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         uint16_t interval, i;
 441         int vec;
 442
 443         if (rte_intr_cap_multiple(intr_handle) &&
 444             dev->data->dev_conf.intr_conf.rxq) {
 445                 if (rte_intr_efd_enable(intr_handle, dev->data->nb_rx_queues))
 446                         return -1;
 447         }
 448
 449         if (rte_intr_dp_is_en(intr_handle) && !intr_handle->intr_vec) {
 450                 intr_handle->intr_vec =
 451                         rte_zmalloc("intr_vec",
 452                                     dev->data->nb_rx_queues * sizeof(int), 0);
 453                 if (!intr_handle->intr_vec) {
 454                         PMD_DRV_LOG(ERR, "Failed to allocate %d rx intr_vec",
 455                                     dev->data->nb_rx_queues);
 456                         return -1;
 457                 }
 458         }
 459
 460         if (!dev->data->dev_conf.intr_conf.rxq ||
 461             !rte_intr_dp_is_en(intr_handle)) {
 462                 /* Rx interrupt disabled, Map interrupt only for writeback */
 463                 vf->nb_msix = 1;
 464                 if (vf->vf_res->vf_cap_flags &
 465                     VIRTCHNL_VF_OFFLOAD_WB_ON_ITR) {
 466                         /* If WB_ON_ITR supports, enable it */
 467                         vf->msix_base = IAVF_RX_VEC_START;
 468                         IAVF_WRITE_REG(hw,
 469                                        IAVF_VFINT_DYN_CTLN1(vf->msix_base - 1),
 470                                        IAVF_VFINT_DYN_CTLN1_ITR_INDX_MASK |
 471                                        IAVF_VFINT_DYN_CTLN1_WB_ON_ITR_MASK);
 472                 } else {
 473                         /* If no WB_ON_ITR offload flags, need to set
 474                          * interrupt for descriptor write back.
 475                          */
 476                         vf->msix_base = IAVF_MISC_VEC_ID;
 477
 478                         /* set ITR to max */
 479                         interval = iavf_calc_itr_interval(
 480                                         IAVF_QUEUE_ITR_INTERVAL_MAX);
 481                         IAVF_WRITE_REG(hw, IAVF_VFINT_DYN_CTL01,
 482                                        IAVF_VFINT_DYN_CTL01_INTENA_MASK |
 483                                        (IAVF_ITR_INDEX_DEFAULT <<
 484                                         IAVF_VFINT_DYN_CTL01_ITR_INDX_SHIFT) |
 485                                        (interval <<
 486                                         IAVF_VFINT_DYN_CTL01_INTERVAL_SHIFT));
 487                 }
 488                 IAVF_WRITE_FLUSH(hw);
 489                 /* map all queues to the same interrupt */
 490                 for (i = 0; i < dev->data->nb_rx_queues; i++)
 491                         vf->rxq_map[vf->msix_base] |= 1 << i;
 492         } else {
 493                 if (!rte_intr_allow_others(intr_handle)) {
 494                         vf->nb_msix = 1;
 495                         vf->msix_base = IAVF_MISC_VEC_ID;
 496                         for (i = 0; i < dev->data->nb_rx_queues; i++) {
 497                                 vf->rxq_map[vf->msix_base] |= 1 << i;
 498                                 intr_handle->intr_vec[i] = IAVF_MISC_VEC_ID;
 499                         }
 500                         PMD_DRV_LOG(DEBUG,
 501                                     "vector %u are mapping to all Rx queues",
 502                                     vf->msix_base);
 503                 } else {
 504                         /* If Rx interrupt is reuquired, and we can use
 505                          * multi interrupts, then the vec is from 1
 506                          */
 507                         vf->nb_msix = RTE_MIN(vf->vf_res->max_vectors,
 508                                               intr_handle->nb_efd);
 509                         vf->msix_base = IAVF_RX_VEC_START;
 510                         vec = IAVF_RX_VEC_START;
 511                         for (i = 0; i < dev->data->nb_rx_queues; i++) {
 512                                 vf->rxq_map[vec] |= 1 << i;
 513                                 intr_handle->intr_vec[i] = vec++;
 514                                 if (vec >= vf->nb_msix)
 515                                         vec = IAVF_RX_VEC_START;
 516                         }
 517                         PMD_DRV_LOG(DEBUG,
 518                                     "%u vectors are mapping to %u Rx queues",
 519                                     vf->nb_msix, dev->data->nb_rx_queues);
 520                 }
 521         }
 522
 523         if (iavf_config_irq_map(adapter)) {
 524                 PMD_DRV_LOG(ERR, "config interrupt mapping failed");
 525                 return -1;
 526         }
 527         return 0;
 528 }
 529
 530 static int
 531 iavf_start_queues(struct rte_eth_dev *dev)
 532 {
 533         struct iavf_rx_queue *rxq;
 534         struct iavf_tx_queue *txq;
 535         int i;
 536
 537         for (i = 0; i < dev->data->nb_tx_queues; i++) {
 538                 txq = dev->data->tx_queues[i];
 539                 if (txq->tx_deferred_start)
 540                         continue;
 541                 if (iavf_dev_tx_queue_start(dev, i) != 0) {
 542                         PMD_DRV_LOG(ERR, "Fail to start queue %u", i);
 543                         return -1;
 544                 }
 545         }
 546
 547         for (i = 0; i < dev->data->nb_rx_queues; i++) {
 548                 rxq = dev->data->rx_queues[i];
 549                 if (rxq->rx_deferred_start)
 550                         continue;
 551                 if (iavf_dev_rx_queue_start(dev, i) != 0) {
 552                         PMD_DRV_LOG(ERR, "Fail to start queue %u", i);
 553                         return -1;
 554                 }
 555         }
 556
 557         return 0;
 558 }
 559
 560 static int
 561 iavf_dev_start(struct rte_eth_dev *dev)
 562 {
 563         struct iavf_adapter *adapter =
 564                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 565         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
 566         struct rte_intr_handle *intr_handle = dev->intr_handle;
 567
 568         PMD_INIT_FUNC_TRACE();
 569
 570         adapter->stopped = 0;
 571
 572         vf->max_pkt_len = dev->data->dev_conf.rxmode.max_rx_pkt_len;
 573         vf->num_queue_pairs = RTE_MAX(dev->data->nb_rx_queues,
 574                                       dev->data->nb_tx_queues);
 575
 576         if (iavf_init_queues(dev) != 0) {
 577                 PMD_DRV_LOG(ERR, "failed to do Queue init");
 578                 return -1;
 579         }
 580
 581         if (iavf_configure_queues(adapter) != 0) {
 582                 PMD_DRV_LOG(ERR, "configure queues failed");
 583                 goto err_queue;
 584         }
 585
 586         if (iavf_config_rx_queues_irqs(dev, intr_handle) != 0) {
 587                 PMD_DRV_LOG(ERR, "configure irq failed");
 588                 goto err_queue;
 589         }
 590         /* re-enable intr again, because efd assign may change */
 591         if (dev->data->dev_conf.intr_conf.rxq != 0) {
 592                 rte_intr_disable(intr_handle);
 593                 rte_intr_enable(intr_handle);
 594         }
 595
 596         /* Set all mac addrs */
 597         iavf_add_del_all_mac_addr(adapter, true);
 598
 599         /* Set all multicast addresses */
 600         iavf_add_del_mc_addr_list(adapter, vf->mc_addrs, vf->mc_addrs_num,
 601                                   true);
 602
 603         if (iavf_start_queues(dev) != 0) {
 604                 PMD_DRV_LOG(ERR, "enable queues failed");
 605                 goto err_mac;
 606         }
 607
 608         return 0;
 609
 610 err_mac:
 611         iavf_add_del_all_mac_addr(adapter, false);
 612 err_queue:
 613         return -1;
 614 }
 615
 616 static int
 617 iavf_dev_stop(struct rte_eth_dev *dev)
 618 {
 619         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
 620         struct iavf_adapter *adapter =
 621                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 622         struct rte_intr_handle *intr_handle = dev->intr_handle;
 623
 624         PMD_INIT_FUNC_TRACE();
 625
 626         if (adapter->stopped == 1)
 627                 return 0;
 628
 629         iavf_stop_queues(dev);
 630
 631         /* Disable the interrupt for Rx */
 632         rte_intr_efd_disable(intr_handle);
 633         /* Rx interrupt vector mapping free */
 634         if (intr_handle->intr_vec) {
 635                 rte_free(intr_handle->intr_vec);
 636                 intr_handle->intr_vec = NULL;
 637         }
 638
 639         /* remove all mac addrs */
 640         iavf_add_del_all_mac_addr(adapter, false);
 641
 642         /* remove all multicast addresses */
 643         iavf_add_del_mc_addr_list(adapter, vf->mc_addrs, vf->mc_addrs_num,
 644                                   false);
 645
 646         adapter->stopped = 1;
 647         dev->data->dev_started = 0;
 648
 649         return 0;
 650 }
 651
 652 static int
 653 iavf_dev_info_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
 654 {
 655         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
 656
 657         dev_info->max_rx_queues = IAVF_MAX_NUM_QUEUES_LV;
 658         dev_info->max_tx_queues = IAVF_MAX_NUM_QUEUES_LV;
 659         dev_info->min_rx_bufsize = IAVF_BUF_SIZE_MIN;
 660         dev_info->max_rx_pktlen = IAVF_FRAME_SIZE_MAX;
 661         dev_info->max_mtu = dev_info->max_rx_pktlen - IAVF_ETH_OVERHEAD;
 662         dev_info->min_mtu = RTE_ETHER_MIN_MTU;
 663         dev_info->hash_key_size = vf->vf_res->rss_key_size;
 664         dev_info->reta_size = vf->vf_res->rss_lut_size;
 665         dev_info->flow_type_rss_offloads = IAVF_RSS_OFFLOAD_ALL;
 666         dev_info->max_mac_addrs = IAVF_NUM_MACADDR_MAX;
 667         dev_info->rx_offload_capa =
 668                 DEV_RX_OFFLOAD_VLAN_STRIP |
 669                 DEV_RX_OFFLOAD_QINQ_STRIP |
 670                 DEV_RX_OFFLOAD_IPV4_CKSUM |
 671                 DEV_RX_OFFLOAD_UDP_CKSUM |
 672                 DEV_RX_OFFLOAD_TCP_CKSUM |
 673                 DEV_RX_OFFLOAD_OUTER_IPV4_CKSUM |
 674                 DEV_RX_OFFLOAD_SCATTER |
 675                 DEV_RX_OFFLOAD_JUMBO_FRAME |
 676                 DEV_RX_OFFLOAD_VLAN_FILTER |
 677                 DEV_RX_OFFLOAD_RSS_HASH;
 678         dev_info->tx_offload_capa =
 679                 DEV_TX_OFFLOAD_VLAN_INSERT |
 680                 DEV_TX_OFFLOAD_QINQ_INSERT |
 681                 DEV_TX_OFFLOAD_IPV4_CKSUM |
 682                 DEV_TX_OFFLOAD_UDP_CKSUM |
 683                 DEV_TX_OFFLOAD_TCP_CKSUM |
 684                 DEV_TX_OFFLOAD_SCTP_CKSUM |
 685                 DEV_TX_OFFLOAD_OUTER_IPV4_CKSUM |
 686                 DEV_TX_OFFLOAD_TCP_TSO |
 687                 DEV_TX_OFFLOAD_VXLAN_TNL_TSO |
 688                 DEV_TX_OFFLOAD_GRE_TNL_TSO |
 689                 DEV_TX_OFFLOAD_IPIP_TNL_TSO |
 690                 DEV_TX_OFFLOAD_GENEVE_TNL_TSO |
 691                 DEV_TX_OFFLOAD_MULTI_SEGS;
 692
 693         dev_info->default_rxconf = (struct rte_eth_rxconf) {
 694                 .rx_free_thresh = IAVF_DEFAULT_RX_FREE_THRESH,
 695                 .rx_drop_en = 0,
 696                 .offloads = 0,
 697         };
 698
 699         dev_info->default_txconf = (struct rte_eth_txconf) {
 700                 .tx_free_thresh = IAVF_DEFAULT_TX_FREE_THRESH,
 701                 .tx_rs_thresh = IAVF_DEFAULT_TX_RS_THRESH,
 702                 .offloads = 0,
 703         };
 704
 705         dev_info->rx_desc_lim = (struct rte_eth_desc_lim) {
 706                 .nb_max = IAVF_MAX_RING_DESC,
 707                 .nb_min = IAVF_MIN_RING_DESC,
 708                 .nb_align = IAVF_ALIGN_RING_DESC,
 709         };
 710
 711         dev_info->tx_desc_lim = (struct rte_eth_desc_lim) {
 712                 .nb_max = IAVF_MAX_RING_DESC,
 713                 .nb_min = IAVF_MIN_RING_DESC,
 714                 .nb_align = IAVF_ALIGN_RING_DESC,
 715         };
 716
 717         return 0;
 718 }
 719
 720 static const uint32_t *
 721 iavf_dev_supported_ptypes_get(struct rte_eth_dev *dev __rte_unused)
 722 {
 723         static const uint32_t ptypes[] = {
 724                 RTE_PTYPE_L2_ETHER,
 725                 RTE_PTYPE_L3_IPV4_EXT_UNKNOWN,
 726                 RTE_PTYPE_L4_FRAG,
 727                 RTE_PTYPE_L4_ICMP,
 728                 RTE_PTYPE_L4_NONFRAG,
 729                 RTE_PTYPE_L4_SCTP,
 730                 RTE_PTYPE_L4_TCP,
 731                 RTE_PTYPE_L4_UDP,
 732                 RTE_PTYPE_UNKNOWN
 733         };
 734         return ptypes;
 735 }
 736
 737 int
 738 iavf_dev_link_update(struct rte_eth_dev *dev,
 739                     __rte_unused int wait_to_complete)
 740 {
 741         struct rte_eth_link new_link;
 742         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
 743
 744         memset(&new_link, 0, sizeof(new_link));
 745
 746         /* Only read status info stored in VF, and the info is updated
 747          *  when receive LINK_CHANGE evnet from PF by Virtchnnl.
 748          */
 749         switch (vf->link_speed) {
 750         case 10:
 751                 new_link.link_speed = ETH_SPEED_NUM_10M;
 752                 break;
 753         case 100:
 754                 new_link.link_speed = ETH_SPEED_NUM_100M;
 755                 break;
 756         case 1000:
 757                 new_link.link_speed = ETH_SPEED_NUM_1G;
 758                 break;
 759         case 10000:
 760                 new_link.link_speed = ETH_SPEED_NUM_10G;
 761                 break;
 762         case 20000:
 763                 new_link.link_speed = ETH_SPEED_NUM_20G;
 764                 break;
 765         case 25000:
 766                 new_link.link_speed = ETH_SPEED_NUM_25G;
 767                 break;
 768         case 40000:
 769                 new_link.link_speed = ETH_SPEED_NUM_40G;
 770                 break;
 771         case 50000:
 772                 new_link.link_speed = ETH_SPEED_NUM_50G;
 773                 break;
 774         case 100000:
 775                 new_link.link_speed = ETH_SPEED_NUM_100G;
 776                 break;
 777         default:
 778                 new_link.link_speed = ETH_SPEED_NUM_NONE;
 779                 break;
 780         }
 781
 782         new_link.link_duplex = ETH_LINK_FULL_DUPLEX;
 783         new_link.link_status = vf->link_up ? ETH_LINK_UP :
 784                                              ETH_LINK_DOWN;
 785         new_link.link_autoneg = !(dev->data->dev_conf.link_speeds &
 786                                 ETH_LINK_SPEED_FIXED);
 787
 788         return rte_eth_linkstatus_set(dev, &new_link);
 789 }
 790
 791 static int
 792 iavf_dev_promiscuous_enable(struct rte_eth_dev *dev)
 793 {
 794         struct iavf_adapter *adapter =
 795                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 796         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
 797
 798         return iavf_config_promisc(adapter,
 799                                   true, vf->promisc_multicast_enabled);
 800 }
 801
 802 static int
 803 iavf_dev_promiscuous_disable(struct rte_eth_dev *dev)
 804 {
 805         struct iavf_adapter *adapter =
 806                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 807         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
 808
 809         return iavf_config_promisc(adapter,
 810                                   false, vf->promisc_multicast_enabled);
 811 }
 812
 813 static int
 814 iavf_dev_allmulticast_enable(struct rte_eth_dev *dev)
 815 {
 816         struct iavf_adapter *adapter =
 817                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 818         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
 819
 820         return iavf_config_promisc(adapter,
 821                                   vf->promisc_unicast_enabled, true);
 822 }
 823
 824 static int
 825 iavf_dev_allmulticast_disable(struct rte_eth_dev *dev)
 826 {
 827         struct iavf_adapter *adapter =
 828                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 829         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
 830
 831         return iavf_config_promisc(adapter,
 832                                   vf->promisc_unicast_enabled, false);
 833 }
 834
 835 static int
 836 iavf_dev_add_mac_addr(struct rte_eth_dev *dev, struct rte_ether_addr *addr,
 837                      __rte_unused uint32_t index,
 838                      __rte_unused uint32_t pool)
 839 {
 840         struct iavf_adapter *adapter =
 841                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 842         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
 843         int err;
 844
 845         if (rte_is_zero_ether_addr(addr)) {
 846                 PMD_DRV_LOG(ERR, "Invalid Ethernet Address");
 847                 return -EINVAL;
 848         }
 849
 850         err = iavf_add_del_eth_addr(adapter, addr, true);
 851         if (err) {
 852                 PMD_DRV_LOG(ERR, "fail to add MAC address");
 853                 return -EIO;
 854         }
 855
 856         vf->mac_num++;
 857
 858         return 0;
 859 }
 860
 861 static void
 862 iavf_dev_del_mac_addr(struct rte_eth_dev *dev, uint32_t index)
 863 {
 864         struct iavf_adapter *adapter =
 865                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 866         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
 867         struct rte_ether_addr *addr;
 868         int err;
 869
 870         addr = &dev->data->mac_addrs[index];
 871
 872         err = iavf_add_del_eth_addr(adapter, addr, false);
 873         if (err)
 874                 PMD_DRV_LOG(ERR, "fail to delete MAC address");
 875
 876         vf->mac_num--;
 877 }
 878
 879 static int
 880 iavf_dev_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
 881 {
 882         struct iavf_adapter *adapter =
 883                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 884         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
 885         int err;
 886
 887         if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_VLAN))
 888                 return -ENOTSUP;
 889
 890         err = iavf_add_del_vlan(adapter, vlan_id, on);
 891         if (err)
 892                 return -EIO;
 893         return 0;
 894 }
 895
 896 static int
 897 iavf_dev_vlan_offload_set(struct rte_eth_dev *dev, int mask)
 898 {
 899         struct iavf_adapter *adapter =
 900                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 901         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
 902         struct rte_eth_conf *dev_conf = &dev->data->dev_conf;
 903         int err;
 904
 905         if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_VLAN))
 906                 return -ENOTSUP;
 907
 908         /* Vlan stripping setting */
 909         if (mask & ETH_VLAN_STRIP_MASK) {
 910                 /* Enable or disable VLAN stripping */
 911                 if (dev_conf->rxmode.offloads & DEV_RX_OFFLOAD_VLAN_STRIP)
 912                         err = iavf_enable_vlan_strip(adapter);
 913                 else
 914                         err = iavf_disable_vlan_strip(adapter);
 915
 916                 if (err)
 917                         return -EIO;
 918         }
 919         return 0;
 920 }
 921
 922 static int
 923 iavf_dev_rss_reta_update(struct rte_eth_dev *dev,
 924                         struct rte_eth_rss_reta_entry64 *reta_conf,
 925                         uint16_t reta_size)
 926 {
 927         struct iavf_adapter *adapter =
 928                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 929         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
 930         uint8_t *lut;
 931         uint16_t i, idx, shift;
 932         int ret;
 933
 934         if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF))
 935                 return -ENOTSUP;
 936
 937         if (reta_size != vf->vf_res->rss_lut_size) {
 938                 PMD_DRV_LOG(ERR, "The size of hash lookup table configured "
 939                         "(%d) doesn't match the number of hardware can "
 940                         "support (%d)", reta_size, vf->vf_res->rss_lut_size);
 941                 return -EINVAL;
 942         }
 943
 944         lut = rte_zmalloc("rss_lut", reta_size, 0);
 945         if (!lut) {
 946                 PMD_DRV_LOG(ERR, "No memory can be allocated");
 947                 return -ENOMEM;
 948         }
 949         /* store the old lut table temporarily */
 950         rte_memcpy(lut, vf->rss_lut, reta_size);
 951
 952         for (i = 0; i < reta_size; i++) {
 953                 idx = i / RTE_RETA_GROUP_SIZE;
 954                 shift = i % RTE_RETA_GROUP_SIZE;
 955                 if (reta_conf[idx].mask & (1ULL << shift))
 956                         lut[i] = reta_conf[idx].reta[shift];
 957         }
 958
 959         rte_memcpy(vf->rss_lut, lut, reta_size);
 960         /* send virtchnnl ops to configure rss*/
 961         ret = iavf_configure_rss_lut(adapter);
 962         if (ret) /* revert back */
 963                 rte_memcpy(vf->rss_lut, lut, reta_size);
 964         rte_free(lut);
 965
 966         return ret;
 967 }
 968
 969 static int
 970 iavf_dev_rss_reta_query(struct rte_eth_dev *dev,
 971                        struct rte_eth_rss_reta_entry64 *reta_conf,
 972                        uint16_t reta_size)
 973 {
 974         struct iavf_adapter *adapter =
 975                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 976         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
 977         uint16_t i, idx, shift;
 978
 979         if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF))
 980                 return -ENOTSUP;
 981
 982         if (reta_size != vf->vf_res->rss_lut_size) {
 983                 PMD_DRV_LOG(ERR, "The size of hash lookup table configured "
 984                         "(%d) doesn't match the number of hardware can "
 985                         "support (%d)", reta_size, vf->vf_res->rss_lut_size);
 986                 return -EINVAL;
 987         }
 988
 989         for (i = 0; i < reta_size; i++) {
 990                 idx = i / RTE_RETA_GROUP_SIZE;
 991                 shift = i % RTE_RETA_GROUP_SIZE;
 992                 if (reta_conf[idx].mask & (1ULL << shift))
 993                         reta_conf[idx].reta[shift] = vf->rss_lut[i];
 994         }
 995
 996         return 0;
 997 }
 998
 999 static int
1000 iavf_dev_rss_hash_update(struct rte_eth_dev *dev,
1001                         struct rte_eth_rss_conf *rss_conf)
1002 {
1003         struct iavf_adapter *adapter =
1004                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1005         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
1006
1007         if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF))
1008                 return -ENOTSUP;
1009
1010         /* HENA setting, it is enabled by default, no change */
1011         if (!rss_conf->rss_key || rss_conf->rss_key_len == 0) {
1012                 PMD_DRV_LOG(DEBUG, "No key to be configured");
1013                 return 0;
1014         } else if (rss_conf->rss_key_len != vf->vf_res->rss_key_size) {
1015                 PMD_DRV_LOG(ERR, "The size of hash key configured "
1016                         "(%d) doesn't match the size of hardware can "
1017                         "support (%d)", rss_conf->rss_key_len,
1018                         vf->vf_res->rss_key_size);
1019                 return -EINVAL;
1020         }
1021
1022         rte_memcpy(vf->rss_key, rss_conf->rss_key, rss_conf->rss_key_len);
1023
1024         return iavf_configure_rss_key(adapter);
1025 }
1026
1027 static int
1028 iavf_dev_rss_hash_conf_get(struct rte_eth_dev *dev,
1029                           struct rte_eth_rss_conf *rss_conf)
1030 {
1031         struct iavf_adapter *adapter =
1032                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1033         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
1034
1035         if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF))
1036                 return -ENOTSUP;
1037
1038          /* Just set it to default value now. */
1039         rss_conf->rss_hf = IAVF_RSS_OFFLOAD_ALL;
1040
1041         if (!rss_conf->rss_key)
1042                 return 0;
1043
1044         rss_conf->rss_key_len = vf->vf_res->rss_key_size;
1045         rte_memcpy(rss_conf->rss_key, vf->rss_key, rss_conf->rss_key_len);
1046
1047         return 0;
1048 }
1049
1050 static int
1051 iavf_dev_mtu_set(struct rte_eth_dev *dev, uint16_t mtu)
1052 {
1053         uint32_t frame_size = mtu + IAVF_ETH_OVERHEAD;
1054         int ret = 0;
1055
1056         if (mtu < RTE_ETHER_MIN_MTU || frame_size > IAVF_FRAME_SIZE_MAX)
1057                 return -EINVAL;
1058
1059         /* mtu setting is forbidden if port is start */
1060         if (dev->data->dev_started) {
1061                 PMD_DRV_LOG(ERR, "port must be stopped before configuration");
1062                 return -EBUSY;
1063         }
1064
1065         if (frame_size > RTE_ETHER_MAX_LEN)
1066                 dev->data->dev_conf.rxmode.offloads |=
1067                                 DEV_RX_OFFLOAD_JUMBO_FRAME;
1068         else
1069                 dev->data->dev_conf.rxmode.offloads &=
1070                                 ~DEV_RX_OFFLOAD_JUMBO_FRAME;
1071
1072         dev->data->dev_conf.rxmode.max_rx_pkt_len = frame_size;
1073
1074         return ret;
1075 }
1076
1077 static int
1078 iavf_dev_set_default_mac_addr(struct rte_eth_dev *dev,
1079                              struct rte_ether_addr *mac_addr)
1080 {
1081         struct iavf_adapter *adapter =
1082                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1083         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(adapter);
1084         struct rte_ether_addr *perm_addr, *old_addr;
1085         int ret;
1086
1087         old_addr = (struct rte_ether_addr *)hw->mac.addr;
1088         perm_addr = (struct rte_ether_addr *)hw->mac.perm_addr;
1089
1090         /* If the MAC address is configured by host, skip the setting */
1091         if (rte_is_valid_assigned_ether_addr(perm_addr))
1092                 return -EPERM;
1093
1094         ret = iavf_add_del_eth_addr(adapter, old_addr, false);
1095         if (ret)
1096                 PMD_DRV_LOG(ERR, "Fail to delete old MAC:"
1097                             " %02X:%02X:%02X:%02X:%02X:%02X",
1098                             old_addr->addr_bytes[0],
1099                             old_addr->addr_bytes[1],
1100                             old_addr->addr_bytes[2],
1101                             old_addr->addr_bytes[3],
1102                             old_addr->addr_bytes[4],
1103                             old_addr->addr_bytes[5]);
1104
1105         ret = iavf_add_del_eth_addr(adapter, mac_addr, true);
1106         if (ret)
1107                 PMD_DRV_LOG(ERR, "Fail to add new MAC:"
1108                             " %02X:%02X:%02X:%02X:%02X:%02X",
1109                             mac_addr->addr_bytes[0],
1110                             mac_addr->addr_bytes[1],
1111                             mac_addr->addr_bytes[2],
1112                             mac_addr->addr_bytes[3],
1113                             mac_addr->addr_bytes[4],
1114                             mac_addr->addr_bytes[5]);
1115
1116         if (ret)
1117                 return -EIO;
1118
1119         rte_ether_addr_copy(mac_addr, (struct rte_ether_addr *)hw->mac.addr);
1120         return 0;
1121 }
1122
1123 static void
1124 iavf_stat_update_48(uint64_t *offset, uint64_t *stat)
1125 {
1126         if (*stat >= *offset)
1127                 *stat = *stat - *offset;
1128         else
1129                 *stat = (uint64_t)((*stat +
1130                         ((uint64_t)1 << IAVF_48_BIT_WIDTH)) - *offset);
1131
1132         *stat &= IAVF_48_BIT_MASK;
1133 }
1134
1135 static void
1136 iavf_stat_update_32(uint64_t *offset, uint64_t *stat)
1137 {
1138         if (*stat >= *offset)
1139                 *stat = (uint64_t)(*stat - *offset);
1140         else
1141                 *stat = (uint64_t)((*stat +
1142                         ((uint64_t)1 << IAVF_32_BIT_WIDTH)) - *offset);
1143 }
1144
1145 static void
1146 iavf_update_stats(struct iavf_vsi *vsi, struct virtchnl_eth_stats *nes)
1147 {
1148         struct virtchnl_eth_stats *oes = &vsi->eth_stats_offset;
1149
1150         iavf_stat_update_48(&oes->rx_bytes, &nes->rx_bytes);
1151         iavf_stat_update_48(&oes->rx_unicast, &nes->rx_unicast);
1152         iavf_stat_update_48(&oes->rx_multicast, &nes->rx_multicast);
1153         iavf_stat_update_48(&oes->rx_broadcast, &nes->rx_broadcast);
1154         iavf_stat_update_32(&oes->rx_discards, &nes->rx_discards);
1155         iavf_stat_update_48(&oes->tx_bytes, &nes->tx_bytes);
1156         iavf_stat_update_48(&oes->tx_unicast, &nes->tx_unicast);
1157         iavf_stat_update_48(&oes->tx_multicast, &nes->tx_multicast);
1158         iavf_stat_update_48(&oes->tx_broadcast, &nes->tx_broadcast);
1159         iavf_stat_update_32(&oes->tx_errors, &nes->tx_errors);
1160         iavf_stat_update_32(&oes->tx_discards, &nes->tx_discards);
1161 }
1162
1163 static int
1164 iavf_dev_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
1165 {
1166         struct iavf_adapter *adapter =
1167                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1168         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
1169         struct iavf_vsi *vsi = &vf->vsi;
1170         struct virtchnl_eth_stats *pstats = NULL;
1171         int ret;
1172
1173         ret = iavf_query_stats(adapter, &pstats);
1174         if (ret == 0) {
1175                 iavf_update_stats(vsi, pstats);
1176                 stats->ipackets = pstats->rx_unicast + pstats->rx_multicast +
1177                                 pstats->rx_broadcast - pstats->rx_discards;
1178                 stats->opackets = pstats->tx_broadcast + pstats->tx_multicast +
1179                                                 pstats->tx_unicast;
1180                 stats->imissed = pstats->rx_discards;
1181                 stats->oerrors = pstats->tx_errors + pstats->tx_discards;
1182                 stats->ibytes = pstats->rx_bytes;
1183                 stats->ibytes -= stats->ipackets * RTE_ETHER_CRC_LEN;
1184                 stats->obytes = pstats->tx_bytes;
1185         } else {
1186                 PMD_DRV_LOG(ERR, "Get statistics failed");
1187         }
1188         return ret;
1189 }
1190
1191 static int
1192 iavf_dev_stats_reset(struct rte_eth_dev *dev)
1193 {
1194         int ret;
1195         struct iavf_adapter *adapter =
1196                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1197         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
1198         struct iavf_vsi *vsi = &vf->vsi;
1199         struct virtchnl_eth_stats *pstats = NULL;
1200
1201         /* read stat values to clear hardware registers */
1202         ret = iavf_query_stats(adapter, &pstats);
1203         if (ret != 0)
1204                 return ret;
1205
1206         /* set stats offset base on current values */
1207         vsi->eth_stats_offset = *pstats;
1208
1209         return 0;
1210 }
1211
1212 static int iavf_dev_xstats_get_names(__rte_unused struct rte_eth_dev *dev,
1213                                       struct rte_eth_xstat_name *xstats_names,
1214                                       __rte_unused unsigned int limit)
1215 {
1216         unsigned int i;
1217
1218         if (xstats_names != NULL)
1219                 for (i = 0; i < IAVF_NB_XSTATS; i++) {
1220                         snprintf(xstats_names[i].name,
1221                                 sizeof(xstats_names[i].name),
1222                                 "%s", rte_iavf_stats_strings[i].name);
1223                 }
1224         return IAVF_NB_XSTATS;
1225 }
1226
1227 static int iavf_dev_xstats_get(struct rte_eth_dev *dev,
1228                                  struct rte_eth_xstat *xstats, unsigned int n)
1229 {
1230         int ret;
1231         unsigned int i;
1232         struct iavf_adapter *adapter =
1233                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1234         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
1235         struct iavf_vsi *vsi = &vf->vsi;
1236         struct virtchnl_eth_stats *pstats = NULL;
1237
1238         if (n < IAVF_NB_XSTATS)
1239                 return IAVF_NB_XSTATS;
1240
1241         ret = iavf_query_stats(adapter, &pstats);
1242         if (ret != 0)
1243                 return 0;
1244
1245         if (!xstats)
1246                 return 0;
1247
1248         iavf_update_stats(vsi, pstats);
1249
1250         /* loop over xstats array and values from pstats */
1251         for (i = 0; i < IAVF_NB_XSTATS; i++) {
1252                 xstats[i].id = i;
1253                 xstats[i].value = *(uint64_t *)(((char *)pstats) +
1254                         rte_iavf_stats_strings[i].offset);
1255         }
1256
1257         return IAVF_NB_XSTATS;
1258 }
1259
1260
1261 static int
1262 iavf_dev_rx_queue_intr_enable(struct rte_eth_dev *dev, uint16_t queue_id)
1263 {
1264         struct iavf_adapter *adapter =
1265                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1266         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1267         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(adapter);
1268         uint16_t msix_intr;
1269
1270         msix_intr = pci_dev->intr_handle.intr_vec[queue_id];
1271         if (msix_intr == IAVF_MISC_VEC_ID) {
1272                 PMD_DRV_LOG(INFO, "MISC is also enabled for control");
1273                 IAVF_WRITE_REG(hw, IAVF_VFINT_DYN_CTL01,
1274                                IAVF_VFINT_DYN_CTL01_INTENA_MASK |
1275                                IAVF_VFINT_DYN_CTL01_CLEARPBA_MASK |
1276                                IAVF_VFINT_DYN_CTL01_ITR_INDX_MASK);
1277         } else {
1278                 IAVF_WRITE_REG(hw,
1279                                IAVF_VFINT_DYN_CTLN1
1280                                 (msix_intr - IAVF_RX_VEC_START),
1281                                IAVF_VFINT_DYN_CTLN1_INTENA_MASK |
1282                                IAVF_VFINT_DYN_CTL01_CLEARPBA_MASK |
1283                                IAVF_VFINT_DYN_CTLN1_ITR_INDX_MASK);
1284         }
1285
1286         IAVF_WRITE_FLUSH(hw);
1287
1288         rte_intr_ack(&pci_dev->intr_handle);
1289
1290         return 0;
1291 }
1292
1293 static int
1294 iavf_dev_rx_queue_intr_disable(struct rte_eth_dev *dev, uint16_t queue_id)
1295 {
1296         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1297         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1298         uint16_t msix_intr;
1299
1300         msix_intr = pci_dev->intr_handle.intr_vec[queue_id];
1301         if (msix_intr == IAVF_MISC_VEC_ID) {
1302                 PMD_DRV_LOG(ERR, "MISC is used for control, cannot disable it");
1303                 return -EIO;
1304         }
1305
1306         IAVF_WRITE_REG(hw,
1307                       IAVF_VFINT_DYN_CTLN1(msix_intr - IAVF_RX_VEC_START),
1308                       0);
1309
1310         IAVF_WRITE_FLUSH(hw);
1311         return 0;
1312 }
1313
1314 static int
1315 iavf_check_vf_reset_done(struct iavf_hw *hw)
1316 {
1317         int i, reset;
1318
1319         for (i = 0; i < IAVF_RESET_WAIT_CNT; i++) {
1320                 reset = IAVF_READ_REG(hw, IAVF_VFGEN_RSTAT) &
1321                         IAVF_VFGEN_RSTAT_VFR_STATE_MASK;
1322                 reset = reset >> IAVF_VFGEN_RSTAT_VFR_STATE_SHIFT;
1323                 if (reset == VIRTCHNL_VFR_VFACTIVE ||
1324                     reset == VIRTCHNL_VFR_COMPLETED)
1325                         break;
1326                 rte_delay_ms(20);
1327         }
1328
1329         if (i >= IAVF_RESET_WAIT_CNT)
1330                 return -1;
1331
1332         return 0;
1333 }
1334
1335 static int
1336 iavf_init_vf(struct rte_eth_dev *dev)
1337 {
1338         int err, bufsz;
1339         struct iavf_adapter *adapter =
1340                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1341         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1342         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
1343
1344         err = iavf_set_mac_type(hw);
1345         if (err) {
1346                 PMD_INIT_LOG(ERR, "set_mac_type failed: %d", err);
1347                 goto err;
1348         }
1349
1350         err = iavf_check_vf_reset_done(hw);
1351         if (err) {
1352                 PMD_INIT_LOG(ERR, "VF is still resetting");
1353                 goto err;
1354         }
1355
1356         iavf_init_adminq_parameter(hw);
1357         err = iavf_init_adminq(hw);
1358         if (err) {
1359                 PMD_INIT_LOG(ERR, "init_adminq failed: %d", err);
1360                 goto err;
1361         }
1362
1363         vf->aq_resp = rte_zmalloc("vf_aq_resp", IAVF_AQ_BUF_SZ, 0);
1364         if (!vf->aq_resp) {
1365                 PMD_INIT_LOG(ERR, "unable to allocate vf_aq_resp memory");
1366                 goto err_aq;
1367         }
1368         if (iavf_check_api_version(adapter) != 0) {
1369                 PMD_INIT_LOG(ERR, "check_api version failed");
1370                 goto err_api;
1371         }
1372
1373         bufsz = sizeof(struct virtchnl_vf_resource) +
1374                 (IAVF_MAX_VF_VSI * sizeof(struct virtchnl_vsi_resource));
1375         vf->vf_res = rte_zmalloc("vf_res", bufsz, 0);
1376         if (!vf->vf_res) {
1377                 PMD_INIT_LOG(ERR, "unable to allocate vf_res memory");
1378                 goto err_api;
1379         }
1380         if (iavf_get_vf_resource(adapter) != 0) {
1381                 PMD_INIT_LOG(ERR, "iavf_get_vf_config failed");
1382                 goto err_alloc;
1383         }
1384         /* Allocate memort for RSS info */
1385         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF) {
1386                 vf->rss_key = rte_zmalloc("rss_key",
1387                                           vf->vf_res->rss_key_size, 0);
1388                 if (!vf->rss_key) {
1389                         PMD_INIT_LOG(ERR, "unable to allocate rss_key memory");
1390                         goto err_rss;
1391                 }
1392                 vf->rss_lut = rte_zmalloc("rss_lut",
1393                                           vf->vf_res->rss_lut_size, 0);
1394                 if (!vf->rss_lut) {
1395                         PMD_INIT_LOG(ERR, "unable to allocate rss_lut memory");
1396                         goto err_rss;
1397                 }
1398         }
1399
1400         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RX_FLEX_DESC) {
1401                 if (iavf_get_supported_rxdid(adapter) != 0) {
1402                         PMD_INIT_LOG(ERR, "failed to do get supported rxdid");
1403                         goto err_rss;
1404                 }
1405         }
1406
1407         return 0;
1408 err_rss:
1409         rte_free(vf->rss_key);
1410         rte_free(vf->rss_lut);
1411 err_alloc:
1412         rte_free(vf->vf_res);
1413         vf->vsi_res = NULL;
1414 err_api:
1415         rte_free(vf->aq_resp);
1416 err_aq:
1417         iavf_shutdown_adminq(hw);
1418 err:
1419         return -1;
1420 }
1421
1422 /* Enable default admin queue interrupt setting */
1423 static inline void
1424 iavf_enable_irq0(struct iavf_hw *hw)
1425 {
1426         /* Enable admin queue interrupt trigger */
1427         IAVF_WRITE_REG(hw, IAVF_VFINT_ICR0_ENA1,
1428                        IAVF_VFINT_ICR0_ENA1_ADMINQ_MASK);
1429
1430         IAVF_WRITE_REG(hw, IAVF_VFINT_DYN_CTL01,
1431                        IAVF_VFINT_DYN_CTL01_INTENA_MASK |
1432                        IAVF_VFINT_DYN_CTL01_CLEARPBA_MASK |
1433                        IAVF_VFINT_DYN_CTL01_ITR_INDX_MASK);
1434
1435         IAVF_WRITE_FLUSH(hw);
1436 }
1437
1438 static inline void
1439 iavf_disable_irq0(struct iavf_hw *hw)
1440 {
1441         /* Disable all interrupt types */
1442         IAVF_WRITE_REG(hw, IAVF_VFINT_ICR0_ENA1, 0);
1443         IAVF_WRITE_REG(hw, IAVF_VFINT_DYN_CTL01,
1444                        IAVF_VFINT_DYN_CTL01_ITR_INDX_MASK);
1445         IAVF_WRITE_FLUSH(hw);
1446 }
1447
1448 static void
1449 iavf_dev_interrupt_handler(void *param)
1450 {
1451         struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
1452         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1453
1454         iavf_disable_irq0(hw);
1455
1456         iavf_handle_virtchnl_msg(dev);
1457
1458         iavf_enable_irq0(hw);
1459 }
1460
1461 static int
1462 iavf_dev_filter_ctrl(struct rte_eth_dev *dev,
1463                      enum rte_filter_type filter_type,
1464                      enum rte_filter_op filter_op,
1465                      void *arg)
1466 {
1467         int ret = 0;
1468
1469         if (!dev)
1470                 return -EINVAL;
1471
1472         switch (filter_type) {
1473         case RTE_ETH_FILTER_GENERIC:
1474                 if (filter_op != RTE_ETH_FILTER_GET)
1475                         return -EINVAL;
1476                 *(const void **)arg = &iavf_flow_ops;
1477                 break;
1478         default:
1479                 PMD_DRV_LOG(WARNING, "Filter type (%d) not supported",
1480                             filter_type);
1481                 ret = -EINVAL;
1482                 break;
1483         }
1484
1485         return ret;
1486 }
1487
1488
1489 static int
1490 iavf_dev_init(struct rte_eth_dev *eth_dev)
1491 {
1492         struct iavf_adapter *adapter =
1493                 IAVF_DEV_PRIVATE_TO_ADAPTER(eth_dev->data->dev_private);
1494         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(adapter);
1495         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
1496         int ret = 0;
1497
1498         PMD_INIT_FUNC_TRACE();
1499
1500         /* assign ops func pointer */
1501         eth_dev->dev_ops = &iavf_eth_dev_ops;
1502         eth_dev->rx_queue_count = iavf_dev_rxq_count;
1503         eth_dev->rx_descriptor_status = iavf_dev_rx_desc_status;
1504         eth_dev->tx_descriptor_status = iavf_dev_tx_desc_status;
1505         eth_dev->rx_pkt_burst = &iavf_recv_pkts;
1506         eth_dev->tx_pkt_burst = &iavf_xmit_pkts;
1507         eth_dev->tx_pkt_prepare = &iavf_prep_pkts;
1508
1509         /* For secondary processes, we don't initialise any further as primary
1510          * has already done this work. Only check if we need a different RX
1511          * and TX function.
1512          */
1513         if (rte_eal_process_type() != RTE_PROC_PRIMARY) {
1514                 iavf_set_rx_function(eth_dev);
1515                 iavf_set_tx_function(eth_dev);
1516                 return 0;
1517         }
1518         rte_eth_copy_pci_info(eth_dev, pci_dev);
1519         eth_dev->data->dev_flags |= RTE_ETH_DEV_AUTOFILL_QUEUE_XSTATS;
1520
1521         hw->vendor_id = pci_dev->id.vendor_id;
1522         hw->device_id = pci_dev->id.device_id;
1523         hw->subsystem_vendor_id = pci_dev->id.subsystem_vendor_id;
1524         hw->subsystem_device_id = pci_dev->id.subsystem_device_id;
1525         hw->bus.bus_id = pci_dev->addr.bus;
1526         hw->bus.device = pci_dev->addr.devid;
1527         hw->bus.func = pci_dev->addr.function;
1528         hw->hw_addr = (void *)pci_dev->mem_resource[0].addr;
1529         hw->back = IAVF_DEV_PRIVATE_TO_ADAPTER(eth_dev->data->dev_private);
1530         adapter->eth_dev = eth_dev;
1531         adapter->stopped = 1;
1532
1533         if (iavf_init_vf(eth_dev) != 0) {
1534                 PMD_INIT_LOG(ERR, "Init vf failed");
1535                 return -1;
1536         }
1537
1538         /* set default ptype table */
1539         adapter->ptype_tbl = iavf_get_default_ptype_table();
1540
1541         /* copy mac addr */
1542         eth_dev->data->mac_addrs = rte_zmalloc(
1543                 "iavf_mac", RTE_ETHER_ADDR_LEN * IAVF_NUM_MACADDR_MAX, 0);
1544         if (!eth_dev->data->mac_addrs) {
1545                 PMD_INIT_LOG(ERR, "Failed to allocate %d bytes needed to"
1546                              " store MAC addresses",
1547                              RTE_ETHER_ADDR_LEN * IAVF_NUM_MACADDR_MAX);
1548                 return -ENOMEM;
1549         }
1550         /* If the MAC address is not configured by host,
1551          * generate a random one.
1552          */
1553         if (!rte_is_valid_assigned_ether_addr(
1554                         (struct rte_ether_addr *)hw->mac.addr))
1555                 rte_eth_random_addr(hw->mac.addr);
1556         rte_ether_addr_copy((struct rte_ether_addr *)hw->mac.addr,
1557                         &eth_dev->data->mac_addrs[0]);
1558
1559         /* register callback func to eal lib */
1560         rte_intr_callback_register(&pci_dev->intr_handle,
1561                                    iavf_dev_interrupt_handler,
1562                                    (void *)eth_dev);
1563
1564         /* enable uio intr after callback register */
1565         rte_intr_enable(&pci_dev->intr_handle);
1566
1567         /* configure and enable device interrupt */
1568         iavf_enable_irq0(hw);
1569
1570         ret = iavf_flow_init(adapter);
1571         if (ret) {
1572                 PMD_INIT_LOG(ERR, "Failed to initialize flow");
1573                 return ret;
1574         }
1575
1576         return 0;
1577 }
1578
1579 static int
1580 iavf_dev_close(struct rte_eth_dev *dev)
1581 {
1582         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1583         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1584         struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
1585         struct iavf_adapter *adapter =
1586                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1587         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
1588         int ret;
1589
1590         if (rte_eal_process_type() != RTE_PROC_PRIMARY)
1591                 return 0;
1592
1593         ret = iavf_dev_stop(dev);
1594
1595         iavf_flow_flush(dev, NULL);
1596         iavf_flow_uninit(adapter);
1597
1598         /*
1599          * disable promiscuous mode before reset vf
1600          * it is a workaround solution when work with kernel driver
1601          * and it is not the normal way
1602          */
1603         if (vf->promisc_unicast_enabled || vf->promisc_multicast_enabled)
1604                 iavf_config_promisc(adapter, false, false);
1605
1606         iavf_shutdown_adminq(hw);
1607         /* disable uio intr before callback unregister */
1608         rte_intr_disable(intr_handle);
1609
1610         /* unregister callback func from eal lib */
1611         rte_intr_callback_unregister(intr_handle,
1612                                      iavf_dev_interrupt_handler, dev);
1613         iavf_disable_irq0(hw);
1614
1615         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF) {
1616                 if (vf->rss_lut) {
1617                         rte_free(vf->rss_lut);
1618                         vf->rss_lut = NULL;
1619                 }
1620                 if (vf->rss_key) {
1621                         rte_free(vf->rss_key);
1622                         vf->rss_key = NULL;
1623                 }
1624         }
1625
1626         rte_free(vf->vf_res);
1627         vf->vsi_res = NULL;
1628         vf->vf_res = NULL;
1629
1630         rte_free(vf->aq_resp);
1631         vf->aq_resp = NULL;
1632
1633         vf->vf_reset = false;
1634
1635         return ret;
1636 }
1637
1638 static int
1639 iavf_dev_uninit(struct rte_eth_dev *dev)
1640 {
1641         if (rte_eal_process_type() != RTE_PROC_PRIMARY)
1642                 return -EPERM;
1643
1644         iavf_dev_close(dev);
1645
1646         return 0;
1647 }
1648
1649 /*
1650  * Reset VF device only to re-initialize resources in PMD layer
1651  */
1652 static int
1653 iavf_dev_reset(struct rte_eth_dev *dev)
1654 {
1655         int ret;
1656
1657         ret = iavf_dev_uninit(dev);
1658         if (ret)
1659                 return ret;
1660
1661         return iavf_dev_init(dev);
1662 }
1663
1664 static int
1665 iavf_dcf_cap_check_handler(__rte_unused const char *key,
1666                            const char *value, __rte_unused void *opaque)
1667 {
1668         if (strcmp(value, "dcf"))
1669                 return -1;
1670
1671         return 0;
1672 }
1673
1674 static int
1675 iavf_dcf_cap_selected(struct rte_devargs *devargs)
1676 {
1677         struct rte_kvargs *kvlist;
1678         const char *key = "cap";
1679         int ret = 0;
1680
1681         if (devargs == NULL)
1682                 return 0;
1683
1684         kvlist = rte_kvargs_parse(devargs->args, NULL);
1685         if (kvlist == NULL)
1686                 return 0;
1687
1688         if (!rte_kvargs_count(kvlist, key))
1689                 goto exit;
1690
1691         /* dcf capability selected when there's a key-value pair: cap=dcf */
1692         if (rte_kvargs_process(kvlist, key,
1693                                iavf_dcf_cap_check_handler, NULL) < 0)
1694                 goto exit;
1695
1696         ret = 1;
1697
1698 exit:
1699         rte_kvargs_free(kvlist);
1700         return ret;
1701 }
1702
1703 static int eth_iavf_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
1704                              struct rte_pci_device *pci_dev)
1705 {
1706         if (iavf_dcf_cap_selected(pci_dev->device.devargs))
1707                 return 1;
1708
1709         return rte_eth_dev_pci_generic_probe(pci_dev,
1710                 sizeof(struct iavf_adapter), iavf_dev_init);
1711 }
1712
1713 static int eth_iavf_pci_remove(struct rte_pci_device *pci_dev)
1714 {
1715         return rte_eth_dev_pci_generic_remove(pci_dev, iavf_dev_uninit);
1716 }
1717
1718 /* Adaptive virtual function driver struct */
1719 static struct rte_pci_driver rte_iavf_pmd = {
1720         .id_table = pci_id_iavf_map,
1721         .drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC,
1722         .probe = eth_iavf_pci_probe,
1723         .remove = eth_iavf_pci_remove,
1724 };
1725
1726 RTE_PMD_REGISTER_PCI(net_iavf, rte_iavf_pmd);
1727 RTE_PMD_REGISTER_PCI_TABLE(net_iavf, pci_id_iavf_map);
1728 RTE_PMD_REGISTER_KMOD_DEP(net_iavf, "* igb_uio | vfio-pci");
1729 RTE_PMD_REGISTER_PARAM_STRING(net_iavf, "cap=dcf");
1730 RTE_LOG_REGISTER(iavf_logtype_init, pmd.net.iavf.init, NOTICE);
1731 RTE_LOG_REGISTER(iavf_logtype_driver, pmd.net.iavf.driver, NOTICE);
1732 #ifdef RTE_LIBRTE_IAVF_DEBUG_RX
1733 RTE_LOG_REGISTER(iavf_logtype_rx, pmd.net.iavf.rx, DEBUG);
1734 #endif
1735 #ifdef RTE_LIBRTE_IAVF_DEBUG_TX
1736 RTE_LOG_REGISTER(iavf_logtype_tx, pmd.net.iavf.tx, DEBUG);
1737 #endif
1738 #ifdef RTE_LIBRTE_IAVF_DEBUG_TX_FREE
1739 RTE_LOG_REGISTER(iavf_logtype_tx_free, pmd.net.iavf.tx_free, DEBUG);
1740 #endif