drivers/net/bnxt/bnxt_ethdev.c

   1 /* SPDX-License-Identifier: BSD-3-Clause
   2  * Copyright(c) 2014-2021 Broadcom
   3  * All rights reserved.
   4  */
   5
   6 #include <inttypes.h>
   7 #include <stdbool.h>
   8
   9 #include <rte_dev.h>
  10 #include <ethdev_driver.h>
  11 #include <ethdev_pci.h>
  12 #include <rte_malloc.h>
  13 #include <rte_cycles.h>
  14 #include <rte_alarm.h>
  15 #include <rte_kvargs.h>
  16 #include <rte_vect.h>
  17
  18 #include "bnxt.h"
  19 #include "bnxt_filter.h"
  20 #include "bnxt_hwrm.h"
  21 #include "bnxt_irq.h"
  22 #include "bnxt_reps.h"
  23 #include "bnxt_ring.h"
  24 #include "bnxt_rxq.h"
  25 #include "bnxt_rxr.h"
  26 #include "bnxt_stats.h"
  27 #include "bnxt_txq.h"
  28 #include "bnxt_txr.h"
  29 #include "bnxt_vnic.h"
  30 #include "hsi_struct_def_dpdk.h"
  31 #include "bnxt_nvm_defs.h"
  32 #include "bnxt_tf_common.h"
  33 #include "ulp_flow_db.h"
  34 #include "rte_pmd_bnxt.h"
  35
  36 #define DRV_MODULE_NAME         "bnxt"
  37 static const char bnxt_version[] =
  38         "Broadcom NetXtreme driver " DRV_MODULE_NAME;
  39
  40 /*
  41  * The set of PCI devices this driver supports
  42  */
  43 static const struct rte_pci_id bnxt_pci_id_map[] = {
  44         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM,
  45                          BROADCOM_DEV_ID_STRATUS_NIC_VF1) },
  46         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM,
  47                          BROADCOM_DEV_ID_STRATUS_NIC_VF2) },
  48         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_STRATUS_NIC) },
  49         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57414_VF) },
  50         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57304_VF) },
  51         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_NS2) },
  52         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57406_VF) },
  53         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57407_MF) },
  54         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_5741X_VF) },
  55         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_5731X_VF) },
  56         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57417_MF) },
  57         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57412) },
  58         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57414) },
  59         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57416_RJ45) },
  60         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57417_RJ45) },
  61         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57412_MF) },
  62         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57317_RJ45) },
  63         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57417_SFP) },
  64         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57416_SFP) },
  65         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57317_SFP) },
  66         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57414_MF) },
  67         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57416_MF) },
  68         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58802) },
  69         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58804) },
  70         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58808) },
  71         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58802_VF) },
  72         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57508) },
  73         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57504) },
  74         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57502) },
  75         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57500_VF1) },
  76         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57500_VF2) },
  77         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57508_MF1) },
  78         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57504_MF1) },
  79         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57502_MF1) },
  80         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57508_MF2) },
  81         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57504_MF2) },
  82         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57502_MF2) },
  83         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58812) },
  84         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58814) },
  85         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58818) },
  86         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58818_VF) },
  87         { .vendor_id = 0, /* sentinel */ },
  88 };
  89
  90 #define BNXT_DEVARG_TRUFLOW     "host-based-truflow"
  91 #define BNXT_DEVARG_FLOW_XSTAT  "flow-xstat"
  92 #define BNXT_DEVARG_MAX_NUM_KFLOWS  "max-num-kflows"
  93 #define BNXT_DEVARG_REPRESENTOR "representor"
  94 #define BNXT_DEVARG_REP_BASED_PF  "rep-based-pf"
  95 #define BNXT_DEVARG_REP_IS_PF  "rep-is-pf"
  96 #define BNXT_DEVARG_REP_Q_R2F  "rep-q-r2f"
  97 #define BNXT_DEVARG_REP_Q_F2R  "rep-q-f2r"
  98 #define BNXT_DEVARG_REP_FC_R2F  "rep-fc-r2f"
  99 #define BNXT_DEVARG_REP_FC_F2R  "rep-fc-f2r"
 100
 101 static const char *const bnxt_dev_args[] = {
 102         BNXT_DEVARG_REPRESENTOR,
 103         BNXT_DEVARG_TRUFLOW,
 104         BNXT_DEVARG_FLOW_XSTAT,
 105         BNXT_DEVARG_MAX_NUM_KFLOWS,
 106         BNXT_DEVARG_REP_BASED_PF,
 107         BNXT_DEVARG_REP_IS_PF,
 108         BNXT_DEVARG_REP_Q_R2F,
 109         BNXT_DEVARG_REP_Q_F2R,
 110         BNXT_DEVARG_REP_FC_R2F,
 111         BNXT_DEVARG_REP_FC_F2R,
 112         NULL
 113 };
 114
 115 /*
 116  * truflow == false to disable the feature
 117  * truflow == true to enable the feature
 118  */
 119 #define BNXT_DEVARG_TRUFLOW_INVALID(truflow)    ((truflow) > 1)
 120
 121 /*
 122  * flow_xstat == false to disable the feature
 123  * flow_xstat == true to enable the feature
 124  */
 125 #define BNXT_DEVARG_FLOW_XSTAT_INVALID(flow_xstat)      ((flow_xstat) > 1)
 126
 127 /*
 128  * rep_is_pf == false to indicate VF representor
 129  * rep_is_pf == true to indicate PF representor
 130  */
 131 #define BNXT_DEVARG_REP_IS_PF_INVALID(rep_is_pf)        ((rep_is_pf) > 1)
 132
 133 /*
 134  * rep_based_pf == Physical index of the PF
 135  */
 136 #define BNXT_DEVARG_REP_BASED_PF_INVALID(rep_based_pf)  ((rep_based_pf) > 15)
 137 /*
 138  * rep_q_r2f == Logical COS Queue index for the rep to endpoint direction
 139  */
 140 #define BNXT_DEVARG_REP_Q_R2F_INVALID(rep_q_r2f)        ((rep_q_r2f) > 3)
 141
 142 /*
 143  * rep_q_f2r == Logical COS Queue index for the endpoint to rep direction
 144  */
 145 #define BNXT_DEVARG_REP_Q_F2R_INVALID(rep_q_f2r)        ((rep_q_f2r) > 3)
 146
 147 /*
 148  * rep_fc_r2f == Flow control for the representor to endpoint direction
 149  */
 150 #define BNXT_DEVARG_REP_FC_R2F_INVALID(rep_fc_r2f)      ((rep_fc_r2f) > 1)
 151
 152 /*
 153  * rep_fc_f2r == Flow control for the endpoint to representor direction
 154  */
 155 #define BNXT_DEVARG_REP_FC_F2R_INVALID(rep_fc_f2r)      ((rep_fc_f2r) > 1)
 156
 157 int bnxt_cfa_code_dynfield_offset = -1;
 158
 159 /*
 160  * max_num_kflows must be >= 32
 161  * and must be a power-of-2 supported value
 162  * return: 1 -> invalid
 163  *         0 -> valid
 164  */
 165 static int bnxt_devarg_max_num_kflow_invalid(uint16_t max_num_kflows)
 166 {
 167         if (max_num_kflows < 32 || !rte_is_power_of_2(max_num_kflows))
 168                 return 1;
 169         return 0;
 170 }
 171
 172 static int bnxt_vlan_offload_set_op(struct rte_eth_dev *dev, int mask);
 173 static int bnxt_dev_uninit(struct rte_eth_dev *eth_dev);
 174 static int bnxt_init_resources(struct bnxt *bp, bool reconfig_dev);
 175 static int bnxt_uninit_resources(struct bnxt *bp, bool reconfig_dev);
 176 static void bnxt_cancel_fw_health_check(struct bnxt *bp);
 177 static int bnxt_restore_vlan_filters(struct bnxt *bp);
 178 static void bnxt_dev_recover(void *arg);
 179 static void bnxt_free_error_recovery_info(struct bnxt *bp);
 180 static void bnxt_free_rep_info(struct bnxt *bp);
 181
 182 int is_bnxt_in_error(struct bnxt *bp)
 183 {
 184         if (bp->flags & BNXT_FLAG_FATAL_ERROR)
 185                 return -EIO;
 186         if (bp->flags & BNXT_FLAG_FW_RESET)
 187                 return -EBUSY;
 188
 189         return 0;
 190 }
 191
 192 /***********************/
 193
 194 /*
 195  * High level utility functions
 196  */
 197
 198 static uint16_t bnxt_rss_ctxts(const struct bnxt *bp)
 199 {
 200         unsigned int num_rss_rings = RTE_MIN(bp->rx_nr_rings,
 201                                              BNXT_RSS_TBL_SIZE_P5);
 202
 203         if (!BNXT_CHIP_P5(bp))
 204                 return 1;
 205
 206         return RTE_ALIGN_MUL_CEIL(num_rss_rings,
 207                                   BNXT_RSS_ENTRIES_PER_CTX_P5) /
 208                                   BNXT_RSS_ENTRIES_PER_CTX_P5;
 209 }
 210
 211 uint16_t bnxt_rss_hash_tbl_size(const struct bnxt *bp)
 212 {
 213         if (!BNXT_CHIP_P5(bp))
 214                 return HW_HASH_INDEX_SIZE;
 215
 216         return bnxt_rss_ctxts(bp) * BNXT_RSS_ENTRIES_PER_CTX_P5;
 217 }
 218
 219 static void bnxt_free_parent_info(struct bnxt *bp)
 220 {
 221         rte_free(bp->parent);
 222         bp->parent = NULL;
 223 }
 224
 225 static void bnxt_free_pf_info(struct bnxt *bp)
 226 {
 227         rte_free(bp->pf);
 228         bp->pf = NULL;
 229 }
 230
 231 static void bnxt_free_link_info(struct bnxt *bp)
 232 {
 233         rte_free(bp->link_info);
 234         bp->link_info = NULL;
 235 }
 236
 237 static void bnxt_free_leds_info(struct bnxt *bp)
 238 {
 239         if (BNXT_VF(bp))
 240                 return;
 241
 242         rte_free(bp->leds);
 243         bp->leds = NULL;
 244 }
 245
 246 static void bnxt_free_flow_stats_info(struct bnxt *bp)
 247 {
 248         rte_free(bp->flow_stat);
 249         bp->flow_stat = NULL;
 250 }
 251
 252 static void bnxt_free_cos_queues(struct bnxt *bp)
 253 {
 254         rte_free(bp->rx_cos_queue);
 255         bp->rx_cos_queue = NULL;
 256         rte_free(bp->tx_cos_queue);
 257         bp->tx_cos_queue = NULL;
 258 }
 259
 260 static void bnxt_free_mem(struct bnxt *bp, bool reconfig)
 261 {
 262         bnxt_free_filter_mem(bp);
 263         bnxt_free_vnic_attributes(bp);
 264         bnxt_free_vnic_mem(bp);
 265
 266         /* tx/rx rings are configured as part of *_queue_setup callbacks.
 267          * If the number of rings change across fw update,
 268          * we don't have much choice except to warn the user.
 269          */
 270         if (!reconfig) {
 271                 bnxt_free_stats(bp);
 272                 bnxt_free_tx_rings(bp);
 273                 bnxt_free_rx_rings(bp);
 274         }
 275         bnxt_free_async_cp_ring(bp);
 276         bnxt_free_rxtx_nq_ring(bp);
 277
 278         rte_free(bp->grp_info);
 279         bp->grp_info = NULL;
 280 }
 281
 282 static int bnxt_alloc_parent_info(struct bnxt *bp)
 283 {
 284         bp->parent = rte_zmalloc("bnxt_parent_info",
 285                                  sizeof(struct bnxt_parent_info), 0);
 286         if (bp->parent == NULL)
 287                 return -ENOMEM;
 288
 289         return 0;
 290 }
 291
 292 static int bnxt_alloc_pf_info(struct bnxt *bp)
 293 {
 294         bp->pf = rte_zmalloc("bnxt_pf_info", sizeof(struct bnxt_pf_info), 0);
 295         if (bp->pf == NULL)
 296                 return -ENOMEM;
 297
 298         return 0;
 299 }
 300
 301 static int bnxt_alloc_link_info(struct bnxt *bp)
 302 {
 303         bp->link_info =
 304                 rte_zmalloc("bnxt_link_info", sizeof(struct bnxt_link_info), 0);
 305         if (bp->link_info == NULL)
 306                 return -ENOMEM;
 307
 308         return 0;
 309 }
 310
 311 static int bnxt_alloc_leds_info(struct bnxt *bp)
 312 {
 313         if (BNXT_VF(bp))
 314                 return 0;
 315
 316         bp->leds = rte_zmalloc("bnxt_leds",
 317                                BNXT_MAX_LED * sizeof(struct bnxt_led_info),
 318                                0);
 319         if (bp->leds == NULL)
 320                 return -ENOMEM;
 321
 322         return 0;
 323 }
 324
 325 static int bnxt_alloc_cos_queues(struct bnxt *bp)
 326 {
 327         bp->rx_cos_queue =
 328                 rte_zmalloc("bnxt_rx_cosq",
 329                             BNXT_COS_QUEUE_COUNT *
 330                             sizeof(struct bnxt_cos_queue_info),
 331                             0);
 332         if (bp->rx_cos_queue == NULL)
 333                 return -ENOMEM;
 334
 335         bp->tx_cos_queue =
 336                 rte_zmalloc("bnxt_tx_cosq",
 337                             BNXT_COS_QUEUE_COUNT *
 338                             sizeof(struct bnxt_cos_queue_info),
 339                             0);
 340         if (bp->tx_cos_queue == NULL)
 341                 return -ENOMEM;
 342
 343         return 0;
 344 }
 345
 346 static int bnxt_alloc_flow_stats_info(struct bnxt *bp)
 347 {
 348         bp->flow_stat = rte_zmalloc("bnxt_flow_xstat",
 349                                     sizeof(struct bnxt_flow_stat_info), 0);
 350         if (bp->flow_stat == NULL)
 351                 return -ENOMEM;
 352
 353         return 0;
 354 }
 355
 356 static int bnxt_alloc_mem(struct bnxt *bp, bool reconfig)
 357 {
 358         int rc;
 359
 360         rc = bnxt_alloc_ring_grps(bp);
 361         if (rc)
 362                 goto alloc_mem_err;
 363
 364         rc = bnxt_alloc_async_ring_struct(bp);
 365         if (rc)
 366                 goto alloc_mem_err;
 367
 368         rc = bnxt_alloc_vnic_mem(bp);
 369         if (rc)
 370                 goto alloc_mem_err;
 371
 372         rc = bnxt_alloc_vnic_attributes(bp);
 373         if (rc)
 374                 goto alloc_mem_err;
 375
 376         rc = bnxt_alloc_filter_mem(bp);
 377         if (rc)
 378                 goto alloc_mem_err;
 379
 380         rc = bnxt_alloc_async_cp_ring(bp);
 381         if (rc)
 382                 goto alloc_mem_err;
 383
 384         rc = bnxt_alloc_rxtx_nq_ring(bp);
 385         if (rc)
 386                 goto alloc_mem_err;
 387
 388         if (BNXT_FLOW_XSTATS_EN(bp)) {
 389                 rc = bnxt_alloc_flow_stats_info(bp);
 390                 if (rc)
 391                         goto alloc_mem_err;
 392         }
 393
 394         return 0;
 395
 396 alloc_mem_err:
 397         bnxt_free_mem(bp, reconfig);
 398         return rc;
 399 }
 400
 401 static int bnxt_setup_one_vnic(struct bnxt *bp, uint16_t vnic_id)
 402 {
 403         struct rte_eth_conf *dev_conf = &bp->eth_dev->data->dev_conf;
 404         struct bnxt_vnic_info *vnic = &bp->vnic_info[vnic_id];
 405         uint64_t rx_offloads = dev_conf->rxmode.offloads;
 406         struct bnxt_rx_queue *rxq;
 407         unsigned int j;
 408         int rc;
 409
 410         rc = bnxt_vnic_grp_alloc(bp, vnic);
 411         if (rc)
 412                 goto err_out;
 413
 414         PMD_DRV_LOG(DEBUG, "vnic[%d] = %p vnic->fw_grp_ids = %p\n",
 415                     vnic_id, vnic, vnic->fw_grp_ids);
 416
 417         rc = bnxt_hwrm_vnic_alloc(bp, vnic);
 418         if (rc)
 419                 goto err_out;
 420
 421         /* Alloc RSS context only if RSS mode is enabled */
 422         if (dev_conf->rxmode.mq_mode & ETH_MQ_RX_RSS) {
 423                 int j, nr_ctxs = bnxt_rss_ctxts(bp);
 424
 425                 /* RSS table size in Thor is 512.
 426                  * Cap max Rx rings to same value
 427                  */
 428                 if (bp->rx_nr_rings > BNXT_RSS_TBL_SIZE_P5) {
 429                         PMD_DRV_LOG(ERR, "RxQ cnt %d > reta_size %d\n",
 430                                     bp->rx_nr_rings, BNXT_RSS_TBL_SIZE_P5);
 431                         goto err_out;
 432                 }
 433
 434                 rc = 0;
 435                 for (j = 0; j < nr_ctxs; j++) {
 436                         rc = bnxt_hwrm_vnic_ctx_alloc(bp, vnic, j);
 437                         if (rc)
 438                                 break;
 439                 }
 440                 if (rc) {
 441                         PMD_DRV_LOG(ERR,
 442                                     "HWRM vnic %d ctx %d alloc failure rc: %x\n",
 443                                     vnic_id, j, rc);
 444                         goto err_out;
 445                 }
 446                 vnic->num_lb_ctxts = nr_ctxs;
 447         }
 448
 449         /*
 450          * Firmware sets pf pair in default vnic cfg. If the VLAN strip
 451          * setting is not available at this time, it will not be
 452          * configured correctly in the CFA.
 453          */
 454         if (rx_offloads & DEV_RX_OFFLOAD_VLAN_STRIP)
 455                 vnic->vlan_strip = true;
 456         else
 457                 vnic->vlan_strip = false;
 458
 459         rc = bnxt_hwrm_vnic_cfg(bp, vnic);
 460         if (rc)
 461                 goto err_out;
 462
 463         rc = bnxt_set_hwrm_vnic_filters(bp, vnic);
 464         if (rc)
 465                 goto err_out;
 466
 467         for (j = 0; j < bp->rx_num_qs_per_vnic; j++) {
 468                 rxq = bp->eth_dev->data->rx_queues[j];
 469
 470                 PMD_DRV_LOG(DEBUG,
 471                             "rxq[%d]->vnic=%p vnic->fw_grp_ids=%p\n",
 472                             j, rxq->vnic, rxq->vnic->fw_grp_ids);
 473
 474                 if (BNXT_HAS_RING_GRPS(bp) && rxq->rx_deferred_start)
 475                         rxq->vnic->fw_grp_ids[j] = INVALID_HW_RING_ID;
 476                 else
 477                         vnic->rx_queue_cnt++;
 478         }
 479
 480         PMD_DRV_LOG(DEBUG, "vnic->rx_queue_cnt = %d\n", vnic->rx_queue_cnt);
 481
 482         rc = bnxt_vnic_rss_configure(bp, vnic);
 483         if (rc)
 484                 goto err_out;
 485
 486         bnxt_hwrm_vnic_plcmode_cfg(bp, vnic);
 487
 488         rc = bnxt_hwrm_vnic_tpa_cfg(bp, vnic,
 489                                     (rx_offloads & DEV_RX_OFFLOAD_TCP_LRO) ?
 490                                     true : false);
 491         if (rc)
 492                 goto err_out;
 493
 494         return 0;
 495 err_out:
 496         PMD_DRV_LOG(ERR, "HWRM vnic %d cfg failure rc: %x\n",
 497                     vnic_id, rc);
 498         return rc;
 499 }
 500
 501 static int bnxt_register_fc_ctx_mem(struct bnxt *bp)
 502 {
 503         int rc = 0;
 504
 505         rc = bnxt_hwrm_ctx_rgtr(bp, bp->flow_stat->rx_fc_in_tbl.dma,
 506                                 &bp->flow_stat->rx_fc_in_tbl.ctx_id);
 507         if (rc)
 508                 return rc;
 509
 510         PMD_DRV_LOG(DEBUG,
 511                     "rx_fc_in_tbl.va = %p rx_fc_in_tbl.dma = %p"
 512                     " rx_fc_in_tbl.ctx_id = %d\n",
 513                     bp->flow_stat->rx_fc_in_tbl.va,
 514                     (void *)((uintptr_t)bp->flow_stat->rx_fc_in_tbl.dma),
 515                     bp->flow_stat->rx_fc_in_tbl.ctx_id);
 516
 517         rc = bnxt_hwrm_ctx_rgtr(bp, bp->flow_stat->rx_fc_out_tbl.dma,
 518                                 &bp->flow_stat->rx_fc_out_tbl.ctx_id);
 519         if (rc)
 520                 return rc;
 521
 522         PMD_DRV_LOG(DEBUG,
 523                     "rx_fc_out_tbl.va = %p rx_fc_out_tbl.dma = %p"
 524                     " rx_fc_out_tbl.ctx_id = %d\n",
 525                     bp->flow_stat->rx_fc_out_tbl.va,
 526                     (void *)((uintptr_t)bp->flow_stat->rx_fc_out_tbl.dma),
 527                     bp->flow_stat->rx_fc_out_tbl.ctx_id);
 528
 529         rc = bnxt_hwrm_ctx_rgtr(bp, bp->flow_stat->tx_fc_in_tbl.dma,
 530                                 &bp->flow_stat->tx_fc_in_tbl.ctx_id);
 531         if (rc)
 532                 return rc;
 533
 534         PMD_DRV_LOG(DEBUG,
 535                     "tx_fc_in_tbl.va = %p tx_fc_in_tbl.dma = %p"
 536                     " tx_fc_in_tbl.ctx_id = %d\n",
 537                     bp->flow_stat->tx_fc_in_tbl.va,
 538                     (void *)((uintptr_t)bp->flow_stat->tx_fc_in_tbl.dma),
 539                     bp->flow_stat->tx_fc_in_tbl.ctx_id);
 540
 541         rc = bnxt_hwrm_ctx_rgtr(bp, bp->flow_stat->tx_fc_out_tbl.dma,
 542                                 &bp->flow_stat->tx_fc_out_tbl.ctx_id);
 543         if (rc)
 544                 return rc;
 545
 546         PMD_DRV_LOG(DEBUG,
 547                     "tx_fc_out_tbl.va = %p tx_fc_out_tbl.dma = %p"
 548                     " tx_fc_out_tbl.ctx_id = %d\n",
 549                     bp->flow_stat->tx_fc_out_tbl.va,
 550                     (void *)((uintptr_t)bp->flow_stat->tx_fc_out_tbl.dma),
 551                     bp->flow_stat->tx_fc_out_tbl.ctx_id);
 552
 553         memset(bp->flow_stat->rx_fc_out_tbl.va,
 554                0,
 555                bp->flow_stat->rx_fc_out_tbl.size);
 556         rc = bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_RX,
 557                                        CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
 558                                        bp->flow_stat->rx_fc_out_tbl.ctx_id,
 559                                        bp->flow_stat->max_fc,
 560                                        true);
 561         if (rc)
 562                 return rc;
 563
 564         memset(bp->flow_stat->tx_fc_out_tbl.va,
 565                0,
 566                bp->flow_stat->tx_fc_out_tbl.size);
 567         rc = bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_TX,
 568                                        CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
 569                                        bp->flow_stat->tx_fc_out_tbl.ctx_id,
 570                                        bp->flow_stat->max_fc,
 571                                        true);
 572
 573         return rc;
 574 }
 575
 576 static int bnxt_alloc_ctx_mem_buf(char *type, size_t size,
 577                                   struct bnxt_ctx_mem_buf_info *ctx)
 578 {
 579         if (!ctx)
 580                 return -EINVAL;
 581
 582         ctx->va = rte_zmalloc(type, size, 0);
 583         if (ctx->va == NULL)
 584                 return -ENOMEM;
 585         rte_mem_lock_page(ctx->va);
 586         ctx->size = size;
 587         ctx->dma = rte_mem_virt2iova(ctx->va);
 588         if (ctx->dma == RTE_BAD_IOVA)
 589                 return -ENOMEM;
 590
 591         return 0;
 592 }
 593
 594 static int bnxt_init_fc_ctx_mem(struct bnxt *bp)
 595 {
 596         struct rte_pci_device *pdev = bp->pdev;
 597         char type[RTE_MEMZONE_NAMESIZE];
 598         uint16_t max_fc;
 599         int rc = 0;
 600
 601         max_fc = bp->flow_stat->max_fc;
 602
 603         sprintf(type, "bnxt_rx_fc_in_" PCI_PRI_FMT, pdev->addr.domain,
 604                 pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
 605         /* 4 bytes for each counter-id */
 606         rc = bnxt_alloc_ctx_mem_buf(type,
 607                                     max_fc * 4,
 608                                     &bp->flow_stat->rx_fc_in_tbl);
 609         if (rc)
 610                 return rc;
 611
 612         sprintf(type, "bnxt_rx_fc_out_" PCI_PRI_FMT, pdev->addr.domain,
 613                 pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
 614         /* 16 bytes for each counter - 8 bytes pkt_count, 8 bytes byte_count */
 615         rc = bnxt_alloc_ctx_mem_buf(type,
 616                                     max_fc * 16,
 617                                     &bp->flow_stat->rx_fc_out_tbl);
 618         if (rc)
 619                 return rc;
 620
 621         sprintf(type, "bnxt_tx_fc_in_" PCI_PRI_FMT, pdev->addr.domain,
 622                 pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
 623         /* 4 bytes for each counter-id */
 624         rc = bnxt_alloc_ctx_mem_buf(type,
 625                                     max_fc * 4,
 626                                     &bp->flow_stat->tx_fc_in_tbl);
 627         if (rc)
 628                 return rc;
 629
 630         sprintf(type, "bnxt_tx_fc_out_" PCI_PRI_FMT, pdev->addr.domain,
 631                 pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
 632         /* 16 bytes for each counter - 8 bytes pkt_count, 8 bytes byte_count */
 633         rc = bnxt_alloc_ctx_mem_buf(type,
 634                                     max_fc * 16,
 635                                     &bp->flow_stat->tx_fc_out_tbl);
 636         if (rc)
 637                 return rc;
 638
 639         rc = bnxt_register_fc_ctx_mem(bp);
 640
 641         return rc;
 642 }
 643
 644 static int bnxt_init_ctx_mem(struct bnxt *bp)
 645 {
 646         int rc = 0;
 647
 648         if (!(bp->fw_cap & BNXT_FW_CAP_ADV_FLOW_COUNTERS) ||
 649             !(BNXT_PF(bp) || BNXT_VF_IS_TRUSTED(bp)) ||
 650             !BNXT_FLOW_XSTATS_EN(bp))
 651                 return 0;
 652
 653         rc = bnxt_hwrm_cfa_counter_qcaps(bp, &bp->flow_stat->max_fc);
 654         if (rc)
 655                 return rc;
 656
 657         rc = bnxt_init_fc_ctx_mem(bp);
 658
 659         return rc;
 660 }
 661
 662 static int bnxt_update_phy_setting(struct bnxt *bp)
 663 {
 664         struct rte_eth_link new;
 665         int rc;
 666
 667         rc = bnxt_get_hwrm_link_config(bp, &new);
 668         if (rc) {
 669                 PMD_DRV_LOG(ERR, "Failed to get link settings\n");
 670                 return rc;
 671         }
 672
 673         /*
 674          * On BCM957508-N2100 adapters, FW will not allow any user other
 675          * than BMC to shutdown the port. bnxt_get_hwrm_link_config() call
 676          * always returns link up. Force phy update always in that case.
 677          */
 678         if (!new.link_status || IS_BNXT_DEV_957508_N2100(bp)) {
 679                 rc = bnxt_set_hwrm_link_config(bp, true);
 680                 if (rc) {
 681                         PMD_DRV_LOG(ERR, "Failed to update PHY settings\n");
 682                         return rc;
 683                 }
 684         }
 685
 686         return rc;
 687 }
 688
 689 static int bnxt_start_nic(struct bnxt *bp)
 690 {
 691         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(bp->eth_dev);
 692         struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
 693         uint32_t intr_vector = 0;
 694         uint32_t queue_id, base = BNXT_MISC_VEC_ID;
 695         uint32_t vec = BNXT_MISC_VEC_ID;
 696         unsigned int i, j;
 697         int rc;
 698
 699         if (bp->eth_dev->data->mtu > RTE_ETHER_MTU) {
 700                 bp->eth_dev->data->dev_conf.rxmode.offloads |=
 701                         DEV_RX_OFFLOAD_JUMBO_FRAME;
 702                 bp->flags |= BNXT_FLAG_JUMBO;
 703         } else {
 704                 bp->eth_dev->data->dev_conf.rxmode.offloads &=
 705                         ~DEV_RX_OFFLOAD_JUMBO_FRAME;
 706                 bp->flags &= ~BNXT_FLAG_JUMBO;
 707         }
 708
 709         /* THOR does not support ring groups.
 710          * But we will use the array to save RSS context IDs.
 711          */
 712         if (BNXT_CHIP_P5(bp))
 713                 bp->max_ring_grps = BNXT_MAX_RSS_CTXTS_P5;
 714
 715         rc = bnxt_alloc_all_hwrm_stat_ctxs(bp);
 716         if (rc) {
 717                 PMD_DRV_LOG(ERR, "HWRM stat ctx alloc failure rc: %x\n", rc);
 718                 goto err_out;
 719         }
 720
 721         rc = bnxt_alloc_hwrm_rings(bp);
 722         if (rc) {
 723                 PMD_DRV_LOG(ERR, "HWRM ring alloc failure rc: %x\n", rc);
 724                 goto err_out;
 725         }
 726
 727         rc = bnxt_alloc_all_hwrm_ring_grps(bp);
 728         if (rc) {
 729                 PMD_DRV_LOG(ERR, "HWRM ring grp alloc failure: %x\n", rc);
 730                 goto err_out;
 731         }
 732
 733         if (!(bp->vnic_cap_flags & BNXT_VNIC_CAP_COS_CLASSIFY))
 734                 goto skip_cosq_cfg;
 735
 736         for (j = 0, i = 0; i < BNXT_COS_QUEUE_COUNT; i++) {
 737                 if (bp->rx_cos_queue[i].id != 0xff) {
 738                         struct bnxt_vnic_info *vnic = &bp->vnic_info[j++];
 739
 740                         if (!vnic) {
 741                                 PMD_DRV_LOG(ERR,
 742                                             "Num pools more than FW profile\n");
 743                                 rc = -EINVAL;
 744                                 goto err_out;
 745                         }
 746                         vnic->cos_queue_id = bp->rx_cos_queue[i].id;
 747                         bp->rx_cosq_cnt++;
 748                 }
 749         }
 750
 751 skip_cosq_cfg:
 752         rc = bnxt_mq_rx_configure(bp);
 753         if (rc) {
 754                 PMD_DRV_LOG(ERR, "MQ mode configure failure rc: %x\n", rc);
 755                 goto err_out;
 756         }
 757
 758         /* default vnic 0 */
 759         rc = bnxt_setup_one_vnic(bp, 0);
 760         if (rc)
 761                 goto err_out;
 762         /* VNIC configuration */
 763         if (BNXT_RFS_NEEDS_VNIC(bp)) {
 764                 for (i = 1; i < bp->nr_vnics; i++) {
 765                         rc = bnxt_setup_one_vnic(bp, i);
 766                         if (rc)
 767                                 goto err_out;
 768                 }
 769         }
 770
 771         rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, &bp->vnic_info[0], 0, NULL);
 772         if (rc) {
 773                 PMD_DRV_LOG(ERR,
 774                         "HWRM cfa l2 rx mask failure rc: %x\n", rc);
 775                 goto err_out;
 776         }
 777
 778         /* check and configure queue intr-vector mapping */
 779         if ((rte_intr_cap_multiple(intr_handle) ||
 780              !RTE_ETH_DEV_SRIOV(bp->eth_dev).active) &&
 781             bp->eth_dev->data->dev_conf.intr_conf.rxq != 0) {
 782                 intr_vector = bp->eth_dev->data->nb_rx_queues;
 783                 PMD_DRV_LOG(DEBUG, "intr_vector = %d\n", intr_vector);
 784                 if (intr_vector > bp->rx_cp_nr_rings) {
 785                         PMD_DRV_LOG(ERR, "At most %d intr queues supported",
 786                                         bp->rx_cp_nr_rings);
 787                         return -ENOTSUP;
 788                 }
 789                 rc = rte_intr_efd_enable(intr_handle, intr_vector);
 790                 if (rc)
 791                         return rc;
 792         }
 793
 794         if (rte_intr_dp_is_en(intr_handle) && !intr_handle->intr_vec) {
 795                 intr_handle->intr_vec =
 796                         rte_zmalloc("intr_vec",
 797                                     bp->eth_dev->data->nb_rx_queues *
 798                                     sizeof(int), 0);
 799                 if (intr_handle->intr_vec == NULL) {
 800                         PMD_DRV_LOG(ERR, "Failed to allocate %d rx_queues"
 801                                 " intr_vec", bp->eth_dev->data->nb_rx_queues);
 802                         rc = -ENOMEM;
 803                         goto err_out;
 804                 }
 805                 PMD_DRV_LOG(DEBUG, "intr_handle->intr_vec = %p "
 806                         "intr_handle->nb_efd = %d intr_handle->max_intr = %d\n",
 807                          intr_handle->intr_vec, intr_handle->nb_efd,
 808                         intr_handle->max_intr);
 809                 for (queue_id = 0; queue_id < bp->eth_dev->data->nb_rx_queues;
 810                      queue_id++) {
 811                         intr_handle->intr_vec[queue_id] =
 812                                                         vec + BNXT_RX_VEC_START;
 813                         if (vec < base + intr_handle->nb_efd - 1)
 814                                 vec++;
 815                 }
 816         }
 817
 818         /* enable uio/vfio intr/eventfd mapping */
 819         rc = rte_intr_enable(intr_handle);
 820 #ifndef RTE_EXEC_ENV_FREEBSD
 821         /* In FreeBSD OS, nic_uio driver does not support interrupts */
 822         if (rc)
 823                 goto err_out;
 824 #endif
 825
 826         rc = bnxt_update_phy_setting(bp);
 827         if (rc)
 828                 goto err_out;
 829
 830         bp->mark_table = rte_zmalloc("bnxt_mark_table", BNXT_MARK_TABLE_SZ, 0);
 831         if (!bp->mark_table)
 832                 PMD_DRV_LOG(ERR, "Allocation of mark table failed\n");
 833
 834         return 0;
 835
 836 err_out:
 837         /* Some of the error status returned by FW may not be from errno.h */
 838         if (rc > 0)
 839                 rc = -EIO;
 840
 841         return rc;
 842 }
 843
 844 static int bnxt_shutdown_nic(struct bnxt *bp)
 845 {
 846         bnxt_free_all_hwrm_resources(bp);
 847         bnxt_free_all_filters(bp);
 848         bnxt_free_all_vnics(bp);
 849         return 0;
 850 }
 851
 852 /*
 853  * Device configuration and status function
 854  */
 855
 856 uint32_t bnxt_get_speed_capabilities(struct bnxt *bp)
 857 {
 858         uint32_t link_speed = 0;
 859         uint32_t speed_capa = 0;
 860
 861         if (bp->link_info == NULL)
 862                 return 0;
 863
 864         link_speed = bp->link_info->support_speeds;
 865
 866         /* If PAM4 is configured, use PAM4 supported speed */
 867         if (link_speed == 0 && bp->link_info->support_pam4_speeds > 0)
 868                 link_speed = bp->link_info->support_pam4_speeds;
 869
 870         if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_LINK_SPEED_100MB)
 871                 speed_capa |= ETH_LINK_SPEED_100M;
 872         if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_100MBHD)
 873                 speed_capa |= ETH_LINK_SPEED_100M_HD;
 874         if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_1GB)
 875                 speed_capa |= ETH_LINK_SPEED_1G;
 876         if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_2_5GB)
 877                 speed_capa |= ETH_LINK_SPEED_2_5G;
 878         if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_10GB)
 879                 speed_capa |= ETH_LINK_SPEED_10G;
 880         if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_20GB)
 881                 speed_capa |= ETH_LINK_SPEED_20G;
 882         if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_25GB)
 883                 speed_capa |= ETH_LINK_SPEED_25G;
 884         if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_40GB)
 885                 speed_capa |= ETH_LINK_SPEED_40G;
 886         if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_50GB)
 887                 speed_capa |= ETH_LINK_SPEED_50G;
 888         if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_100GB)
 889                 speed_capa |= ETH_LINK_SPEED_100G;
 890         if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_PAM4_SPEEDS_50G)
 891                 speed_capa |= ETH_LINK_SPEED_50G;
 892         if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_PAM4_SPEEDS_100G)
 893                 speed_capa |= ETH_LINK_SPEED_100G;
 894         if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_PAM4_SPEEDS_200G)
 895                 speed_capa |= ETH_LINK_SPEED_200G;
 896
 897         if (bp->link_info->auto_mode ==
 898             HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_MODE_NONE)
 899                 speed_capa |= ETH_LINK_SPEED_FIXED;
 900
 901         return speed_capa;
 902 }
 903
 904 static int bnxt_dev_info_get_op(struct rte_eth_dev *eth_dev,
 905                                 struct rte_eth_dev_info *dev_info)
 906 {
 907         struct rte_pci_device *pdev = RTE_DEV_TO_PCI(eth_dev->device);
 908         struct bnxt *bp = eth_dev->data->dev_private;
 909         uint16_t max_vnics, i, j, vpool, vrxq;
 910         unsigned int max_rx_rings;
 911         int rc;
 912
 913         rc = is_bnxt_in_error(bp);
 914         if (rc)
 915                 return rc;
 916
 917         /* MAC Specifics */
 918         dev_info->max_mac_addrs = bp->max_l2_ctx;
 919         dev_info->max_hash_mac_addrs = 0;
 920
 921         /* PF/VF specifics */
 922         if (BNXT_PF(bp))
 923                 dev_info->max_vfs = pdev->max_vfs;
 924
 925         max_rx_rings = bnxt_max_rings(bp);
 926         /* For the sake of symmetry, max_rx_queues = max_tx_queues */
 927         dev_info->max_rx_queues = max_rx_rings;
 928         dev_info->max_tx_queues = max_rx_rings;
 929         dev_info->reta_size = bnxt_rss_hash_tbl_size(bp);
 930         dev_info->hash_key_size = 40;
 931         max_vnics = bp->max_vnics;
 932
 933         /* MTU specifics */
 934         dev_info->min_mtu = RTE_ETHER_MIN_MTU;
 935         dev_info->max_mtu = BNXT_MAX_MTU;
 936
 937         /* Fast path specifics */
 938         dev_info->min_rx_bufsize = 1;
 939         dev_info->max_rx_pktlen = BNXT_MAX_PKT_LEN;
 940
 941         dev_info->rx_offload_capa = BNXT_DEV_RX_OFFLOAD_SUPPORT;
 942         if (bp->flags & BNXT_FLAG_PTP_SUPPORTED)
 943                 dev_info->rx_offload_capa |= DEV_RX_OFFLOAD_TIMESTAMP;
 944         dev_info->tx_queue_offload_capa = DEV_TX_OFFLOAD_MBUF_FAST_FREE;
 945         dev_info->tx_offload_capa = BNXT_DEV_TX_OFFLOAD_SUPPORT |
 946                                     dev_info->tx_queue_offload_capa;
 947         dev_info->flow_type_rss_offloads = BNXT_ETH_RSS_SUPPORT;
 948
 949         dev_info->speed_capa = bnxt_get_speed_capabilities(bp);
 950
 951         /* *INDENT-OFF* */
 952         dev_info->default_rxconf = (struct rte_eth_rxconf) {
 953                 .rx_thresh = {
 954                         .pthresh = 8,
 955                         .hthresh = 8,
 956                         .wthresh = 0,
 957                 },
 958                 .rx_free_thresh = 32,
 959                 .rx_drop_en = BNXT_DEFAULT_RX_DROP_EN,
 960         };
 961
 962         dev_info->default_txconf = (struct rte_eth_txconf) {
 963                 .tx_thresh = {
 964                         .pthresh = 32,
 965                         .hthresh = 0,
 966                         .wthresh = 0,
 967                 },
 968                 .tx_free_thresh = 32,
 969                 .tx_rs_thresh = 32,
 970         };
 971         eth_dev->data->dev_conf.intr_conf.lsc = 1;
 972
 973         eth_dev->data->dev_conf.intr_conf.rxq = 1;
 974         dev_info->rx_desc_lim.nb_min = BNXT_MIN_RING_DESC;
 975         dev_info->rx_desc_lim.nb_max = BNXT_MAX_RX_RING_DESC;
 976         dev_info->tx_desc_lim.nb_min = BNXT_MIN_RING_DESC;
 977         dev_info->tx_desc_lim.nb_max = BNXT_MAX_TX_RING_DESC;
 978
 979         if (BNXT_PF(bp) || BNXT_VF_IS_TRUSTED(bp)) {
 980                 dev_info->switch_info.name = eth_dev->device->name;
 981                 dev_info->switch_info.domain_id = bp->switch_domain_id;
 982                 dev_info->switch_info.port_id =
 983                                 BNXT_PF(bp) ? BNXT_SWITCH_PORT_ID_PF :
 984                                     BNXT_SWITCH_PORT_ID_TRUSTED_VF;
 985         }
 986
 987         /* *INDENT-ON* */
 988
 989         /*
 990          * TODO: default_rxconf, default_txconf, rx_desc_lim, and tx_desc_lim
 991          *       need further investigation.
 992          */
 993
 994         /* VMDq resources */
 995         vpool = 64; /* ETH_64_POOLS */
 996         vrxq = 128; /* ETH_VMDQ_DCB_NUM_QUEUES */
 997         for (i = 0; i < 4; vpool >>= 1, i++) {
 998                 if (max_vnics > vpool) {
 999                         for (j = 0; j < 5; vrxq >>= 1, j++) {
1000                                 if (dev_info->max_rx_queues > vrxq) {
1001                                         if (vpool > vrxq)
1002                                                 vpool = vrxq;
1003                                         goto found;
1004                                 }
1005                         }
1006                         /* Not enough resources to support VMDq */
1007                         break;
1008                 }
1009         }
1010         /* Not enough resources to support VMDq */
1011         vpool = 0;
1012         vrxq = 0;
1013 found:
1014         dev_info->max_vmdq_pools = vpool;
1015         dev_info->vmdq_queue_num = vrxq;
1016
1017         dev_info->vmdq_pool_base = 0;
1018         dev_info->vmdq_queue_base = 0;
1019
1020         return 0;
1021 }
1022
1023 /* Configure the device based on the configuration provided */
1024 static int bnxt_dev_configure_op(struct rte_eth_dev *eth_dev)
1025 {
1026         struct bnxt *bp = eth_dev->data->dev_private;
1027         uint64_t rx_offloads = eth_dev->data->dev_conf.rxmode.offloads;
1028         int rc;
1029
1030         bp->rx_queues = (void *)eth_dev->data->rx_queues;
1031         bp->tx_queues = (void *)eth_dev->data->tx_queues;
1032         bp->tx_nr_rings = eth_dev->data->nb_tx_queues;
1033         bp->rx_nr_rings = eth_dev->data->nb_rx_queues;
1034
1035         rc = is_bnxt_in_error(bp);
1036         if (rc)
1037                 return rc;
1038
1039         if (BNXT_VF(bp) && (bp->flags & BNXT_FLAG_NEW_RM)) {
1040                 rc = bnxt_hwrm_check_vf_rings(bp);
1041                 if (rc) {
1042                         PMD_DRV_LOG(ERR, "HWRM insufficient resources\n");
1043                         return -ENOSPC;
1044                 }
1045
1046                 /* If a resource has already been allocated - in this case
1047                  * it is the async completion ring, free it. Reallocate it after
1048                  * resource reservation. This will ensure the resource counts
1049                  * are calculated correctly.
1050                  */
1051
1052                 pthread_mutex_lock(&bp->def_cp_lock);
1053
1054                 if (!BNXT_HAS_NQ(bp) && bp->async_cp_ring) {
1055                         bnxt_disable_int(bp);
1056                         bnxt_free_cp_ring(bp, bp->async_cp_ring);
1057                 }
1058
1059                 rc = bnxt_hwrm_func_reserve_vf_resc(bp, false);
1060                 if (rc) {
1061                         PMD_DRV_LOG(ERR, "HWRM resource alloc fail:%x\n", rc);
1062                         pthread_mutex_unlock(&bp->def_cp_lock);
1063                         return -ENOSPC;
1064                 }
1065
1066                 if (!BNXT_HAS_NQ(bp) && bp->async_cp_ring) {
1067                         rc = bnxt_alloc_async_cp_ring(bp);
1068                         if (rc) {
1069                                 pthread_mutex_unlock(&bp->def_cp_lock);
1070                                 return rc;
1071                         }
1072                         bnxt_enable_int(bp);
1073                 }
1074
1075                 pthread_mutex_unlock(&bp->def_cp_lock);
1076         }
1077
1078         /* Inherit new configurations */
1079         if (eth_dev->data->nb_rx_queues > bp->max_rx_rings ||
1080             eth_dev->data->nb_tx_queues > bp->max_tx_rings ||
1081             eth_dev->data->nb_rx_queues + eth_dev->data->nb_tx_queues
1082                 + BNXT_NUM_ASYNC_CPR(bp) > bp->max_cp_rings ||
1083             eth_dev->data->nb_rx_queues + eth_dev->data->nb_tx_queues >
1084             bp->max_stat_ctx)
1085                 goto resource_error;
1086
1087         if (BNXT_HAS_RING_GRPS(bp) &&
1088             (uint32_t)(eth_dev->data->nb_rx_queues) > bp->max_ring_grps)
1089                 goto resource_error;
1090
1091         if (!(eth_dev->data->dev_conf.rxmode.mq_mode & ETH_MQ_RX_RSS) &&
1092             bp->max_vnics < eth_dev->data->nb_rx_queues)
1093                 goto resource_error;
1094
1095         bp->rx_cp_nr_rings = bp->rx_nr_rings;
1096         bp->tx_cp_nr_rings = bp->tx_nr_rings;
1097
1098         if (eth_dev->data->dev_conf.rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG)
1099                 rx_offloads |= DEV_RX_OFFLOAD_RSS_HASH;
1100         eth_dev->data->dev_conf.rxmode.offloads = rx_offloads;
1101
1102         if (rx_offloads & DEV_RX_OFFLOAD_JUMBO_FRAME) {
1103                 eth_dev->data->mtu =
1104                         eth_dev->data->dev_conf.rxmode.max_rx_pkt_len -
1105                         RTE_ETHER_HDR_LEN - RTE_ETHER_CRC_LEN - VLAN_TAG_SIZE *
1106                         BNXT_NUM_VLANS;
1107                 bnxt_mtu_set_op(eth_dev, eth_dev->data->mtu);
1108         }
1109         return 0;
1110
1111 resource_error:
1112         PMD_DRV_LOG(ERR,
1113                     "Insufficient resources to support requested config\n");
1114         PMD_DRV_LOG(ERR,
1115                     "Num Queues Requested: Tx %d, Rx %d\n",
1116                     eth_dev->data->nb_tx_queues,
1117                     eth_dev->data->nb_rx_queues);
1118         PMD_DRV_LOG(ERR,
1119                     "MAX: TxQ %d, RxQ %d, CQ %d Stat %d, Grp %d, Vnic %d\n",
1120                     bp->max_tx_rings, bp->max_rx_rings, bp->max_cp_rings,
1121                     bp->max_stat_ctx, bp->max_ring_grps, bp->max_vnics);
1122         return -ENOSPC;
1123 }
1124
1125 void bnxt_print_link_info(struct rte_eth_dev *eth_dev)
1126 {
1127         struct rte_eth_link *link = &eth_dev->data->dev_link;
1128
1129         if (link->link_status)
1130                 PMD_DRV_LOG(INFO, "Port %d Link Up - speed %u Mbps - %s\n",
1131                         eth_dev->data->port_id,
1132                         (uint32_t)link->link_speed,
1133                         (link->link_duplex == ETH_LINK_FULL_DUPLEX) ?
1134                         ("full-duplex") : ("half-duplex\n"));
1135         else
1136                 PMD_DRV_LOG(INFO, "Port %d Link Down\n",
1137                         eth_dev->data->port_id);
1138 }
1139
1140 /*
1141  * Determine whether the current configuration requires support for scattered
1142  * receive; return 1 if scattered receive is required and 0 if not.
1143  */
1144 static int bnxt_scattered_rx(struct rte_eth_dev *eth_dev)
1145 {
1146         uint16_t buf_size;
1147         int i;
1148
1149         if (eth_dev->data->dev_conf.rxmode.offloads & DEV_RX_OFFLOAD_SCATTER)
1150                 return 1;
1151
1152         if (eth_dev->data->dev_conf.rxmode.offloads & DEV_RX_OFFLOAD_TCP_LRO)
1153                 return 1;
1154
1155         for (i = 0; i < eth_dev->data->nb_rx_queues; i++) {
1156                 struct bnxt_rx_queue *rxq = eth_dev->data->rx_queues[i];
1157
1158                 buf_size = (uint16_t)(rte_pktmbuf_data_room_size(rxq->mb_pool) -
1159                                       RTE_PKTMBUF_HEADROOM);
1160                 if (eth_dev->data->dev_conf.rxmode.max_rx_pkt_len > buf_size)
1161                         return 1;
1162         }
1163         return 0;
1164 }
1165
1166 static eth_rx_burst_t
1167 bnxt_receive_function(struct rte_eth_dev *eth_dev)
1168 {
1169         struct bnxt *bp = eth_dev->data->dev_private;
1170
1171         /* Disable vector mode RX for Stingray2 for now */
1172         if (BNXT_CHIP_SR2(bp)) {
1173                 bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
1174                 return bnxt_recv_pkts;
1175         }
1176
1177 #if defined(RTE_ARCH_X86) || defined(RTE_ARCH_ARM64)
1178 #ifndef RTE_LIBRTE_IEEE1588
1179         /*
1180          * Vector mode receive can be enabled only if scatter rx is not
1181          * in use and rx offloads are limited to VLAN stripping and
1182          * CRC stripping.
1183          */
1184         if (!eth_dev->data->scattered_rx &&
1185             !(eth_dev->data->dev_conf.rxmode.offloads &
1186               ~(DEV_RX_OFFLOAD_VLAN_STRIP |
1187                 DEV_RX_OFFLOAD_KEEP_CRC |
1188                 DEV_RX_OFFLOAD_JUMBO_FRAME |
1189                 DEV_RX_OFFLOAD_IPV4_CKSUM |
1190                 DEV_RX_OFFLOAD_UDP_CKSUM |
1191                 DEV_RX_OFFLOAD_TCP_CKSUM |
1192                 DEV_RX_OFFLOAD_OUTER_IPV4_CKSUM |
1193                 DEV_RX_OFFLOAD_OUTER_UDP_CKSUM |
1194                 DEV_RX_OFFLOAD_RSS_HASH |
1195                 DEV_RX_OFFLOAD_VLAN_FILTER)) &&
1196             !BNXT_TRUFLOW_EN(bp) && BNXT_NUM_ASYNC_CPR(bp) &&
1197             rte_vect_get_max_simd_bitwidth() >= RTE_VECT_SIMD_128) {
1198                 PMD_DRV_LOG(INFO, "Using vector mode receive for port %d\n",
1199                             eth_dev->data->port_id);
1200                 bp->flags |= BNXT_FLAG_RX_VECTOR_PKT_MODE;
1201                 return bnxt_recv_pkts_vec;
1202         }
1203         PMD_DRV_LOG(INFO, "Vector mode receive disabled for port %d\n",
1204                     eth_dev->data->port_id);
1205         PMD_DRV_LOG(INFO,
1206                     "Port %d scatter: %d rx offload: %" PRIX64 "\n",
1207                     eth_dev->data->port_id,
1208                     eth_dev->data->scattered_rx,
1209                     eth_dev->data->dev_conf.rxmode.offloads);
1210 #endif
1211 #endif
1212         bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
1213         return bnxt_recv_pkts;
1214 }
1215
1216 static eth_tx_burst_t
1217 bnxt_transmit_function(struct rte_eth_dev *eth_dev)
1218 {
1219         struct bnxt *bp = eth_dev->data->dev_private;
1220
1221         /* Disable vector mode TX for Stingray2 for now */
1222         if (BNXT_CHIP_SR2(bp))
1223                 return bnxt_xmit_pkts;
1224
1225 #if defined(RTE_ARCH_X86) || defined(RTE_ARCH_ARM64)
1226 #ifndef RTE_LIBRTE_IEEE1588
1227         uint64_t offloads = eth_dev->data->dev_conf.txmode.offloads;
1228
1229         /*
1230          * Vector mode transmit can be enabled only if not using scatter rx
1231          * or tx offloads.
1232          */
1233         if (!eth_dev->data->scattered_rx &&
1234             !(offloads & ~DEV_TX_OFFLOAD_MBUF_FAST_FREE) &&
1235             !BNXT_TRUFLOW_EN(bp) &&
1236             rte_vect_get_max_simd_bitwidth() >= RTE_VECT_SIMD_128) {
1237                 PMD_DRV_LOG(INFO, "Using vector mode transmit for port %d\n",
1238                             eth_dev->data->port_id);
1239                 return bnxt_xmit_pkts_vec;
1240         }
1241         PMD_DRV_LOG(INFO, "Vector mode transmit disabled for port %d\n",
1242                     eth_dev->data->port_id);
1243         PMD_DRV_LOG(INFO,
1244                     "Port %d scatter: %d tx offload: %" PRIX64 "\n",
1245                     eth_dev->data->port_id,
1246                     eth_dev->data->scattered_rx,
1247                     offloads);
1248 #endif
1249 #endif
1250         return bnxt_xmit_pkts;
1251 }
1252
1253 static int bnxt_handle_if_change_status(struct bnxt *bp)
1254 {
1255         int rc;
1256
1257         /* Since fw has undergone a reset and lost all contexts,
1258          * set fatal flag to not issue hwrm during cleanup
1259          */
1260         bp->flags |= BNXT_FLAG_FATAL_ERROR;
1261         bnxt_uninit_resources(bp, true);
1262
1263         /* clear fatal flag so that re-init happens */
1264         bp->flags &= ~BNXT_FLAG_FATAL_ERROR;
1265         rc = bnxt_init_resources(bp, true);
1266
1267         bp->flags &= ~BNXT_FLAG_IF_CHANGE_HOT_FW_RESET_DONE;
1268
1269         return rc;
1270 }
1271
1272 static int bnxt_dev_set_link_up_op(struct rte_eth_dev *eth_dev)
1273 {
1274         struct bnxt *bp = eth_dev->data->dev_private;
1275         int rc = 0;
1276
1277         if (!BNXT_SINGLE_PF(bp))
1278                 return -ENOTSUP;
1279
1280         if (!bp->link_info->link_up)
1281                 rc = bnxt_set_hwrm_link_config(bp, true);
1282         if (!rc)
1283                 eth_dev->data->dev_link.link_status = 1;
1284
1285         bnxt_print_link_info(eth_dev);
1286         return rc;
1287 }
1288
1289 static int bnxt_dev_set_link_down_op(struct rte_eth_dev *eth_dev)
1290 {
1291         struct bnxt *bp = eth_dev->data->dev_private;
1292
1293         if (!BNXT_SINGLE_PF(bp))
1294                 return -ENOTSUP;
1295
1296         eth_dev->data->dev_link.link_status = 0;
1297         bnxt_set_hwrm_link_config(bp, false);
1298         bp->link_info->link_up = 0;
1299
1300         return 0;
1301 }
1302
1303 static void bnxt_free_switch_domain(struct bnxt *bp)
1304 {
1305         int rc = 0;
1306
1307         if (!(BNXT_PF(bp) || BNXT_VF_IS_TRUSTED(bp)))
1308                 return;
1309
1310         rc = rte_eth_switch_domain_free(bp->switch_domain_id);
1311         if (rc)
1312                 PMD_DRV_LOG(ERR, "free switch domain:%d fail: %d\n",
1313                             bp->switch_domain_id, rc);
1314 }
1315
1316 static void bnxt_ptp_get_current_time(void *arg)
1317 {
1318         struct bnxt *bp = arg;
1319         struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
1320         int rc;
1321
1322         rc = is_bnxt_in_error(bp);
1323         if (rc)
1324                 return;
1325
1326         if (!ptp)
1327                 return;
1328
1329         bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_CURRENT_TIME,
1330                                 &ptp->current_time);
1331
1332         rc = rte_eal_alarm_set(US_PER_S, bnxt_ptp_get_current_time, (void *)bp);
1333         if (rc != 0) {
1334                 PMD_DRV_LOG(ERR, "Failed to re-schedule PTP alarm\n");
1335                 bp->flags2 &= ~BNXT_FLAGS2_PTP_ALARM_SCHEDULED;
1336         }
1337 }
1338
1339 static int bnxt_schedule_ptp_alarm(struct bnxt *bp)
1340 {
1341         struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
1342         int rc;
1343
1344         if (bp->flags2 & BNXT_FLAGS2_PTP_ALARM_SCHEDULED)
1345                 return 0;
1346
1347         bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_CURRENT_TIME,
1348                                 &ptp->current_time);
1349
1350         rc = rte_eal_alarm_set(US_PER_S, bnxt_ptp_get_current_time, (void *)bp);
1351         return rc;
1352 }
1353
1354 static void bnxt_cancel_ptp_alarm(struct bnxt *bp)
1355 {
1356         if (bp->flags2 & BNXT_FLAGS2_PTP_ALARM_SCHEDULED) {
1357                 rte_eal_alarm_cancel(bnxt_ptp_get_current_time, (void *)bp);
1358                 bp->flags2 &= ~BNXT_FLAGS2_PTP_ALARM_SCHEDULED;
1359         }
1360 }
1361
1362 static void bnxt_ptp_stop(struct bnxt *bp)
1363 {
1364         bnxt_cancel_ptp_alarm(bp);
1365         bp->flags2 &= ~BNXT_FLAGS2_PTP_TIMESYNC_ENABLED;
1366 }
1367
1368 static int bnxt_ptp_start(struct bnxt *bp)
1369 {
1370         int rc;
1371
1372         rc = bnxt_schedule_ptp_alarm(bp);
1373         if (rc != 0) {
1374                 PMD_DRV_LOG(ERR, "Failed to schedule PTP alarm\n");
1375         } else {
1376                 bp->flags2 |= BNXT_FLAGS2_PTP_TIMESYNC_ENABLED;
1377                 bp->flags2 |= BNXT_FLAGS2_PTP_ALARM_SCHEDULED;
1378         }
1379
1380         return rc;
1381 }
1382
1383 static int bnxt_dev_stop(struct rte_eth_dev *eth_dev)
1384 {
1385         struct bnxt *bp = eth_dev->data->dev_private;
1386         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
1387         struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
1388         struct rte_eth_link link;
1389         int ret;
1390
1391         eth_dev->data->dev_started = 0;
1392         eth_dev->data->scattered_rx = 0;
1393
1394         /* Prevent crashes when queues are still in use */
1395         eth_dev->rx_pkt_burst = &bnxt_dummy_recv_pkts;
1396         eth_dev->tx_pkt_burst = &bnxt_dummy_xmit_pkts;
1397
1398         bnxt_disable_int(bp);
1399
1400         /* disable uio/vfio intr/eventfd mapping */
1401         rte_intr_disable(intr_handle);
1402
1403         /* Stop the child representors for this device */
1404         ret = bnxt_rep_stop_all(bp);
1405         if (ret != 0)
1406                 return ret;
1407
1408         /* delete the bnxt ULP port details */
1409         bnxt_ulp_port_deinit(bp);
1410
1411         bnxt_cancel_fw_health_check(bp);
1412
1413         if (BNXT_P5_PTP_TIMESYNC_ENABLED(bp))
1414                 bnxt_cancel_ptp_alarm(bp);
1415
1416         /* Do not bring link down during reset recovery */
1417         if (!is_bnxt_in_error(bp)) {
1418                 bnxt_dev_set_link_down_op(eth_dev);
1419                 /* Wait for link to be reset */
1420                 if (BNXT_SINGLE_PF(bp))
1421                         rte_delay_ms(500);
1422                 /* clear the recorded link status */
1423                 memset(&link, 0, sizeof(link));
1424                 rte_eth_linkstatus_set(eth_dev, &link);
1425         }
1426
1427         /* Clean queue intr-vector mapping */
1428         rte_intr_efd_disable(intr_handle);
1429         if (intr_handle->intr_vec != NULL) {
1430                 rte_free(intr_handle->intr_vec);
1431                 intr_handle->intr_vec = NULL;
1432         }
1433
1434         bnxt_hwrm_port_clr_stats(bp);
1435         bnxt_free_tx_mbufs(bp);
1436         bnxt_free_rx_mbufs(bp);
1437         /* Process any remaining notifications in default completion queue */
1438         bnxt_int_handler(eth_dev);
1439         bnxt_shutdown_nic(bp);
1440         bnxt_hwrm_if_change(bp, false);
1441
1442         rte_free(bp->mark_table);
1443         bp->mark_table = NULL;
1444
1445         bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
1446         bp->rx_cosq_cnt = 0;
1447         /* All filters are deleted on a port stop. */
1448         if (BNXT_FLOW_XSTATS_EN(bp))
1449                 bp->flow_stat->flow_count = 0;
1450
1451         return 0;
1452 }
1453
1454 /* Unload the driver, release resources */
1455 static int bnxt_dev_stop_op(struct rte_eth_dev *eth_dev)
1456 {
1457         struct bnxt *bp = eth_dev->data->dev_private;
1458
1459         pthread_mutex_lock(&bp->err_recovery_lock);
1460         if (bp->flags & BNXT_FLAG_FW_RESET) {
1461                 PMD_DRV_LOG(ERR,
1462                             "Adapter recovering from error..Please retry\n");
1463                 pthread_mutex_unlock(&bp->err_recovery_lock);
1464                 return -EAGAIN;
1465         }
1466         pthread_mutex_unlock(&bp->err_recovery_lock);
1467
1468         return bnxt_dev_stop(eth_dev);
1469 }
1470
1471 static int bnxt_dev_start_op(struct rte_eth_dev *eth_dev)
1472 {
1473         struct bnxt *bp = eth_dev->data->dev_private;
1474         uint64_t rx_offloads = eth_dev->data->dev_conf.rxmode.offloads;
1475         int vlan_mask = 0;
1476         int rc, retry_cnt = BNXT_IF_CHANGE_RETRY_COUNT;
1477
1478         if (!eth_dev->data->nb_tx_queues || !eth_dev->data->nb_rx_queues) {
1479                 PMD_DRV_LOG(ERR, "Queues are not configured yet!\n");
1480                 return -EINVAL;
1481         }
1482
1483         if (bp->rx_cp_nr_rings > RTE_ETHDEV_QUEUE_STAT_CNTRS)
1484                 PMD_DRV_LOG(ERR,
1485                             "RxQ cnt %d > RTE_ETHDEV_QUEUE_STAT_CNTRS %d\n",
1486                             bp->rx_cp_nr_rings, RTE_ETHDEV_QUEUE_STAT_CNTRS);
1487
1488         do {
1489                 rc = bnxt_hwrm_if_change(bp, true);
1490                 if (rc == 0 || rc != -EAGAIN)
1491                         break;
1492
1493                 rte_delay_ms(BNXT_IF_CHANGE_RETRY_INTERVAL);
1494         } while (retry_cnt--);
1495
1496         if (rc)
1497                 return rc;
1498
1499         if (bp->flags & BNXT_FLAG_IF_CHANGE_HOT_FW_RESET_DONE) {
1500                 rc = bnxt_handle_if_change_status(bp);
1501                 if (rc)
1502                         return rc;
1503         }
1504
1505         bnxt_enable_int(bp);
1506
1507         eth_dev->data->scattered_rx = bnxt_scattered_rx(eth_dev);
1508
1509         rc = bnxt_start_nic(bp);
1510         if (rc)
1511                 goto error;
1512
1513         eth_dev->data->dev_started = 1;
1514
1515         bnxt_link_update_op(eth_dev, 1);
1516
1517         if (rx_offloads & DEV_RX_OFFLOAD_VLAN_FILTER)
1518                 vlan_mask |= ETH_VLAN_FILTER_MASK;
1519         if (rx_offloads & DEV_RX_OFFLOAD_VLAN_STRIP)
1520                 vlan_mask |= ETH_VLAN_STRIP_MASK;
1521         rc = bnxt_vlan_offload_set_op(eth_dev, vlan_mask);
1522         if (rc)
1523                 goto error;
1524
1525         /* Initialize bnxt ULP port details */
1526         rc = bnxt_ulp_port_init(bp);
1527         if (rc)
1528                 goto error;
1529
1530         eth_dev->rx_pkt_burst = bnxt_receive_function(eth_dev);
1531         eth_dev->tx_pkt_burst = bnxt_transmit_function(eth_dev);
1532
1533         bnxt_schedule_fw_health_check(bp);
1534
1535         if (BNXT_P5_PTP_TIMESYNC_ENABLED(bp))
1536                 bnxt_schedule_ptp_alarm(bp);
1537
1538         return 0;
1539
1540 error:
1541         bnxt_dev_stop(eth_dev);
1542         return rc;
1543 }
1544
1545 static void
1546 bnxt_uninit_locks(struct bnxt *bp)
1547 {
1548         pthread_mutex_destroy(&bp->flow_lock);
1549         pthread_mutex_destroy(&bp->def_cp_lock);
1550         pthread_mutex_destroy(&bp->health_check_lock);
1551         pthread_mutex_destroy(&bp->err_recovery_lock);
1552         if (bp->rep_info) {
1553                 pthread_mutex_destroy(&bp->rep_info->vfr_lock);
1554                 pthread_mutex_destroy(&bp->rep_info->vfr_start_lock);
1555         }
1556 }
1557
1558 static void bnxt_drv_uninit(struct bnxt *bp)
1559 {
1560         bnxt_free_leds_info(bp);
1561         bnxt_free_cos_queues(bp);
1562         bnxt_free_link_info(bp);
1563         bnxt_free_parent_info(bp);
1564         bnxt_uninit_locks(bp);
1565
1566         rte_memzone_free((const struct rte_memzone *)bp->tx_mem_zone);
1567         bp->tx_mem_zone = NULL;
1568         rte_memzone_free((const struct rte_memzone *)bp->rx_mem_zone);
1569         bp->rx_mem_zone = NULL;
1570
1571         bnxt_free_vf_info(bp);
1572         bnxt_free_pf_info(bp);
1573
1574         rte_free(bp->grp_info);
1575         bp->grp_info = NULL;
1576 }
1577
1578 static int bnxt_dev_close_op(struct rte_eth_dev *eth_dev)
1579 {
1580         struct bnxt *bp = eth_dev->data->dev_private;
1581         int ret = 0;
1582
1583         if (rte_eal_process_type() != RTE_PROC_PRIMARY)
1584                 return 0;
1585
1586         pthread_mutex_lock(&bp->err_recovery_lock);
1587         if (bp->flags & BNXT_FLAG_FW_RESET) {
1588                 PMD_DRV_LOG(ERR,
1589                             "Adapter recovering from error...Please retry\n");
1590                 pthread_mutex_unlock(&bp->err_recovery_lock);
1591                 return -EAGAIN;
1592         }
1593         pthread_mutex_unlock(&bp->err_recovery_lock);
1594
1595         /* cancel the recovery handler before remove dev */
1596         rte_eal_alarm_cancel(bnxt_dev_reset_and_resume, (void *)bp);
1597         rte_eal_alarm_cancel(bnxt_dev_recover, (void *)bp);
1598         bnxt_cancel_fc_thread(bp);
1599
1600         if (eth_dev->data->dev_started)
1601                 ret = bnxt_dev_stop(eth_dev);
1602
1603         bnxt_uninit_resources(bp, false);
1604
1605         bnxt_drv_uninit(bp);
1606
1607         return ret;
1608 }
1609
1610 static void bnxt_mac_addr_remove_op(struct rte_eth_dev *eth_dev,
1611                                     uint32_t index)
1612 {
1613         struct bnxt *bp = eth_dev->data->dev_private;
1614         uint64_t pool_mask = eth_dev->data->mac_pool_sel[index];
1615         struct bnxt_vnic_info *vnic;
1616         struct bnxt_filter_info *filter, *temp_filter;
1617         uint32_t i;
1618
1619         if (is_bnxt_in_error(bp))
1620                 return;
1621
1622         /*
1623          * Loop through all VNICs from the specified filter flow pools to
1624          * remove the corresponding MAC addr filter
1625          */
1626         for (i = 0; i < bp->nr_vnics; i++) {
1627                 if (!(pool_mask & (1ULL << i)))
1628                         continue;
1629
1630                 vnic = &bp->vnic_info[i];
1631                 filter = STAILQ_FIRST(&vnic->filter);
1632                 while (filter) {
1633                         temp_filter = STAILQ_NEXT(filter, next);
1634                         if (filter->mac_index == index) {
1635                                 STAILQ_REMOVE(&vnic->filter, filter,
1636                                                 bnxt_filter_info, next);
1637                                 bnxt_hwrm_clear_l2_filter(bp, filter);
1638                                 bnxt_free_filter(bp, filter);
1639                         }
1640                         filter = temp_filter;
1641                 }
1642         }
1643 }
1644
1645 static int bnxt_add_mac_filter(struct bnxt *bp, struct bnxt_vnic_info *vnic,
1646                                struct rte_ether_addr *mac_addr, uint32_t index,
1647                                uint32_t pool)
1648 {
1649         struct bnxt_filter_info *filter;
1650         int rc = 0;
1651
1652         /* Attach requested MAC address to the new l2_filter */
1653         STAILQ_FOREACH(filter, &vnic->filter, next) {
1654                 if (filter->mac_index == index) {
1655                         PMD_DRV_LOG(DEBUG,
1656                                     "MAC addr already existed for pool %d\n",
1657                                     pool);
1658                         return 0;
1659                 }
1660         }
1661
1662         filter = bnxt_alloc_filter(bp);
1663         if (!filter) {
1664                 PMD_DRV_LOG(ERR, "L2 filter alloc failed\n");
1665                 return -ENODEV;
1666         }
1667
1668         /* bnxt_alloc_filter copies default MAC to filter->l2_addr. So,
1669          * if the MAC that's been programmed now is a different one, then,
1670          * copy that addr to filter->l2_addr
1671          */
1672         if (mac_addr)
1673                 memcpy(filter->l2_addr, mac_addr, RTE_ETHER_ADDR_LEN);
1674         filter->flags |= HWRM_CFA_L2_FILTER_ALLOC_INPUT_FLAGS_OUTERMOST;
1675
1676         rc = bnxt_hwrm_set_l2_filter(bp, vnic->fw_vnic_id, filter);
1677         if (!rc) {
1678                 filter->mac_index = index;
1679                 if (filter->mac_index == 0)
1680                         STAILQ_INSERT_HEAD(&vnic->filter, filter, next);
1681                 else
1682                         STAILQ_INSERT_TAIL(&vnic->filter, filter, next);
1683         } else {
1684                 bnxt_free_filter(bp, filter);
1685         }
1686
1687         return rc;
1688 }
1689
1690 static int bnxt_mac_addr_add_op(struct rte_eth_dev *eth_dev,
1691                                 struct rte_ether_addr *mac_addr,
1692                                 uint32_t index, uint32_t pool)
1693 {
1694         struct bnxt *bp = eth_dev->data->dev_private;
1695         struct bnxt_vnic_info *vnic = &bp->vnic_info[pool];
1696         int rc = 0;
1697
1698         rc = is_bnxt_in_error(bp);
1699         if (rc)
1700                 return rc;
1701
1702         if (BNXT_VF(bp) && !BNXT_VF_IS_TRUSTED(bp)) {
1703                 PMD_DRV_LOG(ERR, "Cannot add MAC address to a VF interface\n");
1704                 return -ENOTSUP;
1705         }
1706
1707         if (!vnic) {
1708                 PMD_DRV_LOG(ERR, "VNIC not found for pool %d!\n", pool);
1709                 return -EINVAL;
1710         }
1711
1712         /* Filter settings will get applied when port is started */
1713         if (!eth_dev->data->dev_started)
1714                 return 0;
1715
1716         rc = bnxt_add_mac_filter(bp, vnic, mac_addr, index, pool);
1717
1718         return rc;
1719 }
1720
1721 int bnxt_link_update_op(struct rte_eth_dev *eth_dev, int wait_to_complete)
1722 {
1723         int rc = 0;
1724         struct bnxt *bp = eth_dev->data->dev_private;
1725         struct rte_eth_link new;
1726         int cnt = wait_to_complete ? BNXT_MAX_LINK_WAIT_CNT :
1727                         BNXT_MIN_LINK_WAIT_CNT;
1728
1729         rc = is_bnxt_in_error(bp);
1730         if (rc)
1731                 return rc;
1732
1733         memset(&new, 0, sizeof(new));
1734
1735         if (bp->link_info == NULL)
1736                 goto out;
1737
1738         do {
1739                 /* Retrieve link info from hardware */
1740                 rc = bnxt_get_hwrm_link_config(bp, &new);
1741                 if (rc) {
1742                         new.link_speed = ETH_LINK_SPEED_100M;
1743                         new.link_duplex = ETH_LINK_FULL_DUPLEX;
1744                         PMD_DRV_LOG(ERR,
1745                                 "Failed to retrieve link rc = 0x%x!\n", rc);
1746                         goto out;
1747                 }
1748
1749                 if (!wait_to_complete || new.link_status)
1750                         break;
1751
1752                 rte_delay_ms(BNXT_LINK_WAIT_INTERVAL);
1753         } while (cnt--);
1754
1755         /* Only single function PF can bring phy down.
1756          * When port is stopped, report link down for VF/MH/NPAR functions.
1757          */
1758         if (!BNXT_SINGLE_PF(bp) && !eth_dev->data->dev_started)
1759                 memset(&new, 0, sizeof(new));
1760
1761 out:
1762         /* Timed out or success */
1763         if (new.link_status != eth_dev->data->dev_link.link_status ||
1764             new.link_speed != eth_dev->data->dev_link.link_speed) {
1765                 rte_eth_linkstatus_set(eth_dev, &new);
1766
1767                 rte_eth_dev_callback_process(eth_dev,
1768                                              RTE_ETH_EVENT_INTR_LSC,
1769                                              NULL);
1770
1771                 bnxt_print_link_info(eth_dev);
1772         }
1773
1774         return rc;
1775 }
1776
1777 static int bnxt_promiscuous_enable_op(struct rte_eth_dev *eth_dev)
1778 {
1779         struct bnxt *bp = eth_dev->data->dev_private;
1780         struct bnxt_vnic_info *vnic;
1781         uint32_t old_flags;
1782         int rc;
1783
1784         rc = is_bnxt_in_error(bp);
1785         if (rc)
1786                 return rc;
1787
1788         /* Filter settings will get applied when port is started */
1789         if (!eth_dev->data->dev_started)
1790                 return 0;
1791
1792         if (bp->vnic_info == NULL)
1793                 return 0;
1794
1795         vnic = BNXT_GET_DEFAULT_VNIC(bp);
1796
1797         old_flags = vnic->flags;
1798         vnic->flags |= BNXT_VNIC_INFO_PROMISC;
1799         rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1800         if (rc != 0)
1801                 vnic->flags = old_flags;
1802
1803         return rc;
1804 }
1805
1806 static int bnxt_promiscuous_disable_op(struct rte_eth_dev *eth_dev)
1807 {
1808         struct bnxt *bp = eth_dev->data->dev_private;
1809         struct bnxt_vnic_info *vnic;
1810         uint32_t old_flags;
1811         int rc;
1812
1813         rc = is_bnxt_in_error(bp);
1814         if (rc)
1815                 return rc;
1816
1817         /* Filter settings will get applied when port is started */
1818         if (!eth_dev->data->dev_started)
1819                 return 0;
1820
1821         if (bp->vnic_info == NULL)
1822                 return 0;
1823
1824         vnic = BNXT_GET_DEFAULT_VNIC(bp);
1825
1826         old_flags = vnic->flags;
1827         vnic->flags &= ~BNXT_VNIC_INFO_PROMISC;
1828         rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1829         if (rc != 0)
1830                 vnic->flags = old_flags;
1831
1832         return rc;
1833 }
1834
1835 static int bnxt_allmulticast_enable_op(struct rte_eth_dev *eth_dev)
1836 {
1837         struct bnxt *bp = eth_dev->data->dev_private;
1838         struct bnxt_vnic_info *vnic;
1839         uint32_t old_flags;
1840         int rc;
1841
1842         rc = is_bnxt_in_error(bp);
1843         if (rc)
1844                 return rc;
1845
1846         /* Filter settings will get applied when port is started */
1847         if (!eth_dev->data->dev_started)
1848                 return 0;
1849
1850         if (bp->vnic_info == NULL)
1851                 return 0;
1852
1853         vnic = BNXT_GET_DEFAULT_VNIC(bp);
1854
1855         old_flags = vnic->flags;
1856         vnic->flags |= BNXT_VNIC_INFO_ALLMULTI;
1857         rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1858         if (rc != 0)
1859                 vnic->flags = old_flags;
1860
1861         return rc;
1862 }
1863
1864 static int bnxt_allmulticast_disable_op(struct rte_eth_dev *eth_dev)
1865 {
1866         struct bnxt *bp = eth_dev->data->dev_private;
1867         struct bnxt_vnic_info *vnic;
1868         uint32_t old_flags;
1869         int rc;
1870
1871         rc = is_bnxt_in_error(bp);
1872         if (rc)
1873                 return rc;
1874
1875         /* Filter settings will get applied when port is started */
1876         if (!eth_dev->data->dev_started)
1877                 return 0;
1878
1879         if (bp->vnic_info == NULL)
1880                 return 0;
1881
1882         vnic = BNXT_GET_DEFAULT_VNIC(bp);
1883
1884         old_flags = vnic->flags;
1885         vnic->flags &= ~BNXT_VNIC_INFO_ALLMULTI;
1886         rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1887         if (rc != 0)
1888                 vnic->flags = old_flags;
1889
1890         return rc;
1891 }
1892
1893 /* Return bnxt_rx_queue pointer corresponding to a given rxq. */
1894 static struct bnxt_rx_queue *bnxt_qid_to_rxq(struct bnxt *bp, uint16_t qid)
1895 {
1896         if (qid >= bp->rx_nr_rings)
1897                 return NULL;
1898
1899         return bp->eth_dev->data->rx_queues[qid];
1900 }
1901
1902 /* Return rxq corresponding to a given rss table ring/group ID. */
1903 static uint16_t bnxt_rss_to_qid(struct bnxt *bp, uint16_t fwr)
1904 {
1905         struct bnxt_rx_queue *rxq;
1906         unsigned int i;
1907
1908         if (!BNXT_HAS_RING_GRPS(bp)) {
1909                 for (i = 0; i < bp->rx_nr_rings; i++) {
1910                         rxq = bp->eth_dev->data->rx_queues[i];
1911                         if (rxq->rx_ring->rx_ring_struct->fw_ring_id == fwr)
1912                                 return rxq->index;
1913                 }
1914         } else {
1915                 for (i = 0; i < bp->rx_nr_rings; i++) {
1916                         if (bp->grp_info[i].fw_grp_id == fwr)
1917                                 return i;
1918                 }
1919         }
1920
1921         return INVALID_HW_RING_ID;
1922 }
1923
1924 static int bnxt_reta_update_op(struct rte_eth_dev *eth_dev,
1925                             struct rte_eth_rss_reta_entry64 *reta_conf,
1926                             uint16_t reta_size)
1927 {
1928         struct bnxt *bp = eth_dev->data->dev_private;
1929         struct rte_eth_conf *dev_conf = &bp->eth_dev->data->dev_conf;
1930         struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
1931         uint16_t tbl_size = bnxt_rss_hash_tbl_size(bp);
1932         uint16_t idx, sft;
1933         int i, rc;
1934
1935         rc = is_bnxt_in_error(bp);
1936         if (rc)
1937                 return rc;
1938
1939         if (!vnic->rss_table)
1940                 return -EINVAL;
1941
1942         if (!(dev_conf->rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG))
1943                 return -EINVAL;
1944
1945         if (reta_size != tbl_size) {
1946                 PMD_DRV_LOG(ERR, "The configured hash table lookup size "
1947                         "(%d) must equal the size supported by the hardware "
1948                         "(%d)\n", reta_size, tbl_size);
1949                 return -EINVAL;
1950         }
1951
1952         for (i = 0; i < reta_size; i++) {
1953                 struct bnxt_rx_queue *rxq;
1954
1955                 idx = i / RTE_RETA_GROUP_SIZE;
1956                 sft = i % RTE_RETA_GROUP_SIZE;
1957
1958                 if (!(reta_conf[idx].mask & (1ULL << sft)))
1959                         continue;
1960
1961                 rxq = bnxt_qid_to_rxq(bp, reta_conf[idx].reta[sft]);
1962                 if (!rxq) {
1963                         PMD_DRV_LOG(ERR, "Invalid ring in reta_conf.\n");
1964                         return -EINVAL;
1965                 }
1966
1967                 if (BNXT_CHIP_P5(bp)) {
1968                         vnic->rss_table[i * 2] =
1969                                 rxq->rx_ring->rx_ring_struct->fw_ring_id;
1970                         vnic->rss_table[i * 2 + 1] =
1971                                 rxq->cp_ring->cp_ring_struct->fw_ring_id;
1972                 } else {
1973                         vnic->rss_table[i] =
1974                             vnic->fw_grp_ids[reta_conf[idx].reta[sft]];
1975                 }
1976         }
1977
1978         rc = bnxt_hwrm_vnic_rss_cfg(bp, vnic);
1979         return rc;
1980 }
1981
1982 static int bnxt_reta_query_op(struct rte_eth_dev *eth_dev,
1983                               struct rte_eth_rss_reta_entry64 *reta_conf,
1984                               uint16_t reta_size)
1985 {
1986         struct bnxt *bp = eth_dev->data->dev_private;
1987         struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
1988         uint16_t tbl_size = bnxt_rss_hash_tbl_size(bp);
1989         uint16_t idx, sft, i;
1990         int rc;
1991
1992         rc = is_bnxt_in_error(bp);
1993         if (rc)
1994                 return rc;
1995
1996         /* Retrieve from the default VNIC */
1997         if (!vnic)
1998                 return -EINVAL;
1999         if (!vnic->rss_table)
2000                 return -EINVAL;
2001
2002         if (reta_size != tbl_size) {
2003                 PMD_DRV_LOG(ERR, "The configured hash table lookup size "
2004                         "(%d) must equal the size supported by the hardware "
2005                         "(%d)\n", reta_size, tbl_size);
2006                 return -EINVAL;
2007         }
2008
2009         for (idx = 0, i = 0; i < reta_size; i++) {
2010                 idx = i / RTE_RETA_GROUP_SIZE;
2011                 sft = i % RTE_RETA_GROUP_SIZE;
2012
2013                 if (reta_conf[idx].mask & (1ULL << sft)) {
2014                         uint16_t qid;
2015
2016                         if (BNXT_CHIP_P5(bp))
2017                                 qid = bnxt_rss_to_qid(bp,
2018                                                       vnic->rss_table[i * 2]);
2019                         else
2020                                 qid = bnxt_rss_to_qid(bp, vnic->rss_table[i]);
2021
2022                         if (qid == INVALID_HW_RING_ID) {
2023                                 PMD_DRV_LOG(ERR, "Inv. entry in rss table.\n");
2024                                 return -EINVAL;
2025                         }
2026                         reta_conf[idx].reta[sft] = qid;
2027                 }
2028         }
2029
2030         return 0;
2031 }
2032
2033 static int bnxt_rss_hash_update_op(struct rte_eth_dev *eth_dev,
2034                                    struct rte_eth_rss_conf *rss_conf)
2035 {
2036         struct bnxt *bp = eth_dev->data->dev_private;
2037         struct rte_eth_conf *dev_conf = &bp->eth_dev->data->dev_conf;
2038         struct bnxt_vnic_info *vnic;
2039         int rc;
2040
2041         rc = is_bnxt_in_error(bp);
2042         if (rc)
2043                 return rc;
2044
2045         /*
2046          * If RSS enablement were different than dev_configure,
2047          * then return -EINVAL
2048          */
2049         if (dev_conf->rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG) {
2050                 if (!rss_conf->rss_hf)
2051                         PMD_DRV_LOG(ERR, "Hash type NONE\n");
2052         } else {
2053                 if (rss_conf->rss_hf & BNXT_ETH_RSS_SUPPORT)
2054                         return -EINVAL;
2055         }
2056
2057         bp->flags |= BNXT_FLAG_UPDATE_HASH;
2058         memcpy(&eth_dev->data->dev_conf.rx_adv_conf.rss_conf,
2059                rss_conf,
2060                sizeof(*rss_conf));
2061
2062         /* Update the default RSS VNIC(s) */
2063         vnic = BNXT_GET_DEFAULT_VNIC(bp);
2064         vnic->hash_type = bnxt_rte_to_hwrm_hash_types(rss_conf->rss_hf);
2065         vnic->hash_mode =
2066                 bnxt_rte_to_hwrm_hash_level(bp, rss_conf->rss_hf,
2067                                             ETH_RSS_LEVEL(rss_conf->rss_hf));
2068
2069         /*
2070          * If hashkey is not specified, use the previously configured
2071          * hashkey
2072          */
2073         if (!rss_conf->rss_key)
2074                 goto rss_config;
2075
2076         if (rss_conf->rss_key_len != HW_HASH_KEY_SIZE) {
2077                 PMD_DRV_LOG(ERR,
2078                             "Invalid hashkey length, should be 16 bytes\n");
2079                 return -EINVAL;
2080         }
2081         memcpy(vnic->rss_hash_key, rss_conf->rss_key, rss_conf->rss_key_len);
2082
2083 rss_config:
2084         rc = bnxt_hwrm_vnic_rss_cfg(bp, vnic);
2085         return rc;
2086 }
2087
2088 static int bnxt_rss_hash_conf_get_op(struct rte_eth_dev *eth_dev,
2089                                      struct rte_eth_rss_conf *rss_conf)
2090 {
2091         struct bnxt *bp = eth_dev->data->dev_private;
2092         struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
2093         int len, rc;
2094         uint32_t hash_types;
2095
2096         rc = is_bnxt_in_error(bp);
2097         if (rc)
2098                 return rc;
2099
2100         /* RSS configuration is the same for all VNICs */
2101         if (vnic && vnic->rss_hash_key) {
2102                 if (rss_conf->rss_key) {
2103                         len = rss_conf->rss_key_len <= HW_HASH_KEY_SIZE ?
2104                               rss_conf->rss_key_len : HW_HASH_KEY_SIZE;
2105                         memcpy(rss_conf->rss_key, vnic->rss_hash_key, len);
2106                 }
2107
2108                 hash_types = vnic->hash_type;
2109                 rss_conf->rss_hf = 0;
2110                 if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV4) {
2111                         rss_conf->rss_hf |= ETH_RSS_IPV4;
2112                         hash_types &= ~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV4;
2113                 }
2114                 if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV4) {
2115                         rss_conf->rss_hf |= ETH_RSS_NONFRAG_IPV4_TCP;
2116                         hash_types &=
2117                                 ~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV4;
2118                 }
2119                 if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV4) {
2120                         rss_conf->rss_hf |= ETH_RSS_NONFRAG_IPV4_UDP;
2121                         hash_types &=
2122                                 ~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV4;
2123                 }
2124                 if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV6) {
2125                         rss_conf->rss_hf |= ETH_RSS_IPV6;
2126                         hash_types &= ~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV6;
2127                 }
2128                 if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV6) {
2129                         rss_conf->rss_hf |= ETH_RSS_NONFRAG_IPV6_TCP;
2130                         hash_types &=
2131                                 ~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV6;
2132                 }
2133                 if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV6) {
2134                         rss_conf->rss_hf |= ETH_RSS_NONFRAG_IPV6_UDP;
2135                         hash_types &=
2136                                 ~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV6;
2137                 }
2138
2139                 rss_conf->rss_hf |=
2140                         bnxt_hwrm_to_rte_rss_level(bp, vnic->hash_mode);
2141
2142                 if (hash_types) {
2143                         PMD_DRV_LOG(ERR,
2144                                 "Unknown RSS config from firmware (%08x), RSS disabled",
2145                                 vnic->hash_type);
2146                         return -ENOTSUP;
2147                 }
2148         } else {
2149                 rss_conf->rss_hf = 0;
2150         }
2151         return 0;
2152 }
2153
2154 static int bnxt_flow_ctrl_get_op(struct rte_eth_dev *dev,
2155                                struct rte_eth_fc_conf *fc_conf)
2156 {
2157         struct bnxt *bp = dev->data->dev_private;
2158         struct rte_eth_link link_info;
2159         int rc;
2160
2161         rc = is_bnxt_in_error(bp);
2162         if (rc)
2163                 return rc;
2164
2165         rc = bnxt_get_hwrm_link_config(bp, &link_info);
2166         if (rc)
2167                 return rc;
2168
2169         memset(fc_conf, 0, sizeof(*fc_conf));
2170         if (bp->link_info->auto_pause)
2171                 fc_conf->autoneg = 1;
2172         switch (bp->link_info->pause) {
2173         case 0:
2174                 fc_conf->mode = RTE_FC_NONE;
2175                 break;
2176         case HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_TX:
2177                 fc_conf->mode = RTE_FC_TX_PAUSE;
2178                 break;
2179         case HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_RX:
2180                 fc_conf->mode = RTE_FC_RX_PAUSE;
2181                 break;
2182         case (HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_TX |
2183                         HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_RX):
2184                 fc_conf->mode = RTE_FC_FULL;
2185                 break;
2186         }
2187         return 0;
2188 }
2189
2190 static int bnxt_flow_ctrl_set_op(struct rte_eth_dev *dev,
2191                                struct rte_eth_fc_conf *fc_conf)
2192 {
2193         struct bnxt *bp = dev->data->dev_private;
2194         int rc;
2195
2196         rc = is_bnxt_in_error(bp);
2197         if (rc)
2198                 return rc;
2199
2200         if (!BNXT_SINGLE_PF(bp) || BNXT_VF(bp)) {
2201                 PMD_DRV_LOG(ERR, "Flow Control Settings cannot be modified\n");
2202                 return -ENOTSUP;
2203         }
2204
2205         switch (fc_conf->mode) {
2206         case RTE_FC_NONE:
2207                 bp->link_info->auto_pause = 0;
2208                 bp->link_info->force_pause = 0;
2209                 break;
2210         case RTE_FC_RX_PAUSE:
2211                 if (fc_conf->autoneg) {
2212                         bp->link_info->auto_pause =
2213                                         HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_RX;
2214                         bp->link_info->force_pause = 0;
2215                 } else {
2216                         bp->link_info->auto_pause = 0;
2217                         bp->link_info->force_pause =
2218                                         HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_RX;
2219                 }
2220                 break;
2221         case RTE_FC_TX_PAUSE:
2222                 if (fc_conf->autoneg) {
2223                         bp->link_info->auto_pause =
2224                                         HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_TX;
2225                         bp->link_info->force_pause = 0;
2226                 } else {
2227                         bp->link_info->auto_pause = 0;
2228                         bp->link_info->force_pause =
2229                                         HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_TX;
2230                 }
2231                 break;
2232         case RTE_FC_FULL:
2233                 if (fc_conf->autoneg) {
2234                         bp->link_info->auto_pause =
2235                                         HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_TX |
2236                                         HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_RX;
2237                         bp->link_info->force_pause = 0;
2238                 } else {
2239                         bp->link_info->auto_pause = 0;
2240                         bp->link_info->force_pause =
2241                                         HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_TX |
2242                                         HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_RX;
2243                 }
2244                 break;
2245         }
2246         return bnxt_set_hwrm_link_config(bp, true);
2247 }
2248
2249 /* Add UDP tunneling port */
2250 static int
2251 bnxt_udp_tunnel_port_add_op(struct rte_eth_dev *eth_dev,
2252                          struct rte_eth_udp_tunnel *udp_tunnel)
2253 {
2254         struct bnxt *bp = eth_dev->data->dev_private;
2255         uint16_t tunnel_type = 0;
2256         int rc = 0;
2257
2258         rc = is_bnxt_in_error(bp);
2259         if (rc)
2260                 return rc;
2261
2262         switch (udp_tunnel->prot_type) {
2263         case RTE_TUNNEL_TYPE_VXLAN:
2264                 if (bp->vxlan_port_cnt) {
2265                         PMD_DRV_LOG(ERR, "Tunnel Port %d already programmed\n",
2266                                 udp_tunnel->udp_port);
2267                         if (bp->vxlan_port != udp_tunnel->udp_port) {
2268                                 PMD_DRV_LOG(ERR, "Only one port allowed\n");
2269                                 return -ENOSPC;
2270                         }
2271                         bp->vxlan_port_cnt++;
2272                         return 0;
2273                 }
2274                 tunnel_type =
2275                         HWRM_TUNNEL_DST_PORT_ALLOC_INPUT_TUNNEL_TYPE_VXLAN;
2276                 bp->vxlan_port_cnt++;
2277                 break;
2278         case RTE_TUNNEL_TYPE_GENEVE:
2279                 if (bp->geneve_port_cnt) {
2280                         PMD_DRV_LOG(ERR, "Tunnel Port %d already programmed\n",
2281                                 udp_tunnel->udp_port);
2282                         if (bp->geneve_port != udp_tunnel->udp_port) {
2283                                 PMD_DRV_LOG(ERR, "Only one port allowed\n");
2284                                 return -ENOSPC;
2285                         }
2286                         bp->geneve_port_cnt++;
2287                         return 0;
2288                 }
2289                 tunnel_type =
2290                         HWRM_TUNNEL_DST_PORT_ALLOC_INPUT_TUNNEL_TYPE_GENEVE;
2291                 bp->geneve_port_cnt++;
2292                 break;
2293         default:
2294                 PMD_DRV_LOG(ERR, "Tunnel type is not supported\n");
2295                 return -ENOTSUP;
2296         }
2297         rc = bnxt_hwrm_tunnel_dst_port_alloc(bp, udp_tunnel->udp_port,
2298                                              tunnel_type);
2299         return rc;
2300 }
2301
2302 static int
2303 bnxt_udp_tunnel_port_del_op(struct rte_eth_dev *eth_dev,
2304                          struct rte_eth_udp_tunnel *udp_tunnel)
2305 {
2306         struct bnxt *bp = eth_dev->data->dev_private;
2307         uint16_t tunnel_type = 0;
2308         uint16_t port = 0;
2309         int rc = 0;
2310
2311         rc = is_bnxt_in_error(bp);
2312         if (rc)
2313                 return rc;
2314
2315         switch (udp_tunnel->prot_type) {
2316         case RTE_TUNNEL_TYPE_VXLAN:
2317                 if (!bp->vxlan_port_cnt) {
2318                         PMD_DRV_LOG(ERR, "No Tunnel port configured yet\n");
2319                         return -EINVAL;
2320                 }
2321                 if (bp->vxlan_port != udp_tunnel->udp_port) {
2322                         PMD_DRV_LOG(ERR, "Req Port: %d. Configured port: %d\n",
2323                                 udp_tunnel->udp_port, bp->vxlan_port);
2324                         return -EINVAL;
2325                 }
2326                 if (--bp->vxlan_port_cnt)
2327                         return 0;
2328
2329                 tunnel_type =
2330                         HWRM_TUNNEL_DST_PORT_FREE_INPUT_TUNNEL_TYPE_VXLAN;
2331                 port = bp->vxlan_fw_dst_port_id;
2332                 break;
2333         case RTE_TUNNEL_TYPE_GENEVE:
2334                 if (!bp->geneve_port_cnt) {
2335                         PMD_DRV_LOG(ERR, "No Tunnel port configured yet\n");
2336                         return -EINVAL;
2337                 }
2338                 if (bp->geneve_port != udp_tunnel->udp_port) {
2339                         PMD_DRV_LOG(ERR, "Req Port: %d. Configured port: %d\n",
2340                                 udp_tunnel->udp_port, bp->geneve_port);
2341                         return -EINVAL;
2342                 }
2343                 if (--bp->geneve_port_cnt)
2344                         return 0;
2345
2346                 tunnel_type =
2347                         HWRM_TUNNEL_DST_PORT_FREE_INPUT_TUNNEL_TYPE_GENEVE;
2348                 port = bp->geneve_fw_dst_port_id;
2349                 break;
2350         default:
2351                 PMD_DRV_LOG(ERR, "Tunnel type is not supported\n");
2352                 return -ENOTSUP;
2353         }
2354
2355         rc = bnxt_hwrm_tunnel_dst_port_free(bp, port, tunnel_type);
2356         return rc;
2357 }
2358
2359 static int bnxt_del_vlan_filter(struct bnxt *bp, uint16_t vlan_id)
2360 {
2361         struct bnxt_filter_info *filter;
2362         struct bnxt_vnic_info *vnic;
2363         int rc = 0;
2364         uint32_t chk = HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN;
2365
2366         vnic = BNXT_GET_DEFAULT_VNIC(bp);
2367         filter = STAILQ_FIRST(&vnic->filter);
2368         while (filter) {
2369                 /* Search for this matching MAC+VLAN filter */
2370                 if (bnxt_vlan_filter_exists(bp, filter, chk, vlan_id)) {
2371                         /* Delete the filter */
2372                         rc = bnxt_hwrm_clear_l2_filter(bp, filter);
2373                         if (rc)
2374                                 return rc;
2375                         STAILQ_REMOVE(&vnic->filter, filter,
2376                                       bnxt_filter_info, next);
2377                         bnxt_free_filter(bp, filter);
2378                         PMD_DRV_LOG(INFO,
2379                                     "Deleted vlan filter for %d\n",
2380                                     vlan_id);
2381                         return 0;
2382                 }
2383                 filter = STAILQ_NEXT(filter, next);
2384         }
2385         return -ENOENT;
2386 }
2387
2388 static int bnxt_add_vlan_filter(struct bnxt *bp, uint16_t vlan_id)
2389 {
2390         struct bnxt_filter_info *filter;
2391         struct bnxt_vnic_info *vnic;
2392         int rc = 0;
2393         uint32_t en = HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN |
2394                 HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN_MASK;
2395         uint32_t chk = HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN;
2396
2397         /* Implementation notes on the use of VNIC in this command:
2398          *
2399          * By default, these filters belong to default vnic for the function.
2400          * Once these filters are set up, only destination VNIC can be modified.
2401          * If the destination VNIC is not specified in this command,
2402          * then the HWRM shall only create an l2 context id.
2403          */
2404
2405         vnic = BNXT_GET_DEFAULT_VNIC(bp);
2406         filter = STAILQ_FIRST(&vnic->filter);
2407         /* Check if the VLAN has already been added */
2408         while (filter) {
2409                 if (bnxt_vlan_filter_exists(bp, filter, chk, vlan_id))
2410                         return -EEXIST;
2411
2412                 filter = STAILQ_NEXT(filter, next);
2413         }
2414
2415         /* No match found. Alloc a fresh filter and issue the L2_FILTER_ALLOC
2416          * command to create MAC+VLAN filter with the right flags, enables set.
2417          */
2418         filter = bnxt_alloc_filter(bp);
2419         if (!filter) {
2420                 PMD_DRV_LOG(ERR,
2421                             "MAC/VLAN filter alloc failed\n");
2422                 return -ENOMEM;
2423         }
2424         /* MAC + VLAN ID filter */
2425         /* If l2_ivlan == 0 and l2_ivlan_mask != 0, only
2426          * untagged packets are received
2427          *
2428          * If l2_ivlan != 0 and l2_ivlan_mask != 0, untagged
2429          * packets and only the programmed vlan's packets are received
2430          */
2431         filter->l2_ivlan = vlan_id;
2432         filter->l2_ivlan_mask = 0x0FFF;
2433         filter->enables |= en;
2434         filter->flags |= HWRM_CFA_L2_FILTER_ALLOC_INPUT_FLAGS_OUTERMOST;
2435
2436         rc = bnxt_hwrm_set_l2_filter(bp, vnic->fw_vnic_id, filter);
2437         if (rc) {
2438                 /* Free the newly allocated filter as we were
2439                  * not able to create the filter in hardware.
2440                  */
2441                 bnxt_free_filter(bp, filter);
2442                 return rc;
2443         }
2444
2445         filter->mac_index = 0;
2446         /* Add this new filter to the list */
2447         if (vlan_id == 0)
2448                 STAILQ_INSERT_HEAD(&vnic->filter, filter, next);
2449         else
2450                 STAILQ_INSERT_TAIL(&vnic->filter, filter, next);
2451
2452         PMD_DRV_LOG(INFO,
2453                     "Added Vlan filter for %d\n", vlan_id);
2454         return rc;
2455 }
2456
2457 static int bnxt_vlan_filter_set_op(struct rte_eth_dev *eth_dev,
2458                 uint16_t vlan_id, int on)
2459 {
2460         struct bnxt *bp = eth_dev->data->dev_private;
2461         int rc;
2462
2463         rc = is_bnxt_in_error(bp);
2464         if (rc)
2465                 return rc;
2466
2467         if (!eth_dev->data->dev_started) {
2468                 PMD_DRV_LOG(ERR, "port must be started before setting vlan\n");
2469                 return -EINVAL;
2470         }
2471
2472         /* These operations apply to ALL existing MAC/VLAN filters */
2473         if (on)
2474                 return bnxt_add_vlan_filter(bp, vlan_id);
2475         else
2476                 return bnxt_del_vlan_filter(bp, vlan_id);
2477 }
2478
2479 static int bnxt_del_dflt_mac_filter(struct bnxt *bp,
2480                                     struct bnxt_vnic_info *vnic)
2481 {
2482         struct bnxt_filter_info *filter;
2483         int rc;
2484
2485         filter = STAILQ_FIRST(&vnic->filter);
2486         while (filter) {
2487                 if (filter->mac_index == 0 &&
2488                     !memcmp(filter->l2_addr, bp->mac_addr,
2489                             RTE_ETHER_ADDR_LEN)) {
2490                         rc = bnxt_hwrm_clear_l2_filter(bp, filter);
2491                         if (!rc) {
2492                                 STAILQ_REMOVE(&vnic->filter, filter,
2493                                               bnxt_filter_info, next);
2494                                 bnxt_free_filter(bp, filter);
2495                         }
2496                         return rc;
2497                 }
2498                 filter = STAILQ_NEXT(filter, next);
2499         }
2500         return 0;
2501 }
2502
2503 static int
2504 bnxt_config_vlan_hw_filter(struct bnxt *bp, uint64_t rx_offloads)
2505 {
2506         struct bnxt_vnic_info *vnic;
2507         unsigned int i;
2508         int rc;
2509
2510         vnic = BNXT_GET_DEFAULT_VNIC(bp);
2511         if (!(rx_offloads & DEV_RX_OFFLOAD_VLAN_FILTER)) {
2512                 /* Remove any VLAN filters programmed */
2513                 for (i = 0; i < RTE_ETHER_MAX_VLAN_ID; i++)
2514                         bnxt_del_vlan_filter(bp, i);
2515
2516                 rc = bnxt_add_mac_filter(bp, vnic, NULL, 0, 0);
2517                 if (rc)
2518                         return rc;
2519         } else {
2520                 /* Default filter will allow packets that match the
2521                  * dest mac. So, it has to be deleted, otherwise, we
2522                  * will endup receiving vlan packets for which the
2523                  * filter is not programmed, when hw-vlan-filter
2524                  * configuration is ON
2525                  */
2526                 bnxt_del_dflt_mac_filter(bp, vnic);
2527                 /* This filter will allow only untagged packets */
2528                 bnxt_add_vlan_filter(bp, 0);
2529         }
2530         PMD_DRV_LOG(DEBUG, "VLAN Filtering: %d\n",
2531                     !!(rx_offloads & DEV_RX_OFFLOAD_VLAN_FILTER));
2532
2533         return 0;
2534 }
2535
2536 static int bnxt_free_one_vnic(struct bnxt *bp, uint16_t vnic_id)
2537 {
2538         struct bnxt_vnic_info *vnic = &bp->vnic_info[vnic_id];
2539         unsigned int i;
2540         int rc;
2541
2542         /* Destroy vnic filters and vnic */
2543         if (bp->eth_dev->data->dev_conf.rxmode.offloads &
2544             DEV_RX_OFFLOAD_VLAN_FILTER) {
2545                 for (i = 0; i < RTE_ETHER_MAX_VLAN_ID; i++)
2546                         bnxt_del_vlan_filter(bp, i);
2547         }
2548         bnxt_del_dflt_mac_filter(bp, vnic);
2549
2550         rc = bnxt_hwrm_vnic_ctx_free(bp, vnic);
2551         if (rc)
2552                 return rc;
2553
2554         rc = bnxt_hwrm_vnic_free(bp, vnic);
2555         if (rc)
2556                 return rc;
2557
2558         rte_free(vnic->fw_grp_ids);
2559         vnic->fw_grp_ids = NULL;
2560
2561         vnic->rx_queue_cnt = 0;
2562
2563         return 0;
2564 }
2565
2566 static int
2567 bnxt_config_vlan_hw_stripping(struct bnxt *bp, uint64_t rx_offloads)
2568 {
2569         struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
2570         int rc;
2571
2572         /* Destroy, recreate and reconfigure the default vnic */
2573         rc = bnxt_free_one_vnic(bp, 0);
2574         if (rc)
2575                 return rc;
2576
2577         /* default vnic 0 */
2578         rc = bnxt_setup_one_vnic(bp, 0);
2579         if (rc)
2580                 return rc;
2581
2582         if (bp->eth_dev->data->dev_conf.rxmode.offloads &
2583             DEV_RX_OFFLOAD_VLAN_FILTER) {
2584                 rc = bnxt_add_vlan_filter(bp, 0);
2585                 if (rc)
2586                         return rc;
2587                 rc = bnxt_restore_vlan_filters(bp);
2588                 if (rc)
2589                         return rc;
2590         } else {
2591                 rc = bnxt_add_mac_filter(bp, vnic, NULL, 0, 0);
2592                 if (rc)
2593                         return rc;
2594         }
2595
2596         rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
2597         if (rc)
2598                 return rc;
2599
2600         PMD_DRV_LOG(DEBUG, "VLAN Strip Offload: %d\n",
2601                     !!(rx_offloads & DEV_RX_OFFLOAD_VLAN_STRIP));
2602
2603         return rc;
2604 }
2605
2606 static int
2607 bnxt_vlan_offload_set_op(struct rte_eth_dev *dev, int mask)
2608 {
2609         uint64_t rx_offloads = dev->data->dev_conf.rxmode.offloads;
2610         struct bnxt *bp = dev->data->dev_private;
2611         int rc;
2612
2613         rc = is_bnxt_in_error(bp);
2614         if (rc)
2615                 return rc;
2616
2617         /* Filter settings will get applied when port is started */
2618         if (!dev->data->dev_started)
2619                 return 0;
2620
2621         if (mask & ETH_VLAN_FILTER_MASK) {
2622                 /* Enable or disable VLAN filtering */
2623                 rc = bnxt_config_vlan_hw_filter(bp, rx_offloads);
2624                 if (rc)
2625                         return rc;
2626         }
2627
2628         if (mask & ETH_VLAN_STRIP_MASK) {
2629                 /* Enable or disable VLAN stripping */
2630                 rc = bnxt_config_vlan_hw_stripping(bp, rx_offloads);
2631                 if (rc)
2632                         return rc;
2633         }
2634
2635         if (mask & ETH_VLAN_EXTEND_MASK) {
2636                 if (rx_offloads & DEV_RX_OFFLOAD_VLAN_EXTEND)
2637                         PMD_DRV_LOG(DEBUG, "Extend VLAN supported\n");
2638                 else
2639                         PMD_DRV_LOG(INFO, "Extend VLAN unsupported\n");
2640         }
2641
2642         return 0;
2643 }
2644
2645 static int
2646 bnxt_vlan_tpid_set_op(struct rte_eth_dev *dev, enum rte_vlan_type vlan_type,
2647                       uint16_t tpid)
2648 {
2649         struct bnxt *bp = dev->data->dev_private;
2650         int qinq = dev->data->dev_conf.rxmode.offloads &
2651                    DEV_RX_OFFLOAD_VLAN_EXTEND;
2652
2653         if (vlan_type != ETH_VLAN_TYPE_INNER &&
2654             vlan_type != ETH_VLAN_TYPE_OUTER) {
2655                 PMD_DRV_LOG(ERR,
2656                             "Unsupported vlan type.");
2657                 return -EINVAL;
2658         }
2659         if (!qinq) {
2660                 PMD_DRV_LOG(ERR,
2661                             "QinQ not enabled. Needs to be ON as we can "
2662                             "accelerate only outer vlan\n");
2663                 return -EINVAL;
2664         }
2665
2666         if (vlan_type == ETH_VLAN_TYPE_OUTER) {
2667                 switch (tpid) {
2668                 case RTE_ETHER_TYPE_QINQ:
2669                         bp->outer_tpid_bd =
2670                                 TX_BD_LONG_CFA_META_VLAN_TPID_TPID88A8;
2671                                 break;
2672                 case RTE_ETHER_TYPE_VLAN:
2673                         bp->outer_tpid_bd =
2674                                 TX_BD_LONG_CFA_META_VLAN_TPID_TPID8100;
2675                                 break;
2676                 case RTE_ETHER_TYPE_QINQ1:
2677                         bp->outer_tpid_bd =
2678                                 TX_BD_LONG_CFA_META_VLAN_TPID_TPID9100;
2679                                 break;
2680                 case RTE_ETHER_TYPE_QINQ2:
2681                         bp->outer_tpid_bd =
2682                                 TX_BD_LONG_CFA_META_VLAN_TPID_TPID9200;
2683                                 break;
2684                 case RTE_ETHER_TYPE_QINQ3:
2685                         bp->outer_tpid_bd =
2686                                  TX_BD_LONG_CFA_META_VLAN_TPID_TPID9300;
2687                                 break;
2688                 default:
2689                         PMD_DRV_LOG(ERR, "Invalid TPID: %x\n", tpid);
2690                         return -EINVAL;
2691                 }
2692                 bp->outer_tpid_bd |= tpid;
2693                 PMD_DRV_LOG(INFO, "outer_tpid_bd = %x\n", bp->outer_tpid_bd);
2694         } else if (vlan_type == ETH_VLAN_TYPE_INNER) {
2695                 PMD_DRV_LOG(ERR,
2696                             "Can accelerate only outer vlan in QinQ\n");
2697                 return -EINVAL;
2698         }
2699
2700         return 0;
2701 }
2702
2703 static int
2704 bnxt_set_default_mac_addr_op(struct rte_eth_dev *dev,
2705                              struct rte_ether_addr *addr)
2706 {
2707         struct bnxt *bp = dev->data->dev_private;
2708         /* Default Filter is tied to VNIC 0 */
2709         struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
2710         int rc;
2711
2712         rc = is_bnxt_in_error(bp);
2713         if (rc)
2714                 return rc;
2715
2716         if (BNXT_VF(bp) && !BNXT_VF_IS_TRUSTED(bp))
2717                 return -EPERM;
2718
2719         if (rte_is_zero_ether_addr(addr))
2720                 return -EINVAL;
2721
2722         /* Filter settings will get applied when port is started */
2723         if (!dev->data->dev_started)
2724                 return 0;
2725
2726         /* Check if the requested MAC is already added */
2727         if (memcmp(addr, bp->mac_addr, RTE_ETHER_ADDR_LEN) == 0)
2728                 return 0;
2729
2730         /* Destroy filter and re-create it */
2731         bnxt_del_dflt_mac_filter(bp, vnic);
2732
2733         memcpy(bp->mac_addr, addr, RTE_ETHER_ADDR_LEN);
2734         if (dev->data->dev_conf.rxmode.offloads & DEV_RX_OFFLOAD_VLAN_FILTER) {
2735                 /* This filter will allow only untagged packets */
2736                 rc = bnxt_add_vlan_filter(bp, 0);
2737         } else {
2738                 rc = bnxt_add_mac_filter(bp, vnic, addr, 0, 0);
2739         }
2740
2741         PMD_DRV_LOG(DEBUG, "Set MAC addr\n");
2742         return rc;
2743 }
2744
2745 static int
2746 bnxt_dev_set_mc_addr_list_op(struct rte_eth_dev *eth_dev,
2747                           struct rte_ether_addr *mc_addr_set,
2748                           uint32_t nb_mc_addr)
2749 {
2750         struct bnxt *bp = eth_dev->data->dev_private;
2751         char *mc_addr_list = (char *)mc_addr_set;
2752         struct bnxt_vnic_info *vnic;
2753         uint32_t off = 0, i = 0;
2754         int rc;
2755
2756         rc = is_bnxt_in_error(bp);
2757         if (rc)
2758                 return rc;
2759
2760         vnic = BNXT_GET_DEFAULT_VNIC(bp);
2761
2762         if (nb_mc_addr > BNXT_MAX_MC_ADDRS) {
2763                 vnic->flags |= BNXT_VNIC_INFO_ALLMULTI;
2764                 goto allmulti;
2765         }
2766
2767         /* TODO Check for Duplicate mcast addresses */
2768         vnic->flags &= ~BNXT_VNIC_INFO_ALLMULTI;
2769         for (i = 0; i < nb_mc_addr; i++) {
2770                 memcpy(vnic->mc_list + off, &mc_addr_list[i],
2771                         RTE_ETHER_ADDR_LEN);
2772                 off += RTE_ETHER_ADDR_LEN;
2773         }
2774
2775         vnic->mc_addr_cnt = i;
2776         if (vnic->mc_addr_cnt)
2777                 vnic->flags |= BNXT_VNIC_INFO_MCAST;
2778         else
2779                 vnic->flags &= ~BNXT_VNIC_INFO_MCAST;
2780
2781 allmulti:
2782         return bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
2783 }
2784
2785 static int
2786 bnxt_fw_version_get(struct rte_eth_dev *dev, char *fw_version, size_t fw_size)
2787 {
2788         struct bnxt *bp = dev->data->dev_private;
2789         uint8_t fw_major = (bp->fw_ver >> 24) & 0xff;
2790         uint8_t fw_minor = (bp->fw_ver >> 16) & 0xff;
2791         uint8_t fw_updt = (bp->fw_ver >> 8) & 0xff;
2792         uint8_t fw_rsvd = bp->fw_ver & 0xff;
2793         int ret;
2794
2795         ret = snprintf(fw_version, fw_size, "%d.%d.%d.%d",
2796                         fw_major, fw_minor, fw_updt, fw_rsvd);
2797
2798         ret += 1; /* add the size of '\0' */
2799         if (fw_size < (uint32_t)ret)
2800                 return ret;
2801         else
2802                 return 0;
2803 }
2804
2805 static void
2806 bnxt_rxq_info_get_op(struct rte_eth_dev *dev, uint16_t queue_id,
2807         struct rte_eth_rxq_info *qinfo)
2808 {
2809         struct bnxt *bp = dev->data->dev_private;
2810         struct bnxt_rx_queue *rxq;
2811
2812         if (is_bnxt_in_error(bp))
2813                 return;
2814
2815         rxq = dev->data->rx_queues[queue_id];
2816
2817         qinfo->mp = rxq->mb_pool;
2818         qinfo->scattered_rx = dev->data->scattered_rx;
2819         qinfo->nb_desc = rxq->nb_rx_desc;
2820
2821         qinfo->conf.rx_free_thresh = rxq->rx_free_thresh;
2822         qinfo->conf.rx_drop_en = rxq->drop_en;
2823         qinfo->conf.rx_deferred_start = rxq->rx_deferred_start;
2824         qinfo->conf.offloads = dev->data->dev_conf.rxmode.offloads;
2825 }
2826
2827 static void
2828 bnxt_txq_info_get_op(struct rte_eth_dev *dev, uint16_t queue_id,
2829         struct rte_eth_txq_info *qinfo)
2830 {
2831         struct bnxt *bp = dev->data->dev_private;
2832         struct bnxt_tx_queue *txq;
2833
2834         if (is_bnxt_in_error(bp))
2835                 return;
2836
2837         txq = dev->data->tx_queues[queue_id];
2838
2839         qinfo->nb_desc = txq->nb_tx_desc;
2840
2841         qinfo->conf.tx_thresh.pthresh = txq->pthresh;
2842         qinfo->conf.tx_thresh.hthresh = txq->hthresh;
2843         qinfo->conf.tx_thresh.wthresh = txq->wthresh;
2844
2845         qinfo->conf.tx_free_thresh = txq->tx_free_thresh;
2846         qinfo->conf.tx_rs_thresh = 0;
2847         qinfo->conf.tx_deferred_start = txq->tx_deferred_start;
2848         qinfo->conf.offloads = txq->offloads;
2849 }
2850
2851 static const struct {
2852         eth_rx_burst_t pkt_burst;
2853         const char *info;
2854 } bnxt_rx_burst_info[] = {
2855         {bnxt_recv_pkts,        "Scalar"},
2856 #if defined(RTE_ARCH_X86)
2857         {bnxt_recv_pkts_vec,    "Vector SSE"},
2858 #elif defined(RTE_ARCH_ARM64)
2859         {bnxt_recv_pkts_vec,    "Vector Neon"},
2860 #endif
2861 };
2862
2863 static int
2864 bnxt_rx_burst_mode_get(struct rte_eth_dev *dev, __rte_unused uint16_t queue_id,
2865                        struct rte_eth_burst_mode *mode)
2866 {
2867         eth_rx_burst_t pkt_burst = dev->rx_pkt_burst;
2868         size_t i;
2869
2870         for (i = 0; i < RTE_DIM(bnxt_rx_burst_info); i++) {
2871                 if (pkt_burst == bnxt_rx_burst_info[i].pkt_burst) {
2872                         snprintf(mode->info, sizeof(mode->info), "%s",
2873                                  bnxt_rx_burst_info[i].info);
2874                         return 0;
2875                 }
2876         }
2877
2878         return -EINVAL;
2879 }
2880
2881 static const struct {
2882         eth_tx_burst_t pkt_burst;
2883         const char *info;
2884 } bnxt_tx_burst_info[] = {
2885         {bnxt_xmit_pkts,        "Scalar"},
2886 #if defined(RTE_ARCH_X86)
2887         {bnxt_xmit_pkts_vec,    "Vector SSE"},
2888 #elif defined(RTE_ARCH_ARM64)
2889         {bnxt_xmit_pkts_vec,    "Vector Neon"},
2890 #endif
2891 };
2892
2893 static int
2894 bnxt_tx_burst_mode_get(struct rte_eth_dev *dev, __rte_unused uint16_t queue_id,
2895                        struct rte_eth_burst_mode *mode)
2896 {
2897         eth_tx_burst_t pkt_burst = dev->tx_pkt_burst;
2898         size_t i;
2899
2900         for (i = 0; i < RTE_DIM(bnxt_tx_burst_info); i++) {
2901                 if (pkt_burst == bnxt_tx_burst_info[i].pkt_burst) {
2902                         snprintf(mode->info, sizeof(mode->info), "%s",
2903                                  bnxt_tx_burst_info[i].info);
2904                         return 0;
2905                 }
2906         }
2907
2908         return -EINVAL;
2909 }
2910
2911 int bnxt_mtu_set_op(struct rte_eth_dev *eth_dev, uint16_t new_mtu)
2912 {
2913         struct bnxt *bp = eth_dev->data->dev_private;
2914         uint32_t new_pkt_size;
2915         uint32_t rc = 0;
2916         uint32_t i;
2917
2918         rc = is_bnxt_in_error(bp);
2919         if (rc)
2920                 return rc;
2921
2922         /* Exit if receive queues are not configured yet */
2923         if (!eth_dev->data->nb_rx_queues)
2924                 return rc;
2925
2926         new_pkt_size = new_mtu + RTE_ETHER_HDR_LEN + RTE_ETHER_CRC_LEN +
2927                        VLAN_TAG_SIZE * BNXT_NUM_VLANS;
2928
2929         /*
2930          * Disallow any MTU change that would require scattered receive support
2931          * if it is not already enabled.
2932          */
2933         if (eth_dev->data->dev_started &&
2934             !eth_dev->data->scattered_rx &&
2935             (new_pkt_size >
2936              eth_dev->data->min_rx_buf_size - RTE_PKTMBUF_HEADROOM)) {
2937                 PMD_DRV_LOG(ERR,
2938                             "MTU change would require scattered rx support. ");
2939                 PMD_DRV_LOG(ERR, "Stop port before changing MTU.\n");
2940                 return -EINVAL;
2941         }
2942
2943         if (new_mtu > RTE_ETHER_MTU) {
2944                 bp->flags |= BNXT_FLAG_JUMBO;
2945                 bp->eth_dev->data->dev_conf.rxmode.offloads |=
2946                         DEV_RX_OFFLOAD_JUMBO_FRAME;
2947         } else {
2948                 bp->eth_dev->data->dev_conf.rxmode.offloads &=
2949                         ~DEV_RX_OFFLOAD_JUMBO_FRAME;
2950                 bp->flags &= ~BNXT_FLAG_JUMBO;
2951         }
2952
2953         /* Is there a change in mtu setting? */
2954         if (eth_dev->data->dev_conf.rxmode.max_rx_pkt_len == new_pkt_size)
2955                 return rc;
2956
2957         for (i = 0; i < bp->nr_vnics; i++) {
2958                 struct bnxt_vnic_info *vnic = &bp->vnic_info[i];
2959                 uint16_t size = 0;
2960
2961                 vnic->mru = BNXT_VNIC_MRU(new_mtu);
2962                 rc = bnxt_hwrm_vnic_cfg(bp, vnic);
2963                 if (rc)
2964                         break;
2965
2966                 size = rte_pktmbuf_data_room_size(bp->rx_queues[0]->mb_pool);
2967                 size -= RTE_PKTMBUF_HEADROOM;
2968
2969                 if (size < new_mtu) {
2970                         rc = bnxt_hwrm_vnic_plcmode_cfg(bp, vnic);
2971                         if (rc)
2972                                 return rc;
2973                 }
2974         }
2975
2976         if (!rc)
2977                 eth_dev->data->dev_conf.rxmode.max_rx_pkt_len = new_pkt_size;
2978
2979         PMD_DRV_LOG(INFO, "New MTU is %d\n", new_mtu);
2980
2981         return rc;
2982 }
2983
2984 static int
2985 bnxt_vlan_pvid_set_op(struct rte_eth_dev *dev, uint16_t pvid, int on)
2986 {
2987         struct bnxt *bp = dev->data->dev_private;
2988         uint16_t vlan = bp->vlan;
2989         int rc;
2990
2991         rc = is_bnxt_in_error(bp);
2992         if (rc)
2993                 return rc;
2994
2995         if (!BNXT_SINGLE_PF(bp) || BNXT_VF(bp)) {
2996                 PMD_DRV_LOG(ERR,
2997                         "PVID cannot be modified for this function\n");
2998                 return -ENOTSUP;
2999         }
3000         bp->vlan = on ? pvid : 0;
3001
3002         rc = bnxt_hwrm_set_default_vlan(bp, 0, 0);
3003         if (rc)
3004                 bp->vlan = vlan;
3005         return rc;
3006 }
3007
3008 static int
3009 bnxt_dev_led_on_op(struct rte_eth_dev *dev)
3010 {
3011         struct bnxt *bp = dev->data->dev_private;
3012         int rc;
3013
3014         rc = is_bnxt_in_error(bp);
3015         if (rc)
3016                 return rc;
3017
3018         return bnxt_hwrm_port_led_cfg(bp, true);
3019 }
3020
3021 static int
3022 bnxt_dev_led_off_op(struct rte_eth_dev *dev)
3023 {
3024         struct bnxt *bp = dev->data->dev_private;
3025         int rc;
3026
3027         rc = is_bnxt_in_error(bp);
3028         if (rc)
3029                 return rc;
3030
3031         return bnxt_hwrm_port_led_cfg(bp, false);
3032 }
3033
3034 static uint32_t
3035 bnxt_rx_queue_count_op(struct rte_eth_dev *dev, uint16_t rx_queue_id)
3036 {
3037         struct bnxt *bp = (struct bnxt *)dev->data->dev_private;
3038         struct bnxt_cp_ring_info *cpr;
3039         uint32_t desc = 0, raw_cons;
3040         struct bnxt_rx_queue *rxq;
3041         struct rx_pkt_cmpl *rxcmp;
3042         int rc;
3043
3044         rc = is_bnxt_in_error(bp);
3045         if (rc)
3046                 return rc;
3047
3048         rxq = dev->data->rx_queues[rx_queue_id];
3049         cpr = rxq->cp_ring;
3050         raw_cons = cpr->cp_raw_cons;
3051
3052         while (1) {
3053                 uint32_t agg_cnt, cons, cmpl_type;
3054
3055                 cons = RING_CMP(cpr->cp_ring_struct, raw_cons);
3056                 rxcmp = (struct rx_pkt_cmpl *)&cpr->cp_desc_ring[cons];
3057
3058                 if (!CMP_VALID(rxcmp, raw_cons, cpr->cp_ring_struct))
3059                         break;
3060
3061                 cmpl_type = CMP_TYPE(rxcmp);
3062
3063                 switch (cmpl_type) {
3064                 case CMPL_BASE_TYPE_RX_L2:
3065                 case CMPL_BASE_TYPE_RX_L2_V2:
3066                         agg_cnt = BNXT_RX_L2_AGG_BUFS(rxcmp);
3067                         raw_cons = raw_cons + CMP_LEN(cmpl_type) + agg_cnt;
3068                         desc++;
3069                         break;
3070
3071                 case CMPL_BASE_TYPE_RX_TPA_END:
3072                         if (BNXT_CHIP_P5(rxq->bp)) {
3073                                 struct rx_tpa_v2_end_cmpl_hi *p5_tpa_end;
3074
3075                                 p5_tpa_end = (void *)rxcmp;
3076                                 agg_cnt = BNXT_TPA_END_AGG_BUFS_TH(p5_tpa_end);
3077                         } else {
3078                                 struct rx_tpa_end_cmpl *tpa_end;
3079
3080                                 tpa_end = (void *)rxcmp;
3081                                 agg_cnt = BNXT_TPA_END_AGG_BUFS(tpa_end);
3082                         }
3083
3084                         raw_cons = raw_cons + CMP_LEN(cmpl_type) + agg_cnt;
3085                         desc++;
3086                         break;
3087
3088                 default:
3089                         raw_cons += CMP_LEN(cmpl_type);
3090                 }
3091         }
3092
3093         return desc;
3094 }
3095
3096 static int
3097 bnxt_rx_descriptor_status_op(void *rx_queue, uint16_t offset)
3098 {
3099         struct bnxt_rx_queue *rxq = rx_queue;
3100         struct bnxt_cp_ring_info *cpr;
3101         struct bnxt_rx_ring_info *rxr;
3102         uint32_t desc, raw_cons;
3103         struct bnxt *bp = rxq->bp;
3104         struct rx_pkt_cmpl *rxcmp;
3105         int rc;
3106
3107         rc = is_bnxt_in_error(bp);
3108         if (rc)
3109                 return rc;
3110
3111         if (offset >= rxq->nb_rx_desc)
3112                 return -EINVAL;
3113
3114         rxr = rxq->rx_ring;
3115         cpr = rxq->cp_ring;
3116
3117         /*
3118          * For the vector receive case, the completion at the requested
3119          * offset can be indexed directly.
3120          */
3121 #if defined(RTE_ARCH_X86) || defined(RTE_ARCH_ARM64)
3122         if (bp->flags & BNXT_FLAG_RX_VECTOR_PKT_MODE) {
3123                 struct rx_pkt_cmpl *rxcmp;
3124                 uint32_t cons;
3125
3126                 /* Check status of completion descriptor. */
3127                 raw_cons = cpr->cp_raw_cons +
3128                            offset * CMP_LEN(CMPL_BASE_TYPE_RX_L2);
3129                 cons = RING_CMP(cpr->cp_ring_struct, raw_cons);
3130                 rxcmp = (struct rx_pkt_cmpl *)&cpr->cp_desc_ring[cons];
3131
3132                 if (CMP_VALID(rxcmp, raw_cons, cpr->cp_ring_struct))
3133                         return RTE_ETH_RX_DESC_DONE;
3134
3135                 /* Check whether rx desc has an mbuf attached. */
3136                 cons = RING_CMP(rxr->rx_ring_struct, raw_cons / 2);
3137                 if (cons >= rxq->rxrearm_start &&
3138                     cons < rxq->rxrearm_start + rxq->rxrearm_nb) {
3139                         return RTE_ETH_RX_DESC_UNAVAIL;
3140                 }
3141
3142                 return RTE_ETH_RX_DESC_AVAIL;
3143         }
3144 #endif
3145
3146         /*
3147          * For the non-vector receive case, scan the completion ring to
3148          * locate the completion descriptor for the requested offset.
3149          */
3150         raw_cons = cpr->cp_raw_cons;
3151         desc = 0;
3152         while (1) {
3153                 uint32_t agg_cnt, cons, cmpl_type;
3154
3155                 cons = RING_CMP(cpr->cp_ring_struct, raw_cons);
3156                 rxcmp = (struct rx_pkt_cmpl *)&cpr->cp_desc_ring[cons];
3157
3158                 if (!CMP_VALID(rxcmp, raw_cons, cpr->cp_ring_struct))
3159                         break;
3160
3161                 cmpl_type = CMP_TYPE(rxcmp);
3162
3163                 switch (cmpl_type) {
3164                 case CMPL_BASE_TYPE_RX_L2:
3165                 case CMPL_BASE_TYPE_RX_L2_V2:
3166                         if (desc == offset) {
3167                                 cons = rxcmp->opaque;
3168                                 if (rxr->rx_buf_ring[cons])
3169                                         return RTE_ETH_RX_DESC_DONE;
3170                                 else
3171                                         return RTE_ETH_RX_DESC_UNAVAIL;
3172                         }
3173                         agg_cnt = BNXT_RX_L2_AGG_BUFS(rxcmp);
3174                         raw_cons = raw_cons + CMP_LEN(cmpl_type) + agg_cnt;
3175                         desc++;
3176                         break;
3177
3178                 case CMPL_BASE_TYPE_RX_TPA_END:
3179                         if (desc == offset)
3180                                 return RTE_ETH_RX_DESC_DONE;
3181
3182                         if (BNXT_CHIP_P5(rxq->bp)) {
3183                                 struct rx_tpa_v2_end_cmpl_hi *p5_tpa_end;
3184
3185                                 p5_tpa_end = (void *)rxcmp;
3186                                 agg_cnt = BNXT_TPA_END_AGG_BUFS_TH(p5_tpa_end);
3187                         } else {
3188                                 struct rx_tpa_end_cmpl *tpa_end;
3189
3190                                 tpa_end = (void *)rxcmp;
3191                                 agg_cnt = BNXT_TPA_END_AGG_BUFS(tpa_end);
3192                         }
3193
3194                         raw_cons = raw_cons + CMP_LEN(cmpl_type) + agg_cnt;
3195                         desc++;
3196                         break;
3197
3198                 default:
3199                         raw_cons += CMP_LEN(cmpl_type);
3200                 }
3201         }
3202
3203         return RTE_ETH_RX_DESC_AVAIL;
3204 }
3205
3206 static int
3207 bnxt_tx_descriptor_status_op(void *tx_queue, uint16_t offset)
3208 {
3209         struct bnxt_tx_queue *txq = (struct bnxt_tx_queue *)tx_queue;
3210         struct bnxt_tx_ring_info *txr;
3211         struct bnxt_cp_ring_info *cpr;
3212         struct rte_mbuf **tx_buf;
3213         struct tx_pkt_cmpl *txcmp;
3214         uint32_t cons, cp_cons;
3215         int rc;
3216
3217         if (!txq)
3218                 return -EINVAL;
3219
3220         rc = is_bnxt_in_error(txq->bp);
3221         if (rc)
3222                 return rc;
3223
3224         cpr = txq->cp_ring;
3225         txr = txq->tx_ring;
3226
3227         if (offset >= txq->nb_tx_desc)
3228                 return -EINVAL;
3229
3230         cons = RING_CMP(cpr->cp_ring_struct, offset);
3231         txcmp = (struct tx_pkt_cmpl *)&cpr->cp_desc_ring[cons];
3232         cp_cons = cpr->cp_raw_cons;
3233
3234         if (cons > cp_cons) {
3235                 if (CMPL_VALID(txcmp, cpr->valid))
3236                         return RTE_ETH_TX_DESC_UNAVAIL;
3237         } else {
3238                 if (CMPL_VALID(txcmp, !cpr->valid))
3239                         return RTE_ETH_TX_DESC_UNAVAIL;
3240         }
3241         tx_buf = &txr->tx_buf_ring[cons];
3242         if (*tx_buf == NULL)
3243                 return RTE_ETH_TX_DESC_DONE;
3244
3245         return RTE_ETH_TX_DESC_FULL;
3246 }
3247
3248 int
3249 bnxt_flow_ops_get_op(struct rte_eth_dev *dev,
3250                      const struct rte_flow_ops **ops)
3251 {
3252         struct bnxt *bp = dev->data->dev_private;
3253         int ret = 0;
3254
3255         if (!bp)
3256                 return -EIO;
3257
3258         if (BNXT_ETH_DEV_IS_REPRESENTOR(dev)) {
3259                 struct bnxt_representor *vfr = dev->data->dev_private;
3260                 bp = vfr->parent_dev->data->dev_private;
3261                 /* parent is deleted while children are still valid */
3262                 if (!bp) {
3263                         PMD_DRV_LOG(DEBUG, "BNXT Port:%d VFR Error\n",
3264                                     dev->data->port_id);
3265                         return -EIO;
3266                 }
3267         }
3268
3269         ret = is_bnxt_in_error(bp);
3270         if (ret)
3271                 return ret;
3272
3273         /* PMD supports thread-safe flow operations.  rte_flow API
3274          * functions can avoid mutex for multi-thread safety.
3275          */
3276         dev->data->dev_flags |= RTE_ETH_DEV_FLOW_OPS_THREAD_SAFE;
3277
3278         if (BNXT_TRUFLOW_EN(bp))
3279                 *ops = &bnxt_ulp_rte_flow_ops;
3280         else
3281                 *ops = &bnxt_flow_ops;
3282
3283         return ret;
3284 }
3285
3286 static const uint32_t *
3287 bnxt_dev_supported_ptypes_get_op(struct rte_eth_dev *dev)
3288 {
3289         static const uint32_t ptypes[] = {
3290                 RTE_PTYPE_L2_ETHER_VLAN,
3291                 RTE_PTYPE_L3_IPV4_EXT_UNKNOWN,
3292                 RTE_PTYPE_L3_IPV6_EXT_UNKNOWN,
3293                 RTE_PTYPE_L4_ICMP,
3294                 RTE_PTYPE_L4_TCP,
3295                 RTE_PTYPE_L4_UDP,
3296                 RTE_PTYPE_INNER_L3_IPV4_EXT_UNKNOWN,
3297                 RTE_PTYPE_INNER_L3_IPV6_EXT_UNKNOWN,
3298                 RTE_PTYPE_INNER_L4_ICMP,
3299                 RTE_PTYPE_INNER_L4_TCP,
3300                 RTE_PTYPE_INNER_L4_UDP,
3301                 RTE_PTYPE_UNKNOWN
3302         };
3303
3304         if (!dev->rx_pkt_burst)
3305                 return NULL;
3306
3307         return ptypes;
3308 }
3309
3310 static int bnxt_map_regs(struct bnxt *bp, uint32_t *reg_arr, int count,
3311                          int reg_win)
3312 {
3313         uint32_t reg_base = *reg_arr & 0xfffff000;
3314         uint32_t win_off;
3315         int i;
3316
3317         for (i = 0; i < count; i++) {
3318                 if ((reg_arr[i] & 0xfffff000) != reg_base)
3319                         return -ERANGE;
3320         }
3321         win_off = BNXT_GRCPF_REG_WINDOW_BASE_OUT + (reg_win - 1) * 4;
3322         rte_write32(reg_base, (uint8_t *)bp->bar0 + win_off);
3323         return 0;
3324 }
3325
3326 static int bnxt_map_ptp_regs(struct bnxt *bp)
3327 {
3328         struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3329         uint32_t *reg_arr;
3330         int rc, i;
3331
3332         reg_arr = ptp->rx_regs;
3333         rc = bnxt_map_regs(bp, reg_arr, BNXT_PTP_RX_REGS, 5);
3334         if (rc)
3335                 return rc;
3336
3337         reg_arr = ptp->tx_regs;
3338         rc = bnxt_map_regs(bp, reg_arr, BNXT_PTP_TX_REGS, 6);
3339         if (rc)
3340                 return rc;
3341
3342         for (i = 0; i < BNXT_PTP_RX_REGS; i++)
3343                 ptp->rx_mapped_regs[i] = 0x5000 + (ptp->rx_regs[i] & 0xfff);
3344
3345         for (i = 0; i < BNXT_PTP_TX_REGS; i++)
3346                 ptp->tx_mapped_regs[i] = 0x6000 + (ptp->tx_regs[i] & 0xfff);
3347
3348         return 0;
3349 }
3350
3351 static void bnxt_unmap_ptp_regs(struct bnxt *bp)
3352 {
3353         rte_write32(0, (uint8_t *)bp->bar0 +
3354                          BNXT_GRCPF_REG_WINDOW_BASE_OUT + 16);
3355         rte_write32(0, (uint8_t *)bp->bar0 +
3356                          BNXT_GRCPF_REG_WINDOW_BASE_OUT + 20);
3357 }
3358
3359 static uint64_t bnxt_cc_read(struct bnxt *bp)
3360 {
3361         uint64_t ns;
3362
3363         ns = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3364                               BNXT_GRCPF_REG_SYNC_TIME));
3365         ns |= (uint64_t)(rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3366                                           BNXT_GRCPF_REG_SYNC_TIME + 4))) << 32;
3367         return ns;
3368 }
3369
3370 static int bnxt_get_tx_ts(struct bnxt *bp, uint64_t *ts)
3371 {
3372         struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3373         uint32_t fifo;
3374
3375         fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3376                                 ptp->tx_mapped_regs[BNXT_PTP_TX_FIFO]));
3377         if (fifo & BNXT_PTP_TX_FIFO_EMPTY)
3378                 return -EAGAIN;
3379
3380         fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3381                                 ptp->tx_mapped_regs[BNXT_PTP_TX_FIFO]));
3382         *ts = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3383                                 ptp->tx_mapped_regs[BNXT_PTP_TX_TS_L]));
3384         *ts |= (uint64_t)rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3385                                 ptp->tx_mapped_regs[BNXT_PTP_TX_TS_H])) << 32;
3386         rte_read32((uint8_t *)bp->bar0 + ptp->tx_mapped_regs[BNXT_PTP_TX_SEQ]);
3387
3388         return 0;
3389 }
3390
3391 static int bnxt_get_rx_ts(struct bnxt *bp, uint64_t *ts)
3392 {
3393         struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3394         struct bnxt_pf_info *pf = bp->pf;
3395         uint16_t port_id;
3396         uint32_t fifo;
3397
3398         fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3399                                 ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO]));
3400         if (!(fifo & BNXT_PTP_RX_FIFO_PENDING))
3401                 return -EAGAIN;
3402
3403         port_id = pf->port_id;
3404         rte_write32(1 << port_id, (uint8_t *)bp->bar0 +
3405                ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO_ADV]);
3406
3407         fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3408                                    ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO]));
3409         if (fifo & BNXT_PTP_RX_FIFO_PENDING) {
3410 /*              bnxt_clr_rx_ts(bp);       TBD  */
3411                 return -EBUSY;
3412         }
3413
3414         *ts = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3415                                 ptp->rx_mapped_regs[BNXT_PTP_RX_TS_L]));
3416         *ts |= (uint64_t)rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3417                                 ptp->rx_mapped_regs[BNXT_PTP_RX_TS_H])) << 32;
3418
3419         return 0;
3420 }
3421
3422 static int
3423 bnxt_timesync_write_time(struct rte_eth_dev *dev, const struct timespec *ts)
3424 {
3425         uint64_t ns;
3426         struct bnxt *bp = dev->data->dev_private;
3427         struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3428
3429         if (!ptp)
3430                 return -ENOTSUP;
3431
3432         ns = rte_timespec_to_ns(ts);
3433         /* Set the timecounters to a new value. */
3434         ptp->tc.nsec = ns;
3435         ptp->tx_tstamp_tc.nsec = ns;
3436         ptp->rx_tstamp_tc.nsec = ns;
3437
3438         return 0;
3439 }
3440
3441 static int
3442 bnxt_timesync_read_time(struct rte_eth_dev *dev, struct timespec *ts)
3443 {
3444         struct bnxt *bp = dev->data->dev_private;
3445         struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3446         uint64_t ns, systime_cycles = 0;
3447         int rc = 0;
3448
3449         if (!ptp)
3450                 return -ENOTSUP;
3451
3452         if (BNXT_CHIP_P5(bp))
3453                 rc = bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_CURRENT_TIME,
3454                                              &systime_cycles);
3455         else
3456                 systime_cycles = bnxt_cc_read(bp);
3457
3458         ns = rte_timecounter_update(&ptp->tc, systime_cycles);
3459         *ts = rte_ns_to_timespec(ns);
3460
3461         return rc;
3462 }
3463 static int
3464 bnxt_timesync_enable(struct rte_eth_dev *dev)
3465 {
3466         struct bnxt *bp = dev->data->dev_private;
3467         struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3468         uint32_t shift = 0;
3469         int rc;
3470
3471         if (!ptp)
3472                 return -ENOTSUP;
3473
3474         ptp->rx_filter = 1;
3475         ptp->tx_tstamp_en = 1;
3476         ptp->rxctl = BNXT_PTP_MSG_EVENTS;
3477
3478         rc = bnxt_hwrm_ptp_cfg(bp);
3479         if (rc)
3480                 return rc;
3481
3482         memset(&ptp->tc, 0, sizeof(struct rte_timecounter));
3483         memset(&ptp->rx_tstamp_tc, 0, sizeof(struct rte_timecounter));
3484         memset(&ptp->tx_tstamp_tc, 0, sizeof(struct rte_timecounter));
3485
3486         ptp->tc.cc_mask = BNXT_CYCLECOUNTER_MASK;
3487         ptp->tc.cc_shift = shift;
3488         ptp->tc.nsec_mask = (1ULL << shift) - 1;
3489
3490         ptp->rx_tstamp_tc.cc_mask = BNXT_CYCLECOUNTER_MASK;
3491         ptp->rx_tstamp_tc.cc_shift = shift;
3492         ptp->rx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
3493
3494         ptp->tx_tstamp_tc.cc_mask = BNXT_CYCLECOUNTER_MASK;
3495         ptp->tx_tstamp_tc.cc_shift = shift;
3496         ptp->tx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
3497
3498         if (!BNXT_CHIP_P5(bp))
3499                 bnxt_map_ptp_regs(bp);
3500         else
3501                 rc = bnxt_ptp_start(bp);
3502
3503         return rc;
3504 }
3505
3506 static int
3507 bnxt_timesync_disable(struct rte_eth_dev *dev)
3508 {
3509         struct bnxt *bp = dev->data->dev_private;
3510         struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3511
3512         if (!ptp)
3513                 return -ENOTSUP;
3514
3515         ptp->rx_filter = 0;
3516         ptp->tx_tstamp_en = 0;
3517         ptp->rxctl = 0;
3518
3519         bnxt_hwrm_ptp_cfg(bp);
3520
3521         if (!BNXT_CHIP_P5(bp))
3522                 bnxt_unmap_ptp_regs(bp);
3523         else
3524                 bnxt_ptp_stop(bp);
3525
3526         return 0;
3527 }
3528
3529 static int
3530 bnxt_timesync_read_rx_timestamp(struct rte_eth_dev *dev,
3531                                  struct timespec *timestamp,
3532                                  uint32_t flags __rte_unused)
3533 {
3534         struct bnxt *bp = dev->data->dev_private;
3535         struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3536         uint64_t rx_tstamp_cycles = 0;
3537         uint64_t ns;
3538
3539         if (!ptp)
3540                 return -ENOTSUP;
3541
3542         if (BNXT_CHIP_P5(bp))
3543                 rx_tstamp_cycles = ptp->rx_timestamp;
3544         else
3545                 bnxt_get_rx_ts(bp, &rx_tstamp_cycles);
3546
3547         ns = rte_timecounter_update(&ptp->rx_tstamp_tc, rx_tstamp_cycles);
3548         *timestamp = rte_ns_to_timespec(ns);
3549         return  0;
3550 }
3551
3552 static int
3553 bnxt_timesync_read_tx_timestamp(struct rte_eth_dev *dev,
3554                                  struct timespec *timestamp)
3555 {
3556         struct bnxt *bp = dev->data->dev_private;
3557         struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3558         uint64_t tx_tstamp_cycles = 0;
3559         uint64_t ns;
3560         int rc = 0;
3561
3562         if (!ptp)
3563                 return -ENOTSUP;
3564
3565         if (BNXT_CHIP_P5(bp))
3566                 rc = bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_PATH_TX,
3567                                              &tx_tstamp_cycles);
3568         else
3569                 rc = bnxt_get_tx_ts(bp, &tx_tstamp_cycles);
3570
3571         ns = rte_timecounter_update(&ptp->tx_tstamp_tc, tx_tstamp_cycles);
3572         *timestamp = rte_ns_to_timespec(ns);
3573
3574         return rc;
3575 }
3576
3577 static int
3578 bnxt_timesync_adjust_time(struct rte_eth_dev *dev, int64_t delta)
3579 {
3580         struct bnxt *bp = dev->data->dev_private;
3581         struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3582
3583         if (!ptp)
3584                 return -ENOTSUP;
3585
3586         ptp->tc.nsec += delta;
3587         ptp->tx_tstamp_tc.nsec += delta;
3588         ptp->rx_tstamp_tc.nsec += delta;
3589
3590         return 0;
3591 }
3592
3593 static int
3594 bnxt_get_eeprom_length_op(struct rte_eth_dev *dev)
3595 {
3596         struct bnxt *bp = dev->data->dev_private;
3597         int rc;
3598         uint32_t dir_entries;
3599         uint32_t entry_length;
3600
3601         rc = is_bnxt_in_error(bp);
3602         if (rc)
3603                 return rc;
3604
3605         PMD_DRV_LOG(INFO, PCI_PRI_FMT "\n",
3606                     bp->pdev->addr.domain, bp->pdev->addr.bus,
3607                     bp->pdev->addr.devid, bp->pdev->addr.function);
3608
3609         rc = bnxt_hwrm_nvm_get_dir_info(bp, &dir_entries, &entry_length);
3610         if (rc != 0)
3611                 return rc;
3612
3613         return dir_entries * entry_length;
3614 }
3615
3616 static int
3617 bnxt_get_eeprom_op(struct rte_eth_dev *dev,
3618                 struct rte_dev_eeprom_info *in_eeprom)
3619 {
3620         struct bnxt *bp = dev->data->dev_private;
3621         uint32_t index;
3622         uint32_t offset;
3623         int rc;
3624
3625         rc = is_bnxt_in_error(bp);
3626         if (rc)
3627                 return rc;
3628
3629         PMD_DRV_LOG(INFO, PCI_PRI_FMT " in_eeprom->offset = %d len = %d\n",
3630                     bp->pdev->addr.domain, bp->pdev->addr.bus,
3631                     bp->pdev->addr.devid, bp->pdev->addr.function,
3632                     in_eeprom->offset, in_eeprom->length);
3633
3634         if (in_eeprom->offset == 0) /* special offset value to get directory */
3635                 return bnxt_get_nvram_directory(bp, in_eeprom->length,
3636                                                 in_eeprom->data);
3637
3638         index = in_eeprom->offset >> 24;
3639         offset = in_eeprom->offset & 0xffffff;
3640
3641         if (index != 0)
3642                 return bnxt_hwrm_get_nvram_item(bp, index - 1, offset,
3643                                            in_eeprom->length, in_eeprom->data);
3644
3645         return 0;
3646 }
3647
3648 static bool bnxt_dir_type_is_ape_bin_format(uint16_t dir_type)
3649 {
3650         switch (dir_type) {
3651         case BNX_DIR_TYPE_CHIMP_PATCH:
3652         case BNX_DIR_TYPE_BOOTCODE:
3653         case BNX_DIR_TYPE_BOOTCODE_2:
3654         case BNX_DIR_TYPE_APE_FW:
3655         case BNX_DIR_TYPE_APE_PATCH:
3656         case BNX_DIR_TYPE_KONG_FW:
3657         case BNX_DIR_TYPE_KONG_PATCH:
3658         case BNX_DIR_TYPE_BONO_FW:
3659         case BNX_DIR_TYPE_BONO_PATCH:
3660                 /* FALLTHROUGH */
3661                 return true;
3662         }
3663
3664         return false;
3665 }
3666
3667 static bool bnxt_dir_type_is_other_exec_format(uint16_t dir_type)
3668 {
3669         switch (dir_type) {
3670         case BNX_DIR_TYPE_AVS:
3671         case BNX_DIR_TYPE_EXP_ROM_MBA:
3672         case BNX_DIR_TYPE_PCIE:
3673         case BNX_DIR_TYPE_TSCF_UCODE:
3674         case BNX_DIR_TYPE_EXT_PHY:
3675         case BNX_DIR_TYPE_CCM:
3676         case BNX_DIR_TYPE_ISCSI_BOOT:
3677         case BNX_DIR_TYPE_ISCSI_BOOT_IPV6:
3678         case BNX_DIR_TYPE_ISCSI_BOOT_IPV4N6:
3679                 /* FALLTHROUGH */
3680                 return true;
3681         }
3682
3683         return false;
3684 }
3685
3686 static bool bnxt_dir_type_is_executable(uint16_t dir_type)
3687 {
3688         return bnxt_dir_type_is_ape_bin_format(dir_type) ||
3689                 bnxt_dir_type_is_other_exec_format(dir_type);
3690 }
3691
3692 static int
3693 bnxt_set_eeprom_op(struct rte_eth_dev *dev,
3694                 struct rte_dev_eeprom_info *in_eeprom)
3695 {
3696         struct bnxt *bp = dev->data->dev_private;
3697         uint8_t index, dir_op;
3698         uint16_t type, ext, ordinal, attr;
3699         int rc;
3700
3701         rc = is_bnxt_in_error(bp);
3702         if (rc)
3703                 return rc;
3704
3705         PMD_DRV_LOG(INFO, PCI_PRI_FMT " in_eeprom->offset = %d len = %d\n",
3706                     bp->pdev->addr.domain, bp->pdev->addr.bus,
3707                     bp->pdev->addr.devid, bp->pdev->addr.function,
3708                     in_eeprom->offset, in_eeprom->length);
3709
3710         if (!BNXT_PF(bp)) {
3711                 PMD_DRV_LOG(ERR, "NVM write not supported from a VF\n");
3712                 return -EINVAL;
3713         }
3714
3715         type = in_eeprom->magic >> 16;
3716
3717         if (type == 0xffff) { /* special value for directory operations */
3718                 index = in_eeprom->magic & 0xff;
3719                 dir_op = in_eeprom->magic >> 8;
3720                 if (index == 0)
3721                         return -EINVAL;
3722                 switch (dir_op) {
3723                 case 0x0e: /* erase */
3724                         if (in_eeprom->offset != ~in_eeprom->magic)
3725                                 return -EINVAL;
3726                         return bnxt_hwrm_erase_nvram_directory(bp, index - 1);
3727                 default:
3728                         return -EINVAL;
3729                 }
3730         }
3731
3732         /* Create or re-write an NVM item: */
3733         if (bnxt_dir_type_is_executable(type) == true)
3734                 return -EOPNOTSUPP;
3735         ext = in_eeprom->magic & 0xffff;
3736         ordinal = in_eeprom->offset >> 16;
3737         attr = in_eeprom->offset & 0xffff;
3738
3739         return bnxt_hwrm_flash_nvram(bp, type, ordinal, ext, attr,
3740                                      in_eeprom->data, in_eeprom->length);
3741 }
3742
3743 /*
3744  * Initialization
3745  */
3746
3747 static const struct eth_dev_ops bnxt_dev_ops = {
3748         .dev_infos_get = bnxt_dev_info_get_op,
3749         .dev_close = bnxt_dev_close_op,
3750         .dev_configure = bnxt_dev_configure_op,
3751         .dev_start = bnxt_dev_start_op,
3752         .dev_stop = bnxt_dev_stop_op,
3753         .dev_set_link_up = bnxt_dev_set_link_up_op,
3754         .dev_set_link_down = bnxt_dev_set_link_down_op,
3755         .stats_get = bnxt_stats_get_op,
3756         .stats_reset = bnxt_stats_reset_op,
3757         .rx_queue_setup = bnxt_rx_queue_setup_op,
3758         .rx_queue_release = bnxt_rx_queue_release_op,
3759         .tx_queue_setup = bnxt_tx_queue_setup_op,
3760         .tx_queue_release = bnxt_tx_queue_release_op,
3761         .rx_queue_intr_enable = bnxt_rx_queue_intr_enable_op,
3762         .rx_queue_intr_disable = bnxt_rx_queue_intr_disable_op,
3763         .reta_update = bnxt_reta_update_op,
3764         .reta_query = bnxt_reta_query_op,
3765         .rss_hash_update = bnxt_rss_hash_update_op,
3766         .rss_hash_conf_get = bnxt_rss_hash_conf_get_op,
3767         .link_update = bnxt_link_update_op,
3768         .promiscuous_enable = bnxt_promiscuous_enable_op,
3769         .promiscuous_disable = bnxt_promiscuous_disable_op,
3770         .allmulticast_enable = bnxt_allmulticast_enable_op,
3771         .allmulticast_disable = bnxt_allmulticast_disable_op,
3772         .mac_addr_add = bnxt_mac_addr_add_op,
3773         .mac_addr_remove = bnxt_mac_addr_remove_op,
3774         .flow_ctrl_get = bnxt_flow_ctrl_get_op,
3775         .flow_ctrl_set = bnxt_flow_ctrl_set_op,
3776         .udp_tunnel_port_add  = bnxt_udp_tunnel_port_add_op,
3777         .udp_tunnel_port_del  = bnxt_udp_tunnel_port_del_op,
3778         .vlan_filter_set = bnxt_vlan_filter_set_op,
3779         .vlan_offload_set = bnxt_vlan_offload_set_op,
3780         .vlan_tpid_set = bnxt_vlan_tpid_set_op,
3781         .vlan_pvid_set = bnxt_vlan_pvid_set_op,
3782         .mtu_set = bnxt_mtu_set_op,
3783         .mac_addr_set = bnxt_set_default_mac_addr_op,
3784         .xstats_get = bnxt_dev_xstats_get_op,
3785         .xstats_get_names = bnxt_dev_xstats_get_names_op,
3786         .xstats_reset = bnxt_dev_xstats_reset_op,
3787         .fw_version_get = bnxt_fw_version_get,
3788         .set_mc_addr_list = bnxt_dev_set_mc_addr_list_op,
3789         .rxq_info_get = bnxt_rxq_info_get_op,
3790         .txq_info_get = bnxt_txq_info_get_op,
3791         .rx_burst_mode_get = bnxt_rx_burst_mode_get,
3792         .tx_burst_mode_get = bnxt_tx_burst_mode_get,
3793         .dev_led_on = bnxt_dev_led_on_op,
3794         .dev_led_off = bnxt_dev_led_off_op,
3795         .rx_queue_start = bnxt_rx_queue_start,
3796         .rx_queue_stop = bnxt_rx_queue_stop,
3797         .tx_queue_start = bnxt_tx_queue_start,
3798         .tx_queue_stop = bnxt_tx_queue_stop,
3799         .flow_ops_get = bnxt_flow_ops_get_op,
3800         .dev_supported_ptypes_get = bnxt_dev_supported_ptypes_get_op,
3801         .get_eeprom_length    = bnxt_get_eeprom_length_op,
3802         .get_eeprom           = bnxt_get_eeprom_op,
3803         .set_eeprom           = bnxt_set_eeprom_op,
3804         .timesync_enable      = bnxt_timesync_enable,
3805         .timesync_disable     = bnxt_timesync_disable,
3806         .timesync_read_time   = bnxt_timesync_read_time,
3807         .timesync_write_time   = bnxt_timesync_write_time,
3808         .timesync_adjust_time = bnxt_timesync_adjust_time,
3809         .timesync_read_rx_timestamp = bnxt_timesync_read_rx_timestamp,
3810         .timesync_read_tx_timestamp = bnxt_timesync_read_tx_timestamp,
3811 };
3812
3813 static uint32_t bnxt_map_reset_regs(struct bnxt *bp, uint32_t reg)
3814 {
3815         uint32_t offset;
3816
3817         /* Only pre-map the reset GRC registers using window 3 */
3818         rte_write32(reg & 0xfffff000, (uint8_t *)bp->bar0 +
3819                     BNXT_GRCPF_REG_WINDOW_BASE_OUT + 8);
3820
3821         offset = BNXT_GRCP_WINDOW_3_BASE + (reg & 0xffc);
3822
3823         return offset;
3824 }
3825
3826 int bnxt_map_fw_health_status_regs(struct bnxt *bp)
3827 {
3828         struct bnxt_error_recovery_info *info = bp->recovery_info;
3829         uint32_t reg_base = 0xffffffff;
3830         int i;
3831
3832         /* Only pre-map the monitoring GRC registers using window 2 */
3833         for (i = 0; i < BNXT_FW_STATUS_REG_CNT; i++) {
3834                 uint32_t reg = info->status_regs[i];
3835
3836                 if (BNXT_FW_STATUS_REG_TYPE(reg) != BNXT_FW_STATUS_REG_TYPE_GRC)
3837                         continue;
3838
3839                 if (reg_base == 0xffffffff)
3840                         reg_base = reg & 0xfffff000;
3841                 if ((reg & 0xfffff000) != reg_base)
3842                         return -ERANGE;
3843
3844                 /* Use mask 0xffc as the Lower 2 bits indicates
3845                  * address space location
3846                  */
3847                 info->mapped_status_regs[i] = BNXT_GRCP_WINDOW_2_BASE +
3848                                                 (reg & 0xffc);
3849         }
3850
3851         if (reg_base == 0xffffffff)
3852                 return 0;
3853
3854         rte_write32(reg_base, (uint8_t *)bp->bar0 +
3855                     BNXT_GRCPF_REG_WINDOW_BASE_OUT + 4);
3856
3857         return 0;
3858 }
3859
3860 static void bnxt_write_fw_reset_reg(struct bnxt *bp, uint32_t index)
3861 {
3862         struct bnxt_error_recovery_info *info = bp->recovery_info;
3863         uint32_t delay = info->delay_after_reset[index];
3864         uint32_t val = info->reset_reg_val[index];
3865         uint32_t reg = info->reset_reg[index];
3866         uint32_t type, offset;
3867         int ret;
3868
3869         type = BNXT_FW_STATUS_REG_TYPE(reg);
3870         offset = BNXT_FW_STATUS_REG_OFF(reg);
3871
3872         switch (type) {
3873         case BNXT_FW_STATUS_REG_TYPE_CFG:
3874                 ret = rte_pci_write_config(bp->pdev, &val, sizeof(val), offset);
3875                 if (ret < 0) {
3876                         PMD_DRV_LOG(ERR, "Failed to write %#x at PCI offset %#x",
3877                                     val, offset);
3878                         return;
3879                 }
3880                 break;
3881         case BNXT_FW_STATUS_REG_TYPE_GRC:
3882                 offset = bnxt_map_reset_regs(bp, offset);
3883                 rte_write32(val, (uint8_t *)bp->bar0 + offset);
3884                 break;
3885         case BNXT_FW_STATUS_REG_TYPE_BAR0:
3886                 rte_write32(val, (uint8_t *)bp->bar0 + offset);
3887                 break;
3888         }
3889         /* wait on a specific interval of time until core reset is complete */
3890         if (delay)
3891                 rte_delay_ms(delay);
3892 }
3893
3894 static void bnxt_dev_cleanup(struct bnxt *bp)
3895 {
3896         bp->eth_dev->data->dev_link.link_status = 0;
3897         bp->link_info->link_up = 0;
3898         if (bp->eth_dev->data->dev_started)
3899                 bnxt_dev_stop(bp->eth_dev);
3900
3901         bnxt_uninit_resources(bp, true);
3902 }
3903
3904 static int
3905 bnxt_check_fw_reset_done(struct bnxt *bp)
3906 {
3907         int timeout = bp->fw_reset_max_msecs;
3908         uint16_t val = 0;
3909         int rc;
3910
3911         do {
3912                 rc = rte_pci_read_config(bp->pdev, &val, sizeof(val), PCI_SUBSYSTEM_ID_OFFSET);
3913                 if (rc < 0) {
3914                         PMD_DRV_LOG(ERR, "Failed to read PCI offset 0x%x", PCI_SUBSYSTEM_ID_OFFSET);
3915                         return rc;
3916                 }
3917                 if (val != 0xffff)
3918                         break;
3919                 rte_delay_ms(1);
3920         } while (timeout--);
3921
3922         if (val == 0xffff) {
3923                 PMD_DRV_LOG(ERR, "Firmware reset aborted, PCI config space invalid\n");
3924                 return -1;
3925         }
3926
3927         return 0;
3928 }
3929
3930 static int bnxt_restore_vlan_filters(struct bnxt *bp)
3931 {
3932         struct rte_eth_dev *dev = bp->eth_dev;
3933         struct rte_vlan_filter_conf *vfc;
3934         int vidx, vbit, rc;
3935         uint16_t vlan_id;
3936
3937         for (vlan_id = 1; vlan_id <= RTE_ETHER_MAX_VLAN_ID; vlan_id++) {
3938                 vfc = &dev->data->vlan_filter_conf;
3939                 vidx = vlan_id / 64;
3940                 vbit = vlan_id % 64;
3941
3942                 /* Each bit corresponds to a VLAN id */
3943                 if (vfc->ids[vidx] & (UINT64_C(1) << vbit)) {
3944                         rc = bnxt_add_vlan_filter(bp, vlan_id);
3945                         if (rc)
3946                                 return rc;
3947                 }
3948         }
3949
3950         return 0;
3951 }
3952
3953 static int bnxt_restore_mac_filters(struct bnxt *bp)
3954 {
3955         struct rte_eth_dev *dev = bp->eth_dev;
3956         struct rte_eth_dev_info dev_info;
3957         struct rte_ether_addr *addr;
3958         uint64_t pool_mask;
3959         uint32_t pool = 0;
3960         uint16_t i;
3961         int rc;
3962
3963         if (BNXT_VF(bp) && !BNXT_VF_IS_TRUSTED(bp))
3964                 return 0;
3965
3966         rc = bnxt_dev_info_get_op(dev, &dev_info);
3967         if (rc)
3968                 return rc;
3969
3970         /* replay MAC address configuration */
3971         for (i = 1; i < dev_info.max_mac_addrs; i++) {
3972                 addr = &dev->data->mac_addrs[i];
3973
3974                 /* skip zero address */
3975                 if (rte_is_zero_ether_addr(addr))
3976                         continue;
3977
3978                 pool = 0;
3979                 pool_mask = dev->data->mac_pool_sel[i];
3980
3981                 do {
3982                         if (pool_mask & 1ULL) {
3983                                 rc = bnxt_mac_addr_add_op(dev, addr, i, pool);
3984                                 if (rc)
3985                                         return rc;
3986                         }
3987                         pool_mask >>= 1;
3988                         pool++;
3989                 } while (pool_mask);
3990         }
3991
3992         return 0;
3993 }
3994
3995 static int bnxt_restore_filters(struct bnxt *bp)
3996 {
3997         struct rte_eth_dev *dev = bp->eth_dev;
3998         int ret = 0;
3999
4000         if (dev->data->all_multicast) {
4001                 ret = bnxt_allmulticast_enable_op(dev);
4002                 if (ret)
4003                         return ret;
4004         }
4005         if (dev->data->promiscuous) {
4006                 ret = bnxt_promiscuous_enable_op(dev);
4007                 if (ret)
4008                         return ret;
4009         }
4010
4011         ret = bnxt_restore_mac_filters(bp);
4012         if (ret)
4013                 return ret;
4014
4015         ret = bnxt_restore_vlan_filters(bp);
4016         /* TODO restore other filters as well */
4017         return ret;
4018 }
4019
4020 static int bnxt_check_fw_ready(struct bnxt *bp)
4021 {
4022         int timeout = bp->fw_reset_max_msecs;
4023         int rc = 0;
4024
4025         do {
4026                 rc = bnxt_hwrm_poll_ver_get(bp);
4027                 if (rc == 0)
4028                         break;
4029                 rte_delay_ms(BNXT_FW_READY_WAIT_INTERVAL);
4030                 timeout -= BNXT_FW_READY_WAIT_INTERVAL;
4031         } while (rc && timeout > 0);
4032
4033         if (rc)
4034                 PMD_DRV_LOG(ERR, "FW is not Ready after reset\n");
4035
4036         return rc;
4037 }
4038
4039 static void bnxt_dev_recover(void *arg)
4040 {
4041         struct bnxt *bp = arg;
4042         int rc = 0;
4043
4044         pthread_mutex_lock(&bp->err_recovery_lock);
4045
4046         if (!bp->fw_reset_min_msecs) {
4047                 rc = bnxt_check_fw_reset_done(bp);
4048                 if (rc)
4049                         goto err;
4050         }
4051
4052         /* Clear Error flag so that device re-init should happen */
4053         bp->flags &= ~BNXT_FLAG_FATAL_ERROR;
4054
4055         rc = bnxt_check_fw_ready(bp);
4056         if (rc)
4057                 goto err;
4058
4059         rc = bnxt_init_resources(bp, true);
4060         if (rc) {
4061                 PMD_DRV_LOG(ERR,
4062                             "Failed to initialize resources after reset\n");
4063                 goto err;
4064         }
4065         /* clear reset flag as the device is initialized now */
4066         bp->flags &= ~BNXT_FLAG_FW_RESET;
4067
4068         rc = bnxt_dev_start_op(bp->eth_dev);
4069         if (rc) {
4070                 PMD_DRV_LOG(ERR, "Failed to start port after reset\n");
4071                 goto err_start;
4072         }
4073
4074         rc = bnxt_restore_filters(bp);
4075         if (rc)
4076                 goto err_start;
4077
4078         PMD_DRV_LOG(INFO, "Recovered from FW reset\n");
4079         pthread_mutex_unlock(&bp->err_recovery_lock);
4080
4081         return;
4082 err_start:
4083         bnxt_dev_stop(bp->eth_dev);
4084 err:
4085         bp->flags |= BNXT_FLAG_FATAL_ERROR;
4086         bnxt_uninit_resources(bp, false);
4087         pthread_mutex_unlock(&bp->err_recovery_lock);
4088         PMD_DRV_LOG(ERR, "Failed to recover from FW reset\n");
4089 }
4090
4091 void bnxt_dev_reset_and_resume(void *arg)
4092 {
4093         struct bnxt *bp = arg;
4094         uint32_t us = US_PER_MS * bp->fw_reset_min_msecs;
4095         uint16_t val = 0;
4096         int rc;
4097
4098         bnxt_dev_cleanup(bp);
4099
4100         bnxt_wait_for_device_shutdown(bp);
4101
4102         /* During some fatal firmware error conditions, the PCI config space
4103          * register 0x2e which normally contains the subsystem ID will become
4104          * 0xffff. This register will revert back to the normal value after
4105          * the chip has completed core reset. If we detect this condition,
4106          * we can poll this config register immediately for the value to revert.
4107          */
4108         if (bp->flags & BNXT_FLAG_FATAL_ERROR) {
4109                 rc = rte_pci_read_config(bp->pdev, &val, sizeof(val), PCI_SUBSYSTEM_ID_OFFSET);
4110                 if (rc < 0) {
4111                         PMD_DRV_LOG(ERR, "Failed to read PCI offset 0x%x", PCI_SUBSYSTEM_ID_OFFSET);
4112                         return;
4113                 }
4114                 if (val == 0xffff) {
4115                         bp->fw_reset_min_msecs = 0;
4116                         us = 1;
4117                 }
4118         }
4119
4120         rc = rte_eal_alarm_set(us, bnxt_dev_recover, (void *)bp);
4121         if (rc)
4122                 PMD_DRV_LOG(ERR, "Error setting recovery alarm");
4123 }
4124
4125 uint32_t bnxt_read_fw_status_reg(struct bnxt *bp, uint32_t index)
4126 {
4127         struct bnxt_error_recovery_info *info = bp->recovery_info;
4128         uint32_t reg = info->status_regs[index];
4129         uint32_t type, offset, val = 0;
4130
4131         type = BNXT_FW_STATUS_REG_TYPE(reg);
4132         offset = BNXT_FW_STATUS_REG_OFF(reg);
4133
4134         switch (type) {
4135         case BNXT_FW_STATUS_REG_TYPE_CFG:
4136                 rte_pci_read_config(bp->pdev, &val, sizeof(val), offset);
4137                 break;
4138         case BNXT_FW_STATUS_REG_TYPE_GRC:
4139                 offset = info->mapped_status_regs[index];
4140                 /* FALLTHROUGH */
4141         case BNXT_FW_STATUS_REG_TYPE_BAR0:
4142                 val = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
4143                                        offset));
4144                 break;
4145         }
4146
4147         return val;
4148 }
4149
4150 static int bnxt_fw_reset_all(struct bnxt *bp)
4151 {
4152         struct bnxt_error_recovery_info *info = bp->recovery_info;
4153         uint32_t i;
4154         int rc = 0;
4155
4156         if (info->flags & BNXT_FLAG_ERROR_RECOVERY_HOST) {
4157                 /* Reset through master function driver */
4158                 for (i = 0; i < info->reg_array_cnt; i++)
4159                         bnxt_write_fw_reset_reg(bp, i);
4160                 /* Wait for time specified by FW after triggering reset */
4161                 rte_delay_ms(info->master_func_wait_period_after_reset);
4162         } else if (info->flags & BNXT_FLAG_ERROR_RECOVERY_CO_CPU) {
4163                 /* Reset with the help of Kong processor */
4164                 rc = bnxt_hwrm_fw_reset(bp);
4165                 if (rc)
4166                         PMD_DRV_LOG(ERR, "Failed to reset FW\n");
4167         }
4168
4169         return rc;
4170 }
4171
4172 static void bnxt_fw_reset_cb(void *arg)
4173 {
4174         struct bnxt *bp = arg;
4175         struct bnxt_error_recovery_info *info = bp->recovery_info;
4176         int rc = 0;
4177
4178         /* Only Master function can do FW reset */
4179         if (bnxt_is_master_func(bp) &&
4180             bnxt_is_recovery_enabled(bp)) {
4181                 rc = bnxt_fw_reset_all(bp);
4182                 if (rc) {
4183                         PMD_DRV_LOG(ERR, "Adapter recovery failed\n");
4184                         return;
4185                 }
4186         }
4187
4188         /* if recovery method is ERROR_RECOVERY_CO_CPU, KONG will send
4189          * EXCEPTION_FATAL_ASYNC event to all the functions
4190          * (including MASTER FUNC). After receiving this Async, all the active
4191          * drivers should treat this case as FW initiated recovery
4192          */
4193         if (info->flags & BNXT_FLAG_ERROR_RECOVERY_HOST) {
4194                 bp->fw_reset_min_msecs = BNXT_MIN_FW_READY_TIMEOUT;
4195                 bp->fw_reset_max_msecs = BNXT_MAX_FW_RESET_TIMEOUT;
4196
4197                 /* To recover from error */
4198                 rte_eal_alarm_set(US_PER_MS, bnxt_dev_reset_and_resume,
4199                                   (void *)bp);
4200         }
4201 }
4202
4203 /* Driver should poll FW heartbeat, reset_counter with the frequency
4204  * advertised by FW in HWRM_ERROR_RECOVERY_QCFG.
4205  * When the driver detects heartbeat stop or change in reset_counter,
4206  * it has to trigger a reset to recover from the error condition.
4207  * A “master PF” is the function who will have the privilege to
4208  * initiate the chimp reset. The master PF will be elected by the
4209  * firmware and will be notified through async message.
4210  */
4211 static void bnxt_check_fw_health(void *arg)
4212 {
4213         struct bnxt *bp = arg;
4214         struct bnxt_error_recovery_info *info = bp->recovery_info;
4215         uint32_t val = 0, wait_msec;
4216
4217         if (!info || !bnxt_is_recovery_enabled(bp) ||
4218             is_bnxt_in_error(bp))
4219                 return;
4220
4221         val = bnxt_read_fw_status_reg(bp, BNXT_FW_HEARTBEAT_CNT_REG);
4222         if (val == info->last_heart_beat)
4223                 goto reset;
4224
4225         info->last_heart_beat = val;
4226
4227         val = bnxt_read_fw_status_reg(bp, BNXT_FW_RECOVERY_CNT_REG);
4228         if (val != info->last_reset_counter)
4229                 goto reset;
4230
4231         info->last_reset_counter = val;
4232
4233         rte_eal_alarm_set(US_PER_MS * info->driver_polling_freq,
4234                           bnxt_check_fw_health, (void *)bp);
4235
4236         return;
4237 reset:
4238         /* Stop DMA to/from device */
4239         bp->flags |= BNXT_FLAG_FATAL_ERROR;
4240         bp->flags |= BNXT_FLAG_FW_RESET;
4241
4242         PMD_DRV_LOG(ERR, "Detected FW dead condition\n");
4243
4244         if (bnxt_is_master_func(bp))
4245                 wait_msec = info->master_func_wait_period;
4246         else
4247                 wait_msec = info->normal_func_wait_period;
4248
4249         rte_eal_alarm_set(US_PER_MS * wait_msec,
4250                           bnxt_fw_reset_cb, (void *)bp);
4251 }
4252
4253 void bnxt_schedule_fw_health_check(struct bnxt *bp)
4254 {
4255         uint32_t polling_freq;
4256
4257         pthread_mutex_lock(&bp->health_check_lock);
4258
4259         if (!bnxt_is_recovery_enabled(bp))
4260                 goto done;
4261
4262         if (bp->flags & BNXT_FLAG_FW_HEALTH_CHECK_SCHEDULED)
4263                 goto done;
4264
4265         polling_freq = bp->recovery_info->driver_polling_freq;
4266
4267         rte_eal_alarm_set(US_PER_MS * polling_freq,
4268                           bnxt_check_fw_health, (void *)bp);
4269         bp->flags |= BNXT_FLAG_FW_HEALTH_CHECK_SCHEDULED;
4270
4271 done:
4272         pthread_mutex_unlock(&bp->health_check_lock);
4273 }
4274
4275 static void bnxt_cancel_fw_health_check(struct bnxt *bp)
4276 {
4277         rte_eal_alarm_cancel(bnxt_check_fw_health, (void *)bp);
4278         bp->flags &= ~BNXT_FLAG_FW_HEALTH_CHECK_SCHEDULED;
4279 }
4280
4281 static bool bnxt_vf_pciid(uint16_t device_id)
4282 {
4283         switch (device_id) {
4284         case BROADCOM_DEV_ID_57304_VF:
4285         case BROADCOM_DEV_ID_57406_VF:
4286         case BROADCOM_DEV_ID_5731X_VF:
4287         case BROADCOM_DEV_ID_5741X_VF:
4288         case BROADCOM_DEV_ID_57414_VF:
4289         case BROADCOM_DEV_ID_STRATUS_NIC_VF1:
4290         case BROADCOM_DEV_ID_STRATUS_NIC_VF2:
4291         case BROADCOM_DEV_ID_58802_VF:
4292         case BROADCOM_DEV_ID_57500_VF1:
4293         case BROADCOM_DEV_ID_57500_VF2:
4294         case BROADCOM_DEV_ID_58818_VF:
4295                 /* FALLTHROUGH */
4296                 return true;
4297         default:
4298                 return false;
4299         }
4300 }
4301
4302 /* Phase 5 device */
4303 static bool bnxt_p5_device(uint16_t device_id)
4304 {
4305         switch (device_id) {
4306         case BROADCOM_DEV_ID_57508:
4307         case BROADCOM_DEV_ID_57504:
4308         case BROADCOM_DEV_ID_57502:
4309         case BROADCOM_DEV_ID_57508_MF1:
4310         case BROADCOM_DEV_ID_57504_MF1:
4311         case BROADCOM_DEV_ID_57502_MF1:
4312         case BROADCOM_DEV_ID_57508_MF2:
4313         case BROADCOM_DEV_ID_57504_MF2:
4314         case BROADCOM_DEV_ID_57502_MF2:
4315         case BROADCOM_DEV_ID_57500_VF1:
4316         case BROADCOM_DEV_ID_57500_VF2:
4317         case BROADCOM_DEV_ID_58812:
4318         case BROADCOM_DEV_ID_58814:
4319         case BROADCOM_DEV_ID_58818:
4320         case BROADCOM_DEV_ID_58818_VF:
4321                 /* FALLTHROUGH */
4322                 return true;
4323         default:
4324                 return false;
4325         }
4326 }
4327
4328 bool bnxt_stratus_device(struct bnxt *bp)
4329 {
4330         uint16_t device_id = bp->pdev->id.device_id;
4331
4332         switch (device_id) {
4333         case BROADCOM_DEV_ID_STRATUS_NIC:
4334         case BROADCOM_DEV_ID_STRATUS_NIC_VF1:
4335         case BROADCOM_DEV_ID_STRATUS_NIC_VF2:
4336                 /* FALLTHROUGH */
4337                 return true;
4338         default:
4339                 return false;
4340         }
4341 }
4342
4343 static int bnxt_map_pci_bars(struct rte_eth_dev *eth_dev)
4344 {
4345         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
4346         struct bnxt *bp = eth_dev->data->dev_private;
4347
4348         /* enable device (incl. PCI PM wakeup), and bus-mastering */
4349         bp->bar0 = (void *)pci_dev->mem_resource[0].addr;
4350         bp->doorbell_base = (void *)pci_dev->mem_resource[2].addr;
4351         if (!bp->bar0 || !bp->doorbell_base) {
4352                 PMD_DRV_LOG(ERR, "Unable to access Hardware\n");
4353                 return -ENODEV;
4354         }
4355
4356         bp->eth_dev = eth_dev;
4357         bp->pdev = pci_dev;
4358
4359         return 0;
4360 }
4361
4362 static int bnxt_alloc_ctx_mem_blk(struct bnxt *bp,
4363                                   struct bnxt_ctx_pg_info *ctx_pg,
4364                                   uint32_t mem_size,
4365                                   const char *suffix,
4366                                   uint16_t idx)
4367 {
4368         struct bnxt_ring_mem_info *rmem = &ctx_pg->ring_mem;
4369         const struct rte_memzone *mz = NULL;
4370         char mz_name[RTE_MEMZONE_NAMESIZE];
4371         rte_iova_t mz_phys_addr;
4372         uint64_t valid_bits = 0;
4373         uint32_t sz;
4374         int i;
4375
4376         if (!mem_size)
4377                 return 0;
4378
4379         rmem->nr_pages = RTE_ALIGN_MUL_CEIL(mem_size, BNXT_PAGE_SIZE) /
4380                          BNXT_PAGE_SIZE;
4381         rmem->page_size = BNXT_PAGE_SIZE;
4382         rmem->pg_arr = ctx_pg->ctx_pg_arr;
4383         rmem->dma_arr = ctx_pg->ctx_dma_arr;
4384         rmem->flags = BNXT_RMEM_VALID_PTE_FLAG;
4385
4386         valid_bits = PTU_PTE_VALID;
4387
4388         if (rmem->nr_pages > 1) {
4389                 snprintf(mz_name, RTE_MEMZONE_NAMESIZE,
4390                          "bnxt_ctx_pg_tbl%s_%x_%d",
4391                          suffix, idx, bp->eth_dev->data->port_id);
4392                 mz_name[RTE_MEMZONE_NAMESIZE - 1] = 0;
4393                 mz = rte_memzone_lookup(mz_name);
4394                 if (!mz) {
4395                         mz = rte_memzone_reserve_aligned(mz_name,
4396                                                 rmem->nr_pages * 8,
4397                                                 SOCKET_ID_ANY,
4398                                                 RTE_MEMZONE_2MB |
4399                                                 RTE_MEMZONE_SIZE_HINT_ONLY |
4400                                                 RTE_MEMZONE_IOVA_CONTIG,
4401                                                 BNXT_PAGE_SIZE);
4402                         if (mz == NULL)
4403                                 return -ENOMEM;
4404                 }
4405
4406                 memset(mz->addr, 0, mz->len);
4407                 mz_phys_addr = mz->iova;
4408
4409                 rmem->pg_tbl = mz->addr;
4410                 rmem->pg_tbl_map = mz_phys_addr;
4411                 rmem->pg_tbl_mz = mz;
4412         }
4413
4414         snprintf(mz_name, RTE_MEMZONE_NAMESIZE, "bnxt_ctx_%s_%x_%d",
4415                  suffix, idx, bp->eth_dev->data->port_id);
4416         mz = rte_memzone_lookup(mz_name);
4417         if (!mz) {
4418                 mz = rte_memzone_reserve_aligned(mz_name,
4419                                                  mem_size,
4420                                                  SOCKET_ID_ANY,
4421                                                  RTE_MEMZONE_1GB |
4422                                                  RTE_MEMZONE_SIZE_HINT_ONLY |
4423                                                  RTE_MEMZONE_IOVA_CONTIG,
4424                                                  BNXT_PAGE_SIZE);
4425                 if (mz == NULL)
4426                         return -ENOMEM;
4427         }
4428
4429         memset(mz->addr, 0, mz->len);
4430         mz_phys_addr = mz->iova;
4431
4432         for (sz = 0, i = 0; sz < mem_size; sz += BNXT_PAGE_SIZE, i++) {
4433                 rmem->pg_arr[i] = ((char *)mz->addr) + sz;
4434                 rmem->dma_arr[i] = mz_phys_addr + sz;
4435
4436                 if (rmem->nr_pages > 1) {
4437                         if (i == rmem->nr_pages - 2 &&
4438                             (rmem->flags & BNXT_RMEM_RING_PTE_FLAG))
4439                                 valid_bits |= PTU_PTE_NEXT_TO_LAST;
4440                         else if (i == rmem->nr_pages - 1 &&
4441                                  (rmem->flags & BNXT_RMEM_RING_PTE_FLAG))
4442                                 valid_bits |= PTU_PTE_LAST;
4443
4444                         rmem->pg_tbl[i] = rte_cpu_to_le_64(rmem->dma_arr[i] |
4445                                                            valid_bits);
4446                 }
4447         }
4448
4449         rmem->mz = mz;
4450         if (rmem->vmem_size)
4451                 rmem->vmem = (void **)mz->addr;
4452         rmem->dma_arr[0] = mz_phys_addr;
4453         return 0;
4454 }
4455
4456 static void bnxt_free_ctx_mem(struct bnxt *bp)
4457 {
4458         int i;
4459
4460         if (!bp->ctx || !(bp->ctx->flags & BNXT_CTX_FLAG_INITED))
4461                 return;
4462
4463         bp->ctx->flags &= ~BNXT_CTX_FLAG_INITED;
4464         rte_memzone_free(bp->ctx->qp_mem.ring_mem.mz);
4465         rte_memzone_free(bp->ctx->srq_mem.ring_mem.mz);
4466         rte_memzone_free(bp->ctx->cq_mem.ring_mem.mz);
4467         rte_memzone_free(bp->ctx->vnic_mem.ring_mem.mz);
4468         rte_memzone_free(bp->ctx->stat_mem.ring_mem.mz);
4469         rte_memzone_free(bp->ctx->qp_mem.ring_mem.pg_tbl_mz);
4470         rte_memzone_free(bp->ctx->srq_mem.ring_mem.pg_tbl_mz);
4471         rte_memzone_free(bp->ctx->cq_mem.ring_mem.pg_tbl_mz);
4472         rte_memzone_free(bp->ctx->vnic_mem.ring_mem.pg_tbl_mz);
4473         rte_memzone_free(bp->ctx->stat_mem.ring_mem.pg_tbl_mz);
4474
4475         for (i = 0; i < bp->ctx->tqm_fp_rings_count + 1; i++) {
4476                 if (bp->ctx->tqm_mem[i])
4477                         rte_memzone_free(bp->ctx->tqm_mem[i]->ring_mem.mz);
4478         }
4479
4480         rte_free(bp->ctx);
4481         bp->ctx = NULL;
4482 }
4483
4484 #define bnxt_roundup(x, y)   ((((x) + ((y) - 1)) / (y)) * (y))
4485
4486 #define min_t(type, x, y) ({                    \
4487         type __min1 = (x);                      \
4488         type __min2 = (y);                      \
4489         __min1 < __min2 ? __min1 : __min2; })
4490
4491 #define max_t(type, x, y) ({                    \
4492         type __max1 = (x);                      \
4493         type __max2 = (y);                      \
4494         __max1 > __max2 ? __max1 : __max2; })
4495
4496 #define clamp_t(type, _x, min, max)     min_t(type, max_t(type, _x, min), max)
4497
4498 int bnxt_alloc_ctx_mem(struct bnxt *bp)
4499 {
4500         struct bnxt_ctx_pg_info *ctx_pg;
4501         struct bnxt_ctx_mem_info *ctx;
4502         uint32_t mem_size, ena, entries;
4503         uint32_t entries_sp, min;
4504         int i, rc;
4505
4506         rc = bnxt_hwrm_func_backing_store_qcaps(bp);
4507         if (rc) {
4508                 PMD_DRV_LOG(ERR, "Query context mem capability failed\n");
4509                 return rc;
4510         }
4511         ctx = bp->ctx;
4512         if (!ctx || (ctx->flags & BNXT_CTX_FLAG_INITED))
4513                 return 0;
4514
4515         ctx_pg = &ctx->qp_mem;
4516         ctx_pg->entries = ctx->qp_min_qp1_entries + ctx->qp_max_l2_entries;
4517         if (ctx->qp_entry_size) {
4518                 mem_size = ctx->qp_entry_size * ctx_pg->entries;
4519                 rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "qp_mem", 0);
4520                 if (rc)
4521                         return rc;
4522         }
4523
4524         ctx_pg = &ctx->srq_mem;
4525         ctx_pg->entries = ctx->srq_max_l2_entries;
4526         if (ctx->srq_entry_size) {
4527                 mem_size = ctx->srq_entry_size * ctx_pg->entries;
4528                 rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "srq_mem", 0);
4529                 if (rc)
4530                         return rc;
4531         }
4532
4533         ctx_pg = &ctx->cq_mem;
4534         ctx_pg->entries = ctx->cq_max_l2_entries;
4535         if (ctx->cq_entry_size) {
4536                 mem_size = ctx->cq_entry_size * ctx_pg->entries;
4537                 rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "cq_mem", 0);
4538                 if (rc)
4539                         return rc;
4540         }
4541
4542         ctx_pg = &ctx->vnic_mem;
4543         ctx_pg->entries = ctx->vnic_max_vnic_entries +
4544                 ctx->vnic_max_ring_table_entries;
4545         if (ctx->vnic_entry_size) {
4546                 mem_size = ctx->vnic_entry_size * ctx_pg->entries;
4547                 rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "vnic_mem", 0);
4548                 if (rc)
4549                         return rc;
4550         }
4551
4552         ctx_pg = &ctx->stat_mem;
4553         ctx_pg->entries = ctx->stat_max_entries;
4554         if (ctx->stat_entry_size) {
4555                 mem_size = ctx->stat_entry_size * ctx_pg->entries;
4556                 rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "stat_mem", 0);
4557                 if (rc)
4558                         return rc;
4559         }
4560
4561         min = ctx->tqm_min_entries_per_ring;
4562
4563         entries_sp = ctx->qp_max_l2_entries +
4564                      ctx->vnic_max_vnic_entries +
4565                      2 * ctx->qp_min_qp1_entries + min;
4566         entries_sp = bnxt_roundup(entries_sp, ctx->tqm_entries_multiple);
4567
4568         entries = ctx->qp_max_l2_entries + ctx->qp_min_qp1_entries;
4569         entries = bnxt_roundup(entries, ctx->tqm_entries_multiple);
4570         entries = clamp_t(uint32_t, entries, min,
4571                           ctx->tqm_max_entries_per_ring);
4572         for (i = 0, ena = 0; i < ctx->tqm_fp_rings_count + 1; i++) {
4573                 /* i=0 is for TQM_SP. i=1 to i=8 applies to RING0 to RING7.
4574                  * i > 8 is other ext rings.
4575                  */
4576                 ctx_pg = ctx->tqm_mem[i];
4577                 ctx_pg->entries = i ? entries : entries_sp;
4578                 if (ctx->tqm_entry_size) {
4579                         mem_size = ctx->tqm_entry_size * ctx_pg->entries;
4580                         rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size,
4581                                                     "tqm_mem", i);
4582                         if (rc)
4583                                 return rc;
4584                 }
4585                 if (i < BNXT_MAX_TQM_LEGACY_RINGS)
4586                         ena |= HWRM_FUNC_BACKING_STORE_CFG_INPUT_ENABLES_TQM_SP << i;
4587                 else
4588                         ena |= HWRM_FUNC_BACKING_STORE_CFG_INPUT_ENABLES_TQM_RING8;
4589         }
4590
4591         ena |= FUNC_BACKING_STORE_CFG_INPUT_DFLT_ENABLES;
4592         rc = bnxt_hwrm_func_backing_store_cfg(bp, ena);
4593         if (rc)
4594                 PMD_DRV_LOG(ERR,
4595                             "Failed to configure context mem: rc = %d\n", rc);
4596         else
4597                 ctx->flags |= BNXT_CTX_FLAG_INITED;
4598
4599         return rc;
4600 }
4601
4602 static int bnxt_alloc_stats_mem(struct bnxt *bp)
4603 {
4604         struct rte_pci_device *pci_dev = bp->pdev;
4605         char mz_name[RTE_MEMZONE_NAMESIZE];
4606         const struct rte_memzone *mz = NULL;
4607         uint32_t total_alloc_len;
4608         rte_iova_t mz_phys_addr;
4609
4610         if (pci_dev->id.device_id == BROADCOM_DEV_ID_NS2)
4611                 return 0;
4612
4613         snprintf(mz_name, RTE_MEMZONE_NAMESIZE,
4614                  "bnxt_" PCI_PRI_FMT "-%s", pci_dev->addr.domain,
4615                  pci_dev->addr.bus, pci_dev->addr.devid,
4616                  pci_dev->addr.function, "rx_port_stats");
4617         mz_name[RTE_MEMZONE_NAMESIZE - 1] = 0;
4618         mz = rte_memzone_lookup(mz_name);
4619         total_alloc_len =
4620                 RTE_CACHE_LINE_ROUNDUP(sizeof(struct rx_port_stats) +
4621                                        sizeof(struct rx_port_stats_ext) + 512);
4622         if (!mz) {
4623                 mz = rte_memzone_reserve(mz_name, total_alloc_len,
4624                                          SOCKET_ID_ANY,
4625                                          RTE_MEMZONE_2MB |
4626                                          RTE_MEMZONE_SIZE_HINT_ONLY |
4627                                          RTE_MEMZONE_IOVA_CONTIG);
4628                 if (mz == NULL)
4629                         return -ENOMEM;
4630         }
4631         memset(mz->addr, 0, mz->len);
4632         mz_phys_addr = mz->iova;
4633
4634         bp->rx_mem_zone = (const void *)mz;
4635         bp->hw_rx_port_stats = mz->addr;
4636         bp->hw_rx_port_stats_map = mz_phys_addr;
4637
4638         snprintf(mz_name, RTE_MEMZONE_NAMESIZE,
4639                  "bnxt_" PCI_PRI_FMT "-%s", pci_dev->addr.domain,
4640                  pci_dev->addr.bus, pci_dev->addr.devid,
4641                  pci_dev->addr.function, "tx_port_stats");
4642         mz_name[RTE_MEMZONE_NAMESIZE - 1] = 0;
4643         mz = rte_memzone_lookup(mz_name);
4644         total_alloc_len =
4645                 RTE_CACHE_LINE_ROUNDUP(sizeof(struct tx_port_stats) +
4646                                        sizeof(struct tx_port_stats_ext) + 512);
4647         if (!mz) {
4648                 mz = rte_memzone_reserve(mz_name,
4649                                          total_alloc_len,
4650                                          SOCKET_ID_ANY,
4651                                          RTE_MEMZONE_2MB |
4652                                          RTE_MEMZONE_SIZE_HINT_ONLY |
4653                                          RTE_MEMZONE_IOVA_CONTIG);
4654                 if (mz == NULL)
4655                         return -ENOMEM;
4656         }
4657         memset(mz->addr, 0, mz->len);
4658         mz_phys_addr = mz->iova;
4659
4660         bp->tx_mem_zone = (const void *)mz;
4661         bp->hw_tx_port_stats = mz->addr;
4662         bp->hw_tx_port_stats_map = mz_phys_addr;
4663         bp->flags |= BNXT_FLAG_PORT_STATS;
4664
4665         /* Display extended statistics if FW supports it */
4666         if (bp->hwrm_spec_code < HWRM_SPEC_CODE_1_8_4 ||
4667             bp->hwrm_spec_code == HWRM_SPEC_CODE_1_9_0 ||
4668             !(bp->flags & BNXT_FLAG_EXT_STATS_SUPPORTED))
4669                 return 0;
4670
4671         bp->hw_rx_port_stats_ext = (void *)
4672                 ((uint8_t *)bp->hw_rx_port_stats +
4673                  sizeof(struct rx_port_stats));
4674         bp->hw_rx_port_stats_ext_map = bp->hw_rx_port_stats_map +
4675                 sizeof(struct rx_port_stats);
4676         bp->flags |= BNXT_FLAG_EXT_RX_PORT_STATS;
4677
4678         if (bp->hwrm_spec_code < HWRM_SPEC_CODE_1_9_2 ||
4679             bp->flags & BNXT_FLAG_EXT_STATS_SUPPORTED) {
4680                 bp->hw_tx_port_stats_ext = (void *)
4681                         ((uint8_t *)bp->hw_tx_port_stats +
4682                          sizeof(struct tx_port_stats));
4683                 bp->hw_tx_port_stats_ext_map =
4684                         bp->hw_tx_port_stats_map +
4685                         sizeof(struct tx_port_stats);
4686                 bp->flags |= BNXT_FLAG_EXT_TX_PORT_STATS;
4687         }
4688
4689         return 0;
4690 }
4691
4692 static int bnxt_setup_mac_addr(struct rte_eth_dev *eth_dev)
4693 {
4694         struct bnxt *bp = eth_dev->data->dev_private;
4695         int rc = 0;
4696
4697         eth_dev->data->mac_addrs = rte_zmalloc("bnxt_mac_addr_tbl",
4698                                                RTE_ETHER_ADDR_LEN *
4699                                                bp->max_l2_ctx,
4700                                                0);
4701         if (eth_dev->data->mac_addrs == NULL) {
4702                 PMD_DRV_LOG(ERR, "Failed to alloc MAC addr tbl\n");
4703                 return -ENOMEM;
4704         }
4705
4706         if (!BNXT_HAS_DFLT_MAC_SET(bp)) {
4707                 if (BNXT_PF(bp))
4708                         return -EINVAL;
4709
4710                 /* Generate a random MAC address, if none was assigned by PF */
4711                 PMD_DRV_LOG(INFO, "VF MAC address not assigned by Host PF\n");
4712                 bnxt_eth_hw_addr_random(bp->mac_addr);
4713                 PMD_DRV_LOG(INFO,
4714                             "Assign random MAC:%02X:%02X:%02X:%02X:%02X:%02X\n",
4715                             bp->mac_addr[0], bp->mac_addr[1], bp->mac_addr[2],
4716                             bp->mac_addr[3], bp->mac_addr[4], bp->mac_addr[5]);
4717
4718                 rc = bnxt_hwrm_set_mac(bp);
4719                 if (rc)
4720                         return rc;
4721         }
4722
4723         /* Copy the permanent MAC from the FUNC_QCAPS response */
4724         memcpy(&eth_dev->data->mac_addrs[0], bp->mac_addr, RTE_ETHER_ADDR_LEN);
4725
4726         return rc;
4727 }
4728
4729 static int bnxt_restore_dflt_mac(struct bnxt *bp)
4730 {
4731         int rc = 0;
4732
4733         /* MAC is already configured in FW */
4734         if (BNXT_HAS_DFLT_MAC_SET(bp))
4735                 return 0;
4736
4737         /* Restore the old MAC configured */
4738         rc = bnxt_hwrm_set_mac(bp);
4739         if (rc)
4740                 PMD_DRV_LOG(ERR, "Failed to restore MAC address\n");
4741
4742         return rc;
4743 }
4744
4745 static void bnxt_config_vf_req_fwd(struct bnxt *bp)
4746 {
4747         if (!BNXT_PF(bp))
4748                 return;
4749
4750         memset(bp->pf->vf_req_fwd, 0, sizeof(bp->pf->vf_req_fwd));
4751
4752         if (!(bp->fw_cap & BNXT_FW_CAP_LINK_ADMIN))
4753                 BNXT_HWRM_CMD_TO_FORWARD(HWRM_PORT_PHY_QCFG);
4754         BNXT_HWRM_CMD_TO_FORWARD(HWRM_FUNC_CFG);
4755         BNXT_HWRM_CMD_TO_FORWARD(HWRM_FUNC_VF_CFG);
4756         BNXT_HWRM_CMD_TO_FORWARD(HWRM_CFA_L2_FILTER_ALLOC);
4757         BNXT_HWRM_CMD_TO_FORWARD(HWRM_OEM_CMD);
4758 }
4759
4760 uint16_t
4761 bnxt_get_svif(uint16_t port_id, bool func_svif,
4762               enum bnxt_ulp_intf_type type)
4763 {
4764         struct rte_eth_dev *eth_dev;
4765         struct bnxt *bp;
4766
4767         eth_dev = &rte_eth_devices[port_id];
4768         if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
4769                 struct bnxt_representor *vfr = eth_dev->data->dev_private;
4770                 if (!vfr)
4771                         return 0;
4772
4773                 if (type == BNXT_ULP_INTF_TYPE_VF_REP)
4774                         return vfr->svif;
4775
4776                 eth_dev = vfr->parent_dev;
4777         }
4778
4779         bp = eth_dev->data->dev_private;
4780
4781         return func_svif ? bp->func_svif : bp->port_svif;
4782 }
4783
4784 uint16_t
4785 bnxt_get_vnic_id(uint16_t port, enum bnxt_ulp_intf_type type)
4786 {
4787         struct rte_eth_dev *eth_dev;
4788         struct bnxt_vnic_info *vnic;
4789         struct bnxt *bp;
4790
4791         eth_dev = &rte_eth_devices[port];
4792         if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
4793                 struct bnxt_representor *vfr = eth_dev->data->dev_private;
4794                 if (!vfr)
4795                         return 0;
4796
4797                 if (type == BNXT_ULP_INTF_TYPE_VF_REP)
4798                         return vfr->dflt_vnic_id;
4799
4800                 eth_dev = vfr->parent_dev;
4801         }
4802
4803         bp = eth_dev->data->dev_private;
4804
4805         vnic = BNXT_GET_DEFAULT_VNIC(bp);
4806
4807         return vnic->fw_vnic_id;
4808 }
4809
4810 uint16_t
4811 bnxt_get_fw_func_id(uint16_t port, enum bnxt_ulp_intf_type type)
4812 {
4813         struct rte_eth_dev *eth_dev;
4814         struct bnxt *bp;
4815
4816         eth_dev = &rte_eth_devices[port];
4817         if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
4818                 struct bnxt_representor *vfr = eth_dev->data->dev_private;
4819                 if (!vfr)
4820                         return 0;
4821
4822                 if (type == BNXT_ULP_INTF_TYPE_VF_REP)
4823                         return vfr->fw_fid;
4824
4825                 eth_dev = vfr->parent_dev;
4826         }
4827
4828         bp = eth_dev->data->dev_private;
4829
4830         return bp->fw_fid;
4831 }
4832
4833 enum bnxt_ulp_intf_type
4834 bnxt_get_interface_type(uint16_t port)
4835 {
4836         struct rte_eth_dev *eth_dev;
4837         struct bnxt *bp;
4838
4839         eth_dev = &rte_eth_devices[port];
4840         if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev))
4841                 return BNXT_ULP_INTF_TYPE_VF_REP;
4842
4843         bp = eth_dev->data->dev_private;
4844         if (BNXT_PF(bp))
4845                 return BNXT_ULP_INTF_TYPE_PF;
4846         else if (BNXT_VF_IS_TRUSTED(bp))
4847                 return BNXT_ULP_INTF_TYPE_TRUSTED_VF;
4848         else if (BNXT_VF(bp))
4849                 return BNXT_ULP_INTF_TYPE_VF;
4850
4851         return BNXT_ULP_INTF_TYPE_INVALID;
4852 }
4853
4854 uint16_t
4855 bnxt_get_phy_port_id(uint16_t port_id)
4856 {
4857         struct bnxt_representor *vfr;
4858         struct rte_eth_dev *eth_dev;
4859         struct bnxt *bp;
4860
4861         eth_dev = &rte_eth_devices[port_id];
4862         if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
4863                 vfr = eth_dev->data->dev_private;
4864                 if (!vfr)
4865                         return 0;
4866
4867                 eth_dev = vfr->parent_dev;
4868         }
4869
4870         bp = eth_dev->data->dev_private;
4871
4872         return BNXT_PF(bp) ? bp->pf->port_id : bp->parent->port_id;
4873 }
4874
4875 uint16_t
4876 bnxt_get_parif(uint16_t port_id, enum bnxt_ulp_intf_type type)
4877 {
4878         struct rte_eth_dev *eth_dev;
4879         struct bnxt *bp;
4880
4881         eth_dev = &rte_eth_devices[port_id];
4882         if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
4883                 struct bnxt_representor *vfr = eth_dev->data->dev_private;
4884                 if (!vfr)
4885                         return 0;
4886
4887                 if (type == BNXT_ULP_INTF_TYPE_VF_REP)
4888                         return vfr->fw_fid - 1;
4889
4890                 eth_dev = vfr->parent_dev;
4891         }
4892
4893         bp = eth_dev->data->dev_private;
4894
4895         return BNXT_PF(bp) ? bp->fw_fid - 1 : bp->parent->fid - 1;
4896 }
4897
4898 uint16_t
4899 bnxt_get_vport(uint16_t port_id)
4900 {
4901         return (1 << bnxt_get_phy_port_id(port_id));
4902 }
4903
4904 static void bnxt_alloc_error_recovery_info(struct bnxt *bp)
4905 {
4906         struct bnxt_error_recovery_info *info = bp->recovery_info;
4907
4908         if (info) {
4909                 if (!(bp->fw_cap & BNXT_FW_CAP_HCOMM_FW_STATUS))
4910                         memset(info, 0, sizeof(*info));
4911                 return;
4912         }
4913
4914         if (!(bp->fw_cap & BNXT_FW_CAP_ERROR_RECOVERY))
4915                 return;
4916
4917         info = rte_zmalloc("bnxt_hwrm_error_recovery_qcfg",
4918                            sizeof(*info), 0);
4919         if (!info)
4920                 bp->fw_cap &= ~BNXT_FW_CAP_ERROR_RECOVERY;
4921
4922         bp->recovery_info = info;
4923 }
4924
4925 static void bnxt_check_fw_status(struct bnxt *bp)
4926 {
4927         uint32_t fw_status;
4928
4929         if (!(bp->recovery_info &&
4930               (bp->fw_cap & BNXT_FW_CAP_HCOMM_FW_STATUS)))
4931                 return;
4932
4933         fw_status = bnxt_read_fw_status_reg(bp, BNXT_FW_STATUS_REG);
4934         if (fw_status != BNXT_FW_STATUS_HEALTHY)
4935                 PMD_DRV_LOG(ERR, "Firmware not responding, status: %#x\n",
4936                             fw_status);
4937 }
4938
4939 static int bnxt_map_hcomm_fw_status_reg(struct bnxt *bp)
4940 {
4941         struct bnxt_error_recovery_info *info = bp->recovery_info;
4942         uint32_t status_loc;
4943         uint32_t sig_ver;
4944
4945         rte_write32(HCOMM_STATUS_STRUCT_LOC, (uint8_t *)bp->bar0 +
4946                     BNXT_GRCPF_REG_WINDOW_BASE_OUT + 4);
4947         sig_ver = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
4948                                    BNXT_GRCP_WINDOW_2_BASE +
4949                                    offsetof(struct hcomm_status,
4950                                             sig_ver)));
4951         /* If the signature is absent, then FW does not support this feature */
4952         if ((sig_ver & HCOMM_STATUS_SIGNATURE_MASK) !=
4953             HCOMM_STATUS_SIGNATURE_VAL)
4954                 return 0;
4955
4956         if (!info) {
4957                 info = rte_zmalloc("bnxt_hwrm_error_recovery_qcfg",
4958                                    sizeof(*info), 0);
4959                 if (!info)
4960                         return -ENOMEM;
4961                 bp->recovery_info = info;
4962         } else {
4963                 memset(info, 0, sizeof(*info));
4964         }
4965
4966         status_loc = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
4967                                       BNXT_GRCP_WINDOW_2_BASE +
4968                                       offsetof(struct hcomm_status,
4969                                                fw_status_loc)));
4970
4971         /* Only pre-map the FW health status GRC register */
4972         if (BNXT_FW_STATUS_REG_TYPE(status_loc) != BNXT_FW_STATUS_REG_TYPE_GRC)
4973                 return 0;
4974
4975         info->status_regs[BNXT_FW_STATUS_REG] = status_loc;
4976         info->mapped_status_regs[BNXT_FW_STATUS_REG] =
4977                 BNXT_GRCP_WINDOW_2_BASE + (status_loc & BNXT_GRCP_OFFSET_MASK);
4978
4979         rte_write32((status_loc & BNXT_GRCP_BASE_MASK), (uint8_t *)bp->bar0 +
4980                     BNXT_GRCPF_REG_WINDOW_BASE_OUT + 4);
4981
4982         bp->fw_cap |= BNXT_FW_CAP_HCOMM_FW_STATUS;
4983
4984         return 0;
4985 }
4986
4987 /* This function gets the FW version along with the
4988  * capabilities(MAX and current) of the function, vnic,
4989  * error recovery, phy and other chip related info
4990  */
4991 static int bnxt_get_config(struct bnxt *bp)
4992 {
4993         uint16_t mtu;
4994         int rc = 0;
4995
4996         bp->fw_cap = 0;
4997
4998         rc = bnxt_map_hcomm_fw_status_reg(bp);
4999         if (rc)
5000                 return rc;
5001
5002         rc = bnxt_hwrm_ver_get(bp, DFLT_HWRM_CMD_TIMEOUT);
5003         if (rc) {
5004                 bnxt_check_fw_status(bp);
5005                 return rc;
5006         }
5007
5008         rc = bnxt_hwrm_func_reset(bp);
5009         if (rc)
5010                 return -EIO;
5011
5012         rc = bnxt_hwrm_vnic_qcaps(bp);
5013         if (rc)
5014                 return rc;
5015
5016         rc = bnxt_hwrm_queue_qportcfg(bp);
5017         if (rc)
5018                 return rc;
5019
5020         /* Get the MAX capabilities for this function.
5021          * This function also allocates context memory for TQM rings and
5022          * informs the firmware about this allocated backing store memory.
5023          */
5024         rc = bnxt_hwrm_func_qcaps(bp);
5025         if (rc)
5026                 return rc;
5027
5028         rc = bnxt_hwrm_func_qcfg(bp, &mtu);
5029         if (rc)
5030                 return rc;
5031
5032         rc = bnxt_hwrm_cfa_adv_flow_mgmt_qcaps(bp);
5033         if (rc)
5034                 return rc;
5035
5036         bnxt_hwrm_port_mac_qcfg(bp);
5037
5038         bnxt_hwrm_parent_pf_qcfg(bp);
5039
5040         bnxt_hwrm_port_phy_qcaps(bp);
5041
5042         bnxt_alloc_error_recovery_info(bp);
5043         /* Get the adapter error recovery support info */
5044         rc = bnxt_hwrm_error_recovery_qcfg(bp);
5045         if (rc)
5046                 bp->fw_cap &= ~BNXT_FW_CAP_ERROR_RECOVERY;
5047
5048         bnxt_hwrm_port_led_qcaps(bp);
5049
5050         return 0;
5051 }
5052
5053 static int
5054 bnxt_init_locks(struct bnxt *bp)
5055 {
5056         int err;
5057
5058         err = pthread_mutex_init(&bp->flow_lock, NULL);
5059         if (err) {
5060                 PMD_DRV_LOG(ERR, "Unable to initialize flow_lock\n");
5061                 return err;
5062         }
5063
5064         err = pthread_mutex_init(&bp->def_cp_lock, NULL);
5065         if (err) {
5066                 PMD_DRV_LOG(ERR, "Unable to initialize def_cp_lock\n");
5067                 return err;
5068         }
5069
5070         err = pthread_mutex_init(&bp->health_check_lock, NULL);
5071         if (err) {
5072                 PMD_DRV_LOG(ERR, "Unable to initialize health_check_lock\n");
5073                 return err;
5074         }
5075
5076         err = pthread_mutex_init(&bp->err_recovery_lock, NULL);
5077         if (err)
5078                 PMD_DRV_LOG(ERR, "Unable to initialize err_recovery_lock\n");
5079
5080         return err;
5081 }
5082
5083 static int bnxt_init_resources(struct bnxt *bp, bool reconfig_dev)
5084 {
5085         int rc = 0;
5086
5087         rc = bnxt_get_config(bp);
5088         if (rc)
5089                 return rc;
5090
5091         if (!reconfig_dev) {
5092                 rc = bnxt_setup_mac_addr(bp->eth_dev);
5093                 if (rc)
5094                         return rc;
5095         } else {
5096                 rc = bnxt_restore_dflt_mac(bp);
5097                 if (rc)
5098                         return rc;
5099         }
5100
5101         bnxt_config_vf_req_fwd(bp);
5102
5103         rc = bnxt_hwrm_func_driver_register(bp);
5104         if (rc) {
5105                 PMD_DRV_LOG(ERR, "Failed to register driver");
5106                 return -EBUSY;
5107         }
5108
5109         if (BNXT_PF(bp)) {
5110                 if (bp->pdev->max_vfs) {
5111                         rc = bnxt_hwrm_allocate_vfs(bp, bp->pdev->max_vfs);
5112                         if (rc) {
5113                                 PMD_DRV_LOG(ERR, "Failed to allocate VFs\n");
5114                                 return rc;
5115                         }
5116                 } else {
5117                         rc = bnxt_hwrm_allocate_pf_only(bp);
5118                         if (rc) {
5119                                 PMD_DRV_LOG(ERR,
5120                                             "Failed to allocate PF resources");
5121                                 return rc;
5122                         }
5123                 }
5124         }
5125
5126         rc = bnxt_alloc_mem(bp, reconfig_dev);
5127         if (rc)
5128                 return rc;
5129
5130         rc = bnxt_setup_int(bp);
5131         if (rc)
5132                 return rc;
5133
5134         rc = bnxt_request_int(bp);
5135         if (rc)
5136                 return rc;
5137
5138         rc = bnxt_init_ctx_mem(bp);
5139         if (rc) {
5140                 PMD_DRV_LOG(ERR, "Failed to init adv_flow_counters\n");
5141                 return rc;
5142         }
5143
5144         return 0;
5145 }
5146
5147 static int
5148 bnxt_parse_devarg_truflow(__rte_unused const char *key,
5149                           const char *value, void *opaque_arg)
5150 {
5151         struct bnxt *bp = opaque_arg;
5152         unsigned long truflow;
5153         char *end = NULL;
5154
5155         if (!value || !opaque_arg) {
5156                 PMD_DRV_LOG(ERR,
5157                             "Invalid parameter passed to truflow devargs.\n");
5158                 return -EINVAL;
5159         }
5160
5161         truflow = strtoul(value, &end, 10);
5162         if (end == NULL || *end != '\0' ||
5163             (truflow == ULONG_MAX && errno == ERANGE)) {
5164                 PMD_DRV_LOG(ERR,
5165                             "Invalid parameter passed to truflow devargs.\n");
5166                 return -EINVAL;
5167         }
5168
5169         if (BNXT_DEVARG_TRUFLOW_INVALID(truflow)) {
5170                 PMD_DRV_LOG(ERR,
5171                             "Invalid value passed to truflow devargs.\n");
5172                 return -EINVAL;
5173         }
5174
5175         if (truflow) {
5176                 bp->flags |= BNXT_FLAG_TRUFLOW_EN;
5177                 PMD_DRV_LOG(INFO, "Host-based truflow feature enabled.\n");
5178         } else {
5179                 bp->flags &= ~BNXT_FLAG_TRUFLOW_EN;
5180                 PMD_DRV_LOG(INFO, "Host-based truflow feature disabled.\n");
5181         }
5182
5183         return 0;
5184 }
5185
5186 static int
5187 bnxt_parse_devarg_flow_xstat(__rte_unused const char *key,
5188                              const char *value, void *opaque_arg)
5189 {
5190         struct bnxt *bp = opaque_arg;
5191         unsigned long flow_xstat;
5192         char *end = NULL;
5193
5194         if (!value || !opaque_arg) {
5195                 PMD_DRV_LOG(ERR,
5196                             "Invalid parameter passed to flow_xstat devarg.\n");
5197                 return -EINVAL;
5198         }
5199
5200         flow_xstat = strtoul(value, &end, 10);
5201         if (end == NULL || *end != '\0' ||
5202             (flow_xstat == ULONG_MAX && errno == ERANGE)) {
5203                 PMD_DRV_LOG(ERR,
5204                             "Invalid parameter passed to flow_xstat devarg.\n");
5205                 return -EINVAL;
5206         }
5207
5208         if (BNXT_DEVARG_FLOW_XSTAT_INVALID(flow_xstat)) {
5209                 PMD_DRV_LOG(ERR,
5210                             "Invalid value passed to flow_xstat devarg.\n");
5211                 return -EINVAL;
5212         }
5213
5214         bp->flags |= BNXT_FLAG_FLOW_XSTATS_EN;
5215         if (BNXT_FLOW_XSTATS_EN(bp))
5216                 PMD_DRV_LOG(INFO, "flow_xstat feature enabled.\n");
5217
5218         return 0;
5219 }
5220
5221 static int
5222 bnxt_parse_devarg_max_num_kflows(__rte_unused const char *key,
5223                                         const char *value, void *opaque_arg)
5224 {
5225         struct bnxt *bp = opaque_arg;
5226         unsigned long max_num_kflows;
5227         char *end = NULL;
5228
5229         if (!value || !opaque_arg) {
5230                 PMD_DRV_LOG(ERR,
5231                         "Invalid parameter passed to max_num_kflows devarg.\n");
5232                 return -EINVAL;
5233         }
5234
5235         max_num_kflows = strtoul(value, &end, 10);
5236         if (end == NULL || *end != '\0' ||
5237                 (max_num_kflows == ULONG_MAX && errno == ERANGE)) {
5238                 PMD_DRV_LOG(ERR,
5239                         "Invalid parameter passed to max_num_kflows devarg.\n");
5240                 return -EINVAL;
5241         }
5242
5243         if (bnxt_devarg_max_num_kflow_invalid(max_num_kflows)) {
5244                 PMD_DRV_LOG(ERR,
5245                         "Invalid value passed to max_num_kflows devarg.\n");
5246                 return -EINVAL;
5247         }
5248
5249         bp->max_num_kflows = max_num_kflows;
5250         if (bp->max_num_kflows)
5251                 PMD_DRV_LOG(INFO, "max_num_kflows set as %ldK.\n",
5252                                 max_num_kflows);
5253
5254         return 0;
5255 }
5256
5257 static int
5258 bnxt_parse_devarg_rep_is_pf(__rte_unused const char *key,
5259                             const char *value, void *opaque_arg)
5260 {
5261         struct bnxt_representor *vfr_bp = opaque_arg;
5262         unsigned long rep_is_pf;
5263         char *end = NULL;
5264
5265         if (!value || !opaque_arg) {
5266                 PMD_DRV_LOG(ERR,
5267                             "Invalid parameter passed to rep_is_pf devargs.\n");
5268                 return -EINVAL;
5269         }
5270
5271         rep_is_pf = strtoul(value, &end, 10);
5272         if (end == NULL || *end != '\0' ||
5273             (rep_is_pf == ULONG_MAX && errno == ERANGE)) {
5274                 PMD_DRV_LOG(ERR,
5275                             "Invalid parameter passed to rep_is_pf devargs.\n");
5276                 return -EINVAL;
5277         }
5278
5279         if (BNXT_DEVARG_REP_IS_PF_INVALID(rep_is_pf)) {
5280                 PMD_DRV_LOG(ERR,
5281                             "Invalid value passed to rep_is_pf devargs.\n");
5282                 return -EINVAL;
5283         }
5284
5285         vfr_bp->flags |= rep_is_pf;
5286         if (BNXT_REP_PF(vfr_bp))
5287                 PMD_DRV_LOG(INFO, "PF representor\n");
5288         else
5289                 PMD_DRV_LOG(INFO, "VF representor\n");
5290
5291         return 0;
5292 }
5293
5294 static int
5295 bnxt_parse_devarg_rep_based_pf(__rte_unused const char *key,
5296                                const char *value, void *opaque_arg)
5297 {
5298         struct bnxt_representor *vfr_bp = opaque_arg;
5299         unsigned long rep_based_pf;
5300         char *end = NULL;
5301
5302         if (!value || !opaque_arg) {
5303                 PMD_DRV_LOG(ERR,
5304                             "Invalid parameter passed to rep_based_pf "
5305                             "devargs.\n");
5306                 return -EINVAL;
5307         }
5308
5309         rep_based_pf = strtoul(value, &end, 10);
5310         if (end == NULL || *end != '\0' ||
5311             (rep_based_pf == ULONG_MAX && errno == ERANGE)) {
5312                 PMD_DRV_LOG(ERR,
5313                             "Invalid parameter passed to rep_based_pf "
5314                             "devargs.\n");
5315                 return -EINVAL;
5316         }
5317
5318         if (BNXT_DEVARG_REP_BASED_PF_INVALID(rep_based_pf)) {
5319                 PMD_DRV_LOG(ERR,
5320                             "Invalid value passed to rep_based_pf devargs.\n");
5321                 return -EINVAL;
5322         }
5323
5324         vfr_bp->rep_based_pf = rep_based_pf;
5325         vfr_bp->flags |= BNXT_REP_BASED_PF_VALID;
5326
5327         PMD_DRV_LOG(INFO, "rep-based-pf = %d\n", vfr_bp->rep_based_pf);
5328
5329         return 0;
5330 }
5331
5332 static int
5333 bnxt_parse_devarg_rep_q_r2f(__rte_unused const char *key,
5334                             const char *value, void *opaque_arg)
5335 {
5336         struct bnxt_representor *vfr_bp = opaque_arg;
5337         unsigned long rep_q_r2f;
5338         char *end = NULL;
5339
5340         if (!value || !opaque_arg) {
5341                 PMD_DRV_LOG(ERR,
5342                             "Invalid parameter passed to rep_q_r2f "
5343                             "devargs.\n");
5344                 return -EINVAL;
5345         }
5346
5347         rep_q_r2f = strtoul(value, &end, 10);
5348         if (end == NULL || *end != '\0' ||
5349             (rep_q_r2f == ULONG_MAX && errno == ERANGE)) {
5350                 PMD_DRV_LOG(ERR,
5351                             "Invalid parameter passed to rep_q_r2f "
5352                             "devargs.\n");
5353                 return -EINVAL;
5354         }
5355
5356         if (BNXT_DEVARG_REP_Q_R2F_INVALID(rep_q_r2f)) {
5357                 PMD_DRV_LOG(ERR,
5358                             "Invalid value passed to rep_q_r2f devargs.\n");
5359                 return -EINVAL;
5360         }
5361
5362         vfr_bp->rep_q_r2f = rep_q_r2f;
5363         vfr_bp->flags |= BNXT_REP_Q_R2F_VALID;
5364         PMD_DRV_LOG(INFO, "rep-q-r2f = %d\n", vfr_bp->rep_q_r2f);
5365
5366         return 0;
5367 }
5368
5369 static int
5370 bnxt_parse_devarg_rep_q_f2r(__rte_unused const char *key,
5371                             const char *value, void *opaque_arg)
5372 {
5373         struct bnxt_representor *vfr_bp = opaque_arg;
5374         unsigned long rep_q_f2r;
5375         char *end = NULL;
5376
5377         if (!value || !opaque_arg) {
5378                 PMD_DRV_LOG(ERR,
5379                             "Invalid parameter passed to rep_q_f2r "
5380                             "devargs.\n");
5381                 return -EINVAL;
5382         }
5383
5384         rep_q_f2r = strtoul(value, &end, 10);
5385         if (end == NULL || *end != '\0' ||
5386             (rep_q_f2r == ULONG_MAX && errno == ERANGE)) {
5387                 PMD_DRV_LOG(ERR,
5388                             "Invalid parameter passed to rep_q_f2r "
5389                             "devargs.\n");
5390                 return -EINVAL;
5391         }
5392
5393         if (BNXT_DEVARG_REP_Q_F2R_INVALID(rep_q_f2r)) {
5394                 PMD_DRV_LOG(ERR,
5395                             "Invalid value passed to rep_q_f2r devargs.\n");
5396                 return -EINVAL;
5397         }
5398
5399         vfr_bp->rep_q_f2r = rep_q_f2r;
5400         vfr_bp->flags |= BNXT_REP_Q_F2R_VALID;
5401         PMD_DRV_LOG(INFO, "rep-q-f2r = %d\n", vfr_bp->rep_q_f2r);
5402
5403         return 0;
5404 }
5405
5406 static int
5407 bnxt_parse_devarg_rep_fc_r2f(__rte_unused const char *key,
5408                              const char *value, void *opaque_arg)
5409 {
5410         struct bnxt_representor *vfr_bp = opaque_arg;
5411         unsigned long rep_fc_r2f;
5412         char *end = NULL;
5413
5414         if (!value || !opaque_arg) {
5415                 PMD_DRV_LOG(ERR,
5416                             "Invalid parameter passed to rep_fc_r2f "
5417                             "devargs.\n");
5418                 return -EINVAL;
5419         }
5420
5421         rep_fc_r2f = strtoul(value, &end, 10);
5422         if (end == NULL || *end != '\0' ||
5423             (rep_fc_r2f == ULONG_MAX && errno == ERANGE)) {
5424                 PMD_DRV_LOG(ERR,
5425                             "Invalid parameter passed to rep_fc_r2f "
5426                             "devargs.\n");
5427                 return -EINVAL;
5428         }
5429
5430         if (BNXT_DEVARG_REP_FC_R2F_INVALID(rep_fc_r2f)) {
5431                 PMD_DRV_LOG(ERR,
5432                             "Invalid value passed to rep_fc_r2f devargs.\n");
5433                 return -EINVAL;
5434         }
5435
5436         vfr_bp->flags |= BNXT_REP_FC_R2F_VALID;
5437         vfr_bp->rep_fc_r2f = rep_fc_r2f;
5438         PMD_DRV_LOG(INFO, "rep-fc-r2f = %lu\n", rep_fc_r2f);
5439
5440         return 0;
5441 }
5442
5443 static int
5444 bnxt_parse_devarg_rep_fc_f2r(__rte_unused const char *key,
5445                              const char *value, void *opaque_arg)
5446 {
5447         struct bnxt_representor *vfr_bp = opaque_arg;
5448         unsigned long rep_fc_f2r;
5449         char *end = NULL;
5450
5451         if (!value || !opaque_arg) {
5452                 PMD_DRV_LOG(ERR,
5453                             "Invalid parameter passed to rep_fc_f2r "
5454                             "devargs.\n");
5455                 return -EINVAL;
5456         }
5457
5458         rep_fc_f2r = strtoul(value, &end, 10);
5459         if (end == NULL || *end != '\0' ||
5460             (rep_fc_f2r == ULONG_MAX && errno == ERANGE)) {
5461                 PMD_DRV_LOG(ERR,
5462                             "Invalid parameter passed to rep_fc_f2r "
5463                             "devargs.\n");
5464                 return -EINVAL;
5465         }
5466
5467         if (BNXT_DEVARG_REP_FC_F2R_INVALID(rep_fc_f2r)) {
5468                 PMD_DRV_LOG(ERR,
5469                             "Invalid value passed to rep_fc_f2r devargs.\n");
5470                 return -EINVAL;
5471         }
5472
5473         vfr_bp->flags |= BNXT_REP_FC_F2R_VALID;
5474         vfr_bp->rep_fc_f2r = rep_fc_f2r;
5475         PMD_DRV_LOG(INFO, "rep-fc-f2r = %lu\n", rep_fc_f2r);
5476
5477         return 0;
5478 }
5479
5480 static int
5481 bnxt_parse_dev_args(struct bnxt *bp, struct rte_devargs *devargs)
5482 {
5483         struct rte_kvargs *kvlist;
5484         int ret;
5485
5486         if (devargs == NULL)
5487                 return 0;
5488
5489         kvlist = rte_kvargs_parse(devargs->args, bnxt_dev_args);
5490         if (kvlist == NULL)
5491                 return -EINVAL;
5492
5493         /*
5494          * Handler for "truflow" devarg.
5495          * Invoked as for ex: "-a 0000:00:0d.0,host-based-truflow=1"
5496          */
5497         ret = rte_kvargs_process(kvlist, BNXT_DEVARG_TRUFLOW,
5498                                  bnxt_parse_devarg_truflow, bp);
5499         if (ret)
5500                 goto err;
5501
5502         /*
5503          * Handler for "flow_xstat" devarg.
5504          * Invoked as for ex: "-a 0000:00:0d.0,flow_xstat=1"
5505          */
5506         ret = rte_kvargs_process(kvlist, BNXT_DEVARG_FLOW_XSTAT,
5507                                  bnxt_parse_devarg_flow_xstat, bp);
5508         if (ret)
5509                 goto err;
5510
5511         /*
5512          * Handler for "max_num_kflows" devarg.
5513          * Invoked as for ex: "-a 000:00:0d.0,max_num_kflows=32"
5514          */
5515         ret = rte_kvargs_process(kvlist, BNXT_DEVARG_MAX_NUM_KFLOWS,
5516                                  bnxt_parse_devarg_max_num_kflows, bp);
5517         if (ret)
5518                 goto err;
5519
5520 err:
5521         rte_kvargs_free(kvlist);
5522         return ret;
5523 }
5524
5525 static int bnxt_alloc_switch_domain(struct bnxt *bp)
5526 {
5527         int rc = 0;
5528
5529         if (BNXT_PF(bp) || BNXT_VF_IS_TRUSTED(bp)) {
5530                 rc = rte_eth_switch_domain_alloc(&bp->switch_domain_id);
5531                 if (rc)
5532                         PMD_DRV_LOG(ERR,
5533                                     "Failed to alloc switch domain: %d\n", rc);
5534                 else
5535                         PMD_DRV_LOG(INFO,
5536                                     "Switch domain allocated %d\n",
5537                                     bp->switch_domain_id);
5538         }
5539
5540         return rc;
5541 }
5542
5543 /* Allocate and initialize various fields in bnxt struct that
5544  * need to be allocated/destroyed only once in the lifetime of the driver
5545  */
5546 static int bnxt_drv_init(struct rte_eth_dev *eth_dev)
5547 {
5548         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
5549         struct bnxt *bp = eth_dev->data->dev_private;
5550         int rc = 0;
5551
5552         bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
5553
5554         if (bnxt_vf_pciid(pci_dev->id.device_id))
5555                 bp->flags |= BNXT_FLAG_VF;
5556
5557         if (bnxt_p5_device(pci_dev->id.device_id))
5558                 bp->flags |= BNXT_FLAG_CHIP_P5;
5559
5560         if (pci_dev->id.device_id == BROADCOM_DEV_ID_58802 ||
5561             pci_dev->id.device_id == BROADCOM_DEV_ID_58804 ||
5562             pci_dev->id.device_id == BROADCOM_DEV_ID_58808 ||
5563             pci_dev->id.device_id == BROADCOM_DEV_ID_58802_VF)
5564                 bp->flags |= BNXT_FLAG_STINGRAY;
5565
5566         if (BNXT_TRUFLOW_EN(bp)) {
5567                 /* extra mbuf field is required to store CFA code from mark */
5568                 static const struct rte_mbuf_dynfield bnxt_cfa_code_dynfield_desc = {
5569                         .name = RTE_PMD_BNXT_CFA_CODE_DYNFIELD_NAME,
5570                         .size = sizeof(bnxt_cfa_code_dynfield_t),
5571                         .align = __alignof__(bnxt_cfa_code_dynfield_t),
5572                 };
5573                 bnxt_cfa_code_dynfield_offset =
5574                         rte_mbuf_dynfield_register(&bnxt_cfa_code_dynfield_desc);
5575                 if (bnxt_cfa_code_dynfield_offset < 0) {
5576                         PMD_DRV_LOG(ERR,
5577                             "Failed to register mbuf field for TruFlow mark\n");
5578                         return -rte_errno;
5579                 }
5580         }
5581
5582         rc = bnxt_map_pci_bars(eth_dev);
5583         if (rc) {
5584                 PMD_DRV_LOG(ERR,
5585                             "Failed to initialize board rc: %x\n", rc);
5586                 return rc;
5587         }
5588
5589         rc = bnxt_alloc_pf_info(bp);
5590         if (rc)
5591                 return rc;
5592
5593         rc = bnxt_alloc_link_info(bp);
5594         if (rc)
5595                 return rc;
5596
5597         rc = bnxt_alloc_parent_info(bp);
5598         if (rc)
5599                 return rc;
5600
5601         rc = bnxt_alloc_hwrm_resources(bp);
5602         if (rc) {
5603                 PMD_DRV_LOG(ERR,
5604                             "Failed to allocate response buffer rc: %x\n", rc);
5605                 return rc;
5606         }
5607         rc = bnxt_alloc_leds_info(bp);
5608         if (rc)
5609                 return rc;
5610
5611         rc = bnxt_alloc_cos_queues(bp);
5612         if (rc)
5613                 return rc;
5614
5615         rc = bnxt_init_locks(bp);
5616         if (rc)
5617                 return rc;
5618
5619         rc = bnxt_alloc_switch_domain(bp);
5620         if (rc)
5621                 return rc;
5622
5623         return rc;
5624 }
5625
5626 static int
5627 bnxt_dev_init(struct rte_eth_dev *eth_dev, void *params __rte_unused)
5628 {
5629         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
5630         static int version_printed;
5631         struct bnxt *bp;
5632         int rc;
5633
5634         if (version_printed++ == 0)
5635                 PMD_DRV_LOG(INFO, "%s\n", bnxt_version);
5636
5637         eth_dev->dev_ops = &bnxt_dev_ops;
5638         eth_dev->rx_queue_count = bnxt_rx_queue_count_op;
5639         eth_dev->rx_descriptor_status = bnxt_rx_descriptor_status_op;
5640         eth_dev->tx_descriptor_status = bnxt_tx_descriptor_status_op;
5641         eth_dev->rx_pkt_burst = &bnxt_recv_pkts;
5642         eth_dev->tx_pkt_burst = &bnxt_xmit_pkts;
5643
5644         /*
5645          * For secondary processes, we don't initialise any further
5646          * as primary has already done this work.
5647          */
5648         if (rte_eal_process_type() != RTE_PROC_PRIMARY)
5649                 return 0;
5650
5651         rte_eth_copy_pci_info(eth_dev, pci_dev);
5652         eth_dev->data->dev_flags |= RTE_ETH_DEV_AUTOFILL_QUEUE_XSTATS;
5653
5654         bp = eth_dev->data->dev_private;
5655
5656         /* Parse dev arguments passed on when starting the DPDK application. */
5657         rc = bnxt_parse_dev_args(bp, pci_dev->device.devargs);
5658         if (rc)
5659                 goto error_free;
5660
5661         rc = bnxt_drv_init(eth_dev);
5662         if (rc)
5663                 goto error_free;
5664
5665         rc = bnxt_init_resources(bp, false);
5666         if (rc)
5667                 goto error_free;
5668
5669         rc = bnxt_alloc_stats_mem(bp);
5670         if (rc)
5671                 goto error_free;
5672
5673         PMD_DRV_LOG(INFO,
5674                     DRV_MODULE_NAME "found at mem %" PRIX64 ", node addr %pM\n",
5675                     pci_dev->mem_resource[0].phys_addr,
5676                     pci_dev->mem_resource[0].addr);
5677
5678         return 0;
5679
5680 error_free:
5681         bnxt_dev_uninit(eth_dev);
5682         return rc;
5683 }
5684
5685
5686 static void bnxt_free_ctx_mem_buf(struct bnxt_ctx_mem_buf_info *ctx)
5687 {
5688         if (!ctx)
5689                 return;
5690
5691         if (ctx->va)
5692                 rte_free(ctx->va);
5693
5694         ctx->va = NULL;
5695         ctx->dma = RTE_BAD_IOVA;
5696         ctx->ctx_id = BNXT_CTX_VAL_INVAL;
5697 }
5698
5699 static void bnxt_unregister_fc_ctx_mem(struct bnxt *bp)
5700 {
5701         bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_RX,
5702                                   CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
5703                                   bp->flow_stat->rx_fc_out_tbl.ctx_id,
5704                                   bp->flow_stat->max_fc,
5705                                   false);
5706
5707         bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_TX,
5708                                   CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
5709                                   bp->flow_stat->tx_fc_out_tbl.ctx_id,
5710                                   bp->flow_stat->max_fc,
5711                                   false);
5712
5713         if (bp->flow_stat->rx_fc_in_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
5714                 bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->rx_fc_in_tbl.ctx_id);
5715         bp->flow_stat->rx_fc_in_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
5716
5717         if (bp->flow_stat->rx_fc_out_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
5718                 bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->rx_fc_out_tbl.ctx_id);
5719         bp->flow_stat->rx_fc_out_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
5720
5721         if (bp->flow_stat->tx_fc_in_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
5722                 bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->tx_fc_in_tbl.ctx_id);
5723         bp->flow_stat->tx_fc_in_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
5724
5725         if (bp->flow_stat->tx_fc_out_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
5726                 bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->tx_fc_out_tbl.ctx_id);
5727         bp->flow_stat->tx_fc_out_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
5728 }
5729
5730 static void bnxt_uninit_fc_ctx_mem(struct bnxt *bp)
5731 {
5732         bnxt_unregister_fc_ctx_mem(bp);
5733
5734         bnxt_free_ctx_mem_buf(&bp->flow_stat->rx_fc_in_tbl);
5735         bnxt_free_ctx_mem_buf(&bp->flow_stat->rx_fc_out_tbl);
5736         bnxt_free_ctx_mem_buf(&bp->flow_stat->tx_fc_in_tbl);
5737         bnxt_free_ctx_mem_buf(&bp->flow_stat->tx_fc_out_tbl);
5738 }
5739
5740 static void bnxt_uninit_ctx_mem(struct bnxt *bp)
5741 {
5742         if (BNXT_FLOW_XSTATS_EN(bp))
5743                 bnxt_uninit_fc_ctx_mem(bp);
5744 }
5745
5746 static void
5747 bnxt_free_error_recovery_info(struct bnxt *bp)
5748 {
5749         rte_free(bp->recovery_info);
5750         bp->recovery_info = NULL;
5751         bp->fw_cap &= ~BNXT_FW_CAP_ERROR_RECOVERY;
5752 }
5753
5754 static int
5755 bnxt_uninit_resources(struct bnxt *bp, bool reconfig_dev)
5756 {
5757         int rc;
5758
5759         bnxt_free_int(bp);
5760         bnxt_free_mem(bp, reconfig_dev);
5761
5762         bnxt_hwrm_func_buf_unrgtr(bp);
5763         if (bp->pf != NULL) {
5764                 rte_free(bp->pf->vf_req_buf);
5765                 bp->pf->vf_req_buf = NULL;
5766         }
5767
5768         rc = bnxt_hwrm_func_driver_unregister(bp, 0);
5769         bp->flags &= ~BNXT_FLAG_REGISTERED;
5770         bnxt_free_ctx_mem(bp);
5771         if (!reconfig_dev) {
5772                 bnxt_free_hwrm_resources(bp);
5773                 bnxt_free_error_recovery_info(bp);
5774         }
5775
5776         bnxt_uninit_ctx_mem(bp);
5777
5778         bnxt_free_flow_stats_info(bp);
5779         if (bp->rep_info != NULL)
5780                 bnxt_free_switch_domain(bp);
5781         bnxt_free_rep_info(bp);
5782         rte_free(bp->ptp_cfg);
5783         bp->ptp_cfg = NULL;
5784         return rc;
5785 }
5786
5787 static int
5788 bnxt_dev_uninit(struct rte_eth_dev *eth_dev)
5789 {
5790         if (rte_eal_process_type() != RTE_PROC_PRIMARY)
5791                 return -EPERM;
5792
5793         PMD_DRV_LOG(DEBUG, "Calling Device uninit\n");
5794
5795         if (eth_dev->state != RTE_ETH_DEV_UNUSED)
5796                 bnxt_dev_close_op(eth_dev);
5797
5798         return 0;
5799 }
5800
5801 static int bnxt_pci_remove_dev_with_reps(struct rte_eth_dev *eth_dev)
5802 {
5803         struct bnxt *bp = eth_dev->data->dev_private;
5804         struct rte_eth_dev *vf_rep_eth_dev;
5805         int ret = 0, i;
5806
5807         if (!bp)
5808                 return -EINVAL;
5809
5810         for (i = 0; i < bp->num_reps; i++) {
5811                 vf_rep_eth_dev = bp->rep_info[i].vfr_eth_dev;
5812                 if (!vf_rep_eth_dev)
5813                         continue;
5814                 PMD_DRV_LOG(DEBUG, "BNXT Port:%d VFR pci remove\n",
5815                             vf_rep_eth_dev->data->port_id);
5816                 rte_eth_dev_destroy(vf_rep_eth_dev, bnxt_representor_uninit);
5817         }
5818         PMD_DRV_LOG(DEBUG, "BNXT Port:%d pci remove\n",
5819                     eth_dev->data->port_id);
5820         ret = rte_eth_dev_destroy(eth_dev, bnxt_dev_uninit);
5821
5822         return ret;
5823 }
5824
5825 static void bnxt_free_rep_info(struct bnxt *bp)
5826 {
5827         rte_free(bp->rep_info);
5828         bp->rep_info = NULL;
5829         rte_free(bp->cfa_code_map);
5830         bp->cfa_code_map = NULL;
5831 }
5832
5833 static int bnxt_init_rep_info(struct bnxt *bp)
5834 {
5835         int i = 0, rc;
5836
5837         if (bp->rep_info)
5838                 return 0;
5839
5840         bp->rep_info = rte_zmalloc("bnxt_rep_info",
5841                                    sizeof(bp->rep_info[0]) * BNXT_MAX_VF_REPS,
5842                                    0);
5843         if (!bp->rep_info) {
5844                 PMD_DRV_LOG(ERR, "Failed to alloc memory for rep info\n");
5845                 return -ENOMEM;
5846         }
5847         bp->cfa_code_map = rte_zmalloc("bnxt_cfa_code_map",
5848                                        sizeof(*bp->cfa_code_map) *
5849                                        BNXT_MAX_CFA_CODE, 0);
5850         if (!bp->cfa_code_map) {
5851                 PMD_DRV_LOG(ERR, "Failed to alloc memory for cfa_code_map\n");
5852                 bnxt_free_rep_info(bp);
5853                 return -ENOMEM;
5854         }
5855
5856         for (i = 0; i < BNXT_MAX_CFA_CODE; i++)
5857                 bp->cfa_code_map[i] = BNXT_VF_IDX_INVALID;
5858
5859         rc = pthread_mutex_init(&bp->rep_info->vfr_lock, NULL);
5860         if (rc) {
5861                 PMD_DRV_LOG(ERR, "Unable to initialize vfr_lock\n");
5862                 bnxt_free_rep_info(bp);
5863                 return rc;
5864         }
5865
5866         rc = pthread_mutex_init(&bp->rep_info->vfr_start_lock, NULL);
5867         if (rc) {
5868                 PMD_DRV_LOG(ERR, "Unable to initialize vfr_start_lock\n");
5869                 bnxt_free_rep_info(bp);
5870                 return rc;
5871         }
5872
5873         return rc;
5874 }
5875
5876 static int bnxt_rep_port_probe(struct rte_pci_device *pci_dev,
5877                                struct rte_eth_devargs *eth_da,
5878                                struct rte_eth_dev *backing_eth_dev,
5879                                const char *dev_args)
5880 {
5881         struct rte_eth_dev *vf_rep_eth_dev;
5882         char name[RTE_ETH_NAME_MAX_LEN];
5883         struct bnxt *backing_bp;
5884         uint16_t num_rep;
5885         int i, ret = 0;
5886         struct rte_kvargs *kvlist = NULL;
5887
5888         if (eth_da->type == RTE_ETH_REPRESENTOR_NONE)
5889                 return 0;
5890         if (eth_da->type != RTE_ETH_REPRESENTOR_VF) {
5891                 PMD_DRV_LOG(ERR, "unsupported representor type %d\n",
5892                             eth_da->type);
5893                 return -ENOTSUP;
5894         }
5895         num_rep = eth_da->nb_representor_ports;
5896         if (num_rep > BNXT_MAX_VF_REPS) {
5897                 PMD_DRV_LOG(ERR, "nb_representor_ports = %d > %d MAX VF REPS\n",
5898                             num_rep, BNXT_MAX_VF_REPS);
5899                 return -EINVAL;
5900         }
5901
5902         if (num_rep >= RTE_MAX_ETHPORTS) {
5903                 PMD_DRV_LOG(ERR,
5904                             "nb_representor_ports = %d > %d MAX ETHPORTS\n",
5905                             num_rep, RTE_MAX_ETHPORTS);
5906                 return -EINVAL;
5907         }
5908
5909         backing_bp = backing_eth_dev->data->dev_private;
5910
5911         if (!(BNXT_PF(backing_bp) || BNXT_VF_IS_TRUSTED(backing_bp))) {
5912                 PMD_DRV_LOG(ERR,
5913                             "Not a PF or trusted VF. No Representor support\n");
5914                 /* Returning an error is not an option.
5915                  * Applications are not handling this correctly
5916                  */
5917                 return 0;
5918         }
5919
5920         if (bnxt_init_rep_info(backing_bp))
5921                 return 0;
5922
5923         for (i = 0; i < num_rep; i++) {
5924                 struct bnxt_representor representor = {
5925                         .vf_id = eth_da->representor_ports[i],
5926                         .switch_domain_id = backing_bp->switch_domain_id,
5927                         .parent_dev = backing_eth_dev
5928                 };
5929
5930                 if (representor.vf_id >= BNXT_MAX_VF_REPS) {
5931                         PMD_DRV_LOG(ERR, "VF-Rep id %d >= %d MAX VF ID\n",
5932                                     representor.vf_id, BNXT_MAX_VF_REPS);
5933                         continue;
5934                 }
5935
5936                 /* representor port net_bdf_port */
5937                 snprintf(name, sizeof(name), "net_%s_representor_%d",
5938                          pci_dev->device.name, eth_da->representor_ports[i]);
5939
5940                 kvlist = rte_kvargs_parse(dev_args, bnxt_dev_args);
5941                 if (kvlist) {
5942                         /*
5943                          * Handler for "rep_is_pf" devarg.
5944                          * Invoked as for ex: "-a 000:00:0d.0,
5945                          * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
5946                          */
5947                         ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_IS_PF,
5948                                                  bnxt_parse_devarg_rep_is_pf,
5949                                                  (void *)&representor);
5950                         if (ret) {
5951                                 ret = -EINVAL;
5952                                 goto err;
5953                         }
5954                         /*
5955                          * Handler for "rep_based_pf" devarg.
5956                          * Invoked as for ex: "-a 000:00:0d.0,
5957                          * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
5958                          */
5959                         ret = rte_kvargs_process(kvlist,
5960                                                  BNXT_DEVARG_REP_BASED_PF,
5961                                                  bnxt_parse_devarg_rep_based_pf,
5962                                                  (void *)&representor);
5963                         if (ret) {
5964                                 ret = -EINVAL;
5965                                 goto err;
5966                         }
5967                         /*
5968                          * Handler for "rep_based_pf" devarg.
5969                          * Invoked as for ex: "-a 000:00:0d.0,
5970                          * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
5971                          */
5972                         ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_Q_R2F,
5973                                                  bnxt_parse_devarg_rep_q_r2f,
5974                                                  (void *)&representor);
5975                         if (ret) {
5976                                 ret = -EINVAL;
5977                                 goto err;
5978                         }
5979                         /*
5980                          * Handler for "rep_based_pf" devarg.
5981                          * Invoked as for ex: "-a 000:00:0d.0,
5982                          * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
5983                          */
5984                         ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_Q_F2R,
5985                                                  bnxt_parse_devarg_rep_q_f2r,
5986                                                  (void *)&representor);
5987                         if (ret) {
5988                                 ret = -EINVAL;
5989                                 goto err;
5990                         }
5991                         /*
5992                          * Handler for "rep_based_pf" devarg.
5993                          * Invoked as for ex: "-a 000:00:0d.0,
5994                          * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
5995                          */
5996                         ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_FC_R2F,
5997                                                  bnxt_parse_devarg_rep_fc_r2f,
5998                                                  (void *)&representor);
5999                         if (ret) {
6000                                 ret = -EINVAL;
6001                                 goto err;
6002                         }
6003                         /*
6004                          * Handler for "rep_based_pf" devarg.
6005                          * Invoked as for ex: "-a 000:00:0d.0,
6006                          * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6007                          */
6008                         ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_FC_F2R,
6009                                                  bnxt_parse_devarg_rep_fc_f2r,
6010                                                  (void *)&representor);
6011                         if (ret) {
6012                                 ret = -EINVAL;
6013                                 goto err;
6014                         }
6015                 }
6016
6017                 ret = rte_eth_dev_create(&pci_dev->device, name,
6018                                          sizeof(struct bnxt_representor),
6019                                          NULL, NULL,
6020                                          bnxt_representor_init,
6021                                          &representor);
6022                 if (ret) {
6023                         PMD_DRV_LOG(ERR, "failed to create bnxt vf "
6024                                     "representor %s.", name);
6025                         goto err;
6026                 }
6027
6028                 vf_rep_eth_dev = rte_eth_dev_allocated(name);
6029                 if (!vf_rep_eth_dev) {
6030                         PMD_DRV_LOG(ERR, "Failed to find the eth_dev"
6031                                     " for VF-Rep: %s.", name);
6032                         ret = -ENODEV;
6033                         goto err;
6034                 }
6035
6036                 PMD_DRV_LOG(DEBUG, "BNXT Port:%d VFR pci probe\n",
6037                             backing_eth_dev->data->port_id);
6038                 backing_bp->rep_info[representor.vf_id].vfr_eth_dev =
6039                                                          vf_rep_eth_dev;
6040                 backing_bp->num_reps++;
6041
6042         }
6043
6044         rte_kvargs_free(kvlist);
6045         return 0;
6046
6047 err:
6048         /* If num_rep > 1, then rollback already created
6049          * ports, since we'll be failing the probe anyway
6050          */
6051         if (num_rep > 1)
6052                 bnxt_pci_remove_dev_with_reps(backing_eth_dev);
6053         rte_errno = -ret;
6054         rte_kvargs_free(kvlist);
6055
6056         return ret;
6057 }
6058
6059 static int bnxt_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
6060                           struct rte_pci_device *pci_dev)
6061 {
6062         struct rte_eth_devargs eth_da = { .nb_representor_ports = 0 };
6063         struct rte_eth_dev *backing_eth_dev;
6064         uint16_t num_rep;
6065         int ret = 0;
6066
6067         if (pci_dev->device.devargs) {
6068                 ret = rte_eth_devargs_parse(pci_dev->device.devargs->args,
6069                                             &eth_da);
6070                 if (ret)
6071                         return ret;
6072         }
6073
6074         num_rep = eth_da.nb_representor_ports;
6075         PMD_DRV_LOG(DEBUG, "nb_representor_ports = %d\n",
6076                     num_rep);
6077
6078         /* We could come here after first level of probe is already invoked
6079          * as part of an application bringup(OVS-DPDK vswitchd), so first check
6080          * for already allocated eth_dev for the backing device (PF/Trusted VF)
6081          */
6082         backing_eth_dev = rte_eth_dev_allocated(pci_dev->device.name);
6083         if (backing_eth_dev == NULL) {
6084                 ret = rte_eth_dev_create(&pci_dev->device, pci_dev->device.name,
6085                                          sizeof(struct bnxt),
6086                                          eth_dev_pci_specific_init, pci_dev,
6087                                          bnxt_dev_init, NULL);
6088
6089                 if (ret || !num_rep)
6090                         return ret;
6091
6092                 backing_eth_dev = rte_eth_dev_allocated(pci_dev->device.name);
6093         }
6094         PMD_DRV_LOG(DEBUG, "BNXT Port:%d pci probe\n",
6095                     backing_eth_dev->data->port_id);
6096
6097         if (!num_rep)
6098                 return ret;
6099
6100         /* probe representor ports now */
6101         ret = bnxt_rep_port_probe(pci_dev, &eth_da, backing_eth_dev,
6102                                   pci_dev->device.devargs->args);
6103
6104         return ret;
6105 }
6106
6107 static int bnxt_pci_remove(struct rte_pci_device *pci_dev)
6108 {
6109         struct rte_eth_dev *eth_dev;
6110
6111         eth_dev = rte_eth_dev_allocated(pci_dev->device.name);
6112         if (!eth_dev)
6113                 return 0; /* Invoked typically only by OVS-DPDK, by the
6114                            * time it comes here the eth_dev is already
6115                            * deleted by rte_eth_dev_close(), so returning
6116                            * +ve value will at least help in proper cleanup
6117                            */
6118
6119         PMD_DRV_LOG(DEBUG, "BNXT Port:%d pci remove\n", eth_dev->data->port_id);
6120         if (rte_eal_process_type() == RTE_PROC_PRIMARY) {
6121                 if (eth_dev->data->dev_flags & RTE_ETH_DEV_REPRESENTOR)
6122                         return rte_eth_dev_destroy(eth_dev,
6123                                                    bnxt_representor_uninit);
6124                 else
6125                         return rte_eth_dev_destroy(eth_dev,
6126                                                    bnxt_dev_uninit);
6127         } else {
6128                 return rte_eth_dev_pci_generic_remove(pci_dev, NULL);
6129         }
6130 }
6131
6132 static struct rte_pci_driver bnxt_rte_pmd = {
6133         .id_table = bnxt_pci_id_map,
6134         .drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC |
6135                         RTE_PCI_DRV_PROBE_AGAIN, /* Needed in case of VF-REPs
6136                                                   * and OVS-DPDK
6137                                                   */
6138         .probe = bnxt_pci_probe,
6139         .remove = bnxt_pci_remove,
6140 };
6141
6142 static bool
6143 is_device_supported(struct rte_eth_dev *dev, struct rte_pci_driver *drv)
6144 {
6145         if (strcmp(dev->device->driver->name, drv->driver.name))
6146                 return false;
6147
6148         return true;
6149 }
6150
6151 bool is_bnxt_supported(struct rte_eth_dev *dev)
6152 {
6153         return is_device_supported(dev, &bnxt_rte_pmd);
6154 }
6155
6156 RTE_LOG_REGISTER(bnxt_logtype_driver, pmd.net.bnxt.driver, NOTICE);
6157 RTE_PMD_REGISTER_PCI(net_bnxt, bnxt_rte_pmd);
6158 RTE_PMD_REGISTER_PCI_TABLE(net_bnxt, bnxt_pci_id_map);
6159 RTE_PMD_REGISTER_KMOD_DEP(net_bnxt, "* igb_uio | uio_pci_generic | vfio-pci");