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         uint16_t num_queue_pairs;
 568         uint16_t index = 0;
 569
 570         PMD_INIT_FUNC_TRACE();
 571
 572         adapter->stopped = 0;
 573
 574         vf->max_pkt_len = dev->data->dev_conf.rxmode.max_rx_pkt_len;
 575         vf->num_queue_pairs = RTE_MAX(dev->data->nb_rx_queues,
 576                                       dev->data->nb_tx_queues);
 577         num_queue_pairs = vf->num_queue_pairs;
 578
 579         if (iavf_init_queues(dev) != 0) {
 580                 PMD_DRV_LOG(ERR, "failed to do Queue init");
 581                 return -1;
 582         }
 583
 584         /* If needed, send configure queues msg multiple times to make the
 585          * adminq buffer length smaller than the 4K limitation.
 586          */
 587         while (num_queue_pairs > IAVF_CFG_Q_NUM_PER_BUF) {
 588                 if (iavf_configure_queues(adapter,
 589                                 IAVF_CFG_Q_NUM_PER_BUF, index) != 0) {
 590                         PMD_DRV_LOG(ERR, "configure queues failed");
 591                         goto err_queue;
 592                 }
 593                 num_queue_pairs -= IAVF_CFG_Q_NUM_PER_BUF;
 594                 index += IAVF_CFG_Q_NUM_PER_BUF;
 595         }
 596
 597         if (iavf_configure_queues(adapter, num_queue_pairs, index) != 0) {
 598                 PMD_DRV_LOG(ERR, "configure queues failed");
 599                 goto err_queue;
 600         }
 601
 602         if (iavf_config_rx_queues_irqs(dev, intr_handle) != 0) {
 603                 PMD_DRV_LOG(ERR, "configure irq failed");
 604                 goto err_queue;
 605         }
 606         /* re-enable intr again, because efd assign may change */
 607         if (dev->data->dev_conf.intr_conf.rxq != 0) {
 608                 rte_intr_disable(intr_handle);
 609                 rte_intr_enable(intr_handle);
 610         }
 611
 612         /* Set all mac addrs */
 613         iavf_add_del_all_mac_addr(adapter, true);
 614
 615         /* Set all multicast addresses */
 616         iavf_add_del_mc_addr_list(adapter, vf->mc_addrs, vf->mc_addrs_num,
 617                                   true);
 618
 619         if (iavf_start_queues(dev) != 0) {
 620                 PMD_DRV_LOG(ERR, "enable queues failed");
 621                 goto err_mac;
 622         }
 623
 624         return 0;
 625
 626 err_mac:
 627         iavf_add_del_all_mac_addr(adapter, false);
 628 err_queue:
 629         return -1;
 630 }
 631
 632 static int
 633 iavf_dev_stop(struct rte_eth_dev *dev)
 634 {
 635         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
 636         struct iavf_adapter *adapter =
 637                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 638         struct rte_intr_handle *intr_handle = dev->intr_handle;
 639
 640         PMD_INIT_FUNC_TRACE();
 641
 642         if (adapter->stopped == 1)
 643                 return 0;
 644
 645         iavf_stop_queues(dev);
 646
 647         /* Disable the interrupt for Rx */
 648         rte_intr_efd_disable(intr_handle);
 649         /* Rx interrupt vector mapping free */
 650         if (intr_handle->intr_vec) {
 651                 rte_free(intr_handle->intr_vec);
 652                 intr_handle->intr_vec = NULL;
 653         }
 654
 655         /* remove all mac addrs */
 656         iavf_add_del_all_mac_addr(adapter, false);
 657
 658         /* remove all multicast addresses */
 659         iavf_add_del_mc_addr_list(adapter, vf->mc_addrs, vf->mc_addrs_num,
 660                                   false);
 661
 662         adapter->stopped = 1;
 663         dev->data->dev_started = 0;
 664
 665         return 0;
 666 }
 667
 668 static int
 669 iavf_dev_info_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
 670 {
 671         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
 672
 673         dev_info->max_rx_queues = IAVF_MAX_NUM_QUEUES_LV;
 674         dev_info->max_tx_queues = IAVF_MAX_NUM_QUEUES_LV;
 675         dev_info->min_rx_bufsize = IAVF_BUF_SIZE_MIN;
 676         dev_info->max_rx_pktlen = IAVF_FRAME_SIZE_MAX;
 677         dev_info->max_mtu = dev_info->max_rx_pktlen - IAVF_ETH_OVERHEAD;
 678         dev_info->min_mtu = RTE_ETHER_MIN_MTU;
 679         dev_info->hash_key_size = vf->vf_res->rss_key_size;
 680         dev_info->reta_size = vf->vf_res->rss_lut_size;
 681         dev_info->flow_type_rss_offloads = IAVF_RSS_OFFLOAD_ALL;
 682         dev_info->max_mac_addrs = IAVF_NUM_MACADDR_MAX;
 683         dev_info->rx_offload_capa =
 684                 DEV_RX_OFFLOAD_VLAN_STRIP |
 685                 DEV_RX_OFFLOAD_QINQ_STRIP |
 686                 DEV_RX_OFFLOAD_IPV4_CKSUM |
 687                 DEV_RX_OFFLOAD_UDP_CKSUM |
 688                 DEV_RX_OFFLOAD_TCP_CKSUM |
 689                 DEV_RX_OFFLOAD_OUTER_IPV4_CKSUM |
 690                 DEV_RX_OFFLOAD_SCATTER |
 691                 DEV_RX_OFFLOAD_JUMBO_FRAME |
 692                 DEV_RX_OFFLOAD_VLAN_FILTER |
 693                 DEV_RX_OFFLOAD_RSS_HASH;
 694         dev_info->tx_offload_capa =
 695                 DEV_TX_OFFLOAD_VLAN_INSERT |
 696                 DEV_TX_OFFLOAD_QINQ_INSERT |
 697                 DEV_TX_OFFLOAD_IPV4_CKSUM |
 698                 DEV_TX_OFFLOAD_UDP_CKSUM |
 699                 DEV_TX_OFFLOAD_TCP_CKSUM |
 700                 DEV_TX_OFFLOAD_SCTP_CKSUM |
 701                 DEV_TX_OFFLOAD_OUTER_IPV4_CKSUM |
 702                 DEV_TX_OFFLOAD_TCP_TSO |
 703                 DEV_TX_OFFLOAD_VXLAN_TNL_TSO |
 704                 DEV_TX_OFFLOAD_GRE_TNL_TSO |
 705                 DEV_TX_OFFLOAD_IPIP_TNL_TSO |
 706                 DEV_TX_OFFLOAD_GENEVE_TNL_TSO |
 707                 DEV_TX_OFFLOAD_MULTI_SEGS;
 708
 709         dev_info->default_rxconf = (struct rte_eth_rxconf) {
 710                 .rx_free_thresh = IAVF_DEFAULT_RX_FREE_THRESH,
 711                 .rx_drop_en = 0,
 712                 .offloads = 0,
 713         };
 714
 715         dev_info->default_txconf = (struct rte_eth_txconf) {
 716                 .tx_free_thresh = IAVF_DEFAULT_TX_FREE_THRESH,
 717                 .tx_rs_thresh = IAVF_DEFAULT_TX_RS_THRESH,
 718                 .offloads = 0,
 719         };
 720
 721         dev_info->rx_desc_lim = (struct rte_eth_desc_lim) {
 722                 .nb_max = IAVF_MAX_RING_DESC,
 723                 .nb_min = IAVF_MIN_RING_DESC,
 724                 .nb_align = IAVF_ALIGN_RING_DESC,
 725         };
 726
 727         dev_info->tx_desc_lim = (struct rte_eth_desc_lim) {
 728                 .nb_max = IAVF_MAX_RING_DESC,
 729                 .nb_min = IAVF_MIN_RING_DESC,
 730                 .nb_align = IAVF_ALIGN_RING_DESC,
 731         };
 732
 733         return 0;
 734 }
 735
 736 static const uint32_t *
 737 iavf_dev_supported_ptypes_get(struct rte_eth_dev *dev __rte_unused)
 738 {
 739         static const uint32_t ptypes[] = {
 740                 RTE_PTYPE_L2_ETHER,
 741                 RTE_PTYPE_L3_IPV4_EXT_UNKNOWN,
 742                 RTE_PTYPE_L4_FRAG,
 743                 RTE_PTYPE_L4_ICMP,
 744                 RTE_PTYPE_L4_NONFRAG,
 745                 RTE_PTYPE_L4_SCTP,
 746                 RTE_PTYPE_L4_TCP,
 747                 RTE_PTYPE_L4_UDP,
 748                 RTE_PTYPE_UNKNOWN
 749         };
 750         return ptypes;
 751 }
 752
 753 int
 754 iavf_dev_link_update(struct rte_eth_dev *dev,
 755                     __rte_unused int wait_to_complete)
 756 {
 757         struct rte_eth_link new_link;
 758         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
 759
 760         memset(&new_link, 0, sizeof(new_link));
 761
 762         /* Only read status info stored in VF, and the info is updated
 763          *  when receive LINK_CHANGE evnet from PF by Virtchnnl.
 764          */
 765         switch (vf->link_speed) {
 766         case 10:
 767                 new_link.link_speed = ETH_SPEED_NUM_10M;
 768                 break;
 769         case 100:
 770                 new_link.link_speed = ETH_SPEED_NUM_100M;
 771                 break;
 772         case 1000:
 773                 new_link.link_speed = ETH_SPEED_NUM_1G;
 774                 break;
 775         case 10000:
 776                 new_link.link_speed = ETH_SPEED_NUM_10G;
 777                 break;
 778         case 20000:
 779                 new_link.link_speed = ETH_SPEED_NUM_20G;
 780                 break;
 781         case 25000:
 782                 new_link.link_speed = ETH_SPEED_NUM_25G;
 783                 break;
 784         case 40000:
 785                 new_link.link_speed = ETH_SPEED_NUM_40G;
 786                 break;
 787         case 50000:
 788                 new_link.link_speed = ETH_SPEED_NUM_50G;
 789                 break;
 790         case 100000:
 791                 new_link.link_speed = ETH_SPEED_NUM_100G;
 792                 break;
 793         default:
 794                 new_link.link_speed = ETH_SPEED_NUM_NONE;
 795                 break;
 796         }
 797
 798         new_link.link_duplex = ETH_LINK_FULL_DUPLEX;
 799         new_link.link_status = vf->link_up ? ETH_LINK_UP :
 800                                              ETH_LINK_DOWN;
 801         new_link.link_autoneg = !(dev->data->dev_conf.link_speeds &
 802                                 ETH_LINK_SPEED_FIXED);
 803
 804         return rte_eth_linkstatus_set(dev, &new_link);
 805 }
 806
 807 static int
 808 iavf_dev_promiscuous_enable(struct rte_eth_dev *dev)
 809 {
 810         struct iavf_adapter *adapter =
 811                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 812         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
 813
 814         return iavf_config_promisc(adapter,
 815                                   true, vf->promisc_multicast_enabled);
 816 }
 817
 818 static int
 819 iavf_dev_promiscuous_disable(struct rte_eth_dev *dev)
 820 {
 821         struct iavf_adapter *adapter =
 822                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 823         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
 824
 825         return iavf_config_promisc(adapter,
 826                                   false, vf->promisc_multicast_enabled);
 827 }
 828
 829 static int
 830 iavf_dev_allmulticast_enable(struct rte_eth_dev *dev)
 831 {
 832         struct iavf_adapter *adapter =
 833                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 834         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
 835
 836         return iavf_config_promisc(adapter,
 837                                   vf->promisc_unicast_enabled, true);
 838 }
 839
 840 static int
 841 iavf_dev_allmulticast_disable(struct rte_eth_dev *dev)
 842 {
 843         struct iavf_adapter *adapter =
 844                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 845         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
 846
 847         return iavf_config_promisc(adapter,
 848                                   vf->promisc_unicast_enabled, false);
 849 }
 850
 851 static int
 852 iavf_dev_add_mac_addr(struct rte_eth_dev *dev, struct rte_ether_addr *addr,
 853                      __rte_unused uint32_t index,
 854                      __rte_unused uint32_t pool)
 855 {
 856         struct iavf_adapter *adapter =
 857                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 858         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
 859         int err;
 860
 861         if (rte_is_zero_ether_addr(addr)) {
 862                 PMD_DRV_LOG(ERR, "Invalid Ethernet Address");
 863                 return -EINVAL;
 864         }
 865
 866         err = iavf_add_del_eth_addr(adapter, addr, true);
 867         if (err) {
 868                 PMD_DRV_LOG(ERR, "fail to add MAC address");
 869                 return -EIO;
 870         }
 871
 872         vf->mac_num++;
 873
 874         return 0;
 875 }
 876
 877 static void
 878 iavf_dev_del_mac_addr(struct rte_eth_dev *dev, uint32_t index)
 879 {
 880         struct iavf_adapter *adapter =
 881                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 882         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
 883         struct rte_ether_addr *addr;
 884         int err;
 885
 886         addr = &dev->data->mac_addrs[index];
 887
 888         err = iavf_add_del_eth_addr(adapter, addr, false);
 889         if (err)
 890                 PMD_DRV_LOG(ERR, "fail to delete MAC address");
 891
 892         vf->mac_num--;
 893 }
 894
 895 static int
 896 iavf_dev_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
 897 {
 898         struct iavf_adapter *adapter =
 899                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 900         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
 901         int err;
 902
 903         if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_VLAN))
 904                 return -ENOTSUP;
 905
 906         err = iavf_add_del_vlan(adapter, vlan_id, on);
 907         if (err)
 908                 return -EIO;
 909         return 0;
 910 }
 911
 912 static int
 913 iavf_dev_vlan_offload_set(struct rte_eth_dev *dev, int mask)
 914 {
 915         struct iavf_adapter *adapter =
 916                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 917         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
 918         struct rte_eth_conf *dev_conf = &dev->data->dev_conf;
 919         int err;
 920
 921         if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_VLAN))
 922                 return -ENOTSUP;
 923
 924         /* Vlan stripping setting */
 925         if (mask & ETH_VLAN_STRIP_MASK) {
 926                 /* Enable or disable VLAN stripping */
 927                 if (dev_conf->rxmode.offloads & DEV_RX_OFFLOAD_VLAN_STRIP)
 928                         err = iavf_enable_vlan_strip(adapter);
 929                 else
 930                         err = iavf_disable_vlan_strip(adapter);
 931
 932                 if (err)
 933                         return -EIO;
 934         }
 935         return 0;
 936 }
 937
 938 static int
 939 iavf_dev_rss_reta_update(struct rte_eth_dev *dev,
 940                         struct rte_eth_rss_reta_entry64 *reta_conf,
 941                         uint16_t reta_size)
 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         uint8_t *lut;
 947         uint16_t i, idx, shift;
 948         int ret;
 949
 950         if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF))
 951                 return -ENOTSUP;
 952
 953         if (reta_size != vf->vf_res->rss_lut_size) {
 954                 PMD_DRV_LOG(ERR, "The size of hash lookup table configured "
 955                         "(%d) doesn't match the number of hardware can "
 956                         "support (%d)", reta_size, vf->vf_res->rss_lut_size);
 957                 return -EINVAL;
 958         }
 959
 960         lut = rte_zmalloc("rss_lut", reta_size, 0);
 961         if (!lut) {
 962                 PMD_DRV_LOG(ERR, "No memory can be allocated");
 963                 return -ENOMEM;
 964         }
 965         /* store the old lut table temporarily */
 966         rte_memcpy(lut, vf->rss_lut, reta_size);
 967
 968         for (i = 0; i < reta_size; i++) {
 969                 idx = i / RTE_RETA_GROUP_SIZE;
 970                 shift = i % RTE_RETA_GROUP_SIZE;
 971                 if (reta_conf[idx].mask & (1ULL << shift))
 972                         lut[i] = reta_conf[idx].reta[shift];
 973         }
 974
 975         rte_memcpy(vf->rss_lut, lut, reta_size);
 976         /* send virtchnnl ops to configure rss*/
 977         ret = iavf_configure_rss_lut(adapter);
 978         if (ret) /* revert back */
 979                 rte_memcpy(vf->rss_lut, lut, reta_size);
 980         rte_free(lut);
 981
 982         return ret;
 983 }
 984
 985 static int
 986 iavf_dev_rss_reta_query(struct rte_eth_dev *dev,
 987                        struct rte_eth_rss_reta_entry64 *reta_conf,
 988                        uint16_t reta_size)
 989 {
 990         struct iavf_adapter *adapter =
 991                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 992         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
 993         uint16_t i, idx, shift;
 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         for (i = 0; i < reta_size; i++) {
1006                 idx = i / RTE_RETA_GROUP_SIZE;
1007                 shift = i % RTE_RETA_GROUP_SIZE;
1008                 if (reta_conf[idx].mask & (1ULL << shift))
1009                         reta_conf[idx].reta[shift] = vf->rss_lut[i];
1010         }
1011
1012         return 0;
1013 }
1014
1015 static int
1016 iavf_dev_rss_hash_update(struct rte_eth_dev *dev,
1017                         struct rte_eth_rss_conf *rss_conf)
1018 {
1019         struct iavf_adapter *adapter =
1020                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1021         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
1022
1023         if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF))
1024                 return -ENOTSUP;
1025
1026         /* HENA setting, it is enabled by default, no change */
1027         if (!rss_conf->rss_key || rss_conf->rss_key_len == 0) {
1028                 PMD_DRV_LOG(DEBUG, "No key to be configured");
1029                 return 0;
1030         } else if (rss_conf->rss_key_len != vf->vf_res->rss_key_size) {
1031                 PMD_DRV_LOG(ERR, "The size of hash key configured "
1032                         "(%d) doesn't match the size of hardware can "
1033                         "support (%d)", rss_conf->rss_key_len,
1034                         vf->vf_res->rss_key_size);
1035                 return -EINVAL;
1036         }
1037
1038         rte_memcpy(vf->rss_key, rss_conf->rss_key, rss_conf->rss_key_len);
1039
1040         return iavf_configure_rss_key(adapter);
1041 }
1042
1043 static int
1044 iavf_dev_rss_hash_conf_get(struct rte_eth_dev *dev,
1045                           struct rte_eth_rss_conf *rss_conf)
1046 {
1047         struct iavf_adapter *adapter =
1048                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1049         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
1050
1051         if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF))
1052                 return -ENOTSUP;
1053
1054          /* Just set it to default value now. */
1055         rss_conf->rss_hf = IAVF_RSS_OFFLOAD_ALL;
1056
1057         if (!rss_conf->rss_key)
1058                 return 0;
1059
1060         rss_conf->rss_key_len = vf->vf_res->rss_key_size;
1061         rte_memcpy(rss_conf->rss_key, vf->rss_key, rss_conf->rss_key_len);
1062
1063         return 0;
1064 }
1065
1066 static int
1067 iavf_dev_mtu_set(struct rte_eth_dev *dev, uint16_t mtu)
1068 {
1069         uint32_t frame_size = mtu + IAVF_ETH_OVERHEAD;
1070         int ret = 0;
1071
1072         if (mtu < RTE_ETHER_MIN_MTU || frame_size > IAVF_FRAME_SIZE_MAX)
1073                 return -EINVAL;
1074
1075         /* mtu setting is forbidden if port is start */
1076         if (dev->data->dev_started) {
1077                 PMD_DRV_LOG(ERR, "port must be stopped before configuration");
1078                 return -EBUSY;
1079         }
1080
1081         if (frame_size > RTE_ETHER_MAX_LEN)
1082                 dev->data->dev_conf.rxmode.offloads |=
1083                                 DEV_RX_OFFLOAD_JUMBO_FRAME;
1084         else
1085                 dev->data->dev_conf.rxmode.offloads &=
1086                                 ~DEV_RX_OFFLOAD_JUMBO_FRAME;
1087
1088         dev->data->dev_conf.rxmode.max_rx_pkt_len = frame_size;
1089
1090         return ret;
1091 }
1092
1093 static int
1094 iavf_dev_set_default_mac_addr(struct rte_eth_dev *dev,
1095                              struct rte_ether_addr *mac_addr)
1096 {
1097         struct iavf_adapter *adapter =
1098                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1099         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(adapter);
1100         struct rte_ether_addr *perm_addr, *old_addr;
1101         int ret;
1102
1103         old_addr = (struct rte_ether_addr *)hw->mac.addr;
1104         perm_addr = (struct rte_ether_addr *)hw->mac.perm_addr;
1105
1106         /* If the MAC address is configured by host, skip the setting */
1107         if (rte_is_valid_assigned_ether_addr(perm_addr))
1108                 return -EPERM;
1109
1110         ret = iavf_add_del_eth_addr(adapter, old_addr, false);
1111         if (ret)
1112                 PMD_DRV_LOG(ERR, "Fail to delete old MAC:"
1113                             " %02X:%02X:%02X:%02X:%02X:%02X",
1114                             old_addr->addr_bytes[0],
1115                             old_addr->addr_bytes[1],
1116                             old_addr->addr_bytes[2],
1117                             old_addr->addr_bytes[3],
1118                             old_addr->addr_bytes[4],
1119                             old_addr->addr_bytes[5]);
1120
1121         ret = iavf_add_del_eth_addr(adapter, mac_addr, true);
1122         if (ret)
1123                 PMD_DRV_LOG(ERR, "Fail to add new MAC:"
1124                             " %02X:%02X:%02X:%02X:%02X:%02X",
1125                             mac_addr->addr_bytes[0],
1126                             mac_addr->addr_bytes[1],
1127                             mac_addr->addr_bytes[2],
1128                             mac_addr->addr_bytes[3],
1129                             mac_addr->addr_bytes[4],
1130                             mac_addr->addr_bytes[5]);
1131
1132         if (ret)
1133                 return -EIO;
1134
1135         rte_ether_addr_copy(mac_addr, (struct rte_ether_addr *)hw->mac.addr);
1136         return 0;
1137 }
1138
1139 static void
1140 iavf_stat_update_48(uint64_t *offset, uint64_t *stat)
1141 {
1142         if (*stat >= *offset)
1143                 *stat = *stat - *offset;
1144         else
1145                 *stat = (uint64_t)((*stat +
1146                         ((uint64_t)1 << IAVF_48_BIT_WIDTH)) - *offset);
1147
1148         *stat &= IAVF_48_BIT_MASK;
1149 }
1150
1151 static void
1152 iavf_stat_update_32(uint64_t *offset, uint64_t *stat)
1153 {
1154         if (*stat >= *offset)
1155                 *stat = (uint64_t)(*stat - *offset);
1156         else
1157                 *stat = (uint64_t)((*stat +
1158                         ((uint64_t)1 << IAVF_32_BIT_WIDTH)) - *offset);
1159 }
1160
1161 static void
1162 iavf_update_stats(struct iavf_vsi *vsi, struct virtchnl_eth_stats *nes)
1163 {
1164         struct virtchnl_eth_stats *oes = &vsi->eth_stats_offset;
1165
1166         iavf_stat_update_48(&oes->rx_bytes, &nes->rx_bytes);
1167         iavf_stat_update_48(&oes->rx_unicast, &nes->rx_unicast);
1168         iavf_stat_update_48(&oes->rx_multicast, &nes->rx_multicast);
1169         iavf_stat_update_48(&oes->rx_broadcast, &nes->rx_broadcast);
1170         iavf_stat_update_32(&oes->rx_discards, &nes->rx_discards);
1171         iavf_stat_update_48(&oes->tx_bytes, &nes->tx_bytes);
1172         iavf_stat_update_48(&oes->tx_unicast, &nes->tx_unicast);
1173         iavf_stat_update_48(&oes->tx_multicast, &nes->tx_multicast);
1174         iavf_stat_update_48(&oes->tx_broadcast, &nes->tx_broadcast);
1175         iavf_stat_update_32(&oes->tx_errors, &nes->tx_errors);
1176         iavf_stat_update_32(&oes->tx_discards, &nes->tx_discards);
1177 }
1178
1179 static int
1180 iavf_dev_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
1181 {
1182         struct iavf_adapter *adapter =
1183                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1184         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
1185         struct iavf_vsi *vsi = &vf->vsi;
1186         struct virtchnl_eth_stats *pstats = NULL;
1187         int ret;
1188
1189         ret = iavf_query_stats(adapter, &pstats);
1190         if (ret == 0) {
1191                 iavf_update_stats(vsi, pstats);
1192                 stats->ipackets = pstats->rx_unicast + pstats->rx_multicast +
1193                                 pstats->rx_broadcast - pstats->rx_discards;
1194                 stats->opackets = pstats->tx_broadcast + pstats->tx_multicast +
1195                                                 pstats->tx_unicast;
1196                 stats->imissed = pstats->rx_discards;
1197                 stats->oerrors = pstats->tx_errors + pstats->tx_discards;
1198                 stats->ibytes = pstats->rx_bytes;
1199                 stats->ibytes -= stats->ipackets * RTE_ETHER_CRC_LEN;
1200                 stats->obytes = pstats->tx_bytes;
1201         } else {
1202                 PMD_DRV_LOG(ERR, "Get statistics failed");
1203         }
1204         return ret;
1205 }
1206
1207 static int
1208 iavf_dev_stats_reset(struct rte_eth_dev *dev)
1209 {
1210         int ret;
1211         struct iavf_adapter *adapter =
1212                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1213         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
1214         struct iavf_vsi *vsi = &vf->vsi;
1215         struct virtchnl_eth_stats *pstats = NULL;
1216
1217         /* read stat values to clear hardware registers */
1218         ret = iavf_query_stats(adapter, &pstats);
1219         if (ret != 0)
1220                 return ret;
1221
1222         /* set stats offset base on current values */
1223         vsi->eth_stats_offset = *pstats;
1224
1225         return 0;
1226 }
1227
1228 static int iavf_dev_xstats_get_names(__rte_unused struct rte_eth_dev *dev,
1229                                       struct rte_eth_xstat_name *xstats_names,
1230                                       __rte_unused unsigned int limit)
1231 {
1232         unsigned int i;
1233
1234         if (xstats_names != NULL)
1235                 for (i = 0; i < IAVF_NB_XSTATS; i++) {
1236                         snprintf(xstats_names[i].name,
1237                                 sizeof(xstats_names[i].name),
1238                                 "%s", rte_iavf_stats_strings[i].name);
1239                 }
1240         return IAVF_NB_XSTATS;
1241 }
1242
1243 static int iavf_dev_xstats_get(struct rte_eth_dev *dev,
1244                                  struct rte_eth_xstat *xstats, unsigned int n)
1245 {
1246         int ret;
1247         unsigned int i;
1248         struct iavf_adapter *adapter =
1249                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1250         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
1251         struct iavf_vsi *vsi = &vf->vsi;
1252         struct virtchnl_eth_stats *pstats = NULL;
1253
1254         if (n < IAVF_NB_XSTATS)
1255                 return IAVF_NB_XSTATS;
1256
1257         ret = iavf_query_stats(adapter, &pstats);
1258         if (ret != 0)
1259                 return 0;
1260
1261         if (!xstats)
1262                 return 0;
1263
1264         iavf_update_stats(vsi, pstats);
1265
1266         /* loop over xstats array and values from pstats */
1267         for (i = 0; i < IAVF_NB_XSTATS; i++) {
1268                 xstats[i].id = i;
1269                 xstats[i].value = *(uint64_t *)(((char *)pstats) +
1270                         rte_iavf_stats_strings[i].offset);
1271         }
1272
1273         return IAVF_NB_XSTATS;
1274 }
1275
1276
1277 static int
1278 iavf_dev_rx_queue_intr_enable(struct rte_eth_dev *dev, uint16_t queue_id)
1279 {
1280         struct iavf_adapter *adapter =
1281                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1282         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1283         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(adapter);
1284         uint16_t msix_intr;
1285
1286         msix_intr = pci_dev->intr_handle.intr_vec[queue_id];
1287         if (msix_intr == IAVF_MISC_VEC_ID) {
1288                 PMD_DRV_LOG(INFO, "MISC is also enabled for control");
1289                 IAVF_WRITE_REG(hw, IAVF_VFINT_DYN_CTL01,
1290                                IAVF_VFINT_DYN_CTL01_INTENA_MASK |
1291                                IAVF_VFINT_DYN_CTL01_CLEARPBA_MASK |
1292                                IAVF_VFINT_DYN_CTL01_ITR_INDX_MASK);
1293         } else {
1294                 IAVF_WRITE_REG(hw,
1295                                IAVF_VFINT_DYN_CTLN1
1296                                 (msix_intr - IAVF_RX_VEC_START),
1297                                IAVF_VFINT_DYN_CTLN1_INTENA_MASK |
1298                                IAVF_VFINT_DYN_CTL01_CLEARPBA_MASK |
1299                                IAVF_VFINT_DYN_CTLN1_ITR_INDX_MASK);
1300         }
1301
1302         IAVF_WRITE_FLUSH(hw);
1303
1304         rte_intr_ack(&pci_dev->intr_handle);
1305
1306         return 0;
1307 }
1308
1309 static int
1310 iavf_dev_rx_queue_intr_disable(struct rte_eth_dev *dev, uint16_t queue_id)
1311 {
1312         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1313         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1314         uint16_t msix_intr;
1315
1316         msix_intr = pci_dev->intr_handle.intr_vec[queue_id];
1317         if (msix_intr == IAVF_MISC_VEC_ID) {
1318                 PMD_DRV_LOG(ERR, "MISC is used for control, cannot disable it");
1319                 return -EIO;
1320         }
1321
1322         IAVF_WRITE_REG(hw,
1323                       IAVF_VFINT_DYN_CTLN1(msix_intr - IAVF_RX_VEC_START),
1324                       0);
1325
1326         IAVF_WRITE_FLUSH(hw);
1327         return 0;
1328 }
1329
1330 static int
1331 iavf_check_vf_reset_done(struct iavf_hw *hw)
1332 {
1333         int i, reset;
1334
1335         for (i = 0; i < IAVF_RESET_WAIT_CNT; i++) {
1336                 reset = IAVF_READ_REG(hw, IAVF_VFGEN_RSTAT) &
1337                         IAVF_VFGEN_RSTAT_VFR_STATE_MASK;
1338                 reset = reset >> IAVF_VFGEN_RSTAT_VFR_STATE_SHIFT;
1339                 if (reset == VIRTCHNL_VFR_VFACTIVE ||
1340                     reset == VIRTCHNL_VFR_COMPLETED)
1341                         break;
1342                 rte_delay_ms(20);
1343         }
1344
1345         if (i >= IAVF_RESET_WAIT_CNT)
1346                 return -1;
1347
1348         return 0;
1349 }
1350
1351 static int
1352 iavf_init_vf(struct rte_eth_dev *dev)
1353 {
1354         int err, bufsz;
1355         struct iavf_adapter *adapter =
1356                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1357         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1358         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
1359
1360         err = iavf_set_mac_type(hw);
1361         if (err) {
1362                 PMD_INIT_LOG(ERR, "set_mac_type failed: %d", err);
1363                 goto err;
1364         }
1365
1366         err = iavf_check_vf_reset_done(hw);
1367         if (err) {
1368                 PMD_INIT_LOG(ERR, "VF is still resetting");
1369                 goto err;
1370         }
1371
1372         iavf_init_adminq_parameter(hw);
1373         err = iavf_init_adminq(hw);
1374         if (err) {
1375                 PMD_INIT_LOG(ERR, "init_adminq failed: %d", err);
1376                 goto err;
1377         }
1378
1379         vf->aq_resp = rte_zmalloc("vf_aq_resp", IAVF_AQ_BUF_SZ, 0);
1380         if (!vf->aq_resp) {
1381                 PMD_INIT_LOG(ERR, "unable to allocate vf_aq_resp memory");
1382                 goto err_aq;
1383         }
1384         if (iavf_check_api_version(adapter) != 0) {
1385                 PMD_INIT_LOG(ERR, "check_api version failed");
1386                 goto err_api;
1387         }
1388
1389         bufsz = sizeof(struct virtchnl_vf_resource) +
1390                 (IAVF_MAX_VF_VSI * sizeof(struct virtchnl_vsi_resource));
1391         vf->vf_res = rte_zmalloc("vf_res", bufsz, 0);
1392         if (!vf->vf_res) {
1393                 PMD_INIT_LOG(ERR, "unable to allocate vf_res memory");
1394                 goto err_api;
1395         }
1396         if (iavf_get_vf_resource(adapter) != 0) {
1397                 PMD_INIT_LOG(ERR, "iavf_get_vf_config failed");
1398                 goto err_alloc;
1399         }
1400         /* Allocate memort for RSS info */
1401         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF) {
1402                 vf->rss_key = rte_zmalloc("rss_key",
1403                                           vf->vf_res->rss_key_size, 0);
1404                 if (!vf->rss_key) {
1405                         PMD_INIT_LOG(ERR, "unable to allocate rss_key memory");
1406                         goto err_rss;
1407                 }
1408                 vf->rss_lut = rte_zmalloc("rss_lut",
1409                                           vf->vf_res->rss_lut_size, 0);
1410                 if (!vf->rss_lut) {
1411                         PMD_INIT_LOG(ERR, "unable to allocate rss_lut memory");
1412                         goto err_rss;
1413                 }
1414         }
1415
1416         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RX_FLEX_DESC) {
1417                 if (iavf_get_supported_rxdid(adapter) != 0) {
1418                         PMD_INIT_LOG(ERR, "failed to do get supported rxdid");
1419                         goto err_rss;
1420                 }
1421         }
1422
1423         return 0;
1424 err_rss:
1425         rte_free(vf->rss_key);
1426         rte_free(vf->rss_lut);
1427 err_alloc:
1428         rte_free(vf->vf_res);
1429         vf->vsi_res = NULL;
1430 err_api:
1431         rte_free(vf->aq_resp);
1432 err_aq:
1433         iavf_shutdown_adminq(hw);
1434 err:
1435         return -1;
1436 }
1437
1438 /* Enable default admin queue interrupt setting */
1439 static inline void
1440 iavf_enable_irq0(struct iavf_hw *hw)
1441 {
1442         /* Enable admin queue interrupt trigger */
1443         IAVF_WRITE_REG(hw, IAVF_VFINT_ICR0_ENA1,
1444                        IAVF_VFINT_ICR0_ENA1_ADMINQ_MASK);
1445
1446         IAVF_WRITE_REG(hw, IAVF_VFINT_DYN_CTL01,
1447                        IAVF_VFINT_DYN_CTL01_INTENA_MASK |
1448                        IAVF_VFINT_DYN_CTL01_CLEARPBA_MASK |
1449                        IAVF_VFINT_DYN_CTL01_ITR_INDX_MASK);
1450
1451         IAVF_WRITE_FLUSH(hw);
1452 }
1453
1454 static inline void
1455 iavf_disable_irq0(struct iavf_hw *hw)
1456 {
1457         /* Disable all interrupt types */
1458         IAVF_WRITE_REG(hw, IAVF_VFINT_ICR0_ENA1, 0);
1459         IAVF_WRITE_REG(hw, IAVF_VFINT_DYN_CTL01,
1460                        IAVF_VFINT_DYN_CTL01_ITR_INDX_MASK);
1461         IAVF_WRITE_FLUSH(hw);
1462 }
1463
1464 static void
1465 iavf_dev_interrupt_handler(void *param)
1466 {
1467         struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
1468         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1469
1470         iavf_disable_irq0(hw);
1471
1472         iavf_handle_virtchnl_msg(dev);
1473
1474         iavf_enable_irq0(hw);
1475 }
1476
1477 static int
1478 iavf_dev_filter_ctrl(struct rte_eth_dev *dev,
1479                      enum rte_filter_type filter_type,
1480                      enum rte_filter_op filter_op,
1481                      void *arg)
1482 {
1483         int ret = 0;
1484
1485         if (!dev)
1486                 return -EINVAL;
1487
1488         switch (filter_type) {
1489         case RTE_ETH_FILTER_GENERIC:
1490                 if (filter_op != RTE_ETH_FILTER_GET)
1491                         return -EINVAL;
1492                 *(const void **)arg = &iavf_flow_ops;
1493                 break;
1494         default:
1495                 PMD_DRV_LOG(WARNING, "Filter type (%d) not supported",
1496                             filter_type);
1497                 ret = -EINVAL;
1498                 break;
1499         }
1500
1501         return ret;
1502 }
1503
1504
1505 static int
1506 iavf_dev_init(struct rte_eth_dev *eth_dev)
1507 {
1508         struct iavf_adapter *adapter =
1509                 IAVF_DEV_PRIVATE_TO_ADAPTER(eth_dev->data->dev_private);
1510         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(adapter);
1511         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
1512         int ret = 0;
1513
1514         PMD_INIT_FUNC_TRACE();
1515
1516         /* assign ops func pointer */
1517         eth_dev->dev_ops = &iavf_eth_dev_ops;
1518         eth_dev->rx_queue_count = iavf_dev_rxq_count;
1519         eth_dev->rx_descriptor_status = iavf_dev_rx_desc_status;
1520         eth_dev->tx_descriptor_status = iavf_dev_tx_desc_status;
1521         eth_dev->rx_pkt_burst = &iavf_recv_pkts;
1522         eth_dev->tx_pkt_burst = &iavf_xmit_pkts;
1523         eth_dev->tx_pkt_prepare = &iavf_prep_pkts;
1524
1525         /* For secondary processes, we don't initialise any further as primary
1526          * has already done this work. Only check if we need a different RX
1527          * and TX function.
1528          */
1529         if (rte_eal_process_type() != RTE_PROC_PRIMARY) {
1530                 iavf_set_rx_function(eth_dev);
1531                 iavf_set_tx_function(eth_dev);
1532                 return 0;
1533         }
1534         rte_eth_copy_pci_info(eth_dev, pci_dev);
1535         eth_dev->data->dev_flags |= RTE_ETH_DEV_AUTOFILL_QUEUE_XSTATS;
1536
1537         hw->vendor_id = pci_dev->id.vendor_id;
1538         hw->device_id = pci_dev->id.device_id;
1539         hw->subsystem_vendor_id = pci_dev->id.subsystem_vendor_id;
1540         hw->subsystem_device_id = pci_dev->id.subsystem_device_id;
1541         hw->bus.bus_id = pci_dev->addr.bus;
1542         hw->bus.device = pci_dev->addr.devid;
1543         hw->bus.func = pci_dev->addr.function;
1544         hw->hw_addr = (void *)pci_dev->mem_resource[0].addr;
1545         hw->back = IAVF_DEV_PRIVATE_TO_ADAPTER(eth_dev->data->dev_private);
1546         adapter->eth_dev = eth_dev;
1547         adapter->stopped = 1;
1548
1549         if (iavf_init_vf(eth_dev) != 0) {
1550                 PMD_INIT_LOG(ERR, "Init vf failed");
1551                 return -1;
1552         }
1553
1554         /* set default ptype table */
1555         adapter->ptype_tbl = iavf_get_default_ptype_table();
1556
1557         /* copy mac addr */
1558         eth_dev->data->mac_addrs = rte_zmalloc(
1559                 "iavf_mac", RTE_ETHER_ADDR_LEN * IAVF_NUM_MACADDR_MAX, 0);
1560         if (!eth_dev->data->mac_addrs) {
1561                 PMD_INIT_LOG(ERR, "Failed to allocate %d bytes needed to"
1562                              " store MAC addresses",
1563                              RTE_ETHER_ADDR_LEN * IAVF_NUM_MACADDR_MAX);
1564                 return -ENOMEM;
1565         }
1566         /* If the MAC address is not configured by host,
1567          * generate a random one.
1568          */
1569         if (!rte_is_valid_assigned_ether_addr(
1570                         (struct rte_ether_addr *)hw->mac.addr))
1571                 rte_eth_random_addr(hw->mac.addr);
1572         rte_ether_addr_copy((struct rte_ether_addr *)hw->mac.addr,
1573                         &eth_dev->data->mac_addrs[0]);
1574
1575         /* register callback func to eal lib */
1576         rte_intr_callback_register(&pci_dev->intr_handle,
1577                                    iavf_dev_interrupt_handler,
1578                                    (void *)eth_dev);
1579
1580         /* enable uio intr after callback register */
1581         rte_intr_enable(&pci_dev->intr_handle);
1582
1583         /* configure and enable device interrupt */
1584         iavf_enable_irq0(hw);
1585
1586         ret = iavf_flow_init(adapter);
1587         if (ret) {
1588                 PMD_INIT_LOG(ERR, "Failed to initialize flow");
1589                 return ret;
1590         }
1591
1592         return 0;
1593 }
1594
1595 static int
1596 iavf_dev_close(struct rte_eth_dev *dev)
1597 {
1598         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1599         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1600         struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
1601         struct iavf_adapter *adapter =
1602                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1603         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
1604         int ret;
1605
1606         if (rte_eal_process_type() != RTE_PROC_PRIMARY)
1607                 return 0;
1608
1609         ret = iavf_dev_stop(dev);
1610
1611         iavf_flow_flush(dev, NULL);
1612         iavf_flow_uninit(adapter);
1613
1614         /*
1615          * disable promiscuous mode before reset vf
1616          * it is a workaround solution when work with kernel driver
1617          * and it is not the normal way
1618          */
1619         if (vf->promisc_unicast_enabled || vf->promisc_multicast_enabled)
1620                 iavf_config_promisc(adapter, false, false);
1621
1622         iavf_shutdown_adminq(hw);
1623         /* disable uio intr before callback unregister */
1624         rte_intr_disable(intr_handle);
1625
1626         /* unregister callback func from eal lib */
1627         rte_intr_callback_unregister(intr_handle,
1628                                      iavf_dev_interrupt_handler, dev);
1629         iavf_disable_irq0(hw);
1630
1631         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF) {
1632                 if (vf->rss_lut) {
1633                         rte_free(vf->rss_lut);
1634                         vf->rss_lut = NULL;
1635                 }
1636                 if (vf->rss_key) {
1637                         rte_free(vf->rss_key);
1638                         vf->rss_key = NULL;
1639                 }
1640         }
1641
1642         rte_free(vf->vf_res);
1643         vf->vsi_res = NULL;
1644         vf->vf_res = NULL;
1645
1646         rte_free(vf->aq_resp);
1647         vf->aq_resp = NULL;
1648
1649         vf->vf_reset = false;
1650
1651         return ret;
1652 }
1653
1654 static int
1655 iavf_dev_uninit(struct rte_eth_dev *dev)
1656 {
1657         if (rte_eal_process_type() != RTE_PROC_PRIMARY)
1658                 return -EPERM;
1659
1660         iavf_dev_close(dev);
1661
1662         return 0;
1663 }
1664
1665 /*
1666  * Reset VF device only to re-initialize resources in PMD layer
1667  */
1668 static int
1669 iavf_dev_reset(struct rte_eth_dev *dev)
1670 {
1671         int ret;
1672
1673         ret = iavf_dev_uninit(dev);
1674         if (ret)
1675                 return ret;
1676
1677         return iavf_dev_init(dev);
1678 }
1679
1680 static int
1681 iavf_dcf_cap_check_handler(__rte_unused const char *key,
1682                            const char *value, __rte_unused void *opaque)
1683 {
1684         if (strcmp(value, "dcf"))
1685                 return -1;
1686
1687         return 0;
1688 }
1689
1690 static int
1691 iavf_dcf_cap_selected(struct rte_devargs *devargs)
1692 {
1693         struct rte_kvargs *kvlist;
1694         const char *key = "cap";
1695         int ret = 0;
1696
1697         if (devargs == NULL)
1698                 return 0;
1699
1700         kvlist = rte_kvargs_parse(devargs->args, NULL);
1701         if (kvlist == NULL)
1702                 return 0;
1703
1704         if (!rte_kvargs_count(kvlist, key))
1705                 goto exit;
1706
1707         /* dcf capability selected when there's a key-value pair: cap=dcf */
1708         if (rte_kvargs_process(kvlist, key,
1709                                iavf_dcf_cap_check_handler, NULL) < 0)
1710                 goto exit;
1711
1712         ret = 1;
1713
1714 exit:
1715         rte_kvargs_free(kvlist);
1716         return ret;
1717 }
1718
1719 static int eth_iavf_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
1720                              struct rte_pci_device *pci_dev)
1721 {
1722         if (iavf_dcf_cap_selected(pci_dev->device.devargs))
1723                 return 1;
1724
1725         return rte_eth_dev_pci_generic_probe(pci_dev,
1726                 sizeof(struct iavf_adapter), iavf_dev_init);
1727 }
1728
1729 static int eth_iavf_pci_remove(struct rte_pci_device *pci_dev)
1730 {
1731         return rte_eth_dev_pci_generic_remove(pci_dev, iavf_dev_uninit);
1732 }
1733
1734 /* Adaptive virtual function driver struct */
1735 static struct rte_pci_driver rte_iavf_pmd = {
1736         .id_table = pci_id_iavf_map,
1737         .drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC,
1738         .probe = eth_iavf_pci_probe,
1739         .remove = eth_iavf_pci_remove,
1740 };
1741
1742 RTE_PMD_REGISTER_PCI(net_iavf, rte_iavf_pmd);
1743 RTE_PMD_REGISTER_PCI_TABLE(net_iavf, pci_id_iavf_map);
1744 RTE_PMD_REGISTER_KMOD_DEP(net_iavf, "* igb_uio | vfio-pci");
1745 RTE_PMD_REGISTER_PARAM_STRING(net_iavf, "cap=dcf");
1746 RTE_LOG_REGISTER(iavf_logtype_init, pmd.net.iavf.init, NOTICE);
1747 RTE_LOG_REGISTER(iavf_logtype_driver, pmd.net.iavf.driver, NOTICE);
1748 #ifdef RTE_LIBRTE_IAVF_DEBUG_RX
1749 RTE_LOG_REGISTER(iavf_logtype_rx, pmd.net.iavf.rx, DEBUG);
1750 #endif
1751 #ifdef RTE_LIBRTE_IAVF_DEBUG_TX
1752 RTE_LOG_REGISTER(iavf_logtype_tx, pmd.net.iavf.tx, DEBUG);
1753 #endif
1754 #ifdef RTE_LIBRTE_IAVF_DEBUG_TX_FREE
1755 RTE_LOG_REGISTER(iavf_logtype_tx_free, pmd.net.iavf.tx_free, DEBUG);
1756 #endif