drivers/net/bnxt/bnxt_ethdev.c

   1 /* SPDX-License-Identifier: BSD-3-Clause
   2  * Copyright(c) 2014-2021 Broadcom
   3  * All rights reserved.
   4  */
   5
   6 #include <inttypes.h>
   7 #include <stdbool.h>
   8
   9 #include <rte_dev.h>
  10 #include <ethdev_driver.h>
  11 #include <ethdev_pci.h>
  12 #include <rte_malloc.h>
  13 #include <rte_cycles.h>
  14 #include <rte_alarm.h>
  15 #include <rte_kvargs.h>
  16 #include <rte_vect.h>
  17
  18 #include "bnxt.h"
  19 #include "bnxt_filter.h"
  20 #include "bnxt_hwrm.h"
  21 #include "bnxt_irq.h"
  22 #include "bnxt_reps.h"
  23 #include "bnxt_ring.h"
  24 #include "bnxt_rxq.h"
  25 #include "bnxt_rxr.h"
  26 #include "bnxt_stats.h"
  27 #include "bnxt_txq.h"
  28 #include "bnxt_txr.h"
  29 #include "bnxt_vnic.h"
  30 #include "hsi_struct_def_dpdk.h"
  31 #include "bnxt_nvm_defs.h"
  32 #include "bnxt_tf_common.h"
  33 #include "ulp_flow_db.h"
  34 #include "rte_pmd_bnxt.h"
  35
  36 #define DRV_MODULE_NAME         "bnxt"
  37 static const char bnxt_version[] =
  38         "Broadcom NetXtreme driver " DRV_MODULE_NAME;
  39
  40 /*
  41  * The set of PCI devices this driver supports
  42  */
  43 static const struct rte_pci_id bnxt_pci_id_map[] = {
  44         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM,
  45                          BROADCOM_DEV_ID_STRATUS_NIC_VF1) },
  46         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM,
  47                          BROADCOM_DEV_ID_STRATUS_NIC_VF2) },
  48         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_STRATUS_NIC) },
  49         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57414_VF) },
  50         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57304_VF) },
  51         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_NS2) },
  52         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57406_VF) },
  53         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57407_MF) },
  54         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_5741X_VF) },
  55         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_5731X_VF) },
  56         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57417_MF) },
  57         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57412) },
  58         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57414) },
  59         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57416_RJ45) },
  60         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57417_RJ45) },
  61         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57412_MF) },
  62         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57317_RJ45) },
  63         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57417_SFP) },
  64         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57416_SFP) },
  65         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57317_SFP) },
  66         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57414_MF) },
  67         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57416_MF) },
  68         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58802) },
  69         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58804) },
  70         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58808) },
  71         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58802_VF) },
  72         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57508) },
  73         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57504) },
  74         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57502) },
  75         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57500_VF1) },
  76         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57500_VF2) },
  77         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57508_MF1) },
  78         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57504_MF1) },
  79         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57502_MF1) },
  80         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57508_MF2) },
  81         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57504_MF2) },
  82         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57502_MF2) },
  83         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58812) },
  84         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58814) },
  85         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58818) },
  86         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58818_VF) },
  87         { .vendor_id = 0, /* sentinel */ },
  88 };
  89
  90 #define BNXT_DEVARG_TRUFLOW     "host-based-truflow"
  91 #define BNXT_DEVARG_FLOW_XSTAT  "flow-xstat"
  92 #define BNXT_DEVARG_MAX_NUM_KFLOWS  "max-num-kflows"
  93 #define BNXT_DEVARG_REPRESENTOR "representor"
  94 #define BNXT_DEVARG_REP_BASED_PF  "rep-based-pf"
  95 #define BNXT_DEVARG_REP_IS_PF  "rep-is-pf"
  96 #define BNXT_DEVARG_REP_Q_R2F  "rep-q-r2f"
  97 #define BNXT_DEVARG_REP_Q_F2R  "rep-q-f2r"
  98 #define BNXT_DEVARG_REP_FC_R2F  "rep-fc-r2f"
  99 #define BNXT_DEVARG_REP_FC_F2R  "rep-fc-f2r"
 100
 101 static const char *const bnxt_dev_args[] = {
 102         BNXT_DEVARG_REPRESENTOR,
 103         BNXT_DEVARG_TRUFLOW,
 104         BNXT_DEVARG_FLOW_XSTAT,
 105         BNXT_DEVARG_MAX_NUM_KFLOWS,
 106         BNXT_DEVARG_REP_BASED_PF,
 107         BNXT_DEVARG_REP_IS_PF,
 108         BNXT_DEVARG_REP_Q_R2F,
 109         BNXT_DEVARG_REP_Q_F2R,
 110         BNXT_DEVARG_REP_FC_R2F,
 111         BNXT_DEVARG_REP_FC_F2R,
 112         NULL
 113 };
 114
 115 /*
 116  * truflow == false to disable the feature
 117  * truflow == true to enable the feature
 118  */
 119 #define BNXT_DEVARG_TRUFLOW_INVALID(truflow)    ((truflow) > 1)
 120
 121 /*
 122  * flow_xstat == false to disable the feature
 123  * flow_xstat == true to enable the feature
 124  */
 125 #define BNXT_DEVARG_FLOW_XSTAT_INVALID(flow_xstat)      ((flow_xstat) > 1)
 126
 127 /*
 128  * rep_is_pf == false to indicate VF representor
 129  * rep_is_pf == true to indicate PF representor
 130  */
 131 #define BNXT_DEVARG_REP_IS_PF_INVALID(rep_is_pf)        ((rep_is_pf) > 1)
 132
 133 /*
 134  * rep_based_pf == Physical index of the PF
 135  */
 136 #define BNXT_DEVARG_REP_BASED_PF_INVALID(rep_based_pf)  ((rep_based_pf) > 15)
 137 /*
 138  * rep_q_r2f == Logical COS Queue index for the rep to endpoint direction
 139  */
 140 #define BNXT_DEVARG_REP_Q_R2F_INVALID(rep_q_r2f)        ((rep_q_r2f) > 3)
 141
 142 /*
 143  * rep_q_f2r == Logical COS Queue index for the endpoint to rep direction
 144  */
 145 #define BNXT_DEVARG_REP_Q_F2R_INVALID(rep_q_f2r)        ((rep_q_f2r) > 3)
 146
 147 /*
 148  * rep_fc_r2f == Flow control for the representor to endpoint direction
 149  */
 150 #define BNXT_DEVARG_REP_FC_R2F_INVALID(rep_fc_r2f)      ((rep_fc_r2f) > 1)
 151
 152 /*
 153  * rep_fc_f2r == Flow control for the endpoint to representor direction
 154  */
 155 #define BNXT_DEVARG_REP_FC_F2R_INVALID(rep_fc_f2r)      ((rep_fc_f2r) > 1)
 156
 157 int bnxt_cfa_code_dynfield_offset = -1;
 158
 159 /*
 160  * max_num_kflows must be >= 32
 161  * and must be a power-of-2 supported value
 162  * return: 1 -> invalid
 163  *         0 -> valid
 164  */
 165 static int bnxt_devarg_max_num_kflow_invalid(uint16_t max_num_kflows)
 166 {
 167         if (max_num_kflows < 32 || !rte_is_power_of_2(max_num_kflows))
 168                 return 1;
 169         return 0;
 170 }
 171
 172 static int bnxt_vlan_offload_set_op(struct rte_eth_dev *dev, int mask);
 173 static int bnxt_dev_uninit(struct rte_eth_dev *eth_dev);
 174 static int bnxt_init_resources(struct bnxt *bp, bool reconfig_dev);
 175 static int bnxt_uninit_resources(struct bnxt *bp, bool reconfig_dev);
 176 static void bnxt_cancel_fw_health_check(struct bnxt *bp);
 177 static int bnxt_restore_vlan_filters(struct bnxt *bp);
 178 static void bnxt_dev_recover(void *arg);
 179 static void bnxt_free_error_recovery_info(struct bnxt *bp);
 180 static void bnxt_free_rep_info(struct bnxt *bp);
 181
 182 int is_bnxt_in_error(struct bnxt *bp)
 183 {
 184         if (bp->flags & BNXT_FLAG_FATAL_ERROR)
 185                 return -EIO;
 186         if (bp->flags & BNXT_FLAG_FW_RESET)
 187                 return -EBUSY;
 188
 189         return 0;
 190 }
 191
 192 /***********************/
 193
 194 /*
 195  * High level utility functions
 196  */
 197
 198 static uint16_t bnxt_rss_ctxts(const struct bnxt *bp)
 199 {
 200         unsigned int num_rss_rings = RTE_MIN(bp->rx_nr_rings,
 201                                              BNXT_RSS_TBL_SIZE_P5);
 202
 203         if (!BNXT_CHIP_P5(bp))
 204                 return 1;
 205
 206         return RTE_ALIGN_MUL_CEIL(num_rss_rings,
 207                                   BNXT_RSS_ENTRIES_PER_CTX_P5) /
 208                                   BNXT_RSS_ENTRIES_PER_CTX_P5;
 209 }
 210
 211 uint16_t bnxt_rss_hash_tbl_size(const struct bnxt *bp)
 212 {
 213         if (!BNXT_CHIP_P5(bp))
 214                 return HW_HASH_INDEX_SIZE;
 215
 216         return bnxt_rss_ctxts(bp) * BNXT_RSS_ENTRIES_PER_CTX_P5;
 217 }
 218
 219 static void bnxt_free_parent_info(struct bnxt *bp)
 220 {
 221         rte_free(bp->parent);
 222 }
 223
 224 static void bnxt_free_pf_info(struct bnxt *bp)
 225 {
 226         rte_free(bp->pf);
 227 }
 228
 229 static void bnxt_free_link_info(struct bnxt *bp)
 230 {
 231         rte_free(bp->link_info);
 232 }
 233
 234 static void bnxt_free_leds_info(struct bnxt *bp)
 235 {
 236         if (BNXT_VF(bp))
 237                 return;
 238
 239         rte_free(bp->leds);
 240         bp->leds = NULL;
 241 }
 242
 243 static void bnxt_free_flow_stats_info(struct bnxt *bp)
 244 {
 245         rte_free(bp->flow_stat);
 246         bp->flow_stat = NULL;
 247 }
 248
 249 static void bnxt_free_cos_queues(struct bnxt *bp)
 250 {
 251         rte_free(bp->rx_cos_queue);
 252         rte_free(bp->tx_cos_queue);
 253 }
 254
 255 static void bnxt_free_mem(struct bnxt *bp, bool reconfig)
 256 {
 257         bnxt_free_filter_mem(bp);
 258         bnxt_free_vnic_attributes(bp);
 259         bnxt_free_vnic_mem(bp);
 260
 261         /* tx/rx rings are configured as part of *_queue_setup callbacks.
 262          * If the number of rings change across fw update,
 263          * we don't have much choice except to warn the user.
 264          */
 265         if (!reconfig) {
 266                 bnxt_free_stats(bp);
 267                 bnxt_free_tx_rings(bp);
 268                 bnxt_free_rx_rings(bp);
 269         }
 270         bnxt_free_async_cp_ring(bp);
 271         bnxt_free_rxtx_nq_ring(bp);
 272
 273         rte_free(bp->grp_info);
 274         bp->grp_info = NULL;
 275 }
 276
 277 static int bnxt_alloc_parent_info(struct bnxt *bp)
 278 {
 279         bp->parent = rte_zmalloc("bnxt_parent_info",
 280                                  sizeof(struct bnxt_parent_info), 0);
 281         if (bp->parent == NULL)
 282                 return -ENOMEM;
 283
 284         return 0;
 285 }
 286
 287 static int bnxt_alloc_pf_info(struct bnxt *bp)
 288 {
 289         bp->pf = rte_zmalloc("bnxt_pf_info", sizeof(struct bnxt_pf_info), 0);
 290         if (bp->pf == NULL)
 291                 return -ENOMEM;
 292
 293         return 0;
 294 }
 295
 296 static int bnxt_alloc_link_info(struct bnxt *bp)
 297 {
 298         bp->link_info =
 299                 rte_zmalloc("bnxt_link_info", sizeof(struct bnxt_link_info), 0);
 300         if (bp->link_info == NULL)
 301                 return -ENOMEM;
 302
 303         return 0;
 304 }
 305
 306 static int bnxt_alloc_leds_info(struct bnxt *bp)
 307 {
 308         if (BNXT_VF(bp))
 309                 return 0;
 310
 311         bp->leds = rte_zmalloc("bnxt_leds",
 312                                BNXT_MAX_LED * sizeof(struct bnxt_led_info),
 313                                0);
 314         if (bp->leds == NULL)
 315                 return -ENOMEM;
 316
 317         return 0;
 318 }
 319
 320 static int bnxt_alloc_cos_queues(struct bnxt *bp)
 321 {
 322         bp->rx_cos_queue =
 323                 rte_zmalloc("bnxt_rx_cosq",
 324                             BNXT_COS_QUEUE_COUNT *
 325                             sizeof(struct bnxt_cos_queue_info),
 326                             0);
 327         if (bp->rx_cos_queue == NULL)
 328                 return -ENOMEM;
 329
 330         bp->tx_cos_queue =
 331                 rte_zmalloc("bnxt_tx_cosq",
 332                             BNXT_COS_QUEUE_COUNT *
 333                             sizeof(struct bnxt_cos_queue_info),
 334                             0);
 335         if (bp->tx_cos_queue == NULL)
 336                 return -ENOMEM;
 337
 338         return 0;
 339 }
 340
 341 static int bnxt_alloc_flow_stats_info(struct bnxt *bp)
 342 {
 343         bp->flow_stat = rte_zmalloc("bnxt_flow_xstat",
 344                                     sizeof(struct bnxt_flow_stat_info), 0);
 345         if (bp->flow_stat == NULL)
 346                 return -ENOMEM;
 347
 348         return 0;
 349 }
 350
 351 static int bnxt_alloc_mem(struct bnxt *bp, bool reconfig)
 352 {
 353         int rc;
 354
 355         rc = bnxt_alloc_ring_grps(bp);
 356         if (rc)
 357                 goto alloc_mem_err;
 358
 359         rc = bnxt_alloc_async_ring_struct(bp);
 360         if (rc)
 361                 goto alloc_mem_err;
 362
 363         rc = bnxt_alloc_vnic_mem(bp);
 364         if (rc)
 365                 goto alloc_mem_err;
 366
 367         rc = bnxt_alloc_vnic_attributes(bp);
 368         if (rc)
 369                 goto alloc_mem_err;
 370
 371         rc = bnxt_alloc_filter_mem(bp);
 372         if (rc)
 373                 goto alloc_mem_err;
 374
 375         rc = bnxt_alloc_async_cp_ring(bp);
 376         if (rc)
 377                 goto alloc_mem_err;
 378
 379         rc = bnxt_alloc_rxtx_nq_ring(bp);
 380         if (rc)
 381                 goto alloc_mem_err;
 382
 383         if (BNXT_FLOW_XSTATS_EN(bp)) {
 384                 rc = bnxt_alloc_flow_stats_info(bp);
 385                 if (rc)
 386                         goto alloc_mem_err;
 387         }
 388
 389         return 0;
 390
 391 alloc_mem_err:
 392         bnxt_free_mem(bp, reconfig);
 393         return rc;
 394 }
 395
 396 static int bnxt_setup_one_vnic(struct bnxt *bp, uint16_t vnic_id)
 397 {
 398         struct rte_eth_conf *dev_conf = &bp->eth_dev->data->dev_conf;
 399         struct bnxt_vnic_info *vnic = &bp->vnic_info[vnic_id];
 400         uint64_t rx_offloads = dev_conf->rxmode.offloads;
 401         struct bnxt_rx_queue *rxq;
 402         unsigned int j;
 403         int rc;
 404
 405         rc = bnxt_vnic_grp_alloc(bp, vnic);
 406         if (rc)
 407                 goto err_out;
 408
 409         PMD_DRV_LOG(DEBUG, "vnic[%d] = %p vnic->fw_grp_ids = %p\n",
 410                     vnic_id, vnic, vnic->fw_grp_ids);
 411
 412         rc = bnxt_hwrm_vnic_alloc(bp, vnic);
 413         if (rc)
 414                 goto err_out;
 415
 416         /* Alloc RSS context only if RSS mode is enabled */
 417         if (dev_conf->rxmode.mq_mode & ETH_MQ_RX_RSS) {
 418                 int j, nr_ctxs = bnxt_rss_ctxts(bp);
 419
 420                 if (bp->rx_nr_rings > BNXT_RSS_TBL_SIZE_P5) {
 421                         PMD_DRV_LOG(ERR, "RxQ cnt %d > reta_size %d\n",
 422                                     bp->rx_nr_rings, BNXT_RSS_TBL_SIZE_P5);
 423                         PMD_DRV_LOG(ERR,
 424                                     "Only queues 0-%d will be in RSS table\n",
 425                                     BNXT_RSS_TBL_SIZE_P5 - 1);
 426                 }
 427
 428                 rc = 0;
 429                 for (j = 0; j < nr_ctxs; j++) {
 430                         rc = bnxt_hwrm_vnic_ctx_alloc(bp, vnic, j);
 431                         if (rc)
 432                                 break;
 433                 }
 434                 if (rc) {
 435                         PMD_DRV_LOG(ERR,
 436                                     "HWRM vnic %d ctx %d alloc failure rc: %x\n",
 437                                     vnic_id, j, rc);
 438                         goto err_out;
 439                 }
 440                 vnic->num_lb_ctxts = nr_ctxs;
 441         }
 442
 443         /*
 444          * Firmware sets pf pair in default vnic cfg. If the VLAN strip
 445          * setting is not available at this time, it will not be
 446          * configured correctly in the CFA.
 447          */
 448         if (rx_offloads & DEV_RX_OFFLOAD_VLAN_STRIP)
 449                 vnic->vlan_strip = true;
 450         else
 451                 vnic->vlan_strip = false;
 452
 453         rc = bnxt_hwrm_vnic_cfg(bp, vnic);
 454         if (rc)
 455                 goto err_out;
 456
 457         rc = bnxt_set_hwrm_vnic_filters(bp, vnic);
 458         if (rc)
 459                 goto err_out;
 460
 461         for (j = 0; j < bp->rx_num_qs_per_vnic; j++) {
 462                 rxq = bp->eth_dev->data->rx_queues[j];
 463
 464                 PMD_DRV_LOG(DEBUG,
 465                             "rxq[%d]->vnic=%p vnic->fw_grp_ids=%p\n",
 466                             j, rxq->vnic, rxq->vnic->fw_grp_ids);
 467
 468                 if (BNXT_HAS_RING_GRPS(bp) && rxq->rx_deferred_start)
 469                         rxq->vnic->fw_grp_ids[j] = INVALID_HW_RING_ID;
 470                 else
 471                         vnic->rx_queue_cnt++;
 472         }
 473
 474         PMD_DRV_LOG(DEBUG, "vnic->rx_queue_cnt = %d\n", vnic->rx_queue_cnt);
 475
 476         rc = bnxt_vnic_rss_configure(bp, vnic);
 477         if (rc)
 478                 goto err_out;
 479
 480         bnxt_hwrm_vnic_plcmode_cfg(bp, vnic);
 481
 482         if (rx_offloads & DEV_RX_OFFLOAD_TCP_LRO)
 483                 bnxt_hwrm_vnic_tpa_cfg(bp, vnic, 1);
 484         else
 485                 bnxt_hwrm_vnic_tpa_cfg(bp, vnic, 0);
 486
 487         return 0;
 488 err_out:
 489         PMD_DRV_LOG(ERR, "HWRM vnic %d cfg failure rc: %x\n",
 490                     vnic_id, rc);
 491         return rc;
 492 }
 493
 494 static int bnxt_register_fc_ctx_mem(struct bnxt *bp)
 495 {
 496         int rc = 0;
 497
 498         rc = bnxt_hwrm_ctx_rgtr(bp, bp->flow_stat->rx_fc_in_tbl.dma,
 499                                 &bp->flow_stat->rx_fc_in_tbl.ctx_id);
 500         if (rc)
 501                 return rc;
 502
 503         PMD_DRV_LOG(DEBUG,
 504                     "rx_fc_in_tbl.va = %p rx_fc_in_tbl.dma = %p"
 505                     " rx_fc_in_tbl.ctx_id = %d\n",
 506                     bp->flow_stat->rx_fc_in_tbl.va,
 507                     (void *)((uintptr_t)bp->flow_stat->rx_fc_in_tbl.dma),
 508                     bp->flow_stat->rx_fc_in_tbl.ctx_id);
 509
 510         rc = bnxt_hwrm_ctx_rgtr(bp, bp->flow_stat->rx_fc_out_tbl.dma,
 511                                 &bp->flow_stat->rx_fc_out_tbl.ctx_id);
 512         if (rc)
 513                 return rc;
 514
 515         PMD_DRV_LOG(DEBUG,
 516                     "rx_fc_out_tbl.va = %p rx_fc_out_tbl.dma = %p"
 517                     " rx_fc_out_tbl.ctx_id = %d\n",
 518                     bp->flow_stat->rx_fc_out_tbl.va,
 519                     (void *)((uintptr_t)bp->flow_stat->rx_fc_out_tbl.dma),
 520                     bp->flow_stat->rx_fc_out_tbl.ctx_id);
 521
 522         rc = bnxt_hwrm_ctx_rgtr(bp, bp->flow_stat->tx_fc_in_tbl.dma,
 523                                 &bp->flow_stat->tx_fc_in_tbl.ctx_id);
 524         if (rc)
 525                 return rc;
 526
 527         PMD_DRV_LOG(DEBUG,
 528                     "tx_fc_in_tbl.va = %p tx_fc_in_tbl.dma = %p"
 529                     " tx_fc_in_tbl.ctx_id = %d\n",
 530                     bp->flow_stat->tx_fc_in_tbl.va,
 531                     (void *)((uintptr_t)bp->flow_stat->tx_fc_in_tbl.dma),
 532                     bp->flow_stat->tx_fc_in_tbl.ctx_id);
 533
 534         rc = bnxt_hwrm_ctx_rgtr(bp, bp->flow_stat->tx_fc_out_tbl.dma,
 535                                 &bp->flow_stat->tx_fc_out_tbl.ctx_id);
 536         if (rc)
 537                 return rc;
 538
 539         PMD_DRV_LOG(DEBUG,
 540                     "tx_fc_out_tbl.va = %p tx_fc_out_tbl.dma = %p"
 541                     " tx_fc_out_tbl.ctx_id = %d\n",
 542                     bp->flow_stat->tx_fc_out_tbl.va,
 543                     (void *)((uintptr_t)bp->flow_stat->tx_fc_out_tbl.dma),
 544                     bp->flow_stat->tx_fc_out_tbl.ctx_id);
 545
 546         memset(bp->flow_stat->rx_fc_out_tbl.va,
 547                0,
 548                bp->flow_stat->rx_fc_out_tbl.size);
 549         rc = bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_RX,
 550                                        CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
 551                                        bp->flow_stat->rx_fc_out_tbl.ctx_id,
 552                                        bp->flow_stat->max_fc,
 553                                        true);
 554         if (rc)
 555                 return rc;
 556
 557         memset(bp->flow_stat->tx_fc_out_tbl.va,
 558                0,
 559                bp->flow_stat->tx_fc_out_tbl.size);
 560         rc = bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_TX,
 561                                        CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
 562                                        bp->flow_stat->tx_fc_out_tbl.ctx_id,
 563                                        bp->flow_stat->max_fc,
 564                                        true);
 565
 566         return rc;
 567 }
 568
 569 static int bnxt_alloc_ctx_mem_buf(char *type, size_t size,
 570                                   struct bnxt_ctx_mem_buf_info *ctx)
 571 {
 572         if (!ctx)
 573                 return -EINVAL;
 574
 575         ctx->va = rte_zmalloc(type, size, 0);
 576         if (ctx->va == NULL)
 577                 return -ENOMEM;
 578         rte_mem_lock_page(ctx->va);
 579         ctx->size = size;
 580         ctx->dma = rte_mem_virt2iova(ctx->va);
 581         if (ctx->dma == RTE_BAD_IOVA)
 582                 return -ENOMEM;
 583
 584         return 0;
 585 }
 586
 587 static int bnxt_init_fc_ctx_mem(struct bnxt *bp)
 588 {
 589         struct rte_pci_device *pdev = bp->pdev;
 590         char type[RTE_MEMZONE_NAMESIZE];
 591         uint16_t max_fc;
 592         int rc = 0;
 593
 594         max_fc = bp->flow_stat->max_fc;
 595
 596         sprintf(type, "bnxt_rx_fc_in_" PCI_PRI_FMT, pdev->addr.domain,
 597                 pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
 598         /* 4 bytes for each counter-id */
 599         rc = bnxt_alloc_ctx_mem_buf(type,
 600                                     max_fc * 4,
 601                                     &bp->flow_stat->rx_fc_in_tbl);
 602         if (rc)
 603                 return rc;
 604
 605         sprintf(type, "bnxt_rx_fc_out_" PCI_PRI_FMT, pdev->addr.domain,
 606                 pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
 607         /* 16 bytes for each counter - 8 bytes pkt_count, 8 bytes byte_count */
 608         rc = bnxt_alloc_ctx_mem_buf(type,
 609                                     max_fc * 16,
 610                                     &bp->flow_stat->rx_fc_out_tbl);
 611         if (rc)
 612                 return rc;
 613
 614         sprintf(type, "bnxt_tx_fc_in_" PCI_PRI_FMT, pdev->addr.domain,
 615                 pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
 616         /* 4 bytes for each counter-id */
 617         rc = bnxt_alloc_ctx_mem_buf(type,
 618                                     max_fc * 4,
 619                                     &bp->flow_stat->tx_fc_in_tbl);
 620         if (rc)
 621                 return rc;
 622
 623         sprintf(type, "bnxt_tx_fc_out_" PCI_PRI_FMT, pdev->addr.domain,
 624                 pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
 625         /* 16 bytes for each counter - 8 bytes pkt_count, 8 bytes byte_count */
 626         rc = bnxt_alloc_ctx_mem_buf(type,
 627                                     max_fc * 16,
 628                                     &bp->flow_stat->tx_fc_out_tbl);
 629         if (rc)
 630                 return rc;
 631
 632         rc = bnxt_register_fc_ctx_mem(bp);
 633
 634         return rc;
 635 }
 636
 637 static int bnxt_init_ctx_mem(struct bnxt *bp)
 638 {
 639         int rc = 0;
 640
 641         if (!(bp->fw_cap & BNXT_FW_CAP_ADV_FLOW_COUNTERS) ||
 642             !(BNXT_PF(bp) || BNXT_VF_IS_TRUSTED(bp)) ||
 643             !BNXT_FLOW_XSTATS_EN(bp))
 644                 return 0;
 645
 646         rc = bnxt_hwrm_cfa_counter_qcaps(bp, &bp->flow_stat->max_fc);
 647         if (rc)
 648                 return rc;
 649
 650         rc = bnxt_init_fc_ctx_mem(bp);
 651
 652         return rc;
 653 }
 654
 655 static int bnxt_update_phy_setting(struct bnxt *bp)
 656 {
 657         struct rte_eth_link new;
 658         int rc;
 659
 660         rc = bnxt_get_hwrm_link_config(bp, &new);
 661         if (rc) {
 662                 PMD_DRV_LOG(ERR, "Failed to get link settings\n");
 663                 return rc;
 664         }
 665
 666         /*
 667          * On BCM957508-N2100 adapters, FW will not allow any user other
 668          * than BMC to shutdown the port. bnxt_get_hwrm_link_config() call
 669          * always returns link up. Force phy update always in that case.
 670          */
 671         if (!new.link_status || IS_BNXT_DEV_957508_N2100(bp)) {
 672                 rc = bnxt_set_hwrm_link_config(bp, true);
 673                 if (rc) {
 674                         PMD_DRV_LOG(ERR, "Failed to update PHY settings\n");
 675                         return rc;
 676                 }
 677         }
 678
 679         return rc;
 680 }
 681
 682 static int bnxt_start_nic(struct bnxt *bp)
 683 {
 684         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(bp->eth_dev);
 685         struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
 686         uint32_t intr_vector = 0;
 687         uint32_t queue_id, base = BNXT_MISC_VEC_ID;
 688         uint32_t vec = BNXT_MISC_VEC_ID;
 689         unsigned int i, j;
 690         int rc;
 691
 692         if (bp->eth_dev->data->mtu > RTE_ETHER_MTU) {
 693                 bp->eth_dev->data->dev_conf.rxmode.offloads |=
 694                         DEV_RX_OFFLOAD_JUMBO_FRAME;
 695                 bp->flags |= BNXT_FLAG_JUMBO;
 696         } else {
 697                 bp->eth_dev->data->dev_conf.rxmode.offloads &=
 698                         ~DEV_RX_OFFLOAD_JUMBO_FRAME;
 699                 bp->flags &= ~BNXT_FLAG_JUMBO;
 700         }
 701
 702         /* THOR does not support ring groups.
 703          * But we will use the array to save RSS context IDs.
 704          */
 705         if (BNXT_CHIP_P5(bp))
 706                 bp->max_ring_grps = BNXT_MAX_RSS_CTXTS_P5;
 707
 708         rc = bnxt_alloc_all_hwrm_stat_ctxs(bp);
 709         if (rc) {
 710                 PMD_DRV_LOG(ERR, "HWRM stat ctx alloc failure rc: %x\n", rc);
 711                 goto err_out;
 712         }
 713
 714         rc = bnxt_alloc_hwrm_rings(bp);
 715         if (rc) {
 716                 PMD_DRV_LOG(ERR, "HWRM ring alloc failure rc: %x\n", rc);
 717                 goto err_out;
 718         }
 719
 720         rc = bnxt_alloc_all_hwrm_ring_grps(bp);
 721         if (rc) {
 722                 PMD_DRV_LOG(ERR, "HWRM ring grp alloc failure: %x\n", rc);
 723                 goto err_out;
 724         }
 725
 726         if (!(bp->vnic_cap_flags & BNXT_VNIC_CAP_COS_CLASSIFY))
 727                 goto skip_cosq_cfg;
 728
 729         for (j = 0, i = 0; i < BNXT_COS_QUEUE_COUNT; i++) {
 730                 if (bp->rx_cos_queue[i].id != 0xff) {
 731                         struct bnxt_vnic_info *vnic = &bp->vnic_info[j++];
 732
 733                         if (!vnic) {
 734                                 PMD_DRV_LOG(ERR,
 735                                             "Num pools more than FW profile\n");
 736                                 rc = -EINVAL;
 737                                 goto err_out;
 738                         }
 739                         vnic->cos_queue_id = bp->rx_cos_queue[i].id;
 740                         bp->rx_cosq_cnt++;
 741                 }
 742         }
 743
 744 skip_cosq_cfg:
 745         rc = bnxt_mq_rx_configure(bp);
 746         if (rc) {
 747                 PMD_DRV_LOG(ERR, "MQ mode configure failure rc: %x\n", rc);
 748                 goto err_out;
 749         }
 750
 751         /* default vnic 0 */
 752         rc = bnxt_setup_one_vnic(bp, 0);
 753         if (rc)
 754                 goto err_out;
 755         /* VNIC configuration */
 756         if (BNXT_RFS_NEEDS_VNIC(bp)) {
 757                 for (i = 1; i < bp->nr_vnics; i++) {
 758                         rc = bnxt_setup_one_vnic(bp, i);
 759                         if (rc)
 760                                 goto err_out;
 761                 }
 762         }
 763
 764         rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, &bp->vnic_info[0], 0, NULL);
 765         if (rc) {
 766                 PMD_DRV_LOG(ERR,
 767                         "HWRM cfa l2 rx mask failure rc: %x\n", rc);
 768                 goto err_out;
 769         }
 770
 771         /* check and configure queue intr-vector mapping */
 772         if ((rte_intr_cap_multiple(intr_handle) ||
 773              !RTE_ETH_DEV_SRIOV(bp->eth_dev).active) &&
 774             bp->eth_dev->data->dev_conf.intr_conf.rxq != 0) {
 775                 intr_vector = bp->eth_dev->data->nb_rx_queues;
 776                 PMD_DRV_LOG(DEBUG, "intr_vector = %d\n", intr_vector);
 777                 if (intr_vector > bp->rx_cp_nr_rings) {
 778                         PMD_DRV_LOG(ERR, "At most %d intr queues supported",
 779                                         bp->rx_cp_nr_rings);
 780                         return -ENOTSUP;
 781                 }
 782                 rc = rte_intr_efd_enable(intr_handle, intr_vector);
 783                 if (rc)
 784                         return rc;
 785         }
 786
 787         if (rte_intr_dp_is_en(intr_handle) && !intr_handle->intr_vec) {
 788                 intr_handle->intr_vec =
 789                         rte_zmalloc("intr_vec",
 790                                     bp->eth_dev->data->nb_rx_queues *
 791                                     sizeof(int), 0);
 792                 if (intr_handle->intr_vec == NULL) {
 793                         PMD_DRV_LOG(ERR, "Failed to allocate %d rx_queues"
 794                                 " intr_vec", bp->eth_dev->data->nb_rx_queues);
 795                         rc = -ENOMEM;
 796                         goto err_disable;
 797                 }
 798                 PMD_DRV_LOG(DEBUG, "intr_handle->intr_vec = %p "
 799                         "intr_handle->nb_efd = %d intr_handle->max_intr = %d\n",
 800                          intr_handle->intr_vec, intr_handle->nb_efd,
 801                         intr_handle->max_intr);
 802                 for (queue_id = 0; queue_id < bp->eth_dev->data->nb_rx_queues;
 803                      queue_id++) {
 804                         intr_handle->intr_vec[queue_id] =
 805                                                         vec + BNXT_RX_VEC_START;
 806                         if (vec < base + intr_handle->nb_efd - 1)
 807                                 vec++;
 808                 }
 809         }
 810
 811         /* enable uio/vfio intr/eventfd mapping */
 812         rc = rte_intr_enable(intr_handle);
 813 #ifndef RTE_EXEC_ENV_FREEBSD
 814         /* In FreeBSD OS, nic_uio driver does not support interrupts */
 815         if (rc)
 816                 goto err_free;
 817 #endif
 818
 819         rc = bnxt_update_phy_setting(bp);
 820         if (rc)
 821                 goto err_free;
 822
 823         bp->mark_table = rte_zmalloc("bnxt_mark_table", BNXT_MARK_TABLE_SZ, 0);
 824         if (!bp->mark_table)
 825                 PMD_DRV_LOG(ERR, "Allocation of mark table failed\n");
 826
 827         return 0;
 828
 829 err_free:
 830         rte_free(intr_handle->intr_vec);
 831 err_disable:
 832         rte_intr_efd_disable(intr_handle);
 833 err_out:
 834         /* Some of the error status returned by FW may not be from errno.h */
 835         if (rc > 0)
 836                 rc = -EIO;
 837
 838         return rc;
 839 }
 840
 841 static int bnxt_shutdown_nic(struct bnxt *bp)
 842 {
 843         bnxt_free_all_hwrm_resources(bp);
 844         bnxt_free_all_filters(bp);
 845         bnxt_free_all_vnics(bp);
 846         return 0;
 847 }
 848
 849 /*
 850  * Device configuration and status function
 851  */
 852
 853 uint32_t bnxt_get_speed_capabilities(struct bnxt *bp)
 854 {
 855         uint32_t link_speed = bp->link_info->support_speeds;
 856         uint32_t speed_capa = 0;
 857
 858         /* If PAM4 is configured, use PAM4 supported speed */
 859         if (link_speed == 0 && bp->link_info->support_pam4_speeds > 0)
 860                 link_speed = bp->link_info->support_pam4_speeds;
 861
 862         if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_LINK_SPEED_100MB)
 863                 speed_capa |= ETH_LINK_SPEED_100M;
 864         if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_100MBHD)
 865                 speed_capa |= ETH_LINK_SPEED_100M_HD;
 866         if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_1GB)
 867                 speed_capa |= ETH_LINK_SPEED_1G;
 868         if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_2_5GB)
 869                 speed_capa |= ETH_LINK_SPEED_2_5G;
 870         if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_10GB)
 871                 speed_capa |= ETH_LINK_SPEED_10G;
 872         if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_20GB)
 873                 speed_capa |= ETH_LINK_SPEED_20G;
 874         if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_25GB)
 875                 speed_capa |= ETH_LINK_SPEED_25G;
 876         if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_40GB)
 877                 speed_capa |= ETH_LINK_SPEED_40G;
 878         if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_50GB)
 879                 speed_capa |= ETH_LINK_SPEED_50G;
 880         if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_100GB)
 881                 speed_capa |= ETH_LINK_SPEED_100G;
 882         if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_PAM4_SPEEDS_50G)
 883                 speed_capa |= ETH_LINK_SPEED_50G;
 884         if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_PAM4_SPEEDS_100G)
 885                 speed_capa |= ETH_LINK_SPEED_100G;
 886         if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_PAM4_SPEEDS_200G)
 887                 speed_capa |= ETH_LINK_SPEED_200G;
 888
 889         if (bp->link_info->auto_mode ==
 890             HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_MODE_NONE)
 891                 speed_capa |= ETH_LINK_SPEED_FIXED;
 892         else
 893                 speed_capa |= ETH_LINK_SPEED_AUTONEG;
 894
 895         return speed_capa;
 896 }
 897
 898 static int bnxt_dev_info_get_op(struct rte_eth_dev *eth_dev,
 899                                 struct rte_eth_dev_info *dev_info)
 900 {
 901         struct rte_pci_device *pdev = RTE_DEV_TO_PCI(eth_dev->device);
 902         struct bnxt *bp = eth_dev->data->dev_private;
 903         uint16_t max_vnics, i, j, vpool, vrxq;
 904         unsigned int max_rx_rings;
 905         int rc;
 906
 907         rc = is_bnxt_in_error(bp);
 908         if (rc)
 909                 return rc;
 910
 911         /* MAC Specifics */
 912         dev_info->max_mac_addrs = bp->max_l2_ctx;
 913         dev_info->max_hash_mac_addrs = 0;
 914
 915         /* PF/VF specifics */
 916         if (BNXT_PF(bp))
 917                 dev_info->max_vfs = pdev->max_vfs;
 918
 919         max_rx_rings = bnxt_max_rings(bp);
 920         /* For the sake of symmetry, max_rx_queues = max_tx_queues */
 921         dev_info->max_rx_queues = max_rx_rings;
 922         dev_info->max_tx_queues = max_rx_rings;
 923         dev_info->reta_size = bnxt_rss_hash_tbl_size(bp);
 924         dev_info->hash_key_size = 40;
 925         max_vnics = bp->max_vnics;
 926
 927         /* MTU specifics */
 928         dev_info->min_mtu = RTE_ETHER_MIN_MTU;
 929         dev_info->max_mtu = BNXT_MAX_MTU;
 930
 931         /* Fast path specifics */
 932         dev_info->min_rx_bufsize = 1;
 933         dev_info->max_rx_pktlen = BNXT_MAX_PKT_LEN;
 934
 935         dev_info->rx_offload_capa = BNXT_DEV_RX_OFFLOAD_SUPPORT;
 936         if (bp->flags & BNXT_FLAG_PTP_SUPPORTED)
 937                 dev_info->rx_offload_capa |= DEV_RX_OFFLOAD_TIMESTAMP;
 938         dev_info->tx_queue_offload_capa = DEV_TX_OFFLOAD_MBUF_FAST_FREE;
 939         dev_info->tx_offload_capa = BNXT_DEV_TX_OFFLOAD_SUPPORT |
 940                                     dev_info->tx_queue_offload_capa;
 941         dev_info->flow_type_rss_offloads = BNXT_ETH_RSS_SUPPORT;
 942
 943         dev_info->speed_capa = bnxt_get_speed_capabilities(bp);
 944
 945         /* *INDENT-OFF* */
 946         dev_info->default_rxconf = (struct rte_eth_rxconf) {
 947                 .rx_thresh = {
 948                         .pthresh = 8,
 949                         .hthresh = 8,
 950                         .wthresh = 0,
 951                 },
 952                 .rx_free_thresh = 32,
 953                 .rx_drop_en = BNXT_DEFAULT_RX_DROP_EN,
 954         };
 955
 956         dev_info->default_txconf = (struct rte_eth_txconf) {
 957                 .tx_thresh = {
 958                         .pthresh = 32,
 959                         .hthresh = 0,
 960                         .wthresh = 0,
 961                 },
 962                 .tx_free_thresh = 32,
 963                 .tx_rs_thresh = 32,
 964         };
 965         eth_dev->data->dev_conf.intr_conf.lsc = 1;
 966
 967         eth_dev->data->dev_conf.intr_conf.rxq = 1;
 968         dev_info->rx_desc_lim.nb_min = BNXT_MIN_RING_DESC;
 969         dev_info->rx_desc_lim.nb_max = BNXT_MAX_RX_RING_DESC;
 970         dev_info->tx_desc_lim.nb_min = BNXT_MIN_RING_DESC;
 971         dev_info->tx_desc_lim.nb_max = BNXT_MAX_TX_RING_DESC;
 972
 973         if (BNXT_PF(bp) || BNXT_VF_IS_TRUSTED(bp)) {
 974                 dev_info->switch_info.name = eth_dev->device->name;
 975                 dev_info->switch_info.domain_id = bp->switch_domain_id;
 976                 dev_info->switch_info.port_id =
 977                                 BNXT_PF(bp) ? BNXT_SWITCH_PORT_ID_PF :
 978                                     BNXT_SWITCH_PORT_ID_TRUSTED_VF;
 979         }
 980
 981         /* *INDENT-ON* */
 982
 983         /*
 984          * TODO: default_rxconf, default_txconf, rx_desc_lim, and tx_desc_lim
 985          *       need further investigation.
 986          */
 987
 988         /* VMDq resources */
 989         vpool = 64; /* ETH_64_POOLS */
 990         vrxq = 128; /* ETH_VMDQ_DCB_NUM_QUEUES */
 991         for (i = 0; i < 4; vpool >>= 1, i++) {
 992                 if (max_vnics > vpool) {
 993                         for (j = 0; j < 5; vrxq >>= 1, j++) {
 994                                 if (dev_info->max_rx_queues > vrxq) {
 995                                         if (vpool > vrxq)
 996                                                 vpool = vrxq;
 997                                         goto found;
 998                                 }
 999                         }
1000                         /* Not enough resources to support VMDq */
1001                         break;
1002                 }
1003         }
1004         /* Not enough resources to support VMDq */
1005         vpool = 0;
1006         vrxq = 0;
1007 found:
1008         dev_info->max_vmdq_pools = vpool;
1009         dev_info->vmdq_queue_num = vrxq;
1010
1011         dev_info->vmdq_pool_base = 0;
1012         dev_info->vmdq_queue_base = 0;
1013
1014         return 0;
1015 }
1016
1017 /* Configure the device based on the configuration provided */
1018 static int bnxt_dev_configure_op(struct rte_eth_dev *eth_dev)
1019 {
1020         struct bnxt *bp = eth_dev->data->dev_private;
1021         uint64_t rx_offloads = eth_dev->data->dev_conf.rxmode.offloads;
1022         int rc;
1023
1024         bp->rx_queues = (void *)eth_dev->data->rx_queues;
1025         bp->tx_queues = (void *)eth_dev->data->tx_queues;
1026         bp->tx_nr_rings = eth_dev->data->nb_tx_queues;
1027         bp->rx_nr_rings = eth_dev->data->nb_rx_queues;
1028
1029         rc = is_bnxt_in_error(bp);
1030         if (rc)
1031                 return rc;
1032
1033         if (BNXT_VF(bp) && (bp->flags & BNXT_FLAG_NEW_RM)) {
1034                 rc = bnxt_hwrm_check_vf_rings(bp);
1035                 if (rc) {
1036                         PMD_DRV_LOG(ERR, "HWRM insufficient resources\n");
1037                         return -ENOSPC;
1038                 }
1039
1040                 /* If a resource has already been allocated - in this case
1041                  * it is the async completion ring, free it. Reallocate it after
1042                  * resource reservation. This will ensure the resource counts
1043                  * are calculated correctly.
1044                  */
1045
1046                 pthread_mutex_lock(&bp->def_cp_lock);
1047
1048                 if (!BNXT_HAS_NQ(bp) && bp->async_cp_ring) {
1049                         bnxt_disable_int(bp);
1050                         bnxt_free_cp_ring(bp, bp->async_cp_ring);
1051                 }
1052
1053                 rc = bnxt_hwrm_func_reserve_vf_resc(bp, false);
1054                 if (rc) {
1055                         PMD_DRV_LOG(ERR, "HWRM resource alloc fail:%x\n", rc);
1056                         pthread_mutex_unlock(&bp->def_cp_lock);
1057                         return -ENOSPC;
1058                 }
1059
1060                 if (!BNXT_HAS_NQ(bp) && bp->async_cp_ring) {
1061                         rc = bnxt_alloc_async_cp_ring(bp);
1062                         if (rc) {
1063                                 pthread_mutex_unlock(&bp->def_cp_lock);
1064                                 return rc;
1065                         }
1066                         bnxt_enable_int(bp);
1067                 }
1068
1069                 pthread_mutex_unlock(&bp->def_cp_lock);
1070         }
1071
1072         /* Inherit new configurations */
1073         if (eth_dev->data->nb_rx_queues > bp->max_rx_rings ||
1074             eth_dev->data->nb_tx_queues > bp->max_tx_rings ||
1075             eth_dev->data->nb_rx_queues + eth_dev->data->nb_tx_queues
1076                 + BNXT_NUM_ASYNC_CPR(bp) > bp->max_cp_rings ||
1077             eth_dev->data->nb_rx_queues + eth_dev->data->nb_tx_queues >
1078             bp->max_stat_ctx)
1079                 goto resource_error;
1080
1081         if (BNXT_HAS_RING_GRPS(bp) &&
1082             (uint32_t)(eth_dev->data->nb_rx_queues) > bp->max_ring_grps)
1083                 goto resource_error;
1084
1085         if (!(eth_dev->data->dev_conf.rxmode.mq_mode & ETH_MQ_RX_RSS) &&
1086             bp->max_vnics < eth_dev->data->nb_rx_queues)
1087                 goto resource_error;
1088
1089         bp->rx_cp_nr_rings = bp->rx_nr_rings;
1090         bp->tx_cp_nr_rings = bp->tx_nr_rings;
1091
1092         if (eth_dev->data->dev_conf.rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG)
1093                 rx_offloads |= DEV_RX_OFFLOAD_RSS_HASH;
1094         eth_dev->data->dev_conf.rxmode.offloads = rx_offloads;
1095
1096         if (rx_offloads & DEV_RX_OFFLOAD_JUMBO_FRAME) {
1097                 eth_dev->data->mtu =
1098                         eth_dev->data->dev_conf.rxmode.max_rx_pkt_len -
1099                         RTE_ETHER_HDR_LEN - RTE_ETHER_CRC_LEN - VLAN_TAG_SIZE *
1100                         BNXT_NUM_VLANS;
1101                 bnxt_mtu_set_op(eth_dev, eth_dev->data->mtu);
1102         }
1103         return 0;
1104
1105 resource_error:
1106         PMD_DRV_LOG(ERR,
1107                     "Insufficient resources to support requested config\n");
1108         PMD_DRV_LOG(ERR,
1109                     "Num Queues Requested: Tx %d, Rx %d\n",
1110                     eth_dev->data->nb_tx_queues,
1111                     eth_dev->data->nb_rx_queues);
1112         PMD_DRV_LOG(ERR,
1113                     "MAX: TxQ %d, RxQ %d, CQ %d Stat %d, Grp %d, Vnic %d\n",
1114                     bp->max_tx_rings, bp->max_rx_rings, bp->max_cp_rings,
1115                     bp->max_stat_ctx, bp->max_ring_grps, bp->max_vnics);
1116         return -ENOSPC;
1117 }
1118
1119 void bnxt_print_link_info(struct rte_eth_dev *eth_dev)
1120 {
1121         struct rte_eth_link *link = &eth_dev->data->dev_link;
1122
1123         if (link->link_status)
1124                 PMD_DRV_LOG(INFO, "Port %d Link Up - speed %u Mbps - %s\n",
1125                         eth_dev->data->port_id,
1126                         (uint32_t)link->link_speed,
1127                         (link->link_duplex == ETH_LINK_FULL_DUPLEX) ?
1128                         ("full-duplex") : ("half-duplex\n"));
1129         else
1130                 PMD_DRV_LOG(INFO, "Port %d Link Down\n",
1131                         eth_dev->data->port_id);
1132 }
1133
1134 /*
1135  * Determine whether the current configuration requires support for scattered
1136  * receive; return 1 if scattered receive is required and 0 if not.
1137  */
1138 static int bnxt_scattered_rx(struct rte_eth_dev *eth_dev)
1139 {
1140         uint16_t buf_size;
1141         int i;
1142
1143         if (eth_dev->data->dev_conf.rxmode.offloads & DEV_RX_OFFLOAD_SCATTER)
1144                 return 1;
1145
1146         if (eth_dev->data->dev_conf.rxmode.offloads & DEV_RX_OFFLOAD_TCP_LRO)
1147                 return 1;
1148
1149         for (i = 0; i < eth_dev->data->nb_rx_queues; i++) {
1150                 struct bnxt_rx_queue *rxq = eth_dev->data->rx_queues[i];
1151
1152                 buf_size = (uint16_t)(rte_pktmbuf_data_room_size(rxq->mb_pool) -
1153                                       RTE_PKTMBUF_HEADROOM);
1154                 if (eth_dev->data->dev_conf.rxmode.max_rx_pkt_len > buf_size)
1155                         return 1;
1156         }
1157         return 0;
1158 }
1159
1160 static eth_rx_burst_t
1161 bnxt_receive_function(struct rte_eth_dev *eth_dev)
1162 {
1163         struct bnxt *bp = eth_dev->data->dev_private;
1164
1165         /* Disable vector mode RX for Stingray2 for now */
1166         if (BNXT_CHIP_SR2(bp)) {
1167                 bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
1168                 return bnxt_recv_pkts;
1169         }
1170
1171 #if defined(RTE_ARCH_X86) || defined(RTE_ARCH_ARM64)
1172 #ifndef RTE_LIBRTE_IEEE1588
1173         /*
1174          * Vector mode receive can be enabled only if scatter rx is not
1175          * in use and rx offloads are limited to VLAN stripping and
1176          * CRC stripping.
1177          */
1178         if (!eth_dev->data->scattered_rx &&
1179             !(eth_dev->data->dev_conf.rxmode.offloads &
1180               ~(DEV_RX_OFFLOAD_VLAN_STRIP |
1181                 DEV_RX_OFFLOAD_KEEP_CRC |
1182                 DEV_RX_OFFLOAD_JUMBO_FRAME |
1183                 DEV_RX_OFFLOAD_IPV4_CKSUM |
1184                 DEV_RX_OFFLOAD_UDP_CKSUM |
1185                 DEV_RX_OFFLOAD_TCP_CKSUM |
1186                 DEV_RX_OFFLOAD_OUTER_IPV4_CKSUM |
1187                 DEV_RX_OFFLOAD_OUTER_UDP_CKSUM |
1188                 DEV_RX_OFFLOAD_RSS_HASH |
1189                 DEV_RX_OFFLOAD_VLAN_FILTER)) &&
1190             !BNXT_TRUFLOW_EN(bp) && BNXT_NUM_ASYNC_CPR(bp) &&
1191             rte_vect_get_max_simd_bitwidth() >= RTE_VECT_SIMD_128) {
1192                 PMD_DRV_LOG(INFO, "Using vector mode receive for port %d\n",
1193                             eth_dev->data->port_id);
1194                 bp->flags |= BNXT_FLAG_RX_VECTOR_PKT_MODE;
1195                 return bnxt_recv_pkts_vec;
1196         }
1197         PMD_DRV_LOG(INFO, "Vector mode receive disabled for port %d\n",
1198                     eth_dev->data->port_id);
1199         PMD_DRV_LOG(INFO,
1200                     "Port %d scatter: %d rx offload: %" PRIX64 "\n",
1201                     eth_dev->data->port_id,
1202                     eth_dev->data->scattered_rx,
1203                     eth_dev->data->dev_conf.rxmode.offloads);
1204 #endif
1205 #endif
1206         bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
1207         return bnxt_recv_pkts;
1208 }
1209
1210 static eth_tx_burst_t
1211 bnxt_transmit_function(struct rte_eth_dev *eth_dev)
1212 {
1213         struct bnxt *bp = eth_dev->data->dev_private;
1214
1215         /* Disable vector mode TX for Stingray2 for now */
1216         if (BNXT_CHIP_SR2(bp))
1217                 return bnxt_xmit_pkts;
1218
1219 #if defined(RTE_ARCH_X86) || defined(RTE_ARCH_ARM64)
1220 #ifndef RTE_LIBRTE_IEEE1588
1221         uint64_t offloads = eth_dev->data->dev_conf.txmode.offloads;
1222
1223         /*
1224          * Vector mode transmit can be enabled only if not using scatter rx
1225          * or tx offloads.
1226          */
1227         if (!eth_dev->data->scattered_rx &&
1228             !(offloads & ~DEV_TX_OFFLOAD_MBUF_FAST_FREE) &&
1229             !BNXT_TRUFLOW_EN(bp) &&
1230             rte_vect_get_max_simd_bitwidth() >= RTE_VECT_SIMD_128) {
1231                 PMD_DRV_LOG(INFO, "Using vector mode transmit for port %d\n",
1232                             eth_dev->data->port_id);
1233                 return bnxt_xmit_pkts_vec;
1234         }
1235         PMD_DRV_LOG(INFO, "Vector mode transmit disabled for port %d\n",
1236                     eth_dev->data->port_id);
1237         PMD_DRV_LOG(INFO,
1238                     "Port %d scatter: %d tx offload: %" PRIX64 "\n",
1239                     eth_dev->data->port_id,
1240                     eth_dev->data->scattered_rx,
1241                     offloads);
1242 #endif
1243 #endif
1244         return bnxt_xmit_pkts;
1245 }
1246
1247 static int bnxt_handle_if_change_status(struct bnxt *bp)
1248 {
1249         int rc;
1250
1251         /* Since fw has undergone a reset and lost all contexts,
1252          * set fatal flag to not issue hwrm during cleanup
1253          */
1254         bp->flags |= BNXT_FLAG_FATAL_ERROR;
1255         bnxt_uninit_resources(bp, true);
1256
1257         /* clear fatal flag so that re-init happens */
1258         bp->flags &= ~BNXT_FLAG_FATAL_ERROR;
1259         rc = bnxt_init_resources(bp, true);
1260
1261         bp->flags &= ~BNXT_FLAG_IF_CHANGE_HOT_FW_RESET_DONE;
1262
1263         return rc;
1264 }
1265
1266 static int bnxt_dev_set_link_up_op(struct rte_eth_dev *eth_dev)
1267 {
1268         struct bnxt *bp = eth_dev->data->dev_private;
1269         int rc = 0;
1270
1271         if (!BNXT_SINGLE_PF(bp))
1272                 return -ENOTSUP;
1273
1274         if (!bp->link_info->link_up)
1275                 rc = bnxt_set_hwrm_link_config(bp, true);
1276         if (!rc)
1277                 eth_dev->data->dev_link.link_status = 1;
1278
1279         bnxt_print_link_info(eth_dev);
1280         return rc;
1281 }
1282
1283 static int bnxt_dev_set_link_down_op(struct rte_eth_dev *eth_dev)
1284 {
1285         struct bnxt *bp = eth_dev->data->dev_private;
1286
1287         if (!BNXT_SINGLE_PF(bp))
1288                 return -ENOTSUP;
1289
1290         eth_dev->data->dev_link.link_status = 0;
1291         bnxt_set_hwrm_link_config(bp, false);
1292         bp->link_info->link_up = 0;
1293
1294         return 0;
1295 }
1296
1297 static void bnxt_free_switch_domain(struct bnxt *bp)
1298 {
1299         int rc = 0;
1300
1301         if (bp->switch_domain_id) {
1302                 rc = rte_eth_switch_domain_free(bp->switch_domain_id);
1303                 if (rc)
1304                         PMD_DRV_LOG(ERR, "free switch domain:%d fail: %d\n",
1305                                     bp->switch_domain_id, rc);
1306         }
1307 }
1308
1309 static void bnxt_ptp_get_current_time(void *arg)
1310 {
1311         struct bnxt *bp = arg;
1312         struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
1313         int rc;
1314
1315         rc = is_bnxt_in_error(bp);
1316         if (rc)
1317                 return;
1318
1319         if (!ptp)
1320                 return;
1321
1322         bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_CURRENT_TIME,
1323                                 &ptp->current_time);
1324
1325         rc = rte_eal_alarm_set(US_PER_S, bnxt_ptp_get_current_time, (void *)bp);
1326         if (rc != 0) {
1327                 PMD_DRV_LOG(ERR, "Failed to re-schedule PTP alarm\n");
1328                 bp->flags2 &= ~BNXT_FLAGS2_PTP_ALARM_SCHEDULED;
1329         }
1330 }
1331
1332 static int bnxt_schedule_ptp_alarm(struct bnxt *bp)
1333 {
1334         struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
1335         int rc;
1336
1337         if (bp->flags2 & BNXT_FLAGS2_PTP_ALARM_SCHEDULED)
1338                 return 0;
1339
1340         bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_CURRENT_TIME,
1341                                 &ptp->current_time);
1342
1343         rc = rte_eal_alarm_set(US_PER_S, bnxt_ptp_get_current_time, (void *)bp);
1344         return rc;
1345 }
1346
1347 static void bnxt_cancel_ptp_alarm(struct bnxt *bp)
1348 {
1349         if (bp->flags2 & BNXT_FLAGS2_PTP_ALARM_SCHEDULED) {
1350                 rte_eal_alarm_cancel(bnxt_ptp_get_current_time, (void *)bp);
1351                 bp->flags2 &= ~BNXT_FLAGS2_PTP_ALARM_SCHEDULED;
1352         }
1353 }
1354
1355 static void bnxt_ptp_stop(struct bnxt *bp)
1356 {
1357         bnxt_cancel_ptp_alarm(bp);
1358         bp->flags2 &= ~BNXT_FLAGS2_PTP_TIMESYNC_ENABLED;
1359 }
1360
1361 static int bnxt_ptp_start(struct bnxt *bp)
1362 {
1363         int rc;
1364
1365         rc = bnxt_schedule_ptp_alarm(bp);
1366         if (rc != 0) {
1367                 PMD_DRV_LOG(ERR, "Failed to schedule PTP alarm\n");
1368         } else {
1369                 bp->flags2 |= BNXT_FLAGS2_PTP_TIMESYNC_ENABLED;
1370                 bp->flags2 |= BNXT_FLAGS2_PTP_ALARM_SCHEDULED;
1371         }
1372
1373         return rc;
1374 }
1375
1376 static int bnxt_dev_stop(struct rte_eth_dev *eth_dev)
1377 {
1378         struct bnxt *bp = eth_dev->data->dev_private;
1379         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
1380         struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
1381         struct rte_eth_link link;
1382         int ret;
1383
1384         eth_dev->data->dev_started = 0;
1385         eth_dev->data->scattered_rx = 0;
1386
1387         /* Prevent crashes when queues are still in use */
1388         eth_dev->rx_pkt_burst = &bnxt_dummy_recv_pkts;
1389         eth_dev->tx_pkt_burst = &bnxt_dummy_xmit_pkts;
1390
1391         bnxt_disable_int(bp);
1392
1393         /* disable uio/vfio intr/eventfd mapping */
1394         rte_intr_disable(intr_handle);
1395
1396         /* Stop the child representors for this device */
1397         ret = bnxt_rep_stop_all(bp);
1398         if (ret != 0)
1399                 return ret;
1400
1401         /* delete the bnxt ULP port details */
1402         bnxt_ulp_port_deinit(bp);
1403
1404         bnxt_cancel_fw_health_check(bp);
1405
1406         if (BNXT_P5_PTP_TIMESYNC_ENABLED(bp))
1407                 bnxt_cancel_ptp_alarm(bp);
1408
1409         /* Do not bring link down during reset recovery */
1410         if (!is_bnxt_in_error(bp)) {
1411                 bnxt_dev_set_link_down_op(eth_dev);
1412                 /* Wait for link to be reset */
1413                 if (BNXT_SINGLE_PF(bp))
1414                         rte_delay_ms(500);
1415                 /* clear the recorded link status */
1416                 memset(&link, 0, sizeof(link));
1417                 rte_eth_linkstatus_set(eth_dev, &link);
1418         }
1419
1420         /* Clean queue intr-vector mapping */
1421         rte_intr_efd_disable(intr_handle);
1422         if (intr_handle->intr_vec != NULL) {
1423                 rte_free(intr_handle->intr_vec);
1424                 intr_handle->intr_vec = NULL;
1425         }
1426
1427         bnxt_hwrm_port_clr_stats(bp);
1428         bnxt_free_tx_mbufs(bp);
1429         bnxt_free_rx_mbufs(bp);
1430         /* Process any remaining notifications in default completion queue */
1431         bnxt_int_handler(eth_dev);
1432         bnxt_shutdown_nic(bp);
1433         bnxt_hwrm_if_change(bp, false);
1434
1435         rte_free(bp->mark_table);
1436         bp->mark_table = NULL;
1437
1438         bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
1439         bp->rx_cosq_cnt = 0;
1440         /* All filters are deleted on a port stop. */
1441         if (BNXT_FLOW_XSTATS_EN(bp))
1442                 bp->flow_stat->flow_count = 0;
1443
1444         return 0;
1445 }
1446
1447 /* Unload the driver, release resources */
1448 static int bnxt_dev_stop_op(struct rte_eth_dev *eth_dev)
1449 {
1450         struct bnxt *bp = eth_dev->data->dev_private;
1451
1452         pthread_mutex_lock(&bp->err_recovery_lock);
1453         if (bp->flags & BNXT_FLAG_FW_RESET) {
1454                 PMD_DRV_LOG(ERR,
1455                             "Adapter recovering from error..Please retry\n");
1456                 pthread_mutex_unlock(&bp->err_recovery_lock);
1457                 return -EAGAIN;
1458         }
1459         pthread_mutex_unlock(&bp->err_recovery_lock);
1460
1461         return bnxt_dev_stop(eth_dev);
1462 }
1463
1464 static int bnxt_dev_start_op(struct rte_eth_dev *eth_dev)
1465 {
1466         struct bnxt *bp = eth_dev->data->dev_private;
1467         uint64_t rx_offloads = eth_dev->data->dev_conf.rxmode.offloads;
1468         int vlan_mask = 0;
1469         int rc, retry_cnt = BNXT_IF_CHANGE_RETRY_COUNT;
1470
1471         if (!eth_dev->data->nb_tx_queues || !eth_dev->data->nb_rx_queues) {
1472                 PMD_DRV_LOG(ERR, "Queues are not configured yet!\n");
1473                 return -EINVAL;
1474         }
1475
1476         if (bp->rx_cp_nr_rings > RTE_ETHDEV_QUEUE_STAT_CNTRS)
1477                 PMD_DRV_LOG(ERR,
1478                             "RxQ cnt %d > RTE_ETHDEV_QUEUE_STAT_CNTRS %d\n",
1479                             bp->rx_cp_nr_rings, RTE_ETHDEV_QUEUE_STAT_CNTRS);
1480
1481         do {
1482                 rc = bnxt_hwrm_if_change(bp, true);
1483                 if (rc == 0 || rc != -EAGAIN)
1484                         break;
1485
1486                 rte_delay_ms(BNXT_IF_CHANGE_RETRY_INTERVAL);
1487         } while (retry_cnt--);
1488
1489         if (rc)
1490                 return rc;
1491
1492         if (bp->flags & BNXT_FLAG_IF_CHANGE_HOT_FW_RESET_DONE) {
1493                 rc = bnxt_handle_if_change_status(bp);
1494                 if (rc)
1495                         return rc;
1496         }
1497
1498         bnxt_enable_int(bp);
1499
1500         eth_dev->data->scattered_rx = bnxt_scattered_rx(eth_dev);
1501
1502         rc = bnxt_start_nic(bp);
1503         if (rc)
1504                 goto error;
1505
1506         eth_dev->data->dev_started = 1;
1507
1508         bnxt_link_update_op(eth_dev, 1);
1509
1510         if (rx_offloads & DEV_RX_OFFLOAD_VLAN_FILTER)
1511                 vlan_mask |= ETH_VLAN_FILTER_MASK;
1512         if (rx_offloads & DEV_RX_OFFLOAD_VLAN_STRIP)
1513                 vlan_mask |= ETH_VLAN_STRIP_MASK;
1514         rc = bnxt_vlan_offload_set_op(eth_dev, vlan_mask);
1515         if (rc)
1516                 goto error;
1517
1518         /* Initialize bnxt ULP port details */
1519         rc = bnxt_ulp_port_init(bp);
1520         if (rc)
1521                 goto error;
1522
1523         eth_dev->rx_pkt_burst = bnxt_receive_function(eth_dev);
1524         eth_dev->tx_pkt_burst = bnxt_transmit_function(eth_dev);
1525
1526         bnxt_schedule_fw_health_check(bp);
1527
1528         if (BNXT_P5_PTP_TIMESYNC_ENABLED(bp))
1529                 bnxt_schedule_ptp_alarm(bp);
1530
1531         return 0;
1532
1533 error:
1534         bnxt_dev_stop(eth_dev);
1535         return rc;
1536 }
1537
1538 static void
1539 bnxt_uninit_locks(struct bnxt *bp)
1540 {
1541         pthread_mutex_destroy(&bp->flow_lock);
1542         pthread_mutex_destroy(&bp->def_cp_lock);
1543         pthread_mutex_destroy(&bp->health_check_lock);
1544         pthread_mutex_destroy(&bp->err_recovery_lock);
1545         if (bp->rep_info) {
1546                 pthread_mutex_destroy(&bp->rep_info->vfr_lock);
1547                 pthread_mutex_destroy(&bp->rep_info->vfr_start_lock);
1548         }
1549 }
1550
1551 static void bnxt_drv_uninit(struct bnxt *bp)
1552 {
1553         bnxt_free_switch_domain(bp);
1554         bnxt_free_leds_info(bp);
1555         bnxt_free_cos_queues(bp);
1556         bnxt_free_link_info(bp);
1557         bnxt_free_pf_info(bp);
1558         bnxt_free_parent_info(bp);
1559         bnxt_uninit_locks(bp);
1560
1561         rte_memzone_free((const struct rte_memzone *)bp->tx_mem_zone);
1562         bp->tx_mem_zone = NULL;
1563         rte_memzone_free((const struct rte_memzone *)bp->rx_mem_zone);
1564         bp->rx_mem_zone = NULL;
1565
1566         bnxt_free_vf_info(bp);
1567
1568         rte_free(bp->grp_info);
1569         bp->grp_info = NULL;
1570 }
1571
1572 static int bnxt_dev_close_op(struct rte_eth_dev *eth_dev)
1573 {
1574         struct bnxt *bp = eth_dev->data->dev_private;
1575         int ret = 0;
1576
1577         if (rte_eal_process_type() != RTE_PROC_PRIMARY)
1578                 return 0;
1579
1580         pthread_mutex_lock(&bp->err_recovery_lock);
1581         if (bp->flags & BNXT_FLAG_FW_RESET) {
1582                 PMD_DRV_LOG(ERR,
1583                             "Adapter recovering from error...Please retry\n");
1584                 pthread_mutex_unlock(&bp->err_recovery_lock);
1585                 return -EAGAIN;
1586         }
1587         pthread_mutex_unlock(&bp->err_recovery_lock);
1588
1589         /* cancel the recovery handler before remove dev */
1590         rte_eal_alarm_cancel(bnxt_dev_reset_and_resume, (void *)bp);
1591         rte_eal_alarm_cancel(bnxt_dev_recover, (void *)bp);
1592         bnxt_cancel_fc_thread(bp);
1593
1594         if (eth_dev->data->dev_started)
1595                 ret = bnxt_dev_stop(eth_dev);
1596
1597         bnxt_uninit_resources(bp, false);
1598
1599         bnxt_drv_uninit(bp);
1600
1601         return ret;
1602 }
1603
1604 static void bnxt_mac_addr_remove_op(struct rte_eth_dev *eth_dev,
1605                                     uint32_t index)
1606 {
1607         struct bnxt *bp = eth_dev->data->dev_private;
1608         uint64_t pool_mask = eth_dev->data->mac_pool_sel[index];
1609         struct bnxt_vnic_info *vnic;
1610         struct bnxt_filter_info *filter, *temp_filter;
1611         uint32_t i;
1612
1613         if (is_bnxt_in_error(bp))
1614                 return;
1615
1616         /*
1617          * Loop through all VNICs from the specified filter flow pools to
1618          * remove the corresponding MAC addr filter
1619          */
1620         for (i = 0; i < bp->nr_vnics; i++) {
1621                 if (!(pool_mask & (1ULL << i)))
1622                         continue;
1623
1624                 vnic = &bp->vnic_info[i];
1625                 filter = STAILQ_FIRST(&vnic->filter);
1626                 while (filter) {
1627                         temp_filter = STAILQ_NEXT(filter, next);
1628                         if (filter->mac_index == index) {
1629                                 STAILQ_REMOVE(&vnic->filter, filter,
1630                                                 bnxt_filter_info, next);
1631                                 bnxt_hwrm_clear_l2_filter(bp, filter);
1632                                 bnxt_free_filter(bp, filter);
1633                         }
1634                         filter = temp_filter;
1635                 }
1636         }
1637 }
1638
1639 static int bnxt_add_mac_filter(struct bnxt *bp, struct bnxt_vnic_info *vnic,
1640                                struct rte_ether_addr *mac_addr, uint32_t index,
1641                                uint32_t pool)
1642 {
1643         struct bnxt_filter_info *filter;
1644         int rc = 0;
1645
1646         /* Attach requested MAC address to the new l2_filter */
1647         STAILQ_FOREACH(filter, &vnic->filter, next) {
1648                 if (filter->mac_index == index) {
1649                         PMD_DRV_LOG(DEBUG,
1650                                     "MAC addr already existed for pool %d\n",
1651                                     pool);
1652                         return 0;
1653                 }
1654         }
1655
1656         filter = bnxt_alloc_filter(bp);
1657         if (!filter) {
1658                 PMD_DRV_LOG(ERR, "L2 filter alloc failed\n");
1659                 return -ENODEV;
1660         }
1661
1662         /* bnxt_alloc_filter copies default MAC to filter->l2_addr. So,
1663          * if the MAC that's been programmed now is a different one, then,
1664          * copy that addr to filter->l2_addr
1665          */
1666         if (mac_addr)
1667                 memcpy(filter->l2_addr, mac_addr, RTE_ETHER_ADDR_LEN);
1668         filter->flags |= HWRM_CFA_L2_FILTER_ALLOC_INPUT_FLAGS_OUTERMOST;
1669
1670         rc = bnxt_hwrm_set_l2_filter(bp, vnic->fw_vnic_id, filter);
1671         if (!rc) {
1672                 filter->mac_index = index;
1673                 if (filter->mac_index == 0)
1674                         STAILQ_INSERT_HEAD(&vnic->filter, filter, next);
1675                 else
1676                         STAILQ_INSERT_TAIL(&vnic->filter, filter, next);
1677         } else {
1678                 bnxt_free_filter(bp, filter);
1679         }
1680
1681         return rc;
1682 }
1683
1684 static int bnxt_mac_addr_add_op(struct rte_eth_dev *eth_dev,
1685                                 struct rte_ether_addr *mac_addr,
1686                                 uint32_t index, uint32_t pool)
1687 {
1688         struct bnxt *bp = eth_dev->data->dev_private;
1689         struct bnxt_vnic_info *vnic = &bp->vnic_info[pool];
1690         int rc = 0;
1691
1692         rc = is_bnxt_in_error(bp);
1693         if (rc)
1694                 return rc;
1695
1696         if (BNXT_VF(bp) && !BNXT_VF_IS_TRUSTED(bp)) {
1697                 PMD_DRV_LOG(ERR, "Cannot add MAC address to a VF interface\n");
1698                 return -ENOTSUP;
1699         }
1700
1701         if (!vnic) {
1702                 PMD_DRV_LOG(ERR, "VNIC not found for pool %d!\n", pool);
1703                 return -EINVAL;
1704         }
1705
1706         /* Filter settings will get applied when port is started */
1707         if (!eth_dev->data->dev_started)
1708                 return 0;
1709
1710         rc = bnxt_add_mac_filter(bp, vnic, mac_addr, index, pool);
1711
1712         return rc;
1713 }
1714
1715 int bnxt_link_update_op(struct rte_eth_dev *eth_dev, int wait_to_complete)
1716 {
1717         int rc = 0;
1718         struct bnxt *bp = eth_dev->data->dev_private;
1719         struct rte_eth_link new;
1720         int cnt = wait_to_complete ? BNXT_MAX_LINK_WAIT_CNT :
1721                         BNXT_MIN_LINK_WAIT_CNT;
1722
1723         rc = is_bnxt_in_error(bp);
1724         if (rc)
1725                 return rc;
1726
1727         memset(&new, 0, sizeof(new));
1728         do {
1729                 /* Retrieve link info from hardware */
1730                 rc = bnxt_get_hwrm_link_config(bp, &new);
1731                 if (rc) {
1732                         new.link_speed = ETH_LINK_SPEED_100M;
1733                         new.link_duplex = ETH_LINK_FULL_DUPLEX;
1734                         PMD_DRV_LOG(ERR,
1735                                 "Failed to retrieve link rc = 0x%x!\n", rc);
1736                         goto out;
1737                 }
1738
1739                 if (!wait_to_complete || new.link_status)
1740                         break;
1741
1742                 rte_delay_ms(BNXT_LINK_WAIT_INTERVAL);
1743         } while (cnt--);
1744
1745         /* Only single function PF can bring phy down.
1746          * When port is stopped, report link down for VF/MH/NPAR functions.
1747          */
1748         if (!BNXT_SINGLE_PF(bp) && !eth_dev->data->dev_started)
1749                 memset(&new, 0, sizeof(new));
1750
1751 out:
1752         /* Timed out or success */
1753         if (new.link_status != eth_dev->data->dev_link.link_status ||
1754             new.link_speed != eth_dev->data->dev_link.link_speed) {
1755                 rte_eth_linkstatus_set(eth_dev, &new);
1756
1757                 rte_eth_dev_callback_process(eth_dev,
1758                                              RTE_ETH_EVENT_INTR_LSC,
1759                                              NULL);
1760
1761                 bnxt_print_link_info(eth_dev);
1762         }
1763
1764         return rc;
1765 }
1766
1767 static int bnxt_promiscuous_enable_op(struct rte_eth_dev *eth_dev)
1768 {
1769         struct bnxt *bp = eth_dev->data->dev_private;
1770         struct bnxt_vnic_info *vnic;
1771         uint32_t old_flags;
1772         int rc;
1773
1774         rc = is_bnxt_in_error(bp);
1775         if (rc)
1776                 return rc;
1777
1778         /* Filter settings will get applied when port is started */
1779         if (!eth_dev->data->dev_started)
1780                 return 0;
1781
1782         if (bp->vnic_info == NULL)
1783                 return 0;
1784
1785         vnic = BNXT_GET_DEFAULT_VNIC(bp);
1786
1787         old_flags = vnic->flags;
1788         vnic->flags |= BNXT_VNIC_INFO_PROMISC;
1789         rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1790         if (rc != 0)
1791                 vnic->flags = old_flags;
1792
1793         return rc;
1794 }
1795
1796 static int bnxt_promiscuous_disable_op(struct rte_eth_dev *eth_dev)
1797 {
1798         struct bnxt *bp = eth_dev->data->dev_private;
1799         struct bnxt_vnic_info *vnic;
1800         uint32_t old_flags;
1801         int rc;
1802
1803         rc = is_bnxt_in_error(bp);
1804         if (rc)
1805                 return rc;
1806
1807         /* Filter settings will get applied when port is started */
1808         if (!eth_dev->data->dev_started)
1809                 return 0;
1810
1811         if (bp->vnic_info == NULL)
1812                 return 0;
1813
1814         vnic = BNXT_GET_DEFAULT_VNIC(bp);
1815
1816         old_flags = vnic->flags;
1817         vnic->flags &= ~BNXT_VNIC_INFO_PROMISC;
1818         rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1819         if (rc != 0)
1820                 vnic->flags = old_flags;
1821
1822         return rc;
1823 }
1824
1825 static int bnxt_allmulticast_enable_op(struct rte_eth_dev *eth_dev)
1826 {
1827         struct bnxt *bp = eth_dev->data->dev_private;
1828         struct bnxt_vnic_info *vnic;
1829         uint32_t old_flags;
1830         int rc;
1831
1832         rc = is_bnxt_in_error(bp);
1833         if (rc)
1834                 return rc;
1835
1836         /* Filter settings will get applied when port is started */
1837         if (!eth_dev->data->dev_started)
1838                 return 0;
1839
1840         if (bp->vnic_info == NULL)
1841                 return 0;
1842
1843         vnic = BNXT_GET_DEFAULT_VNIC(bp);
1844
1845         old_flags = vnic->flags;
1846         vnic->flags |= BNXT_VNIC_INFO_ALLMULTI;
1847         rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1848         if (rc != 0)
1849                 vnic->flags = old_flags;
1850
1851         return rc;
1852 }
1853
1854 static int bnxt_allmulticast_disable_op(struct rte_eth_dev *eth_dev)
1855 {
1856         struct bnxt *bp = eth_dev->data->dev_private;
1857         struct bnxt_vnic_info *vnic;
1858         uint32_t old_flags;
1859         int rc;
1860
1861         rc = is_bnxt_in_error(bp);
1862         if (rc)
1863                 return rc;
1864
1865         /* Filter settings will get applied when port is started */
1866         if (!eth_dev->data->dev_started)
1867                 return 0;
1868
1869         if (bp->vnic_info == NULL)
1870                 return 0;
1871
1872         vnic = BNXT_GET_DEFAULT_VNIC(bp);
1873
1874         old_flags = vnic->flags;
1875         vnic->flags &= ~BNXT_VNIC_INFO_ALLMULTI;
1876         rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1877         if (rc != 0)
1878                 vnic->flags = old_flags;
1879
1880         return rc;
1881 }
1882
1883 /* Return bnxt_rx_queue pointer corresponding to a given rxq. */
1884 static struct bnxt_rx_queue *bnxt_qid_to_rxq(struct bnxt *bp, uint16_t qid)
1885 {
1886         if (qid >= bp->rx_nr_rings)
1887                 return NULL;
1888
1889         return bp->eth_dev->data->rx_queues[qid];
1890 }
1891
1892 /* Return rxq corresponding to a given rss table ring/group ID. */
1893 static uint16_t bnxt_rss_to_qid(struct bnxt *bp, uint16_t fwr)
1894 {
1895         struct bnxt_rx_queue *rxq;
1896         unsigned int i;
1897
1898         if (!BNXT_HAS_RING_GRPS(bp)) {
1899                 for (i = 0; i < bp->rx_nr_rings; i++) {
1900                         rxq = bp->eth_dev->data->rx_queues[i];
1901                         if (rxq->rx_ring->rx_ring_struct->fw_ring_id == fwr)
1902                                 return rxq->index;
1903                 }
1904         } else {
1905                 for (i = 0; i < bp->rx_nr_rings; i++) {
1906                         if (bp->grp_info[i].fw_grp_id == fwr)
1907                                 return i;
1908                 }
1909         }
1910
1911         return INVALID_HW_RING_ID;
1912 }
1913
1914 static int bnxt_reta_update_op(struct rte_eth_dev *eth_dev,
1915                             struct rte_eth_rss_reta_entry64 *reta_conf,
1916                             uint16_t reta_size)
1917 {
1918         struct bnxt *bp = eth_dev->data->dev_private;
1919         struct rte_eth_conf *dev_conf = &bp->eth_dev->data->dev_conf;
1920         struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
1921         uint16_t tbl_size = bnxt_rss_hash_tbl_size(bp);
1922         uint16_t idx, sft;
1923         int i, rc;
1924
1925         rc = is_bnxt_in_error(bp);
1926         if (rc)
1927                 return rc;
1928
1929         if (!vnic->rss_table)
1930                 return -EINVAL;
1931
1932         if (!(dev_conf->rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG))
1933                 return -EINVAL;
1934
1935         if (reta_size != tbl_size) {
1936                 PMD_DRV_LOG(ERR, "The configured hash table lookup size "
1937                         "(%d) must equal the size supported by the hardware "
1938                         "(%d)\n", reta_size, tbl_size);
1939                 return -EINVAL;
1940         }
1941
1942         for (i = 0; i < reta_size; i++) {
1943                 struct bnxt_rx_queue *rxq;
1944
1945                 idx = i / RTE_RETA_GROUP_SIZE;
1946                 sft = i % RTE_RETA_GROUP_SIZE;
1947
1948                 if (!(reta_conf[idx].mask & (1ULL << sft)))
1949                         continue;
1950
1951                 rxq = bnxt_qid_to_rxq(bp, reta_conf[idx].reta[sft]);
1952                 if (!rxq) {
1953                         PMD_DRV_LOG(ERR, "Invalid ring in reta_conf.\n");
1954                         return -EINVAL;
1955                 }
1956
1957                 if (BNXT_CHIP_P5(bp)) {
1958                         vnic->rss_table[i * 2] =
1959                                 rxq->rx_ring->rx_ring_struct->fw_ring_id;
1960                         vnic->rss_table[i * 2 + 1] =
1961                                 rxq->cp_ring->cp_ring_struct->fw_ring_id;
1962                 } else {
1963                         vnic->rss_table[i] =
1964                             vnic->fw_grp_ids[reta_conf[idx].reta[sft]];
1965                 }
1966         }
1967
1968         rc = bnxt_hwrm_vnic_rss_cfg(bp, vnic);
1969         return rc;
1970 }
1971
1972 static int bnxt_reta_query_op(struct rte_eth_dev *eth_dev,
1973                               struct rte_eth_rss_reta_entry64 *reta_conf,
1974                               uint16_t reta_size)
1975 {
1976         struct bnxt *bp = eth_dev->data->dev_private;
1977         struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
1978         uint16_t tbl_size = bnxt_rss_hash_tbl_size(bp);
1979         uint16_t idx, sft, i;
1980         int rc;
1981
1982         rc = is_bnxt_in_error(bp);
1983         if (rc)
1984                 return rc;
1985
1986         /* Retrieve from the default VNIC */
1987         if (!vnic)
1988                 return -EINVAL;
1989         if (!vnic->rss_table)
1990                 return -EINVAL;
1991
1992         if (reta_size != tbl_size) {
1993                 PMD_DRV_LOG(ERR, "The configured hash table lookup size "
1994                         "(%d) must equal the size supported by the hardware "
1995                         "(%d)\n", reta_size, tbl_size);
1996                 return -EINVAL;
1997         }
1998
1999         for (idx = 0, i = 0; i < reta_size; i++) {
2000                 idx = i / RTE_RETA_GROUP_SIZE;
2001                 sft = i % RTE_RETA_GROUP_SIZE;
2002
2003                 if (reta_conf[idx].mask & (1ULL << sft)) {
2004                         uint16_t qid;
2005
2006                         if (BNXT_CHIP_P5(bp))
2007                                 qid = bnxt_rss_to_qid(bp,
2008                                                       vnic->rss_table[i * 2]);
2009                         else
2010                                 qid = bnxt_rss_to_qid(bp, vnic->rss_table[i]);
2011
2012                         if (qid == INVALID_HW_RING_ID) {
2013                                 PMD_DRV_LOG(ERR, "Inv. entry in rss table.\n");
2014                                 return -EINVAL;
2015                         }
2016                         reta_conf[idx].reta[sft] = qid;
2017                 }
2018         }
2019
2020         return 0;
2021 }
2022
2023 static int bnxt_rss_hash_update_op(struct rte_eth_dev *eth_dev,
2024                                    struct rte_eth_rss_conf *rss_conf)
2025 {
2026         struct bnxt *bp = eth_dev->data->dev_private;
2027         struct rte_eth_conf *dev_conf = &bp->eth_dev->data->dev_conf;
2028         struct bnxt_vnic_info *vnic;
2029         int rc;
2030
2031         rc = is_bnxt_in_error(bp);
2032         if (rc)
2033                 return rc;
2034
2035         /*
2036          * If RSS enablement were different than dev_configure,
2037          * then return -EINVAL
2038          */
2039         if (dev_conf->rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG) {
2040                 if (!rss_conf->rss_hf)
2041                         PMD_DRV_LOG(ERR, "Hash type NONE\n");
2042         } else {
2043                 if (rss_conf->rss_hf & BNXT_ETH_RSS_SUPPORT)
2044                         return -EINVAL;
2045         }
2046
2047         bp->flags |= BNXT_FLAG_UPDATE_HASH;
2048         memcpy(&eth_dev->data->dev_conf.rx_adv_conf.rss_conf,
2049                rss_conf,
2050                sizeof(*rss_conf));
2051
2052         /* Update the default RSS VNIC(s) */
2053         vnic = BNXT_GET_DEFAULT_VNIC(bp);
2054         vnic->hash_type = bnxt_rte_to_hwrm_hash_types(rss_conf->rss_hf);
2055         vnic->hash_mode =
2056                 bnxt_rte_to_hwrm_hash_level(bp, rss_conf->rss_hf,
2057                                             ETH_RSS_LEVEL(rss_conf->rss_hf));
2058
2059         /*
2060          * If hashkey is not specified, use the previously configured
2061          * hashkey
2062          */
2063         if (!rss_conf->rss_key)
2064                 goto rss_config;
2065
2066         if (rss_conf->rss_key_len != HW_HASH_KEY_SIZE) {
2067                 PMD_DRV_LOG(ERR,
2068                             "Invalid hashkey length, should be 16 bytes\n");
2069                 return -EINVAL;
2070         }
2071         memcpy(vnic->rss_hash_key, rss_conf->rss_key, rss_conf->rss_key_len);
2072
2073 rss_config:
2074         rc = bnxt_hwrm_vnic_rss_cfg(bp, vnic);
2075         return rc;
2076 }
2077
2078 static int bnxt_rss_hash_conf_get_op(struct rte_eth_dev *eth_dev,
2079                                      struct rte_eth_rss_conf *rss_conf)
2080 {
2081         struct bnxt *bp = eth_dev->data->dev_private;
2082         struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
2083         int len, rc;
2084         uint32_t hash_types;
2085
2086         rc = is_bnxt_in_error(bp);
2087         if (rc)
2088                 return rc;
2089
2090         /* RSS configuration is the same for all VNICs */
2091         if (vnic && vnic->rss_hash_key) {
2092                 if (rss_conf->rss_key) {
2093                         len = rss_conf->rss_key_len <= HW_HASH_KEY_SIZE ?
2094                               rss_conf->rss_key_len : HW_HASH_KEY_SIZE;
2095                         memcpy(rss_conf->rss_key, vnic->rss_hash_key, len);
2096                 }
2097
2098                 hash_types = vnic->hash_type;
2099                 rss_conf->rss_hf = 0;
2100                 if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV4) {
2101                         rss_conf->rss_hf |= ETH_RSS_IPV4;
2102                         hash_types &= ~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV4;
2103                 }
2104                 if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV4) {
2105                         rss_conf->rss_hf |= ETH_RSS_NONFRAG_IPV4_TCP;
2106                         hash_types &=
2107                                 ~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV4;
2108                 }
2109                 if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV4) {
2110                         rss_conf->rss_hf |= ETH_RSS_NONFRAG_IPV4_UDP;
2111                         hash_types &=
2112                                 ~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV4;
2113                 }
2114                 if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV6) {
2115                         rss_conf->rss_hf |= ETH_RSS_IPV6;
2116                         hash_types &= ~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV6;
2117                 }
2118                 if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV6) {
2119                         rss_conf->rss_hf |= ETH_RSS_NONFRAG_IPV6_TCP;
2120                         hash_types &=
2121                                 ~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV6;
2122                 }
2123                 if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV6) {
2124                         rss_conf->rss_hf |= ETH_RSS_NONFRAG_IPV6_UDP;
2125                         hash_types &=
2126                                 ~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV6;
2127                 }
2128
2129                 rss_conf->rss_hf |=
2130                         bnxt_hwrm_to_rte_rss_level(bp, vnic->hash_mode);
2131
2132                 if (hash_types) {
2133                         PMD_DRV_LOG(ERR,
2134                                 "Unknown RSS config from firmware (%08x), RSS disabled",
2135                                 vnic->hash_type);
2136                         return -ENOTSUP;
2137                 }
2138         } else {
2139                 rss_conf->rss_hf = 0;
2140         }
2141         return 0;
2142 }
2143
2144 static int bnxt_flow_ctrl_get_op(struct rte_eth_dev *dev,
2145                                struct rte_eth_fc_conf *fc_conf)
2146 {
2147         struct bnxt *bp = dev->data->dev_private;
2148         struct rte_eth_link link_info;
2149         int rc;
2150
2151         rc = is_bnxt_in_error(bp);
2152         if (rc)
2153                 return rc;
2154
2155         rc = bnxt_get_hwrm_link_config(bp, &link_info);
2156         if (rc)
2157                 return rc;
2158
2159         memset(fc_conf, 0, sizeof(*fc_conf));
2160         if (bp->link_info->auto_pause)
2161                 fc_conf->autoneg = 1;
2162         switch (bp->link_info->pause) {
2163         case 0:
2164                 fc_conf->mode = RTE_FC_NONE;
2165                 break;
2166         case HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_TX:
2167                 fc_conf->mode = RTE_FC_TX_PAUSE;
2168                 break;
2169         case HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_RX:
2170                 fc_conf->mode = RTE_FC_RX_PAUSE;
2171                 break;
2172         case (HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_TX |
2173                         HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_RX):
2174                 fc_conf->mode = RTE_FC_FULL;
2175                 break;
2176         }
2177         return 0;
2178 }
2179
2180 static int bnxt_flow_ctrl_set_op(struct rte_eth_dev *dev,
2181                                struct rte_eth_fc_conf *fc_conf)
2182 {
2183         struct bnxt *bp = dev->data->dev_private;
2184         int rc;
2185
2186         rc = is_bnxt_in_error(bp);
2187         if (rc)
2188                 return rc;
2189
2190         if (!BNXT_SINGLE_PF(bp) || BNXT_VF(bp)) {
2191                 PMD_DRV_LOG(ERR, "Flow Control Settings cannot be modified\n");
2192                 return -ENOTSUP;
2193         }
2194
2195         switch (fc_conf->mode) {
2196         case RTE_FC_NONE:
2197                 bp->link_info->auto_pause = 0;
2198                 bp->link_info->force_pause = 0;
2199                 break;
2200         case RTE_FC_RX_PAUSE:
2201                 if (fc_conf->autoneg) {
2202                         bp->link_info->auto_pause =
2203                                         HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_RX;
2204                         bp->link_info->force_pause = 0;
2205                 } else {
2206                         bp->link_info->auto_pause = 0;
2207                         bp->link_info->force_pause =
2208                                         HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_RX;
2209                 }
2210                 break;
2211         case RTE_FC_TX_PAUSE:
2212                 if (fc_conf->autoneg) {
2213                         bp->link_info->auto_pause =
2214                                         HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_TX;
2215                         bp->link_info->force_pause = 0;
2216                 } else {
2217                         bp->link_info->auto_pause = 0;
2218                         bp->link_info->force_pause =
2219                                         HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_TX;
2220                 }
2221                 break;
2222         case RTE_FC_FULL:
2223                 if (fc_conf->autoneg) {
2224                         bp->link_info->auto_pause =
2225                                         HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_TX |
2226                                         HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_RX;
2227                         bp->link_info->force_pause = 0;
2228                 } else {
2229                         bp->link_info->auto_pause = 0;
2230                         bp->link_info->force_pause =
2231                                         HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_TX |
2232                                         HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_RX;
2233                 }
2234                 break;
2235         }
2236         return bnxt_set_hwrm_link_config(bp, true);
2237 }
2238
2239 /* Add UDP tunneling port */
2240 static int
2241 bnxt_udp_tunnel_port_add_op(struct rte_eth_dev *eth_dev,
2242                          struct rte_eth_udp_tunnel *udp_tunnel)
2243 {
2244         struct bnxt *bp = eth_dev->data->dev_private;
2245         uint16_t tunnel_type = 0;
2246         int rc = 0;
2247
2248         rc = is_bnxt_in_error(bp);
2249         if (rc)
2250                 return rc;
2251
2252         switch (udp_tunnel->prot_type) {
2253         case RTE_TUNNEL_TYPE_VXLAN:
2254                 if (bp->vxlan_port_cnt) {
2255                         PMD_DRV_LOG(ERR, "Tunnel Port %d already programmed\n",
2256                                 udp_tunnel->udp_port);
2257                         if (bp->vxlan_port != udp_tunnel->udp_port) {
2258                                 PMD_DRV_LOG(ERR, "Only one port allowed\n");
2259                                 return -ENOSPC;
2260                         }
2261                         bp->vxlan_port_cnt++;
2262                         return 0;
2263                 }
2264                 tunnel_type =
2265                         HWRM_TUNNEL_DST_PORT_ALLOC_INPUT_TUNNEL_TYPE_VXLAN;
2266                 bp->vxlan_port_cnt++;
2267                 break;
2268         case RTE_TUNNEL_TYPE_GENEVE:
2269                 if (bp->geneve_port_cnt) {
2270                         PMD_DRV_LOG(ERR, "Tunnel Port %d already programmed\n",
2271                                 udp_tunnel->udp_port);
2272                         if (bp->geneve_port != udp_tunnel->udp_port) {
2273                                 PMD_DRV_LOG(ERR, "Only one port allowed\n");
2274                                 return -ENOSPC;
2275                         }
2276                         bp->geneve_port_cnt++;
2277                         return 0;
2278                 }
2279                 tunnel_type =
2280                         HWRM_TUNNEL_DST_PORT_ALLOC_INPUT_TUNNEL_TYPE_GENEVE;
2281                 bp->geneve_port_cnt++;
2282                 break;
2283         default:
2284                 PMD_DRV_LOG(ERR, "Tunnel type is not supported\n");
2285                 return -ENOTSUP;
2286         }
2287         rc = bnxt_hwrm_tunnel_dst_port_alloc(bp, udp_tunnel->udp_port,
2288                                              tunnel_type);
2289         return rc;
2290 }
2291
2292 static int
2293 bnxt_udp_tunnel_port_del_op(struct rte_eth_dev *eth_dev,
2294                          struct rte_eth_udp_tunnel *udp_tunnel)
2295 {
2296         struct bnxt *bp = eth_dev->data->dev_private;
2297         uint16_t tunnel_type = 0;
2298         uint16_t port = 0;
2299         int rc = 0;
2300
2301         rc = is_bnxt_in_error(bp);
2302         if (rc)
2303                 return rc;
2304
2305         switch (udp_tunnel->prot_type) {
2306         case RTE_TUNNEL_TYPE_VXLAN:
2307                 if (!bp->vxlan_port_cnt) {
2308                         PMD_DRV_LOG(ERR, "No Tunnel port configured yet\n");
2309                         return -EINVAL;
2310                 }
2311                 if (bp->vxlan_port != udp_tunnel->udp_port) {
2312                         PMD_DRV_LOG(ERR, "Req Port: %d. Configured port: %d\n",
2313                                 udp_tunnel->udp_port, bp->vxlan_port);
2314                         return -EINVAL;
2315                 }
2316                 if (--bp->vxlan_port_cnt)
2317                         return 0;
2318
2319                 tunnel_type =
2320                         HWRM_TUNNEL_DST_PORT_FREE_INPUT_TUNNEL_TYPE_VXLAN;
2321                 port = bp->vxlan_fw_dst_port_id;
2322                 break;
2323         case RTE_TUNNEL_TYPE_GENEVE:
2324                 if (!bp->geneve_port_cnt) {
2325                         PMD_DRV_LOG(ERR, "No Tunnel port configured yet\n");
2326                         return -EINVAL;
2327                 }
2328                 if (bp->geneve_port != udp_tunnel->udp_port) {
2329                         PMD_DRV_LOG(ERR, "Req Port: %d. Configured port: %d\n",
2330                                 udp_tunnel->udp_port, bp->geneve_port);
2331                         return -EINVAL;
2332                 }
2333                 if (--bp->geneve_port_cnt)
2334                         return 0;
2335
2336                 tunnel_type =
2337                         HWRM_TUNNEL_DST_PORT_FREE_INPUT_TUNNEL_TYPE_GENEVE;
2338                 port = bp->geneve_fw_dst_port_id;
2339                 break;
2340         default:
2341                 PMD_DRV_LOG(ERR, "Tunnel type is not supported\n");
2342                 return -ENOTSUP;
2343         }
2344
2345         rc = bnxt_hwrm_tunnel_dst_port_free(bp, port, tunnel_type);
2346         return rc;
2347 }
2348
2349 static int bnxt_del_vlan_filter(struct bnxt *bp, uint16_t vlan_id)
2350 {
2351         struct bnxt_filter_info *filter;
2352         struct bnxt_vnic_info *vnic;
2353         int rc = 0;
2354         uint32_t chk = HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN;
2355
2356         vnic = BNXT_GET_DEFAULT_VNIC(bp);
2357         filter = STAILQ_FIRST(&vnic->filter);
2358         while (filter) {
2359                 /* Search for this matching MAC+VLAN filter */
2360                 if (bnxt_vlan_filter_exists(bp, filter, chk, vlan_id)) {
2361                         /* Delete the filter */
2362                         rc = bnxt_hwrm_clear_l2_filter(bp, filter);
2363                         if (rc)
2364                                 return rc;
2365                         STAILQ_REMOVE(&vnic->filter, filter,
2366                                       bnxt_filter_info, next);
2367                         bnxt_free_filter(bp, filter);
2368                         PMD_DRV_LOG(INFO,
2369                                     "Deleted vlan filter for %d\n",
2370                                     vlan_id);
2371                         return 0;
2372                 }
2373                 filter = STAILQ_NEXT(filter, next);
2374         }
2375         return -ENOENT;
2376 }
2377
2378 static int bnxt_add_vlan_filter(struct bnxt *bp, uint16_t vlan_id)
2379 {
2380         struct bnxt_filter_info *filter;
2381         struct bnxt_vnic_info *vnic;
2382         int rc = 0;
2383         uint32_t en = HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN |
2384                 HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN_MASK;
2385         uint32_t chk = HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN;
2386
2387         /* Implementation notes on the use of VNIC in this command:
2388          *
2389          * By default, these filters belong to default vnic for the function.
2390          * Once these filters are set up, only destination VNIC can be modified.
2391          * If the destination VNIC is not specified in this command,
2392          * then the HWRM shall only create an l2 context id.
2393          */
2394
2395         vnic = BNXT_GET_DEFAULT_VNIC(bp);
2396         filter = STAILQ_FIRST(&vnic->filter);
2397         /* Check if the VLAN has already been added */
2398         while (filter) {
2399                 if (bnxt_vlan_filter_exists(bp, filter, chk, vlan_id))
2400                         return -EEXIST;
2401
2402                 filter = STAILQ_NEXT(filter, next);
2403         }
2404
2405         /* No match found. Alloc a fresh filter and issue the L2_FILTER_ALLOC
2406          * command to create MAC+VLAN filter with the right flags, enables set.
2407          */
2408         filter = bnxt_alloc_filter(bp);
2409         if (!filter) {
2410                 PMD_DRV_LOG(ERR,
2411                             "MAC/VLAN filter alloc failed\n");
2412                 return -ENOMEM;
2413         }
2414         /* MAC + VLAN ID filter */
2415         /* If l2_ivlan == 0 and l2_ivlan_mask != 0, only
2416          * untagged packets are received
2417          *
2418          * If l2_ivlan != 0 and l2_ivlan_mask != 0, untagged
2419          * packets and only the programmed vlan's packets are received
2420          */
2421         filter->l2_ivlan = vlan_id;
2422         filter->l2_ivlan_mask = 0x0FFF;
2423         filter->enables |= en;
2424         filter->flags |= HWRM_CFA_L2_FILTER_ALLOC_INPUT_FLAGS_OUTERMOST;
2425
2426         rc = bnxt_hwrm_set_l2_filter(bp, vnic->fw_vnic_id, filter);
2427         if (rc) {
2428                 /* Free the newly allocated filter as we were
2429                  * not able to create the filter in hardware.
2430                  */
2431                 bnxt_free_filter(bp, filter);
2432                 return rc;
2433         }
2434
2435         filter->mac_index = 0;
2436         /* Add this new filter to the list */
2437         if (vlan_id == 0)
2438                 STAILQ_INSERT_HEAD(&vnic->filter, filter, next);
2439         else
2440                 STAILQ_INSERT_TAIL(&vnic->filter, filter, next);
2441
2442         PMD_DRV_LOG(INFO,
2443                     "Added Vlan filter for %d\n", vlan_id);
2444         return rc;
2445 }
2446
2447 static int bnxt_vlan_filter_set_op(struct rte_eth_dev *eth_dev,
2448                 uint16_t vlan_id, int on)
2449 {
2450         struct bnxt *bp = eth_dev->data->dev_private;
2451         int rc;
2452
2453         rc = is_bnxt_in_error(bp);
2454         if (rc)
2455                 return rc;
2456
2457         if (!eth_dev->data->dev_started) {
2458                 PMD_DRV_LOG(ERR, "port must be started before setting vlan\n");
2459                 return -EINVAL;
2460         }
2461
2462         /* These operations apply to ALL existing MAC/VLAN filters */
2463         if (on)
2464                 return bnxt_add_vlan_filter(bp, vlan_id);
2465         else
2466                 return bnxt_del_vlan_filter(bp, vlan_id);
2467 }
2468
2469 static int bnxt_del_dflt_mac_filter(struct bnxt *bp,
2470                                     struct bnxt_vnic_info *vnic)
2471 {
2472         struct bnxt_filter_info *filter;
2473         int rc;
2474
2475         filter = STAILQ_FIRST(&vnic->filter);
2476         while (filter) {
2477                 if (filter->mac_index == 0 &&
2478                     !memcmp(filter->l2_addr, bp->mac_addr,
2479                             RTE_ETHER_ADDR_LEN)) {
2480                         rc = bnxt_hwrm_clear_l2_filter(bp, filter);
2481                         if (!rc) {
2482                                 STAILQ_REMOVE(&vnic->filter, filter,
2483                                               bnxt_filter_info, next);
2484                                 bnxt_free_filter(bp, filter);
2485                         }
2486                         return rc;
2487                 }
2488                 filter = STAILQ_NEXT(filter, next);
2489         }
2490         return 0;
2491 }
2492
2493 static int
2494 bnxt_config_vlan_hw_filter(struct bnxt *bp, uint64_t rx_offloads)
2495 {
2496         struct bnxt_vnic_info *vnic;
2497         unsigned int i;
2498         int rc;
2499
2500         vnic = BNXT_GET_DEFAULT_VNIC(bp);
2501         if (!(rx_offloads & DEV_RX_OFFLOAD_VLAN_FILTER)) {
2502                 /* Remove any VLAN filters programmed */
2503                 for (i = 0; i < RTE_ETHER_MAX_VLAN_ID; i++)
2504                         bnxt_del_vlan_filter(bp, i);
2505
2506                 rc = bnxt_add_mac_filter(bp, vnic, NULL, 0, 0);
2507                 if (rc)
2508                         return rc;
2509         } else {
2510                 /* Default filter will allow packets that match the
2511                  * dest mac. So, it has to be deleted, otherwise, we
2512                  * will endup receiving vlan packets for which the
2513                  * filter is not programmed, when hw-vlan-filter
2514                  * configuration is ON
2515                  */
2516                 bnxt_del_dflt_mac_filter(bp, vnic);
2517                 /* This filter will allow only untagged packets */
2518                 bnxt_add_vlan_filter(bp, 0);
2519         }
2520         PMD_DRV_LOG(DEBUG, "VLAN Filtering: %d\n",
2521                     !!(rx_offloads & DEV_RX_OFFLOAD_VLAN_FILTER));
2522
2523         return 0;
2524 }
2525
2526 static int bnxt_free_one_vnic(struct bnxt *bp, uint16_t vnic_id)
2527 {
2528         struct bnxt_vnic_info *vnic = &bp->vnic_info[vnic_id];
2529         unsigned int i;
2530         int rc;
2531
2532         /* Destroy vnic filters and vnic */
2533         if (bp->eth_dev->data->dev_conf.rxmode.offloads &
2534             DEV_RX_OFFLOAD_VLAN_FILTER) {
2535                 for (i = 0; i < RTE_ETHER_MAX_VLAN_ID; i++)
2536                         bnxt_del_vlan_filter(bp, i);
2537         }
2538         bnxt_del_dflt_mac_filter(bp, vnic);
2539
2540         rc = bnxt_hwrm_vnic_ctx_free(bp, vnic);
2541         if (rc)
2542                 return rc;
2543
2544         rc = bnxt_hwrm_vnic_free(bp, vnic);
2545         if (rc)
2546                 return rc;
2547
2548         rte_free(vnic->fw_grp_ids);
2549         vnic->fw_grp_ids = NULL;
2550
2551         vnic->rx_queue_cnt = 0;
2552
2553         return 0;
2554 }
2555
2556 static int
2557 bnxt_config_vlan_hw_stripping(struct bnxt *bp, uint64_t rx_offloads)
2558 {
2559         struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
2560         int rc;
2561
2562         /* Destroy, recreate and reconfigure the default vnic */
2563         rc = bnxt_free_one_vnic(bp, 0);
2564         if (rc)
2565                 return rc;
2566
2567         /* default vnic 0 */
2568         rc = bnxt_setup_one_vnic(bp, 0);
2569         if (rc)
2570                 return rc;
2571
2572         if (bp->eth_dev->data->dev_conf.rxmode.offloads &
2573             DEV_RX_OFFLOAD_VLAN_FILTER) {
2574                 rc = bnxt_add_vlan_filter(bp, 0);
2575                 if (rc)
2576                         return rc;
2577                 rc = bnxt_restore_vlan_filters(bp);
2578                 if (rc)
2579                         return rc;
2580         } else {
2581                 rc = bnxt_add_mac_filter(bp, vnic, NULL, 0, 0);
2582                 if (rc)
2583                         return rc;
2584         }
2585
2586         rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
2587         if (rc)
2588                 return rc;
2589
2590         PMD_DRV_LOG(DEBUG, "VLAN Strip Offload: %d\n",
2591                     !!(rx_offloads & DEV_RX_OFFLOAD_VLAN_STRIP));
2592
2593         return rc;
2594 }
2595
2596 static int
2597 bnxt_vlan_offload_set_op(struct rte_eth_dev *dev, int mask)
2598 {
2599         uint64_t rx_offloads = dev->data->dev_conf.rxmode.offloads;
2600         struct bnxt *bp = dev->data->dev_private;
2601         int rc;
2602
2603         rc = is_bnxt_in_error(bp);
2604         if (rc)
2605                 return rc;
2606
2607         /* Filter settings will get applied when port is started */
2608         if (!dev->data->dev_started)
2609                 return 0;
2610
2611         if (mask & ETH_VLAN_FILTER_MASK) {
2612                 /* Enable or disable VLAN filtering */
2613                 rc = bnxt_config_vlan_hw_filter(bp, rx_offloads);
2614                 if (rc)
2615                         return rc;
2616         }
2617
2618         if (mask & ETH_VLAN_STRIP_MASK) {
2619                 /* Enable or disable VLAN stripping */
2620                 rc = bnxt_config_vlan_hw_stripping(bp, rx_offloads);
2621                 if (rc)
2622                         return rc;
2623         }
2624
2625         if (mask & ETH_VLAN_EXTEND_MASK) {
2626                 if (rx_offloads & DEV_RX_OFFLOAD_VLAN_EXTEND)
2627                         PMD_DRV_LOG(DEBUG, "Extend VLAN supported\n");
2628                 else
2629                         PMD_DRV_LOG(INFO, "Extend VLAN unsupported\n");
2630         }
2631
2632         return 0;
2633 }
2634
2635 static int
2636 bnxt_vlan_tpid_set_op(struct rte_eth_dev *dev, enum rte_vlan_type vlan_type,
2637                       uint16_t tpid)
2638 {
2639         struct bnxt *bp = dev->data->dev_private;
2640         int qinq = dev->data->dev_conf.rxmode.offloads &
2641                    DEV_RX_OFFLOAD_VLAN_EXTEND;
2642
2643         if (vlan_type != ETH_VLAN_TYPE_INNER &&
2644             vlan_type != ETH_VLAN_TYPE_OUTER) {
2645                 PMD_DRV_LOG(ERR,
2646                             "Unsupported vlan type.");
2647                 return -EINVAL;
2648         }
2649         if (!qinq) {
2650                 PMD_DRV_LOG(ERR,
2651                             "QinQ not enabled. Needs to be ON as we can "
2652                             "accelerate only outer vlan\n");
2653                 return -EINVAL;
2654         }
2655
2656         if (vlan_type == ETH_VLAN_TYPE_OUTER) {
2657                 switch (tpid) {
2658                 case RTE_ETHER_TYPE_QINQ:
2659                         bp->outer_tpid_bd =
2660                                 TX_BD_LONG_CFA_META_VLAN_TPID_TPID88A8;
2661                                 break;
2662                 case RTE_ETHER_TYPE_VLAN:
2663                         bp->outer_tpid_bd =
2664                                 TX_BD_LONG_CFA_META_VLAN_TPID_TPID8100;
2665                                 break;
2666                 case RTE_ETHER_TYPE_QINQ1:
2667                         bp->outer_tpid_bd =
2668                                 TX_BD_LONG_CFA_META_VLAN_TPID_TPID9100;
2669                                 break;
2670                 case RTE_ETHER_TYPE_QINQ2:
2671                         bp->outer_tpid_bd =
2672                                 TX_BD_LONG_CFA_META_VLAN_TPID_TPID9200;
2673                                 break;
2674                 case RTE_ETHER_TYPE_QINQ3:
2675                         bp->outer_tpid_bd =
2676                                  TX_BD_LONG_CFA_META_VLAN_TPID_TPID9300;
2677                                 break;
2678                 default:
2679                         PMD_DRV_LOG(ERR, "Invalid TPID: %x\n", tpid);
2680                         return -EINVAL;
2681                 }
2682                 bp->outer_tpid_bd |= tpid;
2683                 PMD_DRV_LOG(INFO, "outer_tpid_bd = %x\n", bp->outer_tpid_bd);
2684         } else if (vlan_type == ETH_VLAN_TYPE_INNER) {
2685                 PMD_DRV_LOG(ERR,
2686                             "Can accelerate only outer vlan in QinQ\n");
2687                 return -EINVAL;
2688         }
2689
2690         return 0;
2691 }
2692
2693 static int
2694 bnxt_set_default_mac_addr_op(struct rte_eth_dev *dev,
2695                              struct rte_ether_addr *addr)
2696 {
2697         struct bnxt *bp = dev->data->dev_private;
2698         /* Default Filter is tied to VNIC 0 */
2699         struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
2700         int rc;
2701
2702         rc = is_bnxt_in_error(bp);
2703         if (rc)
2704                 return rc;
2705
2706         if (BNXT_VF(bp) && !BNXT_VF_IS_TRUSTED(bp))
2707                 return -EPERM;
2708
2709         if (rte_is_zero_ether_addr(addr))
2710                 return -EINVAL;
2711
2712         /* Filter settings will get applied when port is started */
2713         if (!dev->data->dev_started)
2714                 return 0;
2715
2716         /* Check if the requested MAC is already added */
2717         if (memcmp(addr, bp->mac_addr, RTE_ETHER_ADDR_LEN) == 0)
2718                 return 0;
2719
2720         /* Destroy filter and re-create it */
2721         bnxt_del_dflt_mac_filter(bp, vnic);
2722
2723         memcpy(bp->mac_addr, addr, RTE_ETHER_ADDR_LEN);
2724         if (dev->data->dev_conf.rxmode.offloads & DEV_RX_OFFLOAD_VLAN_FILTER) {
2725                 /* This filter will allow only untagged packets */
2726                 rc = bnxt_add_vlan_filter(bp, 0);
2727         } else {
2728                 rc = bnxt_add_mac_filter(bp, vnic, addr, 0, 0);
2729         }
2730
2731         PMD_DRV_LOG(DEBUG, "Set MAC addr\n");
2732         return rc;
2733 }
2734
2735 static int
2736 bnxt_dev_set_mc_addr_list_op(struct rte_eth_dev *eth_dev,
2737                           struct rte_ether_addr *mc_addr_set,
2738                           uint32_t nb_mc_addr)
2739 {
2740         struct bnxt *bp = eth_dev->data->dev_private;
2741         char *mc_addr_list = (char *)mc_addr_set;
2742         struct bnxt_vnic_info *vnic;
2743         uint32_t off = 0, i = 0;
2744         int rc;
2745
2746         rc = is_bnxt_in_error(bp);
2747         if (rc)
2748                 return rc;
2749
2750         vnic = BNXT_GET_DEFAULT_VNIC(bp);
2751
2752         if (nb_mc_addr > BNXT_MAX_MC_ADDRS) {
2753                 vnic->flags |= BNXT_VNIC_INFO_ALLMULTI;
2754                 goto allmulti;
2755         }
2756
2757         /* TODO Check for Duplicate mcast addresses */
2758         vnic->flags &= ~BNXT_VNIC_INFO_ALLMULTI;
2759         for (i = 0; i < nb_mc_addr; i++) {
2760                 memcpy(vnic->mc_list + off, &mc_addr_list[i],
2761                         RTE_ETHER_ADDR_LEN);
2762                 off += RTE_ETHER_ADDR_LEN;
2763         }
2764
2765         vnic->mc_addr_cnt = i;
2766         if (vnic->mc_addr_cnt)
2767                 vnic->flags |= BNXT_VNIC_INFO_MCAST;
2768         else
2769                 vnic->flags &= ~BNXT_VNIC_INFO_MCAST;
2770
2771 allmulti:
2772         return bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
2773 }
2774
2775 static int
2776 bnxt_fw_version_get(struct rte_eth_dev *dev, char *fw_version, size_t fw_size)
2777 {
2778         struct bnxt *bp = dev->data->dev_private;
2779         uint8_t fw_major = (bp->fw_ver >> 24) & 0xff;
2780         uint8_t fw_minor = (bp->fw_ver >> 16) & 0xff;
2781         uint8_t fw_updt = (bp->fw_ver >> 8) & 0xff;
2782         uint8_t fw_rsvd = bp->fw_ver & 0xff;
2783         int ret;
2784
2785         ret = snprintf(fw_version, fw_size, "%d.%d.%d.%d",
2786                         fw_major, fw_minor, fw_updt, fw_rsvd);
2787
2788         ret += 1; /* add the size of '\0' */
2789         if (fw_size < (uint32_t)ret)
2790                 return ret;
2791         else
2792                 return 0;
2793 }
2794
2795 static void
2796 bnxt_rxq_info_get_op(struct rte_eth_dev *dev, uint16_t queue_id,
2797         struct rte_eth_rxq_info *qinfo)
2798 {
2799         struct bnxt *bp = dev->data->dev_private;
2800         struct bnxt_rx_queue *rxq;
2801
2802         if (is_bnxt_in_error(bp))
2803                 return;
2804
2805         rxq = dev->data->rx_queues[queue_id];
2806
2807         qinfo->mp = rxq->mb_pool;
2808         qinfo->scattered_rx = dev->data->scattered_rx;
2809         qinfo->nb_desc = rxq->nb_rx_desc;
2810
2811         qinfo->conf.rx_free_thresh = rxq->rx_free_thresh;
2812         qinfo->conf.rx_drop_en = rxq->drop_en;
2813         qinfo->conf.rx_deferred_start = rxq->rx_deferred_start;
2814         qinfo->conf.offloads = dev->data->dev_conf.rxmode.offloads;
2815 }
2816
2817 static void
2818 bnxt_txq_info_get_op(struct rte_eth_dev *dev, uint16_t queue_id,
2819         struct rte_eth_txq_info *qinfo)
2820 {
2821         struct bnxt *bp = dev->data->dev_private;
2822         struct bnxt_tx_queue *txq;
2823
2824         if (is_bnxt_in_error(bp))
2825                 return;
2826
2827         txq = dev->data->tx_queues[queue_id];
2828
2829         qinfo->nb_desc = txq->nb_tx_desc;
2830
2831         qinfo->conf.tx_thresh.pthresh = txq->pthresh;
2832         qinfo->conf.tx_thresh.hthresh = txq->hthresh;
2833         qinfo->conf.tx_thresh.wthresh = txq->wthresh;
2834
2835         qinfo->conf.tx_free_thresh = txq->tx_free_thresh;
2836         qinfo->conf.tx_rs_thresh = 0;
2837         qinfo->conf.tx_deferred_start = txq->tx_deferred_start;
2838         qinfo->conf.offloads = txq->offloads;
2839 }
2840
2841 static const struct {
2842         eth_rx_burst_t pkt_burst;
2843         const char *info;
2844 } bnxt_rx_burst_info[] = {
2845         {bnxt_recv_pkts,        "Scalar"},
2846 #if defined(RTE_ARCH_X86)
2847         {bnxt_recv_pkts_vec,    "Vector SSE"},
2848 #elif defined(RTE_ARCH_ARM64)
2849         {bnxt_recv_pkts_vec,    "Vector Neon"},
2850 #endif
2851 };
2852
2853 static int
2854 bnxt_rx_burst_mode_get(struct rte_eth_dev *dev, __rte_unused uint16_t queue_id,
2855                        struct rte_eth_burst_mode *mode)
2856 {
2857         eth_rx_burst_t pkt_burst = dev->rx_pkt_burst;
2858         size_t i;
2859
2860         for (i = 0; i < RTE_DIM(bnxt_rx_burst_info); i++) {
2861                 if (pkt_burst == bnxt_rx_burst_info[i].pkt_burst) {
2862                         snprintf(mode->info, sizeof(mode->info), "%s",
2863                                  bnxt_rx_burst_info[i].info);
2864                         return 0;
2865                 }
2866         }
2867
2868         return -EINVAL;
2869 }
2870
2871 static const struct {
2872         eth_tx_burst_t pkt_burst;
2873         const char *info;
2874 } bnxt_tx_burst_info[] = {
2875         {bnxt_xmit_pkts,        "Scalar"},
2876 #if defined(RTE_ARCH_X86)
2877         {bnxt_xmit_pkts_vec,    "Vector SSE"},
2878 #elif defined(RTE_ARCH_ARM64)
2879         {bnxt_xmit_pkts_vec,    "Vector Neon"},
2880 #endif
2881 };
2882
2883 static int
2884 bnxt_tx_burst_mode_get(struct rte_eth_dev *dev, __rte_unused uint16_t queue_id,
2885                        struct rte_eth_burst_mode *mode)
2886 {
2887         eth_tx_burst_t pkt_burst = dev->tx_pkt_burst;
2888         size_t i;
2889
2890         for (i = 0; i < RTE_DIM(bnxt_tx_burst_info); i++) {
2891                 if (pkt_burst == bnxt_tx_burst_info[i].pkt_burst) {
2892                         snprintf(mode->info, sizeof(mode->info), "%s",
2893                                  bnxt_tx_burst_info[i].info);
2894                         return 0;
2895                 }
2896         }
2897
2898         return -EINVAL;
2899 }
2900
2901 int bnxt_mtu_set_op(struct rte_eth_dev *eth_dev, uint16_t new_mtu)
2902 {
2903         struct bnxt *bp = eth_dev->data->dev_private;
2904         uint32_t new_pkt_size;
2905         uint32_t rc = 0;
2906         uint32_t i;
2907
2908         rc = is_bnxt_in_error(bp);
2909         if (rc)
2910                 return rc;
2911
2912         /* Exit if receive queues are not configured yet */
2913         if (!eth_dev->data->nb_rx_queues)
2914                 return rc;
2915
2916         new_pkt_size = new_mtu + RTE_ETHER_HDR_LEN + RTE_ETHER_CRC_LEN +
2917                        VLAN_TAG_SIZE * BNXT_NUM_VLANS;
2918
2919         /*
2920          * Disallow any MTU change that would require scattered receive support
2921          * if it is not already enabled.
2922          */
2923         if (eth_dev->data->dev_started &&
2924             !eth_dev->data->scattered_rx &&
2925             (new_pkt_size >
2926              eth_dev->data->min_rx_buf_size - RTE_PKTMBUF_HEADROOM)) {
2927                 PMD_DRV_LOG(ERR,
2928                             "MTU change would require scattered rx support. ");
2929                 PMD_DRV_LOG(ERR, "Stop port before changing MTU.\n");
2930                 return -EINVAL;
2931         }
2932
2933         if (new_mtu > RTE_ETHER_MTU) {
2934                 bp->flags |= BNXT_FLAG_JUMBO;
2935                 bp->eth_dev->data->dev_conf.rxmode.offloads |=
2936                         DEV_RX_OFFLOAD_JUMBO_FRAME;
2937         } else {
2938                 bp->eth_dev->data->dev_conf.rxmode.offloads &=
2939                         ~DEV_RX_OFFLOAD_JUMBO_FRAME;
2940                 bp->flags &= ~BNXT_FLAG_JUMBO;
2941         }
2942
2943         /* Is there a change in mtu setting? */
2944         if (eth_dev->data->dev_conf.rxmode.max_rx_pkt_len == new_pkt_size)
2945                 return rc;
2946
2947         for (i = 0; i < bp->nr_vnics; i++) {
2948                 struct bnxt_vnic_info *vnic = &bp->vnic_info[i];
2949                 uint16_t size = 0;
2950
2951                 vnic->mru = BNXT_VNIC_MRU(new_mtu);
2952                 rc = bnxt_hwrm_vnic_cfg(bp, vnic);
2953                 if (rc)
2954                         break;
2955
2956                 size = rte_pktmbuf_data_room_size(bp->rx_queues[0]->mb_pool);
2957                 size -= RTE_PKTMBUF_HEADROOM;
2958
2959                 if (size < new_mtu) {
2960                         rc = bnxt_hwrm_vnic_plcmode_cfg(bp, vnic);
2961                         if (rc)
2962                                 return rc;
2963                 }
2964         }
2965
2966         if (!rc)
2967                 eth_dev->data->dev_conf.rxmode.max_rx_pkt_len = new_pkt_size;
2968
2969         PMD_DRV_LOG(INFO, "New MTU is %d\n", new_mtu);
2970
2971         return rc;
2972 }
2973
2974 static int
2975 bnxt_vlan_pvid_set_op(struct rte_eth_dev *dev, uint16_t pvid, int on)
2976 {
2977         struct bnxt *bp = dev->data->dev_private;
2978         uint16_t vlan = bp->vlan;
2979         int rc;
2980
2981         rc = is_bnxt_in_error(bp);
2982         if (rc)
2983                 return rc;
2984
2985         if (!BNXT_SINGLE_PF(bp) || BNXT_VF(bp)) {
2986                 PMD_DRV_LOG(ERR,
2987                         "PVID cannot be modified for this function\n");
2988                 return -ENOTSUP;
2989         }
2990         bp->vlan = on ? pvid : 0;
2991
2992         rc = bnxt_hwrm_set_default_vlan(bp, 0, 0);
2993         if (rc)
2994                 bp->vlan = vlan;
2995         return rc;
2996 }
2997
2998 static int
2999 bnxt_dev_led_on_op(struct rte_eth_dev *dev)
3000 {
3001         struct bnxt *bp = dev->data->dev_private;
3002         int rc;
3003
3004         rc = is_bnxt_in_error(bp);
3005         if (rc)
3006                 return rc;
3007
3008         return bnxt_hwrm_port_led_cfg(bp, true);
3009 }
3010
3011 static int
3012 bnxt_dev_led_off_op(struct rte_eth_dev *dev)
3013 {
3014         struct bnxt *bp = dev->data->dev_private;
3015         int rc;
3016
3017         rc = is_bnxt_in_error(bp);
3018         if (rc)
3019                 return rc;
3020
3021         return bnxt_hwrm_port_led_cfg(bp, false);
3022 }
3023
3024 static uint32_t
3025 bnxt_rx_queue_count_op(struct rte_eth_dev *dev, uint16_t rx_queue_id)
3026 {
3027         struct bnxt *bp = (struct bnxt *)dev->data->dev_private;
3028         struct bnxt_cp_ring_info *cpr;
3029         uint32_t desc = 0, raw_cons;
3030         struct bnxt_rx_queue *rxq;
3031         struct rx_pkt_cmpl *rxcmp;
3032         int rc;
3033
3034         rc = is_bnxt_in_error(bp);
3035         if (rc)
3036                 return rc;
3037
3038         rxq = dev->data->rx_queues[rx_queue_id];
3039         cpr = rxq->cp_ring;
3040         raw_cons = cpr->cp_raw_cons;
3041
3042         while (1) {
3043                 uint32_t agg_cnt, cons, cmpl_type;
3044
3045                 cons = RING_CMP(cpr->cp_ring_struct, raw_cons);
3046                 rxcmp = (struct rx_pkt_cmpl *)&cpr->cp_desc_ring[cons];
3047
3048                 if (!CMP_VALID(rxcmp, raw_cons, cpr->cp_ring_struct))
3049                         break;
3050
3051                 cmpl_type = CMP_TYPE(rxcmp);
3052
3053                 switch (cmpl_type) {
3054                 case CMPL_BASE_TYPE_RX_L2:
3055                 case CMPL_BASE_TYPE_RX_L2_V2:
3056                         agg_cnt = BNXT_RX_L2_AGG_BUFS(rxcmp);
3057                         raw_cons = raw_cons + CMP_LEN(cmpl_type) + agg_cnt;
3058                         desc++;
3059                         break;
3060
3061                 case CMPL_BASE_TYPE_RX_TPA_END:
3062                         if (BNXT_CHIP_P5(rxq->bp)) {
3063                                 struct rx_tpa_v2_end_cmpl_hi *p5_tpa_end;
3064
3065                                 p5_tpa_end = (void *)rxcmp;
3066                                 agg_cnt = BNXT_TPA_END_AGG_BUFS_TH(p5_tpa_end);
3067                         } else {
3068                                 struct rx_tpa_end_cmpl *tpa_end;
3069
3070                                 tpa_end = (void *)rxcmp;
3071                                 agg_cnt = BNXT_TPA_END_AGG_BUFS(tpa_end);
3072                         }
3073
3074                         raw_cons = raw_cons + CMP_LEN(cmpl_type) + agg_cnt;
3075                         desc++;
3076                         break;
3077
3078                 default:
3079                         raw_cons += CMP_LEN(cmpl_type);
3080                 }
3081         }
3082
3083         return desc;
3084 }
3085
3086 static int
3087 bnxt_rx_descriptor_status_op(void *rx_queue, uint16_t offset)
3088 {
3089         struct bnxt_rx_queue *rxq = rx_queue;
3090         struct bnxt_cp_ring_info *cpr;
3091         struct bnxt_rx_ring_info *rxr;
3092         uint32_t desc, raw_cons;
3093         struct bnxt *bp = rxq->bp;
3094         struct rx_pkt_cmpl *rxcmp;
3095         int rc;
3096
3097         rc = is_bnxt_in_error(bp);
3098         if (rc)
3099                 return rc;
3100
3101         if (offset >= rxq->nb_rx_desc)
3102                 return -EINVAL;
3103
3104         rxr = rxq->rx_ring;
3105         cpr = rxq->cp_ring;
3106
3107         /*
3108          * For the vector receive case, the completion at the requested
3109          * offset can be indexed directly.
3110          */
3111 #if defined(RTE_ARCH_X86) || defined(RTE_ARCH_ARM64)
3112         if (bp->flags & BNXT_FLAG_RX_VECTOR_PKT_MODE) {
3113                 struct rx_pkt_cmpl *rxcmp;
3114                 uint32_t cons;
3115
3116                 /* Check status of completion descriptor. */
3117                 raw_cons = cpr->cp_raw_cons +
3118                            offset * CMP_LEN(CMPL_BASE_TYPE_RX_L2);
3119                 cons = RING_CMP(cpr->cp_ring_struct, raw_cons);
3120                 rxcmp = (struct rx_pkt_cmpl *)&cpr->cp_desc_ring[cons];
3121
3122                 if (CMP_VALID(rxcmp, raw_cons, cpr->cp_ring_struct))
3123                         return RTE_ETH_RX_DESC_DONE;
3124
3125                 /* Check whether rx desc has an mbuf attached. */
3126                 cons = RING_CMP(rxr->rx_ring_struct, raw_cons / 2);
3127                 if (cons >= rxq->rxrearm_start &&
3128                     cons < rxq->rxrearm_start + rxq->rxrearm_nb) {
3129                         return RTE_ETH_RX_DESC_UNAVAIL;
3130                 }
3131
3132                 return RTE_ETH_RX_DESC_AVAIL;
3133         }
3134 #endif
3135
3136         /*
3137          * For the non-vector receive case, scan the completion ring to
3138          * locate the completion descriptor for the requested offset.
3139          */
3140         raw_cons = cpr->cp_raw_cons;
3141         desc = 0;
3142         while (1) {
3143                 uint32_t agg_cnt, cons, cmpl_type;
3144
3145                 cons = RING_CMP(cpr->cp_ring_struct, raw_cons);
3146                 rxcmp = (struct rx_pkt_cmpl *)&cpr->cp_desc_ring[cons];
3147
3148                 if (!CMP_VALID(rxcmp, raw_cons, cpr->cp_ring_struct))
3149                         break;
3150
3151                 cmpl_type = CMP_TYPE(rxcmp);
3152
3153                 switch (cmpl_type) {
3154                 case CMPL_BASE_TYPE_RX_L2:
3155                 case CMPL_BASE_TYPE_RX_L2_V2:
3156                         if (desc == offset) {
3157                                 cons = rxcmp->opaque;
3158                                 if (rxr->rx_buf_ring[cons])
3159                                         return RTE_ETH_RX_DESC_DONE;
3160                                 else
3161                                         return RTE_ETH_RX_DESC_UNAVAIL;
3162                         }
3163                         agg_cnt = BNXT_RX_L2_AGG_BUFS(rxcmp);
3164                         raw_cons = raw_cons + CMP_LEN(cmpl_type) + agg_cnt;
3165                         desc++;
3166                         break;
3167
3168                 case CMPL_BASE_TYPE_RX_TPA_END:
3169                         if (desc == offset)
3170                                 return RTE_ETH_RX_DESC_DONE;
3171
3172                         if (BNXT_CHIP_P5(rxq->bp)) {
3173                                 struct rx_tpa_v2_end_cmpl_hi *p5_tpa_end;
3174
3175                                 p5_tpa_end = (void *)rxcmp;
3176                                 agg_cnt = BNXT_TPA_END_AGG_BUFS_TH(p5_tpa_end);
3177                         } else {
3178                                 struct rx_tpa_end_cmpl *tpa_end;
3179
3180                                 tpa_end = (void *)rxcmp;
3181                                 agg_cnt = BNXT_TPA_END_AGG_BUFS(tpa_end);
3182                         }
3183
3184                         raw_cons = raw_cons + CMP_LEN(cmpl_type) + agg_cnt;
3185                         desc++;
3186                         break;
3187
3188                 default:
3189                         raw_cons += CMP_LEN(cmpl_type);
3190                 }
3191         }
3192
3193         return RTE_ETH_RX_DESC_AVAIL;
3194 }
3195
3196 static int
3197 bnxt_tx_descriptor_status_op(void *tx_queue, uint16_t offset)
3198 {
3199         struct bnxt_tx_queue *txq = (struct bnxt_tx_queue *)tx_queue;
3200         struct bnxt_tx_ring_info *txr;
3201         struct bnxt_cp_ring_info *cpr;
3202         struct rte_mbuf **tx_buf;
3203         struct tx_pkt_cmpl *txcmp;
3204         uint32_t cons, cp_cons;
3205         int rc;
3206
3207         if (!txq)
3208                 return -EINVAL;
3209
3210         rc = is_bnxt_in_error(txq->bp);
3211         if (rc)
3212                 return rc;
3213
3214         cpr = txq->cp_ring;
3215         txr = txq->tx_ring;
3216
3217         if (offset >= txq->nb_tx_desc)
3218                 return -EINVAL;
3219
3220         cons = RING_CMP(cpr->cp_ring_struct, offset);
3221         txcmp = (struct tx_pkt_cmpl *)&cpr->cp_desc_ring[cons];
3222         cp_cons = cpr->cp_raw_cons;
3223
3224         if (cons > cp_cons) {
3225                 if (CMPL_VALID(txcmp, cpr->valid))
3226                         return RTE_ETH_TX_DESC_UNAVAIL;
3227         } else {
3228                 if (CMPL_VALID(txcmp, !cpr->valid))
3229                         return RTE_ETH_TX_DESC_UNAVAIL;
3230         }
3231         tx_buf = &txr->tx_buf_ring[cons];
3232         if (*tx_buf == NULL)
3233                 return RTE_ETH_TX_DESC_DONE;
3234
3235         return RTE_ETH_TX_DESC_FULL;
3236 }
3237
3238 int
3239 bnxt_filter_ctrl_op(struct rte_eth_dev *dev,
3240                     enum rte_filter_type filter_type,
3241                     enum rte_filter_op filter_op, void *arg)
3242 {
3243         struct bnxt *bp = dev->data->dev_private;
3244         int ret = 0;
3245
3246         if (!bp)
3247                 return -EIO;
3248
3249         if (BNXT_ETH_DEV_IS_REPRESENTOR(dev)) {
3250                 struct bnxt_representor *vfr = dev->data->dev_private;
3251                 bp = vfr->parent_dev->data->dev_private;
3252                 /* parent is deleted while children are still valid */
3253                 if (!bp) {
3254                         PMD_DRV_LOG(DEBUG, "BNXT Port:%d VFR Error %d:%d\n",
3255                                     dev->data->port_id,
3256                                     filter_type,
3257                                     filter_op);
3258                         return -EIO;
3259                 }
3260         }
3261
3262         ret = is_bnxt_in_error(bp);
3263         if (ret)
3264                 return ret;
3265
3266         switch (filter_type) {
3267         case RTE_ETH_FILTER_GENERIC:
3268                 if (filter_op != RTE_ETH_FILTER_GET)
3269                         return -EINVAL;
3270
3271                 /* PMD supports thread-safe flow operations.  rte_flow API
3272                  * functions can avoid mutex for multi-thread safety.
3273                  */
3274                 dev->data->dev_flags |= RTE_ETH_DEV_FLOW_OPS_THREAD_SAFE;
3275
3276                 if (BNXT_TRUFLOW_EN(bp))
3277                         *(const void **)arg = &bnxt_ulp_rte_flow_ops;
3278                 else
3279                         *(const void **)arg = &bnxt_flow_ops;
3280                 break;
3281         default:
3282                 PMD_DRV_LOG(ERR,
3283                         "Filter type (%d) not supported", filter_type);
3284                 ret = -EINVAL;
3285                 break;
3286         }
3287         return ret;
3288 }
3289
3290 static const uint32_t *
3291 bnxt_dev_supported_ptypes_get_op(struct rte_eth_dev *dev)
3292 {
3293         static const uint32_t ptypes[] = {
3294                 RTE_PTYPE_L2_ETHER_VLAN,
3295                 RTE_PTYPE_L3_IPV4_EXT_UNKNOWN,
3296                 RTE_PTYPE_L3_IPV6_EXT_UNKNOWN,
3297                 RTE_PTYPE_L4_ICMP,
3298                 RTE_PTYPE_L4_TCP,
3299                 RTE_PTYPE_L4_UDP,
3300                 RTE_PTYPE_INNER_L3_IPV4_EXT_UNKNOWN,
3301                 RTE_PTYPE_INNER_L3_IPV6_EXT_UNKNOWN,
3302                 RTE_PTYPE_INNER_L4_ICMP,
3303                 RTE_PTYPE_INNER_L4_TCP,
3304                 RTE_PTYPE_INNER_L4_UDP,
3305                 RTE_PTYPE_UNKNOWN
3306         };
3307
3308         if (!dev->rx_pkt_burst)
3309                 return NULL;
3310
3311         return ptypes;
3312 }
3313
3314 static int bnxt_map_regs(struct bnxt *bp, uint32_t *reg_arr, int count,
3315                          int reg_win)
3316 {
3317         uint32_t reg_base = *reg_arr & 0xfffff000;
3318         uint32_t win_off;
3319         int i;
3320
3321         for (i = 0; i < count; i++) {
3322                 if ((reg_arr[i] & 0xfffff000) != reg_base)
3323                         return -ERANGE;
3324         }
3325         win_off = BNXT_GRCPF_REG_WINDOW_BASE_OUT + (reg_win - 1) * 4;
3326         rte_write32(reg_base, (uint8_t *)bp->bar0 + win_off);
3327         return 0;
3328 }
3329
3330 static int bnxt_map_ptp_regs(struct bnxt *bp)
3331 {
3332         struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3333         uint32_t *reg_arr;
3334         int rc, i;
3335
3336         reg_arr = ptp->rx_regs;
3337         rc = bnxt_map_regs(bp, reg_arr, BNXT_PTP_RX_REGS, 5);
3338         if (rc)
3339                 return rc;
3340
3341         reg_arr = ptp->tx_regs;
3342         rc = bnxt_map_regs(bp, reg_arr, BNXT_PTP_TX_REGS, 6);
3343         if (rc)
3344                 return rc;
3345
3346         for (i = 0; i < BNXT_PTP_RX_REGS; i++)
3347                 ptp->rx_mapped_regs[i] = 0x5000 + (ptp->rx_regs[i] & 0xfff);
3348
3349         for (i = 0; i < BNXT_PTP_TX_REGS; i++)
3350                 ptp->tx_mapped_regs[i] = 0x6000 + (ptp->tx_regs[i] & 0xfff);
3351
3352         return 0;
3353 }
3354
3355 static void bnxt_unmap_ptp_regs(struct bnxt *bp)
3356 {
3357         rte_write32(0, (uint8_t *)bp->bar0 +
3358                          BNXT_GRCPF_REG_WINDOW_BASE_OUT + 16);
3359         rte_write32(0, (uint8_t *)bp->bar0 +
3360                          BNXT_GRCPF_REG_WINDOW_BASE_OUT + 20);
3361 }
3362
3363 static uint64_t bnxt_cc_read(struct bnxt *bp)
3364 {
3365         uint64_t ns;
3366
3367         ns = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3368                               BNXT_GRCPF_REG_SYNC_TIME));
3369         ns |= (uint64_t)(rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3370                                           BNXT_GRCPF_REG_SYNC_TIME + 4))) << 32;
3371         return ns;
3372 }
3373
3374 static int bnxt_get_tx_ts(struct bnxt *bp, uint64_t *ts)
3375 {
3376         struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3377         uint32_t fifo;
3378
3379         fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3380                                 ptp->tx_mapped_regs[BNXT_PTP_TX_FIFO]));
3381         if (fifo & BNXT_PTP_TX_FIFO_EMPTY)
3382                 return -EAGAIN;
3383
3384         fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3385                                 ptp->tx_mapped_regs[BNXT_PTP_TX_FIFO]));
3386         *ts = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3387                                 ptp->tx_mapped_regs[BNXT_PTP_TX_TS_L]));
3388         *ts |= (uint64_t)rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3389                                 ptp->tx_mapped_regs[BNXT_PTP_TX_TS_H])) << 32;
3390         rte_read32((uint8_t *)bp->bar0 + ptp->tx_mapped_regs[BNXT_PTP_TX_SEQ]);
3391
3392         return 0;
3393 }
3394
3395 static int bnxt_get_rx_ts(struct bnxt *bp, uint64_t *ts)
3396 {
3397         struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3398         struct bnxt_pf_info *pf = bp->pf;
3399         uint16_t port_id;
3400         uint32_t fifo;
3401
3402         if (!ptp)
3403                 return -ENODEV;
3404
3405         fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3406                                 ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO]));
3407         if (!(fifo & BNXT_PTP_RX_FIFO_PENDING))
3408                 return -EAGAIN;
3409
3410         port_id = pf->port_id;
3411         rte_write32(1 << port_id, (uint8_t *)bp->bar0 +
3412                ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO_ADV]);
3413
3414         fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3415                                    ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO]));
3416         if (fifo & BNXT_PTP_RX_FIFO_PENDING) {
3417 /*              bnxt_clr_rx_ts(bp);       TBD  */
3418                 return -EBUSY;
3419         }
3420
3421         *ts = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3422                                 ptp->rx_mapped_regs[BNXT_PTP_RX_TS_L]));
3423         *ts |= (uint64_t)rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3424                                 ptp->rx_mapped_regs[BNXT_PTP_RX_TS_H])) << 32;
3425
3426         return 0;
3427 }
3428
3429 static int
3430 bnxt_timesync_write_time(struct rte_eth_dev *dev, const struct timespec *ts)
3431 {
3432         uint64_t ns;
3433         struct bnxt *bp = dev->data->dev_private;
3434         struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3435
3436         if (!ptp)
3437                 return 0;
3438
3439         ns = rte_timespec_to_ns(ts);
3440         /* Set the timecounters to a new value. */
3441         ptp->tc.nsec = ns;
3442         ptp->tx_tstamp_tc.nsec = ns;
3443         ptp->rx_tstamp_tc.nsec = ns;
3444
3445         return 0;
3446 }
3447
3448 static int
3449 bnxt_timesync_read_time(struct rte_eth_dev *dev, struct timespec *ts)
3450 {
3451         struct bnxt *bp = dev->data->dev_private;
3452         struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3453         uint64_t ns, systime_cycles = 0;
3454         int rc = 0;
3455
3456         if (!ptp)
3457                 return 0;
3458
3459         if (BNXT_CHIP_P5(bp))
3460                 rc = bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_CURRENT_TIME,
3461                                              &systime_cycles);
3462         else
3463                 systime_cycles = bnxt_cc_read(bp);
3464
3465         ns = rte_timecounter_update(&ptp->tc, systime_cycles);
3466         *ts = rte_ns_to_timespec(ns);
3467
3468         return rc;
3469 }
3470 static int
3471 bnxt_timesync_enable(struct rte_eth_dev *dev)
3472 {
3473         struct bnxt *bp = dev->data->dev_private;
3474         struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3475         uint32_t shift = 0;
3476         int rc;
3477
3478         if (!ptp)
3479                 return 0;
3480
3481         ptp->rx_filter = 1;
3482         ptp->tx_tstamp_en = 1;
3483         ptp->rxctl = BNXT_PTP_MSG_EVENTS;
3484
3485         rc = bnxt_hwrm_ptp_cfg(bp);
3486         if (rc)
3487                 return rc;
3488
3489         memset(&ptp->tc, 0, sizeof(struct rte_timecounter));
3490         memset(&ptp->rx_tstamp_tc, 0, sizeof(struct rte_timecounter));
3491         memset(&ptp->tx_tstamp_tc, 0, sizeof(struct rte_timecounter));
3492
3493         ptp->tc.cc_mask = BNXT_CYCLECOUNTER_MASK;
3494         ptp->tc.cc_shift = shift;
3495         ptp->tc.nsec_mask = (1ULL << shift) - 1;
3496
3497         ptp->rx_tstamp_tc.cc_mask = BNXT_CYCLECOUNTER_MASK;
3498         ptp->rx_tstamp_tc.cc_shift = shift;
3499         ptp->rx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
3500
3501         ptp->tx_tstamp_tc.cc_mask = BNXT_CYCLECOUNTER_MASK;
3502         ptp->tx_tstamp_tc.cc_shift = shift;
3503         ptp->tx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
3504
3505         if (!BNXT_CHIP_P5(bp))
3506                 bnxt_map_ptp_regs(bp);
3507         else
3508                 rc = bnxt_ptp_start(bp);
3509
3510         return rc;
3511 }
3512
3513 static int
3514 bnxt_timesync_disable(struct rte_eth_dev *dev)
3515 {
3516         struct bnxt *bp = dev->data->dev_private;
3517         struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3518
3519         if (!ptp)
3520                 return 0;
3521
3522         ptp->rx_filter = 0;
3523         ptp->tx_tstamp_en = 0;
3524         ptp->rxctl = 0;
3525
3526         bnxt_hwrm_ptp_cfg(bp);
3527
3528         if (!BNXT_CHIP_P5(bp))
3529                 bnxt_unmap_ptp_regs(bp);
3530         else
3531                 bnxt_ptp_stop(bp);
3532
3533         return 0;
3534 }
3535
3536 static int
3537 bnxt_timesync_read_rx_timestamp(struct rte_eth_dev *dev,
3538                                  struct timespec *timestamp,
3539                                  uint32_t flags __rte_unused)
3540 {
3541         struct bnxt *bp = dev->data->dev_private;
3542         struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3543         uint64_t rx_tstamp_cycles = 0;
3544         uint64_t ns;
3545
3546         if (!ptp)
3547                 return 0;
3548
3549         if (BNXT_CHIP_P5(bp))
3550                 rx_tstamp_cycles = ptp->rx_timestamp;
3551         else
3552                 bnxt_get_rx_ts(bp, &rx_tstamp_cycles);
3553
3554         ns = rte_timecounter_update(&ptp->rx_tstamp_tc, rx_tstamp_cycles);
3555         *timestamp = rte_ns_to_timespec(ns);
3556         return  0;
3557 }
3558
3559 static int
3560 bnxt_timesync_read_tx_timestamp(struct rte_eth_dev *dev,
3561                                  struct timespec *timestamp)
3562 {
3563         struct bnxt *bp = dev->data->dev_private;
3564         struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3565         uint64_t tx_tstamp_cycles = 0;
3566         uint64_t ns;
3567         int rc = 0;
3568
3569         if (!ptp)
3570                 return 0;
3571
3572         if (BNXT_CHIP_P5(bp))
3573                 rc = bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_PATH_TX,
3574                                              &tx_tstamp_cycles);
3575         else
3576                 rc = bnxt_get_tx_ts(bp, &tx_tstamp_cycles);
3577
3578         ns = rte_timecounter_update(&ptp->tx_tstamp_tc, tx_tstamp_cycles);
3579         *timestamp = rte_ns_to_timespec(ns);
3580
3581         return rc;
3582 }
3583
3584 static int
3585 bnxt_timesync_adjust_time(struct rte_eth_dev *dev, int64_t delta)
3586 {
3587         struct bnxt *bp = dev->data->dev_private;
3588         struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3589
3590         if (!ptp)
3591                 return 0;
3592
3593         ptp->tc.nsec += delta;
3594         ptp->tx_tstamp_tc.nsec += delta;
3595         ptp->rx_tstamp_tc.nsec += delta;
3596
3597         return 0;
3598 }
3599
3600 static int
3601 bnxt_get_eeprom_length_op(struct rte_eth_dev *dev)
3602 {
3603         struct bnxt *bp = dev->data->dev_private;
3604         int rc;
3605         uint32_t dir_entries;
3606         uint32_t entry_length;
3607
3608         rc = is_bnxt_in_error(bp);
3609         if (rc)
3610                 return rc;
3611
3612         PMD_DRV_LOG(INFO, PCI_PRI_FMT "\n",
3613                     bp->pdev->addr.domain, bp->pdev->addr.bus,
3614                     bp->pdev->addr.devid, bp->pdev->addr.function);
3615
3616         rc = bnxt_hwrm_nvm_get_dir_info(bp, &dir_entries, &entry_length);
3617         if (rc != 0)
3618                 return rc;
3619
3620         return dir_entries * entry_length;
3621 }
3622
3623 static int
3624 bnxt_get_eeprom_op(struct rte_eth_dev *dev,
3625                 struct rte_dev_eeprom_info *in_eeprom)
3626 {
3627         struct bnxt *bp = dev->data->dev_private;
3628         uint32_t index;
3629         uint32_t offset;
3630         int rc;
3631
3632         rc = is_bnxt_in_error(bp);
3633         if (rc)
3634                 return rc;
3635
3636         PMD_DRV_LOG(INFO, PCI_PRI_FMT " in_eeprom->offset = %d len = %d\n",
3637                     bp->pdev->addr.domain, bp->pdev->addr.bus,
3638                     bp->pdev->addr.devid, bp->pdev->addr.function,
3639                     in_eeprom->offset, in_eeprom->length);
3640
3641         if (in_eeprom->offset == 0) /* special offset value to get directory */
3642                 return bnxt_get_nvram_directory(bp, in_eeprom->length,
3643                                                 in_eeprom->data);
3644
3645         index = in_eeprom->offset >> 24;
3646         offset = in_eeprom->offset & 0xffffff;
3647
3648         if (index != 0)
3649                 return bnxt_hwrm_get_nvram_item(bp, index - 1, offset,
3650                                            in_eeprom->length, in_eeprom->data);
3651
3652         return 0;
3653 }
3654
3655 static bool bnxt_dir_type_is_ape_bin_format(uint16_t dir_type)
3656 {
3657         switch (dir_type) {
3658         case BNX_DIR_TYPE_CHIMP_PATCH:
3659         case BNX_DIR_TYPE_BOOTCODE:
3660         case BNX_DIR_TYPE_BOOTCODE_2:
3661         case BNX_DIR_TYPE_APE_FW:
3662         case BNX_DIR_TYPE_APE_PATCH:
3663         case BNX_DIR_TYPE_KONG_FW:
3664         case BNX_DIR_TYPE_KONG_PATCH:
3665         case BNX_DIR_TYPE_BONO_FW:
3666         case BNX_DIR_TYPE_BONO_PATCH:
3667                 /* FALLTHROUGH */
3668                 return true;
3669         }
3670
3671         return false;
3672 }
3673
3674 static bool bnxt_dir_type_is_other_exec_format(uint16_t dir_type)
3675 {
3676         switch (dir_type) {
3677         case BNX_DIR_TYPE_AVS:
3678         case BNX_DIR_TYPE_EXP_ROM_MBA:
3679         case BNX_DIR_TYPE_PCIE:
3680         case BNX_DIR_TYPE_TSCF_UCODE:
3681         case BNX_DIR_TYPE_EXT_PHY:
3682         case BNX_DIR_TYPE_CCM:
3683         case BNX_DIR_TYPE_ISCSI_BOOT:
3684         case BNX_DIR_TYPE_ISCSI_BOOT_IPV6:
3685         case BNX_DIR_TYPE_ISCSI_BOOT_IPV4N6:
3686                 /* FALLTHROUGH */
3687                 return true;
3688         }
3689
3690         return false;
3691 }
3692
3693 static bool bnxt_dir_type_is_executable(uint16_t dir_type)
3694 {
3695         return bnxt_dir_type_is_ape_bin_format(dir_type) ||
3696                 bnxt_dir_type_is_other_exec_format(dir_type);
3697 }
3698
3699 static int
3700 bnxt_set_eeprom_op(struct rte_eth_dev *dev,
3701                 struct rte_dev_eeprom_info *in_eeprom)
3702 {
3703         struct bnxt *bp = dev->data->dev_private;
3704         uint8_t index, dir_op;
3705         uint16_t type, ext, ordinal, attr;
3706         int rc;
3707
3708         rc = is_bnxt_in_error(bp);
3709         if (rc)
3710                 return rc;
3711
3712         PMD_DRV_LOG(INFO, PCI_PRI_FMT " in_eeprom->offset = %d len = %d\n",
3713                     bp->pdev->addr.domain, bp->pdev->addr.bus,
3714                     bp->pdev->addr.devid, bp->pdev->addr.function,
3715                     in_eeprom->offset, in_eeprom->length);
3716
3717         if (!BNXT_PF(bp)) {
3718                 PMD_DRV_LOG(ERR, "NVM write not supported from a VF\n");
3719                 return -EINVAL;
3720         }
3721
3722         type = in_eeprom->magic >> 16;
3723
3724         if (type == 0xffff) { /* special value for directory operations */
3725                 index = in_eeprom->magic & 0xff;
3726                 dir_op = in_eeprom->magic >> 8;
3727                 if (index == 0)
3728                         return -EINVAL;
3729                 switch (dir_op) {
3730                 case 0x0e: /* erase */
3731                         if (in_eeprom->offset != ~in_eeprom->magic)
3732                                 return -EINVAL;
3733                         return bnxt_hwrm_erase_nvram_directory(bp, index - 1);
3734                 default:
3735                         return -EINVAL;
3736                 }
3737         }
3738
3739         /* Create or re-write an NVM item: */
3740         if (bnxt_dir_type_is_executable(type) == true)
3741                 return -EOPNOTSUPP;
3742         ext = in_eeprom->magic & 0xffff;
3743         ordinal = in_eeprom->offset >> 16;
3744         attr = in_eeprom->offset & 0xffff;
3745
3746         return bnxt_hwrm_flash_nvram(bp, type, ordinal, ext, attr,
3747                                      in_eeprom->data, in_eeprom->length);
3748 }
3749
3750 /*
3751  * Initialization
3752  */
3753
3754 static const struct eth_dev_ops bnxt_dev_ops = {
3755         .dev_infos_get = bnxt_dev_info_get_op,
3756         .dev_close = bnxt_dev_close_op,
3757         .dev_configure = bnxt_dev_configure_op,
3758         .dev_start = bnxt_dev_start_op,
3759         .dev_stop = bnxt_dev_stop_op,
3760         .dev_set_link_up = bnxt_dev_set_link_up_op,
3761         .dev_set_link_down = bnxt_dev_set_link_down_op,
3762         .stats_get = bnxt_stats_get_op,
3763         .stats_reset = bnxt_stats_reset_op,
3764         .rx_queue_setup = bnxt_rx_queue_setup_op,
3765         .rx_queue_release = bnxt_rx_queue_release_op,
3766         .tx_queue_setup = bnxt_tx_queue_setup_op,
3767         .tx_queue_release = bnxt_tx_queue_release_op,
3768         .rx_queue_intr_enable = bnxt_rx_queue_intr_enable_op,
3769         .rx_queue_intr_disable = bnxt_rx_queue_intr_disable_op,
3770         .reta_update = bnxt_reta_update_op,
3771         .reta_query = bnxt_reta_query_op,
3772         .rss_hash_update = bnxt_rss_hash_update_op,
3773         .rss_hash_conf_get = bnxt_rss_hash_conf_get_op,
3774         .link_update = bnxt_link_update_op,
3775         .promiscuous_enable = bnxt_promiscuous_enable_op,
3776         .promiscuous_disable = bnxt_promiscuous_disable_op,
3777         .allmulticast_enable = bnxt_allmulticast_enable_op,
3778         .allmulticast_disable = bnxt_allmulticast_disable_op,
3779         .mac_addr_add = bnxt_mac_addr_add_op,
3780         .mac_addr_remove = bnxt_mac_addr_remove_op,
3781         .flow_ctrl_get = bnxt_flow_ctrl_get_op,
3782         .flow_ctrl_set = bnxt_flow_ctrl_set_op,
3783         .udp_tunnel_port_add  = bnxt_udp_tunnel_port_add_op,
3784         .udp_tunnel_port_del  = bnxt_udp_tunnel_port_del_op,
3785         .vlan_filter_set = bnxt_vlan_filter_set_op,
3786         .vlan_offload_set = bnxt_vlan_offload_set_op,
3787         .vlan_tpid_set = bnxt_vlan_tpid_set_op,
3788         .vlan_pvid_set = bnxt_vlan_pvid_set_op,
3789         .mtu_set = bnxt_mtu_set_op,
3790         .mac_addr_set = bnxt_set_default_mac_addr_op,
3791         .xstats_get = bnxt_dev_xstats_get_op,
3792         .xstats_get_names = bnxt_dev_xstats_get_names_op,
3793         .xstats_reset = bnxt_dev_xstats_reset_op,
3794         .fw_version_get = bnxt_fw_version_get,
3795         .set_mc_addr_list = bnxt_dev_set_mc_addr_list_op,
3796         .rxq_info_get = bnxt_rxq_info_get_op,
3797         .txq_info_get = bnxt_txq_info_get_op,
3798         .rx_burst_mode_get = bnxt_rx_burst_mode_get,
3799         .tx_burst_mode_get = bnxt_tx_burst_mode_get,
3800         .dev_led_on = bnxt_dev_led_on_op,
3801         .dev_led_off = bnxt_dev_led_off_op,
3802         .rx_queue_start = bnxt_rx_queue_start,
3803         .rx_queue_stop = bnxt_rx_queue_stop,
3804         .tx_queue_start = bnxt_tx_queue_start,
3805         .tx_queue_stop = bnxt_tx_queue_stop,
3806         .filter_ctrl = bnxt_filter_ctrl_op,
3807         .dev_supported_ptypes_get = bnxt_dev_supported_ptypes_get_op,
3808         .get_eeprom_length    = bnxt_get_eeprom_length_op,
3809         .get_eeprom           = bnxt_get_eeprom_op,
3810         .set_eeprom           = bnxt_set_eeprom_op,
3811         .timesync_enable      = bnxt_timesync_enable,
3812         .timesync_disable     = bnxt_timesync_disable,
3813         .timesync_read_time   = bnxt_timesync_read_time,
3814         .timesync_write_time   = bnxt_timesync_write_time,
3815         .timesync_adjust_time = bnxt_timesync_adjust_time,
3816         .timesync_read_rx_timestamp = bnxt_timesync_read_rx_timestamp,
3817         .timesync_read_tx_timestamp = bnxt_timesync_read_tx_timestamp,
3818 };
3819
3820 static uint32_t bnxt_map_reset_regs(struct bnxt *bp, uint32_t reg)
3821 {
3822         uint32_t offset;
3823
3824         /* Only pre-map the reset GRC registers using window 3 */
3825         rte_write32(reg & 0xfffff000, (uint8_t *)bp->bar0 +
3826                     BNXT_GRCPF_REG_WINDOW_BASE_OUT + 8);
3827
3828         offset = BNXT_GRCP_WINDOW_3_BASE + (reg & 0xffc);
3829
3830         return offset;
3831 }
3832
3833 int bnxt_map_fw_health_status_regs(struct bnxt *bp)
3834 {
3835         struct bnxt_error_recovery_info *info = bp->recovery_info;
3836         uint32_t reg_base = 0xffffffff;
3837         int i;
3838
3839         /* Only pre-map the monitoring GRC registers using window 2 */
3840         for (i = 0; i < BNXT_FW_STATUS_REG_CNT; i++) {
3841                 uint32_t reg = info->status_regs[i];
3842
3843                 if (BNXT_FW_STATUS_REG_TYPE(reg) != BNXT_FW_STATUS_REG_TYPE_GRC)
3844                         continue;
3845
3846                 if (reg_base == 0xffffffff)
3847                         reg_base = reg & 0xfffff000;
3848                 if ((reg & 0xfffff000) != reg_base)
3849                         return -ERANGE;
3850
3851                 /* Use mask 0xffc as the Lower 2 bits indicates
3852                  * address space location
3853                  */
3854                 info->mapped_status_regs[i] = BNXT_GRCP_WINDOW_2_BASE +
3855                                                 (reg & 0xffc);
3856         }
3857
3858         if (reg_base == 0xffffffff)
3859                 return 0;
3860
3861         rte_write32(reg_base, (uint8_t *)bp->bar0 +
3862                     BNXT_GRCPF_REG_WINDOW_BASE_OUT + 4);
3863
3864         return 0;
3865 }
3866
3867 static void bnxt_write_fw_reset_reg(struct bnxt *bp, uint32_t index)
3868 {
3869         struct bnxt_error_recovery_info *info = bp->recovery_info;
3870         uint32_t delay = info->delay_after_reset[index];
3871         uint32_t val = info->reset_reg_val[index];
3872         uint32_t reg = info->reset_reg[index];
3873         uint32_t type, offset;
3874         int ret;
3875
3876         type = BNXT_FW_STATUS_REG_TYPE(reg);
3877         offset = BNXT_FW_STATUS_REG_OFF(reg);
3878
3879         switch (type) {
3880         case BNXT_FW_STATUS_REG_TYPE_CFG:
3881                 ret = rte_pci_write_config(bp->pdev, &val, sizeof(val), offset);
3882                 if (ret < 0) {
3883                         PMD_DRV_LOG(ERR, "Failed to write %#x at PCI offset %#x",
3884                                     val, offset);
3885                         return;
3886                 }
3887                 break;
3888         case BNXT_FW_STATUS_REG_TYPE_GRC:
3889                 offset = bnxt_map_reset_regs(bp, offset);
3890                 rte_write32(val, (uint8_t *)bp->bar0 + offset);
3891                 break;
3892         case BNXT_FW_STATUS_REG_TYPE_BAR0:
3893                 rte_write32(val, (uint8_t *)bp->bar0 + offset);
3894                 break;
3895         }
3896         /* wait on a specific interval of time until core reset is complete */
3897         if (delay)
3898                 rte_delay_ms(delay);
3899 }
3900
3901 static void bnxt_dev_cleanup(struct bnxt *bp)
3902 {
3903         bp->eth_dev->data->dev_link.link_status = 0;
3904         bp->link_info->link_up = 0;
3905         if (bp->eth_dev->data->dev_started)
3906                 bnxt_dev_stop(bp->eth_dev);
3907
3908         bnxt_uninit_resources(bp, true);
3909 }
3910
3911 static int
3912 bnxt_check_fw_reset_done(struct bnxt *bp)
3913 {
3914         int timeout = bp->fw_reset_max_msecs;
3915         uint16_t val = 0;
3916         int rc;
3917
3918         do {
3919                 rc = rte_pci_read_config(bp->pdev, &val, sizeof(val), PCI_SUBSYSTEM_ID_OFFSET);
3920                 if (rc < 0) {
3921                         PMD_DRV_LOG(ERR, "Failed to read PCI offset 0x%x", PCI_SUBSYSTEM_ID_OFFSET);
3922                         return rc;
3923                 }
3924                 if (val != 0xffff)
3925                         break;
3926                 rte_delay_ms(1);
3927         } while (timeout--);
3928
3929         if (val == 0xffff) {
3930                 PMD_DRV_LOG(ERR, "Firmware reset aborted, PCI config space invalid\n");
3931                 return -1;
3932         }
3933
3934         return 0;
3935 }
3936
3937 static int bnxt_restore_vlan_filters(struct bnxt *bp)
3938 {
3939         struct rte_eth_dev *dev = bp->eth_dev;
3940         struct rte_vlan_filter_conf *vfc;
3941         int vidx, vbit, rc;
3942         uint16_t vlan_id;
3943
3944         for (vlan_id = 1; vlan_id <= RTE_ETHER_MAX_VLAN_ID; vlan_id++) {
3945                 vfc = &dev->data->vlan_filter_conf;
3946                 vidx = vlan_id / 64;
3947                 vbit = vlan_id % 64;
3948
3949                 /* Each bit corresponds to a VLAN id */
3950                 if (vfc->ids[vidx] & (UINT64_C(1) << vbit)) {
3951                         rc = bnxt_add_vlan_filter(bp, vlan_id);
3952                         if (rc)
3953                                 return rc;
3954                 }
3955         }
3956
3957         return 0;
3958 }
3959
3960 static int bnxt_restore_mac_filters(struct bnxt *bp)
3961 {
3962         struct rte_eth_dev *dev = bp->eth_dev;
3963         struct rte_eth_dev_info dev_info;
3964         struct rte_ether_addr *addr;
3965         uint64_t pool_mask;
3966         uint32_t pool = 0;
3967         uint16_t i;
3968         int rc;
3969
3970         if (BNXT_VF(bp) && !BNXT_VF_IS_TRUSTED(bp))
3971                 return 0;
3972
3973         rc = bnxt_dev_info_get_op(dev, &dev_info);
3974         if (rc)
3975                 return rc;
3976
3977         /* replay MAC address configuration */
3978         for (i = 1; i < dev_info.max_mac_addrs; i++) {
3979                 addr = &dev->data->mac_addrs[i];
3980
3981                 /* skip zero address */
3982                 if (rte_is_zero_ether_addr(addr))
3983                         continue;
3984
3985                 pool = 0;
3986                 pool_mask = dev->data->mac_pool_sel[i];
3987
3988                 do {
3989                         if (pool_mask & 1ULL) {
3990                                 rc = bnxt_mac_addr_add_op(dev, addr, i, pool);
3991                                 if (rc)
3992                                         return rc;
3993                         }
3994                         pool_mask >>= 1;
3995                         pool++;
3996                 } while (pool_mask);
3997         }
3998
3999         return 0;
4000 }
4001
4002 static int bnxt_restore_filters(struct bnxt *bp)
4003 {
4004         struct rte_eth_dev *dev = bp->eth_dev;
4005         int ret = 0;
4006
4007         if (dev->data->all_multicast) {
4008                 ret = bnxt_allmulticast_enable_op(dev);
4009                 if (ret)
4010                         return ret;
4011         }
4012         if (dev->data->promiscuous) {
4013                 ret = bnxt_promiscuous_enable_op(dev);
4014                 if (ret)
4015                         return ret;
4016         }
4017
4018         ret = bnxt_restore_mac_filters(bp);
4019         if (ret)
4020                 return ret;
4021
4022         ret = bnxt_restore_vlan_filters(bp);
4023         /* TODO restore other filters as well */
4024         return ret;
4025 }
4026
4027 static int bnxt_check_fw_ready(struct bnxt *bp)
4028 {
4029         int timeout = bp->fw_reset_max_msecs;
4030         int rc = 0;
4031
4032         do {
4033                 rc = bnxt_hwrm_poll_ver_get(bp);
4034                 if (rc == 0)
4035                         break;
4036                 rte_delay_ms(BNXT_FW_READY_WAIT_INTERVAL);
4037                 timeout -= BNXT_FW_READY_WAIT_INTERVAL;
4038         } while (rc && timeout > 0);
4039
4040         if (rc)
4041                 PMD_DRV_LOG(ERR, "FW is not Ready after reset\n");
4042
4043         return rc;
4044 }
4045
4046 static void bnxt_dev_recover(void *arg)
4047 {
4048         struct bnxt *bp = arg;
4049         int rc = 0;
4050
4051         pthread_mutex_lock(&bp->err_recovery_lock);
4052
4053         if (!bp->fw_reset_min_msecs) {
4054                 rc = bnxt_check_fw_reset_done(bp);
4055                 if (rc)
4056                         goto err;
4057         }
4058
4059         /* Clear Error flag so that device re-init should happen */
4060         bp->flags &= ~BNXT_FLAG_FATAL_ERROR;
4061
4062         rc = bnxt_check_fw_ready(bp);
4063         if (rc)
4064                 goto err;
4065
4066         rc = bnxt_init_resources(bp, true);
4067         if (rc) {
4068                 PMD_DRV_LOG(ERR,
4069                             "Failed to initialize resources after reset\n");
4070                 goto err;
4071         }
4072         /* clear reset flag as the device is initialized now */
4073         bp->flags &= ~BNXT_FLAG_FW_RESET;
4074
4075         rc = bnxt_dev_start_op(bp->eth_dev);
4076         if (rc) {
4077                 PMD_DRV_LOG(ERR, "Failed to start port after reset\n");
4078                 goto err_start;
4079         }
4080
4081         rc = bnxt_restore_filters(bp);
4082         if (rc)
4083                 goto err_start;
4084
4085         PMD_DRV_LOG(INFO, "Recovered from FW reset\n");
4086         pthread_mutex_unlock(&bp->err_recovery_lock);
4087
4088         return;
4089 err_start:
4090         bnxt_dev_stop(bp->eth_dev);
4091 err:
4092         bp->flags |= BNXT_FLAG_FATAL_ERROR;
4093         bnxt_uninit_resources(bp, false);
4094         pthread_mutex_unlock(&bp->err_recovery_lock);
4095         PMD_DRV_LOG(ERR, "Failed to recover from FW reset\n");
4096 }
4097
4098 void bnxt_dev_reset_and_resume(void *arg)
4099 {
4100         struct bnxt *bp = arg;
4101         uint32_t us = US_PER_MS * bp->fw_reset_min_msecs;
4102         uint16_t val = 0;
4103         int rc;
4104
4105         bnxt_dev_cleanup(bp);
4106
4107         bnxt_wait_for_device_shutdown(bp);
4108
4109         /* During some fatal firmware error conditions, the PCI config space
4110          * register 0x2e which normally contains the subsystem ID will become
4111          * 0xffff. This register will revert back to the normal value after
4112          * the chip has completed core reset. If we detect this condition,
4113          * we can poll this config register immediately for the value to revert.
4114          */
4115         if (bp->flags & BNXT_FLAG_FATAL_ERROR) {
4116                 rc = rte_pci_read_config(bp->pdev, &val, sizeof(val), PCI_SUBSYSTEM_ID_OFFSET);
4117                 if (rc < 0) {
4118                         PMD_DRV_LOG(ERR, "Failed to read PCI offset 0x%x", PCI_SUBSYSTEM_ID_OFFSET);
4119                         return;
4120                 }
4121                 if (val == 0xffff) {
4122                         bp->fw_reset_min_msecs = 0;
4123                         us = 1;
4124                 }
4125         }
4126
4127         rc = rte_eal_alarm_set(us, bnxt_dev_recover, (void *)bp);
4128         if (rc)
4129                 PMD_DRV_LOG(ERR, "Error setting recovery alarm");
4130 }
4131
4132 uint32_t bnxt_read_fw_status_reg(struct bnxt *bp, uint32_t index)
4133 {
4134         struct bnxt_error_recovery_info *info = bp->recovery_info;
4135         uint32_t reg = info->status_regs[index];
4136         uint32_t type, offset, val = 0;
4137
4138         type = BNXT_FW_STATUS_REG_TYPE(reg);
4139         offset = BNXT_FW_STATUS_REG_OFF(reg);
4140
4141         switch (type) {
4142         case BNXT_FW_STATUS_REG_TYPE_CFG:
4143                 rte_pci_read_config(bp->pdev, &val, sizeof(val), offset);
4144                 break;
4145         case BNXT_FW_STATUS_REG_TYPE_GRC:
4146                 offset = info->mapped_status_regs[index];
4147                 /* FALLTHROUGH */
4148         case BNXT_FW_STATUS_REG_TYPE_BAR0:
4149                 val = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
4150                                        offset));
4151                 break;
4152         }
4153
4154         return val;
4155 }
4156
4157 static int bnxt_fw_reset_all(struct bnxt *bp)
4158 {
4159         struct bnxt_error_recovery_info *info = bp->recovery_info;
4160         uint32_t i;
4161         int rc = 0;
4162
4163         if (info->flags & BNXT_FLAG_ERROR_RECOVERY_HOST) {
4164                 /* Reset through master function driver */
4165                 for (i = 0; i < info->reg_array_cnt; i++)
4166                         bnxt_write_fw_reset_reg(bp, i);
4167                 /* Wait for time specified by FW after triggering reset */
4168                 rte_delay_ms(info->master_func_wait_period_after_reset);
4169         } else if (info->flags & BNXT_FLAG_ERROR_RECOVERY_CO_CPU) {
4170                 /* Reset with the help of Kong processor */
4171                 rc = bnxt_hwrm_fw_reset(bp);
4172                 if (rc)
4173                         PMD_DRV_LOG(ERR, "Failed to reset FW\n");
4174         }
4175
4176         return rc;
4177 }
4178
4179 static void bnxt_fw_reset_cb(void *arg)
4180 {
4181         struct bnxt *bp = arg;
4182         struct bnxt_error_recovery_info *info = bp->recovery_info;
4183         int rc = 0;
4184
4185         /* Only Master function can do FW reset */
4186         if (bnxt_is_master_func(bp) &&
4187             bnxt_is_recovery_enabled(bp)) {
4188                 rc = bnxt_fw_reset_all(bp);
4189                 if (rc) {
4190                         PMD_DRV_LOG(ERR, "Adapter recovery failed\n");
4191                         return;
4192                 }
4193         }
4194
4195         /* if recovery method is ERROR_RECOVERY_CO_CPU, KONG will send
4196          * EXCEPTION_FATAL_ASYNC event to all the functions
4197          * (including MASTER FUNC). After receiving this Async, all the active
4198          * drivers should treat this case as FW initiated recovery
4199          */
4200         if (info->flags & BNXT_FLAG_ERROR_RECOVERY_HOST) {
4201                 bp->fw_reset_min_msecs = BNXT_MIN_FW_READY_TIMEOUT;
4202                 bp->fw_reset_max_msecs = BNXT_MAX_FW_RESET_TIMEOUT;
4203
4204                 /* To recover from error */
4205                 rte_eal_alarm_set(US_PER_MS, bnxt_dev_reset_and_resume,
4206                                   (void *)bp);
4207         }
4208 }
4209
4210 /* Driver should poll FW heartbeat, reset_counter with the frequency
4211  * advertised by FW in HWRM_ERROR_RECOVERY_QCFG.
4212  * When the driver detects heartbeat stop or change in reset_counter,
4213  * it has to trigger a reset to recover from the error condition.
4214  * A “master PF” is the function who will have the privilege to
4215  * initiate the chimp reset. The master PF will be elected by the
4216  * firmware and will be notified through async message.
4217  */
4218 static void bnxt_check_fw_health(void *arg)
4219 {
4220         struct bnxt *bp = arg;
4221         struct bnxt_error_recovery_info *info = bp->recovery_info;
4222         uint32_t val = 0, wait_msec;
4223
4224         if (!info || !bnxt_is_recovery_enabled(bp) ||
4225             is_bnxt_in_error(bp))
4226                 return;
4227
4228         val = bnxt_read_fw_status_reg(bp, BNXT_FW_HEARTBEAT_CNT_REG);
4229         if (val == info->last_heart_beat)
4230                 goto reset;
4231
4232         info->last_heart_beat = val;
4233
4234         val = bnxt_read_fw_status_reg(bp, BNXT_FW_RECOVERY_CNT_REG);
4235         if (val != info->last_reset_counter)
4236                 goto reset;
4237
4238         info->last_reset_counter = val;
4239
4240         rte_eal_alarm_set(US_PER_MS * info->driver_polling_freq,
4241                           bnxt_check_fw_health, (void *)bp);
4242
4243         return;
4244 reset:
4245         /* Stop DMA to/from device */
4246         bp->flags |= BNXT_FLAG_FATAL_ERROR;
4247         bp->flags |= BNXT_FLAG_FW_RESET;
4248
4249         PMD_DRV_LOG(ERR, "Detected FW dead condition\n");
4250
4251         if (bnxt_is_master_func(bp))
4252                 wait_msec = info->master_func_wait_period;
4253         else
4254                 wait_msec = info->normal_func_wait_period;
4255
4256         rte_eal_alarm_set(US_PER_MS * wait_msec,
4257                           bnxt_fw_reset_cb, (void *)bp);
4258 }
4259
4260 void bnxt_schedule_fw_health_check(struct bnxt *bp)
4261 {
4262         uint32_t polling_freq;
4263
4264         pthread_mutex_lock(&bp->health_check_lock);
4265
4266         if (!bnxt_is_recovery_enabled(bp))
4267                 goto done;
4268
4269         if (bp->flags & BNXT_FLAG_FW_HEALTH_CHECK_SCHEDULED)
4270                 goto done;
4271
4272         polling_freq = bp->recovery_info->driver_polling_freq;
4273
4274         rte_eal_alarm_set(US_PER_MS * polling_freq,
4275                           bnxt_check_fw_health, (void *)bp);
4276         bp->flags |= BNXT_FLAG_FW_HEALTH_CHECK_SCHEDULED;
4277
4278 done:
4279         pthread_mutex_unlock(&bp->health_check_lock);
4280 }
4281
4282 static void bnxt_cancel_fw_health_check(struct bnxt *bp)
4283 {
4284         if (!bnxt_is_recovery_enabled(bp))
4285                 return;
4286
4287         rte_eal_alarm_cancel(bnxt_check_fw_health, (void *)bp);
4288         bp->flags &= ~BNXT_FLAG_FW_HEALTH_CHECK_SCHEDULED;
4289 }
4290
4291 static bool bnxt_vf_pciid(uint16_t device_id)
4292 {
4293         switch (device_id) {
4294         case BROADCOM_DEV_ID_57304_VF:
4295         case BROADCOM_DEV_ID_57406_VF:
4296         case BROADCOM_DEV_ID_5731X_VF:
4297         case BROADCOM_DEV_ID_5741X_VF:
4298         case BROADCOM_DEV_ID_57414_VF:
4299         case BROADCOM_DEV_ID_STRATUS_NIC_VF1:
4300         case BROADCOM_DEV_ID_STRATUS_NIC_VF2:
4301         case BROADCOM_DEV_ID_58802_VF:
4302         case BROADCOM_DEV_ID_57500_VF1:
4303         case BROADCOM_DEV_ID_57500_VF2:
4304         case BROADCOM_DEV_ID_58818_VF:
4305                 /* FALLTHROUGH */
4306                 return true;
4307         default:
4308                 return false;
4309         }
4310 }
4311
4312 /* Phase 5 device */
4313 static bool bnxt_p5_device(uint16_t device_id)
4314 {
4315         switch (device_id) {
4316         case BROADCOM_DEV_ID_57508:
4317         case BROADCOM_DEV_ID_57504:
4318         case BROADCOM_DEV_ID_57502:
4319         case BROADCOM_DEV_ID_57508_MF1:
4320         case BROADCOM_DEV_ID_57504_MF1:
4321         case BROADCOM_DEV_ID_57502_MF1:
4322         case BROADCOM_DEV_ID_57508_MF2:
4323         case BROADCOM_DEV_ID_57504_MF2:
4324         case BROADCOM_DEV_ID_57502_MF2:
4325         case BROADCOM_DEV_ID_57500_VF1:
4326         case BROADCOM_DEV_ID_57500_VF2:
4327         case BROADCOM_DEV_ID_58812:
4328         case BROADCOM_DEV_ID_58814:
4329         case BROADCOM_DEV_ID_58818:
4330         case BROADCOM_DEV_ID_58818_VF:
4331                 /* FALLTHROUGH */
4332                 return true;
4333         default:
4334                 return false;
4335         }
4336 }
4337
4338 bool bnxt_stratus_device(struct bnxt *bp)
4339 {
4340         uint16_t device_id = bp->pdev->id.device_id;
4341
4342         switch (device_id) {
4343         case BROADCOM_DEV_ID_STRATUS_NIC:
4344         case BROADCOM_DEV_ID_STRATUS_NIC_VF1:
4345         case BROADCOM_DEV_ID_STRATUS_NIC_VF2:
4346                 /* FALLTHROUGH */
4347                 return true;
4348         default:
4349                 return false;
4350         }
4351 }
4352
4353 static int bnxt_map_pci_bars(struct rte_eth_dev *eth_dev)
4354 {
4355         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
4356         struct bnxt *bp = eth_dev->data->dev_private;
4357
4358         /* enable device (incl. PCI PM wakeup), and bus-mastering */
4359         bp->bar0 = (void *)pci_dev->mem_resource[0].addr;
4360         bp->doorbell_base = (void *)pci_dev->mem_resource[2].addr;
4361         if (!bp->bar0 || !bp->doorbell_base) {
4362                 PMD_DRV_LOG(ERR, "Unable to access Hardware\n");
4363                 return -ENODEV;
4364         }
4365
4366         bp->eth_dev = eth_dev;
4367         bp->pdev = pci_dev;
4368
4369         return 0;
4370 }
4371
4372 static int bnxt_alloc_ctx_mem_blk(struct bnxt *bp,
4373                                   struct bnxt_ctx_pg_info *ctx_pg,
4374                                   uint32_t mem_size,
4375                                   const char *suffix,
4376                                   uint16_t idx)
4377 {
4378         struct bnxt_ring_mem_info *rmem = &ctx_pg->ring_mem;
4379         const struct rte_memzone *mz = NULL;
4380         char mz_name[RTE_MEMZONE_NAMESIZE];
4381         rte_iova_t mz_phys_addr;
4382         uint64_t valid_bits = 0;
4383         uint32_t sz;
4384         int i;
4385
4386         if (!mem_size)
4387                 return 0;
4388
4389         rmem->nr_pages = RTE_ALIGN_MUL_CEIL(mem_size, BNXT_PAGE_SIZE) /
4390                          BNXT_PAGE_SIZE;
4391         rmem->page_size = BNXT_PAGE_SIZE;
4392         rmem->pg_arr = ctx_pg->ctx_pg_arr;
4393         rmem->dma_arr = ctx_pg->ctx_dma_arr;
4394         rmem->flags = BNXT_RMEM_VALID_PTE_FLAG;
4395
4396         valid_bits = PTU_PTE_VALID;
4397
4398         if (rmem->nr_pages > 1) {
4399                 snprintf(mz_name, RTE_MEMZONE_NAMESIZE,
4400                          "bnxt_ctx_pg_tbl%s_%x_%d",
4401                          suffix, idx, bp->eth_dev->data->port_id);
4402                 mz_name[RTE_MEMZONE_NAMESIZE - 1] = 0;
4403                 mz = rte_memzone_lookup(mz_name);
4404                 if (!mz) {
4405                         mz = rte_memzone_reserve_aligned(mz_name,
4406                                                 rmem->nr_pages * 8,
4407                                                 SOCKET_ID_ANY,
4408                                                 RTE_MEMZONE_2MB |
4409                                                 RTE_MEMZONE_SIZE_HINT_ONLY |
4410                                                 RTE_MEMZONE_IOVA_CONTIG,
4411                                                 BNXT_PAGE_SIZE);
4412                         if (mz == NULL)
4413                                 return -ENOMEM;
4414                 }
4415
4416                 memset(mz->addr, 0, mz->len);
4417                 mz_phys_addr = mz->iova;
4418
4419                 rmem->pg_tbl = mz->addr;
4420                 rmem->pg_tbl_map = mz_phys_addr;
4421                 rmem->pg_tbl_mz = mz;
4422         }
4423
4424         snprintf(mz_name, RTE_MEMZONE_NAMESIZE, "bnxt_ctx_%s_%x_%d",
4425                  suffix, idx, bp->eth_dev->data->port_id);
4426         mz = rte_memzone_lookup(mz_name);
4427         if (!mz) {
4428                 mz = rte_memzone_reserve_aligned(mz_name,
4429                                                  mem_size,
4430                                                  SOCKET_ID_ANY,
4431                                                  RTE_MEMZONE_1GB |
4432                                                  RTE_MEMZONE_SIZE_HINT_ONLY |
4433                                                  RTE_MEMZONE_IOVA_CONTIG,
4434                                                  BNXT_PAGE_SIZE);
4435                 if (mz == NULL)
4436                         return -ENOMEM;
4437         }
4438
4439         memset(mz->addr, 0, mz->len);
4440         mz_phys_addr = mz->iova;
4441
4442         for (sz = 0, i = 0; sz < mem_size; sz += BNXT_PAGE_SIZE, i++) {
4443                 rmem->pg_arr[i] = ((char *)mz->addr) + sz;
4444                 rmem->dma_arr[i] = mz_phys_addr + sz;
4445
4446                 if (rmem->nr_pages > 1) {
4447                         if (i == rmem->nr_pages - 2 &&
4448                             (rmem->flags & BNXT_RMEM_RING_PTE_FLAG))
4449                                 valid_bits |= PTU_PTE_NEXT_TO_LAST;
4450                         else if (i == rmem->nr_pages - 1 &&
4451                                  (rmem->flags & BNXT_RMEM_RING_PTE_FLAG))
4452                                 valid_bits |= PTU_PTE_LAST;
4453
4454                         rmem->pg_tbl[i] = rte_cpu_to_le_64(rmem->dma_arr[i] |
4455                                                            valid_bits);
4456                 }
4457         }
4458
4459         rmem->mz = mz;
4460         if (rmem->vmem_size)
4461                 rmem->vmem = (void **)mz->addr;
4462         rmem->dma_arr[0] = mz_phys_addr;
4463         return 0;
4464 }
4465
4466 static void bnxt_free_ctx_mem(struct bnxt *bp)
4467 {
4468         int i;
4469
4470         if (!bp->ctx || !(bp->ctx->flags & BNXT_CTX_FLAG_INITED))
4471                 return;
4472
4473         bp->ctx->flags &= ~BNXT_CTX_FLAG_INITED;
4474         rte_memzone_free(bp->ctx->qp_mem.ring_mem.mz);
4475         rte_memzone_free(bp->ctx->srq_mem.ring_mem.mz);
4476         rte_memzone_free(bp->ctx->cq_mem.ring_mem.mz);
4477         rte_memzone_free(bp->ctx->vnic_mem.ring_mem.mz);
4478         rte_memzone_free(bp->ctx->stat_mem.ring_mem.mz);
4479         rte_memzone_free(bp->ctx->qp_mem.ring_mem.pg_tbl_mz);
4480         rte_memzone_free(bp->ctx->srq_mem.ring_mem.pg_tbl_mz);
4481         rte_memzone_free(bp->ctx->cq_mem.ring_mem.pg_tbl_mz);
4482         rte_memzone_free(bp->ctx->vnic_mem.ring_mem.pg_tbl_mz);
4483         rte_memzone_free(bp->ctx->stat_mem.ring_mem.pg_tbl_mz);
4484
4485         for (i = 0; i < bp->ctx->tqm_fp_rings_count + 1; i++) {
4486                 if (bp->ctx->tqm_mem[i])
4487                         rte_memzone_free(bp->ctx->tqm_mem[i]->ring_mem.mz);
4488         }
4489
4490         rte_free(bp->ctx);
4491         bp->ctx = NULL;
4492 }
4493
4494 #define bnxt_roundup(x, y)   ((((x) + ((y) - 1)) / (y)) * (y))
4495
4496 #define min_t(type, x, y) ({                    \
4497         type __min1 = (x);                      \
4498         type __min2 = (y);                      \
4499         __min1 < __min2 ? __min1 : __min2; })
4500
4501 #define max_t(type, x, y) ({                    \
4502         type __max1 = (x);                      \
4503         type __max2 = (y);                      \
4504         __max1 > __max2 ? __max1 : __max2; })
4505
4506 #define clamp_t(type, _x, min, max)     min_t(type, max_t(type, _x, min), max)
4507
4508 int bnxt_alloc_ctx_mem(struct bnxt *bp)
4509 {
4510         struct bnxt_ctx_pg_info *ctx_pg;
4511         struct bnxt_ctx_mem_info *ctx;
4512         uint32_t mem_size, ena, entries;
4513         uint32_t entries_sp, min;
4514         int i, rc;
4515
4516         rc = bnxt_hwrm_func_backing_store_qcaps(bp);
4517         if (rc) {
4518                 PMD_DRV_LOG(ERR, "Query context mem capability failed\n");
4519                 return rc;
4520         }
4521         ctx = bp->ctx;
4522         if (!ctx || (ctx->flags & BNXT_CTX_FLAG_INITED))
4523                 return 0;
4524
4525         ctx_pg = &ctx->qp_mem;
4526         ctx_pg->entries = ctx->qp_min_qp1_entries + ctx->qp_max_l2_entries;
4527         if (ctx->qp_entry_size) {
4528                 mem_size = ctx->qp_entry_size * ctx_pg->entries;
4529                 rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "qp_mem", 0);
4530                 if (rc)
4531                         return rc;
4532         }
4533
4534         ctx_pg = &ctx->srq_mem;
4535         ctx_pg->entries = ctx->srq_max_l2_entries;
4536         if (ctx->srq_entry_size) {
4537                 mem_size = ctx->srq_entry_size * ctx_pg->entries;
4538                 rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "srq_mem", 0);
4539                 if (rc)
4540                         return rc;
4541         }
4542
4543         ctx_pg = &ctx->cq_mem;
4544         ctx_pg->entries = ctx->cq_max_l2_entries;
4545         if (ctx->cq_entry_size) {
4546                 mem_size = ctx->cq_entry_size * ctx_pg->entries;
4547                 rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "cq_mem", 0);
4548                 if (rc)
4549                         return rc;
4550         }
4551
4552         ctx_pg = &ctx->vnic_mem;
4553         ctx_pg->entries = ctx->vnic_max_vnic_entries +
4554                 ctx->vnic_max_ring_table_entries;
4555         if (ctx->vnic_entry_size) {
4556                 mem_size = ctx->vnic_entry_size * ctx_pg->entries;
4557                 rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "vnic_mem", 0);
4558                 if (rc)
4559                         return rc;
4560         }
4561
4562         ctx_pg = &ctx->stat_mem;
4563         ctx_pg->entries = ctx->stat_max_entries;
4564         if (ctx->stat_entry_size) {
4565                 mem_size = ctx->stat_entry_size * ctx_pg->entries;
4566                 rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "stat_mem", 0);
4567                 if (rc)
4568                         return rc;
4569         }
4570
4571         min = ctx->tqm_min_entries_per_ring;
4572
4573         entries_sp = ctx->qp_max_l2_entries +
4574                      ctx->vnic_max_vnic_entries +
4575                      2 * ctx->qp_min_qp1_entries + min;
4576         entries_sp = bnxt_roundup(entries_sp, ctx->tqm_entries_multiple);
4577
4578         entries = ctx->qp_max_l2_entries + ctx->qp_min_qp1_entries;
4579         entries = bnxt_roundup(entries, ctx->tqm_entries_multiple);
4580         entries = clamp_t(uint32_t, entries, min,
4581                           ctx->tqm_max_entries_per_ring);
4582         for (i = 0, ena = 0; i < ctx->tqm_fp_rings_count + 1; i++) {
4583                 /* i=0 is for TQM_SP. i=1 to i=8 applies to RING0 to RING7.
4584                  * i > 8 is other ext rings.
4585                  */
4586                 ctx_pg = ctx->tqm_mem[i];
4587                 ctx_pg->entries = i ? entries : entries_sp;
4588                 if (ctx->tqm_entry_size) {
4589                         mem_size = ctx->tqm_entry_size * ctx_pg->entries;
4590                         rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size,
4591                                                     "tqm_mem", i);
4592                         if (rc)
4593                                 return rc;
4594                 }
4595                 if (i < BNXT_MAX_TQM_LEGACY_RINGS)
4596                         ena |= HWRM_FUNC_BACKING_STORE_CFG_INPUT_ENABLES_TQM_SP << i;
4597                 else
4598                         ena |= HWRM_FUNC_BACKING_STORE_CFG_INPUT_ENABLES_TQM_RING8;
4599         }
4600
4601         ena |= FUNC_BACKING_STORE_CFG_INPUT_DFLT_ENABLES;
4602         rc = bnxt_hwrm_func_backing_store_cfg(bp, ena);
4603         if (rc)
4604                 PMD_DRV_LOG(ERR,
4605                             "Failed to configure context mem: rc = %d\n", rc);
4606         else
4607                 ctx->flags |= BNXT_CTX_FLAG_INITED;
4608
4609         return rc;
4610 }
4611
4612 static int bnxt_alloc_stats_mem(struct bnxt *bp)
4613 {
4614         struct rte_pci_device *pci_dev = bp->pdev;
4615         char mz_name[RTE_MEMZONE_NAMESIZE];
4616         const struct rte_memzone *mz = NULL;
4617         uint32_t total_alloc_len;
4618         rte_iova_t mz_phys_addr;
4619
4620         if (pci_dev->id.device_id == BROADCOM_DEV_ID_NS2)
4621                 return 0;
4622
4623         snprintf(mz_name, RTE_MEMZONE_NAMESIZE,
4624                  "bnxt_" PCI_PRI_FMT "-%s", pci_dev->addr.domain,
4625                  pci_dev->addr.bus, pci_dev->addr.devid,
4626                  pci_dev->addr.function, "rx_port_stats");
4627         mz_name[RTE_MEMZONE_NAMESIZE - 1] = 0;
4628         mz = rte_memzone_lookup(mz_name);
4629         total_alloc_len =
4630                 RTE_CACHE_LINE_ROUNDUP(sizeof(struct rx_port_stats) +
4631                                        sizeof(struct rx_port_stats_ext) + 512);
4632         if (!mz) {
4633                 mz = rte_memzone_reserve(mz_name, total_alloc_len,
4634                                          SOCKET_ID_ANY,
4635                                          RTE_MEMZONE_2MB |
4636                                          RTE_MEMZONE_SIZE_HINT_ONLY |
4637                                          RTE_MEMZONE_IOVA_CONTIG);
4638                 if (mz == NULL)
4639                         return -ENOMEM;
4640         }
4641         memset(mz->addr, 0, mz->len);
4642         mz_phys_addr = mz->iova;
4643
4644         bp->rx_mem_zone = (const void *)mz;
4645         bp->hw_rx_port_stats = mz->addr;
4646         bp->hw_rx_port_stats_map = mz_phys_addr;
4647
4648         snprintf(mz_name, RTE_MEMZONE_NAMESIZE,
4649                  "bnxt_" PCI_PRI_FMT "-%s", pci_dev->addr.domain,
4650                  pci_dev->addr.bus, pci_dev->addr.devid,
4651                  pci_dev->addr.function, "tx_port_stats");
4652         mz_name[RTE_MEMZONE_NAMESIZE - 1] = 0;
4653         mz = rte_memzone_lookup(mz_name);
4654         total_alloc_len =
4655                 RTE_CACHE_LINE_ROUNDUP(sizeof(struct tx_port_stats) +
4656                                        sizeof(struct tx_port_stats_ext) + 512);
4657         if (!mz) {
4658                 mz = rte_memzone_reserve(mz_name,
4659                                          total_alloc_len,
4660                                          SOCKET_ID_ANY,
4661                                          RTE_MEMZONE_2MB |
4662                                          RTE_MEMZONE_SIZE_HINT_ONLY |
4663                                          RTE_MEMZONE_IOVA_CONTIG);
4664                 if (mz == NULL)
4665                         return -ENOMEM;
4666         }
4667         memset(mz->addr, 0, mz->len);
4668         mz_phys_addr = mz->iova;
4669
4670         bp->tx_mem_zone = (const void *)mz;
4671         bp->hw_tx_port_stats = mz->addr;
4672         bp->hw_tx_port_stats_map = mz_phys_addr;
4673         bp->flags |= BNXT_FLAG_PORT_STATS;
4674
4675         /* Display extended statistics if FW supports it */
4676         if (bp->hwrm_spec_code < HWRM_SPEC_CODE_1_8_4 ||
4677             bp->hwrm_spec_code == HWRM_SPEC_CODE_1_9_0 ||
4678             !(bp->flags & BNXT_FLAG_EXT_STATS_SUPPORTED))
4679                 return 0;
4680
4681         bp->hw_rx_port_stats_ext = (void *)
4682                 ((uint8_t *)bp->hw_rx_port_stats +
4683                  sizeof(struct rx_port_stats));
4684         bp->hw_rx_port_stats_ext_map = bp->hw_rx_port_stats_map +
4685                 sizeof(struct rx_port_stats);
4686         bp->flags |= BNXT_FLAG_EXT_RX_PORT_STATS;
4687
4688         if (bp->hwrm_spec_code < HWRM_SPEC_CODE_1_9_2 ||
4689             bp->flags & BNXT_FLAG_EXT_STATS_SUPPORTED) {
4690                 bp->hw_tx_port_stats_ext = (void *)
4691                         ((uint8_t *)bp->hw_tx_port_stats +
4692                          sizeof(struct tx_port_stats));
4693                 bp->hw_tx_port_stats_ext_map =
4694                         bp->hw_tx_port_stats_map +
4695                         sizeof(struct tx_port_stats);
4696                 bp->flags |= BNXT_FLAG_EXT_TX_PORT_STATS;
4697         }
4698
4699         return 0;
4700 }
4701
4702 static int bnxt_setup_mac_addr(struct rte_eth_dev *eth_dev)
4703 {
4704         struct bnxt *bp = eth_dev->data->dev_private;
4705         int rc = 0;
4706
4707         eth_dev->data->mac_addrs = rte_zmalloc("bnxt_mac_addr_tbl",
4708                                                RTE_ETHER_ADDR_LEN *
4709                                                bp->max_l2_ctx,
4710                                                0);
4711         if (eth_dev->data->mac_addrs == NULL) {
4712                 PMD_DRV_LOG(ERR, "Failed to alloc MAC addr tbl\n");
4713                 return -ENOMEM;
4714         }
4715
4716         if (!BNXT_HAS_DFLT_MAC_SET(bp)) {
4717                 if (BNXT_PF(bp))
4718                         return -EINVAL;
4719
4720                 /* Generate a random MAC address, if none was assigned by PF */
4721                 PMD_DRV_LOG(INFO, "VF MAC address not assigned by Host PF\n");
4722                 bnxt_eth_hw_addr_random(bp->mac_addr);
4723                 PMD_DRV_LOG(INFO,
4724                             "Assign random MAC:%02X:%02X:%02X:%02X:%02X:%02X\n",
4725                             bp->mac_addr[0], bp->mac_addr[1], bp->mac_addr[2],
4726                             bp->mac_addr[3], bp->mac_addr[4], bp->mac_addr[5]);
4727
4728                 rc = bnxt_hwrm_set_mac(bp);
4729                 if (rc)
4730                         return rc;
4731         }
4732
4733         /* Copy the permanent MAC from the FUNC_QCAPS response */
4734         memcpy(&eth_dev->data->mac_addrs[0], bp->mac_addr, RTE_ETHER_ADDR_LEN);
4735
4736         return rc;
4737 }
4738
4739 static int bnxt_restore_dflt_mac(struct bnxt *bp)
4740 {
4741         int rc = 0;
4742
4743         /* MAC is already configured in FW */
4744         if (BNXT_HAS_DFLT_MAC_SET(bp))
4745                 return 0;
4746
4747         /* Restore the old MAC configured */
4748         rc = bnxt_hwrm_set_mac(bp);
4749         if (rc)
4750                 PMD_DRV_LOG(ERR, "Failed to restore MAC address\n");
4751
4752         return rc;
4753 }
4754
4755 static void bnxt_config_vf_req_fwd(struct bnxt *bp)
4756 {
4757         if (!BNXT_PF(bp))
4758                 return;
4759
4760         memset(bp->pf->vf_req_fwd, 0, sizeof(bp->pf->vf_req_fwd));
4761
4762         if (!(bp->fw_cap & BNXT_FW_CAP_LINK_ADMIN))
4763                 BNXT_HWRM_CMD_TO_FORWARD(HWRM_PORT_PHY_QCFG);
4764         BNXT_HWRM_CMD_TO_FORWARD(HWRM_FUNC_CFG);
4765         BNXT_HWRM_CMD_TO_FORWARD(HWRM_FUNC_VF_CFG);
4766         BNXT_HWRM_CMD_TO_FORWARD(HWRM_CFA_L2_FILTER_ALLOC);
4767         BNXT_HWRM_CMD_TO_FORWARD(HWRM_OEM_CMD);
4768 }
4769
4770 uint16_t
4771 bnxt_get_svif(uint16_t port_id, bool func_svif,
4772               enum bnxt_ulp_intf_type type)
4773 {
4774         struct rte_eth_dev *eth_dev;
4775         struct bnxt *bp;
4776
4777         eth_dev = &rte_eth_devices[port_id];
4778         if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
4779                 struct bnxt_representor *vfr = eth_dev->data->dev_private;
4780                 if (!vfr)
4781                         return 0;
4782
4783                 if (type == BNXT_ULP_INTF_TYPE_VF_REP)
4784                         return vfr->svif;
4785
4786                 eth_dev = vfr->parent_dev;
4787         }
4788
4789         bp = eth_dev->data->dev_private;
4790
4791         return func_svif ? bp->func_svif : bp->port_svif;
4792 }
4793
4794 uint16_t
4795 bnxt_get_vnic_id(uint16_t port, enum bnxt_ulp_intf_type type)
4796 {
4797         struct rte_eth_dev *eth_dev;
4798         struct bnxt_vnic_info *vnic;
4799         struct bnxt *bp;
4800
4801         eth_dev = &rte_eth_devices[port];
4802         if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
4803                 struct bnxt_representor *vfr = eth_dev->data->dev_private;
4804                 if (!vfr)
4805                         return 0;
4806
4807                 if (type == BNXT_ULP_INTF_TYPE_VF_REP)
4808                         return vfr->dflt_vnic_id;
4809
4810                 eth_dev = vfr->parent_dev;
4811         }
4812
4813         bp = eth_dev->data->dev_private;
4814
4815         vnic = BNXT_GET_DEFAULT_VNIC(bp);
4816
4817         return vnic->fw_vnic_id;
4818 }
4819
4820 uint16_t
4821 bnxt_get_fw_func_id(uint16_t port, enum bnxt_ulp_intf_type type)
4822 {
4823         struct rte_eth_dev *eth_dev;
4824         struct bnxt *bp;
4825
4826         eth_dev = &rte_eth_devices[port];
4827         if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
4828                 struct bnxt_representor *vfr = eth_dev->data->dev_private;
4829                 if (!vfr)
4830                         return 0;
4831
4832                 if (type == BNXT_ULP_INTF_TYPE_VF_REP)
4833                         return vfr->fw_fid;
4834
4835                 eth_dev = vfr->parent_dev;
4836         }
4837
4838         bp = eth_dev->data->dev_private;
4839
4840         return bp->fw_fid;
4841 }
4842
4843 enum bnxt_ulp_intf_type
4844 bnxt_get_interface_type(uint16_t port)
4845 {
4846         struct rte_eth_dev *eth_dev;
4847         struct bnxt *bp;
4848
4849         eth_dev = &rte_eth_devices[port];
4850         if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev))
4851                 return BNXT_ULP_INTF_TYPE_VF_REP;
4852
4853         bp = eth_dev->data->dev_private;
4854         if (BNXT_PF(bp))
4855                 return BNXT_ULP_INTF_TYPE_PF;
4856         else if (BNXT_VF_IS_TRUSTED(bp))
4857                 return BNXT_ULP_INTF_TYPE_TRUSTED_VF;
4858         else if (BNXT_VF(bp))
4859                 return BNXT_ULP_INTF_TYPE_VF;
4860
4861         return BNXT_ULP_INTF_TYPE_INVALID;
4862 }
4863
4864 uint16_t
4865 bnxt_get_phy_port_id(uint16_t port_id)
4866 {
4867         struct bnxt_representor *vfr;
4868         struct rte_eth_dev *eth_dev;
4869         struct bnxt *bp;
4870
4871         eth_dev = &rte_eth_devices[port_id];
4872         if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
4873                 vfr = eth_dev->data->dev_private;
4874                 if (!vfr)
4875                         return 0;
4876
4877                 eth_dev = vfr->parent_dev;
4878         }
4879
4880         bp = eth_dev->data->dev_private;
4881
4882         return BNXT_PF(bp) ? bp->pf->port_id : bp->parent->port_id;
4883 }
4884
4885 uint16_t
4886 bnxt_get_parif(uint16_t port_id, enum bnxt_ulp_intf_type type)
4887 {
4888         struct rte_eth_dev *eth_dev;
4889         struct bnxt *bp;
4890
4891         eth_dev = &rte_eth_devices[port_id];
4892         if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
4893                 struct bnxt_representor *vfr = eth_dev->data->dev_private;
4894                 if (!vfr)
4895                         return 0;
4896
4897                 if (type == BNXT_ULP_INTF_TYPE_VF_REP)
4898                         return vfr->fw_fid - 1;
4899
4900                 eth_dev = vfr->parent_dev;
4901         }
4902
4903         bp = eth_dev->data->dev_private;
4904
4905         return BNXT_PF(bp) ? bp->fw_fid - 1 : bp->parent->fid - 1;
4906 }
4907
4908 uint16_t
4909 bnxt_get_vport(uint16_t port_id)
4910 {
4911         return (1 << bnxt_get_phy_port_id(port_id));
4912 }
4913
4914 static void bnxt_alloc_error_recovery_info(struct bnxt *bp)
4915 {
4916         struct bnxt_error_recovery_info *info = bp->recovery_info;
4917
4918         if (info) {
4919                 if (!(bp->fw_cap & BNXT_FW_CAP_HCOMM_FW_STATUS))
4920                         memset(info, 0, sizeof(*info));
4921                 return;
4922         }
4923
4924         if (!(bp->fw_cap & BNXT_FW_CAP_ERROR_RECOVERY))
4925                 return;
4926
4927         info = rte_zmalloc("bnxt_hwrm_error_recovery_qcfg",
4928                            sizeof(*info), 0);
4929         if (!info)
4930                 bp->fw_cap &= ~BNXT_FW_CAP_ERROR_RECOVERY;
4931
4932         bp->recovery_info = info;
4933 }
4934
4935 static void bnxt_check_fw_status(struct bnxt *bp)
4936 {
4937         uint32_t fw_status;
4938
4939         if (!(bp->recovery_info &&
4940               (bp->fw_cap & BNXT_FW_CAP_HCOMM_FW_STATUS)))
4941                 return;
4942
4943         fw_status = bnxt_read_fw_status_reg(bp, BNXT_FW_STATUS_REG);
4944         if (fw_status != BNXT_FW_STATUS_HEALTHY)
4945                 PMD_DRV_LOG(ERR, "Firmware not responding, status: %#x\n",
4946                             fw_status);
4947 }
4948
4949 static int bnxt_map_hcomm_fw_status_reg(struct bnxt *bp)
4950 {
4951         struct bnxt_error_recovery_info *info = bp->recovery_info;
4952         uint32_t status_loc;
4953         uint32_t sig_ver;
4954
4955         rte_write32(HCOMM_STATUS_STRUCT_LOC, (uint8_t *)bp->bar0 +
4956                     BNXT_GRCPF_REG_WINDOW_BASE_OUT + 4);
4957         sig_ver = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
4958                                    BNXT_GRCP_WINDOW_2_BASE +
4959                                    offsetof(struct hcomm_status,
4960                                             sig_ver)));
4961         /* If the signature is absent, then FW does not support this feature */
4962         if ((sig_ver & HCOMM_STATUS_SIGNATURE_MASK) !=
4963             HCOMM_STATUS_SIGNATURE_VAL)
4964                 return 0;
4965
4966         if (!info) {
4967                 info = rte_zmalloc("bnxt_hwrm_error_recovery_qcfg",
4968                                    sizeof(*info), 0);
4969                 if (!info)
4970                         return -ENOMEM;
4971                 bp->recovery_info = info;
4972         } else {
4973                 memset(info, 0, sizeof(*info));
4974         }
4975
4976         status_loc = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
4977                                       BNXT_GRCP_WINDOW_2_BASE +
4978                                       offsetof(struct hcomm_status,
4979                                                fw_status_loc)));
4980
4981         /* Only pre-map the FW health status GRC register */
4982         if (BNXT_FW_STATUS_REG_TYPE(status_loc) != BNXT_FW_STATUS_REG_TYPE_GRC)
4983                 return 0;
4984
4985         info->status_regs[BNXT_FW_STATUS_REG] = status_loc;
4986         info->mapped_status_regs[BNXT_FW_STATUS_REG] =
4987                 BNXT_GRCP_WINDOW_2_BASE + (status_loc & BNXT_GRCP_OFFSET_MASK);
4988
4989         rte_write32((status_loc & BNXT_GRCP_BASE_MASK), (uint8_t *)bp->bar0 +
4990                     BNXT_GRCPF_REG_WINDOW_BASE_OUT + 4);
4991
4992         bp->fw_cap |= BNXT_FW_CAP_HCOMM_FW_STATUS;
4993
4994         return 0;
4995 }
4996
4997 /* This function gets the FW version along with the
4998  * capabilities(MAX and current) of the function, vnic,
4999  * error recovery, phy and other chip related info
5000  */
5001 static int bnxt_get_config(struct bnxt *bp)
5002 {
5003         uint16_t mtu;
5004         int rc = 0;
5005
5006         bp->fw_cap = 0;
5007
5008         rc = bnxt_map_hcomm_fw_status_reg(bp);
5009         if (rc)
5010                 return rc;
5011
5012         rc = bnxt_hwrm_ver_get(bp, DFLT_HWRM_CMD_TIMEOUT);
5013         if (rc) {
5014                 bnxt_check_fw_status(bp);
5015                 return rc;
5016         }
5017
5018         rc = bnxt_hwrm_func_reset(bp);
5019         if (rc)
5020                 return -EIO;
5021
5022         rc = bnxt_hwrm_vnic_qcaps(bp);
5023         if (rc)
5024                 return rc;
5025
5026         rc = bnxt_hwrm_queue_qportcfg(bp);
5027         if (rc)
5028                 return rc;
5029
5030         /* Get the MAX capabilities for this function.
5031          * This function also allocates context memory for TQM rings and
5032          * informs the firmware about this allocated backing store memory.
5033          */
5034         rc = bnxt_hwrm_func_qcaps(bp);
5035         if (rc)
5036                 return rc;
5037
5038         rc = bnxt_hwrm_func_qcfg(bp, &mtu);
5039         if (rc)
5040                 return rc;
5041
5042         rc = bnxt_hwrm_cfa_adv_flow_mgmt_qcaps(bp);
5043         if (rc)
5044                 return rc;
5045
5046         bnxt_hwrm_port_mac_qcfg(bp);
5047
5048         bnxt_hwrm_parent_pf_qcfg(bp);
5049
5050         bnxt_hwrm_port_phy_qcaps(bp);
5051
5052         bnxt_alloc_error_recovery_info(bp);
5053         /* Get the adapter error recovery support info */
5054         rc = bnxt_hwrm_error_recovery_qcfg(bp);
5055         if (rc)
5056                 bp->fw_cap &= ~BNXT_FW_CAP_ERROR_RECOVERY;
5057
5058         bnxt_hwrm_port_led_qcaps(bp);
5059
5060         return 0;
5061 }
5062
5063 static int
5064 bnxt_init_locks(struct bnxt *bp)
5065 {
5066         int err;
5067
5068         err = pthread_mutex_init(&bp->flow_lock, NULL);
5069         if (err) {
5070                 PMD_DRV_LOG(ERR, "Unable to initialize flow_lock\n");
5071                 return err;
5072         }
5073
5074         err = pthread_mutex_init(&bp->def_cp_lock, NULL);
5075         if (err) {
5076                 PMD_DRV_LOG(ERR, "Unable to initialize def_cp_lock\n");
5077                 return err;
5078         }
5079
5080         err = pthread_mutex_init(&bp->health_check_lock, NULL);
5081         if (err) {
5082                 PMD_DRV_LOG(ERR, "Unable to initialize health_check_lock\n");
5083                 return err;
5084         }
5085
5086         err = pthread_mutex_init(&bp->err_recovery_lock, NULL);
5087         if (err)
5088                 PMD_DRV_LOG(ERR, "Unable to initialize err_recovery_lock\n");
5089
5090         return err;
5091 }
5092
5093 static int bnxt_init_resources(struct bnxt *bp, bool reconfig_dev)
5094 {
5095         int rc = 0;
5096
5097         rc = bnxt_get_config(bp);
5098         if (rc)
5099                 return rc;
5100
5101         if (!reconfig_dev) {
5102                 rc = bnxt_setup_mac_addr(bp->eth_dev);
5103                 if (rc)
5104                         return rc;
5105         } else {
5106                 rc = bnxt_restore_dflt_mac(bp);
5107                 if (rc)
5108                         return rc;
5109         }
5110
5111         bnxt_config_vf_req_fwd(bp);
5112
5113         rc = bnxt_hwrm_func_driver_register(bp);
5114         if (rc) {
5115                 PMD_DRV_LOG(ERR, "Failed to register driver");
5116                 return -EBUSY;
5117         }
5118
5119         if (BNXT_PF(bp)) {
5120                 if (bp->pdev->max_vfs) {
5121                         rc = bnxt_hwrm_allocate_vfs(bp, bp->pdev->max_vfs);
5122                         if (rc) {
5123                                 PMD_DRV_LOG(ERR, "Failed to allocate VFs\n");
5124                                 return rc;
5125                         }
5126                 } else {
5127                         rc = bnxt_hwrm_allocate_pf_only(bp);
5128                         if (rc) {
5129                                 PMD_DRV_LOG(ERR,
5130                                             "Failed to allocate PF resources");
5131                                 return rc;
5132                         }
5133                 }
5134         }
5135
5136         rc = bnxt_alloc_mem(bp, reconfig_dev);
5137         if (rc)
5138                 return rc;
5139
5140         rc = bnxt_setup_int(bp);
5141         if (rc)
5142                 return rc;
5143
5144         rc = bnxt_request_int(bp);
5145         if (rc)
5146                 return rc;
5147
5148         rc = bnxt_init_ctx_mem(bp);
5149         if (rc) {
5150                 PMD_DRV_LOG(ERR, "Failed to init adv_flow_counters\n");
5151                 return rc;
5152         }
5153
5154         return 0;
5155 }
5156
5157 static int
5158 bnxt_parse_devarg_truflow(__rte_unused const char *key,
5159                           const char *value, void *opaque_arg)
5160 {
5161         struct bnxt *bp = opaque_arg;
5162         unsigned long truflow;
5163         char *end = NULL;
5164
5165         if (!value || !opaque_arg) {
5166                 PMD_DRV_LOG(ERR,
5167                             "Invalid parameter passed to truflow devargs.\n");
5168                 return -EINVAL;
5169         }
5170
5171         truflow = strtoul(value, &end, 10);
5172         if (end == NULL || *end != '\0' ||
5173             (truflow == ULONG_MAX && errno == ERANGE)) {
5174                 PMD_DRV_LOG(ERR,
5175                             "Invalid parameter passed to truflow devargs.\n");
5176                 return -EINVAL;
5177         }
5178
5179         if (BNXT_DEVARG_TRUFLOW_INVALID(truflow)) {
5180                 PMD_DRV_LOG(ERR,
5181                             "Invalid value passed to truflow devargs.\n");
5182                 return -EINVAL;
5183         }
5184
5185         if (truflow) {
5186                 bp->flags |= BNXT_FLAG_TRUFLOW_EN;
5187                 PMD_DRV_LOG(INFO, "Host-based truflow feature enabled.\n");
5188         } else {
5189                 bp->flags &= ~BNXT_FLAG_TRUFLOW_EN;
5190                 PMD_DRV_LOG(INFO, "Host-based truflow feature disabled.\n");
5191         }
5192
5193         return 0;
5194 }
5195
5196 static int
5197 bnxt_parse_devarg_flow_xstat(__rte_unused const char *key,
5198                              const char *value, void *opaque_arg)
5199 {
5200         struct bnxt *bp = opaque_arg;
5201         unsigned long flow_xstat;
5202         char *end = NULL;
5203
5204         if (!value || !opaque_arg) {
5205                 PMD_DRV_LOG(ERR,
5206                             "Invalid parameter passed to flow_xstat devarg.\n");
5207                 return -EINVAL;
5208         }
5209
5210         flow_xstat = strtoul(value, &end, 10);
5211         if (end == NULL || *end != '\0' ||
5212             (flow_xstat == ULONG_MAX && errno == ERANGE)) {
5213                 PMD_DRV_LOG(ERR,
5214                             "Invalid parameter passed to flow_xstat devarg.\n");
5215                 return -EINVAL;
5216         }
5217
5218         if (BNXT_DEVARG_FLOW_XSTAT_INVALID(flow_xstat)) {
5219                 PMD_DRV_LOG(ERR,
5220                             "Invalid value passed to flow_xstat devarg.\n");
5221                 return -EINVAL;
5222         }
5223
5224         bp->flags |= BNXT_FLAG_FLOW_XSTATS_EN;
5225         if (BNXT_FLOW_XSTATS_EN(bp))
5226                 PMD_DRV_LOG(INFO, "flow_xstat feature enabled.\n");
5227
5228         return 0;
5229 }
5230
5231 static int
5232 bnxt_parse_devarg_max_num_kflows(__rte_unused const char *key,
5233                                         const char *value, void *opaque_arg)
5234 {
5235         struct bnxt *bp = opaque_arg;
5236         unsigned long max_num_kflows;
5237         char *end = NULL;
5238
5239         if (!value || !opaque_arg) {
5240                 PMD_DRV_LOG(ERR,
5241                         "Invalid parameter passed to max_num_kflows devarg.\n");
5242                 return -EINVAL;
5243         }
5244
5245         max_num_kflows = strtoul(value, &end, 10);
5246         if (end == NULL || *end != '\0' ||
5247                 (max_num_kflows == ULONG_MAX && errno == ERANGE)) {
5248                 PMD_DRV_LOG(ERR,
5249                         "Invalid parameter passed to max_num_kflows devarg.\n");
5250                 return -EINVAL;
5251         }
5252
5253         if (bnxt_devarg_max_num_kflow_invalid(max_num_kflows)) {
5254                 PMD_DRV_LOG(ERR,
5255                         "Invalid value passed to max_num_kflows devarg.\n");
5256                 return -EINVAL;
5257         }
5258
5259         bp->max_num_kflows = max_num_kflows;
5260         if (bp->max_num_kflows)
5261                 PMD_DRV_LOG(INFO, "max_num_kflows set as %ldK.\n",
5262                                 max_num_kflows);
5263
5264         return 0;
5265 }
5266
5267 static int
5268 bnxt_parse_devarg_rep_is_pf(__rte_unused const char *key,
5269                             const char *value, void *opaque_arg)
5270 {
5271         struct bnxt_representor *vfr_bp = opaque_arg;
5272         unsigned long rep_is_pf;
5273         char *end = NULL;
5274
5275         if (!value || !opaque_arg) {
5276                 PMD_DRV_LOG(ERR,
5277                             "Invalid parameter passed to rep_is_pf devargs.\n");
5278                 return -EINVAL;
5279         }
5280
5281         rep_is_pf = strtoul(value, &end, 10);
5282         if (end == NULL || *end != '\0' ||
5283             (rep_is_pf == ULONG_MAX && errno == ERANGE)) {
5284                 PMD_DRV_LOG(ERR,
5285                             "Invalid parameter passed to rep_is_pf devargs.\n");
5286                 return -EINVAL;
5287         }
5288
5289         if (BNXT_DEVARG_REP_IS_PF_INVALID(rep_is_pf)) {
5290                 PMD_DRV_LOG(ERR,
5291                             "Invalid value passed to rep_is_pf devargs.\n");
5292                 return -EINVAL;
5293         }
5294
5295         vfr_bp->flags |= rep_is_pf;
5296         if (BNXT_REP_PF(vfr_bp))
5297                 PMD_DRV_LOG(INFO, "PF representor\n");
5298         else
5299                 PMD_DRV_LOG(INFO, "VF representor\n");
5300
5301         return 0;
5302 }
5303
5304 static int
5305 bnxt_parse_devarg_rep_based_pf(__rte_unused const char *key,
5306                                const char *value, void *opaque_arg)
5307 {
5308         struct bnxt_representor *vfr_bp = opaque_arg;
5309         unsigned long rep_based_pf;
5310         char *end = NULL;
5311
5312         if (!value || !opaque_arg) {
5313                 PMD_DRV_LOG(ERR,
5314                             "Invalid parameter passed to rep_based_pf "
5315                             "devargs.\n");
5316                 return -EINVAL;
5317         }
5318
5319         rep_based_pf = strtoul(value, &end, 10);
5320         if (end == NULL || *end != '\0' ||
5321             (rep_based_pf == ULONG_MAX && errno == ERANGE)) {
5322                 PMD_DRV_LOG(ERR,
5323                             "Invalid parameter passed to rep_based_pf "
5324                             "devargs.\n");
5325                 return -EINVAL;
5326         }
5327
5328         if (BNXT_DEVARG_REP_BASED_PF_INVALID(rep_based_pf)) {
5329                 PMD_DRV_LOG(ERR,
5330                             "Invalid value passed to rep_based_pf devargs.\n");
5331                 return -EINVAL;
5332         }
5333
5334         vfr_bp->rep_based_pf = rep_based_pf;
5335         vfr_bp->flags |= BNXT_REP_BASED_PF_VALID;
5336
5337         PMD_DRV_LOG(INFO, "rep-based-pf = %d\n", vfr_bp->rep_based_pf);
5338
5339         return 0;
5340 }
5341
5342 static int
5343 bnxt_parse_devarg_rep_q_r2f(__rte_unused const char *key,
5344                             const char *value, void *opaque_arg)
5345 {
5346         struct bnxt_representor *vfr_bp = opaque_arg;
5347         unsigned long rep_q_r2f;
5348         char *end = NULL;
5349
5350         if (!value || !opaque_arg) {
5351                 PMD_DRV_LOG(ERR,
5352                             "Invalid parameter passed to rep_q_r2f "
5353                             "devargs.\n");
5354                 return -EINVAL;
5355         }
5356
5357         rep_q_r2f = strtoul(value, &end, 10);
5358         if (end == NULL || *end != '\0' ||
5359             (rep_q_r2f == ULONG_MAX && errno == ERANGE)) {
5360                 PMD_DRV_LOG(ERR,
5361                             "Invalid parameter passed to rep_q_r2f "
5362                             "devargs.\n");
5363                 return -EINVAL;
5364         }
5365
5366         if (BNXT_DEVARG_REP_Q_R2F_INVALID(rep_q_r2f)) {
5367                 PMD_DRV_LOG(ERR,
5368                             "Invalid value passed to rep_q_r2f devargs.\n");
5369                 return -EINVAL;
5370         }
5371
5372         vfr_bp->rep_q_r2f = rep_q_r2f;
5373         vfr_bp->flags |= BNXT_REP_Q_R2F_VALID;
5374         PMD_DRV_LOG(INFO, "rep-q-r2f = %d\n", vfr_bp->rep_q_r2f);
5375
5376         return 0;
5377 }
5378
5379 static int
5380 bnxt_parse_devarg_rep_q_f2r(__rte_unused const char *key,
5381                             const char *value, void *opaque_arg)
5382 {
5383         struct bnxt_representor *vfr_bp = opaque_arg;
5384         unsigned long rep_q_f2r;
5385         char *end = NULL;
5386
5387         if (!value || !opaque_arg) {
5388                 PMD_DRV_LOG(ERR,
5389                             "Invalid parameter passed to rep_q_f2r "
5390                             "devargs.\n");
5391                 return -EINVAL;
5392         }
5393
5394         rep_q_f2r = strtoul(value, &end, 10);
5395         if (end == NULL || *end != '\0' ||
5396             (rep_q_f2r == ULONG_MAX && errno == ERANGE)) {
5397                 PMD_DRV_LOG(ERR,
5398                             "Invalid parameter passed to rep_q_f2r "
5399                             "devargs.\n");
5400                 return -EINVAL;
5401         }
5402
5403         if (BNXT_DEVARG_REP_Q_F2R_INVALID(rep_q_f2r)) {
5404                 PMD_DRV_LOG(ERR,
5405                             "Invalid value passed to rep_q_f2r devargs.\n");
5406                 return -EINVAL;
5407         }
5408
5409         vfr_bp->rep_q_f2r = rep_q_f2r;
5410         vfr_bp->flags |= BNXT_REP_Q_F2R_VALID;
5411         PMD_DRV_LOG(INFO, "rep-q-f2r = %d\n", vfr_bp->rep_q_f2r);
5412
5413         return 0;
5414 }
5415
5416 static int
5417 bnxt_parse_devarg_rep_fc_r2f(__rte_unused const char *key,
5418                              const char *value, void *opaque_arg)
5419 {
5420         struct bnxt_representor *vfr_bp = opaque_arg;
5421         unsigned long rep_fc_r2f;
5422         char *end = NULL;
5423
5424         if (!value || !opaque_arg) {
5425                 PMD_DRV_LOG(ERR,
5426                             "Invalid parameter passed to rep_fc_r2f "
5427                             "devargs.\n");
5428                 return -EINVAL;
5429         }
5430
5431         rep_fc_r2f = strtoul(value, &end, 10);
5432         if (end == NULL || *end != '\0' ||
5433             (rep_fc_r2f == ULONG_MAX && errno == ERANGE)) {
5434                 PMD_DRV_LOG(ERR,
5435                             "Invalid parameter passed to rep_fc_r2f "
5436                             "devargs.\n");
5437                 return -EINVAL;
5438         }
5439
5440         if (BNXT_DEVARG_REP_FC_R2F_INVALID(rep_fc_r2f)) {
5441                 PMD_DRV_LOG(ERR,
5442                             "Invalid value passed to rep_fc_r2f devargs.\n");
5443                 return -EINVAL;
5444         }
5445
5446         vfr_bp->flags |= BNXT_REP_FC_R2F_VALID;
5447         vfr_bp->rep_fc_r2f = rep_fc_r2f;
5448         PMD_DRV_LOG(INFO, "rep-fc-r2f = %lu\n", rep_fc_r2f);
5449
5450         return 0;
5451 }
5452
5453 static int
5454 bnxt_parse_devarg_rep_fc_f2r(__rte_unused const char *key,
5455                              const char *value, void *opaque_arg)
5456 {
5457         struct bnxt_representor *vfr_bp = opaque_arg;
5458         unsigned long rep_fc_f2r;
5459         char *end = NULL;
5460
5461         if (!value || !opaque_arg) {
5462                 PMD_DRV_LOG(ERR,
5463                             "Invalid parameter passed to rep_fc_f2r "
5464                             "devargs.\n");
5465                 return -EINVAL;
5466         }
5467
5468         rep_fc_f2r = strtoul(value, &end, 10);
5469         if (end == NULL || *end != '\0' ||
5470             (rep_fc_f2r == ULONG_MAX && errno == ERANGE)) {
5471                 PMD_DRV_LOG(ERR,
5472                             "Invalid parameter passed to rep_fc_f2r "
5473                             "devargs.\n");
5474                 return -EINVAL;
5475         }
5476
5477         if (BNXT_DEVARG_REP_FC_F2R_INVALID(rep_fc_f2r)) {
5478                 PMD_DRV_LOG(ERR,
5479                             "Invalid value passed to rep_fc_f2r devargs.\n");
5480                 return -EINVAL;
5481         }
5482
5483         vfr_bp->flags |= BNXT_REP_FC_F2R_VALID;
5484         vfr_bp->rep_fc_f2r = rep_fc_f2r;
5485         PMD_DRV_LOG(INFO, "rep-fc-f2r = %lu\n", rep_fc_f2r);
5486
5487         return 0;
5488 }
5489
5490 static void
5491 bnxt_parse_dev_args(struct bnxt *bp, struct rte_devargs *devargs)
5492 {
5493         struct rte_kvargs *kvlist;
5494
5495         if (devargs == NULL)
5496                 return;
5497
5498         kvlist = rte_kvargs_parse(devargs->args, bnxt_dev_args);
5499         if (kvlist == NULL)
5500                 return;
5501
5502         /*
5503          * Handler for "truflow" devarg.
5504          * Invoked as for ex: "-a 0000:00:0d.0,host-based-truflow=1"
5505          */
5506         rte_kvargs_process(kvlist, BNXT_DEVARG_TRUFLOW,
5507                            bnxt_parse_devarg_truflow, bp);
5508
5509         /*
5510          * Handler for "flow_xstat" devarg.
5511          * Invoked as for ex: "-a 0000:00:0d.0,flow_xstat=1"
5512          */
5513         rte_kvargs_process(kvlist, BNXT_DEVARG_FLOW_XSTAT,
5514                            bnxt_parse_devarg_flow_xstat, bp);
5515
5516         /*
5517          * Handler for "max_num_kflows" devarg.
5518          * Invoked as for ex: "-a 000:00:0d.0,max_num_kflows=32"
5519          */
5520         rte_kvargs_process(kvlist, BNXT_DEVARG_MAX_NUM_KFLOWS,
5521                            bnxt_parse_devarg_max_num_kflows, bp);
5522
5523         rte_kvargs_free(kvlist);
5524 }
5525
5526 static int bnxt_alloc_switch_domain(struct bnxt *bp)
5527 {
5528         int rc = 0;
5529
5530         if (BNXT_PF(bp) || BNXT_VF_IS_TRUSTED(bp)) {
5531                 rc = rte_eth_switch_domain_alloc(&bp->switch_domain_id);
5532                 if (rc)
5533                         PMD_DRV_LOG(ERR,
5534                                     "Failed to alloc switch domain: %d\n", rc);
5535                 else
5536                         PMD_DRV_LOG(INFO,
5537                                     "Switch domain allocated %d\n",
5538                                     bp->switch_domain_id);
5539         }
5540
5541         return rc;
5542 }
5543
5544 /* Allocate and initialize various fields in bnxt struct that
5545  * need to be allocated/destroyed only once in the lifetime of the driver
5546  */
5547 static int bnxt_drv_init(struct rte_eth_dev *eth_dev)
5548 {
5549         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
5550         struct bnxt *bp = eth_dev->data->dev_private;
5551         int rc = 0;
5552
5553         bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
5554
5555         if (bnxt_vf_pciid(pci_dev->id.device_id))
5556                 bp->flags |= BNXT_FLAG_VF;
5557
5558         if (bnxt_p5_device(pci_dev->id.device_id))
5559                 bp->flags |= BNXT_FLAG_CHIP_P5;
5560
5561         if (pci_dev->id.device_id == BROADCOM_DEV_ID_58802 ||
5562             pci_dev->id.device_id == BROADCOM_DEV_ID_58804 ||
5563             pci_dev->id.device_id == BROADCOM_DEV_ID_58808 ||
5564             pci_dev->id.device_id == BROADCOM_DEV_ID_58802_VF)
5565                 bp->flags |= BNXT_FLAG_STINGRAY;
5566
5567         if (BNXT_TRUFLOW_EN(bp)) {
5568                 /* extra mbuf field is required to store CFA code from mark */
5569                 static const struct rte_mbuf_dynfield bnxt_cfa_code_dynfield_desc = {
5570                         .name = RTE_PMD_BNXT_CFA_CODE_DYNFIELD_NAME,
5571                         .size = sizeof(bnxt_cfa_code_dynfield_t),
5572                         .align = __alignof__(bnxt_cfa_code_dynfield_t),
5573                 };
5574                 bnxt_cfa_code_dynfield_offset =
5575                         rte_mbuf_dynfield_register(&bnxt_cfa_code_dynfield_desc);
5576                 if (bnxt_cfa_code_dynfield_offset < 0) {
5577                         PMD_DRV_LOG(ERR,
5578                             "Failed to register mbuf field for TruFlow mark\n");
5579                         return -rte_errno;
5580                 }
5581         }
5582
5583         rc = bnxt_map_pci_bars(eth_dev);
5584         if (rc) {
5585                 PMD_DRV_LOG(ERR,
5586                             "Failed to initialize board rc: %x\n", rc);
5587                 return rc;
5588         }
5589
5590         rc = bnxt_alloc_pf_info(bp);
5591         if (rc)
5592                 return rc;
5593
5594         rc = bnxt_alloc_link_info(bp);
5595         if (rc)
5596                 return rc;
5597
5598         rc = bnxt_alloc_parent_info(bp);
5599         if (rc)
5600                 return rc;
5601
5602         rc = bnxt_alloc_hwrm_resources(bp);
5603         if (rc) {
5604                 PMD_DRV_LOG(ERR,
5605                             "Failed to allocate hwrm resource rc: %x\n", rc);
5606                 return rc;
5607         }
5608         rc = bnxt_alloc_leds_info(bp);
5609         if (rc)
5610                 return rc;
5611
5612         rc = bnxt_alloc_cos_queues(bp);
5613         if (rc)
5614                 return rc;
5615
5616         rc = bnxt_init_locks(bp);
5617         if (rc)
5618                 return rc;
5619
5620         rc = bnxt_alloc_switch_domain(bp);
5621         if (rc)
5622                 return rc;
5623
5624         return rc;
5625 }
5626
5627 static int
5628 bnxt_dev_init(struct rte_eth_dev *eth_dev, void *params __rte_unused)
5629 {
5630         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
5631         static int version_printed;
5632         struct bnxt *bp;
5633         int rc;
5634
5635         if (version_printed++ == 0)
5636                 PMD_DRV_LOG(INFO, "%s\n", bnxt_version);
5637
5638         eth_dev->dev_ops = &bnxt_dev_ops;
5639         eth_dev->rx_queue_count = bnxt_rx_queue_count_op;
5640         eth_dev->rx_descriptor_status = bnxt_rx_descriptor_status_op;
5641         eth_dev->tx_descriptor_status = bnxt_tx_descriptor_status_op;
5642         eth_dev->rx_pkt_burst = &bnxt_recv_pkts;
5643         eth_dev->tx_pkt_burst = &bnxt_xmit_pkts;
5644
5645         /*
5646          * For secondary processes, we don't initialise any further
5647          * as primary has already done this work.
5648          */
5649         if (rte_eal_process_type() != RTE_PROC_PRIMARY)
5650                 return 0;
5651
5652         rte_eth_copy_pci_info(eth_dev, pci_dev);
5653         eth_dev->data->dev_flags |= RTE_ETH_DEV_AUTOFILL_QUEUE_XSTATS;
5654
5655         bp = eth_dev->data->dev_private;
5656
5657         /* Parse dev arguments passed on when starting the DPDK application. */
5658         bnxt_parse_dev_args(bp, pci_dev->device.devargs);
5659
5660         rc = bnxt_drv_init(eth_dev);
5661         if (rc)
5662                 goto error_free;
5663
5664         rc = bnxt_init_resources(bp, false);
5665         if (rc)
5666                 goto error_free;
5667
5668         rc = bnxt_alloc_stats_mem(bp);
5669         if (rc)
5670                 goto error_free;
5671
5672         PMD_DRV_LOG(INFO,
5673                     DRV_MODULE_NAME "found at mem %" PRIX64 ", node addr %pM\n",
5674                     pci_dev->mem_resource[0].phys_addr,
5675                     pci_dev->mem_resource[0].addr);
5676
5677         return 0;
5678
5679 error_free:
5680         bnxt_dev_uninit(eth_dev);
5681         return rc;
5682 }
5683
5684
5685 static void bnxt_free_ctx_mem_buf(struct bnxt_ctx_mem_buf_info *ctx)
5686 {
5687         if (!ctx)
5688                 return;
5689
5690         if (ctx->va)
5691                 rte_free(ctx->va);
5692
5693         ctx->va = NULL;
5694         ctx->dma = RTE_BAD_IOVA;
5695         ctx->ctx_id = BNXT_CTX_VAL_INVAL;
5696 }
5697
5698 static void bnxt_unregister_fc_ctx_mem(struct bnxt *bp)
5699 {
5700         bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_RX,
5701                                   CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
5702                                   bp->flow_stat->rx_fc_out_tbl.ctx_id,
5703                                   bp->flow_stat->max_fc,
5704                                   false);
5705
5706         bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_TX,
5707                                   CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
5708                                   bp->flow_stat->tx_fc_out_tbl.ctx_id,
5709                                   bp->flow_stat->max_fc,
5710                                   false);
5711
5712         if (bp->flow_stat->rx_fc_in_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
5713                 bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->rx_fc_in_tbl.ctx_id);
5714         bp->flow_stat->rx_fc_in_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
5715
5716         if (bp->flow_stat->rx_fc_out_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
5717                 bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->rx_fc_out_tbl.ctx_id);
5718         bp->flow_stat->rx_fc_out_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
5719
5720         if (bp->flow_stat->tx_fc_in_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
5721                 bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->tx_fc_in_tbl.ctx_id);
5722         bp->flow_stat->tx_fc_in_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
5723
5724         if (bp->flow_stat->tx_fc_out_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
5725                 bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->tx_fc_out_tbl.ctx_id);
5726         bp->flow_stat->tx_fc_out_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
5727 }
5728
5729 static void bnxt_uninit_fc_ctx_mem(struct bnxt *bp)
5730 {
5731         bnxt_unregister_fc_ctx_mem(bp);
5732
5733         bnxt_free_ctx_mem_buf(&bp->flow_stat->rx_fc_in_tbl);
5734         bnxt_free_ctx_mem_buf(&bp->flow_stat->rx_fc_out_tbl);
5735         bnxt_free_ctx_mem_buf(&bp->flow_stat->tx_fc_in_tbl);
5736         bnxt_free_ctx_mem_buf(&bp->flow_stat->tx_fc_out_tbl);
5737 }
5738
5739 static void bnxt_uninit_ctx_mem(struct bnxt *bp)
5740 {
5741         if (BNXT_FLOW_XSTATS_EN(bp))
5742                 bnxt_uninit_fc_ctx_mem(bp);
5743 }
5744
5745 static void
5746 bnxt_free_error_recovery_info(struct bnxt *bp)
5747 {
5748         rte_free(bp->recovery_info);
5749         bp->recovery_info = NULL;
5750         bp->fw_cap &= ~BNXT_FW_CAP_ERROR_RECOVERY;
5751 }
5752
5753 static int
5754 bnxt_uninit_resources(struct bnxt *bp, bool reconfig_dev)
5755 {
5756         int rc;
5757
5758         bnxt_free_int(bp);
5759         bnxt_free_mem(bp, reconfig_dev);
5760
5761         bnxt_hwrm_func_buf_unrgtr(bp);
5762         rte_free(bp->pf->vf_req_buf);
5763
5764         rc = bnxt_hwrm_func_driver_unregister(bp, 0);
5765         bp->flags &= ~BNXT_FLAG_REGISTERED;
5766         bnxt_free_ctx_mem(bp);
5767         if (!reconfig_dev) {
5768                 bnxt_free_hwrm_resources(bp);
5769                 bnxt_free_error_recovery_info(bp);
5770         }
5771
5772         bnxt_uninit_ctx_mem(bp);
5773
5774         bnxt_free_flow_stats_info(bp);
5775         bnxt_free_rep_info(bp);
5776         rte_free(bp->ptp_cfg);
5777         bp->ptp_cfg = NULL;
5778         return rc;
5779 }
5780
5781 static int
5782 bnxt_dev_uninit(struct rte_eth_dev *eth_dev)
5783 {
5784         if (rte_eal_process_type() != RTE_PROC_PRIMARY)
5785                 return -EPERM;
5786
5787         PMD_DRV_LOG(DEBUG, "Calling Device uninit\n");
5788
5789         if (eth_dev->state != RTE_ETH_DEV_UNUSED)
5790                 bnxt_dev_close_op(eth_dev);
5791
5792         return 0;
5793 }
5794
5795 static int bnxt_pci_remove_dev_with_reps(struct rte_eth_dev *eth_dev)
5796 {
5797         struct bnxt *bp = eth_dev->data->dev_private;
5798         struct rte_eth_dev *vf_rep_eth_dev;
5799         int ret = 0, i;
5800
5801         if (!bp)
5802                 return -EINVAL;
5803
5804         for (i = 0; i < bp->num_reps; i++) {
5805                 vf_rep_eth_dev = bp->rep_info[i].vfr_eth_dev;
5806                 if (!vf_rep_eth_dev)
5807                         continue;
5808                 PMD_DRV_LOG(DEBUG, "BNXT Port:%d VFR pci remove\n",
5809                             vf_rep_eth_dev->data->port_id);
5810                 rte_eth_dev_destroy(vf_rep_eth_dev, bnxt_representor_uninit);
5811         }
5812         PMD_DRV_LOG(DEBUG, "BNXT Port:%d pci remove\n",
5813                     eth_dev->data->port_id);
5814         ret = rte_eth_dev_destroy(eth_dev, bnxt_dev_uninit);
5815
5816         return ret;
5817 }
5818
5819 static void bnxt_free_rep_info(struct bnxt *bp)
5820 {
5821         rte_free(bp->rep_info);
5822         bp->rep_info = NULL;
5823         rte_free(bp->cfa_code_map);
5824         bp->cfa_code_map = NULL;
5825 }
5826
5827 static int bnxt_init_rep_info(struct bnxt *bp)
5828 {
5829         int i = 0, rc;
5830
5831         if (bp->rep_info)
5832                 return 0;
5833
5834         bp->rep_info = rte_zmalloc("bnxt_rep_info",
5835                                    sizeof(bp->rep_info[0]) * BNXT_MAX_VF_REPS,
5836                                    0);
5837         if (!bp->rep_info) {
5838                 PMD_DRV_LOG(ERR, "Failed to alloc memory for rep info\n");
5839                 return -ENOMEM;
5840         }
5841         bp->cfa_code_map = rte_zmalloc("bnxt_cfa_code_map",
5842                                        sizeof(*bp->cfa_code_map) *
5843                                        BNXT_MAX_CFA_CODE, 0);
5844         if (!bp->cfa_code_map) {
5845                 PMD_DRV_LOG(ERR, "Failed to alloc memory for cfa_code_map\n");
5846                 bnxt_free_rep_info(bp);
5847                 return -ENOMEM;
5848         }
5849
5850         for (i = 0; i < BNXT_MAX_CFA_CODE; i++)
5851                 bp->cfa_code_map[i] = BNXT_VF_IDX_INVALID;
5852
5853         rc = pthread_mutex_init(&bp->rep_info->vfr_lock, NULL);
5854         if (rc) {
5855                 PMD_DRV_LOG(ERR, "Unable to initialize vfr_lock\n");
5856                 bnxt_free_rep_info(bp);
5857                 return rc;
5858         }
5859
5860         rc = pthread_mutex_init(&bp->rep_info->vfr_start_lock, NULL);
5861         if (rc) {
5862                 PMD_DRV_LOG(ERR, "Unable to initialize vfr_start_lock\n");
5863                 bnxt_free_rep_info(bp);
5864                 return rc;
5865         }
5866
5867         return rc;
5868 }
5869
5870 static int bnxt_rep_port_probe(struct rte_pci_device *pci_dev,
5871                                struct rte_eth_devargs *eth_da,
5872                                struct rte_eth_dev *backing_eth_dev,
5873                                const char *dev_args)
5874 {
5875         struct rte_eth_dev *vf_rep_eth_dev;
5876         char name[RTE_ETH_NAME_MAX_LEN];
5877         struct bnxt *backing_bp;
5878         uint16_t num_rep;
5879         int i, ret = 0;
5880         struct rte_kvargs *kvlist = NULL;
5881
5882         if (eth_da->type == RTE_ETH_REPRESENTOR_NONE)
5883                 return 0;
5884         if (eth_da->type != RTE_ETH_REPRESENTOR_VF) {
5885                 PMD_DRV_LOG(ERR, "unsupported representor type %d\n",
5886                             eth_da->type);
5887                 return -ENOTSUP;
5888         }
5889         num_rep = eth_da->nb_representor_ports;
5890         if (num_rep > BNXT_MAX_VF_REPS) {
5891                 PMD_DRV_LOG(ERR, "nb_representor_ports = %d > %d MAX VF REPS\n",
5892                             num_rep, BNXT_MAX_VF_REPS);
5893                 return -EINVAL;
5894         }
5895
5896         if (num_rep >= RTE_MAX_ETHPORTS) {
5897                 PMD_DRV_LOG(ERR,
5898                             "nb_representor_ports = %d > %d MAX ETHPORTS\n",
5899                             num_rep, RTE_MAX_ETHPORTS);
5900                 return -EINVAL;
5901         }
5902
5903         backing_bp = backing_eth_dev->data->dev_private;
5904
5905         if (!(BNXT_PF(backing_bp) || BNXT_VF_IS_TRUSTED(backing_bp))) {
5906                 PMD_DRV_LOG(ERR,
5907                             "Not a PF or trusted VF. No Representor support\n");
5908                 /* Returning an error is not an option.
5909                  * Applications are not handling this correctly
5910                  */
5911                 return 0;
5912         }
5913
5914         if (bnxt_init_rep_info(backing_bp))
5915                 return 0;
5916
5917         for (i = 0; i < num_rep; i++) {
5918                 struct bnxt_representor representor = {
5919                         .vf_id = eth_da->representor_ports[i],
5920                         .switch_domain_id = backing_bp->switch_domain_id,
5921                         .parent_dev = backing_eth_dev
5922                 };
5923
5924                 if (representor.vf_id >= BNXT_MAX_VF_REPS) {
5925                         PMD_DRV_LOG(ERR, "VF-Rep id %d >= %d MAX VF ID\n",
5926                                     representor.vf_id, BNXT_MAX_VF_REPS);
5927                         continue;
5928                 }
5929
5930                 /* representor port net_bdf_port */
5931                 snprintf(name, sizeof(name), "net_%s_representor_%d",
5932                          pci_dev->device.name, eth_da->representor_ports[i]);
5933
5934                 kvlist = rte_kvargs_parse(dev_args, bnxt_dev_args);
5935                 if (kvlist) {
5936                         /*
5937                          * Handler for "rep_is_pf" devarg.
5938                          * Invoked as for ex: "-a 000:00:0d.0,
5939                          * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
5940                          */
5941                         ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_IS_PF,
5942                                                  bnxt_parse_devarg_rep_is_pf,
5943                                                  (void *)&representor);
5944                         if (ret) {
5945                                 ret = -EINVAL;
5946                                 goto err;
5947                         }
5948                         /*
5949                          * Handler for "rep_based_pf" devarg.
5950                          * Invoked as for ex: "-a 000:00:0d.0,
5951                          * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
5952                          */
5953                         ret = rte_kvargs_process(kvlist,
5954                                                  BNXT_DEVARG_REP_BASED_PF,
5955                                                  bnxt_parse_devarg_rep_based_pf,
5956                                                  (void *)&representor);
5957                         if (ret) {
5958                                 ret = -EINVAL;
5959                                 goto err;
5960                         }
5961                         /*
5962                          * Handler for "rep_based_pf" devarg.
5963                          * Invoked as for ex: "-a 000:00:0d.0,
5964                          * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
5965                          */
5966                         ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_Q_R2F,
5967                                                  bnxt_parse_devarg_rep_q_r2f,
5968                                                  (void *)&representor);
5969                         if (ret) {
5970                                 ret = -EINVAL;
5971                                 goto err;
5972                         }
5973                         /*
5974                          * Handler for "rep_based_pf" devarg.
5975                          * Invoked as for ex: "-a 000:00:0d.0,
5976                          * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
5977                          */
5978                         ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_Q_F2R,
5979                                                  bnxt_parse_devarg_rep_q_f2r,
5980                                                  (void *)&representor);
5981                         if (ret) {
5982                                 ret = -EINVAL;
5983                                 goto err;
5984                         }
5985                         /*
5986                          * Handler for "rep_based_pf" devarg.
5987                          * Invoked as for ex: "-a 000:00:0d.0,
5988                          * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
5989                          */
5990                         ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_FC_R2F,
5991                                                  bnxt_parse_devarg_rep_fc_r2f,
5992                                                  (void *)&representor);
5993                         if (ret) {
5994                                 ret = -EINVAL;
5995                                 goto err;
5996                         }
5997                         /*
5998                          * Handler for "rep_based_pf" devarg.
5999                          * Invoked as for ex: "-a 000:00:0d.0,
6000                          * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6001                          */
6002                         ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_FC_F2R,
6003                                                  bnxt_parse_devarg_rep_fc_f2r,
6004                                                  (void *)&representor);
6005                         if (ret) {
6006                                 ret = -EINVAL;
6007                                 goto err;
6008                         }
6009                 }
6010
6011                 ret = rte_eth_dev_create(&pci_dev->device, name,
6012                                          sizeof(struct bnxt_representor),
6013                                          NULL, NULL,
6014                                          bnxt_representor_init,
6015                                          &representor);
6016                 if (ret) {
6017                         PMD_DRV_LOG(ERR, "failed to create bnxt vf "
6018                                     "representor %s.", name);
6019                         goto err;
6020                 }
6021
6022                 vf_rep_eth_dev = rte_eth_dev_allocated(name);
6023                 if (!vf_rep_eth_dev) {
6024                         PMD_DRV_LOG(ERR, "Failed to find the eth_dev"
6025                                     " for VF-Rep: %s.", name);
6026                         ret = -ENODEV;
6027                         goto err;
6028                 }
6029
6030                 PMD_DRV_LOG(DEBUG, "BNXT Port:%d VFR pci probe\n",
6031                             backing_eth_dev->data->port_id);
6032                 backing_bp->rep_info[representor.vf_id].vfr_eth_dev =
6033                                                          vf_rep_eth_dev;
6034                 backing_bp->num_reps++;
6035
6036         }
6037
6038         rte_kvargs_free(kvlist);
6039         return 0;
6040
6041 err:
6042         /* If num_rep > 1, then rollback already created
6043          * ports, since we'll be failing the probe anyway
6044          */
6045         if (num_rep > 1)
6046                 bnxt_pci_remove_dev_with_reps(backing_eth_dev);
6047         rte_errno = -ret;
6048         rte_kvargs_free(kvlist);
6049
6050         return ret;
6051 }
6052
6053 static int bnxt_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
6054                           struct rte_pci_device *pci_dev)
6055 {
6056         struct rte_eth_devargs eth_da = { .nb_representor_ports = 0 };
6057         struct rte_eth_dev *backing_eth_dev;
6058         uint16_t num_rep;
6059         int ret = 0;
6060
6061         if (pci_dev->device.devargs) {
6062                 ret = rte_eth_devargs_parse(pci_dev->device.devargs->args,
6063                                             &eth_da);
6064                 if (ret)
6065                         return ret;
6066         }
6067
6068         num_rep = eth_da.nb_representor_ports;
6069         PMD_DRV_LOG(DEBUG, "nb_representor_ports = %d\n",
6070                     num_rep);
6071
6072         /* We could come here after first level of probe is already invoked
6073          * as part of an application bringup(OVS-DPDK vswitchd), so first check
6074          * for already allocated eth_dev for the backing device (PF/Trusted VF)
6075          */
6076         backing_eth_dev = rte_eth_dev_allocated(pci_dev->device.name);
6077         if (backing_eth_dev == NULL) {
6078                 ret = rte_eth_dev_create(&pci_dev->device, pci_dev->device.name,
6079                                          sizeof(struct bnxt),
6080                                          eth_dev_pci_specific_init, pci_dev,
6081                                          bnxt_dev_init, NULL);
6082
6083                 if (ret || !num_rep)
6084                         return ret;
6085
6086                 backing_eth_dev = rte_eth_dev_allocated(pci_dev->device.name);
6087         }
6088         PMD_DRV_LOG(DEBUG, "BNXT Port:%d pci probe\n",
6089                     backing_eth_dev->data->port_id);
6090
6091         if (!num_rep)
6092                 return ret;
6093
6094         /* probe representor ports now */
6095         ret = bnxt_rep_port_probe(pci_dev, &eth_da, backing_eth_dev,
6096                                   pci_dev->device.devargs->args);
6097
6098         return ret;
6099 }
6100
6101 static int bnxt_pci_remove(struct rte_pci_device *pci_dev)
6102 {
6103         struct rte_eth_dev *eth_dev;
6104
6105         eth_dev = rte_eth_dev_allocated(pci_dev->device.name);
6106         if (!eth_dev)
6107                 return 0; /* Invoked typically only by OVS-DPDK, by the
6108                            * time it comes here the eth_dev is already
6109                            * deleted by rte_eth_dev_close(), so returning
6110                            * +ve value will at least help in proper cleanup
6111                            */
6112
6113         PMD_DRV_LOG(DEBUG, "BNXT Port:%d pci remove\n", eth_dev->data->port_id);
6114         if (rte_eal_process_type() == RTE_PROC_PRIMARY) {
6115                 if (eth_dev->data->dev_flags & RTE_ETH_DEV_REPRESENTOR)
6116                         return rte_eth_dev_destroy(eth_dev,
6117                                                    bnxt_representor_uninit);
6118                 else
6119                         return rte_eth_dev_destroy(eth_dev,
6120                                                    bnxt_dev_uninit);
6121         } else {
6122                 return rte_eth_dev_pci_generic_remove(pci_dev, NULL);
6123         }
6124 }
6125
6126 static struct rte_pci_driver bnxt_rte_pmd = {
6127         .id_table = bnxt_pci_id_map,
6128         .drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC |
6129                         RTE_PCI_DRV_PROBE_AGAIN, /* Needed in case of VF-REPs
6130                                                   * and OVS-DPDK
6131                                                   */
6132         .probe = bnxt_pci_probe,
6133         .remove = bnxt_pci_remove,
6134 };
6135
6136 static bool
6137 is_device_supported(struct rte_eth_dev *dev, struct rte_pci_driver *drv)
6138 {
6139         if (strcmp(dev->device->driver->name, drv->driver.name))
6140                 return false;
6141
6142         return true;
6143 }
6144
6145 bool is_bnxt_supported(struct rte_eth_dev *dev)
6146 {
6147         return is_device_supported(dev, &bnxt_rte_pmd);
6148 }
6149
6150 RTE_LOG_REGISTER(bnxt_logtype_driver, pmd.net.bnxt.driver, NOTICE);
6151 RTE_PMD_REGISTER_PCI(net_bnxt, bnxt_rte_pmd);
6152 RTE_PMD_REGISTER_PCI_TABLE(net_bnxt, bnxt_pci_id_map);
6153 RTE_PMD_REGISTER_KMOD_DEP(net_bnxt, "* igb_uio | uio_pci_generic | vfio-pci");