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