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_ACCUM_STATS "accum-stats"
  91 #define BNXT_DEVARG_FLOW_XSTAT  "flow-xstat"
  92 #define BNXT_DEVARG_MAX_NUM_KFLOWS  "max-num-kflows"
  93 #define BNXT_DEVARG_REPRESENTOR "representor"
  94 #define BNXT_DEVARG_REP_BASED_PF  "rep-based-pf"
  95 #define BNXT_DEVARG_REP_IS_PF  "rep-is-pf"
  96 #define BNXT_DEVARG_REP_Q_R2F  "rep-q-r2f"
  97 #define BNXT_DEVARG_REP_Q_F2R  "rep-q-f2r"
  98 #define BNXT_DEVARG_REP_FC_R2F  "rep-fc-r2f"
  99 #define BNXT_DEVARG_REP_FC_F2R  "rep-fc-f2r"
 100 #define BNXT_DEVARG_APP_ID      "app-id"
 101
 102 static const char *const bnxt_dev_args[] = {
 103         BNXT_DEVARG_REPRESENTOR,
 104         BNXT_DEVARG_ACCUM_STATS,
 105         BNXT_DEVARG_FLOW_XSTAT,
 106         BNXT_DEVARG_MAX_NUM_KFLOWS,
 107         BNXT_DEVARG_REP_BASED_PF,
 108         BNXT_DEVARG_REP_IS_PF,
 109         BNXT_DEVARG_REP_Q_R2F,
 110         BNXT_DEVARG_REP_Q_F2R,
 111         BNXT_DEVARG_REP_FC_R2F,
 112         BNXT_DEVARG_REP_FC_F2R,
 113         BNXT_DEVARG_APP_ID,
 114         NULL
 115 };
 116
 117 /*
 118  * accum-stats == false to disable flow counter accumulation
 119  * accum-stats == true to enable flow counter accumulation
 120  */
 121 #define BNXT_DEVARG_ACCUM_STATS_INVALID(accum_stats)    ((accum_stats) > 1)
 122
 123 /*
 124  * app-id = an non-negative 8-bit number
 125  */
 126 #define BNXT_DEVARG_APP_ID_INVALID(val)                 ((val) > 255)
 127
 128 /*
 129  * flow_xstat == false to disable the feature
 130  * flow_xstat == true to enable the feature
 131  */
 132 #define BNXT_DEVARG_FLOW_XSTAT_INVALID(flow_xstat)      ((flow_xstat) > 1)
 133
 134 /*
 135  * rep_is_pf == false to indicate VF representor
 136  * rep_is_pf == true to indicate PF representor
 137  */
 138 #define BNXT_DEVARG_REP_IS_PF_INVALID(rep_is_pf)        ((rep_is_pf) > 1)
 139
 140 /*
 141  * rep_based_pf == Physical index of the PF
 142  */
 143 #define BNXT_DEVARG_REP_BASED_PF_INVALID(rep_based_pf)  ((rep_based_pf) > 15)
 144 /*
 145  * rep_q_r2f == Logical COS Queue index for the rep to endpoint direction
 146  */
 147 #define BNXT_DEVARG_REP_Q_R2F_INVALID(rep_q_r2f)        ((rep_q_r2f) > 3)
 148
 149 /*
 150  * rep_q_f2r == Logical COS Queue index for the endpoint to rep direction
 151  */
 152 #define BNXT_DEVARG_REP_Q_F2R_INVALID(rep_q_f2r)        ((rep_q_f2r) > 3)
 153
 154 /*
 155  * rep_fc_r2f == Flow control for the representor to endpoint direction
 156  */
 157 #define BNXT_DEVARG_REP_FC_R2F_INVALID(rep_fc_r2f)      ((rep_fc_r2f) > 1)
 158
 159 /*
 160  * rep_fc_f2r == Flow control for the endpoint to representor direction
 161  */
 162 #define BNXT_DEVARG_REP_FC_F2R_INVALID(rep_fc_f2r)      ((rep_fc_f2r) > 1)
 163
 164 int bnxt_cfa_code_dynfield_offset = -1;
 165
 166 /*
 167  * max_num_kflows must be >= 32
 168  * and must be a power-of-2 supported value
 169  * return: 1 -> invalid
 170  *         0 -> valid
 171  */
 172 static int bnxt_devarg_max_num_kflow_invalid(uint16_t max_num_kflows)
 173 {
 174         if (max_num_kflows < 32 || !rte_is_power_of_2(max_num_kflows))
 175                 return 1;
 176         return 0;
 177 }
 178
 179 static int bnxt_vlan_offload_set_op(struct rte_eth_dev *dev, int mask);
 180 static int bnxt_dev_uninit(struct rte_eth_dev *eth_dev);
 181 static int bnxt_init_resources(struct bnxt *bp, bool reconfig_dev);
 182 static int bnxt_uninit_resources(struct bnxt *bp, bool reconfig_dev);
 183 static void bnxt_cancel_fw_health_check(struct bnxt *bp);
 184 static int bnxt_restore_vlan_filters(struct bnxt *bp);
 185 static void bnxt_dev_recover(void *arg);
 186 static void bnxt_free_error_recovery_info(struct bnxt *bp);
 187 static void bnxt_free_rep_info(struct bnxt *bp);
 188
 189 int is_bnxt_in_error(struct bnxt *bp)
 190 {
 191         if (bp->flags & BNXT_FLAG_FATAL_ERROR)
 192                 return -EIO;
 193         if (bp->flags & BNXT_FLAG_FW_RESET)
 194                 return -EBUSY;
 195
 196         return 0;
 197 }
 198
 199 /***********************/
 200
 201 /*
 202  * High level utility functions
 203  */
 204
 205 static uint16_t bnxt_rss_ctxts(const struct bnxt *bp)
 206 {
 207         unsigned int num_rss_rings = RTE_MIN(bp->rx_nr_rings,
 208                                              BNXT_RSS_TBL_SIZE_P5);
 209
 210         if (!BNXT_CHIP_P5(bp))
 211                 return 1;
 212
 213         return RTE_ALIGN_MUL_CEIL(num_rss_rings,
 214                                   BNXT_RSS_ENTRIES_PER_CTX_P5) /
 215                                   BNXT_RSS_ENTRIES_PER_CTX_P5;
 216 }
 217
 218 uint16_t bnxt_rss_hash_tbl_size(const struct bnxt *bp)
 219 {
 220         if (!BNXT_CHIP_P5(bp))
 221                 return HW_HASH_INDEX_SIZE;
 222
 223         return bnxt_rss_ctxts(bp) * BNXT_RSS_ENTRIES_PER_CTX_P5;
 224 }
 225
 226 static void bnxt_free_parent_info(struct bnxt *bp)
 227 {
 228         rte_free(bp->parent);
 229         bp->parent = NULL;
 230 }
 231
 232 static void bnxt_free_pf_info(struct bnxt *bp)
 233 {
 234         rte_free(bp->pf);
 235         bp->pf = NULL;
 236 }
 237
 238 static void bnxt_free_link_info(struct bnxt *bp)
 239 {
 240         rte_free(bp->link_info);
 241         bp->link_info = NULL;
 242 }
 243
 244 static void bnxt_free_leds_info(struct bnxt *bp)
 245 {
 246         if (BNXT_VF(bp))
 247                 return;
 248
 249         rte_free(bp->leds);
 250         bp->leds = NULL;
 251 }
 252
 253 static void bnxt_free_flow_stats_info(struct bnxt *bp)
 254 {
 255         rte_free(bp->flow_stat);
 256         bp->flow_stat = NULL;
 257 }
 258
 259 static void bnxt_free_cos_queues(struct bnxt *bp)
 260 {
 261         rte_free(bp->rx_cos_queue);
 262         bp->rx_cos_queue = NULL;
 263         rte_free(bp->tx_cos_queue);
 264         bp->tx_cos_queue = NULL;
 265 }
 266
 267 static void bnxt_free_mem(struct bnxt *bp, bool reconfig)
 268 {
 269         bnxt_free_filter_mem(bp);
 270         bnxt_free_vnic_attributes(bp);
 271         bnxt_free_vnic_mem(bp);
 272
 273         /* tx/rx rings are configured as part of *_queue_setup callbacks.
 274          * If the number of rings change across fw update,
 275          * we don't have much choice except to warn the user.
 276          */
 277         if (!reconfig) {
 278                 bnxt_free_stats(bp);
 279                 bnxt_free_tx_rings(bp);
 280                 bnxt_free_rx_rings(bp);
 281         }
 282         bnxt_free_async_cp_ring(bp);
 283         bnxt_free_rxtx_nq_ring(bp);
 284
 285         rte_free(bp->grp_info);
 286         bp->grp_info = NULL;
 287 }
 288
 289 static int bnxt_alloc_parent_info(struct bnxt *bp)
 290 {
 291         bp->parent = rte_zmalloc("bnxt_parent_info",
 292                                  sizeof(struct bnxt_parent_info), 0);
 293         if (bp->parent == NULL)
 294                 return -ENOMEM;
 295
 296         return 0;
 297 }
 298
 299 static int bnxt_alloc_pf_info(struct bnxt *bp)
 300 {
 301         bp->pf = rte_zmalloc("bnxt_pf_info", sizeof(struct bnxt_pf_info), 0);
 302         if (bp->pf == NULL)
 303                 return -ENOMEM;
 304
 305         return 0;
 306 }
 307
 308 static int bnxt_alloc_link_info(struct bnxt *bp)
 309 {
 310         bp->link_info =
 311                 rte_zmalloc("bnxt_link_info", sizeof(struct bnxt_link_info), 0);
 312         if (bp->link_info == NULL)
 313                 return -ENOMEM;
 314
 315         return 0;
 316 }
 317
 318 static int bnxt_alloc_leds_info(struct bnxt *bp)
 319 {
 320         if (BNXT_VF(bp))
 321                 return 0;
 322
 323         bp->leds = rte_zmalloc("bnxt_leds",
 324                                BNXT_MAX_LED * sizeof(struct bnxt_led_info),
 325                                0);
 326         if (bp->leds == NULL)
 327                 return -ENOMEM;
 328
 329         return 0;
 330 }
 331
 332 static int bnxt_alloc_cos_queues(struct bnxt *bp)
 333 {
 334         bp->rx_cos_queue =
 335                 rte_zmalloc("bnxt_rx_cosq",
 336                             BNXT_COS_QUEUE_COUNT *
 337                             sizeof(struct bnxt_cos_queue_info),
 338                             0);
 339         if (bp->rx_cos_queue == NULL)
 340                 return -ENOMEM;
 341
 342         bp->tx_cos_queue =
 343                 rte_zmalloc("bnxt_tx_cosq",
 344                             BNXT_COS_QUEUE_COUNT *
 345                             sizeof(struct bnxt_cos_queue_info),
 346                             0);
 347         if (bp->tx_cos_queue == NULL)
 348                 return -ENOMEM;
 349
 350         return 0;
 351 }
 352
 353 static int bnxt_alloc_flow_stats_info(struct bnxt *bp)
 354 {
 355         bp->flow_stat = rte_zmalloc("bnxt_flow_xstat",
 356                                     sizeof(struct bnxt_flow_stat_info), 0);
 357         if (bp->flow_stat == NULL)
 358                 return -ENOMEM;
 359
 360         return 0;
 361 }
 362
 363 static int bnxt_alloc_mem(struct bnxt *bp, bool reconfig)
 364 {
 365         int rc;
 366
 367         rc = bnxt_alloc_ring_grps(bp);
 368         if (rc)
 369                 goto alloc_mem_err;
 370
 371         rc = bnxt_alloc_async_ring_struct(bp);
 372         if (rc)
 373                 goto alloc_mem_err;
 374
 375         rc = bnxt_alloc_vnic_mem(bp);
 376         if (rc)
 377                 goto alloc_mem_err;
 378
 379         rc = bnxt_alloc_vnic_attributes(bp);
 380         if (rc)
 381                 goto alloc_mem_err;
 382
 383         rc = bnxt_alloc_filter_mem(bp);
 384         if (rc)
 385                 goto alloc_mem_err;
 386
 387         rc = bnxt_alloc_async_cp_ring(bp);
 388         if (rc)
 389                 goto alloc_mem_err;
 390
 391         rc = bnxt_alloc_rxtx_nq_ring(bp);
 392         if (rc)
 393                 goto alloc_mem_err;
 394
 395         if (BNXT_FLOW_XSTATS_EN(bp)) {
 396                 rc = bnxt_alloc_flow_stats_info(bp);
 397                 if (rc)
 398                         goto alloc_mem_err;
 399         }
 400
 401         return 0;
 402
 403 alloc_mem_err:
 404         bnxt_free_mem(bp, reconfig);
 405         return rc;
 406 }
 407
 408 static int bnxt_setup_one_vnic(struct bnxt *bp, uint16_t vnic_id)
 409 {
 410         struct rte_eth_conf *dev_conf = &bp->eth_dev->data->dev_conf;
 411         struct bnxt_vnic_info *vnic = &bp->vnic_info[vnic_id];
 412         uint64_t rx_offloads = dev_conf->rxmode.offloads;
 413         struct bnxt_rx_queue *rxq;
 414         unsigned int j;
 415         int rc;
 416
 417         rc = bnxt_vnic_grp_alloc(bp, vnic);
 418         if (rc)
 419                 goto err_out;
 420
 421         PMD_DRV_LOG(DEBUG, "vnic[%d] = %p vnic->fw_grp_ids = %p\n",
 422                     vnic_id, vnic, vnic->fw_grp_ids);
 423
 424         rc = bnxt_hwrm_vnic_alloc(bp, vnic);
 425         if (rc)
 426                 goto err_out;
 427
 428         /* Alloc RSS context only if RSS mode is enabled */
 429         if (dev_conf->rxmode.mq_mode & ETH_MQ_RX_RSS) {
 430                 int j, nr_ctxs = bnxt_rss_ctxts(bp);
 431
 432                 /* RSS table size in Thor is 512.
 433                  * Cap max Rx rings to same value
 434                  */
 435                 if (bp->rx_nr_rings > BNXT_RSS_TBL_SIZE_P5) {
 436                         PMD_DRV_LOG(ERR, "RxQ cnt %d > reta_size %d\n",
 437                                     bp->rx_nr_rings, BNXT_RSS_TBL_SIZE_P5);
 438                         goto err_out;
 439                 }
 440
 441                 rc = 0;
 442                 for (j = 0; j < nr_ctxs; j++) {
 443                         rc = bnxt_hwrm_vnic_ctx_alloc(bp, vnic, j);
 444                         if (rc)
 445                                 break;
 446                 }
 447                 if (rc) {
 448                         PMD_DRV_LOG(ERR,
 449                                     "HWRM vnic %d ctx %d alloc failure rc: %x\n",
 450                                     vnic_id, j, rc);
 451                         goto err_out;
 452                 }
 453                 vnic->num_lb_ctxts = nr_ctxs;
 454         }
 455
 456         /*
 457          * Firmware sets pf pair in default vnic cfg. If the VLAN strip
 458          * setting is not available at this time, it will not be
 459          * configured correctly in the CFA.
 460          */
 461         if (rx_offloads & DEV_RX_OFFLOAD_VLAN_STRIP)
 462                 vnic->vlan_strip = true;
 463         else
 464                 vnic->vlan_strip = false;
 465
 466         rc = bnxt_hwrm_vnic_cfg(bp, vnic);
 467         if (rc)
 468                 goto err_out;
 469
 470         rc = bnxt_set_hwrm_vnic_filters(bp, vnic);
 471         if (rc)
 472                 goto err_out;
 473
 474         for (j = 0; j < bp->rx_num_qs_per_vnic; j++) {
 475                 rxq = bp->eth_dev->data->rx_queues[j];
 476
 477                 PMD_DRV_LOG(DEBUG,
 478                             "rxq[%d]->vnic=%p vnic->fw_grp_ids=%p\n",
 479                             j, rxq->vnic, rxq->vnic->fw_grp_ids);
 480
 481                 if (BNXT_HAS_RING_GRPS(bp) && rxq->rx_deferred_start)
 482                         rxq->vnic->fw_grp_ids[j] = INVALID_HW_RING_ID;
 483                 else
 484                         vnic->rx_queue_cnt++;
 485         }
 486
 487         PMD_DRV_LOG(DEBUG, "vnic->rx_queue_cnt = %d\n", vnic->rx_queue_cnt);
 488
 489         rc = bnxt_vnic_rss_configure(bp, vnic);
 490         if (rc)
 491                 goto err_out;
 492
 493         bnxt_hwrm_vnic_plcmode_cfg(bp, vnic);
 494
 495         rc = bnxt_hwrm_vnic_tpa_cfg(bp, vnic,
 496                                     (rx_offloads & DEV_RX_OFFLOAD_TCP_LRO) ?
 497                                     true : false);
 498         if (rc)
 499                 goto err_out;
 500
 501         return 0;
 502 err_out:
 503         PMD_DRV_LOG(ERR, "HWRM vnic %d cfg failure rc: %x\n",
 504                     vnic_id, rc);
 505         return rc;
 506 }
 507
 508 static int bnxt_register_fc_ctx_mem(struct bnxt *bp)
 509 {
 510         int rc = 0;
 511
 512         rc = bnxt_hwrm_ctx_rgtr(bp, bp->flow_stat->rx_fc_in_tbl.dma,
 513                                 &bp->flow_stat->rx_fc_in_tbl.ctx_id);
 514         if (rc)
 515                 return rc;
 516
 517         PMD_DRV_LOG(DEBUG,
 518                     "rx_fc_in_tbl.va = %p rx_fc_in_tbl.dma = %p"
 519                     " rx_fc_in_tbl.ctx_id = %d\n",
 520                     bp->flow_stat->rx_fc_in_tbl.va,
 521                     (void *)((uintptr_t)bp->flow_stat->rx_fc_in_tbl.dma),
 522                     bp->flow_stat->rx_fc_in_tbl.ctx_id);
 523
 524         rc = bnxt_hwrm_ctx_rgtr(bp, bp->flow_stat->rx_fc_out_tbl.dma,
 525                                 &bp->flow_stat->rx_fc_out_tbl.ctx_id);
 526         if (rc)
 527                 return rc;
 528
 529         PMD_DRV_LOG(DEBUG,
 530                     "rx_fc_out_tbl.va = %p rx_fc_out_tbl.dma = %p"
 531                     " rx_fc_out_tbl.ctx_id = %d\n",
 532                     bp->flow_stat->rx_fc_out_tbl.va,
 533                     (void *)((uintptr_t)bp->flow_stat->rx_fc_out_tbl.dma),
 534                     bp->flow_stat->rx_fc_out_tbl.ctx_id);
 535
 536         rc = bnxt_hwrm_ctx_rgtr(bp, bp->flow_stat->tx_fc_in_tbl.dma,
 537                                 &bp->flow_stat->tx_fc_in_tbl.ctx_id);
 538         if (rc)
 539                 return rc;
 540
 541         PMD_DRV_LOG(DEBUG,
 542                     "tx_fc_in_tbl.va = %p tx_fc_in_tbl.dma = %p"
 543                     " tx_fc_in_tbl.ctx_id = %d\n",
 544                     bp->flow_stat->tx_fc_in_tbl.va,
 545                     (void *)((uintptr_t)bp->flow_stat->tx_fc_in_tbl.dma),
 546                     bp->flow_stat->tx_fc_in_tbl.ctx_id);
 547
 548         rc = bnxt_hwrm_ctx_rgtr(bp, bp->flow_stat->tx_fc_out_tbl.dma,
 549                                 &bp->flow_stat->tx_fc_out_tbl.ctx_id);
 550         if (rc)
 551                 return rc;
 552
 553         PMD_DRV_LOG(DEBUG,
 554                     "tx_fc_out_tbl.va = %p tx_fc_out_tbl.dma = %p"
 555                     " tx_fc_out_tbl.ctx_id = %d\n",
 556                     bp->flow_stat->tx_fc_out_tbl.va,
 557                     (void *)((uintptr_t)bp->flow_stat->tx_fc_out_tbl.dma),
 558                     bp->flow_stat->tx_fc_out_tbl.ctx_id);
 559
 560         memset(bp->flow_stat->rx_fc_out_tbl.va,
 561                0,
 562                bp->flow_stat->rx_fc_out_tbl.size);
 563         rc = bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_RX,
 564                                        CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
 565                                        bp->flow_stat->rx_fc_out_tbl.ctx_id,
 566                                        bp->flow_stat->max_fc,
 567                                        true);
 568         if (rc)
 569                 return rc;
 570
 571         memset(bp->flow_stat->tx_fc_out_tbl.va,
 572                0,
 573                bp->flow_stat->tx_fc_out_tbl.size);
 574         rc = bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_TX,
 575                                        CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
 576                                        bp->flow_stat->tx_fc_out_tbl.ctx_id,
 577                                        bp->flow_stat->max_fc,
 578                                        true);
 579
 580         return rc;
 581 }
 582
 583 static int bnxt_alloc_ctx_mem_buf(struct bnxt *bp, char *type, size_t size,
 584                                   struct bnxt_ctx_mem_buf_info *ctx)
 585 {
 586         if (!ctx)
 587                 return -EINVAL;
 588
 589         ctx->va = rte_zmalloc_socket(type, size, 0,
 590                                      bp->eth_dev->device->numa_node);
 591         if (ctx->va == NULL)
 592                 return -ENOMEM;
 593         rte_mem_lock_page(ctx->va);
 594         ctx->size = size;
 595         ctx->dma = rte_mem_virt2iova(ctx->va);
 596         if (ctx->dma == RTE_BAD_IOVA)
 597                 return -ENOMEM;
 598
 599         return 0;
 600 }
 601
 602 static int bnxt_init_fc_ctx_mem(struct bnxt *bp)
 603 {
 604         struct rte_pci_device *pdev = bp->pdev;
 605         char type[RTE_MEMZONE_NAMESIZE];
 606         uint16_t max_fc;
 607         int rc = 0;
 608
 609         max_fc = bp->flow_stat->max_fc;
 610
 611         sprintf(type, "bnxt_rx_fc_in_" PCI_PRI_FMT, pdev->addr.domain,
 612                 pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
 613         /* 4 bytes for each counter-id */
 614         rc = bnxt_alloc_ctx_mem_buf(bp, type,
 615                                     max_fc * 4,
 616                                     &bp->flow_stat->rx_fc_in_tbl);
 617         if (rc)
 618                 return rc;
 619
 620         sprintf(type, "bnxt_rx_fc_out_" PCI_PRI_FMT, pdev->addr.domain,
 621                 pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
 622         /* 16 bytes for each counter - 8 bytes pkt_count, 8 bytes byte_count */
 623         rc = bnxt_alloc_ctx_mem_buf(bp, type,
 624                                     max_fc * 16,
 625                                     &bp->flow_stat->rx_fc_out_tbl);
 626         if (rc)
 627                 return rc;
 628
 629         sprintf(type, "bnxt_tx_fc_in_" PCI_PRI_FMT, pdev->addr.domain,
 630                 pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
 631         /* 4 bytes for each counter-id */
 632         rc = bnxt_alloc_ctx_mem_buf(bp, type,
 633                                     max_fc * 4,
 634                                     &bp->flow_stat->tx_fc_in_tbl);
 635         if (rc)
 636                 return rc;
 637
 638         sprintf(type, "bnxt_tx_fc_out_" PCI_PRI_FMT, pdev->addr.domain,
 639                 pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
 640         /* 16 bytes for each counter - 8 bytes pkt_count, 8 bytes byte_count */
 641         rc = bnxt_alloc_ctx_mem_buf(bp, type,
 642                                     max_fc * 16,
 643                                     &bp->flow_stat->tx_fc_out_tbl);
 644         if (rc)
 645                 return rc;
 646
 647         rc = bnxt_register_fc_ctx_mem(bp);
 648
 649         return rc;
 650 }
 651
 652 static int bnxt_init_ctx_mem(struct bnxt *bp)
 653 {
 654         int rc = 0;
 655
 656         if (!(bp->fw_cap & BNXT_FW_CAP_ADV_FLOW_COUNTERS) ||
 657             !(BNXT_PF(bp) || BNXT_VF_IS_TRUSTED(bp)) ||
 658             !BNXT_FLOW_XSTATS_EN(bp))
 659                 return 0;
 660
 661         rc = bnxt_hwrm_cfa_counter_qcaps(bp, &bp->flow_stat->max_fc);
 662         if (rc)
 663                 return rc;
 664
 665         rc = bnxt_init_fc_ctx_mem(bp);
 666
 667         return rc;
 668 }
 669
 670 static int bnxt_update_phy_setting(struct bnxt *bp)
 671 {
 672         struct rte_eth_link new;
 673         int rc;
 674
 675         rc = bnxt_get_hwrm_link_config(bp, &new);
 676         if (rc) {
 677                 PMD_DRV_LOG(ERR, "Failed to get link settings\n");
 678                 return rc;
 679         }
 680
 681         /*
 682          * On BCM957508-N2100 adapters, FW will not allow any user other
 683          * than BMC to shutdown the port. bnxt_get_hwrm_link_config() call
 684          * always returns link up. Force phy update always in that case.
 685          */
 686         if (!new.link_status || IS_BNXT_DEV_957508_N2100(bp)) {
 687                 rc = bnxt_set_hwrm_link_config(bp, true);
 688                 if (rc) {
 689                         PMD_DRV_LOG(ERR, "Failed to update PHY settings\n");
 690                         return rc;
 691                 }
 692         }
 693
 694         return rc;
 695 }
 696
 697 static void bnxt_free_prev_ring_stats(struct bnxt *bp)
 698 {
 699         rte_free(bp->prev_rx_ring_stats);
 700         rte_free(bp->prev_tx_ring_stats);
 701
 702         bp->prev_rx_ring_stats = NULL;
 703         bp->prev_tx_ring_stats = NULL;
 704 }
 705
 706 static int bnxt_alloc_prev_ring_stats(struct bnxt *bp)
 707 {
 708         bp->prev_rx_ring_stats =  rte_zmalloc("bnxt_prev_rx_ring_stats",
 709                                               sizeof(struct bnxt_ring_stats) *
 710                                               bp->rx_cp_nr_rings,
 711                                               0);
 712         if (bp->prev_rx_ring_stats == NULL)
 713                 return -ENOMEM;
 714
 715         bp->prev_tx_ring_stats = rte_zmalloc("bnxt_prev_tx_ring_stats",
 716                                              sizeof(struct bnxt_ring_stats) *
 717                                              bp->tx_cp_nr_rings,
 718                                              0);
 719         if (bp->prev_tx_ring_stats == NULL)
 720                 goto error;
 721
 722         return 0;
 723
 724 error:
 725         bnxt_free_prev_ring_stats(bp);
 726         return -ENOMEM;
 727 }
 728
 729 static int bnxt_start_nic(struct bnxt *bp)
 730 {
 731         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(bp->eth_dev);
 732         struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
 733         uint32_t intr_vector = 0;
 734         uint32_t queue_id, base = BNXT_MISC_VEC_ID;
 735         uint32_t vec = BNXT_MISC_VEC_ID;
 736         unsigned int i, j;
 737         int rc;
 738
 739         if (bp->eth_dev->data->mtu > RTE_ETHER_MTU) {
 740                 bp->eth_dev->data->dev_conf.rxmode.offloads |=
 741                         DEV_RX_OFFLOAD_JUMBO_FRAME;
 742                 bp->flags |= BNXT_FLAG_JUMBO;
 743         } else {
 744                 bp->eth_dev->data->dev_conf.rxmode.offloads &=
 745                         ~DEV_RX_OFFLOAD_JUMBO_FRAME;
 746                 bp->flags &= ~BNXT_FLAG_JUMBO;
 747         }
 748
 749         /* THOR does not support ring groups.
 750          * But we will use the array to save RSS context IDs.
 751          */
 752         if (BNXT_CHIP_P5(bp))
 753                 bp->max_ring_grps = BNXT_MAX_RSS_CTXTS_P5;
 754
 755         rc = bnxt_alloc_all_hwrm_stat_ctxs(bp);
 756         if (rc) {
 757                 PMD_DRV_LOG(ERR, "HWRM stat ctx alloc failure rc: %x\n", rc);
 758                 goto err_out;
 759         }
 760
 761         rc = bnxt_alloc_hwrm_rings(bp);
 762         if (rc) {
 763                 PMD_DRV_LOG(ERR, "HWRM ring alloc failure rc: %x\n", rc);
 764                 goto err_out;
 765         }
 766
 767         rc = bnxt_alloc_all_hwrm_ring_grps(bp);
 768         if (rc) {
 769                 PMD_DRV_LOG(ERR, "HWRM ring grp alloc failure: %x\n", rc);
 770                 goto err_out;
 771         }
 772
 773         if (!(bp->vnic_cap_flags & BNXT_VNIC_CAP_COS_CLASSIFY))
 774                 goto skip_cosq_cfg;
 775
 776         for (j = 0, i = 0; i < BNXT_COS_QUEUE_COUNT; i++) {
 777                 if (bp->rx_cos_queue[i].id != 0xff) {
 778                         struct bnxt_vnic_info *vnic = &bp->vnic_info[j++];
 779
 780                         if (!vnic) {
 781                                 PMD_DRV_LOG(ERR,
 782                                             "Num pools more than FW profile\n");
 783                                 rc = -EINVAL;
 784                                 goto err_out;
 785                         }
 786                         vnic->cos_queue_id = bp->rx_cos_queue[i].id;
 787                         bp->rx_cosq_cnt++;
 788                 }
 789         }
 790
 791 skip_cosq_cfg:
 792         rc = bnxt_mq_rx_configure(bp);
 793         if (rc) {
 794                 PMD_DRV_LOG(ERR, "MQ mode configure failure rc: %x\n", rc);
 795                 goto err_out;
 796         }
 797
 798         /* default vnic 0 */
 799         rc = bnxt_setup_one_vnic(bp, 0);
 800         if (rc)
 801                 goto err_out;
 802         /* VNIC configuration */
 803         if (BNXT_RFS_NEEDS_VNIC(bp)) {
 804                 for (i = 1; i < bp->nr_vnics; i++) {
 805                         rc = bnxt_setup_one_vnic(bp, i);
 806                         if (rc)
 807                                 goto err_out;
 808                 }
 809         }
 810
 811         rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, &bp->vnic_info[0], 0, NULL);
 812         if (rc) {
 813                 PMD_DRV_LOG(ERR,
 814                         "HWRM cfa l2 rx mask failure rc: %x\n", rc);
 815                 goto err_out;
 816         }
 817
 818         /* check and configure queue intr-vector mapping */
 819         if ((rte_intr_cap_multiple(intr_handle) ||
 820              !RTE_ETH_DEV_SRIOV(bp->eth_dev).active) &&
 821             bp->eth_dev->data->dev_conf.intr_conf.rxq != 0) {
 822                 intr_vector = bp->eth_dev->data->nb_rx_queues;
 823                 PMD_DRV_LOG(DEBUG, "intr_vector = %d\n", intr_vector);
 824                 if (intr_vector > bp->rx_cp_nr_rings) {
 825                         PMD_DRV_LOG(ERR, "At most %d intr queues supported",
 826                                         bp->rx_cp_nr_rings);
 827                         return -ENOTSUP;
 828                 }
 829                 rc = rte_intr_efd_enable(intr_handle, intr_vector);
 830                 if (rc)
 831                         return rc;
 832         }
 833
 834         if (rte_intr_dp_is_en(intr_handle) && !intr_handle->intr_vec) {
 835                 intr_handle->intr_vec =
 836                         rte_zmalloc("intr_vec",
 837                                     bp->eth_dev->data->nb_rx_queues *
 838                                     sizeof(int), 0);
 839                 if (intr_handle->intr_vec == NULL) {
 840                         PMD_DRV_LOG(ERR, "Failed to allocate %d rx_queues"
 841                                 " intr_vec", bp->eth_dev->data->nb_rx_queues);
 842                         rc = -ENOMEM;
 843                         goto err_out;
 844                 }
 845                 PMD_DRV_LOG(DEBUG, "intr_handle->intr_vec = %p "
 846                         "intr_handle->nb_efd = %d intr_handle->max_intr = %d\n",
 847                          intr_handle->intr_vec, intr_handle->nb_efd,
 848                         intr_handle->max_intr);
 849                 for (queue_id = 0; queue_id < bp->eth_dev->data->nb_rx_queues;
 850                      queue_id++) {
 851                         intr_handle->intr_vec[queue_id] =
 852                                                         vec + BNXT_RX_VEC_START;
 853                         if (vec < base + intr_handle->nb_efd - 1)
 854                                 vec++;
 855                 }
 856         }
 857
 858         /* enable uio/vfio intr/eventfd mapping */
 859         rc = rte_intr_enable(intr_handle);
 860 #ifndef RTE_EXEC_ENV_FREEBSD
 861         /* In FreeBSD OS, nic_uio driver does not support interrupts */
 862         if (rc)
 863                 goto err_out;
 864 #endif
 865
 866         rc = bnxt_update_phy_setting(bp);
 867         if (rc)
 868                 goto err_out;
 869
 870         bp->mark_table = rte_zmalloc("bnxt_mark_table", BNXT_MARK_TABLE_SZ, 0);
 871         if (!bp->mark_table)
 872                 PMD_DRV_LOG(ERR, "Allocation of mark table failed\n");
 873
 874         return 0;
 875
 876 err_out:
 877         /* Some of the error status returned by FW may not be from errno.h */
 878         if (rc > 0)
 879                 rc = -EIO;
 880
 881         return rc;
 882 }
 883
 884 static int bnxt_shutdown_nic(struct bnxt *bp)
 885 {
 886         bnxt_free_all_hwrm_resources(bp);
 887         bnxt_free_all_filters(bp);
 888         bnxt_free_all_vnics(bp);
 889         return 0;
 890 }
 891
 892 /*
 893  * Device configuration and status function
 894  */
 895
 896 uint32_t bnxt_get_speed_capabilities(struct bnxt *bp)
 897 {
 898         uint32_t link_speed = 0;
 899         uint32_t speed_capa = 0;
 900
 901         if (bp->link_info == NULL)
 902                 return 0;
 903
 904         link_speed = bp->link_info->support_speeds;
 905
 906         /* If PAM4 is configured, use PAM4 supported speed */
 907         if (link_speed == 0 && bp->link_info->support_pam4_speeds > 0)
 908                 link_speed = bp->link_info->support_pam4_speeds;
 909
 910         if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_LINK_SPEED_100MB)
 911                 speed_capa |= ETH_LINK_SPEED_100M;
 912         if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_100MBHD)
 913                 speed_capa |= ETH_LINK_SPEED_100M_HD;
 914         if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_1GB)
 915                 speed_capa |= ETH_LINK_SPEED_1G;
 916         if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_2_5GB)
 917                 speed_capa |= ETH_LINK_SPEED_2_5G;
 918         if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_10GB)
 919                 speed_capa |= ETH_LINK_SPEED_10G;
 920         if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_20GB)
 921                 speed_capa |= ETH_LINK_SPEED_20G;
 922         if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_25GB)
 923                 speed_capa |= ETH_LINK_SPEED_25G;
 924         if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_40GB)
 925                 speed_capa |= ETH_LINK_SPEED_40G;
 926         if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_50GB)
 927                 speed_capa |= ETH_LINK_SPEED_50G;
 928         if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_100GB)
 929                 speed_capa |= ETH_LINK_SPEED_100G;
 930         if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_PAM4_SPEEDS_50G)
 931                 speed_capa |= ETH_LINK_SPEED_50G;
 932         if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_PAM4_SPEEDS_100G)
 933                 speed_capa |= ETH_LINK_SPEED_100G;
 934         if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_PAM4_SPEEDS_200G)
 935                 speed_capa |= ETH_LINK_SPEED_200G;
 936
 937         if (bp->link_info->auto_mode ==
 938             HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_MODE_NONE)
 939                 speed_capa |= ETH_LINK_SPEED_FIXED;
 940
 941         return speed_capa;
 942 }
 943
 944 static int bnxt_dev_info_get_op(struct rte_eth_dev *eth_dev,
 945                                 struct rte_eth_dev_info *dev_info)
 946 {
 947         struct rte_pci_device *pdev = RTE_DEV_TO_PCI(eth_dev->device);
 948         struct bnxt *bp = eth_dev->data->dev_private;
 949         uint16_t max_vnics, i, j, vpool, vrxq;
 950         unsigned int max_rx_rings;
 951         int rc;
 952
 953         rc = is_bnxt_in_error(bp);
 954         if (rc)
 955                 return rc;
 956
 957         /* MAC Specifics */
 958         dev_info->max_mac_addrs = bp->max_l2_ctx;
 959         dev_info->max_hash_mac_addrs = 0;
 960
 961         /* PF/VF specifics */
 962         if (BNXT_PF(bp))
 963                 dev_info->max_vfs = pdev->max_vfs;
 964
 965         max_rx_rings = bnxt_max_rings(bp);
 966         /* For the sake of symmetry, max_rx_queues = max_tx_queues */
 967         dev_info->max_rx_queues = max_rx_rings;
 968         dev_info->max_tx_queues = max_rx_rings;
 969         dev_info->reta_size = bnxt_rss_hash_tbl_size(bp);
 970         dev_info->hash_key_size = HW_HASH_KEY_SIZE;
 971         max_vnics = bp->max_vnics;
 972
 973         /* MTU specifics */
 974         dev_info->min_mtu = RTE_ETHER_MIN_MTU;
 975         dev_info->max_mtu = BNXT_MAX_MTU;
 976
 977         /* Fast path specifics */
 978         dev_info->min_rx_bufsize = 1;
 979         dev_info->max_rx_pktlen = BNXT_MAX_PKT_LEN;
 980
 981         dev_info->rx_offload_capa = BNXT_DEV_RX_OFFLOAD_SUPPORT;
 982         if (bp->flags & BNXT_FLAG_PTP_SUPPORTED)
 983                 dev_info->rx_offload_capa |= DEV_RX_OFFLOAD_TIMESTAMP;
 984         dev_info->tx_queue_offload_capa = DEV_TX_OFFLOAD_MBUF_FAST_FREE;
 985         dev_info->tx_offload_capa = BNXT_DEV_TX_OFFLOAD_SUPPORT |
 986                                     dev_info->tx_queue_offload_capa;
 987         dev_info->flow_type_rss_offloads = BNXT_ETH_RSS_SUPPORT;
 988
 989         dev_info->speed_capa = bnxt_get_speed_capabilities(bp);
 990
 991         dev_info->default_rxconf = (struct rte_eth_rxconf) {
 992                 .rx_thresh = {
 993                         .pthresh = 8,
 994                         .hthresh = 8,
 995                         .wthresh = 0,
 996                 },
 997                 .rx_free_thresh = 32,
 998                 .rx_drop_en = BNXT_DEFAULT_RX_DROP_EN,
 999         };
1000
1001         dev_info->default_txconf = (struct rte_eth_txconf) {
1002                 .tx_thresh = {
1003                         .pthresh = 32,
1004                         .hthresh = 0,
1005                         .wthresh = 0,
1006                 },
1007                 .tx_free_thresh = 32,
1008                 .tx_rs_thresh = 32,
1009         };
1010         eth_dev->data->dev_conf.intr_conf.lsc = 1;
1011
1012         eth_dev->data->dev_conf.intr_conf.rxq = 1;
1013         dev_info->rx_desc_lim.nb_min = BNXT_MIN_RING_DESC;
1014         dev_info->rx_desc_lim.nb_max = BNXT_MAX_RX_RING_DESC;
1015         dev_info->tx_desc_lim.nb_min = BNXT_MIN_RING_DESC;
1016         dev_info->tx_desc_lim.nb_max = BNXT_MAX_TX_RING_DESC;
1017
1018         if (BNXT_PF(bp) || BNXT_VF_IS_TRUSTED(bp)) {
1019                 dev_info->switch_info.name = eth_dev->device->name;
1020                 dev_info->switch_info.domain_id = bp->switch_domain_id;
1021                 dev_info->switch_info.port_id =
1022                                 BNXT_PF(bp) ? BNXT_SWITCH_PORT_ID_PF :
1023                                     BNXT_SWITCH_PORT_ID_TRUSTED_VF;
1024         }
1025
1026         /*
1027          * TODO: default_rxconf, default_txconf, rx_desc_lim, and tx_desc_lim
1028          *       need further investigation.
1029          */
1030
1031         /* VMDq resources */
1032         vpool = 64; /* ETH_64_POOLS */
1033         vrxq = 128; /* ETH_VMDQ_DCB_NUM_QUEUES */
1034         for (i = 0; i < 4; vpool >>= 1, i++) {
1035                 if (max_vnics > vpool) {
1036                         for (j = 0; j < 5; vrxq >>= 1, j++) {
1037                                 if (dev_info->max_rx_queues > vrxq) {
1038                                         if (vpool > vrxq)
1039                                                 vpool = vrxq;
1040                                         goto found;
1041                                 }
1042                         }
1043                         /* Not enough resources to support VMDq */
1044                         break;
1045                 }
1046         }
1047         /* Not enough resources to support VMDq */
1048         vpool = 0;
1049         vrxq = 0;
1050 found:
1051         dev_info->max_vmdq_pools = vpool;
1052         dev_info->vmdq_queue_num = vrxq;
1053
1054         dev_info->vmdq_pool_base = 0;
1055         dev_info->vmdq_queue_base = 0;
1056
1057         return 0;
1058 }
1059
1060 /* Configure the device based on the configuration provided */
1061 static int bnxt_dev_configure_op(struct rte_eth_dev *eth_dev)
1062 {
1063         struct bnxt *bp = eth_dev->data->dev_private;
1064         uint64_t rx_offloads = eth_dev->data->dev_conf.rxmode.offloads;
1065         int rc;
1066
1067         bp->rx_queues = (void *)eth_dev->data->rx_queues;
1068         bp->tx_queues = (void *)eth_dev->data->tx_queues;
1069         bp->tx_nr_rings = eth_dev->data->nb_tx_queues;
1070         bp->rx_nr_rings = eth_dev->data->nb_rx_queues;
1071
1072         rc = is_bnxt_in_error(bp);
1073         if (rc)
1074                 return rc;
1075
1076         if (BNXT_VF(bp) && (bp->flags & BNXT_FLAG_NEW_RM)) {
1077                 rc = bnxt_hwrm_check_vf_rings(bp);
1078                 if (rc) {
1079                         PMD_DRV_LOG(ERR, "HWRM insufficient resources\n");
1080                         return -ENOSPC;
1081                 }
1082
1083                 /* If a resource has already been allocated - in this case
1084                  * it is the async completion ring, free it. Reallocate it after
1085                  * resource reservation. This will ensure the resource counts
1086                  * are calculated correctly.
1087                  */
1088
1089                 pthread_mutex_lock(&bp->def_cp_lock);
1090
1091                 if (!BNXT_HAS_NQ(bp) && bp->async_cp_ring) {
1092                         bnxt_disable_int(bp);
1093                         bnxt_free_cp_ring(bp, bp->async_cp_ring);
1094                 }
1095
1096                 rc = bnxt_hwrm_func_reserve_vf_resc(bp, false);
1097                 if (rc) {
1098                         PMD_DRV_LOG(ERR, "HWRM resource alloc fail:%x\n", rc);
1099                         pthread_mutex_unlock(&bp->def_cp_lock);
1100                         return -ENOSPC;
1101                 }
1102
1103                 if (!BNXT_HAS_NQ(bp) && bp->async_cp_ring) {
1104                         rc = bnxt_alloc_async_cp_ring(bp);
1105                         if (rc) {
1106                                 pthread_mutex_unlock(&bp->def_cp_lock);
1107                                 return rc;
1108                         }
1109                         bnxt_enable_int(bp);
1110                 }
1111
1112                 pthread_mutex_unlock(&bp->def_cp_lock);
1113         }
1114
1115         /* Inherit new configurations */
1116         if (eth_dev->data->nb_rx_queues > bp->max_rx_rings ||
1117             eth_dev->data->nb_tx_queues > bp->max_tx_rings ||
1118             eth_dev->data->nb_rx_queues + eth_dev->data->nb_tx_queues
1119                 + BNXT_NUM_ASYNC_CPR(bp) > bp->max_cp_rings ||
1120             eth_dev->data->nb_rx_queues + eth_dev->data->nb_tx_queues >
1121             bp->max_stat_ctx)
1122                 goto resource_error;
1123
1124         if (BNXT_HAS_RING_GRPS(bp) &&
1125             (uint32_t)(eth_dev->data->nb_rx_queues) > bp->max_ring_grps)
1126                 goto resource_error;
1127
1128         if (!(eth_dev->data->dev_conf.rxmode.mq_mode & ETH_MQ_RX_RSS) &&
1129             bp->max_vnics < eth_dev->data->nb_rx_queues)
1130                 goto resource_error;
1131
1132         bp->rx_cp_nr_rings = bp->rx_nr_rings;
1133         bp->tx_cp_nr_rings = bp->tx_nr_rings;
1134
1135         if (eth_dev->data->dev_conf.rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG)
1136                 rx_offloads |= DEV_RX_OFFLOAD_RSS_HASH;
1137         eth_dev->data->dev_conf.rxmode.offloads = rx_offloads;
1138
1139         if (rx_offloads & DEV_RX_OFFLOAD_JUMBO_FRAME) {
1140                 eth_dev->data->mtu =
1141                         eth_dev->data->dev_conf.rxmode.max_rx_pkt_len -
1142                         RTE_ETHER_HDR_LEN - RTE_ETHER_CRC_LEN - VLAN_TAG_SIZE *
1143                         BNXT_NUM_VLANS;
1144                 bnxt_mtu_set_op(eth_dev, eth_dev->data->mtu);
1145         }
1146         return 0;
1147
1148 resource_error:
1149         PMD_DRV_LOG(ERR,
1150                     "Insufficient resources to support requested config\n");
1151         PMD_DRV_LOG(ERR,
1152                     "Num Queues Requested: Tx %d, Rx %d\n",
1153                     eth_dev->data->nb_tx_queues,
1154                     eth_dev->data->nb_rx_queues);
1155         PMD_DRV_LOG(ERR,
1156                     "MAX: TxQ %d, RxQ %d, CQ %d Stat %d, Grp %d, Vnic %d\n",
1157                     bp->max_tx_rings, bp->max_rx_rings, bp->max_cp_rings,
1158                     bp->max_stat_ctx, bp->max_ring_grps, bp->max_vnics);
1159         return -ENOSPC;
1160 }
1161
1162 void bnxt_print_link_info(struct rte_eth_dev *eth_dev)
1163 {
1164         struct rte_eth_link *link = &eth_dev->data->dev_link;
1165
1166         if (link->link_status)
1167                 PMD_DRV_LOG(INFO, "Port %d Link Up - speed %u Mbps - %s\n",
1168                         eth_dev->data->port_id,
1169                         (uint32_t)link->link_speed,
1170                         (link->link_duplex == ETH_LINK_FULL_DUPLEX) ?
1171                         ("full-duplex") : ("half-duplex\n"));
1172         else
1173                 PMD_DRV_LOG(INFO, "Port %d Link Down\n",
1174                         eth_dev->data->port_id);
1175 }
1176
1177 /*
1178  * Determine whether the current configuration requires support for scattered
1179  * receive; return 1 if scattered receive is required and 0 if not.
1180  */
1181 static int bnxt_scattered_rx(struct rte_eth_dev *eth_dev)
1182 {
1183         uint16_t buf_size;
1184         int i;
1185
1186         if (eth_dev->data->dev_conf.rxmode.offloads & DEV_RX_OFFLOAD_SCATTER)
1187                 return 1;
1188
1189         if (eth_dev->data->dev_conf.rxmode.offloads & DEV_RX_OFFLOAD_TCP_LRO)
1190                 return 1;
1191
1192         for (i = 0; i < eth_dev->data->nb_rx_queues; i++) {
1193                 struct bnxt_rx_queue *rxq = eth_dev->data->rx_queues[i];
1194
1195                 buf_size = (uint16_t)(rte_pktmbuf_data_room_size(rxq->mb_pool) -
1196                                       RTE_PKTMBUF_HEADROOM);
1197                 if (eth_dev->data->dev_conf.rxmode.max_rx_pkt_len > buf_size)
1198                         return 1;
1199         }
1200         return 0;
1201 }
1202
1203 static eth_rx_burst_t
1204 bnxt_receive_function(struct rte_eth_dev *eth_dev)
1205 {
1206         struct bnxt *bp = eth_dev->data->dev_private;
1207
1208         /* Disable vector mode RX for Stingray2 for now */
1209         if (BNXT_CHIP_SR2(bp)) {
1210                 bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
1211                 return bnxt_recv_pkts;
1212         }
1213
1214 #if (defined(RTE_ARCH_X86) || defined(RTE_ARCH_ARM64)) && \
1215         !defined(RTE_LIBRTE_IEEE1588)
1216
1217         /* Vector mode receive cannot be enabled if scattered rx is in use. */
1218         if (eth_dev->data->scattered_rx)
1219                 goto use_scalar_rx;
1220
1221         /*
1222          * Vector mode receive cannot be enabled if Truflow is enabled or if
1223          * asynchronous completions and receive completions can be placed in
1224          * the same completion ring.
1225          */
1226         if (BNXT_TRUFLOW_EN(bp) || !BNXT_NUM_ASYNC_CPR(bp))
1227                 goto use_scalar_rx;
1228
1229         /*
1230          * Vector mode receive cannot be enabled if any receive offloads outside
1231          * a limited subset have been enabled.
1232          */
1233         if (eth_dev->data->dev_conf.rxmode.offloads &
1234                 ~(DEV_RX_OFFLOAD_VLAN_STRIP |
1235                   DEV_RX_OFFLOAD_KEEP_CRC |
1236                   DEV_RX_OFFLOAD_JUMBO_FRAME |
1237                   DEV_RX_OFFLOAD_IPV4_CKSUM |
1238                   DEV_RX_OFFLOAD_UDP_CKSUM |
1239                   DEV_RX_OFFLOAD_TCP_CKSUM |
1240                   DEV_RX_OFFLOAD_OUTER_IPV4_CKSUM |
1241                   DEV_RX_OFFLOAD_OUTER_UDP_CKSUM |
1242                   DEV_RX_OFFLOAD_RSS_HASH |
1243                   DEV_RX_OFFLOAD_VLAN_FILTER))
1244                 goto use_scalar_rx;
1245
1246 #if defined(RTE_ARCH_X86) && defined(CC_AVX2_SUPPORT)
1247         if (rte_vect_get_max_simd_bitwidth() >= RTE_VECT_SIMD_256 &&
1248             rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX2) == 1) {
1249                 PMD_DRV_LOG(INFO,
1250                             "Using AVX2 vector mode receive for port %d\n",
1251                             eth_dev->data->port_id);
1252                 bp->flags |= BNXT_FLAG_RX_VECTOR_PKT_MODE;
1253                 return bnxt_recv_pkts_vec_avx2;
1254         }
1255  #endif
1256         if (rte_vect_get_max_simd_bitwidth() >= RTE_VECT_SIMD_128) {
1257                 PMD_DRV_LOG(INFO,
1258                             "Using SSE vector mode receive for port %d\n",
1259                             eth_dev->data->port_id);
1260                 bp->flags |= BNXT_FLAG_RX_VECTOR_PKT_MODE;
1261                 return bnxt_recv_pkts_vec;
1262         }
1263
1264 use_scalar_rx:
1265         PMD_DRV_LOG(INFO, "Vector mode receive disabled for port %d\n",
1266                     eth_dev->data->port_id);
1267         PMD_DRV_LOG(INFO,
1268                     "Port %d scatter: %d rx offload: %" PRIX64 "\n",
1269                     eth_dev->data->port_id,
1270                     eth_dev->data->scattered_rx,
1271                     eth_dev->data->dev_conf.rxmode.offloads);
1272 #endif
1273         bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
1274         return bnxt_recv_pkts;
1275 }
1276
1277 static eth_tx_burst_t
1278 bnxt_transmit_function(struct rte_eth_dev *eth_dev)
1279 {
1280         struct bnxt *bp = eth_dev->data->dev_private;
1281
1282         /* Disable vector mode TX for Stingray2 for now */
1283         if (BNXT_CHIP_SR2(bp))
1284                 return bnxt_xmit_pkts;
1285
1286 #if defined(RTE_ARCH_X86) || defined(RTE_ARCH_ARM64) && \
1287         !defined(RTE_LIBRTE_IEEE1588)
1288         uint64_t offloads = eth_dev->data->dev_conf.txmode.offloads;
1289
1290         /*
1291          * Vector mode transmit can be enabled only if not using scatter rx
1292          * or tx offloads.
1293          */
1294         if (eth_dev->data->scattered_rx ||
1295             (offloads & ~DEV_TX_OFFLOAD_MBUF_FAST_FREE) ||
1296             BNXT_TRUFLOW_EN(bp))
1297                 goto use_scalar_tx;
1298
1299 #if defined(RTE_ARCH_X86) && defined(CC_AVX2_SUPPORT)
1300         if (rte_vect_get_max_simd_bitwidth() >= RTE_VECT_SIMD_256 &&
1301             rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX2) == 1) {
1302                 PMD_DRV_LOG(INFO,
1303                             "Using AVX2 vector mode transmit for port %d\n",
1304                             eth_dev->data->port_id);
1305                 return bnxt_xmit_pkts_vec_avx2;
1306         }
1307 #endif
1308         if (rte_vect_get_max_simd_bitwidth() >= RTE_VECT_SIMD_128) {
1309                 PMD_DRV_LOG(INFO,
1310                             "Using SSE vector mode transmit for port %d\n",
1311                             eth_dev->data->port_id);
1312                 return bnxt_xmit_pkts_vec;
1313         }
1314
1315 use_scalar_tx:
1316         PMD_DRV_LOG(INFO, "Vector mode transmit disabled for port %d\n",
1317                     eth_dev->data->port_id);
1318         PMD_DRV_LOG(INFO,
1319                     "Port %d scatter: %d tx offload: %" PRIX64 "\n",
1320                     eth_dev->data->port_id,
1321                     eth_dev->data->scattered_rx,
1322                     offloads);
1323 #endif
1324         return bnxt_xmit_pkts;
1325 }
1326
1327 static int bnxt_handle_if_change_status(struct bnxt *bp)
1328 {
1329         int rc;
1330
1331         /* Since fw has undergone a reset and lost all contexts,
1332          * set fatal flag to not issue hwrm during cleanup
1333          */
1334         bp->flags |= BNXT_FLAG_FATAL_ERROR;
1335         bnxt_uninit_resources(bp, true);
1336
1337         /* clear fatal flag so that re-init happens */
1338         bp->flags &= ~BNXT_FLAG_FATAL_ERROR;
1339         rc = bnxt_init_resources(bp, true);
1340
1341         bp->flags &= ~BNXT_FLAG_IF_CHANGE_HOT_FW_RESET_DONE;
1342
1343         return rc;
1344 }
1345
1346 static int bnxt_dev_set_link_up_op(struct rte_eth_dev *eth_dev)
1347 {
1348         struct bnxt *bp = eth_dev->data->dev_private;
1349         int rc = 0;
1350
1351         if (!BNXT_SINGLE_PF(bp))
1352                 return -ENOTSUP;
1353
1354         if (!bp->link_info->link_up)
1355                 rc = bnxt_set_hwrm_link_config(bp, true);
1356         if (!rc)
1357                 eth_dev->data->dev_link.link_status = 1;
1358
1359         bnxt_print_link_info(eth_dev);
1360         return rc;
1361 }
1362
1363 static int bnxt_dev_set_link_down_op(struct rte_eth_dev *eth_dev)
1364 {
1365         struct bnxt *bp = eth_dev->data->dev_private;
1366
1367         if (!BNXT_SINGLE_PF(bp))
1368                 return -ENOTSUP;
1369
1370         eth_dev->data->dev_link.link_status = 0;
1371         bnxt_set_hwrm_link_config(bp, false);
1372         bp->link_info->link_up = 0;
1373
1374         return 0;
1375 }
1376
1377 static void bnxt_free_switch_domain(struct bnxt *bp)
1378 {
1379         int rc = 0;
1380
1381         if (!(BNXT_PF(bp) || BNXT_VF_IS_TRUSTED(bp)))
1382                 return;
1383
1384         rc = rte_eth_switch_domain_free(bp->switch_domain_id);
1385         if (rc)
1386                 PMD_DRV_LOG(ERR, "free switch domain:%d fail: %d\n",
1387                             bp->switch_domain_id, rc);
1388 }
1389
1390 static void bnxt_ptp_get_current_time(void *arg)
1391 {
1392         struct bnxt *bp = arg;
1393         struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
1394         int rc;
1395
1396         rc = is_bnxt_in_error(bp);
1397         if (rc)
1398                 return;
1399
1400         if (!ptp)
1401                 return;
1402
1403         bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_CURRENT_TIME,
1404                                 &ptp->current_time);
1405
1406         rc = rte_eal_alarm_set(US_PER_S, bnxt_ptp_get_current_time, (void *)bp);
1407         if (rc != 0) {
1408                 PMD_DRV_LOG(ERR, "Failed to re-schedule PTP alarm\n");
1409                 bp->flags2 &= ~BNXT_FLAGS2_PTP_ALARM_SCHEDULED;
1410         }
1411 }
1412
1413 static int bnxt_schedule_ptp_alarm(struct bnxt *bp)
1414 {
1415         struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
1416         int rc;
1417
1418         if (bp->flags2 & BNXT_FLAGS2_PTP_ALARM_SCHEDULED)
1419                 return 0;
1420
1421         bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_CURRENT_TIME,
1422                                 &ptp->current_time);
1423
1424         rc = rte_eal_alarm_set(US_PER_S, bnxt_ptp_get_current_time, (void *)bp);
1425         return rc;
1426 }
1427
1428 static void bnxt_cancel_ptp_alarm(struct bnxt *bp)
1429 {
1430         if (bp->flags2 & BNXT_FLAGS2_PTP_ALARM_SCHEDULED) {
1431                 rte_eal_alarm_cancel(bnxt_ptp_get_current_time, (void *)bp);
1432                 bp->flags2 &= ~BNXT_FLAGS2_PTP_ALARM_SCHEDULED;
1433         }
1434 }
1435
1436 static void bnxt_ptp_stop(struct bnxt *bp)
1437 {
1438         bnxt_cancel_ptp_alarm(bp);
1439         bp->flags2 &= ~BNXT_FLAGS2_PTP_TIMESYNC_ENABLED;
1440 }
1441
1442 static int bnxt_ptp_start(struct bnxt *bp)
1443 {
1444         int rc;
1445
1446         rc = bnxt_schedule_ptp_alarm(bp);
1447         if (rc != 0) {
1448                 PMD_DRV_LOG(ERR, "Failed to schedule PTP alarm\n");
1449         } else {
1450                 bp->flags2 |= BNXT_FLAGS2_PTP_TIMESYNC_ENABLED;
1451                 bp->flags2 |= BNXT_FLAGS2_PTP_ALARM_SCHEDULED;
1452         }
1453
1454         return rc;
1455 }
1456
1457 static int bnxt_dev_stop(struct rte_eth_dev *eth_dev)
1458 {
1459         struct bnxt *bp = eth_dev->data->dev_private;
1460         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
1461         struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
1462         struct rte_eth_link link;
1463         int ret;
1464
1465         eth_dev->data->dev_started = 0;
1466         eth_dev->data->scattered_rx = 0;
1467
1468         /* Prevent crashes when queues are still in use */
1469         eth_dev->rx_pkt_burst = &bnxt_dummy_recv_pkts;
1470         eth_dev->tx_pkt_burst = &bnxt_dummy_xmit_pkts;
1471
1472         bnxt_disable_int(bp);
1473
1474         /* disable uio/vfio intr/eventfd mapping */
1475         rte_intr_disable(intr_handle);
1476
1477         /* Stop the child representors for this device */
1478         ret = bnxt_rep_stop_all(bp);
1479         if (ret != 0)
1480                 return ret;
1481
1482         /* delete the bnxt ULP port details */
1483         bnxt_ulp_port_deinit(bp);
1484
1485         bnxt_cancel_fw_health_check(bp);
1486
1487         if (BNXT_P5_PTP_TIMESYNC_ENABLED(bp))
1488                 bnxt_cancel_ptp_alarm(bp);
1489
1490         /* Do not bring link down during reset recovery */
1491         if (!is_bnxt_in_error(bp)) {
1492                 bnxt_dev_set_link_down_op(eth_dev);
1493                 /* Wait for link to be reset */
1494                 if (BNXT_SINGLE_PF(bp))
1495                         rte_delay_ms(500);
1496                 /* clear the recorded link status */
1497                 memset(&link, 0, sizeof(link));
1498                 rte_eth_linkstatus_set(eth_dev, &link);
1499         }
1500
1501         /* Clean queue intr-vector mapping */
1502         rte_intr_efd_disable(intr_handle);
1503         if (intr_handle->intr_vec != NULL) {
1504                 rte_free(intr_handle->intr_vec);
1505                 intr_handle->intr_vec = NULL;
1506         }
1507
1508         bnxt_hwrm_port_clr_stats(bp);
1509         bnxt_free_tx_mbufs(bp);
1510         bnxt_free_rx_mbufs(bp);
1511         /* Process any remaining notifications in default completion queue */
1512         bnxt_int_handler(eth_dev);
1513         bnxt_shutdown_nic(bp);
1514         bnxt_hwrm_if_change(bp, false);
1515
1516         bnxt_free_prev_ring_stats(bp);
1517         rte_free(bp->mark_table);
1518         bp->mark_table = NULL;
1519
1520         bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
1521         bp->rx_cosq_cnt = 0;
1522         /* All filters are deleted on a port stop. */
1523         if (BNXT_FLOW_XSTATS_EN(bp))
1524                 bp->flow_stat->flow_count = 0;
1525
1526         return 0;
1527 }
1528
1529 /* Unload the driver, release resources */
1530 static int bnxt_dev_stop_op(struct rte_eth_dev *eth_dev)
1531 {
1532         struct bnxt *bp = eth_dev->data->dev_private;
1533
1534         pthread_mutex_lock(&bp->err_recovery_lock);
1535         if (bp->flags & BNXT_FLAG_FW_RESET) {
1536                 PMD_DRV_LOG(ERR,
1537                             "Adapter recovering from error..Please retry\n");
1538                 pthread_mutex_unlock(&bp->err_recovery_lock);
1539                 return -EAGAIN;
1540         }
1541         pthread_mutex_unlock(&bp->err_recovery_lock);
1542
1543         return bnxt_dev_stop(eth_dev);
1544 }
1545
1546 static int bnxt_dev_start_op(struct rte_eth_dev *eth_dev)
1547 {
1548         struct bnxt *bp = eth_dev->data->dev_private;
1549         uint64_t rx_offloads = eth_dev->data->dev_conf.rxmode.offloads;
1550         int vlan_mask = 0;
1551         int rc, retry_cnt = BNXT_IF_CHANGE_RETRY_COUNT;
1552
1553         if (!eth_dev->data->nb_tx_queues || !eth_dev->data->nb_rx_queues) {
1554                 PMD_DRV_LOG(ERR, "Queues are not configured yet!\n");
1555                 return -EINVAL;
1556         }
1557
1558         if (bp->rx_cp_nr_rings > RTE_ETHDEV_QUEUE_STAT_CNTRS)
1559                 PMD_DRV_LOG(ERR,
1560                             "RxQ cnt %d > RTE_ETHDEV_QUEUE_STAT_CNTRS %d\n",
1561                             bp->rx_cp_nr_rings, RTE_ETHDEV_QUEUE_STAT_CNTRS);
1562
1563         do {
1564                 rc = bnxt_hwrm_if_change(bp, true);
1565                 if (rc == 0 || rc != -EAGAIN)
1566                         break;
1567
1568                 rte_delay_ms(BNXT_IF_CHANGE_RETRY_INTERVAL);
1569         } while (retry_cnt--);
1570
1571         if (rc)
1572                 return rc;
1573
1574         if (bp->flags & BNXT_FLAG_IF_CHANGE_HOT_FW_RESET_DONE) {
1575                 rc = bnxt_handle_if_change_status(bp);
1576                 if (rc)
1577                         return rc;
1578         }
1579
1580         bnxt_enable_int(bp);
1581
1582         eth_dev->data->scattered_rx = bnxt_scattered_rx(eth_dev);
1583
1584         rc = bnxt_start_nic(bp);
1585         if (rc)
1586                 goto error;
1587
1588         rc = bnxt_alloc_prev_ring_stats(bp);
1589         if (rc)
1590                 goto error;
1591
1592         eth_dev->data->dev_started = 1;
1593
1594         bnxt_link_update_op(eth_dev, 1);
1595
1596         if (rx_offloads & DEV_RX_OFFLOAD_VLAN_FILTER)
1597                 vlan_mask |= ETH_VLAN_FILTER_MASK;
1598         if (rx_offloads & DEV_RX_OFFLOAD_VLAN_STRIP)
1599                 vlan_mask |= ETH_VLAN_STRIP_MASK;
1600         rc = bnxt_vlan_offload_set_op(eth_dev, vlan_mask);
1601         if (rc)
1602                 goto error;
1603
1604         /* Initialize bnxt ULP port details */
1605         rc = bnxt_ulp_port_init(bp);
1606         if (rc)
1607                 goto error;
1608
1609         eth_dev->rx_pkt_burst = bnxt_receive_function(eth_dev);
1610         eth_dev->tx_pkt_burst = bnxt_transmit_function(eth_dev);
1611
1612         bnxt_schedule_fw_health_check(bp);
1613
1614         if (BNXT_P5_PTP_TIMESYNC_ENABLED(bp))
1615                 bnxt_schedule_ptp_alarm(bp);
1616
1617         return 0;
1618
1619 error:
1620         bnxt_dev_stop(eth_dev);
1621         return rc;
1622 }
1623
1624 static void
1625 bnxt_uninit_locks(struct bnxt *bp)
1626 {
1627         pthread_mutex_destroy(&bp->flow_lock);
1628         pthread_mutex_destroy(&bp->def_cp_lock);
1629         pthread_mutex_destroy(&bp->health_check_lock);
1630         pthread_mutex_destroy(&bp->err_recovery_lock);
1631         if (bp->rep_info) {
1632                 pthread_mutex_destroy(&bp->rep_info->vfr_lock);
1633                 pthread_mutex_destroy(&bp->rep_info->vfr_start_lock);
1634         }
1635 }
1636
1637 static void bnxt_drv_uninit(struct bnxt *bp)
1638 {
1639         bnxt_free_leds_info(bp);
1640         bnxt_free_cos_queues(bp);
1641         bnxt_free_link_info(bp);
1642         bnxt_free_parent_info(bp);
1643         bnxt_uninit_locks(bp);
1644
1645         rte_memzone_free((const struct rte_memzone *)bp->tx_mem_zone);
1646         bp->tx_mem_zone = NULL;
1647         rte_memzone_free((const struct rte_memzone *)bp->rx_mem_zone);
1648         bp->rx_mem_zone = NULL;
1649
1650         bnxt_free_vf_info(bp);
1651         bnxt_free_pf_info(bp);
1652
1653         rte_free(bp->grp_info);
1654         bp->grp_info = NULL;
1655 }
1656
1657 static int bnxt_dev_close_op(struct rte_eth_dev *eth_dev)
1658 {
1659         struct bnxt *bp = eth_dev->data->dev_private;
1660         int ret = 0;
1661
1662         if (rte_eal_process_type() != RTE_PROC_PRIMARY)
1663                 return 0;
1664
1665         pthread_mutex_lock(&bp->err_recovery_lock);
1666         if (bp->flags & BNXT_FLAG_FW_RESET) {
1667                 PMD_DRV_LOG(ERR,
1668                             "Adapter recovering from error...Please retry\n");
1669                 pthread_mutex_unlock(&bp->err_recovery_lock);
1670                 return -EAGAIN;
1671         }
1672         pthread_mutex_unlock(&bp->err_recovery_lock);
1673
1674         /* cancel the recovery handler before remove dev */
1675         rte_eal_alarm_cancel(bnxt_dev_reset_and_resume, (void *)bp);
1676         rte_eal_alarm_cancel(bnxt_dev_recover, (void *)bp);
1677         bnxt_cancel_fc_thread(bp);
1678
1679         if (eth_dev->data->dev_started)
1680                 ret = bnxt_dev_stop(eth_dev);
1681
1682         bnxt_uninit_resources(bp, false);
1683
1684         bnxt_drv_uninit(bp);
1685
1686         return ret;
1687 }
1688
1689 static void bnxt_mac_addr_remove_op(struct rte_eth_dev *eth_dev,
1690                                     uint32_t index)
1691 {
1692         struct bnxt *bp = eth_dev->data->dev_private;
1693         uint64_t pool_mask = eth_dev->data->mac_pool_sel[index];
1694         struct bnxt_vnic_info *vnic;
1695         struct bnxt_filter_info *filter, *temp_filter;
1696         uint32_t i;
1697
1698         if (is_bnxt_in_error(bp))
1699                 return;
1700
1701         /*
1702          * Loop through all VNICs from the specified filter flow pools to
1703          * remove the corresponding MAC addr filter
1704          */
1705         for (i = 0; i < bp->nr_vnics; i++) {
1706                 if (!(pool_mask & (1ULL << i)))
1707                         continue;
1708
1709                 vnic = &bp->vnic_info[i];
1710                 filter = STAILQ_FIRST(&vnic->filter);
1711                 while (filter) {
1712                         temp_filter = STAILQ_NEXT(filter, next);
1713                         if (filter->mac_index == index) {
1714                                 STAILQ_REMOVE(&vnic->filter, filter,
1715                                                 bnxt_filter_info, next);
1716                                 bnxt_hwrm_clear_l2_filter(bp, filter);
1717                                 bnxt_free_filter(bp, filter);
1718                         }
1719                         filter = temp_filter;
1720                 }
1721         }
1722 }
1723
1724 static int bnxt_add_mac_filter(struct bnxt *bp, struct bnxt_vnic_info *vnic,
1725                                struct rte_ether_addr *mac_addr, uint32_t index,
1726                                uint32_t pool)
1727 {
1728         struct bnxt_filter_info *filter;
1729         int rc = 0;
1730
1731         /* Attach requested MAC address to the new l2_filter */
1732         STAILQ_FOREACH(filter, &vnic->filter, next) {
1733                 if (filter->mac_index == index) {
1734                         PMD_DRV_LOG(DEBUG,
1735                                     "MAC addr already existed for pool %d\n",
1736                                     pool);
1737                         return 0;
1738                 }
1739         }
1740
1741         filter = bnxt_alloc_filter(bp);
1742         if (!filter) {
1743                 PMD_DRV_LOG(ERR, "L2 filter alloc failed\n");
1744                 return -ENODEV;
1745         }
1746
1747         /* bnxt_alloc_filter copies default MAC to filter->l2_addr. So,
1748          * if the MAC that's been programmed now is a different one, then,
1749          * copy that addr to filter->l2_addr
1750          */
1751         if (mac_addr)
1752                 memcpy(filter->l2_addr, mac_addr, RTE_ETHER_ADDR_LEN);
1753         filter->flags |= HWRM_CFA_L2_FILTER_ALLOC_INPUT_FLAGS_OUTERMOST;
1754
1755         rc = bnxt_hwrm_set_l2_filter(bp, vnic->fw_vnic_id, filter);
1756         if (!rc) {
1757                 filter->mac_index = index;
1758                 if (filter->mac_index == 0)
1759                         STAILQ_INSERT_HEAD(&vnic->filter, filter, next);
1760                 else
1761                         STAILQ_INSERT_TAIL(&vnic->filter, filter, next);
1762         } else {
1763                 bnxt_free_filter(bp, filter);
1764         }
1765
1766         return rc;
1767 }
1768
1769 static int bnxt_mac_addr_add_op(struct rte_eth_dev *eth_dev,
1770                                 struct rte_ether_addr *mac_addr,
1771                                 uint32_t index, uint32_t pool)
1772 {
1773         struct bnxt *bp = eth_dev->data->dev_private;
1774         struct bnxt_vnic_info *vnic = &bp->vnic_info[pool];
1775         int rc = 0;
1776
1777         rc = is_bnxt_in_error(bp);
1778         if (rc)
1779                 return rc;
1780
1781         if (BNXT_VF(bp) && !BNXT_VF_IS_TRUSTED(bp)) {
1782                 PMD_DRV_LOG(ERR, "Cannot add MAC address to a VF interface\n");
1783                 return -ENOTSUP;
1784         }
1785
1786         if (!vnic) {
1787                 PMD_DRV_LOG(ERR, "VNIC not found for pool %d!\n", pool);
1788                 return -EINVAL;
1789         }
1790
1791         /* Filter settings will get applied when port is started */
1792         if (!eth_dev->data->dev_started)
1793                 return 0;
1794
1795         rc = bnxt_add_mac_filter(bp, vnic, mac_addr, index, pool);
1796
1797         return rc;
1798 }
1799
1800 int bnxt_link_update_op(struct rte_eth_dev *eth_dev, int wait_to_complete)
1801 {
1802         int rc = 0;
1803         struct bnxt *bp = eth_dev->data->dev_private;
1804         struct rte_eth_link new;
1805         int cnt = wait_to_complete ? BNXT_MAX_LINK_WAIT_CNT :
1806                         BNXT_MIN_LINK_WAIT_CNT;
1807
1808         rc = is_bnxt_in_error(bp);
1809         if (rc)
1810                 return rc;
1811
1812         memset(&new, 0, sizeof(new));
1813
1814         if (bp->link_info == NULL)
1815                 goto out;
1816
1817         do {
1818                 /* Retrieve link info from hardware */
1819                 rc = bnxt_get_hwrm_link_config(bp, &new);
1820                 if (rc) {
1821                         new.link_speed = ETH_LINK_SPEED_100M;
1822                         new.link_duplex = ETH_LINK_FULL_DUPLEX;
1823                         PMD_DRV_LOG(ERR,
1824                                 "Failed to retrieve link rc = 0x%x!\n", rc);
1825                         goto out;
1826                 }
1827
1828                 if (!wait_to_complete || new.link_status)
1829                         break;
1830
1831                 rte_delay_ms(BNXT_LINK_WAIT_INTERVAL);
1832         } while (cnt--);
1833
1834         /* Only single function PF can bring phy down.
1835          * When port is stopped, report link down for VF/MH/NPAR functions.
1836          */
1837         if (!BNXT_SINGLE_PF(bp) && !eth_dev->data->dev_started)
1838                 memset(&new, 0, sizeof(new));
1839
1840 out:
1841         /* Timed out or success */
1842         if (new.link_status != eth_dev->data->dev_link.link_status ||
1843             new.link_speed != eth_dev->data->dev_link.link_speed) {
1844                 rte_eth_linkstatus_set(eth_dev, &new);
1845
1846                 rte_eth_dev_callback_process(eth_dev,
1847                                              RTE_ETH_EVENT_INTR_LSC,
1848                                              NULL);
1849
1850                 bnxt_print_link_info(eth_dev);
1851         }
1852
1853         return rc;
1854 }
1855
1856 static int bnxt_promiscuous_enable_op(struct rte_eth_dev *eth_dev)
1857 {
1858         struct bnxt *bp = eth_dev->data->dev_private;
1859         struct bnxt_vnic_info *vnic;
1860         uint32_t old_flags;
1861         int rc;
1862
1863         rc = is_bnxt_in_error(bp);
1864         if (rc)
1865                 return rc;
1866
1867         /* Filter settings will get applied when port is started */
1868         if (!eth_dev->data->dev_started)
1869                 return 0;
1870
1871         if (bp->vnic_info == NULL)
1872                 return 0;
1873
1874         vnic = BNXT_GET_DEFAULT_VNIC(bp);
1875
1876         old_flags = vnic->flags;
1877         vnic->flags |= BNXT_VNIC_INFO_PROMISC;
1878         rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1879         if (rc != 0)
1880                 vnic->flags = old_flags;
1881
1882         return rc;
1883 }
1884
1885 static int bnxt_promiscuous_disable_op(struct rte_eth_dev *eth_dev)
1886 {
1887         struct bnxt *bp = eth_dev->data->dev_private;
1888         struct bnxt_vnic_info *vnic;
1889         uint32_t old_flags;
1890         int rc;
1891
1892         rc = is_bnxt_in_error(bp);
1893         if (rc)
1894                 return rc;
1895
1896         /* Filter settings will get applied when port is started */
1897         if (!eth_dev->data->dev_started)
1898                 return 0;
1899
1900         if (bp->vnic_info == NULL)
1901                 return 0;
1902
1903         vnic = BNXT_GET_DEFAULT_VNIC(bp);
1904
1905         old_flags = vnic->flags;
1906         vnic->flags &= ~BNXT_VNIC_INFO_PROMISC;
1907         rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1908         if (rc != 0)
1909                 vnic->flags = old_flags;
1910
1911         return rc;
1912 }
1913
1914 static int bnxt_allmulticast_enable_op(struct rte_eth_dev *eth_dev)
1915 {
1916         struct bnxt *bp = eth_dev->data->dev_private;
1917         struct bnxt_vnic_info *vnic;
1918         uint32_t old_flags;
1919         int rc;
1920
1921         rc = is_bnxt_in_error(bp);
1922         if (rc)
1923                 return rc;
1924
1925         /* Filter settings will get applied when port is started */
1926         if (!eth_dev->data->dev_started)
1927                 return 0;
1928
1929         if (bp->vnic_info == NULL)
1930                 return 0;
1931
1932         vnic = BNXT_GET_DEFAULT_VNIC(bp);
1933
1934         old_flags = vnic->flags;
1935         vnic->flags |= BNXT_VNIC_INFO_ALLMULTI;
1936         rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1937         if (rc != 0)
1938                 vnic->flags = old_flags;
1939
1940         return rc;
1941 }
1942
1943 static int bnxt_allmulticast_disable_op(struct rte_eth_dev *eth_dev)
1944 {
1945         struct bnxt *bp = eth_dev->data->dev_private;
1946         struct bnxt_vnic_info *vnic;
1947         uint32_t old_flags;
1948         int rc;
1949
1950         rc = is_bnxt_in_error(bp);
1951         if (rc)
1952                 return rc;
1953
1954         /* Filter settings will get applied when port is started */
1955         if (!eth_dev->data->dev_started)
1956                 return 0;
1957
1958         if (bp->vnic_info == NULL)
1959                 return 0;
1960
1961         vnic = BNXT_GET_DEFAULT_VNIC(bp);
1962
1963         old_flags = vnic->flags;
1964         vnic->flags &= ~BNXT_VNIC_INFO_ALLMULTI;
1965         rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1966         if (rc != 0)
1967                 vnic->flags = old_flags;
1968
1969         return rc;
1970 }
1971
1972 /* Return bnxt_rx_queue pointer corresponding to a given rxq. */
1973 static struct bnxt_rx_queue *bnxt_qid_to_rxq(struct bnxt *bp, uint16_t qid)
1974 {
1975         if (qid >= bp->rx_nr_rings)
1976                 return NULL;
1977
1978         return bp->eth_dev->data->rx_queues[qid];
1979 }
1980
1981 /* Return rxq corresponding to a given rss table ring/group ID. */
1982 static uint16_t bnxt_rss_to_qid(struct bnxt *bp, uint16_t fwr)
1983 {
1984         struct bnxt_rx_queue *rxq;
1985         unsigned int i;
1986
1987         if (!BNXT_HAS_RING_GRPS(bp)) {
1988                 for (i = 0; i < bp->rx_nr_rings; i++) {
1989                         rxq = bp->eth_dev->data->rx_queues[i];
1990                         if (rxq->rx_ring->rx_ring_struct->fw_ring_id == fwr)
1991                                 return rxq->index;
1992                 }
1993         } else {
1994                 for (i = 0; i < bp->rx_nr_rings; i++) {
1995                         if (bp->grp_info[i].fw_grp_id == fwr)
1996                                 return i;
1997                 }
1998         }
1999
2000         return INVALID_HW_RING_ID;
2001 }
2002
2003 static int bnxt_reta_update_op(struct rte_eth_dev *eth_dev,
2004                             struct rte_eth_rss_reta_entry64 *reta_conf,
2005                             uint16_t reta_size)
2006 {
2007         struct bnxt *bp = eth_dev->data->dev_private;
2008         struct rte_eth_conf *dev_conf = &bp->eth_dev->data->dev_conf;
2009         struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
2010         uint16_t tbl_size = bnxt_rss_hash_tbl_size(bp);
2011         uint16_t idx, sft;
2012         int i, rc;
2013
2014         rc = is_bnxt_in_error(bp);
2015         if (rc)
2016                 return rc;
2017
2018         if (!vnic->rss_table)
2019                 return -EINVAL;
2020
2021         if (!(dev_conf->rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG))
2022                 return -EINVAL;
2023
2024         if (reta_size != tbl_size) {
2025                 PMD_DRV_LOG(ERR, "The configured hash table lookup size "
2026                         "(%d) must equal the size supported by the hardware "
2027                         "(%d)\n", reta_size, tbl_size);
2028                 return -EINVAL;
2029         }
2030
2031         for (i = 0; i < reta_size; i++) {
2032                 struct bnxt_rx_queue *rxq;
2033
2034                 idx = i / RTE_RETA_GROUP_SIZE;
2035                 sft = i % RTE_RETA_GROUP_SIZE;
2036
2037                 if (!(reta_conf[idx].mask & (1ULL << sft)))
2038                         continue;
2039
2040                 rxq = bnxt_qid_to_rxq(bp, reta_conf[idx].reta[sft]);
2041                 if (!rxq) {
2042                         PMD_DRV_LOG(ERR, "Invalid ring in reta_conf.\n");
2043                         return -EINVAL;
2044                 }
2045
2046                 if (BNXT_CHIP_P5(bp)) {
2047                         vnic->rss_table[i * 2] =
2048                                 rxq->rx_ring->rx_ring_struct->fw_ring_id;
2049                         vnic->rss_table[i * 2 + 1] =
2050                                 rxq->cp_ring->cp_ring_struct->fw_ring_id;
2051                 } else {
2052                         vnic->rss_table[i] =
2053                             vnic->fw_grp_ids[reta_conf[idx].reta[sft]];
2054                 }
2055         }
2056
2057         rc = bnxt_hwrm_vnic_rss_cfg(bp, vnic);
2058         return rc;
2059 }
2060
2061 static int bnxt_reta_query_op(struct rte_eth_dev *eth_dev,
2062                               struct rte_eth_rss_reta_entry64 *reta_conf,
2063                               uint16_t reta_size)
2064 {
2065         struct bnxt *bp = eth_dev->data->dev_private;
2066         struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
2067         uint16_t tbl_size = bnxt_rss_hash_tbl_size(bp);
2068         uint16_t idx, sft, i;
2069         int rc;
2070
2071         rc = is_bnxt_in_error(bp);
2072         if (rc)
2073                 return rc;
2074
2075         if (!vnic)
2076                 return -EINVAL;
2077         if (!vnic->rss_table)
2078                 return -EINVAL;
2079
2080         if (reta_size != tbl_size) {
2081                 PMD_DRV_LOG(ERR, "The configured hash table lookup size "
2082                         "(%d) must equal the size supported by the hardware "
2083                         "(%d)\n", reta_size, tbl_size);
2084                 return -EINVAL;
2085         }
2086
2087         for (idx = 0, i = 0; i < reta_size; i++) {
2088                 idx = i / RTE_RETA_GROUP_SIZE;
2089                 sft = i % RTE_RETA_GROUP_SIZE;
2090
2091                 if (reta_conf[idx].mask & (1ULL << sft)) {
2092                         uint16_t qid;
2093
2094                         if (BNXT_CHIP_P5(bp))
2095                                 qid = bnxt_rss_to_qid(bp,
2096                                                       vnic->rss_table[i * 2]);
2097                         else
2098                                 qid = bnxt_rss_to_qid(bp, vnic->rss_table[i]);
2099
2100                         if (qid == INVALID_HW_RING_ID) {
2101                                 PMD_DRV_LOG(ERR, "Inv. entry in rss table.\n");
2102                                 return -EINVAL;
2103                         }
2104                         reta_conf[idx].reta[sft] = qid;
2105                 }
2106         }
2107
2108         return 0;
2109 }
2110
2111 static int bnxt_rss_hash_update_op(struct rte_eth_dev *eth_dev,
2112                                    struct rte_eth_rss_conf *rss_conf)
2113 {
2114         struct bnxt *bp = eth_dev->data->dev_private;
2115         struct rte_eth_conf *dev_conf = &bp->eth_dev->data->dev_conf;
2116         struct bnxt_vnic_info *vnic;
2117         int rc;
2118
2119         rc = is_bnxt_in_error(bp);
2120         if (rc)
2121                 return rc;
2122
2123         /*
2124          * If RSS enablement were different than dev_configure,
2125          * then return -EINVAL
2126          */
2127         if (dev_conf->rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG) {
2128                 if (!rss_conf->rss_hf)
2129                         PMD_DRV_LOG(ERR, "Hash type NONE\n");
2130         } else {
2131                 if (rss_conf->rss_hf & BNXT_ETH_RSS_SUPPORT)
2132                         return -EINVAL;
2133         }
2134
2135         bp->flags |= BNXT_FLAG_UPDATE_HASH;
2136         memcpy(&eth_dev->data->dev_conf.rx_adv_conf.rss_conf,
2137                rss_conf,
2138                sizeof(*rss_conf));
2139
2140         /* Update the default RSS VNIC(s) */
2141         vnic = BNXT_GET_DEFAULT_VNIC(bp);
2142         vnic->hash_type = bnxt_rte_to_hwrm_hash_types(rss_conf->rss_hf);
2143         vnic->hash_mode =
2144                 bnxt_rte_to_hwrm_hash_level(bp, rss_conf->rss_hf,
2145                                             ETH_RSS_LEVEL(rss_conf->rss_hf));
2146
2147         /*
2148          * If hashkey is not specified, use the previously configured
2149          * hashkey
2150          */
2151         if (!rss_conf->rss_key)
2152                 goto rss_config;
2153
2154         if (rss_conf->rss_key_len != HW_HASH_KEY_SIZE) {
2155                 PMD_DRV_LOG(ERR,
2156                             "Invalid hashkey length, should be %d bytes\n",
2157                             HW_HASH_KEY_SIZE);
2158                 return -EINVAL;
2159         }
2160         memcpy(vnic->rss_hash_key, rss_conf->rss_key, rss_conf->rss_key_len);
2161
2162 rss_config:
2163         rc = bnxt_hwrm_vnic_rss_cfg(bp, vnic);
2164         return rc;
2165 }
2166
2167 static int bnxt_rss_hash_conf_get_op(struct rte_eth_dev *eth_dev,
2168                                      struct rte_eth_rss_conf *rss_conf)
2169 {
2170         struct bnxt *bp = eth_dev->data->dev_private;
2171         struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
2172         int len, rc;
2173         uint32_t hash_types;
2174
2175         rc = is_bnxt_in_error(bp);
2176         if (rc)
2177                 return rc;
2178
2179         /* RSS configuration is the same for all VNICs */
2180         if (vnic && vnic->rss_hash_key) {
2181                 if (rss_conf->rss_key) {
2182                         len = rss_conf->rss_key_len <= HW_HASH_KEY_SIZE ?
2183                               rss_conf->rss_key_len : HW_HASH_KEY_SIZE;
2184                         memcpy(rss_conf->rss_key, vnic->rss_hash_key, len);
2185                 }
2186
2187                 hash_types = vnic->hash_type;
2188                 rss_conf->rss_hf = 0;
2189                 if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV4) {
2190                         rss_conf->rss_hf |= ETH_RSS_IPV4;
2191                         hash_types &= ~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV4;
2192                 }
2193                 if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV4) {
2194                         rss_conf->rss_hf |= ETH_RSS_NONFRAG_IPV4_TCP;
2195                         hash_types &=
2196                                 ~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV4;
2197                 }
2198                 if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV4) {
2199                         rss_conf->rss_hf |= ETH_RSS_NONFRAG_IPV4_UDP;
2200                         hash_types &=
2201                                 ~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV4;
2202                 }
2203                 if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV6) {
2204                         rss_conf->rss_hf |= ETH_RSS_IPV6;
2205                         hash_types &= ~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV6;
2206                 }
2207                 if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV6) {
2208                         rss_conf->rss_hf |= ETH_RSS_NONFRAG_IPV6_TCP;
2209                         hash_types &=
2210                                 ~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV6;
2211                 }
2212                 if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV6) {
2213                         rss_conf->rss_hf |= ETH_RSS_NONFRAG_IPV6_UDP;
2214                         hash_types &=
2215                                 ~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV6;
2216                 }
2217
2218                 rss_conf->rss_hf |=
2219                         bnxt_hwrm_to_rte_rss_level(bp, vnic->hash_mode);
2220
2221                 if (hash_types) {
2222                         PMD_DRV_LOG(ERR,
2223                                 "Unknown RSS config from firmware (%08x), RSS disabled",
2224                                 vnic->hash_type);
2225                         return -ENOTSUP;
2226                 }
2227         } else {
2228                 rss_conf->rss_hf = 0;
2229         }
2230         return 0;
2231 }
2232
2233 static int bnxt_flow_ctrl_get_op(struct rte_eth_dev *dev,
2234                                struct rte_eth_fc_conf *fc_conf)
2235 {
2236         struct bnxt *bp = dev->data->dev_private;
2237         struct rte_eth_link link_info;
2238         int rc;
2239
2240         rc = is_bnxt_in_error(bp);
2241         if (rc)
2242                 return rc;
2243
2244         rc = bnxt_get_hwrm_link_config(bp, &link_info);
2245         if (rc)
2246                 return rc;
2247
2248         memset(fc_conf, 0, sizeof(*fc_conf));
2249         if (bp->link_info->auto_pause)
2250                 fc_conf->autoneg = 1;
2251         switch (bp->link_info->pause) {
2252         case 0:
2253                 fc_conf->mode = RTE_FC_NONE;
2254                 break;
2255         case HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_TX:
2256                 fc_conf->mode = RTE_FC_TX_PAUSE;
2257                 break;
2258         case HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_RX:
2259                 fc_conf->mode = RTE_FC_RX_PAUSE;
2260                 break;
2261         case (HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_TX |
2262                         HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_RX):
2263                 fc_conf->mode = RTE_FC_FULL;
2264                 break;
2265         }
2266         return 0;
2267 }
2268
2269 static int bnxt_flow_ctrl_set_op(struct rte_eth_dev *dev,
2270                                struct rte_eth_fc_conf *fc_conf)
2271 {
2272         struct bnxt *bp = dev->data->dev_private;
2273         int rc;
2274
2275         rc = is_bnxt_in_error(bp);
2276         if (rc)
2277                 return rc;
2278
2279         if (!BNXT_SINGLE_PF(bp)) {
2280                 PMD_DRV_LOG(ERR,
2281                             "Flow Control Settings cannot be modified on VF or on shared PF\n");
2282                 return -ENOTSUP;
2283         }
2284
2285         switch (fc_conf->mode) {
2286         case RTE_FC_NONE:
2287                 bp->link_info->auto_pause = 0;
2288                 bp->link_info->force_pause = 0;
2289                 break;
2290         case RTE_FC_RX_PAUSE:
2291                 if (fc_conf->autoneg) {
2292                         bp->link_info->auto_pause =
2293                                         HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_RX;
2294                         bp->link_info->force_pause = 0;
2295                 } else {
2296                         bp->link_info->auto_pause = 0;
2297                         bp->link_info->force_pause =
2298                                         HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_RX;
2299                 }
2300                 break;
2301         case RTE_FC_TX_PAUSE:
2302                 if (fc_conf->autoneg) {
2303                         bp->link_info->auto_pause =
2304                                         HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_TX;
2305                         bp->link_info->force_pause = 0;
2306                 } else {
2307                         bp->link_info->auto_pause = 0;
2308                         bp->link_info->force_pause =
2309                                         HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_TX;
2310                 }
2311                 break;
2312         case RTE_FC_FULL:
2313                 if (fc_conf->autoneg) {
2314                         bp->link_info->auto_pause =
2315                                         HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_TX |
2316                                         HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_RX;
2317                         bp->link_info->force_pause = 0;
2318                 } else {
2319                         bp->link_info->auto_pause = 0;
2320                         bp->link_info->force_pause =
2321                                         HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_TX |
2322                                         HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_RX;
2323                 }
2324                 break;
2325         }
2326         return bnxt_set_hwrm_link_config(bp, true);
2327 }
2328
2329 /* Add UDP tunneling port */
2330 static int
2331 bnxt_udp_tunnel_port_add_op(struct rte_eth_dev *eth_dev,
2332                          struct rte_eth_udp_tunnel *udp_tunnel)
2333 {
2334         struct bnxt *bp = eth_dev->data->dev_private;
2335         uint16_t tunnel_type = 0;
2336         int rc = 0;
2337
2338         rc = is_bnxt_in_error(bp);
2339         if (rc)
2340                 return rc;
2341
2342         switch (udp_tunnel->prot_type) {
2343         case RTE_TUNNEL_TYPE_VXLAN:
2344                 if (bp->vxlan_port_cnt) {
2345                         PMD_DRV_LOG(ERR, "Tunnel Port %d already programmed\n",
2346                                 udp_tunnel->udp_port);
2347                         if (bp->vxlan_port != udp_tunnel->udp_port) {
2348                                 PMD_DRV_LOG(ERR, "Only one port allowed\n");
2349                                 return -ENOSPC;
2350                         }
2351                         bp->vxlan_port_cnt++;
2352                         return 0;
2353                 }
2354                 tunnel_type =
2355                         HWRM_TUNNEL_DST_PORT_ALLOC_INPUT_TUNNEL_TYPE_VXLAN;
2356                 bp->vxlan_port_cnt++;
2357                 break;
2358         case RTE_TUNNEL_TYPE_GENEVE:
2359                 if (bp->geneve_port_cnt) {
2360                         PMD_DRV_LOG(ERR, "Tunnel Port %d already programmed\n",
2361                                 udp_tunnel->udp_port);
2362                         if (bp->geneve_port != udp_tunnel->udp_port) {
2363                                 PMD_DRV_LOG(ERR, "Only one port allowed\n");
2364                                 return -ENOSPC;
2365                         }
2366                         bp->geneve_port_cnt++;
2367                         return 0;
2368                 }
2369                 tunnel_type =
2370                         HWRM_TUNNEL_DST_PORT_ALLOC_INPUT_TUNNEL_TYPE_GENEVE;
2371                 bp->geneve_port_cnt++;
2372                 break;
2373         default:
2374                 PMD_DRV_LOG(ERR, "Tunnel type is not supported\n");
2375                 return -ENOTSUP;
2376         }
2377         rc = bnxt_hwrm_tunnel_dst_port_alloc(bp, udp_tunnel->udp_port,
2378                                              tunnel_type);
2379         return rc;
2380 }
2381
2382 static int
2383 bnxt_udp_tunnel_port_del_op(struct rte_eth_dev *eth_dev,
2384                          struct rte_eth_udp_tunnel *udp_tunnel)
2385 {
2386         struct bnxt *bp = eth_dev->data->dev_private;
2387         uint16_t tunnel_type = 0;
2388         uint16_t port = 0;
2389         int rc = 0;
2390
2391         rc = is_bnxt_in_error(bp);
2392         if (rc)
2393                 return rc;
2394
2395         switch (udp_tunnel->prot_type) {
2396         case RTE_TUNNEL_TYPE_VXLAN:
2397                 if (!bp->vxlan_port_cnt) {
2398                         PMD_DRV_LOG(ERR, "No Tunnel port configured yet\n");
2399                         return -EINVAL;
2400                 }
2401                 if (bp->vxlan_port != udp_tunnel->udp_port) {
2402                         PMD_DRV_LOG(ERR, "Req Port: %d. Configured port: %d\n",
2403                                 udp_tunnel->udp_port, bp->vxlan_port);
2404                         return -EINVAL;
2405                 }
2406                 if (--bp->vxlan_port_cnt)
2407                         return 0;
2408
2409                 tunnel_type =
2410                         HWRM_TUNNEL_DST_PORT_FREE_INPUT_TUNNEL_TYPE_VXLAN;
2411                 port = bp->vxlan_fw_dst_port_id;
2412                 break;
2413         case RTE_TUNNEL_TYPE_GENEVE:
2414                 if (!bp->geneve_port_cnt) {
2415                         PMD_DRV_LOG(ERR, "No Tunnel port configured yet\n");
2416                         return -EINVAL;
2417                 }
2418                 if (bp->geneve_port != udp_tunnel->udp_port) {
2419                         PMD_DRV_LOG(ERR, "Req Port: %d. Configured port: %d\n",
2420                                 udp_tunnel->udp_port, bp->geneve_port);
2421                         return -EINVAL;
2422                 }
2423                 if (--bp->geneve_port_cnt)
2424                         return 0;
2425
2426                 tunnel_type =
2427                         HWRM_TUNNEL_DST_PORT_FREE_INPUT_TUNNEL_TYPE_GENEVE;
2428                 port = bp->geneve_fw_dst_port_id;
2429                 break;
2430         default:
2431                 PMD_DRV_LOG(ERR, "Tunnel type is not supported\n");
2432                 return -ENOTSUP;
2433         }
2434
2435         rc = bnxt_hwrm_tunnel_dst_port_free(bp, port, tunnel_type);
2436         return rc;
2437 }
2438
2439 static int bnxt_del_vlan_filter(struct bnxt *bp, uint16_t vlan_id)
2440 {
2441         struct bnxt_filter_info *filter;
2442         struct bnxt_vnic_info *vnic;
2443         int rc = 0;
2444         uint32_t chk = HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN;
2445
2446         vnic = BNXT_GET_DEFAULT_VNIC(bp);
2447         filter = STAILQ_FIRST(&vnic->filter);
2448         while (filter) {
2449                 /* Search for this matching MAC+VLAN filter */
2450                 if (bnxt_vlan_filter_exists(bp, filter, chk, vlan_id)) {
2451                         /* Delete the filter */
2452                         rc = bnxt_hwrm_clear_l2_filter(bp, filter);
2453                         if (rc)
2454                                 return rc;
2455                         STAILQ_REMOVE(&vnic->filter, filter,
2456                                       bnxt_filter_info, next);
2457                         bnxt_free_filter(bp, filter);
2458                         PMD_DRV_LOG(INFO,
2459                                     "Deleted vlan filter for %d\n",
2460                                     vlan_id);
2461                         return 0;
2462                 }
2463                 filter = STAILQ_NEXT(filter, next);
2464         }
2465         return -ENOENT;
2466 }
2467
2468 static int bnxt_add_vlan_filter(struct bnxt *bp, uint16_t vlan_id)
2469 {
2470         struct bnxt_filter_info *filter;
2471         struct bnxt_vnic_info *vnic;
2472         int rc = 0;
2473         uint32_t en = HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN |
2474                 HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN_MASK;
2475         uint32_t chk = HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN;
2476
2477         /* Implementation notes on the use of VNIC in this command:
2478          *
2479          * By default, these filters belong to default vnic for the function.
2480          * Once these filters are set up, only destination VNIC can be modified.
2481          * If the destination VNIC is not specified in this command,
2482          * then the HWRM shall only create an l2 context id.
2483          */
2484
2485         vnic = BNXT_GET_DEFAULT_VNIC(bp);
2486         filter = STAILQ_FIRST(&vnic->filter);
2487         /* Check if the VLAN has already been added */
2488         while (filter) {
2489                 if (bnxt_vlan_filter_exists(bp, filter, chk, vlan_id))
2490                         return -EEXIST;
2491
2492                 filter = STAILQ_NEXT(filter, next);
2493         }
2494
2495         /* No match found. Alloc a fresh filter and issue the L2_FILTER_ALLOC
2496          * command to create MAC+VLAN filter with the right flags, enables set.
2497          */
2498         filter = bnxt_alloc_filter(bp);
2499         if (!filter) {
2500                 PMD_DRV_LOG(ERR,
2501                             "MAC/VLAN filter alloc failed\n");
2502                 return -ENOMEM;
2503         }
2504         /* MAC + VLAN ID filter */
2505         /* If l2_ivlan == 0 and l2_ivlan_mask != 0, only
2506          * untagged packets are received
2507          *
2508          * If l2_ivlan != 0 and l2_ivlan_mask != 0, untagged
2509          * packets and only the programmed vlan's packets are received
2510          */
2511         filter->l2_ivlan = vlan_id;
2512         filter->l2_ivlan_mask = 0x0FFF;
2513         filter->enables |= en;
2514         filter->flags |= HWRM_CFA_L2_FILTER_ALLOC_INPUT_FLAGS_OUTERMOST;
2515
2516         rc = bnxt_hwrm_set_l2_filter(bp, vnic->fw_vnic_id, filter);
2517         if (rc) {
2518                 /* Free the newly allocated filter as we were
2519                  * not able to create the filter in hardware.
2520                  */
2521                 bnxt_free_filter(bp, filter);
2522                 return rc;
2523         }
2524
2525         filter->mac_index = 0;
2526         /* Add this new filter to the list */
2527         if (vlan_id == 0)
2528                 STAILQ_INSERT_HEAD(&vnic->filter, filter, next);
2529         else
2530                 STAILQ_INSERT_TAIL(&vnic->filter, filter, next);
2531
2532         PMD_DRV_LOG(INFO,
2533                     "Added Vlan filter for %d\n", vlan_id);
2534         return rc;
2535 }
2536
2537 static int bnxt_vlan_filter_set_op(struct rte_eth_dev *eth_dev,
2538                 uint16_t vlan_id, int on)
2539 {
2540         struct bnxt *bp = eth_dev->data->dev_private;
2541         int rc;
2542
2543         rc = is_bnxt_in_error(bp);
2544         if (rc)
2545                 return rc;
2546
2547         if (!eth_dev->data->dev_started) {
2548                 PMD_DRV_LOG(ERR, "port must be started before setting vlan\n");
2549                 return -EINVAL;
2550         }
2551
2552         /* These operations apply to ALL existing MAC/VLAN filters */
2553         if (on)
2554                 return bnxt_add_vlan_filter(bp, vlan_id);
2555         else
2556                 return bnxt_del_vlan_filter(bp, vlan_id);
2557 }
2558
2559 static int bnxt_del_dflt_mac_filter(struct bnxt *bp,
2560                                     struct bnxt_vnic_info *vnic)
2561 {
2562         struct bnxt_filter_info *filter;
2563         int rc;
2564
2565         filter = STAILQ_FIRST(&vnic->filter);
2566         while (filter) {
2567                 if (filter->mac_index == 0 &&
2568                     !memcmp(filter->l2_addr, bp->mac_addr,
2569                             RTE_ETHER_ADDR_LEN)) {
2570                         rc = bnxt_hwrm_clear_l2_filter(bp, filter);
2571                         if (!rc) {
2572                                 STAILQ_REMOVE(&vnic->filter, filter,
2573                                               bnxt_filter_info, next);
2574                                 bnxt_free_filter(bp, filter);
2575                         }
2576                         return rc;
2577                 }
2578                 filter = STAILQ_NEXT(filter, next);
2579         }
2580         return 0;
2581 }
2582
2583 static int
2584 bnxt_config_vlan_hw_filter(struct bnxt *bp, uint64_t rx_offloads)
2585 {
2586         struct bnxt_vnic_info *vnic;
2587         unsigned int i;
2588         int rc;
2589
2590         vnic = BNXT_GET_DEFAULT_VNIC(bp);
2591         if (!(rx_offloads & DEV_RX_OFFLOAD_VLAN_FILTER)) {
2592                 /* Remove any VLAN filters programmed */
2593                 for (i = 0; i < RTE_ETHER_MAX_VLAN_ID; i++)
2594                         bnxt_del_vlan_filter(bp, i);
2595
2596                 rc = bnxt_add_mac_filter(bp, vnic, NULL, 0, 0);
2597                 if (rc)
2598                         return rc;
2599         } else {
2600                 /* Default filter will allow packets that match the
2601                  * dest mac. So, it has to be deleted, otherwise, we
2602                  * will endup receiving vlan packets for which the
2603                  * filter is not programmed, when hw-vlan-filter
2604                  * configuration is ON
2605                  */
2606                 bnxt_del_dflt_mac_filter(bp, vnic);
2607                 /* This filter will allow only untagged packets */
2608                 bnxt_add_vlan_filter(bp, 0);
2609         }
2610         PMD_DRV_LOG(DEBUG, "VLAN Filtering: %d\n",
2611                     !!(rx_offloads & DEV_RX_OFFLOAD_VLAN_FILTER));
2612
2613         return 0;
2614 }
2615
2616 static int bnxt_free_one_vnic(struct bnxt *bp, uint16_t vnic_id)
2617 {
2618         struct bnxt_vnic_info *vnic = &bp->vnic_info[vnic_id];
2619         unsigned int i;
2620         int rc;
2621
2622         /* Destroy vnic filters and vnic */
2623         if (bp->eth_dev->data->dev_conf.rxmode.offloads &
2624             DEV_RX_OFFLOAD_VLAN_FILTER) {
2625                 for (i = 0; i < RTE_ETHER_MAX_VLAN_ID; i++)
2626                         bnxt_del_vlan_filter(bp, i);
2627         }
2628         bnxt_del_dflt_mac_filter(bp, vnic);
2629
2630         rc = bnxt_hwrm_vnic_ctx_free(bp, vnic);
2631         if (rc)
2632                 return rc;
2633
2634         rc = bnxt_hwrm_vnic_free(bp, vnic);
2635         if (rc)
2636                 return rc;
2637
2638         rte_free(vnic->fw_grp_ids);
2639         vnic->fw_grp_ids = NULL;
2640
2641         vnic->rx_queue_cnt = 0;
2642
2643         return 0;
2644 }
2645
2646 static int
2647 bnxt_config_vlan_hw_stripping(struct bnxt *bp, uint64_t rx_offloads)
2648 {
2649         struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
2650         int rc;
2651
2652         /* Destroy, recreate and reconfigure the default vnic */
2653         rc = bnxt_free_one_vnic(bp, 0);
2654         if (rc)
2655                 return rc;
2656
2657         /* default vnic 0 */
2658         rc = bnxt_setup_one_vnic(bp, 0);
2659         if (rc)
2660                 return rc;
2661
2662         if (bp->eth_dev->data->dev_conf.rxmode.offloads &
2663             DEV_RX_OFFLOAD_VLAN_FILTER) {
2664                 rc = bnxt_add_vlan_filter(bp, 0);
2665                 if (rc)
2666                         return rc;
2667                 rc = bnxt_restore_vlan_filters(bp);
2668                 if (rc)
2669                         return rc;
2670         } else {
2671                 rc = bnxt_add_mac_filter(bp, vnic, NULL, 0, 0);
2672                 if (rc)
2673                         return rc;
2674         }
2675
2676         rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
2677         if (rc)
2678                 return rc;
2679
2680         PMD_DRV_LOG(DEBUG, "VLAN Strip Offload: %d\n",
2681                     !!(rx_offloads & DEV_RX_OFFLOAD_VLAN_STRIP));
2682
2683         return rc;
2684 }
2685
2686 static int
2687 bnxt_vlan_offload_set_op(struct rte_eth_dev *dev, int mask)
2688 {
2689         uint64_t rx_offloads = dev->data->dev_conf.rxmode.offloads;
2690         struct bnxt *bp = dev->data->dev_private;
2691         int rc;
2692
2693         rc = is_bnxt_in_error(bp);
2694         if (rc)
2695                 return rc;
2696
2697         /* Filter settings will get applied when port is started */
2698         if (!dev->data->dev_started)
2699                 return 0;
2700
2701         if (mask & ETH_VLAN_FILTER_MASK) {
2702                 /* Enable or disable VLAN filtering */
2703                 rc = bnxt_config_vlan_hw_filter(bp, rx_offloads);
2704                 if (rc)
2705                         return rc;
2706         }
2707
2708         if (mask & ETH_VLAN_STRIP_MASK) {
2709                 /* Enable or disable VLAN stripping */
2710                 rc = bnxt_config_vlan_hw_stripping(bp, rx_offloads);
2711                 if (rc)
2712                         return rc;
2713         }
2714
2715         if (mask & ETH_VLAN_EXTEND_MASK) {
2716                 if (rx_offloads & DEV_RX_OFFLOAD_VLAN_EXTEND)
2717                         PMD_DRV_LOG(DEBUG, "Extend VLAN supported\n");
2718                 else
2719                         PMD_DRV_LOG(INFO, "Extend VLAN unsupported\n");
2720         }
2721
2722         return 0;
2723 }
2724
2725 static int
2726 bnxt_vlan_tpid_set_op(struct rte_eth_dev *dev, enum rte_vlan_type vlan_type,
2727                       uint16_t tpid)
2728 {
2729         struct bnxt *bp = dev->data->dev_private;
2730         int qinq = dev->data->dev_conf.rxmode.offloads &
2731                    DEV_RX_OFFLOAD_VLAN_EXTEND;
2732
2733         if (vlan_type != ETH_VLAN_TYPE_INNER &&
2734             vlan_type != ETH_VLAN_TYPE_OUTER) {
2735                 PMD_DRV_LOG(ERR,
2736                             "Unsupported vlan type.");
2737                 return -EINVAL;
2738         }
2739         if (!qinq) {
2740                 PMD_DRV_LOG(ERR,
2741                             "QinQ not enabled. Needs to be ON as we can "
2742                             "accelerate only outer vlan\n");
2743                 return -EINVAL;
2744         }
2745
2746         if (vlan_type == ETH_VLAN_TYPE_OUTER) {
2747                 switch (tpid) {
2748                 case RTE_ETHER_TYPE_QINQ:
2749                         bp->outer_tpid_bd =
2750                                 TX_BD_LONG_CFA_META_VLAN_TPID_TPID88A8;
2751                                 break;
2752                 case RTE_ETHER_TYPE_VLAN:
2753                         bp->outer_tpid_bd =
2754                                 TX_BD_LONG_CFA_META_VLAN_TPID_TPID8100;
2755                                 break;
2756                 case RTE_ETHER_TYPE_QINQ1:
2757                         bp->outer_tpid_bd =
2758                                 TX_BD_LONG_CFA_META_VLAN_TPID_TPID9100;
2759                                 break;
2760                 case RTE_ETHER_TYPE_QINQ2:
2761                         bp->outer_tpid_bd =
2762                                 TX_BD_LONG_CFA_META_VLAN_TPID_TPID9200;
2763                                 break;
2764                 case RTE_ETHER_TYPE_QINQ3:
2765                         bp->outer_tpid_bd =
2766                                  TX_BD_LONG_CFA_META_VLAN_TPID_TPID9300;
2767                                 break;
2768                 default:
2769                         PMD_DRV_LOG(ERR, "Invalid TPID: %x\n", tpid);
2770                         return -EINVAL;
2771                 }
2772                 bp->outer_tpid_bd |= tpid;
2773                 PMD_DRV_LOG(INFO, "outer_tpid_bd = %x\n", bp->outer_tpid_bd);
2774         } else if (vlan_type == ETH_VLAN_TYPE_INNER) {
2775                 PMD_DRV_LOG(ERR,
2776                             "Can accelerate only outer vlan in QinQ\n");
2777                 return -EINVAL;
2778         }
2779
2780         return 0;
2781 }
2782
2783 static int
2784 bnxt_set_default_mac_addr_op(struct rte_eth_dev *dev,
2785                              struct rte_ether_addr *addr)
2786 {
2787         struct bnxt *bp = dev->data->dev_private;
2788         /* Default Filter is tied to VNIC 0 */
2789         struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
2790         int rc;
2791
2792         rc = is_bnxt_in_error(bp);
2793         if (rc)
2794                 return rc;
2795
2796         if (BNXT_VF(bp) && !BNXT_VF_IS_TRUSTED(bp))
2797                 return -EPERM;
2798
2799         if (rte_is_zero_ether_addr(addr))
2800                 return -EINVAL;
2801
2802         /* Filter settings will get applied when port is started */
2803         if (!dev->data->dev_started)
2804                 return 0;
2805
2806         /* Check if the requested MAC is already added */
2807         if (memcmp(addr, bp->mac_addr, RTE_ETHER_ADDR_LEN) == 0)
2808                 return 0;
2809
2810         /* Destroy filter and re-create it */
2811         bnxt_del_dflt_mac_filter(bp, vnic);
2812
2813         memcpy(bp->mac_addr, addr, RTE_ETHER_ADDR_LEN);
2814         if (dev->data->dev_conf.rxmode.offloads & DEV_RX_OFFLOAD_VLAN_FILTER) {
2815                 /* This filter will allow only untagged packets */
2816                 rc = bnxt_add_vlan_filter(bp, 0);
2817         } else {
2818                 rc = bnxt_add_mac_filter(bp, vnic, addr, 0, 0);
2819         }
2820
2821         PMD_DRV_LOG(DEBUG, "Set MAC addr\n");
2822         return rc;
2823 }
2824
2825 static int
2826 bnxt_dev_set_mc_addr_list_op(struct rte_eth_dev *eth_dev,
2827                           struct rte_ether_addr *mc_addr_set,
2828                           uint32_t nb_mc_addr)
2829 {
2830         struct bnxt *bp = eth_dev->data->dev_private;
2831         char *mc_addr_list = (char *)mc_addr_set;
2832         struct bnxt_vnic_info *vnic;
2833         uint32_t off = 0, i = 0;
2834         int rc;
2835
2836         rc = is_bnxt_in_error(bp);
2837         if (rc)
2838                 return rc;
2839
2840         vnic = BNXT_GET_DEFAULT_VNIC(bp);
2841
2842         if (nb_mc_addr > BNXT_MAX_MC_ADDRS) {
2843                 vnic->flags |= BNXT_VNIC_INFO_ALLMULTI;
2844                 goto allmulti;
2845         }
2846
2847         /* TODO Check for Duplicate mcast addresses */
2848         vnic->flags &= ~BNXT_VNIC_INFO_ALLMULTI;
2849         for (i = 0; i < nb_mc_addr; i++) {
2850                 memcpy(vnic->mc_list + off, &mc_addr_list[i],
2851                         RTE_ETHER_ADDR_LEN);
2852                 off += RTE_ETHER_ADDR_LEN;
2853         }
2854
2855         vnic->mc_addr_cnt = i;
2856         if (vnic->mc_addr_cnt)
2857                 vnic->flags |= BNXT_VNIC_INFO_MCAST;
2858         else
2859                 vnic->flags &= ~BNXT_VNIC_INFO_MCAST;
2860
2861 allmulti:
2862         return bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
2863 }
2864
2865 static int
2866 bnxt_fw_version_get(struct rte_eth_dev *dev, char *fw_version, size_t fw_size)
2867 {
2868         struct bnxt *bp = dev->data->dev_private;
2869         uint8_t fw_major = (bp->fw_ver >> 24) & 0xff;
2870         uint8_t fw_minor = (bp->fw_ver >> 16) & 0xff;
2871         uint8_t fw_updt = (bp->fw_ver >> 8) & 0xff;
2872         uint8_t fw_rsvd = bp->fw_ver & 0xff;
2873         int ret;
2874
2875         ret = snprintf(fw_version, fw_size, "%d.%d.%d.%d",
2876                         fw_major, fw_minor, fw_updt, fw_rsvd);
2877         if (ret < 0)
2878                 return -EINVAL;
2879
2880         ret += 1; /* add the size of '\0' */
2881         if (fw_size < (size_t)ret)
2882                 return ret;
2883         else
2884                 return 0;
2885 }
2886
2887 static void
2888 bnxt_rxq_info_get_op(struct rte_eth_dev *dev, uint16_t queue_id,
2889         struct rte_eth_rxq_info *qinfo)
2890 {
2891         struct bnxt *bp = dev->data->dev_private;
2892         struct bnxt_rx_queue *rxq;
2893
2894         if (is_bnxt_in_error(bp))
2895                 return;
2896
2897         rxq = dev->data->rx_queues[queue_id];
2898
2899         qinfo->mp = rxq->mb_pool;
2900         qinfo->scattered_rx = dev->data->scattered_rx;
2901         qinfo->nb_desc = rxq->nb_rx_desc;
2902
2903         qinfo->conf.rx_free_thresh = rxq->rx_free_thresh;
2904         qinfo->conf.rx_drop_en = rxq->drop_en;
2905         qinfo->conf.rx_deferred_start = rxq->rx_deferred_start;
2906         qinfo->conf.offloads = dev->data->dev_conf.rxmode.offloads;
2907 }
2908
2909 static void
2910 bnxt_txq_info_get_op(struct rte_eth_dev *dev, uint16_t queue_id,
2911         struct rte_eth_txq_info *qinfo)
2912 {
2913         struct bnxt *bp = dev->data->dev_private;
2914         struct bnxt_tx_queue *txq;
2915
2916         if (is_bnxt_in_error(bp))
2917                 return;
2918
2919         txq = dev->data->tx_queues[queue_id];
2920
2921         qinfo->nb_desc = txq->nb_tx_desc;
2922
2923         qinfo->conf.tx_thresh.pthresh = txq->pthresh;
2924         qinfo->conf.tx_thresh.hthresh = txq->hthresh;
2925         qinfo->conf.tx_thresh.wthresh = txq->wthresh;
2926
2927         qinfo->conf.tx_free_thresh = txq->tx_free_thresh;
2928         qinfo->conf.tx_rs_thresh = 0;
2929         qinfo->conf.tx_deferred_start = txq->tx_deferred_start;
2930         qinfo->conf.offloads = txq->offloads;
2931 }
2932
2933 static const struct {
2934         eth_rx_burst_t pkt_burst;
2935         const char *info;
2936 } bnxt_rx_burst_info[] = {
2937         {bnxt_recv_pkts,                "Scalar"},
2938 #if defined(RTE_ARCH_X86)
2939         {bnxt_recv_pkts_vec,            "Vector SSE"},
2940 #endif
2941 #if defined(RTE_ARCH_X86) && defined(CC_AVX2_SUPPORT)
2942         {bnxt_recv_pkts_vec_avx2,       "Vector AVX2"},
2943 #endif
2944 #if defined(RTE_ARCH_ARM64)
2945         {bnxt_recv_pkts_vec,            "Vector Neon"},
2946 #endif
2947 };
2948
2949 static int
2950 bnxt_rx_burst_mode_get(struct rte_eth_dev *dev, __rte_unused uint16_t queue_id,
2951                        struct rte_eth_burst_mode *mode)
2952 {
2953         eth_rx_burst_t pkt_burst = dev->rx_pkt_burst;
2954         size_t i;
2955
2956         for (i = 0; i < RTE_DIM(bnxt_rx_burst_info); i++) {
2957                 if (pkt_burst == bnxt_rx_burst_info[i].pkt_burst) {
2958                         snprintf(mode->info, sizeof(mode->info), "%s",
2959                                  bnxt_rx_burst_info[i].info);
2960                         return 0;
2961                 }
2962         }
2963
2964         return -EINVAL;
2965 }
2966
2967 static const struct {
2968         eth_tx_burst_t pkt_burst;
2969         const char *info;
2970 } bnxt_tx_burst_info[] = {
2971         {bnxt_xmit_pkts,                "Scalar"},
2972 #if defined(RTE_ARCH_X86)
2973         {bnxt_xmit_pkts_vec,            "Vector SSE"},
2974 #endif
2975 #if defined(RTE_ARCH_X86) && defined(CC_AVX2_SUPPORT)
2976         {bnxt_xmit_pkts_vec_avx2,       "Vector AVX2"},
2977 #endif
2978 #if defined(RTE_ARCH_ARM64)
2979         {bnxt_xmit_pkts_vec,            "Vector Neon"},
2980 #endif
2981 };
2982
2983 static int
2984 bnxt_tx_burst_mode_get(struct rte_eth_dev *dev, __rte_unused uint16_t queue_id,
2985                        struct rte_eth_burst_mode *mode)
2986 {
2987         eth_tx_burst_t pkt_burst = dev->tx_pkt_burst;
2988         size_t i;
2989
2990         for (i = 0; i < RTE_DIM(bnxt_tx_burst_info); i++) {
2991                 if (pkt_burst == bnxt_tx_burst_info[i].pkt_burst) {
2992                         snprintf(mode->info, sizeof(mode->info), "%s",
2993                                  bnxt_tx_burst_info[i].info);
2994                         return 0;
2995                 }
2996         }
2997
2998         return -EINVAL;
2999 }
3000
3001 int bnxt_mtu_set_op(struct rte_eth_dev *eth_dev, uint16_t new_mtu)
3002 {
3003         struct bnxt *bp = eth_dev->data->dev_private;
3004         uint32_t new_pkt_size;
3005         uint32_t rc = 0;
3006         uint32_t i;
3007
3008         rc = is_bnxt_in_error(bp);
3009         if (rc)
3010                 return rc;
3011
3012         /* Exit if receive queues are not configured yet */
3013         if (!eth_dev->data->nb_rx_queues)
3014                 return rc;
3015
3016         new_pkt_size = new_mtu + RTE_ETHER_HDR_LEN + RTE_ETHER_CRC_LEN +
3017                        VLAN_TAG_SIZE * BNXT_NUM_VLANS;
3018
3019         /*
3020          * Disallow any MTU change that would require scattered receive support
3021          * if it is not already enabled.
3022          */
3023         if (eth_dev->data->dev_started &&
3024             !eth_dev->data->scattered_rx &&
3025             (new_pkt_size >
3026              eth_dev->data->min_rx_buf_size - RTE_PKTMBUF_HEADROOM)) {
3027                 PMD_DRV_LOG(ERR,
3028                             "MTU change would require scattered rx support. ");
3029                 PMD_DRV_LOG(ERR, "Stop port before changing MTU.\n");
3030                 return -EINVAL;
3031         }
3032
3033         if (new_mtu > RTE_ETHER_MTU) {
3034                 bp->flags |= BNXT_FLAG_JUMBO;
3035                 bp->eth_dev->data->dev_conf.rxmode.offloads |=
3036                         DEV_RX_OFFLOAD_JUMBO_FRAME;
3037         } else {
3038                 bp->eth_dev->data->dev_conf.rxmode.offloads &=
3039                         ~DEV_RX_OFFLOAD_JUMBO_FRAME;
3040                 bp->flags &= ~BNXT_FLAG_JUMBO;
3041         }
3042
3043         /* Is there a change in mtu setting? */
3044         if (eth_dev->data->dev_conf.rxmode.max_rx_pkt_len == new_pkt_size)
3045                 return rc;
3046
3047         for (i = 0; i < bp->nr_vnics; i++) {
3048                 struct bnxt_vnic_info *vnic = &bp->vnic_info[i];
3049                 uint16_t size = 0;
3050
3051                 vnic->mru = BNXT_VNIC_MRU(new_mtu);
3052                 rc = bnxt_hwrm_vnic_cfg(bp, vnic);
3053                 if (rc)
3054                         break;
3055
3056                 size = rte_pktmbuf_data_room_size(bp->rx_queues[0]->mb_pool);
3057                 size -= RTE_PKTMBUF_HEADROOM;
3058
3059                 if (size < new_mtu) {
3060                         rc = bnxt_hwrm_vnic_plcmode_cfg(bp, vnic);
3061                         if (rc)
3062                                 return rc;
3063                 }
3064         }
3065
3066         if (!rc)
3067                 eth_dev->data->dev_conf.rxmode.max_rx_pkt_len = new_pkt_size;
3068
3069         PMD_DRV_LOG(INFO, "New MTU is %d\n", new_mtu);
3070
3071         return rc;
3072 }
3073
3074 static int
3075 bnxt_vlan_pvid_set_op(struct rte_eth_dev *dev, uint16_t pvid, int on)
3076 {
3077         struct bnxt *bp = dev->data->dev_private;
3078         uint16_t vlan = bp->vlan;
3079         int rc;
3080
3081         rc = is_bnxt_in_error(bp);
3082         if (rc)
3083                 return rc;
3084
3085         if (!BNXT_SINGLE_PF(bp)) {
3086                 PMD_DRV_LOG(ERR, "PVID cannot be modified on VF or on shared PF\n");
3087                 return -ENOTSUP;
3088         }
3089         bp->vlan = on ? pvid : 0;
3090
3091         rc = bnxt_hwrm_set_default_vlan(bp, 0, 0);
3092         if (rc)
3093                 bp->vlan = vlan;
3094         return rc;
3095 }
3096
3097 static int
3098 bnxt_dev_led_on_op(struct rte_eth_dev *dev)
3099 {
3100         struct bnxt *bp = dev->data->dev_private;
3101         int rc;
3102
3103         rc = is_bnxt_in_error(bp);
3104         if (rc)
3105                 return rc;
3106
3107         return bnxt_hwrm_port_led_cfg(bp, true);
3108 }
3109
3110 static int
3111 bnxt_dev_led_off_op(struct rte_eth_dev *dev)
3112 {
3113         struct bnxt *bp = dev->data->dev_private;
3114         int rc;
3115
3116         rc = is_bnxt_in_error(bp);
3117         if (rc)
3118                 return rc;
3119
3120         return bnxt_hwrm_port_led_cfg(bp, false);
3121 }
3122
3123 static uint32_t
3124 bnxt_rx_queue_count_op(struct rte_eth_dev *dev, uint16_t rx_queue_id)
3125 {
3126         struct bnxt *bp = (struct bnxt *)dev->data->dev_private;
3127         struct bnxt_cp_ring_info *cpr;
3128         uint32_t desc = 0, raw_cons;
3129         struct bnxt_rx_queue *rxq;
3130         struct rx_pkt_cmpl *rxcmp;
3131         int rc;
3132
3133         rc = is_bnxt_in_error(bp);
3134         if (rc)
3135                 return rc;
3136
3137         rxq = dev->data->rx_queues[rx_queue_id];
3138         cpr = rxq->cp_ring;
3139         raw_cons = cpr->cp_raw_cons;
3140
3141         while (1) {
3142                 uint32_t agg_cnt, cons, cmpl_type;
3143
3144                 cons = RING_CMP(cpr->cp_ring_struct, raw_cons);
3145                 rxcmp = (struct rx_pkt_cmpl *)&cpr->cp_desc_ring[cons];
3146
3147                 if (!CMP_VALID(rxcmp, raw_cons, cpr->cp_ring_struct))
3148                         break;
3149
3150                 cmpl_type = CMP_TYPE(rxcmp);
3151
3152                 switch (cmpl_type) {
3153                 case CMPL_BASE_TYPE_RX_L2:
3154                 case CMPL_BASE_TYPE_RX_L2_V2:
3155                         agg_cnt = BNXT_RX_L2_AGG_BUFS(rxcmp);
3156                         raw_cons = raw_cons + CMP_LEN(cmpl_type) + agg_cnt;
3157                         desc++;
3158                         break;
3159
3160                 case CMPL_BASE_TYPE_RX_TPA_END:
3161                         if (BNXT_CHIP_P5(rxq->bp)) {
3162                                 struct rx_tpa_v2_end_cmpl_hi *p5_tpa_end;
3163
3164                                 p5_tpa_end = (void *)rxcmp;
3165                                 agg_cnt = BNXT_TPA_END_AGG_BUFS_TH(p5_tpa_end);
3166                         } else {
3167                                 struct rx_tpa_end_cmpl *tpa_end;
3168
3169                                 tpa_end = (void *)rxcmp;
3170                                 agg_cnt = BNXT_TPA_END_AGG_BUFS(tpa_end);
3171                         }
3172
3173                         raw_cons = raw_cons + CMP_LEN(cmpl_type) + agg_cnt;
3174                         desc++;
3175                         break;
3176
3177                 default:
3178                         raw_cons += CMP_LEN(cmpl_type);
3179                 }
3180         }
3181
3182         return desc;
3183 }
3184
3185 static int
3186 bnxt_rx_descriptor_status_op(void *rx_queue, uint16_t offset)
3187 {
3188         struct bnxt_rx_queue *rxq = rx_queue;
3189         struct bnxt_cp_ring_info *cpr;
3190         struct bnxt_rx_ring_info *rxr;
3191         uint32_t desc, raw_cons;
3192         struct bnxt *bp = rxq->bp;
3193         struct rx_pkt_cmpl *rxcmp;
3194         int rc;
3195
3196         rc = is_bnxt_in_error(bp);
3197         if (rc)
3198                 return rc;
3199
3200         if (offset >= rxq->nb_rx_desc)
3201                 return -EINVAL;
3202
3203         rxr = rxq->rx_ring;
3204         cpr = rxq->cp_ring;
3205
3206         /*
3207          * For the vector receive case, the completion at the requested
3208          * offset can be indexed directly.
3209          */
3210 #if defined(RTE_ARCH_X86) || defined(RTE_ARCH_ARM64)
3211         if (bp->flags & BNXT_FLAG_RX_VECTOR_PKT_MODE) {
3212                 struct rx_pkt_cmpl *rxcmp;
3213                 uint32_t cons;
3214
3215                 /* Check status of completion descriptor. */
3216                 raw_cons = cpr->cp_raw_cons +
3217                            offset * CMP_LEN(CMPL_BASE_TYPE_RX_L2);
3218                 cons = RING_CMP(cpr->cp_ring_struct, raw_cons);
3219                 rxcmp = (struct rx_pkt_cmpl *)&cpr->cp_desc_ring[cons];
3220
3221                 if (CMP_VALID(rxcmp, raw_cons, cpr->cp_ring_struct))
3222                         return RTE_ETH_RX_DESC_DONE;
3223
3224                 /* Check whether rx desc has an mbuf attached. */
3225                 cons = RING_CMP(rxr->rx_ring_struct, raw_cons / 2);
3226                 if (cons >= rxq->rxrearm_start &&
3227                     cons < rxq->rxrearm_start + rxq->rxrearm_nb) {
3228                         return RTE_ETH_RX_DESC_UNAVAIL;
3229                 }
3230
3231                 return RTE_ETH_RX_DESC_AVAIL;
3232         }
3233 #endif
3234
3235         /*
3236          * For the non-vector receive case, scan the completion ring to
3237          * locate the completion descriptor for the requested offset.
3238          */
3239         raw_cons = cpr->cp_raw_cons;
3240         desc = 0;
3241         while (1) {
3242                 uint32_t agg_cnt, cons, cmpl_type;
3243
3244                 cons = RING_CMP(cpr->cp_ring_struct, raw_cons);
3245                 rxcmp = (struct rx_pkt_cmpl *)&cpr->cp_desc_ring[cons];
3246
3247                 if (!CMP_VALID(rxcmp, raw_cons, cpr->cp_ring_struct))
3248                         break;
3249
3250                 cmpl_type = CMP_TYPE(rxcmp);
3251
3252                 switch (cmpl_type) {
3253                 case CMPL_BASE_TYPE_RX_L2:
3254                 case CMPL_BASE_TYPE_RX_L2_V2:
3255                         if (desc == offset) {
3256                                 cons = rxcmp->opaque;
3257                                 if (rxr->rx_buf_ring[cons])
3258                                         return RTE_ETH_RX_DESC_DONE;
3259                                 else
3260                                         return RTE_ETH_RX_DESC_UNAVAIL;
3261                         }
3262                         agg_cnt = BNXT_RX_L2_AGG_BUFS(rxcmp);
3263                         raw_cons = raw_cons + CMP_LEN(cmpl_type) + agg_cnt;
3264                         desc++;
3265                         break;
3266
3267                 case CMPL_BASE_TYPE_RX_TPA_END:
3268                         if (desc == offset)
3269                                 return RTE_ETH_RX_DESC_DONE;
3270
3271                         if (BNXT_CHIP_P5(rxq->bp)) {
3272                                 struct rx_tpa_v2_end_cmpl_hi *p5_tpa_end;
3273
3274                                 p5_tpa_end = (void *)rxcmp;
3275                                 agg_cnt = BNXT_TPA_END_AGG_BUFS_TH(p5_tpa_end);
3276                         } else {
3277                                 struct rx_tpa_end_cmpl *tpa_end;
3278
3279                                 tpa_end = (void *)rxcmp;
3280                                 agg_cnt = BNXT_TPA_END_AGG_BUFS(tpa_end);
3281                         }
3282
3283                         raw_cons = raw_cons + CMP_LEN(cmpl_type) + agg_cnt;
3284                         desc++;
3285                         break;
3286
3287                 default:
3288                         raw_cons += CMP_LEN(cmpl_type);
3289                 }
3290         }
3291
3292         return RTE_ETH_RX_DESC_AVAIL;
3293 }
3294
3295 static int
3296 bnxt_tx_descriptor_status_op(void *tx_queue, uint16_t offset)
3297 {
3298         struct bnxt_tx_queue *txq = (struct bnxt_tx_queue *)tx_queue;
3299         struct bnxt_cp_ring_info *cpr = txq->cp_ring;
3300         uint32_t ring_mask, raw_cons, nb_tx_pkts = 0;
3301         struct bnxt_ring *cp_ring_struct;
3302         struct cmpl_base *cp_desc_ring;
3303         int rc;
3304
3305         rc = is_bnxt_in_error(txq->bp);
3306         if (rc)
3307                 return rc;
3308
3309         if (offset >= txq->nb_tx_desc)
3310                 return -EINVAL;
3311
3312         /* Return "desc done" if descriptor is available for use. */
3313         if (bnxt_tx_bds_in_hw(txq) <= offset)
3314                 return RTE_ETH_TX_DESC_DONE;
3315
3316         raw_cons = cpr->cp_raw_cons;
3317         cp_desc_ring = cpr->cp_desc_ring;
3318         cp_ring_struct = cpr->cp_ring_struct;
3319         ring_mask = cpr->cp_ring_struct->ring_mask;
3320
3321         /* Check to see if hw has posted a completion for the descriptor. */
3322         while (1) {
3323                 struct tx_cmpl *txcmp;
3324                 uint32_t cons;
3325
3326                 cons = RING_CMPL(ring_mask, raw_cons);
3327                 txcmp = (struct tx_cmpl *)&cp_desc_ring[cons];
3328
3329                 if (!CMP_VALID(txcmp, raw_cons, cp_ring_struct))
3330                         break;
3331
3332                 if (CMP_TYPE(txcmp) == TX_CMPL_TYPE_TX_L2)
3333                         nb_tx_pkts += rte_le_to_cpu_32(txcmp->opaque);
3334
3335                 if (nb_tx_pkts > offset)
3336                         return RTE_ETH_TX_DESC_DONE;
3337
3338                 raw_cons = NEXT_RAW_CMP(raw_cons);
3339         }
3340
3341         /* Descriptor is pending transmit, not yet completed by hardware. */
3342         return RTE_ETH_TX_DESC_FULL;
3343 }
3344
3345 int
3346 bnxt_flow_ops_get_op(struct rte_eth_dev *dev,
3347                      const struct rte_flow_ops **ops)
3348 {
3349         struct bnxt *bp = dev->data->dev_private;
3350         int ret = 0;
3351
3352         if (!bp)
3353                 return -EIO;
3354
3355         if (BNXT_ETH_DEV_IS_REPRESENTOR(dev)) {
3356                 struct bnxt_representor *vfr = dev->data->dev_private;
3357                 bp = vfr->parent_dev->data->dev_private;
3358                 /* parent is deleted while children are still valid */
3359                 if (!bp) {
3360                         PMD_DRV_LOG(DEBUG, "BNXT Port:%d VFR Error\n",
3361                                     dev->data->port_id);
3362                         return -EIO;
3363                 }
3364         }
3365
3366         ret = is_bnxt_in_error(bp);
3367         if (ret)
3368                 return ret;
3369
3370         /* PMD supports thread-safe flow operations.  rte_flow API
3371          * functions can avoid mutex for multi-thread safety.
3372          */
3373         dev->data->dev_flags |= RTE_ETH_DEV_FLOW_OPS_THREAD_SAFE;
3374
3375         if (BNXT_TRUFLOW_EN(bp))
3376                 *ops = &bnxt_ulp_rte_flow_ops;
3377         else
3378                 *ops = &bnxt_flow_ops;
3379
3380         return ret;
3381 }
3382
3383 static const uint32_t *
3384 bnxt_dev_supported_ptypes_get_op(struct rte_eth_dev *dev)
3385 {
3386         static const uint32_t ptypes[] = {
3387                 RTE_PTYPE_L2_ETHER_VLAN,
3388                 RTE_PTYPE_L3_IPV4_EXT_UNKNOWN,
3389                 RTE_PTYPE_L3_IPV6_EXT_UNKNOWN,
3390                 RTE_PTYPE_L4_ICMP,
3391                 RTE_PTYPE_L4_TCP,
3392                 RTE_PTYPE_L4_UDP,
3393                 RTE_PTYPE_INNER_L3_IPV4_EXT_UNKNOWN,
3394                 RTE_PTYPE_INNER_L3_IPV6_EXT_UNKNOWN,
3395                 RTE_PTYPE_INNER_L4_ICMP,
3396                 RTE_PTYPE_INNER_L4_TCP,
3397                 RTE_PTYPE_INNER_L4_UDP,
3398                 RTE_PTYPE_UNKNOWN
3399         };
3400
3401         if (!dev->rx_pkt_burst)
3402                 return NULL;
3403
3404         return ptypes;
3405 }
3406
3407 static int bnxt_map_regs(struct bnxt *bp, uint32_t *reg_arr, int count,
3408                          int reg_win)
3409 {
3410         uint32_t reg_base = *reg_arr & 0xfffff000;
3411         uint32_t win_off;
3412         int i;
3413
3414         for (i = 0; i < count; i++) {
3415                 if ((reg_arr[i] & 0xfffff000) != reg_base)
3416                         return -ERANGE;
3417         }
3418         win_off = BNXT_GRCPF_REG_WINDOW_BASE_OUT + (reg_win - 1) * 4;
3419         rte_write32(reg_base, (uint8_t *)bp->bar0 + win_off);
3420         return 0;
3421 }
3422
3423 static int bnxt_map_ptp_regs(struct bnxt *bp)
3424 {
3425         struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3426         uint32_t *reg_arr;
3427         int rc, i;
3428
3429         reg_arr = ptp->rx_regs;
3430         rc = bnxt_map_regs(bp, reg_arr, BNXT_PTP_RX_REGS, 5);
3431         if (rc)
3432                 return rc;
3433
3434         reg_arr = ptp->tx_regs;
3435         rc = bnxt_map_regs(bp, reg_arr, BNXT_PTP_TX_REGS, 6);
3436         if (rc)
3437                 return rc;
3438
3439         for (i = 0; i < BNXT_PTP_RX_REGS; i++)
3440                 ptp->rx_mapped_regs[i] = 0x5000 + (ptp->rx_regs[i] & 0xfff);
3441
3442         for (i = 0; i < BNXT_PTP_TX_REGS; i++)
3443                 ptp->tx_mapped_regs[i] = 0x6000 + (ptp->tx_regs[i] & 0xfff);
3444
3445         return 0;
3446 }
3447
3448 static void bnxt_unmap_ptp_regs(struct bnxt *bp)
3449 {
3450         rte_write32(0, (uint8_t *)bp->bar0 +
3451                          BNXT_GRCPF_REG_WINDOW_BASE_OUT + 16);
3452         rte_write32(0, (uint8_t *)bp->bar0 +
3453                          BNXT_GRCPF_REG_WINDOW_BASE_OUT + 20);
3454 }
3455
3456 static uint64_t bnxt_cc_read(struct bnxt *bp)
3457 {
3458         uint64_t ns;
3459
3460         ns = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3461                               BNXT_GRCPF_REG_SYNC_TIME));
3462         ns |= (uint64_t)(rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3463                                           BNXT_GRCPF_REG_SYNC_TIME + 4))) << 32;
3464         return ns;
3465 }
3466
3467 static int bnxt_get_tx_ts(struct bnxt *bp, uint64_t *ts)
3468 {
3469         struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3470         uint32_t fifo;
3471
3472         fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3473                                 ptp->tx_mapped_regs[BNXT_PTP_TX_FIFO]));
3474         if (fifo & BNXT_PTP_TX_FIFO_EMPTY)
3475                 return -EAGAIN;
3476
3477         fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3478                                 ptp->tx_mapped_regs[BNXT_PTP_TX_FIFO]));
3479         *ts = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3480                                 ptp->tx_mapped_regs[BNXT_PTP_TX_TS_L]));
3481         *ts |= (uint64_t)rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3482                                 ptp->tx_mapped_regs[BNXT_PTP_TX_TS_H])) << 32;
3483         rte_read32((uint8_t *)bp->bar0 + ptp->tx_mapped_regs[BNXT_PTP_TX_SEQ]);
3484
3485         return 0;
3486 }
3487
3488 static int bnxt_clr_rx_ts(struct bnxt *bp, uint64_t *last_ts)
3489 {
3490         struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3491         struct bnxt_pf_info *pf = bp->pf;
3492         uint16_t port_id;
3493         int i = 0;
3494         uint32_t fifo;
3495
3496         if (!ptp || (bp->flags & BNXT_FLAG_CHIP_P5))
3497                 return -EINVAL;
3498
3499         port_id = pf->port_id;
3500         fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3501                                 ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO]));
3502         while ((fifo & BNXT_PTP_RX_FIFO_PENDING) && (i < BNXT_PTP_RX_PND_CNT)) {
3503                 rte_write32(1 << port_id, (uint8_t *)bp->bar0 +
3504                             ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO_ADV]);
3505                 fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3506                                         ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO]));
3507                 *last_ts = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3508                                         ptp->rx_mapped_regs[BNXT_PTP_RX_TS_L]));
3509                 *last_ts |= (uint64_t)rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3510                                         ptp->rx_mapped_regs[BNXT_PTP_RX_TS_H])) << 32;
3511                 i++;
3512         }
3513
3514         if (i >= BNXT_PTP_RX_PND_CNT)
3515                 return -EBUSY;
3516
3517         return 0;
3518 }
3519
3520 static int bnxt_get_rx_ts(struct bnxt *bp, uint64_t *ts)
3521 {
3522         struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3523         struct bnxt_pf_info *pf = bp->pf;
3524         uint16_t port_id;
3525         uint32_t fifo;
3526
3527         fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3528                                 ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO]));
3529         if (!(fifo & BNXT_PTP_RX_FIFO_PENDING))
3530                 return -EAGAIN;
3531
3532         port_id = pf->port_id;
3533         rte_write32(1 << port_id, (uint8_t *)bp->bar0 +
3534                ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO_ADV]);
3535
3536         fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3537                                    ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO]));
3538         if (fifo & BNXT_PTP_RX_FIFO_PENDING)
3539                 return bnxt_clr_rx_ts(bp, ts);
3540
3541         *ts = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3542                                 ptp->rx_mapped_regs[BNXT_PTP_RX_TS_L]));
3543         *ts |= (uint64_t)rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3544                                 ptp->rx_mapped_regs[BNXT_PTP_RX_TS_H])) << 32;
3545
3546         return 0;
3547 }
3548
3549 static int
3550 bnxt_timesync_write_time(struct rte_eth_dev *dev, const struct timespec *ts)
3551 {
3552         uint64_t ns;
3553         struct bnxt *bp = dev->data->dev_private;
3554         struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3555
3556         if (!ptp)
3557                 return -ENOTSUP;
3558
3559         ns = rte_timespec_to_ns(ts);
3560         /* Set the timecounters to a new value. */
3561         ptp->tc.nsec = ns;
3562         ptp->tx_tstamp_tc.nsec = ns;
3563         ptp->rx_tstamp_tc.nsec = ns;
3564
3565         return 0;
3566 }
3567
3568 static int
3569 bnxt_timesync_read_time(struct rte_eth_dev *dev, struct timespec *ts)
3570 {
3571         struct bnxt *bp = dev->data->dev_private;
3572         struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3573         uint64_t ns, systime_cycles = 0;
3574         int rc = 0;
3575
3576         if (!ptp)
3577                 return -ENOTSUP;
3578
3579         if (BNXT_CHIP_P5(bp))
3580                 rc = bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_CURRENT_TIME,
3581                                              &systime_cycles);
3582         else
3583                 systime_cycles = bnxt_cc_read(bp);
3584
3585         ns = rte_timecounter_update(&ptp->tc, systime_cycles);
3586         *ts = rte_ns_to_timespec(ns);
3587
3588         return rc;
3589 }
3590 static int
3591 bnxt_timesync_enable(struct rte_eth_dev *dev)
3592 {
3593         struct bnxt *bp = dev->data->dev_private;
3594         struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3595         uint32_t shift = 0;
3596         int rc;
3597
3598         if (!ptp)
3599                 return -ENOTSUP;
3600
3601         ptp->rx_filter = 1;
3602         ptp->tx_tstamp_en = 1;
3603         ptp->rxctl = BNXT_PTP_MSG_EVENTS;
3604
3605         rc = bnxt_hwrm_ptp_cfg(bp);
3606         if (rc)
3607                 return rc;
3608
3609         memset(&ptp->tc, 0, sizeof(struct rte_timecounter));
3610         memset(&ptp->rx_tstamp_tc, 0, sizeof(struct rte_timecounter));
3611         memset(&ptp->tx_tstamp_tc, 0, sizeof(struct rte_timecounter));
3612
3613         ptp->tc.cc_mask = BNXT_CYCLECOUNTER_MASK;
3614         ptp->tc.cc_shift = shift;
3615         ptp->tc.nsec_mask = (1ULL << shift) - 1;
3616
3617         ptp->rx_tstamp_tc.cc_mask = BNXT_CYCLECOUNTER_MASK;
3618         ptp->rx_tstamp_tc.cc_shift = shift;
3619         ptp->rx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
3620
3621         ptp->tx_tstamp_tc.cc_mask = BNXT_CYCLECOUNTER_MASK;
3622         ptp->tx_tstamp_tc.cc_shift = shift;
3623         ptp->tx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
3624
3625         if (!BNXT_CHIP_P5(bp))
3626                 bnxt_map_ptp_regs(bp);
3627         else
3628                 rc = bnxt_ptp_start(bp);
3629
3630         return rc;
3631 }
3632
3633 static int
3634 bnxt_timesync_disable(struct rte_eth_dev *dev)
3635 {
3636         struct bnxt *bp = dev->data->dev_private;
3637         struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3638
3639         if (!ptp)
3640                 return -ENOTSUP;
3641
3642         ptp->rx_filter = 0;
3643         ptp->tx_tstamp_en = 0;
3644         ptp->rxctl = 0;
3645
3646         bnxt_hwrm_ptp_cfg(bp);
3647
3648         if (!BNXT_CHIP_P5(bp))
3649                 bnxt_unmap_ptp_regs(bp);
3650         else
3651                 bnxt_ptp_stop(bp);
3652
3653         return 0;
3654 }
3655
3656 static int
3657 bnxt_timesync_read_rx_timestamp(struct rte_eth_dev *dev,
3658                                  struct timespec *timestamp,
3659                                  uint32_t flags __rte_unused)
3660 {
3661         struct bnxt *bp = dev->data->dev_private;
3662         struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3663         uint64_t rx_tstamp_cycles = 0;
3664         uint64_t ns;
3665
3666         if (!ptp)
3667                 return -ENOTSUP;
3668
3669         if (BNXT_CHIP_P5(bp))
3670                 rx_tstamp_cycles = ptp->rx_timestamp;
3671         else
3672                 bnxt_get_rx_ts(bp, &rx_tstamp_cycles);
3673
3674         ns = rte_timecounter_update(&ptp->rx_tstamp_tc, rx_tstamp_cycles);
3675         *timestamp = rte_ns_to_timespec(ns);
3676         return  0;
3677 }
3678
3679 static int
3680 bnxt_timesync_read_tx_timestamp(struct rte_eth_dev *dev,
3681                                  struct timespec *timestamp)
3682 {
3683         struct bnxt *bp = dev->data->dev_private;
3684         struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3685         uint64_t tx_tstamp_cycles = 0;
3686         uint64_t ns;
3687         int rc = 0;
3688
3689         if (!ptp)
3690                 return -ENOTSUP;
3691
3692         if (BNXT_CHIP_P5(bp))
3693                 rc = bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_PATH_TX,
3694                                              &tx_tstamp_cycles);
3695         else
3696                 rc = bnxt_get_tx_ts(bp, &tx_tstamp_cycles);
3697
3698         ns = rte_timecounter_update(&ptp->tx_tstamp_tc, tx_tstamp_cycles);
3699         *timestamp = rte_ns_to_timespec(ns);
3700
3701         return rc;
3702 }
3703
3704 static int
3705 bnxt_timesync_adjust_time(struct rte_eth_dev *dev, int64_t delta)
3706 {
3707         struct bnxt *bp = dev->data->dev_private;
3708         struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3709
3710         if (!ptp)
3711                 return -ENOTSUP;
3712
3713         ptp->tc.nsec += delta;
3714         ptp->tx_tstamp_tc.nsec += delta;
3715         ptp->rx_tstamp_tc.nsec += delta;
3716
3717         return 0;
3718 }
3719
3720 static int
3721 bnxt_get_eeprom_length_op(struct rte_eth_dev *dev)
3722 {
3723         struct bnxt *bp = dev->data->dev_private;
3724         int rc;
3725         uint32_t dir_entries;
3726         uint32_t entry_length;
3727
3728         rc = is_bnxt_in_error(bp);
3729         if (rc)
3730                 return rc;
3731
3732         PMD_DRV_LOG(INFO, PCI_PRI_FMT "\n",
3733                     bp->pdev->addr.domain, bp->pdev->addr.bus,
3734                     bp->pdev->addr.devid, bp->pdev->addr.function);
3735
3736         rc = bnxt_hwrm_nvm_get_dir_info(bp, &dir_entries, &entry_length);
3737         if (rc != 0)
3738                 return rc;
3739
3740         return dir_entries * entry_length;
3741 }
3742
3743 static int
3744 bnxt_get_eeprom_op(struct rte_eth_dev *dev,
3745                 struct rte_dev_eeprom_info *in_eeprom)
3746 {
3747         struct bnxt *bp = dev->data->dev_private;
3748         uint32_t index;
3749         uint32_t offset;
3750         int rc;
3751
3752         rc = is_bnxt_in_error(bp);
3753         if (rc)
3754                 return rc;
3755
3756         PMD_DRV_LOG(INFO, PCI_PRI_FMT " in_eeprom->offset = %d len = %d\n",
3757                     bp->pdev->addr.domain, bp->pdev->addr.bus,
3758                     bp->pdev->addr.devid, bp->pdev->addr.function,
3759                     in_eeprom->offset, in_eeprom->length);
3760
3761         if (in_eeprom->offset == 0) /* special offset value to get directory */
3762                 return bnxt_get_nvram_directory(bp, in_eeprom->length,
3763                                                 in_eeprom->data);
3764
3765         index = in_eeprom->offset >> 24;
3766         offset = in_eeprom->offset & 0xffffff;
3767
3768         if (index != 0)
3769                 return bnxt_hwrm_get_nvram_item(bp, index - 1, offset,
3770                                            in_eeprom->length, in_eeprom->data);
3771
3772         return 0;
3773 }
3774
3775 static bool bnxt_dir_type_is_ape_bin_format(uint16_t dir_type)
3776 {
3777         switch (dir_type) {
3778         case BNX_DIR_TYPE_CHIMP_PATCH:
3779         case BNX_DIR_TYPE_BOOTCODE:
3780         case BNX_DIR_TYPE_BOOTCODE_2:
3781         case BNX_DIR_TYPE_APE_FW:
3782         case BNX_DIR_TYPE_APE_PATCH:
3783         case BNX_DIR_TYPE_KONG_FW:
3784         case BNX_DIR_TYPE_KONG_PATCH:
3785         case BNX_DIR_TYPE_BONO_FW:
3786         case BNX_DIR_TYPE_BONO_PATCH:
3787                 /* FALLTHROUGH */
3788                 return true;
3789         }
3790
3791         return false;
3792 }
3793
3794 static bool bnxt_dir_type_is_other_exec_format(uint16_t dir_type)
3795 {
3796         switch (dir_type) {
3797         case BNX_DIR_TYPE_AVS:
3798         case BNX_DIR_TYPE_EXP_ROM_MBA:
3799         case BNX_DIR_TYPE_PCIE:
3800         case BNX_DIR_TYPE_TSCF_UCODE:
3801         case BNX_DIR_TYPE_EXT_PHY:
3802         case BNX_DIR_TYPE_CCM:
3803         case BNX_DIR_TYPE_ISCSI_BOOT:
3804         case BNX_DIR_TYPE_ISCSI_BOOT_IPV6:
3805         case BNX_DIR_TYPE_ISCSI_BOOT_IPV4N6:
3806                 /* FALLTHROUGH */
3807                 return true;
3808         }
3809
3810         return false;
3811 }
3812
3813 static bool bnxt_dir_type_is_executable(uint16_t dir_type)
3814 {
3815         return bnxt_dir_type_is_ape_bin_format(dir_type) ||
3816                 bnxt_dir_type_is_other_exec_format(dir_type);
3817 }
3818
3819 static int
3820 bnxt_set_eeprom_op(struct rte_eth_dev *dev,
3821                 struct rte_dev_eeprom_info *in_eeprom)
3822 {
3823         struct bnxt *bp = dev->data->dev_private;
3824         uint8_t index, dir_op;
3825         uint16_t type, ext, ordinal, attr;
3826         int rc;
3827
3828         rc = is_bnxt_in_error(bp);
3829         if (rc)
3830                 return rc;
3831
3832         PMD_DRV_LOG(INFO, PCI_PRI_FMT " in_eeprom->offset = %d len = %d\n",
3833                     bp->pdev->addr.domain, bp->pdev->addr.bus,
3834                     bp->pdev->addr.devid, bp->pdev->addr.function,
3835                     in_eeprom->offset, in_eeprom->length);
3836
3837         if (!BNXT_PF(bp)) {
3838                 PMD_DRV_LOG(ERR, "NVM write not supported from a VF\n");
3839                 return -EINVAL;
3840         }
3841
3842         type = in_eeprom->magic >> 16;
3843
3844         if (type == 0xffff) { /* special value for directory operations */
3845                 index = in_eeprom->magic & 0xff;
3846                 dir_op = in_eeprom->magic >> 8;
3847                 if (index == 0)
3848                         return -EINVAL;
3849                 switch (dir_op) {
3850                 case 0x0e: /* erase */
3851                         if (in_eeprom->offset != ~in_eeprom->magic)
3852                                 return -EINVAL;
3853                         return bnxt_hwrm_erase_nvram_directory(bp, index - 1);
3854                 default:
3855                         return -EINVAL;
3856                 }
3857         }
3858
3859         /* Create or re-write an NVM item: */
3860         if (bnxt_dir_type_is_executable(type) == true)
3861                 return -EOPNOTSUPP;
3862         ext = in_eeprom->magic & 0xffff;
3863         ordinal = in_eeprom->offset >> 16;
3864         attr = in_eeprom->offset & 0xffff;
3865
3866         return bnxt_hwrm_flash_nvram(bp, type, ordinal, ext, attr,
3867                                      in_eeprom->data, in_eeprom->length);
3868 }
3869
3870 static int bnxt_get_module_info(struct rte_eth_dev *dev,
3871                                 struct rte_eth_dev_module_info *modinfo)
3872 {
3873         uint8_t module_info[SFF_DIAG_SUPPORT_OFFSET + 1];
3874         struct bnxt *bp = dev->data->dev_private;
3875         int rc;
3876
3877         /* No point in going further if phy status indicates
3878          * module is not inserted or if it is powered down or
3879          * if it is of type 10GBase-T
3880          */
3881         if (bp->link_info->module_status >
3882             HWRM_PORT_PHY_QCFG_OUTPUT_MODULE_STATUS_WARNINGMSG) {
3883                 PMD_DRV_LOG(NOTICE, "Port %u : Module is not inserted or is powered down\n",
3884                             dev->data->port_id);
3885                 return -ENOTSUP;
3886         }
3887
3888         /* This feature is not supported in older firmware versions */
3889         if (bp->hwrm_spec_code < 0x10202) {
3890                 PMD_DRV_LOG(NOTICE, "Port %u : Feature is not supported in older firmware\n",
3891                             dev->data->port_id);
3892                 return -ENOTSUP;
3893         }
3894
3895         rc = bnxt_hwrm_read_sfp_module_eeprom_info(bp, I2C_DEV_ADDR_A0, 0, 0,
3896                                                    SFF_DIAG_SUPPORT_OFFSET + 1,
3897                                                    module_info);
3898
3899         if (rc)
3900                 return rc;
3901
3902         switch (module_info[0]) {
3903         case SFF_MODULE_ID_SFP:
3904                 modinfo->type = RTE_ETH_MODULE_SFF_8472;
3905                 modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8472_LEN;
3906                 if (module_info[SFF_DIAG_SUPPORT_OFFSET] == 0)
3907                         modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8436_LEN;
3908                 break;
3909         case SFF_MODULE_ID_QSFP:
3910         case SFF_MODULE_ID_QSFP_PLUS:
3911                 modinfo->type = RTE_ETH_MODULE_SFF_8436;
3912                 modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8436_LEN;
3913                 break;
3914         case SFF_MODULE_ID_QSFP28:
3915                 modinfo->type = RTE_ETH_MODULE_SFF_8636;
3916                 modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8636_MAX_LEN;
3917                 if (module_info[SFF8636_FLATMEM_OFFSET] & SFF8636_FLATMEM_MASK)
3918                         modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8636_LEN;
3919                 break;
3920         default:
3921                 PMD_DRV_LOG(NOTICE, "Port %u : Unsupported module\n", dev->data->port_id);
3922                 return -ENOTSUP;
3923         }
3924
3925         PMD_DRV_LOG(INFO, "Port %u : modinfo->type = %d modinfo->eeprom_len = %d\n",
3926                     dev->data->port_id, modinfo->type, modinfo->eeprom_len);
3927
3928         return 0;
3929 }
3930
3931 static int bnxt_get_module_eeprom(struct rte_eth_dev *dev,
3932                                   struct rte_dev_eeprom_info *info)
3933 {
3934         uint8_t pg_addr[5] = { I2C_DEV_ADDR_A0, I2C_DEV_ADDR_A0 };
3935         uint32_t offset = info->offset, length = info->length;
3936         uint8_t module_info[SFF_DIAG_SUPPORT_OFFSET + 1];
3937         struct bnxt *bp = dev->data->dev_private;
3938         uint8_t *data = info->data;
3939         uint8_t page = offset >> 7;
3940         uint8_t max_pages = 2;
3941         uint8_t opt_pages;
3942         int rc;
3943
3944         rc = bnxt_hwrm_read_sfp_module_eeprom_info(bp, I2C_DEV_ADDR_A0, 0, 0,
3945                                                    SFF_DIAG_SUPPORT_OFFSET + 1,
3946                                                    module_info);
3947         if (rc)
3948                 return rc;
3949
3950         switch (module_info[0]) {
3951         case SFF_MODULE_ID_SFP:
3952                 module_info[SFF_DIAG_SUPPORT_OFFSET] = 0;
3953                 if (module_info[SFF_DIAG_SUPPORT_OFFSET]) {
3954                         pg_addr[2] = I2C_DEV_ADDR_A2;
3955                         pg_addr[3] = I2C_DEV_ADDR_A2;
3956                         max_pages = 4;
3957                 }
3958                 break;
3959         case SFF_MODULE_ID_QSFP28:
3960                 rc = bnxt_hwrm_read_sfp_module_eeprom_info(bp, I2C_DEV_ADDR_A0, 0,
3961                                                            SFF8636_OPT_PAGES_OFFSET,
3962                                                            1, &opt_pages);
3963                 if (rc)
3964                         return rc;
3965
3966                 if (opt_pages & SFF8636_PAGE1_MASK) {
3967                         pg_addr[2] = I2C_DEV_ADDR_A0;
3968                         max_pages = 3;
3969                 }
3970                 if (opt_pages & SFF8636_PAGE2_MASK) {
3971                         pg_addr[3] = I2C_DEV_ADDR_A0;
3972                         max_pages = 4;
3973                 }
3974                 if (~module_info[SFF8636_FLATMEM_OFFSET] & SFF8636_FLATMEM_MASK) {
3975                         pg_addr[4] = I2C_DEV_ADDR_A0;
3976                         max_pages = 5;
3977                 }
3978                 break;
3979         default:
3980                 break;
3981         }
3982
3983         memset(data, 0, length);
3984
3985         offset &= 0xff;
3986         while (length && page < max_pages) {
3987                 uint8_t raw_page = page ? page - 1 : 0;
3988                 uint16_t chunk;
3989
3990                 if (pg_addr[page] == I2C_DEV_ADDR_A2)
3991                         raw_page = 0;
3992                 else if (page)
3993                         offset |= 0x80;
3994                 chunk = RTE_MIN(length, 256 - offset);
3995
3996                 if (pg_addr[page]) {
3997                         rc = bnxt_hwrm_read_sfp_module_eeprom_info(bp, pg_addr[page],
3998                                                                    raw_page, offset,
3999                                                                    chunk, data);
4000                         if (rc)
4001                                 return rc;
4002                 }
4003
4004                 data += chunk;
4005                 length -= chunk;
4006                 offset = 0;
4007                 page += 1 + (chunk > 128);
4008         }
4009
4010         return length ? -EINVAL : 0;
4011 }
4012
4013 /*
4014  * Initialization
4015  */
4016
4017 static const struct eth_dev_ops bnxt_dev_ops = {
4018         .dev_infos_get = bnxt_dev_info_get_op,
4019         .dev_close = bnxt_dev_close_op,
4020         .dev_configure = bnxt_dev_configure_op,
4021         .dev_start = bnxt_dev_start_op,
4022         .dev_stop = bnxt_dev_stop_op,
4023         .dev_set_link_up = bnxt_dev_set_link_up_op,
4024         .dev_set_link_down = bnxt_dev_set_link_down_op,
4025         .stats_get = bnxt_stats_get_op,
4026         .stats_reset = bnxt_stats_reset_op,
4027         .rx_queue_setup = bnxt_rx_queue_setup_op,
4028         .rx_queue_release = bnxt_rx_queue_release_op,
4029         .tx_queue_setup = bnxt_tx_queue_setup_op,
4030         .tx_queue_release = bnxt_tx_queue_release_op,
4031         .rx_queue_intr_enable = bnxt_rx_queue_intr_enable_op,
4032         .rx_queue_intr_disable = bnxt_rx_queue_intr_disable_op,
4033         .reta_update = bnxt_reta_update_op,
4034         .reta_query = bnxt_reta_query_op,
4035         .rss_hash_update = bnxt_rss_hash_update_op,
4036         .rss_hash_conf_get = bnxt_rss_hash_conf_get_op,
4037         .link_update = bnxt_link_update_op,
4038         .promiscuous_enable = bnxt_promiscuous_enable_op,
4039         .promiscuous_disable = bnxt_promiscuous_disable_op,
4040         .allmulticast_enable = bnxt_allmulticast_enable_op,
4041         .allmulticast_disable = bnxt_allmulticast_disable_op,
4042         .mac_addr_add = bnxt_mac_addr_add_op,
4043         .mac_addr_remove = bnxt_mac_addr_remove_op,
4044         .flow_ctrl_get = bnxt_flow_ctrl_get_op,
4045         .flow_ctrl_set = bnxt_flow_ctrl_set_op,
4046         .udp_tunnel_port_add  = bnxt_udp_tunnel_port_add_op,
4047         .udp_tunnel_port_del  = bnxt_udp_tunnel_port_del_op,
4048         .vlan_filter_set = bnxt_vlan_filter_set_op,
4049         .vlan_offload_set = bnxt_vlan_offload_set_op,
4050         .vlan_tpid_set = bnxt_vlan_tpid_set_op,
4051         .vlan_pvid_set = bnxt_vlan_pvid_set_op,
4052         .mtu_set = bnxt_mtu_set_op,
4053         .mac_addr_set = bnxt_set_default_mac_addr_op,
4054         .xstats_get = bnxt_dev_xstats_get_op,
4055         .xstats_get_names = bnxt_dev_xstats_get_names_op,
4056         .xstats_reset = bnxt_dev_xstats_reset_op,
4057         .fw_version_get = bnxt_fw_version_get,
4058         .set_mc_addr_list = bnxt_dev_set_mc_addr_list_op,
4059         .rxq_info_get = bnxt_rxq_info_get_op,
4060         .txq_info_get = bnxt_txq_info_get_op,
4061         .rx_burst_mode_get = bnxt_rx_burst_mode_get,
4062         .tx_burst_mode_get = bnxt_tx_burst_mode_get,
4063         .dev_led_on = bnxt_dev_led_on_op,
4064         .dev_led_off = bnxt_dev_led_off_op,
4065         .rx_queue_start = bnxt_rx_queue_start,
4066         .rx_queue_stop = bnxt_rx_queue_stop,
4067         .tx_queue_start = bnxt_tx_queue_start,
4068         .tx_queue_stop = bnxt_tx_queue_stop,
4069         .flow_ops_get = bnxt_flow_ops_get_op,
4070         .dev_supported_ptypes_get = bnxt_dev_supported_ptypes_get_op,
4071         .get_eeprom_length    = bnxt_get_eeprom_length_op,
4072         .get_eeprom           = bnxt_get_eeprom_op,
4073         .set_eeprom           = bnxt_set_eeprom_op,
4074         .get_module_info = bnxt_get_module_info,
4075         .get_module_eeprom = bnxt_get_module_eeprom,
4076         .timesync_enable      = bnxt_timesync_enable,
4077         .timesync_disable     = bnxt_timesync_disable,
4078         .timesync_read_time   = bnxt_timesync_read_time,
4079         .timesync_write_time   = bnxt_timesync_write_time,
4080         .timesync_adjust_time = bnxt_timesync_adjust_time,
4081         .timesync_read_rx_timestamp = bnxt_timesync_read_rx_timestamp,
4082         .timesync_read_tx_timestamp = bnxt_timesync_read_tx_timestamp,
4083 };
4084
4085 static uint32_t bnxt_map_reset_regs(struct bnxt *bp, uint32_t reg)
4086 {
4087         uint32_t offset;
4088
4089         /* Only pre-map the reset GRC registers using window 3 */
4090         rte_write32(reg & 0xfffff000, (uint8_t *)bp->bar0 +
4091                     BNXT_GRCPF_REG_WINDOW_BASE_OUT + 8);
4092
4093         offset = BNXT_GRCP_WINDOW_3_BASE + (reg & 0xffc);
4094
4095         return offset;
4096 }
4097
4098 int bnxt_map_fw_health_status_regs(struct bnxt *bp)
4099 {
4100         struct bnxt_error_recovery_info *info = bp->recovery_info;
4101         uint32_t reg_base = 0xffffffff;
4102         int i;
4103
4104         /* Only pre-map the monitoring GRC registers using window 2 */
4105         for (i = 0; i < BNXT_FW_STATUS_REG_CNT; i++) {
4106                 uint32_t reg = info->status_regs[i];
4107
4108                 if (BNXT_FW_STATUS_REG_TYPE(reg) != BNXT_FW_STATUS_REG_TYPE_GRC)
4109                         continue;
4110
4111                 if (reg_base == 0xffffffff)
4112                         reg_base = reg & 0xfffff000;
4113                 if ((reg & 0xfffff000) != reg_base)
4114                         return -ERANGE;
4115
4116                 /* Use mask 0xffc as the Lower 2 bits indicates
4117                  * address space location
4118                  */
4119                 info->mapped_status_regs[i] = BNXT_GRCP_WINDOW_2_BASE +
4120                                                 (reg & 0xffc);
4121         }
4122
4123         if (reg_base == 0xffffffff)
4124                 return 0;
4125
4126         rte_write32(reg_base, (uint8_t *)bp->bar0 +
4127                     BNXT_GRCPF_REG_WINDOW_BASE_OUT + 4);
4128
4129         return 0;
4130 }
4131
4132 static void bnxt_write_fw_reset_reg(struct bnxt *bp, uint32_t index)
4133 {
4134         struct bnxt_error_recovery_info *info = bp->recovery_info;
4135         uint32_t delay = info->delay_after_reset[index];
4136         uint32_t val = info->reset_reg_val[index];
4137         uint32_t reg = info->reset_reg[index];
4138         uint32_t type, offset;
4139         int ret;
4140
4141         type = BNXT_FW_STATUS_REG_TYPE(reg);
4142         offset = BNXT_FW_STATUS_REG_OFF(reg);
4143
4144         switch (type) {
4145         case BNXT_FW_STATUS_REG_TYPE_CFG:
4146                 ret = rte_pci_write_config(bp->pdev, &val, sizeof(val), offset);
4147                 if (ret < 0) {
4148                         PMD_DRV_LOG(ERR, "Failed to write %#x at PCI offset %#x",
4149                                     val, offset);
4150                         return;
4151                 }
4152                 break;
4153         case BNXT_FW_STATUS_REG_TYPE_GRC:
4154                 offset = bnxt_map_reset_regs(bp, offset);
4155                 rte_write32(val, (uint8_t *)bp->bar0 + offset);
4156                 break;
4157         case BNXT_FW_STATUS_REG_TYPE_BAR0:
4158                 rte_write32(val, (uint8_t *)bp->bar0 + offset);
4159                 break;
4160         }
4161         /* wait on a specific interval of time until core reset is complete */
4162         if (delay)
4163                 rte_delay_ms(delay);
4164 }
4165
4166 static void bnxt_dev_cleanup(struct bnxt *bp)
4167 {
4168         bp->eth_dev->data->dev_link.link_status = 0;
4169         bp->link_info->link_up = 0;
4170         if (bp->eth_dev->data->dev_started)
4171                 bnxt_dev_stop(bp->eth_dev);
4172
4173         bnxt_uninit_resources(bp, true);
4174 }
4175
4176 static int
4177 bnxt_check_fw_reset_done(struct bnxt *bp)
4178 {
4179         int timeout = bp->fw_reset_max_msecs;
4180         uint16_t val = 0;
4181         int rc;
4182
4183         do {
4184                 rc = rte_pci_read_config(bp->pdev, &val, sizeof(val), PCI_SUBSYSTEM_ID_OFFSET);
4185                 if (rc < 0) {
4186                         PMD_DRV_LOG(ERR, "Failed to read PCI offset 0x%x", PCI_SUBSYSTEM_ID_OFFSET);
4187                         return rc;
4188                 }
4189                 if (val != 0xffff)
4190                         break;
4191                 rte_delay_ms(1);
4192         } while (timeout--);
4193
4194         if (val == 0xffff) {
4195                 PMD_DRV_LOG(ERR, "Firmware reset aborted, PCI config space invalid\n");
4196                 return -1;
4197         }
4198
4199         return 0;
4200 }
4201
4202 static int bnxt_restore_vlan_filters(struct bnxt *bp)
4203 {
4204         struct rte_eth_dev *dev = bp->eth_dev;
4205         struct rte_vlan_filter_conf *vfc;
4206         int vidx, vbit, rc;
4207         uint16_t vlan_id;
4208
4209         for (vlan_id = 1; vlan_id <= RTE_ETHER_MAX_VLAN_ID; vlan_id++) {
4210                 vfc = &dev->data->vlan_filter_conf;
4211                 vidx = vlan_id / 64;
4212                 vbit = vlan_id % 64;
4213
4214                 /* Each bit corresponds to a VLAN id */
4215                 if (vfc->ids[vidx] & (UINT64_C(1) << vbit)) {
4216                         rc = bnxt_add_vlan_filter(bp, vlan_id);
4217                         if (rc)
4218                                 return rc;
4219                 }
4220         }
4221
4222         return 0;
4223 }
4224
4225 static int bnxt_restore_mac_filters(struct bnxt *bp)
4226 {
4227         struct rte_eth_dev *dev = bp->eth_dev;
4228         struct rte_eth_dev_info dev_info;
4229         struct rte_ether_addr *addr;
4230         uint64_t pool_mask;
4231         uint32_t pool = 0;
4232         uint32_t i;
4233         int rc;
4234
4235         if (BNXT_VF(bp) && !BNXT_VF_IS_TRUSTED(bp))
4236                 return 0;
4237
4238         rc = bnxt_dev_info_get_op(dev, &dev_info);
4239         if (rc)
4240                 return rc;
4241
4242         /* replay MAC address configuration */
4243         for (i = 1; i < dev_info.max_mac_addrs; i++) {
4244                 addr = &dev->data->mac_addrs[i];
4245
4246                 /* skip zero address */
4247                 if (rte_is_zero_ether_addr(addr))
4248                         continue;
4249
4250                 pool = 0;
4251                 pool_mask = dev->data->mac_pool_sel[i];
4252
4253                 do {
4254                         if (pool_mask & 1ULL) {
4255                                 rc = bnxt_mac_addr_add_op(dev, addr, i, pool);
4256                                 if (rc)
4257                                         return rc;
4258                         }
4259                         pool_mask >>= 1;
4260                         pool++;
4261                 } while (pool_mask);
4262         }
4263
4264         return 0;
4265 }
4266
4267 static int bnxt_restore_filters(struct bnxt *bp)
4268 {
4269         struct rte_eth_dev *dev = bp->eth_dev;
4270         int ret = 0;
4271
4272         if (dev->data->all_multicast) {
4273                 ret = bnxt_allmulticast_enable_op(dev);
4274                 if (ret)
4275                         return ret;
4276         }
4277         if (dev->data->promiscuous) {
4278                 ret = bnxt_promiscuous_enable_op(dev);
4279                 if (ret)
4280                         return ret;
4281         }
4282
4283         ret = bnxt_restore_mac_filters(bp);
4284         if (ret)
4285                 return ret;
4286
4287         ret = bnxt_restore_vlan_filters(bp);
4288         /* TODO restore other filters as well */
4289         return ret;
4290 }
4291
4292 static int bnxt_check_fw_ready(struct bnxt *bp)
4293 {
4294         int timeout = bp->fw_reset_max_msecs;
4295         int rc = 0;
4296
4297         do {
4298                 rc = bnxt_hwrm_poll_ver_get(bp);
4299                 if (rc == 0)
4300                         break;
4301                 rte_delay_ms(BNXT_FW_READY_WAIT_INTERVAL);
4302                 timeout -= BNXT_FW_READY_WAIT_INTERVAL;
4303         } while (rc && timeout > 0);
4304
4305         if (rc)
4306                 PMD_DRV_LOG(ERR, "FW is not Ready after reset\n");
4307
4308         return rc;
4309 }
4310
4311 static void bnxt_dev_recover(void *arg)
4312 {
4313         struct bnxt *bp = arg;
4314         int rc = 0;
4315
4316         pthread_mutex_lock(&bp->err_recovery_lock);
4317
4318         if (!bp->fw_reset_min_msecs) {
4319                 rc = bnxt_check_fw_reset_done(bp);
4320                 if (rc)
4321                         goto err;
4322         }
4323
4324         /* Clear Error flag so that device re-init should happen */
4325         bp->flags &= ~BNXT_FLAG_FATAL_ERROR;
4326
4327         rc = bnxt_check_fw_ready(bp);
4328         if (rc)
4329                 goto err;
4330
4331         rc = bnxt_init_resources(bp, true);
4332         if (rc) {
4333                 PMD_DRV_LOG(ERR,
4334                             "Failed to initialize resources after reset\n");
4335                 goto err;
4336         }
4337         /* clear reset flag as the device is initialized now */
4338         bp->flags &= ~BNXT_FLAG_FW_RESET;
4339
4340         rc = bnxt_dev_start_op(bp->eth_dev);
4341         if (rc) {
4342                 PMD_DRV_LOG(ERR, "Failed to start port after reset\n");
4343                 goto err_start;
4344         }
4345
4346         rc = bnxt_restore_filters(bp);
4347         if (rc)
4348                 goto err_start;
4349
4350         PMD_DRV_LOG(INFO, "Recovered from FW reset\n");
4351         pthread_mutex_unlock(&bp->err_recovery_lock);
4352
4353         return;
4354 err_start:
4355         bnxt_dev_stop(bp->eth_dev);
4356 err:
4357         bp->flags |= BNXT_FLAG_FATAL_ERROR;
4358         bnxt_uninit_resources(bp, false);
4359         if (bp->eth_dev->data->dev_conf.intr_conf.rmv)
4360                 rte_eth_dev_callback_process(bp->eth_dev,
4361                                              RTE_ETH_EVENT_INTR_RMV,
4362                                              NULL);
4363         pthread_mutex_unlock(&bp->err_recovery_lock);
4364         PMD_DRV_LOG(ERR, "Failed to recover from FW reset\n");
4365 }
4366
4367 void bnxt_dev_reset_and_resume(void *arg)
4368 {
4369         struct bnxt *bp = arg;
4370         uint32_t us = US_PER_MS * bp->fw_reset_min_msecs;
4371         uint16_t val = 0;
4372         int rc;
4373
4374         bnxt_dev_cleanup(bp);
4375
4376         bnxt_wait_for_device_shutdown(bp);
4377
4378         /* During some fatal firmware error conditions, the PCI config space
4379          * register 0x2e which normally contains the subsystem ID will become
4380          * 0xffff. This register will revert back to the normal value after
4381          * the chip has completed core reset. If we detect this condition,
4382          * we can poll this config register immediately for the value to revert.
4383          */
4384         if (bp->flags & BNXT_FLAG_FATAL_ERROR) {
4385                 rc = rte_pci_read_config(bp->pdev, &val, sizeof(val), PCI_SUBSYSTEM_ID_OFFSET);
4386                 if (rc < 0) {
4387                         PMD_DRV_LOG(ERR, "Failed to read PCI offset 0x%x", PCI_SUBSYSTEM_ID_OFFSET);
4388                         return;
4389                 }
4390                 if (val == 0xffff) {
4391                         bp->fw_reset_min_msecs = 0;
4392                         us = 1;
4393                 }
4394         }
4395
4396         rc = rte_eal_alarm_set(us, bnxt_dev_recover, (void *)bp);
4397         if (rc)
4398                 PMD_DRV_LOG(ERR, "Error setting recovery alarm");
4399 }
4400
4401 uint32_t bnxt_read_fw_status_reg(struct bnxt *bp, uint32_t index)
4402 {
4403         struct bnxt_error_recovery_info *info = bp->recovery_info;
4404         uint32_t reg = info->status_regs[index];
4405         uint32_t type, offset, val = 0;
4406         int ret = 0;
4407
4408         type = BNXT_FW_STATUS_REG_TYPE(reg);
4409         offset = BNXT_FW_STATUS_REG_OFF(reg);
4410
4411         switch (type) {
4412         case BNXT_FW_STATUS_REG_TYPE_CFG:
4413                 ret = rte_pci_read_config(bp->pdev, &val, sizeof(val), offset);
4414                 if (ret < 0)
4415                         PMD_DRV_LOG(ERR, "Failed to read PCI offset %#x",
4416                                     offset);
4417                 break;
4418         case BNXT_FW_STATUS_REG_TYPE_GRC:
4419                 offset = info->mapped_status_regs[index];
4420                 /* FALLTHROUGH */
4421         case BNXT_FW_STATUS_REG_TYPE_BAR0:
4422                 val = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
4423                                        offset));
4424                 break;
4425         }
4426
4427         return val;
4428 }
4429
4430 static int bnxt_fw_reset_all(struct bnxt *bp)
4431 {
4432         struct bnxt_error_recovery_info *info = bp->recovery_info;
4433         uint32_t i;
4434         int rc = 0;
4435
4436         if (info->flags & BNXT_FLAG_ERROR_RECOVERY_HOST) {
4437                 /* Reset through master function driver */
4438                 for (i = 0; i < info->reg_array_cnt; i++)
4439                         bnxt_write_fw_reset_reg(bp, i);
4440                 /* Wait for time specified by FW after triggering reset */
4441                 rte_delay_ms(info->master_func_wait_period_after_reset);
4442         } else if (info->flags & BNXT_FLAG_ERROR_RECOVERY_CO_CPU) {
4443                 /* Reset with the help of Kong processor */
4444                 rc = bnxt_hwrm_fw_reset(bp);
4445                 if (rc)
4446                         PMD_DRV_LOG(ERR, "Failed to reset FW\n");
4447         }
4448
4449         return rc;
4450 }
4451
4452 static void bnxt_fw_reset_cb(void *arg)
4453 {
4454         struct bnxt *bp = arg;
4455         struct bnxt_error_recovery_info *info = bp->recovery_info;
4456         int rc = 0;
4457
4458         /* Only Master function can do FW reset */
4459         if (bnxt_is_master_func(bp) &&
4460             bnxt_is_recovery_enabled(bp)) {
4461                 rc = bnxt_fw_reset_all(bp);
4462                 if (rc) {
4463                         PMD_DRV_LOG(ERR, "Adapter recovery failed\n");
4464                         return;
4465                 }
4466         }
4467
4468         /* if recovery method is ERROR_RECOVERY_CO_CPU, KONG will send
4469          * EXCEPTION_FATAL_ASYNC event to all the functions
4470          * (including MASTER FUNC). After receiving this Async, all the active
4471          * drivers should treat this case as FW initiated recovery
4472          */
4473         if (info->flags & BNXT_FLAG_ERROR_RECOVERY_HOST) {
4474                 bp->fw_reset_min_msecs = BNXT_MIN_FW_READY_TIMEOUT;
4475                 bp->fw_reset_max_msecs = BNXT_MAX_FW_RESET_TIMEOUT;
4476
4477                 /* To recover from error */
4478                 rte_eal_alarm_set(US_PER_MS, bnxt_dev_reset_and_resume,
4479                                   (void *)bp);
4480         }
4481 }
4482
4483 /* Driver should poll FW heartbeat, reset_counter with the frequency
4484  * advertised by FW in HWRM_ERROR_RECOVERY_QCFG.
4485  * When the driver detects heartbeat stop or change in reset_counter,
4486  * it has to trigger a reset to recover from the error condition.
4487  * A “master PF” is the function who will have the privilege to
4488  * initiate the chimp reset. The master PF will be elected by the
4489  * firmware and will be notified through async message.
4490  */
4491 static void bnxt_check_fw_health(void *arg)
4492 {
4493         struct bnxt *bp = arg;
4494         struct bnxt_error_recovery_info *info = bp->recovery_info;
4495         uint32_t val = 0, wait_msec;
4496
4497         if (!info || !bnxt_is_recovery_enabled(bp) ||
4498             is_bnxt_in_error(bp))
4499                 return;
4500
4501         val = bnxt_read_fw_status_reg(bp, BNXT_FW_HEARTBEAT_CNT_REG);
4502         if (val == info->last_heart_beat)
4503                 goto reset;
4504
4505         info->last_heart_beat = val;
4506
4507         val = bnxt_read_fw_status_reg(bp, BNXT_FW_RECOVERY_CNT_REG);
4508         if (val != info->last_reset_counter)
4509                 goto reset;
4510
4511         info->last_reset_counter = val;
4512
4513         rte_eal_alarm_set(US_PER_MS * info->driver_polling_freq,
4514                           bnxt_check_fw_health, (void *)bp);
4515
4516         return;
4517 reset:
4518         /* Stop DMA to/from device */
4519         bp->flags |= BNXT_FLAG_FATAL_ERROR;
4520         bp->flags |= BNXT_FLAG_FW_RESET;
4521
4522         bnxt_stop_rxtx(bp);
4523
4524         PMD_DRV_LOG(ERR, "Detected FW dead condition\n");
4525
4526         if (bnxt_is_master_func(bp))
4527                 wait_msec = info->master_func_wait_period;
4528         else
4529                 wait_msec = info->normal_func_wait_period;
4530
4531         rte_eal_alarm_set(US_PER_MS * wait_msec,
4532                           bnxt_fw_reset_cb, (void *)bp);
4533 }
4534
4535 void bnxt_schedule_fw_health_check(struct bnxt *bp)
4536 {
4537         uint32_t polling_freq;
4538
4539         pthread_mutex_lock(&bp->health_check_lock);
4540
4541         if (!bnxt_is_recovery_enabled(bp))
4542                 goto done;
4543
4544         if (bp->flags & BNXT_FLAG_FW_HEALTH_CHECK_SCHEDULED)
4545                 goto done;
4546
4547         polling_freq = bp->recovery_info->driver_polling_freq;
4548
4549         rte_eal_alarm_set(US_PER_MS * polling_freq,
4550                           bnxt_check_fw_health, (void *)bp);
4551         bp->flags |= BNXT_FLAG_FW_HEALTH_CHECK_SCHEDULED;
4552
4553 done:
4554         pthread_mutex_unlock(&bp->health_check_lock);
4555 }
4556
4557 static void bnxt_cancel_fw_health_check(struct bnxt *bp)
4558 {
4559         rte_eal_alarm_cancel(bnxt_check_fw_health, (void *)bp);
4560         bp->flags &= ~BNXT_FLAG_FW_HEALTH_CHECK_SCHEDULED;
4561 }
4562
4563 static bool bnxt_vf_pciid(uint16_t device_id)
4564 {
4565         switch (device_id) {
4566         case BROADCOM_DEV_ID_57304_VF:
4567         case BROADCOM_DEV_ID_57406_VF:
4568         case BROADCOM_DEV_ID_5731X_VF:
4569         case BROADCOM_DEV_ID_5741X_VF:
4570         case BROADCOM_DEV_ID_57414_VF:
4571         case BROADCOM_DEV_ID_STRATUS_NIC_VF1:
4572         case BROADCOM_DEV_ID_STRATUS_NIC_VF2:
4573         case BROADCOM_DEV_ID_58802_VF:
4574         case BROADCOM_DEV_ID_57500_VF1:
4575         case BROADCOM_DEV_ID_57500_VF2:
4576         case BROADCOM_DEV_ID_58818_VF:
4577                 /* FALLTHROUGH */
4578                 return true;
4579         default:
4580                 return false;
4581         }
4582 }
4583
4584 /* Phase 5 device */
4585 static bool bnxt_p5_device(uint16_t device_id)
4586 {
4587         switch (device_id) {
4588         case BROADCOM_DEV_ID_57508:
4589         case BROADCOM_DEV_ID_57504:
4590         case BROADCOM_DEV_ID_57502:
4591         case BROADCOM_DEV_ID_57508_MF1:
4592         case BROADCOM_DEV_ID_57504_MF1:
4593         case BROADCOM_DEV_ID_57502_MF1:
4594         case BROADCOM_DEV_ID_57508_MF2:
4595         case BROADCOM_DEV_ID_57504_MF2:
4596         case BROADCOM_DEV_ID_57502_MF2:
4597         case BROADCOM_DEV_ID_57500_VF1:
4598         case BROADCOM_DEV_ID_57500_VF2:
4599         case BROADCOM_DEV_ID_58812:
4600         case BROADCOM_DEV_ID_58814:
4601         case BROADCOM_DEV_ID_58818:
4602         case BROADCOM_DEV_ID_58818_VF:
4603                 /* FALLTHROUGH */
4604                 return true;
4605         default:
4606                 return false;
4607         }
4608 }
4609
4610 bool bnxt_stratus_device(struct bnxt *bp)
4611 {
4612         uint16_t device_id = bp->pdev->id.device_id;
4613
4614         switch (device_id) {
4615         case BROADCOM_DEV_ID_STRATUS_NIC:
4616         case BROADCOM_DEV_ID_STRATUS_NIC_VF1:
4617         case BROADCOM_DEV_ID_STRATUS_NIC_VF2:
4618                 /* FALLTHROUGH */
4619                 return true;
4620         default:
4621                 return false;
4622         }
4623 }
4624
4625 static int bnxt_map_pci_bars(struct rte_eth_dev *eth_dev)
4626 {
4627         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
4628         struct bnxt *bp = eth_dev->data->dev_private;
4629
4630         /* enable device (incl. PCI PM wakeup), and bus-mastering */
4631         bp->bar0 = (void *)pci_dev->mem_resource[0].addr;
4632         bp->doorbell_base = (void *)pci_dev->mem_resource[2].addr;
4633         if (!bp->bar0 || !bp->doorbell_base) {
4634                 PMD_DRV_LOG(ERR, "Unable to access Hardware\n");
4635                 return -ENODEV;
4636         }
4637
4638         bp->eth_dev = eth_dev;
4639         bp->pdev = pci_dev;
4640
4641         return 0;
4642 }
4643
4644 static int bnxt_alloc_ctx_mem_blk(struct bnxt *bp,
4645                                   struct bnxt_ctx_pg_info *ctx_pg,
4646                                   uint32_t mem_size,
4647                                   const char *suffix,
4648                                   uint16_t idx)
4649 {
4650         struct bnxt_ring_mem_info *rmem = &ctx_pg->ring_mem;
4651         const struct rte_memzone *mz = NULL;
4652         char mz_name[RTE_MEMZONE_NAMESIZE];
4653         rte_iova_t mz_phys_addr;
4654         uint64_t valid_bits = 0;
4655         uint32_t sz;
4656         int i;
4657
4658         if (!mem_size)
4659                 return 0;
4660
4661         rmem->nr_pages = RTE_ALIGN_MUL_CEIL(mem_size, BNXT_PAGE_SIZE) /
4662                          BNXT_PAGE_SIZE;
4663         rmem->page_size = BNXT_PAGE_SIZE;
4664         rmem->pg_arr = ctx_pg->ctx_pg_arr;
4665         rmem->dma_arr = ctx_pg->ctx_dma_arr;
4666         rmem->flags = BNXT_RMEM_VALID_PTE_FLAG;
4667
4668         valid_bits = PTU_PTE_VALID;
4669
4670         if (rmem->nr_pages > 1) {
4671                 snprintf(mz_name, RTE_MEMZONE_NAMESIZE,
4672                          "bnxt_ctx_pg_tbl%s_%x_%d",
4673                          suffix, idx, bp->eth_dev->data->port_id);
4674                 mz_name[RTE_MEMZONE_NAMESIZE - 1] = 0;
4675                 mz = rte_memzone_lookup(mz_name);
4676                 if (!mz) {
4677                         mz = rte_memzone_reserve_aligned(mz_name,
4678                                                 rmem->nr_pages * 8,
4679                                                 bp->eth_dev->device->numa_node,
4680                                                 RTE_MEMZONE_2MB |
4681                                                 RTE_MEMZONE_SIZE_HINT_ONLY |
4682                                                 RTE_MEMZONE_IOVA_CONTIG,
4683                                                 BNXT_PAGE_SIZE);
4684                         if (mz == NULL)
4685                                 return -ENOMEM;
4686                 }
4687
4688                 memset(mz->addr, 0, mz->len);
4689                 mz_phys_addr = mz->iova;
4690
4691                 rmem->pg_tbl = mz->addr;
4692                 rmem->pg_tbl_map = mz_phys_addr;
4693                 rmem->pg_tbl_mz = mz;
4694         }
4695
4696         snprintf(mz_name, RTE_MEMZONE_NAMESIZE, "bnxt_ctx_%s_%x_%d",
4697                  suffix, idx, bp->eth_dev->data->port_id);
4698         mz = rte_memzone_lookup(mz_name);
4699         if (!mz) {
4700                 mz = rte_memzone_reserve_aligned(mz_name,
4701                                                  mem_size,
4702                                                  bp->eth_dev->device->numa_node,
4703                                                  RTE_MEMZONE_1GB |
4704                                                  RTE_MEMZONE_SIZE_HINT_ONLY |
4705                                                  RTE_MEMZONE_IOVA_CONTIG,
4706                                                  BNXT_PAGE_SIZE);
4707                 if (mz == NULL)
4708                         return -ENOMEM;
4709         }
4710
4711         memset(mz->addr, 0, mz->len);
4712         mz_phys_addr = mz->iova;
4713
4714         for (sz = 0, i = 0; sz < mem_size; sz += BNXT_PAGE_SIZE, i++) {
4715                 rmem->pg_arr[i] = ((char *)mz->addr) + sz;
4716                 rmem->dma_arr[i] = mz_phys_addr + sz;
4717
4718                 if (rmem->nr_pages > 1) {
4719                         if (i == rmem->nr_pages - 2 &&
4720                             (rmem->flags & BNXT_RMEM_RING_PTE_FLAG))
4721                                 valid_bits |= PTU_PTE_NEXT_TO_LAST;
4722                         else if (i == rmem->nr_pages - 1 &&
4723                                  (rmem->flags & BNXT_RMEM_RING_PTE_FLAG))
4724                                 valid_bits |= PTU_PTE_LAST;
4725
4726                         rmem->pg_tbl[i] = rte_cpu_to_le_64(rmem->dma_arr[i] |
4727                                                            valid_bits);
4728                 }
4729         }
4730
4731         rmem->mz = mz;
4732         if (rmem->vmem_size)
4733                 rmem->vmem = (void **)mz->addr;
4734         rmem->dma_arr[0] = mz_phys_addr;
4735         return 0;
4736 }
4737
4738 static void bnxt_free_ctx_mem(struct bnxt *bp)
4739 {
4740         int i;
4741
4742         if (!bp->ctx || !(bp->ctx->flags & BNXT_CTX_FLAG_INITED))
4743                 return;
4744
4745         bp->ctx->flags &= ~BNXT_CTX_FLAG_INITED;
4746         rte_memzone_free(bp->ctx->qp_mem.ring_mem.mz);
4747         rte_memzone_free(bp->ctx->srq_mem.ring_mem.mz);
4748         rte_memzone_free(bp->ctx->cq_mem.ring_mem.mz);
4749         rte_memzone_free(bp->ctx->vnic_mem.ring_mem.mz);
4750         rte_memzone_free(bp->ctx->stat_mem.ring_mem.mz);
4751         rte_memzone_free(bp->ctx->qp_mem.ring_mem.pg_tbl_mz);
4752         rte_memzone_free(bp->ctx->srq_mem.ring_mem.pg_tbl_mz);
4753         rte_memzone_free(bp->ctx->cq_mem.ring_mem.pg_tbl_mz);
4754         rte_memzone_free(bp->ctx->vnic_mem.ring_mem.pg_tbl_mz);
4755         rte_memzone_free(bp->ctx->stat_mem.ring_mem.pg_tbl_mz);
4756
4757         for (i = 0; i < bp->ctx->tqm_fp_rings_count + 1; i++) {
4758                 if (bp->ctx->tqm_mem[i])
4759                         rte_memzone_free(bp->ctx->tqm_mem[i]->ring_mem.mz);
4760         }
4761
4762         rte_free(bp->ctx);
4763         bp->ctx = NULL;
4764 }
4765
4766 #define bnxt_roundup(x, y)   ((((x) + ((y) - 1)) / (y)) * (y))
4767
4768 #define min_t(type, x, y) ({                    \
4769         type __min1 = (x);                      \
4770         type __min2 = (y);                      \
4771         __min1 < __min2 ? __min1 : __min2; })
4772
4773 #define max_t(type, x, y) ({                    \
4774         type __max1 = (x);                      \
4775         type __max2 = (y);                      \
4776         __max1 > __max2 ? __max1 : __max2; })
4777
4778 #define clamp_t(type, _x, min, max)     min_t(type, max_t(type, _x, min), max)
4779
4780 int bnxt_alloc_ctx_mem(struct bnxt *bp)
4781 {
4782         struct bnxt_ctx_pg_info *ctx_pg;
4783         struct bnxt_ctx_mem_info *ctx;
4784         uint32_t mem_size, ena, entries;
4785         uint32_t entries_sp, min;
4786         int i, rc;
4787
4788         rc = bnxt_hwrm_func_backing_store_qcaps(bp);
4789         if (rc) {
4790                 PMD_DRV_LOG(ERR, "Query context mem capability failed\n");
4791                 return rc;
4792         }
4793         ctx = bp->ctx;
4794         if (!ctx || (ctx->flags & BNXT_CTX_FLAG_INITED))
4795                 return 0;
4796
4797         ctx_pg = &ctx->qp_mem;
4798         ctx_pg->entries = ctx->qp_min_qp1_entries + ctx->qp_max_l2_entries;
4799         if (ctx->qp_entry_size) {
4800                 mem_size = ctx->qp_entry_size * ctx_pg->entries;
4801                 rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "qp_mem", 0);
4802                 if (rc)
4803                         return rc;
4804         }
4805
4806         ctx_pg = &ctx->srq_mem;
4807         ctx_pg->entries = ctx->srq_max_l2_entries;
4808         if (ctx->srq_entry_size) {
4809                 mem_size = ctx->srq_entry_size * ctx_pg->entries;
4810                 rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "srq_mem", 0);
4811                 if (rc)
4812                         return rc;
4813         }
4814
4815         ctx_pg = &ctx->cq_mem;
4816         ctx_pg->entries = ctx->cq_max_l2_entries;
4817         if (ctx->cq_entry_size) {
4818                 mem_size = ctx->cq_entry_size * ctx_pg->entries;
4819                 rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "cq_mem", 0);
4820                 if (rc)
4821                         return rc;
4822         }
4823
4824         ctx_pg = &ctx->vnic_mem;
4825         ctx_pg->entries = ctx->vnic_max_vnic_entries +
4826                 ctx->vnic_max_ring_table_entries;
4827         if (ctx->vnic_entry_size) {
4828                 mem_size = ctx->vnic_entry_size * ctx_pg->entries;
4829                 rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "vnic_mem", 0);
4830                 if (rc)
4831                         return rc;
4832         }
4833
4834         ctx_pg = &ctx->stat_mem;
4835         ctx_pg->entries = ctx->stat_max_entries;
4836         if (ctx->stat_entry_size) {
4837                 mem_size = ctx->stat_entry_size * ctx_pg->entries;
4838                 rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "stat_mem", 0);
4839                 if (rc)
4840                         return rc;
4841         }
4842
4843         min = ctx->tqm_min_entries_per_ring;
4844
4845         entries_sp = ctx->qp_max_l2_entries +
4846                      ctx->vnic_max_vnic_entries +
4847                      2 * ctx->qp_min_qp1_entries + min;
4848         entries_sp = bnxt_roundup(entries_sp, ctx->tqm_entries_multiple);
4849
4850         entries = ctx->qp_max_l2_entries + ctx->qp_min_qp1_entries;
4851         entries = bnxt_roundup(entries, ctx->tqm_entries_multiple);
4852         entries = clamp_t(uint32_t, entries, min,
4853                           ctx->tqm_max_entries_per_ring);
4854         for (i = 0, ena = 0; i < ctx->tqm_fp_rings_count + 1; i++) {
4855                 /* i=0 is for TQM_SP. i=1 to i=8 applies to RING0 to RING7.
4856                  * i > 8 is other ext rings.
4857                  */
4858                 ctx_pg = ctx->tqm_mem[i];
4859                 ctx_pg->entries = i ? entries : entries_sp;
4860                 if (ctx->tqm_entry_size) {
4861                         mem_size = ctx->tqm_entry_size * ctx_pg->entries;
4862                         rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size,
4863                                                     "tqm_mem", i);
4864                         if (rc)
4865                                 return rc;
4866                 }
4867                 if (i < BNXT_MAX_TQM_LEGACY_RINGS)
4868                         ena |= HWRM_FUNC_BACKING_STORE_CFG_INPUT_ENABLES_TQM_SP << i;
4869                 else
4870                         ena |= HWRM_FUNC_BACKING_STORE_CFG_INPUT_ENABLES_TQM_RING8;
4871         }
4872
4873         ena |= FUNC_BACKING_STORE_CFG_INPUT_DFLT_ENABLES;
4874         rc = bnxt_hwrm_func_backing_store_cfg(bp, ena);
4875         if (rc)
4876                 PMD_DRV_LOG(ERR,
4877                             "Failed to configure context mem: rc = %d\n", rc);
4878         else
4879                 ctx->flags |= BNXT_CTX_FLAG_INITED;
4880
4881         return rc;
4882 }
4883
4884 static int bnxt_alloc_stats_mem(struct bnxt *bp)
4885 {
4886         struct rte_pci_device *pci_dev = bp->pdev;
4887         char mz_name[RTE_MEMZONE_NAMESIZE];
4888         const struct rte_memzone *mz = NULL;
4889         uint32_t total_alloc_len;
4890         rte_iova_t mz_phys_addr;
4891
4892         if (pci_dev->id.device_id == BROADCOM_DEV_ID_NS2)
4893                 return 0;
4894
4895         snprintf(mz_name, RTE_MEMZONE_NAMESIZE,
4896                  "bnxt_" PCI_PRI_FMT "-%s", pci_dev->addr.domain,
4897                  pci_dev->addr.bus, pci_dev->addr.devid,
4898                  pci_dev->addr.function, "rx_port_stats");
4899         mz_name[RTE_MEMZONE_NAMESIZE - 1] = 0;
4900         mz = rte_memzone_lookup(mz_name);
4901         total_alloc_len =
4902                 RTE_CACHE_LINE_ROUNDUP(sizeof(struct rx_port_stats) +
4903                                        sizeof(struct rx_port_stats_ext) + 512);
4904         if (!mz) {
4905                 mz = rte_memzone_reserve(mz_name, total_alloc_len,
4906                                          SOCKET_ID_ANY,
4907                                          RTE_MEMZONE_2MB |
4908                                          RTE_MEMZONE_SIZE_HINT_ONLY |
4909                                          RTE_MEMZONE_IOVA_CONTIG);
4910                 if (mz == NULL)
4911                         return -ENOMEM;
4912         }
4913         memset(mz->addr, 0, mz->len);
4914         mz_phys_addr = mz->iova;
4915
4916         bp->rx_mem_zone = (const void *)mz;
4917         bp->hw_rx_port_stats = mz->addr;
4918         bp->hw_rx_port_stats_map = mz_phys_addr;
4919
4920         snprintf(mz_name, RTE_MEMZONE_NAMESIZE,
4921                  "bnxt_" PCI_PRI_FMT "-%s", pci_dev->addr.domain,
4922                  pci_dev->addr.bus, pci_dev->addr.devid,
4923                  pci_dev->addr.function, "tx_port_stats");
4924         mz_name[RTE_MEMZONE_NAMESIZE - 1] = 0;
4925         mz = rte_memzone_lookup(mz_name);
4926         total_alloc_len =
4927                 RTE_CACHE_LINE_ROUNDUP(sizeof(struct tx_port_stats) +
4928                                        sizeof(struct tx_port_stats_ext) + 512);
4929         if (!mz) {
4930                 mz = rte_memzone_reserve(mz_name,
4931                                          total_alloc_len,
4932                                          SOCKET_ID_ANY,
4933                                          RTE_MEMZONE_2MB |
4934                                          RTE_MEMZONE_SIZE_HINT_ONLY |
4935                                          RTE_MEMZONE_IOVA_CONTIG);
4936                 if (mz == NULL)
4937                         return -ENOMEM;
4938         }
4939         memset(mz->addr, 0, mz->len);
4940         mz_phys_addr = mz->iova;
4941
4942         bp->tx_mem_zone = (const void *)mz;
4943         bp->hw_tx_port_stats = mz->addr;
4944         bp->hw_tx_port_stats_map = mz_phys_addr;
4945         bp->flags |= BNXT_FLAG_PORT_STATS;
4946
4947         /* Display extended statistics if FW supports it */
4948         if (bp->hwrm_spec_code < HWRM_SPEC_CODE_1_8_4 ||
4949             bp->hwrm_spec_code == HWRM_SPEC_CODE_1_9_0 ||
4950             !(bp->flags & BNXT_FLAG_EXT_STATS_SUPPORTED))
4951                 return 0;
4952
4953         bp->hw_rx_port_stats_ext = (void *)
4954                 ((uint8_t *)bp->hw_rx_port_stats +
4955                  sizeof(struct rx_port_stats));
4956         bp->hw_rx_port_stats_ext_map = bp->hw_rx_port_stats_map +
4957                 sizeof(struct rx_port_stats);
4958         bp->flags |= BNXT_FLAG_EXT_RX_PORT_STATS;
4959
4960         if (bp->hwrm_spec_code < HWRM_SPEC_CODE_1_9_2 ||
4961             bp->flags & BNXT_FLAG_EXT_STATS_SUPPORTED) {
4962                 bp->hw_tx_port_stats_ext = (void *)
4963                         ((uint8_t *)bp->hw_tx_port_stats +
4964                          sizeof(struct tx_port_stats));
4965                 bp->hw_tx_port_stats_ext_map =
4966                         bp->hw_tx_port_stats_map +
4967                         sizeof(struct tx_port_stats);
4968                 bp->flags |= BNXT_FLAG_EXT_TX_PORT_STATS;
4969         }
4970
4971         return 0;
4972 }
4973
4974 static int bnxt_setup_mac_addr(struct rte_eth_dev *eth_dev)
4975 {
4976         struct bnxt *bp = eth_dev->data->dev_private;
4977         int rc = 0;
4978
4979         eth_dev->data->mac_addrs = rte_zmalloc("bnxt_mac_addr_tbl",
4980                                                RTE_ETHER_ADDR_LEN *
4981                                                bp->max_l2_ctx,
4982                                                0);
4983         if (eth_dev->data->mac_addrs == NULL) {
4984                 PMD_DRV_LOG(ERR, "Failed to alloc MAC addr tbl\n");
4985                 return -ENOMEM;
4986         }
4987
4988         if (!BNXT_HAS_DFLT_MAC_SET(bp)) {
4989                 if (BNXT_PF(bp))
4990                         return -EINVAL;
4991
4992                 /* Generate a random MAC address, if none was assigned by PF */
4993                 PMD_DRV_LOG(INFO, "VF MAC address not assigned by Host PF\n");
4994                 bnxt_eth_hw_addr_random(bp->mac_addr);
4995                 PMD_DRV_LOG(INFO,
4996                             "Assign random MAC:%02X:%02X:%02X:%02X:%02X:%02X\n",
4997                             bp->mac_addr[0], bp->mac_addr[1], bp->mac_addr[2],
4998                             bp->mac_addr[3], bp->mac_addr[4], bp->mac_addr[5]);
4999
5000                 rc = bnxt_hwrm_set_mac(bp);
5001                 if (rc)
5002                         return rc;
5003         }
5004
5005         /* Copy the permanent MAC from the FUNC_QCAPS response */
5006         memcpy(&eth_dev->data->mac_addrs[0], bp->mac_addr, RTE_ETHER_ADDR_LEN);
5007
5008         return rc;
5009 }
5010
5011 static int bnxt_restore_dflt_mac(struct bnxt *bp)
5012 {
5013         int rc = 0;
5014
5015         /* MAC is already configured in FW */
5016         if (BNXT_HAS_DFLT_MAC_SET(bp))
5017                 return 0;
5018
5019         /* Restore the old MAC configured */
5020         rc = bnxt_hwrm_set_mac(bp);
5021         if (rc)
5022                 PMD_DRV_LOG(ERR, "Failed to restore MAC address\n");
5023
5024         return rc;
5025 }
5026
5027 static void bnxt_config_vf_req_fwd(struct bnxt *bp)
5028 {
5029         if (!BNXT_PF(bp))
5030                 return;
5031
5032         memset(bp->pf->vf_req_fwd, 0, sizeof(bp->pf->vf_req_fwd));
5033
5034         if (!(bp->fw_cap & BNXT_FW_CAP_LINK_ADMIN))
5035                 BNXT_HWRM_CMD_TO_FORWARD(HWRM_PORT_PHY_QCFG);
5036         BNXT_HWRM_CMD_TO_FORWARD(HWRM_FUNC_CFG);
5037         BNXT_HWRM_CMD_TO_FORWARD(HWRM_FUNC_VF_CFG);
5038         BNXT_HWRM_CMD_TO_FORWARD(HWRM_CFA_L2_FILTER_ALLOC);
5039         BNXT_HWRM_CMD_TO_FORWARD(HWRM_OEM_CMD);
5040 }
5041
5042 struct bnxt *
5043 bnxt_get_bp(uint16_t port)
5044 {
5045         struct bnxt *bp;
5046         struct rte_eth_dev *dev;
5047
5048         if (!rte_eth_dev_is_valid_port(port)) {
5049                 PMD_DRV_LOG(ERR, "Invalid port %d\n", port);
5050                 return NULL;
5051         }
5052
5053         dev = &rte_eth_devices[port];
5054         if (!is_bnxt_supported(dev)) {
5055                 PMD_DRV_LOG(ERR, "Device %d not supported\n", port);
5056                 return NULL;
5057         }
5058
5059         bp = (struct bnxt *)dev->data->dev_private;
5060         if (!BNXT_TRUFLOW_EN(bp)) {
5061                 PMD_DRV_LOG(ERR, "TRUFLOW not enabled\n");
5062                 return NULL;
5063         }
5064
5065         return bp;
5066 }
5067
5068 uint16_t
5069 bnxt_get_svif(uint16_t port_id, bool func_svif,
5070               enum bnxt_ulp_intf_type type)
5071 {
5072         struct rte_eth_dev *eth_dev;
5073         struct bnxt *bp;
5074
5075         eth_dev = &rte_eth_devices[port_id];
5076         if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
5077                 struct bnxt_representor *vfr = eth_dev->data->dev_private;
5078                 if (!vfr)
5079                         return 0;
5080
5081                 if (type == BNXT_ULP_INTF_TYPE_VF_REP)
5082                         return vfr->svif;
5083
5084                 eth_dev = vfr->parent_dev;
5085         }
5086
5087         bp = eth_dev->data->dev_private;
5088
5089         return func_svif ? bp->func_svif : bp->port_svif;
5090 }
5091
5092 void
5093 bnxt_get_iface_mac(uint16_t port, enum bnxt_ulp_intf_type type,
5094                    uint8_t *mac, uint8_t *parent_mac)
5095 {
5096         struct rte_eth_dev *eth_dev;
5097         struct bnxt *bp;
5098
5099         if (type != BNXT_ULP_INTF_TYPE_TRUSTED_VF &&
5100             type != BNXT_ULP_INTF_TYPE_PF)
5101                 return;
5102
5103         eth_dev = &rte_eth_devices[port];
5104         bp = eth_dev->data->dev_private;
5105         memcpy(mac, bp->mac_addr, RTE_ETHER_ADDR_LEN);
5106
5107         if (type == BNXT_ULP_INTF_TYPE_TRUSTED_VF)
5108                 memcpy(parent_mac, bp->parent->mac_addr, RTE_ETHER_ADDR_LEN);
5109 }
5110
5111 uint16_t
5112 bnxt_get_parent_vnic_id(uint16_t port, enum bnxt_ulp_intf_type type)
5113 {
5114         struct rte_eth_dev *eth_dev;
5115         struct bnxt *bp;
5116
5117         if (type != BNXT_ULP_INTF_TYPE_TRUSTED_VF)
5118                 return 0;
5119
5120         eth_dev = &rte_eth_devices[port];
5121         bp = eth_dev->data->dev_private;
5122
5123         return bp->parent->vnic;
5124 }
5125 uint16_t
5126 bnxt_get_vnic_id(uint16_t port, enum bnxt_ulp_intf_type type)
5127 {
5128         struct rte_eth_dev *eth_dev;
5129         struct bnxt_vnic_info *vnic;
5130         struct bnxt *bp;
5131
5132         eth_dev = &rte_eth_devices[port];
5133         if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
5134                 struct bnxt_representor *vfr = eth_dev->data->dev_private;
5135                 if (!vfr)
5136                         return 0;
5137
5138                 if (type == BNXT_ULP_INTF_TYPE_VF_REP)
5139                         return vfr->dflt_vnic_id;
5140
5141                 eth_dev = vfr->parent_dev;
5142         }
5143
5144         bp = eth_dev->data->dev_private;
5145
5146         vnic = BNXT_GET_DEFAULT_VNIC(bp);
5147
5148         return vnic->fw_vnic_id;
5149 }
5150
5151 uint16_t
5152 bnxt_get_fw_func_id(uint16_t port, enum bnxt_ulp_intf_type type)
5153 {
5154         struct rte_eth_dev *eth_dev;
5155         struct bnxt *bp;
5156
5157         eth_dev = &rte_eth_devices[port];
5158         if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
5159                 struct bnxt_representor *vfr = eth_dev->data->dev_private;
5160                 if (!vfr)
5161                         return 0;
5162
5163                 if (type == BNXT_ULP_INTF_TYPE_VF_REP)
5164                         return vfr->fw_fid;
5165
5166                 eth_dev = vfr->parent_dev;
5167         }
5168
5169         bp = eth_dev->data->dev_private;
5170
5171         return bp->fw_fid;
5172 }
5173
5174 enum bnxt_ulp_intf_type
5175 bnxt_get_interface_type(uint16_t port)
5176 {
5177         struct rte_eth_dev *eth_dev;
5178         struct bnxt *bp;
5179
5180         eth_dev = &rte_eth_devices[port];
5181         if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev))
5182                 return BNXT_ULP_INTF_TYPE_VF_REP;
5183
5184         bp = eth_dev->data->dev_private;
5185         if (BNXT_PF(bp))
5186                 return BNXT_ULP_INTF_TYPE_PF;
5187         else if (BNXT_VF_IS_TRUSTED(bp))
5188                 return BNXT_ULP_INTF_TYPE_TRUSTED_VF;
5189         else if (BNXT_VF(bp))
5190                 return BNXT_ULP_INTF_TYPE_VF;
5191
5192         return BNXT_ULP_INTF_TYPE_INVALID;
5193 }
5194
5195 uint16_t
5196 bnxt_get_phy_port_id(uint16_t port_id)
5197 {
5198         struct bnxt_representor *vfr;
5199         struct rte_eth_dev *eth_dev;
5200         struct bnxt *bp;
5201
5202         eth_dev = &rte_eth_devices[port_id];
5203         if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
5204                 vfr = eth_dev->data->dev_private;
5205                 if (!vfr)
5206                         return 0;
5207
5208                 eth_dev = vfr->parent_dev;
5209         }
5210
5211         bp = eth_dev->data->dev_private;
5212
5213         return BNXT_PF(bp) ? bp->pf->port_id : bp->parent->port_id;
5214 }
5215
5216 uint16_t
5217 bnxt_get_parif(uint16_t port_id, enum bnxt_ulp_intf_type type)
5218 {
5219         struct rte_eth_dev *eth_dev;
5220         struct bnxt *bp;
5221
5222         eth_dev = &rte_eth_devices[port_id];
5223         if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
5224                 struct bnxt_representor *vfr = eth_dev->data->dev_private;
5225                 if (!vfr)
5226                         return 0;
5227
5228                 if (type == BNXT_ULP_INTF_TYPE_VF_REP)
5229                         return vfr->fw_fid - 1;
5230
5231                 eth_dev = vfr->parent_dev;
5232         }
5233
5234         bp = eth_dev->data->dev_private;
5235
5236         return BNXT_PF(bp) ? bp->fw_fid - 1 : bp->parent->fid - 1;
5237 }
5238
5239 uint16_t
5240 bnxt_get_vport(uint16_t port_id)
5241 {
5242         return (1 << bnxt_get_phy_port_id(port_id));
5243 }
5244
5245 static void bnxt_alloc_error_recovery_info(struct bnxt *bp)
5246 {
5247         struct bnxt_error_recovery_info *info = bp->recovery_info;
5248
5249         if (info) {
5250                 if (!(bp->fw_cap & BNXT_FW_CAP_HCOMM_FW_STATUS))
5251                         memset(info, 0, sizeof(*info));
5252                 return;
5253         }
5254
5255         if (!(bp->fw_cap & BNXT_FW_CAP_ERROR_RECOVERY))
5256                 return;
5257
5258         info = rte_zmalloc("bnxt_hwrm_error_recovery_qcfg",
5259                            sizeof(*info), 0);
5260         if (!info)
5261                 bp->fw_cap &= ~BNXT_FW_CAP_ERROR_RECOVERY;
5262
5263         bp->recovery_info = info;
5264 }
5265
5266 static void bnxt_check_fw_status(struct bnxt *bp)
5267 {
5268         uint32_t fw_status;
5269
5270         if (!(bp->recovery_info &&
5271               (bp->fw_cap & BNXT_FW_CAP_HCOMM_FW_STATUS)))
5272                 return;
5273
5274         fw_status = bnxt_read_fw_status_reg(bp, BNXT_FW_STATUS_REG);
5275         if (fw_status != BNXT_FW_STATUS_HEALTHY)
5276                 PMD_DRV_LOG(ERR, "Firmware not responding, status: %#x\n",
5277                             fw_status);
5278 }
5279
5280 static int bnxt_map_hcomm_fw_status_reg(struct bnxt *bp)
5281 {
5282         struct bnxt_error_recovery_info *info = bp->recovery_info;
5283         uint32_t status_loc;
5284         uint32_t sig_ver;
5285
5286         rte_write32(HCOMM_STATUS_STRUCT_LOC, (uint8_t *)bp->bar0 +
5287                     BNXT_GRCPF_REG_WINDOW_BASE_OUT + 4);
5288         sig_ver = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
5289                                    BNXT_GRCP_WINDOW_2_BASE +
5290                                    offsetof(struct hcomm_status,
5291                                             sig_ver)));
5292         /* If the signature is absent, then FW does not support this feature */
5293         if ((sig_ver & HCOMM_STATUS_SIGNATURE_MASK) !=
5294             HCOMM_STATUS_SIGNATURE_VAL)
5295                 return 0;
5296
5297         if (!info) {
5298                 info = rte_zmalloc("bnxt_hwrm_error_recovery_qcfg",
5299                                    sizeof(*info), 0);
5300                 if (!info)
5301                         return -ENOMEM;
5302                 bp->recovery_info = info;
5303         } else {
5304                 memset(info, 0, sizeof(*info));
5305         }
5306
5307         status_loc = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
5308                                       BNXT_GRCP_WINDOW_2_BASE +
5309                                       offsetof(struct hcomm_status,
5310                                                fw_status_loc)));
5311
5312         /* Only pre-map the FW health status GRC register */
5313         if (BNXT_FW_STATUS_REG_TYPE(status_loc) != BNXT_FW_STATUS_REG_TYPE_GRC)
5314                 return 0;
5315
5316         info->status_regs[BNXT_FW_STATUS_REG] = status_loc;
5317         info->mapped_status_regs[BNXT_FW_STATUS_REG] =
5318                 BNXT_GRCP_WINDOW_2_BASE + (status_loc & BNXT_GRCP_OFFSET_MASK);
5319
5320         rte_write32((status_loc & BNXT_GRCP_BASE_MASK), (uint8_t *)bp->bar0 +
5321                     BNXT_GRCPF_REG_WINDOW_BASE_OUT + 4);
5322
5323         bp->fw_cap |= BNXT_FW_CAP_HCOMM_FW_STATUS;
5324
5325         return 0;
5326 }
5327
5328 /* This function gets the FW version along with the
5329  * capabilities(MAX and current) of the function, vnic,
5330  * error recovery, phy and other chip related info
5331  */
5332 static int bnxt_get_config(struct bnxt *bp)
5333 {
5334         uint16_t mtu;
5335         int rc = 0;
5336
5337         bp->fw_cap = 0;
5338
5339         rc = bnxt_map_hcomm_fw_status_reg(bp);
5340         if (rc)
5341                 return rc;
5342
5343         rc = bnxt_hwrm_ver_get(bp, DFLT_HWRM_CMD_TIMEOUT);
5344         if (rc) {
5345                 bnxt_check_fw_status(bp);
5346                 return rc;
5347         }
5348
5349         rc = bnxt_hwrm_func_reset(bp);
5350         if (rc)
5351                 return -EIO;
5352
5353         rc = bnxt_hwrm_vnic_qcaps(bp);
5354         if (rc)
5355                 return rc;
5356
5357         rc = bnxt_hwrm_queue_qportcfg(bp);
5358         if (rc)
5359                 return rc;
5360
5361         /* Get the MAX capabilities for this function.
5362          * This function also allocates context memory for TQM rings and
5363          * informs the firmware about this allocated backing store memory.
5364          */
5365         rc = bnxt_hwrm_func_qcaps(bp);
5366         if (rc)
5367                 return rc;
5368
5369         rc = bnxt_hwrm_func_qcfg(bp, &mtu);
5370         if (rc)
5371                 return rc;
5372
5373         rc = bnxt_hwrm_cfa_adv_flow_mgmt_qcaps(bp);
5374         if (rc)
5375                 return rc;
5376
5377         bnxt_hwrm_port_mac_qcfg(bp);
5378
5379         bnxt_hwrm_parent_pf_qcfg(bp);
5380
5381         bnxt_hwrm_port_phy_qcaps(bp);
5382
5383         bnxt_alloc_error_recovery_info(bp);
5384         /* Get the adapter error recovery support info */
5385         rc = bnxt_hwrm_error_recovery_qcfg(bp);
5386         if (rc)
5387                 bp->fw_cap &= ~BNXT_FW_CAP_ERROR_RECOVERY;
5388
5389         bnxt_hwrm_port_led_qcaps(bp);
5390
5391         return 0;
5392 }
5393
5394 static int
5395 bnxt_init_locks(struct bnxt *bp)
5396 {
5397         int err;
5398
5399         err = pthread_mutex_init(&bp->flow_lock, NULL);
5400         if (err) {
5401                 PMD_DRV_LOG(ERR, "Unable to initialize flow_lock\n");
5402                 return err;
5403         }
5404
5405         err = pthread_mutex_init(&bp->def_cp_lock, NULL);
5406         if (err) {
5407                 PMD_DRV_LOG(ERR, "Unable to initialize def_cp_lock\n");
5408                 return err;
5409         }
5410
5411         err = pthread_mutex_init(&bp->health_check_lock, NULL);
5412         if (err) {
5413                 PMD_DRV_LOG(ERR, "Unable to initialize health_check_lock\n");
5414                 return err;
5415         }
5416
5417         err = pthread_mutex_init(&bp->err_recovery_lock, NULL);
5418         if (err)
5419                 PMD_DRV_LOG(ERR, "Unable to initialize err_recovery_lock\n");
5420
5421         return err;
5422 }
5423
5424 static int bnxt_init_resources(struct bnxt *bp, bool reconfig_dev)
5425 {
5426         int rc = 0;
5427
5428         rc = bnxt_get_config(bp);
5429         if (rc)
5430                 return rc;
5431
5432         if (!reconfig_dev) {
5433                 rc = bnxt_setup_mac_addr(bp->eth_dev);
5434                 if (rc)
5435                         return rc;
5436         } else {
5437                 rc = bnxt_restore_dflt_mac(bp);
5438                 if (rc)
5439                         return rc;
5440         }
5441
5442         bnxt_config_vf_req_fwd(bp);
5443
5444         rc = bnxt_hwrm_func_driver_register(bp);
5445         if (rc) {
5446                 PMD_DRV_LOG(ERR, "Failed to register driver");
5447                 return -EBUSY;
5448         }
5449
5450         if (BNXT_PF(bp)) {
5451                 if (bp->pdev->max_vfs) {
5452                         rc = bnxt_hwrm_allocate_vfs(bp, bp->pdev->max_vfs);
5453                         if (rc) {
5454                                 PMD_DRV_LOG(ERR, "Failed to allocate VFs\n");
5455                                 return rc;
5456                         }
5457                 } else {
5458                         rc = bnxt_hwrm_allocate_pf_only(bp);
5459                         if (rc) {
5460                                 PMD_DRV_LOG(ERR,
5461                                             "Failed to allocate PF resources");
5462                                 return rc;
5463                         }
5464                 }
5465         }
5466
5467         rc = bnxt_alloc_mem(bp, reconfig_dev);
5468         if (rc)
5469                 return rc;
5470
5471         rc = bnxt_setup_int(bp);
5472         if (rc)
5473                 return rc;
5474
5475         rc = bnxt_request_int(bp);
5476         if (rc)
5477                 return rc;
5478
5479         rc = bnxt_init_ctx_mem(bp);
5480         if (rc) {
5481                 PMD_DRV_LOG(ERR, "Failed to init adv_flow_counters\n");
5482                 return rc;
5483         }
5484
5485         return 0;
5486 }
5487
5488 static int
5489 bnxt_parse_devarg_accum_stats(__rte_unused const char *key,
5490                               const char *value, void *opaque_arg)
5491 {
5492         struct bnxt *bp = opaque_arg;
5493         unsigned long accum_stats;
5494         char *end = NULL;
5495
5496         if (!value || !opaque_arg) {
5497                 PMD_DRV_LOG(ERR,
5498                             "Invalid parameter passed to accum-stats devargs.\n");
5499                 return -EINVAL;
5500         }
5501
5502         accum_stats = strtoul(value, &end, 10);
5503         if (end == NULL || *end != '\0' ||
5504             (accum_stats == ULONG_MAX && errno == ERANGE)) {
5505                 PMD_DRV_LOG(ERR,
5506                             "Invalid parameter passed to accum-stats devargs.\n");
5507                 return -EINVAL;
5508         }
5509
5510         if (BNXT_DEVARG_ACCUM_STATS_INVALID(accum_stats)) {
5511                 PMD_DRV_LOG(ERR,
5512                             "Invalid value passed to accum-stats devargs.\n");
5513                 return -EINVAL;
5514         }
5515
5516         if (accum_stats) {
5517                 bp->flags2 |= BNXT_FLAGS2_ACCUM_STATS_EN;
5518                 PMD_DRV_LOG(INFO, "Host-based accum-stats feature enabled.\n");
5519         } else {
5520                 bp->flags2 &= ~BNXT_FLAGS2_ACCUM_STATS_EN;
5521                 PMD_DRV_LOG(INFO, "Host-based accum-stats feature disabled.\n");
5522         }
5523
5524         return 0;
5525 }
5526
5527 static int
5528 bnxt_parse_devarg_flow_xstat(__rte_unused const char *key,
5529                              const char *value, void *opaque_arg)
5530 {
5531         struct bnxt *bp = opaque_arg;
5532         unsigned long flow_xstat;
5533         char *end = NULL;
5534
5535         if (!value || !opaque_arg) {
5536                 PMD_DRV_LOG(ERR,
5537                             "Invalid parameter passed to flow_xstat devarg.\n");
5538                 return -EINVAL;
5539         }
5540
5541         flow_xstat = strtoul(value, &end, 10);
5542         if (end == NULL || *end != '\0' ||
5543             (flow_xstat == ULONG_MAX && errno == ERANGE)) {
5544                 PMD_DRV_LOG(ERR,
5545                             "Invalid parameter passed to flow_xstat devarg.\n");
5546                 return -EINVAL;
5547         }
5548
5549         if (BNXT_DEVARG_FLOW_XSTAT_INVALID(flow_xstat)) {
5550                 PMD_DRV_LOG(ERR,
5551                             "Invalid value passed to flow_xstat devarg.\n");
5552                 return -EINVAL;
5553         }
5554
5555         bp->flags |= BNXT_FLAG_FLOW_XSTATS_EN;
5556         if (BNXT_FLOW_XSTATS_EN(bp))
5557                 PMD_DRV_LOG(INFO, "flow_xstat feature enabled.\n");
5558
5559         return 0;
5560 }
5561
5562 static int
5563 bnxt_parse_devarg_max_num_kflows(__rte_unused const char *key,
5564                                         const char *value, void *opaque_arg)
5565 {
5566         struct bnxt *bp = opaque_arg;
5567         unsigned long max_num_kflows;
5568         char *end = NULL;
5569
5570         if (!value || !opaque_arg) {
5571                 PMD_DRV_LOG(ERR,
5572                         "Invalid parameter passed to max_num_kflows devarg.\n");
5573                 return -EINVAL;
5574         }
5575
5576         max_num_kflows = strtoul(value, &end, 10);
5577         if (end == NULL || *end != '\0' ||
5578                 (max_num_kflows == ULONG_MAX && errno == ERANGE)) {
5579                 PMD_DRV_LOG(ERR,
5580                         "Invalid parameter passed to max_num_kflows devarg.\n");
5581                 return -EINVAL;
5582         }
5583
5584         if (bnxt_devarg_max_num_kflow_invalid(max_num_kflows)) {
5585                 PMD_DRV_LOG(ERR,
5586                         "Invalid value passed to max_num_kflows devarg.\n");
5587                 return -EINVAL;
5588         }
5589
5590         bp->max_num_kflows = max_num_kflows;
5591         if (bp->max_num_kflows)
5592                 PMD_DRV_LOG(INFO, "max_num_kflows set as %ldK.\n",
5593                                 max_num_kflows);
5594
5595         return 0;
5596 }
5597
5598 static int
5599 bnxt_parse_devarg_app_id(__rte_unused const char *key,
5600                                  const char *value, void *opaque_arg)
5601 {
5602         struct bnxt *bp = opaque_arg;
5603         unsigned long app_id;
5604         char *end = NULL;
5605
5606         if (!value || !opaque_arg) {
5607                 PMD_DRV_LOG(ERR,
5608                             "Invalid parameter passed to app-id "
5609                             "devargs.\n");
5610                 return -EINVAL;
5611         }
5612
5613         app_id = strtoul(value, &end, 10);
5614         if (end == NULL || *end != '\0' ||
5615             (app_id == ULONG_MAX && errno == ERANGE)) {
5616                 PMD_DRV_LOG(ERR,
5617                             "Invalid parameter passed to app_id "
5618                             "devargs.\n");
5619                 return -EINVAL;
5620         }
5621
5622         if (BNXT_DEVARG_APP_ID_INVALID(app_id)) {
5623                 PMD_DRV_LOG(ERR, "Invalid app-id(%d) devargs.\n",
5624                             (uint16_t)app_id);
5625                 return -EINVAL;
5626         }
5627
5628         bp->app_id = app_id;
5629         PMD_DRV_LOG(INFO, "app-id=%d feature enabled.\n", (uint16_t)app_id);
5630
5631         return 0;
5632 }
5633
5634 static int
5635 bnxt_parse_devarg_rep_is_pf(__rte_unused const char *key,
5636                             const char *value, void *opaque_arg)
5637 {
5638         struct bnxt_representor *vfr_bp = opaque_arg;
5639         unsigned long rep_is_pf;
5640         char *end = NULL;
5641
5642         if (!value || !opaque_arg) {
5643                 PMD_DRV_LOG(ERR,
5644                             "Invalid parameter passed to rep_is_pf devargs.\n");
5645                 return -EINVAL;
5646         }
5647
5648         rep_is_pf = strtoul(value, &end, 10);
5649         if (end == NULL || *end != '\0' ||
5650             (rep_is_pf == ULONG_MAX && errno == ERANGE)) {
5651                 PMD_DRV_LOG(ERR,
5652                             "Invalid parameter passed to rep_is_pf devargs.\n");
5653                 return -EINVAL;
5654         }
5655
5656         if (BNXT_DEVARG_REP_IS_PF_INVALID(rep_is_pf)) {
5657                 PMD_DRV_LOG(ERR,
5658                             "Invalid value passed to rep_is_pf devargs.\n");
5659                 return -EINVAL;
5660         }
5661
5662         vfr_bp->flags |= rep_is_pf;
5663         if (BNXT_REP_PF(vfr_bp))
5664                 PMD_DRV_LOG(INFO, "PF representor\n");
5665         else
5666                 PMD_DRV_LOG(INFO, "VF representor\n");
5667
5668         return 0;
5669 }
5670
5671 static int
5672 bnxt_parse_devarg_rep_based_pf(__rte_unused const char *key,
5673                                const char *value, void *opaque_arg)
5674 {
5675         struct bnxt_representor *vfr_bp = opaque_arg;
5676         unsigned long rep_based_pf;
5677         char *end = NULL;
5678
5679         if (!value || !opaque_arg) {
5680                 PMD_DRV_LOG(ERR,
5681                             "Invalid parameter passed to rep_based_pf "
5682                             "devargs.\n");
5683                 return -EINVAL;
5684         }
5685
5686         rep_based_pf = strtoul(value, &end, 10);
5687         if (end == NULL || *end != '\0' ||
5688             (rep_based_pf == ULONG_MAX && errno == ERANGE)) {
5689                 PMD_DRV_LOG(ERR,
5690                             "Invalid parameter passed to rep_based_pf "
5691                             "devargs.\n");
5692                 return -EINVAL;
5693         }
5694
5695         if (BNXT_DEVARG_REP_BASED_PF_INVALID(rep_based_pf)) {
5696                 PMD_DRV_LOG(ERR,
5697                             "Invalid value passed to rep_based_pf devargs.\n");
5698                 return -EINVAL;
5699         }
5700
5701         vfr_bp->rep_based_pf = rep_based_pf;
5702         vfr_bp->flags |= BNXT_REP_BASED_PF_VALID;
5703
5704         PMD_DRV_LOG(INFO, "rep-based-pf = %d\n", vfr_bp->rep_based_pf);
5705
5706         return 0;
5707 }
5708
5709 static int
5710 bnxt_parse_devarg_rep_q_r2f(__rte_unused const char *key,
5711                             const char *value, void *opaque_arg)
5712 {
5713         struct bnxt_representor *vfr_bp = opaque_arg;
5714         unsigned long rep_q_r2f;
5715         char *end = NULL;
5716
5717         if (!value || !opaque_arg) {
5718                 PMD_DRV_LOG(ERR,
5719                             "Invalid parameter passed to rep_q_r2f "
5720                             "devargs.\n");
5721                 return -EINVAL;
5722         }
5723
5724         rep_q_r2f = strtoul(value, &end, 10);
5725         if (end == NULL || *end != '\0' ||
5726             (rep_q_r2f == ULONG_MAX && errno == ERANGE)) {
5727                 PMD_DRV_LOG(ERR,
5728                             "Invalid parameter passed to rep_q_r2f "
5729                             "devargs.\n");
5730                 return -EINVAL;
5731         }
5732
5733         if (BNXT_DEVARG_REP_Q_R2F_INVALID(rep_q_r2f)) {
5734                 PMD_DRV_LOG(ERR,
5735                             "Invalid value passed to rep_q_r2f devargs.\n");
5736                 return -EINVAL;
5737         }
5738
5739         vfr_bp->rep_q_r2f = rep_q_r2f;
5740         vfr_bp->flags |= BNXT_REP_Q_R2F_VALID;
5741         PMD_DRV_LOG(INFO, "rep-q-r2f = %d\n", vfr_bp->rep_q_r2f);
5742
5743         return 0;
5744 }
5745
5746 static int
5747 bnxt_parse_devarg_rep_q_f2r(__rte_unused const char *key,
5748                             const char *value, void *opaque_arg)
5749 {
5750         struct bnxt_representor *vfr_bp = opaque_arg;
5751         unsigned long rep_q_f2r;
5752         char *end = NULL;
5753
5754         if (!value || !opaque_arg) {
5755                 PMD_DRV_LOG(ERR,
5756                             "Invalid parameter passed to rep_q_f2r "
5757                             "devargs.\n");
5758                 return -EINVAL;
5759         }
5760
5761         rep_q_f2r = strtoul(value, &end, 10);
5762         if (end == NULL || *end != '\0' ||
5763             (rep_q_f2r == ULONG_MAX && errno == ERANGE)) {
5764                 PMD_DRV_LOG(ERR,
5765                             "Invalid parameter passed to rep_q_f2r "
5766                             "devargs.\n");
5767                 return -EINVAL;
5768         }
5769
5770         if (BNXT_DEVARG_REP_Q_F2R_INVALID(rep_q_f2r)) {
5771                 PMD_DRV_LOG(ERR,
5772                             "Invalid value passed to rep_q_f2r devargs.\n");
5773                 return -EINVAL;
5774         }
5775
5776         vfr_bp->rep_q_f2r = rep_q_f2r;
5777         vfr_bp->flags |= BNXT_REP_Q_F2R_VALID;
5778         PMD_DRV_LOG(INFO, "rep-q-f2r = %d\n", vfr_bp->rep_q_f2r);
5779
5780         return 0;
5781 }
5782
5783 static int
5784 bnxt_parse_devarg_rep_fc_r2f(__rte_unused const char *key,
5785                              const char *value, void *opaque_arg)
5786 {
5787         struct bnxt_representor *vfr_bp = opaque_arg;
5788         unsigned long rep_fc_r2f;
5789         char *end = NULL;
5790
5791         if (!value || !opaque_arg) {
5792                 PMD_DRV_LOG(ERR,
5793                             "Invalid parameter passed to rep_fc_r2f "
5794                             "devargs.\n");
5795                 return -EINVAL;
5796         }
5797
5798         rep_fc_r2f = strtoul(value, &end, 10);
5799         if (end == NULL || *end != '\0' ||
5800             (rep_fc_r2f == ULONG_MAX && errno == ERANGE)) {
5801                 PMD_DRV_LOG(ERR,
5802                             "Invalid parameter passed to rep_fc_r2f "
5803                             "devargs.\n");
5804                 return -EINVAL;
5805         }
5806
5807         if (BNXT_DEVARG_REP_FC_R2F_INVALID(rep_fc_r2f)) {
5808                 PMD_DRV_LOG(ERR,
5809                             "Invalid value passed to rep_fc_r2f devargs.\n");
5810                 return -EINVAL;
5811         }
5812
5813         vfr_bp->flags |= BNXT_REP_FC_R2F_VALID;
5814         vfr_bp->rep_fc_r2f = rep_fc_r2f;
5815         PMD_DRV_LOG(INFO, "rep-fc-r2f = %lu\n", rep_fc_r2f);
5816
5817         return 0;
5818 }
5819
5820 static int
5821 bnxt_parse_devarg_rep_fc_f2r(__rte_unused const char *key,
5822                              const char *value, void *opaque_arg)
5823 {
5824         struct bnxt_representor *vfr_bp = opaque_arg;
5825         unsigned long rep_fc_f2r;
5826         char *end = NULL;
5827
5828         if (!value || !opaque_arg) {
5829                 PMD_DRV_LOG(ERR,
5830                             "Invalid parameter passed to rep_fc_f2r "
5831                             "devargs.\n");
5832                 return -EINVAL;
5833         }
5834
5835         rep_fc_f2r = strtoul(value, &end, 10);
5836         if (end == NULL || *end != '\0' ||
5837             (rep_fc_f2r == ULONG_MAX && errno == ERANGE)) {
5838                 PMD_DRV_LOG(ERR,
5839                             "Invalid parameter passed to rep_fc_f2r "
5840                             "devargs.\n");
5841                 return -EINVAL;
5842         }
5843
5844         if (BNXT_DEVARG_REP_FC_F2R_INVALID(rep_fc_f2r)) {
5845                 PMD_DRV_LOG(ERR,
5846                             "Invalid value passed to rep_fc_f2r devargs.\n");
5847                 return -EINVAL;
5848         }
5849
5850         vfr_bp->flags |= BNXT_REP_FC_F2R_VALID;
5851         vfr_bp->rep_fc_f2r = rep_fc_f2r;
5852         PMD_DRV_LOG(INFO, "rep-fc-f2r = %lu\n", rep_fc_f2r);
5853
5854         return 0;
5855 }
5856
5857 static int
5858 bnxt_parse_dev_args(struct bnxt *bp, struct rte_devargs *devargs)
5859 {
5860         struct rte_kvargs *kvlist;
5861         int ret;
5862
5863         if (devargs == NULL)
5864                 return 0;
5865
5866         kvlist = rte_kvargs_parse(devargs->args, bnxt_dev_args);
5867         if (kvlist == NULL)
5868                 return -EINVAL;
5869
5870         /*
5871          * Handler for "flow_xstat" devarg.
5872          * Invoked as for ex: "-a 0000:00:0d.0,flow_xstat=1"
5873          */
5874         ret = rte_kvargs_process(kvlist, BNXT_DEVARG_FLOW_XSTAT,
5875                                  bnxt_parse_devarg_flow_xstat, bp);
5876         if (ret)
5877                 goto err;
5878
5879         /*
5880          * Handler for "accum-stats" devarg.
5881          * Invoked as for ex: "-a 0000:00:0d.0,accum-stats=1"
5882          */
5883         rte_kvargs_process(kvlist, BNXT_DEVARG_ACCUM_STATS,
5884                            bnxt_parse_devarg_accum_stats, bp);
5885         /*
5886          * Handler for "max_num_kflows" devarg.
5887          * Invoked as for ex: "-a 000:00:0d.0,max_num_kflows=32"
5888          */
5889         ret = rte_kvargs_process(kvlist, BNXT_DEVARG_MAX_NUM_KFLOWS,
5890                                  bnxt_parse_devarg_max_num_kflows, bp);
5891         if (ret)
5892                 goto err;
5893
5894 err:
5895         /*
5896          * Handler for "app-id" devarg.
5897          * Invoked as for ex: "-a 000:00:0d.0,app-id=1"
5898          */
5899         rte_kvargs_process(kvlist, BNXT_DEVARG_APP_ID,
5900                            bnxt_parse_devarg_app_id, bp);
5901
5902         rte_kvargs_free(kvlist);
5903         return ret;
5904 }
5905
5906 static int bnxt_alloc_switch_domain(struct bnxt *bp)
5907 {
5908         int rc = 0;
5909
5910         if (BNXT_PF(bp) || BNXT_VF_IS_TRUSTED(bp)) {
5911                 rc = rte_eth_switch_domain_alloc(&bp->switch_domain_id);
5912                 if (rc)
5913                         PMD_DRV_LOG(ERR,
5914                                     "Failed to alloc switch domain: %d\n", rc);
5915                 else
5916                         PMD_DRV_LOG(INFO,
5917                                     "Switch domain allocated %d\n",
5918                                     bp->switch_domain_id);
5919         }
5920
5921         return rc;
5922 }
5923
5924 /* Allocate and initialize various fields in bnxt struct that
5925  * need to be allocated/destroyed only once in the lifetime of the driver
5926  */
5927 static int bnxt_drv_init(struct rte_eth_dev *eth_dev)
5928 {
5929         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
5930         struct bnxt *bp = eth_dev->data->dev_private;
5931         int rc = 0;
5932
5933         bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
5934
5935         if (bnxt_vf_pciid(pci_dev->id.device_id))
5936                 bp->flags |= BNXT_FLAG_VF;
5937
5938         if (bnxt_p5_device(pci_dev->id.device_id))
5939                 bp->flags |= BNXT_FLAG_CHIP_P5;
5940
5941         if (pci_dev->id.device_id == BROADCOM_DEV_ID_58802 ||
5942             pci_dev->id.device_id == BROADCOM_DEV_ID_58804 ||
5943             pci_dev->id.device_id == BROADCOM_DEV_ID_58808 ||
5944             pci_dev->id.device_id == BROADCOM_DEV_ID_58802_VF)
5945                 bp->flags |= BNXT_FLAG_STINGRAY;
5946
5947         if (BNXT_TRUFLOW_EN(bp)) {
5948                 /* extra mbuf field is required to store CFA code from mark */
5949                 static const struct rte_mbuf_dynfield bnxt_cfa_code_dynfield_desc = {
5950                         .name = RTE_PMD_BNXT_CFA_CODE_DYNFIELD_NAME,
5951                         .size = sizeof(bnxt_cfa_code_dynfield_t),
5952                         .align = __alignof__(bnxt_cfa_code_dynfield_t),
5953                 };
5954                 bnxt_cfa_code_dynfield_offset =
5955                         rte_mbuf_dynfield_register(&bnxt_cfa_code_dynfield_desc);
5956                 if (bnxt_cfa_code_dynfield_offset < 0) {
5957                         PMD_DRV_LOG(ERR,
5958                             "Failed to register mbuf field for TruFlow mark\n");
5959                         return -rte_errno;
5960                 }
5961         }
5962
5963         rc = bnxt_map_pci_bars(eth_dev);
5964         if (rc) {
5965                 PMD_DRV_LOG(ERR,
5966                             "Failed to initialize board rc: %x\n", rc);
5967                 return rc;
5968         }
5969
5970         rc = bnxt_alloc_pf_info(bp);
5971         if (rc)
5972                 return rc;
5973
5974         rc = bnxt_alloc_link_info(bp);
5975         if (rc)
5976                 return rc;
5977
5978         rc = bnxt_alloc_parent_info(bp);
5979         if (rc)
5980                 return rc;
5981
5982         rc = bnxt_alloc_hwrm_resources(bp);
5983         if (rc) {
5984                 PMD_DRV_LOG(ERR,
5985                             "Failed to allocate response buffer rc: %x\n", rc);
5986                 return rc;
5987         }
5988         rc = bnxt_alloc_leds_info(bp);
5989         if (rc)
5990                 return rc;
5991
5992         rc = bnxt_alloc_cos_queues(bp);
5993         if (rc)
5994                 return rc;
5995
5996         rc = bnxt_init_locks(bp);
5997         if (rc)
5998                 return rc;
5999
6000         rc = bnxt_alloc_switch_domain(bp);
6001         if (rc)
6002                 return rc;
6003
6004         return rc;
6005 }
6006
6007 static int
6008 bnxt_dev_init(struct rte_eth_dev *eth_dev, void *params __rte_unused)
6009 {
6010         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
6011         static int version_printed;
6012         struct bnxt *bp;
6013         int rc;
6014
6015         if (version_printed++ == 0)
6016                 PMD_DRV_LOG(INFO, "%s\n", bnxt_version);
6017
6018         eth_dev->dev_ops = &bnxt_dev_ops;
6019         eth_dev->rx_queue_count = bnxt_rx_queue_count_op;
6020         eth_dev->rx_descriptor_status = bnxt_rx_descriptor_status_op;
6021         eth_dev->tx_descriptor_status = bnxt_tx_descriptor_status_op;
6022         eth_dev->rx_pkt_burst = &bnxt_recv_pkts;
6023         eth_dev->tx_pkt_burst = &bnxt_xmit_pkts;
6024
6025         /*
6026          * For secondary processes, we don't initialise any further
6027          * as primary has already done this work.
6028          */
6029         if (rte_eal_process_type() != RTE_PROC_PRIMARY)
6030                 return 0;
6031
6032         rte_eth_copy_pci_info(eth_dev, pci_dev);
6033         eth_dev->data->dev_flags |= RTE_ETH_DEV_AUTOFILL_QUEUE_XSTATS;
6034
6035         bp = eth_dev->data->dev_private;
6036
6037         /* Parse dev arguments passed on when starting the DPDK application. */
6038         rc = bnxt_parse_dev_args(bp, pci_dev->device.devargs);
6039         if (rc)
6040                 goto error_free;
6041
6042         rc = bnxt_drv_init(eth_dev);
6043         if (rc)
6044                 goto error_free;
6045
6046         rc = bnxt_init_resources(bp, false);
6047         if (rc)
6048                 goto error_free;
6049
6050         rc = bnxt_alloc_stats_mem(bp);
6051         if (rc)
6052                 goto error_free;
6053
6054         PMD_DRV_LOG(INFO,
6055                     "Found %s device at mem %" PRIX64 ", node addr %pM\n",
6056                     DRV_MODULE_NAME,
6057                     pci_dev->mem_resource[0].phys_addr,
6058                     pci_dev->mem_resource[0].addr);
6059
6060         return 0;
6061
6062 error_free:
6063         bnxt_dev_uninit(eth_dev);
6064         return rc;
6065 }
6066
6067
6068 static void bnxt_free_ctx_mem_buf(struct bnxt_ctx_mem_buf_info *ctx)
6069 {
6070         if (!ctx)
6071                 return;
6072
6073         if (ctx->va)
6074                 rte_free(ctx->va);
6075
6076         ctx->va = NULL;
6077         ctx->dma = RTE_BAD_IOVA;
6078         ctx->ctx_id = BNXT_CTX_VAL_INVAL;
6079 }
6080
6081 static void bnxt_unregister_fc_ctx_mem(struct bnxt *bp)
6082 {
6083         bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_RX,
6084                                   CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
6085                                   bp->flow_stat->rx_fc_out_tbl.ctx_id,
6086                                   bp->flow_stat->max_fc,
6087                                   false);
6088
6089         bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_TX,
6090                                   CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
6091                                   bp->flow_stat->tx_fc_out_tbl.ctx_id,
6092                                   bp->flow_stat->max_fc,
6093                                   false);
6094
6095         if (bp->flow_stat->rx_fc_in_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
6096                 bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->rx_fc_in_tbl.ctx_id);
6097         bp->flow_stat->rx_fc_in_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
6098
6099         if (bp->flow_stat->rx_fc_out_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
6100                 bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->rx_fc_out_tbl.ctx_id);
6101         bp->flow_stat->rx_fc_out_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
6102
6103         if (bp->flow_stat->tx_fc_in_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
6104                 bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->tx_fc_in_tbl.ctx_id);
6105         bp->flow_stat->tx_fc_in_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
6106
6107         if (bp->flow_stat->tx_fc_out_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
6108                 bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->tx_fc_out_tbl.ctx_id);
6109         bp->flow_stat->tx_fc_out_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
6110 }
6111
6112 static void bnxt_uninit_fc_ctx_mem(struct bnxt *bp)
6113 {
6114         bnxt_unregister_fc_ctx_mem(bp);
6115
6116         bnxt_free_ctx_mem_buf(&bp->flow_stat->rx_fc_in_tbl);
6117         bnxt_free_ctx_mem_buf(&bp->flow_stat->rx_fc_out_tbl);
6118         bnxt_free_ctx_mem_buf(&bp->flow_stat->tx_fc_in_tbl);
6119         bnxt_free_ctx_mem_buf(&bp->flow_stat->tx_fc_out_tbl);
6120 }
6121
6122 static void bnxt_uninit_ctx_mem(struct bnxt *bp)
6123 {
6124         if (BNXT_FLOW_XSTATS_EN(bp))
6125                 bnxt_uninit_fc_ctx_mem(bp);
6126 }
6127
6128 static void
6129 bnxt_free_error_recovery_info(struct bnxt *bp)
6130 {
6131         rte_free(bp->recovery_info);
6132         bp->recovery_info = NULL;
6133         bp->fw_cap &= ~BNXT_FW_CAP_ERROR_RECOVERY;
6134 }
6135
6136 static int
6137 bnxt_uninit_resources(struct bnxt *bp, bool reconfig_dev)
6138 {
6139         int rc;
6140
6141         bnxt_free_int(bp);
6142         bnxt_free_mem(bp, reconfig_dev);
6143
6144         bnxt_hwrm_func_buf_unrgtr(bp);
6145         if (bp->pf != NULL) {
6146                 rte_free(bp->pf->vf_req_buf);
6147                 bp->pf->vf_req_buf = NULL;
6148         }
6149
6150         rc = bnxt_hwrm_func_driver_unregister(bp, 0);
6151         bp->flags &= ~BNXT_FLAG_REGISTERED;
6152         bnxt_free_ctx_mem(bp);
6153         if (!reconfig_dev) {
6154                 bnxt_free_hwrm_resources(bp);
6155                 bnxt_free_error_recovery_info(bp);
6156         }
6157
6158         bnxt_uninit_ctx_mem(bp);
6159
6160         bnxt_free_flow_stats_info(bp);
6161         if (bp->rep_info != NULL)
6162                 bnxt_free_switch_domain(bp);
6163         bnxt_free_rep_info(bp);
6164         rte_free(bp->ptp_cfg);
6165         bp->ptp_cfg = NULL;
6166         return rc;
6167 }
6168
6169 static int
6170 bnxt_dev_uninit(struct rte_eth_dev *eth_dev)
6171 {
6172         if (rte_eal_process_type() != RTE_PROC_PRIMARY)
6173                 return -EPERM;
6174
6175         PMD_DRV_LOG(DEBUG, "Calling Device uninit\n");
6176
6177         if (eth_dev->state != RTE_ETH_DEV_UNUSED)
6178                 bnxt_dev_close_op(eth_dev);
6179
6180         return 0;
6181 }
6182
6183 static int bnxt_pci_remove_dev_with_reps(struct rte_eth_dev *eth_dev)
6184 {
6185         struct bnxt *bp = eth_dev->data->dev_private;
6186         struct rte_eth_dev *vf_rep_eth_dev;
6187         int ret = 0, i;
6188
6189         if (!bp)
6190                 return -EINVAL;
6191
6192         for (i = 0; i < bp->num_reps; i++) {
6193                 vf_rep_eth_dev = bp->rep_info[i].vfr_eth_dev;
6194                 if (!vf_rep_eth_dev)
6195                         continue;
6196                 PMD_DRV_LOG(DEBUG, "BNXT Port:%d VFR pci remove\n",
6197                             vf_rep_eth_dev->data->port_id);
6198                 rte_eth_dev_destroy(vf_rep_eth_dev, bnxt_representor_uninit);
6199         }
6200         PMD_DRV_LOG(DEBUG, "BNXT Port:%d pci remove\n",
6201                     eth_dev->data->port_id);
6202         ret = rte_eth_dev_destroy(eth_dev, bnxt_dev_uninit);
6203
6204         return ret;
6205 }
6206
6207 static void bnxt_free_rep_info(struct bnxt *bp)
6208 {
6209         rte_free(bp->rep_info);
6210         bp->rep_info = NULL;
6211         rte_free(bp->cfa_code_map);
6212         bp->cfa_code_map = NULL;
6213 }
6214
6215 static int bnxt_init_rep_info(struct bnxt *bp)
6216 {
6217         int i = 0, rc;
6218
6219         if (bp->rep_info)
6220                 return 0;
6221
6222         bp->rep_info = rte_zmalloc("bnxt_rep_info",
6223                                    sizeof(bp->rep_info[0]) * BNXT_MAX_VF_REPS,
6224                                    0);
6225         if (!bp->rep_info) {
6226                 PMD_DRV_LOG(ERR, "Failed to alloc memory for rep info\n");
6227                 return -ENOMEM;
6228         }
6229         bp->cfa_code_map = rte_zmalloc("bnxt_cfa_code_map",
6230                                        sizeof(*bp->cfa_code_map) *
6231                                        BNXT_MAX_CFA_CODE, 0);
6232         if (!bp->cfa_code_map) {
6233                 PMD_DRV_LOG(ERR, "Failed to alloc memory for cfa_code_map\n");
6234                 bnxt_free_rep_info(bp);
6235                 return -ENOMEM;
6236         }
6237
6238         for (i = 0; i < BNXT_MAX_CFA_CODE; i++)
6239                 bp->cfa_code_map[i] = BNXT_VF_IDX_INVALID;
6240
6241         rc = pthread_mutex_init(&bp->rep_info->vfr_lock, NULL);
6242         if (rc) {
6243                 PMD_DRV_LOG(ERR, "Unable to initialize vfr_lock\n");
6244                 bnxt_free_rep_info(bp);
6245                 return rc;
6246         }
6247
6248         rc = pthread_mutex_init(&bp->rep_info->vfr_start_lock, NULL);
6249         if (rc) {
6250                 PMD_DRV_LOG(ERR, "Unable to initialize vfr_start_lock\n");
6251                 bnxt_free_rep_info(bp);
6252                 return rc;
6253         }
6254
6255         return rc;
6256 }
6257
6258 static int bnxt_rep_port_probe(struct rte_pci_device *pci_dev,
6259                                struct rte_eth_devargs *eth_da,
6260                                struct rte_eth_dev *backing_eth_dev,
6261                                const char *dev_args)
6262 {
6263         struct rte_eth_dev *vf_rep_eth_dev;
6264         char name[RTE_ETH_NAME_MAX_LEN];
6265         struct bnxt *backing_bp;
6266         uint16_t num_rep;
6267         int i, ret = 0;
6268         struct rte_kvargs *kvlist = NULL;
6269
6270         if (eth_da->type == RTE_ETH_REPRESENTOR_NONE)
6271                 return 0;
6272         if (eth_da->type != RTE_ETH_REPRESENTOR_VF) {
6273                 PMD_DRV_LOG(ERR, "unsupported representor type %d\n",
6274                             eth_da->type);
6275                 return -ENOTSUP;
6276         }
6277         num_rep = eth_da->nb_representor_ports;
6278         if (num_rep > BNXT_MAX_VF_REPS) {
6279                 PMD_DRV_LOG(ERR, "nb_representor_ports = %d > %d MAX VF REPS\n",
6280                             num_rep, BNXT_MAX_VF_REPS);
6281                 return -EINVAL;
6282         }
6283
6284         if (num_rep >= RTE_MAX_ETHPORTS) {
6285                 PMD_DRV_LOG(ERR,
6286                             "nb_representor_ports = %d > %d MAX ETHPORTS\n",
6287                             num_rep, RTE_MAX_ETHPORTS);
6288                 return -EINVAL;
6289         }
6290
6291         backing_bp = backing_eth_dev->data->dev_private;
6292
6293         if (!(BNXT_PF(backing_bp) || BNXT_VF_IS_TRUSTED(backing_bp))) {
6294                 PMD_DRV_LOG(ERR,
6295                             "Not a PF or trusted VF. No Representor support\n");
6296                 /* Returning an error is not an option.
6297                  * Applications are not handling this correctly
6298                  */
6299                 return 0;
6300         }
6301
6302         if (bnxt_init_rep_info(backing_bp))
6303                 return 0;
6304
6305         for (i = 0; i < num_rep; i++) {
6306                 struct bnxt_representor representor = {
6307                         .vf_id = eth_da->representor_ports[i],
6308                         .switch_domain_id = backing_bp->switch_domain_id,
6309                         .parent_dev = backing_eth_dev
6310                 };
6311
6312                 if (representor.vf_id >= BNXT_MAX_VF_REPS) {
6313                         PMD_DRV_LOG(ERR, "VF-Rep id %d >= %d MAX VF ID\n",
6314                                     representor.vf_id, BNXT_MAX_VF_REPS);
6315                         continue;
6316                 }
6317
6318                 /* representor port net_bdf_port */
6319                 snprintf(name, sizeof(name), "net_%s_representor_%d",
6320                          pci_dev->device.name, eth_da->representor_ports[i]);
6321
6322                 kvlist = rte_kvargs_parse(dev_args, bnxt_dev_args);
6323                 if (kvlist) {
6324                         /*
6325                          * Handler for "rep_is_pf" devarg.
6326                          * Invoked as for ex: "-a 000:00:0d.0,
6327                          * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6328                          */
6329                         ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_IS_PF,
6330                                                  bnxt_parse_devarg_rep_is_pf,
6331                                                  (void *)&representor);
6332                         if (ret) {
6333                                 ret = -EINVAL;
6334                                 goto err;
6335                         }
6336                         /*
6337                          * Handler for "rep_based_pf" devarg.
6338                          * Invoked as for ex: "-a 000:00:0d.0,
6339                          * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6340                          */
6341                         ret = rte_kvargs_process(kvlist,
6342                                                  BNXT_DEVARG_REP_BASED_PF,
6343                                                  bnxt_parse_devarg_rep_based_pf,
6344                                                  (void *)&representor);
6345                         if (ret) {
6346                                 ret = -EINVAL;
6347                                 goto err;
6348                         }
6349                         /*
6350                          * Handler for "rep_based_pf" devarg.
6351                          * Invoked as for ex: "-a 000:00:0d.0,
6352                          * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6353                          */
6354                         ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_Q_R2F,
6355                                                  bnxt_parse_devarg_rep_q_r2f,
6356                                                  (void *)&representor);
6357                         if (ret) {
6358                                 ret = -EINVAL;
6359                                 goto err;
6360                         }
6361                         /*
6362                          * Handler for "rep_based_pf" devarg.
6363                          * Invoked as for ex: "-a 000:00:0d.0,
6364                          * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6365                          */
6366                         ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_Q_F2R,
6367                                                  bnxt_parse_devarg_rep_q_f2r,
6368                                                  (void *)&representor);
6369                         if (ret) {
6370                                 ret = -EINVAL;
6371                                 goto err;
6372                         }
6373                         /*
6374                          * Handler for "rep_based_pf" devarg.
6375                          * Invoked as for ex: "-a 000:00:0d.0,
6376                          * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6377                          */
6378                         ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_FC_R2F,
6379                                                  bnxt_parse_devarg_rep_fc_r2f,
6380                                                  (void *)&representor);
6381                         if (ret) {
6382                                 ret = -EINVAL;
6383                                 goto err;
6384                         }
6385                         /*
6386                          * Handler for "rep_based_pf" devarg.
6387                          * Invoked as for ex: "-a 000:00:0d.0,
6388                          * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6389                          */
6390                         ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_FC_F2R,
6391                                                  bnxt_parse_devarg_rep_fc_f2r,
6392                                                  (void *)&representor);
6393                         if (ret) {
6394                                 ret = -EINVAL;
6395                                 goto err;
6396                         }
6397                 }
6398
6399                 ret = rte_eth_dev_create(&pci_dev->device, name,
6400                                          sizeof(struct bnxt_representor),
6401                                          NULL, NULL,
6402                                          bnxt_representor_init,
6403                                          &representor);
6404                 if (ret) {
6405                         PMD_DRV_LOG(ERR, "failed to create bnxt vf "
6406                                     "representor %s.", name);
6407                         goto err;
6408                 }
6409
6410                 vf_rep_eth_dev = rte_eth_dev_allocated(name);
6411                 if (!vf_rep_eth_dev) {
6412                         PMD_DRV_LOG(ERR, "Failed to find the eth_dev"
6413                                     " for VF-Rep: %s.", name);
6414                         ret = -ENODEV;
6415                         goto err;
6416                 }
6417
6418                 PMD_DRV_LOG(DEBUG, "BNXT Port:%d VFR pci probe\n",
6419                             backing_eth_dev->data->port_id);
6420                 backing_bp->rep_info[representor.vf_id].vfr_eth_dev =
6421                                                          vf_rep_eth_dev;
6422                 backing_bp->num_reps++;
6423
6424         }
6425
6426         rte_kvargs_free(kvlist);
6427         return 0;
6428
6429 err:
6430         /* If num_rep > 1, then rollback already created
6431          * ports, since we'll be failing the probe anyway
6432          */
6433         if (num_rep > 1)
6434                 bnxt_pci_remove_dev_with_reps(backing_eth_dev);
6435         rte_errno = -ret;
6436         rte_kvargs_free(kvlist);
6437
6438         return ret;
6439 }
6440
6441 static int bnxt_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
6442                           struct rte_pci_device *pci_dev)
6443 {
6444         struct rte_eth_devargs eth_da = { .nb_representor_ports = 0 };
6445         struct rte_eth_dev *backing_eth_dev;
6446         uint16_t num_rep;
6447         int ret = 0;
6448
6449         if (pci_dev->device.devargs) {
6450                 ret = rte_eth_devargs_parse(pci_dev->device.devargs->args,
6451                                             &eth_da);
6452                 if (ret)
6453                         return ret;
6454         }
6455
6456         num_rep = eth_da.nb_representor_ports;
6457         PMD_DRV_LOG(DEBUG, "nb_representor_ports = %d\n",
6458                     num_rep);
6459
6460         /* We could come here after first level of probe is already invoked
6461          * as part of an application bringup(OVS-DPDK vswitchd), so first check
6462          * for already allocated eth_dev for the backing device (PF/Trusted VF)
6463          */
6464         backing_eth_dev = rte_eth_dev_allocated(pci_dev->device.name);
6465         if (backing_eth_dev == NULL) {
6466                 ret = rte_eth_dev_create(&pci_dev->device, pci_dev->device.name,
6467                                          sizeof(struct bnxt),
6468                                          eth_dev_pci_specific_init, pci_dev,
6469                                          bnxt_dev_init, NULL);
6470
6471                 if (ret || !num_rep)
6472                         return ret;
6473
6474                 backing_eth_dev = rte_eth_dev_allocated(pci_dev->device.name);
6475         }
6476         PMD_DRV_LOG(DEBUG, "BNXT Port:%d pci probe\n",
6477                     backing_eth_dev->data->port_id);
6478
6479         if (!num_rep)
6480                 return ret;
6481
6482         /* probe representor ports now */
6483         ret = bnxt_rep_port_probe(pci_dev, &eth_da, backing_eth_dev,
6484                                   pci_dev->device.devargs->args);
6485
6486         return ret;
6487 }
6488
6489 static int bnxt_pci_remove(struct rte_pci_device *pci_dev)
6490 {
6491         struct rte_eth_dev *eth_dev;
6492
6493         eth_dev = rte_eth_dev_allocated(pci_dev->device.name);
6494         if (!eth_dev)
6495                 return 0; /* Invoked typically only by OVS-DPDK, by the
6496                            * time it comes here the eth_dev is already
6497                            * deleted by rte_eth_dev_close(), so returning
6498                            * +ve value will at least help in proper cleanup
6499                            */
6500
6501         PMD_DRV_LOG(DEBUG, "BNXT Port:%d pci remove\n", eth_dev->data->port_id);
6502         if (rte_eal_process_type() == RTE_PROC_PRIMARY) {
6503                 if (eth_dev->data->dev_flags & RTE_ETH_DEV_REPRESENTOR)
6504                         return rte_eth_dev_destroy(eth_dev,
6505                                                    bnxt_representor_uninit);
6506                 else
6507                         return rte_eth_dev_destroy(eth_dev,
6508                                                    bnxt_dev_uninit);
6509         } else {
6510                 return rte_eth_dev_pci_generic_remove(pci_dev, NULL);
6511         }
6512 }
6513
6514 static struct rte_pci_driver bnxt_rte_pmd = {
6515         .id_table = bnxt_pci_id_map,
6516         .drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC |
6517                         RTE_PCI_DRV_INTR_RMV |
6518                         RTE_PCI_DRV_PROBE_AGAIN, /* Needed in case of VF-REPs
6519                                                   * and OVS-DPDK
6520                                                   */
6521         .probe = bnxt_pci_probe,
6522         .remove = bnxt_pci_remove,
6523 };
6524
6525 static bool
6526 is_device_supported(struct rte_eth_dev *dev, struct rte_pci_driver *drv)
6527 {
6528         if (strcmp(dev->device->driver->name, drv->driver.name))
6529                 return false;
6530
6531         return true;
6532 }
6533
6534 bool is_bnxt_supported(struct rte_eth_dev *dev)
6535 {
6536         return is_device_supported(dev, &bnxt_rte_pmd);
6537 }
6538
6539 RTE_LOG_REGISTER_SUFFIX(bnxt_logtype_driver, driver, NOTICE);
6540 RTE_PMD_REGISTER_PCI(net_bnxt, bnxt_rte_pmd);
6541 RTE_PMD_REGISTER_PCI_TABLE(net_bnxt, bnxt_pci_id_map);
6542 RTE_PMD_REGISTER_KMOD_DEP(net_bnxt, "* igb_uio | uio_pci_generic | vfio-pci");