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 <ethdev_driver.h>
  23 #include <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 #include "rte_pmd_iavf.h"
  32
  33 /* devargs */
  34 #define IAVF_PROTO_XTR_ARG         "proto_xtr"
  35
  36 static const char * const iavf_valid_args[] = {
  37         IAVF_PROTO_XTR_ARG,
  38         NULL
  39 };
  40
  41 static const struct rte_mbuf_dynfield iavf_proto_xtr_metadata_param = {
  42         .name = "intel_pmd_dynfield_proto_xtr_metadata",
  43         .size = sizeof(uint32_t),
  44         .align = __alignof__(uint32_t),
  45         .flags = 0,
  46 };
  47
  48 struct iavf_proto_xtr_ol {
  49         const struct rte_mbuf_dynflag param;
  50         uint64_t *ol_flag;
  51         bool required;
  52 };
  53
  54 static struct iavf_proto_xtr_ol iavf_proto_xtr_params[] = {
  55         [IAVF_PROTO_XTR_VLAN] = {
  56                 .param = { .name = "intel_pmd_dynflag_proto_xtr_vlan" },
  57                 .ol_flag = &rte_pmd_ifd_dynflag_proto_xtr_vlan_mask },
  58         [IAVF_PROTO_XTR_IPV4] = {
  59                 .param = { .name = "intel_pmd_dynflag_proto_xtr_ipv4" },
  60                 .ol_flag = &rte_pmd_ifd_dynflag_proto_xtr_ipv4_mask },
  61         [IAVF_PROTO_XTR_IPV6] = {
  62                 .param = { .name = "intel_pmd_dynflag_proto_xtr_ipv6" },
  63                 .ol_flag = &rte_pmd_ifd_dynflag_proto_xtr_ipv6_mask },
  64         [IAVF_PROTO_XTR_IPV6_FLOW] = {
  65                 .param = { .name = "intel_pmd_dynflag_proto_xtr_ipv6_flow" },
  66                 .ol_flag = &rte_pmd_ifd_dynflag_proto_xtr_ipv6_flow_mask },
  67         [IAVF_PROTO_XTR_TCP] = {
  68                 .param = { .name = "intel_pmd_dynflag_proto_xtr_tcp" },
  69                 .ol_flag = &rte_pmd_ifd_dynflag_proto_xtr_tcp_mask },
  70         [IAVF_PROTO_XTR_IP_OFFSET] = {
  71                 .param = { .name = "intel_pmd_dynflag_proto_xtr_ip_offset" },
  72                 .ol_flag = &rte_pmd_ifd_dynflag_proto_xtr_ip_offset_mask },
  73 };
  74
  75 static int iavf_dev_configure(struct rte_eth_dev *dev);
  76 static int iavf_dev_start(struct rte_eth_dev *dev);
  77 static int iavf_dev_stop(struct rte_eth_dev *dev);
  78 static int iavf_dev_close(struct rte_eth_dev *dev);
  79 static int iavf_dev_reset(struct rte_eth_dev *dev);
  80 static int iavf_dev_info_get(struct rte_eth_dev *dev,
  81                              struct rte_eth_dev_info *dev_info);
  82 static const uint32_t *iavf_dev_supported_ptypes_get(struct rte_eth_dev *dev);
  83 static int iavf_dev_stats_get(struct rte_eth_dev *dev,
  84                              struct rte_eth_stats *stats);
  85 static int iavf_dev_stats_reset(struct rte_eth_dev *dev);
  86 static int iavf_dev_xstats_get(struct rte_eth_dev *dev,
  87                                  struct rte_eth_xstat *xstats, unsigned int n);
  88 static int iavf_dev_xstats_get_names(struct rte_eth_dev *dev,
  89                                        struct rte_eth_xstat_name *xstats_names,
  90                                        unsigned int limit);
  91 static int iavf_dev_promiscuous_enable(struct rte_eth_dev *dev);
  92 static int iavf_dev_promiscuous_disable(struct rte_eth_dev *dev);
  93 static int iavf_dev_allmulticast_enable(struct rte_eth_dev *dev);
  94 static int iavf_dev_allmulticast_disable(struct rte_eth_dev *dev);
  95 static int iavf_dev_add_mac_addr(struct rte_eth_dev *dev,
  96                                 struct rte_ether_addr *addr,
  97                                 uint32_t index,
  98                                 uint32_t pool);
  99 static void iavf_dev_del_mac_addr(struct rte_eth_dev *dev, uint32_t index);
 100 static int iavf_dev_vlan_filter_set(struct rte_eth_dev *dev,
 101                                    uint16_t vlan_id, int on);
 102 static int iavf_dev_vlan_offload_set(struct rte_eth_dev *dev, int mask);
 103 static int iavf_dev_rss_reta_update(struct rte_eth_dev *dev,
 104                                    struct rte_eth_rss_reta_entry64 *reta_conf,
 105                                    uint16_t reta_size);
 106 static int iavf_dev_rss_reta_query(struct rte_eth_dev *dev,
 107                                   struct rte_eth_rss_reta_entry64 *reta_conf,
 108                                   uint16_t reta_size);
 109 static int iavf_dev_rss_hash_update(struct rte_eth_dev *dev,
 110                                    struct rte_eth_rss_conf *rss_conf);
 111 static int iavf_dev_rss_hash_conf_get(struct rte_eth_dev *dev,
 112                                      struct rte_eth_rss_conf *rss_conf);
 113 static int iavf_dev_mtu_set(struct rte_eth_dev *dev, uint16_t mtu);
 114 static int iavf_dev_set_default_mac_addr(struct rte_eth_dev *dev,
 115                                          struct rte_ether_addr *mac_addr);
 116 static int iavf_dev_rx_queue_intr_enable(struct rte_eth_dev *dev,
 117                                         uint16_t queue_id);
 118 static int iavf_dev_rx_queue_intr_disable(struct rte_eth_dev *dev,
 119                                          uint16_t queue_id);
 120 static int iavf_dev_filter_ctrl(struct rte_eth_dev *dev,
 121                      enum rte_filter_type filter_type,
 122                      enum rte_filter_op filter_op,
 123                      void *arg);
 124 static int iavf_set_mc_addr_list(struct rte_eth_dev *dev,
 125                         struct rte_ether_addr *mc_addrs,
 126                         uint32_t mc_addrs_num);
 127
 128 static const struct rte_pci_id pci_id_iavf_map[] = {
 129         { RTE_PCI_DEVICE(IAVF_INTEL_VENDOR_ID, IAVF_DEV_ID_ADAPTIVE_VF) },
 130         { .vendor_id = 0, /* sentinel */ },
 131 };
 132
 133 struct rte_iavf_xstats_name_off {
 134         char name[RTE_ETH_XSTATS_NAME_SIZE];
 135         unsigned int offset;
 136 };
 137
 138 static const struct rte_iavf_xstats_name_off rte_iavf_stats_strings[] = {
 139         {"rx_bytes", offsetof(struct iavf_eth_stats, rx_bytes)},
 140         {"rx_unicast_packets", offsetof(struct iavf_eth_stats, rx_unicast)},
 141         {"rx_multicast_packets", offsetof(struct iavf_eth_stats, rx_multicast)},
 142         {"rx_broadcast_packets", offsetof(struct iavf_eth_stats, rx_broadcast)},
 143         {"rx_dropped_packets", offsetof(struct iavf_eth_stats, rx_discards)},
 144         {"rx_unknown_protocol_packets", offsetof(struct iavf_eth_stats,
 145                 rx_unknown_protocol)},
 146         {"tx_bytes", offsetof(struct iavf_eth_stats, tx_bytes)},
 147         {"tx_unicast_packets", offsetof(struct iavf_eth_stats, tx_unicast)},
 148         {"tx_multicast_packets", offsetof(struct iavf_eth_stats, tx_multicast)},
 149         {"tx_broadcast_packets", offsetof(struct iavf_eth_stats, tx_broadcast)},
 150         {"tx_dropped_packets", offsetof(struct iavf_eth_stats, tx_discards)},
 151         {"tx_error_packets", offsetof(struct iavf_eth_stats, tx_errors)},
 152 };
 153
 154 #define IAVF_NB_XSTATS (sizeof(rte_iavf_stats_strings) / \
 155                 sizeof(rte_iavf_stats_strings[0]))
 156
 157 static const struct eth_dev_ops iavf_eth_dev_ops = {
 158         .dev_configure              = iavf_dev_configure,
 159         .dev_start                  = iavf_dev_start,
 160         .dev_stop                   = iavf_dev_stop,
 161         .dev_close                  = iavf_dev_close,
 162         .dev_reset                  = iavf_dev_reset,
 163         .dev_infos_get              = iavf_dev_info_get,
 164         .dev_supported_ptypes_get   = iavf_dev_supported_ptypes_get,
 165         .link_update                = iavf_dev_link_update,
 166         .stats_get                  = iavf_dev_stats_get,
 167         .stats_reset                = iavf_dev_stats_reset,
 168         .xstats_get                 = iavf_dev_xstats_get,
 169         .xstats_get_names           = iavf_dev_xstats_get_names,
 170         .xstats_reset               = iavf_dev_stats_reset,
 171         .promiscuous_enable         = iavf_dev_promiscuous_enable,
 172         .promiscuous_disable        = iavf_dev_promiscuous_disable,
 173         .allmulticast_enable        = iavf_dev_allmulticast_enable,
 174         .allmulticast_disable       = iavf_dev_allmulticast_disable,
 175         .mac_addr_add               = iavf_dev_add_mac_addr,
 176         .mac_addr_remove            = iavf_dev_del_mac_addr,
 177         .set_mc_addr_list                       = iavf_set_mc_addr_list,
 178         .vlan_filter_set            = iavf_dev_vlan_filter_set,
 179         .vlan_offload_set           = iavf_dev_vlan_offload_set,
 180         .rx_queue_start             = iavf_dev_rx_queue_start,
 181         .rx_queue_stop              = iavf_dev_rx_queue_stop,
 182         .tx_queue_start             = iavf_dev_tx_queue_start,
 183         .tx_queue_stop              = iavf_dev_tx_queue_stop,
 184         .rx_queue_setup             = iavf_dev_rx_queue_setup,
 185         .rx_queue_release           = iavf_dev_rx_queue_release,
 186         .tx_queue_setup             = iavf_dev_tx_queue_setup,
 187         .tx_queue_release           = iavf_dev_tx_queue_release,
 188         .mac_addr_set               = iavf_dev_set_default_mac_addr,
 189         .reta_update                = iavf_dev_rss_reta_update,
 190         .reta_query                 = iavf_dev_rss_reta_query,
 191         .rss_hash_update            = iavf_dev_rss_hash_update,
 192         .rss_hash_conf_get          = iavf_dev_rss_hash_conf_get,
 193         .rxq_info_get               = iavf_dev_rxq_info_get,
 194         .txq_info_get               = iavf_dev_txq_info_get,
 195         .mtu_set                    = iavf_dev_mtu_set,
 196         .rx_queue_intr_enable       = iavf_dev_rx_queue_intr_enable,
 197         .rx_queue_intr_disable      = iavf_dev_rx_queue_intr_disable,
 198         .filter_ctrl                = iavf_dev_filter_ctrl,
 199         .tx_done_cleanup            = iavf_dev_tx_done_cleanup,
 200 };
 201
 202 static int
 203 iavf_set_mc_addr_list(struct rte_eth_dev *dev,
 204                         struct rte_ether_addr *mc_addrs,
 205                         uint32_t mc_addrs_num)
 206 {
 207         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
 208         struct iavf_adapter *adapter =
 209                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 210         int err, ret;
 211
 212         if (mc_addrs_num > IAVF_NUM_MACADDR_MAX) {
 213                 PMD_DRV_LOG(ERR,
 214                             "can't add more than a limited number (%u) of addresses.",
 215                             (uint32_t)IAVF_NUM_MACADDR_MAX);
 216                 return -EINVAL;
 217         }
 218
 219         /* flush previous addresses */
 220         err = iavf_add_del_mc_addr_list(adapter, vf->mc_addrs, vf->mc_addrs_num,
 221                                         false);
 222         if (err)
 223                 return err;
 224
 225         /* add new ones */
 226         err = iavf_add_del_mc_addr_list(adapter, mc_addrs, mc_addrs_num, true);
 227
 228         if (err) {
 229                 /* if adding mac address list fails, should add the previous
 230                  * addresses back.
 231                  */
 232                 ret = iavf_add_del_mc_addr_list(adapter, vf->mc_addrs,
 233                                                 vf->mc_addrs_num, true);
 234                 if (ret)
 235                         return ret;
 236         } else {
 237                 vf->mc_addrs_num = mc_addrs_num;
 238                 memcpy(vf->mc_addrs,
 239                        mc_addrs, mc_addrs_num * sizeof(*mc_addrs));
 240         }
 241
 242         return err;
 243 }
 244
 245 static int
 246 iavf_init_rss(struct iavf_adapter *adapter)
 247 {
 248         struct iavf_info *vf =  IAVF_DEV_PRIVATE_TO_VF(adapter);
 249         struct rte_eth_rss_conf *rss_conf;
 250         uint16_t i, j, nb_q;
 251         int ret;
 252
 253         rss_conf = &adapter->eth_dev->data->dev_conf.rx_adv_conf.rss_conf;
 254         nb_q = RTE_MIN(adapter->eth_dev->data->nb_rx_queues,
 255                        vf->max_rss_qregion);
 256
 257         if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF)) {
 258                 PMD_DRV_LOG(DEBUG, "RSS is not supported");
 259                 return -ENOTSUP;
 260         }
 261         if (adapter->eth_dev->data->dev_conf.rxmode.mq_mode != ETH_MQ_RX_RSS) {
 262                 PMD_DRV_LOG(WARNING, "RSS is enabled by PF by default");
 263                 /* set all lut items to default queue */
 264                 for (i = 0; i < vf->vf_res->rss_lut_size; i++)
 265                         vf->rss_lut[i] = 0;
 266                 ret = iavf_configure_rss_lut(adapter);
 267                 return ret;
 268         }
 269
 270         /* configure RSS key */
 271         if (!rss_conf->rss_key) {
 272                 /* Calculate the default hash key */
 273                 for (i = 0; i <= vf->vf_res->rss_key_size; i++)
 274                         vf->rss_key[i] = (uint8_t)rte_rand();
 275         } else
 276                 rte_memcpy(vf->rss_key, rss_conf->rss_key,
 277                            RTE_MIN(rss_conf->rss_key_len,
 278                                    vf->vf_res->rss_key_size));
 279
 280         /* init RSS LUT table */
 281         for (i = 0, j = 0; i < vf->vf_res->rss_lut_size; i++, j++) {
 282                 if (j >= nb_q)
 283                         j = 0;
 284                 vf->rss_lut[i] = j;
 285         }
 286         /* send virtchnnl ops to configure rss*/
 287         ret = iavf_configure_rss_lut(adapter);
 288         if (ret)
 289                 return ret;
 290         ret = iavf_configure_rss_key(adapter);
 291         if (ret)
 292                 return ret;
 293
 294         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_ADV_RSS_PF) {
 295                 /* Set RSS hash configuration based on rss_conf->rss_hf. */
 296                 ret = iavf_rss_hash_set(adapter, rss_conf->rss_hf, true);
 297                 if (ret) {
 298                         PMD_DRV_LOG(ERR, "fail to set default RSS");
 299                         return ret;
 300                 }
 301         }
 302
 303         return 0;
 304 }
 305
 306 static int
 307 iavf_queues_req_reset(struct rte_eth_dev *dev, uint16_t num)
 308 {
 309         struct iavf_adapter *ad =
 310                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 311         struct iavf_info *vf =  IAVF_DEV_PRIVATE_TO_VF(ad);
 312         int ret;
 313
 314         ret = iavf_request_queues(ad, num);
 315         if (ret) {
 316                 PMD_DRV_LOG(ERR, "request queues from PF failed");
 317                 return ret;
 318         }
 319         PMD_DRV_LOG(INFO, "change queue pairs from %u to %u",
 320                         vf->vsi_res->num_queue_pairs, num);
 321
 322         ret = iavf_dev_reset(dev);
 323         if (ret) {
 324                 PMD_DRV_LOG(ERR, "vf reset failed");
 325                 return ret;
 326         }
 327
 328         return 0;
 329 }
 330
 331 static int
 332 iavf_dev_vlan_insert_set(struct rte_eth_dev *dev)
 333 {
 334         struct iavf_adapter *adapter =
 335                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 336         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
 337         bool enable;
 338
 339         if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_VLAN_V2))
 340                 return 0;
 341
 342         enable = !!(dev->data->dev_conf.txmode.offloads &
 343                     DEV_TX_OFFLOAD_VLAN_INSERT);
 344         iavf_config_vlan_insert_v2(adapter, enable);
 345
 346         return 0;
 347 }
 348
 349 static int
 350 iavf_dev_init_vlan(struct rte_eth_dev *dev)
 351 {
 352         int err;
 353
 354         err = iavf_dev_vlan_offload_set(dev,
 355                                         ETH_VLAN_STRIP_MASK |
 356                                         ETH_QINQ_STRIP_MASK |
 357                                         ETH_VLAN_FILTER_MASK |
 358                                         ETH_VLAN_EXTEND_MASK);
 359         if (err) {
 360                 PMD_DRV_LOG(ERR, "Failed to update vlan offload");
 361                 return err;
 362         }
 363
 364         err = iavf_dev_vlan_insert_set(dev);
 365         if (err)
 366                 PMD_DRV_LOG(ERR, "Failed to update vlan insertion");
 367
 368         return err;
 369 }
 370
 371 static int
 372 iavf_dev_configure(struct rte_eth_dev *dev)
 373 {
 374         struct iavf_adapter *ad =
 375                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 376         struct iavf_info *vf =  IAVF_DEV_PRIVATE_TO_VF(ad);
 377         uint16_t num_queue_pairs = RTE_MAX(dev->data->nb_rx_queues,
 378                 dev->data->nb_tx_queues);
 379         int ret;
 380
 381         ad->rx_bulk_alloc_allowed = true;
 382         /* Initialize to TRUE. If any of Rx queues doesn't meet the
 383          * vector Rx/Tx preconditions, it will be reset.
 384          */
 385         ad->rx_vec_allowed = true;
 386         ad->tx_vec_allowed = true;
 387
 388         if (dev->data->dev_conf.rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG)
 389                 dev->data->dev_conf.rxmode.offloads |= DEV_RX_OFFLOAD_RSS_HASH;
 390
 391         /* Large VF setting */
 392         if (num_queue_pairs > IAVF_MAX_NUM_QUEUES_DFLT) {
 393                 if (!(vf->vf_res->vf_cap_flags &
 394                                 VIRTCHNL_VF_LARGE_NUM_QPAIRS)) {
 395                         PMD_DRV_LOG(ERR, "large VF is not supported");
 396                         return -1;
 397                 }
 398
 399                 if (num_queue_pairs > IAVF_MAX_NUM_QUEUES_LV) {
 400                         PMD_DRV_LOG(ERR, "queue pairs number cannot be larger than %u",
 401                                 IAVF_MAX_NUM_QUEUES_LV);
 402                         return -1;
 403                 }
 404
 405                 ret = iavf_queues_req_reset(dev, num_queue_pairs);
 406                 if (ret)
 407                         return ret;
 408
 409                 ret = iavf_get_max_rss_queue_region(ad);
 410                 if (ret) {
 411                         PMD_INIT_LOG(ERR, "get max rss queue region failed");
 412                         return ret;
 413                 }
 414
 415                 vf->lv_enabled = true;
 416         } else {
 417                 /* Check if large VF is already enabled. If so, disable and
 418                  * release redundant queue resource.
 419                  * Or check if enough queue pairs. If not, request them from PF.
 420                  */
 421                 if (vf->lv_enabled ||
 422                     num_queue_pairs > vf->vsi_res->num_queue_pairs) {
 423                         ret = iavf_queues_req_reset(dev, num_queue_pairs);
 424                         if (ret)
 425                                 return ret;
 426
 427                         vf->lv_enabled = false;
 428                 }
 429                 /* if large VF is not required, use default rss queue region */
 430                 vf->max_rss_qregion = IAVF_MAX_NUM_QUEUES_DFLT;
 431         }
 432
 433         ret = iavf_dev_init_vlan(dev);
 434         if (ret)
 435                 PMD_DRV_LOG(ERR, "configure VLAN failed: %d", ret);
 436
 437         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF) {
 438                 if (iavf_init_rss(ad) != 0) {
 439                         PMD_DRV_LOG(ERR, "configure rss failed");
 440                         return -1;
 441                 }
 442         }
 443         return 0;
 444 }
 445
 446 static int
 447 iavf_init_rxq(struct rte_eth_dev *dev, struct iavf_rx_queue *rxq)
 448 {
 449         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(dev->data->dev_private);
 450         struct rte_eth_dev_data *dev_data = dev->data;
 451         uint16_t buf_size, max_pkt_len, len;
 452
 453         buf_size = rte_pktmbuf_data_room_size(rxq->mp) - RTE_PKTMBUF_HEADROOM;
 454
 455         /* Calculate the maximum packet length allowed */
 456         len = rxq->rx_buf_len * IAVF_MAX_CHAINED_RX_BUFFERS;
 457         max_pkt_len = RTE_MIN(len, dev->data->dev_conf.rxmode.max_rx_pkt_len);
 458
 459         /* Check if the jumbo frame and maximum packet length are set
 460          * correctly.
 461          */
 462         if (dev->data->dev_conf.rxmode.offloads & DEV_RX_OFFLOAD_JUMBO_FRAME) {
 463                 if (max_pkt_len <= IAVF_ETH_MAX_LEN ||
 464                     max_pkt_len > IAVF_FRAME_SIZE_MAX) {
 465                         PMD_DRV_LOG(ERR, "maximum packet length must be "
 466                                     "larger than %u and smaller than %u, "
 467                                     "as jumbo frame is enabled",
 468                                     (uint32_t)IAVF_ETH_MAX_LEN,
 469                                     (uint32_t)IAVF_FRAME_SIZE_MAX);
 470                         return -EINVAL;
 471                 }
 472         } else {
 473                 if (max_pkt_len < RTE_ETHER_MIN_LEN ||
 474                     max_pkt_len > IAVF_ETH_MAX_LEN) {
 475                         PMD_DRV_LOG(ERR, "maximum packet length must be "
 476                                     "larger than %u and smaller than %u, "
 477                                     "as jumbo frame is disabled",
 478                                     (uint32_t)RTE_ETHER_MIN_LEN,
 479                                     (uint32_t)IAVF_ETH_MAX_LEN);
 480                         return -EINVAL;
 481                 }
 482         }
 483
 484         rxq->max_pkt_len = max_pkt_len;
 485         if ((dev_data->dev_conf.rxmode.offloads & DEV_RX_OFFLOAD_SCATTER) ||
 486             rxq->max_pkt_len > buf_size) {
 487                 dev_data->scattered_rx = 1;
 488         }
 489         IAVF_PCI_REG_WRITE(rxq->qrx_tail, rxq->nb_rx_desc - 1);
 490         IAVF_WRITE_FLUSH(hw);
 491
 492         return 0;
 493 }
 494
 495 static int
 496 iavf_init_queues(struct rte_eth_dev *dev)
 497 {
 498         struct iavf_rx_queue **rxq =
 499                 (struct iavf_rx_queue **)dev->data->rx_queues;
 500         int i, ret = IAVF_SUCCESS;
 501
 502         for (i = 0; i < dev->data->nb_rx_queues; i++) {
 503                 if (!rxq[i] || !rxq[i]->q_set)
 504                         continue;
 505                 ret = iavf_init_rxq(dev, rxq[i]);
 506                 if (ret != IAVF_SUCCESS)
 507                         break;
 508         }
 509         /* set rx/tx function to vector/scatter/single-segment
 510          * according to parameters
 511          */
 512         iavf_set_rx_function(dev);
 513         iavf_set_tx_function(dev);
 514
 515         return ret;
 516 }
 517
 518 static int iavf_config_rx_queues_irqs(struct rte_eth_dev *dev,
 519                                      struct rte_intr_handle *intr_handle)
 520 {
 521         struct iavf_adapter *adapter =
 522                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 523         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
 524         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(adapter);
 525         struct iavf_qv_map *qv_map;
 526         uint16_t interval, i;
 527         int vec;
 528
 529         if (rte_intr_cap_multiple(intr_handle) &&
 530             dev->data->dev_conf.intr_conf.rxq) {
 531                 if (rte_intr_efd_enable(intr_handle, dev->data->nb_rx_queues))
 532                         return -1;
 533         }
 534
 535         if (rte_intr_dp_is_en(intr_handle) && !intr_handle->intr_vec) {
 536                 intr_handle->intr_vec =
 537                         rte_zmalloc("intr_vec",
 538                                     dev->data->nb_rx_queues * sizeof(int), 0);
 539                 if (!intr_handle->intr_vec) {
 540                         PMD_DRV_LOG(ERR, "Failed to allocate %d rx intr_vec",
 541                                     dev->data->nb_rx_queues);
 542                         return -1;
 543                 }
 544         }
 545
 546         qv_map = rte_zmalloc("qv_map",
 547                 dev->data->nb_rx_queues * sizeof(struct iavf_qv_map), 0);
 548         if (!qv_map) {
 549                 PMD_DRV_LOG(ERR, "Failed to allocate %d queue-vector map",
 550                                 dev->data->nb_rx_queues);
 551                 return -1;
 552         }
 553
 554         if (!dev->data->dev_conf.intr_conf.rxq ||
 555             !rte_intr_dp_is_en(intr_handle)) {
 556                 /* Rx interrupt disabled, Map interrupt only for writeback */
 557                 vf->nb_msix = 1;
 558                 if (vf->vf_res->vf_cap_flags &
 559                     VIRTCHNL_VF_OFFLOAD_WB_ON_ITR) {
 560                         /* If WB_ON_ITR supports, enable it */
 561                         vf->msix_base = IAVF_RX_VEC_START;
 562                         /* Set the ITR for index zero, to 2us to make sure that
 563                          * we leave time for aggregation to occur, but don't
 564                          * increase latency dramatically.
 565                          */
 566                         IAVF_WRITE_REG(hw,
 567                                        IAVF_VFINT_DYN_CTLN1(vf->msix_base - 1),
 568                                        (0 << IAVF_VFINT_DYN_CTLN1_ITR_INDX_SHIFT) |
 569                                        IAVF_VFINT_DYN_CTLN1_WB_ON_ITR_MASK |
 570                                        (2UL << IAVF_VFINT_DYN_CTLN1_INTERVAL_SHIFT));
 571                         /* debug - check for success! the return value
 572                          * should be 2, offset is 0x2800
 573                          */
 574                         /* IAVF_READ_REG(hw, IAVF_VFINT_ITRN1(0, 0)); */
 575                 } else {
 576                         /* If no WB_ON_ITR offload flags, need to set
 577                          * interrupt for descriptor write back.
 578                          */
 579                         vf->msix_base = IAVF_MISC_VEC_ID;
 580
 581                         /* set ITR to max */
 582                         interval = iavf_calc_itr_interval(
 583                                         IAVF_QUEUE_ITR_INTERVAL_MAX);
 584                         IAVF_WRITE_REG(hw, IAVF_VFINT_DYN_CTL01,
 585                                        IAVF_VFINT_DYN_CTL01_INTENA_MASK |
 586                                        (IAVF_ITR_INDEX_DEFAULT <<
 587                                         IAVF_VFINT_DYN_CTL01_ITR_INDX_SHIFT) |
 588                                        (interval <<
 589                                         IAVF_VFINT_DYN_CTL01_INTERVAL_SHIFT));
 590                 }
 591                 IAVF_WRITE_FLUSH(hw);
 592                 /* map all queues to the same interrupt */
 593                 for (i = 0; i < dev->data->nb_rx_queues; i++) {
 594                         qv_map[i].queue_id = i;
 595                         qv_map[i].vector_id = vf->msix_base;
 596                 }
 597                 vf->qv_map = qv_map;
 598         } else {
 599                 if (!rte_intr_allow_others(intr_handle)) {
 600                         vf->nb_msix = 1;
 601                         vf->msix_base = IAVF_MISC_VEC_ID;
 602                         for (i = 0; i < dev->data->nb_rx_queues; i++) {
 603                                 qv_map[i].queue_id = i;
 604                                 qv_map[i].vector_id = vf->msix_base;
 605                                 intr_handle->intr_vec[i] = IAVF_MISC_VEC_ID;
 606                         }
 607                         vf->qv_map = qv_map;
 608                         PMD_DRV_LOG(DEBUG,
 609                                     "vector %u are mapping to all Rx queues",
 610                                     vf->msix_base);
 611                 } else {
 612                         /* If Rx interrupt is reuquired, and we can use
 613                          * multi interrupts, then the vec is from 1
 614                          */
 615                         vf->nb_msix = RTE_MIN(intr_handle->nb_efd,
 616                                  (uint16_t)(vf->vf_res->max_vectors - 1));
 617                         vf->msix_base = IAVF_RX_VEC_START;
 618                         vec = IAVF_RX_VEC_START;
 619                         for (i = 0; i < dev->data->nb_rx_queues; i++) {
 620                                 qv_map[i].queue_id = i;
 621                                 qv_map[i].vector_id = vec;
 622                                 intr_handle->intr_vec[i] = vec++;
 623                                 if (vec >= vf->nb_msix + IAVF_RX_VEC_START)
 624                                         vec = IAVF_RX_VEC_START;
 625                         }
 626                         vf->qv_map = qv_map;
 627                         PMD_DRV_LOG(DEBUG,
 628                                     "%u vectors are mapping to %u Rx queues",
 629                                     vf->nb_msix, dev->data->nb_rx_queues);
 630                 }
 631         }
 632
 633         if (!vf->lv_enabled) {
 634                 if (iavf_config_irq_map(adapter)) {
 635                         PMD_DRV_LOG(ERR, "config interrupt mapping failed");
 636                         return -1;
 637                 }
 638         } else {
 639                 uint16_t num_qv_maps = dev->data->nb_rx_queues;
 640                 uint16_t index = 0;
 641
 642                 while (num_qv_maps > IAVF_IRQ_MAP_NUM_PER_BUF) {
 643                         if (iavf_config_irq_map_lv(adapter,
 644                                         IAVF_IRQ_MAP_NUM_PER_BUF, index)) {
 645                                 PMD_DRV_LOG(ERR, "config interrupt mapping for large VF failed");
 646                                 return -1;
 647                         }
 648                         num_qv_maps -= IAVF_IRQ_MAP_NUM_PER_BUF;
 649                         index += IAVF_IRQ_MAP_NUM_PER_BUF;
 650                 }
 651
 652                 if (iavf_config_irq_map_lv(adapter, num_qv_maps, index)) {
 653                         PMD_DRV_LOG(ERR, "config interrupt mapping for large VF failed");
 654                         return -1;
 655                 }
 656         }
 657         return 0;
 658 }
 659
 660 static int
 661 iavf_start_queues(struct rte_eth_dev *dev)
 662 {
 663         struct iavf_rx_queue *rxq;
 664         struct iavf_tx_queue *txq;
 665         int i;
 666
 667         for (i = 0; i < dev->data->nb_tx_queues; i++) {
 668                 txq = dev->data->tx_queues[i];
 669                 if (txq->tx_deferred_start)
 670                         continue;
 671                 if (iavf_dev_tx_queue_start(dev, i) != 0) {
 672                         PMD_DRV_LOG(ERR, "Fail to start queue %u", i);
 673                         return -1;
 674                 }
 675         }
 676
 677         for (i = 0; i < dev->data->nb_rx_queues; i++) {
 678                 rxq = dev->data->rx_queues[i];
 679                 if (rxq->rx_deferred_start)
 680                         continue;
 681                 if (iavf_dev_rx_queue_start(dev, i) != 0) {
 682                         PMD_DRV_LOG(ERR, "Fail to start queue %u", i);
 683                         return -1;
 684                 }
 685         }
 686
 687         return 0;
 688 }
 689
 690 static int
 691 iavf_dev_start(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(dev->data->dev_private);
 696         struct rte_intr_handle *intr_handle = dev->intr_handle;
 697         uint16_t num_queue_pairs;
 698         uint16_t index = 0;
 699
 700         PMD_INIT_FUNC_TRACE();
 701
 702         adapter->stopped = 0;
 703
 704         vf->max_pkt_len = dev->data->dev_conf.rxmode.max_rx_pkt_len;
 705         vf->num_queue_pairs = RTE_MAX(dev->data->nb_rx_queues,
 706                                       dev->data->nb_tx_queues);
 707         num_queue_pairs = vf->num_queue_pairs;
 708
 709         if (iavf_init_queues(dev) != 0) {
 710                 PMD_DRV_LOG(ERR, "failed to do Queue init");
 711                 return -1;
 712         }
 713
 714         /* If needed, send configure queues msg multiple times to make the
 715          * adminq buffer length smaller than the 4K limitation.
 716          */
 717         while (num_queue_pairs > IAVF_CFG_Q_NUM_PER_BUF) {
 718                 if (iavf_configure_queues(adapter,
 719                                 IAVF_CFG_Q_NUM_PER_BUF, index) != 0) {
 720                         PMD_DRV_LOG(ERR, "configure queues failed");
 721                         goto err_queue;
 722                 }
 723                 num_queue_pairs -= IAVF_CFG_Q_NUM_PER_BUF;
 724                 index += IAVF_CFG_Q_NUM_PER_BUF;
 725         }
 726
 727         if (iavf_configure_queues(adapter, num_queue_pairs, index) != 0) {
 728                 PMD_DRV_LOG(ERR, "configure queues failed");
 729                 goto err_queue;
 730         }
 731
 732         if (iavf_config_rx_queues_irqs(dev, intr_handle) != 0) {
 733                 PMD_DRV_LOG(ERR, "configure irq failed");
 734                 goto err_queue;
 735         }
 736         /* re-enable intr again, because efd assign may change */
 737         if (dev->data->dev_conf.intr_conf.rxq != 0) {
 738                 rte_intr_disable(intr_handle);
 739                 rte_intr_enable(intr_handle);
 740         }
 741
 742         /* Set all mac addrs */
 743         iavf_add_del_all_mac_addr(adapter, true);
 744
 745         /* Set all multicast addresses */
 746         iavf_add_del_mc_addr_list(adapter, vf->mc_addrs, vf->mc_addrs_num,
 747                                   true);
 748
 749         if (iavf_start_queues(dev) != 0) {
 750                 PMD_DRV_LOG(ERR, "enable queues failed");
 751                 goto err_mac;
 752         }
 753
 754         return 0;
 755
 756 err_mac:
 757         iavf_add_del_all_mac_addr(adapter, false);
 758 err_queue:
 759         return -1;
 760 }
 761
 762 static int
 763 iavf_dev_stop(struct rte_eth_dev *dev)
 764 {
 765         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
 766         struct iavf_adapter *adapter =
 767                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 768         struct rte_intr_handle *intr_handle = dev->intr_handle;
 769
 770         PMD_INIT_FUNC_TRACE();
 771
 772         if (adapter->stopped == 1)
 773                 return 0;
 774
 775         iavf_stop_queues(dev);
 776
 777         /* Disable the interrupt for Rx */
 778         rte_intr_efd_disable(intr_handle);
 779         /* Rx interrupt vector mapping free */
 780         if (intr_handle->intr_vec) {
 781                 rte_free(intr_handle->intr_vec);
 782                 intr_handle->intr_vec = NULL;
 783         }
 784
 785         /* remove all mac addrs */
 786         iavf_add_del_all_mac_addr(adapter, false);
 787
 788         /* remove all multicast addresses */
 789         iavf_add_del_mc_addr_list(adapter, vf->mc_addrs, vf->mc_addrs_num,
 790                                   false);
 791
 792         adapter->stopped = 1;
 793         dev->data->dev_started = 0;
 794
 795         return 0;
 796 }
 797
 798 static int
 799 iavf_dev_info_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
 800 {
 801         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
 802
 803         dev_info->max_rx_queues = IAVF_MAX_NUM_QUEUES_LV;
 804         dev_info->max_tx_queues = IAVF_MAX_NUM_QUEUES_LV;
 805         dev_info->min_rx_bufsize = IAVF_BUF_SIZE_MIN;
 806         dev_info->max_rx_pktlen = IAVF_FRAME_SIZE_MAX;
 807         dev_info->max_mtu = dev_info->max_rx_pktlen - IAVF_ETH_OVERHEAD;
 808         dev_info->min_mtu = RTE_ETHER_MIN_MTU;
 809         dev_info->hash_key_size = vf->vf_res->rss_key_size;
 810         dev_info->reta_size = vf->vf_res->rss_lut_size;
 811         dev_info->flow_type_rss_offloads = IAVF_RSS_OFFLOAD_ALL;
 812         dev_info->max_mac_addrs = IAVF_NUM_MACADDR_MAX;
 813         dev_info->rx_offload_capa =
 814                 DEV_RX_OFFLOAD_VLAN_STRIP |
 815                 DEV_RX_OFFLOAD_QINQ_STRIP |
 816                 DEV_RX_OFFLOAD_IPV4_CKSUM |
 817                 DEV_RX_OFFLOAD_UDP_CKSUM |
 818                 DEV_RX_OFFLOAD_TCP_CKSUM |
 819                 DEV_RX_OFFLOAD_OUTER_IPV4_CKSUM |
 820                 DEV_RX_OFFLOAD_SCATTER |
 821                 DEV_RX_OFFLOAD_JUMBO_FRAME |
 822                 DEV_RX_OFFLOAD_VLAN_FILTER |
 823                 DEV_RX_OFFLOAD_RSS_HASH;
 824
 825         dev_info->tx_offload_capa =
 826                 DEV_TX_OFFLOAD_VLAN_INSERT |
 827                 DEV_TX_OFFLOAD_QINQ_INSERT |
 828                 DEV_TX_OFFLOAD_IPV4_CKSUM |
 829                 DEV_TX_OFFLOAD_UDP_CKSUM |
 830                 DEV_TX_OFFLOAD_TCP_CKSUM |
 831                 DEV_TX_OFFLOAD_SCTP_CKSUM |
 832                 DEV_TX_OFFLOAD_OUTER_IPV4_CKSUM |
 833                 DEV_TX_OFFLOAD_TCP_TSO |
 834                 DEV_TX_OFFLOAD_VXLAN_TNL_TSO |
 835                 DEV_TX_OFFLOAD_GRE_TNL_TSO |
 836                 DEV_TX_OFFLOAD_IPIP_TNL_TSO |
 837                 DEV_TX_OFFLOAD_GENEVE_TNL_TSO |
 838                 DEV_TX_OFFLOAD_MULTI_SEGS |
 839                 DEV_TX_OFFLOAD_MBUF_FAST_FREE;
 840
 841         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_CRC)
 842                 dev_info->rx_offload_capa |= DEV_RX_OFFLOAD_KEEP_CRC;
 843
 844         dev_info->default_rxconf = (struct rte_eth_rxconf) {
 845                 .rx_free_thresh = IAVF_DEFAULT_RX_FREE_THRESH,
 846                 .rx_drop_en = 0,
 847                 .offloads = 0,
 848         };
 849
 850         dev_info->default_txconf = (struct rte_eth_txconf) {
 851                 .tx_free_thresh = IAVF_DEFAULT_TX_FREE_THRESH,
 852                 .tx_rs_thresh = IAVF_DEFAULT_TX_RS_THRESH,
 853                 .offloads = 0,
 854         };
 855
 856         dev_info->rx_desc_lim = (struct rte_eth_desc_lim) {
 857                 .nb_max = IAVF_MAX_RING_DESC,
 858                 .nb_min = IAVF_MIN_RING_DESC,
 859                 .nb_align = IAVF_ALIGN_RING_DESC,
 860         };
 861
 862         dev_info->tx_desc_lim = (struct rte_eth_desc_lim) {
 863                 .nb_max = IAVF_MAX_RING_DESC,
 864                 .nb_min = IAVF_MIN_RING_DESC,
 865                 .nb_align = IAVF_ALIGN_RING_DESC,
 866         };
 867
 868         return 0;
 869 }
 870
 871 static const uint32_t *
 872 iavf_dev_supported_ptypes_get(struct rte_eth_dev *dev __rte_unused)
 873 {
 874         static const uint32_t ptypes[] = {
 875                 RTE_PTYPE_L2_ETHER,
 876                 RTE_PTYPE_L3_IPV4_EXT_UNKNOWN,
 877                 RTE_PTYPE_L4_FRAG,
 878                 RTE_PTYPE_L4_ICMP,
 879                 RTE_PTYPE_L4_NONFRAG,
 880                 RTE_PTYPE_L4_SCTP,
 881                 RTE_PTYPE_L4_TCP,
 882                 RTE_PTYPE_L4_UDP,
 883                 RTE_PTYPE_UNKNOWN
 884         };
 885         return ptypes;
 886 }
 887
 888 int
 889 iavf_dev_link_update(struct rte_eth_dev *dev,
 890                     __rte_unused int wait_to_complete)
 891 {
 892         struct rte_eth_link new_link;
 893         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
 894
 895         memset(&new_link, 0, sizeof(new_link));
 896
 897         /* Only read status info stored in VF, and the info is updated
 898          *  when receive LINK_CHANGE evnet from PF by Virtchnnl.
 899          */
 900         switch (vf->link_speed) {
 901         case 10:
 902                 new_link.link_speed = ETH_SPEED_NUM_10M;
 903                 break;
 904         case 100:
 905                 new_link.link_speed = ETH_SPEED_NUM_100M;
 906                 break;
 907         case 1000:
 908                 new_link.link_speed = ETH_SPEED_NUM_1G;
 909                 break;
 910         case 10000:
 911                 new_link.link_speed = ETH_SPEED_NUM_10G;
 912                 break;
 913         case 20000:
 914                 new_link.link_speed = ETH_SPEED_NUM_20G;
 915                 break;
 916         case 25000:
 917                 new_link.link_speed = ETH_SPEED_NUM_25G;
 918                 break;
 919         case 40000:
 920                 new_link.link_speed = ETH_SPEED_NUM_40G;
 921                 break;
 922         case 50000:
 923                 new_link.link_speed = ETH_SPEED_NUM_50G;
 924                 break;
 925         case 100000:
 926                 new_link.link_speed = ETH_SPEED_NUM_100G;
 927                 break;
 928         default:
 929                 new_link.link_speed = ETH_SPEED_NUM_NONE;
 930                 break;
 931         }
 932
 933         new_link.link_duplex = ETH_LINK_FULL_DUPLEX;
 934         new_link.link_status = vf->link_up ? ETH_LINK_UP :
 935                                              ETH_LINK_DOWN;
 936         new_link.link_autoneg = !(dev->data->dev_conf.link_speeds &
 937                                 ETH_LINK_SPEED_FIXED);
 938
 939         return rte_eth_linkstatus_set(dev, &new_link);
 940 }
 941
 942 static int
 943 iavf_dev_promiscuous_enable(struct rte_eth_dev *dev)
 944 {
 945         struct iavf_adapter *adapter =
 946                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 947         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
 948
 949         return iavf_config_promisc(adapter,
 950                                   true, vf->promisc_multicast_enabled);
 951 }
 952
 953 static int
 954 iavf_dev_promiscuous_disable(struct rte_eth_dev *dev)
 955 {
 956         struct iavf_adapter *adapter =
 957                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 958         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
 959
 960         return iavf_config_promisc(adapter,
 961                                   false, vf->promisc_multicast_enabled);
 962 }
 963
 964 static int
 965 iavf_dev_allmulticast_enable(struct rte_eth_dev *dev)
 966 {
 967         struct iavf_adapter *adapter =
 968                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 969         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
 970
 971         return iavf_config_promisc(adapter,
 972                                   vf->promisc_unicast_enabled, true);
 973 }
 974
 975 static int
 976 iavf_dev_allmulticast_disable(struct rte_eth_dev *dev)
 977 {
 978         struct iavf_adapter *adapter =
 979                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 980         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
 981
 982         return iavf_config_promisc(adapter,
 983                                   vf->promisc_unicast_enabled, false);
 984 }
 985
 986 static int
 987 iavf_dev_add_mac_addr(struct rte_eth_dev *dev, struct rte_ether_addr *addr,
 988                      __rte_unused uint32_t index,
 989                      __rte_unused uint32_t pool)
 990 {
 991         struct iavf_adapter *adapter =
 992                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 993         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
 994         int err;
 995
 996         if (rte_is_zero_ether_addr(addr)) {
 997                 PMD_DRV_LOG(ERR, "Invalid Ethernet Address");
 998                 return -EINVAL;
 999         }
1000
1001         err = iavf_add_del_eth_addr(adapter, addr, true);
1002         if (err) {
1003                 PMD_DRV_LOG(ERR, "fail to add MAC address");
1004                 return -EIO;
1005         }
1006
1007         vf->mac_num++;
1008
1009         return 0;
1010 }
1011
1012 static void
1013 iavf_dev_del_mac_addr(struct rte_eth_dev *dev, uint32_t index)
1014 {
1015         struct iavf_adapter *adapter =
1016                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1017         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
1018         struct rte_ether_addr *addr;
1019         int err;
1020
1021         addr = &dev->data->mac_addrs[index];
1022
1023         err = iavf_add_del_eth_addr(adapter, addr, false);
1024         if (err)
1025                 PMD_DRV_LOG(ERR, "fail to delete MAC address");
1026
1027         vf->mac_num--;
1028 }
1029
1030 static int
1031 iavf_dev_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
1032 {
1033         struct iavf_adapter *adapter =
1034                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1035         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
1036         int err;
1037
1038         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_VLAN_V2) {
1039                 err = iavf_add_del_vlan_v2(adapter, vlan_id, on);
1040                 if (err)
1041                         return -EIO;
1042                 return 0;
1043         }
1044
1045         if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_VLAN))
1046                 return -ENOTSUP;
1047
1048         err = iavf_add_del_vlan(adapter, vlan_id, on);
1049         if (err)
1050                 return -EIO;
1051         return 0;
1052 }
1053
1054 static void
1055 iavf_iterate_vlan_filters_v2(struct rte_eth_dev *dev, bool enable)
1056 {
1057         struct rte_vlan_filter_conf *vfc = &dev->data->vlan_filter_conf;
1058         struct iavf_adapter *adapter =
1059                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1060         uint32_t i, j;
1061         uint64_t ids;
1062
1063         for (i = 0; i < RTE_DIM(vfc->ids); i++) {
1064                 if (vfc->ids[i] == 0)
1065                         continue;
1066
1067                 ids = vfc->ids[i];
1068                 for (j = 0; ids != 0 && j < 64; j++, ids >>= 1) {
1069                         if (ids & 1)
1070                                 iavf_add_del_vlan_v2(adapter,
1071                                                      64 * i + j, enable);
1072                 }
1073         }
1074 }
1075
1076 static int
1077 iavf_dev_vlan_offload_set_v2(struct rte_eth_dev *dev, int mask)
1078 {
1079         struct rte_eth_rxmode *rxmode = &dev->data->dev_conf.rxmode;
1080         struct iavf_adapter *adapter =
1081                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1082         bool enable;
1083         int err;
1084
1085         if (mask & ETH_VLAN_FILTER_MASK) {
1086                 enable = !!(rxmode->offloads & DEV_RX_OFFLOAD_VLAN_FILTER);
1087
1088                 iavf_iterate_vlan_filters_v2(dev, enable);
1089         }
1090
1091         if (mask & ETH_VLAN_STRIP_MASK) {
1092                 enable = !!(rxmode->offloads & DEV_RX_OFFLOAD_VLAN_STRIP);
1093
1094                 err = iavf_config_vlan_strip_v2(adapter, enable);
1095                 /* If not support, the stripping is already disabled by PF */
1096                 if (err == -ENOTSUP && !enable)
1097                         err = 0;
1098                 if (err)
1099                         return -EIO;
1100         }
1101
1102         return 0;
1103 }
1104
1105 static int
1106 iavf_dev_vlan_offload_set(struct rte_eth_dev *dev, int mask)
1107 {
1108         struct iavf_adapter *adapter =
1109                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1110         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
1111         struct rte_eth_conf *dev_conf = &dev->data->dev_conf;
1112         int err;
1113
1114         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_VLAN_V2)
1115                 return iavf_dev_vlan_offload_set_v2(dev, mask);
1116
1117         if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_VLAN))
1118                 return -ENOTSUP;
1119
1120         /* Vlan stripping setting */
1121         if (mask & ETH_VLAN_STRIP_MASK) {
1122                 /* Enable or disable VLAN stripping */
1123                 if (dev_conf->rxmode.offloads & DEV_RX_OFFLOAD_VLAN_STRIP)
1124                         err = iavf_enable_vlan_strip(adapter);
1125                 else
1126                         err = iavf_disable_vlan_strip(adapter);
1127
1128                 if (err)
1129                         return -EIO;
1130         }
1131         return 0;
1132 }
1133
1134 static int
1135 iavf_dev_rss_reta_update(struct rte_eth_dev *dev,
1136                         struct rte_eth_rss_reta_entry64 *reta_conf,
1137                         uint16_t reta_size)
1138 {
1139         struct iavf_adapter *adapter =
1140                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1141         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
1142         uint8_t *lut;
1143         uint16_t i, idx, shift;
1144         int ret;
1145
1146         if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF))
1147                 return -ENOTSUP;
1148
1149         if (reta_size != vf->vf_res->rss_lut_size) {
1150                 PMD_DRV_LOG(ERR, "The size of hash lookup table configured "
1151                         "(%d) doesn't match the number of hardware can "
1152                         "support (%d)", reta_size, vf->vf_res->rss_lut_size);
1153                 return -EINVAL;
1154         }
1155
1156         lut = rte_zmalloc("rss_lut", reta_size, 0);
1157         if (!lut) {
1158                 PMD_DRV_LOG(ERR, "No memory can be allocated");
1159                 return -ENOMEM;
1160         }
1161         /* store the old lut table temporarily */
1162         rte_memcpy(lut, vf->rss_lut, reta_size);
1163
1164         for (i = 0; i < reta_size; i++) {
1165                 idx = i / RTE_RETA_GROUP_SIZE;
1166                 shift = i % RTE_RETA_GROUP_SIZE;
1167                 if (reta_conf[idx].mask & (1ULL << shift))
1168                         lut[i] = reta_conf[idx].reta[shift];
1169         }
1170
1171         rte_memcpy(vf->rss_lut, lut, reta_size);
1172         /* send virtchnnl ops to configure rss*/
1173         ret = iavf_configure_rss_lut(adapter);
1174         if (ret) /* revert back */
1175                 rte_memcpy(vf->rss_lut, lut, reta_size);
1176         rte_free(lut);
1177
1178         return ret;
1179 }
1180
1181 static int
1182 iavf_dev_rss_reta_query(struct rte_eth_dev *dev,
1183                        struct rte_eth_rss_reta_entry64 *reta_conf,
1184                        uint16_t reta_size)
1185 {
1186         struct iavf_adapter *adapter =
1187                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1188         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
1189         uint16_t i, idx, shift;
1190
1191         if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF))
1192                 return -ENOTSUP;
1193
1194         if (reta_size != vf->vf_res->rss_lut_size) {
1195                 PMD_DRV_LOG(ERR, "The size of hash lookup table configured "
1196                         "(%d) doesn't match the number of hardware can "
1197                         "support (%d)", reta_size, vf->vf_res->rss_lut_size);
1198                 return -EINVAL;
1199         }
1200
1201         for (i = 0; i < reta_size; i++) {
1202                 idx = i / RTE_RETA_GROUP_SIZE;
1203                 shift = i % RTE_RETA_GROUP_SIZE;
1204                 if (reta_conf[idx].mask & (1ULL << shift))
1205                         reta_conf[idx].reta[shift] = vf->rss_lut[i];
1206         }
1207
1208         return 0;
1209 }
1210
1211 static int
1212 iavf_set_rss_key(struct iavf_adapter *adapter, uint8_t *key, uint8_t key_len)
1213 {
1214         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
1215
1216         /* HENA setting, it is enabled by default, no change */
1217         if (!key || key_len == 0) {
1218                 PMD_DRV_LOG(DEBUG, "No key to be configured");
1219                 return 0;
1220         } else if (key_len != vf->vf_res->rss_key_size) {
1221                 PMD_DRV_LOG(ERR, "The size of hash key configured "
1222                         "(%d) doesn't match the size of hardware can "
1223                         "support (%d)", key_len,
1224                         vf->vf_res->rss_key_size);
1225                 return -EINVAL;
1226         }
1227
1228         rte_memcpy(vf->rss_key, key, key_len);
1229
1230         return iavf_configure_rss_key(adapter);
1231 }
1232
1233 static int
1234 iavf_dev_rss_hash_update(struct rte_eth_dev *dev,
1235                         struct rte_eth_rss_conf *rss_conf)
1236 {
1237         struct iavf_adapter *adapter =
1238                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1239         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
1240         int ret;
1241
1242         adapter->eth_dev->data->dev_conf.rx_adv_conf.rss_conf = *rss_conf;
1243
1244         if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF))
1245                 return -ENOTSUP;
1246
1247         /* Set hash key. */
1248         ret = iavf_set_rss_key(adapter, rss_conf->rss_key,
1249                                rss_conf->rss_key_len);
1250         if (ret)
1251                 return ret;
1252
1253         if (rss_conf->rss_hf == 0)
1254                 return 0;
1255
1256         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_ADV_RSS_PF) {
1257                 /* Clear existing RSS. */
1258                 ret = iavf_set_hena(adapter, 0);
1259
1260                 /* It is a workaround, temporarily allow error to be returned
1261                  * due to possible lack of PF handling for hena = 0.
1262                  */
1263                 if (ret)
1264                         PMD_DRV_LOG(WARNING, "fail to clean existing RSS,"
1265                                     "lack PF support");
1266
1267                 /* Set new RSS configuration. */
1268                 ret = iavf_rss_hash_set(adapter, rss_conf->rss_hf, true);
1269                 if (ret) {
1270                         PMD_DRV_LOG(ERR, "fail to set new RSS");
1271                         return ret;
1272                 }
1273         }
1274
1275         return 0;
1276 }
1277
1278 static int
1279 iavf_dev_rss_hash_conf_get(struct rte_eth_dev *dev,
1280                           struct rte_eth_rss_conf *rss_conf)
1281 {
1282         struct iavf_adapter *adapter =
1283                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1284         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
1285
1286         if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF))
1287                 return -ENOTSUP;
1288
1289         rss_conf->rss_hf = vf->rss_hf;
1290
1291         if (!rss_conf->rss_key)
1292                 return 0;
1293
1294         rss_conf->rss_key_len = vf->vf_res->rss_key_size;
1295         rte_memcpy(rss_conf->rss_key, vf->rss_key, rss_conf->rss_key_len);
1296
1297         return 0;
1298 }
1299
1300 static int
1301 iavf_dev_mtu_set(struct rte_eth_dev *dev, uint16_t mtu)
1302 {
1303         uint32_t frame_size = mtu + IAVF_ETH_OVERHEAD;
1304         int ret = 0;
1305
1306         if (mtu < RTE_ETHER_MIN_MTU || frame_size > IAVF_FRAME_SIZE_MAX)
1307                 return -EINVAL;
1308
1309         /* mtu setting is forbidden if port is start */
1310         if (dev->data->dev_started) {
1311                 PMD_DRV_LOG(ERR, "port must be stopped before configuration");
1312                 return -EBUSY;
1313         }
1314
1315         if (frame_size > IAVF_ETH_MAX_LEN)
1316                 dev->data->dev_conf.rxmode.offloads |=
1317                                 DEV_RX_OFFLOAD_JUMBO_FRAME;
1318         else
1319                 dev->data->dev_conf.rxmode.offloads &=
1320                                 ~DEV_RX_OFFLOAD_JUMBO_FRAME;
1321
1322         dev->data->dev_conf.rxmode.max_rx_pkt_len = frame_size;
1323
1324         return ret;
1325 }
1326
1327 static int
1328 iavf_dev_set_default_mac_addr(struct rte_eth_dev *dev,
1329                              struct rte_ether_addr *mac_addr)
1330 {
1331         struct iavf_adapter *adapter =
1332                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1333         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(adapter);
1334         struct rte_ether_addr *perm_addr, *old_addr;
1335         int ret;
1336
1337         old_addr = (struct rte_ether_addr *)hw->mac.addr;
1338         perm_addr = (struct rte_ether_addr *)hw->mac.perm_addr;
1339
1340         /* If the MAC address is configured by host, skip the setting */
1341         if (rte_is_valid_assigned_ether_addr(perm_addr))
1342                 return -EPERM;
1343
1344         ret = iavf_add_del_eth_addr(adapter, old_addr, false);
1345         if (ret)
1346                 PMD_DRV_LOG(ERR, "Fail to delete old MAC:"
1347                             " %02X:%02X:%02X:%02X:%02X:%02X",
1348                             old_addr->addr_bytes[0],
1349                             old_addr->addr_bytes[1],
1350                             old_addr->addr_bytes[2],
1351                             old_addr->addr_bytes[3],
1352                             old_addr->addr_bytes[4],
1353                             old_addr->addr_bytes[5]);
1354
1355         ret = iavf_add_del_eth_addr(adapter, mac_addr, true);
1356         if (ret)
1357                 PMD_DRV_LOG(ERR, "Fail to add new MAC:"
1358                             " %02X:%02X:%02X:%02X:%02X:%02X",
1359                             mac_addr->addr_bytes[0],
1360                             mac_addr->addr_bytes[1],
1361                             mac_addr->addr_bytes[2],
1362                             mac_addr->addr_bytes[3],
1363                             mac_addr->addr_bytes[4],
1364                             mac_addr->addr_bytes[5]);
1365
1366         if (ret)
1367                 return -EIO;
1368
1369         rte_ether_addr_copy(mac_addr, (struct rte_ether_addr *)hw->mac.addr);
1370         return 0;
1371 }
1372
1373 static void
1374 iavf_stat_update_48(uint64_t *offset, uint64_t *stat)
1375 {
1376         if (*stat >= *offset)
1377                 *stat = *stat - *offset;
1378         else
1379                 *stat = (uint64_t)((*stat +
1380                         ((uint64_t)1 << IAVF_48_BIT_WIDTH)) - *offset);
1381
1382         *stat &= IAVF_48_BIT_MASK;
1383 }
1384
1385 static void
1386 iavf_stat_update_32(uint64_t *offset, uint64_t *stat)
1387 {
1388         if (*stat >= *offset)
1389                 *stat = (uint64_t)(*stat - *offset);
1390         else
1391                 *stat = (uint64_t)((*stat +
1392                         ((uint64_t)1 << IAVF_32_BIT_WIDTH)) - *offset);
1393 }
1394
1395 static void
1396 iavf_update_stats(struct iavf_vsi *vsi, struct virtchnl_eth_stats *nes)
1397 {
1398         struct virtchnl_eth_stats *oes = &vsi->eth_stats_offset;
1399
1400         iavf_stat_update_48(&oes->rx_bytes, &nes->rx_bytes);
1401         iavf_stat_update_48(&oes->rx_unicast, &nes->rx_unicast);
1402         iavf_stat_update_48(&oes->rx_multicast, &nes->rx_multicast);
1403         iavf_stat_update_48(&oes->rx_broadcast, &nes->rx_broadcast);
1404         iavf_stat_update_32(&oes->rx_discards, &nes->rx_discards);
1405         iavf_stat_update_48(&oes->tx_bytes, &nes->tx_bytes);
1406         iavf_stat_update_48(&oes->tx_unicast, &nes->tx_unicast);
1407         iavf_stat_update_48(&oes->tx_multicast, &nes->tx_multicast);
1408         iavf_stat_update_48(&oes->tx_broadcast, &nes->tx_broadcast);
1409         iavf_stat_update_32(&oes->tx_errors, &nes->tx_errors);
1410         iavf_stat_update_32(&oes->tx_discards, &nes->tx_discards);
1411 }
1412
1413 static int
1414 iavf_dev_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
1415 {
1416         struct iavf_adapter *adapter =
1417                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1418         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
1419         struct iavf_vsi *vsi = &vf->vsi;
1420         struct virtchnl_eth_stats *pstats = NULL;
1421         int ret;
1422
1423         ret = iavf_query_stats(adapter, &pstats);
1424         if (ret == 0) {
1425                 uint8_t crc_stats_len = (dev->data->dev_conf.rxmode.offloads &
1426                                          DEV_RX_OFFLOAD_KEEP_CRC) ? 0 :
1427                                          RTE_ETHER_CRC_LEN;
1428                 iavf_update_stats(vsi, pstats);
1429                 stats->ipackets = pstats->rx_unicast + pstats->rx_multicast +
1430                                 pstats->rx_broadcast - pstats->rx_discards;
1431                 stats->opackets = pstats->tx_broadcast + pstats->tx_multicast +
1432                                                 pstats->tx_unicast;
1433                 stats->imissed = pstats->rx_discards;
1434                 stats->oerrors = pstats->tx_errors + pstats->tx_discards;
1435                 stats->ibytes = pstats->rx_bytes;
1436                 stats->ibytes -= stats->ipackets * crc_stats_len;
1437                 stats->obytes = pstats->tx_bytes;
1438         } else {
1439                 PMD_DRV_LOG(ERR, "Get statistics failed");
1440         }
1441         return ret;
1442 }
1443
1444 static int
1445 iavf_dev_stats_reset(struct rte_eth_dev *dev)
1446 {
1447         int ret;
1448         struct iavf_adapter *adapter =
1449                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1450         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
1451         struct iavf_vsi *vsi = &vf->vsi;
1452         struct virtchnl_eth_stats *pstats = NULL;
1453
1454         /* read stat values to clear hardware registers */
1455         ret = iavf_query_stats(adapter, &pstats);
1456         if (ret != 0)
1457                 return ret;
1458
1459         /* set stats offset base on current values */
1460         vsi->eth_stats_offset = *pstats;
1461
1462         return 0;
1463 }
1464
1465 static int iavf_dev_xstats_get_names(__rte_unused struct rte_eth_dev *dev,
1466                                       struct rte_eth_xstat_name *xstats_names,
1467                                       __rte_unused unsigned int limit)
1468 {
1469         unsigned int i;
1470
1471         if (xstats_names != NULL)
1472                 for (i = 0; i < IAVF_NB_XSTATS; i++) {
1473                         snprintf(xstats_names[i].name,
1474                                 sizeof(xstats_names[i].name),
1475                                 "%s", rte_iavf_stats_strings[i].name);
1476                 }
1477         return IAVF_NB_XSTATS;
1478 }
1479
1480 static int iavf_dev_xstats_get(struct rte_eth_dev *dev,
1481                                  struct rte_eth_xstat *xstats, unsigned int n)
1482 {
1483         int ret;
1484         unsigned int i;
1485         struct iavf_adapter *adapter =
1486                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1487         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
1488         struct iavf_vsi *vsi = &vf->vsi;
1489         struct virtchnl_eth_stats *pstats = NULL;
1490
1491         if (n < IAVF_NB_XSTATS)
1492                 return IAVF_NB_XSTATS;
1493
1494         ret = iavf_query_stats(adapter, &pstats);
1495         if (ret != 0)
1496                 return 0;
1497
1498         if (!xstats)
1499                 return 0;
1500
1501         iavf_update_stats(vsi, pstats);
1502
1503         /* loop over xstats array and values from pstats */
1504         for (i = 0; i < IAVF_NB_XSTATS; i++) {
1505                 xstats[i].id = i;
1506                 xstats[i].value = *(uint64_t *)(((char *)pstats) +
1507                         rte_iavf_stats_strings[i].offset);
1508         }
1509
1510         return IAVF_NB_XSTATS;
1511 }
1512
1513
1514 static int
1515 iavf_dev_rx_queue_intr_enable(struct rte_eth_dev *dev, uint16_t queue_id)
1516 {
1517         struct iavf_adapter *adapter =
1518                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1519         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1520         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(adapter);
1521         uint16_t msix_intr;
1522
1523         msix_intr = pci_dev->intr_handle.intr_vec[queue_id];
1524         if (msix_intr == IAVF_MISC_VEC_ID) {
1525                 PMD_DRV_LOG(INFO, "MISC is also enabled for control");
1526                 IAVF_WRITE_REG(hw, IAVF_VFINT_DYN_CTL01,
1527                                IAVF_VFINT_DYN_CTL01_INTENA_MASK |
1528                                IAVF_VFINT_DYN_CTL01_CLEARPBA_MASK |
1529                                IAVF_VFINT_DYN_CTL01_ITR_INDX_MASK);
1530         } else {
1531                 IAVF_WRITE_REG(hw,
1532                                IAVF_VFINT_DYN_CTLN1
1533                                 (msix_intr - IAVF_RX_VEC_START),
1534                                IAVF_VFINT_DYN_CTLN1_INTENA_MASK |
1535                                IAVF_VFINT_DYN_CTL01_CLEARPBA_MASK |
1536                                IAVF_VFINT_DYN_CTLN1_ITR_INDX_MASK);
1537         }
1538
1539         IAVF_WRITE_FLUSH(hw);
1540
1541         rte_intr_ack(&pci_dev->intr_handle);
1542
1543         return 0;
1544 }
1545
1546 static int
1547 iavf_dev_rx_queue_intr_disable(struct rte_eth_dev *dev, uint16_t queue_id)
1548 {
1549         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1550         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1551         uint16_t msix_intr;
1552
1553         msix_intr = pci_dev->intr_handle.intr_vec[queue_id];
1554         if (msix_intr == IAVF_MISC_VEC_ID) {
1555                 PMD_DRV_LOG(ERR, "MISC is used for control, cannot disable it");
1556                 return -EIO;
1557         }
1558
1559         IAVF_WRITE_REG(hw,
1560                       IAVF_VFINT_DYN_CTLN1(msix_intr - IAVF_RX_VEC_START),
1561                       0);
1562
1563         IAVF_WRITE_FLUSH(hw);
1564         return 0;
1565 }
1566
1567 static int
1568 iavf_check_vf_reset_done(struct iavf_hw *hw)
1569 {
1570         int i, reset;
1571
1572         for (i = 0; i < IAVF_RESET_WAIT_CNT; i++) {
1573                 reset = IAVF_READ_REG(hw, IAVF_VFGEN_RSTAT) &
1574                         IAVF_VFGEN_RSTAT_VFR_STATE_MASK;
1575                 reset = reset >> IAVF_VFGEN_RSTAT_VFR_STATE_SHIFT;
1576                 if (reset == VIRTCHNL_VFR_VFACTIVE ||
1577                     reset == VIRTCHNL_VFR_COMPLETED)
1578                         break;
1579                 rte_delay_ms(20);
1580         }
1581
1582         if (i >= IAVF_RESET_WAIT_CNT)
1583                 return -1;
1584
1585         return 0;
1586 }
1587
1588 static int
1589 iavf_lookup_proto_xtr_type(const char *flex_name)
1590 {
1591         static struct {
1592                 const char *name;
1593                 enum iavf_proto_xtr_type type;
1594         } xtr_type_map[] = {
1595                 { "vlan",      IAVF_PROTO_XTR_VLAN      },
1596                 { "ipv4",      IAVF_PROTO_XTR_IPV4      },
1597                 { "ipv6",      IAVF_PROTO_XTR_IPV6      },
1598                 { "ipv6_flow", IAVF_PROTO_XTR_IPV6_FLOW },
1599                 { "tcp",       IAVF_PROTO_XTR_TCP       },
1600                 { "ip_offset", IAVF_PROTO_XTR_IP_OFFSET },
1601         };
1602         uint32_t i;
1603
1604         for (i = 0; i < RTE_DIM(xtr_type_map); i++) {
1605                 if (strcmp(flex_name, xtr_type_map[i].name) == 0)
1606                         return xtr_type_map[i].type;
1607         }
1608
1609         PMD_DRV_LOG(ERR, "wrong proto_xtr type, "
1610                     "it should be: vlan|ipv4|ipv6|ipv6_flow|tcp|ip_offset");
1611
1612         return -1;
1613 }
1614
1615 /**
1616  * Parse elem, the elem could be single number/range or '(' ')' group
1617  * 1) A single number elem, it's just a simple digit. e.g. 9
1618  * 2) A single range elem, two digits with a '-' between. e.g. 2-6
1619  * 3) A group elem, combines multiple 1) or 2) with '( )'. e.g (0,2-4,6)
1620  *    Within group elem, '-' used for a range separator;
1621  *                       ',' used for a single number.
1622  */
1623 static int
1624 iavf_parse_queue_set(const char *input, int xtr_type,
1625                      struct iavf_devargs *devargs)
1626 {
1627         const char *str = input;
1628         char *end = NULL;
1629         uint32_t min, max;
1630         uint32_t idx;
1631
1632         while (isblank(*str))
1633                 str++;
1634
1635         if (!isdigit(*str) && *str != '(')
1636                 return -1;
1637
1638         /* process single number or single range of number */
1639         if (*str != '(') {
1640                 errno = 0;
1641                 idx = strtoul(str, &end, 10);
1642                 if (errno || !end || idx >= IAVF_MAX_QUEUE_NUM)
1643                         return -1;
1644
1645                 while (isblank(*end))
1646                         end++;
1647
1648                 min = idx;
1649                 max = idx;
1650
1651                 /* process single <number>-<number> */
1652                 if (*end == '-') {
1653                         end++;
1654                         while (isblank(*end))
1655                                 end++;
1656                         if (!isdigit(*end))
1657                                 return -1;
1658
1659                         errno = 0;
1660                         idx = strtoul(end, &end, 10);
1661                         if (errno || !end || idx >= IAVF_MAX_QUEUE_NUM)
1662                                 return -1;
1663
1664                         max = idx;
1665                         while (isblank(*end))
1666                                 end++;
1667                 }
1668
1669                 if (*end != ':')
1670                         return -1;
1671
1672                 for (idx = RTE_MIN(min, max);
1673                      idx <= RTE_MAX(min, max); idx++)
1674                         devargs->proto_xtr[idx] = xtr_type;
1675
1676                 return 0;
1677         }
1678
1679         /* process set within bracket */
1680         str++;
1681         while (isblank(*str))
1682                 str++;
1683         if (*str == '\0')
1684                 return -1;
1685
1686         min = IAVF_MAX_QUEUE_NUM;
1687         do {
1688                 /* go ahead to the first digit */
1689                 while (isblank(*str))
1690                         str++;
1691                 if (!isdigit(*str))
1692                         return -1;
1693
1694                 /* get the digit value */
1695                 errno = 0;
1696                 idx = strtoul(str, &end, 10);
1697                 if (errno || !end || idx >= IAVF_MAX_QUEUE_NUM)
1698                         return -1;
1699
1700                 /* go ahead to separator '-',',' and ')' */
1701                 while (isblank(*end))
1702                         end++;
1703                 if (*end == '-') {
1704                         if (min == IAVF_MAX_QUEUE_NUM)
1705                                 min = idx;
1706                         else /* avoid continuous '-' */
1707                                 return -1;
1708                 } else if (*end == ',' || *end == ')') {
1709                         max = idx;
1710                         if (min == IAVF_MAX_QUEUE_NUM)
1711                                 min = idx;
1712
1713                         for (idx = RTE_MIN(min, max);
1714                              idx <= RTE_MAX(min, max); idx++)
1715                                 devargs->proto_xtr[idx] = xtr_type;
1716
1717                         min = IAVF_MAX_QUEUE_NUM;
1718                 } else {
1719                         return -1;
1720                 }
1721
1722                 str = end + 1;
1723         } while (*end != ')' && *end != '\0');
1724
1725         return 0;
1726 }
1727
1728 static int
1729 iavf_parse_queue_proto_xtr(const char *queues, struct iavf_devargs *devargs)
1730 {
1731         const char *queue_start;
1732         uint32_t idx;
1733         int xtr_type;
1734         char flex_name[32];
1735
1736         while (isblank(*queues))
1737                 queues++;
1738
1739         if (*queues != '[') {
1740                 xtr_type = iavf_lookup_proto_xtr_type(queues);
1741                 if (xtr_type < 0)
1742                         return -1;
1743
1744                 devargs->proto_xtr_dflt = xtr_type;
1745
1746                 return 0;
1747         }
1748
1749         queues++;
1750         do {
1751                 while (isblank(*queues))
1752                         queues++;
1753                 if (*queues == '\0')
1754                         return -1;
1755
1756                 queue_start = queues;
1757
1758                 /* go across a complete bracket */
1759                 if (*queue_start == '(') {
1760                         queues += strcspn(queues, ")");
1761                         if (*queues != ')')
1762                                 return -1;
1763                 }
1764
1765                 /* scan the separator ':' */
1766                 queues += strcspn(queues, ":");
1767                 if (*queues++ != ':')
1768                         return -1;
1769                 while (isblank(*queues))
1770                         queues++;
1771
1772                 for (idx = 0; ; idx++) {
1773                         if (isblank(queues[idx]) ||
1774                             queues[idx] == ',' ||
1775                             queues[idx] == ']' ||
1776                             queues[idx] == '\0')
1777                                 break;
1778
1779                         if (idx > sizeof(flex_name) - 2)
1780                                 return -1;
1781
1782                         flex_name[idx] = queues[idx];
1783                 }
1784                 flex_name[idx] = '\0';
1785                 xtr_type = iavf_lookup_proto_xtr_type(flex_name);
1786                 if (xtr_type < 0)
1787                         return -1;
1788
1789                 queues += idx;
1790
1791                 while (isblank(*queues) || *queues == ',' || *queues == ']')
1792                         queues++;
1793
1794                 if (iavf_parse_queue_set(queue_start, xtr_type, devargs) < 0)
1795                         return -1;
1796         } while (*queues != '\0');
1797
1798         return 0;
1799 }
1800
1801 static int
1802 iavf_handle_proto_xtr_arg(__rte_unused const char *key, const char *value,
1803                           void *extra_args)
1804 {
1805         struct iavf_devargs *devargs = extra_args;
1806
1807         if (!value || !extra_args)
1808                 return -EINVAL;
1809
1810         if (iavf_parse_queue_proto_xtr(value, devargs) < 0) {
1811                 PMD_DRV_LOG(ERR, "the proto_xtr's parameter is wrong : '%s'",
1812                             value);
1813                 return -1;
1814         }
1815
1816         return 0;
1817 }
1818
1819 static int iavf_parse_devargs(struct rte_eth_dev *dev)
1820 {
1821         struct iavf_adapter *ad =
1822                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1823         struct rte_devargs *devargs = dev->device->devargs;
1824         struct rte_kvargs *kvlist;
1825         int ret;
1826
1827         if (!devargs)
1828                 return 0;
1829
1830         kvlist = rte_kvargs_parse(devargs->args, iavf_valid_args);
1831         if (!kvlist) {
1832                 PMD_INIT_LOG(ERR, "invalid kvargs key\n");
1833                 return -EINVAL;
1834         }
1835
1836         ad->devargs.proto_xtr_dflt = IAVF_PROTO_XTR_NONE;
1837         memset(ad->devargs.proto_xtr, IAVF_PROTO_XTR_NONE,
1838                sizeof(ad->devargs.proto_xtr));
1839
1840         ret = rte_kvargs_process(kvlist, IAVF_PROTO_XTR_ARG,
1841                                  &iavf_handle_proto_xtr_arg, &ad->devargs);
1842         if (ret)
1843                 goto bail;
1844
1845 bail:
1846         rte_kvargs_free(kvlist);
1847         return ret;
1848 }
1849
1850 static void
1851 iavf_init_proto_xtr(struct rte_eth_dev *dev)
1852 {
1853         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
1854         struct iavf_adapter *ad =
1855                         IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1856         const struct iavf_proto_xtr_ol *xtr_ol;
1857         bool proto_xtr_enable = false;
1858         int offset;
1859         uint16_t i;
1860
1861         vf->proto_xtr = rte_zmalloc("vf proto xtr",
1862                                     vf->vsi_res->num_queue_pairs, 0);
1863         if (unlikely(!(vf->proto_xtr))) {
1864                 PMD_DRV_LOG(ERR, "no memory for setting up proto_xtr's table");
1865                 return;
1866         }
1867
1868         for (i = 0; i < vf->vsi_res->num_queue_pairs; i++) {
1869                 vf->proto_xtr[i] = ad->devargs.proto_xtr[i] !=
1870                                         IAVF_PROTO_XTR_NONE ?
1871                                         ad->devargs.proto_xtr[i] :
1872                                         ad->devargs.proto_xtr_dflt;
1873
1874                 if (vf->proto_xtr[i] != IAVF_PROTO_XTR_NONE) {
1875                         uint8_t type = vf->proto_xtr[i];
1876
1877                         iavf_proto_xtr_params[type].required = true;
1878                         proto_xtr_enable = true;
1879                 }
1880         }
1881
1882         if (likely(!proto_xtr_enable))
1883                 return;
1884
1885         offset = rte_mbuf_dynfield_register(&iavf_proto_xtr_metadata_param);
1886         if (unlikely(offset == -1)) {
1887                 PMD_DRV_LOG(ERR,
1888                             "failed to extract protocol metadata, error %d",
1889                             -rte_errno);
1890                 return;
1891         }
1892
1893         PMD_DRV_LOG(DEBUG,
1894                     "proto_xtr metadata offset in mbuf is : %d",
1895                     offset);
1896         rte_pmd_ifd_dynfield_proto_xtr_metadata_offs = offset;
1897
1898         for (i = 0; i < RTE_DIM(iavf_proto_xtr_params); i++) {
1899                 xtr_ol = &iavf_proto_xtr_params[i];
1900
1901                 uint8_t rxdid = iavf_proto_xtr_type_to_rxdid((uint8_t)i);
1902
1903                 if (!xtr_ol->required)
1904                         continue;
1905
1906                 if (!(vf->supported_rxdid & BIT(rxdid))) {
1907                         PMD_DRV_LOG(ERR,
1908                                     "rxdid[%u] is not supported in hardware",
1909                                     rxdid);
1910                         rte_pmd_ifd_dynfield_proto_xtr_metadata_offs = -1;
1911                         break;
1912                 }
1913
1914                 offset = rte_mbuf_dynflag_register(&xtr_ol->param);
1915                 if (unlikely(offset == -1)) {
1916                         PMD_DRV_LOG(ERR,
1917                                     "failed to register proto_xtr offload '%s', error %d",
1918                                     xtr_ol->param.name, -rte_errno);
1919
1920                         rte_pmd_ifd_dynfield_proto_xtr_metadata_offs = -1;
1921                         break;
1922                 }
1923
1924                 PMD_DRV_LOG(DEBUG,
1925                             "proto_xtr offload '%s' offset in mbuf is : %d",
1926                             xtr_ol->param.name, offset);
1927                 *xtr_ol->ol_flag = 1ULL << offset;
1928         }
1929 }
1930
1931 static int
1932 iavf_init_vf(struct rte_eth_dev *dev)
1933 {
1934         int err, bufsz;
1935         struct iavf_adapter *adapter =
1936                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1937         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1938         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
1939
1940         err = iavf_parse_devargs(dev);
1941         if (err) {
1942                 PMD_INIT_LOG(ERR, "Failed to parse devargs");
1943                 goto err;
1944         }
1945
1946         err = iavf_set_mac_type(hw);
1947         if (err) {
1948                 PMD_INIT_LOG(ERR, "set_mac_type failed: %d", err);
1949                 goto err;
1950         }
1951
1952         err = iavf_check_vf_reset_done(hw);
1953         if (err) {
1954                 PMD_INIT_LOG(ERR, "VF is still resetting");
1955                 goto err;
1956         }
1957
1958         iavf_init_adminq_parameter(hw);
1959         err = iavf_init_adminq(hw);
1960         if (err) {
1961                 PMD_INIT_LOG(ERR, "init_adminq failed: %d", err);
1962                 goto err;
1963         }
1964
1965         vf->aq_resp = rte_zmalloc("vf_aq_resp", IAVF_AQ_BUF_SZ, 0);
1966         if (!vf->aq_resp) {
1967                 PMD_INIT_LOG(ERR, "unable to allocate vf_aq_resp memory");
1968                 goto err_aq;
1969         }
1970         if (iavf_check_api_version(adapter) != 0) {
1971                 PMD_INIT_LOG(ERR, "check_api version failed");
1972                 goto err_api;
1973         }
1974
1975         bufsz = sizeof(struct virtchnl_vf_resource) +
1976                 (IAVF_MAX_VF_VSI * sizeof(struct virtchnl_vsi_resource));
1977         vf->vf_res = rte_zmalloc("vf_res", bufsz, 0);
1978         if (!vf->vf_res) {
1979                 PMD_INIT_LOG(ERR, "unable to allocate vf_res memory");
1980                 goto err_api;
1981         }
1982         if (iavf_get_vf_resource(adapter) != 0) {
1983                 PMD_INIT_LOG(ERR, "iavf_get_vf_config failed");
1984                 goto err_alloc;
1985         }
1986         /* Allocate memort for RSS info */
1987         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF) {
1988                 vf->rss_key = rte_zmalloc("rss_key",
1989                                           vf->vf_res->rss_key_size, 0);
1990                 if (!vf->rss_key) {
1991                         PMD_INIT_LOG(ERR, "unable to allocate rss_key memory");
1992                         goto err_rss;
1993                 }
1994                 vf->rss_lut = rte_zmalloc("rss_lut",
1995                                           vf->vf_res->rss_lut_size, 0);
1996                 if (!vf->rss_lut) {
1997                         PMD_INIT_LOG(ERR, "unable to allocate rss_lut memory");
1998                         goto err_rss;
1999                 }
2000         }
2001
2002         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RX_FLEX_DESC) {
2003                 if (iavf_get_supported_rxdid(adapter) != 0) {
2004                         PMD_INIT_LOG(ERR, "failed to do get supported rxdid");
2005                         goto err_rss;
2006                 }
2007         }
2008
2009         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_VLAN_V2) {
2010                 if (iavf_get_vlan_offload_caps_v2(adapter) != 0) {
2011                         PMD_INIT_LOG(ERR, "failed to do get VLAN offload v2 capabilities");
2012                         goto err_rss;
2013                 }
2014         }
2015
2016         iavf_init_proto_xtr(dev);
2017
2018         return 0;
2019 err_rss:
2020         rte_free(vf->rss_key);
2021         rte_free(vf->rss_lut);
2022 err_alloc:
2023         rte_free(vf->vf_res);
2024         vf->vsi_res = NULL;
2025 err_api:
2026         rte_free(vf->aq_resp);
2027 err_aq:
2028         iavf_shutdown_adminq(hw);
2029 err:
2030         return -1;
2031 }
2032
2033 /* Enable default admin queue interrupt setting */
2034 static inline void
2035 iavf_enable_irq0(struct iavf_hw *hw)
2036 {
2037         /* Enable admin queue interrupt trigger */
2038         IAVF_WRITE_REG(hw, IAVF_VFINT_ICR0_ENA1,
2039                        IAVF_VFINT_ICR0_ENA1_ADMINQ_MASK);
2040
2041         IAVF_WRITE_REG(hw, IAVF_VFINT_DYN_CTL01,
2042                        IAVF_VFINT_DYN_CTL01_INTENA_MASK |
2043                        IAVF_VFINT_DYN_CTL01_CLEARPBA_MASK |
2044                        IAVF_VFINT_DYN_CTL01_ITR_INDX_MASK);
2045
2046         IAVF_WRITE_FLUSH(hw);
2047 }
2048
2049 static inline void
2050 iavf_disable_irq0(struct iavf_hw *hw)
2051 {
2052         /* Disable all interrupt types */
2053         IAVF_WRITE_REG(hw, IAVF_VFINT_ICR0_ENA1, 0);
2054         IAVF_WRITE_REG(hw, IAVF_VFINT_DYN_CTL01,
2055                        IAVF_VFINT_DYN_CTL01_ITR_INDX_MASK);
2056         IAVF_WRITE_FLUSH(hw);
2057 }
2058
2059 static void
2060 iavf_dev_interrupt_handler(void *param)
2061 {
2062         struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
2063         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2064
2065         iavf_disable_irq0(hw);
2066
2067         iavf_handle_virtchnl_msg(dev);
2068
2069         iavf_enable_irq0(hw);
2070 }
2071
2072 static int
2073 iavf_dev_filter_ctrl(struct rte_eth_dev *dev,
2074                      enum rte_filter_type filter_type,
2075                      enum rte_filter_op filter_op,
2076                      void *arg)
2077 {
2078         int ret = 0;
2079
2080         if (!dev)
2081                 return -EINVAL;
2082
2083         switch (filter_type) {
2084         case RTE_ETH_FILTER_GENERIC:
2085                 if (filter_op != RTE_ETH_FILTER_GET)
2086                         return -EINVAL;
2087                 *(const void **)arg = &iavf_flow_ops;
2088                 break;
2089         default:
2090                 PMD_DRV_LOG(WARNING, "Filter type (%d) not supported",
2091                             filter_type);
2092                 ret = -EINVAL;
2093                 break;
2094         }
2095
2096         return ret;
2097 }
2098
2099 static void
2100 iavf_default_rss_disable(struct iavf_adapter *adapter)
2101 {
2102         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
2103         int ret = 0;
2104
2105         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_ADV_RSS_PF) {
2106                 /* Set hena = 0 to ask PF to cleanup all existing RSS. */
2107                 ret = iavf_set_hena(adapter, 0);
2108                 if (ret)
2109                         /* It is a workaround, temporarily allow error to be
2110                          * returned due to possible lack of PF handling for
2111                          * hena = 0.
2112                          */
2113                         PMD_INIT_LOG(WARNING, "fail to disable default RSS,"
2114                                     "lack PF support");
2115         }
2116 }
2117
2118 static int
2119 iavf_dev_init(struct rte_eth_dev *eth_dev)
2120 {
2121         struct iavf_adapter *adapter =
2122                 IAVF_DEV_PRIVATE_TO_ADAPTER(eth_dev->data->dev_private);
2123         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(adapter);
2124         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
2125         int ret = 0;
2126
2127         PMD_INIT_FUNC_TRACE();
2128
2129         /* assign ops func pointer */
2130         eth_dev->dev_ops = &iavf_eth_dev_ops;
2131         eth_dev->rx_queue_count = iavf_dev_rxq_count;
2132         eth_dev->rx_descriptor_status = iavf_dev_rx_desc_status;
2133         eth_dev->tx_descriptor_status = iavf_dev_tx_desc_status;
2134         eth_dev->rx_pkt_burst = &iavf_recv_pkts;
2135         eth_dev->tx_pkt_burst = &iavf_xmit_pkts;
2136         eth_dev->tx_pkt_prepare = &iavf_prep_pkts;
2137
2138         /* For secondary processes, we don't initialise any further as primary
2139          * has already done this work. Only check if we need a different RX
2140          * and TX function.
2141          */
2142         if (rte_eal_process_type() != RTE_PROC_PRIMARY) {
2143                 iavf_set_rx_function(eth_dev);
2144                 iavf_set_tx_function(eth_dev);
2145                 return 0;
2146         }
2147         rte_eth_copy_pci_info(eth_dev, pci_dev);
2148         eth_dev->data->dev_flags |= RTE_ETH_DEV_AUTOFILL_QUEUE_XSTATS;
2149
2150         hw->vendor_id = pci_dev->id.vendor_id;
2151         hw->device_id = pci_dev->id.device_id;
2152         hw->subsystem_vendor_id = pci_dev->id.subsystem_vendor_id;
2153         hw->subsystem_device_id = pci_dev->id.subsystem_device_id;
2154         hw->bus.bus_id = pci_dev->addr.bus;
2155         hw->bus.device = pci_dev->addr.devid;
2156         hw->bus.func = pci_dev->addr.function;
2157         hw->hw_addr = (void *)pci_dev->mem_resource[0].addr;
2158         hw->back = IAVF_DEV_PRIVATE_TO_ADAPTER(eth_dev->data->dev_private);
2159         adapter->eth_dev = eth_dev;
2160         adapter->stopped = 1;
2161
2162         if (iavf_init_vf(eth_dev) != 0) {
2163                 PMD_INIT_LOG(ERR, "Init vf failed");
2164                 return -1;
2165         }
2166
2167         /* set default ptype table */
2168         adapter->ptype_tbl = iavf_get_default_ptype_table();
2169
2170         /* copy mac addr */
2171         eth_dev->data->mac_addrs = rte_zmalloc(
2172                 "iavf_mac", RTE_ETHER_ADDR_LEN * IAVF_NUM_MACADDR_MAX, 0);
2173         if (!eth_dev->data->mac_addrs) {
2174                 PMD_INIT_LOG(ERR, "Failed to allocate %d bytes needed to"
2175                              " store MAC addresses",
2176                              RTE_ETHER_ADDR_LEN * IAVF_NUM_MACADDR_MAX);
2177                 return -ENOMEM;
2178         }
2179         /* If the MAC address is not configured by host,
2180          * generate a random one.
2181          */
2182         if (!rte_is_valid_assigned_ether_addr(
2183                         (struct rte_ether_addr *)hw->mac.addr))
2184                 rte_eth_random_addr(hw->mac.addr);
2185         rte_ether_addr_copy((struct rte_ether_addr *)hw->mac.addr,
2186                         &eth_dev->data->mac_addrs[0]);
2187
2188         /* register callback func to eal lib */
2189         rte_intr_callback_register(&pci_dev->intr_handle,
2190                                    iavf_dev_interrupt_handler,
2191                                    (void *)eth_dev);
2192
2193         /* enable uio intr after callback register */
2194         rte_intr_enable(&pci_dev->intr_handle);
2195
2196         /* configure and enable device interrupt */
2197         iavf_enable_irq0(hw);
2198
2199         ret = iavf_flow_init(adapter);
2200         if (ret) {
2201                 PMD_INIT_LOG(ERR, "Failed to initialize flow");
2202                 return ret;
2203         }
2204
2205         iavf_default_rss_disable(adapter);
2206
2207         return 0;
2208 }
2209
2210 static int
2211 iavf_dev_close(struct rte_eth_dev *dev)
2212 {
2213         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2214         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
2215         struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
2216         struct iavf_adapter *adapter =
2217                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
2218         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
2219         int ret;
2220
2221         if (rte_eal_process_type() != RTE_PROC_PRIMARY)
2222                 return 0;
2223
2224         ret = iavf_dev_stop(dev);
2225
2226         iavf_flow_flush(dev, NULL);
2227         iavf_flow_uninit(adapter);
2228
2229         /*
2230          * disable promiscuous mode before reset vf
2231          * it is a workaround solution when work with kernel driver
2232          * and it is not the normal way
2233          */
2234         if (vf->promisc_unicast_enabled || vf->promisc_multicast_enabled)
2235                 iavf_config_promisc(adapter, false, false);
2236
2237         iavf_shutdown_adminq(hw);
2238         /* disable uio intr before callback unregister */
2239         rte_intr_disable(intr_handle);
2240
2241         /* unregister callback func from eal lib */
2242         rte_intr_callback_unregister(intr_handle,
2243                                      iavf_dev_interrupt_handler, dev);
2244         iavf_disable_irq0(hw);
2245
2246         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF) {
2247                 if (vf->rss_lut) {
2248                         rte_free(vf->rss_lut);
2249                         vf->rss_lut = NULL;
2250                 }
2251                 if (vf->rss_key) {
2252                         rte_free(vf->rss_key);
2253                         vf->rss_key = NULL;
2254                 }
2255         }
2256
2257         rte_free(vf->vf_res);
2258         vf->vsi_res = NULL;
2259         vf->vf_res = NULL;
2260
2261         rte_free(vf->aq_resp);
2262         vf->aq_resp = NULL;
2263
2264         vf->vf_reset = false;
2265
2266         return ret;
2267 }
2268
2269 static int
2270 iavf_dev_uninit(struct rte_eth_dev *dev)
2271 {
2272         if (rte_eal_process_type() != RTE_PROC_PRIMARY)
2273                 return -EPERM;
2274
2275         iavf_dev_close(dev);
2276
2277         return 0;
2278 }
2279
2280 /*
2281  * Reset VF device only to re-initialize resources in PMD layer
2282  */
2283 static int
2284 iavf_dev_reset(struct rte_eth_dev *dev)
2285 {
2286         int ret;
2287
2288         ret = iavf_dev_uninit(dev);
2289         if (ret)
2290                 return ret;
2291
2292         return iavf_dev_init(dev);
2293 }
2294
2295 static int
2296 iavf_dcf_cap_check_handler(__rte_unused const char *key,
2297                            const char *value, __rte_unused void *opaque)
2298 {
2299         if (strcmp(value, "dcf"))
2300                 return -1;
2301
2302         return 0;
2303 }
2304
2305 static int
2306 iavf_dcf_cap_selected(struct rte_devargs *devargs)
2307 {
2308         struct rte_kvargs *kvlist;
2309         const char *key = "cap";
2310         int ret = 0;
2311
2312         if (devargs == NULL)
2313                 return 0;
2314
2315         kvlist = rte_kvargs_parse(devargs->args, NULL);
2316         if (kvlist == NULL)
2317                 return 0;
2318
2319         if (!rte_kvargs_count(kvlist, key))
2320                 goto exit;
2321
2322         /* dcf capability selected when there's a key-value pair: cap=dcf */
2323         if (rte_kvargs_process(kvlist, key,
2324                                iavf_dcf_cap_check_handler, NULL) < 0)
2325                 goto exit;
2326
2327         ret = 1;
2328
2329 exit:
2330         rte_kvargs_free(kvlist);
2331         return ret;
2332 }
2333
2334 static int eth_iavf_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
2335                              struct rte_pci_device *pci_dev)
2336 {
2337         if (iavf_dcf_cap_selected(pci_dev->device.devargs))
2338                 return 1;
2339
2340         return rte_eth_dev_pci_generic_probe(pci_dev,
2341                 sizeof(struct iavf_adapter), iavf_dev_init);
2342 }
2343
2344 static int eth_iavf_pci_remove(struct rte_pci_device *pci_dev)
2345 {
2346         return rte_eth_dev_pci_generic_remove(pci_dev, iavf_dev_uninit);
2347 }
2348
2349 /* Adaptive virtual function driver struct */
2350 static struct rte_pci_driver rte_iavf_pmd = {
2351         .id_table = pci_id_iavf_map,
2352         .drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC,
2353         .probe = eth_iavf_pci_probe,
2354         .remove = eth_iavf_pci_remove,
2355 };
2356
2357 RTE_PMD_REGISTER_PCI(net_iavf, rte_iavf_pmd);
2358 RTE_PMD_REGISTER_PCI_TABLE(net_iavf, pci_id_iavf_map);
2359 RTE_PMD_REGISTER_KMOD_DEP(net_iavf, "* igb_uio | vfio-pci");
2360 RTE_PMD_REGISTER_PARAM_STRING(net_iavf, "cap=dcf");
2361 RTE_LOG_REGISTER(iavf_logtype_init, pmd.net.iavf.init, NOTICE);
2362 RTE_LOG_REGISTER(iavf_logtype_driver, pmd.net.iavf.driver, NOTICE);
2363 #ifdef RTE_LIBRTE_IAVF_DEBUG_RX
2364 RTE_LOG_REGISTER(iavf_logtype_rx, pmd.net.iavf.rx, DEBUG);
2365 #endif
2366 #ifdef RTE_LIBRTE_IAVF_DEBUG_TX
2367 RTE_LOG_REGISTER(iavf_logtype_tx, pmd.net.iavf.tx, DEBUG);
2368 #endif
2369 #ifdef RTE_LIBRTE_IAVF_DEBUG_TX_FREE
2370 RTE_LOG_REGISTER(iavf_logtype_tx_free, pmd.net.iavf.tx_free, DEBUG);
2371 #endif