drivers/net/iavf/iavf_ethdev.c

   1 /* SPDX-License-Identifier: BSD-3-Clause
   2  * Copyright(c) 2017 Intel Corporation
   3  */
   4
   5 #include <sys/queue.h>
   6 #include <stdio.h>
   7 #include <errno.h>
   8 #include <stdint.h>
   9 #include <string.h>
  10 #include <unistd.h>
  11 #include <stdarg.h>
  12 #include <inttypes.h>
  13 #include <rte_byteorder.h>
  14 #include <rte_common.h>
  15
  16 #include <rte_interrupts.h>
  17 #include <rte_debug.h>
  18 #include <rte_pci.h>
  19 #include <rte_atomic.h>
  20 #include <rte_eal.h>
  21 #include <rte_ether.h>
  22 #include <rte_ethdev_driver.h>
  23 #include <rte_ethdev_pci.h>
  24 #include <rte_malloc.h>
  25 #include <rte_memzone.h>
  26 #include <rte_dev.h>
  27
  28 #include "iavf.h"
  29 #include "iavf_rxtx.h"
  30 #include "iavf_generic_flow.h"
  31
  32 static int iavf_dev_configure(struct rte_eth_dev *dev);
  33 static int iavf_dev_start(struct rte_eth_dev *dev);
  34 static void iavf_dev_stop(struct rte_eth_dev *dev);
  35 static void iavf_dev_close(struct rte_eth_dev *dev);
  36 static int iavf_dev_reset(struct rte_eth_dev *dev);
  37 static int iavf_dev_info_get(struct rte_eth_dev *dev,
  38                              struct rte_eth_dev_info *dev_info);
  39 static const uint32_t *iavf_dev_supported_ptypes_get(struct rte_eth_dev *dev);
  40 static int iavf_dev_stats_get(struct rte_eth_dev *dev,
  41                              struct rte_eth_stats *stats);
  42 static int iavf_dev_stats_reset(struct rte_eth_dev *dev);
  43 static int iavf_dev_promiscuous_enable(struct rte_eth_dev *dev);
  44 static int iavf_dev_promiscuous_disable(struct rte_eth_dev *dev);
  45 static int iavf_dev_allmulticast_enable(struct rte_eth_dev *dev);
  46 static int iavf_dev_allmulticast_disable(struct rte_eth_dev *dev);
  47 static int iavf_dev_add_mac_addr(struct rte_eth_dev *dev,
  48                                 struct rte_ether_addr *addr,
  49                                 uint32_t index,
  50                                 uint32_t pool);
  51 static void iavf_dev_del_mac_addr(struct rte_eth_dev *dev, uint32_t index);
  52 static int iavf_dev_vlan_filter_set(struct rte_eth_dev *dev,
  53                                    uint16_t vlan_id, int on);
  54 static int iavf_dev_vlan_offload_set(struct rte_eth_dev *dev, int mask);
  55 static int iavf_dev_rss_reta_update(struct rte_eth_dev *dev,
  56                                    struct rte_eth_rss_reta_entry64 *reta_conf,
  57                                    uint16_t reta_size);
  58 static int iavf_dev_rss_reta_query(struct rte_eth_dev *dev,
  59                                   struct rte_eth_rss_reta_entry64 *reta_conf,
  60                                   uint16_t reta_size);
  61 static int iavf_dev_rss_hash_update(struct rte_eth_dev *dev,
  62                                    struct rte_eth_rss_conf *rss_conf);
  63 static int iavf_dev_rss_hash_conf_get(struct rte_eth_dev *dev,
  64                                      struct rte_eth_rss_conf *rss_conf);
  65 static int iavf_dev_mtu_set(struct rte_eth_dev *dev, uint16_t mtu);
  66 static int iavf_dev_set_default_mac_addr(struct rte_eth_dev *dev,
  67                                          struct rte_ether_addr *mac_addr);
  68 static int iavf_dev_rx_queue_intr_enable(struct rte_eth_dev *dev,
  69                                         uint16_t queue_id);
  70 static int iavf_dev_rx_queue_intr_disable(struct rte_eth_dev *dev,
  71                                          uint16_t queue_id);
  72 static int iavf_dev_filter_ctrl(struct rte_eth_dev *dev,
  73                      enum rte_filter_type filter_type,
  74                      enum rte_filter_op filter_op,
  75                      void *arg);
  76
  77
  78 int iavf_logtype_init;
  79 int iavf_logtype_driver;
  80
  81 #ifdef RTE_LIBRTE_IAVF_DEBUG_RX
  82 int iavf_logtype_rx;
  83 #endif
  84 #ifdef RTE_LIBRTE_IAVF_DEBUG_TX
  85 int iavf_logtype_tx;
  86 #endif
  87 #ifdef RTE_LIBRTE_IAVF_DEBUG_TX_FREE
  88 int iavf_logtype_tx_free;
  89 #endif
  90
  91 static const struct rte_pci_id pci_id_iavf_map[] = {
  92         { RTE_PCI_DEVICE(IAVF_INTEL_VENDOR_ID, IAVF_DEV_ID_ADAPTIVE_VF) },
  93         { .vendor_id = 0, /* sentinel */ },
  94 };
  95
  96 static const struct eth_dev_ops iavf_eth_dev_ops = {
  97         .dev_configure              = iavf_dev_configure,
  98         .dev_start                  = iavf_dev_start,
  99         .dev_stop                   = iavf_dev_stop,
 100         .dev_close                  = iavf_dev_close,
 101         .dev_reset                  = iavf_dev_reset,
 102         .dev_infos_get              = iavf_dev_info_get,
 103         .dev_supported_ptypes_get   = iavf_dev_supported_ptypes_get,
 104         .link_update                = iavf_dev_link_update,
 105         .stats_get                  = iavf_dev_stats_get,
 106         .stats_reset                = iavf_dev_stats_reset,
 107         .promiscuous_enable         = iavf_dev_promiscuous_enable,
 108         .promiscuous_disable        = iavf_dev_promiscuous_disable,
 109         .allmulticast_enable        = iavf_dev_allmulticast_enable,
 110         .allmulticast_disable       = iavf_dev_allmulticast_disable,
 111         .mac_addr_add               = iavf_dev_add_mac_addr,
 112         .mac_addr_remove            = iavf_dev_del_mac_addr,
 113         .vlan_filter_set            = iavf_dev_vlan_filter_set,
 114         .vlan_offload_set           = iavf_dev_vlan_offload_set,
 115         .rx_queue_start             = iavf_dev_rx_queue_start,
 116         .rx_queue_stop              = iavf_dev_rx_queue_stop,
 117         .tx_queue_start             = iavf_dev_tx_queue_start,
 118         .tx_queue_stop              = iavf_dev_tx_queue_stop,
 119         .rx_queue_setup             = iavf_dev_rx_queue_setup,
 120         .rx_queue_release           = iavf_dev_rx_queue_release,
 121         .tx_queue_setup             = iavf_dev_tx_queue_setup,
 122         .tx_queue_release           = iavf_dev_tx_queue_release,
 123         .mac_addr_set               = iavf_dev_set_default_mac_addr,
 124         .reta_update                = iavf_dev_rss_reta_update,
 125         .reta_query                 = iavf_dev_rss_reta_query,
 126         .rss_hash_update            = iavf_dev_rss_hash_update,
 127         .rss_hash_conf_get          = iavf_dev_rss_hash_conf_get,
 128         .rxq_info_get               = iavf_dev_rxq_info_get,
 129         .txq_info_get               = iavf_dev_txq_info_get,
 130         .rx_queue_count             = iavf_dev_rxq_count,
 131         .rx_descriptor_status       = iavf_dev_rx_desc_status,
 132         .tx_descriptor_status       = iavf_dev_tx_desc_status,
 133         .mtu_set                    = iavf_dev_mtu_set,
 134         .rx_queue_intr_enable       = iavf_dev_rx_queue_intr_enable,
 135         .rx_queue_intr_disable      = iavf_dev_rx_queue_intr_disable,
 136         .filter_ctrl                = iavf_dev_filter_ctrl,
 137 };
 138
 139 static int
 140 iavf_dev_configure(struct rte_eth_dev *dev)
 141 {
 142         struct iavf_adapter *ad =
 143                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 144         struct iavf_info *vf =  IAVF_DEV_PRIVATE_TO_VF(ad);
 145         struct rte_eth_conf *dev_conf = &dev->data->dev_conf;
 146
 147         ad->rx_bulk_alloc_allowed = true;
 148         /* Initialize to TRUE. If any of Rx queues doesn't meet the
 149          * vector Rx/Tx preconditions, it will be reset.
 150          */
 151         ad->rx_vec_allowed = true;
 152         ad->tx_vec_allowed = true;
 153
 154         if (dev->data->dev_conf.rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG)
 155                 dev->data->dev_conf.rxmode.offloads |= DEV_RX_OFFLOAD_RSS_HASH;
 156
 157         /* Vlan stripping setting */
 158         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_VLAN) {
 159                 if (dev_conf->rxmode.offloads & DEV_RX_OFFLOAD_VLAN_STRIP)
 160                         iavf_enable_vlan_strip(ad);
 161                 else
 162                         iavf_disable_vlan_strip(ad);
 163         }
 164         return 0;
 165 }
 166
 167 static int
 168 iavf_init_rss(struct iavf_adapter *adapter)
 169 {
 170         struct iavf_info *vf =  IAVF_DEV_PRIVATE_TO_VF(adapter);
 171         struct rte_eth_rss_conf *rss_conf;
 172         uint8_t i, j, nb_q;
 173         int ret;
 174
 175         rss_conf = &adapter->eth_dev->data->dev_conf.rx_adv_conf.rss_conf;
 176         nb_q = RTE_MIN(adapter->eth_dev->data->nb_rx_queues,
 177                        IAVF_MAX_NUM_QUEUES);
 178
 179         if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF)) {
 180                 PMD_DRV_LOG(DEBUG, "RSS is not supported");
 181                 return -ENOTSUP;
 182         }
 183         if (adapter->eth_dev->data->dev_conf.rxmode.mq_mode != ETH_MQ_RX_RSS) {
 184                 PMD_DRV_LOG(WARNING, "RSS is enabled by PF by default");
 185                 /* set all lut items to default queue */
 186                 for (i = 0; i < vf->vf_res->rss_lut_size; i++)
 187                         vf->rss_lut[i] = 0;
 188                 ret = iavf_configure_rss_lut(adapter);
 189                 return ret;
 190         }
 191
 192         /* In IAVF, RSS enablement is set by PF driver. It is not supported
 193          * to set based on rss_conf->rss_hf.
 194          */
 195
 196         /* configure RSS key */
 197         if (!rss_conf->rss_key) {
 198                 /* Calculate the default hash key */
 199                 for (i = 0; i <= vf->vf_res->rss_key_size; i++)
 200                         vf->rss_key[i] = (uint8_t)rte_rand();
 201         } else
 202                 rte_memcpy(vf->rss_key, rss_conf->rss_key,
 203                            RTE_MIN(rss_conf->rss_key_len,
 204                                    vf->vf_res->rss_key_size));
 205
 206         /* init RSS LUT table */
 207         for (i = 0, j = 0; i < vf->vf_res->rss_lut_size; i++, j++) {
 208                 if (j >= nb_q)
 209                         j = 0;
 210                 vf->rss_lut[i] = j;
 211         }
 212         /* send virtchnnl ops to configure rss*/
 213         ret = iavf_configure_rss_lut(adapter);
 214         if (ret)
 215                 return ret;
 216         ret = iavf_configure_rss_key(adapter);
 217         if (ret)
 218                 return ret;
 219
 220         return 0;
 221 }
 222
 223 static int
 224 iavf_init_rxq(struct rte_eth_dev *dev, struct iavf_rx_queue *rxq)
 225 {
 226         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(dev->data->dev_private);
 227         struct rte_eth_dev_data *dev_data = dev->data;
 228         uint16_t buf_size, max_pkt_len, len;
 229
 230         buf_size = rte_pktmbuf_data_room_size(rxq->mp) - RTE_PKTMBUF_HEADROOM;
 231
 232         /* Calculate the maximum packet length allowed */
 233         len = rxq->rx_buf_len * IAVF_MAX_CHAINED_RX_BUFFERS;
 234         max_pkt_len = RTE_MIN(len, dev->data->dev_conf.rxmode.max_rx_pkt_len);
 235
 236         /* Check if the jumbo frame and maximum packet length are set
 237          * correctly.
 238          */
 239         if (dev->data->dev_conf.rxmode.offloads & DEV_RX_OFFLOAD_JUMBO_FRAME) {
 240                 if (max_pkt_len <= RTE_ETHER_MAX_LEN ||
 241                     max_pkt_len > IAVF_FRAME_SIZE_MAX) {
 242                         PMD_DRV_LOG(ERR, "maximum packet length must be "
 243                                     "larger than %u and smaller than %u, "
 244                                     "as jumbo frame is enabled",
 245                                     (uint32_t)RTE_ETHER_MAX_LEN,
 246                                     (uint32_t)IAVF_FRAME_SIZE_MAX);
 247                         return -EINVAL;
 248                 }
 249         } else {
 250                 if (max_pkt_len < RTE_ETHER_MIN_LEN ||
 251                     max_pkt_len > RTE_ETHER_MAX_LEN) {
 252                         PMD_DRV_LOG(ERR, "maximum packet length must be "
 253                                     "larger than %u and smaller than %u, "
 254                                     "as jumbo frame is disabled",
 255                                     (uint32_t)RTE_ETHER_MIN_LEN,
 256                                     (uint32_t)RTE_ETHER_MAX_LEN);
 257                         return -EINVAL;
 258                 }
 259         }
 260
 261         rxq->max_pkt_len = max_pkt_len;
 262         if ((dev_data->dev_conf.rxmode.offloads & DEV_RX_OFFLOAD_SCATTER) ||
 263             (rxq->max_pkt_len + 2 * IAVF_VLAN_TAG_SIZE) > buf_size) {
 264                 dev_data->scattered_rx = 1;
 265         }
 266         IAVF_PCI_REG_WRITE(rxq->qrx_tail, rxq->nb_rx_desc - 1);
 267         IAVF_WRITE_FLUSH(hw);
 268
 269         return 0;
 270 }
 271
 272 static int
 273 iavf_init_queues(struct rte_eth_dev *dev)
 274 {
 275         struct iavf_rx_queue **rxq =
 276                 (struct iavf_rx_queue **)dev->data->rx_queues;
 277         int i, ret = IAVF_SUCCESS;
 278
 279         for (i = 0; i < dev->data->nb_rx_queues; i++) {
 280                 if (!rxq[i] || !rxq[i]->q_set)
 281                         continue;
 282                 ret = iavf_init_rxq(dev, rxq[i]);
 283                 if (ret != IAVF_SUCCESS)
 284                         break;
 285         }
 286         /* set rx/tx function to vector/scatter/single-segment
 287          * according to parameters
 288          */
 289         iavf_set_rx_function(dev);
 290         iavf_set_tx_function(dev);
 291
 292         return ret;
 293 }
 294
 295 static int iavf_config_rx_queues_irqs(struct rte_eth_dev *dev,
 296                                      struct rte_intr_handle *intr_handle)
 297 {
 298         struct iavf_adapter *adapter =
 299                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 300         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
 301         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(adapter);
 302         uint16_t interval, i;
 303         int vec;
 304
 305         if (rte_intr_cap_multiple(intr_handle) &&
 306             dev->data->dev_conf.intr_conf.rxq) {
 307                 if (rte_intr_efd_enable(intr_handle, dev->data->nb_rx_queues))
 308                         return -1;
 309         }
 310
 311         if (rte_intr_dp_is_en(intr_handle) && !intr_handle->intr_vec) {
 312                 intr_handle->intr_vec =
 313                         rte_zmalloc("intr_vec",
 314                                     dev->data->nb_rx_queues * sizeof(int), 0);
 315                 if (!intr_handle->intr_vec) {
 316                         PMD_DRV_LOG(ERR, "Failed to allocate %d rx intr_vec",
 317                                     dev->data->nb_rx_queues);
 318                         return -1;
 319                 }
 320         }
 321
 322         if (!dev->data->dev_conf.intr_conf.rxq ||
 323             !rte_intr_dp_is_en(intr_handle)) {
 324                 /* Rx interrupt disabled, Map interrupt only for writeback */
 325                 vf->nb_msix = 1;
 326                 if (vf->vf_res->vf_cap_flags &
 327                     VIRTCHNL_VF_OFFLOAD_WB_ON_ITR) {
 328                         /* If WB_ON_ITR supports, enable it */
 329                         vf->msix_base = IAVF_RX_VEC_START;
 330                         IAVF_WRITE_REG(hw,
 331                                        IAVF_VFINT_DYN_CTLN1(vf->msix_base - 1),
 332                                        IAVF_VFINT_DYN_CTLN1_ITR_INDX_MASK |
 333                                        IAVF_VFINT_DYN_CTLN1_WB_ON_ITR_MASK);
 334                 } else {
 335                         /* If no WB_ON_ITR offload flags, need to set
 336                          * interrupt for descriptor write back.
 337                          */
 338                         vf->msix_base = IAVF_MISC_VEC_ID;
 339
 340                         /* set ITR to max */
 341                         interval = iavf_calc_itr_interval(
 342                                         IAVF_QUEUE_ITR_INTERVAL_MAX);
 343                         IAVF_WRITE_REG(hw, IAVF_VFINT_DYN_CTL01,
 344                                        IAVF_VFINT_DYN_CTL01_INTENA_MASK |
 345                                        (IAVF_ITR_INDEX_DEFAULT <<
 346                                         IAVF_VFINT_DYN_CTL01_ITR_INDX_SHIFT) |
 347                                        (interval <<
 348                                         IAVF_VFINT_DYN_CTL01_INTERVAL_SHIFT));
 349                 }
 350                 IAVF_WRITE_FLUSH(hw);
 351                 /* map all queues to the same interrupt */
 352                 for (i = 0; i < dev->data->nb_rx_queues; i++)
 353                         vf->rxq_map[vf->msix_base] |= 1 << i;
 354         } else {
 355                 if (!rte_intr_allow_others(intr_handle)) {
 356                         vf->nb_msix = 1;
 357                         vf->msix_base = IAVF_MISC_VEC_ID;
 358                         for (i = 0; i < dev->data->nb_rx_queues; i++) {
 359                                 vf->rxq_map[vf->msix_base] |= 1 << i;
 360                                 intr_handle->intr_vec[i] = IAVF_MISC_VEC_ID;
 361                         }
 362                         PMD_DRV_LOG(DEBUG,
 363                                     "vector %u are mapping to all Rx queues",
 364                                     vf->msix_base);
 365                 } else {
 366                         /* If Rx interrupt is reuquired, and we can use
 367                          * multi interrupts, then the vec is from 1
 368                          */
 369                         vf->nb_msix = RTE_MIN(vf->vf_res->max_vectors,
 370                                               intr_handle->nb_efd);
 371                         vf->msix_base = IAVF_RX_VEC_START;
 372                         vec = IAVF_RX_VEC_START;
 373                         for (i = 0; i < dev->data->nb_rx_queues; i++) {
 374                                 vf->rxq_map[vec] |= 1 << i;
 375                                 intr_handle->intr_vec[i] = vec++;
 376                                 if (vec >= vf->nb_msix)
 377                                         vec = IAVF_RX_VEC_START;
 378                         }
 379                         PMD_DRV_LOG(DEBUG,
 380                                     "%u vectors are mapping to %u Rx queues",
 381                                     vf->nb_msix, dev->data->nb_rx_queues);
 382                 }
 383         }
 384
 385         if (iavf_config_irq_map(adapter)) {
 386                 PMD_DRV_LOG(ERR, "config interrupt mapping failed");
 387                 return -1;
 388         }
 389         return 0;
 390 }
 391
 392 static int
 393 iavf_start_queues(struct rte_eth_dev *dev)
 394 {
 395         struct iavf_rx_queue *rxq;
 396         struct iavf_tx_queue *txq;
 397         int i;
 398
 399         for (i = 0; i < dev->data->nb_tx_queues; i++) {
 400                 txq = dev->data->tx_queues[i];
 401                 if (txq->tx_deferred_start)
 402                         continue;
 403                 if (iavf_dev_tx_queue_start(dev, i) != 0) {
 404                         PMD_DRV_LOG(ERR, "Fail to start queue %u", i);
 405                         return -1;
 406                 }
 407         }
 408
 409         for (i = 0; i < dev->data->nb_rx_queues; i++) {
 410                 rxq = dev->data->rx_queues[i];
 411                 if (rxq->rx_deferred_start)
 412                         continue;
 413                 if (iavf_dev_rx_queue_start(dev, i) != 0) {
 414                         PMD_DRV_LOG(ERR, "Fail to start queue %u", i);
 415                         return -1;
 416                 }
 417         }
 418
 419         return 0;
 420 }
 421
 422 static int
 423 iavf_dev_start(struct rte_eth_dev *dev)
 424 {
 425         struct iavf_adapter *adapter =
 426                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 427         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
 428         struct rte_intr_handle *intr_handle = dev->intr_handle;
 429
 430         PMD_INIT_FUNC_TRACE();
 431
 432         adapter->stopped = 0;
 433
 434         vf->max_pkt_len = dev->data->dev_conf.rxmode.max_rx_pkt_len;
 435         vf->num_queue_pairs = RTE_MAX(dev->data->nb_rx_queues,
 436                                       dev->data->nb_tx_queues);
 437
 438         if (iavf_init_queues(dev) != 0) {
 439                 PMD_DRV_LOG(ERR, "failed to do Queue init");
 440                 return -1;
 441         }
 442
 443         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF) {
 444                 if (iavf_init_rss(adapter) != 0) {
 445                         PMD_DRV_LOG(ERR, "configure rss failed");
 446                         goto err_rss;
 447                 }
 448         }
 449
 450         if (iavf_configure_queues(adapter) != 0) {
 451                 PMD_DRV_LOG(ERR, "configure queues failed");
 452                 goto err_queue;
 453         }
 454
 455         if (iavf_config_rx_queues_irqs(dev, intr_handle) != 0) {
 456                 PMD_DRV_LOG(ERR, "configure irq failed");
 457                 goto err_queue;
 458         }
 459         /* re-enable intr again, because efd assign may change */
 460         if (dev->data->dev_conf.intr_conf.rxq != 0) {
 461                 rte_intr_disable(intr_handle);
 462                 rte_intr_enable(intr_handle);
 463         }
 464
 465         /* Set all mac addrs */
 466         iavf_add_del_all_mac_addr(adapter, true);
 467
 468         if (iavf_start_queues(dev) != 0) {
 469                 PMD_DRV_LOG(ERR, "enable queues failed");
 470                 goto err_mac;
 471         }
 472
 473         return 0;
 474
 475 err_mac:
 476         iavf_add_del_all_mac_addr(adapter, false);
 477 err_queue:
 478 err_rss:
 479         return -1;
 480 }
 481
 482 static void
 483 iavf_dev_stop(struct rte_eth_dev *dev)
 484 {
 485         struct iavf_adapter *adapter =
 486                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 487         struct rte_intr_handle *intr_handle = dev->intr_handle;
 488
 489         PMD_INIT_FUNC_TRACE();
 490
 491         if (adapter->stopped == 1)
 492                 return;
 493
 494         iavf_stop_queues(dev);
 495
 496         /* Disable the interrupt for Rx */
 497         rte_intr_efd_disable(intr_handle);
 498         /* Rx interrupt vector mapping free */
 499         if (intr_handle->intr_vec) {
 500                 rte_free(intr_handle->intr_vec);
 501                 intr_handle->intr_vec = NULL;
 502         }
 503
 504         /* remove all mac addrs */
 505         iavf_add_del_all_mac_addr(adapter, false);
 506         adapter->stopped = 1;
 507 }
 508
 509 static int
 510 iavf_dev_info_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
 511 {
 512         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
 513
 514         dev_info->max_rx_queues = vf->vsi_res->num_queue_pairs;
 515         dev_info->max_tx_queues = vf->vsi_res->num_queue_pairs;
 516         dev_info->min_rx_bufsize = IAVF_BUF_SIZE_MIN;
 517         dev_info->max_rx_pktlen = IAVF_FRAME_SIZE_MAX;
 518         dev_info->hash_key_size = vf->vf_res->rss_key_size;
 519         dev_info->reta_size = vf->vf_res->rss_lut_size;
 520         dev_info->flow_type_rss_offloads = IAVF_RSS_OFFLOAD_ALL;
 521         dev_info->max_mac_addrs = IAVF_NUM_MACADDR_MAX;
 522         dev_info->rx_offload_capa =
 523                 DEV_RX_OFFLOAD_VLAN_STRIP |
 524                 DEV_RX_OFFLOAD_QINQ_STRIP |
 525                 DEV_RX_OFFLOAD_IPV4_CKSUM |
 526                 DEV_RX_OFFLOAD_UDP_CKSUM |
 527                 DEV_RX_OFFLOAD_TCP_CKSUM |
 528                 DEV_RX_OFFLOAD_OUTER_IPV4_CKSUM |
 529                 DEV_RX_OFFLOAD_SCATTER |
 530                 DEV_RX_OFFLOAD_JUMBO_FRAME |
 531                 DEV_RX_OFFLOAD_VLAN_FILTER |
 532                 DEV_RX_OFFLOAD_RSS_HASH;
 533         dev_info->tx_offload_capa =
 534                 DEV_TX_OFFLOAD_VLAN_INSERT |
 535                 DEV_TX_OFFLOAD_QINQ_INSERT |
 536                 DEV_TX_OFFLOAD_IPV4_CKSUM |
 537                 DEV_TX_OFFLOAD_UDP_CKSUM |
 538                 DEV_TX_OFFLOAD_TCP_CKSUM |
 539                 DEV_TX_OFFLOAD_SCTP_CKSUM |
 540                 DEV_TX_OFFLOAD_OUTER_IPV4_CKSUM |
 541                 DEV_TX_OFFLOAD_TCP_TSO |
 542                 DEV_TX_OFFLOAD_VXLAN_TNL_TSO |
 543                 DEV_TX_OFFLOAD_GRE_TNL_TSO |
 544                 DEV_TX_OFFLOAD_IPIP_TNL_TSO |
 545                 DEV_TX_OFFLOAD_GENEVE_TNL_TSO |
 546                 DEV_TX_OFFLOAD_MULTI_SEGS;
 547
 548         dev_info->default_rxconf = (struct rte_eth_rxconf) {
 549                 .rx_free_thresh = IAVF_DEFAULT_RX_FREE_THRESH,
 550                 .rx_drop_en = 0,
 551                 .offloads = 0,
 552         };
 553
 554         dev_info->default_txconf = (struct rte_eth_txconf) {
 555                 .tx_free_thresh = IAVF_DEFAULT_TX_FREE_THRESH,
 556                 .tx_rs_thresh = IAVF_DEFAULT_TX_RS_THRESH,
 557                 .offloads = 0,
 558         };
 559
 560         dev_info->rx_desc_lim = (struct rte_eth_desc_lim) {
 561                 .nb_max = IAVF_MAX_RING_DESC,
 562                 .nb_min = IAVF_MIN_RING_DESC,
 563                 .nb_align = IAVF_ALIGN_RING_DESC,
 564         };
 565
 566         dev_info->tx_desc_lim = (struct rte_eth_desc_lim) {
 567                 .nb_max = IAVF_MAX_RING_DESC,
 568                 .nb_min = IAVF_MIN_RING_DESC,
 569                 .nb_align = IAVF_ALIGN_RING_DESC,
 570         };
 571
 572         return 0;
 573 }
 574
 575 static const uint32_t *
 576 iavf_dev_supported_ptypes_get(struct rte_eth_dev *dev __rte_unused)
 577 {
 578         static const uint32_t ptypes[] = {
 579                 RTE_PTYPE_L2_ETHER,
 580                 RTE_PTYPE_L3_IPV4_EXT_UNKNOWN,
 581                 RTE_PTYPE_L4_FRAG,
 582                 RTE_PTYPE_L4_ICMP,
 583                 RTE_PTYPE_L4_NONFRAG,
 584                 RTE_PTYPE_L4_SCTP,
 585                 RTE_PTYPE_L4_TCP,
 586                 RTE_PTYPE_L4_UDP,
 587                 RTE_PTYPE_UNKNOWN
 588         };
 589         return ptypes;
 590 }
 591
 592 int
 593 iavf_dev_link_update(struct rte_eth_dev *dev,
 594                     __rte_unused int wait_to_complete)
 595 {
 596         struct rte_eth_link new_link;
 597         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
 598
 599         /* Only read status info stored in VF, and the info is updated
 600          *  when receive LINK_CHANGE evnet from PF by Virtchnnl.
 601          */
 602         switch (vf->link_speed) {
 603         case 10:
 604                 new_link.link_speed = ETH_SPEED_NUM_10M;
 605                 break;
 606         case 100:
 607                 new_link.link_speed = ETH_SPEED_NUM_100M;
 608                 break;
 609         case 1000:
 610                 new_link.link_speed = ETH_SPEED_NUM_1G;
 611                 break;
 612         case 10000:
 613                 new_link.link_speed = ETH_SPEED_NUM_10G;
 614                 break;
 615         case 20000:
 616                 new_link.link_speed = ETH_SPEED_NUM_20G;
 617                 break;
 618         case 25000:
 619                 new_link.link_speed = ETH_SPEED_NUM_25G;
 620                 break;
 621         case 40000:
 622                 new_link.link_speed = ETH_SPEED_NUM_40G;
 623                 break;
 624         case 50000:
 625                 new_link.link_speed = ETH_SPEED_NUM_50G;
 626                 break;
 627         case 100000:
 628                 new_link.link_speed = ETH_SPEED_NUM_100G;
 629                 break;
 630         default:
 631                 new_link.link_speed = ETH_SPEED_NUM_NONE;
 632                 break;
 633         }
 634
 635         new_link.link_duplex = ETH_LINK_FULL_DUPLEX;
 636         new_link.link_status = vf->link_up ? ETH_LINK_UP :
 637                                              ETH_LINK_DOWN;
 638         new_link.link_autoneg = !(dev->data->dev_conf.link_speeds &
 639                                 ETH_LINK_SPEED_FIXED);
 640
 641         if (rte_atomic64_cmpset((uint64_t *)&dev->data->dev_link,
 642                                 *(uint64_t *)&dev->data->dev_link,
 643                                 *(uint64_t *)&new_link) == 0)
 644                 return -1;
 645
 646         return 0;
 647 }
 648
 649 static int
 650 iavf_dev_promiscuous_enable(struct rte_eth_dev *dev)
 651 {
 652         struct iavf_adapter *adapter =
 653                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 654         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
 655         int ret;
 656
 657         if (vf->promisc_unicast_enabled)
 658                 return 0;
 659
 660         ret = iavf_config_promisc(adapter, true, vf->promisc_multicast_enabled);
 661         if (!ret)
 662                 vf->promisc_unicast_enabled = true;
 663         else
 664                 ret = -EAGAIN;
 665
 666         return ret;
 667 }
 668
 669 static int
 670 iavf_dev_promiscuous_disable(struct rte_eth_dev *dev)
 671 {
 672         struct iavf_adapter *adapter =
 673                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 674         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
 675         int ret;
 676
 677         if (!vf->promisc_unicast_enabled)
 678                 return 0;
 679
 680         ret = iavf_config_promisc(adapter, false,
 681                                   vf->promisc_multicast_enabled);
 682         if (!ret)
 683                 vf->promisc_unicast_enabled = false;
 684         else
 685                 ret = -EAGAIN;
 686
 687         return ret;
 688 }
 689
 690 static int
 691 iavf_dev_allmulticast_enable(struct rte_eth_dev *dev)
 692 {
 693         struct iavf_adapter *adapter =
 694                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 695         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
 696         int ret;
 697
 698         if (vf->promisc_multicast_enabled)
 699                 return 0;
 700
 701         ret = iavf_config_promisc(adapter, vf->promisc_unicast_enabled, true);
 702         if (!ret)
 703                 vf->promisc_multicast_enabled = true;
 704         else
 705                 ret = -EAGAIN;
 706
 707         return ret;
 708 }
 709
 710 static int
 711 iavf_dev_allmulticast_disable(struct rte_eth_dev *dev)
 712 {
 713         struct iavf_adapter *adapter =
 714                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 715         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
 716         int ret;
 717
 718         if (!vf->promisc_multicast_enabled)
 719                 return 0;
 720
 721         ret = iavf_config_promisc(adapter, vf->promisc_unicast_enabled, false);
 722         if (!ret)
 723                 vf->promisc_multicast_enabled = false;
 724         else
 725                 ret = -EAGAIN;
 726
 727         return ret;
 728 }
 729
 730 static int
 731 iavf_dev_add_mac_addr(struct rte_eth_dev *dev, struct rte_ether_addr *addr,
 732                      __rte_unused uint32_t index,
 733                      __rte_unused uint32_t pool)
 734 {
 735         struct iavf_adapter *adapter =
 736                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 737         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
 738         int err;
 739
 740         if (rte_is_zero_ether_addr(addr)) {
 741                 PMD_DRV_LOG(ERR, "Invalid Ethernet Address");
 742                 return -EINVAL;
 743         }
 744
 745         err = iavf_add_del_eth_addr(adapter, addr, true);
 746         if (err) {
 747                 PMD_DRV_LOG(ERR, "fail to add MAC address");
 748                 return -EIO;
 749         }
 750
 751         vf->mac_num++;
 752
 753         return 0;
 754 }
 755
 756 static void
 757 iavf_dev_del_mac_addr(struct rte_eth_dev *dev, uint32_t index)
 758 {
 759         struct iavf_adapter *adapter =
 760                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 761         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
 762         struct rte_ether_addr *addr;
 763         int err;
 764
 765         addr = &dev->data->mac_addrs[index];
 766
 767         err = iavf_add_del_eth_addr(adapter, addr, false);
 768         if (err)
 769                 PMD_DRV_LOG(ERR, "fail to delete MAC address");
 770
 771         vf->mac_num--;
 772 }
 773
 774 static int
 775 iavf_dev_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
 776 {
 777         struct iavf_adapter *adapter =
 778                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 779         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
 780         int err;
 781
 782         if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_VLAN))
 783                 return -ENOTSUP;
 784
 785         err = iavf_add_del_vlan(adapter, vlan_id, on);
 786         if (err)
 787                 return -EIO;
 788         return 0;
 789 }
 790
 791 static int
 792 iavf_dev_vlan_offload_set(struct rte_eth_dev *dev, int mask)
 793 {
 794         struct iavf_adapter *adapter =
 795                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 796         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
 797         struct rte_eth_conf *dev_conf = &dev->data->dev_conf;
 798         int err;
 799
 800         if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_VLAN))
 801                 return -ENOTSUP;
 802
 803         /* Vlan stripping setting */
 804         if (mask & ETH_VLAN_STRIP_MASK) {
 805                 /* Enable or disable VLAN stripping */
 806                 if (dev_conf->rxmode.offloads & DEV_RX_OFFLOAD_VLAN_STRIP)
 807                         err = iavf_enable_vlan_strip(adapter);
 808                 else
 809                         err = iavf_disable_vlan_strip(adapter);
 810
 811                 if (err)
 812                         return -EIO;
 813         }
 814         return 0;
 815 }
 816
 817 static int
 818 iavf_dev_rss_reta_update(struct rte_eth_dev *dev,
 819                         struct rte_eth_rss_reta_entry64 *reta_conf,
 820                         uint16_t reta_size)
 821 {
 822         struct iavf_adapter *adapter =
 823                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 824         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
 825         uint8_t *lut;
 826         uint16_t i, idx, shift;
 827         int ret;
 828
 829         if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF))
 830                 return -ENOTSUP;
 831
 832         if (reta_size != vf->vf_res->rss_lut_size) {
 833                 PMD_DRV_LOG(ERR, "The size of hash lookup table configured "
 834                         "(%d) doesn't match the number of hardware can "
 835                         "support (%d)", reta_size, vf->vf_res->rss_lut_size);
 836                 return -EINVAL;
 837         }
 838
 839         lut = rte_zmalloc("rss_lut", reta_size, 0);
 840         if (!lut) {
 841                 PMD_DRV_LOG(ERR, "No memory can be allocated");
 842                 return -ENOMEM;
 843         }
 844         /* store the old lut table temporarily */
 845         rte_memcpy(lut, vf->rss_lut, reta_size);
 846
 847         for (i = 0; i < reta_size; i++) {
 848                 idx = i / RTE_RETA_GROUP_SIZE;
 849                 shift = i % RTE_RETA_GROUP_SIZE;
 850                 if (reta_conf[idx].mask & (1ULL << shift))
 851                         lut[i] = reta_conf[idx].reta[shift];
 852         }
 853
 854         rte_memcpy(vf->rss_lut, lut, reta_size);
 855         /* send virtchnnl ops to configure rss*/
 856         ret = iavf_configure_rss_lut(adapter);
 857         if (ret) /* revert back */
 858                 rte_memcpy(vf->rss_lut, lut, reta_size);
 859         rte_free(lut);
 860
 861         return ret;
 862 }
 863
 864 static int
 865 iavf_dev_rss_reta_query(struct rte_eth_dev *dev,
 866                        struct rte_eth_rss_reta_entry64 *reta_conf,
 867                        uint16_t reta_size)
 868 {
 869         struct iavf_adapter *adapter =
 870                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 871         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
 872         uint16_t i, idx, shift;
 873
 874         if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF))
 875                 return -ENOTSUP;
 876
 877         if (reta_size != vf->vf_res->rss_lut_size) {
 878                 PMD_DRV_LOG(ERR, "The size of hash lookup table configured "
 879                         "(%d) doesn't match the number of hardware can "
 880                         "support (%d)", reta_size, vf->vf_res->rss_lut_size);
 881                 return -EINVAL;
 882         }
 883
 884         for (i = 0; i < reta_size; i++) {
 885                 idx = i / RTE_RETA_GROUP_SIZE;
 886                 shift = i % RTE_RETA_GROUP_SIZE;
 887                 if (reta_conf[idx].mask & (1ULL << shift))
 888                         reta_conf[idx].reta[shift] = vf->rss_lut[i];
 889         }
 890
 891         return 0;
 892 }
 893
 894 static int
 895 iavf_dev_rss_hash_update(struct rte_eth_dev *dev,
 896                         struct rte_eth_rss_conf *rss_conf)
 897 {
 898         struct iavf_adapter *adapter =
 899                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 900         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
 901
 902         if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF))
 903                 return -ENOTSUP;
 904
 905         /* HENA setting, it is enabled by default, no change */
 906         if (!rss_conf->rss_key || rss_conf->rss_key_len == 0) {
 907                 PMD_DRV_LOG(DEBUG, "No key to be configured");
 908                 return 0;
 909         } else if (rss_conf->rss_key_len != vf->vf_res->rss_key_size) {
 910                 PMD_DRV_LOG(ERR, "The size of hash key configured "
 911                         "(%d) doesn't match the size of hardware can "
 912                         "support (%d)", rss_conf->rss_key_len,
 913                         vf->vf_res->rss_key_size);
 914                 return -EINVAL;
 915         }
 916
 917         rte_memcpy(vf->rss_key, rss_conf->rss_key, rss_conf->rss_key_len);
 918
 919         return iavf_configure_rss_key(adapter);
 920 }
 921
 922 static int
 923 iavf_dev_rss_hash_conf_get(struct rte_eth_dev *dev,
 924                           struct rte_eth_rss_conf *rss_conf)
 925 {
 926         struct iavf_adapter *adapter =
 927                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 928         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
 929
 930         if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF))
 931                 return -ENOTSUP;
 932
 933          /* Just set it to default value now. */
 934         rss_conf->rss_hf = IAVF_RSS_OFFLOAD_ALL;
 935
 936         if (!rss_conf->rss_key)
 937                 return 0;
 938
 939         rss_conf->rss_key_len = vf->vf_res->rss_key_size;
 940         rte_memcpy(rss_conf->rss_key, vf->rss_key, rss_conf->rss_key_len);
 941
 942         return 0;
 943 }
 944
 945 static int
 946 iavf_dev_mtu_set(struct rte_eth_dev *dev, uint16_t mtu)
 947 {
 948         uint32_t frame_size = mtu + IAVF_ETH_OVERHEAD;
 949         int ret = 0;
 950
 951         if (mtu < RTE_ETHER_MIN_MTU || frame_size > IAVF_FRAME_SIZE_MAX)
 952                 return -EINVAL;
 953
 954         /* mtu setting is forbidden if port is start */
 955         if (dev->data->dev_started) {
 956                 PMD_DRV_LOG(ERR, "port must be stopped before configuration");
 957                 return -EBUSY;
 958         }
 959
 960         if (frame_size > RTE_ETHER_MAX_LEN)
 961                 dev->data->dev_conf.rxmode.offloads |=
 962                                 DEV_RX_OFFLOAD_JUMBO_FRAME;
 963         else
 964                 dev->data->dev_conf.rxmode.offloads &=
 965                                 ~DEV_RX_OFFLOAD_JUMBO_FRAME;
 966
 967         dev->data->dev_conf.rxmode.max_rx_pkt_len = frame_size;
 968
 969         return ret;
 970 }
 971
 972 static int
 973 iavf_dev_set_default_mac_addr(struct rte_eth_dev *dev,
 974                              struct rte_ether_addr *mac_addr)
 975 {
 976         struct iavf_adapter *adapter =
 977                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 978         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(adapter);
 979         struct rte_ether_addr *perm_addr, *old_addr;
 980         int ret;
 981
 982         old_addr = (struct rte_ether_addr *)hw->mac.addr;
 983         perm_addr = (struct rte_ether_addr *)hw->mac.perm_addr;
 984
 985         if (rte_is_same_ether_addr(mac_addr, old_addr))
 986                 return 0;
 987
 988         /* If the MAC address is configured by host, skip the setting */
 989         if (rte_is_valid_assigned_ether_addr(perm_addr))
 990                 return -EPERM;
 991
 992         ret = iavf_add_del_eth_addr(adapter, old_addr, false);
 993         if (ret)
 994                 PMD_DRV_LOG(ERR, "Fail to delete old MAC:"
 995                             " %02X:%02X:%02X:%02X:%02X:%02X",
 996                             old_addr->addr_bytes[0],
 997                             old_addr->addr_bytes[1],
 998                             old_addr->addr_bytes[2],
 999                             old_addr->addr_bytes[3],
1000                             old_addr->addr_bytes[4],
1001                             old_addr->addr_bytes[5]);
1002
1003         ret = iavf_add_del_eth_addr(adapter, mac_addr, true);
1004         if (ret)
1005                 PMD_DRV_LOG(ERR, "Fail to add new MAC:"
1006                             " %02X:%02X:%02X:%02X:%02X:%02X",
1007                             mac_addr->addr_bytes[0],
1008                             mac_addr->addr_bytes[1],
1009                             mac_addr->addr_bytes[2],
1010                             mac_addr->addr_bytes[3],
1011                             mac_addr->addr_bytes[4],
1012                             mac_addr->addr_bytes[5]);
1013
1014         if (ret)
1015                 return -EIO;
1016
1017         rte_ether_addr_copy(mac_addr, (struct rte_ether_addr *)hw->mac.addr);
1018         return 0;
1019 }
1020
1021 static void
1022 iavf_stat_update_48(uint64_t *offset, uint64_t *stat)
1023 {
1024         if (*stat >= *offset)
1025                 *stat = *stat - *offset;
1026         else
1027                 *stat = (uint64_t)((*stat +
1028                         ((uint64_t)1 << IAVF_48_BIT_WIDTH)) - *offset);
1029
1030         *stat &= IAVF_48_BIT_MASK;
1031 }
1032
1033 static void
1034 iavf_stat_update_32(uint64_t *offset, uint64_t *stat)
1035 {
1036         if (*stat >= *offset)
1037                 *stat = (uint64_t)(*stat - *offset);
1038         else
1039                 *stat = (uint64_t)((*stat +
1040                         ((uint64_t)1 << IAVF_32_BIT_WIDTH)) - *offset);
1041 }
1042
1043 static void
1044 iavf_update_stats(struct iavf_vsi *vsi, struct virtchnl_eth_stats *nes)
1045 {
1046         struct virtchnl_eth_stats *oes = &vsi->eth_stats_offset;
1047
1048         iavf_stat_update_48(&oes->rx_bytes, &nes->rx_bytes);
1049         iavf_stat_update_48(&oes->rx_unicast, &nes->rx_unicast);
1050         iavf_stat_update_48(&oes->rx_multicast, &nes->rx_multicast);
1051         iavf_stat_update_48(&oes->rx_broadcast, &nes->rx_broadcast);
1052         iavf_stat_update_32(&oes->rx_discards, &nes->rx_discards);
1053         iavf_stat_update_48(&oes->tx_bytes, &nes->tx_bytes);
1054         iavf_stat_update_48(&oes->tx_unicast, &nes->tx_unicast);
1055         iavf_stat_update_48(&oes->tx_multicast, &nes->tx_multicast);
1056         iavf_stat_update_48(&oes->tx_broadcast, &nes->tx_broadcast);
1057         iavf_stat_update_32(&oes->tx_errors, &nes->tx_errors);
1058         iavf_stat_update_32(&oes->tx_discards, &nes->tx_discards);
1059 }
1060
1061 static int
1062 iavf_dev_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
1063 {
1064         struct iavf_adapter *adapter =
1065                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1066         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
1067         struct iavf_vsi *vsi = &vf->vsi;
1068         struct virtchnl_eth_stats *pstats = NULL;
1069         int ret;
1070
1071         ret = iavf_query_stats(adapter, &pstats);
1072         if (ret == 0) {
1073                 iavf_update_stats(vsi, pstats);
1074                 stats->ipackets = pstats->rx_unicast + pstats->rx_multicast +
1075                                 pstats->rx_broadcast - pstats->rx_discards;
1076                 stats->opackets = pstats->tx_broadcast + pstats->tx_multicast +
1077                                                 pstats->tx_unicast;
1078                 stats->imissed = pstats->rx_discards;
1079                 stats->oerrors = pstats->tx_errors + pstats->tx_discards;
1080                 stats->ibytes = pstats->rx_bytes;
1081                 stats->ibytes -= stats->ipackets * RTE_ETHER_CRC_LEN;
1082                 stats->obytes = pstats->tx_bytes;
1083         } else {
1084                 PMD_DRV_LOG(ERR, "Get statistics failed");
1085         }
1086         return ret;
1087 }
1088
1089 static int
1090 iavf_dev_stats_reset(struct rte_eth_dev *dev)
1091 {
1092         int ret;
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
1099         /* read stat values to clear hardware registers */
1100         ret = iavf_query_stats(adapter, &pstats);
1101         if (ret != 0)
1102                 return ret;
1103
1104         /* set stats offset base on current values */
1105         vsi->eth_stats_offset = *pstats;
1106
1107         return 0;
1108 }
1109
1110 static int
1111 iavf_dev_rx_queue_intr_enable(struct rte_eth_dev *dev, uint16_t queue_id)
1112 {
1113         struct iavf_adapter *adapter =
1114                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1115         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1116         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(adapter);
1117         uint16_t msix_intr;
1118
1119         msix_intr = pci_dev->intr_handle.intr_vec[queue_id];
1120         if (msix_intr == IAVF_MISC_VEC_ID) {
1121                 PMD_DRV_LOG(INFO, "MISC is also enabled for control");
1122                 IAVF_WRITE_REG(hw, IAVF_VFINT_DYN_CTL01,
1123                                IAVF_VFINT_DYN_CTL01_INTENA_MASK |
1124                                IAVF_VFINT_DYN_CTL01_CLEARPBA_MASK |
1125                                IAVF_VFINT_DYN_CTL01_ITR_INDX_MASK);
1126         } else {
1127                 IAVF_WRITE_REG(hw,
1128                                IAVF_VFINT_DYN_CTLN1
1129                                 (msix_intr - IAVF_RX_VEC_START),
1130                                IAVF_VFINT_DYN_CTLN1_INTENA_MASK |
1131                                IAVF_VFINT_DYN_CTL01_CLEARPBA_MASK |
1132                                IAVF_VFINT_DYN_CTLN1_ITR_INDX_MASK);
1133         }
1134
1135         IAVF_WRITE_FLUSH(hw);
1136
1137         rte_intr_ack(&pci_dev->intr_handle);
1138
1139         return 0;
1140 }
1141
1142 static int
1143 iavf_dev_rx_queue_intr_disable(struct rte_eth_dev *dev, uint16_t queue_id)
1144 {
1145         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1146         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(dev->data->dev_private);
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(ERR, "MISC is used for control, cannot disable it");
1152                 return -EIO;
1153         }
1154
1155         IAVF_WRITE_REG(hw,
1156                       IAVF_VFINT_DYN_CTLN1(msix_intr - IAVF_RX_VEC_START),
1157                       0);
1158
1159         IAVF_WRITE_FLUSH(hw);
1160         return 0;
1161 }
1162
1163 static int
1164 iavf_check_vf_reset_done(struct iavf_hw *hw)
1165 {
1166         int i, reset;
1167
1168         for (i = 0; i < IAVF_RESET_WAIT_CNT; i++) {
1169                 reset = IAVF_READ_REG(hw, IAVF_VFGEN_RSTAT) &
1170                         IAVF_VFGEN_RSTAT_VFR_STATE_MASK;
1171                 reset = reset >> IAVF_VFGEN_RSTAT_VFR_STATE_SHIFT;
1172                 if (reset == VIRTCHNL_VFR_VFACTIVE ||
1173                     reset == VIRTCHNL_VFR_COMPLETED)
1174                         break;
1175                 rte_delay_ms(20);
1176         }
1177
1178         if (i >= IAVF_RESET_WAIT_CNT)
1179                 return -1;
1180
1181         return 0;
1182 }
1183
1184 static int
1185 iavf_init_vf(struct rte_eth_dev *dev)
1186 {
1187         int err, bufsz;
1188         struct iavf_adapter *adapter =
1189                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1190         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1191         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
1192
1193         err = iavf_set_mac_type(hw);
1194         if (err) {
1195                 PMD_INIT_LOG(ERR, "set_mac_type failed: %d", err);
1196                 goto err;
1197         }
1198
1199         err = iavf_check_vf_reset_done(hw);
1200         if (err) {
1201                 PMD_INIT_LOG(ERR, "VF is still resetting");
1202                 goto err;
1203         }
1204
1205         iavf_init_adminq_parameter(hw);
1206         err = iavf_init_adminq(hw);
1207         if (err) {
1208                 PMD_INIT_LOG(ERR, "init_adminq failed: %d", err);
1209                 goto err;
1210         }
1211
1212         vf->aq_resp = rte_zmalloc("vf_aq_resp", IAVF_AQ_BUF_SZ, 0);
1213         if (!vf->aq_resp) {
1214                 PMD_INIT_LOG(ERR, "unable to allocate vf_aq_resp memory");
1215                 goto err_aq;
1216         }
1217         if (iavf_check_api_version(adapter) != 0) {
1218                 PMD_INIT_LOG(ERR, "check_api version failed");
1219                 goto err_api;
1220         }
1221
1222         bufsz = sizeof(struct virtchnl_vf_resource) +
1223                 (IAVF_MAX_VF_VSI * sizeof(struct virtchnl_vsi_resource));
1224         vf->vf_res = rte_zmalloc("vf_res", bufsz, 0);
1225         if (!vf->vf_res) {
1226                 PMD_INIT_LOG(ERR, "unable to allocate vf_res memory");
1227                 goto err_api;
1228         }
1229         if (iavf_get_vf_resource(adapter) != 0) {
1230                 PMD_INIT_LOG(ERR, "iavf_get_vf_config failed");
1231                 goto err_alloc;
1232         }
1233         /* Allocate memort for RSS info */
1234         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF) {
1235                 vf->rss_key = rte_zmalloc("rss_key",
1236                                           vf->vf_res->rss_key_size, 0);
1237                 if (!vf->rss_key) {
1238                         PMD_INIT_LOG(ERR, "unable to allocate rss_key memory");
1239                         goto err_rss;
1240                 }
1241                 vf->rss_lut = rte_zmalloc("rss_lut",
1242                                           vf->vf_res->rss_lut_size, 0);
1243                 if (!vf->rss_lut) {
1244                         PMD_INIT_LOG(ERR, "unable to allocate rss_lut memory");
1245                         goto err_rss;
1246                 }
1247         }
1248
1249         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RX_FLEX_DESC) {
1250                 if (iavf_get_supported_rxdid(adapter) != 0) {
1251                         PMD_INIT_LOG(ERR, "failed to do get supported rxdid");
1252                         goto err_rss;
1253                 }
1254         }
1255
1256         return 0;
1257 err_rss:
1258         rte_free(vf->rss_key);
1259         rte_free(vf->rss_lut);
1260 err_alloc:
1261         rte_free(vf->vf_res);
1262         vf->vsi_res = NULL;
1263 err_api:
1264         rte_free(vf->aq_resp);
1265 err_aq:
1266         iavf_shutdown_adminq(hw);
1267 err:
1268         return -1;
1269 }
1270
1271 /* Enable default admin queue interrupt setting */
1272 static inline void
1273 iavf_enable_irq0(struct iavf_hw *hw)
1274 {
1275         /* Enable admin queue interrupt trigger */
1276         IAVF_WRITE_REG(hw, IAVF_VFINT_ICR0_ENA1,
1277                        IAVF_VFINT_ICR0_ENA1_ADMINQ_MASK);
1278
1279         IAVF_WRITE_REG(hw, IAVF_VFINT_DYN_CTL01,
1280                        IAVF_VFINT_DYN_CTL01_INTENA_MASK |
1281                        IAVF_VFINT_DYN_CTL01_CLEARPBA_MASK |
1282                        IAVF_VFINT_DYN_CTL01_ITR_INDX_MASK);
1283
1284         IAVF_WRITE_FLUSH(hw);
1285 }
1286
1287 static inline void
1288 iavf_disable_irq0(struct iavf_hw *hw)
1289 {
1290         /* Disable all interrupt types */
1291         IAVF_WRITE_REG(hw, IAVF_VFINT_ICR0_ENA1, 0);
1292         IAVF_WRITE_REG(hw, IAVF_VFINT_DYN_CTL01,
1293                        IAVF_VFINT_DYN_CTL01_ITR_INDX_MASK);
1294         IAVF_WRITE_FLUSH(hw);
1295 }
1296
1297 static void
1298 iavf_dev_interrupt_handler(void *param)
1299 {
1300         struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
1301         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1302
1303         iavf_disable_irq0(hw);
1304
1305         iavf_handle_virtchnl_msg(dev);
1306
1307         iavf_enable_irq0(hw);
1308 }
1309
1310 static int
1311 iavf_dev_filter_ctrl(struct rte_eth_dev *dev,
1312                      enum rte_filter_type filter_type,
1313                      enum rte_filter_op filter_op,
1314                      void *arg)
1315 {
1316         int ret = 0;
1317
1318         if (!dev)
1319                 return -EINVAL;
1320
1321         switch (filter_type) {
1322         case RTE_ETH_FILTER_GENERIC:
1323                 if (filter_op != RTE_ETH_FILTER_GET)
1324                         return -EINVAL;
1325                 *(const void **)arg = &iavf_flow_ops;
1326                 break;
1327         default:
1328                 PMD_DRV_LOG(WARNING, "Filter type (%d) not supported",
1329                             filter_type);
1330                 ret = -EINVAL;
1331                 break;
1332         }
1333
1334         return ret;
1335 }
1336
1337
1338 static int
1339 iavf_dev_init(struct rte_eth_dev *eth_dev)
1340 {
1341         struct iavf_adapter *adapter =
1342                 IAVF_DEV_PRIVATE_TO_ADAPTER(eth_dev->data->dev_private);
1343         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(adapter);
1344         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
1345         int ret = 0;
1346
1347         PMD_INIT_FUNC_TRACE();
1348
1349         /* assign ops func pointer */
1350         eth_dev->dev_ops = &iavf_eth_dev_ops;
1351         eth_dev->rx_pkt_burst = &iavf_recv_pkts;
1352         eth_dev->tx_pkt_burst = &iavf_xmit_pkts;
1353         eth_dev->tx_pkt_prepare = &iavf_prep_pkts;
1354
1355         /* For secondary processes, we don't initialise any further as primary
1356          * has already done this work. Only check if we need a different RX
1357          * and TX function.
1358          */
1359         if (rte_eal_process_type() != RTE_PROC_PRIMARY) {
1360                 iavf_set_rx_function(eth_dev);
1361                 iavf_set_tx_function(eth_dev);
1362                 return 0;
1363         }
1364         rte_eth_copy_pci_info(eth_dev, pci_dev);
1365
1366         hw->vendor_id = pci_dev->id.vendor_id;
1367         hw->device_id = pci_dev->id.device_id;
1368         hw->subsystem_vendor_id = pci_dev->id.subsystem_vendor_id;
1369         hw->subsystem_device_id = pci_dev->id.subsystem_device_id;
1370         hw->bus.bus_id = pci_dev->addr.bus;
1371         hw->bus.device = pci_dev->addr.devid;
1372         hw->bus.func = pci_dev->addr.function;
1373         hw->hw_addr = (void *)pci_dev->mem_resource[0].addr;
1374         hw->back = IAVF_DEV_PRIVATE_TO_ADAPTER(eth_dev->data->dev_private);
1375         adapter->eth_dev = eth_dev;
1376         adapter->stopped = 1;
1377
1378         if (iavf_init_vf(eth_dev) != 0) {
1379                 PMD_INIT_LOG(ERR, "Init vf failed");
1380                 return -1;
1381         }
1382
1383         /* set default ptype table */
1384         adapter->ptype_tbl = iavf_get_default_ptype_table();
1385
1386         /* copy mac addr */
1387         eth_dev->data->mac_addrs = rte_zmalloc(
1388                 "iavf_mac", RTE_ETHER_ADDR_LEN * IAVF_NUM_MACADDR_MAX, 0);
1389         if (!eth_dev->data->mac_addrs) {
1390                 PMD_INIT_LOG(ERR, "Failed to allocate %d bytes needed to"
1391                              " store MAC addresses",
1392                              RTE_ETHER_ADDR_LEN * IAVF_NUM_MACADDR_MAX);
1393                 return -ENOMEM;
1394         }
1395         /* If the MAC address is not configured by host,
1396          * generate a random one.
1397          */
1398         if (!rte_is_valid_assigned_ether_addr(
1399                         (struct rte_ether_addr *)hw->mac.addr))
1400                 rte_eth_random_addr(hw->mac.addr);
1401         rte_ether_addr_copy((struct rte_ether_addr *)hw->mac.addr,
1402                         &eth_dev->data->mac_addrs[0]);
1403
1404         /* register callback func to eal lib */
1405         rte_intr_callback_register(&pci_dev->intr_handle,
1406                                    iavf_dev_interrupt_handler,
1407                                    (void *)eth_dev);
1408
1409         /* enable uio intr after callback register */
1410         rte_intr_enable(&pci_dev->intr_handle);
1411
1412         /* configure and enable device interrupt */
1413         iavf_enable_irq0(hw);
1414
1415         ret = iavf_flow_init(adapter);
1416         if (ret) {
1417                 PMD_INIT_LOG(ERR, "Failed to initialize flow");
1418                 return ret;
1419         }
1420
1421         return 0;
1422 }
1423
1424 static void
1425 iavf_dev_close(struct rte_eth_dev *dev)
1426 {
1427         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1428         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1429         struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
1430         struct iavf_adapter *adapter =
1431                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1432
1433         iavf_dev_stop(dev);
1434         iavf_shutdown_adminq(hw);
1435         /* disable uio intr before callback unregister */
1436         rte_intr_disable(intr_handle);
1437
1438         /* unregister callback func from eal lib */
1439         rte_intr_callback_unregister(intr_handle,
1440                                      iavf_dev_interrupt_handler, dev);
1441         iavf_disable_irq0(hw);
1442
1443         iavf_flow_uninit(adapter);
1444 }
1445
1446 static int
1447 iavf_dev_uninit(struct rte_eth_dev *dev)
1448 {
1449         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
1450
1451         if (rte_eal_process_type() != RTE_PROC_PRIMARY)
1452                 return -EPERM;
1453
1454         dev->dev_ops = NULL;
1455         dev->rx_pkt_burst = NULL;
1456         dev->tx_pkt_burst = NULL;
1457         iavf_dev_close(dev);
1458
1459         rte_free(vf->vf_res);
1460         vf->vsi_res = NULL;
1461         vf->vf_res = NULL;
1462
1463         rte_free(vf->aq_resp);
1464         vf->aq_resp = NULL;
1465
1466         if (vf->rss_lut) {
1467                 rte_free(vf->rss_lut);
1468                 vf->rss_lut = NULL;
1469         }
1470         if (vf->rss_key) {
1471                 rte_free(vf->rss_key);
1472                 vf->rss_key = NULL;
1473         }
1474
1475         return 0;
1476 }
1477
1478 /*
1479  * Reset VF device only to re-initialize resources in PMD layer
1480  */
1481 static int
1482 iavf_dev_reset(struct rte_eth_dev *dev)
1483 {
1484         int ret;
1485
1486         ret = iavf_dev_uninit(dev);
1487         if (ret)
1488                 return ret;
1489
1490         return iavf_dev_init(dev);
1491 }
1492
1493 static int
1494 iavf_dcf_cap_check_handler(__rte_unused const char *key,
1495                            const char *value, __rte_unused void *opaque)
1496 {
1497         if (strcmp(value, "dcf"))
1498                 return -1;
1499
1500         return 0;
1501 }
1502
1503 static int
1504 iavf_dcf_cap_selected(struct rte_devargs *devargs)
1505 {
1506         struct rte_kvargs *kvlist;
1507         const char *key = "cap";
1508         int ret = 0;
1509
1510         if (devargs == NULL)
1511                 return 0;
1512
1513         kvlist = rte_kvargs_parse(devargs->args, NULL);
1514         if (kvlist == NULL)
1515                 return 0;
1516
1517         if (!rte_kvargs_count(kvlist, key))
1518                 goto exit;
1519
1520         /* dcf capability selected when there's a key-value pair: cap=dcf */
1521         if (rte_kvargs_process(kvlist, key,
1522                                iavf_dcf_cap_check_handler, NULL) < 0)
1523                 goto exit;
1524
1525         ret = 1;
1526
1527 exit:
1528         rte_kvargs_free(kvlist);
1529         return ret;
1530 }
1531
1532 static int eth_iavf_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
1533                              struct rte_pci_device *pci_dev)
1534 {
1535         if (iavf_dcf_cap_selected(pci_dev->device.devargs))
1536                 return 1;
1537
1538         return rte_eth_dev_pci_generic_probe(pci_dev,
1539                 sizeof(struct iavf_adapter), iavf_dev_init);
1540 }
1541
1542 static int eth_iavf_pci_remove(struct rte_pci_device *pci_dev)
1543 {
1544         return rte_eth_dev_pci_generic_remove(pci_dev, iavf_dev_uninit);
1545 }
1546
1547 /* Adaptive virtual function driver struct */
1548 static struct rte_pci_driver rte_iavf_pmd = {
1549         .id_table = pci_id_iavf_map,
1550         .drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC,
1551         .probe = eth_iavf_pci_probe,
1552         .remove = eth_iavf_pci_remove,
1553 };
1554
1555 RTE_PMD_REGISTER_PCI(net_iavf, rte_iavf_pmd);
1556 RTE_PMD_REGISTER_PCI_TABLE(net_iavf, pci_id_iavf_map);
1557 RTE_PMD_REGISTER_KMOD_DEP(net_iavf, "* igb_uio | vfio-pci");
1558 RTE_PMD_REGISTER_PARAM_STRING(net_iavf, "cap=dcf");
1559 RTE_INIT(iavf_init_log)
1560 {
1561         iavf_logtype_init = rte_log_register("pmd.net.iavf.init");
1562         if (iavf_logtype_init >= 0)
1563                 rte_log_set_level(iavf_logtype_init, RTE_LOG_NOTICE);
1564         iavf_logtype_driver = rte_log_register("pmd.net.iavf.driver");
1565         if (iavf_logtype_driver >= 0)
1566                 rte_log_set_level(iavf_logtype_driver, RTE_LOG_NOTICE);
1567
1568 #ifdef RTE_LIBRTE_IAVF_DEBUG_RX
1569         iavf_logtype_rx = rte_log_register("pmd.net.iavf.rx");
1570         if (iavf_logtype_rx >= 0)
1571                 rte_log_set_level(iavf_logtype_rx, RTE_LOG_DEBUG);
1572 #endif
1573
1574 #ifdef RTE_LIBRTE_IAVF_DEBUG_TX
1575         iavf_logtype_tx = rte_log_register("pmd.net.iavf.tx");
1576         if (iavf_logtype_tx >= 0)
1577                 rte_log_set_level(iavf_logtype_tx, RTE_LOG_DEBUG);
1578 #endif
1579
1580 #ifdef RTE_LIBRTE_IAVF_DEBUG_TX_FREE
1581         iavf_logtype_tx_free = rte_log_register("pmd.net.iavf.tx_free");
1582         if (iavf_logtype_tx_free >= 0)
1583                 rte_log_set_level(iavf_logtype_tx_free, RTE_LOG_DEBUG);
1584 #endif
1585 }