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 static int iavf_set_mc_addr_list(struct rte_eth_dev *dev,
  77                         struct rte_ether_addr *mc_addrs,
  78                         uint32_t mc_addrs_num);
  79
  80 static const struct rte_pci_id pci_id_iavf_map[] = {
  81         { RTE_PCI_DEVICE(IAVF_INTEL_VENDOR_ID, IAVF_DEV_ID_ADAPTIVE_VF) },
  82         { .vendor_id = 0, /* sentinel */ },
  83 };
  84
  85 static const struct eth_dev_ops iavf_eth_dev_ops = {
  86         .dev_configure              = iavf_dev_configure,
  87         .dev_start                  = iavf_dev_start,
  88         .dev_stop                   = iavf_dev_stop,
  89         .dev_close                  = iavf_dev_close,
  90         .dev_reset                  = iavf_dev_reset,
  91         .dev_infos_get              = iavf_dev_info_get,
  92         .dev_supported_ptypes_get   = iavf_dev_supported_ptypes_get,
  93         .link_update                = iavf_dev_link_update,
  94         .stats_get                  = iavf_dev_stats_get,
  95         .stats_reset                = iavf_dev_stats_reset,
  96         .promiscuous_enable         = iavf_dev_promiscuous_enable,
  97         .promiscuous_disable        = iavf_dev_promiscuous_disable,
  98         .allmulticast_enable        = iavf_dev_allmulticast_enable,
  99         .allmulticast_disable       = iavf_dev_allmulticast_disable,
 100         .mac_addr_add               = iavf_dev_add_mac_addr,
 101         .mac_addr_remove            = iavf_dev_del_mac_addr,
 102         .set_mc_addr_list                       = iavf_set_mc_addr_list,
 103         .vlan_filter_set            = iavf_dev_vlan_filter_set,
 104         .vlan_offload_set           = iavf_dev_vlan_offload_set,
 105         .rx_queue_start             = iavf_dev_rx_queue_start,
 106         .rx_queue_stop              = iavf_dev_rx_queue_stop,
 107         .tx_queue_start             = iavf_dev_tx_queue_start,
 108         .tx_queue_stop              = iavf_dev_tx_queue_stop,
 109         .rx_queue_setup             = iavf_dev_rx_queue_setup,
 110         .rx_queue_release           = iavf_dev_rx_queue_release,
 111         .tx_queue_setup             = iavf_dev_tx_queue_setup,
 112         .tx_queue_release           = iavf_dev_tx_queue_release,
 113         .mac_addr_set               = iavf_dev_set_default_mac_addr,
 114         .reta_update                = iavf_dev_rss_reta_update,
 115         .reta_query                 = iavf_dev_rss_reta_query,
 116         .rss_hash_update            = iavf_dev_rss_hash_update,
 117         .rss_hash_conf_get          = iavf_dev_rss_hash_conf_get,
 118         .rxq_info_get               = iavf_dev_rxq_info_get,
 119         .txq_info_get               = iavf_dev_txq_info_get,
 120         .rx_queue_count             = iavf_dev_rxq_count,
 121         .rx_descriptor_status       = iavf_dev_rx_desc_status,
 122         .tx_descriptor_status       = iavf_dev_tx_desc_status,
 123         .mtu_set                    = iavf_dev_mtu_set,
 124         .rx_queue_intr_enable       = iavf_dev_rx_queue_intr_enable,
 125         .rx_queue_intr_disable      = iavf_dev_rx_queue_intr_disable,
 126         .filter_ctrl                = iavf_dev_filter_ctrl,
 127 };
 128
 129 static int
 130 iavf_set_mc_addr_list(struct rte_eth_dev *dev,
 131                         struct rte_ether_addr *mc_addrs,
 132                         uint32_t mc_addrs_num)
 133 {
 134         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
 135         struct iavf_adapter *adapter =
 136                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 137         int err;
 138
 139         /* flush previous addresses */
 140         err = iavf_add_del_mc_addr_list(adapter, vf->mc_addrs, vf->mc_addrs_num,
 141                                         false);
 142         if (err)
 143                 return err;
 144
 145         vf->mc_addrs_num = 0;
 146
 147         /* add new ones */
 148         err = iavf_add_del_mc_addr_list(adapter, mc_addrs, mc_addrs_num, true);
 149         if (err)
 150                 return err;
 151
 152         vf->mc_addrs_num = mc_addrs_num;
 153         memcpy(vf->mc_addrs, mc_addrs, mc_addrs_num * sizeof(*mc_addrs));
 154
 155         return 0;
 156 }
 157
 158 static int
 159 iavf_init_rss(struct iavf_adapter *adapter)
 160 {
 161         struct iavf_info *vf =  IAVF_DEV_PRIVATE_TO_VF(adapter);
 162         struct rte_eth_rss_conf *rss_conf;
 163         uint8_t i, j, nb_q;
 164         int ret;
 165
 166         rss_conf = &adapter->eth_dev->data->dev_conf.rx_adv_conf.rss_conf;
 167         nb_q = RTE_MIN(adapter->eth_dev->data->nb_rx_queues,
 168                        IAVF_MAX_NUM_QUEUES);
 169
 170         if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF)) {
 171                 PMD_DRV_LOG(DEBUG, "RSS is not supported");
 172                 return -ENOTSUP;
 173         }
 174         if (adapter->eth_dev->data->dev_conf.rxmode.mq_mode != ETH_MQ_RX_RSS) {
 175                 PMD_DRV_LOG(WARNING, "RSS is enabled by PF by default");
 176                 /* set all lut items to default queue */
 177                 for (i = 0; i < vf->vf_res->rss_lut_size; i++)
 178                         vf->rss_lut[i] = 0;
 179                 ret = iavf_configure_rss_lut(adapter);
 180                 return ret;
 181         }
 182
 183         /* In IAVF, RSS enablement is set by PF driver. It is not supported
 184          * to set based on rss_conf->rss_hf.
 185          */
 186
 187         /* configure RSS key */
 188         if (!rss_conf->rss_key) {
 189                 /* Calculate the default hash key */
 190                 for (i = 0; i <= vf->vf_res->rss_key_size; i++)
 191                         vf->rss_key[i] = (uint8_t)rte_rand();
 192         } else
 193                 rte_memcpy(vf->rss_key, rss_conf->rss_key,
 194                            RTE_MIN(rss_conf->rss_key_len,
 195                                    vf->vf_res->rss_key_size));
 196
 197         /* init RSS LUT table */
 198         for (i = 0, j = 0; i < vf->vf_res->rss_lut_size; i++, j++) {
 199                 if (j >= nb_q)
 200                         j = 0;
 201                 vf->rss_lut[i] = j;
 202         }
 203         /* send virtchnnl ops to configure rss*/
 204         ret = iavf_configure_rss_lut(adapter);
 205         if (ret)
 206                 return ret;
 207         ret = iavf_configure_rss_key(adapter);
 208         if (ret)
 209                 return ret;
 210
 211         return 0;
 212 }
 213
 214 static int
 215 iavf_dev_configure(struct rte_eth_dev *dev)
 216 {
 217         struct iavf_adapter *ad =
 218                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 219         struct iavf_info *vf =  IAVF_DEV_PRIVATE_TO_VF(ad);
 220         struct rte_eth_conf *dev_conf = &dev->data->dev_conf;
 221
 222         ad->rx_bulk_alloc_allowed = true;
 223         /* Initialize to TRUE. If any of Rx queues doesn't meet the
 224          * vector Rx/Tx preconditions, it will be reset.
 225          */
 226         ad->rx_vec_allowed = true;
 227         ad->tx_vec_allowed = true;
 228
 229         if (dev->data->dev_conf.rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG)
 230                 dev->data->dev_conf.rxmode.offloads |= DEV_RX_OFFLOAD_RSS_HASH;
 231
 232         /* Vlan stripping setting */
 233         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_VLAN) {
 234                 if (dev_conf->rxmode.offloads & DEV_RX_OFFLOAD_VLAN_STRIP)
 235                         iavf_enable_vlan_strip(ad);
 236                 else
 237                         iavf_disable_vlan_strip(ad);
 238         }
 239
 240         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF) {
 241                 if (iavf_init_rss(ad) != 0) {
 242                         PMD_DRV_LOG(ERR, "configure rss failed");
 243                         return -1;
 244                 }
 245         }
 246         return 0;
 247 }
 248
 249 static int
 250 iavf_init_rxq(struct rte_eth_dev *dev, struct iavf_rx_queue *rxq)
 251 {
 252         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(dev->data->dev_private);
 253         struct rte_eth_dev_data *dev_data = dev->data;
 254         uint16_t buf_size, max_pkt_len, len;
 255
 256         buf_size = rte_pktmbuf_data_room_size(rxq->mp) - RTE_PKTMBUF_HEADROOM;
 257
 258         /* Calculate the maximum packet length allowed */
 259         len = rxq->rx_buf_len * IAVF_MAX_CHAINED_RX_BUFFERS;
 260         max_pkt_len = RTE_MIN(len, dev->data->dev_conf.rxmode.max_rx_pkt_len);
 261
 262         /* Check if the jumbo frame and maximum packet length are set
 263          * correctly.
 264          */
 265         if (dev->data->dev_conf.rxmode.offloads & DEV_RX_OFFLOAD_JUMBO_FRAME) {
 266                 if (max_pkt_len <= RTE_ETHER_MAX_LEN ||
 267                     max_pkt_len > IAVF_FRAME_SIZE_MAX) {
 268                         PMD_DRV_LOG(ERR, "maximum packet length must be "
 269                                     "larger than %u and smaller than %u, "
 270                                     "as jumbo frame is enabled",
 271                                     (uint32_t)RTE_ETHER_MAX_LEN,
 272                                     (uint32_t)IAVF_FRAME_SIZE_MAX);
 273                         return -EINVAL;
 274                 }
 275         } else {
 276                 if (max_pkt_len < RTE_ETHER_MIN_LEN ||
 277                     max_pkt_len > RTE_ETHER_MAX_LEN) {
 278                         PMD_DRV_LOG(ERR, "maximum packet length must be "
 279                                     "larger than %u and smaller than %u, "
 280                                     "as jumbo frame is disabled",
 281                                     (uint32_t)RTE_ETHER_MIN_LEN,
 282                                     (uint32_t)RTE_ETHER_MAX_LEN);
 283                         return -EINVAL;
 284                 }
 285         }
 286
 287         rxq->max_pkt_len = max_pkt_len;
 288         if ((dev_data->dev_conf.rxmode.offloads & DEV_RX_OFFLOAD_SCATTER) ||
 289             rxq->max_pkt_len > buf_size) {
 290                 dev_data->scattered_rx = 1;
 291         }
 292         IAVF_PCI_REG_WRITE(rxq->qrx_tail, rxq->nb_rx_desc - 1);
 293         IAVF_WRITE_FLUSH(hw);
 294
 295         return 0;
 296 }
 297
 298 static int
 299 iavf_init_queues(struct rte_eth_dev *dev)
 300 {
 301         struct iavf_rx_queue **rxq =
 302                 (struct iavf_rx_queue **)dev->data->rx_queues;
 303         int i, ret = IAVF_SUCCESS;
 304
 305         for (i = 0; i < dev->data->nb_rx_queues; i++) {
 306                 if (!rxq[i] || !rxq[i]->q_set)
 307                         continue;
 308                 ret = iavf_init_rxq(dev, rxq[i]);
 309                 if (ret != IAVF_SUCCESS)
 310                         break;
 311         }
 312         /* set rx/tx function to vector/scatter/single-segment
 313          * according to parameters
 314          */
 315         iavf_set_rx_function(dev);
 316         iavf_set_tx_function(dev);
 317
 318         return ret;
 319 }
 320
 321 static int iavf_config_rx_queues_irqs(struct rte_eth_dev *dev,
 322                                      struct rte_intr_handle *intr_handle)
 323 {
 324         struct iavf_adapter *adapter =
 325                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 326         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
 327         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(adapter);
 328         uint16_t interval, i;
 329         int vec;
 330
 331         if (rte_intr_cap_multiple(intr_handle) &&
 332             dev->data->dev_conf.intr_conf.rxq) {
 333                 if (rte_intr_efd_enable(intr_handle, dev->data->nb_rx_queues))
 334                         return -1;
 335         }
 336
 337         if (rte_intr_dp_is_en(intr_handle) && !intr_handle->intr_vec) {
 338                 intr_handle->intr_vec =
 339                         rte_zmalloc("intr_vec",
 340                                     dev->data->nb_rx_queues * sizeof(int), 0);
 341                 if (!intr_handle->intr_vec) {
 342                         PMD_DRV_LOG(ERR, "Failed to allocate %d rx intr_vec",
 343                                     dev->data->nb_rx_queues);
 344                         return -1;
 345                 }
 346         }
 347
 348         if (!dev->data->dev_conf.intr_conf.rxq ||
 349             !rte_intr_dp_is_en(intr_handle)) {
 350                 /* Rx interrupt disabled, Map interrupt only for writeback */
 351                 vf->nb_msix = 1;
 352                 if (vf->vf_res->vf_cap_flags &
 353                     VIRTCHNL_VF_OFFLOAD_WB_ON_ITR) {
 354                         /* If WB_ON_ITR supports, enable it */
 355                         vf->msix_base = IAVF_RX_VEC_START;
 356                         IAVF_WRITE_REG(hw,
 357                                        IAVF_VFINT_DYN_CTLN1(vf->msix_base - 1),
 358                                        IAVF_VFINT_DYN_CTLN1_ITR_INDX_MASK |
 359                                        IAVF_VFINT_DYN_CTLN1_WB_ON_ITR_MASK);
 360                 } else {
 361                         /* If no WB_ON_ITR offload flags, need to set
 362                          * interrupt for descriptor write back.
 363                          */
 364                         vf->msix_base = IAVF_MISC_VEC_ID;
 365
 366                         /* set ITR to max */
 367                         interval = iavf_calc_itr_interval(
 368                                         IAVF_QUEUE_ITR_INTERVAL_MAX);
 369                         IAVF_WRITE_REG(hw, IAVF_VFINT_DYN_CTL01,
 370                                        IAVF_VFINT_DYN_CTL01_INTENA_MASK |
 371                                        (IAVF_ITR_INDEX_DEFAULT <<
 372                                         IAVF_VFINT_DYN_CTL01_ITR_INDX_SHIFT) |
 373                                        (interval <<
 374                                         IAVF_VFINT_DYN_CTL01_INTERVAL_SHIFT));
 375                 }
 376                 IAVF_WRITE_FLUSH(hw);
 377                 /* map all queues to the same interrupt */
 378                 for (i = 0; i < dev->data->nb_rx_queues; i++)
 379                         vf->rxq_map[vf->msix_base] |= 1 << i;
 380         } else {
 381                 if (!rte_intr_allow_others(intr_handle)) {
 382                         vf->nb_msix = 1;
 383                         vf->msix_base = IAVF_MISC_VEC_ID;
 384                         for (i = 0; i < dev->data->nb_rx_queues; i++) {
 385                                 vf->rxq_map[vf->msix_base] |= 1 << i;
 386                                 intr_handle->intr_vec[i] = IAVF_MISC_VEC_ID;
 387                         }
 388                         PMD_DRV_LOG(DEBUG,
 389                                     "vector %u are mapping to all Rx queues",
 390                                     vf->msix_base);
 391                 } else {
 392                         /* If Rx interrupt is reuquired, and we can use
 393                          * multi interrupts, then the vec is from 1
 394                          */
 395                         vf->nb_msix = RTE_MIN(vf->vf_res->max_vectors,
 396                                               intr_handle->nb_efd);
 397                         vf->msix_base = IAVF_RX_VEC_START;
 398                         vec = IAVF_RX_VEC_START;
 399                         for (i = 0; i < dev->data->nb_rx_queues; i++) {
 400                                 vf->rxq_map[vec] |= 1 << i;
 401                                 intr_handle->intr_vec[i] = vec++;
 402                                 if (vec >= vf->nb_msix)
 403                                         vec = IAVF_RX_VEC_START;
 404                         }
 405                         PMD_DRV_LOG(DEBUG,
 406                                     "%u vectors are mapping to %u Rx queues",
 407                                     vf->nb_msix, dev->data->nb_rx_queues);
 408                 }
 409         }
 410
 411         if (iavf_config_irq_map(adapter)) {
 412                 PMD_DRV_LOG(ERR, "config interrupt mapping failed");
 413                 return -1;
 414         }
 415         return 0;
 416 }
 417
 418 static int
 419 iavf_start_queues(struct rte_eth_dev *dev)
 420 {
 421         struct iavf_rx_queue *rxq;
 422         struct iavf_tx_queue *txq;
 423         int i;
 424
 425         for (i = 0; i < dev->data->nb_tx_queues; i++) {
 426                 txq = dev->data->tx_queues[i];
 427                 if (txq->tx_deferred_start)
 428                         continue;
 429                 if (iavf_dev_tx_queue_start(dev, i) != 0) {
 430                         PMD_DRV_LOG(ERR, "Fail to start queue %u", i);
 431                         return -1;
 432                 }
 433         }
 434
 435         for (i = 0; i < dev->data->nb_rx_queues; i++) {
 436                 rxq = dev->data->rx_queues[i];
 437                 if (rxq->rx_deferred_start)
 438                         continue;
 439                 if (iavf_dev_rx_queue_start(dev, i) != 0) {
 440                         PMD_DRV_LOG(ERR, "Fail to start queue %u", i);
 441                         return -1;
 442                 }
 443         }
 444
 445         return 0;
 446 }
 447
 448 static int
 449 iavf_dev_start(struct rte_eth_dev *dev)
 450 {
 451         struct iavf_adapter *adapter =
 452                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 453         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
 454         struct rte_intr_handle *intr_handle = dev->intr_handle;
 455
 456         PMD_INIT_FUNC_TRACE();
 457
 458         adapter->stopped = 0;
 459
 460         vf->max_pkt_len = dev->data->dev_conf.rxmode.max_rx_pkt_len;
 461         vf->num_queue_pairs = RTE_MAX(dev->data->nb_rx_queues,
 462                                       dev->data->nb_tx_queues);
 463
 464         if (iavf_init_queues(dev) != 0) {
 465                 PMD_DRV_LOG(ERR, "failed to do Queue init");
 466                 return -1;
 467         }
 468
 469         if (iavf_configure_queues(adapter) != 0) {
 470                 PMD_DRV_LOG(ERR, "configure queues failed");
 471                 goto err_queue;
 472         }
 473
 474         if (iavf_config_rx_queues_irqs(dev, intr_handle) != 0) {
 475                 PMD_DRV_LOG(ERR, "configure irq failed");
 476                 goto err_queue;
 477         }
 478         /* re-enable intr again, because efd assign may change */
 479         if (dev->data->dev_conf.intr_conf.rxq != 0) {
 480                 rte_intr_disable(intr_handle);
 481                 rte_intr_enable(intr_handle);
 482         }
 483
 484         /* Set all mac addrs */
 485         iavf_add_del_all_mac_addr(adapter, true);
 486
 487         /* Set all multicast addresses */
 488         iavf_add_del_mc_addr_list(adapter, vf->mc_addrs, vf->mc_addrs_num,
 489                                   true);
 490
 491         if (iavf_start_queues(dev) != 0) {
 492                 PMD_DRV_LOG(ERR, "enable queues failed");
 493                 goto err_mac;
 494         }
 495
 496         return 0;
 497
 498 err_mac:
 499         iavf_add_del_all_mac_addr(adapter, false);
 500 err_queue:
 501         return -1;
 502 }
 503
 504 static void
 505 iavf_dev_stop(struct rte_eth_dev *dev)
 506 {
 507         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
 508         struct iavf_adapter *adapter =
 509                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 510         struct rte_intr_handle *intr_handle = dev->intr_handle;
 511
 512         PMD_INIT_FUNC_TRACE();
 513
 514         if (adapter->stopped == 1)
 515                 return;
 516
 517         iavf_stop_queues(dev);
 518
 519         /* Disable the interrupt for Rx */
 520         rte_intr_efd_disable(intr_handle);
 521         /* Rx interrupt vector mapping free */
 522         if (intr_handle->intr_vec) {
 523                 rte_free(intr_handle->intr_vec);
 524                 intr_handle->intr_vec = NULL;
 525         }
 526
 527         /* remove all mac addrs */
 528         iavf_add_del_all_mac_addr(adapter, false);
 529
 530         /* remove all multicast addresses */
 531         iavf_add_del_mc_addr_list(adapter, vf->mc_addrs, vf->mc_addrs_num,
 532                                   false);
 533
 534         adapter->stopped = 1;
 535 }
 536
 537 static int
 538 iavf_dev_info_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
 539 {
 540         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
 541
 542         dev_info->max_rx_queues = vf->vsi_res->num_queue_pairs;
 543         dev_info->max_tx_queues = vf->vsi_res->num_queue_pairs;
 544         dev_info->min_rx_bufsize = IAVF_BUF_SIZE_MIN;
 545         dev_info->max_rx_pktlen = IAVF_FRAME_SIZE_MAX;
 546         dev_info->hash_key_size = vf->vf_res->rss_key_size;
 547         dev_info->reta_size = vf->vf_res->rss_lut_size;
 548         dev_info->flow_type_rss_offloads = IAVF_RSS_OFFLOAD_ALL;
 549         dev_info->max_mac_addrs = IAVF_NUM_MACADDR_MAX;
 550         dev_info->rx_offload_capa =
 551                 DEV_RX_OFFLOAD_VLAN_STRIP |
 552                 DEV_RX_OFFLOAD_QINQ_STRIP |
 553                 DEV_RX_OFFLOAD_IPV4_CKSUM |
 554                 DEV_RX_OFFLOAD_UDP_CKSUM |
 555                 DEV_RX_OFFLOAD_TCP_CKSUM |
 556                 DEV_RX_OFFLOAD_OUTER_IPV4_CKSUM |
 557                 DEV_RX_OFFLOAD_SCATTER |
 558                 DEV_RX_OFFLOAD_JUMBO_FRAME |
 559                 DEV_RX_OFFLOAD_VLAN_FILTER |
 560                 DEV_RX_OFFLOAD_RSS_HASH;
 561         dev_info->tx_offload_capa =
 562                 DEV_TX_OFFLOAD_VLAN_INSERT |
 563                 DEV_TX_OFFLOAD_QINQ_INSERT |
 564                 DEV_TX_OFFLOAD_IPV4_CKSUM |
 565                 DEV_TX_OFFLOAD_UDP_CKSUM |
 566                 DEV_TX_OFFLOAD_TCP_CKSUM |
 567                 DEV_TX_OFFLOAD_SCTP_CKSUM |
 568                 DEV_TX_OFFLOAD_OUTER_IPV4_CKSUM |
 569                 DEV_TX_OFFLOAD_TCP_TSO |
 570                 DEV_TX_OFFLOAD_VXLAN_TNL_TSO |
 571                 DEV_TX_OFFLOAD_GRE_TNL_TSO |
 572                 DEV_TX_OFFLOAD_IPIP_TNL_TSO |
 573                 DEV_TX_OFFLOAD_GENEVE_TNL_TSO |
 574                 DEV_TX_OFFLOAD_MULTI_SEGS;
 575
 576         dev_info->default_rxconf = (struct rte_eth_rxconf) {
 577                 .rx_free_thresh = IAVF_DEFAULT_RX_FREE_THRESH,
 578                 .rx_drop_en = 0,
 579                 .offloads = 0,
 580         };
 581
 582         dev_info->default_txconf = (struct rte_eth_txconf) {
 583                 .tx_free_thresh = IAVF_DEFAULT_TX_FREE_THRESH,
 584                 .tx_rs_thresh = IAVF_DEFAULT_TX_RS_THRESH,
 585                 .offloads = 0,
 586         };
 587
 588         dev_info->rx_desc_lim = (struct rte_eth_desc_lim) {
 589                 .nb_max = IAVF_MAX_RING_DESC,
 590                 .nb_min = IAVF_MIN_RING_DESC,
 591                 .nb_align = IAVF_ALIGN_RING_DESC,
 592         };
 593
 594         dev_info->tx_desc_lim = (struct rte_eth_desc_lim) {
 595                 .nb_max = IAVF_MAX_RING_DESC,
 596                 .nb_min = IAVF_MIN_RING_DESC,
 597                 .nb_align = IAVF_ALIGN_RING_DESC,
 598         };
 599
 600         return 0;
 601 }
 602
 603 static const uint32_t *
 604 iavf_dev_supported_ptypes_get(struct rte_eth_dev *dev __rte_unused)
 605 {
 606         static const uint32_t ptypes[] = {
 607                 RTE_PTYPE_L2_ETHER,
 608                 RTE_PTYPE_L3_IPV4_EXT_UNKNOWN,
 609                 RTE_PTYPE_L4_FRAG,
 610                 RTE_PTYPE_L4_ICMP,
 611                 RTE_PTYPE_L4_NONFRAG,
 612                 RTE_PTYPE_L4_SCTP,
 613                 RTE_PTYPE_L4_TCP,
 614                 RTE_PTYPE_L4_UDP,
 615                 RTE_PTYPE_UNKNOWN
 616         };
 617         return ptypes;
 618 }
 619
 620 int
 621 iavf_dev_link_update(struct rte_eth_dev *dev,
 622                     __rte_unused int wait_to_complete)
 623 {
 624         struct rte_eth_link new_link;
 625         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
 626
 627         memset(&new_link, 0, sizeof(new_link));
 628
 629         /* Only read status info stored in VF, and the info is updated
 630          *  when receive LINK_CHANGE evnet from PF by Virtchnnl.
 631          */
 632         switch (vf->link_speed) {
 633         case 10:
 634                 new_link.link_speed = ETH_SPEED_NUM_10M;
 635                 break;
 636         case 100:
 637                 new_link.link_speed = ETH_SPEED_NUM_100M;
 638                 break;
 639         case 1000:
 640                 new_link.link_speed = ETH_SPEED_NUM_1G;
 641                 break;
 642         case 10000:
 643                 new_link.link_speed = ETH_SPEED_NUM_10G;
 644                 break;
 645         case 20000:
 646                 new_link.link_speed = ETH_SPEED_NUM_20G;
 647                 break;
 648         case 25000:
 649                 new_link.link_speed = ETH_SPEED_NUM_25G;
 650                 break;
 651         case 40000:
 652                 new_link.link_speed = ETH_SPEED_NUM_40G;
 653                 break;
 654         case 50000:
 655                 new_link.link_speed = ETH_SPEED_NUM_50G;
 656                 break;
 657         case 100000:
 658                 new_link.link_speed = ETH_SPEED_NUM_100G;
 659                 break;
 660         default:
 661                 new_link.link_speed = ETH_SPEED_NUM_NONE;
 662                 break;
 663         }
 664
 665         new_link.link_duplex = ETH_LINK_FULL_DUPLEX;
 666         new_link.link_status = vf->link_up ? ETH_LINK_UP :
 667                                              ETH_LINK_DOWN;
 668         new_link.link_autoneg = !(dev->data->dev_conf.link_speeds &
 669                                 ETH_LINK_SPEED_FIXED);
 670
 671         if (rte_atomic64_cmpset((uint64_t *)&dev->data->dev_link,
 672                                 *(uint64_t *)&dev->data->dev_link,
 673                                 *(uint64_t *)&new_link) == 0)
 674                 return -1;
 675
 676         return 0;
 677 }
 678
 679 static int
 680 iavf_dev_promiscuous_enable(struct rte_eth_dev *dev)
 681 {
 682         struct iavf_adapter *adapter =
 683                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 684         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
 685         int ret;
 686
 687         if (vf->promisc_unicast_enabled)
 688                 return 0;
 689
 690         ret = iavf_config_promisc(adapter, true, vf->promisc_multicast_enabled);
 691         if (!ret)
 692                 vf->promisc_unicast_enabled = true;
 693         else
 694                 ret = -EAGAIN;
 695
 696         return ret;
 697 }
 698
 699 static int
 700 iavf_dev_promiscuous_disable(struct rte_eth_dev *dev)
 701 {
 702         struct iavf_adapter *adapter =
 703                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 704         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
 705         int ret;
 706
 707         if (!vf->promisc_unicast_enabled)
 708                 return 0;
 709
 710         ret = iavf_config_promisc(adapter, false,
 711                                   vf->promisc_multicast_enabled);
 712         if (!ret)
 713                 vf->promisc_unicast_enabled = false;
 714         else
 715                 ret = -EAGAIN;
 716
 717         return ret;
 718 }
 719
 720 static int
 721 iavf_dev_allmulticast_enable(struct rte_eth_dev *dev)
 722 {
 723         struct iavf_adapter *adapter =
 724                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 725         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
 726         int ret;
 727
 728         if (vf->promisc_multicast_enabled)
 729                 return 0;
 730
 731         ret = iavf_config_promisc(adapter, vf->promisc_unicast_enabled, true);
 732         if (!ret)
 733                 vf->promisc_multicast_enabled = true;
 734         else
 735                 ret = -EAGAIN;
 736
 737         return ret;
 738 }
 739
 740 static int
 741 iavf_dev_allmulticast_disable(struct rte_eth_dev *dev)
 742 {
 743         struct iavf_adapter *adapter =
 744                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 745         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
 746         int ret;
 747
 748         if (!vf->promisc_multicast_enabled)
 749                 return 0;
 750
 751         ret = iavf_config_promisc(adapter, vf->promisc_unicast_enabled, false);
 752         if (!ret)
 753                 vf->promisc_multicast_enabled = false;
 754         else
 755                 ret = -EAGAIN;
 756
 757         return ret;
 758 }
 759
 760 static int
 761 iavf_dev_add_mac_addr(struct rte_eth_dev *dev, struct rte_ether_addr *addr,
 762                      __rte_unused uint32_t index,
 763                      __rte_unused uint32_t pool)
 764 {
 765         struct iavf_adapter *adapter =
 766                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 767         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
 768         int err;
 769
 770         if (rte_is_zero_ether_addr(addr)) {
 771                 PMD_DRV_LOG(ERR, "Invalid Ethernet Address");
 772                 return -EINVAL;
 773         }
 774
 775         err = iavf_add_del_eth_addr(adapter, addr, true);
 776         if (err) {
 777                 PMD_DRV_LOG(ERR, "fail to add MAC address");
 778                 return -EIO;
 779         }
 780
 781         vf->mac_num++;
 782
 783         return 0;
 784 }
 785
 786 static void
 787 iavf_dev_del_mac_addr(struct rte_eth_dev *dev, uint32_t index)
 788 {
 789         struct iavf_adapter *adapter =
 790                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 791         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
 792         struct rte_ether_addr *addr;
 793         int err;
 794
 795         addr = &dev->data->mac_addrs[index];
 796
 797         err = iavf_add_del_eth_addr(adapter, addr, false);
 798         if (err)
 799                 PMD_DRV_LOG(ERR, "fail to delete MAC address");
 800
 801         vf->mac_num--;
 802 }
 803
 804 static int
 805 iavf_dev_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
 806 {
 807         struct iavf_adapter *adapter =
 808                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 809         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
 810         int err;
 811
 812         if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_VLAN))
 813                 return -ENOTSUP;
 814
 815         err = iavf_add_del_vlan(adapter, vlan_id, on);
 816         if (err)
 817                 return -EIO;
 818         return 0;
 819 }
 820
 821 static int
 822 iavf_dev_vlan_offload_set(struct rte_eth_dev *dev, int mask)
 823 {
 824         struct iavf_adapter *adapter =
 825                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 826         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
 827         struct rte_eth_conf *dev_conf = &dev->data->dev_conf;
 828         int err;
 829
 830         if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_VLAN))
 831                 return -ENOTSUP;
 832
 833         /* Vlan stripping setting */
 834         if (mask & ETH_VLAN_STRIP_MASK) {
 835                 /* Enable or disable VLAN stripping */
 836                 if (dev_conf->rxmode.offloads & DEV_RX_OFFLOAD_VLAN_STRIP)
 837                         err = iavf_enable_vlan_strip(adapter);
 838                 else
 839                         err = iavf_disable_vlan_strip(adapter);
 840
 841                 if (err)
 842                         return -EIO;
 843         }
 844         return 0;
 845 }
 846
 847 static int
 848 iavf_dev_rss_reta_update(struct rte_eth_dev *dev,
 849                         struct rte_eth_rss_reta_entry64 *reta_conf,
 850                         uint16_t reta_size)
 851 {
 852         struct iavf_adapter *adapter =
 853                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 854         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
 855         uint8_t *lut;
 856         uint16_t i, idx, shift;
 857         int ret;
 858
 859         if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF))
 860                 return -ENOTSUP;
 861
 862         if (reta_size != vf->vf_res->rss_lut_size) {
 863                 PMD_DRV_LOG(ERR, "The size of hash lookup table configured "
 864                         "(%d) doesn't match the number of hardware can "
 865                         "support (%d)", reta_size, vf->vf_res->rss_lut_size);
 866                 return -EINVAL;
 867         }
 868
 869         lut = rte_zmalloc("rss_lut", reta_size, 0);
 870         if (!lut) {
 871                 PMD_DRV_LOG(ERR, "No memory can be allocated");
 872                 return -ENOMEM;
 873         }
 874         /* store the old lut table temporarily */
 875         rte_memcpy(lut, vf->rss_lut, reta_size);
 876
 877         for (i = 0; i < reta_size; i++) {
 878                 idx = i / RTE_RETA_GROUP_SIZE;
 879                 shift = i % RTE_RETA_GROUP_SIZE;
 880                 if (reta_conf[idx].mask & (1ULL << shift))
 881                         lut[i] = reta_conf[idx].reta[shift];
 882         }
 883
 884         rte_memcpy(vf->rss_lut, lut, reta_size);
 885         /* send virtchnnl ops to configure rss*/
 886         ret = iavf_configure_rss_lut(adapter);
 887         if (ret) /* revert back */
 888                 rte_memcpy(vf->rss_lut, lut, reta_size);
 889         rte_free(lut);
 890
 891         return ret;
 892 }
 893
 894 static int
 895 iavf_dev_rss_reta_query(struct rte_eth_dev *dev,
 896                        struct rte_eth_rss_reta_entry64 *reta_conf,
 897                        uint16_t reta_size)
 898 {
 899         struct iavf_adapter *adapter =
 900                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 901         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
 902         uint16_t i, idx, shift;
 903
 904         if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF))
 905                 return -ENOTSUP;
 906
 907         if (reta_size != vf->vf_res->rss_lut_size) {
 908                 PMD_DRV_LOG(ERR, "The size of hash lookup table configured "
 909                         "(%d) doesn't match the number of hardware can "
 910                         "support (%d)", reta_size, vf->vf_res->rss_lut_size);
 911                 return -EINVAL;
 912         }
 913
 914         for (i = 0; i < reta_size; i++) {
 915                 idx = i / RTE_RETA_GROUP_SIZE;
 916                 shift = i % RTE_RETA_GROUP_SIZE;
 917                 if (reta_conf[idx].mask & (1ULL << shift))
 918                         reta_conf[idx].reta[shift] = vf->rss_lut[i];
 919         }
 920
 921         return 0;
 922 }
 923
 924 static int
 925 iavf_dev_rss_hash_update(struct rte_eth_dev *dev,
 926                         struct rte_eth_rss_conf *rss_conf)
 927 {
 928         struct iavf_adapter *adapter =
 929                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 930         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
 931
 932         if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF))
 933                 return -ENOTSUP;
 934
 935         /* HENA setting, it is enabled by default, no change */
 936         if (!rss_conf->rss_key || rss_conf->rss_key_len == 0) {
 937                 PMD_DRV_LOG(DEBUG, "No key to be configured");
 938                 return 0;
 939         } else if (rss_conf->rss_key_len != vf->vf_res->rss_key_size) {
 940                 PMD_DRV_LOG(ERR, "The size of hash key configured "
 941                         "(%d) doesn't match the size of hardware can "
 942                         "support (%d)", rss_conf->rss_key_len,
 943                         vf->vf_res->rss_key_size);
 944                 return -EINVAL;
 945         }
 946
 947         rte_memcpy(vf->rss_key, rss_conf->rss_key, rss_conf->rss_key_len);
 948
 949         return iavf_configure_rss_key(adapter);
 950 }
 951
 952 static int
 953 iavf_dev_rss_hash_conf_get(struct rte_eth_dev *dev,
 954                           struct rte_eth_rss_conf *rss_conf)
 955 {
 956         struct iavf_adapter *adapter =
 957                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 958         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
 959
 960         if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF))
 961                 return -ENOTSUP;
 962
 963          /* Just set it to default value now. */
 964         rss_conf->rss_hf = IAVF_RSS_OFFLOAD_ALL;
 965
 966         if (!rss_conf->rss_key)
 967                 return 0;
 968
 969         rss_conf->rss_key_len = vf->vf_res->rss_key_size;
 970         rte_memcpy(rss_conf->rss_key, vf->rss_key, rss_conf->rss_key_len);
 971
 972         return 0;
 973 }
 974
 975 static int
 976 iavf_dev_mtu_set(struct rte_eth_dev *dev, uint16_t mtu)
 977 {
 978         uint32_t frame_size = mtu + IAVF_ETH_OVERHEAD;
 979         int ret = 0;
 980
 981         if (mtu < RTE_ETHER_MIN_MTU || frame_size > IAVF_FRAME_SIZE_MAX)
 982                 return -EINVAL;
 983
 984         /* mtu setting is forbidden if port is start */
 985         if (dev->data->dev_started) {
 986                 PMD_DRV_LOG(ERR, "port must be stopped before configuration");
 987                 return -EBUSY;
 988         }
 989
 990         if (frame_size > RTE_ETHER_MAX_LEN)
 991                 dev->data->dev_conf.rxmode.offloads |=
 992                                 DEV_RX_OFFLOAD_JUMBO_FRAME;
 993         else
 994                 dev->data->dev_conf.rxmode.offloads &=
 995                                 ~DEV_RX_OFFLOAD_JUMBO_FRAME;
 996
 997         dev->data->dev_conf.rxmode.max_rx_pkt_len = frame_size;
 998
 999         return ret;
1000 }
1001
1002 static int
1003 iavf_dev_set_default_mac_addr(struct rte_eth_dev *dev,
1004                              struct rte_ether_addr *mac_addr)
1005 {
1006         struct iavf_adapter *adapter =
1007                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1008         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(adapter);
1009         struct rte_ether_addr *perm_addr, *old_addr;
1010         int ret;
1011
1012         old_addr = (struct rte_ether_addr *)hw->mac.addr;
1013         perm_addr = (struct rte_ether_addr *)hw->mac.perm_addr;
1014
1015         if (rte_is_same_ether_addr(mac_addr, old_addr))
1016                 return 0;
1017
1018         /* If the MAC address is configured by host, skip the setting */
1019         if (rte_is_valid_assigned_ether_addr(perm_addr))
1020                 return -EPERM;
1021
1022         ret = iavf_add_del_eth_addr(adapter, old_addr, false);
1023         if (ret)
1024                 PMD_DRV_LOG(ERR, "Fail to delete old MAC:"
1025                             " %02X:%02X:%02X:%02X:%02X:%02X",
1026                             old_addr->addr_bytes[0],
1027                             old_addr->addr_bytes[1],
1028                             old_addr->addr_bytes[2],
1029                             old_addr->addr_bytes[3],
1030                             old_addr->addr_bytes[4],
1031                             old_addr->addr_bytes[5]);
1032
1033         ret = iavf_add_del_eth_addr(adapter, mac_addr, true);
1034         if (ret)
1035                 PMD_DRV_LOG(ERR, "Fail to add new MAC:"
1036                             " %02X:%02X:%02X:%02X:%02X:%02X",
1037                             mac_addr->addr_bytes[0],
1038                             mac_addr->addr_bytes[1],
1039                             mac_addr->addr_bytes[2],
1040                             mac_addr->addr_bytes[3],
1041                             mac_addr->addr_bytes[4],
1042                             mac_addr->addr_bytes[5]);
1043
1044         if (ret)
1045                 return -EIO;
1046
1047         rte_ether_addr_copy(mac_addr, (struct rte_ether_addr *)hw->mac.addr);
1048         return 0;
1049 }
1050
1051 static void
1052 iavf_stat_update_48(uint64_t *offset, uint64_t *stat)
1053 {
1054         if (*stat >= *offset)
1055                 *stat = *stat - *offset;
1056         else
1057                 *stat = (uint64_t)((*stat +
1058                         ((uint64_t)1 << IAVF_48_BIT_WIDTH)) - *offset);
1059
1060         *stat &= IAVF_48_BIT_MASK;
1061 }
1062
1063 static void
1064 iavf_stat_update_32(uint64_t *offset, uint64_t *stat)
1065 {
1066         if (*stat >= *offset)
1067                 *stat = (uint64_t)(*stat - *offset);
1068         else
1069                 *stat = (uint64_t)((*stat +
1070                         ((uint64_t)1 << IAVF_32_BIT_WIDTH)) - *offset);
1071 }
1072
1073 static void
1074 iavf_update_stats(struct iavf_vsi *vsi, struct virtchnl_eth_stats *nes)
1075 {
1076         struct virtchnl_eth_stats *oes = &vsi->eth_stats_offset;
1077
1078         iavf_stat_update_48(&oes->rx_bytes, &nes->rx_bytes);
1079         iavf_stat_update_48(&oes->rx_unicast, &nes->rx_unicast);
1080         iavf_stat_update_48(&oes->rx_multicast, &nes->rx_multicast);
1081         iavf_stat_update_48(&oes->rx_broadcast, &nes->rx_broadcast);
1082         iavf_stat_update_32(&oes->rx_discards, &nes->rx_discards);
1083         iavf_stat_update_48(&oes->tx_bytes, &nes->tx_bytes);
1084         iavf_stat_update_48(&oes->tx_unicast, &nes->tx_unicast);
1085         iavf_stat_update_48(&oes->tx_multicast, &nes->tx_multicast);
1086         iavf_stat_update_48(&oes->tx_broadcast, &nes->tx_broadcast);
1087         iavf_stat_update_32(&oes->tx_errors, &nes->tx_errors);
1088         iavf_stat_update_32(&oes->tx_discards, &nes->tx_discards);
1089 }
1090
1091 static int
1092 iavf_dev_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
1093 {
1094         struct iavf_adapter *adapter =
1095                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1096         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
1097         struct iavf_vsi *vsi = &vf->vsi;
1098         struct virtchnl_eth_stats *pstats = NULL;
1099         int ret;
1100
1101         ret = iavf_query_stats(adapter, &pstats);
1102         if (ret == 0) {
1103                 iavf_update_stats(vsi, pstats);
1104                 stats->ipackets = pstats->rx_unicast + pstats->rx_multicast +
1105                                 pstats->rx_broadcast - pstats->rx_discards;
1106                 stats->opackets = pstats->tx_broadcast + pstats->tx_multicast +
1107                                                 pstats->tx_unicast;
1108                 stats->imissed = pstats->rx_discards;
1109                 stats->oerrors = pstats->tx_errors + pstats->tx_discards;
1110                 stats->ibytes = pstats->rx_bytes;
1111                 stats->ibytes -= stats->ipackets * RTE_ETHER_CRC_LEN;
1112                 stats->obytes = pstats->tx_bytes;
1113         } else {
1114                 PMD_DRV_LOG(ERR, "Get statistics failed");
1115         }
1116         return ret;
1117 }
1118
1119 static int
1120 iavf_dev_stats_reset(struct rte_eth_dev *dev)
1121 {
1122         int ret;
1123         struct iavf_adapter *adapter =
1124                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1125         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
1126         struct iavf_vsi *vsi = &vf->vsi;
1127         struct virtchnl_eth_stats *pstats = NULL;
1128
1129         /* read stat values to clear hardware registers */
1130         ret = iavf_query_stats(adapter, &pstats);
1131         if (ret != 0)
1132                 return ret;
1133
1134         /* set stats offset base on current values */
1135         vsi->eth_stats_offset = *pstats;
1136
1137         return 0;
1138 }
1139
1140 static int
1141 iavf_dev_rx_queue_intr_enable(struct rte_eth_dev *dev, uint16_t queue_id)
1142 {
1143         struct iavf_adapter *adapter =
1144                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1145         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1146         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(adapter);
1147         uint16_t msix_intr;
1148
1149         msix_intr = pci_dev->intr_handle.intr_vec[queue_id];
1150         if (msix_intr == IAVF_MISC_VEC_ID) {
1151                 PMD_DRV_LOG(INFO, "MISC is also enabled for control");
1152                 IAVF_WRITE_REG(hw, IAVF_VFINT_DYN_CTL01,
1153                                IAVF_VFINT_DYN_CTL01_INTENA_MASK |
1154                                IAVF_VFINT_DYN_CTL01_CLEARPBA_MASK |
1155                                IAVF_VFINT_DYN_CTL01_ITR_INDX_MASK);
1156         } else {
1157                 IAVF_WRITE_REG(hw,
1158                                IAVF_VFINT_DYN_CTLN1
1159                                 (msix_intr - IAVF_RX_VEC_START),
1160                                IAVF_VFINT_DYN_CTLN1_INTENA_MASK |
1161                                IAVF_VFINT_DYN_CTL01_CLEARPBA_MASK |
1162                                IAVF_VFINT_DYN_CTLN1_ITR_INDX_MASK);
1163         }
1164
1165         IAVF_WRITE_FLUSH(hw);
1166
1167         rte_intr_ack(&pci_dev->intr_handle);
1168
1169         return 0;
1170 }
1171
1172 static int
1173 iavf_dev_rx_queue_intr_disable(struct rte_eth_dev *dev, uint16_t queue_id)
1174 {
1175         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1176         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1177         uint16_t msix_intr;
1178
1179         msix_intr = pci_dev->intr_handle.intr_vec[queue_id];
1180         if (msix_intr == IAVF_MISC_VEC_ID) {
1181                 PMD_DRV_LOG(ERR, "MISC is used for control, cannot disable it");
1182                 return -EIO;
1183         }
1184
1185         IAVF_WRITE_REG(hw,
1186                       IAVF_VFINT_DYN_CTLN1(msix_intr - IAVF_RX_VEC_START),
1187                       0);
1188
1189         IAVF_WRITE_FLUSH(hw);
1190         return 0;
1191 }
1192
1193 static int
1194 iavf_check_vf_reset_done(struct iavf_hw *hw)
1195 {
1196         int i, reset;
1197
1198         for (i = 0; i < IAVF_RESET_WAIT_CNT; i++) {
1199                 reset = IAVF_READ_REG(hw, IAVF_VFGEN_RSTAT) &
1200                         IAVF_VFGEN_RSTAT_VFR_STATE_MASK;
1201                 reset = reset >> IAVF_VFGEN_RSTAT_VFR_STATE_SHIFT;
1202                 if (reset == VIRTCHNL_VFR_VFACTIVE ||
1203                     reset == VIRTCHNL_VFR_COMPLETED)
1204                         break;
1205                 rte_delay_ms(20);
1206         }
1207
1208         if (i >= IAVF_RESET_WAIT_CNT)
1209                 return -1;
1210
1211         return 0;
1212 }
1213
1214 static int
1215 iavf_init_vf(struct rte_eth_dev *dev)
1216 {
1217         int err, bufsz;
1218         struct iavf_adapter *adapter =
1219                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1220         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1221         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
1222
1223         err = iavf_set_mac_type(hw);
1224         if (err) {
1225                 PMD_INIT_LOG(ERR, "set_mac_type failed: %d", err);
1226                 goto err;
1227         }
1228
1229         err = iavf_check_vf_reset_done(hw);
1230         if (err) {
1231                 PMD_INIT_LOG(ERR, "VF is still resetting");
1232                 goto err;
1233         }
1234
1235         iavf_init_adminq_parameter(hw);
1236         err = iavf_init_adminq(hw);
1237         if (err) {
1238                 PMD_INIT_LOG(ERR, "init_adminq failed: %d", err);
1239                 goto err;
1240         }
1241
1242         vf->aq_resp = rte_zmalloc("vf_aq_resp", IAVF_AQ_BUF_SZ, 0);
1243         if (!vf->aq_resp) {
1244                 PMD_INIT_LOG(ERR, "unable to allocate vf_aq_resp memory");
1245                 goto err_aq;
1246         }
1247         if (iavf_check_api_version(adapter) != 0) {
1248                 PMD_INIT_LOG(ERR, "check_api version failed");
1249                 goto err_api;
1250         }
1251
1252         bufsz = sizeof(struct virtchnl_vf_resource) +
1253                 (IAVF_MAX_VF_VSI * sizeof(struct virtchnl_vsi_resource));
1254         vf->vf_res = rte_zmalloc("vf_res", bufsz, 0);
1255         if (!vf->vf_res) {
1256                 PMD_INIT_LOG(ERR, "unable to allocate vf_res memory");
1257                 goto err_api;
1258         }
1259         if (iavf_get_vf_resource(adapter) != 0) {
1260                 PMD_INIT_LOG(ERR, "iavf_get_vf_config failed");
1261                 goto err_alloc;
1262         }
1263         /* Allocate memort for RSS info */
1264         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF) {
1265                 vf->rss_key = rte_zmalloc("rss_key",
1266                                           vf->vf_res->rss_key_size, 0);
1267                 if (!vf->rss_key) {
1268                         PMD_INIT_LOG(ERR, "unable to allocate rss_key memory");
1269                         goto err_rss;
1270                 }
1271                 vf->rss_lut = rte_zmalloc("rss_lut",
1272                                           vf->vf_res->rss_lut_size, 0);
1273                 if (!vf->rss_lut) {
1274                         PMD_INIT_LOG(ERR, "unable to allocate rss_lut memory");
1275                         goto err_rss;
1276                 }
1277         }
1278
1279         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RX_FLEX_DESC) {
1280                 if (iavf_get_supported_rxdid(adapter) != 0) {
1281                         PMD_INIT_LOG(ERR, "failed to do get supported rxdid");
1282                         goto err_rss;
1283                 }
1284         }
1285
1286         return 0;
1287 err_rss:
1288         rte_free(vf->rss_key);
1289         rte_free(vf->rss_lut);
1290 err_alloc:
1291         rte_free(vf->vf_res);
1292         vf->vsi_res = NULL;
1293 err_api:
1294         rte_free(vf->aq_resp);
1295 err_aq:
1296         iavf_shutdown_adminq(hw);
1297 err:
1298         return -1;
1299 }
1300
1301 /* Enable default admin queue interrupt setting */
1302 static inline void
1303 iavf_enable_irq0(struct iavf_hw *hw)
1304 {
1305         /* Enable admin queue interrupt trigger */
1306         IAVF_WRITE_REG(hw, IAVF_VFINT_ICR0_ENA1,
1307                        IAVF_VFINT_ICR0_ENA1_ADMINQ_MASK);
1308
1309         IAVF_WRITE_REG(hw, IAVF_VFINT_DYN_CTL01,
1310                        IAVF_VFINT_DYN_CTL01_INTENA_MASK |
1311                        IAVF_VFINT_DYN_CTL01_CLEARPBA_MASK |
1312                        IAVF_VFINT_DYN_CTL01_ITR_INDX_MASK);
1313
1314         IAVF_WRITE_FLUSH(hw);
1315 }
1316
1317 static inline void
1318 iavf_disable_irq0(struct iavf_hw *hw)
1319 {
1320         /* Disable all interrupt types */
1321         IAVF_WRITE_REG(hw, IAVF_VFINT_ICR0_ENA1, 0);
1322         IAVF_WRITE_REG(hw, IAVF_VFINT_DYN_CTL01,
1323                        IAVF_VFINT_DYN_CTL01_ITR_INDX_MASK);
1324         IAVF_WRITE_FLUSH(hw);
1325 }
1326
1327 static void
1328 iavf_dev_interrupt_handler(void *param)
1329 {
1330         struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
1331         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1332
1333         iavf_disable_irq0(hw);
1334
1335         iavf_handle_virtchnl_msg(dev);
1336
1337         iavf_enable_irq0(hw);
1338 }
1339
1340 static int
1341 iavf_dev_filter_ctrl(struct rte_eth_dev *dev,
1342                      enum rte_filter_type filter_type,
1343                      enum rte_filter_op filter_op,
1344                      void *arg)
1345 {
1346         int ret = 0;
1347
1348         if (!dev)
1349                 return -EINVAL;
1350
1351         switch (filter_type) {
1352         case RTE_ETH_FILTER_GENERIC:
1353                 if (filter_op != RTE_ETH_FILTER_GET)
1354                         return -EINVAL;
1355                 *(const void **)arg = &iavf_flow_ops;
1356                 break;
1357         default:
1358                 PMD_DRV_LOG(WARNING, "Filter type (%d) not supported",
1359                             filter_type);
1360                 ret = -EINVAL;
1361                 break;
1362         }
1363
1364         return ret;
1365 }
1366
1367
1368 static int
1369 iavf_dev_init(struct rte_eth_dev *eth_dev)
1370 {
1371         struct iavf_adapter *adapter =
1372                 IAVF_DEV_PRIVATE_TO_ADAPTER(eth_dev->data->dev_private);
1373         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(adapter);
1374         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
1375         int ret = 0;
1376
1377         PMD_INIT_FUNC_TRACE();
1378
1379         /* assign ops func pointer */
1380         eth_dev->dev_ops = &iavf_eth_dev_ops;
1381         eth_dev->rx_pkt_burst = &iavf_recv_pkts;
1382         eth_dev->tx_pkt_burst = &iavf_xmit_pkts;
1383         eth_dev->tx_pkt_prepare = &iavf_prep_pkts;
1384
1385         /* For secondary processes, we don't initialise any further as primary
1386          * has already done this work. Only check if we need a different RX
1387          * and TX function.
1388          */
1389         if (rte_eal_process_type() != RTE_PROC_PRIMARY) {
1390                 iavf_set_rx_function(eth_dev);
1391                 iavf_set_tx_function(eth_dev);
1392                 return 0;
1393         }
1394         rte_eth_copy_pci_info(eth_dev, pci_dev);
1395
1396         hw->vendor_id = pci_dev->id.vendor_id;
1397         hw->device_id = pci_dev->id.device_id;
1398         hw->subsystem_vendor_id = pci_dev->id.subsystem_vendor_id;
1399         hw->subsystem_device_id = pci_dev->id.subsystem_device_id;
1400         hw->bus.bus_id = pci_dev->addr.bus;
1401         hw->bus.device = pci_dev->addr.devid;
1402         hw->bus.func = pci_dev->addr.function;
1403         hw->hw_addr = (void *)pci_dev->mem_resource[0].addr;
1404         hw->back = IAVF_DEV_PRIVATE_TO_ADAPTER(eth_dev->data->dev_private);
1405         adapter->eth_dev = eth_dev;
1406         adapter->stopped = 1;
1407
1408         /* Pass the information to the rte_eth_dev_close() that it should also
1409          * release the private port resources.
1410          */
1411         eth_dev->data->dev_flags |= RTE_ETH_DEV_CLOSE_REMOVE;
1412
1413         if (iavf_init_vf(eth_dev) != 0) {
1414                 PMD_INIT_LOG(ERR, "Init vf failed");
1415                 return -1;
1416         }
1417
1418         /* set default ptype table */
1419         adapter->ptype_tbl = iavf_get_default_ptype_table();
1420
1421         /* copy mac addr */
1422         eth_dev->data->mac_addrs = rte_zmalloc(
1423                 "iavf_mac", RTE_ETHER_ADDR_LEN * IAVF_NUM_MACADDR_MAX, 0);
1424         if (!eth_dev->data->mac_addrs) {
1425                 PMD_INIT_LOG(ERR, "Failed to allocate %d bytes needed to"
1426                              " store MAC addresses",
1427                              RTE_ETHER_ADDR_LEN * IAVF_NUM_MACADDR_MAX);
1428                 return -ENOMEM;
1429         }
1430         /* If the MAC address is not configured by host,
1431          * generate a random one.
1432          */
1433         if (!rte_is_valid_assigned_ether_addr(
1434                         (struct rte_ether_addr *)hw->mac.addr))
1435                 rte_eth_random_addr(hw->mac.addr);
1436         rte_ether_addr_copy((struct rte_ether_addr *)hw->mac.addr,
1437                         &eth_dev->data->mac_addrs[0]);
1438
1439         /* register callback func to eal lib */
1440         rte_intr_callback_register(&pci_dev->intr_handle,
1441                                    iavf_dev_interrupt_handler,
1442                                    (void *)eth_dev);
1443
1444         /* enable uio intr after callback register */
1445         rte_intr_enable(&pci_dev->intr_handle);
1446
1447         /* configure and enable device interrupt */
1448         iavf_enable_irq0(hw);
1449
1450         ret = iavf_flow_init(adapter);
1451         if (ret) {
1452                 PMD_INIT_LOG(ERR, "Failed to initialize flow");
1453                 return ret;
1454         }
1455
1456         return 0;
1457 }
1458
1459 static void
1460 iavf_dev_close(struct rte_eth_dev *dev)
1461 {
1462         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1463         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1464         struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
1465         struct iavf_adapter *adapter =
1466                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1467         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
1468
1469         iavf_dev_stop(dev);
1470         iavf_flow_flush(dev, NULL);
1471         iavf_flow_uninit(adapter);
1472         iavf_shutdown_adminq(hw);
1473         /* disable uio intr before callback unregister */
1474         rte_intr_disable(intr_handle);
1475
1476         /* unregister callback func from eal lib */
1477         rte_intr_callback_unregister(intr_handle,
1478                                      iavf_dev_interrupt_handler, dev);
1479         iavf_disable_irq0(hw);
1480
1481         dev->dev_ops = NULL;
1482         dev->rx_pkt_burst = NULL;
1483         dev->tx_pkt_burst = NULL;
1484
1485         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF) {
1486                 if (vf->rss_lut) {
1487                         rte_free(vf->rss_lut);
1488                         vf->rss_lut = NULL;
1489                 }
1490                 if (vf->rss_key) {
1491                         rte_free(vf->rss_key);
1492                         vf->rss_key = NULL;
1493                 }
1494         }
1495
1496         rte_free(vf->vf_res);
1497         vf->vsi_res = NULL;
1498         vf->vf_res = NULL;
1499
1500         rte_free(vf->aq_resp);
1501         vf->aq_resp = NULL;
1502 }
1503
1504 static int
1505 iavf_dev_uninit(struct rte_eth_dev *dev)
1506 {
1507         if (rte_eal_process_type() != RTE_PROC_PRIMARY)
1508                 return -EPERM;
1509
1510         iavf_dev_close(dev);
1511
1512         return 0;
1513 }
1514
1515 /*
1516  * Reset VF device only to re-initialize resources in PMD layer
1517  */
1518 static int
1519 iavf_dev_reset(struct rte_eth_dev *dev)
1520 {
1521         int ret;
1522
1523         ret = iavf_dev_uninit(dev);
1524         if (ret)
1525                 return ret;
1526
1527         return iavf_dev_init(dev);
1528 }
1529
1530 static int
1531 iavf_dcf_cap_check_handler(__rte_unused const char *key,
1532                            const char *value, __rte_unused void *opaque)
1533 {
1534         if (strcmp(value, "dcf"))
1535                 return -1;
1536
1537         return 0;
1538 }
1539
1540 static int
1541 iavf_dcf_cap_selected(struct rte_devargs *devargs)
1542 {
1543         struct rte_kvargs *kvlist;
1544         const char *key = "cap";
1545         int ret = 0;
1546
1547         if (devargs == NULL)
1548                 return 0;
1549
1550         kvlist = rte_kvargs_parse(devargs->args, NULL);
1551         if (kvlist == NULL)
1552                 return 0;
1553
1554         if (!rte_kvargs_count(kvlist, key))
1555                 goto exit;
1556
1557         /* dcf capability selected when there's a key-value pair: cap=dcf */
1558         if (rte_kvargs_process(kvlist, key,
1559                                iavf_dcf_cap_check_handler, NULL) < 0)
1560                 goto exit;
1561
1562         ret = 1;
1563
1564 exit:
1565         rte_kvargs_free(kvlist);
1566         return ret;
1567 }
1568
1569 static int eth_iavf_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
1570                              struct rte_pci_device *pci_dev)
1571 {
1572         if (iavf_dcf_cap_selected(pci_dev->device.devargs))
1573                 return 1;
1574
1575         return rte_eth_dev_pci_generic_probe(pci_dev,
1576                 sizeof(struct iavf_adapter), iavf_dev_init);
1577 }
1578
1579 static int eth_iavf_pci_remove(struct rte_pci_device *pci_dev)
1580 {
1581         return rte_eth_dev_pci_generic_remove(pci_dev, iavf_dev_uninit);
1582 }
1583
1584 /* Adaptive virtual function driver struct */
1585 static struct rte_pci_driver rte_iavf_pmd = {
1586         .id_table = pci_id_iavf_map,
1587         .drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC,
1588         .probe = eth_iavf_pci_probe,
1589         .remove = eth_iavf_pci_remove,
1590 };
1591
1592 RTE_PMD_REGISTER_PCI(net_iavf, rte_iavf_pmd);
1593 RTE_PMD_REGISTER_PCI_TABLE(net_iavf, pci_id_iavf_map);
1594 RTE_PMD_REGISTER_KMOD_DEP(net_iavf, "* igb_uio | vfio-pci");
1595 RTE_PMD_REGISTER_PARAM_STRING(net_iavf, "cap=dcf");
1596 RTE_LOG_REGISTER(iavf_logtype_init, pmd.net.iavf.init, NOTICE);
1597 RTE_LOG_REGISTER(iavf_logtype_driver, pmd.net.iavf.driver, NOTICE);
1598 #ifdef RTE_LIBRTE_IAVF_DEBUG_RX
1599 RTE_LOG_REGISTER(iavf_logtype_rx, pmd.net.iavf.rx, DEBUG);
1600 #endif
1601 #ifdef RTE_LIBRTE_IAVF_DEBUG_TX
1602 RTE_LOG_REGISTER(iavf_logtype_tx, pmd.net.iavf.tx, DEBUG);
1603 #endif
1604 #ifdef RTE_LIBRTE_IAVF_DEBUG_TX_FREE
1605 RTE_LOG_REGISTER(iavf_logtype_tx_free, pmd.net.iavf.tx_free, DEBUG);
1606 #endif