drivers/net/bnxt/bnxt_ethdev.c

   1 /* SPDX-License-Identifier: BSD-3-Clause
   2  * Copyright(c) 2014-2018 Broadcom
   3  * All rights reserved.
   4  */
   5
   6 #include <inttypes.h>
   7 #include <stdbool.h>
   8
   9 #include <rte_dev.h>
  10 #include <rte_ethdev_driver.h>
  11 #include <rte_ethdev_pci.h>
  12 #include <rte_malloc.h>
  13 #include <rte_cycles.h>
  14 #include <rte_alarm.h>
  15 #include <rte_kvargs.h>
  16 #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_57301) },
  51         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57302) },
  52         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57304_PF) },
  53         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57304_VF) },
  54         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_NS2) },
  55         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57402) },
  56         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57404) },
  57         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57406_PF) },
  58         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57406_VF) },
  59         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57402_MF) },
  60         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57407_RJ45) },
  61         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57404_MF) },
  62         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57406_MF) },
  63         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57407_SFP) },
  64         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57407_MF) },
  65         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_5741X_VF) },
  66         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_5731X_VF) },
  67         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57314) },
  68         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57417_MF) },
  69         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57311) },
  70         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57312) },
  71         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57412) },
  72         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57414) },
  73         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57416_RJ45) },
  74         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57417_RJ45) },
  75         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57412_MF) },
  76         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57317_RJ45) },
  77         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57417_SFP) },
  78         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57416_SFP) },
  79         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57317_SFP) },
  80         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57414_MF) },
  81         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57416_MF) },
  82         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58802) },
  83         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58804) },
  84         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58808) },
  85         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58802_VF) },
  86         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57508) },
  87         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57504) },
  88         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57502) },
  89         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57500_VF1) },
  90         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57500_VF2) },
  91         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57508_MF1) },
  92         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57504_MF1) },
  93         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57502_MF1) },
  94         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57508_MF2) },
  95         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57504_MF2) },
  96         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57502_MF2) },
  97         { .vendor_id = 0, /* sentinel */ },
  98 };
  99
 100 #define BNXT_DEVARG_TRUFLOW     "host-based-truflow"
 101 #define BNXT_DEVARG_FLOW_XSTAT  "flow-xstat"
 102 #define BNXT_DEVARG_MAX_NUM_KFLOWS  "max-num-kflows"
 103 #define BNXT_DEVARG_REPRESENTOR "representor"
 104 #define BNXT_DEVARG_REP_BASED_PF  "rep-based-pf"
 105 #define BNXT_DEVARG_REP_IS_PF  "rep-is-pf"
 106 #define BNXT_DEVARG_REP_Q_R2F  "rep-q-r2f"
 107 #define BNXT_DEVARG_REP_Q_F2R  "rep-q-f2r"
 108 #define BNXT_DEVARG_REP_FC_R2F  "rep-fc-r2f"
 109 #define BNXT_DEVARG_REP_FC_F2R  "rep-fc-f2r"
 110
 111 static const char *const bnxt_dev_args[] = {
 112         BNXT_DEVARG_REPRESENTOR,
 113         BNXT_DEVARG_TRUFLOW,
 114         BNXT_DEVARG_FLOW_XSTAT,
 115         BNXT_DEVARG_MAX_NUM_KFLOWS,
 116         BNXT_DEVARG_REP_BASED_PF,
 117         BNXT_DEVARG_REP_IS_PF,
 118         BNXT_DEVARG_REP_Q_R2F,
 119         BNXT_DEVARG_REP_Q_F2R,
 120         BNXT_DEVARG_REP_FC_R2F,
 121         BNXT_DEVARG_REP_FC_F2R,
 122         NULL
 123 };
 124
 125 /*
 126  * truflow == false to disable the feature
 127  * truflow == true to enable the feature
 128  */
 129 #define BNXT_DEVARG_TRUFLOW_INVALID(truflow)    ((truflow) > 1)
 130
 131 /*
 132  * flow_xstat == false to disable the feature
 133  * flow_xstat == true to enable the feature
 134  */
 135 #define BNXT_DEVARG_FLOW_XSTAT_INVALID(flow_xstat)      ((flow_xstat) > 1)
 136
 137 /*
 138  * rep_is_pf == false to indicate VF representor
 139  * rep_is_pf == true to indicate PF representor
 140  */
 141 #define BNXT_DEVARG_REP_IS_PF_INVALID(rep_is_pf)        ((rep_is_pf) > 1)
 142
 143 /*
 144  * rep_based_pf == Physical index of the PF
 145  */
 146 #define BNXT_DEVARG_REP_BASED_PF_INVALID(rep_based_pf)  ((rep_based_pf) > 15)
 147 /*
 148  * rep_q_r2f == Logical COS Queue index for the rep to endpoint direction
 149  */
 150 #define BNXT_DEVARG_REP_Q_R2F_INVALID(rep_q_r2f)        ((rep_q_r2f) > 3)
 151
 152 /*
 153  * rep_q_f2r == Logical COS Queue index for the endpoint to rep direction
 154  */
 155 #define BNXT_DEVARG_REP_Q_F2R_INVALID(rep_q_f2r)        ((rep_q_f2r) > 3)
 156
 157 /*
 158  * rep_fc_r2f == Flow control for the representor to endpoint direction
 159  */
 160 #define BNXT_DEVARG_REP_FC_R2F_INVALID(rep_fc_r2f)      ((rep_fc_r2f) > 1)
 161
 162 /*
 163  * rep_fc_f2r == Flow control for the endpoint to representor direction
 164  */
 165 #define BNXT_DEVARG_REP_FC_F2R_INVALID(rep_fc_f2r)      ((rep_fc_f2r) > 1)
 166
 167 int bnxt_cfa_code_dynfield_offset = -1;
 168
 169 /*
 170  * max_num_kflows must be >= 32
 171  * and must be a power-of-2 supported value
 172  * return: 1 -> invalid
 173  *         0 -> valid
 174  */
 175 static int bnxt_devarg_max_num_kflow_invalid(uint16_t max_num_kflows)
 176 {
 177         if (max_num_kflows < 32 || !rte_is_power_of_2(max_num_kflows))
 178                 return 1;
 179         return 0;
 180 }
 181
 182 static int bnxt_vlan_offload_set_op(struct rte_eth_dev *dev, int mask);
 183 static int bnxt_dev_uninit(struct rte_eth_dev *eth_dev);
 184 static int bnxt_init_resources(struct bnxt *bp, bool reconfig_dev);
 185 static int bnxt_uninit_resources(struct bnxt *bp, bool reconfig_dev);
 186 static void bnxt_cancel_fw_health_check(struct bnxt *bp);
 187 static int bnxt_restore_vlan_filters(struct bnxt *bp);
 188 static void bnxt_dev_recover(void *arg);
 189 static void bnxt_free_error_recovery_info(struct bnxt *bp);
 190 static void bnxt_free_rep_info(struct bnxt *bp);
 191
 192 int is_bnxt_in_error(struct bnxt *bp)
 193 {
 194         if (bp->flags & BNXT_FLAG_FATAL_ERROR)
 195                 return -EIO;
 196         if (bp->flags & BNXT_FLAG_FW_RESET)
 197                 return -EBUSY;
 198
 199         return 0;
 200 }
 201
 202 /***********************/
 203
 204 /*
 205  * High level utility functions
 206  */
 207
 208 static uint16_t bnxt_rss_ctxts(const struct bnxt *bp)
 209 {
 210         unsigned int num_rss_rings = RTE_MIN(bp->rx_nr_rings,
 211                                              BNXT_RSS_TBL_SIZE_P5);
 212
 213         if (!BNXT_CHIP_P5(bp))
 214                 return 1;
 215
 216         return RTE_ALIGN_MUL_CEIL(num_rss_rings,
 217                                   BNXT_RSS_ENTRIES_PER_CTX_P5) /
 218                                   BNXT_RSS_ENTRIES_PER_CTX_P5;
 219 }
 220
 221 uint16_t bnxt_rss_hash_tbl_size(const struct bnxt *bp)
 222 {
 223         if (!BNXT_CHIP_P5(bp))
 224                 return HW_HASH_INDEX_SIZE;
 225
 226         return bnxt_rss_ctxts(bp) * BNXT_RSS_ENTRIES_PER_CTX_P5;
 227 }
 228
 229 static void bnxt_free_parent_info(struct bnxt *bp)
 230 {
 231         rte_free(bp->parent);
 232 }
 233
 234 static void bnxt_free_pf_info(struct bnxt *bp)
 235 {
 236         rte_free(bp->pf);
 237 }
 238
 239 static void bnxt_free_link_info(struct bnxt *bp)
 240 {
 241         rte_free(bp->link_info);
 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         rte_free(bp->tx_cos_queue);
 263 }
 264
 265 static void bnxt_free_mem(struct bnxt *bp, bool reconfig)
 266 {
 267         bnxt_free_filter_mem(bp);
 268         bnxt_free_vnic_attributes(bp);
 269         bnxt_free_vnic_mem(bp);
 270
 271         /* tx/rx rings are configured as part of *_queue_setup callbacks.
 272          * If the number of rings change across fw update,
 273          * we don't have much choice except to warn the user.
 274          */
 275         if (!reconfig) {
 276                 bnxt_free_stats(bp);
 277                 bnxt_free_tx_rings(bp);
 278                 bnxt_free_rx_rings(bp);
 279         }
 280         bnxt_free_async_cp_ring(bp);
 281         bnxt_free_rxtx_nq_ring(bp);
 282
 283         rte_free(bp->grp_info);
 284         bp->grp_info = NULL;
 285 }
 286
 287 static int bnxt_alloc_parent_info(struct bnxt *bp)
 288 {
 289         bp->parent = rte_zmalloc("bnxt_parent_info",
 290                                  sizeof(struct bnxt_parent_info), 0);
 291         if (bp->parent == NULL)
 292                 return -ENOMEM;
 293
 294         return 0;
 295 }
 296
 297 static int bnxt_alloc_pf_info(struct bnxt *bp)
 298 {
 299         bp->pf = rte_zmalloc("bnxt_pf_info", sizeof(struct bnxt_pf_info), 0);
 300         if (bp->pf == NULL)
 301                 return -ENOMEM;
 302
 303         return 0;
 304 }
 305
 306 static int bnxt_alloc_link_info(struct bnxt *bp)
 307 {
 308         bp->link_info =
 309                 rte_zmalloc("bnxt_link_info", sizeof(struct bnxt_link_info), 0);
 310         if (bp->link_info == NULL)
 311                 return -ENOMEM;
 312
 313         return 0;
 314 }
 315
 316 static int bnxt_alloc_leds_info(struct bnxt *bp)
 317 {
 318         if (BNXT_VF(bp))
 319                 return 0;
 320
 321         bp->leds = rte_zmalloc("bnxt_leds",
 322                                BNXT_MAX_LED * sizeof(struct bnxt_led_info),
 323                                0);
 324         if (bp->leds == NULL)
 325                 return -ENOMEM;
 326
 327         return 0;
 328 }
 329
 330 static int bnxt_alloc_cos_queues(struct bnxt *bp)
 331 {
 332         bp->rx_cos_queue =
 333                 rte_zmalloc("bnxt_rx_cosq",
 334                             BNXT_COS_QUEUE_COUNT *
 335                             sizeof(struct bnxt_cos_queue_info),
 336                             0);
 337         if (bp->rx_cos_queue == NULL)
 338                 return -ENOMEM;
 339
 340         bp->tx_cos_queue =
 341                 rte_zmalloc("bnxt_tx_cosq",
 342                             BNXT_COS_QUEUE_COUNT *
 343                             sizeof(struct bnxt_cos_queue_info),
 344                             0);
 345         if (bp->tx_cos_queue == NULL)
 346                 return -ENOMEM;
 347
 348         return 0;
 349 }
 350
 351 static int bnxt_alloc_flow_stats_info(struct bnxt *bp)
 352 {
 353         bp->flow_stat = rte_zmalloc("bnxt_flow_xstat",
 354                                     sizeof(struct bnxt_flow_stat_info), 0);
 355         if (bp->flow_stat == NULL)
 356                 return -ENOMEM;
 357
 358         return 0;
 359 }
 360
 361 static int bnxt_alloc_mem(struct bnxt *bp, bool reconfig)
 362 {
 363         int rc;
 364
 365         rc = bnxt_alloc_ring_grps(bp);
 366         if (rc)
 367                 goto alloc_mem_err;
 368
 369         rc = bnxt_alloc_async_ring_struct(bp);
 370         if (rc)
 371                 goto alloc_mem_err;
 372
 373         rc = bnxt_alloc_vnic_mem(bp);
 374         if (rc)
 375                 goto alloc_mem_err;
 376
 377         rc = bnxt_alloc_vnic_attributes(bp);
 378         if (rc)
 379                 goto alloc_mem_err;
 380
 381         rc = bnxt_alloc_filter_mem(bp);
 382         if (rc)
 383                 goto alloc_mem_err;
 384
 385         rc = bnxt_alloc_async_cp_ring(bp);
 386         if (rc)
 387                 goto alloc_mem_err;
 388
 389         rc = bnxt_alloc_rxtx_nq_ring(bp);
 390         if (rc)
 391                 goto alloc_mem_err;
 392
 393         if (BNXT_FLOW_XSTATS_EN(bp)) {
 394                 rc = bnxt_alloc_flow_stats_info(bp);
 395                 if (rc)
 396                         goto alloc_mem_err;
 397         }
 398
 399         return 0;
 400
 401 alloc_mem_err:
 402         bnxt_free_mem(bp, reconfig);
 403         return rc;
 404 }
 405
 406 static int bnxt_setup_one_vnic(struct bnxt *bp, uint16_t vnic_id)
 407 {
 408         struct rte_eth_conf *dev_conf = &bp->eth_dev->data->dev_conf;
 409         struct bnxt_vnic_info *vnic = &bp->vnic_info[vnic_id];
 410         uint64_t rx_offloads = dev_conf->rxmode.offloads;
 411         struct bnxt_rx_queue *rxq;
 412         unsigned int j;
 413         int rc;
 414
 415         rc = bnxt_vnic_grp_alloc(bp, vnic);
 416         if (rc)
 417                 goto err_out;
 418
 419         PMD_DRV_LOG(DEBUG, "vnic[%d] = %p vnic->fw_grp_ids = %p\n",
 420                     vnic_id, vnic, vnic->fw_grp_ids);
 421
 422         rc = bnxt_hwrm_vnic_alloc(bp, vnic);
 423         if (rc)
 424                 goto err_out;
 425
 426         /* Alloc RSS context only if RSS mode is enabled */
 427         if (dev_conf->rxmode.mq_mode & ETH_MQ_RX_RSS) {
 428                 int j, nr_ctxs = bnxt_rss_ctxts(bp);
 429
 430                 if (bp->rx_nr_rings > BNXT_RSS_TBL_SIZE_P5) {
 431                         PMD_DRV_LOG(ERR, "RxQ cnt %d > reta_size %d\n",
 432                                     bp->rx_nr_rings, BNXT_RSS_TBL_SIZE_P5);
 433                         PMD_DRV_LOG(ERR,
 434                                     "Only queues 0-%d will be in RSS table\n",
 435                                     BNXT_RSS_TBL_SIZE_P5 - 1);
 436                 }
 437
 438                 rc = 0;
 439                 for (j = 0; j < nr_ctxs; j++) {
 440                         rc = bnxt_hwrm_vnic_ctx_alloc(bp, vnic, j);
 441                         if (rc)
 442                                 break;
 443                 }
 444                 if (rc) {
 445                         PMD_DRV_LOG(ERR,
 446                                     "HWRM vnic %d ctx %d alloc failure rc: %x\n",
 447                                     vnic_id, j, rc);
 448                         goto err_out;
 449                 }
 450                 vnic->num_lb_ctxts = nr_ctxs;
 451         }
 452
 453         /*
 454          * Firmware sets pf pair in default vnic cfg. If the VLAN strip
 455          * setting is not available at this time, it will not be
 456          * configured correctly in the CFA.
 457          */
 458         if (rx_offloads & DEV_RX_OFFLOAD_VLAN_STRIP)
 459                 vnic->vlan_strip = true;
 460         else
 461                 vnic->vlan_strip = false;
 462
 463         rc = bnxt_hwrm_vnic_cfg(bp, vnic);
 464         if (rc)
 465                 goto err_out;
 466
 467         rc = bnxt_set_hwrm_vnic_filters(bp, vnic);
 468         if (rc)
 469                 goto err_out;
 470
 471         for (j = 0; j < bp->rx_num_qs_per_vnic; j++) {
 472                 rxq = bp->eth_dev->data->rx_queues[j];
 473
 474                 PMD_DRV_LOG(DEBUG,
 475                             "rxq[%d]->vnic=%p vnic->fw_grp_ids=%p\n",
 476                             j, rxq->vnic, rxq->vnic->fw_grp_ids);
 477
 478                 if (BNXT_HAS_RING_GRPS(bp) && rxq->rx_deferred_start)
 479                         rxq->vnic->fw_grp_ids[j] = INVALID_HW_RING_ID;
 480                 else
 481                         vnic->rx_queue_cnt++;
 482         }
 483
 484         PMD_DRV_LOG(DEBUG, "vnic->rx_queue_cnt = %d\n", vnic->rx_queue_cnt);
 485
 486         rc = bnxt_vnic_rss_configure(bp, vnic);
 487         if (rc)
 488                 goto err_out;
 489
 490         bnxt_hwrm_vnic_plcmode_cfg(bp, vnic);
 491
 492         if (rx_offloads & DEV_RX_OFFLOAD_TCP_LRO)
 493                 bnxt_hwrm_vnic_tpa_cfg(bp, vnic, 1);
 494         else
 495                 bnxt_hwrm_vnic_tpa_cfg(bp, vnic, 0);
 496
 497         return 0;
 498 err_out:
 499         PMD_DRV_LOG(ERR, "HWRM vnic %d cfg failure rc: %x\n",
 500                     vnic_id, rc);
 501         return rc;
 502 }
 503
 504 static int bnxt_register_fc_ctx_mem(struct bnxt *bp)
 505 {
 506         int rc = 0;
 507
 508         rc = bnxt_hwrm_ctx_rgtr(bp, bp->flow_stat->rx_fc_in_tbl.dma,
 509                                 &bp->flow_stat->rx_fc_in_tbl.ctx_id);
 510         if (rc)
 511                 return rc;
 512
 513         PMD_DRV_LOG(DEBUG,
 514                     "rx_fc_in_tbl.va = %p rx_fc_in_tbl.dma = %p"
 515                     " rx_fc_in_tbl.ctx_id = %d\n",
 516                     bp->flow_stat->rx_fc_in_tbl.va,
 517                     (void *)((uintptr_t)bp->flow_stat->rx_fc_in_tbl.dma),
 518                     bp->flow_stat->rx_fc_in_tbl.ctx_id);
 519
 520         rc = bnxt_hwrm_ctx_rgtr(bp, bp->flow_stat->rx_fc_out_tbl.dma,
 521                                 &bp->flow_stat->rx_fc_out_tbl.ctx_id);
 522         if (rc)
 523                 return rc;
 524
 525         PMD_DRV_LOG(DEBUG,
 526                     "rx_fc_out_tbl.va = %p rx_fc_out_tbl.dma = %p"
 527                     " rx_fc_out_tbl.ctx_id = %d\n",
 528                     bp->flow_stat->rx_fc_out_tbl.va,
 529                     (void *)((uintptr_t)bp->flow_stat->rx_fc_out_tbl.dma),
 530                     bp->flow_stat->rx_fc_out_tbl.ctx_id);
 531
 532         rc = bnxt_hwrm_ctx_rgtr(bp, bp->flow_stat->tx_fc_in_tbl.dma,
 533                                 &bp->flow_stat->tx_fc_in_tbl.ctx_id);
 534         if (rc)
 535                 return rc;
 536
 537         PMD_DRV_LOG(DEBUG,
 538                     "tx_fc_in_tbl.va = %p tx_fc_in_tbl.dma = %p"
 539                     " tx_fc_in_tbl.ctx_id = %d\n",
 540                     bp->flow_stat->tx_fc_in_tbl.va,
 541                     (void *)((uintptr_t)bp->flow_stat->tx_fc_in_tbl.dma),
 542                     bp->flow_stat->tx_fc_in_tbl.ctx_id);
 543
 544         rc = bnxt_hwrm_ctx_rgtr(bp, bp->flow_stat->tx_fc_out_tbl.dma,
 545                                 &bp->flow_stat->tx_fc_out_tbl.ctx_id);
 546         if (rc)
 547                 return rc;
 548
 549         PMD_DRV_LOG(DEBUG,
 550                     "tx_fc_out_tbl.va = %p tx_fc_out_tbl.dma = %p"
 551                     " tx_fc_out_tbl.ctx_id = %d\n",
 552                     bp->flow_stat->tx_fc_out_tbl.va,
 553                     (void *)((uintptr_t)bp->flow_stat->tx_fc_out_tbl.dma),
 554                     bp->flow_stat->tx_fc_out_tbl.ctx_id);
 555
 556         memset(bp->flow_stat->rx_fc_out_tbl.va,
 557                0,
 558                bp->flow_stat->rx_fc_out_tbl.size);
 559         rc = bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_RX,
 560                                        CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
 561                                        bp->flow_stat->rx_fc_out_tbl.ctx_id,
 562                                        bp->flow_stat->max_fc,
 563                                        true);
 564         if (rc)
 565                 return rc;
 566
 567         memset(bp->flow_stat->tx_fc_out_tbl.va,
 568                0,
 569                bp->flow_stat->tx_fc_out_tbl.size);
 570         rc = bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_TX,
 571                                        CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
 572                                        bp->flow_stat->tx_fc_out_tbl.ctx_id,
 573                                        bp->flow_stat->max_fc,
 574                                        true);
 575
 576         return rc;
 577 }
 578
 579 static int bnxt_alloc_ctx_mem_buf(char *type, size_t size,
 580                                   struct bnxt_ctx_mem_buf_info *ctx)
 581 {
 582         if (!ctx)
 583                 return -EINVAL;
 584
 585         ctx->va = rte_zmalloc(type, size, 0);
 586         if (ctx->va == NULL)
 587                 return -ENOMEM;
 588         rte_mem_lock_page(ctx->va);
 589         ctx->size = size;
 590         ctx->dma = rte_mem_virt2iova(ctx->va);
 591         if (ctx->dma == RTE_BAD_IOVA)
 592                 return -ENOMEM;
 593
 594         return 0;
 595 }
 596
 597 static int bnxt_init_fc_ctx_mem(struct bnxt *bp)
 598 {
 599         struct rte_pci_device *pdev = bp->pdev;
 600         char type[RTE_MEMZONE_NAMESIZE];
 601         uint16_t max_fc;
 602         int rc = 0;
 603
 604         max_fc = bp->flow_stat->max_fc;
 605
 606         sprintf(type, "bnxt_rx_fc_in_" PCI_PRI_FMT, pdev->addr.domain,
 607                 pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
 608         /* 4 bytes for each counter-id */
 609         rc = bnxt_alloc_ctx_mem_buf(type,
 610                                     max_fc * 4,
 611                                     &bp->flow_stat->rx_fc_in_tbl);
 612         if (rc)
 613                 return rc;
 614
 615         sprintf(type, "bnxt_rx_fc_out_" PCI_PRI_FMT, pdev->addr.domain,
 616                 pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
 617         /* 16 bytes for each counter - 8 bytes pkt_count, 8 bytes byte_count */
 618         rc = bnxt_alloc_ctx_mem_buf(type,
 619                                     max_fc * 16,
 620                                     &bp->flow_stat->rx_fc_out_tbl);
 621         if (rc)
 622                 return rc;
 623
 624         sprintf(type, "bnxt_tx_fc_in_" PCI_PRI_FMT, pdev->addr.domain,
 625                 pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
 626         /* 4 bytes for each counter-id */
 627         rc = bnxt_alloc_ctx_mem_buf(type,
 628                                     max_fc * 4,
 629                                     &bp->flow_stat->tx_fc_in_tbl);
 630         if (rc)
 631                 return rc;
 632
 633         sprintf(type, "bnxt_tx_fc_out_" PCI_PRI_FMT, pdev->addr.domain,
 634                 pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
 635         /* 16 bytes for each counter - 8 bytes pkt_count, 8 bytes byte_count */
 636         rc = bnxt_alloc_ctx_mem_buf(type,
 637                                     max_fc * 16,
 638                                     &bp->flow_stat->tx_fc_out_tbl);
 639         if (rc)
 640                 return rc;
 641
 642         rc = bnxt_register_fc_ctx_mem(bp);
 643
 644         return rc;
 645 }
 646
 647 static int bnxt_init_ctx_mem(struct bnxt *bp)
 648 {
 649         int rc = 0;
 650
 651         if (!(bp->fw_cap & BNXT_FW_CAP_ADV_FLOW_COUNTERS) ||
 652             !(BNXT_PF(bp) || BNXT_VF_IS_TRUSTED(bp)) ||
 653             !BNXT_FLOW_XSTATS_EN(bp))
 654                 return 0;
 655
 656         rc = bnxt_hwrm_cfa_counter_qcaps(bp, &bp->flow_stat->max_fc);
 657         if (rc)
 658                 return rc;
 659
 660         rc = bnxt_init_fc_ctx_mem(bp);
 661
 662         return rc;
 663 }
 664
 665 static int bnxt_update_phy_setting(struct bnxt *bp)
 666 {
 667         struct rte_eth_link new;
 668         int rc;
 669
 670         rc = bnxt_get_hwrm_link_config(bp, &new);
 671         if (rc) {
 672                 PMD_DRV_LOG(ERR, "Failed to get link settings\n");
 673                 return rc;
 674         }
 675
 676         /*
 677          * On BCM957508-N2100 adapters, FW will not allow any user other
 678          * than BMC to shutdown the port. bnxt_get_hwrm_link_config() call
 679          * always returns link up. Force phy update always in that case.
 680          */
 681         if (!new.link_status || IS_BNXT_DEV_957508_N2100(bp)) {
 682                 rc = bnxt_set_hwrm_link_config(bp, true);
 683                 if (rc) {
 684                         PMD_DRV_LOG(ERR, "Failed to update PHY settings\n");
 685                         return rc;
 686                 }
 687         }
 688
 689         return rc;
 690 }
 691
 692 static int bnxt_init_chip(struct bnxt *bp)
 693 {
 694         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(bp->eth_dev);
 695         struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
 696         uint32_t intr_vector = 0;
 697         uint32_t queue_id, base = BNXT_MISC_VEC_ID;
 698         uint32_t vec = BNXT_MISC_VEC_ID;
 699         unsigned int i, j;
 700         int rc;
 701
 702         if (bp->eth_dev->data->mtu > RTE_ETHER_MTU) {
 703                 bp->eth_dev->data->dev_conf.rxmode.offloads |=
 704                         DEV_RX_OFFLOAD_JUMBO_FRAME;
 705                 bp->flags |= BNXT_FLAG_JUMBO;
 706         } else {
 707                 bp->eth_dev->data->dev_conf.rxmode.offloads &=
 708                         ~DEV_RX_OFFLOAD_JUMBO_FRAME;
 709                 bp->flags &= ~BNXT_FLAG_JUMBO;
 710         }
 711
 712         /* THOR does not support ring groups.
 713          * But we will use the array to save RSS context IDs.
 714          */
 715         if (BNXT_CHIP_P5(bp))
 716                 bp->max_ring_grps = BNXT_MAX_RSS_CTXTS_P5;
 717
 718         rc = bnxt_alloc_all_hwrm_stat_ctxs(bp);
 719         if (rc) {
 720                 PMD_DRV_LOG(ERR, "HWRM stat ctx alloc failure rc: %x\n", rc);
 721                 goto err_out;
 722         }
 723
 724         rc = bnxt_alloc_hwrm_rings(bp);
 725         if (rc) {
 726                 PMD_DRV_LOG(ERR, "HWRM ring alloc failure rc: %x\n", rc);
 727                 goto err_out;
 728         }
 729
 730         rc = bnxt_alloc_all_hwrm_ring_grps(bp);
 731         if (rc) {
 732                 PMD_DRV_LOG(ERR, "HWRM ring grp alloc failure: %x\n", rc);
 733                 goto err_out;
 734         }
 735
 736         if (!(bp->vnic_cap_flags & BNXT_VNIC_CAP_COS_CLASSIFY))
 737                 goto skip_cosq_cfg;
 738
 739         for (j = 0, i = 0; i < BNXT_COS_QUEUE_COUNT; i++) {
 740                 if (bp->rx_cos_queue[i].id != 0xff) {
 741                         struct bnxt_vnic_info *vnic = &bp->vnic_info[j++];
 742
 743                         if (!vnic) {
 744                                 PMD_DRV_LOG(ERR,
 745                                             "Num pools more than FW profile\n");
 746                                 rc = -EINVAL;
 747                                 goto err_out;
 748                         }
 749                         vnic->cos_queue_id = bp->rx_cos_queue[i].id;
 750                         bp->rx_cosq_cnt++;
 751                 }
 752         }
 753
 754 skip_cosq_cfg:
 755         rc = bnxt_mq_rx_configure(bp);
 756         if (rc) {
 757                 PMD_DRV_LOG(ERR, "MQ mode configure failure rc: %x\n", rc);
 758                 goto err_out;
 759         }
 760
 761         /* VNIC configuration */
 762         for (i = 0; i < bp->nr_vnics; i++) {
 763                 rc = bnxt_setup_one_vnic(bp, i);
 764                 if (rc)
 765                         goto err_out;
 766         }
 767
 768         rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, &bp->vnic_info[0], 0, NULL);
 769         if (rc) {
 770                 PMD_DRV_LOG(ERR,
 771                         "HWRM cfa l2 rx mask failure rc: %x\n", rc);
 772                 goto err_out;
 773         }
 774
 775         /* check and configure queue intr-vector mapping */
 776         if ((rte_intr_cap_multiple(intr_handle) ||
 777              !RTE_ETH_DEV_SRIOV(bp->eth_dev).active) &&
 778             bp->eth_dev->data->dev_conf.intr_conf.rxq != 0) {
 779                 intr_vector = bp->eth_dev->data->nb_rx_queues;
 780                 PMD_DRV_LOG(DEBUG, "intr_vector = %d\n", intr_vector);
 781                 if (intr_vector > bp->rx_cp_nr_rings) {
 782                         PMD_DRV_LOG(ERR, "At most %d intr queues supported",
 783                                         bp->rx_cp_nr_rings);
 784                         return -ENOTSUP;
 785                 }
 786                 rc = rte_intr_efd_enable(intr_handle, intr_vector);
 787                 if (rc)
 788                         return rc;
 789         }
 790
 791         if (rte_intr_dp_is_en(intr_handle) && !intr_handle->intr_vec) {
 792                 intr_handle->intr_vec =
 793                         rte_zmalloc("intr_vec",
 794                                     bp->eth_dev->data->nb_rx_queues *
 795                                     sizeof(int), 0);
 796                 if (intr_handle->intr_vec == NULL) {
 797                         PMD_DRV_LOG(ERR, "Failed to allocate %d rx_queues"
 798                                 " intr_vec", bp->eth_dev->data->nb_rx_queues);
 799                         rc = -ENOMEM;
 800                         goto err_disable;
 801                 }
 802                 PMD_DRV_LOG(DEBUG, "intr_handle->intr_vec = %p "
 803                         "intr_handle->nb_efd = %d intr_handle->max_intr = %d\n",
 804                          intr_handle->intr_vec, intr_handle->nb_efd,
 805                         intr_handle->max_intr);
 806                 for (queue_id = 0; queue_id < bp->eth_dev->data->nb_rx_queues;
 807                      queue_id++) {
 808                         intr_handle->intr_vec[queue_id] =
 809                                                         vec + BNXT_RX_VEC_START;
 810                         if (vec < base + intr_handle->nb_efd - 1)
 811                                 vec++;
 812                 }
 813         }
 814
 815         /* enable uio/vfio intr/eventfd mapping */
 816         rc = rte_intr_enable(intr_handle);
 817 #ifndef RTE_EXEC_ENV_FREEBSD
 818         /* In FreeBSD OS, nic_uio driver does not support interrupts */
 819         if (rc)
 820                 goto err_free;
 821 #endif
 822
 823         rc = bnxt_update_phy_setting(bp);
 824         if (rc)
 825                 goto err_free;
 826
 827         bp->mark_table = rte_zmalloc("bnxt_mark_table", BNXT_MARK_TABLE_SZ, 0);
 828         if (!bp->mark_table)
 829                 PMD_DRV_LOG(ERR, "Allocation of mark table failed\n");
 830
 831         return 0;
 832
 833 err_free:
 834         rte_free(intr_handle->intr_vec);
 835 err_disable:
 836         rte_intr_efd_disable(intr_handle);
 837 err_out:
 838         /* Some of the error status returned by FW may not be from errno.h */
 839         if (rc > 0)
 840                 rc = -EIO;
 841
 842         return rc;
 843 }
 844
 845 static int bnxt_shutdown_nic(struct bnxt *bp)
 846 {
 847         bnxt_free_all_hwrm_resources(bp);
 848         bnxt_free_all_filters(bp);
 849         bnxt_free_all_vnics(bp);
 850         return 0;
 851 }
 852
 853 /*
 854  * Device configuration and status function
 855  */
 856
 857 uint32_t bnxt_get_speed_capabilities(struct bnxt *bp)
 858 {
 859         uint32_t link_speed = bp->link_info->support_speeds;
 860         uint32_t speed_capa = 0;
 861
 862         /* If PAM4 is configured, use PAM4 supported speed */
 863         if (link_speed == 0 && bp->link_info->support_pam4_speeds > 0)
 864                 link_speed = bp->link_info->support_pam4_speeds;
 865
 866         if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_LINK_SPEED_100MB)
 867                 speed_capa |= ETH_LINK_SPEED_100M;
 868         if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_100MBHD)
 869                 speed_capa |= ETH_LINK_SPEED_100M_HD;
 870         if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_1GB)
 871                 speed_capa |= ETH_LINK_SPEED_1G;
 872         if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_2_5GB)
 873                 speed_capa |= ETH_LINK_SPEED_2_5G;
 874         if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_10GB)
 875                 speed_capa |= ETH_LINK_SPEED_10G;
 876         if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_20GB)
 877                 speed_capa |= ETH_LINK_SPEED_20G;
 878         if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_25GB)
 879                 speed_capa |= ETH_LINK_SPEED_25G;
 880         if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_40GB)
 881                 speed_capa |= ETH_LINK_SPEED_40G;
 882         if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_50GB)
 883                 speed_capa |= ETH_LINK_SPEED_50G;
 884         if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_100GB)
 885                 speed_capa |= ETH_LINK_SPEED_100G;
 886         if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_PAM4_SPEEDS_50G)
 887                 speed_capa |= ETH_LINK_SPEED_50G;
 888         if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_PAM4_SPEEDS_100G)
 889                 speed_capa |= ETH_LINK_SPEED_100G;
 890         if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_PAM4_SPEEDS_200G)
 891                 speed_capa |= ETH_LINK_SPEED_200G;
 892
 893         if (bp->link_info->auto_mode ==
 894             HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_MODE_NONE)
 895                 speed_capa |= ETH_LINK_SPEED_FIXED;
 896         else
 897                 speed_capa |= ETH_LINK_SPEED_AUTONEG;
 898
 899         return speed_capa;
 900 }
 901
 902 static int bnxt_dev_info_get_op(struct rte_eth_dev *eth_dev,
 903                                 struct rte_eth_dev_info *dev_info)
 904 {
 905         struct rte_pci_device *pdev = RTE_DEV_TO_PCI(eth_dev->device);
 906         struct bnxt *bp = eth_dev->data->dev_private;
 907         uint16_t max_vnics, i, j, vpool, vrxq;
 908         unsigned int max_rx_rings;
 909         int rc;
 910
 911         rc = is_bnxt_in_error(bp);
 912         if (rc)
 913                 return rc;
 914
 915         /* MAC Specifics */
 916         dev_info->max_mac_addrs = bp->max_l2_ctx;
 917         dev_info->max_hash_mac_addrs = 0;
 918
 919         /* PF/VF specifics */
 920         if (BNXT_PF(bp))
 921                 dev_info->max_vfs = pdev->max_vfs;
 922
 923         max_rx_rings = bnxt_max_rings(bp);
 924         /* For the sake of symmetry, max_rx_queues = max_tx_queues */
 925         dev_info->max_rx_queues = max_rx_rings;
 926         dev_info->max_tx_queues = max_rx_rings;
 927         dev_info->reta_size = bnxt_rss_hash_tbl_size(bp);
 928         dev_info->hash_key_size = 40;
 929         max_vnics = bp->max_vnics;
 930
 931         /* MTU specifics */
 932         dev_info->min_mtu = RTE_ETHER_MIN_MTU;
 933         dev_info->max_mtu = BNXT_MAX_MTU;
 934
 935         /* Fast path specifics */
 936         dev_info->min_rx_bufsize = 1;
 937         dev_info->max_rx_pktlen = BNXT_MAX_PKT_LEN;
 938
 939         dev_info->rx_offload_capa = BNXT_DEV_RX_OFFLOAD_SUPPORT;
 940         if (bp->flags & BNXT_FLAG_PTP_SUPPORTED)
 941                 dev_info->rx_offload_capa |= DEV_RX_OFFLOAD_TIMESTAMP;
 942         dev_info->tx_queue_offload_capa = DEV_TX_OFFLOAD_MBUF_FAST_FREE;
 943         dev_info->tx_offload_capa = BNXT_DEV_TX_OFFLOAD_SUPPORT |
 944                                     dev_info->tx_queue_offload_capa;
 945         dev_info->flow_type_rss_offloads = BNXT_ETH_RSS_SUPPORT;
 946
 947         dev_info->speed_capa = bnxt_get_speed_capabilities(bp);
 948
 949         /* *INDENT-OFF* */
 950         dev_info->default_rxconf = (struct rte_eth_rxconf) {
 951                 .rx_thresh = {
 952                         .pthresh = 8,
 953                         .hthresh = 8,
 954                         .wthresh = 0,
 955                 },
 956                 .rx_free_thresh = 32,
 957                 .rx_drop_en = BNXT_DEFAULT_RX_DROP_EN,
 958         };
 959
 960         dev_info->default_txconf = (struct rte_eth_txconf) {
 961                 .tx_thresh = {
 962                         .pthresh = 32,
 963                         .hthresh = 0,
 964                         .wthresh = 0,
 965                 },
 966                 .tx_free_thresh = 32,
 967                 .tx_rs_thresh = 32,
 968         };
 969         eth_dev->data->dev_conf.intr_conf.lsc = 1;
 970
 971         eth_dev->data->dev_conf.intr_conf.rxq = 1;
 972         dev_info->rx_desc_lim.nb_min = BNXT_MIN_RING_DESC;
 973         dev_info->rx_desc_lim.nb_max = BNXT_MAX_RX_RING_DESC;
 974         dev_info->tx_desc_lim.nb_min = BNXT_MIN_RING_DESC;
 975         dev_info->tx_desc_lim.nb_max = BNXT_MAX_TX_RING_DESC;
 976
 977         if (BNXT_PF(bp) || BNXT_VF_IS_TRUSTED(bp)) {
 978                 dev_info->switch_info.name = eth_dev->device->name;
 979                 dev_info->switch_info.domain_id = bp->switch_domain_id;
 980                 dev_info->switch_info.port_id =
 981                                 BNXT_PF(bp) ? BNXT_SWITCH_PORT_ID_PF :
 982                                     BNXT_SWITCH_PORT_ID_TRUSTED_VF;
 983         }
 984
 985         /* *INDENT-ON* */
 986
 987         /*
 988          * TODO: default_rxconf, default_txconf, rx_desc_lim, and tx_desc_lim
 989          *       need further investigation.
 990          */
 991
 992         /* VMDq resources */
 993         vpool = 64; /* ETH_64_POOLS */
 994         vrxq = 128; /* ETH_VMDQ_DCB_NUM_QUEUES */
 995         for (i = 0; i < 4; vpool >>= 1, i++) {
 996                 if (max_vnics > vpool) {
 997                         for (j = 0; j < 5; vrxq >>= 1, j++) {
 998                                 if (dev_info->max_rx_queues > vrxq) {
 999                                         if (vpool > vrxq)
1000                                                 vpool = vrxq;
1001                                         goto found;
1002                                 }
1003                         }
1004                         /* Not enough resources to support VMDq */
1005                         break;
1006                 }
1007         }
1008         /* Not enough resources to support VMDq */
1009         vpool = 0;
1010         vrxq = 0;
1011 found:
1012         dev_info->max_vmdq_pools = vpool;
1013         dev_info->vmdq_queue_num = vrxq;
1014
1015         dev_info->vmdq_pool_base = 0;
1016         dev_info->vmdq_queue_base = 0;
1017
1018         return 0;
1019 }
1020
1021 /* Configure the device based on the configuration provided */
1022 static int bnxt_dev_configure_op(struct rte_eth_dev *eth_dev)
1023 {
1024         struct bnxt *bp = eth_dev->data->dev_private;
1025         uint64_t rx_offloads = eth_dev->data->dev_conf.rxmode.offloads;
1026         int rc;
1027
1028         bp->rx_queues = (void *)eth_dev->data->rx_queues;
1029         bp->tx_queues = (void *)eth_dev->data->tx_queues;
1030         bp->tx_nr_rings = eth_dev->data->nb_tx_queues;
1031         bp->rx_nr_rings = eth_dev->data->nb_rx_queues;
1032
1033         rc = is_bnxt_in_error(bp);
1034         if (rc)
1035                 return rc;
1036
1037         if (BNXT_VF(bp) && (bp->flags & BNXT_FLAG_NEW_RM)) {
1038                 rc = bnxt_hwrm_check_vf_rings(bp);
1039                 if (rc) {
1040                         PMD_DRV_LOG(ERR, "HWRM insufficient resources\n");
1041                         return -ENOSPC;
1042                 }
1043
1044                 /* If a resource has already been allocated - in this case
1045                  * it is the async completion ring, free it. Reallocate it after
1046                  * resource reservation. This will ensure the resource counts
1047                  * are calculated correctly.
1048                  */
1049
1050                 pthread_mutex_lock(&bp->def_cp_lock);
1051
1052                 if (!BNXT_HAS_NQ(bp) && bp->async_cp_ring) {
1053                         bnxt_disable_int(bp);
1054                         bnxt_free_cp_ring(bp, bp->async_cp_ring);
1055                 }
1056
1057                 rc = bnxt_hwrm_func_reserve_vf_resc(bp, false);
1058                 if (rc) {
1059                         PMD_DRV_LOG(ERR, "HWRM resource alloc fail:%x\n", rc);
1060                         pthread_mutex_unlock(&bp->def_cp_lock);
1061                         return -ENOSPC;
1062                 }
1063
1064                 if (!BNXT_HAS_NQ(bp) && bp->async_cp_ring) {
1065                         rc = bnxt_alloc_async_cp_ring(bp);
1066                         if (rc) {
1067                                 pthread_mutex_unlock(&bp->def_cp_lock);
1068                                 return rc;
1069                         }
1070                         bnxt_enable_int(bp);
1071                 }
1072
1073                 pthread_mutex_unlock(&bp->def_cp_lock);
1074         } else {
1075                 /* legacy driver needs to get updated values */
1076                 rc = bnxt_hwrm_func_qcaps(bp);
1077                 if (rc) {
1078                         PMD_DRV_LOG(ERR, "hwrm func qcaps fail:%d\n", rc);
1079                         return rc;
1080                 }
1081         }
1082
1083         /* Inherit new configurations */
1084         if (eth_dev->data->nb_rx_queues > bp->max_rx_rings ||
1085             eth_dev->data->nb_tx_queues > bp->max_tx_rings ||
1086             eth_dev->data->nb_rx_queues + eth_dev->data->nb_tx_queues
1087                 + BNXT_NUM_ASYNC_CPR(bp) > bp->max_cp_rings ||
1088             eth_dev->data->nb_rx_queues + eth_dev->data->nb_tx_queues >
1089             bp->max_stat_ctx)
1090                 goto resource_error;
1091
1092         if (BNXT_HAS_RING_GRPS(bp) &&
1093             (uint32_t)(eth_dev->data->nb_rx_queues) > bp->max_ring_grps)
1094                 goto resource_error;
1095
1096         if (!(eth_dev->data->dev_conf.rxmode.mq_mode & ETH_MQ_RX_RSS) &&
1097             bp->max_vnics < eth_dev->data->nb_rx_queues)
1098                 goto resource_error;
1099
1100         bp->rx_cp_nr_rings = bp->rx_nr_rings;
1101         bp->tx_cp_nr_rings = bp->tx_nr_rings;
1102
1103         if (eth_dev->data->dev_conf.rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG)
1104                 rx_offloads |= DEV_RX_OFFLOAD_RSS_HASH;
1105         eth_dev->data->dev_conf.rxmode.offloads = rx_offloads;
1106
1107         if (rx_offloads & DEV_RX_OFFLOAD_JUMBO_FRAME) {
1108                 eth_dev->data->mtu =
1109                         eth_dev->data->dev_conf.rxmode.max_rx_pkt_len -
1110                         RTE_ETHER_HDR_LEN - RTE_ETHER_CRC_LEN - VLAN_TAG_SIZE *
1111                         BNXT_NUM_VLANS;
1112                 bnxt_mtu_set_op(eth_dev, eth_dev->data->mtu);
1113         }
1114         return 0;
1115
1116 resource_error:
1117         PMD_DRV_LOG(ERR,
1118                     "Insufficient resources to support requested config\n");
1119         PMD_DRV_LOG(ERR,
1120                     "Num Queues Requested: Tx %d, Rx %d\n",
1121                     eth_dev->data->nb_tx_queues,
1122                     eth_dev->data->nb_rx_queues);
1123         PMD_DRV_LOG(ERR,
1124                     "MAX: TxQ %d, RxQ %d, CQ %d Stat %d, Grp %d, Vnic %d\n",
1125                     bp->max_tx_rings, bp->max_rx_rings, bp->max_cp_rings,
1126                     bp->max_stat_ctx, bp->max_ring_grps, bp->max_vnics);
1127         return -ENOSPC;
1128 }
1129
1130 void bnxt_print_link_info(struct rte_eth_dev *eth_dev)
1131 {
1132         struct rte_eth_link *link = &eth_dev->data->dev_link;
1133
1134         if (link->link_status)
1135                 PMD_DRV_LOG(INFO, "Port %d Link Up - speed %u Mbps - %s\n",
1136                         eth_dev->data->port_id,
1137                         (uint32_t)link->link_speed,
1138                         (link->link_duplex == ETH_LINK_FULL_DUPLEX) ?
1139                         ("full-duplex") : ("half-duplex\n"));
1140         else
1141                 PMD_DRV_LOG(INFO, "Port %d Link Down\n",
1142                         eth_dev->data->port_id);
1143 }
1144
1145 /*
1146  * Determine whether the current configuration requires support for scattered
1147  * receive; return 1 if scattered receive is required and 0 if not.
1148  */
1149 static int bnxt_scattered_rx(struct rte_eth_dev *eth_dev)
1150 {
1151         uint16_t buf_size;
1152         int i;
1153
1154         if (eth_dev->data->dev_conf.rxmode.offloads & DEV_RX_OFFLOAD_SCATTER)
1155                 return 1;
1156
1157         for (i = 0; i < eth_dev->data->nb_rx_queues; i++) {
1158                 struct bnxt_rx_queue *rxq = eth_dev->data->rx_queues[i];
1159
1160                 buf_size = (uint16_t)(rte_pktmbuf_data_room_size(rxq->mb_pool) -
1161                                       RTE_PKTMBUF_HEADROOM);
1162                 if (eth_dev->data->dev_conf.rxmode.max_rx_pkt_len > buf_size)
1163                         return 1;
1164         }
1165         return 0;
1166 }
1167
1168 static eth_rx_burst_t
1169 bnxt_receive_function(struct rte_eth_dev *eth_dev)
1170 {
1171         struct bnxt *bp = eth_dev->data->dev_private;
1172
1173 #if defined(RTE_ARCH_X86) || defined(RTE_ARCH_ARM64)
1174 #ifndef RTE_LIBRTE_IEEE1588
1175         /*
1176          * Vector mode receive can be enabled only if scatter rx is not
1177          * in use and rx offloads are limited to VLAN stripping and
1178          * CRC stripping.
1179          */
1180         if (!eth_dev->data->scattered_rx &&
1181             !(eth_dev->data->dev_conf.rxmode.offloads &
1182               ~(DEV_RX_OFFLOAD_VLAN_STRIP |
1183                 DEV_RX_OFFLOAD_KEEP_CRC |
1184                 DEV_RX_OFFLOAD_JUMBO_FRAME |
1185                 DEV_RX_OFFLOAD_IPV4_CKSUM |
1186                 DEV_RX_OFFLOAD_UDP_CKSUM |
1187                 DEV_RX_OFFLOAD_TCP_CKSUM |
1188                 DEV_RX_OFFLOAD_OUTER_IPV4_CKSUM |
1189                 DEV_RX_OFFLOAD_RSS_HASH |
1190                 DEV_RX_OFFLOAD_VLAN_FILTER)) &&
1191             !BNXT_TRUFLOW_EN(bp) && BNXT_NUM_ASYNC_CPR(bp) &&
1192             rte_vect_get_max_simd_bitwidth() >= RTE_VECT_SIMD_128) {
1193                 PMD_DRV_LOG(INFO, "Using vector mode receive for port %d\n",
1194                             eth_dev->data->port_id);
1195                 bp->flags |= BNXT_FLAG_RX_VECTOR_PKT_MODE;
1196                 return bnxt_recv_pkts_vec;
1197         }
1198         PMD_DRV_LOG(INFO, "Vector mode receive disabled for port %d\n",
1199                     eth_dev->data->port_id);
1200         PMD_DRV_LOG(INFO,
1201                     "Port %d scatter: %d rx offload: %" PRIX64 "\n",
1202                     eth_dev->data->port_id,
1203                     eth_dev->data->scattered_rx,
1204                     eth_dev->data->dev_conf.rxmode.offloads);
1205 #endif
1206 #endif
1207         bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
1208         return bnxt_recv_pkts;
1209 }
1210
1211 static eth_tx_burst_t
1212 bnxt_transmit_function(__rte_unused struct rte_eth_dev *eth_dev)
1213 {
1214 #if defined(RTE_ARCH_X86) || defined(RTE_ARCH_ARM64)
1215 #ifndef RTE_LIBRTE_IEEE1588
1216         uint64_t offloads = eth_dev->data->dev_conf.txmode.offloads;
1217         struct bnxt *bp = eth_dev->data->dev_private;
1218
1219         /*
1220          * Vector mode transmit can be enabled only if not using scatter rx
1221          * or tx offloads.
1222          */
1223         if (!eth_dev->data->scattered_rx &&
1224             !(offloads & ~DEV_TX_OFFLOAD_MBUF_FAST_FREE) &&
1225             !BNXT_TRUFLOW_EN(bp) &&
1226             rte_vect_get_max_simd_bitwidth() >= RTE_VECT_SIMD_128) {
1227                 PMD_DRV_LOG(INFO, "Using vector mode transmit for port %d\n",
1228                             eth_dev->data->port_id);
1229                 return bnxt_xmit_pkts_vec;
1230         }
1231         PMD_DRV_LOG(INFO, "Vector mode transmit disabled for port %d\n",
1232                     eth_dev->data->port_id);
1233         PMD_DRV_LOG(INFO,
1234                     "Port %d scatter: %d tx offload: %" PRIX64 "\n",
1235                     eth_dev->data->port_id,
1236                     eth_dev->data->scattered_rx,
1237                     offloads);
1238 #endif
1239 #endif
1240         return bnxt_xmit_pkts;
1241 }
1242
1243 static int bnxt_handle_if_change_status(struct bnxt *bp)
1244 {
1245         int rc;
1246
1247         /* Since fw has undergone a reset and lost all contexts,
1248          * set fatal flag to not issue hwrm during cleanup
1249          */
1250         bp->flags |= BNXT_FLAG_FATAL_ERROR;
1251         bnxt_uninit_resources(bp, true);
1252
1253         /* clear fatal flag so that re-init happens */
1254         bp->flags &= ~BNXT_FLAG_FATAL_ERROR;
1255         rc = bnxt_init_resources(bp, true);
1256
1257         bp->flags &= ~BNXT_FLAG_IF_CHANGE_HOT_FW_RESET_DONE;
1258
1259         return rc;
1260 }
1261
1262 static int bnxt_dev_start_op(struct rte_eth_dev *eth_dev)
1263 {
1264         struct bnxt *bp = eth_dev->data->dev_private;
1265         uint64_t rx_offloads = eth_dev->data->dev_conf.rxmode.offloads;
1266         int vlan_mask = 0;
1267         int rc, retry_cnt = BNXT_IF_CHANGE_RETRY_COUNT;
1268
1269         if (!eth_dev->data->nb_tx_queues || !eth_dev->data->nb_rx_queues) {
1270                 PMD_DRV_LOG(ERR, "Queues are not configured yet!\n");
1271                 return -EINVAL;
1272         }
1273
1274         if (bp->rx_cp_nr_rings > RTE_ETHDEV_QUEUE_STAT_CNTRS) {
1275                 PMD_DRV_LOG(ERR,
1276                         "RxQ cnt %d > RTE_ETHDEV_QUEUE_STAT_CNTRS %d\n",
1277                         bp->rx_cp_nr_rings, RTE_ETHDEV_QUEUE_STAT_CNTRS);
1278         }
1279
1280         do {
1281                 rc = bnxt_hwrm_if_change(bp, true);
1282                 if (rc == 0 || rc != -EAGAIN)
1283                         break;
1284
1285                 rte_delay_ms(BNXT_IF_CHANGE_RETRY_INTERVAL);
1286         } while (retry_cnt--);
1287
1288         if (rc)
1289                 return rc;
1290
1291         if (bp->flags & BNXT_FLAG_IF_CHANGE_HOT_FW_RESET_DONE) {
1292                 rc = bnxt_handle_if_change_status(bp);
1293                 if (rc)
1294                         return rc;
1295         }
1296
1297         bnxt_enable_int(bp);
1298
1299         rc = bnxt_init_chip(bp);
1300         if (rc)
1301                 goto error;
1302
1303         eth_dev->data->scattered_rx = bnxt_scattered_rx(eth_dev);
1304         eth_dev->data->dev_started = 1;
1305
1306         bnxt_link_update_op(eth_dev, 1);
1307
1308         if (rx_offloads & DEV_RX_OFFLOAD_VLAN_FILTER)
1309                 vlan_mask |= ETH_VLAN_FILTER_MASK;
1310         if (rx_offloads & DEV_RX_OFFLOAD_VLAN_STRIP)
1311                 vlan_mask |= ETH_VLAN_STRIP_MASK;
1312         rc = bnxt_vlan_offload_set_op(eth_dev, vlan_mask);
1313         if (rc)
1314                 goto error;
1315
1316         /* Initialize bnxt ULP port details */
1317         rc = bnxt_ulp_port_init(bp);
1318         if (rc)
1319                 goto error;
1320
1321         eth_dev->rx_pkt_burst = bnxt_receive_function(eth_dev);
1322         eth_dev->tx_pkt_burst = bnxt_transmit_function(eth_dev);
1323
1324         bnxt_schedule_fw_health_check(bp);
1325
1326         return 0;
1327
1328 error:
1329         bnxt_shutdown_nic(bp);
1330         bnxt_free_tx_mbufs(bp);
1331         bnxt_free_rx_mbufs(bp);
1332         bnxt_hwrm_if_change(bp, false);
1333         eth_dev->data->dev_started = 0;
1334         return rc;
1335 }
1336
1337 static int bnxt_dev_set_link_up_op(struct rte_eth_dev *eth_dev)
1338 {
1339         struct bnxt *bp = eth_dev->data->dev_private;
1340         int rc = 0;
1341
1342         if (!bp->link_info->link_up)
1343                 rc = bnxt_set_hwrm_link_config(bp, true);
1344         if (!rc)
1345                 eth_dev->data->dev_link.link_status = 1;
1346
1347         bnxt_print_link_info(eth_dev);
1348         return rc;
1349 }
1350
1351 static int bnxt_dev_set_link_down_op(struct rte_eth_dev *eth_dev)
1352 {
1353         struct bnxt *bp = eth_dev->data->dev_private;
1354
1355         eth_dev->data->dev_link.link_status = 0;
1356         bnxt_set_hwrm_link_config(bp, false);
1357         bp->link_info->link_up = 0;
1358
1359         return 0;
1360 }
1361
1362 static void bnxt_free_switch_domain(struct bnxt *bp)
1363 {
1364         int rc = 0;
1365
1366         if (bp->switch_domain_id) {
1367                 rc = rte_eth_switch_domain_free(bp->switch_domain_id);
1368                 if (rc)
1369                         PMD_DRV_LOG(ERR, "free switch domain:%d fail: %d\n",
1370                                     bp->switch_domain_id, rc);
1371         }
1372 }
1373
1374 /* Unload the driver, release resources */
1375 static int bnxt_dev_stop_op(struct rte_eth_dev *eth_dev)
1376 {
1377         struct bnxt *bp = eth_dev->data->dev_private;
1378         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
1379         struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
1380         struct rte_eth_link link;
1381         int ret;
1382
1383         eth_dev->data->dev_started = 0;
1384         eth_dev->data->scattered_rx = 0;
1385
1386         /* Prevent crashes when queues are still in use */
1387         eth_dev->rx_pkt_burst = &bnxt_dummy_recv_pkts;
1388         eth_dev->tx_pkt_burst = &bnxt_dummy_xmit_pkts;
1389
1390         bnxt_disable_int(bp);
1391
1392         /* disable uio/vfio intr/eventfd mapping */
1393         rte_intr_disable(intr_handle);
1394
1395         /* Stop the child representors for this device */
1396         ret = bnxt_rep_stop_all(bp);
1397         if (ret != 0)
1398                 return ret;
1399
1400         /* delete the bnxt ULP port details */
1401         bnxt_ulp_port_deinit(bp);
1402
1403         bnxt_cancel_fw_health_check(bp);
1404
1405         /* Do not bring link down during reset recovery */
1406         if (!is_bnxt_in_error(bp)) {
1407                 bnxt_dev_set_link_down_op(eth_dev);
1408                 /* Wait for link to be reset */
1409                 if (BNXT_SINGLE_PF(bp))
1410                         rte_delay_ms(500);
1411                 /* clear the recorded link status */
1412                 memset(&link, 0, sizeof(link));
1413                 rte_eth_linkstatus_set(eth_dev, &link);
1414         }
1415
1416         /* Clean queue intr-vector mapping */
1417         rte_intr_efd_disable(intr_handle);
1418         if (intr_handle->intr_vec != NULL) {
1419                 rte_free(intr_handle->intr_vec);
1420                 intr_handle->intr_vec = NULL;
1421         }
1422
1423         bnxt_hwrm_port_clr_stats(bp);
1424         bnxt_free_tx_mbufs(bp);
1425         bnxt_free_rx_mbufs(bp);
1426         /* Process any remaining notifications in default completion queue */
1427         bnxt_int_handler(eth_dev);
1428         bnxt_shutdown_nic(bp);
1429         bnxt_hwrm_if_change(bp, false);
1430
1431         rte_free(bp->mark_table);
1432         bp->mark_table = NULL;
1433
1434         bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
1435         bp->rx_cosq_cnt = 0;
1436         /* All filters are deleted on a port stop. */
1437         if (BNXT_FLOW_XSTATS_EN(bp))
1438                 bp->flow_stat->flow_count = 0;
1439
1440         return 0;
1441 }
1442
1443 static int bnxt_dev_close_op(struct rte_eth_dev *eth_dev)
1444 {
1445         struct bnxt *bp = eth_dev->data->dev_private;
1446         int ret = 0;
1447
1448         if (rte_eal_process_type() != RTE_PROC_PRIMARY)
1449                 return 0;
1450
1451         /* cancel the recovery handler before remove dev */
1452         rte_eal_alarm_cancel(bnxt_dev_reset_and_resume, (void *)bp);
1453         rte_eal_alarm_cancel(bnxt_dev_recover, (void *)bp);
1454         bnxt_cancel_fc_thread(bp);
1455
1456         if (eth_dev->data->dev_started)
1457                 ret = bnxt_dev_stop_op(eth_dev);
1458
1459         bnxt_free_switch_domain(bp);
1460
1461         bnxt_uninit_resources(bp, false);
1462
1463         bnxt_free_leds_info(bp);
1464         bnxt_free_cos_queues(bp);
1465         bnxt_free_link_info(bp);
1466         bnxt_free_pf_info(bp);
1467         bnxt_free_parent_info(bp);
1468
1469         rte_memzone_free((const struct rte_memzone *)bp->tx_mem_zone);
1470         bp->tx_mem_zone = NULL;
1471         rte_memzone_free((const struct rte_memzone *)bp->rx_mem_zone);
1472         bp->rx_mem_zone = NULL;
1473
1474         bnxt_hwrm_free_vf_info(bp);
1475
1476         rte_free(bp->grp_info);
1477         bp->grp_info = NULL;
1478
1479         return ret;
1480 }
1481
1482 static void bnxt_mac_addr_remove_op(struct rte_eth_dev *eth_dev,
1483                                     uint32_t index)
1484 {
1485         struct bnxt *bp = eth_dev->data->dev_private;
1486         uint64_t pool_mask = eth_dev->data->mac_pool_sel[index];
1487         struct bnxt_vnic_info *vnic;
1488         struct bnxt_filter_info *filter, *temp_filter;
1489         uint32_t i;
1490
1491         if (is_bnxt_in_error(bp))
1492                 return;
1493
1494         /*
1495          * Loop through all VNICs from the specified filter flow pools to
1496          * remove the corresponding MAC addr filter
1497          */
1498         for (i = 0; i < bp->nr_vnics; i++) {
1499                 if (!(pool_mask & (1ULL << i)))
1500                         continue;
1501
1502                 vnic = &bp->vnic_info[i];
1503                 filter = STAILQ_FIRST(&vnic->filter);
1504                 while (filter) {
1505                         temp_filter = STAILQ_NEXT(filter, next);
1506                         if (filter->mac_index == index) {
1507                                 STAILQ_REMOVE(&vnic->filter, filter,
1508                                                 bnxt_filter_info, next);
1509                                 bnxt_hwrm_clear_l2_filter(bp, filter);
1510                                 bnxt_free_filter(bp, filter);
1511                         }
1512                         filter = temp_filter;
1513                 }
1514         }
1515 }
1516
1517 static int bnxt_add_mac_filter(struct bnxt *bp, struct bnxt_vnic_info *vnic,
1518                                struct rte_ether_addr *mac_addr, uint32_t index,
1519                                uint32_t pool)
1520 {
1521         struct bnxt_filter_info *filter;
1522         int rc = 0;
1523
1524         /* Attach requested MAC address to the new l2_filter */
1525         STAILQ_FOREACH(filter, &vnic->filter, next) {
1526                 if (filter->mac_index == index) {
1527                         PMD_DRV_LOG(DEBUG,
1528                                     "MAC addr already existed for pool %d\n",
1529                                     pool);
1530                         return 0;
1531                 }
1532         }
1533
1534         filter = bnxt_alloc_filter(bp);
1535         if (!filter) {
1536                 PMD_DRV_LOG(ERR, "L2 filter alloc failed\n");
1537                 return -ENODEV;
1538         }
1539
1540         /* bnxt_alloc_filter copies default MAC to filter->l2_addr. So,
1541          * if the MAC that's been programmed now is a different one, then,
1542          * copy that addr to filter->l2_addr
1543          */
1544         if (mac_addr)
1545                 memcpy(filter->l2_addr, mac_addr, RTE_ETHER_ADDR_LEN);
1546         filter->flags |= HWRM_CFA_L2_FILTER_ALLOC_INPUT_FLAGS_OUTERMOST;
1547
1548         rc = bnxt_hwrm_set_l2_filter(bp, vnic->fw_vnic_id, filter);
1549         if (!rc) {
1550                 filter->mac_index = index;
1551                 if (filter->mac_index == 0)
1552                         STAILQ_INSERT_HEAD(&vnic->filter, filter, next);
1553                 else
1554                         STAILQ_INSERT_TAIL(&vnic->filter, filter, next);
1555         } else {
1556                 bnxt_free_filter(bp, filter);
1557         }
1558
1559         return rc;
1560 }
1561
1562 static int bnxt_mac_addr_add_op(struct rte_eth_dev *eth_dev,
1563                                 struct rte_ether_addr *mac_addr,
1564                                 uint32_t index, uint32_t pool)
1565 {
1566         struct bnxt *bp = eth_dev->data->dev_private;
1567         struct bnxt_vnic_info *vnic = &bp->vnic_info[pool];
1568         int rc = 0;
1569
1570         rc = is_bnxt_in_error(bp);
1571         if (rc)
1572                 return rc;
1573
1574         if (BNXT_VF(bp) && !BNXT_VF_IS_TRUSTED(bp)) {
1575                 PMD_DRV_LOG(ERR, "Cannot add MAC address to a VF interface\n");
1576                 return -ENOTSUP;
1577         }
1578
1579         if (!vnic) {
1580                 PMD_DRV_LOG(ERR, "VNIC not found for pool %d!\n", pool);
1581                 return -EINVAL;
1582         }
1583
1584         /* Filter settings will get applied when port is started */
1585         if (!eth_dev->data->dev_started)
1586                 return 0;
1587
1588         rc = bnxt_add_mac_filter(bp, vnic, mac_addr, index, pool);
1589
1590         return rc;
1591 }
1592
1593 int bnxt_link_update_op(struct rte_eth_dev *eth_dev, int wait_to_complete)
1594 {
1595         int rc = 0;
1596         struct bnxt *bp = eth_dev->data->dev_private;
1597         struct rte_eth_link new;
1598         int cnt = wait_to_complete ? BNXT_MAX_LINK_WAIT_CNT :
1599                         BNXT_MIN_LINK_WAIT_CNT;
1600
1601         rc = is_bnxt_in_error(bp);
1602         if (rc)
1603                 return rc;
1604
1605         memset(&new, 0, sizeof(new));
1606         do {
1607                 /* Retrieve link info from hardware */
1608                 rc = bnxt_get_hwrm_link_config(bp, &new);
1609                 if (rc) {
1610                         new.link_speed = ETH_LINK_SPEED_100M;
1611                         new.link_duplex = ETH_LINK_FULL_DUPLEX;
1612                         PMD_DRV_LOG(ERR,
1613                                 "Failed to retrieve link rc = 0x%x!\n", rc);
1614                         goto out;
1615                 }
1616
1617                 if (!wait_to_complete || new.link_status)
1618                         break;
1619
1620                 rte_delay_ms(BNXT_LINK_WAIT_INTERVAL);
1621         } while (cnt--);
1622
1623         /* Only single function PF can bring phy down.
1624          * When port is stopped, report link down for VF/MH/NPAR functions.
1625          */
1626         if (!BNXT_SINGLE_PF(bp) && !eth_dev->data->dev_started)
1627                 memset(&new, 0, sizeof(new));
1628
1629 out:
1630         /* Timed out or success */
1631         if (new.link_status != eth_dev->data->dev_link.link_status ||
1632         new.link_speed != eth_dev->data->dev_link.link_speed) {
1633                 rte_eth_linkstatus_set(eth_dev, &new);
1634
1635                 rte_eth_dev_callback_process(eth_dev,
1636                                              RTE_ETH_EVENT_INTR_LSC,
1637                                              NULL);
1638
1639                 bnxt_print_link_info(eth_dev);
1640         }
1641
1642         return rc;
1643 }
1644
1645 static int bnxt_promiscuous_enable_op(struct rte_eth_dev *eth_dev)
1646 {
1647         struct bnxt *bp = eth_dev->data->dev_private;
1648         struct bnxt_vnic_info *vnic;
1649         uint32_t old_flags;
1650         int rc;
1651
1652         rc = is_bnxt_in_error(bp);
1653         if (rc)
1654                 return rc;
1655
1656         /* Filter settings will get applied when port is started */
1657         if (!eth_dev->data->dev_started)
1658                 return 0;
1659
1660         if (bp->vnic_info == NULL)
1661                 return 0;
1662
1663         vnic = BNXT_GET_DEFAULT_VNIC(bp);
1664
1665         old_flags = vnic->flags;
1666         vnic->flags |= BNXT_VNIC_INFO_PROMISC;
1667         rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1668         if (rc != 0)
1669                 vnic->flags = old_flags;
1670
1671         return rc;
1672 }
1673
1674 static int bnxt_promiscuous_disable_op(struct rte_eth_dev *eth_dev)
1675 {
1676         struct bnxt *bp = eth_dev->data->dev_private;
1677         struct bnxt_vnic_info *vnic;
1678         uint32_t old_flags;
1679         int rc;
1680
1681         rc = is_bnxt_in_error(bp);
1682         if (rc)
1683                 return rc;
1684
1685         /* Filter settings will get applied when port is started */
1686         if (!eth_dev->data->dev_started)
1687                 return 0;
1688
1689         if (bp->vnic_info == NULL)
1690                 return 0;
1691
1692         vnic = BNXT_GET_DEFAULT_VNIC(bp);
1693
1694         old_flags = vnic->flags;
1695         vnic->flags &= ~BNXT_VNIC_INFO_PROMISC;
1696         rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1697         if (rc != 0)
1698                 vnic->flags = old_flags;
1699
1700         return rc;
1701 }
1702
1703 static int bnxt_allmulticast_enable_op(struct rte_eth_dev *eth_dev)
1704 {
1705         struct bnxt *bp = eth_dev->data->dev_private;
1706         struct bnxt_vnic_info *vnic;
1707         uint32_t old_flags;
1708         int rc;
1709
1710         rc = is_bnxt_in_error(bp);
1711         if (rc)
1712                 return rc;
1713
1714         /* Filter settings will get applied when port is started */
1715         if (!eth_dev->data->dev_started)
1716                 return 0;
1717
1718         if (bp->vnic_info == NULL)
1719                 return 0;
1720
1721         vnic = BNXT_GET_DEFAULT_VNIC(bp);
1722
1723         old_flags = vnic->flags;
1724         vnic->flags |= BNXT_VNIC_INFO_ALLMULTI;
1725         rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1726         if (rc != 0)
1727                 vnic->flags = old_flags;
1728
1729         return rc;
1730 }
1731
1732 static int bnxt_allmulticast_disable_op(struct rte_eth_dev *eth_dev)
1733 {
1734         struct bnxt *bp = eth_dev->data->dev_private;
1735         struct bnxt_vnic_info *vnic;
1736         uint32_t old_flags;
1737         int rc;
1738
1739         rc = is_bnxt_in_error(bp);
1740         if (rc)
1741                 return rc;
1742
1743         /* Filter settings will get applied when port is started */
1744         if (!eth_dev->data->dev_started)
1745                 return 0;
1746
1747         if (bp->vnic_info == NULL)
1748                 return 0;
1749
1750         vnic = BNXT_GET_DEFAULT_VNIC(bp);
1751
1752         old_flags = vnic->flags;
1753         vnic->flags &= ~BNXT_VNIC_INFO_ALLMULTI;
1754         rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1755         if (rc != 0)
1756                 vnic->flags = old_flags;
1757
1758         return rc;
1759 }
1760
1761 /* Return bnxt_rx_queue pointer corresponding to a given rxq. */
1762 static struct bnxt_rx_queue *bnxt_qid_to_rxq(struct bnxt *bp, uint16_t qid)
1763 {
1764         if (qid >= bp->rx_nr_rings)
1765                 return NULL;
1766
1767         return bp->eth_dev->data->rx_queues[qid];
1768 }
1769
1770 /* Return rxq corresponding to a given rss table ring/group ID. */
1771 static uint16_t bnxt_rss_to_qid(struct bnxt *bp, uint16_t fwr)
1772 {
1773         struct bnxt_rx_queue *rxq;
1774         unsigned int i;
1775
1776         if (!BNXT_HAS_RING_GRPS(bp)) {
1777                 for (i = 0; i < bp->rx_nr_rings; i++) {
1778                         rxq = bp->eth_dev->data->rx_queues[i];
1779                         if (rxq->rx_ring->rx_ring_struct->fw_ring_id == fwr)
1780                                 return rxq->index;
1781                 }
1782         } else {
1783                 for (i = 0; i < bp->rx_nr_rings; i++) {
1784                         if (bp->grp_info[i].fw_grp_id == fwr)
1785                                 return i;
1786                 }
1787         }
1788
1789         return INVALID_HW_RING_ID;
1790 }
1791
1792 static int bnxt_reta_update_op(struct rte_eth_dev *eth_dev,
1793                             struct rte_eth_rss_reta_entry64 *reta_conf,
1794                             uint16_t reta_size)
1795 {
1796         struct bnxt *bp = eth_dev->data->dev_private;
1797         struct rte_eth_conf *dev_conf = &bp->eth_dev->data->dev_conf;
1798         struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
1799         uint16_t tbl_size = bnxt_rss_hash_tbl_size(bp);
1800         uint16_t idx, sft;
1801         int i, rc;
1802
1803         rc = is_bnxt_in_error(bp);
1804         if (rc)
1805                 return rc;
1806
1807         if (!vnic->rss_table)
1808                 return -EINVAL;
1809
1810         if (!(dev_conf->rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG))
1811                 return -EINVAL;
1812
1813         if (reta_size != tbl_size) {
1814                 PMD_DRV_LOG(ERR, "The configured hash table lookup size "
1815                         "(%d) must equal the size supported by the hardware "
1816                         "(%d)\n", reta_size, tbl_size);
1817                 return -EINVAL;
1818         }
1819
1820         for (i = 0; i < reta_size; i++) {
1821                 struct bnxt_rx_queue *rxq;
1822
1823                 idx = i / RTE_RETA_GROUP_SIZE;
1824                 sft = i % RTE_RETA_GROUP_SIZE;
1825
1826                 if (!(reta_conf[idx].mask & (1ULL << sft)))
1827                         continue;
1828
1829                 rxq = bnxt_qid_to_rxq(bp, reta_conf[idx].reta[sft]);
1830                 if (!rxq) {
1831                         PMD_DRV_LOG(ERR, "Invalid ring in reta_conf.\n");
1832                         return -EINVAL;
1833                 }
1834
1835                 if (BNXT_CHIP_P5(bp)) {
1836                         vnic->rss_table[i * 2] =
1837                                 rxq->rx_ring->rx_ring_struct->fw_ring_id;
1838                         vnic->rss_table[i * 2 + 1] =
1839                                 rxq->cp_ring->cp_ring_struct->fw_ring_id;
1840                 } else {
1841                         vnic->rss_table[i] =
1842                             vnic->fw_grp_ids[reta_conf[idx].reta[sft]];
1843                 }
1844         }
1845
1846         rc = bnxt_hwrm_vnic_rss_cfg(bp, vnic);
1847         return rc;
1848 }
1849
1850 static int bnxt_reta_query_op(struct rte_eth_dev *eth_dev,
1851                               struct rte_eth_rss_reta_entry64 *reta_conf,
1852                               uint16_t reta_size)
1853 {
1854         struct bnxt *bp = eth_dev->data->dev_private;
1855         struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
1856         uint16_t tbl_size = bnxt_rss_hash_tbl_size(bp);
1857         uint16_t idx, sft, i;
1858         int rc;
1859
1860         rc = is_bnxt_in_error(bp);
1861         if (rc)
1862                 return rc;
1863
1864         /* Retrieve from the default VNIC */
1865         if (!vnic)
1866                 return -EINVAL;
1867         if (!vnic->rss_table)
1868                 return -EINVAL;
1869
1870         if (reta_size != tbl_size) {
1871                 PMD_DRV_LOG(ERR, "The configured hash table lookup size "
1872                         "(%d) must equal the size supported by the hardware "
1873                         "(%d)\n", reta_size, tbl_size);
1874                 return -EINVAL;
1875         }
1876
1877         for (idx = 0, i = 0; i < reta_size; i++) {
1878                 idx = i / RTE_RETA_GROUP_SIZE;
1879                 sft = i % RTE_RETA_GROUP_SIZE;
1880
1881                 if (reta_conf[idx].mask & (1ULL << sft)) {
1882                         uint16_t qid;
1883
1884                         if (BNXT_CHIP_P5(bp))
1885                                 qid = bnxt_rss_to_qid(bp,
1886                                                       vnic->rss_table[i * 2]);
1887                         else
1888                                 qid = bnxt_rss_to_qid(bp, vnic->rss_table[i]);
1889
1890                         if (qid == INVALID_HW_RING_ID) {
1891                                 PMD_DRV_LOG(ERR, "Inv. entry in rss table.\n");
1892                                 return -EINVAL;
1893                         }
1894                         reta_conf[idx].reta[sft] = qid;
1895                 }
1896         }
1897
1898         return 0;
1899 }
1900
1901 static int bnxt_rss_hash_update_op(struct rte_eth_dev *eth_dev,
1902                                    struct rte_eth_rss_conf *rss_conf)
1903 {
1904         struct bnxt *bp = eth_dev->data->dev_private;
1905         struct rte_eth_conf *dev_conf = &bp->eth_dev->data->dev_conf;
1906         struct bnxt_vnic_info *vnic;
1907         int rc;
1908
1909         rc = is_bnxt_in_error(bp);
1910         if (rc)
1911                 return rc;
1912
1913         /*
1914          * If RSS enablement were different than dev_configure,
1915          * then return -EINVAL
1916          */
1917         if (dev_conf->rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG) {
1918                 if (!rss_conf->rss_hf)
1919                         PMD_DRV_LOG(ERR, "Hash type NONE\n");
1920         } else {
1921                 if (rss_conf->rss_hf & BNXT_ETH_RSS_SUPPORT)
1922                         return -EINVAL;
1923         }
1924
1925         bp->flags |= BNXT_FLAG_UPDATE_HASH;
1926         memcpy(&eth_dev->data->dev_conf.rx_adv_conf.rss_conf,
1927                rss_conf,
1928                sizeof(*rss_conf));
1929
1930         /* Update the default RSS VNIC(s) */
1931         vnic = BNXT_GET_DEFAULT_VNIC(bp);
1932         vnic->hash_type = bnxt_rte_to_hwrm_hash_types(rss_conf->rss_hf);
1933         vnic->hash_mode =
1934                 bnxt_rte_to_hwrm_hash_level(bp, rss_conf->rss_hf,
1935                                             ETH_RSS_LEVEL(rss_conf->rss_hf));
1936
1937         /*
1938          * If hashkey is not specified, use the previously configured
1939          * hashkey
1940          */
1941         if (!rss_conf->rss_key)
1942                 goto rss_config;
1943
1944         if (rss_conf->rss_key_len != HW_HASH_KEY_SIZE) {
1945                 PMD_DRV_LOG(ERR,
1946                             "Invalid hashkey length, should be 16 bytes\n");
1947                 return -EINVAL;
1948         }
1949         memcpy(vnic->rss_hash_key, rss_conf->rss_key, rss_conf->rss_key_len);
1950
1951 rss_config:
1952         rc = bnxt_hwrm_vnic_rss_cfg(bp, vnic);
1953         return rc;
1954 }
1955
1956 static int bnxt_rss_hash_conf_get_op(struct rte_eth_dev *eth_dev,
1957                                      struct rte_eth_rss_conf *rss_conf)
1958 {
1959         struct bnxt *bp = eth_dev->data->dev_private;
1960         struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
1961         int len, rc;
1962         uint32_t hash_types;
1963
1964         rc = is_bnxt_in_error(bp);
1965         if (rc)
1966                 return rc;
1967
1968         /* RSS configuration is the same for all VNICs */
1969         if (vnic && vnic->rss_hash_key) {
1970                 if (rss_conf->rss_key) {
1971                         len = rss_conf->rss_key_len <= HW_HASH_KEY_SIZE ?
1972                               rss_conf->rss_key_len : HW_HASH_KEY_SIZE;
1973                         memcpy(rss_conf->rss_key, vnic->rss_hash_key, len);
1974                 }
1975
1976                 hash_types = vnic->hash_type;
1977                 rss_conf->rss_hf = 0;
1978                 if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV4) {
1979                         rss_conf->rss_hf |= ETH_RSS_IPV4;
1980                         hash_types &= ~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV4;
1981                 }
1982                 if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV4) {
1983                         rss_conf->rss_hf |= ETH_RSS_NONFRAG_IPV4_TCP;
1984                         hash_types &=
1985                                 ~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV4;
1986                 }
1987                 if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV4) {
1988                         rss_conf->rss_hf |= ETH_RSS_NONFRAG_IPV4_UDP;
1989                         hash_types &=
1990                                 ~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV4;
1991                 }
1992                 if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV6) {
1993                         rss_conf->rss_hf |= ETH_RSS_IPV6;
1994                         hash_types &= ~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV6;
1995                 }
1996                 if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV6) {
1997                         rss_conf->rss_hf |= ETH_RSS_NONFRAG_IPV6_TCP;
1998                         hash_types &=
1999                                 ~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV6;
2000                 }
2001                 if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV6) {
2002                         rss_conf->rss_hf |= ETH_RSS_NONFRAG_IPV6_UDP;
2003                         hash_types &=
2004                                 ~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV6;
2005                 }
2006
2007                 rss_conf->rss_hf |=
2008                         bnxt_hwrm_to_rte_rss_level(bp, vnic->hash_mode);
2009
2010                 if (hash_types) {
2011                         PMD_DRV_LOG(ERR,
2012                                 "Unknown RSS config from firmware (%08x), RSS disabled",
2013                                 vnic->hash_type);
2014                         return -ENOTSUP;
2015                 }
2016         } else {
2017                 rss_conf->rss_hf = 0;
2018         }
2019         return 0;
2020 }
2021
2022 static int bnxt_flow_ctrl_get_op(struct rte_eth_dev *dev,
2023                                struct rte_eth_fc_conf *fc_conf)
2024 {
2025         struct bnxt *bp = dev->data->dev_private;
2026         struct rte_eth_link link_info;
2027         int rc;
2028
2029         rc = is_bnxt_in_error(bp);
2030         if (rc)
2031                 return rc;
2032
2033         rc = bnxt_get_hwrm_link_config(bp, &link_info);
2034         if (rc)
2035                 return rc;
2036
2037         memset(fc_conf, 0, sizeof(*fc_conf));
2038         if (bp->link_info->auto_pause)
2039                 fc_conf->autoneg = 1;
2040         switch (bp->link_info->pause) {
2041         case 0:
2042                 fc_conf->mode = RTE_FC_NONE;
2043                 break;
2044         case HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_TX:
2045                 fc_conf->mode = RTE_FC_TX_PAUSE;
2046                 break;
2047         case HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_RX:
2048                 fc_conf->mode = RTE_FC_RX_PAUSE;
2049                 break;
2050         case (HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_TX |
2051                         HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_RX):
2052                 fc_conf->mode = RTE_FC_FULL;
2053                 break;
2054         }
2055         return 0;
2056 }
2057
2058 static int bnxt_flow_ctrl_set_op(struct rte_eth_dev *dev,
2059                                struct rte_eth_fc_conf *fc_conf)
2060 {
2061         struct bnxt *bp = dev->data->dev_private;
2062         int rc;
2063
2064         rc = is_bnxt_in_error(bp);
2065         if (rc)
2066                 return rc;
2067
2068         if (!BNXT_SINGLE_PF(bp) || BNXT_VF(bp)) {
2069                 PMD_DRV_LOG(ERR, "Flow Control Settings cannot be modified\n");
2070                 return -ENOTSUP;
2071         }
2072
2073         switch (fc_conf->mode) {
2074         case RTE_FC_NONE:
2075                 bp->link_info->auto_pause = 0;
2076                 bp->link_info->force_pause = 0;
2077                 break;
2078         case RTE_FC_RX_PAUSE:
2079                 if (fc_conf->autoneg) {
2080                         bp->link_info->auto_pause =
2081                                         HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_RX;
2082                         bp->link_info->force_pause = 0;
2083                 } else {
2084                         bp->link_info->auto_pause = 0;
2085                         bp->link_info->force_pause =
2086                                         HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_RX;
2087                 }
2088                 break;
2089         case RTE_FC_TX_PAUSE:
2090                 if (fc_conf->autoneg) {
2091                         bp->link_info->auto_pause =
2092                                         HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_TX;
2093                         bp->link_info->force_pause = 0;
2094                 } else {
2095                         bp->link_info->auto_pause = 0;
2096                         bp->link_info->force_pause =
2097                                         HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_TX;
2098                 }
2099                 break;
2100         case RTE_FC_FULL:
2101                 if (fc_conf->autoneg) {
2102                         bp->link_info->auto_pause =
2103                                         HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_TX |
2104                                         HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_RX;
2105                         bp->link_info->force_pause = 0;
2106                 } else {
2107                         bp->link_info->auto_pause = 0;
2108                         bp->link_info->force_pause =
2109                                         HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_TX |
2110                                         HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_RX;
2111                 }
2112                 break;
2113         }
2114         return bnxt_set_hwrm_link_config(bp, true);
2115 }
2116
2117 /* Add UDP tunneling port */
2118 static int
2119 bnxt_udp_tunnel_port_add_op(struct rte_eth_dev *eth_dev,
2120                          struct rte_eth_udp_tunnel *udp_tunnel)
2121 {
2122         struct bnxt *bp = eth_dev->data->dev_private;
2123         uint16_t tunnel_type = 0;
2124         int rc = 0;
2125
2126         rc = is_bnxt_in_error(bp);
2127         if (rc)
2128                 return rc;
2129
2130         switch (udp_tunnel->prot_type) {
2131         case RTE_TUNNEL_TYPE_VXLAN:
2132                 if (bp->vxlan_port_cnt) {
2133                         PMD_DRV_LOG(ERR, "Tunnel Port %d already programmed\n",
2134                                 udp_tunnel->udp_port);
2135                         if (bp->vxlan_port != udp_tunnel->udp_port) {
2136                                 PMD_DRV_LOG(ERR, "Only one port allowed\n");
2137                                 return -ENOSPC;
2138                         }
2139                         bp->vxlan_port_cnt++;
2140                         return 0;
2141                 }
2142                 tunnel_type =
2143                         HWRM_TUNNEL_DST_PORT_ALLOC_INPUT_TUNNEL_TYPE_VXLAN;
2144                 bp->vxlan_port_cnt++;
2145                 break;
2146         case RTE_TUNNEL_TYPE_GENEVE:
2147                 if (bp->geneve_port_cnt) {
2148                         PMD_DRV_LOG(ERR, "Tunnel Port %d already programmed\n",
2149                                 udp_tunnel->udp_port);
2150                         if (bp->geneve_port != udp_tunnel->udp_port) {
2151                                 PMD_DRV_LOG(ERR, "Only one port allowed\n");
2152                                 return -ENOSPC;
2153                         }
2154                         bp->geneve_port_cnt++;
2155                         return 0;
2156                 }
2157                 tunnel_type =
2158                         HWRM_TUNNEL_DST_PORT_ALLOC_INPUT_TUNNEL_TYPE_GENEVE;
2159                 bp->geneve_port_cnt++;
2160                 break;
2161         default:
2162                 PMD_DRV_LOG(ERR, "Tunnel type is not supported\n");
2163                 return -ENOTSUP;
2164         }
2165         rc = bnxt_hwrm_tunnel_dst_port_alloc(bp, udp_tunnel->udp_port,
2166                                              tunnel_type);
2167         return rc;
2168 }
2169
2170 static int
2171 bnxt_udp_tunnel_port_del_op(struct rte_eth_dev *eth_dev,
2172                          struct rte_eth_udp_tunnel *udp_tunnel)
2173 {
2174         struct bnxt *bp = eth_dev->data->dev_private;
2175         uint16_t tunnel_type = 0;
2176         uint16_t port = 0;
2177         int rc = 0;
2178
2179         rc = is_bnxt_in_error(bp);
2180         if (rc)
2181                 return rc;
2182
2183         switch (udp_tunnel->prot_type) {
2184         case RTE_TUNNEL_TYPE_VXLAN:
2185                 if (!bp->vxlan_port_cnt) {
2186                         PMD_DRV_LOG(ERR, "No Tunnel port configured yet\n");
2187                         return -EINVAL;
2188                 }
2189                 if (bp->vxlan_port != udp_tunnel->udp_port) {
2190                         PMD_DRV_LOG(ERR, "Req Port: %d. Configured port: %d\n",
2191                                 udp_tunnel->udp_port, bp->vxlan_port);
2192                         return -EINVAL;
2193                 }
2194                 if (--bp->vxlan_port_cnt)
2195                         return 0;
2196
2197                 tunnel_type =
2198                         HWRM_TUNNEL_DST_PORT_FREE_INPUT_TUNNEL_TYPE_VXLAN;
2199                 port = bp->vxlan_fw_dst_port_id;
2200                 break;
2201         case RTE_TUNNEL_TYPE_GENEVE:
2202                 if (!bp->geneve_port_cnt) {
2203                         PMD_DRV_LOG(ERR, "No Tunnel port configured yet\n");
2204                         return -EINVAL;
2205                 }
2206                 if (bp->geneve_port != udp_tunnel->udp_port) {
2207                         PMD_DRV_LOG(ERR, "Req Port: %d. Configured port: %d\n",
2208                                 udp_tunnel->udp_port, bp->geneve_port);
2209                         return -EINVAL;
2210                 }
2211                 if (--bp->geneve_port_cnt)
2212                         return 0;
2213
2214                 tunnel_type =
2215                         HWRM_TUNNEL_DST_PORT_FREE_INPUT_TUNNEL_TYPE_GENEVE;
2216                 port = bp->geneve_fw_dst_port_id;
2217                 break;
2218         default:
2219                 PMD_DRV_LOG(ERR, "Tunnel type is not supported\n");
2220                 return -ENOTSUP;
2221         }
2222
2223         rc = bnxt_hwrm_tunnel_dst_port_free(bp, port, tunnel_type);
2224         return rc;
2225 }
2226
2227 static int bnxt_del_vlan_filter(struct bnxt *bp, uint16_t vlan_id)
2228 {
2229         struct bnxt_filter_info *filter;
2230         struct bnxt_vnic_info *vnic;
2231         int rc = 0;
2232         uint32_t chk = HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN;
2233
2234         vnic = BNXT_GET_DEFAULT_VNIC(bp);
2235         filter = STAILQ_FIRST(&vnic->filter);
2236         while (filter) {
2237                 /* Search for this matching MAC+VLAN filter */
2238                 if (bnxt_vlan_filter_exists(bp, filter, chk, vlan_id)) {
2239                         /* Delete the filter */
2240                         rc = bnxt_hwrm_clear_l2_filter(bp, filter);
2241                         if (rc)
2242                                 return rc;
2243                         STAILQ_REMOVE(&vnic->filter, filter,
2244                                       bnxt_filter_info, next);
2245                         bnxt_free_filter(bp, filter);
2246                         PMD_DRV_LOG(INFO,
2247                                     "Deleted vlan filter for %d\n",
2248                                     vlan_id);
2249                         return 0;
2250                 }
2251                 filter = STAILQ_NEXT(filter, next);
2252         }
2253         return -ENOENT;
2254 }
2255
2256 static int bnxt_add_vlan_filter(struct bnxt *bp, uint16_t vlan_id)
2257 {
2258         struct bnxt_filter_info *filter;
2259         struct bnxt_vnic_info *vnic;
2260         int rc = 0;
2261         uint32_t en = HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN |
2262                 HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN_MASK;
2263         uint32_t chk = HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN;
2264
2265         /* Implementation notes on the use of VNIC in this command:
2266          *
2267          * By default, these filters belong to default vnic for the function.
2268          * Once these filters are set up, only destination VNIC can be modified.
2269          * If the destination VNIC is not specified in this command,
2270          * then the HWRM shall only create an l2 context id.
2271          */
2272
2273         vnic = BNXT_GET_DEFAULT_VNIC(bp);
2274         filter = STAILQ_FIRST(&vnic->filter);
2275         /* Check if the VLAN has already been added */
2276         while (filter) {
2277                 if (bnxt_vlan_filter_exists(bp, filter, chk, vlan_id))
2278                         return -EEXIST;
2279
2280                 filter = STAILQ_NEXT(filter, next);
2281         }
2282
2283         /* No match found. Alloc a fresh filter and issue the L2_FILTER_ALLOC
2284          * command to create MAC+VLAN filter with the right flags, enables set.
2285          */
2286         filter = bnxt_alloc_filter(bp);
2287         if (!filter) {
2288                 PMD_DRV_LOG(ERR,
2289                             "MAC/VLAN filter alloc failed\n");
2290                 return -ENOMEM;
2291         }
2292         /* MAC + VLAN ID filter */
2293         /* If l2_ivlan == 0 and l2_ivlan_mask != 0, only
2294          * untagged packets are received
2295          *
2296          * If l2_ivlan != 0 and l2_ivlan_mask != 0, untagged
2297          * packets and only the programmed vlan's packets are received
2298          */
2299         filter->l2_ivlan = vlan_id;
2300         filter->l2_ivlan_mask = 0x0FFF;
2301         filter->enables |= en;
2302         filter->flags |= HWRM_CFA_L2_FILTER_ALLOC_INPUT_FLAGS_OUTERMOST;
2303
2304         rc = bnxt_hwrm_set_l2_filter(bp, vnic->fw_vnic_id, filter);
2305         if (rc) {
2306                 /* Free the newly allocated filter as we were
2307                  * not able to create the filter in hardware.
2308                  */
2309                 bnxt_free_filter(bp, filter);
2310                 return rc;
2311         }
2312
2313         filter->mac_index = 0;
2314         /* Add this new filter to the list */
2315         if (vlan_id == 0)
2316                 STAILQ_INSERT_HEAD(&vnic->filter, filter, next);
2317         else
2318                 STAILQ_INSERT_TAIL(&vnic->filter, filter, next);
2319
2320         PMD_DRV_LOG(INFO,
2321                     "Added Vlan filter for %d\n", vlan_id);
2322         return rc;
2323 }
2324
2325 static int bnxt_vlan_filter_set_op(struct rte_eth_dev *eth_dev,
2326                 uint16_t vlan_id, int on)
2327 {
2328         struct bnxt *bp = eth_dev->data->dev_private;
2329         int rc;
2330
2331         rc = is_bnxt_in_error(bp);
2332         if (rc)
2333                 return rc;
2334
2335         if (!eth_dev->data->dev_started) {
2336                 PMD_DRV_LOG(ERR, "port must be started before setting vlan\n");
2337                 return -EINVAL;
2338         }
2339
2340         /* These operations apply to ALL existing MAC/VLAN filters */
2341         if (on)
2342                 return bnxt_add_vlan_filter(bp, vlan_id);
2343         else
2344                 return bnxt_del_vlan_filter(bp, vlan_id);
2345 }
2346
2347 static int bnxt_del_dflt_mac_filter(struct bnxt *bp,
2348                                     struct bnxt_vnic_info *vnic)
2349 {
2350         struct bnxt_filter_info *filter;
2351         int rc;
2352
2353         filter = STAILQ_FIRST(&vnic->filter);
2354         while (filter) {
2355                 if (filter->mac_index == 0 &&
2356                     !memcmp(filter->l2_addr, bp->mac_addr,
2357                             RTE_ETHER_ADDR_LEN)) {
2358                         rc = bnxt_hwrm_clear_l2_filter(bp, filter);
2359                         if (!rc) {
2360                                 STAILQ_REMOVE(&vnic->filter, filter,
2361                                               bnxt_filter_info, next);
2362                                 bnxt_free_filter(bp, filter);
2363                         }
2364                         return rc;
2365                 }
2366                 filter = STAILQ_NEXT(filter, next);
2367         }
2368         return 0;
2369 }
2370
2371 static int
2372 bnxt_config_vlan_hw_filter(struct bnxt *bp, uint64_t rx_offloads)
2373 {
2374         struct bnxt_vnic_info *vnic;
2375         unsigned int i;
2376         int rc;
2377
2378         vnic = BNXT_GET_DEFAULT_VNIC(bp);
2379         if (!(rx_offloads & DEV_RX_OFFLOAD_VLAN_FILTER)) {
2380                 /* Remove any VLAN filters programmed */
2381                 for (i = 0; i < RTE_ETHER_MAX_VLAN_ID; i++)
2382                         bnxt_del_vlan_filter(bp, i);
2383
2384                 rc = bnxt_add_mac_filter(bp, vnic, NULL, 0, 0);
2385                 if (rc)
2386                         return rc;
2387         } else {
2388                 /* Default filter will allow packets that match the
2389                  * dest mac. So, it has to be deleted, otherwise, we
2390                  * will endup receiving vlan packets for which the
2391                  * filter is not programmed, when hw-vlan-filter
2392                  * configuration is ON
2393                  */
2394                 bnxt_del_dflt_mac_filter(bp, vnic);
2395                 /* This filter will allow only untagged packets */
2396                 bnxt_add_vlan_filter(bp, 0);
2397         }
2398         PMD_DRV_LOG(DEBUG, "VLAN Filtering: %d\n",
2399                     !!(rx_offloads & DEV_RX_OFFLOAD_VLAN_FILTER));
2400
2401         return 0;
2402 }
2403
2404 static int bnxt_free_one_vnic(struct bnxt *bp, uint16_t vnic_id)
2405 {
2406         struct bnxt_vnic_info *vnic = &bp->vnic_info[vnic_id];
2407         unsigned int i;
2408         int rc;
2409
2410         /* Destroy vnic filters and vnic */
2411         if (bp->eth_dev->data->dev_conf.rxmode.offloads &
2412             DEV_RX_OFFLOAD_VLAN_FILTER) {
2413                 for (i = 0; i < RTE_ETHER_MAX_VLAN_ID; i++)
2414                         bnxt_del_vlan_filter(bp, i);
2415         }
2416         bnxt_del_dflt_mac_filter(bp, vnic);
2417
2418         rc = bnxt_hwrm_vnic_free(bp, vnic);
2419         if (rc)
2420                 return rc;
2421
2422         rte_free(vnic->fw_grp_ids);
2423         vnic->fw_grp_ids = NULL;
2424
2425         vnic->rx_queue_cnt = 0;
2426
2427         return 0;
2428 }
2429
2430 static int
2431 bnxt_config_vlan_hw_stripping(struct bnxt *bp, uint64_t rx_offloads)
2432 {
2433         struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
2434         int rc;
2435
2436         /* Destroy, recreate and reconfigure the default vnic */
2437         rc = bnxt_free_one_vnic(bp, 0);
2438         if (rc)
2439                 return rc;
2440
2441         /* default vnic 0 */
2442         rc = bnxt_setup_one_vnic(bp, 0);
2443         if (rc)
2444                 return rc;
2445
2446         if (bp->eth_dev->data->dev_conf.rxmode.offloads &
2447             DEV_RX_OFFLOAD_VLAN_FILTER) {
2448                 rc = bnxt_add_vlan_filter(bp, 0);
2449                 if (rc)
2450                         return rc;
2451                 rc = bnxt_restore_vlan_filters(bp);
2452                 if (rc)
2453                         return rc;
2454         } else {
2455                 rc = bnxt_add_mac_filter(bp, vnic, NULL, 0, 0);
2456                 if (rc)
2457                         return rc;
2458         }
2459
2460         rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
2461         if (rc)
2462                 return rc;
2463
2464         PMD_DRV_LOG(DEBUG, "VLAN Strip Offload: %d\n",
2465                     !!(rx_offloads & DEV_RX_OFFLOAD_VLAN_STRIP));
2466
2467         return rc;
2468 }
2469
2470 static int
2471 bnxt_vlan_offload_set_op(struct rte_eth_dev *dev, int mask)
2472 {
2473         uint64_t rx_offloads = dev->data->dev_conf.rxmode.offloads;
2474         struct bnxt *bp = dev->data->dev_private;
2475         int rc;
2476
2477         rc = is_bnxt_in_error(bp);
2478         if (rc)
2479                 return rc;
2480
2481         /* Filter settings will get applied when port is started */
2482         if (!dev->data->dev_started)
2483                 return 0;
2484
2485         if (mask & ETH_VLAN_FILTER_MASK) {
2486                 /* Enable or disable VLAN filtering */
2487                 rc = bnxt_config_vlan_hw_filter(bp, rx_offloads);
2488                 if (rc)
2489                         return rc;
2490         }
2491
2492         if (mask & ETH_VLAN_STRIP_MASK) {
2493                 /* Enable or disable VLAN stripping */
2494                 rc = bnxt_config_vlan_hw_stripping(bp, rx_offloads);
2495                 if (rc)
2496                         return rc;
2497         }
2498
2499         if (mask & ETH_VLAN_EXTEND_MASK) {
2500                 if (rx_offloads & DEV_RX_OFFLOAD_VLAN_EXTEND)
2501                         PMD_DRV_LOG(DEBUG, "Extend VLAN supported\n");
2502                 else
2503                         PMD_DRV_LOG(INFO, "Extend VLAN unsupported\n");
2504         }
2505
2506         return 0;
2507 }
2508
2509 static int
2510 bnxt_vlan_tpid_set_op(struct rte_eth_dev *dev, enum rte_vlan_type vlan_type,
2511                       uint16_t tpid)
2512 {
2513         struct bnxt *bp = dev->data->dev_private;
2514         int qinq = dev->data->dev_conf.rxmode.offloads &
2515                    DEV_RX_OFFLOAD_VLAN_EXTEND;
2516
2517         if (vlan_type != ETH_VLAN_TYPE_INNER &&
2518             vlan_type != ETH_VLAN_TYPE_OUTER) {
2519                 PMD_DRV_LOG(ERR,
2520                             "Unsupported vlan type.");
2521                 return -EINVAL;
2522         }
2523         if (!qinq) {
2524                 PMD_DRV_LOG(ERR,
2525                             "QinQ not enabled. Needs to be ON as we can "
2526                             "accelerate only outer vlan\n");
2527                 return -EINVAL;
2528         }
2529
2530         if (vlan_type == ETH_VLAN_TYPE_OUTER) {
2531                 switch (tpid) {
2532                 case RTE_ETHER_TYPE_QINQ:
2533                         bp->outer_tpid_bd =
2534                                 TX_BD_LONG_CFA_META_VLAN_TPID_TPID88A8;
2535                                 break;
2536                 case RTE_ETHER_TYPE_VLAN:
2537                         bp->outer_tpid_bd =
2538                                 TX_BD_LONG_CFA_META_VLAN_TPID_TPID8100;
2539                                 break;
2540                 case RTE_ETHER_TYPE_QINQ1:
2541                         bp->outer_tpid_bd =
2542                                 TX_BD_LONG_CFA_META_VLAN_TPID_TPID9100;
2543                                 break;
2544                 case RTE_ETHER_TYPE_QINQ2:
2545                         bp->outer_tpid_bd =
2546                                 TX_BD_LONG_CFA_META_VLAN_TPID_TPID9200;
2547                                 break;
2548                 case RTE_ETHER_TYPE_QINQ3:
2549                         bp->outer_tpid_bd =
2550                                  TX_BD_LONG_CFA_META_VLAN_TPID_TPID9300;
2551                                 break;
2552                 default:
2553                         PMD_DRV_LOG(ERR, "Invalid TPID: %x\n", tpid);
2554                         return -EINVAL;
2555                 }
2556                 bp->outer_tpid_bd |= tpid;
2557                 PMD_DRV_LOG(INFO, "outer_tpid_bd = %x\n", bp->outer_tpid_bd);
2558         } else if (vlan_type == ETH_VLAN_TYPE_INNER) {
2559                 PMD_DRV_LOG(ERR,
2560                             "Can accelerate only outer vlan in QinQ\n");
2561                 return -EINVAL;
2562         }
2563
2564         return 0;
2565 }
2566
2567 static int
2568 bnxt_set_default_mac_addr_op(struct rte_eth_dev *dev,
2569                              struct rte_ether_addr *addr)
2570 {
2571         struct bnxt *bp = dev->data->dev_private;
2572         /* Default Filter is tied to VNIC 0 */
2573         struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
2574         int rc;
2575
2576         rc = is_bnxt_in_error(bp);
2577         if (rc)
2578                 return rc;
2579
2580         if (BNXT_VF(bp) && !BNXT_VF_IS_TRUSTED(bp))
2581                 return -EPERM;
2582
2583         if (rte_is_zero_ether_addr(addr))
2584                 return -EINVAL;
2585
2586         /* Filter settings will get applied when port is started */
2587         if (!dev->data->dev_started)
2588                 return 0;
2589
2590         /* Check if the requested MAC is already added */
2591         if (memcmp(addr, bp->mac_addr, RTE_ETHER_ADDR_LEN) == 0)
2592                 return 0;
2593
2594         /* Destroy filter and re-create it */
2595         bnxt_del_dflt_mac_filter(bp, vnic);
2596
2597         memcpy(bp->mac_addr, addr, RTE_ETHER_ADDR_LEN);
2598         if (dev->data->dev_conf.rxmode.offloads & DEV_RX_OFFLOAD_VLAN_FILTER) {
2599                 /* This filter will allow only untagged packets */
2600                 rc = bnxt_add_vlan_filter(bp, 0);
2601         } else {
2602                 rc = bnxt_add_mac_filter(bp, vnic, addr, 0, 0);
2603         }
2604
2605         PMD_DRV_LOG(DEBUG, "Set MAC addr\n");
2606         return rc;
2607 }
2608
2609 static int
2610 bnxt_dev_set_mc_addr_list_op(struct rte_eth_dev *eth_dev,
2611                           struct rte_ether_addr *mc_addr_set,
2612                           uint32_t nb_mc_addr)
2613 {
2614         struct bnxt *bp = eth_dev->data->dev_private;
2615         char *mc_addr_list = (char *)mc_addr_set;
2616         struct bnxt_vnic_info *vnic;
2617         uint32_t off = 0, i = 0;
2618         int rc;
2619
2620         rc = is_bnxt_in_error(bp);
2621         if (rc)
2622                 return rc;
2623
2624         vnic = BNXT_GET_DEFAULT_VNIC(bp);
2625
2626         if (nb_mc_addr > BNXT_MAX_MC_ADDRS) {
2627                 vnic->flags |= BNXT_VNIC_INFO_ALLMULTI;
2628                 goto allmulti;
2629         }
2630
2631         /* TODO Check for Duplicate mcast addresses */
2632         vnic->flags &= ~BNXT_VNIC_INFO_ALLMULTI;
2633         for (i = 0; i < nb_mc_addr; i++) {
2634                 memcpy(vnic->mc_list + off, &mc_addr_list[i],
2635                         RTE_ETHER_ADDR_LEN);
2636                 off += RTE_ETHER_ADDR_LEN;
2637         }
2638
2639         vnic->mc_addr_cnt = i;
2640         if (vnic->mc_addr_cnt)
2641                 vnic->flags |= BNXT_VNIC_INFO_MCAST;
2642         else
2643                 vnic->flags &= ~BNXT_VNIC_INFO_MCAST;
2644
2645 allmulti:
2646         return bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
2647 }
2648
2649 static int
2650 bnxt_fw_version_get(struct rte_eth_dev *dev, char *fw_version, size_t fw_size)
2651 {
2652         struct bnxt *bp = dev->data->dev_private;
2653         uint8_t fw_major = (bp->fw_ver >> 24) & 0xff;
2654         uint8_t fw_minor = (bp->fw_ver >> 16) & 0xff;
2655         uint8_t fw_updt = (bp->fw_ver >> 8) & 0xff;
2656         uint8_t fw_rsvd = bp->fw_ver & 0xff;
2657         int ret;
2658
2659         ret = snprintf(fw_version, fw_size, "%d.%d.%d.%d",
2660                         fw_major, fw_minor, fw_updt, fw_rsvd);
2661
2662         ret += 1; /* add the size of '\0' */
2663         if (fw_size < (uint32_t)ret)
2664                 return ret;
2665         else
2666                 return 0;
2667 }
2668
2669 static void
2670 bnxt_rxq_info_get_op(struct rte_eth_dev *dev, uint16_t queue_id,
2671         struct rte_eth_rxq_info *qinfo)
2672 {
2673         struct bnxt *bp = dev->data->dev_private;
2674         struct bnxt_rx_queue *rxq;
2675
2676         if (is_bnxt_in_error(bp))
2677                 return;
2678
2679         rxq = dev->data->rx_queues[queue_id];
2680
2681         qinfo->mp = rxq->mb_pool;
2682         qinfo->scattered_rx = dev->data->scattered_rx;
2683         qinfo->nb_desc = rxq->nb_rx_desc;
2684
2685         qinfo->conf.rx_free_thresh = rxq->rx_free_thresh;
2686         qinfo->conf.rx_drop_en = rxq->drop_en;
2687         qinfo->conf.rx_deferred_start = rxq->rx_deferred_start;
2688         qinfo->conf.offloads = dev->data->dev_conf.rxmode.offloads;
2689 }
2690
2691 static void
2692 bnxt_txq_info_get_op(struct rte_eth_dev *dev, uint16_t queue_id,
2693         struct rte_eth_txq_info *qinfo)
2694 {
2695         struct bnxt *bp = dev->data->dev_private;
2696         struct bnxt_tx_queue *txq;
2697
2698         if (is_bnxt_in_error(bp))
2699                 return;
2700
2701         txq = dev->data->tx_queues[queue_id];
2702
2703         qinfo->nb_desc = txq->nb_tx_desc;
2704
2705         qinfo->conf.tx_thresh.pthresh = txq->pthresh;
2706         qinfo->conf.tx_thresh.hthresh = txq->hthresh;
2707         qinfo->conf.tx_thresh.wthresh = txq->wthresh;
2708
2709         qinfo->conf.tx_free_thresh = txq->tx_free_thresh;
2710         qinfo->conf.tx_rs_thresh = 0;
2711         qinfo->conf.tx_deferred_start = txq->tx_deferred_start;
2712         qinfo->conf.offloads = txq->offloads;
2713 }
2714
2715 static const struct {
2716         eth_rx_burst_t pkt_burst;
2717         const char *info;
2718 } bnxt_rx_burst_info[] = {
2719         {bnxt_recv_pkts,        "Scalar"},
2720 #if defined(RTE_ARCH_X86)
2721         {bnxt_recv_pkts_vec,    "Vector SSE"},
2722 #elif defined(RTE_ARCH_ARM64)
2723         {bnxt_recv_pkts_vec,    "Vector Neon"},
2724 #endif
2725 };
2726
2727 static int
2728 bnxt_rx_burst_mode_get(struct rte_eth_dev *dev, __rte_unused uint16_t queue_id,
2729                        struct rte_eth_burst_mode *mode)
2730 {
2731         eth_rx_burst_t pkt_burst = dev->rx_pkt_burst;
2732         size_t i;
2733
2734         for (i = 0; i < RTE_DIM(bnxt_rx_burst_info); i++) {
2735                 if (pkt_burst == bnxt_rx_burst_info[i].pkt_burst) {
2736                         snprintf(mode->info, sizeof(mode->info), "%s",
2737                                  bnxt_rx_burst_info[i].info);
2738                         return 0;
2739                 }
2740         }
2741
2742         return -EINVAL;
2743 }
2744
2745 static const struct {
2746         eth_tx_burst_t pkt_burst;
2747         const char *info;
2748 } bnxt_tx_burst_info[] = {
2749         {bnxt_xmit_pkts,        "Scalar"},
2750 #if defined(RTE_ARCH_X86)
2751         {bnxt_xmit_pkts_vec,    "Vector SSE"},
2752 #elif defined(RTE_ARCH_ARM64)
2753         {bnxt_xmit_pkts_vec,    "Vector Neon"},
2754 #endif
2755 };
2756
2757 static int
2758 bnxt_tx_burst_mode_get(struct rte_eth_dev *dev, __rte_unused uint16_t queue_id,
2759                        struct rte_eth_burst_mode *mode)
2760 {
2761         eth_tx_burst_t pkt_burst = dev->tx_pkt_burst;
2762         size_t i;
2763
2764         for (i = 0; i < RTE_DIM(bnxt_tx_burst_info); i++) {
2765                 if (pkt_burst == bnxt_tx_burst_info[i].pkt_burst) {
2766                         snprintf(mode->info, sizeof(mode->info), "%s",
2767                                  bnxt_tx_burst_info[i].info);
2768                         return 0;
2769                 }
2770         }
2771
2772         return -EINVAL;
2773 }
2774
2775 int bnxt_mtu_set_op(struct rte_eth_dev *eth_dev, uint16_t new_mtu)
2776 {
2777         struct bnxt *bp = eth_dev->data->dev_private;
2778         uint32_t new_pkt_size;
2779         uint32_t rc = 0;
2780         uint32_t i;
2781
2782         rc = is_bnxt_in_error(bp);
2783         if (rc)
2784                 return rc;
2785
2786         /* Exit if receive queues are not configured yet */
2787         if (!eth_dev->data->nb_rx_queues)
2788                 return rc;
2789
2790         new_pkt_size = new_mtu + RTE_ETHER_HDR_LEN + RTE_ETHER_CRC_LEN +
2791                        VLAN_TAG_SIZE * BNXT_NUM_VLANS;
2792
2793         /*
2794          * Disallow any MTU change that would require scattered receive support
2795          * if it is not already enabled.
2796          */
2797         if (eth_dev->data->dev_started &&
2798             !eth_dev->data->scattered_rx &&
2799             (new_pkt_size >
2800              eth_dev->data->min_rx_buf_size - RTE_PKTMBUF_HEADROOM)) {
2801                 PMD_DRV_LOG(ERR,
2802                             "MTU change would require scattered rx support. ");
2803                 PMD_DRV_LOG(ERR, "Stop port before changing MTU.\n");
2804                 return -EINVAL;
2805         }
2806
2807         if (new_mtu > RTE_ETHER_MTU) {
2808                 bp->flags |= BNXT_FLAG_JUMBO;
2809                 bp->eth_dev->data->dev_conf.rxmode.offloads |=
2810                         DEV_RX_OFFLOAD_JUMBO_FRAME;
2811         } else {
2812                 bp->eth_dev->data->dev_conf.rxmode.offloads &=
2813                         ~DEV_RX_OFFLOAD_JUMBO_FRAME;
2814                 bp->flags &= ~BNXT_FLAG_JUMBO;
2815         }
2816
2817         /* Is there a change in mtu setting? */
2818         if (eth_dev->data->dev_conf.rxmode.max_rx_pkt_len == new_pkt_size)
2819                 return rc;
2820
2821         for (i = 0; i < bp->nr_vnics; i++) {
2822                 struct bnxt_vnic_info *vnic = &bp->vnic_info[i];
2823                 uint16_t size = 0;
2824
2825                 vnic->mru = BNXT_VNIC_MRU(new_mtu);
2826                 rc = bnxt_hwrm_vnic_cfg(bp, vnic);
2827                 if (rc)
2828                         break;
2829
2830                 size = rte_pktmbuf_data_room_size(bp->rx_queues[0]->mb_pool);
2831                 size -= RTE_PKTMBUF_HEADROOM;
2832
2833                 if (size < new_mtu) {
2834                         rc = bnxt_hwrm_vnic_plcmode_cfg(bp, vnic);
2835                         if (rc)
2836                                 return rc;
2837                 }
2838         }
2839
2840         if (!rc)
2841                 eth_dev->data->dev_conf.rxmode.max_rx_pkt_len = new_pkt_size;
2842
2843         PMD_DRV_LOG(INFO, "New MTU is %d\n", new_mtu);
2844
2845         return rc;
2846 }
2847
2848 static int
2849 bnxt_vlan_pvid_set_op(struct rte_eth_dev *dev, uint16_t pvid, int on)
2850 {
2851         struct bnxt *bp = dev->data->dev_private;
2852         uint16_t vlan = bp->vlan;
2853         int rc;
2854
2855         rc = is_bnxt_in_error(bp);
2856         if (rc)
2857                 return rc;
2858
2859         if (!BNXT_SINGLE_PF(bp) || BNXT_VF(bp)) {
2860                 PMD_DRV_LOG(ERR,
2861                         "PVID cannot be modified for this function\n");
2862                 return -ENOTSUP;
2863         }
2864         bp->vlan = on ? pvid : 0;
2865
2866         rc = bnxt_hwrm_set_default_vlan(bp, 0, 0);
2867         if (rc)
2868                 bp->vlan = vlan;
2869         return rc;
2870 }
2871
2872 static int
2873 bnxt_dev_led_on_op(struct rte_eth_dev *dev)
2874 {
2875         struct bnxt *bp = dev->data->dev_private;
2876         int rc;
2877
2878         rc = is_bnxt_in_error(bp);
2879         if (rc)
2880                 return rc;
2881
2882         return bnxt_hwrm_port_led_cfg(bp, true);
2883 }
2884
2885 static int
2886 bnxt_dev_led_off_op(struct rte_eth_dev *dev)
2887 {
2888         struct bnxt *bp = dev->data->dev_private;
2889         int rc;
2890
2891         rc = is_bnxt_in_error(bp);
2892         if (rc)
2893                 return rc;
2894
2895         return bnxt_hwrm_port_led_cfg(bp, false);
2896 }
2897
2898 static uint32_t
2899 bnxt_rx_queue_count_op(struct rte_eth_dev *dev, uint16_t rx_queue_id)
2900 {
2901         struct bnxt *bp = (struct bnxt *)dev->data->dev_private;
2902         uint32_t desc = 0, raw_cons = 0, cons;
2903         struct bnxt_cp_ring_info *cpr;
2904         struct bnxt_rx_queue *rxq;
2905         struct rx_pkt_cmpl *rxcmp;
2906         int rc;
2907
2908         rc = is_bnxt_in_error(bp);
2909         if (rc)
2910                 return rc;
2911
2912         rxq = dev->data->rx_queues[rx_queue_id];
2913         cpr = rxq->cp_ring;
2914         raw_cons = cpr->cp_raw_cons;
2915
2916         while (1) {
2917                 cons = RING_CMP(cpr->cp_ring_struct, raw_cons);
2918                 rte_prefetch0(&cpr->cp_desc_ring[cons]);
2919                 rxcmp = (struct rx_pkt_cmpl *)&cpr->cp_desc_ring[cons];
2920
2921                 if (!CMP_VALID(rxcmp, raw_cons, cpr->cp_ring_struct)) {
2922                         break;
2923                 } else {
2924                         raw_cons++;
2925                         desc++;
2926                 }
2927         }
2928
2929         return desc;
2930 }
2931
2932 static int
2933 bnxt_rx_descriptor_status_op(void *rx_queue, uint16_t offset)
2934 {
2935         struct bnxt_rx_queue *rxq = (struct bnxt_rx_queue *)rx_queue;
2936         struct bnxt_rx_ring_info *rxr;
2937         struct bnxt_cp_ring_info *cpr;
2938         struct rte_mbuf *rx_buf;
2939         struct rx_pkt_cmpl *rxcmp;
2940         uint32_t cons, cp_cons;
2941         int rc;
2942
2943         if (!rxq)
2944                 return -EINVAL;
2945
2946         rc = is_bnxt_in_error(rxq->bp);
2947         if (rc)
2948                 return rc;
2949
2950         cpr = rxq->cp_ring;
2951         rxr = rxq->rx_ring;
2952
2953         if (offset >= rxq->nb_rx_desc)
2954                 return -EINVAL;
2955
2956         cons = RING_CMP(cpr->cp_ring_struct, offset);
2957         cp_cons = cpr->cp_raw_cons;
2958         rxcmp = (struct rx_pkt_cmpl *)&cpr->cp_desc_ring[cons];
2959
2960         if (cons > cp_cons) {
2961                 if (CMPL_VALID(rxcmp, cpr->valid))
2962                         return RTE_ETH_RX_DESC_DONE;
2963         } else {
2964                 if (CMPL_VALID(rxcmp, !cpr->valid))
2965                         return RTE_ETH_RX_DESC_DONE;
2966         }
2967         rx_buf = rxr->rx_buf_ring[cons];
2968         if (rx_buf == NULL || rx_buf == &rxq->fake_mbuf)
2969                 return RTE_ETH_RX_DESC_UNAVAIL;
2970
2971
2972         return RTE_ETH_RX_DESC_AVAIL;
2973 }
2974
2975 static int
2976 bnxt_tx_descriptor_status_op(void *tx_queue, uint16_t offset)
2977 {
2978         struct bnxt_tx_queue *txq = (struct bnxt_tx_queue *)tx_queue;
2979         struct bnxt_tx_ring_info *txr;
2980         struct bnxt_cp_ring_info *cpr;
2981         struct bnxt_sw_tx_bd *tx_buf;
2982         struct tx_pkt_cmpl *txcmp;
2983         uint32_t cons, cp_cons;
2984         int rc;
2985
2986         if (!txq)
2987                 return -EINVAL;
2988
2989         rc = is_bnxt_in_error(txq->bp);
2990         if (rc)
2991                 return rc;
2992
2993         cpr = txq->cp_ring;
2994         txr = txq->tx_ring;
2995
2996         if (offset >= txq->nb_tx_desc)
2997                 return -EINVAL;
2998
2999         cons = RING_CMP(cpr->cp_ring_struct, offset);
3000         txcmp = (struct tx_pkt_cmpl *)&cpr->cp_desc_ring[cons];
3001         cp_cons = cpr->cp_raw_cons;
3002
3003         if (cons > cp_cons) {
3004                 if (CMPL_VALID(txcmp, cpr->valid))
3005                         return RTE_ETH_TX_DESC_UNAVAIL;
3006         } else {
3007                 if (CMPL_VALID(txcmp, !cpr->valid))
3008                         return RTE_ETH_TX_DESC_UNAVAIL;
3009         }
3010         tx_buf = &txr->tx_buf_ring[cons];
3011         if (tx_buf->mbuf == NULL)
3012                 return RTE_ETH_TX_DESC_DONE;
3013
3014         return RTE_ETH_TX_DESC_FULL;
3015 }
3016
3017 int
3018 bnxt_filter_ctrl_op(struct rte_eth_dev *dev,
3019                     enum rte_filter_type filter_type,
3020                     enum rte_filter_op filter_op, void *arg)
3021 {
3022         struct bnxt *bp = dev->data->dev_private;
3023         int ret = 0;
3024
3025         if (!bp)
3026                 return -EIO;
3027
3028         if (BNXT_ETH_DEV_IS_REPRESENTOR(dev)) {
3029                 struct bnxt_representor *vfr = dev->data->dev_private;
3030                 bp = vfr->parent_dev->data->dev_private;
3031                 /* parent is deleted while children are still valid */
3032                 if (!bp) {
3033                         PMD_DRV_LOG(DEBUG, "BNXT Port:%d VFR Error %d:%d\n",
3034                                     dev->data->port_id,
3035                                     filter_type,
3036                                     filter_op);
3037                         return -EIO;
3038                 }
3039         }
3040
3041         ret = is_bnxt_in_error(bp);
3042         if (ret)
3043                 return ret;
3044
3045         switch (filter_type) {
3046         case RTE_ETH_FILTER_GENERIC:
3047                 if (filter_op != RTE_ETH_FILTER_GET)
3048                         return -EINVAL;
3049
3050                 /* PMD supports thread-safe flow operations.  rte_flow API
3051                  * functions can avoid mutex for multi-thread safety.
3052                  */
3053                 dev->data->dev_flags |= RTE_ETH_DEV_FLOW_OPS_THREAD_SAFE;
3054
3055                 if (BNXT_TRUFLOW_EN(bp))
3056                         *(const void **)arg = &bnxt_ulp_rte_flow_ops;
3057                 else
3058                         *(const void **)arg = &bnxt_flow_ops;
3059                 break;
3060         default:
3061                 PMD_DRV_LOG(ERR,
3062                         "Filter type (%d) not supported", filter_type);
3063                 ret = -EINVAL;
3064                 break;
3065         }
3066         return ret;
3067 }
3068
3069 static const uint32_t *
3070 bnxt_dev_supported_ptypes_get_op(struct rte_eth_dev *dev)
3071 {
3072         static const uint32_t ptypes[] = {
3073                 RTE_PTYPE_L2_ETHER_VLAN,
3074                 RTE_PTYPE_L3_IPV4_EXT_UNKNOWN,
3075                 RTE_PTYPE_L3_IPV6_EXT_UNKNOWN,
3076                 RTE_PTYPE_L4_ICMP,
3077                 RTE_PTYPE_L4_TCP,
3078                 RTE_PTYPE_L4_UDP,
3079                 RTE_PTYPE_INNER_L3_IPV4_EXT_UNKNOWN,
3080                 RTE_PTYPE_INNER_L3_IPV6_EXT_UNKNOWN,
3081                 RTE_PTYPE_INNER_L4_ICMP,
3082                 RTE_PTYPE_INNER_L4_TCP,
3083                 RTE_PTYPE_INNER_L4_UDP,
3084                 RTE_PTYPE_UNKNOWN
3085         };
3086
3087         if (!dev->rx_pkt_burst)
3088                 return NULL;
3089
3090         return ptypes;
3091 }
3092
3093 static int bnxt_map_regs(struct bnxt *bp, uint32_t *reg_arr, int count,
3094                          int reg_win)
3095 {
3096         uint32_t reg_base = *reg_arr & 0xfffff000;
3097         uint32_t win_off;
3098         int i;
3099
3100         for (i = 0; i < count; i++) {
3101                 if ((reg_arr[i] & 0xfffff000) != reg_base)
3102                         return -ERANGE;
3103         }
3104         win_off = BNXT_GRCPF_REG_WINDOW_BASE_OUT + (reg_win - 1) * 4;
3105         rte_write32(reg_base, (uint8_t *)bp->bar0 + win_off);
3106         return 0;
3107 }
3108
3109 static int bnxt_map_ptp_regs(struct bnxt *bp)
3110 {
3111         struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3112         uint32_t *reg_arr;
3113         int rc, i;
3114
3115         reg_arr = ptp->rx_regs;
3116         rc = bnxt_map_regs(bp, reg_arr, BNXT_PTP_RX_REGS, 5);
3117         if (rc)
3118                 return rc;
3119
3120         reg_arr = ptp->tx_regs;
3121         rc = bnxt_map_regs(bp, reg_arr, BNXT_PTP_TX_REGS, 6);
3122         if (rc)
3123                 return rc;
3124
3125         for (i = 0; i < BNXT_PTP_RX_REGS; i++)
3126                 ptp->rx_mapped_regs[i] = 0x5000 + (ptp->rx_regs[i] & 0xfff);
3127
3128         for (i = 0; i < BNXT_PTP_TX_REGS; i++)
3129                 ptp->tx_mapped_regs[i] = 0x6000 + (ptp->tx_regs[i] & 0xfff);
3130
3131         return 0;
3132 }
3133
3134 static void bnxt_unmap_ptp_regs(struct bnxt *bp)
3135 {
3136         rte_write32(0, (uint8_t *)bp->bar0 +
3137                          BNXT_GRCPF_REG_WINDOW_BASE_OUT + 16);
3138         rte_write32(0, (uint8_t *)bp->bar0 +
3139                          BNXT_GRCPF_REG_WINDOW_BASE_OUT + 20);
3140 }
3141
3142 static uint64_t bnxt_cc_read(struct bnxt *bp)
3143 {
3144         uint64_t ns;
3145
3146         ns = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3147                               BNXT_GRCPF_REG_SYNC_TIME));
3148         ns |= (uint64_t)(rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3149                                           BNXT_GRCPF_REG_SYNC_TIME + 4))) << 32;
3150         return ns;
3151 }
3152
3153 static int bnxt_get_tx_ts(struct bnxt *bp, uint64_t *ts)
3154 {
3155         struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3156         uint32_t fifo;
3157
3158         fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3159                                 ptp->tx_mapped_regs[BNXT_PTP_TX_FIFO]));
3160         if (fifo & BNXT_PTP_TX_FIFO_EMPTY)
3161                 return -EAGAIN;
3162
3163         fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3164                                 ptp->tx_mapped_regs[BNXT_PTP_TX_FIFO]));
3165         *ts = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3166                                 ptp->tx_mapped_regs[BNXT_PTP_TX_TS_L]));
3167         *ts |= (uint64_t)rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3168                                 ptp->tx_mapped_regs[BNXT_PTP_TX_TS_H])) << 32;
3169
3170         return 0;
3171 }
3172
3173 static int bnxt_get_rx_ts(struct bnxt *bp, uint64_t *ts)
3174 {
3175         struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3176         struct bnxt_pf_info *pf = bp->pf;
3177         uint16_t port_id;
3178         uint32_t fifo;
3179
3180         if (!ptp)
3181                 return -ENODEV;
3182
3183         fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3184                                 ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO]));
3185         if (!(fifo & BNXT_PTP_RX_FIFO_PENDING))
3186                 return -EAGAIN;
3187
3188         port_id = pf->port_id;
3189         rte_write32(1 << port_id, (uint8_t *)bp->bar0 +
3190                ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO_ADV]);
3191
3192         fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3193                                    ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO]));
3194         if (fifo & BNXT_PTP_RX_FIFO_PENDING) {
3195 /*              bnxt_clr_rx_ts(bp);       TBD  */
3196                 return -EBUSY;
3197         }
3198
3199         *ts = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3200                                 ptp->rx_mapped_regs[BNXT_PTP_RX_TS_L]));
3201         *ts |= (uint64_t)rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3202                                 ptp->rx_mapped_regs[BNXT_PTP_RX_TS_H])) << 32;
3203
3204         return 0;
3205 }
3206
3207 static int
3208 bnxt_timesync_write_time(struct rte_eth_dev *dev, const struct timespec *ts)
3209 {
3210         uint64_t ns;
3211         struct bnxt *bp = dev->data->dev_private;
3212         struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3213
3214         if (!ptp)
3215                 return 0;
3216
3217         ns = rte_timespec_to_ns(ts);
3218         /* Set the timecounters to a new value. */
3219         ptp->tc.nsec = ns;
3220
3221         return 0;
3222 }
3223
3224 static int
3225 bnxt_timesync_read_time(struct rte_eth_dev *dev, struct timespec *ts)
3226 {
3227         struct bnxt *bp = dev->data->dev_private;
3228         struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3229         uint64_t ns, systime_cycles = 0;
3230         int rc = 0;
3231
3232         if (!ptp)
3233                 return 0;
3234
3235         if (BNXT_CHIP_P5(bp))
3236                 rc = bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_CURRENT_TIME,
3237                                              &systime_cycles);
3238         else
3239                 systime_cycles = bnxt_cc_read(bp);
3240
3241         ns = rte_timecounter_update(&ptp->tc, systime_cycles);
3242         *ts = rte_ns_to_timespec(ns);
3243
3244         return rc;
3245 }
3246 static int
3247 bnxt_timesync_enable(struct rte_eth_dev *dev)
3248 {
3249         struct bnxt *bp = dev->data->dev_private;
3250         struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3251         uint32_t shift = 0;
3252         int rc;
3253
3254         if (!ptp)
3255                 return 0;
3256
3257         ptp->rx_filter = 1;
3258         ptp->tx_tstamp_en = 1;
3259         ptp->rxctl = BNXT_PTP_MSG_EVENTS;
3260
3261         rc = bnxt_hwrm_ptp_cfg(bp);
3262         if (rc)
3263                 return rc;
3264
3265         memset(&ptp->tc, 0, sizeof(struct rte_timecounter));
3266         memset(&ptp->rx_tstamp_tc, 0, sizeof(struct rte_timecounter));
3267         memset(&ptp->tx_tstamp_tc, 0, sizeof(struct rte_timecounter));
3268
3269         ptp->tc.cc_mask = BNXT_CYCLECOUNTER_MASK;
3270         ptp->tc.cc_shift = shift;
3271         ptp->tc.nsec_mask = (1ULL << shift) - 1;
3272
3273         ptp->rx_tstamp_tc.cc_mask = BNXT_CYCLECOUNTER_MASK;
3274         ptp->rx_tstamp_tc.cc_shift = shift;
3275         ptp->rx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
3276
3277         ptp->tx_tstamp_tc.cc_mask = BNXT_CYCLECOUNTER_MASK;
3278         ptp->tx_tstamp_tc.cc_shift = shift;
3279         ptp->tx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
3280
3281         if (!BNXT_CHIP_P5(bp))
3282                 bnxt_map_ptp_regs(bp);
3283
3284         return 0;
3285 }
3286
3287 static int
3288 bnxt_timesync_disable(struct rte_eth_dev *dev)
3289 {
3290         struct bnxt *bp = dev->data->dev_private;
3291         struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3292
3293         if (!ptp)
3294                 return 0;
3295
3296         ptp->rx_filter = 0;
3297         ptp->tx_tstamp_en = 0;
3298         ptp->rxctl = 0;
3299
3300         bnxt_hwrm_ptp_cfg(bp);
3301
3302         if (!BNXT_CHIP_P5(bp))
3303                 bnxt_unmap_ptp_regs(bp);
3304
3305         return 0;
3306 }
3307
3308 static int
3309 bnxt_timesync_read_rx_timestamp(struct rte_eth_dev *dev,
3310                                  struct timespec *timestamp,
3311                                  uint32_t flags __rte_unused)
3312 {
3313         struct bnxt *bp = dev->data->dev_private;
3314         struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3315         uint64_t rx_tstamp_cycles = 0;
3316         uint64_t ns;
3317
3318         if (!ptp)
3319                 return 0;
3320
3321         if (BNXT_CHIP_P5(bp))
3322                 rx_tstamp_cycles = ptp->rx_timestamp;
3323         else
3324                 bnxt_get_rx_ts(bp, &rx_tstamp_cycles);
3325
3326         ns = rte_timecounter_update(&ptp->rx_tstamp_tc, rx_tstamp_cycles);
3327         *timestamp = rte_ns_to_timespec(ns);
3328         return  0;
3329 }
3330
3331 static int
3332 bnxt_timesync_read_tx_timestamp(struct rte_eth_dev *dev,
3333                                  struct timespec *timestamp)
3334 {
3335         struct bnxt *bp = dev->data->dev_private;
3336         struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3337         uint64_t tx_tstamp_cycles = 0;
3338         uint64_t ns;
3339         int rc = 0;
3340
3341         if (!ptp)
3342                 return 0;
3343
3344         if (BNXT_CHIP_P5(bp))
3345                 rc = bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_PATH_TX,
3346                                              &tx_tstamp_cycles);
3347         else
3348                 rc = bnxt_get_tx_ts(bp, &tx_tstamp_cycles);
3349
3350         ns = rte_timecounter_update(&ptp->tx_tstamp_tc, tx_tstamp_cycles);
3351         *timestamp = rte_ns_to_timespec(ns);
3352
3353         return rc;
3354 }
3355
3356 static int
3357 bnxt_timesync_adjust_time(struct rte_eth_dev *dev, int64_t delta)
3358 {
3359         struct bnxt *bp = dev->data->dev_private;
3360         struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3361
3362         if (!ptp)
3363                 return 0;
3364
3365         ptp->tc.nsec += delta;
3366
3367         return 0;
3368 }
3369
3370 static int
3371 bnxt_get_eeprom_length_op(struct rte_eth_dev *dev)
3372 {
3373         struct bnxt *bp = dev->data->dev_private;
3374         int rc;
3375         uint32_t dir_entries;
3376         uint32_t entry_length;
3377
3378         rc = is_bnxt_in_error(bp);
3379         if (rc)
3380                 return rc;
3381
3382         PMD_DRV_LOG(INFO, PCI_PRI_FMT "\n",
3383                     bp->pdev->addr.domain, bp->pdev->addr.bus,
3384                     bp->pdev->addr.devid, bp->pdev->addr.function);
3385
3386         rc = bnxt_hwrm_nvm_get_dir_info(bp, &dir_entries, &entry_length);
3387         if (rc != 0)
3388                 return rc;
3389
3390         return dir_entries * entry_length;
3391 }
3392
3393 static int
3394 bnxt_get_eeprom_op(struct rte_eth_dev *dev,
3395                 struct rte_dev_eeprom_info *in_eeprom)
3396 {
3397         struct bnxt *bp = dev->data->dev_private;
3398         uint32_t index;
3399         uint32_t offset;
3400         int rc;
3401
3402         rc = is_bnxt_in_error(bp);
3403         if (rc)
3404                 return rc;
3405
3406         PMD_DRV_LOG(INFO, PCI_PRI_FMT " in_eeprom->offset = %d len = %d\n",
3407                     bp->pdev->addr.domain, bp->pdev->addr.bus,
3408                     bp->pdev->addr.devid, bp->pdev->addr.function,
3409                     in_eeprom->offset, in_eeprom->length);
3410
3411         if (in_eeprom->offset == 0) /* special offset value to get directory */
3412                 return bnxt_get_nvram_directory(bp, in_eeprom->length,
3413                                                 in_eeprom->data);
3414
3415         index = in_eeprom->offset >> 24;
3416         offset = in_eeprom->offset & 0xffffff;
3417
3418         if (index != 0)
3419                 return bnxt_hwrm_get_nvram_item(bp, index - 1, offset,
3420                                            in_eeprom->length, in_eeprom->data);
3421
3422         return 0;
3423 }
3424
3425 static bool bnxt_dir_type_is_ape_bin_format(uint16_t dir_type)
3426 {
3427         switch (dir_type) {
3428         case BNX_DIR_TYPE_CHIMP_PATCH:
3429         case BNX_DIR_TYPE_BOOTCODE:
3430         case BNX_DIR_TYPE_BOOTCODE_2:
3431         case BNX_DIR_TYPE_APE_FW:
3432         case BNX_DIR_TYPE_APE_PATCH:
3433         case BNX_DIR_TYPE_KONG_FW:
3434         case BNX_DIR_TYPE_KONG_PATCH:
3435         case BNX_DIR_TYPE_BONO_FW:
3436         case BNX_DIR_TYPE_BONO_PATCH:
3437                 /* FALLTHROUGH */
3438                 return true;
3439         }
3440
3441         return false;
3442 }
3443
3444 static bool bnxt_dir_type_is_other_exec_format(uint16_t dir_type)
3445 {
3446         switch (dir_type) {
3447         case BNX_DIR_TYPE_AVS:
3448         case BNX_DIR_TYPE_EXP_ROM_MBA:
3449         case BNX_DIR_TYPE_PCIE:
3450         case BNX_DIR_TYPE_TSCF_UCODE:
3451         case BNX_DIR_TYPE_EXT_PHY:
3452         case BNX_DIR_TYPE_CCM:
3453         case BNX_DIR_TYPE_ISCSI_BOOT:
3454         case BNX_DIR_TYPE_ISCSI_BOOT_IPV6:
3455         case BNX_DIR_TYPE_ISCSI_BOOT_IPV4N6:
3456                 /* FALLTHROUGH */
3457                 return true;
3458         }
3459
3460         return false;
3461 }
3462
3463 static bool bnxt_dir_type_is_executable(uint16_t dir_type)
3464 {
3465         return bnxt_dir_type_is_ape_bin_format(dir_type) ||
3466                 bnxt_dir_type_is_other_exec_format(dir_type);
3467 }
3468
3469 static int
3470 bnxt_set_eeprom_op(struct rte_eth_dev *dev,
3471                 struct rte_dev_eeprom_info *in_eeprom)
3472 {
3473         struct bnxt *bp = dev->data->dev_private;
3474         uint8_t index, dir_op;
3475         uint16_t type, ext, ordinal, attr;
3476         int rc;
3477
3478         rc = is_bnxt_in_error(bp);
3479         if (rc)
3480                 return rc;
3481
3482         PMD_DRV_LOG(INFO, PCI_PRI_FMT " in_eeprom->offset = %d len = %d\n",
3483                     bp->pdev->addr.domain, bp->pdev->addr.bus,
3484                     bp->pdev->addr.devid, bp->pdev->addr.function,
3485                     in_eeprom->offset, in_eeprom->length);
3486
3487         if (!BNXT_PF(bp)) {
3488                 PMD_DRV_LOG(ERR, "NVM write not supported from a VF\n");
3489                 return -EINVAL;
3490         }
3491
3492         type = in_eeprom->magic >> 16;
3493
3494         if (type == 0xffff) { /* special value for directory operations */
3495                 index = in_eeprom->magic & 0xff;
3496                 dir_op = in_eeprom->magic >> 8;
3497                 if (index == 0)
3498                         return -EINVAL;
3499                 switch (dir_op) {
3500                 case 0x0e: /* erase */
3501                         if (in_eeprom->offset != ~in_eeprom->magic)
3502                                 return -EINVAL;
3503                         return bnxt_hwrm_erase_nvram_directory(bp, index - 1);
3504                 default:
3505                         return -EINVAL;
3506                 }
3507         }
3508
3509         /* Create or re-write an NVM item: */
3510         if (bnxt_dir_type_is_executable(type) == true)
3511                 return -EOPNOTSUPP;
3512         ext = in_eeprom->magic & 0xffff;
3513         ordinal = in_eeprom->offset >> 16;
3514         attr = in_eeprom->offset & 0xffff;
3515
3516         return bnxt_hwrm_flash_nvram(bp, type, ordinal, ext, attr,
3517                                      in_eeprom->data, in_eeprom->length);
3518 }
3519
3520 /*
3521  * Initialization
3522  */
3523
3524 static const struct eth_dev_ops bnxt_dev_ops = {
3525         .dev_infos_get = bnxt_dev_info_get_op,
3526         .dev_close = bnxt_dev_close_op,
3527         .dev_configure = bnxt_dev_configure_op,
3528         .dev_start = bnxt_dev_start_op,
3529         .dev_stop = bnxt_dev_stop_op,
3530         .dev_set_link_up = bnxt_dev_set_link_up_op,
3531         .dev_set_link_down = bnxt_dev_set_link_down_op,
3532         .stats_get = bnxt_stats_get_op,
3533         .stats_reset = bnxt_stats_reset_op,
3534         .rx_queue_setup = bnxt_rx_queue_setup_op,
3535         .rx_queue_release = bnxt_rx_queue_release_op,
3536         .tx_queue_setup = bnxt_tx_queue_setup_op,
3537         .tx_queue_release = bnxt_tx_queue_release_op,
3538         .rx_queue_intr_enable = bnxt_rx_queue_intr_enable_op,
3539         .rx_queue_intr_disable = bnxt_rx_queue_intr_disable_op,
3540         .reta_update = bnxt_reta_update_op,
3541         .reta_query = bnxt_reta_query_op,
3542         .rss_hash_update = bnxt_rss_hash_update_op,
3543         .rss_hash_conf_get = bnxt_rss_hash_conf_get_op,
3544         .link_update = bnxt_link_update_op,
3545         .promiscuous_enable = bnxt_promiscuous_enable_op,
3546         .promiscuous_disable = bnxt_promiscuous_disable_op,
3547         .allmulticast_enable = bnxt_allmulticast_enable_op,
3548         .allmulticast_disable = bnxt_allmulticast_disable_op,
3549         .mac_addr_add = bnxt_mac_addr_add_op,
3550         .mac_addr_remove = bnxt_mac_addr_remove_op,
3551         .flow_ctrl_get = bnxt_flow_ctrl_get_op,
3552         .flow_ctrl_set = bnxt_flow_ctrl_set_op,
3553         .udp_tunnel_port_add  = bnxt_udp_tunnel_port_add_op,
3554         .udp_tunnel_port_del  = bnxt_udp_tunnel_port_del_op,
3555         .vlan_filter_set = bnxt_vlan_filter_set_op,
3556         .vlan_offload_set = bnxt_vlan_offload_set_op,
3557         .vlan_tpid_set = bnxt_vlan_tpid_set_op,
3558         .vlan_pvid_set = bnxt_vlan_pvid_set_op,
3559         .mtu_set = bnxt_mtu_set_op,
3560         .mac_addr_set = bnxt_set_default_mac_addr_op,
3561         .xstats_get = bnxt_dev_xstats_get_op,
3562         .xstats_get_names = bnxt_dev_xstats_get_names_op,
3563         .xstats_reset = bnxt_dev_xstats_reset_op,
3564         .fw_version_get = bnxt_fw_version_get,
3565         .set_mc_addr_list = bnxt_dev_set_mc_addr_list_op,
3566         .rxq_info_get = bnxt_rxq_info_get_op,
3567         .txq_info_get = bnxt_txq_info_get_op,
3568         .rx_burst_mode_get = bnxt_rx_burst_mode_get,
3569         .tx_burst_mode_get = bnxt_tx_burst_mode_get,
3570         .dev_led_on = bnxt_dev_led_on_op,
3571         .dev_led_off = bnxt_dev_led_off_op,
3572         .rx_queue_start = bnxt_rx_queue_start,
3573         .rx_queue_stop = bnxt_rx_queue_stop,
3574         .tx_queue_start = bnxt_tx_queue_start,
3575         .tx_queue_stop = bnxt_tx_queue_stop,
3576         .filter_ctrl = bnxt_filter_ctrl_op,
3577         .dev_supported_ptypes_get = bnxt_dev_supported_ptypes_get_op,
3578         .get_eeprom_length    = bnxt_get_eeprom_length_op,
3579         .get_eeprom           = bnxt_get_eeprom_op,
3580         .set_eeprom           = bnxt_set_eeprom_op,
3581         .timesync_enable      = bnxt_timesync_enable,
3582         .timesync_disable     = bnxt_timesync_disable,
3583         .timesync_read_time   = bnxt_timesync_read_time,
3584         .timesync_write_time   = bnxt_timesync_write_time,
3585         .timesync_adjust_time = bnxt_timesync_adjust_time,
3586         .timesync_read_rx_timestamp = bnxt_timesync_read_rx_timestamp,
3587         .timesync_read_tx_timestamp = bnxt_timesync_read_tx_timestamp,
3588 };
3589
3590 static uint32_t bnxt_map_reset_regs(struct bnxt *bp, uint32_t reg)
3591 {
3592         uint32_t offset;
3593
3594         /* Only pre-map the reset GRC registers using window 3 */
3595         rte_write32(reg & 0xfffff000, (uint8_t *)bp->bar0 +
3596                     BNXT_GRCPF_REG_WINDOW_BASE_OUT + 8);
3597
3598         offset = BNXT_GRCP_WINDOW_3_BASE + (reg & 0xffc);
3599
3600         return offset;
3601 }
3602
3603 int bnxt_map_fw_health_status_regs(struct bnxt *bp)
3604 {
3605         struct bnxt_error_recovery_info *info = bp->recovery_info;
3606         uint32_t reg_base = 0xffffffff;
3607         int i;
3608
3609         /* Only pre-map the monitoring GRC registers using window 2 */
3610         for (i = 0; i < BNXT_FW_STATUS_REG_CNT; i++) {
3611                 uint32_t reg = info->status_regs[i];
3612
3613                 if (BNXT_FW_STATUS_REG_TYPE(reg) != BNXT_FW_STATUS_REG_TYPE_GRC)
3614                         continue;
3615
3616                 if (reg_base == 0xffffffff)
3617                         reg_base = reg & 0xfffff000;
3618                 if ((reg & 0xfffff000) != reg_base)
3619                         return -ERANGE;
3620
3621                 /* Use mask 0xffc as the Lower 2 bits indicates
3622                  * address space location
3623                  */
3624                 info->mapped_status_regs[i] = BNXT_GRCP_WINDOW_2_BASE +
3625                                                 (reg & 0xffc);
3626         }
3627
3628         if (reg_base == 0xffffffff)
3629                 return 0;
3630
3631         rte_write32(reg_base, (uint8_t *)bp->bar0 +
3632                     BNXT_GRCPF_REG_WINDOW_BASE_OUT + 4);
3633
3634         return 0;
3635 }
3636
3637 static void bnxt_write_fw_reset_reg(struct bnxt *bp, uint32_t index)
3638 {
3639         struct bnxt_error_recovery_info *info = bp->recovery_info;
3640         uint32_t delay = info->delay_after_reset[index];
3641         uint32_t val = info->reset_reg_val[index];
3642         uint32_t reg = info->reset_reg[index];
3643         uint32_t type, offset;
3644
3645         type = BNXT_FW_STATUS_REG_TYPE(reg);
3646         offset = BNXT_FW_STATUS_REG_OFF(reg);
3647
3648         switch (type) {
3649         case BNXT_FW_STATUS_REG_TYPE_CFG:
3650                 rte_pci_write_config(bp->pdev, &val, sizeof(val), offset);
3651                 break;
3652         case BNXT_FW_STATUS_REG_TYPE_GRC:
3653                 offset = bnxt_map_reset_regs(bp, offset);
3654                 rte_write32(val, (uint8_t *)bp->bar0 + offset);
3655                 break;
3656         case BNXT_FW_STATUS_REG_TYPE_BAR0:
3657                 rte_write32(val, (uint8_t *)bp->bar0 + offset);
3658                 break;
3659         }
3660         /* wait on a specific interval of time until core reset is complete */
3661         if (delay)
3662                 rte_delay_ms(delay);
3663 }
3664
3665 static void bnxt_dev_cleanup(struct bnxt *bp)
3666 {
3667         bp->eth_dev->data->dev_link.link_status = 0;
3668         bp->link_info->link_up = 0;
3669         if (bp->eth_dev->data->dev_started)
3670                 bnxt_dev_stop_op(bp->eth_dev);
3671
3672         bnxt_uninit_resources(bp, true);
3673 }
3674
3675 static int bnxt_restore_vlan_filters(struct bnxt *bp)
3676 {
3677         struct rte_eth_dev *dev = bp->eth_dev;
3678         struct rte_vlan_filter_conf *vfc;
3679         int vidx, vbit, rc;
3680         uint16_t vlan_id;
3681
3682         for (vlan_id = 1; vlan_id <= RTE_ETHER_MAX_VLAN_ID; vlan_id++) {
3683                 vfc = &dev->data->vlan_filter_conf;
3684                 vidx = vlan_id / 64;
3685                 vbit = vlan_id % 64;
3686
3687                 /* Each bit corresponds to a VLAN id */
3688                 if (vfc->ids[vidx] & (UINT64_C(1) << vbit)) {
3689                         rc = bnxt_add_vlan_filter(bp, vlan_id);
3690                         if (rc)
3691                                 return rc;
3692                 }
3693         }
3694
3695         return 0;
3696 }
3697
3698 static int bnxt_restore_mac_filters(struct bnxt *bp)
3699 {
3700         struct rte_eth_dev *dev = bp->eth_dev;
3701         struct rte_eth_dev_info dev_info;
3702         struct rte_ether_addr *addr;
3703         uint64_t pool_mask;
3704         uint32_t pool = 0;
3705         uint16_t i;
3706         int rc;
3707
3708         if (BNXT_VF(bp) && !BNXT_VF_IS_TRUSTED(bp))
3709                 return 0;
3710
3711         rc = bnxt_dev_info_get_op(dev, &dev_info);
3712         if (rc)
3713                 return rc;
3714
3715         /* replay MAC address configuration */
3716         for (i = 1; i < dev_info.max_mac_addrs; i++) {
3717                 addr = &dev->data->mac_addrs[i];
3718
3719                 /* skip zero address */
3720                 if (rte_is_zero_ether_addr(addr))
3721                         continue;
3722
3723                 pool = 0;
3724                 pool_mask = dev->data->mac_pool_sel[i];
3725
3726                 do {
3727                         if (pool_mask & 1ULL) {
3728                                 rc = bnxt_mac_addr_add_op(dev, addr, i, pool);
3729                                 if (rc)
3730                                         return rc;
3731                         }
3732                         pool_mask >>= 1;
3733                         pool++;
3734                 } while (pool_mask);
3735         }
3736
3737         return 0;
3738 }
3739
3740 static int bnxt_restore_filters(struct bnxt *bp)
3741 {
3742         struct rte_eth_dev *dev = bp->eth_dev;
3743         int ret = 0;
3744
3745         if (dev->data->all_multicast) {
3746                 ret = bnxt_allmulticast_enable_op(dev);
3747                 if (ret)
3748                         return ret;
3749         }
3750         if (dev->data->promiscuous) {
3751                 ret = bnxt_promiscuous_enable_op(dev);
3752                 if (ret)
3753                         return ret;
3754         }
3755
3756         ret = bnxt_restore_mac_filters(bp);
3757         if (ret)
3758                 return ret;
3759
3760         ret = bnxt_restore_vlan_filters(bp);
3761         /* TODO restore other filters as well */
3762         return ret;
3763 }
3764
3765 static void bnxt_dev_recover(void *arg)
3766 {
3767         struct bnxt *bp = arg;
3768         int timeout = bp->fw_reset_max_msecs;
3769         int rc = 0;
3770
3771         /* Clear Error flag so that device re-init should happen */
3772         bp->flags &= ~BNXT_FLAG_FATAL_ERROR;
3773
3774         do {
3775                 rc = bnxt_hwrm_ver_get(bp, SHORT_HWRM_CMD_TIMEOUT);
3776                 if (rc == 0)
3777                         break;
3778                 rte_delay_ms(BNXT_FW_READY_WAIT_INTERVAL);
3779                 timeout -= BNXT_FW_READY_WAIT_INTERVAL;
3780         } while (rc && timeout);
3781
3782         if (rc) {
3783                 PMD_DRV_LOG(ERR, "FW is not Ready after reset\n");
3784                 goto err;
3785         }
3786
3787         rc = bnxt_init_resources(bp, true);
3788         if (rc) {
3789                 PMD_DRV_LOG(ERR,
3790                             "Failed to initialize resources after reset\n");
3791                 goto err;
3792         }
3793         /* clear reset flag as the device is initialized now */
3794         bp->flags &= ~BNXT_FLAG_FW_RESET;
3795
3796         rc = bnxt_dev_start_op(bp->eth_dev);
3797         if (rc) {
3798                 PMD_DRV_LOG(ERR, "Failed to start port after reset\n");
3799                 goto err_start;
3800         }
3801
3802         rc = bnxt_restore_filters(bp);
3803         if (rc)
3804                 goto err_start;
3805
3806         PMD_DRV_LOG(INFO, "Recovered from FW reset\n");
3807         return;
3808 err_start:
3809         bnxt_dev_stop_op(bp->eth_dev);
3810 err:
3811         bp->flags |= BNXT_FLAG_FATAL_ERROR;
3812         bnxt_uninit_resources(bp, false);
3813         PMD_DRV_LOG(ERR, "Failed to recover from FW reset\n");
3814 }
3815
3816 void bnxt_dev_reset_and_resume(void *arg)
3817 {
3818         struct bnxt *bp = arg;
3819         int rc;
3820
3821         bnxt_dev_cleanup(bp);
3822
3823         bnxt_wait_for_device_shutdown(bp);
3824
3825         rc = rte_eal_alarm_set(US_PER_MS * bp->fw_reset_min_msecs,
3826                                bnxt_dev_recover, (void *)bp);
3827         if (rc)
3828                 PMD_DRV_LOG(ERR, "Error setting recovery alarm");
3829 }
3830
3831 uint32_t bnxt_read_fw_status_reg(struct bnxt *bp, uint32_t index)
3832 {
3833         struct bnxt_error_recovery_info *info = bp->recovery_info;
3834         uint32_t reg = info->status_regs[index];
3835         uint32_t type, offset, val = 0;
3836
3837         type = BNXT_FW_STATUS_REG_TYPE(reg);
3838         offset = BNXT_FW_STATUS_REG_OFF(reg);
3839
3840         switch (type) {
3841         case BNXT_FW_STATUS_REG_TYPE_CFG:
3842                 rte_pci_read_config(bp->pdev, &val, sizeof(val), offset);
3843                 break;
3844         case BNXT_FW_STATUS_REG_TYPE_GRC:
3845                 offset = info->mapped_status_regs[index];
3846                 /* FALLTHROUGH */
3847         case BNXT_FW_STATUS_REG_TYPE_BAR0:
3848                 val = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3849                                        offset));
3850                 break;
3851         }
3852
3853         return val;
3854 }
3855
3856 static int bnxt_fw_reset_all(struct bnxt *bp)
3857 {
3858         struct bnxt_error_recovery_info *info = bp->recovery_info;
3859         uint32_t i;
3860         int rc = 0;
3861
3862         if (info->flags & BNXT_FLAG_ERROR_RECOVERY_HOST) {
3863                 /* Reset through master function driver */
3864                 for (i = 0; i < info->reg_array_cnt; i++)
3865                         bnxt_write_fw_reset_reg(bp, i);
3866                 /* Wait for time specified by FW after triggering reset */
3867                 rte_delay_ms(info->master_func_wait_period_after_reset);
3868         } else if (info->flags & BNXT_FLAG_ERROR_RECOVERY_CO_CPU) {
3869                 /* Reset with the help of Kong processor */
3870                 rc = bnxt_hwrm_fw_reset(bp);
3871                 if (rc)
3872                         PMD_DRV_LOG(ERR, "Failed to reset FW\n");
3873         }
3874
3875         return rc;
3876 }
3877
3878 static void bnxt_fw_reset_cb(void *arg)
3879 {
3880         struct bnxt *bp = arg;
3881         struct bnxt_error_recovery_info *info = bp->recovery_info;
3882         int rc = 0;
3883
3884         /* Only Master function can do FW reset */
3885         if (bnxt_is_master_func(bp) &&
3886             bnxt_is_recovery_enabled(bp)) {
3887                 rc = bnxt_fw_reset_all(bp);
3888                 if (rc) {
3889                         PMD_DRV_LOG(ERR, "Adapter recovery failed\n");
3890                         return;
3891                 }
3892         }
3893
3894         /* if recovery method is ERROR_RECOVERY_CO_CPU, KONG will send
3895          * EXCEPTION_FATAL_ASYNC event to all the functions
3896          * (including MASTER FUNC). After receiving this Async, all the active
3897          * drivers should treat this case as FW initiated recovery
3898          */
3899         if (info->flags & BNXT_FLAG_ERROR_RECOVERY_HOST) {
3900                 bp->fw_reset_min_msecs = BNXT_MIN_FW_READY_TIMEOUT;
3901                 bp->fw_reset_max_msecs = BNXT_MAX_FW_RESET_TIMEOUT;
3902
3903                 /* To recover from error */
3904                 rte_eal_alarm_set(US_PER_MS, bnxt_dev_reset_and_resume,
3905                                   (void *)bp);
3906         }
3907 }
3908
3909 /* Driver should poll FW heartbeat, reset_counter with the frequency
3910  * advertised by FW in HWRM_ERROR_RECOVERY_QCFG.
3911  * When the driver detects heartbeat stop or change in reset_counter,
3912  * it has to trigger a reset to recover from the error condition.
3913  * A “master PF” is the function who will have the privilege to
3914  * initiate the chimp reset. The master PF will be elected by the
3915  * firmware and will be notified through async message.
3916  */
3917 static void bnxt_check_fw_health(void *arg)
3918 {
3919         struct bnxt *bp = arg;
3920         struct bnxt_error_recovery_info *info = bp->recovery_info;
3921         uint32_t val = 0, wait_msec;
3922
3923         if (!info || !bnxt_is_recovery_enabled(bp) ||
3924             is_bnxt_in_error(bp))
3925                 return;
3926
3927         val = bnxt_read_fw_status_reg(bp, BNXT_FW_HEARTBEAT_CNT_REG);
3928         if (val == info->last_heart_beat)
3929                 goto reset;
3930
3931         info->last_heart_beat = val;
3932
3933         val = bnxt_read_fw_status_reg(bp, BNXT_FW_RECOVERY_CNT_REG);
3934         if (val != info->last_reset_counter)
3935                 goto reset;
3936
3937         info->last_reset_counter = val;
3938
3939         rte_eal_alarm_set(US_PER_MS * info->driver_polling_freq,
3940                           bnxt_check_fw_health, (void *)bp);
3941
3942         return;
3943 reset:
3944         /* Stop DMA to/from device */
3945         bp->flags |= BNXT_FLAG_FATAL_ERROR;
3946         bp->flags |= BNXT_FLAG_FW_RESET;
3947
3948         PMD_DRV_LOG(ERR, "Detected FW dead condition\n");
3949
3950         if (bnxt_is_master_func(bp))
3951                 wait_msec = info->master_func_wait_period;
3952         else
3953                 wait_msec = info->normal_func_wait_period;
3954
3955         rte_eal_alarm_set(US_PER_MS * wait_msec,
3956                           bnxt_fw_reset_cb, (void *)bp);
3957 }
3958
3959 void bnxt_schedule_fw_health_check(struct bnxt *bp)
3960 {
3961         uint32_t polling_freq;
3962
3963         pthread_mutex_lock(&bp->health_check_lock);
3964
3965         if (!bnxt_is_recovery_enabled(bp))
3966                 goto done;
3967
3968         if (bp->flags & BNXT_FLAG_FW_HEALTH_CHECK_SCHEDULED)
3969                 goto done;
3970
3971         polling_freq = bp->recovery_info->driver_polling_freq;
3972
3973         rte_eal_alarm_set(US_PER_MS * polling_freq,
3974                           bnxt_check_fw_health, (void *)bp);
3975         bp->flags |= BNXT_FLAG_FW_HEALTH_CHECK_SCHEDULED;
3976
3977 done:
3978         pthread_mutex_unlock(&bp->health_check_lock);
3979 }
3980
3981 static void bnxt_cancel_fw_health_check(struct bnxt *bp)
3982 {
3983         if (!bnxt_is_recovery_enabled(bp))
3984                 return;
3985
3986         rte_eal_alarm_cancel(bnxt_check_fw_health, (void *)bp);
3987         bp->flags &= ~BNXT_FLAG_FW_HEALTH_CHECK_SCHEDULED;
3988 }
3989
3990 static bool bnxt_vf_pciid(uint16_t device_id)
3991 {
3992         switch (device_id) {
3993         case BROADCOM_DEV_ID_57304_VF:
3994         case BROADCOM_DEV_ID_57406_VF:
3995         case BROADCOM_DEV_ID_5731X_VF:
3996         case BROADCOM_DEV_ID_5741X_VF:
3997         case BROADCOM_DEV_ID_57414_VF:
3998         case BROADCOM_DEV_ID_STRATUS_NIC_VF1:
3999         case BROADCOM_DEV_ID_STRATUS_NIC_VF2:
4000         case BROADCOM_DEV_ID_58802_VF:
4001         case BROADCOM_DEV_ID_57500_VF1:
4002         case BROADCOM_DEV_ID_57500_VF2:
4003                 /* FALLTHROUGH */
4004                 return true;
4005         default:
4006                 return false;
4007         }
4008 }
4009
4010 /* Phase 5 device */
4011 static bool bnxt_p5_device(uint16_t device_id)
4012 {
4013         switch (device_id) {
4014         case BROADCOM_DEV_ID_57508:
4015         case BROADCOM_DEV_ID_57504:
4016         case BROADCOM_DEV_ID_57502:
4017         case BROADCOM_DEV_ID_57508_MF1:
4018         case BROADCOM_DEV_ID_57504_MF1:
4019         case BROADCOM_DEV_ID_57502_MF1:
4020         case BROADCOM_DEV_ID_57508_MF2:
4021         case BROADCOM_DEV_ID_57504_MF2:
4022         case BROADCOM_DEV_ID_57502_MF2:
4023         case BROADCOM_DEV_ID_57500_VF1:
4024         case BROADCOM_DEV_ID_57500_VF2:
4025                 /* FALLTHROUGH */
4026                 return true;
4027         default:
4028                 return false;
4029         }
4030 }
4031
4032 bool bnxt_stratus_device(struct bnxt *bp)
4033 {
4034         uint16_t device_id = bp->pdev->id.device_id;
4035
4036         switch (device_id) {
4037         case BROADCOM_DEV_ID_STRATUS_NIC:
4038         case BROADCOM_DEV_ID_STRATUS_NIC_VF1:
4039         case BROADCOM_DEV_ID_STRATUS_NIC_VF2:
4040                 /* FALLTHROUGH */
4041                 return true;
4042         default:
4043                 return false;
4044         }
4045 }
4046
4047 static int bnxt_init_board(struct rte_eth_dev *eth_dev)
4048 {
4049         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
4050         struct bnxt *bp = eth_dev->data->dev_private;
4051
4052         /* enable device (incl. PCI PM wakeup), and bus-mastering */
4053         bp->bar0 = (void *)pci_dev->mem_resource[0].addr;
4054         bp->doorbell_base = (void *)pci_dev->mem_resource[2].addr;
4055         if (!bp->bar0 || !bp->doorbell_base) {
4056                 PMD_DRV_LOG(ERR, "Unable to access Hardware\n");
4057                 return -ENODEV;
4058         }
4059
4060         bp->eth_dev = eth_dev;
4061         bp->pdev = pci_dev;
4062
4063         return 0;
4064 }
4065
4066 static int bnxt_alloc_ctx_mem_blk(struct bnxt *bp,
4067                                   struct bnxt_ctx_pg_info *ctx_pg,
4068                                   uint32_t mem_size,
4069                                   const char *suffix,
4070                                   uint16_t idx)
4071 {
4072         struct bnxt_ring_mem_info *rmem = &ctx_pg->ring_mem;
4073         const struct rte_memzone *mz = NULL;
4074         char mz_name[RTE_MEMZONE_NAMESIZE];
4075         rte_iova_t mz_phys_addr;
4076         uint64_t valid_bits = 0;
4077         uint32_t sz;
4078         int i;
4079
4080         if (!mem_size)
4081                 return 0;
4082
4083         rmem->nr_pages = RTE_ALIGN_MUL_CEIL(mem_size, BNXT_PAGE_SIZE) /
4084                          BNXT_PAGE_SIZE;
4085         rmem->page_size = BNXT_PAGE_SIZE;
4086         rmem->pg_arr = ctx_pg->ctx_pg_arr;
4087         rmem->dma_arr = ctx_pg->ctx_dma_arr;
4088         rmem->flags = BNXT_RMEM_VALID_PTE_FLAG;
4089
4090         valid_bits = PTU_PTE_VALID;
4091
4092         if (rmem->nr_pages > 1) {
4093                 snprintf(mz_name, RTE_MEMZONE_NAMESIZE,
4094                          "bnxt_ctx_pg_tbl%s_%x_%d",
4095                          suffix, idx, bp->eth_dev->data->port_id);
4096                 mz_name[RTE_MEMZONE_NAMESIZE - 1] = 0;
4097                 mz = rte_memzone_lookup(mz_name);
4098                 if (!mz) {
4099                         mz = rte_memzone_reserve_aligned(mz_name,
4100                                                 rmem->nr_pages * 8,
4101                                                 SOCKET_ID_ANY,
4102                                                 RTE_MEMZONE_2MB |
4103                                                 RTE_MEMZONE_SIZE_HINT_ONLY |
4104                                                 RTE_MEMZONE_IOVA_CONTIG,
4105                                                 BNXT_PAGE_SIZE);
4106                         if (mz == NULL)
4107                                 return -ENOMEM;
4108                 }
4109
4110                 memset(mz->addr, 0, mz->len);
4111                 mz_phys_addr = mz->iova;
4112
4113                 rmem->pg_tbl = mz->addr;
4114                 rmem->pg_tbl_map = mz_phys_addr;
4115                 rmem->pg_tbl_mz = mz;
4116         }
4117
4118         snprintf(mz_name, RTE_MEMZONE_NAMESIZE, "bnxt_ctx_%s_%x_%d",
4119                  suffix, idx, bp->eth_dev->data->port_id);
4120         mz = rte_memzone_lookup(mz_name);
4121         if (!mz) {
4122                 mz = rte_memzone_reserve_aligned(mz_name,
4123                                                  mem_size,
4124                                                  SOCKET_ID_ANY,
4125                                                  RTE_MEMZONE_1GB |
4126                                                  RTE_MEMZONE_SIZE_HINT_ONLY |
4127                                                  RTE_MEMZONE_IOVA_CONTIG,
4128                                                  BNXT_PAGE_SIZE);
4129                 if (mz == NULL)
4130                         return -ENOMEM;
4131         }
4132
4133         memset(mz->addr, 0, mz->len);
4134         mz_phys_addr = mz->iova;
4135
4136         for (sz = 0, i = 0; sz < mem_size; sz += BNXT_PAGE_SIZE, i++) {
4137                 rmem->pg_arr[i] = ((char *)mz->addr) + sz;
4138                 rmem->dma_arr[i] = mz_phys_addr + sz;
4139
4140                 if (rmem->nr_pages > 1) {
4141                         if (i == rmem->nr_pages - 2 &&
4142                             (rmem->flags & BNXT_RMEM_RING_PTE_FLAG))
4143                                 valid_bits |= PTU_PTE_NEXT_TO_LAST;
4144                         else if (i == rmem->nr_pages - 1 &&
4145                                  (rmem->flags & BNXT_RMEM_RING_PTE_FLAG))
4146                                 valid_bits |= PTU_PTE_LAST;
4147
4148                         rmem->pg_tbl[i] = rte_cpu_to_le_64(rmem->dma_arr[i] |
4149                                                            valid_bits);
4150                 }
4151         }
4152
4153         rmem->mz = mz;
4154         if (rmem->vmem_size)
4155                 rmem->vmem = (void **)mz->addr;
4156         rmem->dma_arr[0] = mz_phys_addr;
4157         return 0;
4158 }
4159
4160 static void bnxt_free_ctx_mem(struct bnxt *bp)
4161 {
4162         int i;
4163
4164         if (!bp->ctx || !(bp->ctx->flags & BNXT_CTX_FLAG_INITED))
4165                 return;
4166
4167         bp->ctx->flags &= ~BNXT_CTX_FLAG_INITED;
4168         rte_memzone_free(bp->ctx->qp_mem.ring_mem.mz);
4169         rte_memzone_free(bp->ctx->srq_mem.ring_mem.mz);
4170         rte_memzone_free(bp->ctx->cq_mem.ring_mem.mz);
4171         rte_memzone_free(bp->ctx->vnic_mem.ring_mem.mz);
4172         rte_memzone_free(bp->ctx->stat_mem.ring_mem.mz);
4173         rte_memzone_free(bp->ctx->qp_mem.ring_mem.pg_tbl_mz);
4174         rte_memzone_free(bp->ctx->srq_mem.ring_mem.pg_tbl_mz);
4175         rte_memzone_free(bp->ctx->cq_mem.ring_mem.pg_tbl_mz);
4176         rte_memzone_free(bp->ctx->vnic_mem.ring_mem.pg_tbl_mz);
4177         rte_memzone_free(bp->ctx->stat_mem.ring_mem.pg_tbl_mz);
4178
4179         for (i = 0; i < bp->ctx->tqm_fp_rings_count + 1; i++) {
4180                 if (bp->ctx->tqm_mem[i])
4181                         rte_memzone_free(bp->ctx->tqm_mem[i]->ring_mem.mz);
4182         }
4183
4184         rte_free(bp->ctx);
4185         bp->ctx = NULL;
4186 }
4187
4188 #define bnxt_roundup(x, y)   ((((x) + ((y) - 1)) / (y)) * (y))
4189
4190 #define min_t(type, x, y) ({                    \
4191         type __min1 = (x);                      \
4192         type __min2 = (y);                      \
4193         __min1 < __min2 ? __min1 : __min2; })
4194
4195 #define max_t(type, x, y) ({                    \
4196         type __max1 = (x);                      \
4197         type __max2 = (y);                      \
4198         __max1 > __max2 ? __max1 : __max2; })
4199
4200 #define clamp_t(type, _x, min, max)     min_t(type, max_t(type, _x, min), max)
4201
4202 int bnxt_alloc_ctx_mem(struct bnxt *bp)
4203 {
4204         struct bnxt_ctx_pg_info *ctx_pg;
4205         struct bnxt_ctx_mem_info *ctx;
4206         uint32_t mem_size, ena, entries;
4207         uint32_t entries_sp, min;
4208         int i, rc;
4209
4210         rc = bnxt_hwrm_func_backing_store_qcaps(bp);
4211         if (rc) {
4212                 PMD_DRV_LOG(ERR, "Query context mem capability failed\n");
4213                 return rc;
4214         }
4215         ctx = bp->ctx;
4216         if (!ctx || (ctx->flags & BNXT_CTX_FLAG_INITED))
4217                 return 0;
4218
4219         ctx_pg = &ctx->qp_mem;
4220         ctx_pg->entries = ctx->qp_min_qp1_entries + ctx->qp_max_l2_entries;
4221         mem_size = ctx->qp_entry_size * ctx_pg->entries;
4222         rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "qp_mem", 0);
4223         if (rc)
4224                 return rc;
4225
4226         ctx_pg = &ctx->srq_mem;
4227         ctx_pg->entries = ctx->srq_max_l2_entries;
4228         mem_size = ctx->srq_entry_size * ctx_pg->entries;
4229         rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "srq_mem", 0);
4230         if (rc)
4231                 return rc;
4232
4233         ctx_pg = &ctx->cq_mem;
4234         ctx_pg->entries = ctx->cq_max_l2_entries;
4235         mem_size = ctx->cq_entry_size * ctx_pg->entries;
4236         rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "cq_mem", 0);
4237         if (rc)
4238                 return rc;
4239
4240         ctx_pg = &ctx->vnic_mem;
4241         ctx_pg->entries = ctx->vnic_max_vnic_entries +
4242                 ctx->vnic_max_ring_table_entries;
4243         mem_size = ctx->vnic_entry_size * ctx_pg->entries;
4244         rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "vnic_mem", 0);
4245         if (rc)
4246                 return rc;
4247
4248         ctx_pg = &ctx->stat_mem;
4249         ctx_pg->entries = ctx->stat_max_entries;
4250         mem_size = ctx->stat_entry_size * ctx_pg->entries;
4251         rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "stat_mem", 0);
4252         if (rc)
4253                 return rc;
4254
4255         min = ctx->tqm_min_entries_per_ring;
4256
4257         entries_sp = ctx->qp_max_l2_entries +
4258                      ctx->vnic_max_vnic_entries +
4259                      2 * ctx->qp_min_qp1_entries + min;
4260         entries_sp = bnxt_roundup(entries_sp, ctx->tqm_entries_multiple);
4261
4262         entries = ctx->qp_max_l2_entries + ctx->qp_min_qp1_entries;
4263         entries = bnxt_roundup(entries, ctx->tqm_entries_multiple);
4264         entries = clamp_t(uint32_t, entries, min,
4265                           ctx->tqm_max_entries_per_ring);
4266         for (i = 0, ena = 0; i < ctx->tqm_fp_rings_count + 1; i++) {
4267                 ctx_pg = ctx->tqm_mem[i];
4268                 ctx_pg->entries = i ? entries : entries_sp;
4269                 mem_size = ctx->tqm_entry_size * ctx_pg->entries;
4270                 rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "tqm_mem", i);
4271                 if (rc)
4272                         return rc;
4273                 ena |= HWRM_FUNC_BACKING_STORE_CFG_INPUT_ENABLES_TQM_SP << i;
4274         }
4275
4276         ena |= FUNC_BACKING_STORE_CFG_INPUT_DFLT_ENABLES;
4277         rc = bnxt_hwrm_func_backing_store_cfg(bp, ena);
4278         if (rc)
4279                 PMD_DRV_LOG(ERR,
4280                             "Failed to configure context mem: rc = %d\n", rc);
4281         else
4282                 ctx->flags |= BNXT_CTX_FLAG_INITED;
4283
4284         return rc;
4285 }
4286
4287 static int bnxt_alloc_stats_mem(struct bnxt *bp)
4288 {
4289         struct rte_pci_device *pci_dev = bp->pdev;
4290         char mz_name[RTE_MEMZONE_NAMESIZE];
4291         const struct rte_memzone *mz = NULL;
4292         uint32_t total_alloc_len;
4293         rte_iova_t mz_phys_addr;
4294
4295         if (pci_dev->id.device_id == BROADCOM_DEV_ID_NS2)
4296                 return 0;
4297
4298         snprintf(mz_name, RTE_MEMZONE_NAMESIZE,
4299                  "bnxt_" PCI_PRI_FMT "-%s", pci_dev->addr.domain,
4300                  pci_dev->addr.bus, pci_dev->addr.devid,
4301                  pci_dev->addr.function, "rx_port_stats");
4302         mz_name[RTE_MEMZONE_NAMESIZE - 1] = 0;
4303         mz = rte_memzone_lookup(mz_name);
4304         total_alloc_len =
4305                 RTE_CACHE_LINE_ROUNDUP(sizeof(struct rx_port_stats) +
4306                                        sizeof(struct rx_port_stats_ext) + 512);
4307         if (!mz) {
4308                 mz = rte_memzone_reserve(mz_name, total_alloc_len,
4309                                          SOCKET_ID_ANY,
4310                                          RTE_MEMZONE_2MB |
4311                                          RTE_MEMZONE_SIZE_HINT_ONLY |
4312                                          RTE_MEMZONE_IOVA_CONTIG);
4313                 if (mz == NULL)
4314                         return -ENOMEM;
4315         }
4316         memset(mz->addr, 0, mz->len);
4317         mz_phys_addr = mz->iova;
4318
4319         bp->rx_mem_zone = (const void *)mz;
4320         bp->hw_rx_port_stats = mz->addr;
4321         bp->hw_rx_port_stats_map = mz_phys_addr;
4322
4323         snprintf(mz_name, RTE_MEMZONE_NAMESIZE,
4324                  "bnxt_" PCI_PRI_FMT "-%s", pci_dev->addr.domain,
4325                  pci_dev->addr.bus, pci_dev->addr.devid,
4326                  pci_dev->addr.function, "tx_port_stats");
4327         mz_name[RTE_MEMZONE_NAMESIZE - 1] = 0;
4328         mz = rte_memzone_lookup(mz_name);
4329         total_alloc_len =
4330                 RTE_CACHE_LINE_ROUNDUP(sizeof(struct tx_port_stats) +
4331                                        sizeof(struct tx_port_stats_ext) + 512);
4332         if (!mz) {
4333                 mz = rte_memzone_reserve(mz_name,
4334                                          total_alloc_len,
4335                                          SOCKET_ID_ANY,
4336                                          RTE_MEMZONE_2MB |
4337                                          RTE_MEMZONE_SIZE_HINT_ONLY |
4338                                          RTE_MEMZONE_IOVA_CONTIG);
4339                 if (mz == NULL)
4340                         return -ENOMEM;
4341         }
4342         memset(mz->addr, 0, mz->len);
4343         mz_phys_addr = mz->iova;
4344
4345         bp->tx_mem_zone = (const void *)mz;
4346         bp->hw_tx_port_stats = mz->addr;
4347         bp->hw_tx_port_stats_map = mz_phys_addr;
4348         bp->flags |= BNXT_FLAG_PORT_STATS;
4349
4350         /* Display extended statistics if FW supports it */
4351         if (bp->hwrm_spec_code < HWRM_SPEC_CODE_1_8_4 ||
4352             bp->hwrm_spec_code == HWRM_SPEC_CODE_1_9_0 ||
4353             !(bp->flags & BNXT_FLAG_EXT_STATS_SUPPORTED))
4354                 return 0;
4355
4356         bp->hw_rx_port_stats_ext = (void *)
4357                 ((uint8_t *)bp->hw_rx_port_stats +
4358                  sizeof(struct rx_port_stats));
4359         bp->hw_rx_port_stats_ext_map = bp->hw_rx_port_stats_map +
4360                 sizeof(struct rx_port_stats);
4361         bp->flags |= BNXT_FLAG_EXT_RX_PORT_STATS;
4362
4363         if (bp->hwrm_spec_code < HWRM_SPEC_CODE_1_9_2 ||
4364             bp->flags & BNXT_FLAG_EXT_STATS_SUPPORTED) {
4365                 bp->hw_tx_port_stats_ext = (void *)
4366                         ((uint8_t *)bp->hw_tx_port_stats +
4367                          sizeof(struct tx_port_stats));
4368                 bp->hw_tx_port_stats_ext_map =
4369                         bp->hw_tx_port_stats_map +
4370                         sizeof(struct tx_port_stats);
4371                 bp->flags |= BNXT_FLAG_EXT_TX_PORT_STATS;
4372         }
4373
4374         return 0;
4375 }
4376
4377 static int bnxt_setup_mac_addr(struct rte_eth_dev *eth_dev)
4378 {
4379         struct bnxt *bp = eth_dev->data->dev_private;
4380         int rc = 0;
4381
4382         eth_dev->data->mac_addrs = rte_zmalloc("bnxt_mac_addr_tbl",
4383                                                RTE_ETHER_ADDR_LEN *
4384                                                bp->max_l2_ctx,
4385                                                0);
4386         if (eth_dev->data->mac_addrs == NULL) {
4387                 PMD_DRV_LOG(ERR, "Failed to alloc MAC addr tbl\n");
4388                 return -ENOMEM;
4389         }
4390
4391         if (!BNXT_HAS_DFLT_MAC_SET(bp)) {
4392                 if (BNXT_PF(bp))
4393                         return -EINVAL;
4394
4395                 /* Generate a random MAC address, if none was assigned by PF */
4396                 PMD_DRV_LOG(INFO, "VF MAC address not assigned by Host PF\n");
4397                 bnxt_eth_hw_addr_random(bp->mac_addr);
4398                 PMD_DRV_LOG(INFO,
4399                             "Assign random MAC:%02X:%02X:%02X:%02X:%02X:%02X\n",
4400                             bp->mac_addr[0], bp->mac_addr[1], bp->mac_addr[2],
4401                             bp->mac_addr[3], bp->mac_addr[4], bp->mac_addr[5]);
4402
4403                 rc = bnxt_hwrm_set_mac(bp);
4404                 if (rc)
4405                         return rc;
4406         }
4407
4408         /* Copy the permanent MAC from the FUNC_QCAPS response */
4409         memcpy(&eth_dev->data->mac_addrs[0], bp->mac_addr, RTE_ETHER_ADDR_LEN);
4410
4411         return rc;
4412 }
4413
4414 static int bnxt_restore_dflt_mac(struct bnxt *bp)
4415 {
4416         int rc = 0;
4417
4418         /* MAC is already configured in FW */
4419         if (BNXT_HAS_DFLT_MAC_SET(bp))
4420                 return 0;
4421
4422         /* Restore the old MAC configured */
4423         rc = bnxt_hwrm_set_mac(bp);
4424         if (rc)
4425                 PMD_DRV_LOG(ERR, "Failed to restore MAC address\n");
4426
4427         return rc;
4428 }
4429
4430 static void bnxt_config_vf_req_fwd(struct bnxt *bp)
4431 {
4432         if (!BNXT_PF(bp))
4433                 return;
4434
4435         memset(bp->pf->vf_req_fwd, 0, sizeof(bp->pf->vf_req_fwd));
4436
4437         if (!(bp->fw_cap & BNXT_FW_CAP_LINK_ADMIN))
4438                 BNXT_HWRM_CMD_TO_FORWARD(HWRM_PORT_PHY_QCFG);
4439         BNXT_HWRM_CMD_TO_FORWARD(HWRM_FUNC_CFG);
4440         BNXT_HWRM_CMD_TO_FORWARD(HWRM_FUNC_VF_CFG);
4441         BNXT_HWRM_CMD_TO_FORWARD(HWRM_CFA_L2_FILTER_ALLOC);
4442         BNXT_HWRM_CMD_TO_FORWARD(HWRM_OEM_CMD);
4443 }
4444
4445 uint16_t
4446 bnxt_get_svif(uint16_t port_id, bool func_svif,
4447               enum bnxt_ulp_intf_type type)
4448 {
4449         struct rte_eth_dev *eth_dev;
4450         struct bnxt *bp;
4451
4452         eth_dev = &rte_eth_devices[port_id];
4453         if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
4454                 struct bnxt_representor *vfr = eth_dev->data->dev_private;
4455                 if (!vfr)
4456                         return 0;
4457
4458                 if (type == BNXT_ULP_INTF_TYPE_VF_REP)
4459                         return vfr->svif;
4460
4461                 eth_dev = vfr->parent_dev;
4462         }
4463
4464         bp = eth_dev->data->dev_private;
4465
4466         return func_svif ? bp->func_svif : bp->port_svif;
4467 }
4468
4469 uint16_t
4470 bnxt_get_vnic_id(uint16_t port, enum bnxt_ulp_intf_type type)
4471 {
4472         struct rte_eth_dev *eth_dev;
4473         struct bnxt_vnic_info *vnic;
4474         struct bnxt *bp;
4475
4476         eth_dev = &rte_eth_devices[port];
4477         if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
4478                 struct bnxt_representor *vfr = eth_dev->data->dev_private;
4479                 if (!vfr)
4480                         return 0;
4481
4482                 if (type == BNXT_ULP_INTF_TYPE_VF_REP)
4483                         return vfr->dflt_vnic_id;
4484
4485                 eth_dev = vfr->parent_dev;
4486         }
4487
4488         bp = eth_dev->data->dev_private;
4489
4490         vnic = BNXT_GET_DEFAULT_VNIC(bp);
4491
4492         return vnic->fw_vnic_id;
4493 }
4494
4495 uint16_t
4496 bnxt_get_fw_func_id(uint16_t port, enum bnxt_ulp_intf_type type)
4497 {
4498         struct rte_eth_dev *eth_dev;
4499         struct bnxt *bp;
4500
4501         eth_dev = &rte_eth_devices[port];
4502         if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
4503                 struct bnxt_representor *vfr = eth_dev->data->dev_private;
4504                 if (!vfr)
4505                         return 0;
4506
4507                 if (type == BNXT_ULP_INTF_TYPE_VF_REP)
4508                         return vfr->fw_fid;
4509
4510                 eth_dev = vfr->parent_dev;
4511         }
4512
4513         bp = eth_dev->data->dev_private;
4514
4515         return bp->fw_fid;
4516 }
4517
4518 enum bnxt_ulp_intf_type
4519 bnxt_get_interface_type(uint16_t port)
4520 {
4521         struct rte_eth_dev *eth_dev;
4522         struct bnxt *bp;
4523
4524         eth_dev = &rte_eth_devices[port];
4525         if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev))
4526                 return BNXT_ULP_INTF_TYPE_VF_REP;
4527
4528         bp = eth_dev->data->dev_private;
4529         if (BNXT_PF(bp))
4530                 return BNXT_ULP_INTF_TYPE_PF;
4531         else if (BNXT_VF_IS_TRUSTED(bp))
4532                 return BNXT_ULP_INTF_TYPE_TRUSTED_VF;
4533         else if (BNXT_VF(bp))
4534                 return BNXT_ULP_INTF_TYPE_VF;
4535
4536         return BNXT_ULP_INTF_TYPE_INVALID;
4537 }
4538
4539 uint16_t
4540 bnxt_get_phy_port_id(uint16_t port_id)
4541 {
4542         struct bnxt_representor *vfr;
4543         struct rte_eth_dev *eth_dev;
4544         struct bnxt *bp;
4545
4546         eth_dev = &rte_eth_devices[port_id];
4547         if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
4548                 vfr = eth_dev->data->dev_private;
4549                 if (!vfr)
4550                         return 0;
4551
4552                 eth_dev = vfr->parent_dev;
4553         }
4554
4555         bp = eth_dev->data->dev_private;
4556
4557         return BNXT_PF(bp) ? bp->pf->port_id : bp->parent->port_id;
4558 }
4559
4560 uint16_t
4561 bnxt_get_parif(uint16_t port_id, enum bnxt_ulp_intf_type type)
4562 {
4563         struct rte_eth_dev *eth_dev;
4564         struct bnxt *bp;
4565
4566         eth_dev = &rte_eth_devices[port_id];
4567         if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
4568                 struct bnxt_representor *vfr = eth_dev->data->dev_private;
4569                 if (!vfr)
4570                         return 0;
4571
4572                 if (type == BNXT_ULP_INTF_TYPE_VF_REP)
4573                         return vfr->fw_fid - 1;
4574
4575                 eth_dev = vfr->parent_dev;
4576         }
4577
4578         bp = eth_dev->data->dev_private;
4579
4580         return BNXT_PF(bp) ? bp->fw_fid - 1 : bp->parent->fid - 1;
4581 }
4582
4583 uint16_t
4584 bnxt_get_vport(uint16_t port_id)
4585 {
4586         return (1 << bnxt_get_phy_port_id(port_id));
4587 }
4588
4589 static void bnxt_alloc_error_recovery_info(struct bnxt *bp)
4590 {
4591         struct bnxt_error_recovery_info *info = bp->recovery_info;
4592
4593         if (info) {
4594                 if (!(bp->fw_cap & BNXT_FW_CAP_HCOMM_FW_STATUS))
4595                         memset(info, 0, sizeof(*info));
4596                 return;
4597         }
4598
4599         if (!(bp->fw_cap & BNXT_FW_CAP_ERROR_RECOVERY))
4600                 return;
4601
4602         info = rte_zmalloc("bnxt_hwrm_error_recovery_qcfg",
4603                            sizeof(*info), 0);
4604         if (!info)
4605                 bp->fw_cap &= ~BNXT_FW_CAP_ERROR_RECOVERY;
4606
4607         bp->recovery_info = info;
4608 }
4609
4610 static void bnxt_check_fw_status(struct bnxt *bp)
4611 {
4612         uint32_t fw_status;
4613
4614         if (!(bp->recovery_info &&
4615               (bp->fw_cap & BNXT_FW_CAP_HCOMM_FW_STATUS)))
4616                 return;
4617
4618         fw_status = bnxt_read_fw_status_reg(bp, BNXT_FW_STATUS_REG);
4619         if (fw_status != BNXT_FW_STATUS_HEALTHY)
4620                 PMD_DRV_LOG(ERR, "Firmware not responding, status: %#x\n",
4621                             fw_status);
4622 }
4623
4624 static int bnxt_map_hcomm_fw_status_reg(struct bnxt *bp)
4625 {
4626         struct bnxt_error_recovery_info *info = bp->recovery_info;
4627         uint32_t status_loc;
4628         uint32_t sig_ver;
4629
4630         rte_write32(HCOMM_STATUS_STRUCT_LOC, (uint8_t *)bp->bar0 +
4631                     BNXT_GRCPF_REG_WINDOW_BASE_OUT + 4);
4632         sig_ver = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
4633                                    BNXT_GRCP_WINDOW_2_BASE +
4634                                    offsetof(struct hcomm_status,
4635                                             sig_ver)));
4636         /* If the signature is absent, then FW does not support this feature */
4637         if ((sig_ver & HCOMM_STATUS_SIGNATURE_MASK) !=
4638             HCOMM_STATUS_SIGNATURE_VAL)
4639                 return 0;
4640
4641         if (!info) {
4642                 info = rte_zmalloc("bnxt_hwrm_error_recovery_qcfg",
4643                                    sizeof(*info), 0);
4644                 if (!info)
4645                         return -ENOMEM;
4646                 bp->recovery_info = info;
4647         } else {
4648                 memset(info, 0, sizeof(*info));
4649         }
4650
4651         status_loc = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
4652                                       BNXT_GRCP_WINDOW_2_BASE +
4653                                       offsetof(struct hcomm_status,
4654                                                fw_status_loc)));
4655
4656         /* Only pre-map the FW health status GRC register */
4657         if (BNXT_FW_STATUS_REG_TYPE(status_loc) != BNXT_FW_STATUS_REG_TYPE_GRC)
4658                 return 0;
4659
4660         info->status_regs[BNXT_FW_STATUS_REG] = status_loc;
4661         info->mapped_status_regs[BNXT_FW_STATUS_REG] =
4662                 BNXT_GRCP_WINDOW_2_BASE + (status_loc & BNXT_GRCP_OFFSET_MASK);
4663
4664         rte_write32((status_loc & BNXT_GRCP_BASE_MASK), (uint8_t *)bp->bar0 +
4665                     BNXT_GRCPF_REG_WINDOW_BASE_OUT + 4);
4666
4667         bp->fw_cap |= BNXT_FW_CAP_HCOMM_FW_STATUS;
4668
4669         return 0;
4670 }
4671
4672 static int bnxt_init_fw(struct bnxt *bp)
4673 {
4674         uint16_t mtu;
4675         int rc = 0;
4676
4677         bp->fw_cap = 0;
4678
4679         rc = bnxt_map_hcomm_fw_status_reg(bp);
4680         if (rc)
4681                 return rc;
4682
4683         rc = bnxt_hwrm_ver_get(bp, DFLT_HWRM_CMD_TIMEOUT);
4684         if (rc) {
4685                 bnxt_check_fw_status(bp);
4686                 return rc;
4687         }
4688
4689         rc = bnxt_hwrm_func_reset(bp);
4690         if (rc)
4691                 return -EIO;
4692
4693         rc = bnxt_hwrm_vnic_qcaps(bp);
4694         if (rc)
4695                 return rc;
4696
4697         rc = bnxt_hwrm_queue_qportcfg(bp);
4698         if (rc)
4699                 return rc;
4700
4701         /* Get the MAX capabilities for this function.
4702          * This function also allocates context memory for TQM rings and
4703          * informs the firmware about this allocated backing store memory.
4704          */
4705         rc = bnxt_hwrm_func_qcaps(bp);
4706         if (rc)
4707                 return rc;
4708
4709         rc = bnxt_hwrm_func_qcfg(bp, &mtu);
4710         if (rc)
4711                 return rc;
4712
4713         bnxt_hwrm_port_mac_qcfg(bp);
4714
4715         bnxt_hwrm_parent_pf_qcfg(bp);
4716
4717         bnxt_hwrm_port_phy_qcaps(bp);
4718
4719         bnxt_alloc_error_recovery_info(bp);
4720         /* Get the adapter error recovery support info */
4721         rc = bnxt_hwrm_error_recovery_qcfg(bp);
4722         if (rc)
4723                 bp->fw_cap &= ~BNXT_FW_CAP_ERROR_RECOVERY;
4724
4725         bnxt_hwrm_port_led_qcaps(bp);
4726
4727         return 0;
4728 }
4729
4730 static int
4731 bnxt_init_locks(struct bnxt *bp)
4732 {
4733         int err;
4734
4735         err = pthread_mutex_init(&bp->flow_lock, NULL);
4736         if (err) {
4737                 PMD_DRV_LOG(ERR, "Unable to initialize flow_lock\n");
4738                 return err;
4739         }
4740
4741         err = pthread_mutex_init(&bp->def_cp_lock, NULL);
4742         if (err) {
4743                 PMD_DRV_LOG(ERR, "Unable to initialize def_cp_lock\n");
4744                 return err;
4745         }
4746
4747         err = pthread_mutex_init(&bp->health_check_lock, NULL);
4748         if (err)
4749                 PMD_DRV_LOG(ERR, "Unable to initialize health_check_lock\n");
4750         return err;
4751 }
4752
4753 static int bnxt_init_resources(struct bnxt *bp, bool reconfig_dev)
4754 {
4755         int rc = 0;
4756
4757         rc = bnxt_init_fw(bp);
4758         if (rc)
4759                 return rc;
4760
4761         if (!reconfig_dev) {
4762                 rc = bnxt_setup_mac_addr(bp->eth_dev);
4763                 if (rc)
4764                         return rc;
4765         } else {
4766                 rc = bnxt_restore_dflt_mac(bp);
4767                 if (rc)
4768                         return rc;
4769         }
4770
4771         bnxt_config_vf_req_fwd(bp);
4772
4773         rc = bnxt_hwrm_func_driver_register(bp);
4774         if (rc) {
4775                 PMD_DRV_LOG(ERR, "Failed to register driver");
4776                 return -EBUSY;
4777         }
4778
4779         if (BNXT_PF(bp)) {
4780                 if (bp->pdev->max_vfs) {
4781                         rc = bnxt_hwrm_allocate_vfs(bp, bp->pdev->max_vfs);
4782                         if (rc) {
4783                                 PMD_DRV_LOG(ERR, "Failed to allocate VFs\n");
4784                                 return rc;
4785                         }
4786                 } else {
4787                         rc = bnxt_hwrm_allocate_pf_only(bp);
4788                         if (rc) {
4789                                 PMD_DRV_LOG(ERR,
4790                                             "Failed to allocate PF resources");
4791                                 return rc;
4792                         }
4793                 }
4794         }
4795
4796         rc = bnxt_alloc_mem(bp, reconfig_dev);
4797         if (rc)
4798                 return rc;
4799
4800         rc = bnxt_setup_int(bp);
4801         if (rc)
4802                 return rc;
4803
4804         rc = bnxt_request_int(bp);
4805         if (rc)
4806                 return rc;
4807
4808         rc = bnxt_init_ctx_mem(bp);
4809         if (rc) {
4810                 PMD_DRV_LOG(ERR, "Failed to init adv_flow_counters\n");
4811                 return rc;
4812         }
4813
4814         rc = bnxt_init_locks(bp);
4815         if (rc)
4816                 return rc;
4817
4818         return 0;
4819 }
4820
4821 static int
4822 bnxt_parse_devarg_truflow(__rte_unused const char *key,
4823                           const char *value, void *opaque_arg)
4824 {
4825         struct bnxt *bp = opaque_arg;
4826         unsigned long truflow;
4827         char *end = NULL;
4828
4829         if (!value || !opaque_arg) {
4830                 PMD_DRV_LOG(ERR,
4831                             "Invalid parameter passed to truflow devargs.\n");
4832                 return -EINVAL;
4833         }
4834
4835         truflow = strtoul(value, &end, 10);
4836         if (end == NULL || *end != '\0' ||
4837             (truflow == ULONG_MAX && errno == ERANGE)) {
4838                 PMD_DRV_LOG(ERR,
4839                             "Invalid parameter passed to truflow devargs.\n");
4840                 return -EINVAL;
4841         }
4842
4843         if (BNXT_DEVARG_TRUFLOW_INVALID(truflow)) {
4844                 PMD_DRV_LOG(ERR,
4845                             "Invalid value passed to truflow devargs.\n");
4846                 return -EINVAL;
4847         }
4848
4849         if (truflow) {
4850                 bp->flags |= BNXT_FLAG_TRUFLOW_EN;
4851                 PMD_DRV_LOG(INFO, "Host-based truflow feature enabled.\n");
4852         } else {
4853                 bp->flags &= ~BNXT_FLAG_TRUFLOW_EN;
4854                 PMD_DRV_LOG(INFO, "Host-based truflow feature disabled.\n");
4855         }
4856
4857         return 0;
4858 }
4859
4860 static int
4861 bnxt_parse_devarg_flow_xstat(__rte_unused const char *key,
4862                              const char *value, void *opaque_arg)
4863 {
4864         struct bnxt *bp = opaque_arg;
4865         unsigned long flow_xstat;
4866         char *end = NULL;
4867
4868         if (!value || !opaque_arg) {
4869                 PMD_DRV_LOG(ERR,
4870                             "Invalid parameter passed to flow_xstat devarg.\n");
4871                 return -EINVAL;
4872         }
4873
4874         flow_xstat = strtoul(value, &end, 10);
4875         if (end == NULL || *end != '\0' ||
4876             (flow_xstat == ULONG_MAX && errno == ERANGE)) {
4877                 PMD_DRV_LOG(ERR,
4878                             "Invalid parameter passed to flow_xstat devarg.\n");
4879                 return -EINVAL;
4880         }
4881
4882         if (BNXT_DEVARG_FLOW_XSTAT_INVALID(flow_xstat)) {
4883                 PMD_DRV_LOG(ERR,
4884                             "Invalid value passed to flow_xstat devarg.\n");
4885                 return -EINVAL;
4886         }
4887
4888         bp->flags |= BNXT_FLAG_FLOW_XSTATS_EN;
4889         if (BNXT_FLOW_XSTATS_EN(bp))
4890                 PMD_DRV_LOG(INFO, "flow_xstat feature enabled.\n");
4891
4892         return 0;
4893 }
4894
4895 static int
4896 bnxt_parse_devarg_max_num_kflows(__rte_unused const char *key,
4897                                         const char *value, void *opaque_arg)
4898 {
4899         struct bnxt *bp = opaque_arg;
4900         unsigned long max_num_kflows;
4901         char *end = NULL;
4902
4903         if (!value || !opaque_arg) {
4904                 PMD_DRV_LOG(ERR,
4905                         "Invalid parameter passed to max_num_kflows devarg.\n");
4906                 return -EINVAL;
4907         }
4908
4909         max_num_kflows = strtoul(value, &end, 10);
4910         if (end == NULL || *end != '\0' ||
4911                 (max_num_kflows == ULONG_MAX && errno == ERANGE)) {
4912                 PMD_DRV_LOG(ERR,
4913                         "Invalid parameter passed to max_num_kflows devarg.\n");
4914                 return -EINVAL;
4915         }
4916
4917         if (bnxt_devarg_max_num_kflow_invalid(max_num_kflows)) {
4918                 PMD_DRV_LOG(ERR,
4919                         "Invalid value passed to max_num_kflows devarg.\n");
4920                 return -EINVAL;
4921         }
4922
4923         bp->max_num_kflows = max_num_kflows;
4924         if (bp->max_num_kflows)
4925                 PMD_DRV_LOG(INFO, "max_num_kflows set as %ldK.\n",
4926                                 max_num_kflows);
4927
4928         return 0;
4929 }
4930
4931 static int
4932 bnxt_parse_devarg_rep_is_pf(__rte_unused const char *key,
4933                             const char *value, void *opaque_arg)
4934 {
4935         struct bnxt_representor *vfr_bp = opaque_arg;
4936         unsigned long rep_is_pf;
4937         char *end = NULL;
4938
4939         if (!value || !opaque_arg) {
4940                 PMD_DRV_LOG(ERR,
4941                             "Invalid parameter passed to rep_is_pf devargs.\n");
4942                 return -EINVAL;
4943         }
4944
4945         rep_is_pf = strtoul(value, &end, 10);
4946         if (end == NULL || *end != '\0' ||
4947             (rep_is_pf == ULONG_MAX && errno == ERANGE)) {
4948                 PMD_DRV_LOG(ERR,
4949                             "Invalid parameter passed to rep_is_pf devargs.\n");
4950                 return -EINVAL;
4951         }
4952
4953         if (BNXT_DEVARG_REP_IS_PF_INVALID(rep_is_pf)) {
4954                 PMD_DRV_LOG(ERR,
4955                             "Invalid value passed to rep_is_pf devargs.\n");
4956                 return -EINVAL;
4957         }
4958
4959         vfr_bp->flags |= rep_is_pf;
4960         if (BNXT_REP_PF(vfr_bp))
4961                 PMD_DRV_LOG(INFO, "PF representor\n");
4962         else
4963                 PMD_DRV_LOG(INFO, "VF representor\n");
4964
4965         return 0;
4966 }
4967
4968 static int
4969 bnxt_parse_devarg_rep_based_pf(__rte_unused const char *key,
4970                                const char *value, void *opaque_arg)
4971 {
4972         struct bnxt_representor *vfr_bp = opaque_arg;
4973         unsigned long rep_based_pf;
4974         char *end = NULL;
4975
4976         if (!value || !opaque_arg) {
4977                 PMD_DRV_LOG(ERR,
4978                             "Invalid parameter passed to rep_based_pf "
4979                             "devargs.\n");
4980                 return -EINVAL;
4981         }
4982
4983         rep_based_pf = strtoul(value, &end, 10);
4984         if (end == NULL || *end != '\0' ||
4985             (rep_based_pf == ULONG_MAX && errno == ERANGE)) {
4986                 PMD_DRV_LOG(ERR,
4987                             "Invalid parameter passed to rep_based_pf "
4988                             "devargs.\n");
4989                 return -EINVAL;
4990         }
4991
4992         if (BNXT_DEVARG_REP_BASED_PF_INVALID(rep_based_pf)) {
4993                 PMD_DRV_LOG(ERR,
4994                             "Invalid value passed to rep_based_pf devargs.\n");
4995                 return -EINVAL;
4996         }
4997
4998         vfr_bp->rep_based_pf = rep_based_pf;
4999         vfr_bp->flags |= BNXT_REP_BASED_PF_VALID;
5000
5001         PMD_DRV_LOG(INFO, "rep-based-pf = %d\n", vfr_bp->rep_based_pf);
5002
5003         return 0;
5004 }
5005
5006 static int
5007 bnxt_parse_devarg_rep_q_r2f(__rte_unused const char *key,
5008                             const char *value, void *opaque_arg)
5009 {
5010         struct bnxt_representor *vfr_bp = opaque_arg;
5011         unsigned long rep_q_r2f;
5012         char *end = NULL;
5013
5014         if (!value || !opaque_arg) {
5015                 PMD_DRV_LOG(ERR,
5016                             "Invalid parameter passed to rep_q_r2f "
5017                             "devargs.\n");
5018                 return -EINVAL;
5019         }
5020
5021         rep_q_r2f = strtoul(value, &end, 10);
5022         if (end == NULL || *end != '\0' ||
5023             (rep_q_r2f == ULONG_MAX && errno == ERANGE)) {
5024                 PMD_DRV_LOG(ERR,
5025                             "Invalid parameter passed to rep_q_r2f "
5026                             "devargs.\n");
5027                 return -EINVAL;
5028         }
5029
5030         if (BNXT_DEVARG_REP_Q_R2F_INVALID(rep_q_r2f)) {
5031                 PMD_DRV_LOG(ERR,
5032                             "Invalid value passed to rep_q_r2f devargs.\n");
5033                 return -EINVAL;
5034         }
5035
5036         vfr_bp->rep_q_r2f = rep_q_r2f;
5037         vfr_bp->flags |= BNXT_REP_Q_R2F_VALID;
5038         PMD_DRV_LOG(INFO, "rep-q-r2f = %d\n", vfr_bp->rep_q_r2f);
5039
5040         return 0;
5041 }
5042
5043 static int
5044 bnxt_parse_devarg_rep_q_f2r(__rte_unused const char *key,
5045                             const char *value, void *opaque_arg)
5046 {
5047         struct bnxt_representor *vfr_bp = opaque_arg;
5048         unsigned long rep_q_f2r;
5049         char *end = NULL;
5050
5051         if (!value || !opaque_arg) {
5052                 PMD_DRV_LOG(ERR,
5053                             "Invalid parameter passed to rep_q_f2r "
5054                             "devargs.\n");
5055                 return -EINVAL;
5056         }
5057
5058         rep_q_f2r = strtoul(value, &end, 10);
5059         if (end == NULL || *end != '\0' ||
5060             (rep_q_f2r == ULONG_MAX && errno == ERANGE)) {
5061                 PMD_DRV_LOG(ERR,
5062                             "Invalid parameter passed to rep_q_f2r "
5063                             "devargs.\n");
5064                 return -EINVAL;
5065         }
5066
5067         if (BNXT_DEVARG_REP_Q_F2R_INVALID(rep_q_f2r)) {
5068                 PMD_DRV_LOG(ERR,
5069                             "Invalid value passed to rep_q_f2r devargs.\n");
5070                 return -EINVAL;
5071         }
5072
5073         vfr_bp->rep_q_f2r = rep_q_f2r;
5074         vfr_bp->flags |= BNXT_REP_Q_F2R_VALID;
5075         PMD_DRV_LOG(INFO, "rep-q-f2r = %d\n", vfr_bp->rep_q_f2r);
5076
5077         return 0;
5078 }
5079
5080 static int
5081 bnxt_parse_devarg_rep_fc_r2f(__rte_unused const char *key,
5082                              const char *value, void *opaque_arg)
5083 {
5084         struct bnxt_representor *vfr_bp = opaque_arg;
5085         unsigned long rep_fc_r2f;
5086         char *end = NULL;
5087
5088         if (!value || !opaque_arg) {
5089                 PMD_DRV_LOG(ERR,
5090                             "Invalid parameter passed to rep_fc_r2f "
5091                             "devargs.\n");
5092                 return -EINVAL;
5093         }
5094
5095         rep_fc_r2f = strtoul(value, &end, 10);
5096         if (end == NULL || *end != '\0' ||
5097             (rep_fc_r2f == ULONG_MAX && errno == ERANGE)) {
5098                 PMD_DRV_LOG(ERR,
5099                             "Invalid parameter passed to rep_fc_r2f "
5100                             "devargs.\n");
5101                 return -EINVAL;
5102         }
5103
5104         if (BNXT_DEVARG_REP_FC_R2F_INVALID(rep_fc_r2f)) {
5105                 PMD_DRV_LOG(ERR,
5106                             "Invalid value passed to rep_fc_r2f devargs.\n");
5107                 return -EINVAL;
5108         }
5109
5110         vfr_bp->flags |= BNXT_REP_FC_R2F_VALID;
5111         vfr_bp->rep_fc_r2f = rep_fc_r2f;
5112         PMD_DRV_LOG(INFO, "rep-fc-r2f = %lu\n", rep_fc_r2f);
5113
5114         return 0;
5115 }
5116
5117 static int
5118 bnxt_parse_devarg_rep_fc_f2r(__rte_unused const char *key,
5119                              const char *value, void *opaque_arg)
5120 {
5121         struct bnxt_representor *vfr_bp = opaque_arg;
5122         unsigned long rep_fc_f2r;
5123         char *end = NULL;
5124
5125         if (!value || !opaque_arg) {
5126                 PMD_DRV_LOG(ERR,
5127                             "Invalid parameter passed to rep_fc_f2r "
5128                             "devargs.\n");
5129                 return -EINVAL;
5130         }
5131
5132         rep_fc_f2r = strtoul(value, &end, 10);
5133         if (end == NULL || *end != '\0' ||
5134             (rep_fc_f2r == ULONG_MAX && errno == ERANGE)) {
5135                 PMD_DRV_LOG(ERR,
5136                             "Invalid parameter passed to rep_fc_f2r "
5137                             "devargs.\n");
5138                 return -EINVAL;
5139         }
5140
5141         if (BNXT_DEVARG_REP_FC_F2R_INVALID(rep_fc_f2r)) {
5142                 PMD_DRV_LOG(ERR,
5143                             "Invalid value passed to rep_fc_f2r devargs.\n");
5144                 return -EINVAL;
5145         }
5146
5147         vfr_bp->flags |= BNXT_REP_FC_F2R_VALID;
5148         vfr_bp->rep_fc_f2r = rep_fc_f2r;
5149         PMD_DRV_LOG(INFO, "rep-fc-f2r = %lu\n", rep_fc_f2r);
5150
5151         return 0;
5152 }
5153
5154 static void
5155 bnxt_parse_dev_args(struct bnxt *bp, struct rte_devargs *devargs)
5156 {
5157         struct rte_kvargs *kvlist;
5158
5159         if (devargs == NULL)
5160                 return;
5161
5162         kvlist = rte_kvargs_parse(devargs->args, bnxt_dev_args);
5163         if (kvlist == NULL)
5164                 return;
5165
5166         /*
5167          * Handler for "truflow" devarg.
5168          * Invoked as for ex: "-a 0000:00:0d.0,host-based-truflow=1"
5169          */
5170         rte_kvargs_process(kvlist, BNXT_DEVARG_TRUFLOW,
5171                            bnxt_parse_devarg_truflow, bp);
5172
5173         /*
5174          * Handler for "flow_xstat" devarg.
5175          * Invoked as for ex: "-a 0000:00:0d.0,flow_xstat=1"
5176          */
5177         rte_kvargs_process(kvlist, BNXT_DEVARG_FLOW_XSTAT,
5178                            bnxt_parse_devarg_flow_xstat, bp);
5179
5180         /*
5181          * Handler for "max_num_kflows" devarg.
5182          * Invoked as for ex: "-a 000:00:0d.0,max_num_kflows=32"
5183          */
5184         rte_kvargs_process(kvlist, BNXT_DEVARG_MAX_NUM_KFLOWS,
5185                            bnxt_parse_devarg_max_num_kflows, bp);
5186
5187         rte_kvargs_free(kvlist);
5188 }
5189
5190 static int bnxt_alloc_switch_domain(struct bnxt *bp)
5191 {
5192         int rc = 0;
5193
5194         if (BNXT_PF(bp) || BNXT_VF_IS_TRUSTED(bp)) {
5195                 rc = rte_eth_switch_domain_alloc(&bp->switch_domain_id);
5196                 if (rc)
5197                         PMD_DRV_LOG(ERR,
5198                                     "Failed to alloc switch domain: %d\n", rc);
5199                 else
5200                         PMD_DRV_LOG(INFO,
5201                                     "Switch domain allocated %d\n",
5202                                     bp->switch_domain_id);
5203         }
5204
5205         return rc;
5206 }
5207
5208 static int
5209 bnxt_dev_init(struct rte_eth_dev *eth_dev, void *params __rte_unused)
5210 {
5211         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
5212         static int version_printed;
5213         struct bnxt *bp;
5214         int rc;
5215
5216         if (version_printed++ == 0)
5217                 PMD_DRV_LOG(INFO, "%s\n", bnxt_version);
5218
5219         eth_dev->dev_ops = &bnxt_dev_ops;
5220         eth_dev->rx_queue_count = bnxt_rx_queue_count_op;
5221         eth_dev->rx_descriptor_status = bnxt_rx_descriptor_status_op;
5222         eth_dev->tx_descriptor_status = bnxt_tx_descriptor_status_op;
5223         eth_dev->rx_pkt_burst = &bnxt_recv_pkts;
5224         eth_dev->tx_pkt_burst = &bnxt_xmit_pkts;
5225
5226         /*
5227          * For secondary processes, we don't initialise any further
5228          * as primary has already done this work.
5229          */
5230         if (rte_eal_process_type() != RTE_PROC_PRIMARY)
5231                 return 0;
5232
5233         rte_eth_copy_pci_info(eth_dev, pci_dev);
5234         eth_dev->data->dev_flags |= RTE_ETH_DEV_AUTOFILL_QUEUE_XSTATS;
5235
5236         bp = eth_dev->data->dev_private;
5237
5238         /* Parse dev arguments passed on when starting the DPDK application. */
5239         bnxt_parse_dev_args(bp, pci_dev->device.devargs);
5240
5241         bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
5242
5243         if (bnxt_vf_pciid(pci_dev->id.device_id))
5244                 bp->flags |= BNXT_FLAG_VF;
5245
5246         if (bnxt_p5_device(pci_dev->id.device_id))
5247                 bp->flags |= BNXT_FLAG_CHIP_P5;
5248
5249         if (pci_dev->id.device_id == BROADCOM_DEV_ID_58802 ||
5250             pci_dev->id.device_id == BROADCOM_DEV_ID_58804 ||
5251             pci_dev->id.device_id == BROADCOM_DEV_ID_58808 ||
5252             pci_dev->id.device_id == BROADCOM_DEV_ID_58802_VF)
5253                 bp->flags |= BNXT_FLAG_STINGRAY;
5254
5255         if (BNXT_TRUFLOW_EN(bp)) {
5256                 /* extra mbuf field is required to store CFA code from mark */
5257                 static const struct rte_mbuf_dynfield bnxt_cfa_code_dynfield_desc = {
5258                         .name = RTE_PMD_BNXT_CFA_CODE_DYNFIELD_NAME,
5259                         .size = sizeof(bnxt_cfa_code_dynfield_t),
5260                         .align = __alignof__(bnxt_cfa_code_dynfield_t),
5261                 };
5262                 bnxt_cfa_code_dynfield_offset =
5263                         rte_mbuf_dynfield_register(&bnxt_cfa_code_dynfield_desc);
5264                 if (bnxt_cfa_code_dynfield_offset < 0) {
5265                         PMD_DRV_LOG(ERR,
5266                             "Failed to register mbuf field for TruFlow mark\n");
5267                         return -rte_errno;
5268                 }
5269         }
5270
5271         rc = bnxt_init_board(eth_dev);
5272         if (rc) {
5273                 PMD_DRV_LOG(ERR,
5274                             "Failed to initialize board rc: %x\n", rc);
5275                 return rc;
5276         }
5277
5278         rc = bnxt_alloc_pf_info(bp);
5279         if (rc)
5280                 goto error_free;
5281
5282         rc = bnxt_alloc_link_info(bp);
5283         if (rc)
5284                 goto error_free;
5285
5286         rc = bnxt_alloc_parent_info(bp);
5287         if (rc)
5288                 goto error_free;
5289
5290         rc = bnxt_alloc_hwrm_resources(bp);
5291         if (rc) {
5292                 PMD_DRV_LOG(ERR,
5293                             "Failed to allocate hwrm resource rc: %x\n", rc);
5294                 goto error_free;
5295         }
5296         rc = bnxt_alloc_leds_info(bp);
5297         if (rc)
5298                 goto error_free;
5299
5300         rc = bnxt_alloc_cos_queues(bp);
5301         if (rc)
5302                 goto error_free;
5303
5304         rc = bnxt_init_resources(bp, false);
5305         if (rc)
5306                 goto error_free;
5307
5308         rc = bnxt_alloc_stats_mem(bp);
5309         if (rc)
5310                 goto error_free;
5311
5312         bnxt_alloc_switch_domain(bp);
5313
5314         PMD_DRV_LOG(INFO,
5315                     DRV_MODULE_NAME "found at mem %" PRIX64 ", node addr %pM\n",
5316                     pci_dev->mem_resource[0].phys_addr,
5317                     pci_dev->mem_resource[0].addr);
5318
5319         return 0;
5320
5321 error_free:
5322         bnxt_dev_uninit(eth_dev);
5323         return rc;
5324 }
5325
5326
5327 static void bnxt_free_ctx_mem_buf(struct bnxt_ctx_mem_buf_info *ctx)
5328 {
5329         if (!ctx)
5330                 return;
5331
5332         if (ctx->va)
5333                 rte_free(ctx->va);
5334
5335         ctx->va = NULL;
5336         ctx->dma = RTE_BAD_IOVA;
5337         ctx->ctx_id = BNXT_CTX_VAL_INVAL;
5338 }
5339
5340 static void bnxt_unregister_fc_ctx_mem(struct bnxt *bp)
5341 {
5342         bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_RX,
5343                                   CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
5344                                   bp->flow_stat->rx_fc_out_tbl.ctx_id,
5345                                   bp->flow_stat->max_fc,
5346                                   false);
5347
5348         bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_TX,
5349                                   CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
5350                                   bp->flow_stat->tx_fc_out_tbl.ctx_id,
5351                                   bp->flow_stat->max_fc,
5352                                   false);
5353
5354         if (bp->flow_stat->rx_fc_in_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
5355                 bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->rx_fc_in_tbl.ctx_id);
5356         bp->flow_stat->rx_fc_in_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
5357
5358         if (bp->flow_stat->rx_fc_out_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
5359                 bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->rx_fc_out_tbl.ctx_id);
5360         bp->flow_stat->rx_fc_out_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
5361
5362         if (bp->flow_stat->tx_fc_in_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
5363                 bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->tx_fc_in_tbl.ctx_id);
5364         bp->flow_stat->tx_fc_in_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
5365
5366         if (bp->flow_stat->tx_fc_out_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
5367                 bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->tx_fc_out_tbl.ctx_id);
5368         bp->flow_stat->tx_fc_out_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
5369 }
5370
5371 static void bnxt_uninit_fc_ctx_mem(struct bnxt *bp)
5372 {
5373         bnxt_unregister_fc_ctx_mem(bp);
5374
5375         bnxt_free_ctx_mem_buf(&bp->flow_stat->rx_fc_in_tbl);
5376         bnxt_free_ctx_mem_buf(&bp->flow_stat->rx_fc_out_tbl);
5377         bnxt_free_ctx_mem_buf(&bp->flow_stat->tx_fc_in_tbl);
5378         bnxt_free_ctx_mem_buf(&bp->flow_stat->tx_fc_out_tbl);
5379 }
5380
5381 static void bnxt_uninit_ctx_mem(struct bnxt *bp)
5382 {
5383         if (BNXT_FLOW_XSTATS_EN(bp))
5384                 bnxt_uninit_fc_ctx_mem(bp);
5385 }
5386
5387 static void
5388 bnxt_free_error_recovery_info(struct bnxt *bp)
5389 {
5390         rte_free(bp->recovery_info);
5391         bp->recovery_info = NULL;
5392         bp->fw_cap &= ~BNXT_FW_CAP_ERROR_RECOVERY;
5393 }
5394
5395 static void
5396 bnxt_uninit_locks(struct bnxt *bp)
5397 {
5398         pthread_mutex_destroy(&bp->flow_lock);
5399         pthread_mutex_destroy(&bp->def_cp_lock);
5400         pthread_mutex_destroy(&bp->health_check_lock);
5401         if (bp->rep_info) {
5402                 pthread_mutex_destroy(&bp->rep_info->vfr_lock);
5403                 pthread_mutex_destroy(&bp->rep_info->vfr_start_lock);
5404         }
5405 }
5406
5407 static int
5408 bnxt_uninit_resources(struct bnxt *bp, bool reconfig_dev)
5409 {
5410         int rc;
5411
5412         bnxt_free_int(bp);
5413         bnxt_free_mem(bp, reconfig_dev);
5414
5415         bnxt_hwrm_func_buf_unrgtr(bp);
5416         rte_free(bp->pf->vf_req_buf);
5417
5418         rc = bnxt_hwrm_func_driver_unregister(bp, 0);
5419         bp->flags &= ~BNXT_FLAG_REGISTERED;
5420         bnxt_free_ctx_mem(bp);
5421         if (!reconfig_dev) {
5422                 bnxt_free_hwrm_resources(bp);
5423                 bnxt_free_error_recovery_info(bp);
5424         }
5425
5426         bnxt_uninit_ctx_mem(bp);
5427
5428         bnxt_uninit_locks(bp);
5429         bnxt_free_flow_stats_info(bp);
5430         bnxt_free_rep_info(bp);
5431         rte_free(bp->ptp_cfg);
5432         bp->ptp_cfg = NULL;
5433         return rc;
5434 }
5435
5436 static int
5437 bnxt_dev_uninit(struct rte_eth_dev *eth_dev)
5438 {
5439         if (rte_eal_process_type() != RTE_PROC_PRIMARY)
5440                 return -EPERM;
5441
5442         PMD_DRV_LOG(DEBUG, "Calling Device uninit\n");
5443
5444         if (eth_dev->state != RTE_ETH_DEV_UNUSED)
5445                 bnxt_dev_close_op(eth_dev);
5446
5447         return 0;
5448 }
5449
5450 static int bnxt_pci_remove_dev_with_reps(struct rte_eth_dev *eth_dev)
5451 {
5452         struct bnxt *bp = eth_dev->data->dev_private;
5453         struct rte_eth_dev *vf_rep_eth_dev;
5454         int ret = 0, i;
5455
5456         if (!bp)
5457                 return -EINVAL;
5458
5459         for (i = 0; i < bp->num_reps; i++) {
5460                 vf_rep_eth_dev = bp->rep_info[i].vfr_eth_dev;
5461                 if (!vf_rep_eth_dev)
5462                         continue;
5463                 PMD_DRV_LOG(DEBUG, "BNXT Port:%d VFR pci remove\n",
5464                             vf_rep_eth_dev->data->port_id);
5465                 rte_eth_dev_destroy(vf_rep_eth_dev, bnxt_representor_uninit);
5466         }
5467         PMD_DRV_LOG(DEBUG, "BNXT Port:%d pci remove\n",
5468                     eth_dev->data->port_id);
5469         ret = rte_eth_dev_destroy(eth_dev, bnxt_dev_uninit);
5470
5471         return ret;
5472 }
5473
5474 static void bnxt_free_rep_info(struct bnxt *bp)
5475 {
5476         rte_free(bp->rep_info);
5477         bp->rep_info = NULL;
5478         rte_free(bp->cfa_code_map);
5479         bp->cfa_code_map = NULL;
5480 }
5481
5482 static int bnxt_init_rep_info(struct bnxt *bp)
5483 {
5484         int i = 0, rc;
5485
5486         if (bp->rep_info)
5487                 return 0;
5488
5489         bp->rep_info = rte_zmalloc("bnxt_rep_info",
5490                                    sizeof(bp->rep_info[0]) * BNXT_MAX_VF_REPS,
5491                                    0);
5492         if (!bp->rep_info) {
5493                 PMD_DRV_LOG(ERR, "Failed to alloc memory for rep info\n");
5494                 return -ENOMEM;
5495         }
5496         bp->cfa_code_map = rte_zmalloc("bnxt_cfa_code_map",
5497                                        sizeof(*bp->cfa_code_map) *
5498                                        BNXT_MAX_CFA_CODE, 0);
5499         if (!bp->cfa_code_map) {
5500                 PMD_DRV_LOG(ERR, "Failed to alloc memory for cfa_code_map\n");
5501                 bnxt_free_rep_info(bp);
5502                 return -ENOMEM;
5503         }
5504
5505         for (i = 0; i < BNXT_MAX_CFA_CODE; i++)
5506                 bp->cfa_code_map[i] = BNXT_VF_IDX_INVALID;
5507
5508         rc = pthread_mutex_init(&bp->rep_info->vfr_lock, NULL);
5509         if (rc) {
5510                 PMD_DRV_LOG(ERR, "Unable to initialize vfr_lock\n");
5511                 bnxt_free_rep_info(bp);
5512                 return rc;
5513         }
5514
5515         rc = pthread_mutex_init(&bp->rep_info->vfr_start_lock, NULL);
5516         if (rc) {
5517                 PMD_DRV_LOG(ERR, "Unable to initialize vfr_start_lock\n");
5518                 bnxt_free_rep_info(bp);
5519                 return rc;
5520         }
5521
5522         return rc;
5523 }
5524
5525 static int bnxt_rep_port_probe(struct rte_pci_device *pci_dev,
5526                                struct rte_eth_devargs *eth_da,
5527                                struct rte_eth_dev *backing_eth_dev,
5528                                const char *dev_args)
5529 {
5530         struct rte_eth_dev *vf_rep_eth_dev;
5531         char name[RTE_ETH_NAME_MAX_LEN];
5532         struct bnxt *backing_bp;
5533         uint16_t num_rep;
5534         int i, ret = 0;
5535         struct rte_kvargs *kvlist = NULL;
5536
5537         num_rep = eth_da->nb_representor_ports;
5538         if (num_rep > BNXT_MAX_VF_REPS) {
5539                 PMD_DRV_LOG(ERR, "nb_representor_ports = %d > %d MAX VF REPS\n",
5540                             num_rep, BNXT_MAX_VF_REPS);
5541                 return -EINVAL;
5542         }
5543
5544         if (num_rep >= RTE_MAX_ETHPORTS) {
5545                 PMD_DRV_LOG(ERR,
5546                             "nb_representor_ports = %d > %d MAX ETHPORTS\n",
5547                             num_rep, RTE_MAX_ETHPORTS);
5548                 return -EINVAL;
5549         }
5550
5551         backing_bp = backing_eth_dev->data->dev_private;
5552
5553         if (!(BNXT_PF(backing_bp) || BNXT_VF_IS_TRUSTED(backing_bp))) {
5554                 PMD_DRV_LOG(ERR,
5555                             "Not a PF or trusted VF. No Representor support\n");
5556                 /* Returning an error is not an option.
5557                  * Applications are not handling this correctly
5558                  */
5559                 return 0;
5560         }
5561
5562         if (bnxt_init_rep_info(backing_bp))
5563                 return 0;
5564
5565         for (i = 0; i < num_rep; i++) {
5566                 struct bnxt_representor representor = {
5567                         .vf_id = eth_da->representor_ports[i],
5568                         .switch_domain_id = backing_bp->switch_domain_id,
5569                         .parent_dev = backing_eth_dev
5570                 };
5571
5572                 if (representor.vf_id >= BNXT_MAX_VF_REPS) {
5573                         PMD_DRV_LOG(ERR, "VF-Rep id %d >= %d MAX VF ID\n",
5574                                     representor.vf_id, BNXT_MAX_VF_REPS);
5575                         continue;
5576                 }
5577
5578                 /* representor port net_bdf_port */
5579                 snprintf(name, sizeof(name), "net_%s_representor_%d",
5580                          pci_dev->device.name, eth_da->representor_ports[i]);
5581
5582                 kvlist = rte_kvargs_parse(dev_args, bnxt_dev_args);
5583                 if (kvlist) {
5584                         /*
5585                          * Handler for "rep_is_pf" devarg.
5586                          * Invoked as for ex: "-a 000:00:0d.0,
5587                          * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
5588                          */
5589                         ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_IS_PF,
5590                                                  bnxt_parse_devarg_rep_is_pf,
5591                                                  (void *)&representor);
5592                         if (ret) {
5593                                 ret = -EINVAL;
5594                                 goto err;
5595                         }
5596                         /*
5597                          * Handler for "rep_based_pf" devarg.
5598                          * Invoked as for ex: "-a 000:00:0d.0,
5599                          * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
5600                          */
5601                         ret = rte_kvargs_process(kvlist,
5602                                                  BNXT_DEVARG_REP_BASED_PF,
5603                                                  bnxt_parse_devarg_rep_based_pf,
5604                                                  (void *)&representor);
5605                         if (ret) {
5606                                 ret = -EINVAL;
5607                                 goto err;
5608                         }
5609                         /*
5610                          * Handler for "rep_based_pf" devarg.
5611                          * Invoked as for ex: "-a 000:00:0d.0,
5612                          * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
5613                          */
5614                         ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_Q_R2F,
5615                                                  bnxt_parse_devarg_rep_q_r2f,
5616                                                  (void *)&representor);
5617                         if (ret) {
5618                                 ret = -EINVAL;
5619                                 goto err;
5620                         }
5621                         /*
5622                          * Handler for "rep_based_pf" devarg.
5623                          * Invoked as for ex: "-a 000:00:0d.0,
5624                          * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
5625                          */
5626                         ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_Q_F2R,
5627                                                  bnxt_parse_devarg_rep_q_f2r,
5628                                                  (void *)&representor);
5629                         if (ret) {
5630                                 ret = -EINVAL;
5631                                 goto err;
5632                         }
5633                         /*
5634                          * Handler for "rep_based_pf" devarg.
5635                          * Invoked as for ex: "-a 000:00:0d.0,
5636                          * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
5637                          */
5638                         ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_FC_R2F,
5639                                                  bnxt_parse_devarg_rep_fc_r2f,
5640                                                  (void *)&representor);
5641                         if (ret) {
5642                                 ret = -EINVAL;
5643                                 goto err;
5644                         }
5645                         /*
5646                          * Handler for "rep_based_pf" devarg.
5647                          * Invoked as for ex: "-a 000:00:0d.0,
5648                          * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
5649                          */
5650                         ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_FC_F2R,
5651                                                  bnxt_parse_devarg_rep_fc_f2r,
5652                                                  (void *)&representor);
5653                         if (ret) {
5654                                 ret = -EINVAL;
5655                                 goto err;
5656                         }
5657                 }
5658
5659                 ret = rte_eth_dev_create(&pci_dev->device, name,
5660                                          sizeof(struct bnxt_representor),
5661                                          NULL, NULL,
5662                                          bnxt_representor_init,
5663                                          &representor);
5664                 if (ret) {
5665                         PMD_DRV_LOG(ERR, "failed to create bnxt vf "
5666                                     "representor %s.", name);
5667                         goto err;
5668                 }
5669
5670                 vf_rep_eth_dev = rte_eth_dev_allocated(name);
5671                 if (!vf_rep_eth_dev) {
5672                         PMD_DRV_LOG(ERR, "Failed to find the eth_dev"
5673                                     " for VF-Rep: %s.", name);
5674                         ret = -ENODEV;
5675                         goto err;
5676                 }
5677
5678                 PMD_DRV_LOG(DEBUG, "BNXT Port:%d VFR pci probe\n",
5679                             backing_eth_dev->data->port_id);
5680                 backing_bp->rep_info[representor.vf_id].vfr_eth_dev =
5681                                                          vf_rep_eth_dev;
5682                 backing_bp->num_reps++;
5683
5684         }
5685
5686         rte_kvargs_free(kvlist);
5687         return 0;
5688
5689 err:
5690         /* If num_rep > 1, then rollback already created
5691          * ports, since we'll be failing the probe anyway
5692          */
5693         if (num_rep > 1)
5694                 bnxt_pci_remove_dev_with_reps(backing_eth_dev);
5695         rte_errno = -ret;
5696         rte_kvargs_free(kvlist);
5697
5698         return ret;
5699 }
5700
5701 static int bnxt_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
5702                           struct rte_pci_device *pci_dev)
5703 {
5704         struct rte_eth_devargs eth_da = { .nb_representor_ports = 0 };
5705         struct rte_eth_dev *backing_eth_dev;
5706         uint16_t num_rep;
5707         int ret = 0;
5708
5709         if (pci_dev->device.devargs) {
5710                 ret = rte_eth_devargs_parse(pci_dev->device.devargs->args,
5711                                             &eth_da);
5712                 if (ret)
5713                         return ret;
5714         }
5715
5716         num_rep = eth_da.nb_representor_ports;
5717         PMD_DRV_LOG(DEBUG, "nb_representor_ports = %d\n",
5718                     num_rep);
5719
5720         /* We could come here after first level of probe is already invoked
5721          * as part of an application bringup(OVS-DPDK vswitchd), so first check
5722          * for already allocated eth_dev for the backing device (PF/Trusted VF)
5723          */
5724         backing_eth_dev = rte_eth_dev_allocated(pci_dev->device.name);
5725         if (backing_eth_dev == NULL) {
5726                 ret = rte_eth_dev_create(&pci_dev->device, pci_dev->device.name,
5727                                          sizeof(struct bnxt),
5728                                          eth_dev_pci_specific_init, pci_dev,
5729                                          bnxt_dev_init, NULL);
5730
5731                 if (ret || !num_rep)
5732                         return ret;
5733
5734                 backing_eth_dev = rte_eth_dev_allocated(pci_dev->device.name);
5735         }
5736         PMD_DRV_LOG(DEBUG, "BNXT Port:%d pci probe\n",
5737                     backing_eth_dev->data->port_id);
5738
5739         if (!num_rep)
5740                 return ret;
5741
5742         /* probe representor ports now */
5743         ret = bnxt_rep_port_probe(pci_dev, &eth_da, backing_eth_dev,
5744                                   pci_dev->device.devargs->args);
5745
5746         return ret;
5747 }
5748
5749 static int bnxt_pci_remove(struct rte_pci_device *pci_dev)
5750 {
5751         struct rte_eth_dev *eth_dev;
5752
5753         eth_dev = rte_eth_dev_allocated(pci_dev->device.name);
5754         if (!eth_dev)
5755                 return 0; /* Invoked typically only by OVS-DPDK, by the
5756                            * time it comes here the eth_dev is already
5757                            * deleted by rte_eth_dev_close(), so returning
5758                            * +ve value will at least help in proper cleanup
5759                            */
5760
5761         PMD_DRV_LOG(DEBUG, "BNXT Port:%d pci remove\n", eth_dev->data->port_id);
5762         if (rte_eal_process_type() == RTE_PROC_PRIMARY) {
5763                 if (eth_dev->data->dev_flags & RTE_ETH_DEV_REPRESENTOR)
5764                         return rte_eth_dev_destroy(eth_dev,
5765                                                    bnxt_representor_uninit);
5766                 else
5767                         return rte_eth_dev_destroy(eth_dev,
5768                                                    bnxt_dev_uninit);
5769         } else {
5770                 return rte_eth_dev_pci_generic_remove(pci_dev, NULL);
5771         }
5772 }
5773
5774 static struct rte_pci_driver bnxt_rte_pmd = {
5775         .id_table = bnxt_pci_id_map,
5776         .drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC |
5777                         RTE_PCI_DRV_PROBE_AGAIN, /* Needed in case of VF-REPs
5778                                                   * and OVS-DPDK
5779                                                   */
5780         .probe = bnxt_pci_probe,
5781         .remove = bnxt_pci_remove,
5782 };
5783
5784 static bool
5785 is_device_supported(struct rte_eth_dev *dev, struct rte_pci_driver *drv)
5786 {
5787         if (strcmp(dev->device->driver->name, drv->driver.name))
5788                 return false;
5789
5790         return true;
5791 }
5792
5793 bool is_bnxt_supported(struct rte_eth_dev *dev)
5794 {
5795         return is_device_supported(dev, &bnxt_rte_pmd);
5796 }
5797
5798 RTE_LOG_REGISTER(bnxt_logtype_driver, pmd.net.bnxt.driver, NOTICE);
5799 RTE_PMD_REGISTER_PCI(net_bnxt, bnxt_rte_pmd);
5800 RTE_PMD_REGISTER_PCI_TABLE(net_bnxt, bnxt_pci_id_map);
5801 RTE_PMD_REGISTER_KMOD_DEP(net_bnxt, "* igb_uio | uio_pci_generic | vfio-pci");