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