drivers/net/iavf/iavf_ethdev.c

   1 /* SPDX-License-Identifier: BSD-3-Clause
   2  * Copyright(c) 2017 Intel Corporation
   3  */
   4
   5 #include <sys/queue.h>
   6 #include <stdio.h>
   7 #include <errno.h>
   8 #include <stdint.h>
   9 #include <string.h>
  10 #include <unistd.h>
  11 #include <stdarg.h>
  12 #include <inttypes.h>
  13 #include <rte_byteorder.h>
  14 #include <rte_common.h>
  15
  16 #include <rte_interrupts.h>
  17 #include <rte_debug.h>
  18 #include <rte_pci.h>
  19 #include <rte_atomic.h>
  20 #include <rte_eal.h>
  21 #include <rte_ether.h>
  22 #include <rte_ethdev_driver.h>
  23 #include <rte_ethdev_pci.h>
  24 #include <rte_malloc.h>
  25 #include <rte_memzone.h>
  26 #include <rte_dev.h>
  27
  28 #include "iavf.h"
  29 #include "iavf_rxtx.h"
  30 #include "iavf_generic_flow.h"
  31
  32 static int iavf_dev_configure(struct rte_eth_dev *dev);
  33 static int iavf_dev_start(struct rte_eth_dev *dev);
  34 static int iavf_dev_stop(struct rte_eth_dev *dev);
  35 static int iavf_dev_close(struct rte_eth_dev *dev);
  36 static int iavf_dev_reset(struct rte_eth_dev *dev);
  37 static int iavf_dev_info_get(struct rte_eth_dev *dev,
  38                              struct rte_eth_dev_info *dev_info);
  39 static const uint32_t *iavf_dev_supported_ptypes_get(struct rte_eth_dev *dev);
  40 static int iavf_dev_stats_get(struct rte_eth_dev *dev,
  41                              struct rte_eth_stats *stats);
  42 static int iavf_dev_stats_reset(struct rte_eth_dev *dev);
  43 static int iavf_dev_xstats_get(struct rte_eth_dev *dev,
  44                                  struct rte_eth_xstat *xstats, unsigned int n);
  45 static int iavf_dev_xstats_get_names(struct rte_eth_dev *dev,
  46                                        struct rte_eth_xstat_name *xstats_names,
  47                                        unsigned int limit);
  48 static int iavf_dev_promiscuous_enable(struct rte_eth_dev *dev);
  49 static int iavf_dev_promiscuous_disable(struct rte_eth_dev *dev);
  50 static int iavf_dev_allmulticast_enable(struct rte_eth_dev *dev);
  51 static int iavf_dev_allmulticast_disable(struct rte_eth_dev *dev);
  52 static int iavf_dev_add_mac_addr(struct rte_eth_dev *dev,
  53                                 struct rte_ether_addr *addr,
  54                                 uint32_t index,
  55                                 uint32_t pool);
  56 static void iavf_dev_del_mac_addr(struct rte_eth_dev *dev, uint32_t index);
  57 static int iavf_dev_vlan_filter_set(struct rte_eth_dev *dev,
  58                                    uint16_t vlan_id, int on);
  59 static int iavf_dev_vlan_offload_set(struct rte_eth_dev *dev, int mask);
  60 static int iavf_dev_rss_reta_update(struct rte_eth_dev *dev,
  61                                    struct rte_eth_rss_reta_entry64 *reta_conf,
  62                                    uint16_t reta_size);
  63 static int iavf_dev_rss_reta_query(struct rte_eth_dev *dev,
  64                                   struct rte_eth_rss_reta_entry64 *reta_conf,
  65                                   uint16_t reta_size);
  66 static int iavf_dev_rss_hash_update(struct rte_eth_dev *dev,
  67                                    struct rte_eth_rss_conf *rss_conf);
  68 static int iavf_dev_rss_hash_conf_get(struct rte_eth_dev *dev,
  69                                      struct rte_eth_rss_conf *rss_conf);
  70 static int iavf_dev_mtu_set(struct rte_eth_dev *dev, uint16_t mtu);
  71 static int iavf_dev_set_default_mac_addr(struct rte_eth_dev *dev,
  72                                          struct rte_ether_addr *mac_addr);
  73 static int iavf_dev_rx_queue_intr_enable(struct rte_eth_dev *dev,
  74                                         uint16_t queue_id);
  75 static int iavf_dev_rx_queue_intr_disable(struct rte_eth_dev *dev,
  76                                          uint16_t queue_id);
  77 static int iavf_dev_filter_ctrl(struct rte_eth_dev *dev,
  78                      enum rte_filter_type filter_type,
  79                      enum rte_filter_op filter_op,
  80                      void *arg);
  81 static int iavf_set_mc_addr_list(struct rte_eth_dev *dev,
  82                         struct rte_ether_addr *mc_addrs,
  83                         uint32_t mc_addrs_num);
  84
  85 static const struct rte_pci_id pci_id_iavf_map[] = {
  86         { RTE_PCI_DEVICE(IAVF_INTEL_VENDOR_ID, IAVF_DEV_ID_ADAPTIVE_VF) },
  87         { .vendor_id = 0, /* sentinel */ },
  88 };
  89
  90 struct rte_iavf_xstats_name_off {
  91         char name[RTE_ETH_XSTATS_NAME_SIZE];
  92         unsigned int offset;
  93 };
  94
  95 static const struct rte_iavf_xstats_name_off rte_iavf_stats_strings[] = {
  96         {"rx_bytes", offsetof(struct iavf_eth_stats, rx_bytes)},
  97         {"rx_unicast_packets", offsetof(struct iavf_eth_stats, rx_unicast)},
  98         {"rx_multicast_packets", offsetof(struct iavf_eth_stats, rx_multicast)},
  99         {"rx_broadcast_packets", offsetof(struct iavf_eth_stats, rx_broadcast)},
 100         {"rx_dropped_packets", offsetof(struct iavf_eth_stats, rx_discards)},
 101         {"rx_unknown_protocol_packets", offsetof(struct iavf_eth_stats,
 102                 rx_unknown_protocol)},
 103         {"tx_bytes", offsetof(struct iavf_eth_stats, tx_bytes)},
 104         {"tx_unicast_packets", offsetof(struct iavf_eth_stats, tx_unicast)},
 105         {"tx_multicast_packets", offsetof(struct iavf_eth_stats, tx_multicast)},
 106         {"tx_broadcast_packets", offsetof(struct iavf_eth_stats, tx_broadcast)},
 107         {"tx_dropped_packets", offsetof(struct iavf_eth_stats, tx_discards)},
 108         {"tx_error_packets", offsetof(struct iavf_eth_stats, tx_errors)},
 109 };
 110
 111 #define IAVF_NB_XSTATS (sizeof(rte_iavf_stats_strings) / \
 112                 sizeof(rte_iavf_stats_strings[0]))
 113
 114 static const struct eth_dev_ops iavf_eth_dev_ops = {
 115         .dev_configure              = iavf_dev_configure,
 116         .dev_start                  = iavf_dev_start,
 117         .dev_stop                   = iavf_dev_stop,
 118         .dev_close                  = iavf_dev_close,
 119         .dev_reset                  = iavf_dev_reset,
 120         .dev_infos_get              = iavf_dev_info_get,
 121         .dev_supported_ptypes_get   = iavf_dev_supported_ptypes_get,
 122         .link_update                = iavf_dev_link_update,
 123         .stats_get                  = iavf_dev_stats_get,
 124         .stats_reset                = iavf_dev_stats_reset,
 125         .xstats_get                 = iavf_dev_xstats_get,
 126         .xstats_get_names           = iavf_dev_xstats_get_names,
 127         .xstats_reset               = iavf_dev_stats_reset,
 128         .promiscuous_enable         = iavf_dev_promiscuous_enable,
 129         .promiscuous_disable        = iavf_dev_promiscuous_disable,
 130         .allmulticast_enable        = iavf_dev_allmulticast_enable,
 131         .allmulticast_disable       = iavf_dev_allmulticast_disable,
 132         .mac_addr_add               = iavf_dev_add_mac_addr,
 133         .mac_addr_remove            = iavf_dev_del_mac_addr,
 134         .set_mc_addr_list                       = iavf_set_mc_addr_list,
 135         .vlan_filter_set            = iavf_dev_vlan_filter_set,
 136         .vlan_offload_set           = iavf_dev_vlan_offload_set,
 137         .rx_queue_start             = iavf_dev_rx_queue_start,
 138         .rx_queue_stop              = iavf_dev_rx_queue_stop,
 139         .tx_queue_start             = iavf_dev_tx_queue_start,
 140         .tx_queue_stop              = iavf_dev_tx_queue_stop,
 141         .rx_queue_setup             = iavf_dev_rx_queue_setup,
 142         .rx_queue_release           = iavf_dev_rx_queue_release,
 143         .tx_queue_setup             = iavf_dev_tx_queue_setup,
 144         .tx_queue_release           = iavf_dev_tx_queue_release,
 145         .mac_addr_set               = iavf_dev_set_default_mac_addr,
 146         .reta_update                = iavf_dev_rss_reta_update,
 147         .reta_query                 = iavf_dev_rss_reta_query,
 148         .rss_hash_update            = iavf_dev_rss_hash_update,
 149         .rss_hash_conf_get          = iavf_dev_rss_hash_conf_get,
 150         .rxq_info_get               = iavf_dev_rxq_info_get,
 151         .txq_info_get               = iavf_dev_txq_info_get,
 152         .mtu_set                    = iavf_dev_mtu_set,
 153         .rx_queue_intr_enable       = iavf_dev_rx_queue_intr_enable,
 154         .rx_queue_intr_disable      = iavf_dev_rx_queue_intr_disable,
 155         .filter_ctrl                = iavf_dev_filter_ctrl,
 156         .tx_done_cleanup            = iavf_dev_tx_done_cleanup,
 157 };
 158
 159 static int
 160 iavf_set_mc_addr_list(struct rte_eth_dev *dev,
 161                         struct rte_ether_addr *mc_addrs,
 162                         uint32_t mc_addrs_num)
 163 {
 164         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
 165         struct iavf_adapter *adapter =
 166                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 167         int err, ret;
 168
 169         if (mc_addrs_num > IAVF_NUM_MACADDR_MAX) {
 170                 PMD_DRV_LOG(ERR,
 171                             "can't add more than a limited number (%u) of addresses.",
 172                             (uint32_t)IAVF_NUM_MACADDR_MAX);
 173                 return -EINVAL;
 174         }
 175
 176         /* flush previous addresses */
 177         err = iavf_add_del_mc_addr_list(adapter, vf->mc_addrs, vf->mc_addrs_num,
 178                                         false);
 179         if (err)
 180                 return err;
 181
 182         /* add new ones */
 183         err = iavf_add_del_mc_addr_list(adapter, mc_addrs, mc_addrs_num, true);
 184
 185         if (err) {
 186                 /* if adding mac address list fails, should add the previous
 187                  * addresses back.
 188                  */
 189                 ret = iavf_add_del_mc_addr_list(adapter, vf->mc_addrs,
 190                                                 vf->mc_addrs_num, true);
 191                 if (ret)
 192                         return ret;
 193         } else {
 194                 vf->mc_addrs_num = mc_addrs_num;
 195                 memcpy(vf->mc_addrs,
 196                        mc_addrs, mc_addrs_num * sizeof(*mc_addrs));
 197         }
 198
 199         return err;
 200 }
 201
 202 static int
 203 iavf_init_rss(struct iavf_adapter *adapter)
 204 {
 205         struct iavf_info *vf =  IAVF_DEV_PRIVATE_TO_VF(adapter);
 206         struct rte_eth_rss_conf *rss_conf;
 207         uint16_t i, j, nb_q;
 208         int ret;
 209
 210         rss_conf = &adapter->eth_dev->data->dev_conf.rx_adv_conf.rss_conf;
 211         nb_q = RTE_MIN(adapter->eth_dev->data->nb_rx_queues,
 212                        vf->max_rss_qregion);
 213
 214         if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF)) {
 215                 PMD_DRV_LOG(DEBUG, "RSS is not supported");
 216                 return -ENOTSUP;
 217         }
 218         if (adapter->eth_dev->data->dev_conf.rxmode.mq_mode != ETH_MQ_RX_RSS) {
 219                 PMD_DRV_LOG(WARNING, "RSS is enabled by PF by default");
 220                 /* set all lut items to default queue */
 221                 for (i = 0; i < vf->vf_res->rss_lut_size; i++)
 222                         vf->rss_lut[i] = 0;
 223                 ret = iavf_configure_rss_lut(adapter);
 224                 return ret;
 225         }
 226
 227         /* In IAVF, RSS enablement is set by PF driver. It is not supported
 228          * to set based on rss_conf->rss_hf.
 229          */
 230
 231         /* configure RSS key */
 232         if (!rss_conf->rss_key) {
 233                 /* Calculate the default hash key */
 234                 for (i = 0; i <= vf->vf_res->rss_key_size; i++)
 235                         vf->rss_key[i] = (uint8_t)rte_rand();
 236         } else
 237                 rte_memcpy(vf->rss_key, rss_conf->rss_key,
 238                            RTE_MIN(rss_conf->rss_key_len,
 239                                    vf->vf_res->rss_key_size));
 240
 241         /* init RSS LUT table */
 242         for (i = 0, j = 0; i < vf->vf_res->rss_lut_size; i++, j++) {
 243                 if (j >= nb_q)
 244                         j = 0;
 245                 vf->rss_lut[i] = j;
 246         }
 247         /* send virtchnnl ops to configure rss*/
 248         ret = iavf_configure_rss_lut(adapter);
 249         if (ret)
 250                 return ret;
 251         ret = iavf_configure_rss_key(adapter);
 252         if (ret)
 253                 return ret;
 254
 255         return 0;
 256 }
 257
 258 static int
 259 iavf_queues_req_reset(struct rte_eth_dev *dev, uint16_t num)
 260 {
 261         struct iavf_adapter *ad =
 262                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 263         struct iavf_info *vf =  IAVF_DEV_PRIVATE_TO_VF(ad);
 264         int ret;
 265
 266         ret = iavf_request_queues(ad, num);
 267         if (ret) {
 268                 PMD_DRV_LOG(ERR, "request queues from PF failed");
 269                 return ret;
 270         }
 271         PMD_DRV_LOG(INFO, "change queue pairs from %u to %u",
 272                         vf->vsi_res->num_queue_pairs, num);
 273
 274         ret = iavf_dev_reset(dev);
 275         if (ret) {
 276                 PMD_DRV_LOG(ERR, "vf reset failed");
 277                 return ret;
 278         }
 279
 280         return 0;
 281 }
 282
 283 static int
 284 iavf_dev_configure(struct rte_eth_dev *dev)
 285 {
 286         struct iavf_adapter *ad =
 287                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 288         struct iavf_info *vf =  IAVF_DEV_PRIVATE_TO_VF(ad);
 289         struct rte_eth_conf *dev_conf = &dev->data->dev_conf;
 290         uint16_t num_queue_pairs = RTE_MAX(dev->data->nb_rx_queues,
 291                 dev->data->nb_tx_queues);
 292         int ret;
 293
 294         ad->rx_bulk_alloc_allowed = true;
 295         /* Initialize to TRUE. If any of Rx queues doesn't meet the
 296          * vector Rx/Tx preconditions, it will be reset.
 297          */
 298         ad->rx_vec_allowed = true;
 299         ad->tx_vec_allowed = true;
 300
 301         if (dev->data->dev_conf.rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG)
 302                 dev->data->dev_conf.rxmode.offloads |= DEV_RX_OFFLOAD_RSS_HASH;
 303
 304         /* Large VF setting */
 305         if (num_queue_pairs > IAVF_MAX_NUM_QUEUES_DFLT) {
 306                 if (!(vf->vf_res->vf_cap_flags &
 307                                 VIRTCHNL_VF_LARGE_NUM_QPAIRS)) {
 308                         PMD_DRV_LOG(ERR, "large VF is not supported");
 309                         return -1;
 310                 }
 311
 312                 if (num_queue_pairs > IAVF_MAX_NUM_QUEUES_LV) {
 313                         PMD_DRV_LOG(ERR, "queue pairs number cannot be larger than %u",
 314                                 IAVF_MAX_NUM_QUEUES_LV);
 315                         return -1;
 316                 }
 317
 318                 ret = iavf_queues_req_reset(dev, num_queue_pairs);
 319                 if (ret)
 320                         return ret;
 321
 322                 ret = iavf_get_max_rss_queue_region(ad);
 323                 if (ret) {
 324                         PMD_INIT_LOG(ERR, "get max rss queue region failed");
 325                         return ret;
 326                 }
 327
 328                 vf->lv_enabled = true;
 329         } else {
 330                 /* Check if large VF is already enabled. If so, disable and
 331                  * release redundant queue resource.
 332                  */
 333                 if (vf->lv_enabled) {
 334                         ret = iavf_queues_req_reset(dev, num_queue_pairs);
 335                         if (ret)
 336                                 return ret;
 337
 338                         vf->lv_enabled = false;
 339                 }
 340                 /* if large VF is not required, use default rss queue region */
 341                 vf->max_rss_qregion = IAVF_MAX_NUM_QUEUES_DFLT;
 342         }
 343
 344         /* Vlan stripping setting */
 345         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_VLAN) {
 346                 if (dev_conf->rxmode.offloads & DEV_RX_OFFLOAD_VLAN_STRIP)
 347                         iavf_enable_vlan_strip(ad);
 348                 else
 349                         iavf_disable_vlan_strip(ad);
 350         }
 351
 352         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF) {
 353                 if (iavf_init_rss(ad) != 0) {
 354                         PMD_DRV_LOG(ERR, "configure rss failed");
 355                         return -1;
 356                 }
 357         }
 358         return 0;
 359 }
 360
 361 static int
 362 iavf_init_rxq(struct rte_eth_dev *dev, struct iavf_rx_queue *rxq)
 363 {
 364         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(dev->data->dev_private);
 365         struct rte_eth_dev_data *dev_data = dev->data;
 366         uint16_t buf_size, max_pkt_len, len;
 367
 368         buf_size = rte_pktmbuf_data_room_size(rxq->mp) - RTE_PKTMBUF_HEADROOM;
 369
 370         /* Calculate the maximum packet length allowed */
 371         len = rxq->rx_buf_len * IAVF_MAX_CHAINED_RX_BUFFERS;
 372         max_pkt_len = RTE_MIN(len, dev->data->dev_conf.rxmode.max_rx_pkt_len);
 373
 374         /* Check if the jumbo frame and maximum packet length are set
 375          * correctly.
 376          */
 377         if (dev->data->dev_conf.rxmode.offloads & DEV_RX_OFFLOAD_JUMBO_FRAME) {
 378                 if (max_pkt_len <= RTE_ETHER_MAX_LEN ||
 379                     max_pkt_len > IAVF_FRAME_SIZE_MAX) {
 380                         PMD_DRV_LOG(ERR, "maximum packet length must be "
 381                                     "larger than %u and smaller than %u, "
 382                                     "as jumbo frame is enabled",
 383                                     (uint32_t)RTE_ETHER_MAX_LEN,
 384                                     (uint32_t)IAVF_FRAME_SIZE_MAX);
 385                         return -EINVAL;
 386                 }
 387         } else {
 388                 if (max_pkt_len < RTE_ETHER_MIN_LEN ||
 389                     max_pkt_len > RTE_ETHER_MAX_LEN) {
 390                         PMD_DRV_LOG(ERR, "maximum packet length must be "
 391                                     "larger than %u and smaller than %u, "
 392                                     "as jumbo frame is disabled",
 393                                     (uint32_t)RTE_ETHER_MIN_LEN,
 394                                     (uint32_t)RTE_ETHER_MAX_LEN);
 395                         return -EINVAL;
 396                 }
 397         }
 398
 399         rxq->max_pkt_len = max_pkt_len;
 400         if ((dev_data->dev_conf.rxmode.offloads & DEV_RX_OFFLOAD_SCATTER) ||
 401             rxq->max_pkt_len > buf_size) {
 402                 dev_data->scattered_rx = 1;
 403         }
 404         IAVF_PCI_REG_WRITE(rxq->qrx_tail, rxq->nb_rx_desc - 1);
 405         IAVF_WRITE_FLUSH(hw);
 406
 407         return 0;
 408 }
 409
 410 static int
 411 iavf_init_queues(struct rte_eth_dev *dev)
 412 {
 413         struct iavf_rx_queue **rxq =
 414                 (struct iavf_rx_queue **)dev->data->rx_queues;
 415         int i, ret = IAVF_SUCCESS;
 416
 417         for (i = 0; i < dev->data->nb_rx_queues; i++) {
 418                 if (!rxq[i] || !rxq[i]->q_set)
 419                         continue;
 420                 ret = iavf_init_rxq(dev, rxq[i]);
 421                 if (ret != IAVF_SUCCESS)
 422                         break;
 423         }
 424         /* set rx/tx function to vector/scatter/single-segment
 425          * according to parameters
 426          */
 427         iavf_set_rx_function(dev);
 428         iavf_set_tx_function(dev);
 429
 430         return ret;
 431 }
 432
 433 static int iavf_config_rx_queues_irqs(struct rte_eth_dev *dev,
 434                                      struct rte_intr_handle *intr_handle)
 435 {
 436         struct iavf_adapter *adapter =
 437                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 438         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
 439         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(adapter);
 440         struct iavf_qv_map *qv_map;
 441         uint16_t interval, i;
 442         int vec;
 443
 444         if (rte_intr_cap_multiple(intr_handle) &&
 445             dev->data->dev_conf.intr_conf.rxq) {
 446                 if (rte_intr_efd_enable(intr_handle, dev->data->nb_rx_queues))
 447                         return -1;
 448         }
 449
 450         if (rte_intr_dp_is_en(intr_handle) && !intr_handle->intr_vec) {
 451                 intr_handle->intr_vec =
 452                         rte_zmalloc("intr_vec",
 453                                     dev->data->nb_rx_queues * sizeof(int), 0);
 454                 if (!intr_handle->intr_vec) {
 455                         PMD_DRV_LOG(ERR, "Failed to allocate %d rx intr_vec",
 456                                     dev->data->nb_rx_queues);
 457                         return -1;
 458                 }
 459         }
 460
 461         qv_map = rte_zmalloc("qv_map",
 462                 dev->data->nb_rx_queues * sizeof(struct iavf_qv_map), 0);
 463         if (!qv_map) {
 464                 PMD_DRV_LOG(ERR, "Failed to allocate %d queue-vector map",
 465                                 dev->data->nb_rx_queues);
 466                 return -1;
 467         }
 468
 469         if (!dev->data->dev_conf.intr_conf.rxq ||
 470             !rte_intr_dp_is_en(intr_handle)) {
 471                 /* Rx interrupt disabled, Map interrupt only for writeback */
 472                 vf->nb_msix = 1;
 473                 if (vf->vf_res->vf_cap_flags &
 474                     VIRTCHNL_VF_OFFLOAD_WB_ON_ITR) {
 475                         /* If WB_ON_ITR supports, enable it */
 476                         vf->msix_base = IAVF_RX_VEC_START;
 477                         /* Set the ITR for index zero, to 2us to make sure that
 478                          * we leave time for aggregation to occur, but don't
 479                          * increase latency dramatically.
 480                          */
 481                         IAVF_WRITE_REG(hw,
 482                                        IAVF_VFINT_DYN_CTLN1(vf->msix_base - 1),
 483                                        (0 << IAVF_VFINT_DYN_CTLN1_ITR_INDX_SHIFT) |
 484                                        IAVF_VFINT_DYN_CTLN1_WB_ON_ITR_MASK |
 485                                        (2UL << IAVF_VFINT_DYN_CTLN1_INTERVAL_SHIFT));
 486                         /* debug - check for success! the return value
 487                          * should be 2, offset is 0x2800
 488                          */
 489                         /* IAVF_READ_REG(hw, IAVF_VFINT_ITRN1(0, 0)); */
 490                 } else {
 491                         /* If no WB_ON_ITR offload flags, need to set
 492                          * interrupt for descriptor write back.
 493                          */
 494                         vf->msix_base = IAVF_MISC_VEC_ID;
 495
 496                         /* set ITR to max */
 497                         interval = iavf_calc_itr_interval(
 498                                         IAVF_QUEUE_ITR_INTERVAL_MAX);
 499                         IAVF_WRITE_REG(hw, IAVF_VFINT_DYN_CTL01,
 500                                        IAVF_VFINT_DYN_CTL01_INTENA_MASK |
 501                                        (IAVF_ITR_INDEX_DEFAULT <<
 502                                         IAVF_VFINT_DYN_CTL01_ITR_INDX_SHIFT) |
 503                                        (interval <<
 504                                         IAVF_VFINT_DYN_CTL01_INTERVAL_SHIFT));
 505                 }
 506                 IAVF_WRITE_FLUSH(hw);
 507                 /* map all queues to the same interrupt */
 508                 for (i = 0; i < dev->data->nb_rx_queues; i++) {
 509                         qv_map[i].queue_id = i;
 510                         qv_map[i].vector_id = vf->msix_base;
 511                 }
 512                 vf->qv_map = qv_map;
 513         } else {
 514                 if (!rte_intr_allow_others(intr_handle)) {
 515                         vf->nb_msix = 1;
 516                         vf->msix_base = IAVF_MISC_VEC_ID;
 517                         for (i = 0; i < dev->data->nb_rx_queues; i++) {
 518                                 qv_map[i].queue_id = i;
 519                                 qv_map[i].vector_id = vf->msix_base;
 520                                 intr_handle->intr_vec[i] = IAVF_MISC_VEC_ID;
 521                         }
 522                         vf->qv_map = qv_map;
 523                         PMD_DRV_LOG(DEBUG,
 524                                     "vector %u are mapping to all Rx queues",
 525                                     vf->msix_base);
 526                 } else {
 527                         /* If Rx interrupt is reuquired, and we can use
 528                          * multi interrupts, then the vec is from 1
 529                          */
 530                         vf->nb_msix = RTE_MIN(vf->vf_res->max_vectors,
 531                                               intr_handle->nb_efd);
 532                         vf->msix_base = IAVF_RX_VEC_START;
 533                         vec = IAVF_RX_VEC_START;
 534                         for (i = 0; i < dev->data->nb_rx_queues; i++) {
 535                                 qv_map[i].queue_id = i;
 536                                 qv_map[i].vector_id = vec;
 537                                 intr_handle->intr_vec[i] = vec++;
 538                                 if (vec >= vf->nb_msix)
 539                                         vec = IAVF_RX_VEC_START;
 540                         }
 541                         vf->qv_map = qv_map;
 542                         PMD_DRV_LOG(DEBUG,
 543                                     "%u vectors are mapping to %u Rx queues",
 544                                     vf->nb_msix, dev->data->nb_rx_queues);
 545                 }
 546         }
 547
 548         if (!vf->lv_enabled) {
 549                 if (iavf_config_irq_map(adapter)) {
 550                         PMD_DRV_LOG(ERR, "config interrupt mapping failed");
 551                         return -1;
 552                 }
 553         } else {
 554                 uint16_t num_qv_maps = dev->data->nb_rx_queues;
 555                 uint16_t index = 0;
 556
 557                 while (num_qv_maps > IAVF_IRQ_MAP_NUM_PER_BUF) {
 558                         if (iavf_config_irq_map_lv(adapter,
 559                                         IAVF_IRQ_MAP_NUM_PER_BUF, index)) {
 560                                 PMD_DRV_LOG(ERR, "config interrupt mapping for large VF failed");
 561                                 return -1;
 562                         }
 563                         num_qv_maps -= IAVF_IRQ_MAP_NUM_PER_BUF;
 564                         index += IAVF_IRQ_MAP_NUM_PER_BUF;
 565                 }
 566
 567                 if (iavf_config_irq_map_lv(adapter, num_qv_maps, index)) {
 568                         PMD_DRV_LOG(ERR, "config interrupt mapping for large VF failed");
 569                         return -1;
 570                 }
 571         }
 572         return 0;
 573 }
 574
 575 static int
 576 iavf_start_queues(struct rte_eth_dev *dev)
 577 {
 578         struct iavf_rx_queue *rxq;
 579         struct iavf_tx_queue *txq;
 580         int i;
 581
 582         for (i = 0; i < dev->data->nb_tx_queues; i++) {
 583                 txq = dev->data->tx_queues[i];
 584                 if (txq->tx_deferred_start)
 585                         continue;
 586                 if (iavf_dev_tx_queue_start(dev, i) != 0) {
 587                         PMD_DRV_LOG(ERR, "Fail to start queue %u", i);
 588                         return -1;
 589                 }
 590         }
 591
 592         for (i = 0; i < dev->data->nb_rx_queues; i++) {
 593                 rxq = dev->data->rx_queues[i];
 594                 if (rxq->rx_deferred_start)
 595                         continue;
 596                 if (iavf_dev_rx_queue_start(dev, i) != 0) {
 597                         PMD_DRV_LOG(ERR, "Fail to start queue %u", i);
 598                         return -1;
 599                 }
 600         }
 601
 602         return 0;
 603 }
 604
 605 static int
 606 iavf_dev_start(struct rte_eth_dev *dev)
 607 {
 608         struct iavf_adapter *adapter =
 609                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 610         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
 611         struct rte_intr_handle *intr_handle = dev->intr_handle;
 612         uint16_t num_queue_pairs;
 613         uint16_t index = 0;
 614
 615         PMD_INIT_FUNC_TRACE();
 616
 617         adapter->stopped = 0;
 618
 619         vf->max_pkt_len = dev->data->dev_conf.rxmode.max_rx_pkt_len;
 620         vf->num_queue_pairs = RTE_MAX(dev->data->nb_rx_queues,
 621                                       dev->data->nb_tx_queues);
 622         num_queue_pairs = vf->num_queue_pairs;
 623
 624         if (iavf_init_queues(dev) != 0) {
 625                 PMD_DRV_LOG(ERR, "failed to do Queue init");
 626                 return -1;
 627         }
 628
 629         /* If needed, send configure queues msg multiple times to make the
 630          * adminq buffer length smaller than the 4K limitation.
 631          */
 632         while (num_queue_pairs > IAVF_CFG_Q_NUM_PER_BUF) {
 633                 if (iavf_configure_queues(adapter,
 634                                 IAVF_CFG_Q_NUM_PER_BUF, index) != 0) {
 635                         PMD_DRV_LOG(ERR, "configure queues failed");
 636                         goto err_queue;
 637                 }
 638                 num_queue_pairs -= IAVF_CFG_Q_NUM_PER_BUF;
 639                 index += IAVF_CFG_Q_NUM_PER_BUF;
 640         }
 641
 642         if (iavf_configure_queues(adapter, num_queue_pairs, index) != 0) {
 643                 PMD_DRV_LOG(ERR, "configure queues failed");
 644                 goto err_queue;
 645         }
 646
 647         if (iavf_config_rx_queues_irqs(dev, intr_handle) != 0) {
 648                 PMD_DRV_LOG(ERR, "configure irq failed");
 649                 goto err_queue;
 650         }
 651         /* re-enable intr again, because efd assign may change */
 652         if (dev->data->dev_conf.intr_conf.rxq != 0) {
 653                 rte_intr_disable(intr_handle);
 654                 rte_intr_enable(intr_handle);
 655         }
 656
 657         /* Set all mac addrs */
 658         iavf_add_del_all_mac_addr(adapter, true);
 659
 660         /* Set all multicast addresses */
 661         iavf_add_del_mc_addr_list(adapter, vf->mc_addrs, vf->mc_addrs_num,
 662                                   true);
 663
 664         if (iavf_start_queues(dev) != 0) {
 665                 PMD_DRV_LOG(ERR, "enable queues failed");
 666                 goto err_mac;
 667         }
 668
 669         return 0;
 670
 671 err_mac:
 672         iavf_add_del_all_mac_addr(adapter, false);
 673 err_queue:
 674         return -1;
 675 }
 676
 677 static int
 678 iavf_dev_stop(struct rte_eth_dev *dev)
 679 {
 680         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
 681         struct iavf_adapter *adapter =
 682                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 683         struct rte_intr_handle *intr_handle = dev->intr_handle;
 684
 685         PMD_INIT_FUNC_TRACE();
 686
 687         if (adapter->stopped == 1)
 688                 return 0;
 689
 690         iavf_stop_queues(dev);
 691
 692         /* Disable the interrupt for Rx */
 693         rte_intr_efd_disable(intr_handle);
 694         /* Rx interrupt vector mapping free */
 695         if (intr_handle->intr_vec) {
 696                 rte_free(intr_handle->intr_vec);
 697                 intr_handle->intr_vec = NULL;
 698         }
 699
 700         /* remove all mac addrs */
 701         iavf_add_del_all_mac_addr(adapter, false);
 702
 703         /* remove all multicast addresses */
 704         iavf_add_del_mc_addr_list(adapter, vf->mc_addrs, vf->mc_addrs_num,
 705                                   false);
 706
 707         adapter->stopped = 1;
 708         dev->data->dev_started = 0;
 709
 710         return 0;
 711 }
 712
 713 static int
 714 iavf_dev_info_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
 715 {
 716         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
 717
 718         dev_info->max_rx_queues = IAVF_MAX_NUM_QUEUES_LV;
 719         dev_info->max_tx_queues = IAVF_MAX_NUM_QUEUES_LV;
 720         dev_info->min_rx_bufsize = IAVF_BUF_SIZE_MIN;
 721         dev_info->max_rx_pktlen = IAVF_FRAME_SIZE_MAX;
 722         dev_info->max_mtu = dev_info->max_rx_pktlen - IAVF_ETH_OVERHEAD;
 723         dev_info->min_mtu = RTE_ETHER_MIN_MTU;
 724         dev_info->hash_key_size = vf->vf_res->rss_key_size;
 725         dev_info->reta_size = vf->vf_res->rss_lut_size;
 726         dev_info->flow_type_rss_offloads = IAVF_RSS_OFFLOAD_ALL;
 727         dev_info->max_mac_addrs = IAVF_NUM_MACADDR_MAX;
 728         dev_info->rx_offload_capa =
 729                 DEV_RX_OFFLOAD_VLAN_STRIP |
 730                 DEV_RX_OFFLOAD_QINQ_STRIP |
 731                 DEV_RX_OFFLOAD_IPV4_CKSUM |
 732                 DEV_RX_OFFLOAD_UDP_CKSUM |
 733                 DEV_RX_OFFLOAD_TCP_CKSUM |
 734                 DEV_RX_OFFLOAD_OUTER_IPV4_CKSUM |
 735                 DEV_RX_OFFLOAD_SCATTER |
 736                 DEV_RX_OFFLOAD_JUMBO_FRAME |
 737                 DEV_RX_OFFLOAD_VLAN_FILTER |
 738                 DEV_RX_OFFLOAD_RSS_HASH;
 739         dev_info->tx_offload_capa =
 740                 DEV_TX_OFFLOAD_VLAN_INSERT |
 741                 DEV_TX_OFFLOAD_QINQ_INSERT |
 742                 DEV_TX_OFFLOAD_IPV4_CKSUM |
 743                 DEV_TX_OFFLOAD_UDP_CKSUM |
 744                 DEV_TX_OFFLOAD_TCP_CKSUM |
 745                 DEV_TX_OFFLOAD_SCTP_CKSUM |
 746                 DEV_TX_OFFLOAD_OUTER_IPV4_CKSUM |
 747                 DEV_TX_OFFLOAD_TCP_TSO |
 748                 DEV_TX_OFFLOAD_VXLAN_TNL_TSO |
 749                 DEV_TX_OFFLOAD_GRE_TNL_TSO |
 750                 DEV_TX_OFFLOAD_IPIP_TNL_TSO |
 751                 DEV_TX_OFFLOAD_GENEVE_TNL_TSO |
 752                 DEV_TX_OFFLOAD_MULTI_SEGS;
 753
 754         dev_info->default_rxconf = (struct rte_eth_rxconf) {
 755                 .rx_free_thresh = IAVF_DEFAULT_RX_FREE_THRESH,
 756                 .rx_drop_en = 0,
 757                 .offloads = 0,
 758         };
 759
 760         dev_info->default_txconf = (struct rte_eth_txconf) {
 761                 .tx_free_thresh = IAVF_DEFAULT_TX_FREE_THRESH,
 762                 .tx_rs_thresh = IAVF_DEFAULT_TX_RS_THRESH,
 763                 .offloads = 0,
 764         };
 765
 766         dev_info->rx_desc_lim = (struct rte_eth_desc_lim) {
 767                 .nb_max = IAVF_MAX_RING_DESC,
 768                 .nb_min = IAVF_MIN_RING_DESC,
 769                 .nb_align = IAVF_ALIGN_RING_DESC,
 770         };
 771
 772         dev_info->tx_desc_lim = (struct rte_eth_desc_lim) {
 773                 .nb_max = IAVF_MAX_RING_DESC,
 774                 .nb_min = IAVF_MIN_RING_DESC,
 775                 .nb_align = IAVF_ALIGN_RING_DESC,
 776         };
 777
 778         return 0;
 779 }
 780
 781 static const uint32_t *
 782 iavf_dev_supported_ptypes_get(struct rte_eth_dev *dev __rte_unused)
 783 {
 784         static const uint32_t ptypes[] = {
 785                 RTE_PTYPE_L2_ETHER,
 786                 RTE_PTYPE_L3_IPV4_EXT_UNKNOWN,
 787                 RTE_PTYPE_L4_FRAG,
 788                 RTE_PTYPE_L4_ICMP,
 789                 RTE_PTYPE_L4_NONFRAG,
 790                 RTE_PTYPE_L4_SCTP,
 791                 RTE_PTYPE_L4_TCP,
 792                 RTE_PTYPE_L4_UDP,
 793                 RTE_PTYPE_UNKNOWN
 794         };
 795         return ptypes;
 796 }
 797
 798 int
 799 iavf_dev_link_update(struct rte_eth_dev *dev,
 800                     __rte_unused int wait_to_complete)
 801 {
 802         struct rte_eth_link new_link;
 803         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
 804
 805         memset(&new_link, 0, sizeof(new_link));
 806
 807         /* Only read status info stored in VF, and the info is updated
 808          *  when receive LINK_CHANGE evnet from PF by Virtchnnl.
 809          */
 810         switch (vf->link_speed) {
 811         case 10:
 812                 new_link.link_speed = ETH_SPEED_NUM_10M;
 813                 break;
 814         case 100:
 815                 new_link.link_speed = ETH_SPEED_NUM_100M;
 816                 break;
 817         case 1000:
 818                 new_link.link_speed = ETH_SPEED_NUM_1G;
 819                 break;
 820         case 10000:
 821                 new_link.link_speed = ETH_SPEED_NUM_10G;
 822                 break;
 823         case 20000:
 824                 new_link.link_speed = ETH_SPEED_NUM_20G;
 825                 break;
 826         case 25000:
 827                 new_link.link_speed = ETH_SPEED_NUM_25G;
 828                 break;
 829         case 40000:
 830                 new_link.link_speed = ETH_SPEED_NUM_40G;
 831                 break;
 832         case 50000:
 833                 new_link.link_speed = ETH_SPEED_NUM_50G;
 834                 break;
 835         case 100000:
 836                 new_link.link_speed = ETH_SPEED_NUM_100G;
 837                 break;
 838         default:
 839                 new_link.link_speed = ETH_SPEED_NUM_NONE;
 840                 break;
 841         }
 842
 843         new_link.link_duplex = ETH_LINK_FULL_DUPLEX;
 844         new_link.link_status = vf->link_up ? ETH_LINK_UP :
 845                                              ETH_LINK_DOWN;
 846         new_link.link_autoneg = !(dev->data->dev_conf.link_speeds &
 847                                 ETH_LINK_SPEED_FIXED);
 848
 849         return rte_eth_linkstatus_set(dev, &new_link);
 850 }
 851
 852 static int
 853 iavf_dev_promiscuous_enable(struct rte_eth_dev *dev)
 854 {
 855         struct iavf_adapter *adapter =
 856                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 857         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
 858
 859         return iavf_config_promisc(adapter,
 860                                   true, vf->promisc_multicast_enabled);
 861 }
 862
 863 static int
 864 iavf_dev_promiscuous_disable(struct rte_eth_dev *dev)
 865 {
 866         struct iavf_adapter *adapter =
 867                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 868         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
 869
 870         return iavf_config_promisc(adapter,
 871                                   false, vf->promisc_multicast_enabled);
 872 }
 873
 874 static int
 875 iavf_dev_allmulticast_enable(struct rte_eth_dev *dev)
 876 {
 877         struct iavf_adapter *adapter =
 878                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 879         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
 880
 881         return iavf_config_promisc(adapter,
 882                                   vf->promisc_unicast_enabled, true);
 883 }
 884
 885 static int
 886 iavf_dev_allmulticast_disable(struct rte_eth_dev *dev)
 887 {
 888         struct iavf_adapter *adapter =
 889                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 890         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
 891
 892         return iavf_config_promisc(adapter,
 893                                   vf->promisc_unicast_enabled, false);
 894 }
 895
 896 static int
 897 iavf_dev_add_mac_addr(struct rte_eth_dev *dev, struct rte_ether_addr *addr,
 898                      __rte_unused uint32_t index,
 899                      __rte_unused uint32_t pool)
 900 {
 901         struct iavf_adapter *adapter =
 902                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 903         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
 904         int err;
 905
 906         if (rte_is_zero_ether_addr(addr)) {
 907                 PMD_DRV_LOG(ERR, "Invalid Ethernet Address");
 908                 return -EINVAL;
 909         }
 910
 911         err = iavf_add_del_eth_addr(adapter, addr, true);
 912         if (err) {
 913                 PMD_DRV_LOG(ERR, "fail to add MAC address");
 914                 return -EIO;
 915         }
 916
 917         vf->mac_num++;
 918
 919         return 0;
 920 }
 921
 922 static void
 923 iavf_dev_del_mac_addr(struct rte_eth_dev *dev, uint32_t index)
 924 {
 925         struct iavf_adapter *adapter =
 926                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 927         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
 928         struct rte_ether_addr *addr;
 929         int err;
 930
 931         addr = &dev->data->mac_addrs[index];
 932
 933         err = iavf_add_del_eth_addr(adapter, addr, false);
 934         if (err)
 935                 PMD_DRV_LOG(ERR, "fail to delete MAC address");
 936
 937         vf->mac_num--;
 938 }
 939
 940 static int
 941 iavf_dev_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
 942 {
 943         struct iavf_adapter *adapter =
 944                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 945         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
 946         int err;
 947
 948         if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_VLAN))
 949                 return -ENOTSUP;
 950
 951         err = iavf_add_del_vlan(adapter, vlan_id, on);
 952         if (err)
 953                 return -EIO;
 954         return 0;
 955 }
 956
 957 static int
 958 iavf_dev_vlan_offload_set(struct rte_eth_dev *dev, int mask)
 959 {
 960         struct iavf_adapter *adapter =
 961                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 962         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
 963         struct rte_eth_conf *dev_conf = &dev->data->dev_conf;
 964         int err;
 965
 966         if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_VLAN))
 967                 return -ENOTSUP;
 968
 969         /* Vlan stripping setting */
 970         if (mask & ETH_VLAN_STRIP_MASK) {
 971                 /* Enable or disable VLAN stripping */
 972                 if (dev_conf->rxmode.offloads & DEV_RX_OFFLOAD_VLAN_STRIP)
 973                         err = iavf_enable_vlan_strip(adapter);
 974                 else
 975                         err = iavf_disable_vlan_strip(adapter);
 976
 977                 if (err)
 978                         return -EIO;
 979         }
 980         return 0;
 981 }
 982
 983 static int
 984 iavf_dev_rss_reta_update(struct rte_eth_dev *dev,
 985                         struct rte_eth_rss_reta_entry64 *reta_conf,
 986                         uint16_t reta_size)
 987 {
 988         struct iavf_adapter *adapter =
 989                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 990         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
 991         uint8_t *lut;
 992         uint16_t i, idx, shift;
 993         int ret;
 994
 995         if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF))
 996                 return -ENOTSUP;
 997
 998         if (reta_size != vf->vf_res->rss_lut_size) {
 999                 PMD_DRV_LOG(ERR, "The size of hash lookup table configured "
1000                         "(%d) doesn't match the number of hardware can "
1001                         "support (%d)", reta_size, vf->vf_res->rss_lut_size);
1002                 return -EINVAL;
1003         }
1004
1005         lut = rte_zmalloc("rss_lut", reta_size, 0);
1006         if (!lut) {
1007                 PMD_DRV_LOG(ERR, "No memory can be allocated");
1008                 return -ENOMEM;
1009         }
1010         /* store the old lut table temporarily */
1011         rte_memcpy(lut, vf->rss_lut, reta_size);
1012
1013         for (i = 0; i < reta_size; i++) {
1014                 idx = i / RTE_RETA_GROUP_SIZE;
1015                 shift = i % RTE_RETA_GROUP_SIZE;
1016                 if (reta_conf[idx].mask & (1ULL << shift))
1017                         lut[i] = reta_conf[idx].reta[shift];
1018         }
1019
1020         rte_memcpy(vf->rss_lut, lut, reta_size);
1021         /* send virtchnnl ops to configure rss*/
1022         ret = iavf_configure_rss_lut(adapter);
1023         if (ret) /* revert back */
1024                 rte_memcpy(vf->rss_lut, lut, reta_size);
1025         rte_free(lut);
1026
1027         return ret;
1028 }
1029
1030 static int
1031 iavf_dev_rss_reta_query(struct rte_eth_dev *dev,
1032                        struct rte_eth_rss_reta_entry64 *reta_conf,
1033                        uint16_t reta_size)
1034 {
1035         struct iavf_adapter *adapter =
1036                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1037         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
1038         uint16_t i, idx, shift;
1039
1040         if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF))
1041                 return -ENOTSUP;
1042
1043         if (reta_size != vf->vf_res->rss_lut_size) {
1044                 PMD_DRV_LOG(ERR, "The size of hash lookup table configured "
1045                         "(%d) doesn't match the number of hardware can "
1046                         "support (%d)", reta_size, vf->vf_res->rss_lut_size);
1047                 return -EINVAL;
1048         }
1049
1050         for (i = 0; i < reta_size; i++) {
1051                 idx = i / RTE_RETA_GROUP_SIZE;
1052                 shift = i % RTE_RETA_GROUP_SIZE;
1053                 if (reta_conf[idx].mask & (1ULL << shift))
1054                         reta_conf[idx].reta[shift] = vf->rss_lut[i];
1055         }
1056
1057         return 0;
1058 }
1059
1060 static int
1061 iavf_dev_rss_hash_update(struct rte_eth_dev *dev,
1062                         struct rte_eth_rss_conf *rss_conf)
1063 {
1064         struct iavf_adapter *adapter =
1065                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1066         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
1067
1068         if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF))
1069                 return -ENOTSUP;
1070
1071         /* HENA setting, it is enabled by default, no change */
1072         if (!rss_conf->rss_key || rss_conf->rss_key_len == 0) {
1073                 PMD_DRV_LOG(DEBUG, "No key to be configured");
1074                 return 0;
1075         } else if (rss_conf->rss_key_len != vf->vf_res->rss_key_size) {
1076                 PMD_DRV_LOG(ERR, "The size of hash key configured "
1077                         "(%d) doesn't match the size of hardware can "
1078                         "support (%d)", rss_conf->rss_key_len,
1079                         vf->vf_res->rss_key_size);
1080                 return -EINVAL;
1081         }
1082
1083         rte_memcpy(vf->rss_key, rss_conf->rss_key, rss_conf->rss_key_len);
1084
1085         return iavf_configure_rss_key(adapter);
1086 }
1087
1088 static int
1089 iavf_dev_rss_hash_conf_get(struct rte_eth_dev *dev,
1090                           struct rte_eth_rss_conf *rss_conf)
1091 {
1092         struct iavf_adapter *adapter =
1093                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1094         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
1095
1096         if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF))
1097                 return -ENOTSUP;
1098
1099          /* Just set it to default value now. */
1100         rss_conf->rss_hf = IAVF_RSS_OFFLOAD_ALL;
1101
1102         if (!rss_conf->rss_key)
1103                 return 0;
1104
1105         rss_conf->rss_key_len = vf->vf_res->rss_key_size;
1106         rte_memcpy(rss_conf->rss_key, vf->rss_key, rss_conf->rss_key_len);
1107
1108         return 0;
1109 }
1110
1111 static int
1112 iavf_dev_mtu_set(struct rte_eth_dev *dev, uint16_t mtu)
1113 {
1114         uint32_t frame_size = mtu + IAVF_ETH_OVERHEAD;
1115         int ret = 0;
1116
1117         if (mtu < RTE_ETHER_MIN_MTU || frame_size > IAVF_FRAME_SIZE_MAX)
1118                 return -EINVAL;
1119
1120         /* mtu setting is forbidden if port is start */
1121         if (dev->data->dev_started) {
1122                 PMD_DRV_LOG(ERR, "port must be stopped before configuration");
1123                 return -EBUSY;
1124         }
1125
1126         if (frame_size > RTE_ETHER_MAX_LEN)
1127                 dev->data->dev_conf.rxmode.offloads |=
1128                                 DEV_RX_OFFLOAD_JUMBO_FRAME;
1129         else
1130                 dev->data->dev_conf.rxmode.offloads &=
1131                                 ~DEV_RX_OFFLOAD_JUMBO_FRAME;
1132
1133         dev->data->dev_conf.rxmode.max_rx_pkt_len = frame_size;
1134
1135         return ret;
1136 }
1137
1138 static int
1139 iavf_dev_set_default_mac_addr(struct rte_eth_dev *dev,
1140                              struct rte_ether_addr *mac_addr)
1141 {
1142         struct iavf_adapter *adapter =
1143                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1144         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(adapter);
1145         struct rte_ether_addr *perm_addr, *old_addr;
1146         int ret;
1147
1148         old_addr = (struct rte_ether_addr *)hw->mac.addr;
1149         perm_addr = (struct rte_ether_addr *)hw->mac.perm_addr;
1150
1151         /* If the MAC address is configured by host, skip the setting */
1152         if (rte_is_valid_assigned_ether_addr(perm_addr))
1153                 return -EPERM;
1154
1155         ret = iavf_add_del_eth_addr(adapter, old_addr, false);
1156         if (ret)
1157                 PMD_DRV_LOG(ERR, "Fail to delete old MAC:"
1158                             " %02X:%02X:%02X:%02X:%02X:%02X",
1159                             old_addr->addr_bytes[0],
1160                             old_addr->addr_bytes[1],
1161                             old_addr->addr_bytes[2],
1162                             old_addr->addr_bytes[3],
1163                             old_addr->addr_bytes[4],
1164                             old_addr->addr_bytes[5]);
1165
1166         ret = iavf_add_del_eth_addr(adapter, mac_addr, true);
1167         if (ret)
1168                 PMD_DRV_LOG(ERR, "Fail to add new MAC:"
1169                             " %02X:%02X:%02X:%02X:%02X:%02X",
1170                             mac_addr->addr_bytes[0],
1171                             mac_addr->addr_bytes[1],
1172                             mac_addr->addr_bytes[2],
1173                             mac_addr->addr_bytes[3],
1174                             mac_addr->addr_bytes[4],
1175                             mac_addr->addr_bytes[5]);
1176
1177         if (ret)
1178                 return -EIO;
1179
1180         rte_ether_addr_copy(mac_addr, (struct rte_ether_addr *)hw->mac.addr);
1181         return 0;
1182 }
1183
1184 static void
1185 iavf_stat_update_48(uint64_t *offset, uint64_t *stat)
1186 {
1187         if (*stat >= *offset)
1188                 *stat = *stat - *offset;
1189         else
1190                 *stat = (uint64_t)((*stat +
1191                         ((uint64_t)1 << IAVF_48_BIT_WIDTH)) - *offset);
1192
1193         *stat &= IAVF_48_BIT_MASK;
1194 }
1195
1196 static void
1197 iavf_stat_update_32(uint64_t *offset, uint64_t *stat)
1198 {
1199         if (*stat >= *offset)
1200                 *stat = (uint64_t)(*stat - *offset);
1201         else
1202                 *stat = (uint64_t)((*stat +
1203                         ((uint64_t)1 << IAVF_32_BIT_WIDTH)) - *offset);
1204 }
1205
1206 static void
1207 iavf_update_stats(struct iavf_vsi *vsi, struct virtchnl_eth_stats *nes)
1208 {
1209         struct virtchnl_eth_stats *oes = &vsi->eth_stats_offset;
1210
1211         iavf_stat_update_48(&oes->rx_bytes, &nes->rx_bytes);
1212         iavf_stat_update_48(&oes->rx_unicast, &nes->rx_unicast);
1213         iavf_stat_update_48(&oes->rx_multicast, &nes->rx_multicast);
1214         iavf_stat_update_48(&oes->rx_broadcast, &nes->rx_broadcast);
1215         iavf_stat_update_32(&oes->rx_discards, &nes->rx_discards);
1216         iavf_stat_update_48(&oes->tx_bytes, &nes->tx_bytes);
1217         iavf_stat_update_48(&oes->tx_unicast, &nes->tx_unicast);
1218         iavf_stat_update_48(&oes->tx_multicast, &nes->tx_multicast);
1219         iavf_stat_update_48(&oes->tx_broadcast, &nes->tx_broadcast);
1220         iavf_stat_update_32(&oes->tx_errors, &nes->tx_errors);
1221         iavf_stat_update_32(&oes->tx_discards, &nes->tx_discards);
1222 }
1223
1224 static int
1225 iavf_dev_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
1226 {
1227         struct iavf_adapter *adapter =
1228                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1229         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
1230         struct iavf_vsi *vsi = &vf->vsi;
1231         struct virtchnl_eth_stats *pstats = NULL;
1232         int ret;
1233
1234         ret = iavf_query_stats(adapter, &pstats);
1235         if (ret == 0) {
1236                 iavf_update_stats(vsi, pstats);
1237                 stats->ipackets = pstats->rx_unicast + pstats->rx_multicast +
1238                                 pstats->rx_broadcast - pstats->rx_discards;
1239                 stats->opackets = pstats->tx_broadcast + pstats->tx_multicast +
1240                                                 pstats->tx_unicast;
1241                 stats->imissed = pstats->rx_discards;
1242                 stats->oerrors = pstats->tx_errors + pstats->tx_discards;
1243                 stats->ibytes = pstats->rx_bytes;
1244                 stats->ibytes -= stats->ipackets * RTE_ETHER_CRC_LEN;
1245                 stats->obytes = pstats->tx_bytes;
1246         } else {
1247                 PMD_DRV_LOG(ERR, "Get statistics failed");
1248         }
1249         return ret;
1250 }
1251
1252 static int
1253 iavf_dev_stats_reset(struct rte_eth_dev *dev)
1254 {
1255         int ret;
1256         struct iavf_adapter *adapter =
1257                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1258         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
1259         struct iavf_vsi *vsi = &vf->vsi;
1260         struct virtchnl_eth_stats *pstats = NULL;
1261
1262         /* read stat values to clear hardware registers */
1263         ret = iavf_query_stats(adapter, &pstats);
1264         if (ret != 0)
1265                 return ret;
1266
1267         /* set stats offset base on current values */
1268         vsi->eth_stats_offset = *pstats;
1269
1270         return 0;
1271 }
1272
1273 static int iavf_dev_xstats_get_names(__rte_unused struct rte_eth_dev *dev,
1274                                       struct rte_eth_xstat_name *xstats_names,
1275                                       __rte_unused unsigned int limit)
1276 {
1277         unsigned int i;
1278
1279         if (xstats_names != NULL)
1280                 for (i = 0; i < IAVF_NB_XSTATS; i++) {
1281                         snprintf(xstats_names[i].name,
1282                                 sizeof(xstats_names[i].name),
1283                                 "%s", rte_iavf_stats_strings[i].name);
1284                 }
1285         return IAVF_NB_XSTATS;
1286 }
1287
1288 static int iavf_dev_xstats_get(struct rte_eth_dev *dev,
1289                                  struct rte_eth_xstat *xstats, unsigned int n)
1290 {
1291         int ret;
1292         unsigned int i;
1293         struct iavf_adapter *adapter =
1294                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1295         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
1296         struct iavf_vsi *vsi = &vf->vsi;
1297         struct virtchnl_eth_stats *pstats = NULL;
1298
1299         if (n < IAVF_NB_XSTATS)
1300                 return IAVF_NB_XSTATS;
1301
1302         ret = iavf_query_stats(adapter, &pstats);
1303         if (ret != 0)
1304                 return 0;
1305
1306         if (!xstats)
1307                 return 0;
1308
1309         iavf_update_stats(vsi, pstats);
1310
1311         /* loop over xstats array and values from pstats */
1312         for (i = 0; i < IAVF_NB_XSTATS; i++) {
1313                 xstats[i].id = i;
1314                 xstats[i].value = *(uint64_t *)(((char *)pstats) +
1315                         rte_iavf_stats_strings[i].offset);
1316         }
1317
1318         return IAVF_NB_XSTATS;
1319 }
1320
1321
1322 static int
1323 iavf_dev_rx_queue_intr_enable(struct rte_eth_dev *dev, uint16_t queue_id)
1324 {
1325         struct iavf_adapter *adapter =
1326                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1327         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1328         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(adapter);
1329         uint16_t msix_intr;
1330
1331         msix_intr = pci_dev->intr_handle.intr_vec[queue_id];
1332         if (msix_intr == IAVF_MISC_VEC_ID) {
1333                 PMD_DRV_LOG(INFO, "MISC is also enabled for control");
1334                 IAVF_WRITE_REG(hw, IAVF_VFINT_DYN_CTL01,
1335                                IAVF_VFINT_DYN_CTL01_INTENA_MASK |
1336                                IAVF_VFINT_DYN_CTL01_CLEARPBA_MASK |
1337                                IAVF_VFINT_DYN_CTL01_ITR_INDX_MASK);
1338         } else {
1339                 IAVF_WRITE_REG(hw,
1340                                IAVF_VFINT_DYN_CTLN1
1341                                 (msix_intr - IAVF_RX_VEC_START),
1342                                IAVF_VFINT_DYN_CTLN1_INTENA_MASK |
1343                                IAVF_VFINT_DYN_CTL01_CLEARPBA_MASK |
1344                                IAVF_VFINT_DYN_CTLN1_ITR_INDX_MASK);
1345         }
1346
1347         IAVF_WRITE_FLUSH(hw);
1348
1349         rte_intr_ack(&pci_dev->intr_handle);
1350
1351         return 0;
1352 }
1353
1354 static int
1355 iavf_dev_rx_queue_intr_disable(struct rte_eth_dev *dev, uint16_t queue_id)
1356 {
1357         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1358         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1359         uint16_t msix_intr;
1360
1361         msix_intr = pci_dev->intr_handle.intr_vec[queue_id];
1362         if (msix_intr == IAVF_MISC_VEC_ID) {
1363                 PMD_DRV_LOG(ERR, "MISC is used for control, cannot disable it");
1364                 return -EIO;
1365         }
1366
1367         IAVF_WRITE_REG(hw,
1368                       IAVF_VFINT_DYN_CTLN1(msix_intr - IAVF_RX_VEC_START),
1369                       0);
1370
1371         IAVF_WRITE_FLUSH(hw);
1372         return 0;
1373 }
1374
1375 static int
1376 iavf_check_vf_reset_done(struct iavf_hw *hw)
1377 {
1378         int i, reset;
1379
1380         for (i = 0; i < IAVF_RESET_WAIT_CNT; i++) {
1381                 reset = IAVF_READ_REG(hw, IAVF_VFGEN_RSTAT) &
1382                         IAVF_VFGEN_RSTAT_VFR_STATE_MASK;
1383                 reset = reset >> IAVF_VFGEN_RSTAT_VFR_STATE_SHIFT;
1384                 if (reset == VIRTCHNL_VFR_VFACTIVE ||
1385                     reset == VIRTCHNL_VFR_COMPLETED)
1386                         break;
1387                 rte_delay_ms(20);
1388         }
1389
1390         if (i >= IAVF_RESET_WAIT_CNT)
1391                 return -1;
1392
1393         return 0;
1394 }
1395
1396 static int
1397 iavf_init_vf(struct rte_eth_dev *dev)
1398 {
1399         int err, bufsz;
1400         struct iavf_adapter *adapter =
1401                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1402         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1403         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
1404
1405         err = iavf_set_mac_type(hw);
1406         if (err) {
1407                 PMD_INIT_LOG(ERR, "set_mac_type failed: %d", err);
1408                 goto err;
1409         }
1410
1411         err = iavf_check_vf_reset_done(hw);
1412         if (err) {
1413                 PMD_INIT_LOG(ERR, "VF is still resetting");
1414                 goto err;
1415         }
1416
1417         iavf_init_adminq_parameter(hw);
1418         err = iavf_init_adminq(hw);
1419         if (err) {
1420                 PMD_INIT_LOG(ERR, "init_adminq failed: %d", err);
1421                 goto err;
1422         }
1423
1424         vf->aq_resp = rte_zmalloc("vf_aq_resp", IAVF_AQ_BUF_SZ, 0);
1425         if (!vf->aq_resp) {
1426                 PMD_INIT_LOG(ERR, "unable to allocate vf_aq_resp memory");
1427                 goto err_aq;
1428         }
1429         if (iavf_check_api_version(adapter) != 0) {
1430                 PMD_INIT_LOG(ERR, "check_api version failed");
1431                 goto err_api;
1432         }
1433
1434         bufsz = sizeof(struct virtchnl_vf_resource) +
1435                 (IAVF_MAX_VF_VSI * sizeof(struct virtchnl_vsi_resource));
1436         vf->vf_res = rte_zmalloc("vf_res", bufsz, 0);
1437         if (!vf->vf_res) {
1438                 PMD_INIT_LOG(ERR, "unable to allocate vf_res memory");
1439                 goto err_api;
1440         }
1441         if (iavf_get_vf_resource(adapter) != 0) {
1442                 PMD_INIT_LOG(ERR, "iavf_get_vf_config failed");
1443                 goto err_alloc;
1444         }
1445         /* Allocate memort for RSS info */
1446         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF) {
1447                 vf->rss_key = rte_zmalloc("rss_key",
1448                                           vf->vf_res->rss_key_size, 0);
1449                 if (!vf->rss_key) {
1450                         PMD_INIT_LOG(ERR, "unable to allocate rss_key memory");
1451                         goto err_rss;
1452                 }
1453                 vf->rss_lut = rte_zmalloc("rss_lut",
1454                                           vf->vf_res->rss_lut_size, 0);
1455                 if (!vf->rss_lut) {
1456                         PMD_INIT_LOG(ERR, "unable to allocate rss_lut memory");
1457                         goto err_rss;
1458                 }
1459         }
1460
1461         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RX_FLEX_DESC) {
1462                 if (iavf_get_supported_rxdid(adapter) != 0) {
1463                         PMD_INIT_LOG(ERR, "failed to do get supported rxdid");
1464                         goto err_rss;
1465                 }
1466         }
1467
1468         return 0;
1469 err_rss:
1470         rte_free(vf->rss_key);
1471         rte_free(vf->rss_lut);
1472 err_alloc:
1473         rte_free(vf->vf_res);
1474         vf->vsi_res = NULL;
1475 err_api:
1476         rte_free(vf->aq_resp);
1477 err_aq:
1478         iavf_shutdown_adminq(hw);
1479 err:
1480         return -1;
1481 }
1482
1483 /* Enable default admin queue interrupt setting */
1484 static inline void
1485 iavf_enable_irq0(struct iavf_hw *hw)
1486 {
1487         /* Enable admin queue interrupt trigger */
1488         IAVF_WRITE_REG(hw, IAVF_VFINT_ICR0_ENA1,
1489                        IAVF_VFINT_ICR0_ENA1_ADMINQ_MASK);
1490
1491         IAVF_WRITE_REG(hw, IAVF_VFINT_DYN_CTL01,
1492                        IAVF_VFINT_DYN_CTL01_INTENA_MASK |
1493                        IAVF_VFINT_DYN_CTL01_CLEARPBA_MASK |
1494                        IAVF_VFINT_DYN_CTL01_ITR_INDX_MASK);
1495
1496         IAVF_WRITE_FLUSH(hw);
1497 }
1498
1499 static inline void
1500 iavf_disable_irq0(struct iavf_hw *hw)
1501 {
1502         /* Disable all interrupt types */
1503         IAVF_WRITE_REG(hw, IAVF_VFINT_ICR0_ENA1, 0);
1504         IAVF_WRITE_REG(hw, IAVF_VFINT_DYN_CTL01,
1505                        IAVF_VFINT_DYN_CTL01_ITR_INDX_MASK);
1506         IAVF_WRITE_FLUSH(hw);
1507 }
1508
1509 static void
1510 iavf_dev_interrupt_handler(void *param)
1511 {
1512         struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
1513         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1514
1515         iavf_disable_irq0(hw);
1516
1517         iavf_handle_virtchnl_msg(dev);
1518
1519         iavf_enable_irq0(hw);
1520 }
1521
1522 static int
1523 iavf_dev_filter_ctrl(struct rte_eth_dev *dev,
1524                      enum rte_filter_type filter_type,
1525                      enum rte_filter_op filter_op,
1526                      void *arg)
1527 {
1528         int ret = 0;
1529
1530         if (!dev)
1531                 return -EINVAL;
1532
1533         switch (filter_type) {
1534         case RTE_ETH_FILTER_GENERIC:
1535                 if (filter_op != RTE_ETH_FILTER_GET)
1536                         return -EINVAL;
1537                 *(const void **)arg = &iavf_flow_ops;
1538                 break;
1539         default:
1540                 PMD_DRV_LOG(WARNING, "Filter type (%d) not supported",
1541                             filter_type);
1542                 ret = -EINVAL;
1543                 break;
1544         }
1545
1546         return ret;
1547 }
1548
1549
1550 static int
1551 iavf_dev_init(struct rte_eth_dev *eth_dev)
1552 {
1553         struct iavf_adapter *adapter =
1554                 IAVF_DEV_PRIVATE_TO_ADAPTER(eth_dev->data->dev_private);
1555         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(adapter);
1556         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
1557         int ret = 0;
1558
1559         PMD_INIT_FUNC_TRACE();
1560
1561         /* assign ops func pointer */
1562         eth_dev->dev_ops = &iavf_eth_dev_ops;
1563         eth_dev->rx_queue_count = iavf_dev_rxq_count;
1564         eth_dev->rx_descriptor_status = iavf_dev_rx_desc_status;
1565         eth_dev->tx_descriptor_status = iavf_dev_tx_desc_status;
1566         eth_dev->rx_pkt_burst = &iavf_recv_pkts;
1567         eth_dev->tx_pkt_burst = &iavf_xmit_pkts;
1568         eth_dev->tx_pkt_prepare = &iavf_prep_pkts;
1569
1570         /* For secondary processes, we don't initialise any further as primary
1571          * has already done this work. Only check if we need a different RX
1572          * and TX function.
1573          */
1574         if (rte_eal_process_type() != RTE_PROC_PRIMARY) {
1575                 iavf_set_rx_function(eth_dev);
1576                 iavf_set_tx_function(eth_dev);
1577                 return 0;
1578         }
1579         rte_eth_copy_pci_info(eth_dev, pci_dev);
1580         eth_dev->data->dev_flags |= RTE_ETH_DEV_AUTOFILL_QUEUE_XSTATS;
1581
1582         hw->vendor_id = pci_dev->id.vendor_id;
1583         hw->device_id = pci_dev->id.device_id;
1584         hw->subsystem_vendor_id = pci_dev->id.subsystem_vendor_id;
1585         hw->subsystem_device_id = pci_dev->id.subsystem_device_id;
1586         hw->bus.bus_id = pci_dev->addr.bus;
1587         hw->bus.device = pci_dev->addr.devid;
1588         hw->bus.func = pci_dev->addr.function;
1589         hw->hw_addr = (void *)pci_dev->mem_resource[0].addr;
1590         hw->back = IAVF_DEV_PRIVATE_TO_ADAPTER(eth_dev->data->dev_private);
1591         adapter->eth_dev = eth_dev;
1592         adapter->stopped = 1;
1593
1594         if (iavf_init_vf(eth_dev) != 0) {
1595                 PMD_INIT_LOG(ERR, "Init vf failed");
1596                 return -1;
1597         }
1598
1599         /* set default ptype table */
1600         adapter->ptype_tbl = iavf_get_default_ptype_table();
1601
1602         /* copy mac addr */
1603         eth_dev->data->mac_addrs = rte_zmalloc(
1604                 "iavf_mac", RTE_ETHER_ADDR_LEN * IAVF_NUM_MACADDR_MAX, 0);
1605         if (!eth_dev->data->mac_addrs) {
1606                 PMD_INIT_LOG(ERR, "Failed to allocate %d bytes needed to"
1607                              " store MAC addresses",
1608                              RTE_ETHER_ADDR_LEN * IAVF_NUM_MACADDR_MAX);
1609                 return -ENOMEM;
1610         }
1611         /* If the MAC address is not configured by host,
1612          * generate a random one.
1613          */
1614         if (!rte_is_valid_assigned_ether_addr(
1615                         (struct rte_ether_addr *)hw->mac.addr))
1616                 rte_eth_random_addr(hw->mac.addr);
1617         rte_ether_addr_copy((struct rte_ether_addr *)hw->mac.addr,
1618                         &eth_dev->data->mac_addrs[0]);
1619
1620         /* register callback func to eal lib */
1621         rte_intr_callback_register(&pci_dev->intr_handle,
1622                                    iavf_dev_interrupt_handler,
1623                                    (void *)eth_dev);
1624
1625         /* enable uio intr after callback register */
1626         rte_intr_enable(&pci_dev->intr_handle);
1627
1628         /* configure and enable device interrupt */
1629         iavf_enable_irq0(hw);
1630
1631         ret = iavf_flow_init(adapter);
1632         if (ret) {
1633                 PMD_INIT_LOG(ERR, "Failed to initialize flow");
1634                 return ret;
1635         }
1636
1637         return 0;
1638 }
1639
1640 static int
1641 iavf_dev_close(struct rte_eth_dev *dev)
1642 {
1643         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1644         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1645         struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
1646         struct iavf_adapter *adapter =
1647                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1648         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
1649         int ret;
1650
1651         if (rte_eal_process_type() != RTE_PROC_PRIMARY)
1652                 return 0;
1653
1654         ret = iavf_dev_stop(dev);
1655
1656         iavf_flow_flush(dev, NULL);
1657         iavf_flow_uninit(adapter);
1658
1659         /*
1660          * disable promiscuous mode before reset vf
1661          * it is a workaround solution when work with kernel driver
1662          * and it is not the normal way
1663          */
1664         if (vf->promisc_unicast_enabled || vf->promisc_multicast_enabled)
1665                 iavf_config_promisc(adapter, false, false);
1666
1667         iavf_shutdown_adminq(hw);
1668         /* disable uio intr before callback unregister */
1669         rte_intr_disable(intr_handle);
1670
1671         /* unregister callback func from eal lib */
1672         rte_intr_callback_unregister(intr_handle,
1673                                      iavf_dev_interrupt_handler, dev);
1674         iavf_disable_irq0(hw);
1675
1676         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF) {
1677                 if (vf->rss_lut) {
1678                         rte_free(vf->rss_lut);
1679                         vf->rss_lut = NULL;
1680                 }
1681                 if (vf->rss_key) {
1682                         rte_free(vf->rss_key);
1683                         vf->rss_key = NULL;
1684                 }
1685         }
1686
1687         rte_free(vf->vf_res);
1688         vf->vsi_res = NULL;
1689         vf->vf_res = NULL;
1690
1691         rte_free(vf->aq_resp);
1692         vf->aq_resp = NULL;
1693
1694         vf->vf_reset = false;
1695
1696         return ret;
1697 }
1698
1699 static int
1700 iavf_dev_uninit(struct rte_eth_dev *dev)
1701 {
1702         if (rte_eal_process_type() != RTE_PROC_PRIMARY)
1703                 return -EPERM;
1704
1705         iavf_dev_close(dev);
1706
1707         return 0;
1708 }
1709
1710 /*
1711  * Reset VF device only to re-initialize resources in PMD layer
1712  */
1713 static int
1714 iavf_dev_reset(struct rte_eth_dev *dev)
1715 {
1716         int ret;
1717
1718         ret = iavf_dev_uninit(dev);
1719         if (ret)
1720                 return ret;
1721
1722         return iavf_dev_init(dev);
1723 }
1724
1725 static int
1726 iavf_dcf_cap_check_handler(__rte_unused const char *key,
1727                            const char *value, __rte_unused void *opaque)
1728 {
1729         if (strcmp(value, "dcf"))
1730                 return -1;
1731
1732         return 0;
1733 }
1734
1735 static int
1736 iavf_dcf_cap_selected(struct rte_devargs *devargs)
1737 {
1738         struct rte_kvargs *kvlist;
1739         const char *key = "cap";
1740         int ret = 0;
1741
1742         if (devargs == NULL)
1743                 return 0;
1744
1745         kvlist = rte_kvargs_parse(devargs->args, NULL);
1746         if (kvlist == NULL)
1747                 return 0;
1748
1749         if (!rte_kvargs_count(kvlist, key))
1750                 goto exit;
1751
1752         /* dcf capability selected when there's a key-value pair: cap=dcf */
1753         if (rte_kvargs_process(kvlist, key,
1754                                iavf_dcf_cap_check_handler, NULL) < 0)
1755                 goto exit;
1756
1757         ret = 1;
1758
1759 exit:
1760         rte_kvargs_free(kvlist);
1761         return ret;
1762 }
1763
1764 static int eth_iavf_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
1765                              struct rte_pci_device *pci_dev)
1766 {
1767         if (iavf_dcf_cap_selected(pci_dev->device.devargs))
1768                 return 1;
1769
1770         return rte_eth_dev_pci_generic_probe(pci_dev,
1771                 sizeof(struct iavf_adapter), iavf_dev_init);
1772 }
1773
1774 static int eth_iavf_pci_remove(struct rte_pci_device *pci_dev)
1775 {
1776         return rte_eth_dev_pci_generic_remove(pci_dev, iavf_dev_uninit);
1777 }
1778
1779 /* Adaptive virtual function driver struct */
1780 static struct rte_pci_driver rte_iavf_pmd = {
1781         .id_table = pci_id_iavf_map,
1782         .drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC,
1783         .probe = eth_iavf_pci_probe,
1784         .remove = eth_iavf_pci_remove,
1785 };
1786
1787 RTE_PMD_REGISTER_PCI(net_iavf, rte_iavf_pmd);
1788 RTE_PMD_REGISTER_PCI_TABLE(net_iavf, pci_id_iavf_map);
1789 RTE_PMD_REGISTER_KMOD_DEP(net_iavf, "* igb_uio | vfio-pci");
1790 RTE_PMD_REGISTER_PARAM_STRING(net_iavf, "cap=dcf");
1791 RTE_LOG_REGISTER(iavf_logtype_init, pmd.net.iavf.init, NOTICE);
1792 RTE_LOG_REGISTER(iavf_logtype_driver, pmd.net.iavf.driver, NOTICE);
1793 #ifdef RTE_LIBRTE_IAVF_DEBUG_RX
1794 RTE_LOG_REGISTER(iavf_logtype_rx, pmd.net.iavf.rx, DEBUG);
1795 #endif
1796 #ifdef RTE_LIBRTE_IAVF_DEBUG_TX
1797 RTE_LOG_REGISTER(iavf_logtype_tx, pmd.net.iavf.tx, DEBUG);
1798 #endif
1799 #ifdef RTE_LIBRTE_IAVF_DEBUG_TX_FREE
1800 RTE_LOG_REGISTER(iavf_logtype_tx_free, pmd.net.iavf.tx_free, DEBUG);
1801 #endif