MASK;
1615 	rc = bnxt_vlan_offload_set_op(eth_dev, vlan_mask);
1616 	if (rc)
1617 		goto error;
1618 
1619 	/* Initialize bnxt ULP port details */
1620 	rc = bnxt_ulp_port_init(bp);
1621 	if (rc)
1622 		goto error;
1623 
1624 	eth_dev->rx_pkt_burst = bnxt_receive_function(eth_dev);
1625 	eth_dev->tx_pkt_burst = bnxt_transmit_function(eth_dev);
1626 
1627 	bnxt_schedule_fw_health_check(bp);
1628 
1629 	if (BNXT_P5_PTP_TIMESYNC_ENABLED(bp))
1630 		bnxt_schedule_ptp_alarm(bp);
1631 
1632 	return 0;
1633 
1634 error:
1635 	bnxt_dev_stop(eth_dev);
1636 	return rc;
1637 }
1638 
1639 static void
1640 bnxt_uninit_locks(struct bnxt *bp)
1641 {
1642 	pthread_mutex_destroy(&bp->flow_lock);
1643 	pthread_mutex_destroy(&bp->def_cp_lock);
1644 	pthread_mutex_destroy(&bp->health_check_lock);
1645 	pthread_mutex_destroy(&bp->err_recovery_lock);
1646 	if (bp->rep_info) {
1647 		pthread_mutex_destroy(&bp->rep_info->vfr_lock);
1648 		pthread_mutex_destroy(&bp->rep_info->vfr_start_lock);
1649 	}
1650 }
1651 
1652 static void bnxt_drv_uninit(struct bnxt *bp)
1653 {
1654 	bnxt_free_leds_info(bp);
1655 	bnxt_free_cos_queues(bp);
1656 	bnxt_free_link_info(bp);
1657 	bnxt_free_parent_info(bp);
1658 	bnxt_uninit_locks(bp);
1659 
1660 	rte_memzone_free((const struct rte_memzone *)bp->tx_mem_zone);
1661 	bp->tx_mem_zone = NULL;
1662 	rte_memzone_free((const struct rte_memzone *)bp->rx_mem_zone);
1663 	bp->rx_mem_zone = NULL;
1664 
1665 	bnxt_free_vf_info(bp);
1666 	bnxt_free_pf_info(bp);
1667 
1668 	rte_free(bp->grp_info);
1669 	bp->grp_info = NULL;
1670 }
1671 
1672 static int bnxt_dev_close_op(struct rte_eth_dev *eth_dev)
1673 {
1674 	struct bnxt *bp = eth_dev->data->dev_private;
1675 	int ret = 0;
1676 
1677 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
1678 		return 0;
1679 
1680 	pthread_mutex_lock(&bp->err_recovery_lock);
1681 	if (bp->flags & BNXT_FLAG_FW_RESET) {
1682 		PMD_DRV_LOG(ERR,
1683 			    "Adapter recovering from error...Please retry\n");
1684 		pthread_mutex_unlock(&bp->err_recovery_lock);
1685 		return -EAGAIN;
1686 	}
1687 	pthread_mutex_unlock(&bp->err_recovery_lock);
1688 
1689 	/* cancel the recovery handler before remove dev */
1690 	rte_eal_alarm_cancel(bnxt_dev_reset_and_resume, (void *)bp);
1691 	rte_eal_alarm_cancel(bnxt_dev_recover, (void *)bp);
1692 	bnxt_cancel_fc_thread(bp);
1693 
1694 	if (eth_dev->data->dev_started)
1695 		ret = bnxt_dev_stop(eth_dev);
1696 
1697 	bnxt_uninit_resources(bp, false);
1698 
1699 	bnxt_drv_uninit(bp);
1700 
1701 	return ret;
1702 }
1703 
1704 static void bnxt_mac_addr_remove_op(struct rte_eth_dev *eth_dev,
1705 				    uint32_t index)
1706 {
1707 	struct bnxt *bp = eth_dev->data->dev_private;
1708 	uint64_t pool_mask = eth_dev->data->mac_pool_sel[index];
1709 	struct bnxt_vnic_info *vnic;
1710 	struct bnxt_filter_info *filter, *temp_filter;
1711 	uint32_t i;
1712 
1713 	if (is_bnxt_in_error(bp))
1714 		return;
1715 
1716 	/*
1717 	 * Loop through all VNICs from the specified filter flow pools to
1718 	 * remove the corresponding MAC addr filter
1719 	 */
1720 	for (i = 0; i < bp->nr_vnics; i++) {
1721 		if (!(pool_mask & (1ULL << i)))
1722 			continue;
1723 
1724 		vnic = &bp->vnic_info[i];
1725 		filter = STAILQ_FIRST(&vnic->filter);
1726 		while (filter) {
1727 			temp_filter = STAILQ_NEXT(filter, next);
1728 			if (filter->mac_index == index) {
1729 				STAILQ_REMOVE(&vnic->filter, filter,
1730 						bnxt_filter_info, next);
1731 				bnxt_hwrm_clear_l2_filter(bp, filter);
1732 				bnxt_free_filter(bp, filter);
1733 			}
1734 			filter = temp_filter;
1735 		}
1736 	}
1737 }
1738 
1739 static int bnxt_add_mac_filter(struct bnxt *bp, struct bnxt_vnic_info *vnic,
1740 			       struct rte_ether_addr *mac_addr, uint32_t index,
1741 			       uint32_t pool)
1742 {
1743 	struct bnxt_filter_info *filter;
1744 	int rc = 0;
1745 
1746 	/* Attach requested MAC address to the new l2_filter */
1747 	STAILQ_FOREACH(filter, &vnic->filter, next) {
1748 		if (filter->mac_index == index) {
1749 			PMD_DRV_LOG(DEBUG,
1750 				    "MAC addr already existed for pool %d\n",
1751 				    pool);
1752 			return 0;
1753 		}
1754 	}
1755 
1756 	filter = bnxt_alloc_filter(bp);
1757 	if (!filter) {
1758 		PMD_DRV_LOG(ERR, "L2 filter alloc failed\n");
1759 		return -ENODEV;
1760 	}
1761 
1762 	/* bnxt_alloc_filter copies default MAC to filter->l2_addr. So,
1763 	 * if the MAC that's been programmed now is a different one, then,
1764 	 * copy that addr to filter->l2_addr
1765 	 */
1766 	if (mac_addr)
1767 		memcpy(filter->l2_addr, mac_addr, RTE_ETHER_ADDR_LEN);
1768 	filter->flags |= HWRM_CFA_L2_FILTER_ALLOC_INPUT_FLAGS_OUTERMOST;
1769 
1770 	rc = bnxt_hwrm_set_l2_filter(bp, vnic->fw_vnic_id, filter);
1771 	if (!rc) {
1772 		filter->mac_index = index;
1773 		if (filter->mac_index == 0)
1774 			STAILQ_INSERT_HEAD(&vnic->filter, filter, next);
1775 		else
1776 			STAILQ_INSERT_TAIL(&vnic->filter, filter, next);
1777 	} else {
1778 		bnxt_free_filter(bp, filter);
1779 	}
1780 
1781 	return rc;
1782 }
1783 
1784 static int bnxt_mac_addr_add_op(struct rte_eth_dev *eth_dev,
1785 				struct rte_ether_addr *mac_addr,
1786 				uint32_t index, uint32_t pool)
1787 {
1788 	struct bnxt *bp = eth_dev->data->dev_private;
1789 	struct bnxt_vnic_info *vnic = &bp->vnic_info[pool];
1790 	int rc = 0;
1791 
1792 	rc = is_bnxt_in_error(bp);
1793 	if (rc)
1794 		return rc;
1795 
1796 	if (BNXT_VF(bp) && !BNXT_VF_IS_TRUSTED(bp)) {
1797 		PMD_DRV_LOG(ERR, "Cannot add MAC address to a VF interface\n");
1798 		return -ENOTSUP;
1799 	}
1800 
1801 	if (!vnic) {
1802 		PMD_DRV_LOG(ERR, "VNIC not found for pool %d!\n", pool);
1803 		return -EINVAL;
1804 	}
1805 
1806 	/* Filter settings will get applied when port is started */
1807 	if (!eth_dev->data->dev_started)
1808 		return 0;
1809 
1810 	rc = bnxt_add_mac_filter(bp, vnic, mac_addr, index, pool);
1811 
1812 	return rc;
1813 }
1814 
1815 int bnxt_link_update_op(struct rte_eth_dev *eth_dev, int wait_to_complete)
1816 {
1817 	int rc = 0;
1818 	struct bnxt *bp = eth_dev->data->dev_private;
1819 	struct rte_eth_link new;
1820 	int cnt = wait_to_complete ? BNXT_MAX_LINK_WAIT_CNT :
1821 			BNXT_MIN_LINK_WAIT_CNT;
1822 
1823 	rc = is_bnxt_in_error(bp);
1824 	if (rc)
1825 		return rc;
1826 
1827 	memset(&new, 0, sizeof(new));
1828 
1829 	if (bp->link_info == NULL)
1830 		goto out;
1831 
1832 	do {
1833 		/* Retrieve link info from hardware */
1834 		rc = bnxt_get_hwrm_link_config(bp, &new);
1835 		if (rc) {
1836 			new.link_speed = RTE_ETH_LINK_SPEED_100M;
1837 			new.link_duplex = RTE_ETH_LINK_FULL_DUPLEX;
1838 			PMD_DRV_LOG(ERR,
1839 				"Failed to retrieve link rc = 0x%x!\n", rc);
1840 			goto out;
1841 		}
1842 
1843 		if (!wait_to_complete || new.link_status)
1844 			break;
1845 
1846 		rte_delay_ms(BNXT_LINK_WAIT_INTERVAL);
1847 	} while (cnt--);
1848 
1849 	/* Only single function PF can bring phy down.
1850 	 * When port is stopped, report link down for VF/MH/NPAR functions.
1851 	 */
1852 	if (!BNXT_SINGLE_PF(bp) && !eth_dev->data->dev_started)
1853 		memset(&new, 0, sizeof(new));
1854 
1855 out:
1856 	/* Timed out or success */
1857 	if (new.link_status != eth_dev->data->dev_link.link_status ||
1858 	    new.link_speed != eth_dev->data->dev_link.link_speed) {
1859 		rte_eth_linkstatus_set(eth_dev, &new);
1860 		bnxt_print_link_info(eth_dev);
1861 	}
1862 
1863 	return rc;
1864 }
1865 
1866 static int bnxt_promiscuous_enable_op(struct rte_eth_dev *eth_dev)
1867 {
1868 	struct bnxt *bp = eth_dev->data->dev_private;
1869 	struct bnxt_vnic_info *vnic;
1870 	uint32_t old_flags;
1871 	int rc;
1872 
1873 	rc = is_bnxt_in_error(bp);
1874 	if (rc)
1875 		return rc;
1876 
1877 	/* Filter settings will get applied when port is started */
1878 	if (!eth_dev->data->dev_started)
1879 		return 0;
1880 
1881 	if (bp->vnic_info == NULL)
1882 		return 0;
1883 
1884 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
1885 
1886 	old_flags = vnic->flags;
1887 	vnic->flags |= BNXT_VNIC_INFO_PROMISC;
1888 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1889 	if (rc != 0)
1890 		vnic->flags = old_flags;
1891 
1892 	return rc;
1893 }
1894 
1895 static int bnxt_promiscuous_disable_op(struct rte_eth_dev *eth_dev)
1896 {
1897 	struct bnxt *bp = eth_dev->data->dev_private;
1898 	struct bnxt_vnic_info *vnic;
1899 	uint32_t old_flags;
1900 	int rc;
1901 
1902 	rc = is_bnxt_in_error(bp);
1903 	if (rc)
1904 		return rc;
1905 
1906 	/* Filter settings will get applied when port is started */
1907 	if (!eth_dev->data->dev_started)
1908 		return 0;
1909 
1910 	if (bp->vnic_info == NULL)
1911 		return 0;
1912 
1913 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
1914 
1915 	old_flags = vnic->flags;
1916 	vnic->flags &= ~BNXT_VNIC_INFO_PROMISC;
1917 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1918 	if (rc != 0)
1919 		vnic->flags = old_flags;
1920 
1921 	return rc;
1922 }
1923 
1924 static int bnxt_allmulticast_enable_op(struct rte_eth_dev *eth_dev)
1925 {
1926 	struct bnxt *bp = eth_dev->data->dev_private;
1927 	struct bnxt_vnic_info *vnic;
1928 	uint32_t old_flags;
1929 	int rc;
1930 
1931 	rc = is_bnxt_in_error(bp);
1932 	if (rc)
1933 		return rc;
1934 
1935 	/* Filter settings will get applied when port is started */
1936 	if (!eth_dev->data->dev_started)
1937 		return 0;
1938 
1939 	if (bp->vnic_info == NULL)
1940 		return 0;
1941 
1942 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
1943 
1944 	old_flags = vnic->flags;
1945 	vnic->flags |= BNXT_VNIC_INFO_ALLMULTI;
1946 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1947 	if (rc != 0)
1948 		vnic->flags = old_flags;
1949 
1950 	return rc;
1951 }
1952 
1953 static int bnxt_allmulticast_disable_op(struct rte_eth_dev *eth_dev)
1954 {
1955 	struct bnxt *bp = eth_dev->data->dev_private;
1956 	struct bnxt_vnic_info *vnic;
1957 	uint32_t old_flags;
1958 	int rc;
1959 
1960 	rc = is_bnxt_in_error(bp);
1961 	if (rc)
1962 		return rc;
1963 
1964 	/* Filter settings will get applied when port is started */
1965 	if (!eth_dev->data->dev_started)
1966 		return 0;
1967 
1968 	if (bp->vnic_info == NULL)
1969 		return 0;
1970 
1971 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
1972 
1973 	old_flags = vnic->flags;
1974 	vnic->flags &= ~BNXT_VNIC_INFO_ALLMULTI;
1975 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1976 	if (rc != 0)
1977 		vnic->flags = old_flags;
1978 
1979 	return rc;
1980 }
1981 
1982 /* Return bnxt_rx_queue pointer corresponding to a given rxq. */
1983 static struct bnxt_rx_queue *bnxt_qid_to_rxq(struct bnxt *bp, uint16_t qid)
1984 {
1985 	if (qid >= bp->rx_nr_rings)
1986 		return NULL;
1987 
1988 	return bp->eth_dev->data->rx_queues[qid];
1989 }
1990 
1991 /* Return rxq corresponding to a given rss table ring/group ID. */
1992 static uint16_t bnxt_rss_to_qid(struct bnxt *bp, uint16_t fwr)
1993 {
1994 	struct bnxt_rx_queue *rxq;
1995 	unsigned int i;
1996 
1997 	if (!BNXT_HAS_RING_GRPS(bp)) {
1998 		for (i = 0; i < bp->rx_nr_rings; i++) {
1999 			rxq = bp->eth_dev->data->rx_queues[i];
2000 			if (rxq->rx_ring->rx_ring_struct->fw_ring_id == fwr)
2001 				return rxq->index;
2002 		}
2003 	} else {
2004 		for (i = 0; i < bp->rx_nr_rings; i++) {
2005 			if (bp->grp_info[i].fw_grp_id == fwr)
2006 				return i;
2007 		}
2008 	}
2009 
2010 	return INVALID_HW_RING_ID;
2011 }
2012 
2013 static int bnxt_reta_update_op(struct rte_eth_dev *eth_dev,
2014 			    struct rte_eth_rss_reta_entry64 *reta_conf,
2015 			    uint16_t reta_size)
2016 {
2017 	struct bnxt *bp = eth_dev->data->dev_private;
2018 	struct rte_eth_conf *dev_conf = &bp->eth_dev->data->dev_conf;
2019 	struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
2020 	uint16_t tbl_size = bnxt_rss_hash_tbl_size(bp);
2021 	uint16_t idx, sft;
2022 	int i, rc;
2023 
2024 	rc = is_bnxt_in_error(bp);
2025 	if (rc)
2026 		return rc;
2027 
2028 	if (!vnic->rss_table)
2029 		return -EINVAL;
2030 
2031 	if (!(dev_conf->rxmode.mq_mode & RTE_ETH_MQ_RX_RSS_FLAG))
2032 		return -EINVAL;
2033 
2034 	if (reta_size != tbl_size) {
2035 		PMD_DRV_LOG(ERR, "The configured hash table lookup size "
2036 			"(%d) must equal the size supported by the hardware "
2037 			"(%d)\n", reta_size, tbl_size);
2038 		return -EINVAL;
2039 	}
2040 
2041 	for (i = 0; i < reta_size; i++) {
2042 		struct bnxt_rx_queue *rxq;
2043 
2044 		idx = i / RTE_ETH_RETA_GROUP_SIZE;
2045 		sft = i % RTE_ETH_RETA_GROUP_SIZE;
2046 
2047 		if (!(reta_conf[idx].mask & (1ULL << sft)))
2048 			continue;
2049 
2050 		rxq = bnxt_qid_to_rxq(bp, reta_conf[idx].reta[sft]);
2051 		if (!rxq) {
2052 			PMD_DRV_LOG(ERR, "Invalid ring in reta_conf.\n");
2053 			return -EINVAL;
2054 		}
2055 
2056 		if (BNXT_CHIP_P5(bp)) {
2057 			vnic->rss_table[i * 2] =
2058 				rxq->rx_ring->rx_ring_struct->fw_ring_id;
2059 			vnic->rss_table[i * 2 + 1] =
2060 				rxq->cp_ring->cp_ring_struct->fw_ring_id;
2061 		} else {
2062 			vnic->rss_table[i] =
2063 			    vnic->fw_grp_ids[reta_conf[idx].reta[sft]];
2064 		}
2065 	}
2066 
2067 	rc = bnxt_hwrm_vnic_rss_cfg(bp, vnic);
2068 	return rc;
2069 }
2070 
2071 static int bnxt_reta_query_op(struct rte_eth_dev *eth_dev,
2072 			      struct rte_eth_rss_reta_entry64 *reta_conf,
2073 			      uint16_t reta_size)
2074 {
2075 	struct bnxt *bp = eth_dev->data->dev_private;
2076 	struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
2077 	uint16_t tbl_size = bnxt_rss_hash_tbl_size(bp);
2078 	uint16_t idx, sft, i;
2079 	int rc;
2080 
2081 	rc = is_bnxt_in_error(bp);
2082 	if (rc)
2083 		return rc;
2084 
2085 	if (!vnic)
2086 		return -EINVAL;
2087 	if (!vnic->rss_table)
2088 		return -EINVAL;
2089 
2090 	if (reta_size != tbl_size) {
2091 		PMD_DRV_LOG(ERR, "The configured hash table lookup size "
2092 			"(%d) must equal the size supported by the hardware "
2093 			"(%d)\n", reta_size, tbl_size);
2094 		return -EINVAL;
2095 	}
2096 
2097 	for (idx = 0, i = 0; i < reta_size; i++) {
2098 		idx = i / RTE_ETH_RETA_GROUP_SIZE;
2099 		sft = i % RTE_ETH_RETA_GROUP_SIZE;
2100 
2101 		if (reta_conf[idx].mask & (1ULL << sft)) {
2102 			uint16_t qid;
2103 
2104 			if (BNXT_CHIP_P5(bp))
2105 				qid = bnxt_rss_to_qid(bp,
2106 						      vnic->rss_table[i * 2]);
2107 			else
2108 				qid = bnxt_rss_to_qid(bp, vnic->rss_table[i]);
2109 
2110 			if (qid == INVALID_HW_RING_ID) {
2111 				PMD_DRV_LOG(ERR, "Inv. entry in rss table.\n");
2112 				return -EINVAL;
2113 			}
2114 			reta_conf[idx].reta[sft] = qid;
2115 		}
2116 	}
2117 
2118 	return 0;
2119 }
2120 
2121 static int bnxt_rss_hash_update_op(struct rte_eth_dev *eth_dev,
2122 				   struct rte_eth_rss_conf *rss_conf)
2123 {
2124 	struct bnxt *bp = eth_dev->data->dev_private;
2125 	struct rte_eth_conf *dev_conf = &bp->eth_dev->data->dev_conf;
2126 	struct bnxt_vnic_info *vnic;
2127 	int rc;
2128 
2129 	rc = is_bnxt_in_error(bp);
2130 	if (rc)
2131 		return rc;
2132 
2133 	/*
2134 	 * If RSS enablement were different than dev_configure,
2135 	 * then return -EINVAL
2136 	 */
2137 	if (dev_conf->rxmode.mq_mode & RTE_ETH_MQ_RX_RSS_FLAG) {
2138 		if (!rss_conf->rss_hf)
2139 			PMD_DRV_LOG(ERR, "Hash type NONE\n");
2140 	} else {
2141 		if (rss_conf->rss_hf & BNXT_ETH_RSS_SUPPORT)
2142 			return -EINVAL;
2143 	}
2144 
2145 	bp->flags |= BNXT_FLAG_UPDATE_HASH;
2146 	memcpy(&eth_dev->data->dev_conf.rx_adv_conf.rss_conf,
2147 	       rss_conf,
2148 	       sizeof(*rss_conf));
2149 
2150 	/* Update the default RSS VNIC(s) */
2151 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
2152 	vnic->hash_type = bnxt_rte_to_hwrm_hash_types(rss_conf->rss_hf);
2153 	vnic->hash_mode =
2154 		bnxt_rte_to_hwrm_hash_level(bp, rss_conf->rss_hf,
2155 					    RTE_ETH_RSS_LEVEL(rss_conf->rss_hf));
2156 
2157 	/*
2158 	 * If hashkey is not specified, use the previously configured
2159 	 * hashkey
2160 	 */
2161 	if (!rss_conf->rss_key)
2162 		goto rss_config;
2163 
2164 	if (rss_conf->rss_key_len != HW_HASH_KEY_SIZE) {
2165 		PMD_DRV_LOG(ERR,
2166 			    "Invalid hashkey length, should be %d bytes\n",
2167 			    HW_HASH_KEY_SIZE);
2168 		return -EINVAL;
2169 	}
2170 	memcpy(vnic->rss_hash_key, rss_conf->rss_key, rss_conf->rss_key_len);
2171 
2172 rss_config:
2173 	rc = bnxt_hwrm_vnic_rss_cfg(bp, vnic);
2174 	return rc;
2175 }
2176 
2177 static int bnxt_rss_hash_conf_get_op(struct rte_eth_dev *eth_dev,
2178 				     struct rte_eth_rss_conf *rss_conf)
2179 {
2180 	struct bnxt *bp = eth_dev->data->dev_private;
2181 	struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
2182 	int len, rc;
2183 	uint32_t hash_types;
2184 
2185 	rc = is_bnxt_in_error(bp);
2186 	if (rc)
2187 		return rc;
2188 
2189 	/* RSS configuration is the same for all VNICs */
2190 	if (vnic && vnic->rss_hash_key) {
2191 		if (rss_conf->rss_key) {
2192 			len = rss_conf->rss_key_len <= HW_HASH_KEY_SIZE ?
2193 			      rss_conf->rss_key_len : HW_HASH_KEY_SIZE;
2194 			memcpy(rss_conf->rss_key, vnic->rss_hash_key, len);
2195 		}
2196 
2197 		hash_types = vnic->hash_type;
2198 		rss_conf->rss_hf = 0;
2199 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV4) {
2200 			rss_conf->rss_hf |= RTE_ETH_RSS_IPV4;
2201 			hash_types &= ~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV4;
2202 		}
2203 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV4) {
2204 			rss_conf->rss_hf |= RTE_ETH_RSS_NONFRAG_IPV4_TCP;
2205 			hash_types &=
2206 				~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV4;
2207 		}
2208 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV4) {
2209 			rss_conf->rss_hf |= RTE_ETH_RSS_NONFRAG_IPV4_UDP;
2210 			hash_types &=
2211 				~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV4;
2212 		}
2213 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV6) {
2214 			rss_conf->rss_hf |= RTE_ETH_RSS_IPV6;
2215 			hash_types &= ~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV6;
2216 		}
2217 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV6) {
2218 			rss_conf->rss_hf |= RTE_ETH_RSS_NONFRAG_IPV6_TCP;
2219 			hash_types &=
2220 				~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV6;
2221 		}
2222 		if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV6) {
2223 			rss_conf->rss_hf |= RTE_ETH_RSS_NONFRAG_IPV6_UDP;
2224 			hash_types &=
2225 				~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV6;
2226 		}
2227 
2228 		rss_conf->rss_hf |=
2229 			bnxt_hwrm_to_rte_rss_level(bp, vnic->hash_mode);
2230 
2231 		if (hash_types) {
2232 			PMD_DRV_LOG(ERR,
2233 				"Unknown RSS config from firmware (%08x), RSS disabled",
2234 				vnic->hash_type);
2235 			return -ENOTSUP;
2236 		}
2237 	} else {
2238 		rss_conf->rss_hf = 0;
2239 	}
2240 	return 0;
2241 }
2242 
2243 static int bnxt_flow_ctrl_get_op(struct rte_eth_dev *dev,
2244 			       struct rte_eth_fc_conf *fc_conf)
2245 {
2246 	struct bnxt *bp = dev->data->dev_private;
2247 	struct rte_eth_link link_info;
2248 	int rc;
2249 
2250 	rc = is_bnxt_in_error(bp);
2251 	if (rc)
2252 		return rc;
2253 
2254 	rc = bnxt_get_hwrm_link_config(bp, &link_info);
2255 	if (rc)
2256 		return rc;
2257 
2258 	memset(fc_conf, 0, sizeof(*fc_conf));
2259 	if (bp->link_info->auto_pause)
2260 		fc_conf->autoneg = 1;
2261 	switch (bp->link_info->pause) {
2262 	case 0:
2263 		fc_conf->mode = RTE_ETH_FC_NONE;
2264 		break;
2265 	case HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_TX:
2266 		fc_conf->mode = RTE_ETH_FC_TX_PAUSE;
2267 		break;
2268 	case HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_RX:
2269 		fc_conf->mode = RTE_ETH_FC_RX_PAUSE;
2270 		break;
2271 	case (HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_TX |
2272 			HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_RX):
2273 		fc_conf->mode = RTE_ETH_FC_FULL;
2274 		break;
2275 	}
2276 	return 0;
2277 }
2278 
2279 static int bnxt_flow_ctrl_set_op(struct rte_eth_dev *dev,
2280 			       struct rte_eth_fc_conf *fc_conf)
2281 {
2282 	struct bnxt *bp = dev->data->dev_private;
2283 	int rc;
2284 
2285 	rc = is_bnxt_in_error(bp);
2286 	if (rc)
2287 		return rc;
2288 
2289 	if (!BNXT_SINGLE_PF(bp)) {
2290 		PMD_DRV_LOG(ERR,
2291 			    "Flow Control Settings cannot be modified on VF or on shared PF\n");
2292 		return -ENOTSUP;
2293 	}
2294 
2295 	switch (fc_conf->mode) {
2296 	case RTE_ETH_FC_NONE:
2297 		bp->link_info->auto_pause = 0;
2298 		bp->link_info->force_pause = 0;
2299 		break;
2300 	case RTE_ETH_FC_RX_PAUSE:
2301 		if (fc_conf->autoneg) {
2302 			bp->link_info->auto_pause =
2303 					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_RX;
2304 			bp->link_info->force_pause = 0;
2305 		} else {
2306 			bp->link_info->auto_pause = 0;
2307 			bp->link_info->force_pause =
2308 					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_RX;
2309 		}
2310 		break;
2311 	case RTE_ETH_FC_TX_PAUSE:
2312 		if (fc_conf->autoneg) {
2313 			bp->link_info->auto_pause =
2314 					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_TX;
2315 			bp->link_info->force_pause = 0;
2316 		} else {
2317 			bp->link_info->auto_pause = 0;
2318 			bp->link_info->force_pause =
2319 					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_TX;
2320 		}
2321 		break;
2322 	case RTE_ETH_FC_FULL:
2323 		if (fc_conf->autoneg) {
2324 			bp->link_info->auto_pause =
2325 					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_TX |
2326 					HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_RX;
2327 			bp->link_info->force_pause = 0;
2328 		} else {
2329 			bp->link_info->auto_pause = 0;
2330 			bp->link_info->force_pause =
2331 					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_TX |
2332 					HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_RX;
2333 		}
2334 		break;
2335 	}
2336 	return bnxt_set_hwrm_link_config(bp, true);
2337 }
2338 
2339 /* Add UDP tunneling port */
2340 static int
2341 bnxt_udp_tunnel_port_add_op(struct rte_eth_dev *eth_dev,
2342 			 struct rte_eth_udp_tunnel *udp_tunnel)
2343 {
2344 	struct bnxt *bp = eth_dev->data->dev_private;
2345 	uint16_t tunnel_type = 0;
2346 	int rc = 0;
2347 
2348 	rc = is_bnxt_in_error(bp);
2349 	if (rc)
2350 		return rc;
2351 
2352 	switch (udp_tunnel->prot_type) {
2353 	case RTE_ETH_TUNNEL_TYPE_VXLAN:
2354 		if (bp->vxlan_port_cnt) {
2355 			PMD_DRV_LOG(ERR, "Tunnel Port %d already programmed\n",
2356 				udp_tunnel->udp_port);
2357 			if (bp->vxlan_port != udp_tunnel->udp_port) {
2358 				PMD_DRV_LOG(ERR, "Only one port allowed\n");
2359 				return -ENOSPC;
2360 			}
2361 			bp->vxlan_port_cnt++;
2362 			return 0;
2363 		}
2364 		tunnel_type =
2365 			HWRM_TUNNEL_DST_PORT_ALLOC_INPUT_TUNNEL_TYPE_VXLAN;
2366 		break;
2367 	case RTE_ETH_TUNNEL_TYPE_GENEVE:
2368 		if (bp->geneve_port_cnt) {
2369 			PMD_DRV_LOG(ERR, "Tunnel Port %d already programmed\n",
2370 				udp_tunnel->udp_port);
2371 			if (bp->geneve_port != udp_tunnel->udp_port) {
2372 				PMD_DRV_LOG(ERR, "Only one port allowed\n");
2373 				return -ENOSPC;
2374 			}
2375 			bp->geneve_port_cnt++;
2376 			return 0;
2377 		}
2378 		tunnel_type =
2379 			HWRM_TUNNEL_DST_PORT_ALLOC_INPUT_TUNNEL_TYPE_GENEVE;
2380 		break;
2381 	default:
2382 		PMD_DRV_LOG(ERR, "Tunnel type is not supported\n");
2383 		return -ENOTSUP;
2384 	}
2385 	rc = bnxt_hwrm_tunnel_dst_port_alloc(bp, udp_tunnel->udp_port,
2386 					     tunnel_type);
2387 
2388 	if (rc != 0)
2389 		return rc;
2390 
2391 	if (tunnel_type ==
2392 	    HWRM_TUNNEL_DST_PORT_ALLOC_INPUT_TUNNEL_TYPE_VXLAN)
2393 		bp->vxlan_port_cnt++;
2394 
2395 	if (tunnel_type ==
2396 	    HWRM_TUNNEL_DST_PORT_ALLOC_INPUT_TUNNEL_TYPE_GENEVE)
2397 		bp->geneve_port_cnt++;
2398 
2399 	return rc;
2400 }
2401 
2402 static int
2403 bnxt_udp_tunnel_port_del_op(struct rte_eth_dev *eth_dev,
2404 			 struct rte_eth_udp_tunnel *udp_tunnel)
2405 {
2406 	struct bnxt *bp = eth_dev->data->dev_private;
2407 	uint16_t tunnel_type = 0;
2408 	uint16_t port = 0;
2409 	int rc = 0;
2410 
2411 	rc = is_bnxt_in_error(bp);
2412 	if (rc)
2413 		return rc;
2414 
2415 	switch (udp_tunnel->prot_type) {
2416 	case RTE_ETH_TUNNEL_TYPE_VXLAN:
2417 		if (!bp->vxlan_port_cnt) {
2418 			PMD_DRV_LOG(ERR, "No Tunnel port configured yet\n");
2419 			return -EINVAL;
2420 		}
2421 		if (bp->vxlan_port != udp_tunnel->udp_port) {
2422 			PMD_DRV_LOG(ERR, "Req Port: %d. Configured port: %d\n",
2423 				udp_tunnel->udp_port, bp->vxlan_port);
2424 			return -EINVAL;
2425 		}
2426 		if (--bp->vxlan_port_cnt)
2427 			return 0;
2428 
2429 		tunnel_type =
2430 			HWRM_TUNNEL_DST_PORT_FREE_INPUT_TUNNEL_TYPE_VXLAN;
2431 		port = bp->vxlan_fw_dst_port_id;
2432 		break;
2433 	case RTE_ETH_TUNNEL_TYPE_GENEVE:
2434 		if (!bp->geneve_port_cnt) {
2435 			PMD_DRV_LOG(ERR, "No Tunnel port configured yet\n");
2436 			return -EINVAL;
2437 		}
2438 		if (bp->geneve_port != udp_tunnel->udp_port) {
2439 			PMD_DRV_LOG(ERR, "Req Port: %d. Configured port: %d\n",
2440 				udp_tunnel->udp_port, bp->geneve_port);
2441 			return -EINVAL;
2442 		}
2443 		if (--bp->geneve_port_cnt)
2444 			return 0;
2445 
2446 		tunnel_type =
2447 			HWRM_TUNNEL_DST_PORT_FREE_INPUT_TUNNEL_TYPE_GENEVE;
2448 		port = bp->geneve_fw_dst_port_id;
2449 		break;
2450 	default:
2451 		PMD_DRV_LOG(ERR, "Tunnel type is not supported\n");
2452 		return -ENOTSUP;
2453 	}
2454 
2455 	rc = bnxt_hwrm_tunnel_dst_port_free(bp, port, tunnel_type);
2456 	return rc;
2457 }
2458 
2459 static int bnxt_del_vlan_filter(struct bnxt *bp, uint16_t vlan_id)
2460 {
2461 	struct bnxt_filter_info *filter;
2462 	struct bnxt_vnic_info *vnic;
2463 	int rc = 0;
2464 	uint32_t chk = HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN;
2465 
2466 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
2467 	filter = STAILQ_FIRST(&vnic->filter);
2468 	while (filter) {
2469 		/* Search for this matching MAC+VLAN filter */
2470 		if (bnxt_vlan_filter_exists(bp, filter, chk, vlan_id)) {
2471 			/* Delete the filter */
2472 			rc = bnxt_hwrm_clear_l2_filter(bp, filter);
2473 			if (rc)
2474 				return rc;
2475 			STAILQ_REMOVE(&vnic->filter, filter,
2476 				      bnxt_filter_info, next);
2477 			bnxt_free_filter(bp, filter);
2478 			PMD_DRV_LOG(INFO,
2479 				    "Deleted vlan filter for %d\n",
2480 				    vlan_id);
2481 			return 0;
2482 		}
2483 		filter = STAILQ_NEXT(filter, next);
2484 	}
2485 	return -ENOENT;
2486 }
2487 
2488 static int bnxt_add_vlan_filter(struct bnxt *bp, uint16_t vlan_id)
2489 {
2490 	struct bnxt_filter_info *filter;
2491 	struct bnxt_vnic_info *vnic;
2492 	int rc = 0;
2493 	uint32_t en = HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN |
2494 		HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN_MASK;
2495 	uint32_t chk = HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN;
2496 
2497 	/* Implementation notes on the use of VNIC in this command:
2498 	 *
2499 	 * By default, these filters belong to default vnic for the function.
2500 	 * Once these filters are set up, only destination VNIC can be modified.
2501 	 * If the destination VNIC is not specified in this command,
2502 	 * then the HWRM shall only create an l2 context id.
2503 	 */
2504 
2505 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
2506 	filter = STAILQ_FIRST(&vnic->filter);
2507 	/* Check if the VLAN has already been added */
2508 	while (filter) {
2509 		if (bnxt_vlan_filter_exists(bp, filter, chk, vlan_id))
2510 			return -EEXIST;
2511 
2512 		filter = STAILQ_NEXT(filter, next);
2513 	}
2514 
2515 	/* No match found. Alloc a fresh filter and issue the L2_FILTER_ALLOC
2516 	 * command to create MAC+VLAN filter with the right flags, enables set.
2517 	 */
2518 	filter = bnxt_alloc_filter(bp);
2519 	if (!filter) {
2520 		PMD_DRV_LOG(ERR,
2521 			    "MAC/VLAN filter alloc failed\n");
2522 		return -ENOMEM;
2523 	}
2524 	/* MAC + VLAN ID filter */
2525 	/* If l2_ivlan == 0 and l2_ivlan_mask != 0, only
2526 	 * untagged packets are received
2527 	 *
2528 	 * If l2_ivlan != 0 and l2_ivlan_mask != 0, untagged
2529 	 * packets and only the programmed vlan's packets are received
2530 	 */
2531 	filter->l2_ivlan = vlan_id;
2532 	filter->l2_ivlan_mask = 0x0FFF;
2533 	filter->enables |= en;
2534 	filter->flags |= HWRM_CFA_L2_FILTER_ALLOC_INPUT_FLAGS_OUTERMOST;
2535 
2536 	rc = bnxt_hwrm_set_l2_filter(bp, vnic->fw_vnic_id, filter);
2537 	if (rc) {
2538 		/* Free the newly allocated filter as we were
2539 		 * not able to create the filter in hardware.
2540 		 */
2541 		bnxt_free_filter(bp, filter);
2542 		return rc;
2543 	}
2544 
2545 	filter->mac_index = 0;
2546 	/* Add this new filter to the list */
2547 	if (vlan_id == 0)
2548 		STAILQ_INSERT_HEAD(&vnic->filter, filter, next);
2549 	else
2550 		STAILQ_INSERT_TAIL(&vnic->filter, filter, next);
2551 
2552 	PMD_DRV_LOG(INFO,
2553 		    "Added Vlan filter for %d\n", vlan_id);
2554 	return rc;
2555 }
2556 
2557 static int bnxt_vlan_filter_set_op(struct rte_eth_dev *eth_dev,
2558 		uint16_t vlan_id, int on)
2559 {
2560 	struct bnxt *bp = eth_dev->data->dev_private;
2561 	int rc;
2562 
2563 	rc = is_bnxt_in_error(bp);
2564 	if (rc)
2565 		return rc;
2566 
2567 	if (!eth_dev->data->dev_started) {
2568 		PMD_DRV_LOG(ERR, "port must be started before setting vlan\n");
2569 		return -EINVAL;
2570 	}
2571 
2572 	/* These operations apply to ALL existing MAC/VLAN filters */
2573 	if (on)
2574 		return bnxt_add_vlan_filter(bp, vlan_id);
2575 	else
2576 		return bnxt_del_vlan_filter(bp, vlan_id);
2577 }
2578 
2579 static int bnxt_del_dflt_mac_filter(struct bnxt *bp,
2580 				    struct bnxt_vnic_info *vnic)
2581 {
2582 	struct bnxt_filter_info *filter;
2583 	int rc;
2584 
2585 	filter = STAILQ_FIRST(&vnic->filter);
2586 	while (filter) {
2587 		if (filter->mac_index == 0 &&
2588 		    !memcmp(filter->l2_addr, bp->mac_addr,
2589 			    RTE_ETHER_ADDR_LEN)) {
2590 			rc = bnxt_hwrm_clear_l2_filter(bp, filter);
2591 			if (!rc) {
2592 				STAILQ_REMOVE(&vnic->filter, filter,
2593 					      bnxt_filter_info, next);
2594 				bnxt_free_filter(bp, filter);
2595 			}
2596 			return rc;
2597 		}
2598 		filter = STAILQ_NEXT(filter, next);
2599 	}
2600 	return 0;
2601 }
2602 
2603 static int
2604 bnxt_config_vlan_hw_filter(struct bnxt *bp, uint64_t rx_offloads)
2605 {
2606 	struct bnxt_vnic_info *vnic;
2607 	unsigned int i;
2608 	int rc;
2609 
2610 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
2611 	if (!(rx_offloads & RTE_ETH_RX_OFFLOAD_VLAN_FILTER)) {
2612 		/* Remove any VLAN filters programmed */
2613 		for (i = 0; i < RTE_ETHER_MAX_VLAN_ID; i++)
2614 			bnxt_del_vlan_filter(bp, i);
2615 
2616 		rc = bnxt_add_mac_filter(bp, vnic, NULL, 0, 0);
2617 		if (rc)
2618 			return rc;
2619 	} else {
2620 		/* Default filter will allow packets that match the
2621 		 * dest mac. So, it has to be deleted, otherwise, we
2622 		 * will endup receiving vlan packets for which the
2623 		 * filter is not programmed, when hw-vlan-filter
2624 		 * configuration is ON
2625 		 */
2626 		bnxt_del_dflt_mac_filter(bp, vnic);
2627 		/* This filter will allow only untagged packets */
2628 		bnxt_add_vlan_filter(bp, 0);
2629 	}
2630 	PMD_DRV_LOG(DEBUG, "VLAN Filtering: %d\n",
2631 		    !!(rx_offloads & RTE_ETH_RX_OFFLOAD_VLAN_FILTER));
2632 
2633 	return 0;
2634 }
2635 
2636 static int bnxt_free_one_vnic(struct bnxt *bp, uint16_t vnic_id)
2637 {
2638 	struct bnxt_vnic_info *vnic = &bp->vnic_info[vnic_id];
2639 	unsigned int i;
2640 	int rc;
2641 
2642 	/* Destroy vnic filters and vnic */
2643 	if (bp->eth_dev->data->dev_conf.rxmode.offloads &
2644 	    RTE_ETH_RX_OFFLOAD_VLAN_FILTER) {
2645 		for (i = 0; i < RTE_ETHER_MAX_VLAN_ID; i++)
2646 			bnxt_del_vlan_filter(bp, i);
2647 	}
2648 	bnxt_del_dflt_mac_filter(bp, vnic);
2649 
2650 	rc = bnxt_hwrm_vnic_ctx_free(bp, vnic);
2651 	if (rc)
2652 		return rc;
2653 
2654 	rc = bnxt_hwrm_vnic_free(bp, vnic);
2655 	if (rc)
2656 		return rc;
2657 
2658 	rte_free(vnic->fw_grp_ids);
2659 	vnic->fw_grp_ids = NULL;
2660 
2661 	vnic->rx_queue_cnt = 0;
2662 
2663 	return 0;
2664 }
2665 
2666 static int
2667 bnxt_config_vlan_hw_stripping(struct bnxt *bp, uint64_t rx_offloads)
2668 {
2669 	struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
2670 	int rc;
2671 
2672 	/* Destroy, recreate and reconfigure the default vnic */
2673 	rc = bnxt_free_one_vnic(bp, 0);
2674 	if (rc)
2675 		return rc;
2676 
2677 	/* default vnic 0 */
2678 	rc = bnxt_setup_one_vnic(bp, 0);
2679 	if (rc)
2680 		return rc;
2681 
2682 	if (bp->eth_dev->data->dev_conf.rxmode.offloads &
2683 	    RTE_ETH_RX_OFFLOAD_VLAN_FILTER) {
2684 		rc = bnxt_add_vlan_filter(bp, 0);
2685 		if (rc)
2686 			return rc;
2687 		rc = bnxt_restore_vlan_filters(bp);
2688 		if (rc)
2689 			return rc;
2690 	} else {
2691 		rc = bnxt_add_mac_filter(bp, vnic, NULL, 0, 0);
2692 		if (rc)
2693 			return rc;
2694 	}
2695 
2696 	rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
2697 	if (rc)
2698 		return rc;
2699 
2700 	PMD_DRV_LOG(DEBUG, "VLAN Strip Offload: %d\n",
2701 		    !!(rx_offloads & RTE_ETH_RX_OFFLOAD_VLAN_STRIP));
2702 
2703 	return rc;
2704 }
2705 
2706 static int
2707 bnxt_vlan_offload_set_op(struct rte_eth_dev *dev, int mask)
2708 {
2709 	uint64_t rx_offloads = dev->data->dev_conf.rxmode.offloads;
2710 	struct bnxt *bp = dev->data->dev_private;
2711 	int rc;
2712 
2713 	rc = is_bnxt_in_error(bp);
2714 	if (rc)
2715 		return rc;
2716 
2717 	/* Filter settings will get applied when port is started */
2718 	if (!dev->data->dev_started)
2719 		return 0;
2720 
2721 	if (mask & RTE_ETH_VLAN_FILTER_MASK) {
2722 		/* Enable or disable VLAN filtering */
2723 		rc = bnxt_config_vlan_hw_filter(bp, rx_offloads);
2724 		if (rc)
2725 			return rc;
2726 	}
2727 
2728 	if (mask & RTE_ETH_VLAN_STRIP_MASK) {
2729 		/* Enable or disable VLAN stripping */
2730 		rc = bnxt_config_vlan_hw_stripping(bp, rx_offloads);
2731 		if (rc)
2732 			return rc;
2733 	}
2734 
2735 	if (mask & RTE_ETH_VLAN_EXTEND_MASK) {
2736 		if (rx_offloads & RTE_ETH_RX_OFFLOAD_VLAN_EXTEND)
2737 			PMD_DRV_LOG(DEBUG, "Extend VLAN supported\n");
2738 		else
2739 			PMD_DRV_LOG(INFO, "Extend VLAN unsupported\n");
2740 	}
2741 
2742 	return 0;
2743 }
2744 
2745 static int
2746 bnxt_vlan_tpid_set_op(struct rte_eth_dev *dev, enum rte_vlan_type vlan_type,
2747 		      uint16_t tpid)
2748 {
2749 	struct bnxt *bp = dev->data->dev_private;
2750 	int qinq = dev->data->dev_conf.rxmode.offloads &
2751 		   RTE_ETH_RX_OFFLOAD_VLAN_EXTEND;
2752 
2753 	if (vlan_type != RTE_ETH_VLAN_TYPE_INNER &&
2754 	    vlan_type != RTE_ETH_VLAN_TYPE_OUTER) {
2755 		PMD_DRV_LOG(ERR,
2756 			    "Unsupported vlan type.");
2757 		return -EINVAL;
2758 	}
2759 	if (!qinq) {
2760 		PMD_DRV_LOG(ERR,
2761 			    "QinQ not enabled. Needs to be ON as we can "
2762 			    "accelerate only outer vlan\n");
2763 		return -EINVAL;
2764 	}
2765 
2766 	if (vlan_type == RTE_ETH_VLAN_TYPE_OUTER) {
2767 		switch (tpid) {
2768 		case RTE_ETHER_TYPE_QINQ:
2769 			bp->outer_tpid_bd =
2770 				TX_BD_LONG_CFA_META_VLAN_TPID_TPID88A8;
2771 				break;
2772 		case RTE_ETHER_TYPE_VLAN:
2773 			bp->outer_tpid_bd =
2774 				TX_BD_LONG_CFA_META_VLAN_TPID_TPID8100;
2775 				break;
2776 		case RTE_ETHER_TYPE_QINQ1:
2777 			bp->outer_tpid_bd =
2778 				TX_BD_LONG_CFA_META_VLAN_TPID_TPID9100;
2779 				break;
2780 		case RTE_ETHER_TYPE_QINQ2:
2781 			bp->outer_tpid_bd =
2782 				TX_BD_LONG_CFA_META_VLAN_TPID_TPID9200;
2783 				break;
2784 		case RTE_ETHER_TYPE_QINQ3:
2785 			bp->outer_tpid_bd =
2786 				 TX_BD_LONG_CFA_META_VLAN_TPID_TPID9300;
2787 				break;
2788 		default:
2789 			PMD_DRV_LOG(ERR, "Invalid TPID: %x\n", tpid);
2790 			return -EINVAL;
2791 		}
2792 		bp->outer_tpid_bd |= tpid;
2793 		PMD_DRV_LOG(INFO, "outer_tpid_bd = %x\n", bp->outer_tpid_bd);
2794 	} else if (vlan_type == RTE_ETH_VLAN_TYPE_INNER) {
2795 		PMD_DRV_LOG(ERR,
2796 			    "Can accelerate only outer vlan in QinQ\n");
2797 		return -EINVAL;
2798 	}
2799 
2800 	return 0;
2801 }
2802 
2803 static int
2804 bnxt_set_default_mac_addr_op(struct rte_eth_dev *dev,
2805 			     struct rte_ether_addr *addr)
2806 {
2807 	struct bnxt *bp = dev->data->dev_private;
2808 	/* Default Filter is tied to VNIC 0 */
2809 	struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
2810 	int rc;
2811 
2812 	rc = is_bnxt_in_error(bp);
2813 	if (rc)
2814 		return rc;
2815 
2816 	if (BNXT_VF(bp) && !BNXT_VF_IS_TRUSTED(bp))
2817 		return -EPERM;
2818 
2819 	if (rte_is_zero_ether_addr(addr))
2820 		return -EINVAL;
2821 
2822 	/* Filter settings will get applied when port is started */
2823 	if (!dev->data->dev_started)
2824 		return 0;
2825 
2826 	/* Check if the requested MAC is already added */
2827 	if (memcmp(addr, bp->mac_addr, RTE_ETHER_ADDR_LEN) == 0)
2828 		return 0;
2829 
2830 	/* Destroy filter and re-create it */
2831 	bnxt_del_dflt_mac_filter(bp, vnic);
2832 
2833 	memcpy(bp->mac_addr, addr, RTE_ETHER_ADDR_LEN);
2834 	if (dev->data->dev_conf.rxmode.offloads & RTE_ETH_RX_OFFLOAD_VLAN_FILTER) {
2835 		/* This filter will allow only untagged packets */
2836 		rc = bnxt_add_vlan_filter(bp, 0);
2837 	} else {
2838 		rc = bnxt_add_mac_filter(bp, vnic, addr, 0, 0);
2839 	}
2840 
2841 	PMD_DRV_LOG(DEBUG, "Set MAC addr\n");
2842 	return rc;
2843 }
2844 
2845 static int
2846 bnxt_dev_set_mc_addr_list_op(struct rte_eth_dev *eth_dev,
2847 			  struct rte_ether_addr *mc_addr_set,
2848 			  uint32_t nb_mc_addr)
2849 {
2850 	struct bnxt *bp = eth_dev->data->dev_private;
2851 	char *mc_addr_list = (char *)mc_addr_set;
2852 	struct bnxt_vnic_info *vnic;
2853 	uint32_t off = 0, i = 0;
2854 	int rc;
2855 
2856 	rc = is_bnxt_in_error(bp);
2857 	if (rc)
2858 		return rc;
2859 
2860 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
2861 
2862 	if (nb_mc_addr > BNXT_MAX_MC_ADDRS) {
2863 		vnic->flags |= BNXT_VNIC_INFO_ALLMULTI;
2864 		goto allmulti;
2865 	}
2866 
2867 	/* TODO Check for Duplicate mcast addresses */
2868 	vnic->flags &= ~BNXT_VNIC_INFO_ALLMULTI;
2869 	for (i = 0; i < nb_mc_addr; i++) {
2870 		memcpy(vnic->mc_list + off, &mc_addr_list[i],
2871 			RTE_ETHER_ADDR_LEN);
2872 		off += RTE_ETHER_ADDR_LEN;
2873 	}
2874 
2875 	vnic->mc_addr_cnt = i;
2876 	if (vnic->mc_addr_cnt)
2877 		vnic->flags |= BNXT_VNIC_INFO_MCAST;
2878 	else
2879 		vnic->flags &= ~BNXT_VNIC_INFO_MCAST;
2880 
2881 allmulti:
2882 	return bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
2883 }
2884 
2885 static int
2886 bnxt_fw_version_get(struct rte_eth_dev *dev, char *fw_version, size_t fw_size)
2887 {
2888 	struct bnxt *bp = dev->data->dev_private;
2889 	uint8_t fw_major = (bp->fw_ver >> 24) & 0xff;
2890 	uint8_t fw_minor = (bp->fw_ver >> 16) & 0xff;
2891 	uint8_t fw_updt = (bp->fw_ver >> 8) & 0xff;
2892 	uint8_t fw_rsvd = bp->fw_ver & 0xff;
2893 	int ret;
2894 
2895 	ret = snprintf(fw_version, fw_size, "%d.%d.%d.%d",
2896 			fw_major, fw_minor, fw_updt, fw_rsvd);
2897 	if (ret < 0)
2898 		return -EINVAL;
2899 
2900 	ret += 1; /* add the size of '\0' */
2901 	if (fw_size < (size_t)ret)
2902 		return ret;
2903 	else
2904 		return 0;
2905 }
2906 
2907 static void
2908 bnxt_rxq_info_get_op(struct rte_eth_dev *dev, uint16_t queue_id,
2909 	struct rte_eth_rxq_info *qinfo)
2910 {
2911 	struct bnxt *bp = dev->data->dev_private;
2912 	struct bnxt_rx_queue *rxq;
2913 
2914 	if (is_bnxt_in_error(bp))
2915 		return;
2916 
2917 	rxq = dev->data->rx_queues[queue_id];
2918 
2919 	qinfo->mp = rxq->mb_pool;
2920 	qinfo->scattered_rx = dev->data->scattered_rx;
2921 	qinfo->nb_desc = rxq->nb_rx_desc;
2922 
2923 	qinfo->conf.rx_free_thresh = rxq->rx_free_thresh;
2924 	qinfo->conf.rx_drop_en = rxq->drop_en;
2925 	qinfo->conf.rx_deferred_start = rxq->rx_deferred_start;
2926 	qinfo->conf.offloads = dev->data->dev_conf.rxmode.offloads;
2927 }
2928 
2929 static void
2930 bnxt_txq_info_get_op(struct rte_eth_dev *dev, uint16_t queue_id,
2931 	struct rte_eth_txq_info *qinfo)
2932 {
2933 	struct bnxt *bp = dev->data->dev_private;
2934 	struct bnxt_tx_queue *txq;
2935 
2936 	if (is_bnxt_in_error(bp))
2937 		return;
2938 
2939 	txq = dev->data->tx_queues[queue_id];
2940 
2941 	qinfo->nb_desc = txq->nb_tx_desc;
2942 
2943 	qinfo->conf.tx_thresh.pthresh = txq->pthresh;
2944 	qinfo->conf.tx_thresh.hthresh = txq->hthresh;
2945 	qinfo->conf.tx_thresh.wthresh = txq->wthresh;
2946 
2947 	qinfo->conf.tx_free_thresh = txq->tx_free_thresh;
2948 	qinfo->conf.tx_rs_thresh = 0;
2949 	qinfo->conf.tx_deferred_start = txq->tx_deferred_start;
2950 	qinfo->conf.offloads = txq->offloads;
2951 }
2952 
2953 static const struct {
2954 	eth_rx_burst_t pkt_burst;
2955 	const char *info;
2956 } bnxt_rx_burst_info[] = {
2957 	{bnxt_recv_pkts,		"Scalar"},
2958 #if defined(RTE_ARCH_X86)
2959 	{bnxt_recv_pkts_vec,		"Vector SSE"},
2960 #endif
2961 #if defined(RTE_ARCH_X86) && defined(CC_AVX2_SUPPORT)
2962 	{bnxt_recv_pkts_vec_avx2,	"Vector AVX2"},
2963 #endif
2964 #if defined(RTE_ARCH_ARM64)
2965 	{bnxt_recv_pkts_vec,		"Vector Neon"},
2966 #endif
2967 };
2968 
2969 static int
2970 bnxt_rx_burst_mode_get(struct rte_eth_dev *dev, __rte_unused uint16_t queue_id,
2971 		       struct rte_eth_burst_mode *mode)
2972 {
2973 	eth_rx_burst_t pkt_burst = dev->rx_pkt_burst;
2974 	size_t i;
2975 
2976 	for (i = 0; i < RTE_DIM(bnxt_rx_burst_info); i++) {
2977 		if (pkt_burst == bnxt_rx_burst_info[i].pkt_burst) {
2978 			snprintf(mode->info, sizeof(mode->info), "%s",
2979 				 bnxt_rx_burst_info[i].info);
2980 			return 0;
2981 		}
2982 	}
2983 
2984 	return -EINVAL;
2985 }
2986 
2987 static const struct {
2988 	eth_tx_burst_t pkt_burst;
2989 	const char *info;
2990 } bnxt_tx_burst_info[] = {
2991 	{bnxt_xmit_pkts,		"Scalar"},
2992 #if defined(RTE_ARCH_X86)
2993 	{bnxt_xmit_pkts_vec,		"Vector SSE"},
2994 #endif
2995 #if defined(RTE_ARCH_X86) && defined(CC_AVX2_SUPPORT)
2996 	{bnxt_xmit_pkts_vec_avx2,	"Vector AVX2"},
2997 #endif
2998 #if defined(RTE_ARCH_ARM64)
2999 	{bnxt_xmit_pkts_vec,		"Vector Neon"},
3000 #endif
3001 };
3002 
3003 static int
3004 bnxt_tx_burst_mode_get(struct rte_eth_dev *dev, __rte_unused uint16_t queue_id,
3005 		       struct rte_eth_burst_mode *mode)
3006 {
3007 	eth_tx_burst_t pkt_burst = dev->tx_pkt_burst;
3008 	size_t i;
3009 
3010 	for (i = 0; i < RTE_DIM(bnxt_tx_burst_info); i++) {
3011 		if (pkt_burst == bnxt_tx_burst_info[i].pkt_burst) {
3012 			snprintf(mode->info, sizeof(mode->info), "%s",
3013 				 bnxt_tx_burst_info[i].info);
3014 			return 0;
3015 		}
3016 	}
3017 
3018 	return -EINVAL;
3019 }
3020 
3021 int bnxt_mtu_set_op(struct rte_eth_dev *eth_dev, uint16_t new_mtu)
3022 {
3023 	uint32_t overhead = BNXT_MAX_PKT_LEN - BNXT_MAX_MTU;
3024 	struct bnxt *bp = eth_dev->data->dev_private;
3025 	uint32_t new_pkt_size;
3026 	uint32_t rc;
3027 	uint32_t i;
3028 
3029 	rc = is_bnxt_in_error(bp);
3030 	if (rc)
3031 		return rc;
3032 
3033 	/* Exit if receive queues are not configured yet */
3034 	if (!eth_dev->data->nb_rx_queues)
3035 		return rc;
3036 
3037 	new_pkt_size = new_mtu + overhead;
3038 
3039 	/*
3040 	 * Disallow any MTU change that would require scattered receive support
3041 	 * if it is not already enabled.
3042 	 */
3043 	if (eth_dev->data->dev_started &&
3044 	    !eth_dev->data->scattered_rx &&
3045 	    (new_pkt_size >
3046 	     eth_dev->data->min_rx_buf_size - RTE_PKTMBUF_HEADROOM)) {
3047 		PMD_DRV_LOG(ERR,
3048 			    "MTU change would require scattered rx support. ");
3049 		PMD_DRV_LOG(ERR, "Stop port before changing MTU.\n");
3050 		return -EINVAL;
3051 	}
3052 
3053 	if (new_mtu > RTE_ETHER_MTU)
3054 		bp->flags |= BNXT_FLAG_JUMBO;
3055 	else
3056 		bp->flags &= ~BNXT_FLAG_JUMBO;
3057 
3058 	/* Is there a change in mtu setting? */
3059 	if (eth_dev->data->mtu == new_mtu)
3060 		return rc;
3061 
3062 	for (i = 0; i < bp->nr_vnics; i++) {
3063 		struct bnxt_vnic_info *vnic = &bp->vnic_info[i];
3064 		uint16_t size = 0;
3065 
3066 		vnic->mru = BNXT_VNIC_MRU(new_mtu);
3067 		rc = bnxt_hwrm_vnic_cfg(bp, vnic);
3068 		if (rc)
3069 			break;
3070 
3071 		size = rte_pktmbuf_data_room_size(bp->rx_queues[0]->mb_pool);
3072 		size -= RTE_PKTMBUF_HEADROOM;
3073 
3074 		if (size < new_mtu) {
3075 			rc = bnxt_hwrm_vnic_plcmode_cfg(bp, vnic);
3076 			if (rc)
3077 				return rc;
3078 		}
3079 	}
3080 
3081 	if (bnxt_hwrm_config_host_mtu(bp))
3082 		PMD_DRV_LOG(WARNING, "Failed to configure host MTU\n");
3083 
3084 	PMD_DRV_LOG(INFO, "New MTU is %d\n", new_mtu);
3085 
3086 	return rc;
3087 }
3088 
3089 static int
3090 bnxt_vlan_pvid_set_op(struct rte_eth_dev *dev, uint16_t pvid, int on)
3091 {
3092 	struct bnxt *bp = dev->data->dev_private;
3093 	uint16_t vlan = bp->vlan;
3094 	int rc;
3095 
3096 	rc = is_bnxt_in_error(bp);
3097 	if (rc)
3098 		return rc;
3099 
3100 	if (!BNXT_SINGLE_PF(bp)) {
3101 		PMD_DRV_LOG(ERR, "PVID cannot be modified on VF or on shared PF\n");
3102 		return -ENOTSUP;
3103 	}
3104 	bp->vlan = on ? pvid : 0;
3105 
3106 	rc = bnxt_hwrm_set_default_vlan(bp, 0, 0);
3107 	if (rc)
3108 		bp->vlan = vlan;
3109 	return rc;
3110 }
3111 
3112 static int
3113 bnxt_dev_led_on_op(struct rte_eth_dev *dev)
3114 {
3115 	struct bnxt *bp = dev->data->dev_private;
3116 	int rc;
3117 
3118 	rc = is_bnxt_in_error(bp);
3119 	if (rc)
3120 		return rc;
3121 
3122 	return bnxt_hwrm_port_led_cfg(bp, true);
3123 }
3124 
3125 static int
3126 bnxt_dev_led_off_op(struct rte_eth_dev *dev)
3127 {
3128 	struct bnxt *bp = dev->data->dev_private;
3129 	int rc;
3130 
3131 	rc = is_bnxt_in_error(bp);
3132 	if (rc)
3133 		return rc;
3134 
3135 	return bnxt_hwrm_port_led_cfg(bp, false);
3136 }
3137 
3138 static uint32_t
3139 bnxt_rx_queue_count_op(void *rx_queue)
3140 {
3141 	struct bnxt *bp;
3142 	struct bnxt_cp_ring_info *cpr;
3143 	uint32_t desc = 0, raw_cons, cp_ring_size;
3144 	struct bnxt_rx_queue *rxq;
3145 	struct rx_pkt_cmpl *rxcmp;
3146 	int rc;
3147 
3148 	rxq = rx_queue;
3149 	bp = rxq->bp;
3150 
3151 	rc = is_bnxt_in_error(bp);
3152 	if (rc)
3153 		return rc;
3154 
3155 	cpr = rxq->cp_ring;
3156 	raw_cons = cpr->cp_raw_cons;
3157 	cp_ring_size = cpr->cp_ring_struct->ring_size;
3158 
3159 	while (1) {
3160 		uint32_t agg_cnt, cons, cmpl_type;
3161 
3162 		cons = RING_CMP(cpr->cp_ring_struct, raw_cons);
3163 		rxcmp = (struct rx_pkt_cmpl *)&cpr->cp_desc_ring[cons];
3164 
3165 		if (!bnxt_cpr_cmp_valid(rxcmp, raw_cons, cp_ring_size))
3166 			break;
3167 
3168 		cmpl_type = CMP_TYPE(rxcmp);
3169 
3170 		switch (cmpl_type) {
3171 		case CMPL_BASE_TYPE_RX_L2:
3172 		case CMPL_BASE_TYPE_RX_L2_V2:
3173 			agg_cnt = BNXT_RX_L2_AGG_BUFS(rxcmp);
3174 			raw_cons = raw_cons + CMP_LEN(cmpl_type) + agg_cnt;
3175 			desc++;
3176 			break;
3177 
3178 		case CMPL_BASE_TYPE_RX_TPA_END:
3179 			if (BNXT_CHIP_P5(rxq->bp)) {
3180 				struct rx_tpa_v2_end_cmpl_hi *p5_tpa_end;
3181 
3182 				p5_tpa_end = (void *)rxcmp;
3183 				agg_cnt = BNXT_TPA_END_AGG_BUFS_TH(p5_tpa_end);
3184 			} else {
3185 				struct rx_tpa_end_cmpl *tpa_end;
3186 
3187 				tpa_end = (void *)rxcmp;
3188 				agg_cnt = BNXT_TPA_END_AGG_BUFS(tpa_end);
3189 			}
3190 
3191 			raw_cons = raw_cons + CMP_LEN(cmpl_type) + agg_cnt;
3192 			desc++;
3193 			break;
3194 
3195 		default:
3196 			raw_cons += CMP_LEN(cmpl_type);
3197 		}
3198 	}
3199 
3200 	return desc;
3201 }
3202 
3203 static int
3204 bnxt_rx_descriptor_status_op(void *rx_queue, uint16_t offset)
3205 {
3206 	struct bnxt_rx_queue *rxq = rx_queue;
3207 	struct bnxt_cp_ring_info *cpr;
3208 	struct bnxt_rx_ring_info *rxr;
3209 	uint32_t desc, raw_cons, cp_ring_size;
3210 	struct bnxt *bp = rxq->bp;
3211 	struct rx_pkt_cmpl *rxcmp;
3212 	int rc;
3213 
3214 	rc = is_bnxt_in_error(bp);
3215 	if (rc)
3216 		return rc;
3217 
3218 	if (offset >= rxq->nb_rx_desc)
3219 		return -EINVAL;
3220 
3221 	rxr = rxq->rx_ring;
3222 	cpr = rxq->cp_ring;
3223 	cp_ring_size = cpr->cp_ring_struct->ring_size;
3224 
3225 	/*
3226 	 * For the vector receive case, the completion at the requested
3227 	 * offset can be indexed directly.
3228 	 */
3229 #if defined(RTE_ARCH_X86) || defined(RTE_ARCH_ARM64)
3230 	if (bp->flags & BNXT_FLAG_RX_VECTOR_PKT_MODE) {
3231 		struct rx_pkt_cmpl *rxcmp;
3232 		uint32_t cons;
3233 
3234 		/* Check status of completion descriptor. */
3235 		raw_cons = cpr->cp_raw_cons +
3236 			   offset * CMP_LEN(CMPL_BASE_TYPE_RX_L2);
3237 		cons = RING_CMP(cpr->cp_ring_struct, raw_cons);
3238 		rxcmp = (struct rx_pkt_cmpl *)&cpr->cp_desc_ring[cons];
3239 
3240 		if (bnxt_cpr_cmp_valid(rxcmp, raw_cons, cp_ring_size))
3241 			return RTE_ETH_RX_DESC_DONE;
3242 
3243 		/* Check whether rx desc has an mbuf attached. */
3244 		cons = RING_CMP(rxr->rx_ring_struct, raw_cons / 2);
3245 		if (cons >= rxq->rxrearm_start &&
3246 		    cons < rxq->rxrearm_start + rxq->rxrearm_nb) {
3247 			return RTE_ETH_RX_DESC_UNAVAIL;
3248 		}
3249 
3250 		return RTE_ETH_RX_DESC_AVAIL;
3251 	}
3252 #endif
3253 
3254 	/*
3255 	 * For the non-vector receive case, scan the completion ring to
3256 	 * locate the completion descriptor for the requested offset.
3257 	 */
3258 	raw_cons = cpr->cp_raw_cons;
3259 	desc = 0;
3260 	while (1) {
3261 		uint32_t agg_cnt, cons, cmpl_type;
3262 
3263 		cons = RING_CMP(cpr->cp_ring_struct, raw_cons);
3264 		rxcmp = (struct rx_pkt_cmpl *)&cpr->cp_desc_ring[cons];
3265 
3266 		if (!bnxt_cpr_cmp_valid(rxcmp, raw_cons, cp_ring_size))
3267 			break;
3268 
3269 		cmpl_type = CMP_TYPE(rxcmp);
3270 
3271 		switch (cmpl_type) {
3272 		case CMPL_BASE_TYPE_RX_L2:
3273 		case CMPL_BASE_TYPE_RX_L2_V2:
3274 			if (desc == offset) {
3275 				cons = rxcmp->opaque;
3276 				if (rxr->rx_buf_ring[cons])
3277 					return RTE_ETH_RX_DESC_DONE;
3278 				else
3279 					return RTE_ETH_RX_DESC_UNAVAIL;
3280 			}
3281 			agg_cnt = BNXT_RX_L2_AGG_BUFS(rxcmp);
3282 			raw_cons = raw_cons + CMP_LEN(cmpl_type) + agg_cnt;
3283 			desc++;
3284 			break;
3285 
3286 		case CMPL_BASE_TYPE_RX_TPA_END:
3287 			if (desc == offset)
3288 				return RTE_ETH_RX_DESC_DONE;
3289 
3290 			if (BNXT_CHIP_P5(rxq->bp)) {
3291 				struct rx_tpa_v2_end_cmpl_hi *p5_tpa_end;
3292 
3293 				p5_tpa_end = (void *)rxcmp;
3294 				agg_cnt = BNXT_TPA_END_AGG_BUFS_TH(p5_tpa_end);
3295 			} else {
3296 				struct rx_tpa_end_cmpl *tpa_end;
3297 
3298 				tpa_end = (void *)rxcmp;
3299 				agg_cnt = BNXT_TPA_END_AGG_BUFS(tpa_end);
3300 			}
3301 
3302 			raw_cons = raw_cons + CMP_LEN(cmpl_type) + agg_cnt;
3303 			desc++;
3304 			break;
3305 
3306 		default:
3307 			raw_cons += CMP_LEN(cmpl_type);
3308 		}
3309 	}
3310 
3311 	return RTE_ETH_RX_DESC_AVAIL;
3312 }
3313 
3314 static int
3315 bnxt_tx_descriptor_status_op(void *tx_queue, uint16_t offset)
3316 {
3317 	struct bnxt_tx_queue *txq = (struct bnxt_tx_queue *)tx_queue;
3318 	struct bnxt_cp_ring_info *cpr = txq->cp_ring;
3319 	uint32_t ring_mask, raw_cons, nb_tx_pkts = 0;
3320 	struct cmpl_base *cp_desc_ring;
3321 	int rc;
3322 
3323 	rc = is_bnxt_in_error(txq->bp);
3324 	if (rc)
3325 		return rc;
3326 
3327 	if (offset >= txq->nb_tx_desc)
3328 		return -EINVAL;
3329 
3330 	/* Return "desc done" if descriptor is available for use. */
3331 	if (bnxt_tx_bds_in_hw(txq) <= offset)
3332 		return RTE_ETH_TX_DESC_DONE;
3333 
3334 	raw_cons = cpr->cp_raw_cons;
3335 	cp_desc_ring = cpr->cp_desc_ring;
3336 	ring_mask = cpr->cp_ring_struct->ring_mask;
3337 
3338 	/* Check to see if hw has posted a completion for the descriptor. */
3339 	while (1) {
3340 		struct tx_cmpl *txcmp;
3341 		uint32_t cons;
3342 
3343 		cons = RING_CMPL(ring_mask, raw_cons);
3344 		txcmp = (struct tx_cmpl *)&cp_desc_ring[cons];
3345 
3346 		if (!bnxt_cpr_cmp_valid(txcmp, raw_cons, ring_mask + 1))
3347 			break;
3348 
3349 		if (CMP_TYPE(txcmp) == TX_CMPL_TYPE_TX_L2)
3350 			nb_tx_pkts += rte_le_to_cpu_32(txcmp->opaque);
3351 
3352 		if (nb_tx_pkts > offset)
3353 			return RTE_ETH_TX_DESC_DONE;
3354 
3355 		raw_cons = NEXT_RAW_CMP(raw_cons);
3356 	}
3357 
3358 	/* Descriptor is pending transmit, not yet completed by hardware. */
3359 	return RTE_ETH_TX_DESC_FULL;
3360 }
3361 
3362 int
3363 bnxt_flow_ops_get_op(struct rte_eth_dev *dev,
3364 		     const struct rte_flow_ops **ops)
3365 {
3366 	struct bnxt *bp = dev->data->dev_private;
3367 	int ret = 0;
3368 
3369 	if (!bp)
3370 		return -EIO;
3371 
3372 	if (BNXT_ETH_DEV_IS_REPRESENTOR(dev)) {
3373 		struct bnxt_representor *vfr = dev->data->dev_private;
3374 		bp = vfr->parent_dev->data->dev_private;
3375 		/* parent is deleted while children are still valid */
3376 		if (!bp) {
3377 			PMD_DRV_LOG(DEBUG, "BNXT Port:%d VFR Error\n",
3378 				    dev->data->port_id);
3379 			return -EIO;
3380 		}
3381 	}
3382 
3383 	ret = is_bnxt_in_error(bp);
3384 	if (ret)
3385 		return ret;
3386 
3387 	/* PMD supports thread-safe flow operations.  rte_flow API
3388 	 * functions can avoid mutex for multi-thread safety.
3389 	 */
3390 	dev->data->dev_flags |= RTE_ETH_DEV_FLOW_OPS_THREAD_SAFE;
3391 
3392 	if (BNXT_TRUFLOW_EN(bp))
3393 		*ops = &bnxt_ulp_rte_flow_ops;
3394 	else
3395 		*ops = &bnxt_flow_ops;
3396 
3397 	return ret;
3398 }
3399 
3400 static const uint32_t *
3401 bnxt_dev_supported_ptypes_get_op(struct rte_eth_dev *dev)
3402 {
3403 	static const uint32_t ptypes[] = {
3404 		RTE_PTYPE_L2_ETHER_VLAN,
3405 		RTE_PTYPE_L3_IPV4_EXT_UNKNOWN,
3406 		RTE_PTYPE_L3_IPV6_EXT_UNKNOWN,
3407 		RTE_PTYPE_L4_ICMP,
3408 		RTE_PTYPE_L4_TCP,
3409 		RTE_PTYPE_L4_UDP,
3410 		RTE_PTYPE_INNER_L3_IPV4_EXT_UNKNOWN,
3411 		RTE_PTYPE_INNER_L3_IPV6_EXT_UNKNOWN,
3412 		RTE_PTYPE_INNER_L4_ICMP,
3413 		RTE_PTYPE_INNER_L4_TCP,
3414 		RTE_PTYPE_INNER_L4_UDP,
3415 		RTE_PTYPE_UNKNOWN
3416 	};
3417 
3418 	if (!dev->rx_pkt_burst)
3419 		return NULL;
3420 
3421 	return ptypes;
3422 }
3423 
3424 static int bnxt_map_regs(struct bnxt *bp, uint32_t *reg_arr, int count,
3425 			 int reg_win)
3426 {
3427 	uint32_t reg_base = *reg_arr & 0xfffff000;
3428 	uint32_t win_off;
3429 	int i;
3430 
3431 	for (i = 0; i < count; i++) {
3432 		if ((reg_arr[i] & 0xfffff000) != reg_base)
3433 			return -ERANGE;
3434 	}
3435 	win_off = BNXT_GRCPF_REG_WINDOW_BASE_OUT + (reg_win - 1) * 4;
3436 	rte_write32(reg_base, (uint8_t *)bp->bar0 + win_off);
3437 	return 0;
3438 }
3439 
3440 static int bnxt_map_ptp_regs(struct bnxt *bp)
3441 {
3442 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3443 	uint32_t *reg_arr;
3444 	int rc, i;
3445 
3446 	reg_arr = ptp->rx_regs;
3447 	rc = bnxt_map_regs(bp, reg_arr, BNXT_PTP_RX_REGS, 5);
3448 	if (rc)
3449 		return rc;
3450 
3451 	reg_arr = ptp->tx_regs;
3452 	rc = bnxt_map_regs(bp, reg_arr, BNXT_PTP_TX_REGS, 6);
3453 	if (rc)
3454 		return rc;
3455 
3456 	for (i = 0; i < BNXT_PTP_RX_REGS; i++)
3457 		ptp->rx_mapped_regs[i] = 0x5000 + (ptp->rx_regs[i] & 0xfff);
3458 
3459 	for (i = 0; i < BNXT_PTP_TX_REGS; i++)
3460 		ptp->tx_mapped_regs[i] = 0x6000 + (ptp->tx_regs[i] & 0xfff);
3461 
3462 	return 0;
3463 }
3464 
3465 static void bnxt_unmap_ptp_regs(struct bnxt *bp)
3466 {
3467 	rte_write32(0, (uint8_t *)bp->bar0 +
3468 			 BNXT_GRCPF_REG_WINDOW_BASE_OUT + 16);
3469 	rte_write32(0, (uint8_t *)bp->bar0 +
3470 			 BNXT_GRCPF_REG_WINDOW_BASE_OUT + 20);
3471 }
3472 
3473 static uint64_t bnxt_cc_read(struct bnxt *bp)
3474 {
3475 	uint64_t ns;
3476 
3477 	ns = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3478 			      BNXT_GRCPF_REG_SYNC_TIME));
3479 	ns |= (uint64_t)(rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3480 					  BNXT_GRCPF_REG_SYNC_TIME + 4))) << 32;
3481 	return ns;
3482 }
3483 
3484 static int bnxt_get_tx_ts(struct bnxt *bp, uint64_t *ts)
3485 {
3486 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3487 	uint32_t fifo;
3488 
3489 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3490 				ptp->tx_mapped_regs[BNXT_PTP_TX_FIFO]));
3491 	if (fifo & BNXT_PTP_TX_FIFO_EMPTY)
3492 		return -EAGAIN;
3493 
3494 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3495 				ptp->tx_mapped_regs[BNXT_PTP_TX_FIFO]));
3496 	*ts = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3497 				ptp->tx_mapped_regs[BNXT_PTP_TX_TS_L]));
3498 	*ts |= (uint64_t)rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3499 				ptp->tx_mapped_regs[BNXT_PTP_TX_TS_H])) << 32;
3500 	rte_read32((uint8_t *)bp->bar0 + ptp->tx_mapped_regs[BNXT_PTP_TX_SEQ]);
3501 
3502 	return 0;
3503 }
3504 
3505 static int bnxt_clr_rx_ts(struct bnxt *bp, uint64_t *last_ts)
3506 {
3507 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3508 	struct bnxt_pf_info *pf = bp->pf;
3509 	uint16_t port_id;
3510 	int i = 0;
3511 	uint32_t fifo;
3512 
3513 	if (!ptp || (bp->flags & BNXT_FLAG_CHIP_P5))
3514 		return -EINVAL;
3515 
3516 	port_id = pf->port_id;
3517 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3518 				ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO]));
3519 	while ((fifo & BNXT_PTP_RX_FIFO_PENDING) && (i < BNXT_PTP_RX_PND_CNT)) {
3520 		rte_write32(1 << port_id, (uint8_t *)bp->bar0 +
3521 			    ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO_ADV]);
3522 		fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3523 					ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO]));
3524 		*last_ts = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3525 					ptp->rx_mapped_regs[BNXT_PTP_RX_TS_L]));
3526 		*last_ts |= (uint64_t)rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3527 					ptp->rx_mapped_regs[BNXT_PTP_RX_TS_H])) << 32;
3528 		i++;
3529 	}
3530 
3531 	if (i >= BNXT_PTP_RX_PND_CNT)
3532 		return -EBUSY;
3533 
3534 	return 0;
3535 }
3536 
3537 static int bnxt_get_rx_ts(struct bnxt *bp, uint64_t *ts)
3538 {
3539 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3540 	struct bnxt_pf_info *pf = bp->pf;
3541 	uint16_t port_id;
3542 	uint32_t fifo;
3543 
3544 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3545 				ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO]));
3546 	if (!(fifo & BNXT_PTP_RX_FIFO_PENDING))
3547 		return -EAGAIN;
3548 
3549 	port_id = pf->port_id;
3550 	rte_write32(1 << port_id, (uint8_t *)bp->bar0 +
3551 	       ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO_ADV]);
3552 
3553 	fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3554 				   ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO]));
3555 	if (fifo & BNXT_PTP_RX_FIFO_PENDING)
3556 		return bnxt_clr_rx_ts(bp, ts);
3557 
3558 	*ts = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3559 				ptp->rx_mapped_regs[BNXT_PTP_RX_TS_L]));
3560 	*ts |= (uint64_t)rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3561 				ptp->rx_mapped_regs[BNXT_PTP_RX_TS_H])) << 32;
3562 
3563 	return 0;
3564 }
3565 
3566 static int
3567 bnxt_timesync_write_time(struct rte_eth_dev *dev, const struct timespec *ts)
3568 {
3569 	uint64_t ns;
3570 	struct bnxt *bp = dev->data->dev_private;
3571 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3572 
3573 	if (!ptp)
3574 		return -ENOTSUP;
3575 
3576 	ns = rte_timespec_to_ns(ts);
3577 	/* Set the timecounters to a new value. */
3578 	ptp->tc.nsec = ns;
3579 	ptp->tx_tstamp_tc.nsec = ns;
3580 	ptp->rx_tstamp_tc.nsec = ns;
3581 
3582 	return 0;
3583 }
3584 
3585 static int
3586 bnxt_timesync_read_time(struct rte_eth_dev *dev, struct timespec *ts)
3587 {
3588 	struct bnxt *bp = dev->data->dev_private;
3589 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3590 	uint64_t ns, systime_cycles = 0;
3591 	int rc = 0;
3592 
3593 	if (!ptp)
3594 		return -ENOTSUP;
3595 
3596 	if (BNXT_CHIP_P5(bp))
3597 		rc = bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_CURRENT_TIME,
3598 					     &systime_cycles);
3599 	else
3600 		systime_cycles = bnxt_cc_read(bp);
3601 
3602 	ns = rte_timecounter_update(&ptp->tc, systime_cycles);
3603 	*ts = rte_ns_to_timespec(ns);
3604 
3605 	return rc;
3606 }
3607 static int
3608 bnxt_timesync_enable(struct rte_eth_dev *dev)
3609 {
3610 	struct bnxt *bp = dev->data->dev_private;
3611 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3612 	uint32_t shift = 0;
3613 	int rc;
3614 
3615 	if (!ptp)
3616 		return -ENOTSUP;
3617 
3618 	ptp->rx_filter = 1;
3619 	ptp->tx_tstamp_en = 1;
3620 	ptp->rxctl = BNXT_PTP_MSG_EVENTS;
3621 
3622 	rc = bnxt_hwrm_ptp_cfg(bp);
3623 	if (rc)
3624 		return rc;
3625 
3626 	memset(&ptp->tc, 0, sizeof(struct rte_timecounter));
3627 	memset(&ptp->rx_tstamp_tc, 0, sizeof(struct rte_timecounter));
3628 	memset(&ptp->tx_tstamp_tc, 0, sizeof(struct rte_timecounter));
3629 
3630 	ptp->tc.cc_mask = BNXT_CYCLECOUNTER_MASK;
3631 	ptp->tc.cc_shift = shift;
3632 	ptp->tc.nsec_mask = (1ULL << shift) - 1;
3633 
3634 	ptp->rx_tstamp_tc.cc_mask = BNXT_CYCLECOUNTER_MASK;
3635 	ptp->rx_tstamp_tc.cc_shift = shift;
3636 	ptp->rx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
3637 
3638 	ptp->tx_tstamp_tc.cc_mask = BNXT_CYCLECOUNTER_MASK;
3639 	ptp->tx_tstamp_tc.cc_shift = shift;
3640 	ptp->tx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
3641 
3642 	if (!BNXT_CHIP_P5(bp))
3643 		bnxt_map_ptp_regs(bp);
3644 	else
3645 		rc = bnxt_ptp_start(bp);
3646 
3647 	return rc;
3648 }
3649 
3650 static int
3651 bnxt_timesync_disable(struct rte_eth_dev *dev)
3652 {
3653 	struct bnxt *bp = dev->data->dev_private;
3654 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3655 
3656 	if (!ptp)
3657 		return -ENOTSUP;
3658 
3659 	ptp->rx_filter = 0;
3660 	ptp->tx_tstamp_en = 0;
3661 	ptp->rxctl = 0;
3662 
3663 	bnxt_hwrm_ptp_cfg(bp);
3664 
3665 	if (!BNXT_CHIP_P5(bp))
3666 		bnxt_unmap_ptp_regs(bp);
3667 	else
3668 		bnxt_ptp_stop(bp);
3669 
3670 	return 0;
3671 }
3672 
3673 static int
3674 bnxt_timesync_read_rx_timestamp(struct rte_eth_dev *dev,
3675 				 struct timespec *timestamp,
3676 				 uint32_t flags __rte_unused)
3677 {
3678 	struct bnxt *bp = dev->data->dev_private;
3679 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3680 	uint64_t rx_tstamp_cycles = 0;
3681 	uint64_t ns;
3682 
3683 	if (!ptp)
3684 		return -ENOTSUP;
3685 
3686 	if (BNXT_CHIP_P5(bp))
3687 		rx_tstamp_cycles = ptp->rx_timestamp;
3688 	else
3689 		bnxt_get_rx_ts(bp, &rx_tstamp_cycles);
3690 
3691 	ns = rte_timecounter_update(&ptp->rx_tstamp_tc, rx_tstamp_cycles);
3692 	*timestamp = rte_ns_to_timespec(ns);
3693 	return  0;
3694 }
3695 
3696 static int
3697 bnxt_timesync_read_tx_timestamp(struct rte_eth_dev *dev,
3698 				 struct timespec *timestamp)
3699 {
3700 	struct bnxt *bp = dev->data->dev_private;
3701 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3702 	uint64_t tx_tstamp_cycles = 0;
3703 	uint64_t ns;
3704 	int rc = 0;
3705 
3706 	if (!ptp)
3707 		return -ENOTSUP;
3708 
3709 	if (BNXT_CHIP_P5(bp))
3710 		rc = bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_PATH_TX,
3711 					     &tx_tstamp_cycles);
3712 	else
3713 		rc = bnxt_get_tx_ts(bp, &tx_tstamp_cycles);
3714 
3715 	ns = rte_timecounter_update(&ptp->tx_tstamp_tc, tx_tstamp_cycles);
3716 	*timestamp = rte_ns_to_timespec(ns);
3717 
3718 	return rc;
3719 }
3720 
3721 static int
3722 bnxt_timesync_adjust_time(struct rte_eth_dev *dev, int64_t delta)
3723 {
3724 	struct bnxt *bp = dev->data->dev_private;
3725 	struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3726 
3727 	if (!ptp)
3728 		return -ENOTSUP;
3729 
3730 	ptp->tc.nsec += delta;
3731 	ptp->tx_tstamp_tc.nsec += delta;
3732 	ptp->rx_tstamp_tc.nsec += delta;
3733 
3734 	return 0;
3735 }
3736 
3737 static int
3738 bnxt_get_eeprom_length_op(struct rte_eth_dev *dev)
3739 {
3740 	struct bnxt *bp = dev->data->dev_private;
3741 	int rc;
3742 	uint32_t dir_entries;
3743 	uint32_t entry_length;
3744 
3745 	rc = is_bnxt_in_error(bp);
3746 	if (rc)
3747 		return rc;
3748 
3749 	PMD_DRV_LOG(INFO, PCI_PRI_FMT "\n",
3750 		    bp->pdev->addr.domain, bp->pdev->addr.bus,
3751 		    bp->pdev->addr.devid, bp->pdev->addr.function);
3752 
3753 	rc = bnxt_hwrm_nvm_get_dir_info(bp, &dir_entries, &entry_length);
3754 	if (rc != 0)
3755 		return rc;
3756 
3757 	return dir_entries * entry_length;
3758 }
3759 
3760 static int
3761 bnxt_get_eeprom_op(struct rte_eth_dev *dev,
3762 		struct rte_dev_eeprom_info *in_eeprom)
3763 {
3764 	struct bnxt *bp = dev->data->dev_private;
3765 	uint32_t index;
3766 	uint32_t offset;
3767 	int rc;
3768 
3769 	rc = is_bnxt_in_error(bp);
3770 	if (rc)
3771 		return rc;
3772 
3773 	PMD_DRV_LOG(INFO, PCI_PRI_FMT " in_eeprom->offset = %d len = %d\n",
3774 		    bp->pdev->addr.domain, bp->pdev->addr.bus,
3775 		    bp->pdev->addr.devid, bp->pdev->addr.function,
3776 		    in_eeprom->offset, in_eeprom->length);
3777 
3778 	if (in_eeprom->offset == 0) /* special offset value to get directory */
3779 		return bnxt_get_nvram_directory(bp, in_eeprom->length,
3780 						in_eeprom->data);
3781 
3782 	index = in_eeprom->offset >> 24;
3783 	offset = in_eeprom->offset & 0xffffff;
3784 
3785 	if (index != 0)
3786 		return bnxt_hwrm_get_nvram_item(bp, index - 1, offset,
3787 					   in_eeprom->length, in_eeprom->data);
3788 
3789 	return 0;
3790 }
3791 
3792 static bool bnxt_dir_type_is_ape_bin_format(uint16_t dir_type)
3793 {
3794 	switch (dir_type) {
3795 	case BNX_DIR_TYPE_CHIMP_PATCH:
3796 	case BNX_DIR_TYPE_BOOTCODE:
3797 	case BNX_DIR_TYPE_BOOTCODE_2:
3798 	case BNX_DIR_TYPE_APE_FW:
3799 	case BNX_DIR_TYPE_APE_PATCH:
3800 	case BNX_DIR_TYPE_KONG_FW:
3801 	case BNX_DIR_TYPE_KONG_PATCH:
3802 	case BNX_DIR_TYPE_BONO_FW:
3803 	case BNX_DIR_TYPE_BONO_PATCH:
3804 		/* FALLTHROUGH */
3805 		return true;
3806 	}
3807 
3808 	return false;
3809 }
3810 
3811 static bool bnxt_dir_type_is_other_exec_format(uint16_t dir_type)
3812 {
3813 	switch (dir_type) {
3814 	case BNX_DIR_TYPE_AVS:
3815 	case BNX_DIR_TYPE_EXP_ROM_MBA:
3816 	case BNX_DIR_TYPE_PCIE:
3817 	case BNX_DIR_TYPE_TSCF_UCODE:
3818 	case BNX_DIR_TYPE_EXT_PHY:
3819 	case BNX_DIR_TYPE_CCM:
3820 	case BNX_DIR_TYPE_ISCSI_BOOT:
3821 	case BNX_DIR_TYPE_ISCSI_BOOT_IPV6:
3822 	case BNX_DIR_TYPE_ISCSI_BOOT_IPV4N6:
3823 		/* FALLTHROUGH */
3824 		return true;
3825 	}
3826 
3827 	return false;
3828 }
3829 
3830 static bool bnxt_dir_type_is_executable(uint16_t dir_type)
3831 {
3832 	return bnxt_dir_type_is_ape_bin_format(dir_type) ||
3833 		bnxt_dir_type_is_other_exec_format(dir_type);
3834 }
3835 
3836 static int
3837 bnxt_set_eeprom_op(struct rte_eth_dev *dev,
3838 		struct rte_dev_eeprom_info *in_eeprom)
3839 {
3840 	struct bnxt *bp = dev->data->dev_private;
3841 	uint8_t index, dir_op;
3842 	uint16_t type, ext, ordinal, attr;
3843 	int rc;
3844 
3845 	rc = is_bnxt_in_error(bp);
3846 	if (rc)
3847 		return rc;
3848 
3849 	PMD_DRV_LOG(INFO, PCI_PRI_FMT " in_eeprom->offset = %d len = %d\n",
3850 		    bp->pdev->addr.domain, bp->pdev->addr.bus,
3851 		    bp->pdev->addr.devid, bp->pdev->addr.function,
3852 		    in_eeprom->offset, in_eeprom->length);
3853 
3854 	if (!BNXT_PF(bp)) {
3855 		PMD_DRV_LOG(ERR, "NVM write not supported from a VF\n");
3856 		return -EINVAL;
3857 	}
3858 
3859 	type = in_eeprom->magic >> 16;
3860 
3861 	if (type == 0xffff) { /* special value for directory operations */
3862 		index = in_eeprom->magic & 0xff;
3863 		dir_op = in_eeprom->magic >> 8;
3864 		if (index == 0)
3865 			return -EINVAL;
3866 		switch (dir_op) {
3867 		case 0x0e: /* erase */
3868 			if (in_eeprom->offset != ~in_eeprom->magic)
3869 				return -EINVAL;
3870 			return bnxt_hwrm_erase_nvram_directory(bp, index - 1);
3871 		default:
3872 			return -EINVAL;
3873 		}
3874 	}
3875 
3876 	/* Create or re-write an NVM item: */
3877 	if (bnxt_dir_type_is_executable(type) == true)
3878 		return -EOPNOTSUPP;
3879 	ext = in_eeprom->magic & 0xffff;
3880 	ordinal = in_eeprom->offset >> 16;
3881 	attr = in_eeprom->offset & 0xffff;
3882 
3883 	return bnxt_hwrm_flash_nvram(bp, type, ordinal, ext, attr,
3884 				     in_eeprom->data, in_eeprom->length);
3885 }
3886 
3887 static int bnxt_get_module_info(struct rte_eth_dev *dev,
3888 				struct rte_eth_dev_module_info *modinfo)
3889 {
3890 	uint8_t module_info[SFF_DIAG_SUPPORT_OFFSET + 1];
3891 	struct bnxt *bp = dev->data->dev_private;
3892 	int rc;
3893 
3894 	/* No point in going further if phy status indicates
3895 	 * module is not inserted or if it is powered down or
3896 	 * if it is of type 10GBase-T
3897 	 */
3898 	if (bp->link_info->module_status >
3899 	    HWRM_PORT_PHY_QCFG_OUTPUT_MODULE_STATUS_WARNINGMSG) {
3900 		PMD_DRV_LOG(NOTICE, "Port %u : Module is not inserted or is powered down\n",
3901 			    dev->data->port_id);
3902 		return -ENOTSUP;
3903 	}
3904 
3905 	/* This feature is not supported in older firmware versions */
3906 	if (bp->hwrm_spec_code < 0x10202) {
3907 		PMD_DRV_LOG(NOTICE, "Port %u : Feature is not supported in older firmware\n",
3908 			    dev->data->port_id);
3909 		return -ENOTSUP;
3910 	}
3911 
3912 	rc = bnxt_hwrm_read_sfp_module_eeprom_info(bp, I2C_DEV_ADDR_A0, 0, 0,
3913 						   SFF_DIAG_SUPPORT_OFFSET + 1,
3914 						   module_info);
3915 
3916 	if (rc)
3917 		return rc;
3918 
3919 	switch (module_info[0]) {
3920 	case SFF_MODULE_ID_SFP:
3921 		modinfo->type = RTE_ETH_MODULE_SFF_8472;
3922 		modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8472_LEN;
3923 		if (module_info[SFF_DIAG_SUPPORT_OFFSET] == 0)
3924 			modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8436_LEN;
3925 		break;
3926 	case SFF_MODULE_ID_QSFP:
3927 	case SFF_MODULE_ID_QSFP_PLUS:
3928 		modinfo->type = RTE_ETH_MODULE_SFF_8436;
3929 		modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8436_LEN;
3930 		break;
3931 	case SFF_MODULE_ID_QSFP28:
3932 		modinfo->type = RTE_ETH_MODULE_SFF_8636;
3933 		modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8636_MAX_LEN;
3934 		if (module_info[SFF8636_FLATMEM_OFFSET] & SFF8636_FLATMEM_MASK)
3935 			modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8636_LEN;
3936 		break;
3937 	default:
3938 		PMD_DRV_LOG(NOTICE, "Port %u : Unsupported module\n", dev->data->port_id);
3939 		return -ENOTSUP;
3940 	}
3941 
3942 	PMD_DRV_LOG(INFO, "Port %u : modinfo->type = %d modinfo->eeprom_len = %d\n",
3943 		    dev->data->port_id, modinfo->type, modinfo->eeprom_len);
3944 
3945 	return 0;
3946 }
3947 
3948 static int bnxt_get_module_eeprom(struct rte_eth_dev *dev,
3949 				  struct rte_dev_eeprom_info *info)
3950 {
3951 	uint8_t pg_addr[5] = { I2C_DEV_ADDR_A0, I2C_DEV_ADDR_A0 };
3952 	uint32_t offset = info->offset, length = info->length;
3953 	uint8_t module_info[SFF_DIAG_SUPPORT_OFFSET + 1];
3954 	struct bnxt *bp = dev->data->dev_private;
3955 	uint8_t *data = info->data;
3956 	uint8_t page = offset >> 7;
3957 	uint8_t max_pages = 2;
3958 	uint8_t opt_pages;
3959 	int rc;
3960 
3961 	rc = bnxt_hwrm_read_sfp_module_eeprom_info(bp, I2C_DEV_ADDR_A0, 0, 0,
3962 						   SFF_DIAG_SUPPORT_OFFSET + 1,
3963 						   module_info);
3964 	if (rc)
3965 		return rc;
3966 
3967 	switch (module_info[0]) {
3968 	case SFF_MODULE_ID_SFP:
3969 		module_info[SFF_DIAG_SUPPORT_OFFSET] = 0;
3970 		if (module_info[SFF_DIAG_SUPPORT_OFFSET]) {
3971 			pg_addr[2] = I2C_DEV_ADDR_A2;
3972 			pg_addr[3] = I2C_DEV_ADDR_A2;
3973 			max_pages = 4;
3974 		}
3975 		break;
3976 	case SFF_MODULE_ID_QSFP28:
3977 		rc = bnxt_hwrm_read_sfp_module_eeprom_info(bp, I2C_DEV_ADDR_A0, 0,
3978 							   SFF8636_OPT_PAGES_OFFSET,
3979 							   1, &opt_pages);
3980 		if (rc)
3981 			return rc;
3982 
3983 		if (opt_pages & SFF8636_PAGE1_MASK) {
3984 			pg_addr[2] = I2C_DEV_ADDR_A0;
3985 			max_pages = 3;
3986 		}
3987 		if (opt_pages & SFF8636_PAGE2_MASK) {
3988 			pg_addr[3] = I2C_DEV_ADDR_A0;
3989 			max_pages = 4;
3990 		}
3991 		if (~module_info[SFF8636_FLATMEM_OFFSET] & SFF8636_FLATMEM_MASK) {
3992 			pg_addr[4] = I2C_DEV_ADDR_A0;
3993 			max_pages = 5;
3994 		}
3995 		break;
3996 	default:
3997 		break;
3998 	}
3999 
4000 	memset(data, 0, length);
4001 
4002 	offset &= 0xff;
4003 	while (length && page < max_pages) {
4004 		uint8_t raw_page = page ? page - 1 : 0;
4005 		uint16_t chunk;
4006 
4007 		if (pg_addr[page] == I2C_DEV_ADDR_A2)
4008 			raw_page = 0;
4009 		else if (page)
4010 			offset |= 0x80;
4011 		chunk = RTE_MIN(length, 256 - offset);
4012 
4013 		if (pg_addr[page]) {
4014 			rc = bnxt_hwrm_read_sfp_module_eeprom_info(bp, pg_addr[page],
4015 								   raw_page, offset,
4016 								   chunk, data);
4017 			if (rc)
4018 				return rc;
4019 		}
4020 
4021 		data += chunk;
4022 		length -= chunk;
4023 		offset = 0;
4024 		page += 1 + (chunk > 128);
4025 	}
4026 
4027 	return length ? -EINVAL : 0;
4028 }
4029 
4030 /*
4031  * Initialization
4032  */
4033 
4034 static const struct eth_dev_ops bnxt_dev_ops = {
4035 	.dev_infos_get = bnxt_dev_info_get_op,
4036 	.dev_close = bnxt_dev_close_op,
4037 	.dev_configure = bnxt_dev_configure_op,
4038 	.dev_start = bnxt_dev_start_op,
4039 	.dev_stop = bnxt_dev_stop_op,
4040 	.dev_set_link_up = bnxt_dev_set_link_up_op,
4041 	.dev_set_link_down = bnxt_dev_set_link_down_op,
4042 	.stats_get = bnxt_stats_get_op,
4043 	.stats_reset = bnxt_stats_reset_op,
4044 	.rx_queue_setup = bnxt_rx_queue_setup_op,
4045 	.rx_queue_release = bnxt_rx_queue_release_op,
4046 	.tx_queue_setup = bnxt_tx_queue_setup_op,
4047 	.tx_queue_release = bnxt_tx_queue_release_op,
4048 	.rx_queue_intr_enable = bnxt_rx_queue_intr_enable_op,
4049 	.rx_queue_intr_disable = bnxt_rx_queue_intr_disable_op,
4050 	.reta_update = bnxt_reta_update_op,
4051 	.reta_query = bnxt_reta_query_op,
4052 	.rss_hash_update = bnxt_rss_hash_update_op,
4053 	.rss_hash_conf_get = bnxt_rss_hash_conf_get_op,
4054 	.link_update = bnxt_link_update_op,
4055 	.promiscuous_enable = bnxt_promiscuous_enable_op,
4056 	.promiscuous_disable = bnxt_promiscuous_disable_op,
4057 	.allmulticast_enable = bnxt_allmulticast_enable_op,
4058 	.allmulticast_disable = bnxt_allmulticast_disable_op,
4059 	.mac_addr_add = bnxt_mac_addr_add_op,
4060 	.mac_addr_remove = bnxt_mac_addr_remove_op,
4061 	.flow_ctrl_get = bnxt_flow_ctrl_get_op,
4062 	.flow_ctrl_set = bnxt_flow_ctrl_set_op,
4063 	.udp_tunnel_port_add  = bnxt_udp_tunnel_port_add_op,
4064 	.udp_tunnel_port_del  = bnxt_udp_tunnel_port_del_op,
4065 	.vlan_filter_set = bnxt_vlan_filter_set_op,
4066 	.vlan_offload_set = bnxt_vlan_offload_set_op,
4067 	.vlan_tpid_set = bnxt_vlan_tpid_set_op,
4068 	.vlan_pvid_set = bnxt_vlan_pvid_set_op,
4069 	.mtu_set = bnxt_mtu_set_op,
4070 	.mac_addr_set = bnxt_set_default_mac_addr_op,
4071 	.xstats_get = bnxt_dev_xstats_get_op,
4072 	.xstats_get_names = bnxt_dev_xstats_get_names_op,
4073 	.xstats_reset = bnxt_dev_xstats_reset_op,
4074 	.fw_version_get = bnxt_fw_version_get,
4075 	.set_mc_addr_list = bnxt_dev_set_mc_addr_list_op,
4076 	.rxq_info_get = bnxt_rxq_info_get_op,
4077 	.txq_info_get = bnxt_txq_info_get_op,
4078 	.rx_burst_mode_get = bnxt_rx_burst_mode_get,
4079 	.tx_burst_mode_get = bnxt_tx_burst_mode_get,
4080 	.dev_led_on = bnxt_dev_led_on_op,
4081 	.dev_led_off = bnxt_dev_led_off_op,
4082 	.rx_queue_start = bnxt_rx_queue_start,
4083 	.rx_queue_stop = bnxt_rx_queue_stop,
4084 	.tx_queue_start = bnxt_tx_queue_start,
4085 	.tx_queue_stop = bnxt_tx_queue_stop,
4086 	.flow_ops_get = bnxt_flow_ops_get_op,
4087 	.dev_supported_ptypes_get = bnxt_dev_supported_ptypes_get_op,
4088 	.get_eeprom_length    = bnxt_get_eeprom_length_op,
4089 	.get_eeprom           = bnxt_get_eeprom_op,
4090 	.set_eeprom           = bnxt_set_eeprom_op,
4091 	.get_module_info = bnxt_get_module_info,
4092 	.get_module_eeprom = bnxt_get_module_eeprom,
4093 	.timesync_enable      = bnxt_timesync_enable,
4094 	.timesync_disable     = bnxt_timesync_disable,
4095 	.timesync_read_time   = bnxt_timesync_read_time,
4096 	.timesync_write_time   = bnxt_timesync_write_time,
4097 	.timesync_adjust_time = bnxt_timesync_adjust_time,
4098 	.timesync_read_rx_timestamp = bnxt_timesync_read_rx_timestamp,
4099 	.timesync_read_tx_timestamp = bnxt_timesync_read_tx_timestamp,
4100 };
4101 
4102 static uint32_t bnxt_map_reset_regs(struct bnxt *bp, uint32_t reg)
4103 {
4104 	uint32_t offset;
4105 
4106 	/* Only pre-map the reset GRC registers using window 3 */
4107 	rte_write32(reg & 0xfffff000, (uint8_t *)bp->bar0 +
4108 		    BNXT_GRCPF_REG_WINDOW_BASE_OUT + 8);
4109 
4110 	offset = BNXT_GRCP_WINDOW_3_BASE + (reg & 0xffc);
4111 
4112 	return offset;
4113 }
4114 
4115 int bnxt_map_fw_health_status_regs(struct bnxt *bp)
4116 {
4117 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4118 	uint32_t reg_base = 0xffffffff;
4119 	int i;
4120 
4121 	/* Only pre-map the monitoring GRC registers using window 2 */
4122 	for (i = 0; i < BNXT_FW_STATUS_REG_CNT; i++) {
4123 		uint32_t reg = info->status_regs[i];
4124 
4125 		if (BNXT_FW_STATUS_REG_TYPE(reg) != BNXT_FW_STATUS_REG_TYPE_GRC)
4126 			continue;
4127 
4128 		if (reg_base == 0xffffffff)
4129 			reg_base = reg & 0xfffff000;
4130 		if ((reg & 0xfffff000) != reg_base)
4131 			return -ERANGE;
4132 
4133 		/* Use mask 0xffc as the Lower 2 bits indicates
4134 		 * address space location
4135 		 */
4136 		info->mapped_status_regs[i] = BNXT_GRCP_WINDOW_2_BASE +
4137 						(reg & 0xffc);
4138 	}
4139 
4140 	if (reg_base == 0xffffffff)
4141 		return 0;
4142 
4143 	rte_write32(reg_base, (uint8_t *)bp->bar0 +
4144 		    BNXT_GRCPF_REG_WINDOW_BASE_OUT + 4);
4145 
4146 	return 0;
4147 }
4148 
4149 static void bnxt_write_fw_reset_reg(struct bnxt *bp, uint32_t index)
4150 {
4151 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4152 	uint32_t delay = info->delay_after_reset[index];
4153 	uint32_t val = info->reset_reg_val[index];
4154 	uint32_t reg = info->reset_reg[index];
4155 	uint32_t type, offset;
4156 	int ret;
4157 
4158 	type = BNXT_FW_STATUS_REG_TYPE(reg);
4159 	offset = BNXT_FW_STATUS_REG_OFF(reg);
4160 
4161 	switch (type) {
4162 	case BNXT_FW_STATUS_REG_TYPE_CFG:
4163 		ret = rte_pci_write_config(bp->pdev, &val, sizeof(val), offset);
4164 		if (ret < 0) {
4165 			PMD_DRV_LOG(ERR, "Failed to write %#x at PCI offset %#x",
4166 				    val, offset);
4167 			return;
4168 		}
4169 		break;
4170 	case BNXT_FW_STATUS_REG_TYPE_GRC:
4171 		offset = bnxt_map_reset_regs(bp, offset);
4172 		rte_write32(val, (uint8_t *)bp->bar0 + offset);
4173 		break;
4174 	case BNXT_FW_STATUS_REG_TYPE_BAR0:
4175 		rte_write32(val, (uint8_t *)bp->bar0 + offset);
4176 		break;
4177 	}
4178 	/* wait on a specific interval of time until core reset is complete */
4179 	if (delay)
4180 		rte_delay_ms(delay);
4181 }
4182 
4183 static void bnxt_dev_cleanup(struct bnxt *bp)
4184 {
4185 	bp->eth_dev->data->dev_link.link_status = 0;
4186 	bp->link_info->link_up = 0;
4187 	if (bp->eth_dev->data->dev_started)
4188 		bnxt_dev_stop(bp->eth_dev);
4189 
4190 	bnxt_uninit_resources(bp, true);
4191 }
4192 
4193 static int
4194 bnxt_check_fw_reset_done(struct bnxt *bp)
4195 {
4196 	int timeout = bp->fw_reset_max_msecs;
4197 	uint16_t val = 0;
4198 	int rc;
4199 
4200 	do {
4201 		rc = rte_pci_read_config(bp->pdev, &val, sizeof(val), PCI_SUBSYSTEM_ID_OFFSET);
4202 		if (rc < 0) {
4203 			PMD_DRV_LOG(ERR, "Failed to read PCI offset 0x%x", PCI_SUBSYSTEM_ID_OFFSET);
4204 			return rc;
4205 		}
4206 		if (val != 0xffff)
4207 			break;
4208 		rte_delay_ms(1);
4209 	} while (timeout--);
4210 
4211 	if (val == 0xffff) {
4212 		PMD_DRV_LOG(ERR, "Firmware reset aborted, PCI config space invalid\n");
4213 		return -1;
4214 	}
4215 
4216 	return 0;
4217 }
4218 
4219 static int bnxt_restore_vlan_filters(struct bnxt *bp)
4220 {
4221 	struct rte_eth_dev *dev = bp->eth_dev;
4222 	struct rte_vlan_filter_conf *vfc;
4223 	int vidx, vbit, rc;
4224 	uint16_t vlan_id;
4225 
4226 	for (vlan_id = 1; vlan_id <= RTE_ETHER_MAX_VLAN_ID; vlan_id++) {
4227 		vfc = &dev->data->vlan_filter_conf;
4228 		vidx = vlan_id / 64;
4229 		vbit = vlan_id % 64;
4230 
4231 		/* Each bit corresponds to a VLAN id */
4232 		if (vfc->ids[vidx] & (UINT64_C(1) << vbit)) {
4233 			rc = bnxt_add_vlan_filter(bp, vlan_id);
4234 			if (rc)
4235 				return rc;
4236 		}
4237 	}
4238 
4239 	return 0;
4240 }
4241 
4242 static int bnxt_restore_mac_filters(struct bnxt *bp)
4243 {
4244 	struct rte_eth_dev *dev = bp->eth_dev;
4245 	struct rte_eth_dev_info dev_info;
4246 	struct rte_ether_addr *addr;
4247 	uint64_t pool_mask;
4248 	uint32_t pool = 0;
4249 	uint32_t i;
4250 	int rc;
4251 
4252 	if (BNXT_VF(bp) && !BNXT_VF_IS_TRUSTED(bp))
4253 		return 0;
4254 
4255 	rc = bnxt_dev_info_get_op(dev, &dev_info);
4256 	if (rc)
4257 		return rc;
4258 
4259 	/* replay MAC address configuration */
4260 	for (i = 1; i < dev_info.max_mac_addrs; i++) {
4261 		addr = &dev->data->mac_addrs[i];
4262 
4263 		/* skip zero address */
4264 		if (rte_is_zero_ether_addr(addr))
4265 			continue;
4266 
4267 		pool = 0;
4268 		pool_mask = dev->data->mac_pool_sel[i];
4269 
4270 		do {
4271 			if (pool_mask & 1ULL) {
4272 				rc = bnxt_mac_addr_add_op(dev, addr, i, pool);
4273 				if (rc)
4274 					return rc;
4275 			}
4276 			pool_mask >>= 1;
4277 			pool++;
4278 		} while (pool_mask);
4279 	}
4280 
4281 	return 0;
4282 }
4283 
4284 static int bnxt_restore_filters(struct bnxt *bp)
4285 {
4286 	struct rte_eth_dev *dev = bp->eth_dev;
4287 	int ret = 0;
4288 
4289 	if (dev->data->all_multicast) {
4290 		ret = bnxt_allmulticast_enable_op(dev);
4291 		if (ret)
4292 			return ret;
4293 	}
4294 	if (dev->data->promiscuous) {
4295 		ret = bnxt_promiscuous_enable_op(dev);
4296 		if (ret)
4297 			return ret;
4298 	}
4299 
4300 	ret = bnxt_restore_mac_filters(bp);
4301 	if (ret)
4302 		return ret;
4303 
4304 	ret = bnxt_restore_vlan_filters(bp);
4305 	/* TODO restore other filters as well */
4306 	return ret;
4307 }
4308 
4309 static int bnxt_check_fw_ready(struct bnxt *bp)
4310 {
4311 	int timeout = bp->fw_reset_max_msecs;
4312 	int rc = 0;
4313 
4314 	do {
4315 		rc = bnxt_hwrm_poll_ver_get(bp);
4316 		if (rc == 0)
4317 			break;
4318 		rte_delay_ms(BNXT_FW_READY_WAIT_INTERVAL);
4319 		timeout -= BNXT_FW_READY_WAIT_INTERVAL;
4320 	} while (rc && timeout > 0);
4321 
4322 	if (rc)
4323 		PMD_DRV_LOG(ERR, "FW is not Ready after reset\n");
4324 
4325 	return rc;
4326 }
4327 
4328 static void bnxt_dev_recover(void *arg)
4329 {
4330 	struct bnxt *bp = arg;
4331 	int rc = 0;
4332 
4333 	pthread_mutex_lock(&bp->err_recovery_lock);
4334 
4335 	if (!bp->fw_reset_min_msecs) {
4336 		rc = bnxt_check_fw_reset_done(bp);
4337 		if (rc)
4338 			goto err;
4339 	}
4340 
4341 	/* Clear Error flag so that device re-init should happen */
4342 	bp->flags &= ~BNXT_FLAG_FATAL_ERROR;
4343 
4344 	rc = bnxt_check_fw_ready(bp);
4345 	if (rc)
4346 		goto err;
4347 
4348 	rc = bnxt_init_resources(bp, true);
4349 	if (rc) {
4350 		PMD_DRV_LOG(ERR,
4351 			    "Failed to initialize resources after reset\n");
4352 		goto err;
4353 	}
4354 	/* clear reset flag as the device is initialized now */
4355 	bp->flags &= ~BNXT_FLAG_FW_RESET;
4356 
4357 	rc = bnxt_dev_start_op(bp->eth_dev);
4358 	if (rc) {
4359 		PMD_DRV_LOG(ERR, "Failed to start port after reset\n");
4360 		goto err_start;
4361 	}
4362 
4363 	rc = bnxt_restore_filters(bp);
4364 	if (rc)
4365 		goto err_start;
4366 
4367 	PMD_DRV_LOG(INFO, "Recovered from FW reset\n");
4368 	pthread_mutex_unlock(&bp->err_recovery_lock);
4369 
4370 	return;
4371 err_start:
4372 	bnxt_dev_stop(bp->eth_dev);
4373 err:
4374 	bp->flags |= BNXT_FLAG_FATAL_ERROR;
4375 	bnxt_uninit_resources(bp, false);
4376 	if (bp->eth_dev->data->dev_conf.intr_conf.rmv)
4377 		rte_eth_dev_callback_process(bp->eth_dev,
4378 					     RTE_ETH_EVENT_INTR_RMV,
4379 					     NULL);
4380 	pthread_mutex_unlock(&bp->err_recovery_lock);
4381 	PMD_DRV_LOG(ERR, "Failed to recover from FW reset\n");
4382 }
4383 
4384 void bnxt_dev_reset_and_resume(void *arg)
4385 {
4386 	struct bnxt *bp = arg;
4387 	uint32_t us = US_PER_MS * bp->fw_reset_min_msecs;
4388 	uint16_t val = 0;
4389 	int rc;
4390 
4391 	bnxt_dev_cleanup(bp);
4392 
4393 	bnxt_wait_for_device_shutdown(bp);
4394 
4395 	/* During some fatal firmware error conditions, the PCI config space
4396 	 * register 0x2e which normally contains the subsystem ID will become
4397 	 * 0xffff. This register will revert back to the normal value after
4398 	 * the chip has completed core reset. If we detect this condition,
4399 	 * we can poll this config register immediately for the value to revert.
4400 	 */
4401 	if (bp->flags & BNXT_FLAG_FATAL_ERROR) {
4402 		rc = rte_pci_read_config(bp->pdev, &val, sizeof(val), PCI_SUBSYSTEM_ID_OFFSET);
4403 		if (rc < 0) {
4404 			PMD_DRV_LOG(ERR, "Failed to read PCI offset 0x%x", PCI_SUBSYSTEM_ID_OFFSET);
4405 			return;
4406 		}
4407 		if (val == 0xffff) {
4408 			bp->fw_reset_min_msecs = 0;
4409 			us = 1;
4410 		}
4411 	}
4412 
4413 	rc = rte_eal_alarm_set(us, bnxt_dev_recover, (void *)bp);
4414 	if (rc)
4415 		PMD_DRV_LOG(ERR, "Error setting recovery alarm");
4416 }
4417 
4418 uint32_t bnxt_read_fw_status_reg(struct bnxt *bp, uint32_t index)
4419 {
4420 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4421 	uint32_t reg = info->status_regs[index];
4422 	uint32_t type, offset, val = 0;
4423 	int ret = 0;
4424 
4425 	type = BNXT_FW_STATUS_REG_TYPE(reg);
4426 	offset = BNXT_FW_STATUS_REG_OFF(reg);
4427 
4428 	switch (type) {
4429 	case BNXT_FW_STATUS_REG_TYPE_CFG:
4430 		ret = rte_pci_read_config(bp->pdev, &val, sizeof(val), offset);
4431 		if (ret < 0)
4432 			PMD_DRV_LOG(ERR, "Failed to read PCI offset %#x",
4433 				    offset);
4434 		break;
4435 	case BNXT_FW_STATUS_REG_TYPE_GRC:
4436 		offset = info->mapped_status_regs[index];
4437 		/* FALLTHROUGH */
4438 	case BNXT_FW_STATUS_REG_TYPE_BAR0:
4439 		val = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
4440 				       offset));
4441 		break;
4442 	}
4443 
4444 	return val;
4445 }
4446 
4447 static int bnxt_fw_reset_all(struct bnxt *bp)
4448 {
4449 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4450 	uint32_t i;
4451 	int rc = 0;
4452 
4453 	if (info->flags & BNXT_FLAG_ERROR_RECOVERY_HOST) {
4454 		/* Reset through primary function driver */
4455 		for (i = 0; i < info->reg_array_cnt; i++)
4456 			bnxt_write_fw_reset_reg(bp, i);
4457 		/* Wait for time specified by FW after triggering reset */
4458 		rte_delay_ms(info->primary_func_wait_period_after_reset);
4459 	} else if (info->flags & BNXT_FLAG_ERROR_RECOVERY_CO_CPU) {
4460 		/* Reset with the help of Kong processor */
4461 		rc = bnxt_hwrm_fw_reset(bp);
4462 		if (rc)
4463 			PMD_DRV_LOG(ERR, "Failed to reset FW\n");
4464 	}
4465 
4466 	return rc;
4467 }
4468 
4469 static void bnxt_fw_reset_cb(void *arg)
4470 {
4471 	struct bnxt *bp = arg;
4472 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4473 	int rc = 0;
4474 
4475 	/* Only Primary function can do FW reset */
4476 	if (bnxt_is_primary_func(bp) &&
4477 	    bnxt_is_recovery_enabled(bp)) {
4478 		rc = bnxt_fw_reset_all(bp);
4479 		if (rc) {
4480 			PMD_DRV_LOG(ERR, "Adapter recovery failed\n");
4481 			return;
4482 		}
4483 	}
4484 
4485 	/* if recovery method is ERROR_RECOVERY_CO_CPU, KONG will send
4486 	 * EXCEPTION_FATAL_ASYNC event to all the functions
4487 	 * (including MASTER FUNC). After receiving this Async, all the active
4488 	 * drivers should treat this case as FW initiated recovery
4489 	 */
4490 	if (info->flags & BNXT_FLAG_ERROR_RECOVERY_HOST) {
4491 		bp->fw_reset_min_msecs = BNXT_MIN_FW_READY_TIMEOUT;
4492 		bp->fw_reset_max_msecs = BNXT_MAX_FW_RESET_TIMEOUT;
4493 
4494 		/* To recover from error */
4495 		rte_eal_alarm_set(US_PER_MS, bnxt_dev_reset_and_resume,
4496 				  (void *)bp);
4497 	}
4498 }
4499 
4500 /* Driver should poll FW heartbeat, reset_counter with the frequency
4501  * advertised by FW in HWRM_ERROR_RECOVERY_QCFG.
4502  * When the driver detects heartbeat stop or change in reset_counter,
4503  * it has to trigger a reset to recover from the error condition.
4504  * A “primary function” is the function who will have the privilege to
4505  * initiate the chimp reset. The primary function will be elected by the
4506  * firmware and will be notified through async message.
4507  */
4508 static void bnxt_check_fw_health(void *arg)
4509 {
4510 	struct bnxt *bp = arg;
4511 	struct bnxt_error_recovery_info *info = bp->recovery_info;
4512 	uint32_t val = 0, wait_msec;
4513 
4514 	if (!info || !bnxt_is_recovery_enabled(bp) ||
4515 	    is_bnxt_in_error(bp))
4516 		return;
4517 
4518 	val = bnxt_read_fw_status_reg(bp, BNXT_FW_HEARTBEAT_CNT_REG);
4519 	if (val == info->last_heart_beat)
4520 		goto reset;
4521 
4522 	info->last_heart_beat = val;
4523 
4524 	val = bnxt_read_fw_status_reg(bp, BNXT_FW_RECOVERY_CNT_REG);
4525 	if (val != info->last_reset_counter)
4526 		goto reset;
4527 
4528 	info->last_reset_counter = val;
4529 
4530 	rte_eal_alarm_set(US_PER_MS * info->driver_polling_freq,
4531 			  bnxt_check_fw_health, (void *)bp);
4532 
4533 	return;
4534 reset:
4535 	/* Stop DMA to/from device */
4536 	bp->flags |= BNXT_FLAG_FATAL_ERROR;
4537 	bp->flags |= BNXT_FLAG_FW_RESET;
4538 
4539 	bnxt_stop_rxtx(bp);
4540 
4541 	PMD_DRV_LOG(ERR, "Detected FW dead condition\n");
4542 
4543 	if (bnxt_is_primary_func(bp))
4544 		wait_msec = info->primary_func_wait_period;
4545 	else
4546 		wait_msec = info->normal_func_wait_period;
4547 
4548 	rte_eal_alarm_set(US_PER_MS * wait_msec,
4549 			  bnxt_fw_reset_cb, (void *)bp);
4550 }
4551 
4552 void bnxt_schedule_fw_health_check(struct bnxt *bp)
4553 {
4554 	uint32_t polling_freq;
4555 
4556 	pthread_mutex_lock(&bp->health_check_lock);
4557 
4558 	if (!bnxt_is_recovery_enabled(bp))
4559 		goto done;
4560 
4561 	if (bp->flags & BNXT_FLAG_FW_HEALTH_CHECK_SCHEDULED)
4562 		goto done;
4563 
4564 	polling_freq = bp->recovery_info->driver_polling_freq;
4565 
4566 	rte_eal_alarm_set(US_PER_MS * polling_freq,
4567 			  bnxt_check_fw_health, (void *)bp);
4568 	bp->flags |= BNXT_FLAG_FW_HEALTH_CHECK_SCHEDULED;
4569 
4570 done:
4571 	pthread_mutex_unlock(&bp->health_check_lock);
4572 }
4573 
4574 static void bnxt_cancel_fw_health_check(struct bnxt *bp)
4575 {
4576 	rte_eal_alarm_cancel(bnxt_check_fw_health, (void *)bp);
4577 	bp->flags &= ~BNXT_FLAG_FW_HEALTH_CHECK_SCHEDULED;
4578 }
4579 
4580 static bool bnxt_vf_pciid(uint16_t device_id)
4581 {
4582 	switch (device_id) {
4583 	case BROADCOM_DEV_ID_57304_VF:
4584 	case BROADCOM_DEV_ID_57406_VF:
4585 	case BROADCOM_DEV_ID_5731X_VF:
4586 	case BROADCOM_DEV_ID_5741X_VF:
4587 	case BROADCOM_DEV_ID_57414_VF:
4588 	case BROADCOM_DEV_ID_STRATUS_NIC_VF1:
4589 	case BROADCOM_DEV_ID_STRATUS_NIC_VF2:
4590 	case BROADCOM_DEV_ID_58802_VF:
4591 	case BROADCOM_DEV_ID_57500_VF1:
4592 	case BROADCOM_DEV_ID_57500_VF2:
4593 	case BROADCOM_DEV_ID_58818_VF:
4594 		/* FALLTHROUGH */
4595 		return true;
4596 	default:
4597 		return false;
4598 	}
4599 }
4600 
4601 /* Phase 5 device */
4602 static bool bnxt_p5_device(uint16_t device_id)
4603 {
4604 	switch (device_id) {
4605 	case BROADCOM_DEV_ID_57508:
4606 	case BROADCOM_DEV_ID_57504:
4607 	case BROADCOM_DEV_ID_57502:
4608 	case BROADCOM_DEV_ID_57508_MF1:
4609 	case BROADCOM_DEV_ID_57504_MF1:
4610 	case BROADCOM_DEV_ID_57502_MF1:
4611 	case BROADCOM_DEV_ID_57508_MF2:
4612 	case BROADCOM_DEV_ID_57504_MF2:
4613 	case BROADCOM_DEV_ID_57502_MF2:
4614 	case BROADCOM_DEV_ID_57500_VF1:
4615 	case BROADCOM_DEV_ID_57500_VF2:
4616 	case BROADCOM_DEV_ID_58812:
4617 	case BROADCOM_DEV_ID_58814:
4618 	case BROADCOM_DEV_ID_58818:
4619 	case BROADCOM_DEV_ID_58818_VF:
4620 		/* FALLTHROUGH */
4621 		return true;
4622 	default:
4623 		return false;
4624 	}
4625 }
4626 
4627 bool bnxt_stratus_device(struct bnxt *bp)
4628 {
4629 	uint16_t device_id = bp->pdev->id.device_id;
4630 
4631 	switch (device_id) {
4632 	case BROADCOM_DEV_ID_STRATUS_NIC:
4633 	case BROADCOM_DEV_ID_STRATUS_NIC_VF1:
4634 	case BROADCOM_DEV_ID_STRATUS_NIC_VF2:
4635 		/* FALLTHROUGH */
4636 		return true;
4637 	default:
4638 		return false;
4639 	}
4640 }
4641 
4642 static int bnxt_map_pci_bars(struct rte_eth_dev *eth_dev)
4643 {
4644 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
4645 	struct bnxt *bp = eth_dev->data->dev_private;
4646 
4647 	/* enable device (incl. PCI PM wakeup), and bus-mastering */
4648 	bp->bar0 = (void *)pci_dev->mem_resource[0].addr;
4649 	bp->doorbell_base = (void *)pci_dev->mem_resource[2].addr;
4650 	if (!bp->bar0 || !bp->doorbell_base) {
4651 		PMD_DRV_LOG(ERR, "Unable to access Hardware\n");
4652 		return -ENODEV;
4653 	}
4654 
4655 	bp->eth_dev = eth_dev;
4656 	bp->pdev = pci_dev;
4657 
4658 	return 0;
4659 }
4660 
4661 static int bnxt_alloc_ctx_mem_blk(struct bnxt *bp,
4662 				  struct bnxt_ctx_pg_info *ctx_pg,
4663 				  uint32_t mem_size,
4664 				  const char *suffix,
4665 				  uint16_t idx)
4666 {
4667 	struct bnxt_ring_mem_info *rmem = &ctx_pg->ring_mem;
4668 	const struct rte_memzone *mz = NULL;
4669 	char mz_name[RTE_MEMZONE_NAMESIZE];
4670 	rte_iova_t mz_phys_addr;
4671 	uint64_t valid_bits = 0;
4672 	uint32_t sz;
4673 	int i;
4674 
4675 	if (!mem_size)
4676 		return 0;
4677 
4678 	rmem->nr_pages = RTE_ALIGN_MUL_CEIL(mem_size, BNXT_PAGE_SIZE) /
4679 			 BNXT_PAGE_SIZE;
4680 	rmem->page_size = BNXT_PAGE_SIZE;
4681 	rmem->pg_arr = ctx_pg->ctx_pg_arr;
4682 	rmem->dma_arr = ctx_pg->ctx_dma_arr;
4683 	rmem->flags = BNXT_RMEM_VALID_PTE_FLAG;
4684 
4685 	valid_bits = PTU_PTE_VALID;
4686 
4687 	if (rmem->nr_pages > 1) {
4688 		snprintf(mz_name, RTE_MEMZONE_NAMESIZE,
4689 			 "bnxt_ctx_pg_tbl%s_%x_%d",
4690 			 suffix, idx, bp->eth_dev->data->port_id);
4691 		mz_name[RTE_MEMZONE_NAMESIZE - 1] = 0;
4692 		mz = rte_memzone_lookup(mz_name);
4693 		if (!mz) {
4694 			mz = rte_memzone_reserve_aligned(mz_name,
4695 						rmem->nr_pages * 8,
4696 						bp->eth_dev->device->numa_node,
4697 						RTE_MEMZONE_2MB |
4698 						RTE_MEMZONE_SIZE_HINT_ONLY |
4699 						RTE_MEMZONE_IOVA_CONTIG,
4700 						BNXT_PAGE_SIZE);
4701 			if (mz == NULL)
4702 				return -ENOMEM;
4703 		}
4704 
4705 		memset(mz->addr, 0, mz->len);
4706 		mz_phys_addr = mz->iova;
4707 
4708 		rmem->pg_tbl = mz->addr;
4709 		rmem->pg_tbl_map = mz_phys_addr;
4710 		rmem->pg_tbl_mz = mz;
4711 	}
4712 
4713 	snprintf(mz_name, RTE_MEMZONE_NAMESIZE, "bnxt_ctx_%s_%x_%d",
4714 		 suffix, idx, bp->eth_dev->data->port_id);
4715 	mz = rte_memzone_lookup(mz_name);
4716 	if (!mz) {
4717 		mz = rte_memzone_reserve_aligned(mz_name,
4718 						 mem_size,
4719 						 bp->eth_dev->device->numa_node,
4720 						 RTE_MEMZONE_1GB |
4721 						 RTE_MEMZONE_SIZE_HINT_ONLY |
4722 						 RTE_MEMZONE_IOVA_CONTIG,
4723 						 BNXT_PAGE_SIZE);
4724 		if (mz == NULL)
4725 			return -ENOMEM;
4726 	}
4727 
4728 	memset(mz->addr, 0, mz->len);
4729 	mz_phys_addr = mz->iova;
4730 
4731 	for (sz = 0, i = 0; sz < mem_size; sz += BNXT_PAGE_SIZE, i++) {
4732 		rmem->pg_arr[i] = ((char *)mz->addr) + sz;
4733 		rmem->dma_arr[i] = mz_phys_addr + sz;
4734 
4735 		if (rmem->nr_pages > 1) {
4736 			if (i == rmem->nr_pages - 2 &&
4737 			    (rmem->flags & BNXT_RMEM_RING_PTE_FLAG))
4738 				valid_bits |= PTU_PTE_NEXT_TO_LAST;
4739 			else if (i == rmem->nr_pages - 1 &&
4740 				 (rmem->flags & BNXT_RMEM_RING_PTE_FLAG))
4741 				valid_bits |= PTU_PTE_LAST;
4742 
4743 			rmem->pg_tbl[i] = rte_cpu_to_le_64(rmem->dma_arr[i] |
4744 							   valid_bits);
4745 		}
4746 	}
4747 
4748 	rmem->mz = mz;
4749 	if (rmem->vmem_size)
4750 		rmem->vmem = (void **)mz->addr;
4751 	rmem->dma_arr[0] = mz_phys_addr;
4752 	return 0;
4753 }
4754 
4755 static void bnxt_free_ctx_mem(struct bnxt *bp)
4756 {
4757 	int i;
4758 
4759 	if (!bp->ctx || !(bp->ctx->flags & BNXT_CTX_FLAG_INITED))
4760 		return;
4761 
4762 	bp->ctx->flags &= ~BNXT_CTX_FLAG_INITED;
4763 	rte_memzone_free(bp->ctx->qp_mem.ring_mem.mz);
4764 	rte_memzone_free(bp->ctx->srq_mem.ring_mem.mz);
4765 	rte_memzone_free(bp->ctx->cq_mem.ring_mem.mz);
4766 	rte_memzone_free(bp->ctx->vnic_mem.ring_mem.mz);
4767 	rte_memzone_free(bp->ctx->stat_mem.ring_mem.mz);
4768 	rte_memzone_free(bp->ctx->qp_mem.ring_mem.pg_tbl_mz);
4769 	rte_memzone_free(bp->ctx->srq_mem.ring_mem.pg_tbl_mz);
4770 	rte_memzone_free(bp->ctx->cq_mem.ring_mem.pg_tbl_mz);
4771 	rte_memzone_free(bp->ctx->vnic_mem.ring_mem.pg_tbl_mz);
4772 	rte_memzone_free(bp->ctx->stat_mem.ring_mem.pg_tbl_mz);
4773 
4774 	for (i = 0; i < bp->ctx->tqm_fp_rings_count + 1; i++) {
4775 		if (bp->ctx->tqm_mem[i])
4776 			rte_memzone_free(bp->ctx->tqm_mem[i]->ring_mem.mz);
4777 	}
4778 
4779 	rte_free(bp->ctx);
4780 	bp->ctx = NULL;
4781 }
4782 
4783 #define bnxt_roundup(x, y)   ((((x) + ((y) - 1)) / (y)) * (y))
4784 
4785 #define min_t(type, x, y) ({                    \
4786 	type __min1 = (x);                      \
4787 	type __min2 = (y);                      \
4788 	__min1 < __min2 ? __min1 : __min2; })
4789 
4790 #define max_t(type, x, y) ({                    \
4791 	type __max1 = (x);                      \
4792 	type __max2 = (y);                      \
4793 	__max1 > __max2 ? __max1 : __max2; })
4794 
4795 #define clamp_t(type, _x, min, max)     min_t(type, max_t(type, _x, min), max)
4796 
4797 int bnxt_alloc_ctx_mem(struct bnxt *bp)
4798 {
4799 	struct bnxt_ctx_pg_info *ctx_pg;
4800 	struct bnxt_ctx_mem_info *ctx;
4801 	uint32_t mem_size, ena, entries;
4802 	uint32_t entries_sp, min;
4803 	int i, rc;
4804 
4805 	rc = bnxt_hwrm_func_backing_store_qcaps(bp);
4806 	if (rc) {
4807 		PMD_DRV_LOG(ERR, "Query context mem capability failed\n");
4808 		return rc;
4809 	}
4810 	ctx = bp->ctx;
4811 	if (!ctx || (ctx->flags & BNXT_CTX_FLAG_INITED))
4812 		return 0;
4813 
4814 	ctx_pg = &ctx->qp_mem;
4815 	ctx_pg->entries = ctx->qp_min_qp1_entries + ctx->qp_max_l2_entries;
4816 	if (ctx->qp_entry_size) {
4817 		mem_size = ctx->qp_entry_size * ctx_pg->entries;
4818 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "qp_mem", 0);
4819 		if (rc)
4820 			return rc;
4821 	}
4822 
4823 	ctx_pg = &ctx->srq_mem;
4824 	ctx_pg->entries = ctx->srq_max_l2_entries;
4825 	if (ctx->srq_entry_size) {
4826 		mem_size = ctx->srq_entry_size * ctx_pg->entries;
4827 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "srq_mem", 0);
4828 		if (rc)
4829 			return rc;
4830 	}
4831 
4832 	ctx_pg = &ctx->cq_mem;
4833 	ctx_pg->entries = ctx->cq_max_l2_entries;
4834 	if (ctx->cq_entry_size) {
4835 		mem_size = ctx->cq_entry_size * ctx_pg->entries;
4836 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "cq_mem", 0);
4837 		if (rc)
4838 			return rc;
4839 	}
4840 
4841 	ctx_pg = &ctx->vnic_mem;
4842 	ctx_pg->entries = ctx->vnic_max_vnic_entries +
4843 		ctx->vnic_max_ring_table_entries;
4844 	if (ctx->vnic_entry_size) {
4845 		mem_size = ctx->vnic_entry_size * ctx_pg->entries;
4846 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "vnic_mem", 0);
4847 		if (rc)
4848 			return rc;
4849 	}
4850 
4851 	ctx_pg = &ctx->stat_mem;
4852 	ctx_pg->entries = ctx->stat_max_entries;
4853 	if (ctx->stat_entry_size) {
4854 		mem_size = ctx->stat_entry_size * ctx_pg->entries;
4855 		rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "stat_mem", 0);
4856 		if (rc)
4857 			return rc;
4858 	}
4859 
4860 	min = ctx->tqm_min_entries_per_ring;
4861 
4862 	entries_sp = ctx->qp_max_l2_entries +
4863 		     ctx->vnic_max_vnic_entries +
4864 		     2 * ctx->qp_min_qp1_entries + min;
4865 	entries_sp = bnxt_roundup(entries_sp, ctx->tqm_entries_multiple);
4866 
4867 	entries = ctx->qp_max_l2_entries + ctx->qp_min_qp1_entries;
4868 	entries = bnxt_roundup(entries, ctx->tqm_entries_multiple);
4869 	entries = clamp_t(uint32_t, entries, min,
4870 			  ctx->tqm_max_entries_per_ring);
4871 	for (i = 0, ena = 0; i < ctx->tqm_fp_rings_count + 1; i++) {
4872 		/* i=0 is for TQM_SP. i=1 to i=8 applies to RING0 to RING7.
4873 		 * i > 8 is other ext rings.
4874 		 */
4875 		ctx_pg = ctx->tqm_mem[i];
4876 		ctx_pg->entries = i ? entries : entries_sp;
4877 		if (ctx->tqm_entry_size) {
4878 			mem_size = ctx->tqm_entry_size * ctx_pg->entries;
4879 			rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size,
4880 						    "tqm_mem", i);
4881 			if (rc)
4882 				return rc;
4883 		}
4884 		if (i < BNXT_MAX_TQM_LEGACY_RINGS)
4885 			ena |= HWRM_FUNC_BACKING_STORE_CFG_INPUT_ENABLES_TQM_SP << i;
4886 		else
4887 			ena |= HWRM_FUNC_BACKING_STORE_CFG_INPUT_ENABLES_TQM_RING8;
4888 	}
4889 
4890 	ena |= FUNC_BACKING_STORE_CFG_INPUT_DFLT_ENABLES;
4891 	rc = bnxt_hwrm_func_backing_store_cfg(bp, ena);
4892 	if (rc)
4893 		PMD_DRV_LOG(ERR,
4894 			    "Failed to configure context mem: rc = %d\n", rc);
4895 	else
4896 		ctx->flags |= BNXT_CTX_FLAG_INITED;
4897 
4898 	return rc;
4899 }
4900 
4901 static int bnxt_alloc_stats_mem(struct bnxt *bp)
4902 {
4903 	struct rte_pci_device *pci_dev = bp->pdev;
4904 	char mz_name[RTE_MEMZONE_NAMESIZE];
4905 	const struct rte_memzone *mz = NULL;
4906 	uint32_t total_alloc_len;
4907 	rte_iova_t mz_phys_addr;
4908 
4909 	if (pci_dev->id.device_id == BROADCOM_DEV_ID_NS2)
4910 		return 0;
4911 
4912 	snprintf(mz_name, RTE_MEMZONE_NAMESIZE,
4913 		 "bnxt_" PCI_PRI_FMT "-%s", pci_dev->addr.domain,
4914 		 pci_dev->addr.bus, pci_dev->addr.devid,
4915 		 pci_dev->addr.function, "rx_port_stats");
4916 	mz_name[RTE_MEMZONE_NAMESIZE - 1] = 0;
4917 	mz = rte_memzone_lookup(mz_name);
4918 	total_alloc_len =
4919 		RTE_CACHE_LINE_ROUNDUP(sizeof(struct rx_port_stats) +
4920 				       sizeof(struct rx_port_stats_ext) + 512);
4921 	if (!mz) {
4922 		mz = rte_memzone_reserve(mz_name, total_alloc_len,
4923 					 SOCKET_ID_ANY,
4924 					 RTE_MEMZONE_2MB |
4925 					 RTE_MEMZONE_SIZE_HINT_ONLY |
4926 					 RTE_MEMZONE_IOVA_CONTIG);
4927 		if (mz == NULL)
4928 			return -ENOMEM;
4929 	}
4930 	memset(mz->addr, 0, mz->len);
4931 	mz_phys_addr = mz->iova;
4932 
4933 	bp->rx_mem_zone = (const void *)mz;
4934 	bp->hw_rx_port_stats = mz->addr;
4935 	bp->hw_rx_port_stats_map = mz_phys_addr;
4936 
4937 	snprintf(mz_name, RTE_MEMZONE_NAMESIZE,
4938 		 "bnxt_" PCI_PRI_FMT "-%s", pci_dev->addr.domain,
4939 		 pci_dev->addr.bus, pci_dev->addr.devid,
4940 		 pci_dev->addr.function, "tx_port_stats");
4941 	mz_name[RTE_MEMZONE_NAMESIZE - 1] = 0;
4942 	mz = rte_memzone_lookup(mz_name);
4943 	total_alloc_len =
4944 		RTE_CACHE_LINE_ROUNDUP(sizeof(struct tx_port_stats) +
4945 				       sizeof(struct tx_port_stats_ext) + 512);
4946 	if (!mz) {
4947 		mz = rte_memzone_reserve(mz_name,
4948 					 total_alloc_len,
4949 					 SOCKET_ID_ANY,
4950 					 RTE_MEMZONE_2MB |
4951 					 RTE_MEMZONE_SIZE_HINT_ONLY |
4952 					 RTE_MEMZONE_IOVA_CONTIG);
4953 		if (mz == NULL)
4954 			return -ENOMEM;
4955 	}
4956 	memset(mz->addr, 0, mz->len);
4957 	mz_phys_addr = mz->iova;
4958 
4959 	bp->tx_mem_zone = (const void *)mz;
4960 	bp->hw_tx_port_stats = mz->addr;
4961 	bp->hw_tx_port_stats_map = mz_phys_addr;
4962 	bp->flags |= BNXT_FLAG_PORT_STATS;
4963 
4964 	/* Display extended statistics if FW supports it */
4965 	if (bp->hwrm_spec_code < HWRM_SPEC_CODE_1_8_4 ||
4966 	    bp->hwrm_spec_code == HWRM_SPEC_CODE_1_9_0 ||
4967 	    !(bp->flags & BNXT_FLAG_EXT_STATS_SUPPORTED))
4968 		return 0;
4969 
4970 	bp->hw_rx_port_stats_ext = (void *)
4971 		((uint8_t *)bp->hw_rx_port_stats +
4972 		 sizeof(struct rx_port_stats));
4973 	bp->hw_rx_port_stats_ext_map = bp->hw_rx_port_stats_map +
4974 		sizeof(struct rx_port_stats);
4975 	bp->flags |= BNXT_FLAG_EXT_RX_PORT_STATS;
4976 
4977 	if (bp->hwrm_spec_code < HWRM_SPEC_CODE_1_9_2 ||
4978 	    bp->flags & BNXT_FLAG_EXT_STATS_SUPPORTED) {
4979 		bp->hw_tx_port_stats_ext = (void *)
4980 			((uint8_t *)bp->hw_tx_port_stats +
4981 			 sizeof(struct tx_port_stats));
4982 		bp->hw_tx_port_stats_ext_map =
4983 			bp->hw_tx_port_stats_map +
4984 			sizeof(struct tx_port_stats);
4985 		bp->flags |= BNXT_FLAG_EXT_TX_PORT_STATS;
4986 	}
4987 
4988 	return 0;
4989 }
4990 
4991 static int bnxt_setup_mac_addr(struct rte_eth_dev *eth_dev)
4992 {
4993 	struct bnxt *bp = eth_dev->data->dev_private;
4994 	int rc = 0;
4995 
4996 	eth_dev->data->mac_addrs = rte_zmalloc("bnxt_mac_addr_tbl",
4997 					       RTE_ETHER_ADDR_LEN *
4998 					       bp->max_l2_ctx,
4999 					       0);
5000 	if (eth_dev->data->mac_addrs == NULL) {
5001 		PMD_DRV_LOG(ERR, "Failed to alloc MAC addr tbl\n");
5002 		return -ENOMEM;
5003 	}
5004 
5005 	if (!BNXT_HAS_DFLT_MAC_SET(bp)) {
5006 		if (BNXT_PF(bp))
5007 			return -EINVAL;
5008 
5009 		/* Generate a random MAC address, if none was assigned by PF */
5010 		PMD_DRV_LOG(INFO, "VF MAC address not assigned by Host PF\n");
5011 		bnxt_eth_hw_addr_random(bp->mac_addr);
5012 		PMD_DRV_LOG(INFO,
5013 			    "Assign random MAC:" RTE_ETHER_ADDR_PRT_FMT "\n",
5014 			    bp->mac_addr[0], bp->mac_addr[1], bp->mac_addr[2],
5015 			    bp->mac_addr[3], bp->mac_addr[4], bp->mac_addr[5]);
5016 
5017 		rc = bnxt_hwrm_set_mac(bp);
5018 		if (rc)
5019 			return rc;
5020 	}
5021 
5022 	/* Copy the permanent MAC from the FUNC_QCAPS response */
5023 	memcpy(&eth_dev->data->mac_addrs[0], bp->mac_addr, RTE_ETHER_ADDR_LEN);
5024 
5025 	return rc;
5026 }
5027 
5028 static int bnxt_restore_dflt_mac(struct bnxt *bp)
5029 {
5030 	int rc = 0;
5031 
5032 	/* MAC is already configured in FW */
5033 	if (BNXT_HAS_DFLT_MAC_SET(bp))
5034 		return 0;
5035 
5036 	/* Restore the old MAC configured */
5037 	rc = bnxt_hwrm_set_mac(bp);
5038 	if (rc)
5039 		PMD_DRV_LOG(ERR, "Failed to restore MAC address\n");
5040 
5041 	return rc;
5042 }
5043 
5044 static void bnxt_config_vf_req_fwd(struct bnxt *bp)
5045 {
5046 	if (!BNXT_PF(bp))
5047 		return;
5048 
5049 	memset(bp->pf->vf_req_fwd, 0, sizeof(bp->pf->vf_req_fwd));
5050 
5051 	if (!(bp->fw_cap & BNXT_FW_CAP_LINK_ADMIN))
5052 		BNXT_HWRM_CMD_TO_FORWARD(HWRM_PORT_PHY_QCFG);
5053 	BNXT_HWRM_CMD_TO_FORWARD(HWRM_FUNC_CFG);
5054 	BNXT_HWRM_CMD_TO_FORWARD(HWRM_FUNC_VF_CFG);
5055 	BNXT_HWRM_CMD_TO_FORWARD(HWRM_CFA_L2_FILTER_ALLOC);
5056 	BNXT_HWRM_CMD_TO_FORWARD(HWRM_OEM_CMD);
5057 }
5058 
5059 struct bnxt *
5060 bnxt_get_bp(uint16_t port)
5061 {
5062 	struct bnxt *bp;
5063 	struct rte_eth_dev *dev;
5064 
5065 	if (!rte_eth_dev_is_valid_port(port)) {
5066 		PMD_DRV_LOG(ERR, "Invalid port %d\n", port);
5067 		return NULL;
5068 	}
5069 
5070 	dev = &rte_eth_devices[port];
5071 	if (!is_bnxt_supported(dev)) {
5072 		PMD_DRV_LOG(ERR, "Device %d not supported\n", port);
5073 		return NULL;
5074 	}
5075 
5076 	bp = (struct bnxt *)dev->data->dev_private;
5077 	if (!BNXT_TRUFLOW_EN(bp)) {
5078 		PMD_DRV_LOG(ERR, "TRUFLOW not enabled\n");
5079 		return NULL;
5080 	}
5081 
5082 	return bp;
5083 }
5084 
5085 uint16_t
5086 bnxt_get_svif(uint16_t port_id, bool func_svif,
5087 	      enum bnxt_ulp_intf_type type)
5088 {
5089 	struct rte_eth_dev *eth_dev;
5090 	struct bnxt *bp;
5091 
5092 	eth_dev = &rte_eth_devices[port_id];
5093 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
5094 		struct bnxt_representor *vfr = eth_dev->data->dev_private;
5095 		if (!vfr)
5096 			return 0;
5097 
5098 		if (type == BNXT_ULP_INTF_TYPE_VF_REP)
5099 			return vfr->svif;
5100 
5101 		eth_dev = vfr->parent_dev;
5102 	}
5103 
5104 	bp = eth_dev->data->dev_private;
5105 
5106 	return func_svif ? bp->func_svif : bp->port_svif;
5107 }
5108 
5109 void
5110 bnxt_get_iface_mac(uint16_t port, enum bnxt_ulp_intf_type type,
5111 		   uint8_t *mac, uint8_t *parent_mac)
5112 {
5113 	struct rte_eth_dev *eth_dev;
5114 	struct bnxt *bp;
5115 
5116 	if (type != BNXT_ULP_INTF_TYPE_TRUSTED_VF &&
5117 	    type != BNXT_ULP_INTF_TYPE_PF)
5118 		return;
5119 
5120 	eth_dev = &rte_eth_devices[port];
5121 	bp = eth_dev->data->dev_private;
5122 	memcpy(mac, bp->mac_addr, RTE_ETHER_ADDR_LEN);
5123 
5124 	if (type == BNXT_ULP_INTF_TYPE_TRUSTED_VF)
5125 		memcpy(parent_mac, bp->parent->mac_addr, RTE_ETHER_ADDR_LEN);
5126 }
5127 
5128 uint16_t
5129 bnxt_get_parent_vnic_id(uint16_t port, enum bnxt_ulp_intf_type type)
5130 {
5131 	struct rte_eth_dev *eth_dev;
5132 	struct bnxt *bp;
5133 
5134 	if (type != BNXT_ULP_INTF_TYPE_TRUSTED_VF)
5135 		return 0;
5136 
5137 	eth_dev = &rte_eth_devices[port];
5138 	bp = eth_dev->data->dev_private;
5139 
5140 	return bp->parent->vnic;
5141 }
5142 uint16_t
5143 bnxt_get_vnic_id(uint16_t port, enum bnxt_ulp_intf_type type)
5144 {
5145 	struct rte_eth_dev *eth_dev;
5146 	struct bnxt_vnic_info *vnic;
5147 	struct bnxt *bp;
5148 
5149 	eth_dev = &rte_eth_devices[port];
5150 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
5151 		struct bnxt_representor *vfr = eth_dev->data->dev_private;
5152 		if (!vfr)
5153 			return 0;
5154 
5155 		if (type == BNXT_ULP_INTF_TYPE_VF_REP)
5156 			return vfr->dflt_vnic_id;
5157 
5158 		eth_dev = vfr->parent_dev;
5159 	}
5160 
5161 	bp = eth_dev->data->dev_private;
5162 
5163 	vnic = BNXT_GET_DEFAULT_VNIC(bp);
5164 
5165 	return vnic->fw_vnic_id;
5166 }
5167 
5168 uint16_t
5169 bnxt_get_fw_func_id(uint16_t port, enum bnxt_ulp_intf_type type)
5170 {
5171 	struct rte_eth_dev *eth_dev;
5172 	struct bnxt *bp;
5173 
5174 	eth_dev = &rte_eth_devices[port];
5175 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
5176 		struct bnxt_representor *vfr = eth_dev->data->dev_private;
5177 		if (!vfr)
5178 			return 0;
5179 
5180 		if (type == BNXT_ULP_INTF_TYPE_VF_REP)
5181 			return vfr->fw_fid;
5182 
5183 		eth_dev = vfr->parent_dev;
5184 	}
5185 
5186 	bp = eth_dev->data->dev_private;
5187 
5188 	return bp->fw_fid;
5189 }
5190 
5191 enum bnxt_ulp_intf_type
5192 bnxt_get_interface_type(uint16_t port)
5193 {
5194 	struct rte_eth_dev *eth_dev;
5195 	struct bnxt *bp;
5196 
5197 	eth_dev = &rte_eth_devices[port];
5198 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev))
5199 		return BNXT_ULP_INTF_TYPE_VF_REP;
5200 
5201 	bp = eth_dev->data->dev_private;
5202 	if (BNXT_PF(bp))
5203 		return BNXT_ULP_INTF_TYPE_PF;
5204 	else if (BNXT_VF_IS_TRUSTED(bp))
5205 		return BNXT_ULP_INTF_TYPE_TRUSTED_VF;
5206 	else if (BNXT_VF(bp))
5207 		return BNXT_ULP_INTF_TYPE_VF;
5208 
5209 	return BNXT_ULP_INTF_TYPE_INVALID;
5210 }
5211 
5212 uint16_t
5213 bnxt_get_phy_port_id(uint16_t port_id)
5214 {
5215 	struct bnxt_representor *vfr;
5216 	struct rte_eth_dev *eth_dev;
5217 	struct bnxt *bp;
5218 
5219 	eth_dev = &rte_eth_devices[port_id];
5220 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
5221 		vfr = eth_dev->data->dev_private;
5222 		if (!vfr)
5223 			return 0;
5224 
5225 		eth_dev = vfr->parent_dev;
5226 	}
5227 
5228 	bp = eth_dev->data->dev_private;
5229 
5230 	return BNXT_PF(bp) ? bp->pf->port_id : bp->parent->port_id;
5231 }
5232 
5233 uint16_t
5234 bnxt_get_parif(uint16_t port_id, enum bnxt_ulp_intf_type type)
5235 {
5236 	struct rte_eth_dev *eth_dev;
5237 	struct bnxt *bp;
5238 
5239 	eth_dev = &rte_eth_devices[port_id];
5240 	if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
5241 		struct bnxt_representor *vfr = eth_dev->data->dev_private;
5242 		if (!vfr)
5243 			return 0;
5244 
5245 		if (type == BNXT_ULP_INTF_TYPE_VF_REP)
5246 			return vfr->fw_fid - 1;
5247 
5248 		eth_dev = vfr->parent_dev;
5249 	}
5250 
5251 	bp = eth_dev->data->dev_private;
5252 
5253 	return BNXT_PF(bp) ? bp->fw_fid - 1 : bp->parent->fid - 1;
5254 }
5255 
5256 uint16_t
5257 bnxt_get_vport(uint16_t port_id)
5258 {
5259 	return (1 << bnxt_get_phy_port_id(port_id));
5260 }
5261 
5262 static void bnxt_alloc_error_recovery_info(struct bnxt *bp)
5263 {
5264 	struct bnxt_error_recovery_info *info = bp->recovery_info;
5265 
5266 	if (info) {
5267 		if (!(bp->fw_cap & BNXT_FW_CAP_HCOMM_FW_STATUS))
5268 			memset(info, 0, sizeof(*info));
5269 		return;
5270 	}
5271 
5272 	if (!(bp->fw_cap & BNXT_FW_CAP_ERROR_RECOVERY))
5273 		return;
5274 
5275 	info = rte_zmalloc("bnxt_hwrm_error_recovery_qcfg",
5276 			   sizeof(*info), 0);
5277 	if (!info)
5278 		bp->fw_cap &= ~BNXT_FW_CAP_ERROR_RECOVERY;
5279 
5280 	bp->recovery_info = info;
5281 }
5282 
5283 static void bnxt_check_fw_status(struct bnxt *bp)
5284 {
5285 	uint32_t fw_status;
5286 
5287 	if (!(bp->recovery_info &&
5288 	      (bp->fw_cap & BNXT_FW_CAP_HCOMM_FW_STATUS)))
5289 		return;
5290 
5291 	fw_status = bnxt_read_fw_status_reg(bp, BNXT_FW_STATUS_REG);
5292 	if (fw_status != BNXT_FW_STATUS_HEALTHY)
5293 		PMD_DRV_LOG(ERR, "Firmware not responding, status: %#x\n",
5294 			    fw_status);
5295 }
5296 
5297 static int bnxt_map_hcomm_fw_status_reg(struct bnxt *bp)
5298 {
5299 	struct bnxt_error_recovery_info *info = bp->recovery_info;
5300 	uint32_t status_loc;
5301 	uint32_t sig_ver;
5302 
5303 	rte_write32(HCOMM_STATUS_STRUCT_LOC, (uint8_t *)bp->bar0 +
5304 		    BNXT_GRCPF_REG_WINDOW_BASE_OUT + 4);
5305 	sig_ver = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
5306 				   BNXT_GRCP_WINDOW_2_BASE +
5307 				   offsetof(struct hcomm_status,
5308 					    sig_ver)));
5309 	/* If the signature is absent, then FW does not support this feature */
5310 	if ((sig_ver & HCOMM_STATUS_SIGNATURE_MASK) !=
5311 	    HCOMM_STATUS_SIGNATURE_VAL)
5312 		return 0;
5313 
5314 	if (!info) {
5315 		info = rte_zmalloc("bnxt_hwrm_error_recovery_qcfg",
5316 				   sizeof(*info), 0);
5317 		if (!info)
5318 			return -ENOMEM;
5319 		bp->recovery_info = info;
5320 	} else {
5321 		memset(info, 0, sizeof(*info));
5322 	}
5323 
5324 	status_loc = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
5325 				      BNXT_GRCP_WINDOW_2_BASE +
5326 				      offsetof(struct hcomm_status,
5327 					       fw_status_loc)));
5328 
5329 	/* Only pre-map the FW health status GRC register */
5330 	if (BNXT_FW_STATUS_REG_TYPE(status_loc) != BNXT_FW_STATUS_REG_TYPE_GRC)
5331 		return 0;
5332 
5333 	info->status_regs[BNXT_FW_STATUS_REG] = status_loc;
5334 	info->mapped_status_regs[BNXT_FW_STATUS_REG] =
5335 		BNXT_GRCP_WINDOW_2_BASE + (status_loc & BNXT_GRCP_OFFSET_MASK);
5336 
5337 	rte_write32((status_loc & BNXT_GRCP_BASE_MASK), (uint8_t *)bp->bar0 +
5338 		    BNXT_GRCPF_REG_WINDOW_BASE_OUT + 4);
5339 
5340 	bp->fw_cap |= BNXT_FW_CAP_HCOMM_FW_STATUS;
5341 
5342 	return 0;
5343 }
5344 
5345 /* This function gets the FW version along with the
5346  * capabilities(MAX and current) of the function, vnic,
5347  * error recovery, phy and other chip related info
5348  */
5349 static int bnxt_get_config(struct bnxt *bp)
5350 {
5351 	uint16_t mtu;
5352 	int rc = 0;
5353 
5354 	bp->fw_cap = 0;
5355 
5356 	rc = bnxt_map_hcomm_fw_status_reg(bp);
5357 	if (rc)
5358 		return rc;
5359 
5360 	rc = bnxt_hwrm_ver_get(bp, DFLT_HWRM_CMD_TIMEOUT);
5361 	if (rc) {
5362 		bnxt_check_fw_status(bp);
5363 		return rc;
5364 	}
5365 
5366 	rc = bnxt_hwrm_func_reset(bp);
5367 	if (rc)
5368 		return -EIO;
5369 
5370 	rc = bnxt_hwrm_vnic_qcaps(bp);
5371 	if (rc)
5372 		return rc;
5373 
5374 	rc = bnxt_hwrm_queue_qportcfg(bp);
5375 	if (rc)
5376 		return rc;
5377 
5378 	/* Get the MAX capabilities for this function.
5379 	 * This function also allocates context memory for TQM rings and
5380 	 * informs the firmware about this allocated backing store memory.
5381 	 */
5382 	rc = bnxt_hwrm_func_qcaps(bp);
5383 	if (rc)
5384 		return rc;
5385 
5386 	rc = bnxt_hwrm_func_qcfg(bp, &mtu);
5387 	if (rc)
5388 		return rc;
5389 
5390 	rc = bnxt_hwrm_cfa_adv_flow_mgmt_qcaps(bp);
5391 	if (rc)
5392 		return rc;
5393 
5394 	bnxt_hwrm_port_mac_qcfg(bp);
5395 
5396 	bnxt_hwrm_parent_pf_qcfg(bp);
5397 
5398 	bnxt_hwrm_port_phy_qcaps(bp);
5399 
5400 	bnxt_alloc_error_recovery_info(bp);
5401 	/* Get the adapter error recovery support info */
5402 	rc = bnxt_hwrm_error_recovery_qcfg(bp);
5403 	if (rc)
5404 		bp->fw_cap &= ~BNXT_FW_CAP_ERROR_RECOVERY;
5405 
5406 	bnxt_hwrm_port_led_qcaps(bp);
5407 
5408 	return 0;
5409 }
5410 
5411 static int
5412 bnxt_init_locks(struct bnxt *bp)
5413 {
5414 	int err;
5415 
5416 	err = pthread_mutex_init(&bp->flow_lock, NULL);
5417 	if (err) {
5418 		PMD_DRV_LOG(ERR, "Unable to initialize flow_lock\n");
5419 		return err;
5420 	}
5421 
5422 	err = pthread_mutex_init(&bp->def_cp_lock, NULL);
5423 	if (err) {
5424 		PMD_DRV_LOG(ERR, "Unable to initialize def_cp_lock\n");
5425 		return err;
5426 	}
5427 
5428 	err = pthread_mutex_init(&bp->health_check_lock, NULL);
5429 	if (err) {
5430 		PMD_DRV_LOG(ERR, "Unable to initialize health_check_lock\n");
5431 		return err;
5432 	}
5433 
5434 	err = pthread_mutex_init(&bp->err_recovery_lock, NULL);
5435 	if (err)
5436 		PMD_DRV_LOG(ERR, "Unable to initialize err_recovery_lock\n");
5437 
5438 	return err;
5439 }
5440 
5441 static int bnxt_init_resources(struct bnxt *bp, bool reconfig_dev)
5442 {
5443 	int rc = 0;
5444 
5445 	rc = bnxt_get_config(bp);
5446 	if (rc)
5447 		return rc;
5448 
5449 	if (!reconfig_dev) {
5450 		rc = bnxt_setup_mac_addr(bp->eth_dev);
5451 		if (rc)
5452 			return rc;
5453 	} else {
5454 		rc = bnxt_restore_dflt_mac(bp);
5455 		if (rc)
5456 			return rc;
5457 	}
5458 
5459 	bnxt_config_vf_req_fwd(bp);
5460 
5461 	rc = bnxt_hwrm_func_driver_register(bp);
5462 	if (rc) {
5463 		PMD_DRV_LOG(ERR, "Failed to register driver");
5464 		return -EBUSY;
5465 	}
5466 
5467 	if (BNXT_PF(bp)) {
5468 		if (bp->pdev->max_vfs) {
5469 			rc = bnxt_hwrm_allocate_vfs(bp, bp->pdev->max_vfs);
5470 			if (rc) {
5471 				PMD_DRV_LOG(ERR, "Failed to allocate VFs\n");
5472 				return rc;
5473 			}
5474 		} else {
5475 			rc = bnxt_hwrm_allocate_pf_only(bp);
5476 			if (rc) {
5477 				PMD_DRV_LOG(ERR,
5478 					    "Failed to allocate PF resources");
5479 				return rc;
5480 			}
5481 		}
5482 	}
5483 
5484 	rc = bnxt_alloc_mem(bp, reconfig_dev);
5485 	if (rc)
5486 		return rc;
5487 
5488 	rc = bnxt_setup_int(bp);
5489 	if (rc)
5490 		return rc;
5491 
5492 	rc = bnxt_request_int(bp);
5493 	if (rc)
5494 		return rc;
5495 
5496 	rc = bnxt_init_ctx_mem(bp);
5497 	if (rc) {
5498 		PMD_DRV_LOG(ERR, "Failed to init adv_flow_counters\n");
5499 		return rc;
5500 	}
5501 
5502 	return 0;
5503 }
5504 
5505 static int
5506 bnxt_parse_devarg_accum_stats(__rte_unused const char *key,
5507 			      const char *value, void *opaque_arg)
5508 {
5509 	struct bnxt *bp = opaque_arg;
5510 	unsigned long accum_stats;
5511 	char *end = NULL;
5512 
5513 	if (!value || !opaque_arg) {
5514 		PMD_DRV_LOG(ERR,
5515 			    "Invalid parameter passed to accum-stats devargs.\n");
5516 		return -EINVAL;
5517 	}
5518 
5519 	accum_stats = strtoul(value, &end, 10);
5520 	if (end == NULL || *end != '\0' ||
5521 	    (accum_stats == ULONG_MAX && errno == ERANGE)) {
5522 		PMD_DRV_LOG(ERR,
5523 			    "Invalid parameter passed to accum-stats devargs.\n");
5524 		return -EINVAL;
5525 	}
5526 
5527 	if (BNXT_DEVARG_ACCUM_STATS_INVALID(accum_stats)) {
5528 		PMD_DRV_LOG(ERR,
5529 			    "Invalid value passed to accum-stats devargs.\n");
5530 		return -EINVAL;
5531 	}
5532 
5533 	if (accum_stats) {
5534 		bp->flags2 |= BNXT_FLAGS2_ACCUM_STATS_EN;
5535 		PMD_DRV_LOG(INFO, "Host-based accum-stats feature enabled.\n");
5536 	} else {
5537 		bp->flags2 &= ~BNXT_FLAGS2_ACCUM_STATS_EN;
5538 		PMD_DRV_LOG(INFO, "Host-based accum-stats feature disabled.\n");
5539 	}
5540 
5541 	return 0;
5542 }
5543 
5544 static int
5545 bnxt_parse_devarg_flow_xstat(__rte_unused const char *key,
5546 			     const char *value, void *opaque_arg)
5547 {
5548 	struct bnxt *bp = opaque_arg;
5549 	unsigned long flow_xstat;
5550 	char *end = NULL;
5551 
5552 	if (!value || !opaque_arg) {
5553 		PMD_DRV_LOG(ERR,
5554 			    "Invalid parameter passed to flow_xstat devarg.\n");
5555 		return -EINVAL;
5556 	}
5557 
5558 	flow_xstat = strtoul(value, &end, 10);
5559 	if (end == NULL || *end != '\0' ||
5560 	    (flow_xstat == ULONG_MAX && errno == ERANGE)) {
5561 		PMD_DRV_LOG(ERR,
5562 			    "Invalid parameter passed to flow_xstat devarg.\n");
5563 		return -EINVAL;
5564 	}
5565 
5566 	if (BNXT_DEVARG_FLOW_XSTAT_INVALID(flow_xstat)) {
5567 		PMD_DRV_LOG(ERR,
5568 			    "Invalid value passed to flow_xstat devarg.\n");
5569 		return -EINVAL;
5570 	}
5571 
5572 	bp->flags |= BNXT_FLAG_FLOW_XSTATS_EN;
5573 	if (BNXT_FLOW_XSTATS_EN(bp))
5574 		PMD_DRV_LOG(INFO, "flow_xstat feature enabled.\n");
5575 
5576 	return 0;
5577 }
5578 
5579 static int
5580 bnxt_parse_devarg_max_num_kflows(__rte_unused const char *key,
5581 					const char *value, void *opaque_arg)
5582 {
5583 	struct bnxt *bp = opaque_arg;
5584 	unsigned long max_num_kflows;
5585 	char *end = NULL;
5586 
5587 	if (!value || !opaque_arg) {
5588 		PMD_DRV_LOG(ERR,
5589 			"Invalid parameter passed to max_num_kflows devarg.\n");
5590 		return -EINVAL;
5591 	}
5592 
5593 	max_num_kflows = strtoul(value, &end, 10);
5594 	if (end == NULL || *end != '\0' ||
5595 		(max_num_kflows == ULONG_MAX && errno == ERANGE)) {
5596 		PMD_DRV_LOG(ERR,
5597 			"Invalid parameter passed to max_num_kflows devarg.\n");
5598 		return -EINVAL;
5599 	}
5600 
5601 	if (bnxt_devarg_max_num_kflow_invalid(max_num_kflows)) {
5602 		PMD_DRV_LOG(ERR,
5603 			"Invalid value passed to max_num_kflows devarg.\n");
5604 		return -EINVAL;
5605 	}
5606 
5607 	bp->max_num_kflows = max_num_kflows;
5608 	if (bp->max_num_kflows)
5609 		PMD_DRV_LOG(INFO, "max_num_kflows set as %ldK.\n",
5610 				max_num_kflows);
5611 
5612 	return 0;
5613 }
5614 
5615 static int
5616 bnxt_parse_devarg_app_id(__rte_unused const char *key,
5617 				 const char *value, void *opaque_arg)
5618 {
5619 	struct bnxt *bp = opaque_arg;
5620 	unsigned long app_id;
5621 	char *end = NULL;
5622 
5623 	if (!value || !opaque_arg) {
5624 		PMD_DRV_LOG(ERR,
5625 			    "Invalid parameter passed to app-id "
5626 			    "devargs.\n");
5627 		return -EINVAL;
5628 	}
5629 
5630 	app_id = strtoul(value, &end, 10);
5631 	if (end == NULL || *end != '\0' ||
5632 	    (app_id == ULONG_MAX && errno == ERANGE)) {
5633 		PMD_DRV_LOG(ERR,
5634 			    "Invalid parameter passed to app_id "
5635 			    "devargs.\n");
5636 		return -EINVAL;
5637 	}
5638 
5639 	if (BNXT_DEVARG_APP_ID_INVALID(app_id)) {
5640 		PMD_DRV_LOG(ERR, "Invalid app-id(%d) devargs.\n",
5641 			    (uint16_t)app_id);
5642 		return -EINVAL;
5643 	}
5644 
5645 	bp->app_id = app_id;
5646 	PMD_DRV_LOG(INFO, "app-id=%d feature enabled.\n", (uint16_t)app_id);
5647 
5648 	return 0;
5649 }
5650 
5651 static int
5652 bnxt_parse_devarg_rep_is_pf(__rte_unused const char *key,
5653 			    const char *value, void *opaque_arg)
5654 {
5655 	struct bnxt_representor *vfr_bp = opaque_arg;
5656 	unsigned long rep_is_pf;
5657 	char *end = NULL;
5658 
5659 	if (!value || !opaque_arg) {
5660 		PMD_DRV_LOG(ERR,
5661 			    "Invalid parameter passed to rep_is_pf devargs.\n");
5662 		return -EINVAL;
5663 	}
5664 
5665 	rep_is_pf = strtoul(value, &end, 10);
5666 	if (end == NULL || *end != '\0' ||
5667 	    (rep_is_pf == ULONG_MAX && errno == ERANGE)) {
5668 		PMD_DRV_LOG(ERR,
5669 			    "Invalid parameter passed to rep_is_pf devargs.\n");
5670 		return -EINVAL;
5671 	}
5672 
5673 	if (BNXT_DEVARG_REP_IS_PF_INVALID(rep_is_pf)) {
5674 		PMD_DRV_LOG(ERR,
5675 			    "Invalid value passed to rep_is_pf devargs.\n");
5676 		return -EINVAL;
5677 	}
5678 
5679 	vfr_bp->flags |= rep_is_pf;
5680 	if (BNXT_REP_PF(vfr_bp))
5681 		PMD_DRV_LOG(INFO, "PF representor\n");
5682 	else
5683 		PMD_DRV_LOG(INFO, "VF representor\n");
5684 
5685 	return 0;
5686 }
5687 
5688 static int
5689 bnxt_parse_devarg_rep_based_pf(__rte_unused const char *key,
5690 			       const char *value, void *opaque_arg)
5691 {
5692 	struct bnxt_representor *vfr_bp = opaque_arg;
5693 	unsigned long rep_based_pf;
5694 	char *end = NULL;
5695 
5696 	if (!value || !opaque_arg) {
5697 		PMD_DRV_LOG(ERR,
5698 			    "Invalid parameter passed to rep_based_pf "
5699 			    "devargs.\n");
5700 		return -EINVAL;
5701 	}
5702 
5703 	rep_based_pf = strtoul(value, &end, 10);
5704 	if (end == NULL || *end != '\0' ||
5705 	    (rep_based_pf == ULONG_MAX && errno == ERANGE)) {
5706 		PMD_DRV_LOG(ERR,
5707 			    "Invalid parameter passed to rep_based_pf "
5708 			    "devargs.\n");
5709 		return -EINVAL;
5710 	}
5711 
5712 	if (BNXT_DEVARG_REP_BASED_PF_INVALID(rep_based_pf)) {
5713 		PMD_DRV_LOG(ERR,
5714 			    "Invalid value passed to rep_based_pf devargs.\n");
5715 		return -EINVAL;
5716 	}
5717 
5718 	vfr_bp->rep_based_pf = rep_based_pf;
5719 	vfr_bp->flags |= BNXT_REP_BASED_PF_VALID;
5720 
5721 	PMD_DRV_LOG(INFO, "rep-based-pf = %d\n", vfr_bp->rep_based_pf);
5722 
5723 	return 0;
5724 }
5725 
5726 static int
5727 bnxt_parse_devarg_rep_q_r2f(__rte_unused const char *key,
5728 			    const char *value, void *opaque_arg)
5729 {
5730 	struct bnxt_representor *vfr_bp = opaque_arg;
5731 	unsigned long rep_q_r2f;
5732 	char *end = NULL;
5733 
5734 	if (!value || !opaque_arg) {
5735 		PMD_DRV_LOG(ERR,
5736 			    "Invalid parameter passed to rep_q_r2f "
5737 			    "devargs.\n");
5738 		return -EINVAL;
5739 	}
5740 
5741 	rep_q_r2f = strtoul(value, &end, 10);
5742 	if (end == NULL || *end != '\0' ||
5743 	    (rep_q_r2f == ULONG_MAX && errno == ERANGE)) {
5744 		PMD_DRV_LOG(ERR,
5745 			    "Invalid parameter passed to rep_q_r2f "
5746 			    "devargs.\n");
5747 		return -EINVAL;
5748 	}
5749 
5750 	if (BNXT_DEVARG_REP_Q_R2F_INVALID(rep_q_r2f)) {
5751 		PMD_DRV_LOG(ERR,
5752 			    "Invalid value passed to rep_q_r2f devargs.\n");
5753 		return -EINVAL;
5754 	}
5755 
5756 	vfr_bp->rep_q_r2f = rep_q_r2f;
5757 	vfr_bp->flags |= BNXT_REP_Q_R2F_VALID;
5758 	PMD_DRV_LOG(INFO, "rep-q-r2f = %d\n", vfr_bp->rep_q_r2f);
5759 
5760 	return 0;
5761 }
5762 
5763 static int
5764 bnxt_parse_devarg_rep_q_f2r(__rte_unused const char *key,
5765 			    const char *value, void *opaque_arg)
5766 {
5767 	struct bnxt_representor *vfr_bp = opaque_arg;
5768 	unsigned long rep_q_f2r;
5769 	char *end = NULL;
5770 
5771 	if (!value || !opaque_arg) {
5772 		PMD_DRV_LOG(ERR,
5773 			    "Invalid parameter passed to rep_q_f2r "
5774 			    "devargs.\n");
5775 		return -EINVAL;
5776 	}
5777 
5778 	rep_q_f2r = strtoul(value, &end, 10);
5779 	if (end == NULL || *end != '\0' ||
5780 	    (rep_q_f2r == ULONG_MAX && errno == ERANGE)) {
5781 		PMD_DRV_LOG(ERR,
5782 			    "Invalid parameter passed to rep_q_f2r "
5783 			    "devargs.\n");
5784 		return -EINVAL;
5785 	}
5786 
5787 	if (BNXT_DEVARG_REP_Q_F2R_INVALID(rep_q_f2r)) {
5788 		PMD_DRV_LOG(ERR,
5789 			    "Invalid value passed to rep_q_f2r devargs.\n");
5790 		return -EINVAL;
5791 	}
5792 
5793 	vfr_bp->rep_q_f2r = rep_q_f2r;
5794 	vfr_bp->flags |= BNXT_REP_Q_F2R_VALID;
5795 	PMD_DRV_LOG(INFO, "rep-q-f2r = %d\n", vfr_bp->rep_q_f2r);
5796 
5797 	return 0;
5798 }
5799 
5800 static int
5801 bnxt_parse_devarg_rep_fc_r2f(__rte_unused const char *key,
5802 			     const char *value, void *opaque_arg)
5803 {
5804 	struct bnxt_representor *vfr_bp = opaque_arg;
5805 	unsigned long rep_fc_r2f;
5806 	char *end = NULL;
5807 
5808 	if (!value || !opaque_arg) {
5809 		PMD_DRV_LOG(ERR,
5810 			    "Invalid parameter passed to rep_fc_r2f "
5811 			    "devargs.\n");
5812 		return -EINVAL;
5813 	}
5814 
5815 	rep_fc_r2f = strtoul(value, &end, 10);
5816 	if (end == NULL || *end != '\0' ||
5817 	    (rep_fc_r2f == ULONG_MAX && errno == ERANGE)) {
5818 		PMD_DRV_LOG(ERR,
5819 			    "Invalid parameter passed to rep_fc_r2f "
5820 			    "devargs.\n");
5821 		return -EINVAL;
5822 	}
5823 
5824 	if (BNXT_DEVARG_REP_FC_R2F_INVALID(rep_fc_r2f)) {
5825 		PMD_DRV_LOG(ERR,
5826 			    "Invalid value passed to rep_fc_r2f devargs.\n");
5827 		return -EINVAL;
5828 	}
5829 
5830 	vfr_bp->flags |= BNXT_REP_FC_R2F_VALID;
5831 	vfr_bp->rep_fc_r2f = rep_fc_r2f;
5832 	PMD_DRV_LOG(INFO, "rep-fc-r2f = %lu\n", rep_fc_r2f);
5833 
5834 	return 0;
5835 }
5836 
5837 static int
5838 bnxt_parse_devarg_rep_fc_f2r(__rte_unused const char *key,
5839 			     const char *value, void *opaque_arg)
5840 {
5841 	struct bnxt_representor *vfr_bp = opaque_arg;
5842 	unsigned long rep_fc_f2r;
5843 	char *end = NULL;
5844 
5845 	if (!value || !opaque_arg) {
5846 		PMD_DRV_LOG(ERR,
5847 			    "Invalid parameter passed to rep_fc_f2r "
5848 			    "devargs.\n");
5849 		return -EINVAL;
5850 	}
5851 
5852 	rep_fc_f2r = strtoul(value, &end, 10);
5853 	if (end == NULL || *end != '\0' ||
5854 	    (rep_fc_f2r == ULONG_MAX && errno == ERANGE)) {
5855 		PMD_DRV_LOG(ERR,
5856 			    "Invalid parameter passed to rep_fc_f2r "
5857 			    "devargs.\n");
5858 		return -EINVAL;
5859 	}
5860 
5861 	if (BNXT_DEVARG_REP_FC_F2R_INVALID(rep_fc_f2r)) {
5862 		PMD_DRV_LOG(ERR,
5863 			    "Invalid value passed to rep_fc_f2r devargs.\n");
5864 		return -EINVAL;
5865 	}
5866 
5867 	vfr_bp->flags |= BNXT_REP_FC_F2R_VALID;
5868 	vfr_bp->rep_fc_f2r = rep_fc_f2r;
5869 	PMD_DRV_LOG(INFO, "rep-fc-f2r = %lu\n", rep_fc_f2r);
5870 
5871 	return 0;
5872 }
5873 
5874 static int
5875 bnxt_parse_dev_args(struct bnxt *bp, struct rte_devargs *devargs)
5876 {
5877 	struct rte_kvargs *kvlist;
5878 	int ret;
5879 
5880 	if (devargs == NULL)
5881 		return 0;
5882 
5883 	kvlist = rte_kvargs_parse(devargs->args, bnxt_dev_args);
5884 	if (kvlist == NULL)
5885 		return -EINVAL;
5886 
5887 	/*
5888 	 * Handler for "flow_xstat" devarg.
5889 	 * Invoked as for ex: "-a 0000:00:0d.0,flow_xstat=1"
5890 	 */
5891 	ret = rte_kvargs_process(kvlist, BNXT_DEVARG_FLOW_XSTAT,
5892 				 bnxt_parse_devarg_flow_xstat, bp);
5893 	if (ret)
5894 		goto err;
5895 
5896 	/*
5897 	 * Handler for "accum-stats" devarg.
5898 	 * Invoked as for ex: "-a 0000:00:0d.0,accum-stats=1"
5899 	 */
5900 	rte_kvargs_process(kvlist, BNXT_DEVARG_ACCUM_STATS,
5901 			   bnxt_parse_devarg_accum_stats, bp);
5902 	/*
5903 	 * Handler for "max_num_kflows" devarg.
5904 	 * Invoked as for ex: "-a 000:00:0d.0,max_num_kflows=32"
5905 	 */
5906 	ret = rte_kvargs_process(kvlist, BNXT_DEVARG_MAX_NUM_KFLOWS,
5907 				 bnxt_parse_devarg_max_num_kflows, bp);
5908 	if (ret)
5909 		goto err;
5910 
5911 err:
5912 	/*
5913 	 * Handler for "app-id" devarg.
5914 	 * Invoked as for ex: "-a 000:00:0d.0,app-id=1"
5915 	 */
5916 	rte_kvargs_process(kvlist, BNXT_DEVARG_APP_ID,
5917 			   bnxt_parse_devarg_app_id, bp);
5918 
5919 	rte_kvargs_free(kvlist);
5920 	return ret;
5921 }
5922 
5923 static int bnxt_alloc_switch_domain(struct bnxt *bp)
5924 {
5925 	int rc = 0;
5926 
5927 	if (BNXT_PF(bp) || BNXT_VF_IS_TRUSTED(bp)) {
5928 		rc = rte_eth_switch_domain_alloc(&bp->switch_domain_id);
5929 		if (rc)
5930 			PMD_DRV_LOG(ERR,
5931 				    "Failed to alloc switch domain: %d\n", rc);
5932 		else
5933 			PMD_DRV_LOG(INFO,
5934 				    "Switch domain allocated %d\n",
5935 				    bp->switch_domain_id);
5936 	}
5937 
5938 	return rc;
5939 }
5940 
5941 /* Allocate and initialize various fields in bnxt struct that
5942  * need to be allocated/destroyed only once in the lifetime of the driver
5943  */
5944 static int bnxt_drv_init(struct rte_eth_dev *eth_dev)
5945 {
5946 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
5947 	struct bnxt *bp = eth_dev->data->dev_private;
5948 	int rc = 0;
5949 
5950 	bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
5951 
5952 	if (bnxt_vf_pciid(pci_dev->id.device_id))
5953 		bp->flags |= BNXT_FLAG_VF;
5954 
5955 	if (bnxt_p5_device(pci_dev->id.device_id))
5956 		bp->flags |= BNXT_FLAG_CHIP_P5;
5957 
5958 	if (pci_dev->id.device_id == BROADCOM_DEV_ID_58802 ||
5959 	    pci_dev->id.device_id == BROADCOM_DEV_ID_58804 ||
5960 	    pci_dev->id.device_id == BROADCOM_DEV_ID_58808 ||
5961 	    pci_dev->id.device_id == BROADCOM_DEV_ID_58802_VF)
5962 		bp->flags |= BNXT_FLAG_STINGRAY;
5963 
5964 	if (BNXT_TRUFLOW_EN(bp)) {
5965 		/* extra mbuf field is required to store CFA code from mark */
5966 		static const struct rte_mbuf_dynfield bnxt_cfa_code_dynfield_desc = {
5967 			.name = RTE_PMD_BNXT_CFA_CODE_DYNFIELD_NAME,
5968 			.size = sizeof(bnxt_cfa_code_dynfield_t),
5969 			.align = __alignof__(bnxt_cfa_code_dynfield_t),
5970 		};
5971 		bnxt_cfa_code_dynfield_offset =
5972 			rte_mbuf_dynfield_register(&bnxt_cfa_code_dynfield_desc);
5973 		if (bnxt_cfa_code_dynfield_offset < 0) {
5974 			PMD_DRV_LOG(ERR,
5975 			    "Failed to register mbuf field for TruFlow mark\n");
5976 			return -rte_errno;
5977 		}
5978 	}
5979 
5980 	rc = bnxt_map_pci_bars(eth_dev);
5981 	if (rc) {
5982 		PMD_DRV_LOG(ERR,
5983 			    "Failed to initialize board rc: %x\n", rc);
5984 		return rc;
5985 	}
5986 
5987 	rc = bnxt_alloc_pf_info(bp);
5988 	if (rc)
5989 		return rc;
5990 
5991 	rc = bnxt_alloc_link_info(bp);
5992 	if (rc)
5993 		return rc;
5994 
5995 	rc = bnxt_alloc_parent_info(bp);
5996 	if (rc)
5997 		return rc;
5998 
5999 	rc = bnxt_alloc_hwrm_resources(bp);
6000 	if (rc) {
6001 		PMD_DRV_LOG(ERR,
6002 			    "Failed to allocate response buffer rc: %x\n", rc);
6003 		return rc;
6004 	}
6005 	rc = bnxt_alloc_leds_info(bp);
6006 	if (rc)
6007 		return rc;
6008 
6009 	rc = bnxt_alloc_cos_queues(bp);
6010 	if (rc)
6011 		return rc;
6012 
6013 	rc = bnxt_init_locks(bp);
6014 	if (rc)
6015 		return rc;
6016 
6017 	rc = bnxt_alloc_switch_domain(bp);
6018 	if (rc)
6019 		return rc;
6020 
6021 	return rc;
6022 }
6023 
6024 static int
6025 bnxt_dev_init(struct rte_eth_dev *eth_dev, void *params __rte_unused)
6026 {
6027 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
6028 	static int version_printed;
6029 	struct bnxt *bp;
6030 	int rc;
6031 
6032 	if (version_printed++ == 0)
6033 		PMD_DRV_LOG(INFO, "%s\n", bnxt_version);
6034 
6035 	eth_dev->dev_ops = &bnxt_dev_ops;
6036 	eth_dev->rx_queue_count = bnxt_rx_queue_count_op;
6037 	eth_dev->rx_descriptor_status = bnxt_rx_descriptor_status_op;
6038 	eth_dev->tx_descriptor_status = bnxt_tx_descriptor_status_op;
6039 	eth_dev->rx_pkt_burst = &bnxt_recv_pkts;
6040 	eth_dev->tx_pkt_burst = &bnxt_xmit_pkts;
6041 
6042 	/*
6043 	 * For secondary processes, we don't initialise any further
6044 	 * as primary has already done this work.
6045 	 */
6046 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
6047 		return 0;
6048 
6049 	rte_eth_copy_pci_info(eth_dev, pci_dev);
6050 	eth_dev->data->dev_flags |= RTE_ETH_DEV_AUTOFILL_QUEUE_XSTATS;
6051 
6052 	bp = eth_dev->data->dev_private;
6053 
6054 	/* Parse dev arguments passed on when starting the DPDK application. */
6055 	rc = bnxt_parse_dev_args(bp, pci_dev->device.devargs);
6056 	if (rc)
6057 		goto error_free;
6058 
6059 	rc = bnxt_drv_init(eth_dev);
6060 	if (rc)
6061 		goto error_free;
6062 
6063 	rc = bnxt_init_resources(bp, false);
6064 	if (rc)
6065 		goto error_free;
6066 
6067 	rc = bnxt_alloc_stats_mem(bp);
6068 	if (rc)
6069 		goto error_free;
6070 
6071 	PMD_DRV_LOG(INFO,
6072 		    "Found %s device at mem %" PRIX64 ", node addr %pM\n",
6073 		    DRV_MODULE_NAME,
6074 		    pci_dev->mem_resource[0].phys_addr,
6075 		    pci_dev->mem_resource[0].addr);
6076 
6077 	return 0;
6078 
6079 error_free:
6080 	bnxt_dev_uninit(eth_dev);
6081 	return rc;
6082 }
6083 
6084 
6085 static void bnxt_free_ctx_mem_buf(struct bnxt_ctx_mem_buf_info *ctx)
6086 {
6087 	if (!ctx)
6088 		return;
6089 
6090 	if (ctx->va)
6091 		rte_free(ctx->va);
6092 
6093 	ctx->va = NULL;
6094 	ctx->dma = RTE_BAD_IOVA;
6095 	ctx->ctx_id = BNXT_CTX_VAL_INVAL;
6096 }
6097 
6098 static void bnxt_unregister_fc_ctx_mem(struct bnxt *bp)
6099 {
6100 	bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_RX,
6101 				  CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
6102 				  bp->flow_stat->rx_fc_out_tbl.ctx_id,
6103 				  bp->flow_stat->max_fc,
6104 				  false);
6105 
6106 	bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_TX,
6107 				  CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
6108 				  bp->flow_stat->tx_fc_out_tbl.ctx_id,
6109 				  bp->flow_stat->max_fc,
6110 				  false);
6111 
6112 	if (bp->flow_stat->rx_fc_in_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
6113 		bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->rx_fc_in_tbl.ctx_id);
6114 	bp->flow_stat->rx_fc_in_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
6115 
6116 	if (bp->flow_stat->rx_fc_out_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
6117 		bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->rx_fc_out_tbl.ctx_id);
6118 	bp->flow_stat->rx_fc_out_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
6119 
6120 	if (bp->flow_stat->tx_fc_in_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
6121 		bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->tx_fc_in_tbl.ctx_id);
6122 	bp->flow_stat->tx_fc_in_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
6123 
6124 	if (bp->flow_stat->tx_fc_out_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
6125 		bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->tx_fc_out_tbl.ctx_id);
6126 	bp->flow_stat->tx_fc_out_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
6127 }
6128 
6129 static void bnxt_uninit_fc_ctx_mem(struct bnxt *bp)
6130 {
6131 	bnxt_unregister_fc_ctx_mem(bp);
6132 
6133 	bnxt_free_ctx_mem_buf(&bp->flow_stat->rx_fc_in_tbl);
6134 	bnxt_free_ctx_mem_buf(&bp->flow_stat->rx_fc_out_tbl);
6135 	bnxt_free_ctx_mem_buf(&bp->flow_stat->tx_fc_in_tbl);
6136 	bnxt_free_ctx_mem_buf(&bp->flow_stat->tx_fc_out_tbl);
6137 }
6138 
6139 static void bnxt_uninit_ctx_mem(struct bnxt *bp)
6140 {
6141 	if (BNXT_FLOW_XSTATS_EN(bp))
6142 		bnxt_uninit_fc_ctx_mem(bp);
6143 }
6144 
6145 static void
6146 bnxt_free_error_recovery_info(struct bnxt *bp)
6147 {
6148 	rte_free(bp->recovery_info);
6149 	bp->recovery_info = NULL;
6150 	bp->fw_cap &= ~BNXT_FW_CAP_ERROR_RECOVERY;
6151 }
6152 
6153 static int
6154 bnxt_uninit_resources(struct bnxt *bp, bool reconfig_dev)
6155 {
6156 	int rc;
6157 
6158 	bnxt_free_int(bp);
6159 	bnxt_free_mem(bp, reconfig_dev);
6160 
6161 	bnxt_hwrm_func_buf_unrgtr(bp);
6162 	if (bp->pf != NULL) {
6163 		rte_free(bp->pf->vf_req_buf);
6164 		bp->pf->vf_req_buf = NULL;
6165 	}
6166 
6167 	rc = bnxt_hwrm_func_driver_unregister(bp);
6168 	bp->flags &= ~BNXT_FLAG_REGISTERED;
6169 	bnxt_free_ctx_mem(bp);
6170 	if (!reconfig_dev) {
6171 		bnxt_free_hwrm_resources(bp);
6172 		bnxt_free_error_recovery_info(bp);
6173 	}
6174 
6175 	bnxt_uninit_ctx_mem(bp);
6176 
6177 	bnxt_free_flow_stats_info(bp);
6178 	if (bp->rep_info != NULL)
6179 		bnxt_free_switch_domain(bp);
6180 	bnxt_free_rep_info(bp);
6181 	rte_free(bp->ptp_cfg);
6182 	bp->ptp_cfg = NULL;
6183 	return rc;
6184 }
6185 
6186 static int
6187 bnxt_dev_uninit(struct rte_eth_dev *eth_dev)
6188 {
6189 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
6190 		return -EPERM;
6191 
6192 	PMD_DRV_LOG(DEBUG, "Calling Device uninit\n");
6193 
6194 	if (eth_dev->state != RTE_ETH_DEV_UNUSED)
6195 		bnxt_dev_close_op(eth_dev);
6196 
6197 	return 0;
6198 }
6199 
6200 static int bnxt_pci_remove_dev_with_reps(struct rte_eth_dev *eth_dev)
6201 {
6202 	struct bnxt *bp = eth_dev->data->dev_private;
6203 	struct rte_eth_dev *vf_rep_eth_dev;
6204 	int ret = 0, i;
6205 
6206 	if (!bp)
6207 		return -EINVAL;
6208 
6209 	for (i = 0; i < bp->num_reps; i++) {
6210 		vf_rep_eth_dev = bp->rep_info[i].vfr_eth_dev;
6211 		if (!vf_rep_eth_dev)
6212 			continue;
6213 		PMD_DRV_LOG(DEBUG, "BNXT Port:%d VFR pci remove\n",
6214 			    vf_rep_eth_dev->data->port_id);
6215 		rte_eth_dev_destroy(vf_rep_eth_dev, bnxt_representor_uninit);
6216 	}
6217 	PMD_DRV_LOG(DEBUG, "BNXT Port:%d pci remove\n",
6218 		    eth_dev->data->port_id);
6219 	ret = rte_eth_dev_destroy(eth_dev, bnxt_dev_uninit);
6220 
6221 	return ret;
6222 }
6223 
6224 static void bnxt_free_rep_info(struct bnxt *bp)
6225 {
6226 	rte_free(bp->rep_info);
6227 	bp->rep_info = NULL;
6228 	rte_free(bp->cfa_code_map);
6229 	bp->cfa_code_map = NULL;
6230 }
6231 
6232 static int bnxt_init_rep_info(struct bnxt *bp)
6233 {
6234 	int i = 0, rc;
6235 
6236 	if (bp->rep_info)
6237 		return 0;
6238 
6239 	bp->rep_info = rte_zmalloc("bnxt_rep_info",
6240 				   sizeof(bp->rep_info[0]) * BNXT_MAX_VF_REPS,
6241 				   0);
6242 	if (!bp->rep_info) {
6243 		PMD_DRV_LOG(ERR, "Failed to alloc memory for rep info\n");
6244 		return -ENOMEM;
6245 	}
6246 	bp->cfa_code_map = rte_zmalloc("bnxt_cfa_code_map",
6247 				       sizeof(*bp->cfa_code_map) *
6248 				       BNXT_MAX_CFA_CODE, 0);
6249 	if (!bp->cfa_code_map) {
6250 		PMD_DRV_LOG(ERR, "Failed to alloc memory for cfa_code_map\n");
6251 		bnxt_free_rep_info(bp);
6252 		return -ENOMEM;
6253 	}
6254 
6255 	for (i = 0; i < BNXT_MAX_CFA_CODE; i++)
6256 		bp->cfa_code_map[i] = BNXT_VF_IDX_INVALID;
6257 
6258 	rc = pthread_mutex_init(&bp->rep_info->vfr_lock, NULL);
6259 	if (rc) {
6260 		PMD_DRV_LOG(ERR, "Unable to initialize vfr_lock\n");
6261 		bnxt_free_rep_info(bp);
6262 		return rc;
6263 	}
6264 
6265 	rc = pthread_mutex_init(&bp->rep_info->vfr_start_lock, NULL);
6266 	if (rc) {
6267 		PMD_DRV_LOG(ERR, "Unable to initialize vfr_start_lock\n");
6268 		bnxt_free_rep_info(bp);
6269 		return rc;
6270 	}
6271 
6272 	return rc;
6273 }
6274 
6275 static int bnxt_rep_port_probe(struct rte_pci_device *pci_dev,
6276 			       struct rte_eth_devargs *eth_da,
6277 			       struct rte_eth_dev *backing_eth_dev,
6278 			       const char *dev_args)
6279 {
6280 	struct rte_eth_dev *vf_rep_eth_dev;
6281 	char name[RTE_ETH_NAME_MAX_LEN];
6282 	struct bnxt *backing_bp;
6283 	uint16_t num_rep;
6284 	int i, ret = 0;
6285 	struct rte_kvargs *kvlist = NULL;
6286 
6287 	if (eth_da->type == RTE_ETH_REPRESENTOR_NONE)
6288 		return 0;
6289 	if (eth_da->type != RTE_ETH_REPRESENTOR_VF) {
6290 		PMD_DRV_LOG(ERR, "unsupported representor type %d\n",
6291 			    eth_da->type);
6292 		return -ENOTSUP;
6293 	}
6294 	num_rep = eth_da->nb_representor_ports;
6295 	if (num_rep > BNXT_MAX_VF_REPS) {
6296 		PMD_DRV_LOG(ERR, "nb_representor_ports = %d > %d MAX VF REPS\n",
6297 			    num_rep, BNXT_MAX_VF_REPS);
6298 		return -EINVAL;
6299 	}
6300 
6301 	if (num_rep >= RTE_MAX_ETHPORTS) {
6302 		PMD_DRV_LOG(ERR,
6303 			    "nb_representor_ports = %d > %d MAX ETHPORTS\n",
6304 			    num_rep, RTE_MAX_ETHPORTS);
6305 		return -EINVAL;
6306 	}
6307 
6308 	backing_bp = backing_eth_dev->data->dev_private;
6309 
6310 	if (!(BNXT_PF(backing_bp) || BNXT_VF_IS_TRUSTED(backing_bp))) {
6311 		PMD_DRV_LOG(ERR,
6312 			    "Not a PF or trusted VF. No Representor support\n");
6313 		/* Returning an error is not an option.
6314 		 * Applications are not handling this correctly
6315 		 */
6316 		return 0;
6317 	}
6318 
6319 	if (bnxt_init_rep_info(backing_bp))
6320 		return 0;
6321 
6322 	for (i = 0; i < num_rep; i++) {
6323 		struct bnxt_representor representor = {
6324 			.vf_id = eth_da->representor_ports[i],
6325 			.switch_domain_id = backing_bp->switch_domain_id,
6326 			.parent_dev = backing_eth_dev
6327 		};
6328 
6329 		if (representor.vf_id >= BNXT_MAX_VF_REPS) {
6330 			PMD_DRV_LOG(ERR, "VF-Rep id %d >= %d MAX VF ID\n",
6331 				    representor.vf_id, BNXT_MAX_VF_REPS);
6332 			continue;
6333 		}
6334 
6335 		/* representor port net_bdf_port */
6336 		snprintf(name, sizeof(name), "net_%s_representor_%d",
6337 			 pci_dev->device.name, eth_da->representor_ports[i]);
6338 
6339 		kvlist = rte_kvargs_parse(dev_args, bnxt_dev_args);
6340 		if (kvlist) {
6341 			/*
6342 			 * Handler for "rep_is_pf" devarg.
6343 			 * Invoked as for ex: "-a 000:00:0d.0,
6344 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6345 			 */
6346 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_IS_PF,
6347 						 bnxt_parse_devarg_rep_is_pf,
6348 						 (void *)&representor);
6349 			if (ret) {
6350 				ret = -EINVAL;
6351 				goto err;
6352 			}
6353 			/*
6354 			 * Handler for "rep_based_pf" devarg.
6355 			 * Invoked as for ex: "-a 000:00:0d.0,
6356 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6357 			 */
6358 			ret = rte_kvargs_process(kvlist,
6359 						 BNXT_DEVARG_REP_BASED_PF,
6360 						 bnxt_parse_devarg_rep_based_pf,
6361 						 (void *)&representor);
6362 			if (ret) {
6363 				ret = -EINVAL;
6364 				goto err;
6365 			}
6366 			/*
6367 			 * Handler for "rep_based_pf" devarg.
6368 			 * Invoked as for ex: "-a 000:00:0d.0,
6369 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6370 			 */
6371 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_Q_R2F,
6372 						 bnxt_parse_devarg_rep_q_r2f,
6373 						 (void *)&representor);
6374 			if (ret) {
6375 				ret = -EINVAL;
6376 				goto err;
6377 			}
6378 			/*
6379 			 * Handler for "rep_based_pf" devarg.
6380 			 * Invoked as for ex: "-a 000:00:0d.0,
6381 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6382 			 */
6383 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_Q_F2R,
6384 						 bnxt_parse_devarg_rep_q_f2r,
6385 						 (void *)&representor);
6386 			if (ret) {
6387 				ret = -EINVAL;
6388 				goto err;
6389 			}
6390 			/*
6391 			 * Handler for "rep_based_pf" devarg.
6392 			 * Invoked as for ex: "-a 000:00:0d.0,
6393 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6394 			 */
6395 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_FC_R2F,
6396 						 bnxt_parse_devarg_rep_fc_r2f,
6397 						 (void *)&representor);
6398 			if (ret) {
6399 				ret = -EINVAL;
6400 				goto err;
6401 			}
6402 			/*
6403 			 * Handler for "rep_based_pf" devarg.
6404 			 * Invoked as for ex: "-a 000:00:0d.0,
6405 			 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6406 			 */
6407 			ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_FC_F2R,
6408 						 bnxt_parse_devarg_rep_fc_f2r,
6409 						 (void *)&representor);
6410 			if (ret) {
6411 				ret = -EINVAL;
6412 				goto err;
6413 			}
6414 		}
6415 
6416 		ret = rte_eth_dev_create(&pci_dev->device, name,
6417 					 sizeof(struct bnxt_representor),
6418 					 NULL, NULL,
6419 					 bnxt_representor_init,
6420 					 &representor);
6421 		if (ret) {
6422 			PMD_DRV_LOG(ERR, "failed to create bnxt vf "
6423 				    "representor %s.", name);
6424 			goto err;
6425 		}
6426 
6427 		vf_rep_eth_dev = rte_eth_dev_allocated(name);
6428 		if (!vf_rep_eth_dev) {
6429 			PMD_DRV_LOG(ERR, "Failed to find the eth_dev"
6430 				    " for VF-Rep: %s.", name);
6431 			ret = -ENODEV;
6432 			goto err;
6433 		}
6434 
6435 		PMD_DRV_LOG(DEBUG, "BNXT Port:%d VFR pci probe\n",
6436 			    backing_eth_dev->data->port_id);
6437 		backing_bp->rep_info[representor.vf_id].vfr_eth_dev =
6438 							 vf_rep_eth_dev;
6439 		backing_bp->num_reps++;
6440 
6441 	}
6442 
6443 	rte_kvargs_free(kvlist);
6444 	return 0;
6445 
6446 err:
6447 	/* If num_rep > 1, then rollback already created
6448 	 * ports, since we'll be failing the probe anyway
6449 	 */
6450 	if (num_rep > 1)
6451 		bnxt_pci_remove_dev_with_reps(backing_eth_dev);
6452 	rte_errno = -ret;
6453 	rte_kvargs_free(kvlist);
6454 
6455 	return ret;
6456 }
6457 
6458 static int bnxt_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
6459 			  struct rte_pci_device *pci_dev)
6460 {
6461 	struct rte_eth_devargs eth_da = { .nb_representor_ports = 0 };
6462 	struct rte_eth_dev *backing_eth_dev;
6463 	uint16_t num_rep;
6464 	int ret = 0;
6465 
6466 	if (pci_dev->device.devargs) {
6467 		ret = rte_eth_devargs_parse(pci_dev->device.devargs->args,
6468 					    &eth_da);
6469 		if (ret)
6470 			return ret;
6471 	}
6472 
6473 	num_rep = eth_da.nb_representor_ports;
6474 	PMD_DRV_LOG(DEBUG, "nb_representor_ports = %d\n",
6475 		    num_rep);
6476 
6477 	/* We could come here after first level of probe is already invoked
6478 	 * as part of an application bringup(OVS-DPDK vswitchd), so first check
6479 	 * for already allocated eth_dev for the backing device (PF/Trusted VF)
6480 	 */
6481 	backing_eth_dev = rte_eth_dev_allocated(pci_dev->device.name);
6482 	if (backing_eth_dev == NULL) {
6483 		ret = rte_eth_dev_create(&pci_dev->device, pci_dev->device.name,
6484 					 sizeof(struct bnxt),
6485 					 eth_dev_pci_specific_init, pci_dev,
6486 					 bnxt_dev_init, NULL);
6487 
6488 		if (ret || !num_rep)
6489 			return ret;
6490 
6491 		backing_eth_dev = rte_eth_dev_allocated(pci_dev->device.name);
6492 	}
6493 	PMD_DRV_LOG(DEBUG, "BNXT Port:%d pci probe\n",
6494 		    backing_eth_dev->data->port_id);
6495 
6496 	if (!num_rep)
6497 		return ret;
6498 
6499 	/* probe representor ports now */
6500 	ret = bnxt_rep_port_probe(pci_dev, &eth_da, backing_eth_dev,
6501 				  pci_dev->device.devargs->args);
6502 
6503 	return ret;
6504 }
6505 
6506 static int bnxt_pci_remove(struct rte_pci_device *pci_dev)
6507 {
6508 	struct rte_eth_dev *eth_dev;
6509 
6510 	eth_dev = rte_eth_dev_allocated(pci_dev->device.name);
6511 	if (!eth_dev)
6512 		return 0; /* Invoked typically only by OVS-DPDK, by the
6513 			   * time it comes here the eth_dev is already
6514 			   * deleted by rte_eth_dev_close(), so returning
6515 			   * +ve value will at least help in proper cleanup
6516 			   */
6517 
6518 	PMD_DRV_LOG(DEBUG, "BNXT Port:%d pci remove\n", eth_dev->data->port_id);
6519 	if (rte_eal_process_type() == RTE_PROC_PRIMARY) {
6520 		if (eth_dev->data->dev_flags & RTE_ETH_DEV_REPRESENTOR)
6521 			return rte_eth_dev_destroy(eth_dev,
6522 						   bnxt_representor_uninit);
6523 		else
6524 			return rte_eth_dev_destroy(eth_dev,
6525 						   bnxt_dev_uninit);
6526 	} else {
6527 		return rte_eth_dev_pci_generic_remove(pci_dev, NULL);
6528 	}
6529 }
6530 
6531 static struct rte_pci_driver bnxt_rte_pmd = {
6532 	.id_table = bnxt_pci_id_map,
6533 	.drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC |
6534 			RTE_PCI_DRV_INTR_RMV |
6535 			RTE_PCI_DRV_PROBE_AGAIN, /* Needed in case of VF-REPs
6536 						  * and OVS-DPDK
6537 						  */
6538 	.probe = bnxt_pci_probe,
6539 	.remove = bnxt_pci_remove,
6540 };
6541 
6542 static bool
6543 is_device_supported(struct rte_eth_dev *dev, struct rte_pci_driver *drv)
6544 {
6545 	if (strcmp(dev->device->driver->name, drv->driver.name))
6546 		return false;
6547 
6548 	return true;
6549 }
6550 
6551 bool is_bnxt_supported(struct rte_eth_dev *dev)
6552 {
6553 	return is_device_supported(dev, &bnxt_rte_pmd);
6554 }
6555 
6556 RTE_LOG_REGISTER_SUFFIX(bnxt_logtype_driver, driver, NOTICE);
6557 RTE_PMD_REGISTER_PCI(net_bnxt, bnxt_rte_pmd);
6558 RTE_PMD_REGISTER_PCI_TABLE(net_bnxt, bnxt_pci_id_map);
6559 
6560