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 void iavf_dev_stop(struct rte_eth_dev *dev);
  35 static void 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_promiscuous_enable(struct rte_eth_dev *dev);
  44 static int iavf_dev_promiscuous_disable(struct rte_eth_dev *dev);
  45 static int iavf_dev_allmulticast_enable(struct rte_eth_dev *dev);
  46 static int iavf_dev_allmulticast_disable(struct rte_eth_dev *dev);
  47 static int iavf_dev_add_mac_addr(struct rte_eth_dev *dev,
  48                                 struct rte_ether_addr *addr,
  49                                 uint32_t index,
  50                                 uint32_t pool);
  51 static void iavf_dev_del_mac_addr(struct rte_eth_dev *dev, uint32_t index);
  52 static int iavf_dev_vlan_filter_set(struct rte_eth_dev *dev,
  53                                    uint16_t vlan_id, int on);
  54 static int iavf_dev_vlan_offload_set(struct rte_eth_dev *dev, int mask);
  55 static int iavf_dev_rss_reta_update(struct rte_eth_dev *dev,
  56                                    struct rte_eth_rss_reta_entry64 *reta_conf,
  57                                    uint16_t reta_size);
  58 static int iavf_dev_rss_reta_query(struct rte_eth_dev *dev,
  59                                   struct rte_eth_rss_reta_entry64 *reta_conf,
  60                                   uint16_t reta_size);
  61 static int iavf_dev_rss_hash_update(struct rte_eth_dev *dev,
  62                                    struct rte_eth_rss_conf *rss_conf);
  63 static int iavf_dev_rss_hash_conf_get(struct rte_eth_dev *dev,
  64                                      struct rte_eth_rss_conf *rss_conf);
  65 static int iavf_dev_mtu_set(struct rte_eth_dev *dev, uint16_t mtu);
  66 static int iavf_dev_set_default_mac_addr(struct rte_eth_dev *dev,
  67                                          struct rte_ether_addr *mac_addr);
  68 static int iavf_dev_rx_queue_intr_enable(struct rte_eth_dev *dev,
  69                                         uint16_t queue_id);
  70 static int iavf_dev_rx_queue_intr_disable(struct rte_eth_dev *dev,
  71                                          uint16_t queue_id);
  72 static int iavf_dev_filter_ctrl(struct rte_eth_dev *dev,
  73                      enum rte_filter_type filter_type,
  74                      enum rte_filter_op filter_op,
  75                      void *arg);
  76
  77 static const struct rte_pci_id pci_id_iavf_map[] = {
  78         { RTE_PCI_DEVICE(IAVF_INTEL_VENDOR_ID, IAVF_DEV_ID_ADAPTIVE_VF) },
  79         { .vendor_id = 0, /* sentinel */ },
  80 };
  81
  82 static const struct eth_dev_ops iavf_eth_dev_ops = {
  83         .dev_configure              = iavf_dev_configure,
  84         .dev_start                  = iavf_dev_start,
  85         .dev_stop                   = iavf_dev_stop,
  86         .dev_close                  = iavf_dev_close,
  87         .dev_reset                  = iavf_dev_reset,
  88         .dev_infos_get              = iavf_dev_info_get,
  89         .dev_supported_ptypes_get   = iavf_dev_supported_ptypes_get,
  90         .link_update                = iavf_dev_link_update,
  91         .stats_get                  = iavf_dev_stats_get,
  92         .stats_reset                = iavf_dev_stats_reset,
  93         .promiscuous_enable         = iavf_dev_promiscuous_enable,
  94         .promiscuous_disable        = iavf_dev_promiscuous_disable,
  95         .allmulticast_enable        = iavf_dev_allmulticast_enable,
  96         .allmulticast_disable       = iavf_dev_allmulticast_disable,
  97         .mac_addr_add               = iavf_dev_add_mac_addr,
  98         .mac_addr_remove            = iavf_dev_del_mac_addr,
  99         .vlan_filter_set            = iavf_dev_vlan_filter_set,
 100         .vlan_offload_set           = iavf_dev_vlan_offload_set,
 101         .rx_queue_start             = iavf_dev_rx_queue_start,
 102         .rx_queue_stop              = iavf_dev_rx_queue_stop,
 103         .tx_queue_start             = iavf_dev_tx_queue_start,
 104         .tx_queue_stop              = iavf_dev_tx_queue_stop,
 105         .rx_queue_setup             = iavf_dev_rx_queue_setup,
 106         .rx_queue_release           = iavf_dev_rx_queue_release,
 107         .tx_queue_setup             = iavf_dev_tx_queue_setup,
 108         .tx_queue_release           = iavf_dev_tx_queue_release,
 109         .mac_addr_set               = iavf_dev_set_default_mac_addr,
 110         .reta_update                = iavf_dev_rss_reta_update,
 111         .reta_query                 = iavf_dev_rss_reta_query,
 112         .rss_hash_update            = iavf_dev_rss_hash_update,
 113         .rss_hash_conf_get          = iavf_dev_rss_hash_conf_get,
 114         .rxq_info_get               = iavf_dev_rxq_info_get,
 115         .txq_info_get               = iavf_dev_txq_info_get,
 116         .rx_queue_count             = iavf_dev_rxq_count,
 117         .rx_descriptor_status       = iavf_dev_rx_desc_status,
 118         .tx_descriptor_status       = iavf_dev_tx_desc_status,
 119         .mtu_set                    = iavf_dev_mtu_set,
 120         .rx_queue_intr_enable       = iavf_dev_rx_queue_intr_enable,
 121         .rx_queue_intr_disable      = iavf_dev_rx_queue_intr_disable,
 122         .filter_ctrl                = iavf_dev_filter_ctrl,
 123 };
 124
 125 static int
 126 iavf_dev_configure(struct rte_eth_dev *dev)
 127 {
 128         struct iavf_adapter *ad =
 129                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 130         struct iavf_info *vf =  IAVF_DEV_PRIVATE_TO_VF(ad);
 131         struct rte_eth_conf *dev_conf = &dev->data->dev_conf;
 132
 133         ad->rx_bulk_alloc_allowed = true;
 134         /* Initialize to TRUE. If any of Rx queues doesn't meet the
 135          * vector Rx/Tx preconditions, it will be reset.
 136          */
 137         ad->rx_vec_allowed = true;
 138         ad->tx_vec_allowed = true;
 139
 140         if (dev->data->dev_conf.rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG)
 141                 dev->data->dev_conf.rxmode.offloads |= DEV_RX_OFFLOAD_RSS_HASH;
 142
 143         /* Vlan stripping setting */
 144         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_VLAN) {
 145                 if (dev_conf->rxmode.offloads & DEV_RX_OFFLOAD_VLAN_STRIP)
 146                         iavf_enable_vlan_strip(ad);
 147                 else
 148                         iavf_disable_vlan_strip(ad);
 149         }
 150         return 0;
 151 }
 152
 153 static int
 154 iavf_init_rss(struct iavf_adapter *adapter)
 155 {
 156         struct iavf_info *vf =  IAVF_DEV_PRIVATE_TO_VF(adapter);
 157         struct rte_eth_rss_conf *rss_conf;
 158         uint8_t i, j, nb_q;
 159         int ret;
 160
 161         rss_conf = &adapter->eth_dev->data->dev_conf.rx_adv_conf.rss_conf;
 162         nb_q = RTE_MIN(adapter->eth_dev->data->nb_rx_queues,
 163                        IAVF_MAX_NUM_QUEUES);
 164
 165         if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF)) {
 166                 PMD_DRV_LOG(DEBUG, "RSS is not supported");
 167                 return -ENOTSUP;
 168         }
 169         if (adapter->eth_dev->data->dev_conf.rxmode.mq_mode != ETH_MQ_RX_RSS) {
 170                 PMD_DRV_LOG(WARNING, "RSS is enabled by PF by default");
 171                 /* set all lut items to default queue */
 172                 for (i = 0; i < vf->vf_res->rss_lut_size; i++)
 173                         vf->rss_lut[i] = 0;
 174                 ret = iavf_configure_rss_lut(adapter);
 175                 return ret;
 176         }
 177
 178         /* In IAVF, RSS enablement is set by PF driver. It is not supported
 179          * to set based on rss_conf->rss_hf.
 180          */
 181
 182         /* configure RSS key */
 183         if (!rss_conf->rss_key) {
 184                 /* Calculate the default hash key */
 185                 for (i = 0; i <= vf->vf_res->rss_key_size; i++)
 186                         vf->rss_key[i] = (uint8_t)rte_rand();
 187         } else
 188                 rte_memcpy(vf->rss_key, rss_conf->rss_key,
 189                            RTE_MIN(rss_conf->rss_key_len,
 190                                    vf->vf_res->rss_key_size));
 191
 192         /* init RSS LUT table */
 193         for (i = 0, j = 0; i < vf->vf_res->rss_lut_size; i++, j++) {
 194                 if (j >= nb_q)
 195                         j = 0;
 196                 vf->rss_lut[i] = j;
 197         }
 198         /* send virtchnnl ops to configure rss*/
 199         ret = iavf_configure_rss_lut(adapter);
 200         if (ret)
 201                 return ret;
 202         ret = iavf_configure_rss_key(adapter);
 203         if (ret)
 204                 return ret;
 205
 206         return 0;
 207 }
 208
 209 static int
 210 iavf_init_rxq(struct rte_eth_dev *dev, struct iavf_rx_queue *rxq)
 211 {
 212         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(dev->data->dev_private);
 213         struct rte_eth_dev_data *dev_data = dev->data;
 214         uint16_t buf_size, max_pkt_len, len;
 215
 216         buf_size = rte_pktmbuf_data_room_size(rxq->mp) - RTE_PKTMBUF_HEADROOM;
 217
 218         /* Calculate the maximum packet length allowed */
 219         len = rxq->rx_buf_len * IAVF_MAX_CHAINED_RX_BUFFERS;
 220         max_pkt_len = RTE_MIN(len, dev->data->dev_conf.rxmode.max_rx_pkt_len);
 221
 222         /* Check if the jumbo frame and maximum packet length are set
 223          * correctly.
 224          */
 225         if (dev->data->dev_conf.rxmode.offloads & DEV_RX_OFFLOAD_JUMBO_FRAME) {
 226                 if (max_pkt_len <= RTE_ETHER_MAX_LEN ||
 227                     max_pkt_len > IAVF_FRAME_SIZE_MAX) {
 228                         PMD_DRV_LOG(ERR, "maximum packet length must be "
 229                                     "larger than %u and smaller than %u, "
 230                                     "as jumbo frame is enabled",
 231                                     (uint32_t)RTE_ETHER_MAX_LEN,
 232                                     (uint32_t)IAVF_FRAME_SIZE_MAX);
 233                         return -EINVAL;
 234                 }
 235         } else {
 236                 if (max_pkt_len < RTE_ETHER_MIN_LEN ||
 237                     max_pkt_len > RTE_ETHER_MAX_LEN) {
 238                         PMD_DRV_LOG(ERR, "maximum packet length must be "
 239                                     "larger than %u and smaller than %u, "
 240                                     "as jumbo frame is disabled",
 241                                     (uint32_t)RTE_ETHER_MIN_LEN,
 242                                     (uint32_t)RTE_ETHER_MAX_LEN);
 243                         return -EINVAL;
 244                 }
 245         }
 246
 247         rxq->max_pkt_len = max_pkt_len;
 248         if ((dev_data->dev_conf.rxmode.offloads & DEV_RX_OFFLOAD_SCATTER) ||
 249             (rxq->max_pkt_len + 2 * IAVF_VLAN_TAG_SIZE) > buf_size) {
 250                 dev_data->scattered_rx = 1;
 251         }
 252         IAVF_PCI_REG_WRITE(rxq->qrx_tail, rxq->nb_rx_desc - 1);
 253         IAVF_WRITE_FLUSH(hw);
 254
 255         return 0;
 256 }
 257
 258 static int
 259 iavf_init_queues(struct rte_eth_dev *dev)
 260 {
 261         struct iavf_rx_queue **rxq =
 262                 (struct iavf_rx_queue **)dev->data->rx_queues;
 263         int i, ret = IAVF_SUCCESS;
 264
 265         for (i = 0; i < dev->data->nb_rx_queues; i++) {
 266                 if (!rxq[i] || !rxq[i]->q_set)
 267                         continue;
 268                 ret = iavf_init_rxq(dev, rxq[i]);
 269                 if (ret != IAVF_SUCCESS)
 270                         break;
 271         }
 272         /* set rx/tx function to vector/scatter/single-segment
 273          * according to parameters
 274          */
 275         iavf_set_rx_function(dev);
 276         iavf_set_tx_function(dev);
 277
 278         return ret;
 279 }
 280
 281 static int iavf_config_rx_queues_irqs(struct rte_eth_dev *dev,
 282                                      struct rte_intr_handle *intr_handle)
 283 {
 284         struct iavf_adapter *adapter =
 285                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 286         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
 287         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(adapter);
 288         uint16_t interval, i;
 289         int vec;
 290
 291         if (rte_intr_cap_multiple(intr_handle) &&
 292             dev->data->dev_conf.intr_conf.rxq) {
 293                 if (rte_intr_efd_enable(intr_handle, dev->data->nb_rx_queues))
 294                         return -1;
 295         }
 296
 297         if (rte_intr_dp_is_en(intr_handle) && !intr_handle->intr_vec) {
 298                 intr_handle->intr_vec =
 299                         rte_zmalloc("intr_vec",
 300                                     dev->data->nb_rx_queues * sizeof(int), 0);
 301                 if (!intr_handle->intr_vec) {
 302                         PMD_DRV_LOG(ERR, "Failed to allocate %d rx intr_vec",
 303                                     dev->data->nb_rx_queues);
 304                         return -1;
 305                 }
 306         }
 307
 308         if (!dev->data->dev_conf.intr_conf.rxq ||
 309             !rte_intr_dp_is_en(intr_handle)) {
 310                 /* Rx interrupt disabled, Map interrupt only for writeback */
 311                 vf->nb_msix = 1;
 312                 if (vf->vf_res->vf_cap_flags &
 313                     VIRTCHNL_VF_OFFLOAD_WB_ON_ITR) {
 314                         /* If WB_ON_ITR supports, enable it */
 315                         vf->msix_base = IAVF_RX_VEC_START;
 316                         IAVF_WRITE_REG(hw,
 317                                        IAVF_VFINT_DYN_CTLN1(vf->msix_base - 1),
 318                                        IAVF_VFINT_DYN_CTLN1_ITR_INDX_MASK |
 319                                        IAVF_VFINT_DYN_CTLN1_WB_ON_ITR_MASK);
 320                 } else {
 321                         /* If no WB_ON_ITR offload flags, need to set
 322                          * interrupt for descriptor write back.
 323                          */
 324                         vf->msix_base = IAVF_MISC_VEC_ID;
 325
 326                         /* set ITR to max */
 327                         interval = iavf_calc_itr_interval(
 328                                         IAVF_QUEUE_ITR_INTERVAL_MAX);
 329                         IAVF_WRITE_REG(hw, IAVF_VFINT_DYN_CTL01,
 330                                        IAVF_VFINT_DYN_CTL01_INTENA_MASK |
 331                                        (IAVF_ITR_INDEX_DEFAULT <<
 332                                         IAVF_VFINT_DYN_CTL01_ITR_INDX_SHIFT) |
 333                                        (interval <<
 334                                         IAVF_VFINT_DYN_CTL01_INTERVAL_SHIFT));
 335                 }
 336                 IAVF_WRITE_FLUSH(hw);
 337                 /* map all queues to the same interrupt */
 338                 for (i = 0; i < dev->data->nb_rx_queues; i++)
 339                         vf->rxq_map[vf->msix_base] |= 1 << i;
 340         } else {
 341                 if (!rte_intr_allow_others(intr_handle)) {
 342                         vf->nb_msix = 1;
 343                         vf->msix_base = IAVF_MISC_VEC_ID;
 344                         for (i = 0; i < dev->data->nb_rx_queues; i++) {
 345                                 vf->rxq_map[vf->msix_base] |= 1 << i;
 346                                 intr_handle->intr_vec[i] = IAVF_MISC_VEC_ID;
 347                         }
 348                         PMD_DRV_LOG(DEBUG,
 349                                     "vector %u are mapping to all Rx queues",
 350                                     vf->msix_base);
 351                 } else {
 352                         /* If Rx interrupt is reuquired, and we can use
 353                          * multi interrupts, then the vec is from 1
 354                          */
 355                         vf->nb_msix = RTE_MIN(vf->vf_res->max_vectors,
 356                                               intr_handle->nb_efd);
 357                         vf->msix_base = IAVF_RX_VEC_START;
 358                         vec = IAVF_RX_VEC_START;
 359                         for (i = 0; i < dev->data->nb_rx_queues; i++) {
 360                                 vf->rxq_map[vec] |= 1 << i;
 361                                 intr_handle->intr_vec[i] = vec++;
 362                                 if (vec >= vf->nb_msix)
 363                                         vec = IAVF_RX_VEC_START;
 364                         }
 365                         PMD_DRV_LOG(DEBUG,
 366                                     "%u vectors are mapping to %u Rx queues",
 367                                     vf->nb_msix, dev->data->nb_rx_queues);
 368                 }
 369         }
 370
 371         if (iavf_config_irq_map(adapter)) {
 372                 PMD_DRV_LOG(ERR, "config interrupt mapping failed");
 373                 return -1;
 374         }
 375         return 0;
 376 }
 377
 378 static int
 379 iavf_start_queues(struct rte_eth_dev *dev)
 380 {
 381         struct iavf_rx_queue *rxq;
 382         struct iavf_tx_queue *txq;
 383         int i;
 384
 385         for (i = 0; i < dev->data->nb_tx_queues; i++) {
 386                 txq = dev->data->tx_queues[i];
 387                 if (txq->tx_deferred_start)
 388                         continue;
 389                 if (iavf_dev_tx_queue_start(dev, i) != 0) {
 390                         PMD_DRV_LOG(ERR, "Fail to start queue %u", i);
 391                         return -1;
 392                 }
 393         }
 394
 395         for (i = 0; i < dev->data->nb_rx_queues; i++) {
 396                 rxq = dev->data->rx_queues[i];
 397                 if (rxq->rx_deferred_start)
 398                         continue;
 399                 if (iavf_dev_rx_queue_start(dev, i) != 0) {
 400                         PMD_DRV_LOG(ERR, "Fail to start queue %u", i);
 401                         return -1;
 402                 }
 403         }
 404
 405         return 0;
 406 }
 407
 408 static int
 409 iavf_dev_start(struct rte_eth_dev *dev)
 410 {
 411         struct iavf_adapter *adapter =
 412                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 413         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
 414         struct rte_intr_handle *intr_handle = dev->intr_handle;
 415
 416         PMD_INIT_FUNC_TRACE();
 417
 418         adapter->stopped = 0;
 419
 420         vf->max_pkt_len = dev->data->dev_conf.rxmode.max_rx_pkt_len;
 421         vf->num_queue_pairs = RTE_MAX(dev->data->nb_rx_queues,
 422                                       dev->data->nb_tx_queues);
 423
 424         if (iavf_init_queues(dev) != 0) {
 425                 PMD_DRV_LOG(ERR, "failed to do Queue init");
 426                 return -1;
 427         }
 428
 429         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF) {
 430                 if (iavf_init_rss(adapter) != 0) {
 431                         PMD_DRV_LOG(ERR, "configure rss failed");
 432                         goto err_rss;
 433                 }
 434         }
 435
 436         if (iavf_configure_queues(adapter) != 0) {
 437                 PMD_DRV_LOG(ERR, "configure queues failed");
 438                 goto err_queue;
 439         }
 440
 441         if (iavf_config_rx_queues_irqs(dev, intr_handle) != 0) {
 442                 PMD_DRV_LOG(ERR, "configure irq failed");
 443                 goto err_queue;
 444         }
 445         /* re-enable intr again, because efd assign may change */
 446         if (dev->data->dev_conf.intr_conf.rxq != 0) {
 447                 rte_intr_disable(intr_handle);
 448                 rte_intr_enable(intr_handle);
 449         }
 450
 451         /* Set all mac addrs */
 452         iavf_add_del_all_mac_addr(adapter, true);
 453
 454         if (iavf_start_queues(dev) != 0) {
 455                 PMD_DRV_LOG(ERR, "enable queues failed");
 456                 goto err_mac;
 457         }
 458
 459         return 0;
 460
 461 err_mac:
 462         iavf_add_del_all_mac_addr(adapter, false);
 463 err_queue:
 464 err_rss:
 465         return -1;
 466 }
 467
 468 static void
 469 iavf_dev_stop(struct rte_eth_dev *dev)
 470 {
 471         struct iavf_adapter *adapter =
 472                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 473         struct rte_intr_handle *intr_handle = dev->intr_handle;
 474
 475         PMD_INIT_FUNC_TRACE();
 476
 477         if (adapter->stopped == 1)
 478                 return;
 479
 480         iavf_stop_queues(dev);
 481
 482         /* Disable the interrupt for Rx */
 483         rte_intr_efd_disable(intr_handle);
 484         /* Rx interrupt vector mapping free */
 485         if (intr_handle->intr_vec) {
 486                 rte_free(intr_handle->intr_vec);
 487                 intr_handle->intr_vec = NULL;
 488         }
 489
 490         /* remove all mac addrs */
 491         iavf_add_del_all_mac_addr(adapter, false);
 492         adapter->stopped = 1;
 493 }
 494
 495 static int
 496 iavf_dev_info_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
 497 {
 498         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
 499
 500         dev_info->max_rx_queues = vf->vsi_res->num_queue_pairs;
 501         dev_info->max_tx_queues = vf->vsi_res->num_queue_pairs;
 502         dev_info->min_rx_bufsize = IAVF_BUF_SIZE_MIN;
 503         dev_info->max_rx_pktlen = IAVF_FRAME_SIZE_MAX;
 504         dev_info->hash_key_size = vf->vf_res->rss_key_size;
 505         dev_info->reta_size = vf->vf_res->rss_lut_size;
 506         dev_info->flow_type_rss_offloads = IAVF_RSS_OFFLOAD_ALL;
 507         dev_info->max_mac_addrs = IAVF_NUM_MACADDR_MAX;
 508         dev_info->rx_offload_capa =
 509                 DEV_RX_OFFLOAD_VLAN_STRIP |
 510                 DEV_RX_OFFLOAD_QINQ_STRIP |
 511                 DEV_RX_OFFLOAD_IPV4_CKSUM |
 512                 DEV_RX_OFFLOAD_UDP_CKSUM |
 513                 DEV_RX_OFFLOAD_TCP_CKSUM |
 514                 DEV_RX_OFFLOAD_OUTER_IPV4_CKSUM |
 515                 DEV_RX_OFFLOAD_SCATTER |
 516                 DEV_RX_OFFLOAD_JUMBO_FRAME |
 517                 DEV_RX_OFFLOAD_VLAN_FILTER |
 518                 DEV_RX_OFFLOAD_RSS_HASH;
 519         dev_info->tx_offload_capa =
 520                 DEV_TX_OFFLOAD_VLAN_INSERT |
 521                 DEV_TX_OFFLOAD_QINQ_INSERT |
 522                 DEV_TX_OFFLOAD_IPV4_CKSUM |
 523                 DEV_TX_OFFLOAD_UDP_CKSUM |
 524                 DEV_TX_OFFLOAD_TCP_CKSUM |
 525                 DEV_TX_OFFLOAD_SCTP_CKSUM |
 526                 DEV_TX_OFFLOAD_OUTER_IPV4_CKSUM |
 527                 DEV_TX_OFFLOAD_TCP_TSO |
 528                 DEV_TX_OFFLOAD_VXLAN_TNL_TSO |
 529                 DEV_TX_OFFLOAD_GRE_TNL_TSO |
 530                 DEV_TX_OFFLOAD_IPIP_TNL_TSO |
 531                 DEV_TX_OFFLOAD_GENEVE_TNL_TSO |
 532                 DEV_TX_OFFLOAD_MULTI_SEGS;
 533
 534         dev_info->default_rxconf = (struct rte_eth_rxconf) {
 535                 .rx_free_thresh = IAVF_DEFAULT_RX_FREE_THRESH,
 536                 .rx_drop_en = 0,
 537                 .offloads = 0,
 538         };
 539
 540         dev_info->default_txconf = (struct rte_eth_txconf) {
 541                 .tx_free_thresh = IAVF_DEFAULT_TX_FREE_THRESH,
 542                 .tx_rs_thresh = IAVF_DEFAULT_TX_RS_THRESH,
 543                 .offloads = 0,
 544         };
 545
 546         dev_info->rx_desc_lim = (struct rte_eth_desc_lim) {
 547                 .nb_max = IAVF_MAX_RING_DESC,
 548                 .nb_min = IAVF_MIN_RING_DESC,
 549                 .nb_align = IAVF_ALIGN_RING_DESC,
 550         };
 551
 552         dev_info->tx_desc_lim = (struct rte_eth_desc_lim) {
 553                 .nb_max = IAVF_MAX_RING_DESC,
 554                 .nb_min = IAVF_MIN_RING_DESC,
 555                 .nb_align = IAVF_ALIGN_RING_DESC,
 556         };
 557
 558         return 0;
 559 }
 560
 561 static const uint32_t *
 562 iavf_dev_supported_ptypes_get(struct rte_eth_dev *dev __rte_unused)
 563 {
 564         static const uint32_t ptypes[] = {
 565                 RTE_PTYPE_L2_ETHER,
 566                 RTE_PTYPE_L3_IPV4_EXT_UNKNOWN,
 567                 RTE_PTYPE_L4_FRAG,
 568                 RTE_PTYPE_L4_ICMP,
 569                 RTE_PTYPE_L4_NONFRAG,
 570                 RTE_PTYPE_L4_SCTP,
 571                 RTE_PTYPE_L4_TCP,
 572                 RTE_PTYPE_L4_UDP,
 573                 RTE_PTYPE_UNKNOWN
 574         };
 575         return ptypes;
 576 }
 577
 578 int
 579 iavf_dev_link_update(struct rte_eth_dev *dev,
 580                     __rte_unused int wait_to_complete)
 581 {
 582         struct rte_eth_link new_link;
 583         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
 584
 585         /* Only read status info stored in VF, and the info is updated
 586          *  when receive LINK_CHANGE evnet from PF by Virtchnnl.
 587          */
 588         switch (vf->link_speed) {
 589         case 10:
 590                 new_link.link_speed = ETH_SPEED_NUM_10M;
 591                 break;
 592         case 100:
 593                 new_link.link_speed = ETH_SPEED_NUM_100M;
 594                 break;
 595         case 1000:
 596                 new_link.link_speed = ETH_SPEED_NUM_1G;
 597                 break;
 598         case 10000:
 599                 new_link.link_speed = ETH_SPEED_NUM_10G;
 600                 break;
 601         case 20000:
 602                 new_link.link_speed = ETH_SPEED_NUM_20G;
 603                 break;
 604         case 25000:
 605                 new_link.link_speed = ETH_SPEED_NUM_25G;
 606                 break;
 607         case 40000:
 608                 new_link.link_speed = ETH_SPEED_NUM_40G;
 609                 break;
 610         case 50000:
 611                 new_link.link_speed = ETH_SPEED_NUM_50G;
 612                 break;
 613         case 100000:
 614                 new_link.link_speed = ETH_SPEED_NUM_100G;
 615                 break;
 616         default:
 617                 new_link.link_speed = ETH_SPEED_NUM_NONE;
 618                 break;
 619         }
 620
 621         new_link.link_duplex = ETH_LINK_FULL_DUPLEX;
 622         new_link.link_status = vf->link_up ? ETH_LINK_UP :
 623                                              ETH_LINK_DOWN;
 624         new_link.link_autoneg = !(dev->data->dev_conf.link_speeds &
 625                                 ETH_LINK_SPEED_FIXED);
 626
 627         if (rte_atomic64_cmpset((uint64_t *)&dev->data->dev_link,
 628                                 *(uint64_t *)&dev->data->dev_link,
 629                                 *(uint64_t *)&new_link) == 0)
 630                 return -1;
 631
 632         return 0;
 633 }
 634
 635 static int
 636 iavf_dev_promiscuous_enable(struct rte_eth_dev *dev)
 637 {
 638         struct iavf_adapter *adapter =
 639                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 640         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
 641         int ret;
 642
 643         if (vf->promisc_unicast_enabled)
 644                 return 0;
 645
 646         ret = iavf_config_promisc(adapter, true, vf->promisc_multicast_enabled);
 647         if (!ret)
 648                 vf->promisc_unicast_enabled = true;
 649         else
 650                 ret = -EAGAIN;
 651
 652         return ret;
 653 }
 654
 655 static int
 656 iavf_dev_promiscuous_disable(struct rte_eth_dev *dev)
 657 {
 658         struct iavf_adapter *adapter =
 659                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 660         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
 661         int ret;
 662
 663         if (!vf->promisc_unicast_enabled)
 664                 return 0;
 665
 666         ret = iavf_config_promisc(adapter, false,
 667                                   vf->promisc_multicast_enabled);
 668         if (!ret)
 669                 vf->promisc_unicast_enabled = false;
 670         else
 671                 ret = -EAGAIN;
 672
 673         return ret;
 674 }
 675
 676 static int
 677 iavf_dev_allmulticast_enable(struct rte_eth_dev *dev)
 678 {
 679         struct iavf_adapter *adapter =
 680                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 681         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
 682         int ret;
 683
 684         if (vf->promisc_multicast_enabled)
 685                 return 0;
 686
 687         ret = iavf_config_promisc(adapter, vf->promisc_unicast_enabled, true);
 688         if (!ret)
 689                 vf->promisc_multicast_enabled = true;
 690         else
 691                 ret = -EAGAIN;
 692
 693         return ret;
 694 }
 695
 696 static int
 697 iavf_dev_allmulticast_disable(struct rte_eth_dev *dev)
 698 {
 699         struct iavf_adapter *adapter =
 700                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 701         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
 702         int ret;
 703
 704         if (!vf->promisc_multicast_enabled)
 705                 return 0;
 706
 707         ret = iavf_config_promisc(adapter, vf->promisc_unicast_enabled, false);
 708         if (!ret)
 709                 vf->promisc_multicast_enabled = false;
 710         else
 711                 ret = -EAGAIN;
 712
 713         return ret;
 714 }
 715
 716 static int
 717 iavf_dev_add_mac_addr(struct rte_eth_dev *dev, struct rte_ether_addr *addr,
 718                      __rte_unused uint32_t index,
 719                      __rte_unused uint32_t pool)
 720 {
 721         struct iavf_adapter *adapter =
 722                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 723         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
 724         int err;
 725
 726         if (rte_is_zero_ether_addr(addr)) {
 727                 PMD_DRV_LOG(ERR, "Invalid Ethernet Address");
 728                 return -EINVAL;
 729         }
 730
 731         err = iavf_add_del_eth_addr(adapter, addr, true);
 732         if (err) {
 733                 PMD_DRV_LOG(ERR, "fail to add MAC address");
 734                 return -EIO;
 735         }
 736
 737         vf->mac_num++;
 738
 739         return 0;
 740 }
 741
 742 static void
 743 iavf_dev_del_mac_addr(struct rte_eth_dev *dev, uint32_t index)
 744 {
 745         struct iavf_adapter *adapter =
 746                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 747         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
 748         struct rte_ether_addr *addr;
 749         int err;
 750
 751         addr = &dev->data->mac_addrs[index];
 752
 753         err = iavf_add_del_eth_addr(adapter, addr, false);
 754         if (err)
 755                 PMD_DRV_LOG(ERR, "fail to delete MAC address");
 756
 757         vf->mac_num--;
 758 }
 759
 760 static int
 761 iavf_dev_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
 762 {
 763         struct iavf_adapter *adapter =
 764                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 765         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
 766         int err;
 767
 768         if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_VLAN))
 769                 return -ENOTSUP;
 770
 771         err = iavf_add_del_vlan(adapter, vlan_id, on);
 772         if (err)
 773                 return -EIO;
 774         return 0;
 775 }
 776
 777 static int
 778 iavf_dev_vlan_offload_set(struct rte_eth_dev *dev, int mask)
 779 {
 780         struct iavf_adapter *adapter =
 781                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 782         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
 783         struct rte_eth_conf *dev_conf = &dev->data->dev_conf;
 784         int err;
 785
 786         if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_VLAN))
 787                 return -ENOTSUP;
 788
 789         /* Vlan stripping setting */
 790         if (mask & ETH_VLAN_STRIP_MASK) {
 791                 /* Enable or disable VLAN stripping */
 792                 if (dev_conf->rxmode.offloads & DEV_RX_OFFLOAD_VLAN_STRIP)
 793                         err = iavf_enable_vlan_strip(adapter);
 794                 else
 795                         err = iavf_disable_vlan_strip(adapter);
 796
 797                 if (err)
 798                         return -EIO;
 799         }
 800         return 0;
 801 }
 802
 803 static int
 804 iavf_dev_rss_reta_update(struct rte_eth_dev *dev,
 805                         struct rte_eth_rss_reta_entry64 *reta_conf,
 806                         uint16_t reta_size)
 807 {
 808         struct iavf_adapter *adapter =
 809                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 810         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
 811         uint8_t *lut;
 812         uint16_t i, idx, shift;
 813         int ret;
 814
 815         if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF))
 816                 return -ENOTSUP;
 817
 818         if (reta_size != vf->vf_res->rss_lut_size) {
 819                 PMD_DRV_LOG(ERR, "The size of hash lookup table configured "
 820                         "(%d) doesn't match the number of hardware can "
 821                         "support (%d)", reta_size, vf->vf_res->rss_lut_size);
 822                 return -EINVAL;
 823         }
 824
 825         lut = rte_zmalloc("rss_lut", reta_size, 0);
 826         if (!lut) {
 827                 PMD_DRV_LOG(ERR, "No memory can be allocated");
 828                 return -ENOMEM;
 829         }
 830         /* store the old lut table temporarily */
 831         rte_memcpy(lut, vf->rss_lut, reta_size);
 832
 833         for (i = 0; i < reta_size; i++) {
 834                 idx = i / RTE_RETA_GROUP_SIZE;
 835                 shift = i % RTE_RETA_GROUP_SIZE;
 836                 if (reta_conf[idx].mask & (1ULL << shift))
 837                         lut[i] = reta_conf[idx].reta[shift];
 838         }
 839
 840         rte_memcpy(vf->rss_lut, lut, reta_size);
 841         /* send virtchnnl ops to configure rss*/
 842         ret = iavf_configure_rss_lut(adapter);
 843         if (ret) /* revert back */
 844                 rte_memcpy(vf->rss_lut, lut, reta_size);
 845         rte_free(lut);
 846
 847         return ret;
 848 }
 849
 850 static int
 851 iavf_dev_rss_reta_query(struct rte_eth_dev *dev,
 852                        struct rte_eth_rss_reta_entry64 *reta_conf,
 853                        uint16_t reta_size)
 854 {
 855         struct iavf_adapter *adapter =
 856                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 857         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
 858         uint16_t i, idx, shift;
 859
 860         if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF))
 861                 return -ENOTSUP;
 862
 863         if (reta_size != vf->vf_res->rss_lut_size) {
 864                 PMD_DRV_LOG(ERR, "The size of hash lookup table configured "
 865                         "(%d) doesn't match the number of hardware can "
 866                         "support (%d)", reta_size, vf->vf_res->rss_lut_size);
 867                 return -EINVAL;
 868         }
 869
 870         for (i = 0; i < reta_size; i++) {
 871                 idx = i / RTE_RETA_GROUP_SIZE;
 872                 shift = i % RTE_RETA_GROUP_SIZE;
 873                 if (reta_conf[idx].mask & (1ULL << shift))
 874                         reta_conf[idx].reta[shift] = vf->rss_lut[i];
 875         }
 876
 877         return 0;
 878 }
 879
 880 static int
 881 iavf_dev_rss_hash_update(struct rte_eth_dev *dev,
 882                         struct rte_eth_rss_conf *rss_conf)
 883 {
 884         struct iavf_adapter *adapter =
 885                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 886         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
 887
 888         if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF))
 889                 return -ENOTSUP;
 890
 891         /* HENA setting, it is enabled by default, no change */
 892         if (!rss_conf->rss_key || rss_conf->rss_key_len == 0) {
 893                 PMD_DRV_LOG(DEBUG, "No key to be configured");
 894                 return 0;
 895         } else if (rss_conf->rss_key_len != vf->vf_res->rss_key_size) {
 896                 PMD_DRV_LOG(ERR, "The size of hash key configured "
 897                         "(%d) doesn't match the size of hardware can "
 898                         "support (%d)", rss_conf->rss_key_len,
 899                         vf->vf_res->rss_key_size);
 900                 return -EINVAL;
 901         }
 902
 903         rte_memcpy(vf->rss_key, rss_conf->rss_key, rss_conf->rss_key_len);
 904
 905         return iavf_configure_rss_key(adapter);
 906 }
 907
 908 static int
 909 iavf_dev_rss_hash_conf_get(struct rte_eth_dev *dev,
 910                           struct rte_eth_rss_conf *rss_conf)
 911 {
 912         struct iavf_adapter *adapter =
 913                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 914         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
 915
 916         if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF))
 917                 return -ENOTSUP;
 918
 919          /* Just set it to default value now. */
 920         rss_conf->rss_hf = IAVF_RSS_OFFLOAD_ALL;
 921
 922         if (!rss_conf->rss_key)
 923                 return 0;
 924
 925         rss_conf->rss_key_len = vf->vf_res->rss_key_size;
 926         rte_memcpy(rss_conf->rss_key, vf->rss_key, rss_conf->rss_key_len);
 927
 928         return 0;
 929 }
 930
 931 static int
 932 iavf_dev_mtu_set(struct rte_eth_dev *dev, uint16_t mtu)
 933 {
 934         uint32_t frame_size = mtu + IAVF_ETH_OVERHEAD;
 935         int ret = 0;
 936
 937         if (mtu < RTE_ETHER_MIN_MTU || frame_size > IAVF_FRAME_SIZE_MAX)
 938                 return -EINVAL;
 939
 940         /* mtu setting is forbidden if port is start */
 941         if (dev->data->dev_started) {
 942                 PMD_DRV_LOG(ERR, "port must be stopped before configuration");
 943                 return -EBUSY;
 944         }
 945
 946         if (frame_size > RTE_ETHER_MAX_LEN)
 947                 dev->data->dev_conf.rxmode.offloads |=
 948                                 DEV_RX_OFFLOAD_JUMBO_FRAME;
 949         else
 950                 dev->data->dev_conf.rxmode.offloads &=
 951                                 ~DEV_RX_OFFLOAD_JUMBO_FRAME;
 952
 953         dev->data->dev_conf.rxmode.max_rx_pkt_len = frame_size;
 954
 955         return ret;
 956 }
 957
 958 static int
 959 iavf_dev_set_default_mac_addr(struct rte_eth_dev *dev,
 960                              struct rte_ether_addr *mac_addr)
 961 {
 962         struct iavf_adapter *adapter =
 963                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 964         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(adapter);
 965         struct rte_ether_addr *perm_addr, *old_addr;
 966         int ret;
 967
 968         old_addr = (struct rte_ether_addr *)hw->mac.addr;
 969         perm_addr = (struct rte_ether_addr *)hw->mac.perm_addr;
 970
 971         if (rte_is_same_ether_addr(mac_addr, old_addr))
 972                 return 0;
 973
 974         /* If the MAC address is configured by host, skip the setting */
 975         if (rte_is_valid_assigned_ether_addr(perm_addr))
 976                 return -EPERM;
 977
 978         ret = iavf_add_del_eth_addr(adapter, old_addr, false);
 979         if (ret)
 980                 PMD_DRV_LOG(ERR, "Fail to delete old MAC:"
 981                             " %02X:%02X:%02X:%02X:%02X:%02X",
 982                             old_addr->addr_bytes[0],
 983                             old_addr->addr_bytes[1],
 984                             old_addr->addr_bytes[2],
 985                             old_addr->addr_bytes[3],
 986                             old_addr->addr_bytes[4],
 987                             old_addr->addr_bytes[5]);
 988
 989         ret = iavf_add_del_eth_addr(adapter, mac_addr, true);
 990         if (ret)
 991                 PMD_DRV_LOG(ERR, "Fail to add new MAC:"
 992                             " %02X:%02X:%02X:%02X:%02X:%02X",
 993                             mac_addr->addr_bytes[0],
 994                             mac_addr->addr_bytes[1],
 995                             mac_addr->addr_bytes[2],
 996                             mac_addr->addr_bytes[3],
 997                             mac_addr->addr_bytes[4],
 998                             mac_addr->addr_bytes[5]);
 999
1000         if (ret)
1001                 return -EIO;
1002
1003         rte_ether_addr_copy(mac_addr, (struct rte_ether_addr *)hw->mac.addr);
1004         return 0;
1005 }
1006
1007 static void
1008 iavf_stat_update_48(uint64_t *offset, uint64_t *stat)
1009 {
1010         if (*stat >= *offset)
1011                 *stat = *stat - *offset;
1012         else
1013                 *stat = (uint64_t)((*stat +
1014                         ((uint64_t)1 << IAVF_48_BIT_WIDTH)) - *offset);
1015
1016         *stat &= IAVF_48_BIT_MASK;
1017 }
1018
1019 static void
1020 iavf_stat_update_32(uint64_t *offset, uint64_t *stat)
1021 {
1022         if (*stat >= *offset)
1023                 *stat = (uint64_t)(*stat - *offset);
1024         else
1025                 *stat = (uint64_t)((*stat +
1026                         ((uint64_t)1 << IAVF_32_BIT_WIDTH)) - *offset);
1027 }
1028
1029 static void
1030 iavf_update_stats(struct iavf_vsi *vsi, struct virtchnl_eth_stats *nes)
1031 {
1032         struct virtchnl_eth_stats *oes = &vsi->eth_stats_offset;
1033
1034         iavf_stat_update_48(&oes->rx_bytes, &nes->rx_bytes);
1035         iavf_stat_update_48(&oes->rx_unicast, &nes->rx_unicast);
1036         iavf_stat_update_48(&oes->rx_multicast, &nes->rx_multicast);
1037         iavf_stat_update_48(&oes->rx_broadcast, &nes->rx_broadcast);
1038         iavf_stat_update_32(&oes->rx_discards, &nes->rx_discards);
1039         iavf_stat_update_48(&oes->tx_bytes, &nes->tx_bytes);
1040         iavf_stat_update_48(&oes->tx_unicast, &nes->tx_unicast);
1041         iavf_stat_update_48(&oes->tx_multicast, &nes->tx_multicast);
1042         iavf_stat_update_48(&oes->tx_broadcast, &nes->tx_broadcast);
1043         iavf_stat_update_32(&oes->tx_errors, &nes->tx_errors);
1044         iavf_stat_update_32(&oes->tx_discards, &nes->tx_discards);
1045 }
1046
1047 static int
1048 iavf_dev_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
1049 {
1050         struct iavf_adapter *adapter =
1051                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1052         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
1053         struct iavf_vsi *vsi = &vf->vsi;
1054         struct virtchnl_eth_stats *pstats = NULL;
1055         int ret;
1056
1057         ret = iavf_query_stats(adapter, &pstats);
1058         if (ret == 0) {
1059                 iavf_update_stats(vsi, pstats);
1060                 stats->ipackets = pstats->rx_unicast + pstats->rx_multicast +
1061                                 pstats->rx_broadcast - pstats->rx_discards;
1062                 stats->opackets = pstats->tx_broadcast + pstats->tx_multicast +
1063                                                 pstats->tx_unicast;
1064                 stats->imissed = pstats->rx_discards;
1065                 stats->oerrors = pstats->tx_errors + pstats->tx_discards;
1066                 stats->ibytes = pstats->rx_bytes;
1067                 stats->ibytes -= stats->ipackets * RTE_ETHER_CRC_LEN;
1068                 stats->obytes = pstats->tx_bytes;
1069         } else {
1070                 PMD_DRV_LOG(ERR, "Get statistics failed");
1071         }
1072         return ret;
1073 }
1074
1075 static int
1076 iavf_dev_stats_reset(struct rte_eth_dev *dev)
1077 {
1078         int ret;
1079         struct iavf_adapter *adapter =
1080                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1081         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
1082         struct iavf_vsi *vsi = &vf->vsi;
1083         struct virtchnl_eth_stats *pstats = NULL;
1084
1085         /* read stat values to clear hardware registers */
1086         ret = iavf_query_stats(adapter, &pstats);
1087         if (ret != 0)
1088                 return ret;
1089
1090         /* set stats offset base on current values */
1091         vsi->eth_stats_offset = *pstats;
1092
1093         return 0;
1094 }
1095
1096 static int
1097 iavf_dev_rx_queue_intr_enable(struct rte_eth_dev *dev, uint16_t queue_id)
1098 {
1099         struct iavf_adapter *adapter =
1100                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1101         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1102         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(adapter);
1103         uint16_t msix_intr;
1104
1105         msix_intr = pci_dev->intr_handle.intr_vec[queue_id];
1106         if (msix_intr == IAVF_MISC_VEC_ID) {
1107                 PMD_DRV_LOG(INFO, "MISC is also enabled for control");
1108                 IAVF_WRITE_REG(hw, IAVF_VFINT_DYN_CTL01,
1109                                IAVF_VFINT_DYN_CTL01_INTENA_MASK |
1110                                IAVF_VFINT_DYN_CTL01_CLEARPBA_MASK |
1111                                IAVF_VFINT_DYN_CTL01_ITR_INDX_MASK);
1112         } else {
1113                 IAVF_WRITE_REG(hw,
1114                                IAVF_VFINT_DYN_CTLN1
1115                                 (msix_intr - IAVF_RX_VEC_START),
1116                                IAVF_VFINT_DYN_CTLN1_INTENA_MASK |
1117                                IAVF_VFINT_DYN_CTL01_CLEARPBA_MASK |
1118                                IAVF_VFINT_DYN_CTLN1_ITR_INDX_MASK);
1119         }
1120
1121         IAVF_WRITE_FLUSH(hw);
1122
1123         rte_intr_ack(&pci_dev->intr_handle);
1124
1125         return 0;
1126 }
1127
1128 static int
1129 iavf_dev_rx_queue_intr_disable(struct rte_eth_dev *dev, uint16_t queue_id)
1130 {
1131         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1132         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1133         uint16_t msix_intr;
1134
1135         msix_intr = pci_dev->intr_handle.intr_vec[queue_id];
1136         if (msix_intr == IAVF_MISC_VEC_ID) {
1137                 PMD_DRV_LOG(ERR, "MISC is used for control, cannot disable it");
1138                 return -EIO;
1139         }
1140
1141         IAVF_WRITE_REG(hw,
1142                       IAVF_VFINT_DYN_CTLN1(msix_intr - IAVF_RX_VEC_START),
1143                       0);
1144
1145         IAVF_WRITE_FLUSH(hw);
1146         return 0;
1147 }
1148
1149 static int
1150 iavf_check_vf_reset_done(struct iavf_hw *hw)
1151 {
1152         int i, reset;
1153
1154         for (i = 0; i < IAVF_RESET_WAIT_CNT; i++) {
1155                 reset = IAVF_READ_REG(hw, IAVF_VFGEN_RSTAT) &
1156                         IAVF_VFGEN_RSTAT_VFR_STATE_MASK;
1157                 reset = reset >> IAVF_VFGEN_RSTAT_VFR_STATE_SHIFT;
1158                 if (reset == VIRTCHNL_VFR_VFACTIVE ||
1159                     reset == VIRTCHNL_VFR_COMPLETED)
1160                         break;
1161                 rte_delay_ms(20);
1162         }
1163
1164         if (i >= IAVF_RESET_WAIT_CNT)
1165                 return -1;
1166
1167         return 0;
1168 }
1169
1170 static int
1171 iavf_init_vf(struct rte_eth_dev *dev)
1172 {
1173         int err, bufsz;
1174         struct iavf_adapter *adapter =
1175                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1176         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1177         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
1178
1179         err = iavf_set_mac_type(hw);
1180         if (err) {
1181                 PMD_INIT_LOG(ERR, "set_mac_type failed: %d", err);
1182                 goto err;
1183         }
1184
1185         err = iavf_check_vf_reset_done(hw);
1186         if (err) {
1187                 PMD_INIT_LOG(ERR, "VF is still resetting");
1188                 goto err;
1189         }
1190
1191         iavf_init_adminq_parameter(hw);
1192         err = iavf_init_adminq(hw);
1193         if (err) {
1194                 PMD_INIT_LOG(ERR, "init_adminq failed: %d", err);
1195                 goto err;
1196         }
1197
1198         vf->aq_resp = rte_zmalloc("vf_aq_resp", IAVF_AQ_BUF_SZ, 0);
1199         if (!vf->aq_resp) {
1200                 PMD_INIT_LOG(ERR, "unable to allocate vf_aq_resp memory");
1201                 goto err_aq;
1202         }
1203         if (iavf_check_api_version(adapter) != 0) {
1204                 PMD_INIT_LOG(ERR, "check_api version failed");
1205                 goto err_api;
1206         }
1207
1208         bufsz = sizeof(struct virtchnl_vf_resource) +
1209                 (IAVF_MAX_VF_VSI * sizeof(struct virtchnl_vsi_resource));
1210         vf->vf_res = rte_zmalloc("vf_res", bufsz, 0);
1211         if (!vf->vf_res) {
1212                 PMD_INIT_LOG(ERR, "unable to allocate vf_res memory");
1213                 goto err_api;
1214         }
1215         if (iavf_get_vf_resource(adapter) != 0) {
1216                 PMD_INIT_LOG(ERR, "iavf_get_vf_config failed");
1217                 goto err_alloc;
1218         }
1219         /* Allocate memort for RSS info */
1220         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF) {
1221                 vf->rss_key = rte_zmalloc("rss_key",
1222                                           vf->vf_res->rss_key_size, 0);
1223                 if (!vf->rss_key) {
1224                         PMD_INIT_LOG(ERR, "unable to allocate rss_key memory");
1225                         goto err_rss;
1226                 }
1227                 vf->rss_lut = rte_zmalloc("rss_lut",
1228                                           vf->vf_res->rss_lut_size, 0);
1229                 if (!vf->rss_lut) {
1230                         PMD_INIT_LOG(ERR, "unable to allocate rss_lut memory");
1231                         goto err_rss;
1232                 }
1233         }
1234
1235         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RX_FLEX_DESC) {
1236                 if (iavf_get_supported_rxdid(adapter) != 0) {
1237                         PMD_INIT_LOG(ERR, "failed to do get supported rxdid");
1238                         goto err_rss;
1239                 }
1240         }
1241
1242         return 0;
1243 err_rss:
1244         rte_free(vf->rss_key);
1245         rte_free(vf->rss_lut);
1246 err_alloc:
1247         rte_free(vf->vf_res);
1248         vf->vsi_res = NULL;
1249 err_api:
1250         rte_free(vf->aq_resp);
1251 err_aq:
1252         iavf_shutdown_adminq(hw);
1253 err:
1254         return -1;
1255 }
1256
1257 /* Enable default admin queue interrupt setting */
1258 static inline void
1259 iavf_enable_irq0(struct iavf_hw *hw)
1260 {
1261         /* Enable admin queue interrupt trigger */
1262         IAVF_WRITE_REG(hw, IAVF_VFINT_ICR0_ENA1,
1263                        IAVF_VFINT_ICR0_ENA1_ADMINQ_MASK);
1264
1265         IAVF_WRITE_REG(hw, IAVF_VFINT_DYN_CTL01,
1266                        IAVF_VFINT_DYN_CTL01_INTENA_MASK |
1267                        IAVF_VFINT_DYN_CTL01_CLEARPBA_MASK |
1268                        IAVF_VFINT_DYN_CTL01_ITR_INDX_MASK);
1269
1270         IAVF_WRITE_FLUSH(hw);
1271 }
1272
1273 static inline void
1274 iavf_disable_irq0(struct iavf_hw *hw)
1275 {
1276         /* Disable all interrupt types */
1277         IAVF_WRITE_REG(hw, IAVF_VFINT_ICR0_ENA1, 0);
1278         IAVF_WRITE_REG(hw, IAVF_VFINT_DYN_CTL01,
1279                        IAVF_VFINT_DYN_CTL01_ITR_INDX_MASK);
1280         IAVF_WRITE_FLUSH(hw);
1281 }
1282
1283 static void
1284 iavf_dev_interrupt_handler(void *param)
1285 {
1286         struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
1287         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1288
1289         iavf_disable_irq0(hw);
1290
1291         iavf_handle_virtchnl_msg(dev);
1292
1293         iavf_enable_irq0(hw);
1294 }
1295
1296 static int
1297 iavf_dev_filter_ctrl(struct rte_eth_dev *dev,
1298                      enum rte_filter_type filter_type,
1299                      enum rte_filter_op filter_op,
1300                      void *arg)
1301 {
1302         int ret = 0;
1303
1304         if (!dev)
1305                 return -EINVAL;
1306
1307         switch (filter_type) {
1308         case RTE_ETH_FILTER_GENERIC:
1309                 if (filter_op != RTE_ETH_FILTER_GET)
1310                         return -EINVAL;
1311                 *(const void **)arg = &iavf_flow_ops;
1312                 break;
1313         default:
1314                 PMD_DRV_LOG(WARNING, "Filter type (%d) not supported",
1315                             filter_type);
1316                 ret = -EINVAL;
1317                 break;
1318         }
1319
1320         return ret;
1321 }
1322
1323
1324 static int
1325 iavf_dev_init(struct rte_eth_dev *eth_dev)
1326 {
1327         struct iavf_adapter *adapter =
1328                 IAVF_DEV_PRIVATE_TO_ADAPTER(eth_dev->data->dev_private);
1329         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(adapter);
1330         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
1331         int ret = 0;
1332
1333         PMD_INIT_FUNC_TRACE();
1334
1335         /* assign ops func pointer */
1336         eth_dev->dev_ops = &iavf_eth_dev_ops;
1337         eth_dev->rx_pkt_burst = &iavf_recv_pkts;
1338         eth_dev->tx_pkt_burst = &iavf_xmit_pkts;
1339         eth_dev->tx_pkt_prepare = &iavf_prep_pkts;
1340
1341         /* For secondary processes, we don't initialise any further as primary
1342          * has already done this work. Only check if we need a different RX
1343          * and TX function.
1344          */
1345         if (rte_eal_process_type() != RTE_PROC_PRIMARY) {
1346                 iavf_set_rx_function(eth_dev);
1347                 iavf_set_tx_function(eth_dev);
1348                 return 0;
1349         }
1350         rte_eth_copy_pci_info(eth_dev, pci_dev);
1351
1352         hw->vendor_id = pci_dev->id.vendor_id;
1353         hw->device_id = pci_dev->id.device_id;
1354         hw->subsystem_vendor_id = pci_dev->id.subsystem_vendor_id;
1355         hw->subsystem_device_id = pci_dev->id.subsystem_device_id;
1356         hw->bus.bus_id = pci_dev->addr.bus;
1357         hw->bus.device = pci_dev->addr.devid;
1358         hw->bus.func = pci_dev->addr.function;
1359         hw->hw_addr = (void *)pci_dev->mem_resource[0].addr;
1360         hw->back = IAVF_DEV_PRIVATE_TO_ADAPTER(eth_dev->data->dev_private);
1361         adapter->eth_dev = eth_dev;
1362         adapter->stopped = 1;
1363
1364         if (iavf_init_vf(eth_dev) != 0) {
1365                 PMD_INIT_LOG(ERR, "Init vf failed");
1366                 return -1;
1367         }
1368
1369         /* set default ptype table */
1370         adapter->ptype_tbl = iavf_get_default_ptype_table();
1371
1372         /* copy mac addr */
1373         eth_dev->data->mac_addrs = rte_zmalloc(
1374                 "iavf_mac", RTE_ETHER_ADDR_LEN * IAVF_NUM_MACADDR_MAX, 0);
1375         if (!eth_dev->data->mac_addrs) {
1376                 PMD_INIT_LOG(ERR, "Failed to allocate %d bytes needed to"
1377                              " store MAC addresses",
1378                              RTE_ETHER_ADDR_LEN * IAVF_NUM_MACADDR_MAX);
1379                 return -ENOMEM;
1380         }
1381         /* If the MAC address is not configured by host,
1382          * generate a random one.
1383          */
1384         if (!rte_is_valid_assigned_ether_addr(
1385                         (struct rte_ether_addr *)hw->mac.addr))
1386                 rte_eth_random_addr(hw->mac.addr);
1387         rte_ether_addr_copy((struct rte_ether_addr *)hw->mac.addr,
1388                         &eth_dev->data->mac_addrs[0]);
1389
1390         /* register callback func to eal lib */
1391         rte_intr_callback_register(&pci_dev->intr_handle,
1392                                    iavf_dev_interrupt_handler,
1393                                    (void *)eth_dev);
1394
1395         /* enable uio intr after callback register */
1396         rte_intr_enable(&pci_dev->intr_handle);
1397
1398         /* configure and enable device interrupt */
1399         iavf_enable_irq0(hw);
1400
1401         ret = iavf_flow_init(adapter);
1402         if (ret) {
1403                 PMD_INIT_LOG(ERR, "Failed to initialize flow");
1404                 return ret;
1405         }
1406
1407         return 0;
1408 }
1409
1410 static void
1411 iavf_dev_close(struct rte_eth_dev *dev)
1412 {
1413         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1414         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1415         struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
1416         struct iavf_adapter *adapter =
1417                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1418
1419         iavf_dev_stop(dev);
1420         iavf_flow_flush(dev, NULL);
1421         iavf_flow_uninit(adapter);
1422         iavf_shutdown_adminq(hw);
1423         /* disable uio intr before callback unregister */
1424         rte_intr_disable(intr_handle);
1425
1426         /* unregister callback func from eal lib */
1427         rte_intr_callback_unregister(intr_handle,
1428                                      iavf_dev_interrupt_handler, dev);
1429         iavf_disable_irq0(hw);
1430 }
1431
1432 static int
1433 iavf_dev_uninit(struct rte_eth_dev *dev)
1434 {
1435         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
1436
1437         if (rte_eal_process_type() != RTE_PROC_PRIMARY)
1438                 return -EPERM;
1439
1440         dev->dev_ops = NULL;
1441         dev->rx_pkt_burst = NULL;
1442         dev->tx_pkt_burst = NULL;
1443         iavf_dev_close(dev);
1444
1445         rte_free(vf->vf_res);
1446         vf->vsi_res = NULL;
1447         vf->vf_res = NULL;
1448
1449         rte_free(vf->aq_resp);
1450         vf->aq_resp = NULL;
1451
1452         if (vf->rss_lut) {
1453                 rte_free(vf->rss_lut);
1454                 vf->rss_lut = NULL;
1455         }
1456         if (vf->rss_key) {
1457                 rte_free(vf->rss_key);
1458                 vf->rss_key = NULL;
1459         }
1460
1461         return 0;
1462 }
1463
1464 /*
1465  * Reset VF device only to re-initialize resources in PMD layer
1466  */
1467 static int
1468 iavf_dev_reset(struct rte_eth_dev *dev)
1469 {
1470         int ret;
1471
1472         ret = iavf_dev_uninit(dev);
1473         if (ret)
1474                 return ret;
1475
1476         return iavf_dev_init(dev);
1477 }
1478
1479 static int
1480 iavf_dcf_cap_check_handler(__rte_unused const char *key,
1481                            const char *value, __rte_unused void *opaque)
1482 {
1483         if (strcmp(value, "dcf"))
1484                 return -1;
1485
1486         return 0;
1487 }
1488
1489 static int
1490 iavf_dcf_cap_selected(struct rte_devargs *devargs)
1491 {
1492         struct rte_kvargs *kvlist;
1493         const char *key = "cap";
1494         int ret = 0;
1495
1496         if (devargs == NULL)
1497                 return 0;
1498
1499         kvlist = rte_kvargs_parse(devargs->args, NULL);
1500         if (kvlist == NULL)
1501                 return 0;
1502
1503         if (!rte_kvargs_count(kvlist, key))
1504                 goto exit;
1505
1506         /* dcf capability selected when there's a key-value pair: cap=dcf */
1507         if (rte_kvargs_process(kvlist, key,
1508                                iavf_dcf_cap_check_handler, NULL) < 0)
1509                 goto exit;
1510
1511         ret = 1;
1512
1513 exit:
1514         rte_kvargs_free(kvlist);
1515         return ret;
1516 }
1517
1518 static int eth_iavf_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
1519                              struct rte_pci_device *pci_dev)
1520 {
1521         if (iavf_dcf_cap_selected(pci_dev->device.devargs))
1522                 return 1;
1523
1524         return rte_eth_dev_pci_generic_probe(pci_dev,
1525                 sizeof(struct iavf_adapter), iavf_dev_init);
1526 }
1527
1528 static int eth_iavf_pci_remove(struct rte_pci_device *pci_dev)
1529 {
1530         return rte_eth_dev_pci_generic_remove(pci_dev, iavf_dev_uninit);
1531 }
1532
1533 /* Adaptive virtual function driver struct */
1534 static struct rte_pci_driver rte_iavf_pmd = {
1535         .id_table = pci_id_iavf_map,
1536         .drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC,
1537         .probe = eth_iavf_pci_probe,
1538         .remove = eth_iavf_pci_remove,
1539 };
1540
1541 RTE_PMD_REGISTER_PCI(net_iavf, rte_iavf_pmd);
1542 RTE_PMD_REGISTER_PCI_TABLE(net_iavf, pci_id_iavf_map);
1543 RTE_PMD_REGISTER_KMOD_DEP(net_iavf, "* igb_uio | vfio-pci");
1544 RTE_PMD_REGISTER_PARAM_STRING(net_iavf, "cap=dcf");
1545 RTE_LOG_REGISTER(iavf_logtype_init, pmd.net.iavf.init, NOTICE);
1546 RTE_LOG_REGISTER(iavf_logtype_driver, pmd.net.iavf.driver, NOTICE);
1547 #ifdef RTE_LIBRTE_IAVF_DEBUG_RX
1548 RTE_LOG_REGISTER(iavf_logtype_rx, pmd.net.iavf.rx, DEBUG);
1549 #endif
1550 #ifdef RTE_LIBRTE_IAVF_DEBUG_TX
1551 RTE_LOG_REGISTER(iavf_logtype_tx, pmd.net.iavf.tx, DEBUG);
1552 #endif
1553 #ifdef RTE_LIBRTE_IAVF_DEBUG_TX_FREE
1554 RTE_LOG_REGISTER(iavf_logtype_tx_free, pmd.net.iavf.tx_free, DEBUG);
1555 #endif