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