drivers/net/bnxt/bnxt_ethdev.c

   1 /* SPDX-License-Identifier: BSD-3-Clause
   2  * Copyright(c) 2014-2018 Broadcom
   3  * All rights reserved.
   4  */
   5
   6 #include <inttypes.h>
   7 #include <stdbool.h>
   8
   9 #include <rte_dev.h>
  10 #include <rte_ethdev_driver.h>
  11 #include <rte_ethdev_pci.h>
  12 #include <rte_malloc.h>
  13 #include <rte_cycles.h>
  14 #include <rte_alarm.h>
  15
  16 #include "bnxt.h"
  17 #include "bnxt_filter.h"
  18 #include "bnxt_hwrm.h"
  19 #include "bnxt_irq.h"
  20 #include "bnxt_ring.h"
  21 #include "bnxt_rxq.h"
  22 #include "bnxt_rxr.h"
  23 #include "bnxt_stats.h"
  24 #include "bnxt_txq.h"
  25 #include "bnxt_txr.h"
  26 #include "bnxt_vnic.h"
  27 #include "hsi_struct_def_dpdk.h"
  28 #include "bnxt_nvm_defs.h"
  29
  30 #define DRV_MODULE_NAME         "bnxt"
  31 static const char bnxt_version[] =
  32         "Broadcom NetXtreme driver " DRV_MODULE_NAME;
  33 int bnxt_logtype_driver;
  34
  35 /*
  36  * The set of PCI devices this driver supports
  37  */
  38 static const struct rte_pci_id bnxt_pci_id_map[] = {
  39         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM,
  40                          BROADCOM_DEV_ID_STRATUS_NIC_VF1) },
  41         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM,
  42                          BROADCOM_DEV_ID_STRATUS_NIC_VF2) },
  43         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_STRATUS_NIC) },
  44         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57414_VF) },
  45         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57301) },
  46         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57302) },
  47         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57304_PF) },
  48         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57304_VF) },
  49         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_NS2) },
  50         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57402) },
  51         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57404) },
  52         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57406_PF) },
  53         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57406_VF) },
  54         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57402_MF) },
  55         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57407_RJ45) },
  56         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57404_MF) },
  57         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57406_MF) },
  58         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57407_SFP) },
  59         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57407_MF) },
  60         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_5741X_VF) },
  61         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_5731X_VF) },
  62         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57314) },
  63         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57417_MF) },
  64         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57311) },
  65         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57312) },
  66         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57412) },
  67         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57414) },
  68         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57416_RJ45) },
  69         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57417_RJ45) },
  70         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57412_MF) },
  71         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57317_RJ45) },
  72         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57417_SFP) },
  73         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57416_SFP) },
  74         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57317_SFP) },
  75         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57414_MF) },
  76         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57416_MF) },
  77         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58802) },
  78         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58804) },
  79         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58808) },
  80         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58802_VF) },
  81         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57508) },
  82         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57504) },
  83         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57502) },
  84         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57500_VF1) },
  85         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57500_VF2) },
  86         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57508_MF1) },
  87         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57504_MF1) },
  88         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57502_MF1) },
  89         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57508_MF2) },
  90         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57504_MF2) },
  91         { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57502_MF2) },
  92         { .vendor_id = 0, /* sentinel */ },
  93 };
  94
  95 #define BNXT_ETH_RSS_SUPPORT (  \
  96         ETH_RSS_IPV4 |          \
  97         ETH_RSS_NONFRAG_IPV4_TCP |      \
  98         ETH_RSS_NONFRAG_IPV4_UDP |      \
  99         ETH_RSS_IPV6 |          \
 100         ETH_RSS_NONFRAG_IPV6_TCP |      \
 101         ETH_RSS_NONFRAG_IPV6_UDP)
 102
 103 #define BNXT_DEV_TX_OFFLOAD_SUPPORT (DEV_TX_OFFLOAD_VLAN_INSERT | \
 104                                      DEV_TX_OFFLOAD_IPV4_CKSUM | \
 105                                      DEV_TX_OFFLOAD_TCP_CKSUM | \
 106                                      DEV_TX_OFFLOAD_UDP_CKSUM | \
 107                                      DEV_TX_OFFLOAD_TCP_TSO | \
 108                                      DEV_TX_OFFLOAD_OUTER_IPV4_CKSUM | \
 109                                      DEV_TX_OFFLOAD_VXLAN_TNL_TSO | \
 110                                      DEV_TX_OFFLOAD_GRE_TNL_TSO | \
 111                                      DEV_TX_OFFLOAD_IPIP_TNL_TSO | \
 112                                      DEV_TX_OFFLOAD_GENEVE_TNL_TSO | \
 113                                      DEV_TX_OFFLOAD_QINQ_INSERT | \
 114                                      DEV_TX_OFFLOAD_MULTI_SEGS)
 115
 116 #define BNXT_DEV_RX_OFFLOAD_SUPPORT (DEV_RX_OFFLOAD_VLAN_FILTER | \
 117                                      DEV_RX_OFFLOAD_VLAN_STRIP | \
 118                                      DEV_RX_OFFLOAD_IPV4_CKSUM | \
 119                                      DEV_RX_OFFLOAD_UDP_CKSUM | \
 120                                      DEV_RX_OFFLOAD_TCP_CKSUM | \
 121                                      DEV_RX_OFFLOAD_OUTER_IPV4_CKSUM | \
 122                                      DEV_RX_OFFLOAD_JUMBO_FRAME | \
 123                                      DEV_RX_OFFLOAD_KEEP_CRC | \
 124                                      DEV_RX_OFFLOAD_VLAN_EXTEND | \
 125                                      DEV_RX_OFFLOAD_TCP_LRO | \
 126                                      DEV_RX_OFFLOAD_SCATTER | \
 127                                      DEV_RX_OFFLOAD_RSS_HASH)
 128
 129 static int bnxt_vlan_offload_set_op(struct rte_eth_dev *dev, int mask);
 130 static void bnxt_print_link_info(struct rte_eth_dev *eth_dev);
 131 static int bnxt_dev_uninit(struct rte_eth_dev *eth_dev);
 132 static int bnxt_init_resources(struct bnxt *bp, bool reconfig_dev);
 133 static int bnxt_uninit_resources(struct bnxt *bp, bool reconfig_dev);
 134 static void bnxt_cancel_fw_health_check(struct bnxt *bp);
 135
 136 int is_bnxt_in_error(struct bnxt *bp)
 137 {
 138         if (bp->flags & BNXT_FLAG_FATAL_ERROR)
 139                 return -EIO;
 140         if (bp->flags & BNXT_FLAG_FW_RESET)
 141                 return -EBUSY;
 142
 143         return 0;
 144 }
 145
 146 /***********************/
 147
 148 /*
 149  * High level utility functions
 150  */
 151
 152 uint16_t bnxt_rss_ctxts(const struct bnxt *bp)
 153 {
 154         if (!BNXT_CHIP_THOR(bp))
 155                 return 1;
 156
 157         return RTE_ALIGN_MUL_CEIL(bp->rx_nr_rings,
 158                                   BNXT_RSS_ENTRIES_PER_CTX_THOR) /
 159                                     BNXT_RSS_ENTRIES_PER_CTX_THOR;
 160 }
 161
 162 static uint16_t  bnxt_rss_hash_tbl_size(const struct bnxt *bp)
 163 {
 164         if (!BNXT_CHIP_THOR(bp))
 165                 return HW_HASH_INDEX_SIZE;
 166
 167         return bnxt_rss_ctxts(bp) * BNXT_RSS_ENTRIES_PER_CTX_THOR;
 168 }
 169
 170 static void bnxt_free_mem(struct bnxt *bp, bool reconfig)
 171 {
 172         bnxt_free_filter_mem(bp);
 173         bnxt_free_vnic_attributes(bp);
 174         bnxt_free_vnic_mem(bp);
 175
 176         /* tx/rx rings are configured as part of *_queue_setup callbacks.
 177          * If the number of rings change across fw update,
 178          * we don't have much choice except to warn the user.
 179          */
 180         if (!reconfig) {
 181                 bnxt_free_stats(bp);
 182                 bnxt_free_tx_rings(bp);
 183                 bnxt_free_rx_rings(bp);
 184         }
 185         bnxt_free_async_cp_ring(bp);
 186         bnxt_free_rxtx_nq_ring(bp);
 187
 188         rte_free(bp->grp_info);
 189         bp->grp_info = NULL;
 190 }
 191
 192 static int bnxt_alloc_mem(struct bnxt *bp, bool reconfig)
 193 {
 194         int rc;
 195
 196         rc = bnxt_alloc_ring_grps(bp);
 197         if (rc)
 198                 goto alloc_mem_err;
 199
 200         rc = bnxt_alloc_async_ring_struct(bp);
 201         if (rc)
 202                 goto alloc_mem_err;
 203
 204         rc = bnxt_alloc_vnic_mem(bp);
 205         if (rc)
 206                 goto alloc_mem_err;
 207
 208         rc = bnxt_alloc_vnic_attributes(bp);
 209         if (rc)
 210                 goto alloc_mem_err;
 211
 212         rc = bnxt_alloc_filter_mem(bp);
 213         if (rc)
 214                 goto alloc_mem_err;
 215
 216         rc = bnxt_alloc_async_cp_ring(bp);
 217         if (rc)
 218                 goto alloc_mem_err;
 219
 220         rc = bnxt_alloc_rxtx_nq_ring(bp);
 221         if (rc)
 222                 goto alloc_mem_err;
 223
 224         return 0;
 225
 226 alloc_mem_err:
 227         bnxt_free_mem(bp, reconfig);
 228         return rc;
 229 }
 230
 231 static int bnxt_init_chip(struct bnxt *bp)
 232 {
 233         struct bnxt_rx_queue *rxq;
 234         struct rte_eth_link new;
 235         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(bp->eth_dev);
 236         struct rte_eth_conf *dev_conf = &bp->eth_dev->data->dev_conf;
 237         struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
 238         uint64_t rx_offloads = dev_conf->rxmode.offloads;
 239         uint32_t intr_vector = 0;
 240         uint32_t queue_id, base = BNXT_MISC_VEC_ID;
 241         uint32_t vec = BNXT_MISC_VEC_ID;
 242         unsigned int i, j;
 243         int rc;
 244
 245         if (bp->eth_dev->data->mtu > RTE_ETHER_MTU) {
 246                 bp->eth_dev->data->dev_conf.rxmode.offloads |=
 247                         DEV_RX_OFFLOAD_JUMBO_FRAME;
 248                 bp->flags |= BNXT_FLAG_JUMBO;
 249         } else {
 250                 bp->eth_dev->data->dev_conf.rxmode.offloads &=
 251                         ~DEV_RX_OFFLOAD_JUMBO_FRAME;
 252                 bp->flags &= ~BNXT_FLAG_JUMBO;
 253         }
 254
 255         /* THOR does not support ring groups.
 256          * But we will use the array to save RSS context IDs.
 257          */
 258         if (BNXT_CHIP_THOR(bp))
 259                 bp->max_ring_grps = BNXT_MAX_RSS_CTXTS_THOR;
 260
 261         rc = bnxt_alloc_all_hwrm_stat_ctxs(bp);
 262         if (rc) {
 263                 PMD_DRV_LOG(ERR, "HWRM stat ctx alloc failure rc: %x\n", rc);
 264                 goto err_out;
 265         }
 266
 267         rc = bnxt_alloc_hwrm_rings(bp);
 268         if (rc) {
 269                 PMD_DRV_LOG(ERR, "HWRM ring alloc failure rc: %x\n", rc);
 270                 goto err_out;
 271         }
 272
 273         rc = bnxt_alloc_all_hwrm_ring_grps(bp);
 274         if (rc) {
 275                 PMD_DRV_LOG(ERR, "HWRM ring grp alloc failure: %x\n", rc);
 276                 goto err_out;
 277         }
 278
 279         if (!(bp->vnic_cap_flags & BNXT_VNIC_CAP_COS_CLASSIFY))
 280                 goto skip_cosq_cfg;
 281
 282         for (j = 0, i = 0; i < BNXT_COS_QUEUE_COUNT; i++) {
 283                 if (bp->rx_cos_queue[i].id != 0xff) {
 284                         struct bnxt_vnic_info *vnic = &bp->vnic_info[j++];
 285
 286                         if (!vnic) {
 287                                 PMD_DRV_LOG(ERR,
 288                                             "Num pools more than FW profile\n");
 289                                 rc = -EINVAL;
 290                                 goto err_out;
 291                         }
 292                         vnic->cos_queue_id = bp->rx_cos_queue[i].id;
 293                         bp->rx_cosq_cnt++;
 294                 }
 295         }
 296
 297 skip_cosq_cfg:
 298         rc = bnxt_mq_rx_configure(bp);
 299         if (rc) {
 300                 PMD_DRV_LOG(ERR, "MQ mode configure failure rc: %x\n", rc);
 301                 goto err_out;
 302         }
 303
 304         /* VNIC configuration */
 305         for (i = 0; i < bp->nr_vnics; i++) {
 306                 struct rte_eth_conf *dev_conf = &bp->eth_dev->data->dev_conf;
 307                 struct bnxt_vnic_info *vnic = &bp->vnic_info[i];
 308
 309                 rc = bnxt_vnic_grp_alloc(bp, vnic);
 310                 if (rc)
 311                         goto err_out;
 312
 313                 PMD_DRV_LOG(DEBUG, "vnic[%d] = %p vnic->fw_grp_ids = %p\n",
 314                             i, vnic, vnic->fw_grp_ids);
 315
 316                 rc = bnxt_hwrm_vnic_alloc(bp, vnic);
 317                 if (rc) {
 318                         PMD_DRV_LOG(ERR, "HWRM vnic %d alloc failure rc: %x\n",
 319                                 i, rc);
 320                         goto err_out;
 321                 }
 322
 323                 /* Alloc RSS context only if RSS mode is enabled */
 324                 if (dev_conf->rxmode.mq_mode & ETH_MQ_RX_RSS) {
 325                         int j, nr_ctxs = bnxt_rss_ctxts(bp);
 326
 327                         rc = 0;
 328                         for (j = 0; j < nr_ctxs; j++) {
 329                                 rc = bnxt_hwrm_vnic_ctx_alloc(bp, vnic, j);
 330                                 if (rc)
 331                                         break;
 332                         }
 333                         if (rc) {
 334                                 PMD_DRV_LOG(ERR,
 335                                   "HWRM vnic %d ctx %d alloc failure rc: %x\n",
 336                                   i, j, rc);
 337                                 goto err_out;
 338                         }
 339                         vnic->num_lb_ctxts = nr_ctxs;
 340                 }
 341
 342                 /*
 343                  * Firmware sets pf pair in default vnic cfg. If the VLAN strip
 344                  * setting is not available at this time, it will not be
 345                  * configured correctly in the CFA.
 346                  */
 347                 if (rx_offloads & DEV_RX_OFFLOAD_VLAN_STRIP)
 348                         vnic->vlan_strip = true;
 349                 else
 350                         vnic->vlan_strip = false;
 351
 352                 rc = bnxt_hwrm_vnic_cfg(bp, vnic);
 353                 if (rc) {
 354                         PMD_DRV_LOG(ERR, "HWRM vnic %d cfg failure rc: %x\n",
 355                                 i, rc);
 356                         goto err_out;
 357                 }
 358
 359                 rc = bnxt_set_hwrm_vnic_filters(bp, vnic);
 360                 if (rc) {
 361                         PMD_DRV_LOG(ERR,
 362                                 "HWRM vnic %d filter failure rc: %x\n",
 363                                 i, rc);
 364                         goto err_out;
 365                 }
 366
 367                 for (j = 0; j < bp->rx_num_qs_per_vnic; j++) {
 368                         rxq = bp->eth_dev->data->rx_queues[j];
 369
 370                         PMD_DRV_LOG(DEBUG,
 371                                     "rxq[%d]->vnic=%p vnic->fw_grp_ids=%p\n",
 372                                     j, rxq->vnic, rxq->vnic->fw_grp_ids);
 373
 374                         if (BNXT_HAS_RING_GRPS(bp) && rxq->rx_deferred_start)
 375                                 rxq->vnic->fw_grp_ids[j] = INVALID_HW_RING_ID;
 376                 }
 377
 378                 rc = bnxt_vnic_rss_configure(bp, vnic);
 379                 if (rc) {
 380                         PMD_DRV_LOG(ERR,
 381                                     "HWRM vnic set RSS failure rc: %x\n", rc);
 382                         goto err_out;
 383                 }
 384
 385                 bnxt_hwrm_vnic_plcmode_cfg(bp, vnic);
 386
 387                 if (bp->eth_dev->data->dev_conf.rxmode.offloads &
 388                     DEV_RX_OFFLOAD_TCP_LRO)
 389                         bnxt_hwrm_vnic_tpa_cfg(bp, vnic, 1);
 390                 else
 391                         bnxt_hwrm_vnic_tpa_cfg(bp, vnic, 0);
 392         }
 393         rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, &bp->vnic_info[0], 0, NULL);
 394         if (rc) {
 395                 PMD_DRV_LOG(ERR,
 396                         "HWRM cfa l2 rx mask failure rc: %x\n", rc);
 397                 goto err_out;
 398         }
 399
 400         /* check and configure queue intr-vector mapping */
 401         if ((rte_intr_cap_multiple(intr_handle) ||
 402              !RTE_ETH_DEV_SRIOV(bp->eth_dev).active) &&
 403             bp->eth_dev->data->dev_conf.intr_conf.rxq != 0) {
 404                 intr_vector = bp->eth_dev->data->nb_rx_queues;
 405                 PMD_DRV_LOG(DEBUG, "intr_vector = %d\n", intr_vector);
 406                 if (intr_vector > bp->rx_cp_nr_rings) {
 407                         PMD_DRV_LOG(ERR, "At most %d intr queues supported",
 408                                         bp->rx_cp_nr_rings);
 409                         return -ENOTSUP;
 410                 }
 411                 rc = rte_intr_efd_enable(intr_handle, intr_vector);
 412                 if (rc)
 413                         return rc;
 414         }
 415
 416         if (rte_intr_dp_is_en(intr_handle) && !intr_handle->intr_vec) {
 417                 intr_handle->intr_vec =
 418                         rte_zmalloc("intr_vec",
 419                                     bp->eth_dev->data->nb_rx_queues *
 420                                     sizeof(int), 0);
 421                 if (intr_handle->intr_vec == NULL) {
 422                         PMD_DRV_LOG(ERR, "Failed to allocate %d rx_queues"
 423                                 " intr_vec", bp->eth_dev->data->nb_rx_queues);
 424                         rc = -ENOMEM;
 425                         goto err_disable;
 426                 }
 427                 PMD_DRV_LOG(DEBUG, "intr_handle->intr_vec = %p "
 428                         "intr_handle->nb_efd = %d intr_handle->max_intr = %d\n",
 429                          intr_handle->intr_vec, intr_handle->nb_efd,
 430                         intr_handle->max_intr);
 431                 for (queue_id = 0; queue_id < bp->eth_dev->data->nb_rx_queues;
 432                      queue_id++) {
 433                         intr_handle->intr_vec[queue_id] =
 434                                                         vec + BNXT_RX_VEC_START;
 435                         if (vec < base + intr_handle->nb_efd - 1)
 436                                 vec++;
 437                 }
 438         }
 439
 440         /* enable uio/vfio intr/eventfd mapping */
 441         rc = rte_intr_enable(intr_handle);
 442 #ifndef RTE_EXEC_ENV_FREEBSD
 443         /* In FreeBSD OS, nic_uio driver does not support interrupts */
 444         if (rc)
 445                 goto err_free;
 446 #endif
 447
 448         rc = bnxt_get_hwrm_link_config(bp, &new);
 449         if (rc) {
 450                 PMD_DRV_LOG(ERR, "HWRM Get link config failure rc: %x\n", rc);
 451                 goto err_free;
 452         }
 453
 454         if (!bp->link_info.link_up) {
 455                 rc = bnxt_set_hwrm_link_config(bp, true);
 456                 if (rc) {
 457                         PMD_DRV_LOG(ERR,
 458                                 "HWRM link config failure rc: %x\n", rc);
 459                         goto err_free;
 460                 }
 461         }
 462         bnxt_print_link_info(bp->eth_dev);
 463
 464         bp->mark_table = rte_zmalloc("bnxt_mark_table", BNXT_MARK_TABLE_SZ, 0);
 465         if (!bp->mark_table)
 466                 PMD_DRV_LOG(ERR, "Allocation of mark table failed\n");
 467
 468         return 0;
 469
 470 err_free:
 471         rte_free(intr_handle->intr_vec);
 472 err_disable:
 473         rte_intr_efd_disable(intr_handle);
 474 err_out:
 475         /* Some of the error status returned by FW may not be from errno.h */
 476         if (rc > 0)
 477                 rc = -EIO;
 478
 479         return rc;
 480 }
 481
 482 static int bnxt_shutdown_nic(struct bnxt *bp)
 483 {
 484         bnxt_free_all_hwrm_resources(bp);
 485         bnxt_free_all_filters(bp);
 486         bnxt_free_all_vnics(bp);
 487         return 0;
 488 }
 489
 490 /*
 491  * Device configuration and status function
 492  */
 493
 494 static int bnxt_dev_info_get_op(struct rte_eth_dev *eth_dev,
 495                                 struct rte_eth_dev_info *dev_info)
 496 {
 497         struct rte_pci_device *pdev = RTE_DEV_TO_PCI(eth_dev->device);
 498         struct bnxt *bp = eth_dev->data->dev_private;
 499         uint16_t max_vnics, i, j, vpool, vrxq;
 500         unsigned int max_rx_rings;
 501         int rc;
 502
 503         rc = is_bnxt_in_error(bp);
 504         if (rc)
 505                 return rc;
 506
 507         /* MAC Specifics */
 508         dev_info->max_mac_addrs = bp->max_l2_ctx;
 509         dev_info->max_hash_mac_addrs = 0;
 510
 511         /* PF/VF specifics */
 512         if (BNXT_PF(bp))
 513                 dev_info->max_vfs = pdev->max_vfs;
 514
 515         max_rx_rings = BNXT_MAX_RINGS(bp);
 516         /* For the sake of symmetry, max_rx_queues = max_tx_queues */
 517         dev_info->max_rx_queues = max_rx_rings;
 518         dev_info->max_tx_queues = max_rx_rings;
 519         dev_info->reta_size = bnxt_rss_hash_tbl_size(bp);
 520         dev_info->hash_key_size = 40;
 521         max_vnics = bp->max_vnics;
 522
 523         /* MTU specifics */
 524         dev_info->min_mtu = RTE_ETHER_MIN_MTU;
 525         dev_info->max_mtu = BNXT_MAX_MTU;
 526
 527         /* Fast path specifics */
 528         dev_info->min_rx_bufsize = 1;
 529         dev_info->max_rx_pktlen = BNXT_MAX_PKT_LEN;
 530
 531         dev_info->rx_offload_capa = BNXT_DEV_RX_OFFLOAD_SUPPORT;
 532         if (bp->flags & BNXT_FLAG_PTP_SUPPORTED)
 533                 dev_info->rx_offload_capa |= DEV_RX_OFFLOAD_TIMESTAMP;
 534         dev_info->tx_offload_capa = BNXT_DEV_TX_OFFLOAD_SUPPORT;
 535         dev_info->flow_type_rss_offloads = BNXT_ETH_RSS_SUPPORT;
 536
 537         /* *INDENT-OFF* */
 538         dev_info->default_rxconf = (struct rte_eth_rxconf) {
 539                 .rx_thresh = {
 540                         .pthresh = 8,
 541                         .hthresh = 8,
 542                         .wthresh = 0,
 543                 },
 544                 .rx_free_thresh = 32,
 545                 /* If no descriptors available, pkts are dropped by default */
 546                 .rx_drop_en = 1,
 547         };
 548
 549         dev_info->default_txconf = (struct rte_eth_txconf) {
 550                 .tx_thresh = {
 551                         .pthresh = 32,
 552                         .hthresh = 0,
 553                         .wthresh = 0,
 554                 },
 555                 .tx_free_thresh = 32,
 556                 .tx_rs_thresh = 32,
 557         };
 558         eth_dev->data->dev_conf.intr_conf.lsc = 1;
 559
 560         eth_dev->data->dev_conf.intr_conf.rxq = 1;
 561         dev_info->rx_desc_lim.nb_min = BNXT_MIN_RING_DESC;
 562         dev_info->rx_desc_lim.nb_max = BNXT_MAX_RX_RING_DESC;
 563         dev_info->tx_desc_lim.nb_min = BNXT_MIN_RING_DESC;
 564         dev_info->tx_desc_lim.nb_max = BNXT_MAX_TX_RING_DESC;
 565
 566         /* *INDENT-ON* */
 567
 568         /*
 569          * TODO: default_rxconf, default_txconf, rx_desc_lim, and tx_desc_lim
 570          *       need further investigation.
 571          */
 572
 573         /* VMDq resources */
 574         vpool = 64; /* ETH_64_POOLS */
 575         vrxq = 128; /* ETH_VMDQ_DCB_NUM_QUEUES */
 576         for (i = 0; i < 4; vpool >>= 1, i++) {
 577                 if (max_vnics > vpool) {
 578                         for (j = 0; j < 5; vrxq >>= 1, j++) {
 579                                 if (dev_info->max_rx_queues > vrxq) {
 580                                         if (vpool > vrxq)
 581                                                 vpool = vrxq;
 582                                         goto found;
 583                                 }
 584                         }
 585                         /* Not enough resources to support VMDq */
 586                         break;
 587                 }
 588         }
 589         /* Not enough resources to support VMDq */
 590         vpool = 0;
 591         vrxq = 0;
 592 found:
 593         dev_info->max_vmdq_pools = vpool;
 594         dev_info->vmdq_queue_num = vrxq;
 595
 596         dev_info->vmdq_pool_base = 0;
 597         dev_info->vmdq_queue_base = 0;
 598
 599         return 0;
 600 }
 601
 602 /* Configure the device based on the configuration provided */
 603 static int bnxt_dev_configure_op(struct rte_eth_dev *eth_dev)
 604 {
 605         struct bnxt *bp = eth_dev->data->dev_private;
 606         uint64_t rx_offloads = eth_dev->data->dev_conf.rxmode.offloads;
 607         int rc;
 608
 609         bp->rx_queues = (void *)eth_dev->data->rx_queues;
 610         bp->tx_queues = (void *)eth_dev->data->tx_queues;
 611         bp->tx_nr_rings = eth_dev->data->nb_tx_queues;
 612         bp->rx_nr_rings = eth_dev->data->nb_rx_queues;
 613
 614         rc = is_bnxt_in_error(bp);
 615         if (rc)
 616                 return rc;
 617
 618         if (BNXT_VF(bp) && (bp->flags & BNXT_FLAG_NEW_RM)) {
 619                 rc = bnxt_hwrm_check_vf_rings(bp);
 620                 if (rc) {
 621                         PMD_DRV_LOG(ERR, "HWRM insufficient resources\n");
 622                         return -ENOSPC;
 623                 }
 624
 625                 /* If a resource has already been allocated - in this case
 626                  * it is the async completion ring, free it. Reallocate it after
 627                  * resource reservation. This will ensure the resource counts
 628                  * are calculated correctly.
 629                  */
 630
 631                 pthread_mutex_lock(&bp->def_cp_lock);
 632
 633                 if (!BNXT_HAS_NQ(bp) && bp->async_cp_ring) {
 634                         bnxt_disable_int(bp);
 635                         bnxt_free_cp_ring(bp, bp->async_cp_ring);
 636                 }
 637
 638                 rc = bnxt_hwrm_func_reserve_vf_resc(bp, false);
 639                 if (rc) {
 640                         PMD_DRV_LOG(ERR, "HWRM resource alloc fail:%x\n", rc);
 641                         pthread_mutex_unlock(&bp->def_cp_lock);
 642                         return -ENOSPC;
 643                 }
 644
 645                 if (!BNXT_HAS_NQ(bp) && bp->async_cp_ring) {
 646                         rc = bnxt_alloc_async_cp_ring(bp);
 647                         if (rc) {
 648                                 pthread_mutex_unlock(&bp->def_cp_lock);
 649                                 return rc;
 650                         }
 651                         bnxt_enable_int(bp);
 652                 }
 653
 654                 pthread_mutex_unlock(&bp->def_cp_lock);
 655         } else {
 656                 /* legacy driver needs to get updated values */
 657                 rc = bnxt_hwrm_func_qcaps(bp);
 658                 if (rc) {
 659                         PMD_DRV_LOG(ERR, "hwrm func qcaps fail:%d\n", rc);
 660                         return rc;
 661                 }
 662         }
 663
 664         /* Inherit new configurations */
 665         if (eth_dev->data->nb_rx_queues > bp->max_rx_rings ||
 666             eth_dev->data->nb_tx_queues > bp->max_tx_rings ||
 667             eth_dev->data->nb_rx_queues + eth_dev->data->nb_tx_queues
 668                 + BNXT_NUM_ASYNC_CPR(bp) > bp->max_cp_rings ||
 669             eth_dev->data->nb_rx_queues + eth_dev->data->nb_tx_queues >
 670             bp->max_stat_ctx)
 671                 goto resource_error;
 672
 673         if (BNXT_HAS_RING_GRPS(bp) &&
 674             (uint32_t)(eth_dev->data->nb_rx_queues) > bp->max_ring_grps)
 675                 goto resource_error;
 676
 677         if (!(eth_dev->data->dev_conf.rxmode.mq_mode & ETH_MQ_RX_RSS) &&
 678             bp->max_vnics < eth_dev->data->nb_rx_queues)
 679                 goto resource_error;
 680
 681         bp->rx_cp_nr_rings = bp->rx_nr_rings;
 682         bp->tx_cp_nr_rings = bp->tx_nr_rings;
 683
 684         if (eth_dev->data->dev_conf.rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG)
 685                 rx_offloads |= DEV_RX_OFFLOAD_RSS_HASH;
 686         eth_dev->data->dev_conf.rxmode.offloads = rx_offloads;
 687
 688         if (rx_offloads & DEV_RX_OFFLOAD_JUMBO_FRAME) {
 689                 eth_dev->data->mtu =
 690                         eth_dev->data->dev_conf.rxmode.max_rx_pkt_len -
 691                         RTE_ETHER_HDR_LEN - RTE_ETHER_CRC_LEN - VLAN_TAG_SIZE *
 692                         BNXT_NUM_VLANS;
 693                 bnxt_mtu_set_op(eth_dev, eth_dev->data->mtu);
 694         }
 695         return 0;
 696
 697 resource_error:
 698         PMD_DRV_LOG(ERR,
 699                     "Insufficient resources to support requested config\n");
 700         PMD_DRV_LOG(ERR,
 701                     "Num Queues Requested: Tx %d, Rx %d\n",
 702                     eth_dev->data->nb_tx_queues,
 703                     eth_dev->data->nb_rx_queues);
 704         PMD_DRV_LOG(ERR,
 705                     "MAX: TxQ %d, RxQ %d, CQ %d Stat %d, Grp %d, Vnic %d\n",
 706                     bp->max_tx_rings, bp->max_rx_rings, bp->max_cp_rings,
 707                     bp->max_stat_ctx, bp->max_ring_grps, bp->max_vnics);
 708         return -ENOSPC;
 709 }
 710
 711 static void bnxt_print_link_info(struct rte_eth_dev *eth_dev)
 712 {
 713         struct rte_eth_link *link = &eth_dev->data->dev_link;
 714
 715         if (link->link_status)
 716                 PMD_DRV_LOG(INFO, "Port %d Link Up - speed %u Mbps - %s\n",
 717                         eth_dev->data->port_id,
 718                         (uint32_t)link->link_speed,
 719                         (link->link_duplex == ETH_LINK_FULL_DUPLEX) ?
 720                         ("full-duplex") : ("half-duplex\n"));
 721         else
 722                 PMD_DRV_LOG(INFO, "Port %d Link Down\n",
 723                         eth_dev->data->port_id);
 724 }
 725
 726 /*
 727  * Determine whether the current configuration requires support for scattered
 728  * receive; return 1 if scattered receive is required and 0 if not.
 729  */
 730 static int bnxt_scattered_rx(struct rte_eth_dev *eth_dev)
 731 {
 732         uint16_t buf_size;
 733         int i;
 734
 735         if (eth_dev->data->dev_conf.rxmode.offloads & DEV_RX_OFFLOAD_SCATTER)
 736                 return 1;
 737
 738         for (i = 0; i < eth_dev->data->nb_rx_queues; i++) {
 739                 struct bnxt_rx_queue *rxq = eth_dev->data->rx_queues[i];
 740
 741                 buf_size = (uint16_t)(rte_pktmbuf_data_room_size(rxq->mb_pool) -
 742                                       RTE_PKTMBUF_HEADROOM);
 743                 if (eth_dev->data->dev_conf.rxmode.max_rx_pkt_len > buf_size)
 744                         return 1;
 745         }
 746         return 0;
 747 }
 748
 749 static eth_rx_burst_t
 750 bnxt_receive_function(struct rte_eth_dev *eth_dev)
 751 {
 752         struct bnxt *bp = eth_dev->data->dev_private;
 753
 754 #ifdef RTE_ARCH_X86
 755 #ifndef RTE_LIBRTE_IEEE1588
 756         /*
 757          * Vector mode receive can be enabled only if scatter rx is not
 758          * in use and rx offloads are limited to VLAN stripping and
 759          * CRC stripping.
 760          */
 761         if (!eth_dev->data->scattered_rx &&
 762             !(eth_dev->data->dev_conf.rxmode.offloads &
 763               ~(DEV_RX_OFFLOAD_VLAN_STRIP |
 764                 DEV_RX_OFFLOAD_KEEP_CRC |
 765                 DEV_RX_OFFLOAD_JUMBO_FRAME |
 766                 DEV_RX_OFFLOAD_IPV4_CKSUM |
 767                 DEV_RX_OFFLOAD_UDP_CKSUM |
 768                 DEV_RX_OFFLOAD_TCP_CKSUM |
 769                 DEV_RX_OFFLOAD_OUTER_IPV4_CKSUM |
 770                 DEV_RX_OFFLOAD_RSS_HASH |
 771                 DEV_RX_OFFLOAD_VLAN_FILTER))) {
 772                 PMD_DRV_LOG(INFO, "Using vector mode receive for port %d\n",
 773                             eth_dev->data->port_id);
 774                 bp->flags |= BNXT_FLAG_RX_VECTOR_PKT_MODE;
 775                 return bnxt_recv_pkts_vec;
 776         }
 777         PMD_DRV_LOG(INFO, "Vector mode receive disabled for port %d\n",
 778                     eth_dev->data->port_id);
 779         PMD_DRV_LOG(INFO,
 780                     "Port %d scatter: %d rx offload: %" PRIX64 "\n",
 781                     eth_dev->data->port_id,
 782                     eth_dev->data->scattered_rx,
 783                     eth_dev->data->dev_conf.rxmode.offloads);
 784 #endif
 785 #endif
 786         bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
 787         return bnxt_recv_pkts;
 788 }
 789
 790 static eth_tx_burst_t
 791 bnxt_transmit_function(__rte_unused struct rte_eth_dev *eth_dev)
 792 {
 793 #ifdef RTE_ARCH_X86
 794 #ifndef RTE_LIBRTE_IEEE1588
 795         /*
 796          * Vector mode transmit can be enabled only if not using scatter rx
 797          * or tx offloads.
 798          */
 799         if (!eth_dev->data->scattered_rx &&
 800             !eth_dev->data->dev_conf.txmode.offloads) {
 801                 PMD_DRV_LOG(INFO, "Using vector mode transmit for port %d\n",
 802                             eth_dev->data->port_id);
 803                 return bnxt_xmit_pkts_vec;
 804         }
 805         PMD_DRV_LOG(INFO, "Vector mode transmit disabled for port %d\n",
 806                     eth_dev->data->port_id);
 807         PMD_DRV_LOG(INFO,
 808                     "Port %d scatter: %d tx offload: %" PRIX64 "\n",
 809                     eth_dev->data->port_id,
 810                     eth_dev->data->scattered_rx,
 811                     eth_dev->data->dev_conf.txmode.offloads);
 812 #endif
 813 #endif
 814         return bnxt_xmit_pkts;
 815 }
 816
 817 static int bnxt_handle_if_change_status(struct bnxt *bp)
 818 {
 819         int rc;
 820
 821         /* Since fw has undergone a reset and lost all contexts,
 822          * set fatal flag to not issue hwrm during cleanup
 823          */
 824         bp->flags |= BNXT_FLAG_FATAL_ERROR;
 825         bnxt_uninit_resources(bp, true);
 826
 827         /* clear fatal flag so that re-init happens */
 828         bp->flags &= ~BNXT_FLAG_FATAL_ERROR;
 829         rc = bnxt_init_resources(bp, true);
 830
 831         bp->flags &= ~BNXT_FLAG_IF_CHANGE_HOT_FW_RESET_DONE;
 832
 833         return rc;
 834 }
 835
 836 static int bnxt_dev_start_op(struct rte_eth_dev *eth_dev)
 837 {
 838         struct bnxt *bp = eth_dev->data->dev_private;
 839         uint64_t rx_offloads = eth_dev->data->dev_conf.rxmode.offloads;
 840         int vlan_mask = 0;
 841         int rc;
 842
 843         if (!eth_dev->data->nb_tx_queues || !eth_dev->data->nb_rx_queues) {
 844                 PMD_DRV_LOG(ERR, "Queues are not configured yet!\n");
 845                 return -EINVAL;
 846         }
 847
 848         if (bp->rx_cp_nr_rings > RTE_ETHDEV_QUEUE_STAT_CNTRS) {
 849                 PMD_DRV_LOG(ERR,
 850                         "RxQ cnt %d > CONFIG_RTE_ETHDEV_QUEUE_STAT_CNTRS %d\n",
 851                         bp->rx_cp_nr_rings, RTE_ETHDEV_QUEUE_STAT_CNTRS);
 852         }
 853
 854         rc = bnxt_hwrm_if_change(bp, 1);
 855         if (!rc) {
 856                 if (bp->flags & BNXT_FLAG_IF_CHANGE_HOT_FW_RESET_DONE) {
 857                         rc = bnxt_handle_if_change_status(bp);
 858                         if (rc)
 859                                 return rc;
 860                 }
 861         }
 862         bnxt_enable_int(bp);
 863
 864         rc = bnxt_init_chip(bp);
 865         if (rc)
 866                 goto error;
 867
 868         eth_dev->data->scattered_rx = bnxt_scattered_rx(eth_dev);
 869
 870         bnxt_link_update(eth_dev, 1, ETH_LINK_UP);
 871
 872         if (rx_offloads & DEV_RX_OFFLOAD_VLAN_FILTER)
 873                 vlan_mask |= ETH_VLAN_FILTER_MASK;
 874         if (rx_offloads & DEV_RX_OFFLOAD_VLAN_STRIP)
 875                 vlan_mask |= ETH_VLAN_STRIP_MASK;
 876         rc = bnxt_vlan_offload_set_op(eth_dev, vlan_mask);
 877         if (rc)
 878                 goto error;
 879
 880         eth_dev->rx_pkt_burst = bnxt_receive_function(eth_dev);
 881         eth_dev->tx_pkt_burst = bnxt_transmit_function(eth_dev);
 882
 883         bp->flags |= BNXT_FLAG_INIT_DONE;
 884         eth_dev->data->dev_started = 1;
 885         bp->dev_stopped = 0;
 886         pthread_mutex_lock(&bp->def_cp_lock);
 887         bnxt_schedule_fw_health_check(bp);
 888         pthread_mutex_unlock(&bp->def_cp_lock);
 889         return 0;
 890
 891 error:
 892         bnxt_hwrm_if_change(bp, 0);
 893         bnxt_shutdown_nic(bp);
 894         bnxt_free_tx_mbufs(bp);
 895         bnxt_free_rx_mbufs(bp);
 896         return rc;
 897 }
 898
 899 static int bnxt_dev_set_link_up_op(struct rte_eth_dev *eth_dev)
 900 {
 901         struct bnxt *bp = eth_dev->data->dev_private;
 902         int rc = 0;
 903
 904         if (!bp->link_info.link_up)
 905                 rc = bnxt_set_hwrm_link_config(bp, true);
 906         if (!rc)
 907                 eth_dev->data->dev_link.link_status = 1;
 908
 909         bnxt_print_link_info(eth_dev);
 910         return rc;
 911 }
 912
 913 static int bnxt_dev_set_link_down_op(struct rte_eth_dev *eth_dev)
 914 {
 915         struct bnxt *bp = eth_dev->data->dev_private;
 916
 917         eth_dev->data->dev_link.link_status = 0;
 918         bnxt_set_hwrm_link_config(bp, false);
 919         bp->link_info.link_up = 0;
 920
 921         return 0;
 922 }
 923
 924 /* Unload the driver, release resources */
 925 static void bnxt_dev_stop_op(struct rte_eth_dev *eth_dev)
 926 {
 927         struct bnxt *bp = eth_dev->data->dev_private;
 928         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
 929         struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
 930
 931         eth_dev->data->dev_started = 0;
 932         /* Prevent crashes when queues are still in use */
 933         eth_dev->rx_pkt_burst = &bnxt_dummy_recv_pkts;
 934         eth_dev->tx_pkt_burst = &bnxt_dummy_xmit_pkts;
 935
 936         bnxt_disable_int(bp);
 937
 938         /* disable uio/vfio intr/eventfd mapping */
 939         rte_intr_disable(intr_handle);
 940
 941         bnxt_cancel_fw_health_check(bp);
 942
 943         bp->flags &= ~BNXT_FLAG_INIT_DONE;
 944         if (bp->eth_dev->data->dev_started) {
 945                 /* TBD: STOP HW queues DMA */
 946                 eth_dev->data->dev_link.link_status = 0;
 947         }
 948         bnxt_dev_set_link_down_op(eth_dev);
 949
 950         /* Wait for link to be reset and the async notification to process.
 951          * During reset recovery, there is no need to wait
 952          */
 953         if (!is_bnxt_in_error(bp))
 954                 bnxt_link_update(eth_dev, 1, ETH_LINK_DOWN);
 955
 956         /* Clean queue intr-vector mapping */
 957         rte_intr_efd_disable(intr_handle);
 958         if (intr_handle->intr_vec != NULL) {
 959                 rte_free(intr_handle->intr_vec);
 960                 intr_handle->intr_vec = NULL;
 961         }
 962
 963         bnxt_hwrm_port_clr_stats(bp);
 964         bnxt_free_tx_mbufs(bp);
 965         bnxt_free_rx_mbufs(bp);
 966         /* Process any remaining notifications in default completion queue */
 967         bnxt_int_handler(eth_dev);
 968         bnxt_shutdown_nic(bp);
 969         bnxt_hwrm_if_change(bp, 0);
 970         memset(bp->mark_table, 0, BNXT_MARK_TABLE_SZ);
 971         bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
 972         bp->dev_stopped = 1;
 973         bp->rx_cosq_cnt = 0;
 974 }
 975
 976 static void bnxt_dev_close_op(struct rte_eth_dev *eth_dev)
 977 {
 978         struct bnxt *bp = eth_dev->data->dev_private;
 979
 980         if (bp->dev_stopped == 0)
 981                 bnxt_dev_stop_op(eth_dev);
 982
 983         if (eth_dev->data->mac_addrs != NULL) {
 984                 rte_free(eth_dev->data->mac_addrs);
 985                 eth_dev->data->mac_addrs = NULL;
 986         }
 987         if (bp->grp_info != NULL) {
 988                 rte_free(bp->grp_info);
 989                 bp->grp_info = NULL;
 990         }
 991
 992         rte_free(bp->mark_table);
 993         bp->mark_table = NULL;
 994
 995         bnxt_dev_uninit(eth_dev);
 996 }
 997
 998 static void bnxt_mac_addr_remove_op(struct rte_eth_dev *eth_dev,
 999                                     uint32_t index)
1000 {
1001         struct bnxt *bp = eth_dev->data->dev_private;
1002         uint64_t pool_mask = eth_dev->data->mac_pool_sel[index];
1003         struct bnxt_vnic_info *vnic;
1004         struct bnxt_filter_info *filter, *temp_filter;
1005         uint32_t i;
1006
1007         if (is_bnxt_in_error(bp))
1008                 return;
1009
1010         /*
1011          * Loop through all VNICs from the specified filter flow pools to
1012          * remove the corresponding MAC addr filter
1013          */
1014         for (i = 0; i < bp->nr_vnics; i++) {
1015                 if (!(pool_mask & (1ULL << i)))
1016                         continue;
1017
1018                 vnic = &bp->vnic_info[i];
1019                 filter = STAILQ_FIRST(&vnic->filter);
1020                 while (filter) {
1021                         temp_filter = STAILQ_NEXT(filter, next);
1022                         if (filter->mac_index == index) {
1023                                 STAILQ_REMOVE(&vnic->filter, filter,
1024                                                 bnxt_filter_info, next);
1025                                 bnxt_hwrm_clear_l2_filter(bp, filter);
1026                                 bnxt_free_filter(bp, filter);
1027                         }
1028                         filter = temp_filter;
1029                 }
1030         }
1031 }
1032
1033 static int bnxt_add_mac_filter(struct bnxt *bp, struct bnxt_vnic_info *vnic,
1034                                struct rte_ether_addr *mac_addr, uint32_t index,
1035                                uint32_t pool)
1036 {
1037         struct bnxt_filter_info *filter;
1038         int rc = 0;
1039
1040         /* Attach requested MAC address to the new l2_filter */
1041         STAILQ_FOREACH(filter, &vnic->filter, next) {
1042                 if (filter->mac_index == index) {
1043                         PMD_DRV_LOG(DEBUG,
1044                                     "MAC addr already existed for pool %d\n",
1045                                     pool);
1046                         return 0;
1047                 }
1048         }
1049
1050         filter = bnxt_alloc_filter(bp);
1051         if (!filter) {
1052                 PMD_DRV_LOG(ERR, "L2 filter alloc failed\n");
1053                 return -ENODEV;
1054         }
1055
1056         /* bnxt_alloc_filter copies default MAC to filter->l2_addr. So,
1057          * if the MAC that's been programmed now is a different one, then,
1058          * copy that addr to filter->l2_addr
1059          */
1060         if (mac_addr)
1061                 memcpy(filter->l2_addr, mac_addr, RTE_ETHER_ADDR_LEN);
1062         filter->flags |= HWRM_CFA_L2_FILTER_ALLOC_INPUT_FLAGS_OUTERMOST;
1063
1064         rc = bnxt_hwrm_set_l2_filter(bp, vnic->fw_vnic_id, filter);
1065         if (!rc) {
1066                 filter->mac_index = index;
1067                 if (filter->mac_index == 0)
1068                         STAILQ_INSERT_HEAD(&vnic->filter, filter, next);
1069                 else
1070                         STAILQ_INSERT_TAIL(&vnic->filter, filter, next);
1071         } else {
1072                 bnxt_free_filter(bp, filter);
1073         }
1074
1075         return rc;
1076 }
1077
1078 static int bnxt_mac_addr_add_op(struct rte_eth_dev *eth_dev,
1079                                 struct rte_ether_addr *mac_addr,
1080                                 uint32_t index, uint32_t pool)
1081 {
1082         struct bnxt *bp = eth_dev->data->dev_private;
1083         struct bnxt_vnic_info *vnic = &bp->vnic_info[pool];
1084         int rc = 0;
1085
1086         rc = is_bnxt_in_error(bp);
1087         if (rc)
1088                 return rc;
1089
1090         if (BNXT_VF(bp) & !BNXT_VF_IS_TRUSTED(bp)) {
1091                 PMD_DRV_LOG(ERR, "Cannot add MAC address to a VF interface\n");
1092                 return -ENOTSUP;
1093         }
1094
1095         if (!vnic) {
1096                 PMD_DRV_LOG(ERR, "VNIC not found for pool %d!\n", pool);
1097                 return -EINVAL;
1098         }
1099
1100         rc = bnxt_add_mac_filter(bp, vnic, mac_addr, index, pool);
1101
1102         return rc;
1103 }
1104
1105 int bnxt_link_update(struct rte_eth_dev *eth_dev, int wait_to_complete,
1106                      bool exp_link_status)
1107 {
1108         int rc = 0;
1109         struct bnxt *bp = eth_dev->data->dev_private;
1110         struct rte_eth_link new;
1111         int cnt = exp_link_status ? BNXT_LINK_UP_WAIT_CNT :
1112                   BNXT_LINK_DOWN_WAIT_CNT;
1113
1114         rc = is_bnxt_in_error(bp);
1115         if (rc)
1116                 return rc;
1117
1118         memset(&new, 0, sizeof(new));
1119         do {
1120                 /* Retrieve link info from hardware */
1121                 rc = bnxt_get_hwrm_link_config(bp, &new);
1122                 if (rc) {
1123                         new.link_speed = ETH_LINK_SPEED_100M;
1124                         new.link_duplex = ETH_LINK_FULL_DUPLEX;
1125                         PMD_DRV_LOG(ERR,
1126                                 "Failed to retrieve link rc = 0x%x!\n", rc);
1127                         goto out;
1128                 }
1129
1130                 if (!wait_to_complete || new.link_status == exp_link_status)
1131                         break;
1132
1133                 rte_delay_ms(BNXT_LINK_WAIT_INTERVAL);
1134         } while (cnt--);
1135
1136 out:
1137         /* Timed out or success */
1138         if (new.link_status != eth_dev->data->dev_link.link_status ||
1139         new.link_speed != eth_dev->data->dev_link.link_speed) {
1140                 rte_eth_linkstatus_set(eth_dev, &new);
1141
1142                 _rte_eth_dev_callback_process(eth_dev,
1143                                               RTE_ETH_EVENT_INTR_LSC,
1144                                               NULL);
1145
1146                 bnxt_print_link_info(eth_dev);
1147         }
1148
1149         return rc;
1150 }
1151
1152 static int bnxt_link_update_op(struct rte_eth_dev *eth_dev,
1153                                int wait_to_complete)
1154 {
1155         return bnxt_link_update(eth_dev, wait_to_complete, ETH_LINK_UP);
1156 }
1157
1158 static int bnxt_promiscuous_enable_op(struct rte_eth_dev *eth_dev)
1159 {
1160         struct bnxt *bp = eth_dev->data->dev_private;
1161         struct bnxt_vnic_info *vnic;
1162         uint32_t old_flags;
1163         int rc;
1164
1165         rc = is_bnxt_in_error(bp);
1166         if (rc)
1167                 return rc;
1168
1169         if (bp->vnic_info == NULL)
1170                 return 0;
1171
1172         vnic = BNXT_GET_DEFAULT_VNIC(bp);
1173
1174         old_flags = vnic->flags;
1175         vnic->flags |= BNXT_VNIC_INFO_PROMISC;
1176         rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1177         if (rc != 0)
1178                 vnic->flags = old_flags;
1179
1180         return rc;
1181 }
1182
1183 static int bnxt_promiscuous_disable_op(struct rte_eth_dev *eth_dev)
1184 {
1185         struct bnxt *bp = eth_dev->data->dev_private;
1186         struct bnxt_vnic_info *vnic;
1187         uint32_t old_flags;
1188         int rc;
1189
1190         rc = is_bnxt_in_error(bp);
1191         if (rc)
1192                 return rc;
1193
1194         if (bp->vnic_info == NULL)
1195                 return 0;
1196
1197         vnic = BNXT_GET_DEFAULT_VNIC(bp);
1198
1199         old_flags = vnic->flags;
1200         vnic->flags &= ~BNXT_VNIC_INFO_PROMISC;
1201         rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1202         if (rc != 0)
1203                 vnic->flags = old_flags;
1204
1205         return rc;
1206 }
1207
1208 static int bnxt_allmulticast_enable_op(struct rte_eth_dev *eth_dev)
1209 {
1210         struct bnxt *bp = eth_dev->data->dev_private;
1211         struct bnxt_vnic_info *vnic;
1212         uint32_t old_flags;
1213         int rc;
1214
1215         rc = is_bnxt_in_error(bp);
1216         if (rc)
1217                 return rc;
1218
1219         if (bp->vnic_info == NULL)
1220                 return 0;
1221
1222         vnic = BNXT_GET_DEFAULT_VNIC(bp);
1223
1224         old_flags = vnic->flags;
1225         vnic->flags |= BNXT_VNIC_INFO_ALLMULTI;
1226         rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1227         if (rc != 0)
1228                 vnic->flags = old_flags;
1229
1230         return rc;
1231 }
1232
1233 static int bnxt_allmulticast_disable_op(struct rte_eth_dev *eth_dev)
1234 {
1235         struct bnxt *bp = eth_dev->data->dev_private;
1236         struct bnxt_vnic_info *vnic;
1237         uint32_t old_flags;
1238         int rc;
1239
1240         rc = is_bnxt_in_error(bp);
1241         if (rc)
1242                 return rc;
1243
1244         if (bp->vnic_info == NULL)
1245                 return 0;
1246
1247         vnic = BNXT_GET_DEFAULT_VNIC(bp);
1248
1249         old_flags = vnic->flags;
1250         vnic->flags &= ~BNXT_VNIC_INFO_ALLMULTI;
1251         rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1252         if (rc != 0)
1253                 vnic->flags = old_flags;
1254
1255         return rc;
1256 }
1257
1258 /* Return bnxt_rx_queue pointer corresponding to a given rxq. */
1259 static struct bnxt_rx_queue *bnxt_qid_to_rxq(struct bnxt *bp, uint16_t qid)
1260 {
1261         if (qid >= bp->rx_nr_rings)
1262                 return NULL;
1263
1264         return bp->eth_dev->data->rx_queues[qid];
1265 }
1266
1267 /* Return rxq corresponding to a given rss table ring/group ID. */
1268 static uint16_t bnxt_rss_to_qid(struct bnxt *bp, uint16_t fwr)
1269 {
1270         struct bnxt_rx_queue *rxq;
1271         unsigned int i;
1272
1273         if (!BNXT_HAS_RING_GRPS(bp)) {
1274                 for (i = 0; i < bp->rx_nr_rings; i++) {
1275                         rxq = bp->eth_dev->data->rx_queues[i];
1276                         if (rxq->rx_ring->rx_ring_struct->fw_ring_id == fwr)
1277                                 return rxq->index;
1278                 }
1279         } else {
1280                 for (i = 0; i < bp->rx_nr_rings; i++) {
1281                         if (bp->grp_info[i].fw_grp_id == fwr)
1282                                 return i;
1283                 }
1284         }
1285
1286         return INVALID_HW_RING_ID;
1287 }
1288
1289 static int bnxt_reta_update_op(struct rte_eth_dev *eth_dev,
1290                             struct rte_eth_rss_reta_entry64 *reta_conf,
1291                             uint16_t reta_size)
1292 {
1293         struct bnxt *bp = eth_dev->data->dev_private;
1294         struct rte_eth_conf *dev_conf = &bp->eth_dev->data->dev_conf;
1295         struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
1296         uint16_t tbl_size = bnxt_rss_hash_tbl_size(bp);
1297         uint16_t idx, sft;
1298         int i, rc;
1299
1300         rc = is_bnxt_in_error(bp);
1301         if (rc)
1302                 return rc;
1303
1304         if (!vnic->rss_table)
1305                 return -EINVAL;
1306
1307         if (!(dev_conf->rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG))
1308                 return -EINVAL;
1309
1310         if (reta_size != tbl_size) {
1311                 PMD_DRV_LOG(ERR, "The configured hash table lookup size "
1312                         "(%d) must equal the size supported by the hardware "
1313                         "(%d)\n", reta_size, tbl_size);
1314                 return -EINVAL;
1315         }
1316
1317         for (i = 0; i < reta_size; i++) {
1318                 struct bnxt_rx_queue *rxq;
1319
1320                 idx = i / RTE_RETA_GROUP_SIZE;
1321                 sft = i % RTE_RETA_GROUP_SIZE;
1322
1323                 if (!(reta_conf[idx].mask & (1ULL << sft)))
1324                         continue;
1325
1326                 rxq = bnxt_qid_to_rxq(bp, reta_conf[idx].reta[sft]);
1327                 if (!rxq) {
1328                         PMD_DRV_LOG(ERR, "Invalid ring in reta_conf.\n");
1329                         return -EINVAL;
1330                 }
1331
1332                 if (BNXT_CHIP_THOR(bp)) {
1333                         vnic->rss_table[i * 2] =
1334                                 rxq->rx_ring->rx_ring_struct->fw_ring_id;
1335                         vnic->rss_table[i * 2 + 1] =
1336                                 rxq->cp_ring->cp_ring_struct->fw_ring_id;
1337                 } else {
1338                         vnic->rss_table[i] =
1339                             vnic->fw_grp_ids[reta_conf[idx].reta[sft]];
1340                 }
1341         }
1342
1343         bnxt_hwrm_vnic_rss_cfg(bp, vnic);
1344         return 0;
1345 }
1346
1347 static int bnxt_reta_query_op(struct rte_eth_dev *eth_dev,
1348                               struct rte_eth_rss_reta_entry64 *reta_conf,
1349                               uint16_t reta_size)
1350 {
1351         struct bnxt *bp = eth_dev->data->dev_private;
1352         struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
1353         uint16_t tbl_size = bnxt_rss_hash_tbl_size(bp);
1354         uint16_t idx, sft, i;
1355         int rc;
1356
1357         rc = is_bnxt_in_error(bp);
1358         if (rc)
1359                 return rc;
1360
1361         /* Retrieve from the default VNIC */
1362         if (!vnic)
1363                 return -EINVAL;
1364         if (!vnic->rss_table)
1365                 return -EINVAL;
1366
1367         if (reta_size != tbl_size) {
1368                 PMD_DRV_LOG(ERR, "The configured hash table lookup size "
1369                         "(%d) must equal the size supported by the hardware "
1370                         "(%d)\n", reta_size, tbl_size);
1371                 return -EINVAL;
1372         }
1373
1374         for (idx = 0, i = 0; i < reta_size; i++) {
1375                 idx = i / RTE_RETA_GROUP_SIZE;
1376                 sft = i % RTE_RETA_GROUP_SIZE;
1377
1378                 if (reta_conf[idx].mask & (1ULL << sft)) {
1379                         uint16_t qid;
1380
1381                         if (BNXT_CHIP_THOR(bp))
1382                                 qid = bnxt_rss_to_qid(bp,
1383                                                       vnic->rss_table[i * 2]);
1384                         else
1385                                 qid = bnxt_rss_to_qid(bp, vnic->rss_table[i]);
1386
1387                         if (qid == INVALID_HW_RING_ID) {
1388                                 PMD_DRV_LOG(ERR, "Inv. entry in rss table.\n");
1389                                 return -EINVAL;
1390                         }
1391                         reta_conf[idx].reta[sft] = qid;
1392                 }
1393         }
1394
1395         return 0;
1396 }
1397
1398 static int bnxt_rss_hash_update_op(struct rte_eth_dev *eth_dev,
1399                                    struct rte_eth_rss_conf *rss_conf)
1400 {
1401         struct bnxt *bp = eth_dev->data->dev_private;
1402         struct rte_eth_conf *dev_conf = &bp->eth_dev->data->dev_conf;
1403         struct bnxt_vnic_info *vnic;
1404         int rc;
1405
1406         rc = is_bnxt_in_error(bp);
1407         if (rc)
1408                 return rc;
1409
1410         /*
1411          * If RSS enablement were different than dev_configure,
1412          * then return -EINVAL
1413          */
1414         if (dev_conf->rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG) {
1415                 if (!rss_conf->rss_hf)
1416                         PMD_DRV_LOG(ERR, "Hash type NONE\n");
1417         } else {
1418                 if (rss_conf->rss_hf & BNXT_ETH_RSS_SUPPORT)
1419                         return -EINVAL;
1420         }
1421
1422         bp->flags |= BNXT_FLAG_UPDATE_HASH;
1423         memcpy(&bp->rss_conf, rss_conf, sizeof(*rss_conf));
1424
1425         /* Update the default RSS VNIC(s) */
1426         vnic = BNXT_GET_DEFAULT_VNIC(bp);
1427         vnic->hash_type = bnxt_rte_to_hwrm_hash_types(rss_conf->rss_hf);
1428
1429         /*
1430          * If hashkey is not specified, use the previously configured
1431          * hashkey
1432          */
1433         if (!rss_conf->rss_key)
1434                 goto rss_config;
1435
1436         if (rss_conf->rss_key_len != HW_HASH_KEY_SIZE) {
1437                 PMD_DRV_LOG(ERR,
1438                             "Invalid hashkey length, should be 16 bytes\n");
1439                 return -EINVAL;
1440         }
1441         memcpy(vnic->rss_hash_key, rss_conf->rss_key, rss_conf->rss_key_len);
1442
1443 rss_config:
1444         bnxt_hwrm_vnic_rss_cfg(bp, vnic);
1445         return 0;
1446 }
1447
1448 static int bnxt_rss_hash_conf_get_op(struct rte_eth_dev *eth_dev,
1449                                      struct rte_eth_rss_conf *rss_conf)
1450 {
1451         struct bnxt *bp = eth_dev->data->dev_private;
1452         struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
1453         int len, rc;
1454         uint32_t hash_types;
1455
1456         rc = is_bnxt_in_error(bp);
1457         if (rc)
1458                 return rc;
1459
1460         /* RSS configuration is the same for all VNICs */
1461         if (vnic && vnic->rss_hash_key) {
1462                 if (rss_conf->rss_key) {
1463                         len = rss_conf->rss_key_len <= HW_HASH_KEY_SIZE ?
1464                               rss_conf->rss_key_len : HW_HASH_KEY_SIZE;
1465                         memcpy(rss_conf->rss_key, vnic->rss_hash_key, len);
1466                 }
1467
1468                 hash_types = vnic->hash_type;
1469                 rss_conf->rss_hf = 0;
1470                 if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV4) {
1471                         rss_conf->rss_hf |= ETH_RSS_IPV4;
1472                         hash_types &= ~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV4;
1473                 }
1474                 if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV4) {
1475                         rss_conf->rss_hf |= ETH_RSS_NONFRAG_IPV4_TCP;
1476                         hash_types &=
1477                                 ~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV4;
1478                 }
1479                 if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV4) {
1480                         rss_conf->rss_hf |= ETH_RSS_NONFRAG_IPV4_UDP;
1481                         hash_types &=
1482                                 ~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV4;
1483                 }
1484                 if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV6) {
1485                         rss_conf->rss_hf |= ETH_RSS_IPV6;
1486                         hash_types &= ~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV6;
1487                 }
1488                 if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV6) {
1489                         rss_conf->rss_hf |= ETH_RSS_NONFRAG_IPV6_TCP;
1490                         hash_types &=
1491                                 ~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV6;
1492                 }
1493                 if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV6) {
1494                         rss_conf->rss_hf |= ETH_RSS_NONFRAG_IPV6_UDP;
1495                         hash_types &=
1496                                 ~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV6;
1497                 }
1498                 if (hash_types) {
1499                         PMD_DRV_LOG(ERR,
1500                                 "Unknwon RSS config from firmware (%08x), RSS disabled",
1501                                 vnic->hash_type);
1502                         return -ENOTSUP;
1503                 }
1504         } else {
1505                 rss_conf->rss_hf = 0;
1506         }
1507         return 0;
1508 }
1509
1510 static int bnxt_flow_ctrl_get_op(struct rte_eth_dev *dev,
1511                                struct rte_eth_fc_conf *fc_conf)
1512 {
1513         struct bnxt *bp = dev->data->dev_private;
1514         struct rte_eth_link link_info;
1515         int rc;
1516
1517         rc = is_bnxt_in_error(bp);
1518         if (rc)
1519                 return rc;
1520
1521         rc = bnxt_get_hwrm_link_config(bp, &link_info);
1522         if (rc)
1523                 return rc;
1524
1525         memset(fc_conf, 0, sizeof(*fc_conf));
1526         if (bp->link_info.auto_pause)
1527                 fc_conf->autoneg = 1;
1528         switch (bp->link_info.pause) {
1529         case 0:
1530                 fc_conf->mode = RTE_FC_NONE;
1531                 break;
1532         case HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_TX:
1533                 fc_conf->mode = RTE_FC_TX_PAUSE;
1534                 break;
1535         case HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_RX:
1536                 fc_conf->mode = RTE_FC_RX_PAUSE;
1537                 break;
1538         case (HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_TX |
1539                         HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_RX):
1540                 fc_conf->mode = RTE_FC_FULL;
1541                 break;
1542         }
1543         return 0;
1544 }
1545
1546 static int bnxt_flow_ctrl_set_op(struct rte_eth_dev *dev,
1547                                struct rte_eth_fc_conf *fc_conf)
1548 {
1549         struct bnxt *bp = dev->data->dev_private;
1550         int rc;
1551
1552         rc = is_bnxt_in_error(bp);
1553         if (rc)
1554                 return rc;
1555
1556         if (!BNXT_SINGLE_PF(bp) || BNXT_VF(bp)) {
1557                 PMD_DRV_LOG(ERR, "Flow Control Settings cannot be modified\n");
1558                 return -ENOTSUP;
1559         }
1560
1561         switch (fc_conf->mode) {
1562         case RTE_FC_NONE:
1563                 bp->link_info.auto_pause = 0;
1564                 bp->link_info.force_pause = 0;
1565                 break;
1566         case RTE_FC_RX_PAUSE:
1567                 if (fc_conf->autoneg) {
1568                         bp->link_info.auto_pause =
1569                                         HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_RX;
1570                         bp->link_info.force_pause = 0;
1571                 } else {
1572                         bp->link_info.auto_pause = 0;
1573                         bp->link_info.force_pause =
1574                                         HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_RX;
1575                 }
1576                 break;
1577         case RTE_FC_TX_PAUSE:
1578                 if (fc_conf->autoneg) {
1579                         bp->link_info.auto_pause =
1580                                         HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_TX;
1581                         bp->link_info.force_pause = 0;
1582                 } else {
1583                         bp->link_info.auto_pause = 0;
1584                         bp->link_info.force_pause =
1585                                         HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_TX;
1586                 }
1587                 break;
1588         case RTE_FC_FULL:
1589                 if (fc_conf->autoneg) {
1590                         bp->link_info.auto_pause =
1591                                         HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_TX |
1592                                         HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_RX;
1593                         bp->link_info.force_pause = 0;
1594                 } else {
1595                         bp->link_info.auto_pause = 0;
1596                         bp->link_info.force_pause =
1597                                         HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_TX |
1598                                         HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_RX;
1599                 }
1600                 break;
1601         }
1602         return bnxt_set_hwrm_link_config(bp, true);
1603 }
1604
1605 /* Add UDP tunneling port */
1606 static int
1607 bnxt_udp_tunnel_port_add_op(struct rte_eth_dev *eth_dev,
1608                          struct rte_eth_udp_tunnel *udp_tunnel)
1609 {
1610         struct bnxt *bp = eth_dev->data->dev_private;
1611         uint16_t tunnel_type = 0;
1612         int rc = 0;
1613
1614         rc = is_bnxt_in_error(bp);
1615         if (rc)
1616                 return rc;
1617
1618         switch (udp_tunnel->prot_type) {
1619         case RTE_TUNNEL_TYPE_VXLAN:
1620                 if (bp->vxlan_port_cnt) {
1621                         PMD_DRV_LOG(ERR, "Tunnel Port %d already programmed\n",
1622                                 udp_tunnel->udp_port);
1623                         if (bp->vxlan_port != udp_tunnel->udp_port) {
1624                                 PMD_DRV_LOG(ERR, "Only one port allowed\n");
1625                                 return -ENOSPC;
1626                         }
1627                         bp->vxlan_port_cnt++;
1628                         return 0;
1629                 }
1630                 tunnel_type =
1631                         HWRM_TUNNEL_DST_PORT_ALLOC_INPUT_TUNNEL_TYPE_VXLAN;
1632                 bp->vxlan_port_cnt++;
1633                 break;
1634         case RTE_TUNNEL_TYPE_GENEVE:
1635                 if (bp->geneve_port_cnt) {
1636                         PMD_DRV_LOG(ERR, "Tunnel Port %d already programmed\n",
1637                                 udp_tunnel->udp_port);
1638                         if (bp->geneve_port != udp_tunnel->udp_port) {
1639                                 PMD_DRV_LOG(ERR, "Only one port allowed\n");
1640                                 return -ENOSPC;
1641                         }
1642                         bp->geneve_port_cnt++;
1643                         return 0;
1644                 }
1645                 tunnel_type =
1646                         HWRM_TUNNEL_DST_PORT_ALLOC_INPUT_TUNNEL_TYPE_GENEVE;
1647                 bp->geneve_port_cnt++;
1648                 break;
1649         default:
1650                 PMD_DRV_LOG(ERR, "Tunnel type is not supported\n");
1651                 return -ENOTSUP;
1652         }
1653         rc = bnxt_hwrm_tunnel_dst_port_alloc(bp, udp_tunnel->udp_port,
1654                                              tunnel_type);
1655         return rc;
1656 }
1657
1658 static int
1659 bnxt_udp_tunnel_port_del_op(struct rte_eth_dev *eth_dev,
1660                          struct rte_eth_udp_tunnel *udp_tunnel)
1661 {
1662         struct bnxt *bp = eth_dev->data->dev_private;
1663         uint16_t tunnel_type = 0;
1664         uint16_t port = 0;
1665         int rc = 0;
1666
1667         rc = is_bnxt_in_error(bp);
1668         if (rc)
1669                 return rc;
1670
1671         switch (udp_tunnel->prot_type) {
1672         case RTE_TUNNEL_TYPE_VXLAN:
1673                 if (!bp->vxlan_port_cnt) {
1674                         PMD_DRV_LOG(ERR, "No Tunnel port configured yet\n");
1675                         return -EINVAL;
1676                 }
1677                 if (bp->vxlan_port != udp_tunnel->udp_port) {
1678                         PMD_DRV_LOG(ERR, "Req Port: %d. Configured port: %d\n",
1679                                 udp_tunnel->udp_port, bp->vxlan_port);
1680                         return -EINVAL;
1681                 }
1682                 if (--bp->vxlan_port_cnt)
1683                         return 0;
1684
1685                 tunnel_type =
1686                         HWRM_TUNNEL_DST_PORT_FREE_INPUT_TUNNEL_TYPE_VXLAN;
1687                 port = bp->vxlan_fw_dst_port_id;
1688                 break;
1689         case RTE_TUNNEL_TYPE_GENEVE:
1690                 if (!bp->geneve_port_cnt) {
1691                         PMD_DRV_LOG(ERR, "No Tunnel port configured yet\n");
1692                         return -EINVAL;
1693                 }
1694                 if (bp->geneve_port != udp_tunnel->udp_port) {
1695                         PMD_DRV_LOG(ERR, "Req Port: %d. Configured port: %d\n",
1696                                 udp_tunnel->udp_port, bp->geneve_port);
1697                         return -EINVAL;
1698                 }
1699                 if (--bp->geneve_port_cnt)
1700                         return 0;
1701
1702                 tunnel_type =
1703                         HWRM_TUNNEL_DST_PORT_FREE_INPUT_TUNNEL_TYPE_GENEVE;
1704                 port = bp->geneve_fw_dst_port_id;
1705                 break;
1706         default:
1707                 PMD_DRV_LOG(ERR, "Tunnel type is not supported\n");
1708                 return -ENOTSUP;
1709         }
1710
1711         rc = bnxt_hwrm_tunnel_dst_port_free(bp, port, tunnel_type);
1712         if (!rc) {
1713                 if (tunnel_type ==
1714                     HWRM_TUNNEL_DST_PORT_FREE_INPUT_TUNNEL_TYPE_VXLAN)
1715                         bp->vxlan_port = 0;
1716                 if (tunnel_type ==
1717                     HWRM_TUNNEL_DST_PORT_FREE_INPUT_TUNNEL_TYPE_GENEVE)
1718                         bp->geneve_port = 0;
1719         }
1720         return rc;
1721 }
1722
1723 static int bnxt_del_vlan_filter(struct bnxt *bp, uint16_t vlan_id)
1724 {
1725         struct bnxt_filter_info *filter;
1726         struct bnxt_vnic_info *vnic;
1727         int rc = 0;
1728         uint32_t chk = HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN;
1729
1730         vnic = BNXT_GET_DEFAULT_VNIC(bp);
1731         filter = STAILQ_FIRST(&vnic->filter);
1732         while (filter) {
1733                 /* Search for this matching MAC+VLAN filter */
1734                 if (bnxt_vlan_filter_exists(bp, filter, chk, vlan_id)) {
1735                         /* Delete the filter */
1736                         rc = bnxt_hwrm_clear_l2_filter(bp, filter);
1737                         if (rc)
1738                                 return rc;
1739                         STAILQ_REMOVE(&vnic->filter, filter,
1740                                       bnxt_filter_info, next);
1741                         bnxt_free_filter(bp, filter);
1742                         PMD_DRV_LOG(INFO,
1743                                     "Deleted vlan filter for %d\n",
1744                                     vlan_id);
1745                         return 0;
1746                 }
1747                 filter = STAILQ_NEXT(filter, next);
1748         }
1749         return -ENOENT;
1750 }
1751
1752 static int bnxt_add_vlan_filter(struct bnxt *bp, uint16_t vlan_id)
1753 {
1754         struct bnxt_filter_info *filter;
1755         struct bnxt_vnic_info *vnic;
1756         int rc = 0;
1757         uint32_t en = HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN |
1758                 HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN_MASK;
1759         uint32_t chk = HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN;
1760
1761         /* Implementation notes on the use of VNIC in this command:
1762          *
1763          * By default, these filters belong to default vnic for the function.
1764          * Once these filters are set up, only destination VNIC can be modified.
1765          * If the destination VNIC is not specified in this command,
1766          * then the HWRM shall only create an l2 context id.
1767          */
1768
1769         vnic = BNXT_GET_DEFAULT_VNIC(bp);
1770         filter = STAILQ_FIRST(&vnic->filter);
1771         /* Check if the VLAN has already been added */
1772         while (filter) {
1773                 if (bnxt_vlan_filter_exists(bp, filter, chk, vlan_id))
1774                         return -EEXIST;
1775
1776                 filter = STAILQ_NEXT(filter, next);
1777         }
1778
1779         /* No match found. Alloc a fresh filter and issue the L2_FILTER_ALLOC
1780          * command to create MAC+VLAN filter with the right flags, enables set.
1781          */
1782         filter = bnxt_alloc_filter(bp);
1783         if (!filter) {
1784                 PMD_DRV_LOG(ERR,
1785                             "MAC/VLAN filter alloc failed\n");
1786                 return -ENOMEM;
1787         }
1788         /* MAC + VLAN ID filter */
1789         /* If l2_ivlan == 0 and l2_ivlan_mask != 0, only
1790          * untagged packets are received
1791          *
1792          * If l2_ivlan != 0 and l2_ivlan_mask != 0, untagged
1793          * packets and only the programmed vlan's packets are received
1794          */
1795         filter->l2_ivlan = vlan_id;
1796         filter->l2_ivlan_mask = 0x0FFF;
1797         filter->enables |= en;
1798         filter->flags |= HWRM_CFA_L2_FILTER_ALLOC_INPUT_FLAGS_OUTERMOST;
1799
1800         rc = bnxt_hwrm_set_l2_filter(bp, vnic->fw_vnic_id, filter);
1801         if (rc) {
1802                 /* Free the newly allocated filter as we were
1803                  * not able to create the filter in hardware.
1804                  */
1805                 bnxt_free_filter(bp, filter);
1806                 return rc;
1807         }
1808
1809         filter->mac_index = 0;
1810         /* Add this new filter to the list */
1811         if (vlan_id == 0)
1812                 STAILQ_INSERT_HEAD(&vnic->filter, filter, next);
1813         else
1814                 STAILQ_INSERT_TAIL(&vnic->filter, filter, next);
1815
1816         PMD_DRV_LOG(INFO,
1817                     "Added Vlan filter for %d\n", vlan_id);
1818         return rc;
1819 }
1820
1821 static int bnxt_vlan_filter_set_op(struct rte_eth_dev *eth_dev,
1822                 uint16_t vlan_id, int on)
1823 {
1824         struct bnxt *bp = eth_dev->data->dev_private;
1825         int rc;
1826
1827         rc = is_bnxt_in_error(bp);
1828         if (rc)
1829                 return rc;
1830
1831         /* These operations apply to ALL existing MAC/VLAN filters */
1832         if (on)
1833                 return bnxt_add_vlan_filter(bp, vlan_id);
1834         else
1835                 return bnxt_del_vlan_filter(bp, vlan_id);
1836 }
1837
1838 static int bnxt_del_dflt_mac_filter(struct bnxt *bp,
1839                                     struct bnxt_vnic_info *vnic)
1840 {
1841         struct bnxt_filter_info *filter;
1842         int rc;
1843
1844         filter = STAILQ_FIRST(&vnic->filter);
1845         while (filter) {
1846                 if (filter->mac_index == 0 &&
1847                     !memcmp(filter->l2_addr, bp->mac_addr,
1848                             RTE_ETHER_ADDR_LEN)) {
1849                         rc = bnxt_hwrm_clear_l2_filter(bp, filter);
1850                         if (!rc) {
1851                                 STAILQ_REMOVE(&vnic->filter, filter,
1852                                               bnxt_filter_info, next);
1853                                 bnxt_free_filter(bp, filter);
1854                         }
1855                         return rc;
1856                 }
1857                 filter = STAILQ_NEXT(filter, next);
1858         }
1859         return 0;
1860 }
1861
1862 static int
1863 bnxt_vlan_offload_set_op(struct rte_eth_dev *dev, int mask)
1864 {
1865         struct bnxt *bp = dev->data->dev_private;
1866         uint64_t rx_offloads = dev->data->dev_conf.rxmode.offloads;
1867         struct bnxt_vnic_info *vnic;
1868         unsigned int i;
1869         int rc;
1870
1871         rc = is_bnxt_in_error(bp);
1872         if (rc)
1873                 return rc;
1874
1875         vnic = BNXT_GET_DEFAULT_VNIC(bp);
1876         if (!(rx_offloads & DEV_RX_OFFLOAD_VLAN_FILTER)) {
1877                 /* Remove any VLAN filters programmed */
1878                 for (i = 0; i < RTE_ETHER_MAX_VLAN_ID; i++)
1879                         bnxt_del_vlan_filter(bp, i);
1880
1881                 rc = bnxt_add_mac_filter(bp, vnic, NULL, 0, 0);
1882                 if (rc)
1883                         return rc;
1884         } else {
1885                 /* Default filter will allow packets that match the
1886                  * dest mac. So, it has to be deleted, otherwise, we
1887                  * will endup receiving vlan packets for which the
1888                  * filter is not programmed, when hw-vlan-filter
1889                  * configuration is ON
1890                  */
1891                 bnxt_del_dflt_mac_filter(bp, vnic);
1892                 /* This filter will allow only untagged packets */
1893                 bnxt_add_vlan_filter(bp, 0);
1894         }
1895         PMD_DRV_LOG(DEBUG, "VLAN Filtering: %d\n",
1896                     !!(rx_offloads & DEV_RX_OFFLOAD_VLAN_FILTER));
1897
1898         if (mask & ETH_VLAN_STRIP_MASK) {
1899                 /* Enable or disable VLAN stripping */
1900                 for (i = 0; i < bp->nr_vnics; i++) {
1901                         struct bnxt_vnic_info *vnic = &bp->vnic_info[i];
1902                         if (rx_offloads & DEV_RX_OFFLOAD_VLAN_STRIP)
1903                                 vnic->vlan_strip = true;
1904                         else
1905                                 vnic->vlan_strip = false;
1906                         bnxt_hwrm_vnic_cfg(bp, vnic);
1907                 }
1908                 PMD_DRV_LOG(DEBUG, "VLAN Strip Offload: %d\n",
1909                         !!(rx_offloads & DEV_RX_OFFLOAD_VLAN_STRIP));
1910         }
1911
1912         if (mask & ETH_VLAN_EXTEND_MASK) {
1913                 if (rx_offloads & DEV_RX_OFFLOAD_VLAN_EXTEND)
1914                         PMD_DRV_LOG(DEBUG, "Extend VLAN supported\n");
1915                 else
1916                         PMD_DRV_LOG(INFO, "Extend VLAN unsupported\n");
1917         }
1918
1919         return 0;
1920 }
1921
1922 static int
1923 bnxt_vlan_tpid_set_op(struct rte_eth_dev *dev, enum rte_vlan_type vlan_type,
1924                       uint16_t tpid)
1925 {
1926         struct bnxt *bp = dev->data->dev_private;
1927         int qinq = dev->data->dev_conf.rxmode.offloads &
1928                    DEV_RX_OFFLOAD_VLAN_EXTEND;
1929
1930         if (vlan_type != ETH_VLAN_TYPE_INNER &&
1931             vlan_type != ETH_VLAN_TYPE_OUTER) {
1932                 PMD_DRV_LOG(ERR,
1933                             "Unsupported vlan type.");
1934                 return -EINVAL;
1935         }
1936         if (!qinq) {
1937                 PMD_DRV_LOG(ERR,
1938                             "QinQ not enabled. Needs to be ON as we can "
1939                             "accelerate only outer vlan\n");
1940                 return -EINVAL;
1941         }
1942
1943         if (vlan_type == ETH_VLAN_TYPE_OUTER) {
1944                 switch (tpid) {
1945                 case RTE_ETHER_TYPE_QINQ:
1946                         bp->outer_tpid_bd =
1947                                 TX_BD_LONG_CFA_META_VLAN_TPID_TPID88A8;
1948                                 break;
1949                 case RTE_ETHER_TYPE_VLAN:
1950                         bp->outer_tpid_bd =
1951                                 TX_BD_LONG_CFA_META_VLAN_TPID_TPID8100;
1952                                 break;
1953                 case 0x9100:
1954                         bp->outer_tpid_bd =
1955                                 TX_BD_LONG_CFA_META_VLAN_TPID_TPID9100;
1956                                 break;
1957                 case 0x9200:
1958                         bp->outer_tpid_bd =
1959                                 TX_BD_LONG_CFA_META_VLAN_TPID_TPID9200;
1960                                 break;
1961                 case 0x9300:
1962                         bp->outer_tpid_bd =
1963                                  TX_BD_LONG_CFA_META_VLAN_TPID_TPID9300;
1964                                 break;
1965                 default:
1966                         PMD_DRV_LOG(ERR, "Invalid TPID: %x\n", tpid);
1967                         return -EINVAL;
1968                 }
1969                 bp->outer_tpid_bd |= tpid;
1970                 PMD_DRV_LOG(INFO, "outer_tpid_bd = %x\n", bp->outer_tpid_bd);
1971         } else if (vlan_type == ETH_VLAN_TYPE_INNER) {
1972                 PMD_DRV_LOG(ERR,
1973                             "Can accelerate only outer vlan in QinQ\n");
1974                 return -EINVAL;
1975         }
1976
1977         return 0;
1978 }
1979
1980 static int
1981 bnxt_set_default_mac_addr_op(struct rte_eth_dev *dev,
1982                              struct rte_ether_addr *addr)
1983 {
1984         struct bnxt *bp = dev->data->dev_private;
1985         /* Default Filter is tied to VNIC 0 */
1986         struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
1987         struct bnxt_filter_info *filter;
1988         int rc;
1989
1990         rc = is_bnxt_in_error(bp);
1991         if (rc)
1992                 return rc;
1993
1994         if (BNXT_VF(bp) && !BNXT_VF_IS_TRUSTED(bp))
1995                 return -EPERM;
1996
1997         if (rte_is_zero_ether_addr(addr))
1998                 return -EINVAL;
1999
2000         STAILQ_FOREACH(filter, &vnic->filter, next) {
2001                 /* Default Filter is at Index 0 */
2002                 if (filter->mac_index != 0)
2003                         continue;
2004
2005                 memcpy(filter->l2_addr, addr, RTE_ETHER_ADDR_LEN);
2006                 memset(filter->l2_addr_mask, 0xff, RTE_ETHER_ADDR_LEN);
2007                 filter->flags |= HWRM_CFA_L2_FILTER_ALLOC_INPUT_FLAGS_PATH_RX |
2008                         HWRM_CFA_L2_FILTER_ALLOC_INPUT_FLAGS_OUTERMOST;
2009                 filter->enables |=
2010                         HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_ADDR |
2011                         HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_ADDR_MASK;
2012
2013                 rc = bnxt_hwrm_set_l2_filter(bp, vnic->fw_vnic_id, filter);
2014                 if (rc) {
2015                         memcpy(filter->l2_addr, bp->mac_addr,
2016                                RTE_ETHER_ADDR_LEN);
2017                         return rc;
2018                 }
2019
2020                 memcpy(bp->mac_addr, addr, RTE_ETHER_ADDR_LEN);
2021                 PMD_DRV_LOG(DEBUG, "Set MAC addr\n");
2022                 return 0;
2023         }
2024
2025         return 0;
2026 }
2027
2028 static int
2029 bnxt_dev_set_mc_addr_list_op(struct rte_eth_dev *eth_dev,
2030                           struct rte_ether_addr *mc_addr_set,
2031                           uint32_t nb_mc_addr)
2032 {
2033         struct bnxt *bp = eth_dev->data->dev_private;
2034         char *mc_addr_list = (char *)mc_addr_set;
2035         struct bnxt_vnic_info *vnic;
2036         uint32_t off = 0, i = 0;
2037         int rc;
2038
2039         rc = is_bnxt_in_error(bp);
2040         if (rc)
2041                 return rc;
2042
2043         vnic = BNXT_GET_DEFAULT_VNIC(bp);
2044
2045         if (nb_mc_addr > BNXT_MAX_MC_ADDRS) {
2046                 vnic->flags |= BNXT_VNIC_INFO_ALLMULTI;
2047                 goto allmulti;
2048         }
2049
2050         /* TODO Check for Duplicate mcast addresses */
2051         vnic->flags &= ~BNXT_VNIC_INFO_ALLMULTI;
2052         for (i = 0; i < nb_mc_addr; i++) {
2053                 memcpy(vnic->mc_list + off, &mc_addr_list[i],
2054                         RTE_ETHER_ADDR_LEN);
2055                 off += RTE_ETHER_ADDR_LEN;
2056         }
2057
2058         vnic->mc_addr_cnt = i;
2059         if (vnic->mc_addr_cnt)
2060                 vnic->flags |= BNXT_VNIC_INFO_MCAST;
2061         else
2062                 vnic->flags &= ~BNXT_VNIC_INFO_MCAST;
2063
2064 allmulti:
2065         return bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
2066 }
2067
2068 static int
2069 bnxt_fw_version_get(struct rte_eth_dev *dev, char *fw_version, size_t fw_size)
2070 {
2071         struct bnxt *bp = dev->data->dev_private;
2072         uint8_t fw_major = (bp->fw_ver >> 24) & 0xff;
2073         uint8_t fw_minor = (bp->fw_ver >> 16) & 0xff;
2074         uint8_t fw_updt = (bp->fw_ver >> 8) & 0xff;
2075         int ret;
2076
2077         ret = snprintf(fw_version, fw_size, "%d.%d.%d",
2078                         fw_major, fw_minor, fw_updt);
2079
2080         ret += 1; /* add the size of '\0' */
2081         if (fw_size < (uint32_t)ret)
2082                 return ret;
2083         else
2084                 return 0;
2085 }
2086
2087 static void
2088 bnxt_rxq_info_get_op(struct rte_eth_dev *dev, uint16_t queue_id,
2089         struct rte_eth_rxq_info *qinfo)
2090 {
2091         struct bnxt *bp = dev->data->dev_private;
2092         struct bnxt_rx_queue *rxq;
2093
2094         if (is_bnxt_in_error(bp))
2095                 return;
2096
2097         rxq = dev->data->rx_queues[queue_id];
2098
2099         qinfo->mp = rxq->mb_pool;
2100         qinfo->scattered_rx = dev->data->scattered_rx;
2101         qinfo->nb_desc = rxq->nb_rx_desc;
2102
2103         qinfo->conf.rx_free_thresh = rxq->rx_free_thresh;
2104         qinfo->conf.rx_drop_en = 0;
2105         qinfo->conf.rx_deferred_start = rxq->rx_deferred_start;
2106 }
2107
2108 static void
2109 bnxt_txq_info_get_op(struct rte_eth_dev *dev, uint16_t queue_id,
2110         struct rte_eth_txq_info *qinfo)
2111 {
2112         struct bnxt *bp = dev->data->dev_private;
2113         struct bnxt_tx_queue *txq;
2114
2115         if (is_bnxt_in_error(bp))
2116                 return;
2117
2118         txq = dev->data->tx_queues[queue_id];
2119
2120         qinfo->nb_desc = txq->nb_tx_desc;
2121
2122         qinfo->conf.tx_thresh.pthresh = txq->pthresh;
2123         qinfo->conf.tx_thresh.hthresh = txq->hthresh;
2124         qinfo->conf.tx_thresh.wthresh = txq->wthresh;
2125
2126         qinfo->conf.tx_free_thresh = txq->tx_free_thresh;
2127         qinfo->conf.tx_rs_thresh = 0;
2128         qinfo->conf.tx_deferred_start = txq->tx_deferred_start;
2129 }
2130
2131 int bnxt_mtu_set_op(struct rte_eth_dev *eth_dev, uint16_t new_mtu)
2132 {
2133         struct bnxt *bp = eth_dev->data->dev_private;
2134         uint32_t new_pkt_size;
2135         uint32_t rc = 0;
2136         uint32_t i;
2137
2138         rc = is_bnxt_in_error(bp);
2139         if (rc)
2140                 return rc;
2141
2142         /* Exit if receive queues are not configured yet */
2143         if (!eth_dev->data->nb_rx_queues)
2144                 return rc;
2145
2146         new_pkt_size = new_mtu + RTE_ETHER_HDR_LEN + RTE_ETHER_CRC_LEN +
2147                        VLAN_TAG_SIZE * BNXT_NUM_VLANS;
2148
2149 #ifdef RTE_ARCH_X86
2150         /*
2151          * If vector-mode tx/rx is active, disallow any MTU change that would
2152          * require scattered receive support.
2153          */
2154         if (eth_dev->data->dev_started &&
2155             (eth_dev->rx_pkt_burst == bnxt_recv_pkts_vec ||
2156              eth_dev->tx_pkt_burst == bnxt_xmit_pkts_vec) &&
2157             (new_pkt_size >
2158              eth_dev->data->min_rx_buf_size - RTE_PKTMBUF_HEADROOM)) {
2159                 PMD_DRV_LOG(ERR,
2160                             "MTU change would require scattered rx support. ");
2161                 PMD_DRV_LOG(ERR, "Stop port before changing MTU.\n");
2162                 return -EINVAL;
2163         }
2164 #endif
2165
2166         if (new_mtu > RTE_ETHER_MTU) {
2167                 bp->flags |= BNXT_FLAG_JUMBO;
2168                 bp->eth_dev->data->dev_conf.rxmode.offloads |=
2169                         DEV_RX_OFFLOAD_JUMBO_FRAME;
2170         } else {
2171                 bp->eth_dev->data->dev_conf.rxmode.offloads &=
2172                         ~DEV_RX_OFFLOAD_JUMBO_FRAME;
2173                 bp->flags &= ~BNXT_FLAG_JUMBO;
2174         }
2175
2176         /* Is there a change in mtu setting? */
2177         if (eth_dev->data->dev_conf.rxmode.max_rx_pkt_len == new_pkt_size)
2178                 return rc;
2179
2180         for (i = 0; i < bp->nr_vnics; i++) {
2181                 struct bnxt_vnic_info *vnic = &bp->vnic_info[i];
2182                 uint16_t size = 0;
2183
2184                 vnic->mru = BNXT_VNIC_MRU(new_mtu);
2185                 rc = bnxt_hwrm_vnic_cfg(bp, vnic);
2186                 if (rc)
2187                         break;
2188
2189                 size = rte_pktmbuf_data_room_size(bp->rx_queues[0]->mb_pool);
2190                 size -= RTE_PKTMBUF_HEADROOM;
2191
2192                 if (size < new_mtu) {
2193                         rc = bnxt_hwrm_vnic_plcmode_cfg(bp, vnic);
2194                         if (rc)
2195                                 return rc;
2196                 }
2197         }
2198
2199         if (!rc)
2200                 eth_dev->data->dev_conf.rxmode.max_rx_pkt_len = new_pkt_size;
2201
2202         PMD_DRV_LOG(INFO, "New MTU is %d\n", new_mtu);
2203
2204         return rc;
2205 }
2206
2207 static int
2208 bnxt_vlan_pvid_set_op(struct rte_eth_dev *dev, uint16_t pvid, int on)
2209 {
2210         struct bnxt *bp = dev->data->dev_private;
2211         uint16_t vlan = bp->vlan;
2212         int rc;
2213
2214         rc = is_bnxt_in_error(bp);
2215         if (rc)
2216                 return rc;
2217
2218         if (!BNXT_SINGLE_PF(bp) || BNXT_VF(bp)) {
2219                 PMD_DRV_LOG(ERR,
2220                         "PVID cannot be modified for this function\n");
2221                 return -ENOTSUP;
2222         }
2223         bp->vlan = on ? pvid : 0;
2224
2225         rc = bnxt_hwrm_set_default_vlan(bp, 0, 0);
2226         if (rc)
2227                 bp->vlan = vlan;
2228         return rc;
2229 }
2230
2231 static int
2232 bnxt_dev_led_on_op(struct rte_eth_dev *dev)
2233 {
2234         struct bnxt *bp = dev->data->dev_private;
2235         int rc;
2236
2237         rc = is_bnxt_in_error(bp);
2238         if (rc)
2239                 return rc;
2240
2241         return bnxt_hwrm_port_led_cfg(bp, true);
2242 }
2243
2244 static int
2245 bnxt_dev_led_off_op(struct rte_eth_dev *dev)
2246 {
2247         struct bnxt *bp = dev->data->dev_private;
2248         int rc;
2249
2250         rc = is_bnxt_in_error(bp);
2251         if (rc)
2252                 return rc;
2253
2254         return bnxt_hwrm_port_led_cfg(bp, false);
2255 }
2256
2257 static uint32_t
2258 bnxt_rx_queue_count_op(struct rte_eth_dev *dev, uint16_t rx_queue_id)
2259 {
2260         struct bnxt *bp = (struct bnxt *)dev->data->dev_private;
2261         uint32_t desc = 0, raw_cons = 0, cons;
2262         struct bnxt_cp_ring_info *cpr;
2263         struct bnxt_rx_queue *rxq;
2264         struct rx_pkt_cmpl *rxcmp;
2265         int rc;
2266
2267         rc = is_bnxt_in_error(bp);
2268         if (rc)
2269                 return rc;
2270
2271         rxq = dev->data->rx_queues[rx_queue_id];
2272         cpr = rxq->cp_ring;
2273         raw_cons = cpr->cp_raw_cons;
2274
2275         while (1) {
2276                 cons = RING_CMP(cpr->cp_ring_struct, raw_cons);
2277                 rte_prefetch0(&cpr->cp_desc_ring[cons]);
2278                 rxcmp = (struct rx_pkt_cmpl *)&cpr->cp_desc_ring[cons];
2279
2280                 if (!CMP_VALID(rxcmp, raw_cons, cpr->cp_ring_struct)) {
2281                         break;
2282                 } else {
2283                         raw_cons++;
2284                         desc++;
2285                 }
2286         }
2287
2288         return desc;
2289 }
2290
2291 static int
2292 bnxt_rx_descriptor_status_op(void *rx_queue, uint16_t offset)
2293 {
2294         struct bnxt_rx_queue *rxq = (struct bnxt_rx_queue *)rx_queue;
2295         struct bnxt_rx_ring_info *rxr;
2296         struct bnxt_cp_ring_info *cpr;
2297         struct bnxt_sw_rx_bd *rx_buf;
2298         struct rx_pkt_cmpl *rxcmp;
2299         uint32_t cons, cp_cons;
2300         int rc;
2301
2302         if (!rxq)
2303                 return -EINVAL;
2304
2305         rc = is_bnxt_in_error(rxq->bp);
2306         if (rc)
2307                 return rc;
2308
2309         cpr = rxq->cp_ring;
2310         rxr = rxq->rx_ring;
2311
2312         if (offset >= rxq->nb_rx_desc)
2313                 return -EINVAL;
2314
2315         cons = RING_CMP(cpr->cp_ring_struct, offset);
2316         cp_cons = cpr->cp_raw_cons;
2317         rxcmp = (struct rx_pkt_cmpl *)&cpr->cp_desc_ring[cons];
2318
2319         if (cons > cp_cons) {
2320                 if (CMPL_VALID(rxcmp, cpr->valid))
2321                         return RTE_ETH_RX_DESC_DONE;
2322         } else {
2323                 if (CMPL_VALID(rxcmp, !cpr->valid))
2324                         return RTE_ETH_RX_DESC_DONE;
2325         }
2326         rx_buf = &rxr->rx_buf_ring[cons];
2327         if (rx_buf->mbuf == NULL)
2328                 return RTE_ETH_RX_DESC_UNAVAIL;
2329
2330
2331         return RTE_ETH_RX_DESC_AVAIL;
2332 }
2333
2334 static int
2335 bnxt_tx_descriptor_status_op(void *tx_queue, uint16_t offset)
2336 {
2337         struct bnxt_tx_queue *txq = (struct bnxt_tx_queue *)tx_queue;
2338         struct bnxt_tx_ring_info *txr;
2339         struct bnxt_cp_ring_info *cpr;
2340         struct bnxt_sw_tx_bd *tx_buf;
2341         struct tx_pkt_cmpl *txcmp;
2342         uint32_t cons, cp_cons;
2343         int rc;
2344
2345         if (!txq)
2346                 return -EINVAL;
2347
2348         rc = is_bnxt_in_error(txq->bp);
2349         if (rc)
2350                 return rc;
2351
2352         cpr = txq->cp_ring;
2353         txr = txq->tx_ring;
2354
2355         if (offset >= txq->nb_tx_desc)
2356                 return -EINVAL;
2357
2358         cons = RING_CMP(cpr->cp_ring_struct, offset);
2359         txcmp = (struct tx_pkt_cmpl *)&cpr->cp_desc_ring[cons];
2360         cp_cons = cpr->cp_raw_cons;
2361
2362         if (cons > cp_cons) {
2363                 if (CMPL_VALID(txcmp, cpr->valid))
2364                         return RTE_ETH_TX_DESC_UNAVAIL;
2365         } else {
2366                 if (CMPL_VALID(txcmp, !cpr->valid))
2367                         return RTE_ETH_TX_DESC_UNAVAIL;
2368         }
2369         tx_buf = &txr->tx_buf_ring[cons];
2370         if (tx_buf->mbuf == NULL)
2371                 return RTE_ETH_TX_DESC_DONE;
2372
2373         return RTE_ETH_TX_DESC_FULL;
2374 }
2375
2376 static struct bnxt_filter_info *
2377 bnxt_match_and_validate_ether_filter(struct bnxt *bp,
2378                                 struct rte_eth_ethertype_filter *efilter,
2379                                 struct bnxt_vnic_info *vnic0,
2380                                 struct bnxt_vnic_info *vnic,
2381                                 int *ret)
2382 {
2383         struct bnxt_filter_info *mfilter = NULL;
2384         int match = 0;
2385         *ret = 0;
2386
2387         if (efilter->ether_type == RTE_ETHER_TYPE_IPV4 ||
2388                 efilter->ether_type == RTE_ETHER_TYPE_IPV6) {
2389                 PMD_DRV_LOG(ERR, "invalid ether_type(0x%04x) in"
2390                         " ethertype filter.", efilter->ether_type);
2391                 *ret = -EINVAL;
2392                 goto exit;
2393         }
2394         if (efilter->queue >= bp->rx_nr_rings) {
2395                 PMD_DRV_LOG(ERR, "Invalid queue %d\n", efilter->queue);
2396                 *ret = -EINVAL;
2397                 goto exit;
2398         }
2399
2400         vnic0 = BNXT_GET_DEFAULT_VNIC(bp);
2401         vnic = &bp->vnic_info[efilter->queue];
2402         if (vnic == NULL) {
2403                 PMD_DRV_LOG(ERR, "Invalid queue %d\n", efilter->queue);
2404                 *ret = -EINVAL;
2405                 goto exit;
2406         }
2407
2408         if (efilter->flags & RTE_ETHTYPE_FLAGS_DROP) {
2409                 STAILQ_FOREACH(mfilter, &vnic0->filter, next) {
2410                         if ((!memcmp(efilter->mac_addr.addr_bytes,
2411                                      mfilter->l2_addr, RTE_ETHER_ADDR_LEN) &&
2412                              mfilter->flags ==
2413                              HWRM_CFA_NTUPLE_FILTER_ALLOC_INPUT_FLAGS_DROP &&
2414                              mfilter->ethertype == efilter->ether_type)) {
2415                                 match = 1;
2416                                 break;
2417                         }
2418                 }
2419         } else {
2420                 STAILQ_FOREACH(mfilter, &vnic->filter, next)
2421                         if ((!memcmp(efilter->mac_addr.addr_bytes,
2422                                      mfilter->l2_addr, RTE_ETHER_ADDR_LEN) &&
2423                              mfilter->ethertype == efilter->ether_type &&
2424                              mfilter->flags ==
2425                              HWRM_CFA_L2_FILTER_CFG_INPUT_FLAGS_PATH_RX)) {
2426                                 match = 1;
2427                                 break;
2428                         }
2429         }
2430
2431         if (match)
2432                 *ret = -EEXIST;
2433
2434 exit:
2435         return mfilter;
2436 }
2437
2438 static int
2439 bnxt_ethertype_filter(struct rte_eth_dev *dev,
2440                         enum rte_filter_op filter_op,
2441                         void *arg)
2442 {
2443         struct bnxt *bp = dev->data->dev_private;
2444         struct rte_eth_ethertype_filter *efilter =
2445                         (struct rte_eth_ethertype_filter *)arg;
2446         struct bnxt_filter_info *bfilter, *filter1;
2447         struct bnxt_vnic_info *vnic, *vnic0;
2448         int ret;
2449
2450         if (filter_op == RTE_ETH_FILTER_NOP)
2451                 return 0;
2452
2453         if (arg == NULL) {
2454                 PMD_DRV_LOG(ERR, "arg shouldn't be NULL for operation %u.",
2455                             filter_op);
2456                 return -EINVAL;
2457         }
2458
2459         vnic0 = BNXT_GET_DEFAULT_VNIC(bp);
2460         vnic = &bp->vnic_info[efilter->queue];
2461
2462         switch (filter_op) {
2463         case RTE_ETH_FILTER_ADD:
2464                 bnxt_match_and_validate_ether_filter(bp, efilter,
2465                                                         vnic0, vnic, &ret);
2466                 if (ret < 0)
2467                         return ret;
2468
2469                 bfilter = bnxt_get_unused_filter(bp);
2470                 if (bfilter == NULL) {
2471                         PMD_DRV_LOG(ERR,
2472                                 "Not enough resources for a new filter.\n");
2473                         return -ENOMEM;
2474                 }
2475                 bfilter->filter_type = HWRM_CFA_NTUPLE_FILTER;
2476                 memcpy(bfilter->l2_addr, efilter->mac_addr.addr_bytes,
2477                        RTE_ETHER_ADDR_LEN);
2478                 memcpy(bfilter->dst_macaddr, efilter->mac_addr.addr_bytes,
2479                        RTE_ETHER_ADDR_LEN);
2480                 bfilter->enables |= NTUPLE_FLTR_ALLOC_INPUT_EN_DST_MACADDR;
2481                 bfilter->ethertype = efilter->ether_type;
2482                 bfilter->enables |= NTUPLE_FLTR_ALLOC_INPUT_EN_ETHERTYPE;
2483
2484                 filter1 = bnxt_get_l2_filter(bp, bfilter, vnic0);
2485                 if (filter1 == NULL) {
2486                         ret = -EINVAL;
2487                         goto cleanup;
2488                 }
2489                 bfilter->enables |=
2490                         HWRM_CFA_NTUPLE_FILTER_ALLOC_INPUT_ENABLES_L2_FILTER_ID;
2491                 bfilter->fw_l2_filter_id = filter1->fw_l2_filter_id;
2492
2493                 bfilter->dst_id = vnic->fw_vnic_id;
2494
2495                 if (efilter->flags & RTE_ETHTYPE_FLAGS_DROP) {
2496                         bfilter->flags =
2497                                 HWRM_CFA_NTUPLE_FILTER_ALLOC_INPUT_FLAGS_DROP;
2498                 }
2499
2500                 ret = bnxt_hwrm_set_ntuple_filter(bp, bfilter->dst_id, bfilter);
2501                 if (ret)
2502                         goto cleanup;
2503                 STAILQ_INSERT_TAIL(&vnic->filter, bfilter, next);
2504                 break;
2505         case RTE_ETH_FILTER_DELETE:
2506                 filter1 = bnxt_match_and_validate_ether_filter(bp, efilter,
2507                                                         vnic0, vnic, &ret);
2508                 if (ret == -EEXIST) {
2509                         ret = bnxt_hwrm_clear_ntuple_filter(bp, filter1);
2510
2511                         STAILQ_REMOVE(&vnic->filter, filter1, bnxt_filter_info,
2512                                       next);
2513                         bnxt_free_filter(bp, filter1);
2514                 } else if (ret == 0) {
2515                         PMD_DRV_LOG(ERR, "No matching filter found\n");
2516                 }
2517                 break;
2518         default:
2519                 PMD_DRV_LOG(ERR, "unsupported operation %u.", filter_op);
2520                 ret = -EINVAL;
2521                 goto error;
2522         }
2523         return ret;
2524 cleanup:
2525         bnxt_free_filter(bp, bfilter);
2526 error:
2527         return ret;
2528 }
2529
2530 static inline int
2531 parse_ntuple_filter(struct bnxt *bp,
2532                     struct rte_eth_ntuple_filter *nfilter,
2533                     struct bnxt_filter_info *bfilter)
2534 {
2535         uint32_t en = 0;
2536
2537         if (nfilter->queue >= bp->rx_nr_rings) {
2538                 PMD_DRV_LOG(ERR, "Invalid queue %d\n", nfilter->queue);
2539                 return -EINVAL;
2540         }
2541
2542         switch (nfilter->dst_port_mask) {
2543         case UINT16_MAX:
2544                 bfilter->dst_port_mask = -1;
2545                 bfilter->dst_port = nfilter->dst_port;
2546                 en |= NTUPLE_FLTR_ALLOC_INPUT_EN_DST_PORT |
2547                         NTUPLE_FLTR_ALLOC_INPUT_EN_DST_PORT_MASK;
2548                 break;
2549         default:
2550                 PMD_DRV_LOG(ERR, "invalid dst_port mask.");
2551                 return -EINVAL;
2552         }
2553
2554         bfilter->ip_addr_type = NTUPLE_FLTR_ALLOC_INPUT_IP_ADDR_TYPE_IPV4;
2555         en |= NTUPLE_FLTR_ALLOC_IN_EN_IP_PROTO;
2556
2557         switch (nfilter->proto_mask) {
2558         case UINT8_MAX:
2559                 if (nfilter->proto == 17) /* IPPROTO_UDP */
2560                         bfilter->ip_protocol = 17;
2561                 else if (nfilter->proto == 6) /* IPPROTO_TCP */
2562                         bfilter->ip_protocol = 6;
2563                 else
2564                         return -EINVAL;
2565                 en |= NTUPLE_FLTR_ALLOC_IN_EN_IP_PROTO;
2566                 break;
2567         default:
2568                 PMD_DRV_LOG(ERR, "invalid protocol mask.");
2569                 return -EINVAL;
2570         }
2571
2572         switch (nfilter->dst_ip_mask) {
2573         case UINT32_MAX:
2574                 bfilter->dst_ipaddr_mask[0] = -1;
2575                 bfilter->dst_ipaddr[0] = nfilter->dst_ip;
2576                 en |= NTUPLE_FLTR_ALLOC_INPUT_EN_DST_IPADDR |
2577                         NTUPLE_FLTR_ALLOC_INPUT_EN_DST_IPADDR_MASK;
2578                 break;
2579         default:
2580                 PMD_DRV_LOG(ERR, "invalid dst_ip mask.");
2581                 return -EINVAL;
2582         }
2583
2584         switch (nfilter->src_ip_mask) {
2585         case UINT32_MAX:
2586                 bfilter->src_ipaddr_mask[0] = -1;
2587                 bfilter->src_ipaddr[0] = nfilter->src_ip;
2588                 en |= NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_IPADDR |
2589                         NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_IPADDR_MASK;
2590                 break;
2591         default:
2592                 PMD_DRV_LOG(ERR, "invalid src_ip mask.");
2593                 return -EINVAL;
2594         }
2595
2596         switch (nfilter->src_port_mask) {
2597         case UINT16_MAX:
2598                 bfilter->src_port_mask = -1;
2599                 bfilter->src_port = nfilter->src_port;
2600                 en |= NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_PORT |
2601                         NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_PORT_MASK;
2602                 break;
2603         default:
2604                 PMD_DRV_LOG(ERR, "invalid src_port mask.");
2605                 return -EINVAL;
2606         }
2607
2608         bfilter->enables = en;
2609         return 0;
2610 }
2611
2612 static struct bnxt_filter_info*
2613 bnxt_match_ntuple_filter(struct bnxt *bp,
2614                          struct bnxt_filter_info *bfilter,
2615                          struct bnxt_vnic_info **mvnic)
2616 {
2617         struct bnxt_filter_info *mfilter = NULL;
2618         int i;
2619
2620         for (i = bp->nr_vnics - 1; i >= 0; i--) {
2621                 struct bnxt_vnic_info *vnic = &bp->vnic_info[i];
2622                 STAILQ_FOREACH(mfilter, &vnic->filter, next) {
2623                         if (bfilter->src_ipaddr[0] == mfilter->src_ipaddr[0] &&
2624                             bfilter->src_ipaddr_mask[0] ==
2625                             mfilter->src_ipaddr_mask[0] &&
2626                             bfilter->src_port == mfilter->src_port &&
2627                             bfilter->src_port_mask == mfilter->src_port_mask &&
2628                             bfilter->dst_ipaddr[0] == mfilter->dst_ipaddr[0] &&
2629                             bfilter->dst_ipaddr_mask[0] ==
2630                             mfilter->dst_ipaddr_mask[0] &&
2631                             bfilter->dst_port == mfilter->dst_port &&
2632                             bfilter->dst_port_mask == mfilter->dst_port_mask &&
2633                             bfilter->flags == mfilter->flags &&
2634                             bfilter->enables == mfilter->enables) {
2635                                 if (mvnic)
2636                                         *mvnic = vnic;
2637                                 return mfilter;
2638                         }
2639                 }
2640         }
2641         return NULL;
2642 }
2643
2644 static int
2645 bnxt_cfg_ntuple_filter(struct bnxt *bp,
2646                        struct rte_eth_ntuple_filter *nfilter,
2647                        enum rte_filter_op filter_op)
2648 {
2649         struct bnxt_filter_info *bfilter, *mfilter, *filter1;
2650         struct bnxt_vnic_info *vnic, *vnic0, *mvnic;
2651         int ret;
2652
2653         if (nfilter->flags != RTE_5TUPLE_FLAGS) {
2654                 PMD_DRV_LOG(ERR, "only 5tuple is supported.");
2655                 return -EINVAL;
2656         }
2657
2658         if (nfilter->flags & RTE_NTUPLE_FLAGS_TCP_FLAG) {
2659                 PMD_DRV_LOG(ERR, "Ntuple filter: TCP flags not supported\n");
2660                 return -EINVAL;
2661         }
2662
2663         bfilter = bnxt_get_unused_filter(bp);
2664         if (bfilter == NULL) {
2665                 PMD_DRV_LOG(ERR,
2666                         "Not enough resources for a new filter.\n");
2667                 return -ENOMEM;
2668         }
2669         ret = parse_ntuple_filter(bp, nfilter, bfilter);
2670         if (ret < 0)
2671                 goto free_filter;
2672
2673         vnic = &bp->vnic_info[nfilter->queue];
2674         vnic0 = BNXT_GET_DEFAULT_VNIC(bp);
2675         filter1 = STAILQ_FIRST(&vnic0->filter);
2676         if (filter1 == NULL) {
2677                 ret = -EINVAL;
2678                 goto free_filter;
2679         }
2680
2681         bfilter->dst_id = vnic->fw_vnic_id;
2682         bfilter->fw_l2_filter_id = filter1->fw_l2_filter_id;
2683         bfilter->enables |=
2684                 HWRM_CFA_NTUPLE_FILTER_ALLOC_INPUT_ENABLES_L2_FILTER_ID;
2685         bfilter->ethertype = 0x800;
2686         bfilter->enables |= NTUPLE_FLTR_ALLOC_INPUT_EN_ETHERTYPE;
2687
2688         mfilter = bnxt_match_ntuple_filter(bp, bfilter, &mvnic);
2689
2690         if (mfilter != NULL && filter_op == RTE_ETH_FILTER_ADD &&
2691             bfilter->dst_id == mfilter->dst_id) {
2692                 PMD_DRV_LOG(ERR, "filter exists.\n");
2693                 ret = -EEXIST;
2694                 goto free_filter;
2695         } else if (mfilter != NULL && filter_op == RTE_ETH_FILTER_ADD &&
2696                    bfilter->dst_id != mfilter->dst_id) {
2697                 mfilter->dst_id = vnic->fw_vnic_id;
2698                 ret = bnxt_hwrm_set_ntuple_filter(bp, mfilter->dst_id, mfilter);
2699                 STAILQ_REMOVE(&mvnic->filter, mfilter, bnxt_filter_info, next);
2700                 STAILQ_INSERT_TAIL(&vnic->filter, mfilter, next);
2701                 PMD_DRV_LOG(ERR, "filter with matching pattern exists.\n");
2702                 PMD_DRV_LOG(ERR, " Updated it to the new destination queue\n");
2703                 goto free_filter;
2704         }
2705         if (mfilter == NULL && filter_op == RTE_ETH_FILTER_DELETE) {
2706                 PMD_DRV_LOG(ERR, "filter doesn't exist.");
2707                 ret = -ENOENT;
2708                 goto free_filter;
2709         }
2710
2711         if (filter_op == RTE_ETH_FILTER_ADD) {
2712                 bfilter->filter_type = HWRM_CFA_NTUPLE_FILTER;
2713                 ret = bnxt_hwrm_set_ntuple_filter(bp, bfilter->dst_id, bfilter);
2714                 if (ret)
2715                         goto free_filter;
2716                 STAILQ_INSERT_TAIL(&vnic->filter, bfilter, next);
2717         } else {
2718                 if (mfilter == NULL) {
2719                         /* This should not happen. But for Coverity! */
2720                         ret = -ENOENT;
2721                         goto free_filter;
2722                 }
2723                 ret = bnxt_hwrm_clear_ntuple_filter(bp, mfilter);
2724
2725                 STAILQ_REMOVE(&vnic->filter, mfilter, bnxt_filter_info, next);
2726                 bnxt_free_filter(bp, mfilter);
2727                 bnxt_free_filter(bp, bfilter);
2728         }
2729
2730         return 0;
2731 free_filter:
2732         bnxt_free_filter(bp, bfilter);
2733         return ret;
2734 }
2735
2736 static int
2737 bnxt_ntuple_filter(struct rte_eth_dev *dev,
2738                         enum rte_filter_op filter_op,
2739                         void *arg)
2740 {
2741         struct bnxt *bp = dev->data->dev_private;
2742         int ret;
2743
2744         if (filter_op == RTE_ETH_FILTER_NOP)
2745                 return 0;
2746
2747         if (arg == NULL) {
2748                 PMD_DRV_LOG(ERR, "arg shouldn't be NULL for operation %u.",
2749                             filter_op);
2750                 return -EINVAL;
2751         }
2752
2753         switch (filter_op) {
2754         case RTE_ETH_FILTER_ADD:
2755                 ret = bnxt_cfg_ntuple_filter(bp,
2756                         (struct rte_eth_ntuple_filter *)arg,
2757                         filter_op);
2758                 break;
2759         case RTE_ETH_FILTER_DELETE:
2760                 ret = bnxt_cfg_ntuple_filter(bp,
2761                         (struct rte_eth_ntuple_filter *)arg,
2762                         filter_op);
2763                 break;
2764         default:
2765                 PMD_DRV_LOG(ERR, "unsupported operation %u.", filter_op);
2766                 ret = -EINVAL;
2767                 break;
2768         }
2769         return ret;
2770 }
2771
2772 static int
2773 bnxt_parse_fdir_filter(struct bnxt *bp,
2774                        struct rte_eth_fdir_filter *fdir,
2775                        struct bnxt_filter_info *filter)
2776 {
2777         enum rte_fdir_mode fdir_mode =
2778                 bp->eth_dev->data->dev_conf.fdir_conf.mode;
2779         struct bnxt_vnic_info *vnic0, *vnic;
2780         struct bnxt_filter_info *filter1;
2781         uint32_t en = 0;
2782         int i;
2783
2784         if (fdir_mode == RTE_FDIR_MODE_PERFECT_TUNNEL)
2785                 return -EINVAL;
2786
2787         filter->l2_ovlan = fdir->input.flow_ext.vlan_tci;
2788         en |= EM_FLOW_ALLOC_INPUT_EN_OVLAN_VID;
2789
2790         switch (fdir->input.flow_type) {
2791         case RTE_ETH_FLOW_IPV4:
2792         case RTE_ETH_FLOW_NONFRAG_IPV4_OTHER:
2793                 /* FALLTHROUGH */
2794                 filter->src_ipaddr[0] = fdir->input.flow.ip4_flow.src_ip;
2795                 en |= NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_IPADDR;
2796                 filter->dst_ipaddr[0] = fdir->input.flow.ip4_flow.dst_ip;
2797                 en |= NTUPLE_FLTR_ALLOC_INPUT_EN_DST_IPADDR;
2798                 filter->ip_protocol = fdir->input.flow.ip4_flow.proto;
2799                 en |= NTUPLE_FLTR_ALLOC_IN_EN_IP_PROTO;
2800                 filter->ip_addr_type =
2801                         NTUPLE_FLTR_ALLOC_INPUT_IP_ADDR_TYPE_IPV4;
2802                 filter->src_ipaddr_mask[0] = 0xffffffff;
2803                 en |= NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_IPADDR_MASK;
2804                 filter->dst_ipaddr_mask[0] = 0xffffffff;
2805                 en |= NTUPLE_FLTR_ALLOC_INPUT_EN_DST_IPADDR_MASK;
2806                 filter->ethertype = 0x800;
2807                 filter->enables |= NTUPLE_FLTR_ALLOC_INPUT_EN_ETHERTYPE;
2808                 break;
2809         case RTE_ETH_FLOW_NONFRAG_IPV4_TCP:
2810                 filter->src_port = fdir->input.flow.tcp4_flow.src_port;
2811                 en |= NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_PORT;
2812                 filter->dst_port = fdir->input.flow.tcp4_flow.dst_port;
2813                 en |= NTUPLE_FLTR_ALLOC_INPUT_EN_DST_PORT;
2814                 filter->dst_port_mask = 0xffff;
2815                 en |= NTUPLE_FLTR_ALLOC_INPUT_EN_DST_PORT_MASK;
2816                 filter->src_port_mask = 0xffff;
2817                 en |= NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_PORT_MASK;
2818                 filter->src_ipaddr[0] = fdir->input.flow.tcp4_flow.ip.src_ip;
2819                 en |= NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_IPADDR;
2820                 filter->dst_ipaddr[0] = fdir->input.flow.tcp4_flow.ip.dst_ip;
2821                 en |= NTUPLE_FLTR_ALLOC_INPUT_EN_DST_IPADDR;
2822                 filter->ip_protocol = 6;
2823                 en |= NTUPLE_FLTR_ALLOC_IN_EN_IP_PROTO;
2824                 filter->ip_addr_type =
2825                         NTUPLE_FLTR_ALLOC_INPUT_IP_ADDR_TYPE_IPV4;
2826                 filter->src_ipaddr_mask[0] = 0xffffffff;
2827                 en |= NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_IPADDR_MASK;
2828                 filter->dst_ipaddr_mask[0] = 0xffffffff;
2829                 en |= NTUPLE_FLTR_ALLOC_INPUT_EN_DST_IPADDR_MASK;
2830                 filter->ethertype = 0x800;
2831                 filter->enables |= NTUPLE_FLTR_ALLOC_INPUT_EN_ETHERTYPE;
2832                 break;
2833         case RTE_ETH_FLOW_NONFRAG_IPV4_UDP:
2834                 filter->src_port = fdir->input.flow.udp4_flow.src_port;
2835                 en |= NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_PORT;
2836                 filter->dst_port = fdir->input.flow.udp4_flow.dst_port;
2837                 en |= NTUPLE_FLTR_ALLOC_INPUT_EN_DST_PORT;
2838                 filter->dst_port_mask = 0xffff;
2839                 en |= NTUPLE_FLTR_ALLOC_INPUT_EN_DST_PORT_MASK;
2840                 filter->src_port_mask = 0xffff;
2841                 en |= NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_PORT_MASK;
2842                 filter->src_ipaddr[0] = fdir->input.flow.udp4_flow.ip.src_ip;
2843                 en |= NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_IPADDR;
2844                 filter->dst_ipaddr[0] = fdir->input.flow.udp4_flow.ip.dst_ip;
2845                 en |= NTUPLE_FLTR_ALLOC_INPUT_EN_DST_IPADDR;
2846                 filter->ip_protocol = 17;
2847                 en |= NTUPLE_FLTR_ALLOC_IN_EN_IP_PROTO;
2848                 filter->ip_addr_type =
2849                         NTUPLE_FLTR_ALLOC_INPUT_IP_ADDR_TYPE_IPV4;
2850                 filter->src_ipaddr_mask[0] = 0xffffffff;
2851                 en |= NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_IPADDR_MASK;
2852                 filter->dst_ipaddr_mask[0] = 0xffffffff;
2853                 en |= NTUPLE_FLTR_ALLOC_INPUT_EN_DST_IPADDR_MASK;
2854                 filter->ethertype = 0x800;
2855                 filter->enables |= NTUPLE_FLTR_ALLOC_INPUT_EN_ETHERTYPE;
2856                 break;
2857         case RTE_ETH_FLOW_IPV6:
2858         case RTE_ETH_FLOW_NONFRAG_IPV6_OTHER:
2859                 /* FALLTHROUGH */
2860                 filter->ip_addr_type =
2861                         NTUPLE_FLTR_ALLOC_INPUT_IP_ADDR_TYPE_IPV6;
2862                 filter->ip_protocol = fdir->input.flow.ipv6_flow.proto;
2863                 en |= NTUPLE_FLTR_ALLOC_IN_EN_IP_PROTO;
2864                 rte_memcpy(filter->src_ipaddr,
2865                            fdir->input.flow.ipv6_flow.src_ip, 16);
2866                 en |= NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_IPADDR;
2867                 rte_memcpy(filter->dst_ipaddr,
2868                            fdir->input.flow.ipv6_flow.dst_ip, 16);
2869                 en |= NTUPLE_FLTR_ALLOC_INPUT_EN_DST_IPADDR;
2870                 memset(filter->dst_ipaddr_mask, 0xff, 16);
2871                 en |= NTUPLE_FLTR_ALLOC_INPUT_EN_DST_IPADDR_MASK;
2872                 memset(filter->src_ipaddr_mask, 0xff, 16);
2873                 en |= NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_IPADDR_MASK;
2874                 filter->ethertype = 0x86dd;
2875                 filter->enables |= NTUPLE_FLTR_ALLOC_INPUT_EN_ETHERTYPE;
2876                 break;
2877         case RTE_ETH_FLOW_NONFRAG_IPV6_TCP:
2878                 filter->src_port = fdir->input.flow.tcp6_flow.src_port;
2879                 en |= NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_PORT;
2880                 filter->dst_port = fdir->input.flow.tcp6_flow.dst_port;
2881                 en |= NTUPLE_FLTR_ALLOC_INPUT_EN_DST_PORT;
2882                 filter->dst_port_mask = 0xffff;
2883                 en |= NTUPLE_FLTR_ALLOC_INPUT_EN_DST_PORT_MASK;
2884                 filter->src_port_mask = 0xffff;
2885                 en |= NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_PORT_MASK;
2886                 filter->ip_addr_type =
2887                         NTUPLE_FLTR_ALLOC_INPUT_IP_ADDR_TYPE_IPV6;
2888                 filter->ip_protocol = fdir->input.flow.tcp6_flow.ip.proto;
2889                 en |= NTUPLE_FLTR_ALLOC_IN_EN_IP_PROTO;
2890                 rte_memcpy(filter->src_ipaddr,
2891                            fdir->input.flow.tcp6_flow.ip.src_ip, 16);
2892                 en |= NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_IPADDR;
2893                 rte_memcpy(filter->dst_ipaddr,
2894                            fdir->input.flow.tcp6_flow.ip.dst_ip, 16);
2895                 en |= NTUPLE_FLTR_ALLOC_INPUT_EN_DST_IPADDR;
2896                 memset(filter->dst_ipaddr_mask, 0xff, 16);
2897                 en |= NTUPLE_FLTR_ALLOC_INPUT_EN_DST_IPADDR_MASK;
2898                 memset(filter->src_ipaddr_mask, 0xff, 16);
2899                 en |= NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_IPADDR_MASK;
2900                 filter->ethertype = 0x86dd;
2901                 filter->enables |= NTUPLE_FLTR_ALLOC_INPUT_EN_ETHERTYPE;
2902                 break;
2903         case RTE_ETH_FLOW_NONFRAG_IPV6_UDP:
2904                 filter->src_port = fdir->input.flow.udp6_flow.src_port;
2905                 en |= NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_PORT;
2906                 filter->dst_port = fdir->input.flow.udp6_flow.dst_port;
2907                 en |= NTUPLE_FLTR_ALLOC_INPUT_EN_DST_PORT;
2908                 filter->dst_port_mask = 0xffff;
2909                 en |= NTUPLE_FLTR_ALLOC_INPUT_EN_DST_PORT_MASK;
2910                 filter->src_port_mask = 0xffff;
2911                 en |= NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_PORT_MASK;
2912                 filter->ip_addr_type =
2913                         NTUPLE_FLTR_ALLOC_INPUT_IP_ADDR_TYPE_IPV6;
2914                 filter->ip_protocol = fdir->input.flow.udp6_flow.ip.proto;
2915                 en |= NTUPLE_FLTR_ALLOC_IN_EN_IP_PROTO;
2916                 rte_memcpy(filter->src_ipaddr,
2917                            fdir->input.flow.udp6_flow.ip.src_ip, 16);
2918                 en |= NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_IPADDR;
2919                 rte_memcpy(filter->dst_ipaddr,
2920                            fdir->input.flow.udp6_flow.ip.dst_ip, 16);
2921                 en |= NTUPLE_FLTR_ALLOC_INPUT_EN_DST_IPADDR;
2922                 memset(filter->dst_ipaddr_mask, 0xff, 16);
2923                 en |= NTUPLE_FLTR_ALLOC_INPUT_EN_DST_IPADDR_MASK;
2924                 memset(filter->src_ipaddr_mask, 0xff, 16);
2925                 en |= NTUPLE_FLTR_ALLOC_INPUT_EN_SRC_IPADDR_MASK;
2926                 filter->ethertype = 0x86dd;
2927                 filter->enables |= NTUPLE_FLTR_ALLOC_INPUT_EN_ETHERTYPE;
2928                 break;
2929         case RTE_ETH_FLOW_L2_PAYLOAD:
2930                 filter->ethertype = fdir->input.flow.l2_flow.ether_type;
2931                 en |= NTUPLE_FLTR_ALLOC_INPUT_EN_ETHERTYPE;
2932                 break;
2933         case RTE_ETH_FLOW_VXLAN:
2934                 if (fdir->action.behavior == RTE_ETH_FDIR_REJECT)
2935                         return -EINVAL;
2936                 filter->vni = fdir->input.flow.tunnel_flow.tunnel_id;
2937                 filter->tunnel_type =
2938                         CFA_NTUPLE_FILTER_ALLOC_REQ_TUNNEL_TYPE_VXLAN;
2939                 en |= HWRM_CFA_NTUPLE_FILTER_ALLOC_INPUT_ENABLES_TUNNEL_TYPE;
2940                 break;
2941         case RTE_ETH_FLOW_NVGRE:
2942                 if (fdir->action.behavior == RTE_ETH_FDIR_REJECT)
2943                         return -EINVAL;
2944                 filter->vni = fdir->input.flow.tunnel_flow.tunnel_id;
2945                 filter->tunnel_type =
2946                         CFA_NTUPLE_FILTER_ALLOC_REQ_TUNNEL_TYPE_NVGRE;
2947                 en |= HWRM_CFA_NTUPLE_FILTER_ALLOC_INPUT_ENABLES_TUNNEL_TYPE;
2948                 break;
2949         case RTE_ETH_FLOW_UNKNOWN:
2950         case RTE_ETH_FLOW_RAW:
2951         case RTE_ETH_FLOW_FRAG_IPV4:
2952         case RTE_ETH_FLOW_NONFRAG_IPV4_SCTP:
2953         case RTE_ETH_FLOW_FRAG_IPV6:
2954         case RTE_ETH_FLOW_NONFRAG_IPV6_SCTP:
2955         case RTE_ETH_FLOW_IPV6_EX:
2956         case RTE_ETH_FLOW_IPV6_TCP_EX:
2957         case RTE_ETH_FLOW_IPV6_UDP_EX:
2958         case RTE_ETH_FLOW_GENEVE:
2959                 /* FALLTHROUGH */
2960         default:
2961                 return -EINVAL;
2962         }
2963
2964         vnic0 = BNXT_GET_DEFAULT_VNIC(bp);
2965         vnic = &bp->vnic_info[fdir->action.rx_queue];
2966         if (vnic == NULL) {
2967                 PMD_DRV_LOG(ERR, "Invalid queue %d\n", fdir->action.rx_queue);
2968                 return -EINVAL;
2969         }
2970
2971         if (fdir_mode == RTE_FDIR_MODE_PERFECT_MAC_VLAN) {
2972                 rte_memcpy(filter->dst_macaddr,
2973                         fdir->input.flow.mac_vlan_flow.mac_addr.addr_bytes, 6);
2974                         en |= NTUPLE_FLTR_ALLOC_INPUT_EN_DST_MACADDR;
2975         }
2976
2977         if (fdir->action.behavior == RTE_ETH_FDIR_REJECT) {
2978                 filter->flags = HWRM_CFA_NTUPLE_FILTER_ALLOC_INPUT_FLAGS_DROP;
2979                 filter1 = STAILQ_FIRST(&vnic0->filter);
2980                 //filter1 = bnxt_get_l2_filter(bp, filter, vnic0);
2981         } else {
2982                 filter->dst_id = vnic->fw_vnic_id;
2983                 for (i = 0; i < RTE_ETHER_ADDR_LEN; i++)
2984                         if (filter->dst_macaddr[i] == 0x00)
2985                                 filter1 = STAILQ_FIRST(&vnic0->filter);
2986                         else
2987                                 filter1 = bnxt_get_l2_filter(bp, filter, vnic);
2988         }
2989
2990         if (filter1 == NULL)
2991                 return -EINVAL;
2992
2993         en |= HWRM_CFA_NTUPLE_FILTER_ALLOC_INPUT_ENABLES_L2_FILTER_ID;
2994         filter->fw_l2_filter_id = filter1->fw_l2_filter_id;
2995
2996         filter->enables = en;
2997
2998         return 0;
2999 }
3000
3001 static struct bnxt_filter_info *
3002 bnxt_match_fdir(struct bnxt *bp, struct bnxt_filter_info *nf,
3003                 struct bnxt_vnic_info **mvnic)
3004 {
3005         struct bnxt_filter_info *mf = NULL;
3006         int i;
3007
3008         for (i = bp->nr_vnics - 1; i >= 0; i--) {
3009                 struct bnxt_vnic_info *vnic = &bp->vnic_info[i];
3010
3011                 STAILQ_FOREACH(mf, &vnic->filter, next) {
3012                         if (mf->filter_type == nf->filter_type &&
3013                             mf->flags == nf->flags &&
3014                             mf->src_port == nf->src_port &&
3015                             mf->src_port_mask == nf->src_port_mask &&
3016                             mf->dst_port == nf->dst_port &&
3017                             mf->dst_port_mask == nf->dst_port_mask &&
3018                             mf->ip_protocol == nf->ip_protocol &&
3019                             mf->ip_addr_type == nf->ip_addr_type &&
3020                             mf->ethertype == nf->ethertype &&
3021                             mf->vni == nf->vni &&
3022                             mf->tunnel_type == nf->tunnel_type &&
3023                             mf->l2_ovlan == nf->l2_ovlan &&
3024                             mf->l2_ovlan_mask == nf->l2_ovlan_mask &&
3025                             mf->l2_ivlan == nf->l2_ivlan &&
3026                             mf->l2_ivlan_mask == nf->l2_ivlan_mask &&
3027                             !memcmp(mf->l2_addr, nf->l2_addr,
3028                                     RTE_ETHER_ADDR_LEN) &&
3029                             !memcmp(mf->l2_addr_mask, nf->l2_addr_mask,
3030                                     RTE_ETHER_ADDR_LEN) &&
3031                             !memcmp(mf->src_macaddr, nf->src_macaddr,
3032                                     RTE_ETHER_ADDR_LEN) &&
3033                             !memcmp(mf->dst_macaddr, nf->dst_macaddr,
3034                                     RTE_ETHER_ADDR_LEN) &&
3035                             !memcmp(mf->src_ipaddr, nf->src_ipaddr,
3036                                     sizeof(nf->src_ipaddr)) &&
3037                             !memcmp(mf->src_ipaddr_mask, nf->src_ipaddr_mask,
3038                                     sizeof(nf->src_ipaddr_mask)) &&
3039                             !memcmp(mf->dst_ipaddr, nf->dst_ipaddr,
3040                                     sizeof(nf->dst_ipaddr)) &&
3041                             !memcmp(mf->dst_ipaddr_mask, nf->dst_ipaddr_mask,
3042                                     sizeof(nf->dst_ipaddr_mask))) {
3043                                 if (mvnic)
3044                                         *mvnic = vnic;
3045                                 return mf;
3046                         }
3047                 }
3048         }
3049         return NULL;
3050 }
3051
3052 static int
3053 bnxt_fdir_filter(struct rte_eth_dev *dev,
3054                  enum rte_filter_op filter_op,
3055                  void *arg)
3056 {
3057         struct bnxt *bp = dev->data->dev_private;
3058         struct rte_eth_fdir_filter *fdir  = (struct rte_eth_fdir_filter *)arg;
3059         struct bnxt_filter_info *filter, *match;
3060         struct bnxt_vnic_info *vnic, *mvnic;
3061         int ret = 0, i;
3062
3063         if (filter_op == RTE_ETH_FILTER_NOP)
3064                 return 0;
3065
3066         if (arg == NULL && filter_op != RTE_ETH_FILTER_FLUSH)
3067                 return -EINVAL;
3068
3069         switch (filter_op) {
3070         case RTE_ETH_FILTER_ADD:
3071         case RTE_ETH_FILTER_DELETE:
3072                 /* FALLTHROUGH */
3073                 filter = bnxt_get_unused_filter(bp);
3074                 if (filter == NULL) {
3075                         PMD_DRV_LOG(ERR,
3076                                 "Not enough resources for a new flow.\n");
3077                         return -ENOMEM;
3078                 }
3079
3080                 ret = bnxt_parse_fdir_filter(bp, fdir, filter);
3081                 if (ret != 0)
3082                         goto free_filter;
3083                 filter->filter_type = HWRM_CFA_NTUPLE_FILTER;
3084
3085                 if (fdir->action.behavior == RTE_ETH_FDIR_REJECT)
3086                         vnic = &bp->vnic_info[0];
3087                 else
3088                         vnic = &bp->vnic_info[fdir->action.rx_queue];
3089
3090                 match = bnxt_match_fdir(bp, filter, &mvnic);
3091                 if (match != NULL && filter_op == RTE_ETH_FILTER_ADD) {
3092                         if (match->dst_id == vnic->fw_vnic_id) {
3093                                 PMD_DRV_LOG(ERR, "Flow already exists.\n");
3094                                 ret = -EEXIST;
3095                                 goto free_filter;
3096                         } else {
3097                                 match->dst_id = vnic->fw_vnic_id;
3098                                 ret = bnxt_hwrm_set_ntuple_filter(bp,
3099                                                                   match->dst_id,
3100                                                                   match);
3101                                 STAILQ_REMOVE(&mvnic->filter, match,
3102                                               bnxt_filter_info, next);
3103                                 STAILQ_INSERT_TAIL(&vnic->filter, match, next);
3104                                 PMD_DRV_LOG(ERR,
3105                                         "Filter with matching pattern exist\n");
3106                                 PMD_DRV_LOG(ERR,
3107                                         "Updated it to new destination q\n");
3108                                 goto free_filter;
3109                         }
3110                 }
3111                 if (match == NULL && filter_op == RTE_ETH_FILTER_DELETE) {
3112                         PMD_DRV_LOG(ERR, "Flow does not exist.\n");
3113                         ret = -ENOENT;
3114                         goto free_filter;
3115                 }
3116
3117                 if (filter_op == RTE_ETH_FILTER_ADD) {
3118                         ret = bnxt_hwrm_set_ntuple_filter(bp,
3119                                                           filter->dst_id,
3120                                                           filter);
3121                         if (ret)
3122                                 goto free_filter;
3123                         STAILQ_INSERT_TAIL(&vnic->filter, filter, next);
3124                 } else {
3125                         ret = bnxt_hwrm_clear_ntuple_filter(bp, match);
3126                         STAILQ_REMOVE(&vnic->filter, match,
3127                                       bnxt_filter_info, next);
3128                         bnxt_free_filter(bp, match);
3129                         bnxt_free_filter(bp, filter);
3130                 }
3131                 break;
3132         case RTE_ETH_FILTER_FLUSH:
3133                 for (i = bp->nr_vnics - 1; i >= 0; i--) {
3134                         struct bnxt_vnic_info *vnic = &bp->vnic_info[i];
3135
3136                         STAILQ_FOREACH(filter, &vnic->filter, next) {
3137                                 if (filter->filter_type ==
3138                                     HWRM_CFA_NTUPLE_FILTER) {
3139                                         ret =
3140                                         bnxt_hwrm_clear_ntuple_filter(bp,
3141                                                                       filter);
3142                                         STAILQ_REMOVE(&vnic->filter, filter,
3143                                                       bnxt_filter_info, next);
3144                                 }
3145                         }
3146                 }
3147                 return ret;
3148         case RTE_ETH_FILTER_UPDATE:
3149         case RTE_ETH_FILTER_STATS:
3150         case RTE_ETH_FILTER_INFO:
3151                 PMD_DRV_LOG(ERR, "operation %u not implemented", filter_op);
3152                 break;
3153         default:
3154                 PMD_DRV_LOG(ERR, "unknown operation %u", filter_op);
3155                 ret = -EINVAL;
3156                 break;
3157         }
3158         return ret;
3159
3160 free_filter:
3161         bnxt_free_filter(bp, filter);
3162         return ret;
3163 }
3164
3165 static int
3166 bnxt_filter_ctrl_op(struct rte_eth_dev *dev,
3167                     enum rte_filter_type filter_type,
3168                     enum rte_filter_op filter_op, void *arg)
3169 {
3170         int ret = 0;
3171
3172         ret = is_bnxt_in_error(dev->data->dev_private);
3173         if (ret)
3174                 return ret;
3175
3176         switch (filter_type) {
3177         case RTE_ETH_FILTER_TUNNEL:
3178                 PMD_DRV_LOG(ERR,
3179                         "filter type: %d: To be implemented\n", filter_type);
3180                 break;
3181         case RTE_ETH_FILTER_FDIR:
3182                 ret = bnxt_fdir_filter(dev, filter_op, arg);
3183                 break;
3184         case RTE_ETH_FILTER_NTUPLE:
3185                 ret = bnxt_ntuple_filter(dev, filter_op, arg);
3186                 break;
3187         case RTE_ETH_FILTER_ETHERTYPE:
3188                 ret = bnxt_ethertype_filter(dev, filter_op, arg);
3189                 break;
3190         case RTE_ETH_FILTER_GENERIC:
3191                 if (filter_op != RTE_ETH_FILTER_GET)
3192                         return -EINVAL;
3193                 *(const void **)arg = &bnxt_flow_ops;
3194                 break;
3195         default:
3196                 PMD_DRV_LOG(ERR,
3197                         "Filter type (%d) not supported", filter_type);
3198                 ret = -EINVAL;
3199                 break;
3200         }
3201         return ret;
3202 }
3203
3204 static const uint32_t *
3205 bnxt_dev_supported_ptypes_get_op(struct rte_eth_dev *dev)
3206 {
3207         static const uint32_t ptypes[] = {
3208                 RTE_PTYPE_L2_ETHER_VLAN,
3209                 RTE_PTYPE_L3_IPV4_EXT_UNKNOWN,
3210                 RTE_PTYPE_L3_IPV6_EXT_UNKNOWN,
3211                 RTE_PTYPE_L4_ICMP,
3212                 RTE_PTYPE_L4_TCP,
3213                 RTE_PTYPE_L4_UDP,
3214                 RTE_PTYPE_INNER_L3_IPV4_EXT_UNKNOWN,
3215                 RTE_PTYPE_INNER_L3_IPV6_EXT_UNKNOWN,
3216                 RTE_PTYPE_INNER_L4_ICMP,
3217                 RTE_PTYPE_INNER_L4_TCP,
3218                 RTE_PTYPE_INNER_L4_UDP,
3219                 RTE_PTYPE_UNKNOWN
3220         };
3221
3222         if (!dev->rx_pkt_burst)
3223                 return NULL;
3224
3225         return ptypes;
3226 }
3227
3228 static int bnxt_map_regs(struct bnxt *bp, uint32_t *reg_arr, int count,
3229                          int reg_win)
3230 {
3231         uint32_t reg_base = *reg_arr & 0xfffff000;
3232         uint32_t win_off;
3233         int i;
3234
3235         for (i = 0; i < count; i++) {
3236                 if ((reg_arr[i] & 0xfffff000) != reg_base)
3237                         return -ERANGE;
3238         }
3239         win_off = BNXT_GRCPF_REG_WINDOW_BASE_OUT + (reg_win - 1) * 4;
3240         rte_write32(reg_base, (uint8_t *)bp->bar0 + win_off);
3241         return 0;
3242 }
3243
3244 static int bnxt_map_ptp_regs(struct bnxt *bp)
3245 {
3246         struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3247         uint32_t *reg_arr;
3248         int rc, i;
3249
3250         reg_arr = ptp->rx_regs;
3251         rc = bnxt_map_regs(bp, reg_arr, BNXT_PTP_RX_REGS, 5);
3252         if (rc)
3253                 return rc;
3254
3255         reg_arr = ptp->tx_regs;
3256         rc = bnxt_map_regs(bp, reg_arr, BNXT_PTP_TX_REGS, 6);
3257         if (rc)
3258                 return rc;
3259
3260         for (i = 0; i < BNXT_PTP_RX_REGS; i++)
3261                 ptp->rx_mapped_regs[i] = 0x5000 + (ptp->rx_regs[i] & 0xfff);
3262
3263         for (i = 0; i < BNXT_PTP_TX_REGS; i++)
3264                 ptp->tx_mapped_regs[i] = 0x6000 + (ptp->tx_regs[i] & 0xfff);
3265
3266         return 0;
3267 }
3268
3269 static void bnxt_unmap_ptp_regs(struct bnxt *bp)
3270 {
3271         rte_write32(0, (uint8_t *)bp->bar0 +
3272                          BNXT_GRCPF_REG_WINDOW_BASE_OUT + 16);
3273         rte_write32(0, (uint8_t *)bp->bar0 +
3274                          BNXT_GRCPF_REG_WINDOW_BASE_OUT + 20);
3275 }
3276
3277 static uint64_t bnxt_cc_read(struct bnxt *bp)
3278 {
3279         uint64_t ns;
3280
3281         ns = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3282                               BNXT_GRCPF_REG_SYNC_TIME));
3283         ns |= (uint64_t)(rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3284                                           BNXT_GRCPF_REG_SYNC_TIME + 4))) << 32;
3285         return ns;
3286 }
3287
3288 static int bnxt_get_tx_ts(struct bnxt *bp, uint64_t *ts)
3289 {
3290         struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3291         uint32_t fifo;
3292
3293         fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3294                                 ptp->tx_mapped_regs[BNXT_PTP_TX_FIFO]));
3295         if (fifo & BNXT_PTP_TX_FIFO_EMPTY)
3296                 return -EAGAIN;
3297
3298         fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3299                                 ptp->tx_mapped_regs[BNXT_PTP_TX_FIFO]));
3300         *ts = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3301                                 ptp->tx_mapped_regs[BNXT_PTP_TX_TS_L]));
3302         *ts |= (uint64_t)rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3303                                 ptp->tx_mapped_regs[BNXT_PTP_TX_TS_H])) << 32;
3304
3305         return 0;
3306 }
3307
3308 static int bnxt_get_rx_ts(struct bnxt *bp, uint64_t *ts)
3309 {
3310         struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3311         struct bnxt_pf_info *pf = &bp->pf;
3312         uint16_t port_id;
3313         uint32_t fifo;
3314
3315         if (!ptp)
3316                 return -ENODEV;
3317
3318         fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3319                                 ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO]));
3320         if (!(fifo & BNXT_PTP_RX_FIFO_PENDING))
3321                 return -EAGAIN;
3322
3323         port_id = pf->port_id;
3324         rte_write32(1 << port_id, (uint8_t *)bp->bar0 +
3325                ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO_ADV]);
3326
3327         fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3328                                    ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO]));
3329         if (fifo & BNXT_PTP_RX_FIFO_PENDING) {
3330 /*              bnxt_clr_rx_ts(bp);       TBD  */
3331                 return -EBUSY;
3332         }
3333
3334         *ts = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3335                                 ptp->rx_mapped_regs[BNXT_PTP_RX_TS_L]));
3336         *ts |= (uint64_t)rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3337                                 ptp->rx_mapped_regs[BNXT_PTP_RX_TS_H])) << 32;
3338
3339         return 0;
3340 }
3341
3342 static int
3343 bnxt_timesync_write_time(struct rte_eth_dev *dev, const struct timespec *ts)
3344 {
3345         uint64_t ns;
3346         struct bnxt *bp = dev->data->dev_private;
3347         struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3348
3349         if (!ptp)
3350                 return 0;
3351
3352         ns = rte_timespec_to_ns(ts);
3353         /* Set the timecounters to a new value. */
3354         ptp->tc.nsec = ns;
3355
3356         return 0;
3357 }
3358
3359 static int
3360 bnxt_timesync_read_time(struct rte_eth_dev *dev, struct timespec *ts)
3361 {
3362         struct bnxt *bp = dev->data->dev_private;
3363         struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3364         uint64_t ns, systime_cycles = 0;
3365         int rc = 0;
3366
3367         if (!ptp)
3368                 return 0;
3369
3370         if (BNXT_CHIP_THOR(bp))
3371                 rc = bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_CURRENT_TIME,
3372                                              &systime_cycles);
3373         else
3374                 systime_cycles = bnxt_cc_read(bp);
3375
3376         ns = rte_timecounter_update(&ptp->tc, systime_cycles);
3377         *ts = rte_ns_to_timespec(ns);
3378
3379         return rc;
3380 }
3381 static int
3382 bnxt_timesync_enable(struct rte_eth_dev *dev)
3383 {
3384         struct bnxt *bp = dev->data->dev_private;
3385         struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3386         uint32_t shift = 0;
3387         int rc;
3388
3389         if (!ptp)
3390                 return 0;
3391
3392         ptp->rx_filter = 1;
3393         ptp->tx_tstamp_en = 1;
3394         ptp->rxctl = BNXT_PTP_MSG_EVENTS;
3395
3396         rc = bnxt_hwrm_ptp_cfg(bp);
3397         if (rc)
3398                 return rc;
3399
3400         memset(&ptp->tc, 0, sizeof(struct rte_timecounter));
3401         memset(&ptp->rx_tstamp_tc, 0, sizeof(struct rte_timecounter));
3402         memset(&ptp->tx_tstamp_tc, 0, sizeof(struct rte_timecounter));
3403
3404         ptp->tc.cc_mask = BNXT_CYCLECOUNTER_MASK;
3405         ptp->tc.cc_shift = shift;
3406         ptp->tc.nsec_mask = (1ULL << shift) - 1;
3407
3408         ptp->rx_tstamp_tc.cc_mask = BNXT_CYCLECOUNTER_MASK;
3409         ptp->rx_tstamp_tc.cc_shift = shift;
3410         ptp->rx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
3411
3412         ptp->tx_tstamp_tc.cc_mask = BNXT_CYCLECOUNTER_MASK;
3413         ptp->tx_tstamp_tc.cc_shift = shift;
3414         ptp->tx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
3415
3416         if (!BNXT_CHIP_THOR(bp))
3417                 bnxt_map_ptp_regs(bp);
3418
3419         return 0;
3420 }
3421
3422 static int
3423 bnxt_timesync_disable(struct rte_eth_dev *dev)
3424 {
3425         struct bnxt *bp = dev->data->dev_private;
3426         struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3427
3428         if (!ptp)
3429                 return 0;
3430
3431         ptp->rx_filter = 0;
3432         ptp->tx_tstamp_en = 0;
3433         ptp->rxctl = 0;
3434
3435         bnxt_hwrm_ptp_cfg(bp);
3436
3437         if (!BNXT_CHIP_THOR(bp))
3438                 bnxt_unmap_ptp_regs(bp);
3439
3440         return 0;
3441 }
3442
3443 static int
3444 bnxt_timesync_read_rx_timestamp(struct rte_eth_dev *dev,
3445                                  struct timespec *timestamp,
3446                                  uint32_t flags __rte_unused)
3447 {
3448         struct bnxt *bp = dev->data->dev_private;
3449         struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3450         uint64_t rx_tstamp_cycles = 0;
3451         uint64_t ns;
3452
3453         if (!ptp)
3454                 return 0;
3455
3456         if (BNXT_CHIP_THOR(bp))
3457                 rx_tstamp_cycles = ptp->rx_timestamp;
3458         else
3459                 bnxt_get_rx_ts(bp, &rx_tstamp_cycles);
3460
3461         ns = rte_timecounter_update(&ptp->rx_tstamp_tc, rx_tstamp_cycles);
3462         *timestamp = rte_ns_to_timespec(ns);
3463         return  0;
3464 }
3465
3466 static int
3467 bnxt_timesync_read_tx_timestamp(struct rte_eth_dev *dev,
3468                                  struct timespec *timestamp)
3469 {
3470         struct bnxt *bp = dev->data->dev_private;
3471         struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3472         uint64_t tx_tstamp_cycles = 0;
3473         uint64_t ns;
3474         int rc = 0;
3475
3476         if (!ptp)
3477                 return 0;
3478
3479         if (BNXT_CHIP_THOR(bp))
3480                 rc = bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_PATH_TX,
3481                                              &tx_tstamp_cycles);
3482         else
3483                 rc = bnxt_get_tx_ts(bp, &tx_tstamp_cycles);
3484
3485         ns = rte_timecounter_update(&ptp->tx_tstamp_tc, tx_tstamp_cycles);
3486         *timestamp = rte_ns_to_timespec(ns);
3487
3488         return rc;
3489 }
3490
3491 static int
3492 bnxt_timesync_adjust_time(struct rte_eth_dev *dev, int64_t delta)
3493 {
3494         struct bnxt *bp = dev->data->dev_private;
3495         struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3496
3497         if (!ptp)
3498                 return 0;
3499
3500         ptp->tc.nsec += delta;
3501
3502         return 0;
3503 }
3504
3505 static int
3506 bnxt_get_eeprom_length_op(struct rte_eth_dev *dev)
3507 {
3508         struct bnxt *bp = dev->data->dev_private;
3509         int rc;
3510         uint32_t dir_entries;
3511         uint32_t entry_length;
3512
3513         rc = is_bnxt_in_error(bp);
3514         if (rc)
3515                 return rc;
3516
3517         PMD_DRV_LOG(INFO, "%04x:%02x:%02x:%02x\n",
3518                 bp->pdev->addr.domain, bp->pdev->addr.bus,
3519                 bp->pdev->addr.devid, bp->pdev->addr.function);
3520
3521         rc = bnxt_hwrm_nvm_get_dir_info(bp, &dir_entries, &entry_length);
3522         if (rc != 0)
3523                 return rc;
3524
3525         return dir_entries * entry_length;
3526 }
3527
3528 static int
3529 bnxt_get_eeprom_op(struct rte_eth_dev *dev,
3530                 struct rte_dev_eeprom_info *in_eeprom)
3531 {
3532         struct bnxt *bp = dev->data->dev_private;
3533         uint32_t index;
3534         uint32_t offset;
3535         int rc;
3536
3537         rc = is_bnxt_in_error(bp);
3538         if (rc)
3539                 return rc;
3540
3541         PMD_DRV_LOG(INFO, "%04x:%02x:%02x:%02x in_eeprom->offset = %d "
3542                 "len = %d\n", bp->pdev->addr.domain,
3543                 bp->pdev->addr.bus, bp->pdev->addr.devid,
3544                 bp->pdev->addr.function, in_eeprom->offset, in_eeprom->length);
3545
3546         if (in_eeprom->offset == 0) /* special offset value to get directory */
3547                 return bnxt_get_nvram_directory(bp, in_eeprom->length,
3548                                                 in_eeprom->data);
3549
3550         index = in_eeprom->offset >> 24;
3551         offset = in_eeprom->offset & 0xffffff;
3552
3553         if (index != 0)
3554                 return bnxt_hwrm_get_nvram_item(bp, index - 1, offset,
3555                                            in_eeprom->length, in_eeprom->data);
3556
3557         return 0;
3558 }
3559
3560 static bool bnxt_dir_type_is_ape_bin_format(uint16_t dir_type)
3561 {
3562         switch (dir_type) {
3563         case BNX_DIR_TYPE_CHIMP_PATCH:
3564         case BNX_DIR_TYPE_BOOTCODE:
3565         case BNX_DIR_TYPE_BOOTCODE_2:
3566         case BNX_DIR_TYPE_APE_FW:
3567         case BNX_DIR_TYPE_APE_PATCH:
3568         case BNX_DIR_TYPE_KONG_FW:
3569         case BNX_DIR_TYPE_KONG_PATCH:
3570         case BNX_DIR_TYPE_BONO_FW:
3571         case BNX_DIR_TYPE_BONO_PATCH:
3572                 /* FALLTHROUGH */
3573                 return true;
3574         }
3575
3576         return false;
3577 }
3578
3579 static bool bnxt_dir_type_is_other_exec_format(uint16_t dir_type)
3580 {
3581         switch (dir_type) {
3582         case BNX_DIR_TYPE_AVS:
3583         case BNX_DIR_TYPE_EXP_ROM_MBA:
3584         case BNX_DIR_TYPE_PCIE:
3585         case BNX_DIR_TYPE_TSCF_UCODE:
3586         case BNX_DIR_TYPE_EXT_PHY:
3587         case BNX_DIR_TYPE_CCM:
3588         case BNX_DIR_TYPE_ISCSI_BOOT:
3589         case BNX_DIR_TYPE_ISCSI_BOOT_IPV6:
3590         case BNX_DIR_TYPE_ISCSI_BOOT_IPV4N6:
3591                 /* FALLTHROUGH */
3592                 return true;
3593         }
3594
3595         return false;
3596 }
3597
3598 static bool bnxt_dir_type_is_executable(uint16_t dir_type)
3599 {
3600         return bnxt_dir_type_is_ape_bin_format(dir_type) ||
3601                 bnxt_dir_type_is_other_exec_format(dir_type);
3602 }
3603
3604 static int
3605 bnxt_set_eeprom_op(struct rte_eth_dev *dev,
3606                 struct rte_dev_eeprom_info *in_eeprom)
3607 {
3608         struct bnxt *bp = dev->data->dev_private;
3609         uint8_t index, dir_op;
3610         uint16_t type, ext, ordinal, attr;
3611         int rc;
3612
3613         rc = is_bnxt_in_error(bp);
3614         if (rc)
3615                 return rc;
3616
3617         PMD_DRV_LOG(INFO, "%04x:%02x:%02x:%02x in_eeprom->offset = %d "
3618                 "len = %d\n", bp->pdev->addr.domain,
3619                 bp->pdev->addr.bus, bp->pdev->addr.devid,
3620                 bp->pdev->addr.function, in_eeprom->offset, in_eeprom->length);
3621
3622         if (!BNXT_PF(bp)) {
3623                 PMD_DRV_LOG(ERR, "NVM write not supported from a VF\n");
3624                 return -EINVAL;
3625         }
3626
3627         type = in_eeprom->magic >> 16;
3628
3629         if (type == 0xffff) { /* special value for directory operations */
3630                 index = in_eeprom->magic & 0xff;
3631                 dir_op = in_eeprom->magic >> 8;
3632                 if (index == 0)
3633                         return -EINVAL;
3634                 switch (dir_op) {
3635                 case 0x0e: /* erase */
3636                         if (in_eeprom->offset != ~in_eeprom->magic)
3637                                 return -EINVAL;
3638                         return bnxt_hwrm_erase_nvram_directory(bp, index - 1);
3639                 default:
3640                         return -EINVAL;
3641                 }
3642         }
3643
3644         /* Create or re-write an NVM item: */
3645         if (bnxt_dir_type_is_executable(type) == true)
3646                 return -EOPNOTSUPP;
3647         ext = in_eeprom->magic & 0xffff;
3648         ordinal = in_eeprom->offset >> 16;
3649         attr = in_eeprom->offset & 0xffff;
3650
3651         return bnxt_hwrm_flash_nvram(bp, type, ordinal, ext, attr,
3652                                      in_eeprom->data, in_eeprom->length);
3653 }
3654
3655 /*
3656  * Initialization
3657  */
3658
3659 static const struct eth_dev_ops bnxt_dev_ops = {
3660         .dev_infos_get = bnxt_dev_info_get_op,
3661         .dev_close = bnxt_dev_close_op,
3662         .dev_configure = bnxt_dev_configure_op,
3663         .dev_start = bnxt_dev_start_op,
3664         .dev_stop = bnxt_dev_stop_op,
3665         .dev_set_link_up = bnxt_dev_set_link_up_op,
3666         .dev_set_link_down = bnxt_dev_set_link_down_op,
3667         .stats_get = bnxt_stats_get_op,
3668         .stats_reset = bnxt_stats_reset_op,
3669         .rx_queue_setup = bnxt_rx_queue_setup_op,
3670         .rx_queue_release = bnxt_rx_queue_release_op,
3671         .tx_queue_setup = bnxt_tx_queue_setup_op,
3672         .tx_queue_release = bnxt_tx_queue_release_op,
3673         .rx_queue_intr_enable = bnxt_rx_queue_intr_enable_op,
3674         .rx_queue_intr_disable = bnxt_rx_queue_intr_disable_op,
3675         .reta_update = bnxt_reta_update_op,
3676         .reta_query = bnxt_reta_query_op,
3677         .rss_hash_update = bnxt_rss_hash_update_op,
3678         .rss_hash_conf_get = bnxt_rss_hash_conf_get_op,
3679         .link_update = bnxt_link_update_op,
3680         .promiscuous_enable = bnxt_promiscuous_enable_op,
3681         .promiscuous_disable = bnxt_promiscuous_disable_op,
3682         .allmulticast_enable = bnxt_allmulticast_enable_op,
3683         .allmulticast_disable = bnxt_allmulticast_disable_op,
3684         .mac_addr_add = bnxt_mac_addr_add_op,
3685         .mac_addr_remove = bnxt_mac_addr_remove_op,
3686         .flow_ctrl_get = bnxt_flow_ctrl_get_op,
3687         .flow_ctrl_set = bnxt_flow_ctrl_set_op,
3688         .udp_tunnel_port_add  = bnxt_udp_tunnel_port_add_op,
3689         .udp_tunnel_port_del  = bnxt_udp_tunnel_port_del_op,
3690         .vlan_filter_set = bnxt_vlan_filter_set_op,
3691         .vlan_offload_set = bnxt_vlan_offload_set_op,
3692         .vlan_tpid_set = bnxt_vlan_tpid_set_op,
3693         .vlan_pvid_set = bnxt_vlan_pvid_set_op,
3694         .mtu_set = bnxt_mtu_set_op,
3695         .mac_addr_set = bnxt_set_default_mac_addr_op,
3696         .xstats_get = bnxt_dev_xstats_get_op,
3697         .xstats_get_names = bnxt_dev_xstats_get_names_op,
3698         .xstats_reset = bnxt_dev_xstats_reset_op,
3699         .fw_version_get = bnxt_fw_version_get,
3700         .set_mc_addr_list = bnxt_dev_set_mc_addr_list_op,
3701         .rxq_info_get = bnxt_rxq_info_get_op,
3702         .txq_info_get = bnxt_txq_info_get_op,
3703         .dev_led_on = bnxt_dev_led_on_op,
3704         .dev_led_off = bnxt_dev_led_off_op,
3705         .xstats_get_by_id = bnxt_dev_xstats_get_by_id_op,
3706         .xstats_get_names_by_id = bnxt_dev_xstats_get_names_by_id_op,
3707         .rx_queue_count = bnxt_rx_queue_count_op,
3708         .rx_descriptor_status = bnxt_rx_descriptor_status_op,
3709         .tx_descriptor_status = bnxt_tx_descriptor_status_op,
3710         .rx_queue_start = bnxt_rx_queue_start,
3711         .rx_queue_stop = bnxt_rx_queue_stop,
3712         .tx_queue_start = bnxt_tx_queue_start,
3713         .tx_queue_stop = bnxt_tx_queue_stop,
3714         .filter_ctrl = bnxt_filter_ctrl_op,
3715         .dev_supported_ptypes_get = bnxt_dev_supported_ptypes_get_op,
3716         .get_eeprom_length    = bnxt_get_eeprom_length_op,
3717         .get_eeprom           = bnxt_get_eeprom_op,
3718         .set_eeprom           = bnxt_set_eeprom_op,
3719         .timesync_enable      = bnxt_timesync_enable,
3720         .timesync_disable     = bnxt_timesync_disable,
3721         .timesync_read_time   = bnxt_timesync_read_time,
3722         .timesync_write_time   = bnxt_timesync_write_time,
3723         .timesync_adjust_time = bnxt_timesync_adjust_time,
3724         .timesync_read_rx_timestamp = bnxt_timesync_read_rx_timestamp,
3725         .timesync_read_tx_timestamp = bnxt_timesync_read_tx_timestamp,
3726 };
3727
3728 static uint32_t bnxt_map_reset_regs(struct bnxt *bp, uint32_t reg)
3729 {
3730         uint32_t offset;
3731
3732         /* Only pre-map the reset GRC registers using window 3 */
3733         rte_write32(reg & 0xfffff000, (uint8_t *)bp->bar0 +
3734                     BNXT_GRCPF_REG_WINDOW_BASE_OUT + 8);
3735
3736         offset = BNXT_GRCP_WINDOW_3_BASE + (reg & 0xffc);
3737
3738         return offset;
3739 }
3740
3741 int bnxt_map_fw_health_status_regs(struct bnxt *bp)
3742 {
3743         struct bnxt_error_recovery_info *info = bp->recovery_info;
3744         uint32_t reg_base = 0xffffffff;
3745         int i;
3746
3747         /* Only pre-map the monitoring GRC registers using window 2 */
3748         for (i = 0; i < BNXT_FW_STATUS_REG_CNT; i++) {
3749                 uint32_t reg = info->status_regs[i];
3750
3751                 if (BNXT_FW_STATUS_REG_TYPE(reg) != BNXT_FW_STATUS_REG_TYPE_GRC)
3752                         continue;
3753
3754                 if (reg_base == 0xffffffff)
3755                         reg_base = reg & 0xfffff000;
3756                 if ((reg & 0xfffff000) != reg_base)
3757                         return -ERANGE;
3758
3759                 /* Use mask 0xffc as the Lower 2 bits indicates
3760                  * address space location
3761                  */
3762                 info->mapped_status_regs[i] = BNXT_GRCP_WINDOW_2_BASE +
3763                                                 (reg & 0xffc);
3764         }
3765
3766         if (reg_base == 0xffffffff)
3767                 return 0;
3768
3769         rte_write32(reg_base, (uint8_t *)bp->bar0 +
3770                     BNXT_GRCPF_REG_WINDOW_BASE_OUT + 4);
3771
3772         return 0;
3773 }
3774
3775 static void bnxt_write_fw_reset_reg(struct bnxt *bp, uint32_t index)
3776 {
3777         struct bnxt_error_recovery_info *info = bp->recovery_info;
3778         uint32_t delay = info->delay_after_reset[index];
3779         uint32_t val = info->reset_reg_val[index];
3780         uint32_t reg = info->reset_reg[index];
3781         uint32_t type, offset;
3782
3783         type = BNXT_FW_STATUS_REG_TYPE(reg);
3784         offset = BNXT_FW_STATUS_REG_OFF(reg);
3785
3786         switch (type) {
3787         case BNXT_FW_STATUS_REG_TYPE_CFG:
3788                 rte_pci_write_config(bp->pdev, &val, sizeof(val), offset);
3789                 break;
3790         case BNXT_FW_STATUS_REG_TYPE_GRC:
3791                 offset = bnxt_map_reset_regs(bp, offset);
3792                 rte_write32(val, (uint8_t *)bp->bar0 + offset);
3793                 break;
3794         case BNXT_FW_STATUS_REG_TYPE_BAR0:
3795                 rte_write32(val, (uint8_t *)bp->bar0 + offset);
3796                 break;
3797         }
3798         /* wait on a specific interval of time until core reset is complete */
3799         if (delay)
3800                 rte_delay_ms(delay);
3801 }
3802
3803 static void bnxt_dev_cleanup(struct bnxt *bp)
3804 {
3805         bnxt_set_hwrm_link_config(bp, false);
3806         bp->link_info.link_up = 0;
3807         if (bp->dev_stopped == 0)
3808                 bnxt_dev_stop_op(bp->eth_dev);
3809
3810         bnxt_uninit_resources(bp, true);
3811 }
3812
3813 static int bnxt_restore_vlan_filters(struct bnxt *bp)
3814 {
3815         struct rte_eth_dev *dev = bp->eth_dev;
3816         struct rte_vlan_filter_conf *vfc;
3817         int vidx, vbit, rc;
3818         uint16_t vlan_id;
3819
3820         for (vlan_id = 1; vlan_id <= RTE_ETHER_MAX_VLAN_ID; vlan_id++) {
3821                 vfc = &dev->data->vlan_filter_conf;
3822                 vidx = vlan_id / 64;
3823                 vbit = vlan_id % 64;
3824
3825                 /* Each bit corresponds to a VLAN id */
3826                 if (vfc->ids[vidx] & (UINT64_C(1) << vbit)) {
3827                         rc = bnxt_add_vlan_filter(bp, vlan_id);
3828                         if (rc)
3829                                 return rc;
3830                 }
3831         }
3832
3833         return 0;
3834 }
3835
3836 static int bnxt_restore_mac_filters(struct bnxt *bp)
3837 {
3838         struct rte_eth_dev *dev = bp->eth_dev;
3839         struct rte_eth_dev_info dev_info;
3840         struct rte_ether_addr *addr;
3841         uint64_t pool_mask;
3842         uint32_t pool = 0;
3843         uint16_t i;
3844         int rc;
3845
3846         if (BNXT_VF(bp) & !BNXT_VF_IS_TRUSTED(bp))
3847                 return 0;
3848
3849         rc = bnxt_dev_info_get_op(dev, &dev_info);
3850         if (rc)
3851                 return rc;
3852
3853         /* replay MAC address configuration */
3854         for (i = 1; i < dev_info.max_mac_addrs; i++) {
3855                 addr = &dev->data->mac_addrs[i];
3856
3857                 /* skip zero address */
3858                 if (rte_is_zero_ether_addr(addr))
3859                         continue;
3860
3861                 pool = 0;
3862                 pool_mask = dev->data->mac_pool_sel[i];
3863
3864                 do {
3865                         if (pool_mask & 1ULL) {
3866                                 rc = bnxt_mac_addr_add_op(dev, addr, i, pool);
3867                                 if (rc)
3868                                         return rc;
3869                         }
3870                         pool_mask >>= 1;
3871                         pool++;
3872                 } while (pool_mask);
3873         }
3874
3875         return 0;
3876 }
3877
3878 static int bnxt_restore_filters(struct bnxt *bp)
3879 {
3880         struct rte_eth_dev *dev = bp->eth_dev;
3881         int ret = 0;
3882
3883         if (dev->data->all_multicast)
3884                 ret = bnxt_allmulticast_enable_op(dev);
3885         if (dev->data->promiscuous)
3886                 ret = bnxt_promiscuous_enable_op(dev);
3887
3888         ret = bnxt_restore_mac_filters(bp);
3889         if (ret)
3890                 return ret;
3891
3892         ret = bnxt_restore_vlan_filters(bp);
3893         /* TODO restore other filters as well */
3894         return ret;
3895 }
3896
3897 static void bnxt_dev_recover(void *arg)
3898 {
3899         struct bnxt *bp = arg;
3900         int timeout = bp->fw_reset_max_msecs;
3901         int rc = 0;
3902
3903         /* Clear Error flag so that device re-init should happen */
3904         bp->flags &= ~BNXT_FLAG_FATAL_ERROR;
3905
3906         do {
3907                 rc = bnxt_hwrm_ver_get(bp);
3908                 if (rc == 0)
3909                         break;
3910                 rte_delay_ms(BNXT_FW_READY_WAIT_INTERVAL);
3911                 timeout -= BNXT_FW_READY_WAIT_INTERVAL;
3912         } while (rc && timeout);
3913
3914         if (rc) {
3915                 PMD_DRV_LOG(ERR, "FW is not Ready after reset\n");
3916                 goto err;
3917         }
3918
3919         rc = bnxt_init_resources(bp, true);
3920         if (rc) {
3921                 PMD_DRV_LOG(ERR,
3922                             "Failed to initialize resources after reset\n");
3923                 goto err;
3924         }
3925         /* clear reset flag as the device is initialized now */
3926         bp->flags &= ~BNXT_FLAG_FW_RESET;
3927
3928         rc = bnxt_dev_start_op(bp->eth_dev);
3929         if (rc) {
3930                 PMD_DRV_LOG(ERR, "Failed to start port after reset\n");
3931                 goto err;
3932         }
3933
3934         rc = bnxt_restore_filters(bp);
3935         if (rc)
3936                 goto err;
3937
3938         PMD_DRV_LOG(INFO, "Recovered from FW reset\n");
3939         return;
3940 err:
3941         bp->flags |= BNXT_FLAG_FATAL_ERROR;
3942         bnxt_uninit_resources(bp, false);
3943         PMD_DRV_LOG(ERR, "Failed to recover from FW reset\n");
3944 }
3945
3946 void bnxt_dev_reset_and_resume(void *arg)
3947 {
3948         struct bnxt *bp = arg;
3949         int rc;
3950
3951         bnxt_dev_cleanup(bp);
3952
3953         bnxt_wait_for_device_shutdown(bp);
3954
3955         rc = rte_eal_alarm_set(US_PER_MS * bp->fw_reset_min_msecs,
3956                                bnxt_dev_recover, (void *)bp);
3957         if (rc)
3958                 PMD_DRV_LOG(ERR, "Error setting recovery alarm");
3959 }
3960
3961 uint32_t bnxt_read_fw_status_reg(struct bnxt *bp, uint32_t index)
3962 {
3963         struct bnxt_error_recovery_info *info = bp->recovery_info;
3964         uint32_t reg = info->status_regs[index];
3965         uint32_t type, offset, val = 0;
3966
3967         type = BNXT_FW_STATUS_REG_TYPE(reg);
3968         offset = BNXT_FW_STATUS_REG_OFF(reg);
3969
3970         switch (type) {
3971         case BNXT_FW_STATUS_REG_TYPE_CFG:
3972                 rte_pci_read_config(bp->pdev, &val, sizeof(val), offset);
3973                 break;
3974         case BNXT_FW_STATUS_REG_TYPE_GRC:
3975                 offset = info->mapped_status_regs[index];
3976                 /* FALLTHROUGH */
3977         case BNXT_FW_STATUS_REG_TYPE_BAR0:
3978                 val = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3979                                        offset));
3980                 break;
3981         }
3982
3983         return val;
3984 }
3985
3986 static int bnxt_fw_reset_all(struct bnxt *bp)
3987 {
3988         struct bnxt_error_recovery_info *info = bp->recovery_info;
3989         uint32_t i;
3990         int rc = 0;
3991
3992         if (info->flags & BNXT_FLAG_ERROR_RECOVERY_HOST) {
3993                 /* Reset through master function driver */
3994                 for (i = 0; i < info->reg_array_cnt; i++)
3995                         bnxt_write_fw_reset_reg(bp, i);
3996                 /* Wait for time specified by FW after triggering reset */
3997                 rte_delay_ms(info->master_func_wait_period_after_reset);
3998         } else if (info->flags & BNXT_FLAG_ERROR_RECOVERY_CO_CPU) {
3999                 /* Reset with the help of Kong processor */
4000                 rc = bnxt_hwrm_fw_reset(bp);
4001                 if (rc)
4002                         PMD_DRV_LOG(ERR, "Failed to reset FW\n");
4003         }
4004
4005         return rc;
4006 }
4007
4008 static void bnxt_fw_reset_cb(void *arg)
4009 {
4010         struct bnxt *bp = arg;
4011         struct bnxt_error_recovery_info *info = bp->recovery_info;
4012         int rc = 0;
4013
4014         /* Only Master function can do FW reset */
4015         if (bnxt_is_master_func(bp) &&
4016             bnxt_is_recovery_enabled(bp)) {
4017                 rc = bnxt_fw_reset_all(bp);
4018                 if (rc) {
4019                         PMD_DRV_LOG(ERR, "Adapter recovery failed\n");
4020                         return;
4021                 }
4022         }
4023
4024         /* if recovery method is ERROR_RECOVERY_CO_CPU, KONG will send
4025          * EXCEPTION_FATAL_ASYNC event to all the functions
4026          * (including MASTER FUNC). After receiving this Async, all the active
4027          * drivers should treat this case as FW initiated recovery
4028          */
4029         if (info->flags & BNXT_FLAG_ERROR_RECOVERY_HOST) {
4030                 bp->fw_reset_min_msecs = BNXT_MIN_FW_READY_TIMEOUT;
4031                 bp->fw_reset_max_msecs = BNXT_MAX_FW_RESET_TIMEOUT;
4032
4033                 /* To recover from error */
4034                 rte_eal_alarm_set(US_PER_MS, bnxt_dev_reset_and_resume,
4035                                   (void *)bp);
4036         }
4037 }
4038
4039 /* Driver should poll FW heartbeat, reset_counter with the frequency
4040  * advertised by FW in HWRM_ERROR_RECOVERY_QCFG.
4041  * When the driver detects heartbeat stop or change in reset_counter,
4042  * it has to trigger a reset to recover from the error condition.
4043  * A “master PF” is the function who will have the privilege to
4044  * initiate the chimp reset. The master PF will be elected by the
4045  * firmware and will be notified through async message.
4046  */
4047 static void bnxt_check_fw_health(void *arg)
4048 {
4049         struct bnxt *bp = arg;
4050         struct bnxt_error_recovery_info *info = bp->recovery_info;
4051         uint32_t val = 0, wait_msec;
4052
4053         if (!info || !bnxt_is_recovery_enabled(bp) ||
4054             is_bnxt_in_error(bp))
4055                 return;
4056
4057         val = bnxt_read_fw_status_reg(bp, BNXT_FW_HEARTBEAT_CNT_REG);
4058         if (val == info->last_heart_beat)
4059                 goto reset;
4060
4061         info->last_heart_beat = val;
4062
4063         val = bnxt_read_fw_status_reg(bp, BNXT_FW_RECOVERY_CNT_REG);
4064         if (val != info->last_reset_counter)
4065                 goto reset;
4066
4067         info->last_reset_counter = val;
4068
4069         rte_eal_alarm_set(US_PER_MS * info->driver_polling_freq,
4070                           bnxt_check_fw_health, (void *)bp);
4071
4072         return;
4073 reset:
4074         /* Stop DMA to/from device */
4075         bp->flags |= BNXT_FLAG_FATAL_ERROR;
4076         bp->flags |= BNXT_FLAG_FW_RESET;
4077
4078         PMD_DRV_LOG(ERR, "Detected FW dead condition\n");
4079
4080         if (bnxt_is_master_func(bp))
4081                 wait_msec = info->master_func_wait_period;
4082         else
4083                 wait_msec = info->normal_func_wait_period;
4084
4085         rte_eal_alarm_set(US_PER_MS * wait_msec,
4086                           bnxt_fw_reset_cb, (void *)bp);
4087 }
4088
4089 void bnxt_schedule_fw_health_check(struct bnxt *bp)
4090 {
4091         uint32_t polling_freq;
4092
4093         if (!bnxt_is_recovery_enabled(bp))
4094                 return;
4095
4096         if (bp->flags & BNXT_FLAG_FW_HEALTH_CHECK_SCHEDULED)
4097                 return;
4098
4099         polling_freq = bp->recovery_info->driver_polling_freq;
4100
4101         rte_eal_alarm_set(US_PER_MS * polling_freq,
4102                           bnxt_check_fw_health, (void *)bp);
4103         bp->flags |= BNXT_FLAG_FW_HEALTH_CHECK_SCHEDULED;
4104 }
4105
4106 static void bnxt_cancel_fw_health_check(struct bnxt *bp)
4107 {
4108         if (!bnxt_is_recovery_enabled(bp))
4109                 return;
4110
4111         rte_eal_alarm_cancel(bnxt_check_fw_health, (void *)bp);
4112         bp->flags &= ~BNXT_FLAG_FW_HEALTH_CHECK_SCHEDULED;
4113 }
4114
4115 static bool bnxt_vf_pciid(uint16_t device_id)
4116 {
4117         switch (device_id) {
4118         case BROADCOM_DEV_ID_57304_VF:
4119         case BROADCOM_DEV_ID_57406_VF:
4120         case BROADCOM_DEV_ID_5731X_VF:
4121         case BROADCOM_DEV_ID_5741X_VF:
4122         case BROADCOM_DEV_ID_57414_VF:
4123         case BROADCOM_DEV_ID_STRATUS_NIC_VF1:
4124         case BROADCOM_DEV_ID_STRATUS_NIC_VF2:
4125         case BROADCOM_DEV_ID_58802_VF:
4126         case BROADCOM_DEV_ID_57500_VF1:
4127         case BROADCOM_DEV_ID_57500_VF2:
4128                 /* FALLTHROUGH */
4129                 return true;
4130         default:
4131                 return false;
4132         }
4133 }
4134
4135 static bool bnxt_thor_device(uint16_t device_id)
4136 {
4137         switch (device_id) {
4138         case BROADCOM_DEV_ID_57508:
4139         case BROADCOM_DEV_ID_57504:
4140         case BROADCOM_DEV_ID_57502:
4141         case BROADCOM_DEV_ID_57508_MF1:
4142         case BROADCOM_DEV_ID_57504_MF1:
4143         case BROADCOM_DEV_ID_57502_MF1:
4144         case BROADCOM_DEV_ID_57508_MF2:
4145         case BROADCOM_DEV_ID_57504_MF2:
4146         case BROADCOM_DEV_ID_57502_MF2:
4147         case BROADCOM_DEV_ID_57500_VF1:
4148         case BROADCOM_DEV_ID_57500_VF2:
4149                 /* FALLTHROUGH */
4150                 return true;
4151         default:
4152                 return false;
4153         }
4154 }
4155
4156 bool bnxt_stratus_device(struct bnxt *bp)
4157 {
4158         uint16_t device_id = bp->pdev->id.device_id;
4159
4160         switch (device_id) {
4161         case BROADCOM_DEV_ID_STRATUS_NIC:
4162         case BROADCOM_DEV_ID_STRATUS_NIC_VF1:
4163         case BROADCOM_DEV_ID_STRATUS_NIC_VF2:
4164                 /* FALLTHROUGH */
4165                 return true;
4166         default:
4167                 return false;
4168         }
4169 }
4170
4171 static int bnxt_init_board(struct rte_eth_dev *eth_dev)
4172 {
4173         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
4174         struct bnxt *bp = eth_dev->data->dev_private;
4175
4176         /* enable device (incl. PCI PM wakeup), and bus-mastering */
4177         bp->bar0 = (void *)pci_dev->mem_resource[0].addr;
4178         bp->doorbell_base = (void *)pci_dev->mem_resource[2].addr;
4179         if (!bp->bar0 || !bp->doorbell_base) {
4180                 PMD_DRV_LOG(ERR, "Unable to access Hardware\n");
4181                 return -ENODEV;
4182         }
4183
4184         bp->eth_dev = eth_dev;
4185         bp->pdev = pci_dev;
4186
4187         return 0;
4188 }
4189
4190 static int bnxt_alloc_ctx_mem_blk(struct bnxt *bp,
4191                                   struct bnxt_ctx_pg_info *ctx_pg,
4192                                   uint32_t mem_size,
4193                                   const char *suffix,
4194                                   uint16_t idx)
4195 {
4196         struct bnxt_ring_mem_info *rmem = &ctx_pg->ring_mem;
4197         const struct rte_memzone *mz = NULL;
4198         char mz_name[RTE_MEMZONE_NAMESIZE];
4199         rte_iova_t mz_phys_addr;
4200         uint64_t valid_bits = 0;
4201         uint32_t sz;
4202         int i;
4203
4204         if (!mem_size)
4205                 return 0;
4206
4207         rmem->nr_pages = RTE_ALIGN_MUL_CEIL(mem_size, BNXT_PAGE_SIZE) /
4208                          BNXT_PAGE_SIZE;
4209         rmem->page_size = BNXT_PAGE_SIZE;
4210         rmem->pg_arr = ctx_pg->ctx_pg_arr;
4211         rmem->dma_arr = ctx_pg->ctx_dma_arr;
4212         rmem->flags = BNXT_RMEM_VALID_PTE_FLAG;
4213
4214         valid_bits = PTU_PTE_VALID;
4215
4216         if (rmem->nr_pages > 1) {
4217                 snprintf(mz_name, RTE_MEMZONE_NAMESIZE,
4218                          "bnxt_ctx_pg_tbl%s_%x_%d",
4219                          suffix, idx, bp->eth_dev->data->port_id);
4220                 mz_name[RTE_MEMZONE_NAMESIZE - 1] = 0;
4221                 mz = rte_memzone_lookup(mz_name);
4222                 if (!mz) {
4223                         mz = rte_memzone_reserve_aligned(mz_name,
4224                                                 rmem->nr_pages * 8,
4225                                                 SOCKET_ID_ANY,
4226                                                 RTE_MEMZONE_2MB |
4227                                                 RTE_MEMZONE_SIZE_HINT_ONLY |
4228                                                 RTE_MEMZONE_IOVA_CONTIG,
4229                                                 BNXT_PAGE_SIZE);
4230                         if (mz == NULL)
4231                                 return -ENOMEM;
4232                 }
4233
4234                 memset(mz->addr, 0, mz->len);
4235                 mz_phys_addr = mz->iova;
4236                 if ((unsigned long)mz->addr == mz_phys_addr) {
4237                         PMD_DRV_LOG(DEBUG,
4238                                     "physical address same as virtual\n");
4239                         PMD_DRV_LOG(DEBUG, "Using rte_mem_virt2iova()\n");
4240                         mz_phys_addr = rte_mem_virt2iova(mz->addr);
4241                         if (mz_phys_addr == RTE_BAD_IOVA) {
4242                                 PMD_DRV_LOG(ERR,
4243                                         "unable to map addr to phys memory\n");
4244                                 return -ENOMEM;
4245                         }
4246                 }
4247                 rte_mem_lock_page(((char *)mz->addr));
4248
4249                 rmem->pg_tbl = mz->addr;
4250                 rmem->pg_tbl_map = mz_phys_addr;
4251                 rmem->pg_tbl_mz = mz;
4252         }
4253
4254         snprintf(mz_name, RTE_MEMZONE_NAMESIZE, "bnxt_ctx_%s_%x_%d",
4255                  suffix, idx, bp->eth_dev->data->port_id);
4256         mz = rte_memzone_lookup(mz_name);
4257         if (!mz) {
4258                 mz = rte_memzone_reserve_aligned(mz_name,
4259                                                  mem_size,
4260                                                  SOCKET_ID_ANY,
4261                                                  RTE_MEMZONE_1GB |
4262                                                  RTE_MEMZONE_SIZE_HINT_ONLY |
4263                                                  RTE_MEMZONE_IOVA_CONTIG,
4264                                                  BNXT_PAGE_SIZE);
4265                 if (mz == NULL)
4266                         return -ENOMEM;
4267         }
4268
4269         memset(mz->addr, 0, mz->len);
4270         mz_phys_addr = mz->iova;
4271         if ((unsigned long)mz->addr == mz_phys_addr) {
4272                 PMD_DRV_LOG(DEBUG,
4273                             "Memzone physical address same as virtual.\n");
4274                 PMD_DRV_LOG(DEBUG, "Using rte_mem_virt2iova()\n");
4275                 for (sz = 0; sz < mem_size; sz += BNXT_PAGE_SIZE)
4276                         rte_mem_lock_page(((char *)mz->addr) + sz);
4277                 mz_phys_addr = rte_mem_virt2iova(mz->addr);
4278                 if (mz_phys_addr == RTE_BAD_IOVA) {
4279                         PMD_DRV_LOG(ERR,
4280                                     "unable to map addr to phys memory\n");
4281                         return -ENOMEM;
4282                 }
4283         }
4284
4285         for (sz = 0, i = 0; sz < mem_size; sz += BNXT_PAGE_SIZE, i++) {
4286                 rte_mem_lock_page(((char *)mz->addr) + sz);
4287                 rmem->pg_arr[i] = ((char *)mz->addr) + sz;
4288                 rmem->dma_arr[i] = mz_phys_addr + sz;
4289
4290                 if (rmem->nr_pages > 1) {
4291                         if (i == rmem->nr_pages - 2 &&
4292                             (rmem->flags & BNXT_RMEM_RING_PTE_FLAG))
4293                                 valid_bits |= PTU_PTE_NEXT_TO_LAST;
4294                         else if (i == rmem->nr_pages - 1 &&
4295                                  (rmem->flags & BNXT_RMEM_RING_PTE_FLAG))
4296                                 valid_bits |= PTU_PTE_LAST;
4297
4298                         rmem->pg_tbl[i] = rte_cpu_to_le_64(rmem->dma_arr[i] |
4299                                                            valid_bits);
4300                 }
4301         }
4302
4303         rmem->mz = mz;
4304         if (rmem->vmem_size)
4305                 rmem->vmem = (void **)mz->addr;
4306         rmem->dma_arr[0] = mz_phys_addr;
4307         return 0;
4308 }
4309
4310 static void bnxt_free_ctx_mem(struct bnxt *bp)
4311 {
4312         int i;
4313
4314         if (!bp->ctx || !(bp->ctx->flags & BNXT_CTX_FLAG_INITED))
4315                 return;
4316
4317         bp->ctx->flags &= ~BNXT_CTX_FLAG_INITED;
4318         rte_memzone_free(bp->ctx->qp_mem.ring_mem.mz);
4319         rte_memzone_free(bp->ctx->srq_mem.ring_mem.mz);
4320         rte_memzone_free(bp->ctx->cq_mem.ring_mem.mz);
4321         rte_memzone_free(bp->ctx->vnic_mem.ring_mem.mz);
4322         rte_memzone_free(bp->ctx->stat_mem.ring_mem.mz);
4323         rte_memzone_free(bp->ctx->qp_mem.ring_mem.pg_tbl_mz);
4324         rte_memzone_free(bp->ctx->srq_mem.ring_mem.pg_tbl_mz);
4325         rte_memzone_free(bp->ctx->cq_mem.ring_mem.pg_tbl_mz);
4326         rte_memzone_free(bp->ctx->vnic_mem.ring_mem.pg_tbl_mz);
4327         rte_memzone_free(bp->ctx->stat_mem.ring_mem.pg_tbl_mz);
4328
4329         for (i = 0; i < BNXT_MAX_Q; i++) {
4330                 if (bp->ctx->tqm_mem[i])
4331                         rte_memzone_free(bp->ctx->tqm_mem[i]->ring_mem.mz);
4332         }
4333
4334         rte_free(bp->ctx);
4335         bp->ctx = NULL;
4336 }
4337
4338 #define bnxt_roundup(x, y)   ((((x) + ((y) - 1)) / (y)) * (y))
4339
4340 #define min_t(type, x, y) ({                    \
4341         type __min1 = (x);                      \
4342         type __min2 = (y);                      \
4343         __min1 < __min2 ? __min1 : __min2; })
4344
4345 #define max_t(type, x, y) ({                    \
4346         type __max1 = (x);                      \
4347         type __max2 = (y);                      \
4348         __max1 > __max2 ? __max1 : __max2; })
4349
4350 #define clamp_t(type, _x, min, max)     min_t(type, max_t(type, _x, min), max)
4351
4352 int bnxt_alloc_ctx_mem(struct bnxt *bp)
4353 {
4354         struct bnxt_ctx_pg_info *ctx_pg;
4355         struct bnxt_ctx_mem_info *ctx;
4356         uint32_t mem_size, ena, entries;
4357         int i, rc;
4358
4359         rc = bnxt_hwrm_func_backing_store_qcaps(bp);
4360         if (rc) {
4361                 PMD_DRV_LOG(ERR, "Query context mem capability failed\n");
4362                 return rc;
4363         }
4364         ctx = bp->ctx;
4365         if (!ctx || (ctx->flags & BNXT_CTX_FLAG_INITED))
4366                 return 0;
4367
4368         ctx_pg = &ctx->qp_mem;
4369         ctx_pg->entries = ctx->qp_min_qp1_entries + ctx->qp_max_l2_entries;
4370         mem_size = ctx->qp_entry_size * ctx_pg->entries;
4371         rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "qp_mem", 0);
4372         if (rc)
4373                 return rc;
4374
4375         ctx_pg = &ctx->srq_mem;
4376         ctx_pg->entries = ctx->srq_max_l2_entries;
4377         mem_size = ctx->srq_entry_size * ctx_pg->entries;
4378         rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "srq_mem", 0);
4379         if (rc)
4380                 return rc;
4381
4382         ctx_pg = &ctx->cq_mem;
4383         ctx_pg->entries = ctx->cq_max_l2_entries;
4384         mem_size = ctx->cq_entry_size * ctx_pg->entries;
4385         rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "cq_mem", 0);
4386         if (rc)
4387                 return rc;
4388
4389         ctx_pg = &ctx->vnic_mem;
4390         ctx_pg->entries = ctx->vnic_max_vnic_entries +
4391                 ctx->vnic_max_ring_table_entries;
4392         mem_size = ctx->vnic_entry_size * ctx_pg->entries;
4393         rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "vnic_mem", 0);
4394         if (rc)
4395                 return rc;
4396
4397         ctx_pg = &ctx->stat_mem;
4398         ctx_pg->entries = ctx->stat_max_entries;
4399         mem_size = ctx->stat_entry_size * ctx_pg->entries;
4400         rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "stat_mem", 0);
4401         if (rc)
4402                 return rc;
4403
4404         entries = ctx->qp_max_l2_entries +
4405                   ctx->vnic_max_vnic_entries +
4406                   ctx->tqm_min_entries_per_ring;
4407         entries = bnxt_roundup(entries, ctx->tqm_entries_multiple);
4408         entries = clamp_t(uint32_t, entries, ctx->tqm_min_entries_per_ring,
4409                           ctx->tqm_max_entries_per_ring);
4410         for (i = 0, ena = 0; i < BNXT_MAX_Q; i++) {
4411                 ctx_pg = ctx->tqm_mem[i];
4412                 /* use min tqm entries for now. */
4413                 ctx_pg->entries = entries;
4414                 mem_size = ctx->tqm_entry_size * ctx_pg->entries;
4415                 rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "tqm_mem", i);
4416                 if (rc)
4417                         return rc;
4418                 ena |= HWRM_FUNC_BACKING_STORE_CFG_INPUT_ENABLES_TQM_SP << i;
4419         }
4420
4421         ena |= FUNC_BACKING_STORE_CFG_INPUT_DFLT_ENABLES;
4422         rc = bnxt_hwrm_func_backing_store_cfg(bp, ena);
4423         if (rc)
4424                 PMD_DRV_LOG(ERR,
4425                             "Failed to configure context mem: rc = %d\n", rc);
4426         else
4427                 ctx->flags |= BNXT_CTX_FLAG_INITED;
4428
4429         return rc;
4430 }
4431
4432 static int bnxt_alloc_stats_mem(struct bnxt *bp)
4433 {
4434         struct rte_pci_device *pci_dev = bp->pdev;
4435         char mz_name[RTE_MEMZONE_NAMESIZE];
4436         const struct rte_memzone *mz = NULL;
4437         uint32_t total_alloc_len;
4438         rte_iova_t mz_phys_addr;
4439
4440         if (pci_dev->id.device_id == BROADCOM_DEV_ID_NS2)
4441                 return 0;
4442
4443         snprintf(mz_name, RTE_MEMZONE_NAMESIZE,
4444                  "bnxt_" PCI_PRI_FMT "-%s", pci_dev->addr.domain,
4445                  pci_dev->addr.bus, pci_dev->addr.devid,
4446                  pci_dev->addr.function, "rx_port_stats");
4447         mz_name[RTE_MEMZONE_NAMESIZE - 1] = 0;
4448         mz = rte_memzone_lookup(mz_name);
4449         total_alloc_len =
4450                 RTE_CACHE_LINE_ROUNDUP(sizeof(struct rx_port_stats) +
4451                                        sizeof(struct rx_port_stats_ext) + 512);
4452         if (!mz) {
4453                 mz = rte_memzone_reserve(mz_name, total_alloc_len,
4454                                          SOCKET_ID_ANY,
4455                                          RTE_MEMZONE_2MB |
4456                                          RTE_MEMZONE_SIZE_HINT_ONLY |
4457                                          RTE_MEMZONE_IOVA_CONTIG);
4458                 if (mz == NULL)
4459                         return -ENOMEM;
4460         }
4461         memset(mz->addr, 0, mz->len);
4462         mz_phys_addr = mz->iova;
4463         if ((unsigned long)mz->addr == mz_phys_addr) {
4464                 PMD_DRV_LOG(DEBUG,
4465                             "Memzone physical address same as virtual.\n");
4466                 PMD_DRV_LOG(DEBUG,
4467                             "Using rte_mem_virt2iova()\n");
4468                 mz_phys_addr = rte_mem_virt2iova(mz->addr);
4469                 if (mz_phys_addr == RTE_BAD_IOVA) {
4470                         PMD_DRV_LOG(ERR,
4471                                     "Can't map address to physical memory\n");
4472                         return -ENOMEM;
4473                 }
4474         }
4475
4476         bp->rx_mem_zone = (const void *)mz;
4477         bp->hw_rx_port_stats = mz->addr;
4478         bp->hw_rx_port_stats_map = mz_phys_addr;
4479
4480         snprintf(mz_name, RTE_MEMZONE_NAMESIZE,
4481                  "bnxt_" PCI_PRI_FMT "-%s", pci_dev->addr.domain,
4482                  pci_dev->addr.bus, pci_dev->addr.devid,
4483                  pci_dev->addr.function, "tx_port_stats");
4484         mz_name[RTE_MEMZONE_NAMESIZE - 1] = 0;
4485         mz = rte_memzone_lookup(mz_name);
4486         total_alloc_len =
4487                 RTE_CACHE_LINE_ROUNDUP(sizeof(struct tx_port_stats) +
4488                                        sizeof(struct tx_port_stats_ext) + 512);
4489         if (!mz) {
4490                 mz = rte_memzone_reserve(mz_name,
4491                                          total_alloc_len,
4492                                          SOCKET_ID_ANY,
4493                                          RTE_MEMZONE_2MB |
4494                                          RTE_MEMZONE_SIZE_HINT_ONLY |
4495                                          RTE_MEMZONE_IOVA_CONTIG);
4496                 if (mz == NULL)
4497                         return -ENOMEM;
4498         }
4499         memset(mz->addr, 0, mz->len);
4500         mz_phys_addr = mz->iova;
4501         if ((unsigned long)mz->addr == mz_phys_addr) {
4502                 PMD_DRV_LOG(DEBUG,
4503                             "Memzone physical address same as virtual\n");
4504                 PMD_DRV_LOG(DEBUG, "Using rte_mem_virt2iova()\n");
4505                 mz_phys_addr = rte_mem_virt2iova(mz->addr);
4506                 if (mz_phys_addr == RTE_BAD_IOVA) {
4507                         PMD_DRV_LOG(ERR,
4508                                     "Can't map address to physical memory\n");
4509                         return -ENOMEM;
4510                 }
4511         }
4512
4513         bp->tx_mem_zone = (const void *)mz;
4514         bp->hw_tx_port_stats = mz->addr;
4515         bp->hw_tx_port_stats_map = mz_phys_addr;
4516         bp->flags |= BNXT_FLAG_PORT_STATS;
4517
4518         /* Display extended statistics if FW supports it */
4519         if (bp->hwrm_spec_code < HWRM_SPEC_CODE_1_8_4 ||
4520             bp->hwrm_spec_code == HWRM_SPEC_CODE_1_9_0 ||
4521             !(bp->flags & BNXT_FLAG_EXT_STATS_SUPPORTED))
4522                 return 0;
4523
4524         bp->hw_rx_port_stats_ext = (void *)
4525                 ((uint8_t *)bp->hw_rx_port_stats +
4526                  sizeof(struct rx_port_stats));
4527         bp->hw_rx_port_stats_ext_map = bp->hw_rx_port_stats_map +
4528                 sizeof(struct rx_port_stats);
4529         bp->flags |= BNXT_FLAG_EXT_RX_PORT_STATS;
4530
4531         if (bp->hwrm_spec_code < HWRM_SPEC_CODE_1_9_2 ||
4532             bp->flags & BNXT_FLAG_EXT_STATS_SUPPORTED) {
4533                 bp->hw_tx_port_stats_ext = (void *)
4534                         ((uint8_t *)bp->hw_tx_port_stats +
4535                          sizeof(struct tx_port_stats));
4536                 bp->hw_tx_port_stats_ext_map =
4537                         bp->hw_tx_port_stats_map +
4538                         sizeof(struct tx_port_stats);
4539                 bp->flags |= BNXT_FLAG_EXT_TX_PORT_STATS;
4540         }
4541
4542         return 0;
4543 }
4544
4545 static int bnxt_setup_mac_addr(struct rte_eth_dev *eth_dev)
4546 {
4547         struct bnxt *bp = eth_dev->data->dev_private;
4548         int rc = 0;
4549
4550         eth_dev->data->mac_addrs = rte_zmalloc("bnxt_mac_addr_tbl",
4551                                                RTE_ETHER_ADDR_LEN *
4552                                                bp->max_l2_ctx,
4553                                                0);
4554         if (eth_dev->data->mac_addrs == NULL) {
4555                 PMD_DRV_LOG(ERR, "Failed to alloc MAC addr tbl\n");
4556                 return -ENOMEM;
4557         }
4558
4559         if (bnxt_check_zero_bytes(bp->dflt_mac_addr, RTE_ETHER_ADDR_LEN)) {
4560                 if (BNXT_PF(bp))
4561                         return -EINVAL;
4562
4563                 /* Generate a random MAC address, if none was assigned by PF */
4564                 PMD_DRV_LOG(INFO, "VF MAC address not assigned by Host PF\n");
4565                 bnxt_eth_hw_addr_random(bp->mac_addr);
4566                 PMD_DRV_LOG(INFO,
4567                             "Assign random MAC:%02X:%02X:%02X:%02X:%02X:%02X\n",
4568                             bp->mac_addr[0], bp->mac_addr[1], bp->mac_addr[2],
4569                             bp->mac_addr[3], bp->mac_addr[4], bp->mac_addr[5]);
4570
4571                 rc = bnxt_hwrm_set_mac(bp);
4572                 if (!rc)
4573                         memcpy(&bp->eth_dev->data->mac_addrs[0], bp->mac_addr,
4574                                RTE_ETHER_ADDR_LEN);
4575                 return rc;
4576         }
4577
4578         /* Copy the permanent MAC from the FUNC_QCAPS response */
4579         memcpy(bp->mac_addr, bp->dflt_mac_addr, RTE_ETHER_ADDR_LEN);
4580         memcpy(&eth_dev->data->mac_addrs[0], bp->mac_addr, RTE_ETHER_ADDR_LEN);
4581
4582         return rc;
4583 }
4584
4585 static int bnxt_restore_dflt_mac(struct bnxt *bp)
4586 {
4587         int rc = 0;
4588
4589         /* MAC is already configured in FW */
4590         if (!bnxt_check_zero_bytes(bp->dflt_mac_addr, RTE_ETHER_ADDR_LEN))
4591                 return 0;
4592
4593         /* Restore the old MAC configured */
4594         rc = bnxt_hwrm_set_mac(bp);
4595         if (rc)
4596                 PMD_DRV_LOG(ERR, "Failed to restore MAC address\n");
4597
4598         return rc;
4599 }
4600
4601 static void bnxt_config_vf_req_fwd(struct bnxt *bp)
4602 {
4603         if (!BNXT_PF(bp))
4604                 return;
4605
4606 #define ALLOW_FUNC(x)   \
4607         { \
4608                 uint32_t arg = (x); \
4609                 bp->pf.vf_req_fwd[((arg) >> 5)] &= \
4610                 ~rte_cpu_to_le_32(1 << ((arg) & 0x1f)); \
4611         }
4612
4613         /* Forward all requests if firmware is new enough */
4614         if (((bp->fw_ver >= ((20 << 24) | (6 << 16) | (100 << 8))) &&
4615              (bp->fw_ver < ((20 << 24) | (7 << 16)))) ||
4616             ((bp->fw_ver >= ((20 << 24) | (8 << 16))))) {
4617                 memset(bp->pf.vf_req_fwd, 0xff, sizeof(bp->pf.vf_req_fwd));
4618         } else {
4619                 PMD_DRV_LOG(WARNING,
4620                             "Firmware too old for VF mailbox functionality\n");
4621                 memset(bp->pf.vf_req_fwd, 0, sizeof(bp->pf.vf_req_fwd));
4622         }
4623
4624         /*
4625          * The following are used for driver cleanup. If we disallow these,
4626          * VF drivers can't clean up cleanly.
4627          */
4628         ALLOW_FUNC(HWRM_FUNC_DRV_UNRGTR);
4629         ALLOW_FUNC(HWRM_VNIC_FREE);
4630         ALLOW_FUNC(HWRM_RING_FREE);
4631         ALLOW_FUNC(HWRM_RING_GRP_FREE);
4632         ALLOW_FUNC(HWRM_VNIC_RSS_COS_LB_CTX_FREE);
4633         ALLOW_FUNC(HWRM_CFA_L2_FILTER_FREE);
4634         ALLOW_FUNC(HWRM_STAT_CTX_FREE);
4635         ALLOW_FUNC(HWRM_PORT_PHY_QCFG);
4636         ALLOW_FUNC(HWRM_VNIC_TPA_CFG);
4637 }
4638
4639 static int bnxt_init_fw(struct bnxt *bp)
4640 {
4641         uint16_t mtu;
4642         int rc = 0;
4643
4644         bp->fw_cap = 0;
4645
4646         rc = bnxt_hwrm_ver_get(bp);
4647         if (rc)
4648                 return rc;
4649
4650         rc = bnxt_hwrm_func_reset(bp);
4651         if (rc)
4652                 return -EIO;
4653
4654         rc = bnxt_hwrm_vnic_qcaps(bp);
4655         if (rc)
4656                 return rc;
4657
4658         rc = bnxt_hwrm_queue_qportcfg(bp);
4659         if (rc)
4660                 return rc;
4661
4662         /* Get the MAX capabilities for this function.
4663          * This function also allocates context memory for TQM rings and
4664          * informs the firmware about this allocated backing store memory.
4665          */
4666         rc = bnxt_hwrm_func_qcaps(bp);
4667         if (rc)
4668                 return rc;
4669
4670         rc = bnxt_hwrm_func_qcfg(bp, &mtu);
4671         if (rc)
4672                 return rc;
4673
4674         rc = bnxt_hwrm_cfa_adv_flow_mgmt_qcaps(bp);
4675         if (rc)
4676                 return rc;
4677
4678         /* Get the adapter error recovery support info */
4679         rc = bnxt_hwrm_error_recovery_qcfg(bp);
4680         if (rc)
4681                 bp->fw_cap &= ~BNXT_FW_CAP_ERROR_RECOVERY;
4682
4683         bnxt_hwrm_port_led_qcaps(bp);
4684
4685         return 0;
4686 }
4687
4688 static int
4689 bnxt_init_locks(struct bnxt *bp)
4690 {
4691         int err;
4692
4693         err = pthread_mutex_init(&bp->flow_lock, NULL);
4694         if (err) {
4695                 PMD_DRV_LOG(ERR, "Unable to initialize flow_lock\n");
4696                 return err;
4697         }
4698
4699         err = pthread_mutex_init(&bp->def_cp_lock, NULL);
4700         if (err)
4701                 PMD_DRV_LOG(ERR, "Unable to initialize def_cp_lock\n");
4702         return err;
4703 }
4704
4705 static int bnxt_init_resources(struct bnxt *bp, bool reconfig_dev)
4706 {
4707         int rc;
4708
4709         rc = bnxt_init_fw(bp);
4710         if (rc)
4711                 return rc;
4712
4713         if (!reconfig_dev) {
4714                 rc = bnxt_setup_mac_addr(bp->eth_dev);
4715                 if (rc)
4716                         return rc;
4717         } else {
4718                 rc = bnxt_restore_dflt_mac(bp);
4719                 if (rc)
4720                         return rc;
4721         }
4722
4723         bnxt_config_vf_req_fwd(bp);
4724
4725         rc = bnxt_hwrm_func_driver_register(bp);
4726         if (rc) {
4727                 PMD_DRV_LOG(ERR, "Failed to register driver");
4728                 return -EBUSY;
4729         }
4730
4731         if (BNXT_PF(bp)) {
4732                 if (bp->pdev->max_vfs) {
4733                         rc = bnxt_hwrm_allocate_vfs(bp, bp->pdev->max_vfs);
4734                         if (rc) {
4735                                 PMD_DRV_LOG(ERR, "Failed to allocate VFs\n");
4736                                 return rc;
4737                         }
4738                 } else {
4739                         rc = bnxt_hwrm_allocate_pf_only(bp);
4740                         if (rc) {
4741                                 PMD_DRV_LOG(ERR,
4742                                             "Failed to allocate PF resources");
4743                                 return rc;
4744                         }
4745                 }
4746         }
4747
4748         rc = bnxt_alloc_mem(bp, reconfig_dev);
4749         if (rc)
4750                 return rc;
4751
4752         rc = bnxt_setup_int(bp);
4753         if (rc)
4754                 return rc;
4755
4756         rc = bnxt_request_int(bp);
4757         if (rc)
4758                 return rc;
4759
4760         rc = bnxt_init_locks(bp);
4761         if (rc)
4762                 return rc;
4763
4764         return 0;
4765 }
4766
4767 static int
4768 bnxt_dev_init(struct rte_eth_dev *eth_dev)
4769 {
4770         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
4771         static int version_printed;
4772         struct bnxt *bp;
4773         int rc;
4774
4775         if (version_printed++ == 0)
4776                 PMD_DRV_LOG(INFO, "%s\n", bnxt_version);
4777
4778         eth_dev->dev_ops = &bnxt_dev_ops;
4779         eth_dev->rx_pkt_burst = &bnxt_recv_pkts;
4780         eth_dev->tx_pkt_burst = &bnxt_xmit_pkts;
4781
4782         /*
4783          * For secondary processes, we don't initialise any further
4784          * as primary has already done this work.
4785          */
4786         if (rte_eal_process_type() != RTE_PROC_PRIMARY)
4787                 return 0;
4788
4789         rte_eth_copy_pci_info(eth_dev, pci_dev);
4790
4791         bp = eth_dev->data->dev_private;
4792
4793         bp->dev_stopped = 1;
4794         bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
4795
4796         if (bnxt_vf_pciid(pci_dev->id.device_id))
4797                 bp->flags |= BNXT_FLAG_VF;
4798
4799         if (bnxt_thor_device(pci_dev->id.device_id))
4800                 bp->flags |= BNXT_FLAG_THOR_CHIP;
4801
4802         if (pci_dev->id.device_id == BROADCOM_DEV_ID_58802 ||
4803             pci_dev->id.device_id == BROADCOM_DEV_ID_58804 ||
4804             pci_dev->id.device_id == BROADCOM_DEV_ID_58808 ||
4805             pci_dev->id.device_id == BROADCOM_DEV_ID_58802_VF)
4806                 bp->flags |= BNXT_FLAG_STINGRAY;
4807
4808         rc = bnxt_init_board(eth_dev);
4809         if (rc) {
4810                 PMD_DRV_LOG(ERR,
4811                             "Failed to initialize board rc: %x\n", rc);
4812                 return rc;
4813         }
4814
4815         rc = bnxt_alloc_hwrm_resources(bp);
4816         if (rc) {
4817                 PMD_DRV_LOG(ERR,
4818                             "Failed to allocate hwrm resource rc: %x\n", rc);
4819                 goto error_free;
4820         }
4821         rc = bnxt_init_resources(bp, false);
4822         if (rc)
4823                 goto error_free;
4824
4825         rc = bnxt_alloc_stats_mem(bp);
4826         if (rc)
4827                 goto error_free;
4828
4829         PMD_DRV_LOG(INFO,
4830                     DRV_MODULE_NAME "found at mem %" PRIX64 ", node addr %pM\n",
4831                     pci_dev->mem_resource[0].phys_addr,
4832                     pci_dev->mem_resource[0].addr);
4833
4834         return 0;
4835
4836 error_free:
4837         bnxt_dev_uninit(eth_dev);
4838         return rc;
4839 }
4840
4841 static void
4842 bnxt_uninit_locks(struct bnxt *bp)
4843 {
4844         pthread_mutex_destroy(&bp->flow_lock);
4845         pthread_mutex_destroy(&bp->def_cp_lock);
4846 }
4847
4848 static int
4849 bnxt_uninit_resources(struct bnxt *bp, bool reconfig_dev)
4850 {
4851         int rc;
4852
4853         bnxt_free_int(bp);
4854         bnxt_free_mem(bp, reconfig_dev);
4855         bnxt_hwrm_func_buf_unrgtr(bp);
4856         rc = bnxt_hwrm_func_driver_unregister(bp, 0);
4857         bp->flags &= ~BNXT_FLAG_REGISTERED;
4858         bnxt_free_ctx_mem(bp);
4859         if (!reconfig_dev) {
4860                 bnxt_free_hwrm_resources(bp);
4861
4862                 if (bp->recovery_info != NULL) {
4863                         rte_free(bp->recovery_info);
4864                         bp->recovery_info = NULL;
4865                 }
4866         }
4867
4868         bnxt_uninit_locks(bp);
4869         rte_free(bp->ptp_cfg);
4870         bp->ptp_cfg = NULL;
4871         return rc;
4872 }
4873
4874 static int
4875 bnxt_dev_uninit(struct rte_eth_dev *eth_dev)
4876 {
4877         struct bnxt *bp = eth_dev->data->dev_private;
4878         int rc;
4879
4880         if (rte_eal_process_type() != RTE_PROC_PRIMARY)
4881                 return -EPERM;
4882
4883         PMD_DRV_LOG(DEBUG, "Calling Device uninit\n");
4884
4885         rc = bnxt_uninit_resources(bp, false);
4886
4887         if (bp->tx_mem_zone) {
4888                 rte_memzone_free((const struct rte_memzone *)bp->tx_mem_zone);
4889                 bp->tx_mem_zone = NULL;
4890         }
4891
4892         if (bp->rx_mem_zone) {
4893                 rte_memzone_free((const struct rte_memzone *)bp->rx_mem_zone);
4894                 bp->rx_mem_zone = NULL;
4895         }
4896
4897         if (bp->dev_stopped == 0)
4898                 bnxt_dev_close_op(eth_dev);
4899         if (bp->pf.vf_info)
4900                 rte_free(bp->pf.vf_info);
4901         eth_dev->dev_ops = NULL;
4902         eth_dev->rx_pkt_burst = NULL;
4903         eth_dev->tx_pkt_burst = NULL;
4904
4905         return rc;
4906 }
4907
4908 static int bnxt_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
4909         struct rte_pci_device *pci_dev)
4910 {
4911         return rte_eth_dev_pci_generic_probe(pci_dev, sizeof(struct bnxt),
4912                 bnxt_dev_init);
4913 }
4914
4915 static int bnxt_pci_remove(struct rte_pci_device *pci_dev)
4916 {
4917         if (rte_eal_process_type() == RTE_PROC_PRIMARY)
4918                 return rte_eth_dev_pci_generic_remove(pci_dev,
4919                                 bnxt_dev_uninit);
4920         else
4921                 return rte_eth_dev_pci_generic_remove(pci_dev, NULL);
4922 }
4923
4924 static struct rte_pci_driver bnxt_rte_pmd = {
4925         .id_table = bnxt_pci_id_map,
4926         .drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC,
4927         .probe = bnxt_pci_probe,
4928         .remove = bnxt_pci_remove,
4929 };
4930
4931 static bool
4932 is_device_supported(struct rte_eth_dev *dev, struct rte_pci_driver *drv)
4933 {
4934         if (strcmp(dev->device->driver->name, drv->driver.name))
4935                 return false;
4936
4937         return true;
4938 }
4939
4940 bool is_bnxt_supported(struct rte_eth_dev *dev)
4941 {
4942         return is_device_supported(dev, &bnxt_rte_pmd);
4943 }
4944
4945 RTE_INIT(bnxt_init_log)
4946 {
4947         bnxt_logtype_driver = rte_log_register("pmd.net.bnxt.driver");
4948         if (bnxt_logtype_driver >= 0)
4949                 rte_log_set_level(bnxt_logtype_driver, RTE_LOG_NOTICE);
4950 }
4951
4952 RTE_PMD_REGISTER_PCI(net_bnxt, bnxt_rte_pmd);
4953 RTE_PMD_REGISTER_PCI_TABLE(net_bnxt, bnxt_pci_id_map);
4954 RTE_PMD_REGISTER_KMOD_DEP(net_bnxt, "* igb_uio | uio_pci_generic | vfio-pci");