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 static int iavf_tm_ops_get(struct rte_eth_dev *dev __rte_unused, void *arg);
 126
 127 static const struct rte_pci_id pci_id_iavf_map[] = {
 128         { RTE_PCI_DEVICE(IAVF_INTEL_VENDOR_ID, IAVF_DEV_ID_ADAPTIVE_VF) },
 129         { RTE_PCI_DEVICE(IAVF_INTEL_VENDOR_ID, IAVF_DEV_ID_VF) },
 130         { RTE_PCI_DEVICE(IAVF_INTEL_VENDOR_ID, IAVF_DEV_ID_VF_HV) },
 131         { RTE_PCI_DEVICE(IAVF_INTEL_VENDOR_ID, IAVF_DEV_ID_X722_VF) },
 132         { RTE_PCI_DEVICE(IAVF_INTEL_VENDOR_ID, IAVF_DEV_ID_X722_A0_VF) },
 133         { .vendor_id = 0, /* sentinel */ },
 134 };
 135
 136 struct rte_iavf_xstats_name_off {
 137         char name[RTE_ETH_XSTATS_NAME_SIZE];
 138         unsigned int offset;
 139 };
 140
 141 static const struct rte_iavf_xstats_name_off rte_iavf_stats_strings[] = {
 142         {"rx_bytes", offsetof(struct iavf_eth_stats, rx_bytes)},
 143         {"rx_unicast_packets", offsetof(struct iavf_eth_stats, rx_unicast)},
 144         {"rx_multicast_packets", offsetof(struct iavf_eth_stats, rx_multicast)},
 145         {"rx_broadcast_packets", offsetof(struct iavf_eth_stats, rx_broadcast)},
 146         {"rx_dropped_packets", offsetof(struct iavf_eth_stats, rx_discards)},
 147         {"rx_unknown_protocol_packets", offsetof(struct iavf_eth_stats,
 148                 rx_unknown_protocol)},
 149         {"tx_bytes", offsetof(struct iavf_eth_stats, tx_bytes)},
 150         {"tx_unicast_packets", offsetof(struct iavf_eth_stats, tx_unicast)},
 151         {"tx_multicast_packets", offsetof(struct iavf_eth_stats, tx_multicast)},
 152         {"tx_broadcast_packets", offsetof(struct iavf_eth_stats, tx_broadcast)},
 153         {"tx_dropped_packets", offsetof(struct iavf_eth_stats, tx_discards)},
 154         {"tx_error_packets", offsetof(struct iavf_eth_stats, tx_errors)},
 155 };
 156
 157 #define IAVF_NB_XSTATS (sizeof(rte_iavf_stats_strings) / \
 158                 sizeof(rte_iavf_stats_strings[0]))
 159
 160 static const struct eth_dev_ops iavf_eth_dev_ops = {
 161         .dev_configure              = iavf_dev_configure,
 162         .dev_start                  = iavf_dev_start,
 163         .dev_stop                   = iavf_dev_stop,
 164         .dev_close                  = iavf_dev_close,
 165         .dev_reset                  = iavf_dev_reset,
 166         .dev_infos_get              = iavf_dev_info_get,
 167         .dev_supported_ptypes_get   = iavf_dev_supported_ptypes_get,
 168         .link_update                = iavf_dev_link_update,
 169         .stats_get                  = iavf_dev_stats_get,
 170         .stats_reset                = iavf_dev_stats_reset,
 171         .xstats_get                 = iavf_dev_xstats_get,
 172         .xstats_get_names           = iavf_dev_xstats_get_names,
 173         .xstats_reset               = iavf_dev_stats_reset,
 174         .promiscuous_enable         = iavf_dev_promiscuous_enable,
 175         .promiscuous_disable        = iavf_dev_promiscuous_disable,
 176         .allmulticast_enable        = iavf_dev_allmulticast_enable,
 177         .allmulticast_disable       = iavf_dev_allmulticast_disable,
 178         .mac_addr_add               = iavf_dev_add_mac_addr,
 179         .mac_addr_remove            = iavf_dev_del_mac_addr,
 180         .set_mc_addr_list                       = iavf_set_mc_addr_list,
 181         .vlan_filter_set            = iavf_dev_vlan_filter_set,
 182         .vlan_offload_set           = iavf_dev_vlan_offload_set,
 183         .rx_queue_start             = iavf_dev_rx_queue_start,
 184         .rx_queue_stop              = iavf_dev_rx_queue_stop,
 185         .tx_queue_start             = iavf_dev_tx_queue_start,
 186         .tx_queue_stop              = iavf_dev_tx_queue_stop,
 187         .rx_queue_setup             = iavf_dev_rx_queue_setup,
 188         .rx_queue_release           = iavf_dev_rx_queue_release,
 189         .tx_queue_setup             = iavf_dev_tx_queue_setup,
 190         .tx_queue_release           = iavf_dev_tx_queue_release,
 191         .mac_addr_set               = iavf_dev_set_default_mac_addr,
 192         .reta_update                = iavf_dev_rss_reta_update,
 193         .reta_query                 = iavf_dev_rss_reta_query,
 194         .rss_hash_update            = iavf_dev_rss_hash_update,
 195         .rss_hash_conf_get          = iavf_dev_rss_hash_conf_get,
 196         .rxq_info_get               = iavf_dev_rxq_info_get,
 197         .txq_info_get               = iavf_dev_txq_info_get,
 198         .mtu_set                    = iavf_dev_mtu_set,
 199         .rx_queue_intr_enable       = iavf_dev_rx_queue_intr_enable,
 200         .rx_queue_intr_disable      = iavf_dev_rx_queue_intr_disable,
 201         .flow_ops_get               = iavf_dev_flow_ops_get,
 202         .tx_done_cleanup            = iavf_dev_tx_done_cleanup,
 203         .get_monitor_addr           = iavf_get_monitor_addr,
 204         .tm_ops_get                 = iavf_tm_ops_get,
 205 };
 206
 207 static int
 208 iavf_tm_ops_get(struct rte_eth_dev *dev __rte_unused,
 209                         void *arg)
 210 {
 211         if (!arg)
 212                 return -EINVAL;
 213
 214         *(const void **)arg = &iavf_tm_ops;
 215
 216         return 0;
 217 }
 218
 219 static int
 220 iavf_set_mc_addr_list(struct rte_eth_dev *dev,
 221                         struct rte_ether_addr *mc_addrs,
 222                         uint32_t mc_addrs_num)
 223 {
 224         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
 225         struct iavf_adapter *adapter =
 226                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 227         int err, ret;
 228
 229         if (mc_addrs_num > IAVF_NUM_MACADDR_MAX) {
 230                 PMD_DRV_LOG(ERR,
 231                             "can't add more than a limited number (%u) of addresses.",
 232                             (uint32_t)IAVF_NUM_MACADDR_MAX);
 233                 return -EINVAL;
 234         }
 235
 236         /* flush previous addresses */
 237         err = iavf_add_del_mc_addr_list(adapter, vf->mc_addrs, vf->mc_addrs_num,
 238                                         false);
 239         if (err)
 240                 return err;
 241
 242         /* add new ones */
 243         err = iavf_add_del_mc_addr_list(adapter, mc_addrs, mc_addrs_num, true);
 244
 245         if (err) {
 246                 /* if adding mac address list fails, should add the previous
 247                  * addresses back.
 248                  */
 249                 ret = iavf_add_del_mc_addr_list(adapter, vf->mc_addrs,
 250                                                 vf->mc_addrs_num, true);
 251                 if (ret)
 252                         return ret;
 253         } else {
 254                 vf->mc_addrs_num = mc_addrs_num;
 255                 memcpy(vf->mc_addrs,
 256                        mc_addrs, mc_addrs_num * sizeof(*mc_addrs));
 257         }
 258
 259         return err;
 260 }
 261
 262 static void
 263 iavf_config_rss_hf(struct iavf_adapter *adapter, uint64_t rss_hf)
 264 {
 265         static const uint64_t map_hena_rss[] = {
 266                 /* IPv4 */
 267                 [IAVF_FILTER_PCTYPE_NONF_UNICAST_IPV4_UDP] =
 268                                 ETH_RSS_NONFRAG_IPV4_UDP,
 269                 [IAVF_FILTER_PCTYPE_NONF_MULTICAST_IPV4_UDP] =
 270                                 ETH_RSS_NONFRAG_IPV4_UDP,
 271                 [IAVF_FILTER_PCTYPE_NONF_IPV4_UDP] =
 272                                 ETH_RSS_NONFRAG_IPV4_UDP,
 273                 [IAVF_FILTER_PCTYPE_NONF_IPV4_TCP_SYN_NO_ACK] =
 274                                 ETH_RSS_NONFRAG_IPV4_TCP,
 275                 [IAVF_FILTER_PCTYPE_NONF_IPV4_TCP] =
 276                                 ETH_RSS_NONFRAG_IPV4_TCP,
 277                 [IAVF_FILTER_PCTYPE_NONF_IPV4_SCTP] =
 278                                 ETH_RSS_NONFRAG_IPV4_SCTP,
 279                 [IAVF_FILTER_PCTYPE_NONF_IPV4_OTHER] =
 280                                 ETH_RSS_NONFRAG_IPV4_OTHER,
 281                 [IAVF_FILTER_PCTYPE_FRAG_IPV4] = ETH_RSS_FRAG_IPV4,
 282
 283                 /* IPv6 */
 284                 [IAVF_FILTER_PCTYPE_NONF_UNICAST_IPV6_UDP] =
 285                                 ETH_RSS_NONFRAG_IPV6_UDP,
 286                 [IAVF_FILTER_PCTYPE_NONF_MULTICAST_IPV6_UDP] =
 287                                 ETH_RSS_NONFRAG_IPV6_UDP,
 288                 [IAVF_FILTER_PCTYPE_NONF_IPV6_UDP] =
 289                                 ETH_RSS_NONFRAG_IPV6_UDP,
 290                 [IAVF_FILTER_PCTYPE_NONF_IPV6_TCP_SYN_NO_ACK] =
 291                                 ETH_RSS_NONFRAG_IPV6_TCP,
 292                 [IAVF_FILTER_PCTYPE_NONF_IPV6_TCP] =
 293                                 ETH_RSS_NONFRAG_IPV6_TCP,
 294                 [IAVF_FILTER_PCTYPE_NONF_IPV6_SCTP] =
 295                                 ETH_RSS_NONFRAG_IPV6_SCTP,
 296                 [IAVF_FILTER_PCTYPE_NONF_IPV6_OTHER] =
 297                                 ETH_RSS_NONFRAG_IPV6_OTHER,
 298                 [IAVF_FILTER_PCTYPE_FRAG_IPV6] = ETH_RSS_FRAG_IPV6,
 299
 300                 /* L2 Payload */
 301                 [IAVF_FILTER_PCTYPE_L2_PAYLOAD] = ETH_RSS_L2_PAYLOAD
 302         };
 303
 304         const uint64_t ipv4_rss = ETH_RSS_NONFRAG_IPV4_UDP |
 305                                   ETH_RSS_NONFRAG_IPV4_TCP |
 306                                   ETH_RSS_NONFRAG_IPV4_SCTP |
 307                                   ETH_RSS_NONFRAG_IPV4_OTHER |
 308                                   ETH_RSS_FRAG_IPV4;
 309
 310         const uint64_t ipv6_rss = ETH_RSS_NONFRAG_IPV6_UDP |
 311                                   ETH_RSS_NONFRAG_IPV6_TCP |
 312                                   ETH_RSS_NONFRAG_IPV6_SCTP |
 313                                   ETH_RSS_NONFRAG_IPV6_OTHER |
 314                                   ETH_RSS_FRAG_IPV6;
 315
 316         struct iavf_info *vf =  IAVF_DEV_PRIVATE_TO_VF(adapter);
 317         uint64_t caps = 0, hena = 0, valid_rss_hf = 0;
 318         uint32_t i;
 319         int ret;
 320
 321         ret = iavf_get_hena_caps(adapter, &caps);
 322         if (ret) {
 323                 /**
 324                  * RSS offload type configuration is not a necessary feature
 325                  * for VF, so here just print a warning and return.
 326                  */
 327                 PMD_DRV_LOG(WARNING,
 328                             "fail to get RSS offload type caps, ret: %d", ret);
 329                 return;
 330         }
 331
 332         /**
 333          * ETH_RSS_IPV4 and ETH_RSS_IPV6 can be considered as 2
 334          * generalizations of all other IPv4 and IPv6 RSS types.
 335          */
 336         if (rss_hf & ETH_RSS_IPV4)
 337                 rss_hf |= ipv4_rss;
 338
 339         if (rss_hf & ETH_RSS_IPV6)
 340                 rss_hf |= ipv6_rss;
 341
 342         RTE_BUILD_BUG_ON(RTE_DIM(map_hena_rss) > sizeof(uint64_t) * CHAR_BIT);
 343
 344         for (i = 0; i < RTE_DIM(map_hena_rss); i++) {
 345                 uint64_t bit = BIT_ULL(i);
 346
 347                 if ((caps & bit) && (map_hena_rss[i] & rss_hf)) {
 348                         valid_rss_hf |= map_hena_rss[i];
 349                         hena |= bit;
 350                 }
 351         }
 352
 353         ret = iavf_set_hena(adapter, hena);
 354         if (ret) {
 355                 /**
 356                  * RSS offload type configuration is not a necessary feature
 357                  * for VF, so here just print a warning and return.
 358                  */
 359                 PMD_DRV_LOG(WARNING,
 360                             "fail to set RSS offload types, ret: %d", ret);
 361                 return;
 362         }
 363
 364         if (valid_rss_hf & ipv4_rss)
 365                 valid_rss_hf |= rss_hf & ETH_RSS_IPV4;
 366
 367         if (valid_rss_hf & ipv6_rss)
 368                 valid_rss_hf |= rss_hf & ETH_RSS_IPV6;
 369
 370         if (rss_hf & ~valid_rss_hf)
 371                 PMD_DRV_LOG(WARNING, "Unsupported rss_hf 0x%" PRIx64,
 372                             rss_hf & ~valid_rss_hf);
 373
 374         vf->rss_hf = valid_rss_hf;
 375 }
 376
 377 static int
 378 iavf_init_rss(struct iavf_adapter *adapter)
 379 {
 380         struct iavf_info *vf =  IAVF_DEV_PRIVATE_TO_VF(adapter);
 381         struct rte_eth_rss_conf *rss_conf;
 382         uint16_t i, j, nb_q;
 383         int ret;
 384
 385         rss_conf = &adapter->eth_dev->data->dev_conf.rx_adv_conf.rss_conf;
 386         nb_q = RTE_MIN(adapter->eth_dev->data->nb_rx_queues,
 387                        vf->max_rss_qregion);
 388
 389         if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF)) {
 390                 PMD_DRV_LOG(DEBUG, "RSS is not supported");
 391                 return -ENOTSUP;
 392         }
 393
 394         /* configure RSS key */
 395         if (!rss_conf->rss_key) {
 396                 /* Calculate the default hash key */
 397                 for (i = 0; i < vf->vf_res->rss_key_size; i++)
 398                         vf->rss_key[i] = (uint8_t)rte_rand();
 399         } else
 400                 rte_memcpy(vf->rss_key, rss_conf->rss_key,
 401                            RTE_MIN(rss_conf->rss_key_len,
 402                                    vf->vf_res->rss_key_size));
 403
 404         /* init RSS LUT table */
 405         for (i = 0, j = 0; i < vf->vf_res->rss_lut_size; i++, j++) {
 406                 if (j >= nb_q)
 407                         j = 0;
 408                 vf->rss_lut[i] = j;
 409         }
 410         /* send virtchnnl ops to configure rss*/
 411         ret = iavf_configure_rss_lut(adapter);
 412         if (ret)
 413                 return ret;
 414         ret = iavf_configure_rss_key(adapter);
 415         if (ret)
 416                 return ret;
 417
 418         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_ADV_RSS_PF) {
 419                 /* Set RSS hash configuration based on rss_conf->rss_hf. */
 420                 ret = iavf_rss_hash_set(adapter, rss_conf->rss_hf, true);
 421                 if (ret) {
 422                         PMD_DRV_LOG(ERR, "fail to set default RSS");
 423                         return ret;
 424                 }
 425         } else {
 426                 iavf_config_rss_hf(adapter, rss_conf->rss_hf);
 427         }
 428
 429         return 0;
 430 }
 431
 432 static int
 433 iavf_queues_req_reset(struct rte_eth_dev *dev, uint16_t num)
 434 {
 435         struct iavf_adapter *ad =
 436                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 437         struct iavf_info *vf =  IAVF_DEV_PRIVATE_TO_VF(ad);
 438         int ret;
 439
 440         ret = iavf_request_queues(ad, num);
 441         if (ret) {
 442                 PMD_DRV_LOG(ERR, "request queues from PF failed");
 443                 return ret;
 444         }
 445         PMD_DRV_LOG(INFO, "change queue pairs from %u to %u",
 446                         vf->vsi_res->num_queue_pairs, num);
 447
 448         ret = iavf_dev_reset(dev);
 449         if (ret) {
 450                 PMD_DRV_LOG(ERR, "vf reset failed");
 451                 return ret;
 452         }
 453
 454         return 0;
 455 }
 456
 457 static int
 458 iavf_dev_vlan_insert_set(struct rte_eth_dev *dev)
 459 {
 460         struct iavf_adapter *adapter =
 461                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 462         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
 463         bool enable;
 464
 465         if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_VLAN_V2))
 466                 return 0;
 467
 468         enable = !!(dev->data->dev_conf.txmode.offloads &
 469                     DEV_TX_OFFLOAD_VLAN_INSERT);
 470         iavf_config_vlan_insert_v2(adapter, enable);
 471
 472         return 0;
 473 }
 474
 475 static int
 476 iavf_dev_init_vlan(struct rte_eth_dev *dev)
 477 {
 478         int err;
 479
 480         err = iavf_dev_vlan_offload_set(dev,
 481                                         ETH_VLAN_STRIP_MASK |
 482                                         ETH_QINQ_STRIP_MASK |
 483                                         ETH_VLAN_FILTER_MASK |
 484                                         ETH_VLAN_EXTEND_MASK);
 485         if (err) {
 486                 PMD_DRV_LOG(ERR, "Failed to update vlan offload");
 487                 return err;
 488         }
 489
 490         err = iavf_dev_vlan_insert_set(dev);
 491         if (err)
 492                 PMD_DRV_LOG(ERR, "Failed to update vlan insertion");
 493
 494         return err;
 495 }
 496
 497 static int
 498 iavf_dev_configure(struct rte_eth_dev *dev)
 499 {
 500         struct iavf_adapter *ad =
 501                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 502         struct iavf_info *vf =  IAVF_DEV_PRIVATE_TO_VF(ad);
 503         uint16_t num_queue_pairs = RTE_MAX(dev->data->nb_rx_queues,
 504                 dev->data->nb_tx_queues);
 505         int ret;
 506
 507         ad->rx_bulk_alloc_allowed = true;
 508         /* Initialize to TRUE. If any of Rx queues doesn't meet the
 509          * vector Rx/Tx preconditions, it will be reset.
 510          */
 511         ad->rx_vec_allowed = true;
 512         ad->tx_vec_allowed = true;
 513
 514         if (dev->data->dev_conf.rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG)
 515                 dev->data->dev_conf.rxmode.offloads |= DEV_RX_OFFLOAD_RSS_HASH;
 516
 517         /* Large VF setting */
 518         if (num_queue_pairs > IAVF_MAX_NUM_QUEUES_DFLT) {
 519                 if (!(vf->vf_res->vf_cap_flags &
 520                                 VIRTCHNL_VF_LARGE_NUM_QPAIRS)) {
 521                         PMD_DRV_LOG(ERR, "large VF is not supported");
 522                         return -1;
 523                 }
 524
 525                 if (num_queue_pairs > IAVF_MAX_NUM_QUEUES_LV) {
 526                         PMD_DRV_LOG(ERR, "queue pairs number cannot be larger than %u",
 527                                 IAVF_MAX_NUM_QUEUES_LV);
 528                         return -1;
 529                 }
 530
 531                 ret = iavf_queues_req_reset(dev, num_queue_pairs);
 532                 if (ret)
 533                         return ret;
 534
 535                 ret = iavf_get_max_rss_queue_region(ad);
 536                 if (ret) {
 537                         PMD_INIT_LOG(ERR, "get max rss queue region failed");
 538                         return ret;
 539                 }
 540
 541                 vf->lv_enabled = true;
 542         } else {
 543                 /* Check if large VF is already enabled. If so, disable and
 544                  * release redundant queue resource.
 545                  * Or check if enough queue pairs. If not, request them from PF.
 546                  */
 547                 if (vf->lv_enabled ||
 548                     num_queue_pairs > vf->vsi_res->num_queue_pairs) {
 549                         ret = iavf_queues_req_reset(dev, num_queue_pairs);
 550                         if (ret)
 551                                 return ret;
 552
 553                         vf->lv_enabled = false;
 554                 }
 555                 /* if large VF is not required, use default rss queue region */
 556                 vf->max_rss_qregion = IAVF_MAX_NUM_QUEUES_DFLT;
 557         }
 558
 559         ret = iavf_dev_init_vlan(dev);
 560         if (ret)
 561                 PMD_DRV_LOG(ERR, "configure VLAN failed: %d", ret);
 562
 563         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF) {
 564                 if (iavf_init_rss(ad) != 0) {
 565                         PMD_DRV_LOG(ERR, "configure rss failed");
 566                         return -1;
 567                 }
 568         }
 569         return 0;
 570 }
 571
 572 static int
 573 iavf_init_rxq(struct rte_eth_dev *dev, struct iavf_rx_queue *rxq)
 574 {
 575         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(dev->data->dev_private);
 576         struct rte_eth_dev_data *dev_data = dev->data;
 577         uint16_t buf_size, max_pkt_len, len;
 578
 579         buf_size = rte_pktmbuf_data_room_size(rxq->mp) - RTE_PKTMBUF_HEADROOM;
 580
 581         /* Calculate the maximum packet length allowed */
 582         len = rxq->rx_buf_len * IAVF_MAX_CHAINED_RX_BUFFERS;
 583         max_pkt_len = RTE_MIN(len, dev->data->dev_conf.rxmode.max_rx_pkt_len);
 584
 585         /* Check if the jumbo frame and maximum packet length are set
 586          * correctly.
 587          */
 588         if (dev->data->dev_conf.rxmode.offloads & DEV_RX_OFFLOAD_JUMBO_FRAME) {
 589                 if (max_pkt_len <= IAVF_ETH_MAX_LEN ||
 590                     max_pkt_len > IAVF_FRAME_SIZE_MAX) {
 591                         PMD_DRV_LOG(ERR, "maximum packet length must be "
 592                                     "larger than %u and smaller than %u, "
 593                                     "as jumbo frame is enabled",
 594                                     (uint32_t)IAVF_ETH_MAX_LEN,
 595                                     (uint32_t)IAVF_FRAME_SIZE_MAX);
 596                         return -EINVAL;
 597                 }
 598         } else {
 599                 if (max_pkt_len < RTE_ETHER_MIN_LEN ||
 600                     max_pkt_len > IAVF_ETH_MAX_LEN) {
 601                         PMD_DRV_LOG(ERR, "maximum packet length must be "
 602                                     "larger than %u and smaller than %u, "
 603                                     "as jumbo frame is disabled",
 604                                     (uint32_t)RTE_ETHER_MIN_LEN,
 605                                     (uint32_t)IAVF_ETH_MAX_LEN);
 606                         return -EINVAL;
 607                 }
 608         }
 609
 610         rxq->max_pkt_len = max_pkt_len;
 611         if ((dev_data->dev_conf.rxmode.offloads & DEV_RX_OFFLOAD_SCATTER) ||
 612             rxq->max_pkt_len > buf_size) {
 613                 dev_data->scattered_rx = 1;
 614         }
 615         IAVF_PCI_REG_WRITE(rxq->qrx_tail, rxq->nb_rx_desc - 1);
 616         IAVF_WRITE_FLUSH(hw);
 617
 618         return 0;
 619 }
 620
 621 static int
 622 iavf_init_queues(struct rte_eth_dev *dev)
 623 {
 624         struct iavf_rx_queue **rxq =
 625                 (struct iavf_rx_queue **)dev->data->rx_queues;
 626         int i, ret = IAVF_SUCCESS;
 627
 628         for (i = 0; i < dev->data->nb_rx_queues; i++) {
 629                 if (!rxq[i] || !rxq[i]->q_set)
 630                         continue;
 631                 ret = iavf_init_rxq(dev, rxq[i]);
 632                 if (ret != IAVF_SUCCESS)
 633                         break;
 634         }
 635         /* set rx/tx function to vector/scatter/single-segment
 636          * according to parameters
 637          */
 638         iavf_set_rx_function(dev);
 639         iavf_set_tx_function(dev);
 640
 641         return ret;
 642 }
 643
 644 static int iavf_config_rx_queues_irqs(struct rte_eth_dev *dev,
 645                                      struct rte_intr_handle *intr_handle)
 646 {
 647         struct iavf_adapter *adapter =
 648                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 649         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
 650         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(adapter);
 651         struct iavf_qv_map *qv_map;
 652         uint16_t interval, i;
 653         int vec;
 654
 655         if (rte_intr_cap_multiple(intr_handle) &&
 656             dev->data->dev_conf.intr_conf.rxq) {
 657                 if (rte_intr_efd_enable(intr_handle, dev->data->nb_rx_queues))
 658                         return -1;
 659         }
 660
 661         if (rte_intr_dp_is_en(intr_handle) && !intr_handle->intr_vec) {
 662                 intr_handle->intr_vec =
 663                         rte_zmalloc("intr_vec",
 664                                     dev->data->nb_rx_queues * sizeof(int), 0);
 665                 if (!intr_handle->intr_vec) {
 666                         PMD_DRV_LOG(ERR, "Failed to allocate %d rx intr_vec",
 667                                     dev->data->nb_rx_queues);
 668                         return -1;
 669                 }
 670         }
 671
 672         qv_map = rte_zmalloc("qv_map",
 673                 dev->data->nb_rx_queues * sizeof(struct iavf_qv_map), 0);
 674         if (!qv_map) {
 675                 PMD_DRV_LOG(ERR, "Failed to allocate %d queue-vector map",
 676                                 dev->data->nb_rx_queues);
 677                 return -1;
 678         }
 679
 680         if (!dev->data->dev_conf.intr_conf.rxq ||
 681             !rte_intr_dp_is_en(intr_handle)) {
 682                 /* Rx interrupt disabled, Map interrupt only for writeback */
 683                 vf->nb_msix = 1;
 684                 if (vf->vf_res->vf_cap_flags &
 685                     VIRTCHNL_VF_OFFLOAD_WB_ON_ITR) {
 686                         /* If WB_ON_ITR supports, enable it */
 687                         vf->msix_base = IAVF_RX_VEC_START;
 688                         /* Set the ITR for index zero, to 2us to make sure that
 689                          * we leave time for aggregation to occur, but don't
 690                          * increase latency dramatically.
 691                          */
 692                         IAVF_WRITE_REG(hw,
 693                                        IAVF_VFINT_DYN_CTLN1(vf->msix_base - 1),
 694                                        (0 << IAVF_VFINT_DYN_CTLN1_ITR_INDX_SHIFT) |
 695                                        IAVF_VFINT_DYN_CTLN1_WB_ON_ITR_MASK |
 696                                        (2UL << IAVF_VFINT_DYN_CTLN1_INTERVAL_SHIFT));
 697                         /* debug - check for success! the return value
 698                          * should be 2, offset is 0x2800
 699                          */
 700                         /* IAVF_READ_REG(hw, IAVF_VFINT_ITRN1(0, 0)); */
 701                 } else {
 702                         /* If no WB_ON_ITR offload flags, need to set
 703                          * interrupt for descriptor write back.
 704                          */
 705                         vf->msix_base = IAVF_MISC_VEC_ID;
 706
 707                         /* set ITR to max */
 708                         interval = iavf_calc_itr_interval(
 709                                         IAVF_QUEUE_ITR_INTERVAL_MAX);
 710                         IAVF_WRITE_REG(hw, IAVF_VFINT_DYN_CTL01,
 711                                        IAVF_VFINT_DYN_CTL01_INTENA_MASK |
 712                                        (IAVF_ITR_INDEX_DEFAULT <<
 713                                         IAVF_VFINT_DYN_CTL01_ITR_INDX_SHIFT) |
 714                                        (interval <<
 715                                         IAVF_VFINT_DYN_CTL01_INTERVAL_SHIFT));
 716                 }
 717                 IAVF_WRITE_FLUSH(hw);
 718                 /* map all queues to the same interrupt */
 719                 for (i = 0; i < dev->data->nb_rx_queues; i++) {
 720                         qv_map[i].queue_id = i;
 721                         qv_map[i].vector_id = vf->msix_base;
 722                 }
 723                 vf->qv_map = qv_map;
 724         } else {
 725                 if (!rte_intr_allow_others(intr_handle)) {
 726                         vf->nb_msix = 1;
 727                         vf->msix_base = IAVF_MISC_VEC_ID;
 728                         for (i = 0; i < dev->data->nb_rx_queues; i++) {
 729                                 qv_map[i].queue_id = i;
 730                                 qv_map[i].vector_id = vf->msix_base;
 731                                 intr_handle->intr_vec[i] = IAVF_MISC_VEC_ID;
 732                         }
 733                         vf->qv_map = qv_map;
 734                         PMD_DRV_LOG(DEBUG,
 735                                     "vector %u are mapping to all Rx queues",
 736                                     vf->msix_base);
 737                 } else {
 738                         /* If Rx interrupt is reuquired, and we can use
 739                          * multi interrupts, then the vec is from 1
 740                          */
 741                         vf->nb_msix = RTE_MIN(intr_handle->nb_efd,
 742                                  (uint16_t)(vf->vf_res->max_vectors - 1));
 743                         vf->msix_base = IAVF_RX_VEC_START;
 744                         vec = IAVF_RX_VEC_START;
 745                         for (i = 0; i < dev->data->nb_rx_queues; i++) {
 746                                 qv_map[i].queue_id = i;
 747                                 qv_map[i].vector_id = vec;
 748                                 intr_handle->intr_vec[i] = vec++;
 749                                 if (vec >= vf->nb_msix + IAVF_RX_VEC_START)
 750                                         vec = IAVF_RX_VEC_START;
 751                         }
 752                         vf->qv_map = qv_map;
 753                         PMD_DRV_LOG(DEBUG,
 754                                     "%u vectors are mapping to %u Rx queues",
 755                                     vf->nb_msix, dev->data->nb_rx_queues);
 756                 }
 757         }
 758
 759         if (!vf->lv_enabled) {
 760                 if (iavf_config_irq_map(adapter)) {
 761                         PMD_DRV_LOG(ERR, "config interrupt mapping failed");
 762                         return -1;
 763                 }
 764         } else {
 765                 uint16_t num_qv_maps = dev->data->nb_rx_queues;
 766                 uint16_t index = 0;
 767
 768                 while (num_qv_maps > IAVF_IRQ_MAP_NUM_PER_BUF) {
 769                         if (iavf_config_irq_map_lv(adapter,
 770                                         IAVF_IRQ_MAP_NUM_PER_BUF, index)) {
 771                                 PMD_DRV_LOG(ERR, "config interrupt mapping for large VF failed");
 772                                 return -1;
 773                         }
 774                         num_qv_maps -= IAVF_IRQ_MAP_NUM_PER_BUF;
 775                         index += IAVF_IRQ_MAP_NUM_PER_BUF;
 776                 }
 777
 778                 if (iavf_config_irq_map_lv(adapter, num_qv_maps, index)) {
 779                         PMD_DRV_LOG(ERR, "config interrupt mapping for large VF failed");
 780                         return -1;
 781                 }
 782         }
 783         return 0;
 784 }
 785
 786 static int
 787 iavf_start_queues(struct rte_eth_dev *dev)
 788 {
 789         struct iavf_rx_queue *rxq;
 790         struct iavf_tx_queue *txq;
 791         int i;
 792
 793         for (i = 0; i < dev->data->nb_tx_queues; i++) {
 794                 txq = dev->data->tx_queues[i];
 795                 if (txq->tx_deferred_start)
 796                         continue;
 797                 if (iavf_dev_tx_queue_start(dev, i) != 0) {
 798                         PMD_DRV_LOG(ERR, "Fail to start queue %u", i);
 799                         return -1;
 800                 }
 801         }
 802
 803         for (i = 0; i < dev->data->nb_rx_queues; i++) {
 804                 rxq = dev->data->rx_queues[i];
 805                 if (rxq->rx_deferred_start)
 806                         continue;
 807                 if (iavf_dev_rx_queue_start(dev, i) != 0) {
 808                         PMD_DRV_LOG(ERR, "Fail to start queue %u", i);
 809                         return -1;
 810                 }
 811         }
 812
 813         return 0;
 814 }
 815
 816 static int
 817 iavf_dev_start(struct rte_eth_dev *dev)
 818 {
 819         struct iavf_adapter *adapter =
 820                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 821         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
 822         struct rte_intr_handle *intr_handle = dev->intr_handle;
 823         uint16_t num_queue_pairs;
 824         uint16_t index = 0;
 825
 826         PMD_INIT_FUNC_TRACE();
 827
 828         adapter->stopped = 0;
 829
 830         vf->max_pkt_len = dev->data->dev_conf.rxmode.max_rx_pkt_len;
 831         vf->num_queue_pairs = RTE_MAX(dev->data->nb_rx_queues,
 832                                       dev->data->nb_tx_queues);
 833         num_queue_pairs = vf->num_queue_pairs;
 834
 835         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_QOS)
 836                 if (iavf_get_qos_cap(adapter)) {
 837                         PMD_INIT_LOG(ERR, "Failed to get qos capability");
 838                         return -1;
 839                 }
 840
 841         if (iavf_init_queues(dev) != 0) {
 842                 PMD_DRV_LOG(ERR, "failed to do Queue init");
 843                 return -1;
 844         }
 845
 846         /* If needed, send configure queues msg multiple times to make the
 847          * adminq buffer length smaller than the 4K limitation.
 848          */
 849         while (num_queue_pairs > IAVF_CFG_Q_NUM_PER_BUF) {
 850                 if (iavf_configure_queues(adapter,
 851                                 IAVF_CFG_Q_NUM_PER_BUF, index) != 0) {
 852                         PMD_DRV_LOG(ERR, "configure queues failed");
 853                         goto err_queue;
 854                 }
 855                 num_queue_pairs -= IAVF_CFG_Q_NUM_PER_BUF;
 856                 index += IAVF_CFG_Q_NUM_PER_BUF;
 857         }
 858
 859         if (iavf_configure_queues(adapter, num_queue_pairs, index) != 0) {
 860                 PMD_DRV_LOG(ERR, "configure queues failed");
 861                 goto err_queue;
 862         }
 863
 864         if (iavf_config_rx_queues_irqs(dev, intr_handle) != 0) {
 865                 PMD_DRV_LOG(ERR, "configure irq failed");
 866                 goto err_queue;
 867         }
 868         /* re-enable intr again, because efd assign may change */
 869         if (dev->data->dev_conf.intr_conf.rxq != 0) {
 870                 rte_intr_disable(intr_handle);
 871                 rte_intr_enable(intr_handle);
 872         }
 873
 874         /* Set all mac addrs */
 875         iavf_add_del_all_mac_addr(adapter, true);
 876
 877         /* Set all multicast addresses */
 878         iavf_add_del_mc_addr_list(adapter, vf->mc_addrs, vf->mc_addrs_num,
 879                                   true);
 880
 881         if (iavf_start_queues(dev) != 0) {
 882                 PMD_DRV_LOG(ERR, "enable queues failed");
 883                 goto err_mac;
 884         }
 885
 886         return 0;
 887
 888 err_mac:
 889         iavf_add_del_all_mac_addr(adapter, false);
 890 err_queue:
 891         return -1;
 892 }
 893
 894 static int
 895 iavf_dev_stop(struct rte_eth_dev *dev)
 896 {
 897         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
 898         struct iavf_adapter *adapter =
 899                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 900         struct rte_intr_handle *intr_handle = dev->intr_handle;
 901
 902         PMD_INIT_FUNC_TRACE();
 903
 904         if (adapter->stopped == 1)
 905                 return 0;
 906
 907         iavf_stop_queues(dev);
 908
 909         /* Disable the interrupt for Rx */
 910         rte_intr_efd_disable(intr_handle);
 911         /* Rx interrupt vector mapping free */
 912         if (intr_handle->intr_vec) {
 913                 rte_free(intr_handle->intr_vec);
 914                 intr_handle->intr_vec = NULL;
 915         }
 916
 917         /* remove all mac addrs */
 918         iavf_add_del_all_mac_addr(adapter, false);
 919
 920         /* remove all multicast addresses */
 921         iavf_add_del_mc_addr_list(adapter, vf->mc_addrs, vf->mc_addrs_num,
 922                                   false);
 923
 924         adapter->stopped = 1;
 925         dev->data->dev_started = 0;
 926
 927         return 0;
 928 }
 929
 930 static int
 931 iavf_dev_info_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
 932 {
 933         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
 934
 935         dev_info->max_rx_queues = IAVF_MAX_NUM_QUEUES_LV;
 936         dev_info->max_tx_queues = IAVF_MAX_NUM_QUEUES_LV;
 937         dev_info->min_rx_bufsize = IAVF_BUF_SIZE_MIN;
 938         dev_info->max_rx_pktlen = IAVF_FRAME_SIZE_MAX;
 939         dev_info->max_mtu = dev_info->max_rx_pktlen - IAVF_ETH_OVERHEAD;
 940         dev_info->min_mtu = RTE_ETHER_MIN_MTU;
 941         dev_info->hash_key_size = vf->vf_res->rss_key_size;
 942         dev_info->reta_size = vf->vf_res->rss_lut_size;
 943         dev_info->flow_type_rss_offloads = IAVF_RSS_OFFLOAD_ALL;
 944         dev_info->max_mac_addrs = IAVF_NUM_MACADDR_MAX;
 945         dev_info->rx_offload_capa =
 946                 DEV_RX_OFFLOAD_VLAN_STRIP |
 947                 DEV_RX_OFFLOAD_QINQ_STRIP |
 948                 DEV_RX_OFFLOAD_IPV4_CKSUM |
 949                 DEV_RX_OFFLOAD_UDP_CKSUM |
 950                 DEV_RX_OFFLOAD_TCP_CKSUM |
 951                 DEV_RX_OFFLOAD_OUTER_IPV4_CKSUM |
 952                 DEV_RX_OFFLOAD_SCATTER |
 953                 DEV_RX_OFFLOAD_JUMBO_FRAME |
 954                 DEV_RX_OFFLOAD_VLAN_FILTER |
 955                 DEV_RX_OFFLOAD_RSS_HASH;
 956
 957         dev_info->tx_offload_capa =
 958                 DEV_TX_OFFLOAD_VLAN_INSERT |
 959                 DEV_TX_OFFLOAD_QINQ_INSERT |
 960                 DEV_TX_OFFLOAD_IPV4_CKSUM |
 961                 DEV_TX_OFFLOAD_UDP_CKSUM |
 962                 DEV_TX_OFFLOAD_TCP_CKSUM |
 963                 DEV_TX_OFFLOAD_SCTP_CKSUM |
 964                 DEV_TX_OFFLOAD_OUTER_IPV4_CKSUM |
 965                 DEV_TX_OFFLOAD_TCP_TSO |
 966                 DEV_TX_OFFLOAD_VXLAN_TNL_TSO |
 967                 DEV_TX_OFFLOAD_GRE_TNL_TSO |
 968                 DEV_TX_OFFLOAD_IPIP_TNL_TSO |
 969                 DEV_TX_OFFLOAD_GENEVE_TNL_TSO |
 970                 DEV_TX_OFFLOAD_MULTI_SEGS |
 971                 DEV_TX_OFFLOAD_MBUF_FAST_FREE;
 972
 973         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_CRC)
 974                 dev_info->rx_offload_capa |= DEV_RX_OFFLOAD_KEEP_CRC;
 975
 976         dev_info->default_rxconf = (struct rte_eth_rxconf) {
 977                 .rx_free_thresh = IAVF_DEFAULT_RX_FREE_THRESH,
 978                 .rx_drop_en = 0,
 979                 .offloads = 0,
 980         };
 981
 982         dev_info->default_txconf = (struct rte_eth_txconf) {
 983                 .tx_free_thresh = IAVF_DEFAULT_TX_FREE_THRESH,
 984                 .tx_rs_thresh = IAVF_DEFAULT_TX_RS_THRESH,
 985                 .offloads = 0,
 986         };
 987
 988         dev_info->rx_desc_lim = (struct rte_eth_desc_lim) {
 989                 .nb_max = IAVF_MAX_RING_DESC,
 990                 .nb_min = IAVF_MIN_RING_DESC,
 991                 .nb_align = IAVF_ALIGN_RING_DESC,
 992         };
 993
 994         dev_info->tx_desc_lim = (struct rte_eth_desc_lim) {
 995                 .nb_max = IAVF_MAX_RING_DESC,
 996                 .nb_min = IAVF_MIN_RING_DESC,
 997                 .nb_align = IAVF_ALIGN_RING_DESC,
 998         };
 999
1000         return 0;
1001 }
1002
1003 static const uint32_t *
1004 iavf_dev_supported_ptypes_get(struct rte_eth_dev *dev __rte_unused)
1005 {
1006         static const uint32_t ptypes[] = {
1007                 RTE_PTYPE_L2_ETHER,
1008                 RTE_PTYPE_L3_IPV4_EXT_UNKNOWN,
1009                 RTE_PTYPE_L4_FRAG,
1010                 RTE_PTYPE_L4_ICMP,
1011                 RTE_PTYPE_L4_NONFRAG,
1012                 RTE_PTYPE_L4_SCTP,
1013                 RTE_PTYPE_L4_TCP,
1014                 RTE_PTYPE_L4_UDP,
1015                 RTE_PTYPE_UNKNOWN
1016         };
1017         return ptypes;
1018 }
1019
1020 int
1021 iavf_dev_link_update(struct rte_eth_dev *dev,
1022                     __rte_unused int wait_to_complete)
1023 {
1024         struct rte_eth_link new_link;
1025         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
1026
1027         memset(&new_link, 0, sizeof(new_link));
1028
1029         /* Only read status info stored in VF, and the info is updated
1030          *  when receive LINK_CHANGE evnet from PF by Virtchnnl.
1031          */
1032         switch (vf->link_speed) {
1033         case 10:
1034                 new_link.link_speed = ETH_SPEED_NUM_10M;
1035                 break;
1036         case 100:
1037                 new_link.link_speed = ETH_SPEED_NUM_100M;
1038                 break;
1039         case 1000:
1040                 new_link.link_speed = ETH_SPEED_NUM_1G;
1041                 break;
1042         case 10000:
1043                 new_link.link_speed = ETH_SPEED_NUM_10G;
1044                 break;
1045         case 20000:
1046                 new_link.link_speed = ETH_SPEED_NUM_20G;
1047                 break;
1048         case 25000:
1049                 new_link.link_speed = ETH_SPEED_NUM_25G;
1050                 break;
1051         case 40000:
1052                 new_link.link_speed = ETH_SPEED_NUM_40G;
1053                 break;
1054         case 50000:
1055                 new_link.link_speed = ETH_SPEED_NUM_50G;
1056                 break;
1057         case 100000:
1058                 new_link.link_speed = ETH_SPEED_NUM_100G;
1059                 break;
1060         default:
1061                 new_link.link_speed = ETH_SPEED_NUM_NONE;
1062                 break;
1063         }
1064
1065         new_link.link_duplex = ETH_LINK_FULL_DUPLEX;
1066         new_link.link_status = vf->link_up ? ETH_LINK_UP :
1067                                              ETH_LINK_DOWN;
1068         new_link.link_autoneg = !(dev->data->dev_conf.link_speeds &
1069                                 ETH_LINK_SPEED_FIXED);
1070
1071         return rte_eth_linkstatus_set(dev, &new_link);
1072 }
1073
1074 static int
1075 iavf_dev_promiscuous_enable(struct rte_eth_dev *dev)
1076 {
1077         struct iavf_adapter *adapter =
1078                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1079         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
1080
1081         return iavf_config_promisc(adapter,
1082                                   true, vf->promisc_multicast_enabled);
1083 }
1084
1085 static int
1086 iavf_dev_promiscuous_disable(struct rte_eth_dev *dev)
1087 {
1088         struct iavf_adapter *adapter =
1089                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1090         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
1091
1092         return iavf_config_promisc(adapter,
1093                                   false, vf->promisc_multicast_enabled);
1094 }
1095
1096 static int
1097 iavf_dev_allmulticast_enable(struct rte_eth_dev *dev)
1098 {
1099         struct iavf_adapter *adapter =
1100                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1101         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
1102
1103         return iavf_config_promisc(adapter,
1104                                   vf->promisc_unicast_enabled, true);
1105 }
1106
1107 static int
1108 iavf_dev_allmulticast_disable(struct rte_eth_dev *dev)
1109 {
1110         struct iavf_adapter *adapter =
1111                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1112         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
1113
1114         return iavf_config_promisc(adapter,
1115                                   vf->promisc_unicast_enabled, false);
1116 }
1117
1118 static int
1119 iavf_dev_add_mac_addr(struct rte_eth_dev *dev, struct rte_ether_addr *addr,
1120                      __rte_unused uint32_t index,
1121                      __rte_unused uint32_t pool)
1122 {
1123         struct iavf_adapter *adapter =
1124                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1125         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
1126         int err;
1127
1128         if (rte_is_zero_ether_addr(addr)) {
1129                 PMD_DRV_LOG(ERR, "Invalid Ethernet Address");
1130                 return -EINVAL;
1131         }
1132
1133         err = iavf_add_del_eth_addr(adapter, addr, true, VIRTCHNL_ETHER_ADDR_EXTRA);
1134         if (err) {
1135                 PMD_DRV_LOG(ERR, "fail to add MAC address");
1136                 return -EIO;
1137         }
1138
1139         vf->mac_num++;
1140
1141         return 0;
1142 }
1143
1144 static void
1145 iavf_dev_del_mac_addr(struct rte_eth_dev *dev, uint32_t index)
1146 {
1147         struct iavf_adapter *adapter =
1148                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1149         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
1150         struct rte_ether_addr *addr;
1151         int err;
1152
1153         addr = &dev->data->mac_addrs[index];
1154
1155         err = iavf_add_del_eth_addr(adapter, addr, false, VIRTCHNL_ETHER_ADDR_EXTRA);
1156         if (err)
1157                 PMD_DRV_LOG(ERR, "fail to delete MAC address");
1158
1159         vf->mac_num--;
1160 }
1161
1162 static int
1163 iavf_dev_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
1164 {
1165         struct iavf_adapter *adapter =
1166                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1167         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
1168         int err;
1169
1170         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_VLAN_V2) {
1171                 err = iavf_add_del_vlan_v2(adapter, vlan_id, on);
1172                 if (err)
1173                         return -EIO;
1174                 return 0;
1175         }
1176
1177         if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_VLAN))
1178                 return -ENOTSUP;
1179
1180         err = iavf_add_del_vlan(adapter, vlan_id, on);
1181         if (err)
1182                 return -EIO;
1183         return 0;
1184 }
1185
1186 static void
1187 iavf_iterate_vlan_filters_v2(struct rte_eth_dev *dev, bool enable)
1188 {
1189         struct rte_vlan_filter_conf *vfc = &dev->data->vlan_filter_conf;
1190         struct iavf_adapter *adapter =
1191                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1192         uint32_t i, j;
1193         uint64_t ids;
1194
1195         for (i = 0; i < RTE_DIM(vfc->ids); i++) {
1196                 if (vfc->ids[i] == 0)
1197                         continue;
1198
1199                 ids = vfc->ids[i];
1200                 for (j = 0; ids != 0 && j < 64; j++, ids >>= 1) {
1201                         if (ids & 1)
1202                                 iavf_add_del_vlan_v2(adapter,
1203                                                      64 * i + j, enable);
1204                 }
1205         }
1206 }
1207
1208 static int
1209 iavf_dev_vlan_offload_set_v2(struct rte_eth_dev *dev, int mask)
1210 {
1211         struct rte_eth_rxmode *rxmode = &dev->data->dev_conf.rxmode;
1212         struct iavf_adapter *adapter =
1213                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1214         bool enable;
1215         int err;
1216
1217         if (mask & ETH_VLAN_FILTER_MASK) {
1218                 enable = !!(rxmode->offloads & DEV_RX_OFFLOAD_VLAN_FILTER);
1219
1220                 iavf_iterate_vlan_filters_v2(dev, enable);
1221         }
1222
1223         if (mask & ETH_VLAN_STRIP_MASK) {
1224                 enable = !!(rxmode->offloads & DEV_RX_OFFLOAD_VLAN_STRIP);
1225
1226                 err = iavf_config_vlan_strip_v2(adapter, enable);
1227                 /* If not support, the stripping is already disabled by PF */
1228                 if (err == -ENOTSUP && !enable)
1229                         err = 0;
1230                 if (err)
1231                         return -EIO;
1232         }
1233
1234         return 0;
1235 }
1236
1237 static int
1238 iavf_dev_vlan_offload_set(struct rte_eth_dev *dev, int mask)
1239 {
1240         struct iavf_adapter *adapter =
1241                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1242         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
1243         struct rte_eth_conf *dev_conf = &dev->data->dev_conf;
1244         int err;
1245
1246         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_VLAN_V2)
1247                 return iavf_dev_vlan_offload_set_v2(dev, mask);
1248
1249         if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_VLAN))
1250                 return -ENOTSUP;
1251
1252         /* Vlan stripping setting */
1253         if (mask & ETH_VLAN_STRIP_MASK) {
1254                 /* Enable or disable VLAN stripping */
1255                 if (dev_conf->rxmode.offloads & DEV_RX_OFFLOAD_VLAN_STRIP)
1256                         err = iavf_enable_vlan_strip(adapter);
1257                 else
1258                         err = iavf_disable_vlan_strip(adapter);
1259
1260                 if (err)
1261                         return -EIO;
1262         }
1263         return 0;
1264 }
1265
1266 static int
1267 iavf_dev_rss_reta_update(struct rte_eth_dev *dev,
1268                         struct rte_eth_rss_reta_entry64 *reta_conf,
1269                         uint16_t reta_size)
1270 {
1271         struct iavf_adapter *adapter =
1272                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1273         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
1274         uint8_t *lut;
1275         uint16_t i, idx, shift;
1276         int ret;
1277
1278         if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF))
1279                 return -ENOTSUP;
1280
1281         if (reta_size != vf->vf_res->rss_lut_size) {
1282                 PMD_DRV_LOG(ERR, "The size of hash lookup table configured "
1283                         "(%d) doesn't match the number of hardware can "
1284                         "support (%d)", reta_size, vf->vf_res->rss_lut_size);
1285                 return -EINVAL;
1286         }
1287
1288         lut = rte_zmalloc("rss_lut", reta_size, 0);
1289         if (!lut) {
1290                 PMD_DRV_LOG(ERR, "No memory can be allocated");
1291                 return -ENOMEM;
1292         }
1293         /* store the old lut table temporarily */
1294         rte_memcpy(lut, vf->rss_lut, reta_size);
1295
1296         for (i = 0; i < reta_size; i++) {
1297                 idx = i / RTE_RETA_GROUP_SIZE;
1298                 shift = i % RTE_RETA_GROUP_SIZE;
1299                 if (reta_conf[idx].mask & (1ULL << shift))
1300                         lut[i] = reta_conf[idx].reta[shift];
1301         }
1302
1303         rte_memcpy(vf->rss_lut, lut, reta_size);
1304         /* send virtchnnl ops to configure rss*/
1305         ret = iavf_configure_rss_lut(adapter);
1306         if (ret) /* revert back */
1307                 rte_memcpy(vf->rss_lut, lut, reta_size);
1308         rte_free(lut);
1309
1310         return ret;
1311 }
1312
1313 static int
1314 iavf_dev_rss_reta_query(struct rte_eth_dev *dev,
1315                        struct rte_eth_rss_reta_entry64 *reta_conf,
1316                        uint16_t reta_size)
1317 {
1318         struct iavf_adapter *adapter =
1319                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1320         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
1321         uint16_t i, idx, shift;
1322
1323         if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF))
1324                 return -ENOTSUP;
1325
1326         if (reta_size != vf->vf_res->rss_lut_size) {
1327                 PMD_DRV_LOG(ERR, "The size of hash lookup table configured "
1328                         "(%d) doesn't match the number of hardware can "
1329                         "support (%d)", reta_size, vf->vf_res->rss_lut_size);
1330                 return -EINVAL;
1331         }
1332
1333         for (i = 0; i < reta_size; i++) {
1334                 idx = i / RTE_RETA_GROUP_SIZE;
1335                 shift = i % RTE_RETA_GROUP_SIZE;
1336                 if (reta_conf[idx].mask & (1ULL << shift))
1337                         reta_conf[idx].reta[shift] = vf->rss_lut[i];
1338         }
1339
1340         return 0;
1341 }
1342
1343 static int
1344 iavf_set_rss_key(struct iavf_adapter *adapter, uint8_t *key, uint8_t key_len)
1345 {
1346         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
1347
1348         /* HENA setting, it is enabled by default, no change */
1349         if (!key || key_len == 0) {
1350                 PMD_DRV_LOG(DEBUG, "No key to be configured");
1351                 return 0;
1352         } else if (key_len != vf->vf_res->rss_key_size) {
1353                 PMD_DRV_LOG(ERR, "The size of hash key configured "
1354                         "(%d) doesn't match the size of hardware can "
1355                         "support (%d)", key_len,
1356                         vf->vf_res->rss_key_size);
1357                 return -EINVAL;
1358         }
1359
1360         rte_memcpy(vf->rss_key, key, key_len);
1361
1362         return iavf_configure_rss_key(adapter);
1363 }
1364
1365 static int
1366 iavf_dev_rss_hash_update(struct rte_eth_dev *dev,
1367                         struct rte_eth_rss_conf *rss_conf)
1368 {
1369         struct iavf_adapter *adapter =
1370                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1371         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
1372         int ret;
1373
1374         adapter->eth_dev->data->dev_conf.rx_adv_conf.rss_conf = *rss_conf;
1375
1376         if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF))
1377                 return -ENOTSUP;
1378
1379         /* Set hash key. */
1380         ret = iavf_set_rss_key(adapter, rss_conf->rss_key,
1381                                rss_conf->rss_key_len);
1382         if (ret)
1383                 return ret;
1384
1385         if (rss_conf->rss_hf == 0) {
1386                 vf->rss_hf = 0;
1387                 ret = iavf_set_hena(adapter, 0);
1388
1389                 /* It is a workaround, temporarily allow error to be returned
1390                  * due to possible lack of PF handling for hena = 0.
1391                  */
1392                 if (ret)
1393                         PMD_DRV_LOG(WARNING, "fail to clean existing RSS, lack PF support");
1394                 return 0;
1395         }
1396
1397         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_ADV_RSS_PF) {
1398                 /* Clear existing RSS. */
1399                 ret = iavf_set_hena(adapter, 0);
1400
1401                 /* It is a workaround, temporarily allow error to be returned
1402                  * due to possible lack of PF handling for hena = 0.
1403                  */
1404                 if (ret)
1405                         PMD_DRV_LOG(WARNING, "fail to clean existing RSS,"
1406                                     "lack PF support");
1407
1408                 /* Set new RSS configuration. */
1409                 ret = iavf_rss_hash_set(adapter, rss_conf->rss_hf, true);
1410                 if (ret) {
1411                         PMD_DRV_LOG(ERR, "fail to set new RSS");
1412                         return ret;
1413                 }
1414         } else {
1415                 iavf_config_rss_hf(adapter, rss_conf->rss_hf);
1416         }
1417
1418         return 0;
1419 }
1420
1421 static int
1422 iavf_dev_rss_hash_conf_get(struct rte_eth_dev *dev,
1423                           struct rte_eth_rss_conf *rss_conf)
1424 {
1425         struct iavf_adapter *adapter =
1426                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1427         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
1428
1429         if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF))
1430                 return -ENOTSUP;
1431
1432         rss_conf->rss_hf = vf->rss_hf;
1433
1434         if (!rss_conf->rss_key)
1435                 return 0;
1436
1437         rss_conf->rss_key_len = vf->vf_res->rss_key_size;
1438         rte_memcpy(rss_conf->rss_key, vf->rss_key, rss_conf->rss_key_len);
1439
1440         return 0;
1441 }
1442
1443 static int
1444 iavf_dev_mtu_set(struct rte_eth_dev *dev, uint16_t mtu)
1445 {
1446         uint32_t frame_size = mtu + IAVF_ETH_OVERHEAD;
1447         int ret = 0;
1448
1449         if (mtu < RTE_ETHER_MIN_MTU || frame_size > IAVF_FRAME_SIZE_MAX)
1450                 return -EINVAL;
1451
1452         /* mtu setting is forbidden if port is start */
1453         if (dev->data->dev_started) {
1454                 PMD_DRV_LOG(ERR, "port must be stopped before configuration");
1455                 return -EBUSY;
1456         }
1457
1458         if (frame_size > IAVF_ETH_MAX_LEN)
1459                 dev->data->dev_conf.rxmode.offloads |=
1460                                 DEV_RX_OFFLOAD_JUMBO_FRAME;
1461         else
1462                 dev->data->dev_conf.rxmode.offloads &=
1463                                 ~DEV_RX_OFFLOAD_JUMBO_FRAME;
1464
1465         dev->data->dev_conf.rxmode.max_rx_pkt_len = frame_size;
1466
1467         return ret;
1468 }
1469
1470 static int
1471 iavf_dev_set_default_mac_addr(struct rte_eth_dev *dev,
1472                              struct rte_ether_addr *mac_addr)
1473 {
1474         struct iavf_adapter *adapter =
1475                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1476         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(adapter);
1477         struct rte_ether_addr *old_addr;
1478         int ret;
1479
1480         old_addr = (struct rte_ether_addr *)hw->mac.addr;
1481
1482         if (rte_is_same_ether_addr(old_addr, mac_addr))
1483                 return 0;
1484
1485         ret = iavf_add_del_eth_addr(adapter, old_addr, false, VIRTCHNL_ETHER_ADDR_PRIMARY);
1486         if (ret)
1487                 PMD_DRV_LOG(ERR, "Fail to delete old MAC:"
1488                             " %02X:%02X:%02X:%02X:%02X:%02X",
1489                             old_addr->addr_bytes[0],
1490                             old_addr->addr_bytes[1],
1491                             old_addr->addr_bytes[2],
1492                             old_addr->addr_bytes[3],
1493                             old_addr->addr_bytes[4],
1494                             old_addr->addr_bytes[5]);
1495
1496         ret = iavf_add_del_eth_addr(adapter, mac_addr, true, VIRTCHNL_ETHER_ADDR_PRIMARY);
1497         if (ret)
1498                 PMD_DRV_LOG(ERR, "Fail to add new MAC:"
1499                             " %02X:%02X:%02X:%02X:%02X:%02X",
1500                             mac_addr->addr_bytes[0],
1501                             mac_addr->addr_bytes[1],
1502                             mac_addr->addr_bytes[2],
1503                             mac_addr->addr_bytes[3],
1504                             mac_addr->addr_bytes[4],
1505                             mac_addr->addr_bytes[5]);
1506
1507         if (ret)
1508                 return -EIO;
1509
1510         rte_ether_addr_copy(mac_addr, (struct rte_ether_addr *)hw->mac.addr);
1511         return 0;
1512 }
1513
1514 static void
1515 iavf_stat_update_48(uint64_t *offset, uint64_t *stat)
1516 {
1517         if (*stat >= *offset)
1518                 *stat = *stat - *offset;
1519         else
1520                 *stat = (uint64_t)((*stat +
1521                         ((uint64_t)1 << IAVF_48_BIT_WIDTH)) - *offset);
1522
1523         *stat &= IAVF_48_BIT_MASK;
1524 }
1525
1526 static void
1527 iavf_stat_update_32(uint64_t *offset, uint64_t *stat)
1528 {
1529         if (*stat >= *offset)
1530                 *stat = (uint64_t)(*stat - *offset);
1531         else
1532                 *stat = (uint64_t)((*stat +
1533                         ((uint64_t)1 << IAVF_32_BIT_WIDTH)) - *offset);
1534 }
1535
1536 static void
1537 iavf_update_stats(struct iavf_vsi *vsi, struct virtchnl_eth_stats *nes)
1538 {
1539         struct virtchnl_eth_stats *oes = &vsi->eth_stats_offset;
1540
1541         iavf_stat_update_48(&oes->rx_bytes, &nes->rx_bytes);
1542         iavf_stat_update_48(&oes->rx_unicast, &nes->rx_unicast);
1543         iavf_stat_update_48(&oes->rx_multicast, &nes->rx_multicast);
1544         iavf_stat_update_48(&oes->rx_broadcast, &nes->rx_broadcast);
1545         iavf_stat_update_32(&oes->rx_discards, &nes->rx_discards);
1546         iavf_stat_update_48(&oes->tx_bytes, &nes->tx_bytes);
1547         iavf_stat_update_48(&oes->tx_unicast, &nes->tx_unicast);
1548         iavf_stat_update_48(&oes->tx_multicast, &nes->tx_multicast);
1549         iavf_stat_update_48(&oes->tx_broadcast, &nes->tx_broadcast);
1550         iavf_stat_update_32(&oes->tx_errors, &nes->tx_errors);
1551         iavf_stat_update_32(&oes->tx_discards, &nes->tx_discards);
1552 }
1553
1554 static int
1555 iavf_dev_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
1556 {
1557         struct iavf_adapter *adapter =
1558                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1559         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
1560         struct iavf_vsi *vsi = &vf->vsi;
1561         struct virtchnl_eth_stats *pstats = NULL;
1562         int ret;
1563
1564         ret = iavf_query_stats(adapter, &pstats);
1565         if (ret == 0) {
1566                 uint8_t crc_stats_len = (dev->data->dev_conf.rxmode.offloads &
1567                                          DEV_RX_OFFLOAD_KEEP_CRC) ? 0 :
1568                                          RTE_ETHER_CRC_LEN;
1569                 iavf_update_stats(vsi, pstats);
1570                 stats->ipackets = pstats->rx_unicast + pstats->rx_multicast +
1571                                 pstats->rx_broadcast - pstats->rx_discards;
1572                 stats->opackets = pstats->tx_broadcast + pstats->tx_multicast +
1573                                                 pstats->tx_unicast;
1574                 stats->imissed = pstats->rx_discards;
1575                 stats->oerrors = pstats->tx_errors + pstats->tx_discards;
1576                 stats->ibytes = pstats->rx_bytes;
1577                 stats->ibytes -= stats->ipackets * crc_stats_len;
1578                 stats->obytes = pstats->tx_bytes;
1579         } else {
1580                 PMD_DRV_LOG(ERR, "Get statistics failed");
1581         }
1582         return ret;
1583 }
1584
1585 static int
1586 iavf_dev_stats_reset(struct rte_eth_dev *dev)
1587 {
1588         int ret;
1589         struct iavf_adapter *adapter =
1590                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1591         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
1592         struct iavf_vsi *vsi = &vf->vsi;
1593         struct virtchnl_eth_stats *pstats = NULL;
1594
1595         /* read stat values to clear hardware registers */
1596         ret = iavf_query_stats(adapter, &pstats);
1597         if (ret != 0)
1598                 return ret;
1599
1600         /* set stats offset base on current values */
1601         vsi->eth_stats_offset = *pstats;
1602
1603         return 0;
1604 }
1605
1606 static int iavf_dev_xstats_get_names(__rte_unused struct rte_eth_dev *dev,
1607                                       struct rte_eth_xstat_name *xstats_names,
1608                                       __rte_unused unsigned int limit)
1609 {
1610         unsigned int i;
1611
1612         if (xstats_names != NULL)
1613                 for (i = 0; i < IAVF_NB_XSTATS; i++) {
1614                         snprintf(xstats_names[i].name,
1615                                 sizeof(xstats_names[i].name),
1616                                 "%s", rte_iavf_stats_strings[i].name);
1617                 }
1618         return IAVF_NB_XSTATS;
1619 }
1620
1621 static int iavf_dev_xstats_get(struct rte_eth_dev *dev,
1622                                  struct rte_eth_xstat *xstats, unsigned int n)
1623 {
1624         int ret;
1625         unsigned int i;
1626         struct iavf_adapter *adapter =
1627                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1628         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
1629         struct iavf_vsi *vsi = &vf->vsi;
1630         struct virtchnl_eth_stats *pstats = NULL;
1631
1632         if (n < IAVF_NB_XSTATS)
1633                 return IAVF_NB_XSTATS;
1634
1635         ret = iavf_query_stats(adapter, &pstats);
1636         if (ret != 0)
1637                 return 0;
1638
1639         if (!xstats)
1640                 return 0;
1641
1642         iavf_update_stats(vsi, pstats);
1643
1644         /* loop over xstats array and values from pstats */
1645         for (i = 0; i < IAVF_NB_XSTATS; i++) {
1646                 xstats[i].id = i;
1647                 xstats[i].value = *(uint64_t *)(((char *)pstats) +
1648                         rte_iavf_stats_strings[i].offset);
1649         }
1650
1651         return IAVF_NB_XSTATS;
1652 }
1653
1654
1655 static int
1656 iavf_dev_rx_queue_intr_enable(struct rte_eth_dev *dev, uint16_t queue_id)
1657 {
1658         struct iavf_adapter *adapter =
1659                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1660         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1661         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(adapter);
1662         uint16_t msix_intr;
1663
1664         msix_intr = pci_dev->intr_handle.intr_vec[queue_id];
1665         if (msix_intr == IAVF_MISC_VEC_ID) {
1666                 PMD_DRV_LOG(INFO, "MISC is also enabled for control");
1667                 IAVF_WRITE_REG(hw, IAVF_VFINT_DYN_CTL01,
1668                                IAVF_VFINT_DYN_CTL01_INTENA_MASK |
1669                                IAVF_VFINT_DYN_CTL01_CLEARPBA_MASK |
1670                                IAVF_VFINT_DYN_CTL01_ITR_INDX_MASK);
1671         } else {
1672                 IAVF_WRITE_REG(hw,
1673                                IAVF_VFINT_DYN_CTLN1
1674                                 (msix_intr - IAVF_RX_VEC_START),
1675                                IAVF_VFINT_DYN_CTLN1_INTENA_MASK |
1676                                IAVF_VFINT_DYN_CTL01_CLEARPBA_MASK |
1677                                IAVF_VFINT_DYN_CTLN1_ITR_INDX_MASK);
1678         }
1679
1680         IAVF_WRITE_FLUSH(hw);
1681
1682         rte_intr_ack(&pci_dev->intr_handle);
1683
1684         return 0;
1685 }
1686
1687 static int
1688 iavf_dev_rx_queue_intr_disable(struct rte_eth_dev *dev, uint16_t queue_id)
1689 {
1690         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1691         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1692         uint16_t msix_intr;
1693
1694         msix_intr = pci_dev->intr_handle.intr_vec[queue_id];
1695         if (msix_intr == IAVF_MISC_VEC_ID) {
1696                 PMD_DRV_LOG(ERR, "MISC is used for control, cannot disable it");
1697                 return -EIO;
1698         }
1699
1700         IAVF_WRITE_REG(hw,
1701                       IAVF_VFINT_DYN_CTLN1(msix_intr - IAVF_RX_VEC_START),
1702                       0);
1703
1704         IAVF_WRITE_FLUSH(hw);
1705         return 0;
1706 }
1707
1708 static int
1709 iavf_check_vf_reset_done(struct iavf_hw *hw)
1710 {
1711         int i, reset;
1712
1713         for (i = 0; i < IAVF_RESET_WAIT_CNT; i++) {
1714                 reset = IAVF_READ_REG(hw, IAVF_VFGEN_RSTAT) &
1715                         IAVF_VFGEN_RSTAT_VFR_STATE_MASK;
1716                 reset = reset >> IAVF_VFGEN_RSTAT_VFR_STATE_SHIFT;
1717                 if (reset == VIRTCHNL_VFR_VFACTIVE ||
1718                     reset == VIRTCHNL_VFR_COMPLETED)
1719                         break;
1720                 rte_delay_ms(20);
1721         }
1722
1723         if (i >= IAVF_RESET_WAIT_CNT)
1724                 return -1;
1725
1726         return 0;
1727 }
1728
1729 static int
1730 iavf_lookup_proto_xtr_type(const char *flex_name)
1731 {
1732         static struct {
1733                 const char *name;
1734                 enum iavf_proto_xtr_type type;
1735         } xtr_type_map[] = {
1736                 { "vlan",      IAVF_PROTO_XTR_VLAN      },
1737                 { "ipv4",      IAVF_PROTO_XTR_IPV4      },
1738                 { "ipv6",      IAVF_PROTO_XTR_IPV6      },
1739                 { "ipv6_flow", IAVF_PROTO_XTR_IPV6_FLOW },
1740                 { "tcp",       IAVF_PROTO_XTR_TCP       },
1741                 { "ip_offset", IAVF_PROTO_XTR_IP_OFFSET },
1742         };
1743         uint32_t i;
1744
1745         for (i = 0; i < RTE_DIM(xtr_type_map); i++) {
1746                 if (strcmp(flex_name, xtr_type_map[i].name) == 0)
1747                         return xtr_type_map[i].type;
1748         }
1749
1750         PMD_DRV_LOG(ERR, "wrong proto_xtr type, "
1751                     "it should be: vlan|ipv4|ipv6|ipv6_flow|tcp|ip_offset");
1752
1753         return -1;
1754 }
1755
1756 /**
1757  * Parse elem, the elem could be single number/range or '(' ')' group
1758  * 1) A single number elem, it's just a simple digit. e.g. 9
1759  * 2) A single range elem, two digits with a '-' between. e.g. 2-6
1760  * 3) A group elem, combines multiple 1) or 2) with '( )'. e.g (0,2-4,6)
1761  *    Within group elem, '-' used for a range separator;
1762  *                       ',' used for a single number.
1763  */
1764 static int
1765 iavf_parse_queue_set(const char *input, int xtr_type,
1766                      struct iavf_devargs *devargs)
1767 {
1768         const char *str = input;
1769         char *end = NULL;
1770         uint32_t min, max;
1771         uint32_t idx;
1772
1773         while (isblank(*str))
1774                 str++;
1775
1776         if (!isdigit(*str) && *str != '(')
1777                 return -1;
1778
1779         /* process single number or single range of number */
1780         if (*str != '(') {
1781                 errno = 0;
1782                 idx = strtoul(str, &end, 10);
1783                 if (errno || !end || idx >= IAVF_MAX_QUEUE_NUM)
1784                         return -1;
1785
1786                 while (isblank(*end))
1787                         end++;
1788
1789                 min = idx;
1790                 max = idx;
1791
1792                 /* process single <number>-<number> */
1793                 if (*end == '-') {
1794                         end++;
1795                         while (isblank(*end))
1796                                 end++;
1797                         if (!isdigit(*end))
1798                                 return -1;
1799
1800                         errno = 0;
1801                         idx = strtoul(end, &end, 10);
1802                         if (errno || !end || idx >= IAVF_MAX_QUEUE_NUM)
1803                                 return -1;
1804
1805                         max = idx;
1806                         while (isblank(*end))
1807                                 end++;
1808                 }
1809
1810                 if (*end != ':')
1811                         return -1;
1812
1813                 for (idx = RTE_MIN(min, max);
1814                      idx <= RTE_MAX(min, max); idx++)
1815                         devargs->proto_xtr[idx] = xtr_type;
1816
1817                 return 0;
1818         }
1819
1820         /* process set within bracket */
1821         str++;
1822         while (isblank(*str))
1823                 str++;
1824         if (*str == '\0')
1825                 return -1;
1826
1827         min = IAVF_MAX_QUEUE_NUM;
1828         do {
1829                 /* go ahead to the first digit */
1830                 while (isblank(*str))
1831                         str++;
1832                 if (!isdigit(*str))
1833                         return -1;
1834
1835                 /* get the digit value */
1836                 errno = 0;
1837                 idx = strtoul(str, &end, 10);
1838                 if (errno || !end || idx >= IAVF_MAX_QUEUE_NUM)
1839                         return -1;
1840
1841                 /* go ahead to separator '-',',' and ')' */
1842                 while (isblank(*end))
1843                         end++;
1844                 if (*end == '-') {
1845                         if (min == IAVF_MAX_QUEUE_NUM)
1846                                 min = idx;
1847                         else /* avoid continuous '-' */
1848                                 return -1;
1849                 } else if (*end == ',' || *end == ')') {
1850                         max = idx;
1851                         if (min == IAVF_MAX_QUEUE_NUM)
1852                                 min = idx;
1853
1854                         for (idx = RTE_MIN(min, max);
1855                              idx <= RTE_MAX(min, max); idx++)
1856                                 devargs->proto_xtr[idx] = xtr_type;
1857
1858                         min = IAVF_MAX_QUEUE_NUM;
1859                 } else {
1860                         return -1;
1861                 }
1862
1863                 str = end + 1;
1864         } while (*end != ')' && *end != '\0');
1865
1866         return 0;
1867 }
1868
1869 static int
1870 iavf_parse_queue_proto_xtr(const char *queues, struct iavf_devargs *devargs)
1871 {
1872         const char *queue_start;
1873         uint32_t idx;
1874         int xtr_type;
1875         char flex_name[32];
1876
1877         while (isblank(*queues))
1878                 queues++;
1879
1880         if (*queues != '[') {
1881                 xtr_type = iavf_lookup_proto_xtr_type(queues);
1882                 if (xtr_type < 0)
1883                         return -1;
1884
1885                 devargs->proto_xtr_dflt = xtr_type;
1886
1887                 return 0;
1888         }
1889
1890         queues++;
1891         do {
1892                 while (isblank(*queues))
1893                         queues++;
1894                 if (*queues == '\0')
1895                         return -1;
1896
1897                 queue_start = queues;
1898
1899                 /* go across a complete bracket */
1900                 if (*queue_start == '(') {
1901                         queues += strcspn(queues, ")");
1902                         if (*queues != ')')
1903                                 return -1;
1904                 }
1905
1906                 /* scan the separator ':' */
1907                 queues += strcspn(queues, ":");
1908                 if (*queues++ != ':')
1909                         return -1;
1910                 while (isblank(*queues))
1911                         queues++;
1912
1913                 for (idx = 0; ; idx++) {
1914                         if (isblank(queues[idx]) ||
1915                             queues[idx] == ',' ||
1916                             queues[idx] == ']' ||
1917                             queues[idx] == '\0')
1918                                 break;
1919
1920                         if (idx > sizeof(flex_name) - 2)
1921                                 return -1;
1922
1923                         flex_name[idx] = queues[idx];
1924                 }
1925                 flex_name[idx] = '\0';
1926                 xtr_type = iavf_lookup_proto_xtr_type(flex_name);
1927                 if (xtr_type < 0)
1928                         return -1;
1929
1930                 queues += idx;
1931
1932                 while (isblank(*queues) || *queues == ',' || *queues == ']')
1933                         queues++;
1934
1935                 if (iavf_parse_queue_set(queue_start, xtr_type, devargs) < 0)
1936                         return -1;
1937         } while (*queues != '\0');
1938
1939         return 0;
1940 }
1941
1942 static int
1943 iavf_handle_proto_xtr_arg(__rte_unused const char *key, const char *value,
1944                           void *extra_args)
1945 {
1946         struct iavf_devargs *devargs = extra_args;
1947
1948         if (!value || !extra_args)
1949                 return -EINVAL;
1950
1951         if (iavf_parse_queue_proto_xtr(value, devargs) < 0) {
1952                 PMD_DRV_LOG(ERR, "the proto_xtr's parameter is wrong : '%s'",
1953                             value);
1954                 return -1;
1955         }
1956
1957         return 0;
1958 }
1959
1960 static int iavf_parse_devargs(struct rte_eth_dev *dev)
1961 {
1962         struct iavf_adapter *ad =
1963                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1964         struct rte_devargs *devargs = dev->device->devargs;
1965         struct rte_kvargs *kvlist;
1966         int ret;
1967
1968         if (!devargs)
1969                 return 0;
1970
1971         kvlist = rte_kvargs_parse(devargs->args, iavf_valid_args);
1972         if (!kvlist) {
1973                 PMD_INIT_LOG(ERR, "invalid kvargs key\n");
1974                 return -EINVAL;
1975         }
1976
1977         ad->devargs.proto_xtr_dflt = IAVF_PROTO_XTR_NONE;
1978         memset(ad->devargs.proto_xtr, IAVF_PROTO_XTR_NONE,
1979                sizeof(ad->devargs.proto_xtr));
1980
1981         ret = rte_kvargs_process(kvlist, IAVF_PROTO_XTR_ARG,
1982                                  &iavf_handle_proto_xtr_arg, &ad->devargs);
1983         if (ret)
1984                 goto bail;
1985
1986 bail:
1987         rte_kvargs_free(kvlist);
1988         return ret;
1989 }
1990
1991 static void
1992 iavf_init_proto_xtr(struct rte_eth_dev *dev)
1993 {
1994         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
1995         struct iavf_adapter *ad =
1996                         IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1997         const struct iavf_proto_xtr_ol *xtr_ol;
1998         bool proto_xtr_enable = false;
1999         int offset;
2000         uint16_t i;
2001
2002         vf->proto_xtr = rte_zmalloc("vf proto xtr",
2003                                     vf->vsi_res->num_queue_pairs, 0);
2004         if (unlikely(!(vf->proto_xtr))) {
2005                 PMD_DRV_LOG(ERR, "no memory for setting up proto_xtr's table");
2006                 return;
2007         }
2008
2009         for (i = 0; i < vf->vsi_res->num_queue_pairs; i++) {
2010                 vf->proto_xtr[i] = ad->devargs.proto_xtr[i] !=
2011                                         IAVF_PROTO_XTR_NONE ?
2012                                         ad->devargs.proto_xtr[i] :
2013                                         ad->devargs.proto_xtr_dflt;
2014
2015                 if (vf->proto_xtr[i] != IAVF_PROTO_XTR_NONE) {
2016                         uint8_t type = vf->proto_xtr[i];
2017
2018                         iavf_proto_xtr_params[type].required = true;
2019                         proto_xtr_enable = true;
2020                 }
2021         }
2022
2023         if (likely(!proto_xtr_enable))
2024                 return;
2025
2026         offset = rte_mbuf_dynfield_register(&iavf_proto_xtr_metadata_param);
2027         if (unlikely(offset == -1)) {
2028                 PMD_DRV_LOG(ERR,
2029                             "failed to extract protocol metadata, error %d",
2030                             -rte_errno);
2031                 return;
2032         }
2033
2034         PMD_DRV_LOG(DEBUG,
2035                     "proto_xtr metadata offset in mbuf is : %d",
2036                     offset);
2037         rte_pmd_ifd_dynfield_proto_xtr_metadata_offs = offset;
2038
2039         for (i = 0; i < RTE_DIM(iavf_proto_xtr_params); i++) {
2040                 xtr_ol = &iavf_proto_xtr_params[i];
2041
2042                 uint8_t rxdid = iavf_proto_xtr_type_to_rxdid((uint8_t)i);
2043
2044                 if (!xtr_ol->required)
2045                         continue;
2046
2047                 if (!(vf->supported_rxdid & BIT(rxdid))) {
2048                         PMD_DRV_LOG(ERR,
2049                                     "rxdid[%u] is not supported in hardware",
2050                                     rxdid);
2051                         rte_pmd_ifd_dynfield_proto_xtr_metadata_offs = -1;
2052                         break;
2053                 }
2054
2055                 offset = rte_mbuf_dynflag_register(&xtr_ol->param);
2056                 if (unlikely(offset == -1)) {
2057                         PMD_DRV_LOG(ERR,
2058                                     "failed to register proto_xtr offload '%s', error %d",
2059                                     xtr_ol->param.name, -rte_errno);
2060
2061                         rte_pmd_ifd_dynfield_proto_xtr_metadata_offs = -1;
2062                         break;
2063                 }
2064
2065                 PMD_DRV_LOG(DEBUG,
2066                             "proto_xtr offload '%s' offset in mbuf is : %d",
2067                             xtr_ol->param.name, offset);
2068                 *xtr_ol->ol_flag = 1ULL << offset;
2069         }
2070 }
2071
2072 static int
2073 iavf_init_vf(struct rte_eth_dev *dev)
2074 {
2075         int err, bufsz;
2076         struct iavf_adapter *adapter =
2077                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
2078         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2079         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
2080
2081         err = iavf_parse_devargs(dev);
2082         if (err) {
2083                 PMD_INIT_LOG(ERR, "Failed to parse devargs");
2084                 goto err;
2085         }
2086
2087         err = iavf_set_mac_type(hw);
2088         if (err) {
2089                 PMD_INIT_LOG(ERR, "set_mac_type failed: %d", err);
2090                 goto err;
2091         }
2092
2093         err = iavf_check_vf_reset_done(hw);
2094         if (err) {
2095                 PMD_INIT_LOG(ERR, "VF is still resetting");
2096                 goto err;
2097         }
2098
2099         iavf_init_adminq_parameter(hw);
2100         err = iavf_init_adminq(hw);
2101         if (err) {
2102                 PMD_INIT_LOG(ERR, "init_adminq failed: %d", err);
2103                 goto err;
2104         }
2105
2106         vf->aq_resp = rte_zmalloc("vf_aq_resp", IAVF_AQ_BUF_SZ, 0);
2107         if (!vf->aq_resp) {
2108                 PMD_INIT_LOG(ERR, "unable to allocate vf_aq_resp memory");
2109                 goto err_aq;
2110         }
2111         if (iavf_check_api_version(adapter) != 0) {
2112                 PMD_INIT_LOG(ERR, "check_api version failed");
2113                 goto err_api;
2114         }
2115
2116         bufsz = sizeof(struct virtchnl_vf_resource) +
2117                 (IAVF_MAX_VF_VSI * sizeof(struct virtchnl_vsi_resource));
2118         vf->vf_res = rte_zmalloc("vf_res", bufsz, 0);
2119         if (!vf->vf_res) {
2120                 PMD_INIT_LOG(ERR, "unable to allocate vf_res memory");
2121                 goto err_api;
2122         }
2123
2124         if (iavf_get_vf_resource(adapter) != 0) {
2125                 PMD_INIT_LOG(ERR, "iavf_get_vf_config failed");
2126                 goto err_alloc;
2127         }
2128         /* Allocate memort for RSS info */
2129         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF) {
2130                 vf->rss_key = rte_zmalloc("rss_key",
2131                                           vf->vf_res->rss_key_size, 0);
2132                 if (!vf->rss_key) {
2133                         PMD_INIT_LOG(ERR, "unable to allocate rss_key memory");
2134                         goto err_rss;
2135                 }
2136                 vf->rss_lut = rte_zmalloc("rss_lut",
2137                                           vf->vf_res->rss_lut_size, 0);
2138                 if (!vf->rss_lut) {
2139                         PMD_INIT_LOG(ERR, "unable to allocate rss_lut memory");
2140                         goto err_rss;
2141                 }
2142         }
2143
2144         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RX_FLEX_DESC) {
2145                 if (iavf_get_supported_rxdid(adapter) != 0) {
2146                         PMD_INIT_LOG(ERR, "failed to do get supported rxdid");
2147                         goto err_rss;
2148                 }
2149         }
2150
2151         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_VLAN_V2) {
2152                 if (iavf_get_vlan_offload_caps_v2(adapter) != 0) {
2153                         PMD_INIT_LOG(ERR, "failed to do get VLAN offload v2 capabilities");
2154                         goto err_rss;
2155                 }
2156         }
2157
2158         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_QOS) {
2159                 bufsz = sizeof(struct virtchnl_qos_cap_list) +
2160                         IAVF_MAX_TRAFFIC_CLASS *
2161                         sizeof(struct virtchnl_qos_cap_elem);
2162                 vf->qos_cap = rte_zmalloc("qos_cap", bufsz, 0);
2163                 if (!vf->qos_cap) {
2164                         PMD_INIT_LOG(ERR, "unable to allocate qos_cap memory");
2165                         goto err_rss;
2166                 }
2167                 iavf_tm_conf_init(dev);
2168         }
2169
2170         iavf_init_proto_xtr(dev);
2171
2172         return 0;
2173 err_rss:
2174         rte_free(vf->rss_key);
2175         rte_free(vf->rss_lut);
2176 err_alloc:
2177         rte_free(vf->qos_cap);
2178         rte_free(vf->vf_res);
2179         vf->vsi_res = NULL;
2180 err_api:
2181         rte_free(vf->aq_resp);
2182 err_aq:
2183         iavf_shutdown_adminq(hw);
2184 err:
2185         return -1;
2186 }
2187
2188 /* Enable default admin queue interrupt setting */
2189 static inline void
2190 iavf_enable_irq0(struct iavf_hw *hw)
2191 {
2192         /* Enable admin queue interrupt trigger */
2193         IAVF_WRITE_REG(hw, IAVF_VFINT_ICR0_ENA1,
2194                        IAVF_VFINT_ICR0_ENA1_ADMINQ_MASK);
2195
2196         IAVF_WRITE_REG(hw, IAVF_VFINT_DYN_CTL01,
2197                        IAVF_VFINT_DYN_CTL01_INTENA_MASK |
2198                        IAVF_VFINT_DYN_CTL01_CLEARPBA_MASK |
2199                        IAVF_VFINT_DYN_CTL01_ITR_INDX_MASK);
2200
2201         IAVF_WRITE_FLUSH(hw);
2202 }
2203
2204 static inline void
2205 iavf_disable_irq0(struct iavf_hw *hw)
2206 {
2207         /* Disable all interrupt types */
2208         IAVF_WRITE_REG(hw, IAVF_VFINT_ICR0_ENA1, 0);
2209         IAVF_WRITE_REG(hw, IAVF_VFINT_DYN_CTL01,
2210                        IAVF_VFINT_DYN_CTL01_ITR_INDX_MASK);
2211         IAVF_WRITE_FLUSH(hw);
2212 }
2213
2214 static void
2215 iavf_dev_interrupt_handler(void *param)
2216 {
2217         struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
2218         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2219
2220         iavf_disable_irq0(hw);
2221
2222         iavf_handle_virtchnl_msg(dev);
2223
2224         iavf_enable_irq0(hw);
2225 }
2226
2227 static int
2228 iavf_dev_flow_ops_get(struct rte_eth_dev *dev,
2229                       const struct rte_flow_ops **ops)
2230 {
2231         if (!dev)
2232                 return -EINVAL;
2233
2234         *ops = &iavf_flow_ops;
2235         return 0;
2236 }
2237
2238 static void
2239 iavf_default_rss_disable(struct iavf_adapter *adapter)
2240 {
2241         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
2242         int ret = 0;
2243
2244         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF) {
2245                 /* Set hena = 0 to ask PF to cleanup all existing RSS. */
2246                 ret = iavf_set_hena(adapter, 0);
2247                 if (ret)
2248                         /* It is a workaround, temporarily allow error to be
2249                          * returned due to possible lack of PF handling for
2250                          * hena = 0.
2251                          */
2252                         PMD_INIT_LOG(WARNING, "fail to disable default RSS,"
2253                                     "lack PF support");
2254         }
2255 }
2256
2257 static int
2258 iavf_dev_init(struct rte_eth_dev *eth_dev)
2259 {
2260         struct iavf_adapter *adapter =
2261                 IAVF_DEV_PRIVATE_TO_ADAPTER(eth_dev->data->dev_private);
2262         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(adapter);
2263         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
2264         int ret = 0;
2265
2266         PMD_INIT_FUNC_TRACE();
2267
2268         /* assign ops func pointer */
2269         eth_dev->dev_ops = &iavf_eth_dev_ops;
2270         eth_dev->rx_queue_count = iavf_dev_rxq_count;
2271         eth_dev->rx_descriptor_status = iavf_dev_rx_desc_status;
2272         eth_dev->tx_descriptor_status = iavf_dev_tx_desc_status;
2273         eth_dev->rx_pkt_burst = &iavf_recv_pkts;
2274         eth_dev->tx_pkt_burst = &iavf_xmit_pkts;
2275         eth_dev->tx_pkt_prepare = &iavf_prep_pkts;
2276
2277         /* For secondary processes, we don't initialise any further as primary
2278          * has already done this work. Only check if we need a different RX
2279          * and TX function.
2280          */
2281         if (rte_eal_process_type() != RTE_PROC_PRIMARY) {
2282                 iavf_set_rx_function(eth_dev);
2283                 iavf_set_tx_function(eth_dev);
2284                 return 0;
2285         }
2286         rte_eth_copy_pci_info(eth_dev, pci_dev);
2287         eth_dev->data->dev_flags |= RTE_ETH_DEV_AUTOFILL_QUEUE_XSTATS;
2288
2289         hw->vendor_id = pci_dev->id.vendor_id;
2290         hw->device_id = pci_dev->id.device_id;
2291         hw->subsystem_vendor_id = pci_dev->id.subsystem_vendor_id;
2292         hw->subsystem_device_id = pci_dev->id.subsystem_device_id;
2293         hw->bus.bus_id = pci_dev->addr.bus;
2294         hw->bus.device = pci_dev->addr.devid;
2295         hw->bus.func = pci_dev->addr.function;
2296         hw->hw_addr = (void *)pci_dev->mem_resource[0].addr;
2297         hw->back = IAVF_DEV_PRIVATE_TO_ADAPTER(eth_dev->data->dev_private);
2298         adapter->eth_dev = eth_dev;
2299         adapter->stopped = 1;
2300
2301         if (iavf_init_vf(eth_dev) != 0) {
2302                 PMD_INIT_LOG(ERR, "Init vf failed");
2303                 return -1;
2304         }
2305
2306         /* set default ptype table */
2307         adapter->ptype_tbl = iavf_get_default_ptype_table();
2308
2309         /* copy mac addr */
2310         eth_dev->data->mac_addrs = rte_zmalloc(
2311                 "iavf_mac", RTE_ETHER_ADDR_LEN * IAVF_NUM_MACADDR_MAX, 0);
2312         if (!eth_dev->data->mac_addrs) {
2313                 PMD_INIT_LOG(ERR, "Failed to allocate %d bytes needed to"
2314                              " store MAC addresses",
2315                              RTE_ETHER_ADDR_LEN * IAVF_NUM_MACADDR_MAX);
2316                 return -ENOMEM;
2317         }
2318         /* If the MAC address is not configured by host,
2319          * generate a random one.
2320          */
2321         if (!rte_is_valid_assigned_ether_addr(
2322                         (struct rte_ether_addr *)hw->mac.addr))
2323                 rte_eth_random_addr(hw->mac.addr);
2324         rte_ether_addr_copy((struct rte_ether_addr *)hw->mac.addr,
2325                         &eth_dev->data->mac_addrs[0]);
2326
2327         /* register callback func to eal lib */
2328         rte_intr_callback_register(&pci_dev->intr_handle,
2329                                    iavf_dev_interrupt_handler,
2330                                    (void *)eth_dev);
2331
2332         /* enable uio intr after callback register */
2333         rte_intr_enable(&pci_dev->intr_handle);
2334
2335         /* configure and enable device interrupt */
2336         iavf_enable_irq0(hw);
2337
2338         ret = iavf_flow_init(adapter);
2339         if (ret) {
2340                 PMD_INIT_LOG(ERR, "Failed to initialize flow");
2341                 return ret;
2342         }
2343
2344         iavf_default_rss_disable(adapter);
2345
2346         return 0;
2347 }
2348
2349 static int
2350 iavf_dev_close(struct rte_eth_dev *dev)
2351 {
2352         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2353         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
2354         struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
2355         struct iavf_adapter *adapter =
2356                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
2357         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
2358         int ret;
2359
2360         if (rte_eal_process_type() != RTE_PROC_PRIMARY)
2361                 return 0;
2362
2363         ret = iavf_dev_stop(dev);
2364
2365         iavf_flow_flush(dev, NULL);
2366         iavf_flow_uninit(adapter);
2367
2368         /*
2369          * disable promiscuous mode before reset vf
2370          * it is a workaround solution when work with kernel driver
2371          * and it is not the normal way
2372          */
2373         if (vf->promisc_unicast_enabled || vf->promisc_multicast_enabled)
2374                 iavf_config_promisc(adapter, false, false);
2375
2376         iavf_shutdown_adminq(hw);
2377         /* disable uio intr before callback unregister */
2378         rte_intr_disable(intr_handle);
2379
2380         /* unregister callback func from eal lib */
2381         rte_intr_callback_unregister(intr_handle,
2382                                      iavf_dev_interrupt_handler, dev);
2383         iavf_disable_irq0(hw);
2384
2385         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_QOS)
2386                 iavf_tm_conf_uninit(dev);
2387
2388         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF) {
2389                 if (vf->rss_lut) {
2390                         rte_free(vf->rss_lut);
2391                         vf->rss_lut = NULL;
2392                 }
2393                 if (vf->rss_key) {
2394                         rte_free(vf->rss_key);
2395                         vf->rss_key = NULL;
2396                 }
2397         }
2398
2399         rte_free(vf->vf_res);
2400         vf->vsi_res = NULL;
2401         vf->vf_res = NULL;
2402
2403         rte_free(vf->aq_resp);
2404         vf->aq_resp = NULL;
2405
2406         /*
2407          * If the VF is reset via VFLR, the device will be knocked out of bus
2408          * master mode, and the driver will fail to recover from the reset. Fix
2409          * this by enabling bus mastering after every reset. In a non-VFLR case,
2410          * the bus master bit will not be disabled, and this call will have no
2411          * effect.
2412          */
2413         if (vf->vf_reset && !rte_pci_set_bus_master(pci_dev, true))
2414                 vf->vf_reset = false;
2415
2416         return ret;
2417 }
2418
2419 static int
2420 iavf_dev_uninit(struct rte_eth_dev *dev)
2421 {
2422         if (rte_eal_process_type() != RTE_PROC_PRIMARY)
2423                 return -EPERM;
2424
2425         iavf_dev_close(dev);
2426
2427         return 0;
2428 }
2429
2430 /*
2431  * Reset VF device only to re-initialize resources in PMD layer
2432  */
2433 static int
2434 iavf_dev_reset(struct rte_eth_dev *dev)
2435 {
2436         int ret;
2437
2438         ret = iavf_dev_uninit(dev);
2439         if (ret)
2440                 return ret;
2441
2442         return iavf_dev_init(dev);
2443 }
2444
2445 static int
2446 iavf_dcf_cap_check_handler(__rte_unused const char *key,
2447                            const char *value, __rte_unused void *opaque)
2448 {
2449         if (strcmp(value, "dcf"))
2450                 return -1;
2451
2452         return 0;
2453 }
2454
2455 static int
2456 iavf_dcf_cap_selected(struct rte_devargs *devargs)
2457 {
2458         struct rte_kvargs *kvlist;
2459         const char *key = "cap";
2460         int ret = 0;
2461
2462         if (devargs == NULL)
2463                 return 0;
2464
2465         kvlist = rte_kvargs_parse(devargs->args, NULL);
2466         if (kvlist == NULL)
2467                 return 0;
2468
2469         if (!rte_kvargs_count(kvlist, key))
2470                 goto exit;
2471
2472         /* dcf capability selected when there's a key-value pair: cap=dcf */
2473         if (rte_kvargs_process(kvlist, key,
2474                                iavf_dcf_cap_check_handler, NULL) < 0)
2475                 goto exit;
2476
2477         ret = 1;
2478
2479 exit:
2480         rte_kvargs_free(kvlist);
2481         return ret;
2482 }
2483
2484 static int
2485 iavf_drv_i40evf_check_handler(__rte_unused const char *key,
2486                               const char *value, __rte_unused void *opaque)
2487 {
2488         if (strcmp(value, "i40evf"))
2489                 return -1;
2490
2491         return 0;
2492 }
2493
2494 static int
2495 iavf_drv_i40evf_selected(struct rte_devargs *devargs, uint16_t device_id)
2496 {
2497         struct rte_kvargs *kvlist;
2498         int ret = 0;
2499
2500         if (device_id != IAVF_DEV_ID_VF &&
2501             device_id != IAVF_DEV_ID_VF_HV &&
2502             device_id != IAVF_DEV_ID_X722_VF &&
2503             device_id != IAVF_DEV_ID_X722_A0_VF)
2504                 return 0;
2505
2506         if (devargs == NULL)
2507                 return 0;
2508
2509         kvlist = rte_kvargs_parse(devargs->args, NULL);
2510         if (kvlist == NULL)
2511                 return 0;
2512
2513         if (!rte_kvargs_count(kvlist, RTE_DEVARGS_KEY_DRIVER))
2514                 goto exit;
2515
2516         /* i40evf driver selected when there's a key-value pair:
2517          * driver=i40evf
2518          */
2519         if (rte_kvargs_process(kvlist, RTE_DEVARGS_KEY_DRIVER,
2520                                iavf_drv_i40evf_check_handler, NULL) < 0)
2521                 goto exit;
2522
2523         ret = 1;
2524
2525 exit:
2526         rte_kvargs_free(kvlist);
2527         return ret;
2528 }
2529
2530 static int eth_iavf_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
2531                              struct rte_pci_device *pci_dev)
2532 {
2533         if (iavf_dcf_cap_selected(pci_dev->device.devargs) ||
2534             iavf_drv_i40evf_selected(pci_dev->device.devargs,
2535                                      pci_dev->id.device_id))
2536                 return 1;
2537
2538         return rte_eth_dev_pci_generic_probe(pci_dev,
2539                 sizeof(struct iavf_adapter), iavf_dev_init);
2540 }
2541
2542 static int eth_iavf_pci_remove(struct rte_pci_device *pci_dev)
2543 {
2544         return rte_eth_dev_pci_generic_remove(pci_dev, iavf_dev_uninit);
2545 }
2546
2547 /* Adaptive virtual function driver struct */
2548 static struct rte_pci_driver rte_iavf_pmd = {
2549         .id_table = pci_id_iavf_map,
2550         .drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC,
2551         .probe = eth_iavf_pci_probe,
2552         .remove = eth_iavf_pci_remove,
2553 };
2554
2555 RTE_PMD_REGISTER_PCI(net_iavf, rte_iavf_pmd);
2556 RTE_PMD_REGISTER_PCI_TABLE(net_iavf, pci_id_iavf_map);
2557 RTE_PMD_REGISTER_KMOD_DEP(net_iavf, "* igb_uio | vfio-pci");
2558 RTE_PMD_REGISTER_PARAM_STRING(net_iavf, "cap=dcf driver=i40evf");
2559 RTE_LOG_REGISTER_SUFFIX(iavf_logtype_init, init, NOTICE);
2560 RTE_LOG_REGISTER_SUFFIX(iavf_logtype_driver, driver, NOTICE);
2561 #ifdef RTE_ETHDEV_DEBUG_RX
2562 RTE_LOG_REGISTER_SUFFIX(iavf_logtype_rx, rx, DEBUG);
2563 #endif
2564 #ifdef RTE_ETHDEV_DEBUG_TX
2565 RTE_LOG_REGISTER_SUFFIX(iavf_logtype_tx, tx, DEBUG);
2566 #endif