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;
 578
 579         buf_size = rte_pktmbuf_data_room_size(rxq->mp) - RTE_PKTMBUF_HEADROOM;
 580
 581         /* Calculate the maximum packet length allowed */
 582         max_pkt_len = RTE_MIN((uint32_t)
 583                         rxq->rx_buf_len * IAVF_MAX_CHAINED_RX_BUFFERS,
 584                         dev->data->dev_conf.rxmode.max_rx_pkt_len);
 585
 586         /* Check if the jumbo frame and maximum packet length are set
 587          * correctly.
 588          */
 589         if (dev->data->dev_conf.rxmode.offloads & DEV_RX_OFFLOAD_JUMBO_FRAME) {
 590                 if (max_pkt_len <= IAVF_ETH_MAX_LEN ||
 591                     max_pkt_len > IAVF_FRAME_SIZE_MAX) {
 592                         PMD_DRV_LOG(ERR, "maximum packet length must be "
 593                                     "larger than %u and smaller than %u, "
 594                                     "as jumbo frame is enabled",
 595                                     (uint32_t)IAVF_ETH_MAX_LEN,
 596                                     (uint32_t)IAVF_FRAME_SIZE_MAX);
 597                         return -EINVAL;
 598                 }
 599         } else {
 600                 if (max_pkt_len < RTE_ETHER_MIN_LEN ||
 601                     max_pkt_len > IAVF_ETH_MAX_LEN) {
 602                         PMD_DRV_LOG(ERR, "maximum packet length must be "
 603                                     "larger than %u and smaller than %u, "
 604                                     "as jumbo frame is disabled",
 605                                     (uint32_t)RTE_ETHER_MIN_LEN,
 606                                     (uint32_t)IAVF_ETH_MAX_LEN);
 607                         return -EINVAL;
 608                 }
 609         }
 610
 611         rxq->max_pkt_len = max_pkt_len;
 612         if ((dev_data->dev_conf.rxmode.offloads & DEV_RX_OFFLOAD_SCATTER) ||
 613             rxq->max_pkt_len > buf_size) {
 614                 dev_data->scattered_rx = 1;
 615         }
 616         IAVF_PCI_REG_WRITE(rxq->qrx_tail, rxq->nb_rx_desc - 1);
 617         IAVF_WRITE_FLUSH(hw);
 618
 619         return 0;
 620 }
 621
 622 static int
 623 iavf_init_queues(struct rte_eth_dev *dev)
 624 {
 625         struct iavf_rx_queue **rxq =
 626                 (struct iavf_rx_queue **)dev->data->rx_queues;
 627         int i, ret = IAVF_SUCCESS;
 628
 629         for (i = 0; i < dev->data->nb_rx_queues; i++) {
 630                 if (!rxq[i] || !rxq[i]->q_set)
 631                         continue;
 632                 ret = iavf_init_rxq(dev, rxq[i]);
 633                 if (ret != IAVF_SUCCESS)
 634                         break;
 635         }
 636         /* set rx/tx function to vector/scatter/single-segment
 637          * according to parameters
 638          */
 639         iavf_set_rx_function(dev);
 640         iavf_set_tx_function(dev);
 641
 642         return ret;
 643 }
 644
 645 static int iavf_config_rx_queues_irqs(struct rte_eth_dev *dev,
 646                                      struct rte_intr_handle *intr_handle)
 647 {
 648         struct iavf_adapter *adapter =
 649                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 650         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
 651         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(adapter);
 652         struct iavf_qv_map *qv_map;
 653         uint16_t interval, i;
 654         int vec;
 655
 656         if (rte_intr_cap_multiple(intr_handle) &&
 657             dev->data->dev_conf.intr_conf.rxq) {
 658                 if (rte_intr_efd_enable(intr_handle, dev->data->nb_rx_queues))
 659                         return -1;
 660         }
 661
 662         if (rte_intr_dp_is_en(intr_handle) && !intr_handle->intr_vec) {
 663                 intr_handle->intr_vec =
 664                         rte_zmalloc("intr_vec",
 665                                     dev->data->nb_rx_queues * sizeof(int), 0);
 666                 if (!intr_handle->intr_vec) {
 667                         PMD_DRV_LOG(ERR, "Failed to allocate %d rx intr_vec",
 668                                     dev->data->nb_rx_queues);
 669                         return -1;
 670                 }
 671         }
 672
 673         qv_map = rte_zmalloc("qv_map",
 674                 dev->data->nb_rx_queues * sizeof(struct iavf_qv_map), 0);
 675         if (!qv_map) {
 676                 PMD_DRV_LOG(ERR, "Failed to allocate %d queue-vector map",
 677                                 dev->data->nb_rx_queues);
 678                 return -1;
 679         }
 680
 681         if (!dev->data->dev_conf.intr_conf.rxq ||
 682             !rte_intr_dp_is_en(intr_handle)) {
 683                 /* Rx interrupt disabled, Map interrupt only for writeback */
 684                 vf->nb_msix = 1;
 685                 if (vf->vf_res->vf_cap_flags &
 686                     VIRTCHNL_VF_OFFLOAD_WB_ON_ITR) {
 687                         /* If WB_ON_ITR supports, enable it */
 688                         vf->msix_base = IAVF_RX_VEC_START;
 689                         /* Set the ITR for index zero, to 2us to make sure that
 690                          * we leave time for aggregation to occur, but don't
 691                          * increase latency dramatically.
 692                          */
 693                         IAVF_WRITE_REG(hw,
 694                                        IAVF_VFINT_DYN_CTLN1(vf->msix_base - 1),
 695                                        (0 << IAVF_VFINT_DYN_CTLN1_ITR_INDX_SHIFT) |
 696                                        IAVF_VFINT_DYN_CTLN1_WB_ON_ITR_MASK |
 697                                        (2UL << IAVF_VFINT_DYN_CTLN1_INTERVAL_SHIFT));
 698                         /* debug - check for success! the return value
 699                          * should be 2, offset is 0x2800
 700                          */
 701                         /* IAVF_READ_REG(hw, IAVF_VFINT_ITRN1(0, 0)); */
 702                 } else {
 703                         /* If no WB_ON_ITR offload flags, need to set
 704                          * interrupt for descriptor write back.
 705                          */
 706                         vf->msix_base = IAVF_MISC_VEC_ID;
 707
 708                         /* set ITR to max */
 709                         interval = iavf_calc_itr_interval(
 710                                         IAVF_QUEUE_ITR_INTERVAL_MAX);
 711                         IAVF_WRITE_REG(hw, IAVF_VFINT_DYN_CTL01,
 712                                        IAVF_VFINT_DYN_CTL01_INTENA_MASK |
 713                                        (IAVF_ITR_INDEX_DEFAULT <<
 714                                         IAVF_VFINT_DYN_CTL01_ITR_INDX_SHIFT) |
 715                                        (interval <<
 716                                         IAVF_VFINT_DYN_CTL01_INTERVAL_SHIFT));
 717                 }
 718                 IAVF_WRITE_FLUSH(hw);
 719                 /* map all queues to the same interrupt */
 720                 for (i = 0; i < dev->data->nb_rx_queues; i++) {
 721                         qv_map[i].queue_id = i;
 722                         qv_map[i].vector_id = vf->msix_base;
 723                 }
 724                 vf->qv_map = qv_map;
 725         } else {
 726                 if (!rte_intr_allow_others(intr_handle)) {
 727                         vf->nb_msix = 1;
 728                         vf->msix_base = IAVF_MISC_VEC_ID;
 729                         for (i = 0; i < dev->data->nb_rx_queues; i++) {
 730                                 qv_map[i].queue_id = i;
 731                                 qv_map[i].vector_id = vf->msix_base;
 732                                 intr_handle->intr_vec[i] = IAVF_MISC_VEC_ID;
 733                         }
 734                         vf->qv_map = qv_map;
 735                         PMD_DRV_LOG(DEBUG,
 736                                     "vector %u are mapping to all Rx queues",
 737                                     vf->msix_base);
 738                 } else {
 739                         /* If Rx interrupt is reuquired, and we can use
 740                          * multi interrupts, then the vec is from 1
 741                          */
 742                         vf->nb_msix = RTE_MIN(intr_handle->nb_efd,
 743                                  (uint16_t)(vf->vf_res->max_vectors - 1));
 744                         vf->msix_base = IAVF_RX_VEC_START;
 745                         vec = IAVF_RX_VEC_START;
 746                         for (i = 0; i < dev->data->nb_rx_queues; i++) {
 747                                 qv_map[i].queue_id = i;
 748                                 qv_map[i].vector_id = vec;
 749                                 intr_handle->intr_vec[i] = vec++;
 750                                 if (vec >= vf->nb_msix + IAVF_RX_VEC_START)
 751                                         vec = IAVF_RX_VEC_START;
 752                         }
 753                         vf->qv_map = qv_map;
 754                         PMD_DRV_LOG(DEBUG,
 755                                     "%u vectors are mapping to %u Rx queues",
 756                                     vf->nb_msix, dev->data->nb_rx_queues);
 757                 }
 758         }
 759
 760         if (!vf->lv_enabled) {
 761                 if (iavf_config_irq_map(adapter)) {
 762                         PMD_DRV_LOG(ERR, "config interrupt mapping failed");
 763                         return -1;
 764                 }
 765         } else {
 766                 uint16_t num_qv_maps = dev->data->nb_rx_queues;
 767                 uint16_t index = 0;
 768
 769                 while (num_qv_maps > IAVF_IRQ_MAP_NUM_PER_BUF) {
 770                         if (iavf_config_irq_map_lv(adapter,
 771                                         IAVF_IRQ_MAP_NUM_PER_BUF, index)) {
 772                                 PMD_DRV_LOG(ERR, "config interrupt mapping for large VF failed");
 773                                 return -1;
 774                         }
 775                         num_qv_maps -= IAVF_IRQ_MAP_NUM_PER_BUF;
 776                         index += IAVF_IRQ_MAP_NUM_PER_BUF;
 777                 }
 778
 779                 if (iavf_config_irq_map_lv(adapter, num_qv_maps, index)) {
 780                         PMD_DRV_LOG(ERR, "config interrupt mapping for large VF failed");
 781                         return -1;
 782                 }
 783         }
 784         return 0;
 785 }
 786
 787 static int
 788 iavf_start_queues(struct rte_eth_dev *dev)
 789 {
 790         struct iavf_rx_queue *rxq;
 791         struct iavf_tx_queue *txq;
 792         int i;
 793
 794         for (i = 0; i < dev->data->nb_tx_queues; i++) {
 795                 txq = dev->data->tx_queues[i];
 796                 if (txq->tx_deferred_start)
 797                         continue;
 798                 if (iavf_dev_tx_queue_start(dev, i) != 0) {
 799                         PMD_DRV_LOG(ERR, "Fail to start queue %u", i);
 800                         return -1;
 801                 }
 802         }
 803
 804         for (i = 0; i < dev->data->nb_rx_queues; i++) {
 805                 rxq = dev->data->rx_queues[i];
 806                 if (rxq->rx_deferred_start)
 807                         continue;
 808                 if (iavf_dev_rx_queue_start(dev, i) != 0) {
 809                         PMD_DRV_LOG(ERR, "Fail to start queue %u", i);
 810                         return -1;
 811                 }
 812         }
 813
 814         return 0;
 815 }
 816
 817 static int
 818 iavf_dev_start(struct rte_eth_dev *dev)
 819 {
 820         struct iavf_adapter *adapter =
 821                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 822         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
 823         struct rte_intr_handle *intr_handle = dev->intr_handle;
 824         uint16_t num_queue_pairs;
 825         uint16_t index = 0;
 826
 827         PMD_INIT_FUNC_TRACE();
 828
 829         adapter->stopped = 0;
 830
 831         vf->max_pkt_len = dev->data->dev_conf.rxmode.max_rx_pkt_len;
 832         vf->num_queue_pairs = RTE_MAX(dev->data->nb_rx_queues,
 833                                       dev->data->nb_tx_queues);
 834         num_queue_pairs = vf->num_queue_pairs;
 835
 836         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_QOS)
 837                 if (iavf_get_qos_cap(adapter)) {
 838                         PMD_INIT_LOG(ERR, "Failed to get qos capability");
 839                         return -1;
 840                 }
 841
 842         if (iavf_init_queues(dev) != 0) {
 843                 PMD_DRV_LOG(ERR, "failed to do Queue init");
 844                 return -1;
 845         }
 846
 847         /* If needed, send configure queues msg multiple times to make the
 848          * adminq buffer length smaller than the 4K limitation.
 849          */
 850         while (num_queue_pairs > IAVF_CFG_Q_NUM_PER_BUF) {
 851                 if (iavf_configure_queues(adapter,
 852                                 IAVF_CFG_Q_NUM_PER_BUF, index) != 0) {
 853                         PMD_DRV_LOG(ERR, "configure queues failed");
 854                         goto err_queue;
 855                 }
 856                 num_queue_pairs -= IAVF_CFG_Q_NUM_PER_BUF;
 857                 index += IAVF_CFG_Q_NUM_PER_BUF;
 858         }
 859
 860         if (iavf_configure_queues(adapter, num_queue_pairs, index) != 0) {
 861                 PMD_DRV_LOG(ERR, "configure queues failed");
 862                 goto err_queue;
 863         }
 864
 865         if (iavf_config_rx_queues_irqs(dev, intr_handle) != 0) {
 866                 PMD_DRV_LOG(ERR, "configure irq failed");
 867                 goto err_queue;
 868         }
 869         /* re-enable intr again, because efd assign may change */
 870         if (dev->data->dev_conf.intr_conf.rxq != 0) {
 871                 rte_intr_disable(intr_handle);
 872                 rte_intr_enable(intr_handle);
 873         }
 874
 875         /* Set all mac addrs */
 876         iavf_add_del_all_mac_addr(adapter, true);
 877
 878         /* Set all multicast addresses */
 879         iavf_add_del_mc_addr_list(adapter, vf->mc_addrs, vf->mc_addrs_num,
 880                                   true);
 881
 882         if (iavf_start_queues(dev) != 0) {
 883                 PMD_DRV_LOG(ERR, "enable queues failed");
 884                 goto err_mac;
 885         }
 886
 887         return 0;
 888
 889 err_mac:
 890         iavf_add_del_all_mac_addr(adapter, false);
 891 err_queue:
 892         return -1;
 893 }
 894
 895 static int
 896 iavf_dev_stop(struct rte_eth_dev *dev)
 897 {
 898         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
 899         struct iavf_adapter *adapter =
 900                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 901         struct rte_intr_handle *intr_handle = dev->intr_handle;
 902
 903         PMD_INIT_FUNC_TRACE();
 904
 905         if (adapter->stopped == 1)
 906                 return 0;
 907
 908         iavf_stop_queues(dev);
 909
 910         /* Disable the interrupt for Rx */
 911         rte_intr_efd_disable(intr_handle);
 912         /* Rx interrupt vector mapping free */
 913         if (intr_handle->intr_vec) {
 914                 rte_free(intr_handle->intr_vec);
 915                 intr_handle->intr_vec = NULL;
 916         }
 917
 918         /* remove all mac addrs */
 919         iavf_add_del_all_mac_addr(adapter, false);
 920
 921         /* remove all multicast addresses */
 922         iavf_add_del_mc_addr_list(adapter, vf->mc_addrs, vf->mc_addrs_num,
 923                                   false);
 924
 925         adapter->stopped = 1;
 926         dev->data->dev_started = 0;
 927
 928         return 0;
 929 }
 930
 931 static int
 932 iavf_dev_info_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
 933 {
 934         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
 935
 936         dev_info->max_rx_queues = IAVF_MAX_NUM_QUEUES_LV;
 937         dev_info->max_tx_queues = IAVF_MAX_NUM_QUEUES_LV;
 938         dev_info->min_rx_bufsize = IAVF_BUF_SIZE_MIN;
 939         dev_info->max_rx_pktlen = IAVF_FRAME_SIZE_MAX;
 940         dev_info->max_mtu = dev_info->max_rx_pktlen - IAVF_ETH_OVERHEAD;
 941         dev_info->min_mtu = RTE_ETHER_MIN_MTU;
 942         dev_info->hash_key_size = vf->vf_res->rss_key_size;
 943         dev_info->reta_size = vf->vf_res->rss_lut_size;
 944         dev_info->flow_type_rss_offloads = IAVF_RSS_OFFLOAD_ALL;
 945         dev_info->max_mac_addrs = IAVF_NUM_MACADDR_MAX;
 946         dev_info->rx_offload_capa =
 947                 DEV_RX_OFFLOAD_VLAN_STRIP |
 948                 DEV_RX_OFFLOAD_QINQ_STRIP |
 949                 DEV_RX_OFFLOAD_IPV4_CKSUM |
 950                 DEV_RX_OFFLOAD_UDP_CKSUM |
 951                 DEV_RX_OFFLOAD_TCP_CKSUM |
 952                 DEV_RX_OFFLOAD_OUTER_IPV4_CKSUM |
 953                 DEV_RX_OFFLOAD_SCATTER |
 954                 DEV_RX_OFFLOAD_JUMBO_FRAME |
 955                 DEV_RX_OFFLOAD_VLAN_FILTER |
 956                 DEV_RX_OFFLOAD_RSS_HASH;
 957
 958         dev_info->tx_offload_capa =
 959                 DEV_TX_OFFLOAD_VLAN_INSERT |
 960                 DEV_TX_OFFLOAD_QINQ_INSERT |
 961                 DEV_TX_OFFLOAD_IPV4_CKSUM |
 962                 DEV_TX_OFFLOAD_UDP_CKSUM |
 963                 DEV_TX_OFFLOAD_TCP_CKSUM |
 964                 DEV_TX_OFFLOAD_SCTP_CKSUM |
 965                 DEV_TX_OFFLOAD_OUTER_IPV4_CKSUM |
 966                 DEV_TX_OFFLOAD_TCP_TSO |
 967                 DEV_TX_OFFLOAD_VXLAN_TNL_TSO |
 968                 DEV_TX_OFFLOAD_GRE_TNL_TSO |
 969                 DEV_TX_OFFLOAD_IPIP_TNL_TSO |
 970                 DEV_TX_OFFLOAD_GENEVE_TNL_TSO |
 971                 DEV_TX_OFFLOAD_MULTI_SEGS |
 972                 DEV_TX_OFFLOAD_MBUF_FAST_FREE;
 973
 974         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_CRC)
 975                 dev_info->rx_offload_capa |= DEV_RX_OFFLOAD_KEEP_CRC;
 976
 977         dev_info->default_rxconf = (struct rte_eth_rxconf) {
 978                 .rx_free_thresh = IAVF_DEFAULT_RX_FREE_THRESH,
 979                 .rx_drop_en = 0,
 980                 .offloads = 0,
 981         };
 982
 983         dev_info->default_txconf = (struct rte_eth_txconf) {
 984                 .tx_free_thresh = IAVF_DEFAULT_TX_FREE_THRESH,
 985                 .tx_rs_thresh = IAVF_DEFAULT_TX_RS_THRESH,
 986                 .offloads = 0,
 987         };
 988
 989         dev_info->rx_desc_lim = (struct rte_eth_desc_lim) {
 990                 .nb_max = IAVF_MAX_RING_DESC,
 991                 .nb_min = IAVF_MIN_RING_DESC,
 992                 .nb_align = IAVF_ALIGN_RING_DESC,
 993         };
 994
 995         dev_info->tx_desc_lim = (struct rte_eth_desc_lim) {
 996                 .nb_max = IAVF_MAX_RING_DESC,
 997                 .nb_min = IAVF_MIN_RING_DESC,
 998                 .nb_align = IAVF_ALIGN_RING_DESC,
 999         };
1000
1001         return 0;
1002 }
1003
1004 static const uint32_t *
1005 iavf_dev_supported_ptypes_get(struct rte_eth_dev *dev __rte_unused)
1006 {
1007         static const uint32_t ptypes[] = {
1008                 RTE_PTYPE_L2_ETHER,
1009                 RTE_PTYPE_L3_IPV4_EXT_UNKNOWN,
1010                 RTE_PTYPE_L4_FRAG,
1011                 RTE_PTYPE_L4_ICMP,
1012                 RTE_PTYPE_L4_NONFRAG,
1013                 RTE_PTYPE_L4_SCTP,
1014                 RTE_PTYPE_L4_TCP,
1015                 RTE_PTYPE_L4_UDP,
1016                 RTE_PTYPE_UNKNOWN
1017         };
1018         return ptypes;
1019 }
1020
1021 int
1022 iavf_dev_link_update(struct rte_eth_dev *dev,
1023                     __rte_unused int wait_to_complete)
1024 {
1025         struct rte_eth_link new_link;
1026         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
1027
1028         memset(&new_link, 0, sizeof(new_link));
1029
1030         /* Only read status info stored in VF, and the info is updated
1031          *  when receive LINK_CHANGE evnet from PF by Virtchnnl.
1032          */
1033         switch (vf->link_speed) {
1034         case 10:
1035                 new_link.link_speed = ETH_SPEED_NUM_10M;
1036                 break;
1037         case 100:
1038                 new_link.link_speed = ETH_SPEED_NUM_100M;
1039                 break;
1040         case 1000:
1041                 new_link.link_speed = ETH_SPEED_NUM_1G;
1042                 break;
1043         case 10000:
1044                 new_link.link_speed = ETH_SPEED_NUM_10G;
1045                 break;
1046         case 20000:
1047                 new_link.link_speed = ETH_SPEED_NUM_20G;
1048                 break;
1049         case 25000:
1050                 new_link.link_speed = ETH_SPEED_NUM_25G;
1051                 break;
1052         case 40000:
1053                 new_link.link_speed = ETH_SPEED_NUM_40G;
1054                 break;
1055         case 50000:
1056                 new_link.link_speed = ETH_SPEED_NUM_50G;
1057                 break;
1058         case 100000:
1059                 new_link.link_speed = ETH_SPEED_NUM_100G;
1060                 break;
1061         default:
1062                 new_link.link_speed = ETH_SPEED_NUM_NONE;
1063                 break;
1064         }
1065
1066         new_link.link_duplex = ETH_LINK_FULL_DUPLEX;
1067         new_link.link_status = vf->link_up ? ETH_LINK_UP :
1068                                              ETH_LINK_DOWN;
1069         new_link.link_autoneg = !(dev->data->dev_conf.link_speeds &
1070                                 ETH_LINK_SPEED_FIXED);
1071
1072         return rte_eth_linkstatus_set(dev, &new_link);
1073 }
1074
1075 static int
1076 iavf_dev_promiscuous_enable(struct rte_eth_dev *dev)
1077 {
1078         struct iavf_adapter *adapter =
1079                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1080         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
1081
1082         return iavf_config_promisc(adapter,
1083                                   true, vf->promisc_multicast_enabled);
1084 }
1085
1086 static int
1087 iavf_dev_promiscuous_disable(struct rte_eth_dev *dev)
1088 {
1089         struct iavf_adapter *adapter =
1090                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1091         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
1092
1093         return iavf_config_promisc(adapter,
1094                                   false, vf->promisc_multicast_enabled);
1095 }
1096
1097 static int
1098 iavf_dev_allmulticast_enable(struct rte_eth_dev *dev)
1099 {
1100         struct iavf_adapter *adapter =
1101                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1102         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
1103
1104         return iavf_config_promisc(adapter,
1105                                   vf->promisc_unicast_enabled, true);
1106 }
1107
1108 static int
1109 iavf_dev_allmulticast_disable(struct rte_eth_dev *dev)
1110 {
1111         struct iavf_adapter *adapter =
1112                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1113         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
1114
1115         return iavf_config_promisc(adapter,
1116                                   vf->promisc_unicast_enabled, false);
1117 }
1118
1119 static int
1120 iavf_dev_add_mac_addr(struct rte_eth_dev *dev, struct rte_ether_addr *addr,
1121                      __rte_unused uint32_t index,
1122                      __rte_unused uint32_t pool)
1123 {
1124         struct iavf_adapter *adapter =
1125                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1126         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
1127         int err;
1128
1129         if (rte_is_zero_ether_addr(addr)) {
1130                 PMD_DRV_LOG(ERR, "Invalid Ethernet Address");
1131                 return -EINVAL;
1132         }
1133
1134         err = iavf_add_del_eth_addr(adapter, addr, true, VIRTCHNL_ETHER_ADDR_EXTRA);
1135         if (err) {
1136                 PMD_DRV_LOG(ERR, "fail to add MAC address");
1137                 return -EIO;
1138         }
1139
1140         vf->mac_num++;
1141
1142         return 0;
1143 }
1144
1145 static void
1146 iavf_dev_del_mac_addr(struct rte_eth_dev *dev, uint32_t index)
1147 {
1148         struct iavf_adapter *adapter =
1149                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1150         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
1151         struct rte_ether_addr *addr;
1152         int err;
1153
1154         addr = &dev->data->mac_addrs[index];
1155
1156         err = iavf_add_del_eth_addr(adapter, addr, false, VIRTCHNL_ETHER_ADDR_EXTRA);
1157         if (err)
1158                 PMD_DRV_LOG(ERR, "fail to delete MAC address");
1159
1160         vf->mac_num--;
1161 }
1162
1163 static int
1164 iavf_dev_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
1165 {
1166         struct iavf_adapter *adapter =
1167                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1168         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
1169         int err;
1170
1171         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_VLAN_V2) {
1172                 err = iavf_add_del_vlan_v2(adapter, vlan_id, on);
1173                 if (err)
1174                         return -EIO;
1175                 return 0;
1176         }
1177
1178         if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_VLAN))
1179                 return -ENOTSUP;
1180
1181         err = iavf_add_del_vlan(adapter, vlan_id, on);
1182         if (err)
1183                 return -EIO;
1184         return 0;
1185 }
1186
1187 static void
1188 iavf_iterate_vlan_filters_v2(struct rte_eth_dev *dev, bool enable)
1189 {
1190         struct rte_vlan_filter_conf *vfc = &dev->data->vlan_filter_conf;
1191         struct iavf_adapter *adapter =
1192                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1193         uint32_t i, j;
1194         uint64_t ids;
1195
1196         for (i = 0; i < RTE_DIM(vfc->ids); i++) {
1197                 if (vfc->ids[i] == 0)
1198                         continue;
1199
1200                 ids = vfc->ids[i];
1201                 for (j = 0; ids != 0 && j < 64; j++, ids >>= 1) {
1202                         if (ids & 1)
1203                                 iavf_add_del_vlan_v2(adapter,
1204                                                      64 * i + j, enable);
1205                 }
1206         }
1207 }
1208
1209 static int
1210 iavf_dev_vlan_offload_set_v2(struct rte_eth_dev *dev, int mask)
1211 {
1212         struct rte_eth_rxmode *rxmode = &dev->data->dev_conf.rxmode;
1213         struct iavf_adapter *adapter =
1214                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1215         bool enable;
1216         int err;
1217
1218         if (mask & ETH_VLAN_FILTER_MASK) {
1219                 enable = !!(rxmode->offloads & DEV_RX_OFFLOAD_VLAN_FILTER);
1220
1221                 iavf_iterate_vlan_filters_v2(dev, enable);
1222         }
1223
1224         if (mask & ETH_VLAN_STRIP_MASK) {
1225                 enable = !!(rxmode->offloads & DEV_RX_OFFLOAD_VLAN_STRIP);
1226
1227                 err = iavf_config_vlan_strip_v2(adapter, enable);
1228                 /* If not support, the stripping is already disabled by PF */
1229                 if (err == -ENOTSUP && !enable)
1230                         err = 0;
1231                 if (err)
1232                         return -EIO;
1233         }
1234
1235         return 0;
1236 }
1237
1238 static int
1239 iavf_dev_vlan_offload_set(struct rte_eth_dev *dev, int mask)
1240 {
1241         struct iavf_adapter *adapter =
1242                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1243         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
1244         struct rte_eth_conf *dev_conf = &dev->data->dev_conf;
1245         int err;
1246
1247         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_VLAN_V2)
1248                 return iavf_dev_vlan_offload_set_v2(dev, mask);
1249
1250         if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_VLAN))
1251                 return -ENOTSUP;
1252
1253         /* Vlan stripping setting */
1254         if (mask & ETH_VLAN_STRIP_MASK) {
1255                 /* Enable or disable VLAN stripping */
1256                 if (dev_conf->rxmode.offloads & DEV_RX_OFFLOAD_VLAN_STRIP)
1257                         err = iavf_enable_vlan_strip(adapter);
1258                 else
1259                         err = iavf_disable_vlan_strip(adapter);
1260
1261                 if (err)
1262                         return -EIO;
1263         }
1264         return 0;
1265 }
1266
1267 static int
1268 iavf_dev_rss_reta_update(struct rte_eth_dev *dev,
1269                         struct rte_eth_rss_reta_entry64 *reta_conf,
1270                         uint16_t reta_size)
1271 {
1272         struct iavf_adapter *adapter =
1273                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1274         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
1275         uint8_t *lut;
1276         uint16_t i, idx, shift;
1277         int ret;
1278
1279         if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF))
1280                 return -ENOTSUP;
1281
1282         if (reta_size != vf->vf_res->rss_lut_size) {
1283                 PMD_DRV_LOG(ERR, "The size of hash lookup table configured "
1284                         "(%d) doesn't match the number of hardware can "
1285                         "support (%d)", reta_size, vf->vf_res->rss_lut_size);
1286                 return -EINVAL;
1287         }
1288
1289         lut = rte_zmalloc("rss_lut", reta_size, 0);
1290         if (!lut) {
1291                 PMD_DRV_LOG(ERR, "No memory can be allocated");
1292                 return -ENOMEM;
1293         }
1294         /* store the old lut table temporarily */
1295         rte_memcpy(lut, vf->rss_lut, reta_size);
1296
1297         for (i = 0; i < reta_size; i++) {
1298                 idx = i / RTE_RETA_GROUP_SIZE;
1299                 shift = i % RTE_RETA_GROUP_SIZE;
1300                 if (reta_conf[idx].mask & (1ULL << shift))
1301                         lut[i] = reta_conf[idx].reta[shift];
1302         }
1303
1304         rte_memcpy(vf->rss_lut, lut, reta_size);
1305         /* send virtchnnl ops to configure rss*/
1306         ret = iavf_configure_rss_lut(adapter);
1307         if (ret) /* revert back */
1308                 rte_memcpy(vf->rss_lut, lut, reta_size);
1309         rte_free(lut);
1310
1311         return ret;
1312 }
1313
1314 static int
1315 iavf_dev_rss_reta_query(struct rte_eth_dev *dev,
1316                        struct rte_eth_rss_reta_entry64 *reta_conf,
1317                        uint16_t reta_size)
1318 {
1319         struct iavf_adapter *adapter =
1320                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1321         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
1322         uint16_t i, idx, shift;
1323
1324         if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF))
1325                 return -ENOTSUP;
1326
1327         if (reta_size != vf->vf_res->rss_lut_size) {
1328                 PMD_DRV_LOG(ERR, "The size of hash lookup table configured "
1329                         "(%d) doesn't match the number of hardware can "
1330                         "support (%d)", reta_size, vf->vf_res->rss_lut_size);
1331                 return -EINVAL;
1332         }
1333
1334         for (i = 0; i < reta_size; i++) {
1335                 idx = i / RTE_RETA_GROUP_SIZE;
1336                 shift = i % RTE_RETA_GROUP_SIZE;
1337                 if (reta_conf[idx].mask & (1ULL << shift))
1338                         reta_conf[idx].reta[shift] = vf->rss_lut[i];
1339         }
1340
1341         return 0;
1342 }
1343
1344 static int
1345 iavf_set_rss_key(struct iavf_adapter *adapter, uint8_t *key, uint8_t key_len)
1346 {
1347         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
1348
1349         /* HENA setting, it is enabled by default, no change */
1350         if (!key || key_len == 0) {
1351                 PMD_DRV_LOG(DEBUG, "No key to be configured");
1352                 return 0;
1353         } else if (key_len != vf->vf_res->rss_key_size) {
1354                 PMD_DRV_LOG(ERR, "The size of hash key configured "
1355                         "(%d) doesn't match the size of hardware can "
1356                         "support (%d)", key_len,
1357                         vf->vf_res->rss_key_size);
1358                 return -EINVAL;
1359         }
1360
1361         rte_memcpy(vf->rss_key, key, key_len);
1362
1363         return iavf_configure_rss_key(adapter);
1364 }
1365
1366 static int
1367 iavf_dev_rss_hash_update(struct rte_eth_dev *dev,
1368                         struct rte_eth_rss_conf *rss_conf)
1369 {
1370         struct iavf_adapter *adapter =
1371                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1372         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
1373         int ret;
1374
1375         adapter->eth_dev->data->dev_conf.rx_adv_conf.rss_conf = *rss_conf;
1376
1377         if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF))
1378                 return -ENOTSUP;
1379
1380         /* Set hash key. */
1381         ret = iavf_set_rss_key(adapter, rss_conf->rss_key,
1382                                rss_conf->rss_key_len);
1383         if (ret)
1384                 return ret;
1385
1386         if (rss_conf->rss_hf == 0) {
1387                 vf->rss_hf = 0;
1388                 ret = iavf_set_hena(adapter, 0);
1389
1390                 /* It is a workaround, temporarily allow error to be returned
1391                  * due to possible lack of PF handling for hena = 0.
1392                  */
1393                 if (ret)
1394                         PMD_DRV_LOG(WARNING, "fail to clean existing RSS, lack PF support");
1395                 return 0;
1396         }
1397
1398         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_ADV_RSS_PF) {
1399                 /* Clear existing RSS. */
1400                 ret = iavf_set_hena(adapter, 0);
1401
1402                 /* It is a workaround, temporarily allow error to be returned
1403                  * due to possible lack of PF handling for hena = 0.
1404                  */
1405                 if (ret)
1406                         PMD_DRV_LOG(WARNING, "fail to clean existing RSS,"
1407                                     "lack PF support");
1408
1409                 /* Set new RSS configuration. */
1410                 ret = iavf_rss_hash_set(adapter, rss_conf->rss_hf, true);
1411                 if (ret) {
1412                         PMD_DRV_LOG(ERR, "fail to set new RSS");
1413                         return ret;
1414                 }
1415         } else {
1416                 iavf_config_rss_hf(adapter, rss_conf->rss_hf);
1417         }
1418
1419         return 0;
1420 }
1421
1422 static int
1423 iavf_dev_rss_hash_conf_get(struct rte_eth_dev *dev,
1424                           struct rte_eth_rss_conf *rss_conf)
1425 {
1426         struct iavf_adapter *adapter =
1427                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1428         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
1429
1430         if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF))
1431                 return -ENOTSUP;
1432
1433         rss_conf->rss_hf = vf->rss_hf;
1434
1435         if (!rss_conf->rss_key)
1436                 return 0;
1437
1438         rss_conf->rss_key_len = vf->vf_res->rss_key_size;
1439         rte_memcpy(rss_conf->rss_key, vf->rss_key, rss_conf->rss_key_len);
1440
1441         return 0;
1442 }
1443
1444 static int
1445 iavf_dev_mtu_set(struct rte_eth_dev *dev, uint16_t mtu)
1446 {
1447         uint32_t frame_size = mtu + IAVF_ETH_OVERHEAD;
1448         int ret = 0;
1449
1450         if (mtu < RTE_ETHER_MIN_MTU || frame_size > IAVF_FRAME_SIZE_MAX)
1451                 return -EINVAL;
1452
1453         /* mtu setting is forbidden if port is start */
1454         if (dev->data->dev_started) {
1455                 PMD_DRV_LOG(ERR, "port must be stopped before configuration");
1456                 return -EBUSY;
1457         }
1458
1459         if (frame_size > IAVF_ETH_MAX_LEN)
1460                 dev->data->dev_conf.rxmode.offloads |=
1461                                 DEV_RX_OFFLOAD_JUMBO_FRAME;
1462         else
1463                 dev->data->dev_conf.rxmode.offloads &=
1464                                 ~DEV_RX_OFFLOAD_JUMBO_FRAME;
1465
1466         dev->data->dev_conf.rxmode.max_rx_pkt_len = frame_size;
1467
1468         return ret;
1469 }
1470
1471 static int
1472 iavf_dev_set_default_mac_addr(struct rte_eth_dev *dev,
1473                              struct rte_ether_addr *mac_addr)
1474 {
1475         struct iavf_adapter *adapter =
1476                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1477         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(adapter);
1478         struct rte_ether_addr *old_addr;
1479         int ret;
1480
1481         old_addr = (struct rte_ether_addr *)hw->mac.addr;
1482
1483         if (rte_is_same_ether_addr(old_addr, mac_addr))
1484                 return 0;
1485
1486         ret = iavf_add_del_eth_addr(adapter, old_addr, false, VIRTCHNL_ETHER_ADDR_PRIMARY);
1487         if (ret)
1488                 PMD_DRV_LOG(ERR, "Fail to delete old MAC:"
1489                             RTE_ETHER_ADDR_PRT_FMT,
1490                             old_addr->addr_bytes[0],
1491                             old_addr->addr_bytes[1],
1492                             old_addr->addr_bytes[2],
1493                             old_addr->addr_bytes[3],
1494                             old_addr->addr_bytes[4],
1495                             old_addr->addr_bytes[5]);
1496
1497         ret = iavf_add_del_eth_addr(adapter, mac_addr, true, VIRTCHNL_ETHER_ADDR_PRIMARY);
1498         if (ret)
1499                 PMD_DRV_LOG(ERR, "Fail to add new MAC:"
1500                             RTE_ETHER_ADDR_PRT_FMT,
1501                             mac_addr->addr_bytes[0],
1502                             mac_addr->addr_bytes[1],
1503                             mac_addr->addr_bytes[2],
1504                             mac_addr->addr_bytes[3],
1505                             mac_addr->addr_bytes[4],
1506                             mac_addr->addr_bytes[5]);
1507
1508         if (ret)
1509                 return -EIO;
1510
1511         rte_ether_addr_copy(mac_addr, (struct rte_ether_addr *)hw->mac.addr);
1512         return 0;
1513 }
1514
1515 static void
1516 iavf_stat_update_48(uint64_t *offset, uint64_t *stat)
1517 {
1518         if (*stat >= *offset)
1519                 *stat = *stat - *offset;
1520         else
1521                 *stat = (uint64_t)((*stat +
1522                         ((uint64_t)1 << IAVF_48_BIT_WIDTH)) - *offset);
1523
1524         *stat &= IAVF_48_BIT_MASK;
1525 }
1526
1527 static void
1528 iavf_stat_update_32(uint64_t *offset, uint64_t *stat)
1529 {
1530         if (*stat >= *offset)
1531                 *stat = (uint64_t)(*stat - *offset);
1532         else
1533                 *stat = (uint64_t)((*stat +
1534                         ((uint64_t)1 << IAVF_32_BIT_WIDTH)) - *offset);
1535 }
1536
1537 static void
1538 iavf_update_stats(struct iavf_vsi *vsi, struct virtchnl_eth_stats *nes)
1539 {
1540         struct virtchnl_eth_stats *oes = &vsi->eth_stats_offset;
1541
1542         iavf_stat_update_48(&oes->rx_bytes, &nes->rx_bytes);
1543         iavf_stat_update_48(&oes->rx_unicast, &nes->rx_unicast);
1544         iavf_stat_update_48(&oes->rx_multicast, &nes->rx_multicast);
1545         iavf_stat_update_48(&oes->rx_broadcast, &nes->rx_broadcast);
1546         iavf_stat_update_32(&oes->rx_discards, &nes->rx_discards);
1547         iavf_stat_update_48(&oes->tx_bytes, &nes->tx_bytes);
1548         iavf_stat_update_48(&oes->tx_unicast, &nes->tx_unicast);
1549         iavf_stat_update_48(&oes->tx_multicast, &nes->tx_multicast);
1550         iavf_stat_update_48(&oes->tx_broadcast, &nes->tx_broadcast);
1551         iavf_stat_update_32(&oes->tx_errors, &nes->tx_errors);
1552         iavf_stat_update_32(&oes->tx_discards, &nes->tx_discards);
1553 }
1554
1555 static int
1556 iavf_dev_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
1557 {
1558         struct iavf_adapter *adapter =
1559                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1560         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
1561         struct iavf_vsi *vsi = &vf->vsi;
1562         struct virtchnl_eth_stats *pstats = NULL;
1563         int ret;
1564
1565         ret = iavf_query_stats(adapter, &pstats);
1566         if (ret == 0) {
1567                 uint8_t crc_stats_len = (dev->data->dev_conf.rxmode.offloads &
1568                                          DEV_RX_OFFLOAD_KEEP_CRC) ? 0 :
1569                                          RTE_ETHER_CRC_LEN;
1570                 iavf_update_stats(vsi, pstats);
1571                 stats->ipackets = pstats->rx_unicast + pstats->rx_multicast +
1572                                 pstats->rx_broadcast - pstats->rx_discards;
1573                 stats->opackets = pstats->tx_broadcast + pstats->tx_multicast +
1574                                                 pstats->tx_unicast;
1575                 stats->imissed = pstats->rx_discards;
1576                 stats->oerrors = pstats->tx_errors + pstats->tx_discards;
1577                 stats->ibytes = pstats->rx_bytes;
1578                 stats->ibytes -= stats->ipackets * crc_stats_len;
1579                 stats->obytes = pstats->tx_bytes;
1580         } else {
1581                 PMD_DRV_LOG(ERR, "Get statistics failed");
1582         }
1583         return ret;
1584 }
1585
1586 static int
1587 iavf_dev_stats_reset(struct rte_eth_dev *dev)
1588 {
1589         int ret;
1590         struct iavf_adapter *adapter =
1591                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1592         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
1593         struct iavf_vsi *vsi = &vf->vsi;
1594         struct virtchnl_eth_stats *pstats = NULL;
1595
1596         /* read stat values to clear hardware registers */
1597         ret = iavf_query_stats(adapter, &pstats);
1598         if (ret != 0)
1599                 return ret;
1600
1601         /* set stats offset base on current values */
1602         vsi->eth_stats_offset = *pstats;
1603
1604         return 0;
1605 }
1606
1607 static int iavf_dev_xstats_get_names(__rte_unused struct rte_eth_dev *dev,
1608                                       struct rte_eth_xstat_name *xstats_names,
1609                                       __rte_unused unsigned int limit)
1610 {
1611         unsigned int i;
1612
1613         if (xstats_names != NULL)
1614                 for (i = 0; i < IAVF_NB_XSTATS; i++) {
1615                         snprintf(xstats_names[i].name,
1616                                 sizeof(xstats_names[i].name),
1617                                 "%s", rte_iavf_stats_strings[i].name);
1618                 }
1619         return IAVF_NB_XSTATS;
1620 }
1621
1622 static int iavf_dev_xstats_get(struct rte_eth_dev *dev,
1623                                  struct rte_eth_xstat *xstats, unsigned int n)
1624 {
1625         int ret;
1626         unsigned int i;
1627         struct iavf_adapter *adapter =
1628                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1629         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
1630         struct iavf_vsi *vsi = &vf->vsi;
1631         struct virtchnl_eth_stats *pstats = NULL;
1632
1633         if (n < IAVF_NB_XSTATS)
1634                 return IAVF_NB_XSTATS;
1635
1636         ret = iavf_query_stats(adapter, &pstats);
1637         if (ret != 0)
1638                 return 0;
1639
1640         if (!xstats)
1641                 return 0;
1642
1643         iavf_update_stats(vsi, pstats);
1644
1645         /* loop over xstats array and values from pstats */
1646         for (i = 0; i < IAVF_NB_XSTATS; i++) {
1647                 xstats[i].id = i;
1648                 xstats[i].value = *(uint64_t *)(((char *)pstats) +
1649                         rte_iavf_stats_strings[i].offset);
1650         }
1651
1652         return IAVF_NB_XSTATS;
1653 }
1654
1655
1656 static int
1657 iavf_dev_rx_queue_intr_enable(struct rte_eth_dev *dev, uint16_t queue_id)
1658 {
1659         struct iavf_adapter *adapter =
1660                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1661         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1662         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(adapter);
1663         uint16_t msix_intr;
1664
1665         msix_intr = pci_dev->intr_handle.intr_vec[queue_id];
1666         if (msix_intr == IAVF_MISC_VEC_ID) {
1667                 PMD_DRV_LOG(INFO, "MISC is also enabled for control");
1668                 IAVF_WRITE_REG(hw, IAVF_VFINT_DYN_CTL01,
1669                                IAVF_VFINT_DYN_CTL01_INTENA_MASK |
1670                                IAVF_VFINT_DYN_CTL01_CLEARPBA_MASK |
1671                                IAVF_VFINT_DYN_CTL01_ITR_INDX_MASK);
1672         } else {
1673                 IAVF_WRITE_REG(hw,
1674                                IAVF_VFINT_DYN_CTLN1
1675                                 (msix_intr - IAVF_RX_VEC_START),
1676                                IAVF_VFINT_DYN_CTLN1_INTENA_MASK |
1677                                IAVF_VFINT_DYN_CTL01_CLEARPBA_MASK |
1678                                IAVF_VFINT_DYN_CTLN1_ITR_INDX_MASK);
1679         }
1680
1681         IAVF_WRITE_FLUSH(hw);
1682
1683         rte_intr_ack(&pci_dev->intr_handle);
1684
1685         return 0;
1686 }
1687
1688 static int
1689 iavf_dev_rx_queue_intr_disable(struct rte_eth_dev *dev, uint16_t queue_id)
1690 {
1691         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1692         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1693         uint16_t msix_intr;
1694
1695         msix_intr = pci_dev->intr_handle.intr_vec[queue_id];
1696         if (msix_intr == IAVF_MISC_VEC_ID) {
1697                 PMD_DRV_LOG(ERR, "MISC is used for control, cannot disable it");
1698                 return -EIO;
1699         }
1700
1701         IAVF_WRITE_REG(hw,
1702                       IAVF_VFINT_DYN_CTLN1(msix_intr - IAVF_RX_VEC_START),
1703                       0);
1704
1705         IAVF_WRITE_FLUSH(hw);
1706         return 0;
1707 }
1708
1709 static int
1710 iavf_check_vf_reset_done(struct iavf_hw *hw)
1711 {
1712         int i, reset;
1713
1714         for (i = 0; i < IAVF_RESET_WAIT_CNT; i++) {
1715                 reset = IAVF_READ_REG(hw, IAVF_VFGEN_RSTAT) &
1716                         IAVF_VFGEN_RSTAT_VFR_STATE_MASK;
1717                 reset = reset >> IAVF_VFGEN_RSTAT_VFR_STATE_SHIFT;
1718                 if (reset == VIRTCHNL_VFR_VFACTIVE ||
1719                     reset == VIRTCHNL_VFR_COMPLETED)
1720                         break;
1721                 rte_delay_ms(20);
1722         }
1723
1724         if (i >= IAVF_RESET_WAIT_CNT)
1725                 return -1;
1726
1727         return 0;
1728 }
1729
1730 static int
1731 iavf_lookup_proto_xtr_type(const char *flex_name)
1732 {
1733         static struct {
1734                 const char *name;
1735                 enum iavf_proto_xtr_type type;
1736         } xtr_type_map[] = {
1737                 { "vlan",      IAVF_PROTO_XTR_VLAN      },
1738                 { "ipv4",      IAVF_PROTO_XTR_IPV4      },
1739                 { "ipv6",      IAVF_PROTO_XTR_IPV6      },
1740                 { "ipv6_flow", IAVF_PROTO_XTR_IPV6_FLOW },
1741                 { "tcp",       IAVF_PROTO_XTR_TCP       },
1742                 { "ip_offset", IAVF_PROTO_XTR_IP_OFFSET },
1743         };
1744         uint32_t i;
1745
1746         for (i = 0; i < RTE_DIM(xtr_type_map); i++) {
1747                 if (strcmp(flex_name, xtr_type_map[i].name) == 0)
1748                         return xtr_type_map[i].type;
1749         }
1750
1751         PMD_DRV_LOG(ERR, "wrong proto_xtr type, "
1752                     "it should be: vlan|ipv4|ipv6|ipv6_flow|tcp|ip_offset");
1753
1754         return -1;
1755 }
1756
1757 /**
1758  * Parse elem, the elem could be single number/range or '(' ')' group
1759  * 1) A single number elem, it's just a simple digit. e.g. 9
1760  * 2) A single range elem, two digits with a '-' between. e.g. 2-6
1761  * 3) A group elem, combines multiple 1) or 2) with '( )'. e.g (0,2-4,6)
1762  *    Within group elem, '-' used for a range separator;
1763  *                       ',' used for a single number.
1764  */
1765 static int
1766 iavf_parse_queue_set(const char *input, int xtr_type,
1767                      struct iavf_devargs *devargs)
1768 {
1769         const char *str = input;
1770         char *end = NULL;
1771         uint32_t min, max;
1772         uint32_t idx;
1773
1774         while (isblank(*str))
1775                 str++;
1776
1777         if (!isdigit(*str) && *str != '(')
1778                 return -1;
1779
1780         /* process single number or single range of number */
1781         if (*str != '(') {
1782                 errno = 0;
1783                 idx = strtoul(str, &end, 10);
1784                 if (errno || !end || idx >= IAVF_MAX_QUEUE_NUM)
1785                         return -1;
1786
1787                 while (isblank(*end))
1788                         end++;
1789
1790                 min = idx;
1791                 max = idx;
1792
1793                 /* process single <number>-<number> */
1794                 if (*end == '-') {
1795                         end++;
1796                         while (isblank(*end))
1797                                 end++;
1798                         if (!isdigit(*end))
1799                                 return -1;
1800
1801                         errno = 0;
1802                         idx = strtoul(end, &end, 10);
1803                         if (errno || !end || idx >= IAVF_MAX_QUEUE_NUM)
1804                                 return -1;
1805
1806                         max = idx;
1807                         while (isblank(*end))
1808                                 end++;
1809                 }
1810
1811                 if (*end != ':')
1812                         return -1;
1813
1814                 for (idx = RTE_MIN(min, max);
1815                      idx <= RTE_MAX(min, max); idx++)
1816                         devargs->proto_xtr[idx] = xtr_type;
1817
1818                 return 0;
1819         }
1820
1821         /* process set within bracket */
1822         str++;
1823         while (isblank(*str))
1824                 str++;
1825         if (*str == '\0')
1826                 return -1;
1827
1828         min = IAVF_MAX_QUEUE_NUM;
1829         do {
1830                 /* go ahead to the first digit */
1831                 while (isblank(*str))
1832                         str++;
1833                 if (!isdigit(*str))
1834                         return -1;
1835
1836                 /* get the digit value */
1837                 errno = 0;
1838                 idx = strtoul(str, &end, 10);
1839                 if (errno || !end || idx >= IAVF_MAX_QUEUE_NUM)
1840                         return -1;
1841
1842                 /* go ahead to separator '-',',' and ')' */
1843                 while (isblank(*end))
1844                         end++;
1845                 if (*end == '-') {
1846                         if (min == IAVF_MAX_QUEUE_NUM)
1847                                 min = idx;
1848                         else /* avoid continuous '-' */
1849                                 return -1;
1850                 } else if (*end == ',' || *end == ')') {
1851                         max = idx;
1852                         if (min == IAVF_MAX_QUEUE_NUM)
1853                                 min = idx;
1854
1855                         for (idx = RTE_MIN(min, max);
1856                              idx <= RTE_MAX(min, max); idx++)
1857                                 devargs->proto_xtr[idx] = xtr_type;
1858
1859                         min = IAVF_MAX_QUEUE_NUM;
1860                 } else {
1861                         return -1;
1862                 }
1863
1864                 str = end + 1;
1865         } while (*end != ')' && *end != '\0');
1866
1867         return 0;
1868 }
1869
1870 static int
1871 iavf_parse_queue_proto_xtr(const char *queues, struct iavf_devargs *devargs)
1872 {
1873         const char *queue_start;
1874         uint32_t idx;
1875         int xtr_type;
1876         char flex_name[32];
1877
1878         while (isblank(*queues))
1879                 queues++;
1880
1881         if (*queues != '[') {
1882                 xtr_type = iavf_lookup_proto_xtr_type(queues);
1883                 if (xtr_type < 0)
1884                         return -1;
1885
1886                 devargs->proto_xtr_dflt = xtr_type;
1887
1888                 return 0;
1889         }
1890
1891         queues++;
1892         do {
1893                 while (isblank(*queues))
1894                         queues++;
1895                 if (*queues == '\0')
1896                         return -1;
1897
1898                 queue_start = queues;
1899
1900                 /* go across a complete bracket */
1901                 if (*queue_start == '(') {
1902                         queues += strcspn(queues, ")");
1903                         if (*queues != ')')
1904                                 return -1;
1905                 }
1906
1907                 /* scan the separator ':' */
1908                 queues += strcspn(queues, ":");
1909                 if (*queues++ != ':')
1910                         return -1;
1911                 while (isblank(*queues))
1912                         queues++;
1913
1914                 for (idx = 0; ; idx++) {
1915                         if (isblank(queues[idx]) ||
1916                             queues[idx] == ',' ||
1917                             queues[idx] == ']' ||
1918                             queues[idx] == '\0')
1919                                 break;
1920
1921                         if (idx > sizeof(flex_name) - 2)
1922                                 return -1;
1923
1924                         flex_name[idx] = queues[idx];
1925                 }
1926                 flex_name[idx] = '\0';
1927                 xtr_type = iavf_lookup_proto_xtr_type(flex_name);
1928                 if (xtr_type < 0)
1929                         return -1;
1930
1931                 queues += idx;
1932
1933                 while (isblank(*queues) || *queues == ',' || *queues == ']')
1934                         queues++;
1935
1936                 if (iavf_parse_queue_set(queue_start, xtr_type, devargs) < 0)
1937                         return -1;
1938         } while (*queues != '\0');
1939
1940         return 0;
1941 }
1942
1943 static int
1944 iavf_handle_proto_xtr_arg(__rte_unused const char *key, const char *value,
1945                           void *extra_args)
1946 {
1947         struct iavf_devargs *devargs = extra_args;
1948
1949         if (!value || !extra_args)
1950                 return -EINVAL;
1951
1952         if (iavf_parse_queue_proto_xtr(value, devargs) < 0) {
1953                 PMD_DRV_LOG(ERR, "the proto_xtr's parameter is wrong : '%s'",
1954                             value);
1955                 return -1;
1956         }
1957
1958         return 0;
1959 }
1960
1961 static int iavf_parse_devargs(struct rte_eth_dev *dev)
1962 {
1963         struct iavf_adapter *ad =
1964                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1965         struct rte_devargs *devargs = dev->device->devargs;
1966         struct rte_kvargs *kvlist;
1967         int ret;
1968
1969         if (!devargs)
1970                 return 0;
1971
1972         kvlist = rte_kvargs_parse(devargs->args, iavf_valid_args);
1973         if (!kvlist) {
1974                 PMD_INIT_LOG(ERR, "invalid kvargs key\n");
1975                 return -EINVAL;
1976         }
1977
1978         ad->devargs.proto_xtr_dflt = IAVF_PROTO_XTR_NONE;
1979         memset(ad->devargs.proto_xtr, IAVF_PROTO_XTR_NONE,
1980                sizeof(ad->devargs.proto_xtr));
1981
1982         ret = rte_kvargs_process(kvlist, IAVF_PROTO_XTR_ARG,
1983                                  &iavf_handle_proto_xtr_arg, &ad->devargs);
1984         if (ret)
1985                 goto bail;
1986
1987 bail:
1988         rte_kvargs_free(kvlist);
1989         return ret;
1990 }
1991
1992 static void
1993 iavf_init_proto_xtr(struct rte_eth_dev *dev)
1994 {
1995         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
1996         struct iavf_adapter *ad =
1997                         IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1998         const struct iavf_proto_xtr_ol *xtr_ol;
1999         bool proto_xtr_enable = false;
2000         int offset;
2001         uint16_t i;
2002
2003         vf->proto_xtr = rte_zmalloc("vf proto xtr",
2004                                     vf->vsi_res->num_queue_pairs, 0);
2005         if (unlikely(!(vf->proto_xtr))) {
2006                 PMD_DRV_LOG(ERR, "no memory for setting up proto_xtr's table");
2007                 return;
2008         }
2009
2010         for (i = 0; i < vf->vsi_res->num_queue_pairs; i++) {
2011                 vf->proto_xtr[i] = ad->devargs.proto_xtr[i] !=
2012                                         IAVF_PROTO_XTR_NONE ?
2013                                         ad->devargs.proto_xtr[i] :
2014                                         ad->devargs.proto_xtr_dflt;
2015
2016                 if (vf->proto_xtr[i] != IAVF_PROTO_XTR_NONE) {
2017                         uint8_t type = vf->proto_xtr[i];
2018
2019                         iavf_proto_xtr_params[type].required = true;
2020                         proto_xtr_enable = true;
2021                 }
2022         }
2023
2024         if (likely(!proto_xtr_enable))
2025                 return;
2026
2027         offset = rte_mbuf_dynfield_register(&iavf_proto_xtr_metadata_param);
2028         if (unlikely(offset == -1)) {
2029                 PMD_DRV_LOG(ERR,
2030                             "failed to extract protocol metadata, error %d",
2031                             -rte_errno);
2032                 return;
2033         }
2034
2035         PMD_DRV_LOG(DEBUG,
2036                     "proto_xtr metadata offset in mbuf is : %d",
2037                     offset);
2038         rte_pmd_ifd_dynfield_proto_xtr_metadata_offs = offset;
2039
2040         for (i = 0; i < RTE_DIM(iavf_proto_xtr_params); i++) {
2041                 xtr_ol = &iavf_proto_xtr_params[i];
2042
2043                 uint8_t rxdid = iavf_proto_xtr_type_to_rxdid((uint8_t)i);
2044
2045                 if (!xtr_ol->required)
2046                         continue;
2047
2048                 if (!(vf->supported_rxdid & BIT(rxdid))) {
2049                         PMD_DRV_LOG(ERR,
2050                                     "rxdid[%u] is not supported in hardware",
2051                                     rxdid);
2052                         rte_pmd_ifd_dynfield_proto_xtr_metadata_offs = -1;
2053                         break;
2054                 }
2055
2056                 offset = rte_mbuf_dynflag_register(&xtr_ol->param);
2057                 if (unlikely(offset == -1)) {
2058                         PMD_DRV_LOG(ERR,
2059                                     "failed to register proto_xtr offload '%s', error %d",
2060                                     xtr_ol->param.name, -rte_errno);
2061
2062                         rte_pmd_ifd_dynfield_proto_xtr_metadata_offs = -1;
2063                         break;
2064                 }
2065
2066                 PMD_DRV_LOG(DEBUG,
2067                             "proto_xtr offload '%s' offset in mbuf is : %d",
2068                             xtr_ol->param.name, offset);
2069                 *xtr_ol->ol_flag = 1ULL << offset;
2070         }
2071 }
2072
2073 static int
2074 iavf_init_vf(struct rte_eth_dev *dev)
2075 {
2076         int err, bufsz;
2077         struct iavf_adapter *adapter =
2078                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
2079         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2080         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
2081
2082         err = iavf_parse_devargs(dev);
2083         if (err) {
2084                 PMD_INIT_LOG(ERR, "Failed to parse devargs");
2085                 goto err;
2086         }
2087
2088         err = iavf_set_mac_type(hw);
2089         if (err) {
2090                 PMD_INIT_LOG(ERR, "set_mac_type failed: %d", err);
2091                 goto err;
2092         }
2093
2094         err = iavf_check_vf_reset_done(hw);
2095         if (err) {
2096                 PMD_INIT_LOG(ERR, "VF is still resetting");
2097                 goto err;
2098         }
2099
2100         iavf_init_adminq_parameter(hw);
2101         err = iavf_init_adminq(hw);
2102         if (err) {
2103                 PMD_INIT_LOG(ERR, "init_adminq failed: %d", err);
2104                 goto err;
2105         }
2106
2107         vf->aq_resp = rte_zmalloc("vf_aq_resp", IAVF_AQ_BUF_SZ, 0);
2108         if (!vf->aq_resp) {
2109                 PMD_INIT_LOG(ERR, "unable to allocate vf_aq_resp memory");
2110                 goto err_aq;
2111         }
2112         if (iavf_check_api_version(adapter) != 0) {
2113                 PMD_INIT_LOG(ERR, "check_api version failed");
2114                 goto err_api;
2115         }
2116
2117         bufsz = sizeof(struct virtchnl_vf_resource) +
2118                 (IAVF_MAX_VF_VSI * sizeof(struct virtchnl_vsi_resource));
2119         vf->vf_res = rte_zmalloc("vf_res", bufsz, 0);
2120         if (!vf->vf_res) {
2121                 PMD_INIT_LOG(ERR, "unable to allocate vf_res memory");
2122                 goto err_api;
2123         }
2124
2125         if (iavf_get_vf_resource(adapter) != 0) {
2126                 PMD_INIT_LOG(ERR, "iavf_get_vf_config failed");
2127                 goto err_alloc;
2128         }
2129         /* Allocate memort for RSS info */
2130         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF) {
2131                 vf->rss_key = rte_zmalloc("rss_key",
2132                                           vf->vf_res->rss_key_size, 0);
2133                 if (!vf->rss_key) {
2134                         PMD_INIT_LOG(ERR, "unable to allocate rss_key memory");
2135                         goto err_rss;
2136                 }
2137                 vf->rss_lut = rte_zmalloc("rss_lut",
2138                                           vf->vf_res->rss_lut_size, 0);
2139                 if (!vf->rss_lut) {
2140                         PMD_INIT_LOG(ERR, "unable to allocate rss_lut memory");
2141                         goto err_rss;
2142                 }
2143         }
2144
2145         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RX_FLEX_DESC) {
2146                 if (iavf_get_supported_rxdid(adapter) != 0) {
2147                         PMD_INIT_LOG(ERR, "failed to do get supported rxdid");
2148                         goto err_rss;
2149                 }
2150         }
2151
2152         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_VLAN_V2) {
2153                 if (iavf_get_vlan_offload_caps_v2(adapter) != 0) {
2154                         PMD_INIT_LOG(ERR, "failed to do get VLAN offload v2 capabilities");
2155                         goto err_rss;
2156                 }
2157         }
2158
2159         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_QOS) {
2160                 bufsz = sizeof(struct virtchnl_qos_cap_list) +
2161                         IAVF_MAX_TRAFFIC_CLASS *
2162                         sizeof(struct virtchnl_qos_cap_elem);
2163                 vf->qos_cap = rte_zmalloc("qos_cap", bufsz, 0);
2164                 if (!vf->qos_cap) {
2165                         PMD_INIT_LOG(ERR, "unable to allocate qos_cap memory");
2166                         goto err_rss;
2167                 }
2168                 iavf_tm_conf_init(dev);
2169         }
2170
2171         iavf_init_proto_xtr(dev);
2172
2173         return 0;
2174 err_rss:
2175         rte_free(vf->rss_key);
2176         rte_free(vf->rss_lut);
2177 err_alloc:
2178         rte_free(vf->qos_cap);
2179         rte_free(vf->vf_res);
2180         vf->vsi_res = NULL;
2181 err_api:
2182         rte_free(vf->aq_resp);
2183 err_aq:
2184         iavf_shutdown_adminq(hw);
2185 err:
2186         return -1;
2187 }
2188
2189 /* Enable default admin queue interrupt setting */
2190 static inline void
2191 iavf_enable_irq0(struct iavf_hw *hw)
2192 {
2193         /* Enable admin queue interrupt trigger */
2194         IAVF_WRITE_REG(hw, IAVF_VFINT_ICR0_ENA1,
2195                        IAVF_VFINT_ICR0_ENA1_ADMINQ_MASK);
2196
2197         IAVF_WRITE_REG(hw, IAVF_VFINT_DYN_CTL01,
2198                        IAVF_VFINT_DYN_CTL01_INTENA_MASK |
2199                        IAVF_VFINT_DYN_CTL01_CLEARPBA_MASK |
2200                        IAVF_VFINT_DYN_CTL01_ITR_INDX_MASK);
2201
2202         IAVF_WRITE_FLUSH(hw);
2203 }
2204
2205 static inline void
2206 iavf_disable_irq0(struct iavf_hw *hw)
2207 {
2208         /* Disable all interrupt types */
2209         IAVF_WRITE_REG(hw, IAVF_VFINT_ICR0_ENA1, 0);
2210         IAVF_WRITE_REG(hw, IAVF_VFINT_DYN_CTL01,
2211                        IAVF_VFINT_DYN_CTL01_ITR_INDX_MASK);
2212         IAVF_WRITE_FLUSH(hw);
2213 }
2214
2215 static void
2216 iavf_dev_interrupt_handler(void *param)
2217 {
2218         struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
2219         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2220
2221         iavf_disable_irq0(hw);
2222
2223         iavf_handle_virtchnl_msg(dev);
2224
2225         iavf_enable_irq0(hw);
2226 }
2227
2228 static int
2229 iavf_dev_flow_ops_get(struct rte_eth_dev *dev,
2230                       const struct rte_flow_ops **ops)
2231 {
2232         if (!dev)
2233                 return -EINVAL;
2234
2235         *ops = &iavf_flow_ops;
2236         return 0;
2237 }
2238
2239 static void
2240 iavf_default_rss_disable(struct iavf_adapter *adapter)
2241 {
2242         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
2243         int ret = 0;
2244
2245         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF) {
2246                 /* Set hena = 0 to ask PF to cleanup all existing RSS. */
2247                 ret = iavf_set_hena(adapter, 0);
2248                 if (ret)
2249                         /* It is a workaround, temporarily allow error to be
2250                          * returned due to possible lack of PF handling for
2251                          * hena = 0.
2252                          */
2253                         PMD_INIT_LOG(WARNING, "fail to disable default RSS,"
2254                                     "lack PF support");
2255         }
2256 }
2257
2258 static int
2259 iavf_dev_init(struct rte_eth_dev *eth_dev)
2260 {
2261         struct iavf_adapter *adapter =
2262                 IAVF_DEV_PRIVATE_TO_ADAPTER(eth_dev->data->dev_private);
2263         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(adapter);
2264         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
2265         int ret = 0;
2266
2267         PMD_INIT_FUNC_TRACE();
2268
2269         /* assign ops func pointer */
2270         eth_dev->dev_ops = &iavf_eth_dev_ops;
2271         eth_dev->rx_queue_count = iavf_dev_rxq_count;
2272         eth_dev->rx_descriptor_status = iavf_dev_rx_desc_status;
2273         eth_dev->tx_descriptor_status = iavf_dev_tx_desc_status;
2274         eth_dev->rx_pkt_burst = &iavf_recv_pkts;
2275         eth_dev->tx_pkt_burst = &iavf_xmit_pkts;
2276         eth_dev->tx_pkt_prepare = &iavf_prep_pkts;
2277
2278         /* For secondary processes, we don't initialise any further as primary
2279          * has already done this work. Only check if we need a different RX
2280          * and TX function.
2281          */
2282         if (rte_eal_process_type() != RTE_PROC_PRIMARY) {
2283                 iavf_set_rx_function(eth_dev);
2284                 iavf_set_tx_function(eth_dev);
2285                 return 0;
2286         }
2287         rte_eth_copy_pci_info(eth_dev, pci_dev);
2288         eth_dev->data->dev_flags |= RTE_ETH_DEV_AUTOFILL_QUEUE_XSTATS;
2289
2290         hw->vendor_id = pci_dev->id.vendor_id;
2291         hw->device_id = pci_dev->id.device_id;
2292         hw->subsystem_vendor_id = pci_dev->id.subsystem_vendor_id;
2293         hw->subsystem_device_id = pci_dev->id.subsystem_device_id;
2294         hw->bus.bus_id = pci_dev->addr.bus;
2295         hw->bus.device = pci_dev->addr.devid;
2296         hw->bus.func = pci_dev->addr.function;
2297         hw->hw_addr = (void *)pci_dev->mem_resource[0].addr;
2298         hw->back = IAVF_DEV_PRIVATE_TO_ADAPTER(eth_dev->data->dev_private);
2299         adapter->eth_dev = eth_dev;
2300         adapter->stopped = 1;
2301
2302         if (iavf_init_vf(eth_dev) != 0) {
2303                 PMD_INIT_LOG(ERR, "Init vf failed");
2304                 return -1;
2305         }
2306
2307         /* set default ptype table */
2308         adapter->ptype_tbl = iavf_get_default_ptype_table();
2309
2310         /* copy mac addr */
2311         eth_dev->data->mac_addrs = rte_zmalloc(
2312                 "iavf_mac", RTE_ETHER_ADDR_LEN * IAVF_NUM_MACADDR_MAX, 0);
2313         if (!eth_dev->data->mac_addrs) {
2314                 PMD_INIT_LOG(ERR, "Failed to allocate %d bytes needed to"
2315                              " store MAC addresses",
2316                              RTE_ETHER_ADDR_LEN * IAVF_NUM_MACADDR_MAX);
2317                 return -ENOMEM;
2318         }
2319         /* If the MAC address is not configured by host,
2320          * generate a random one.
2321          */
2322         if (!rte_is_valid_assigned_ether_addr(
2323                         (struct rte_ether_addr *)hw->mac.addr))
2324                 rte_eth_random_addr(hw->mac.addr);
2325         rte_ether_addr_copy((struct rte_ether_addr *)hw->mac.addr,
2326                         &eth_dev->data->mac_addrs[0]);
2327
2328         /* register callback func to eal lib */
2329         rte_intr_callback_register(&pci_dev->intr_handle,
2330                                    iavf_dev_interrupt_handler,
2331                                    (void *)eth_dev);
2332
2333         /* enable uio intr after callback register */
2334         rte_intr_enable(&pci_dev->intr_handle);
2335
2336         /* configure and enable device interrupt */
2337         iavf_enable_irq0(hw);
2338
2339         ret = iavf_flow_init(adapter);
2340         if (ret) {
2341                 PMD_INIT_LOG(ERR, "Failed to initialize flow");
2342                 return ret;
2343         }
2344
2345         iavf_default_rss_disable(adapter);
2346
2347         return 0;
2348 }
2349
2350 static int
2351 iavf_dev_close(struct rte_eth_dev *dev)
2352 {
2353         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2354         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
2355         struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
2356         struct iavf_adapter *adapter =
2357                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
2358         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
2359         int ret;
2360
2361         if (rte_eal_process_type() != RTE_PROC_PRIMARY)
2362                 return 0;
2363
2364         ret = iavf_dev_stop(dev);
2365
2366         iavf_flow_flush(dev, NULL);
2367         iavf_flow_uninit(adapter);
2368
2369         /*
2370          * disable promiscuous mode before reset vf
2371          * it is a workaround solution when work with kernel driver
2372          * and it is not the normal way
2373          */
2374         if (vf->promisc_unicast_enabled || vf->promisc_multicast_enabled)
2375                 iavf_config_promisc(adapter, false, false);
2376
2377         iavf_shutdown_adminq(hw);
2378         /* disable uio intr before callback unregister */
2379         rte_intr_disable(intr_handle);
2380
2381         /* unregister callback func from eal lib */
2382         rte_intr_callback_unregister(intr_handle,
2383                                      iavf_dev_interrupt_handler, dev);
2384         iavf_disable_irq0(hw);
2385
2386         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_QOS)
2387                 iavf_tm_conf_uninit(dev);
2388
2389         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF) {
2390                 if (vf->rss_lut) {
2391                         rte_free(vf->rss_lut);
2392                         vf->rss_lut = NULL;
2393                 }
2394                 if (vf->rss_key) {
2395                         rte_free(vf->rss_key);
2396                         vf->rss_key = NULL;
2397                 }
2398         }
2399
2400         rte_free(vf->vf_res);
2401         vf->vsi_res = NULL;
2402         vf->vf_res = NULL;
2403
2404         rte_free(vf->aq_resp);
2405         vf->aq_resp = NULL;
2406
2407         /*
2408          * If the VF is reset via VFLR, the device will be knocked out of bus
2409          * master mode, and the driver will fail to recover from the reset. Fix
2410          * this by enabling bus mastering after every reset. In a non-VFLR case,
2411          * the bus master bit will not be disabled, and this call will have no
2412          * effect.
2413          */
2414         if (vf->vf_reset && !rte_pci_set_bus_master(pci_dev, true))
2415                 vf->vf_reset = false;
2416
2417         return ret;
2418 }
2419
2420 static int
2421 iavf_dev_uninit(struct rte_eth_dev *dev)
2422 {
2423         if (rte_eal_process_type() != RTE_PROC_PRIMARY)
2424                 return -EPERM;
2425
2426         iavf_dev_close(dev);
2427
2428         return 0;
2429 }
2430
2431 /*
2432  * Reset VF device only to re-initialize resources in PMD layer
2433  */
2434 static int
2435 iavf_dev_reset(struct rte_eth_dev *dev)
2436 {
2437         int ret;
2438
2439         ret = iavf_dev_uninit(dev);
2440         if (ret)
2441                 return ret;
2442
2443         return iavf_dev_init(dev);
2444 }
2445
2446 static int
2447 iavf_dcf_cap_check_handler(__rte_unused const char *key,
2448                            const char *value, __rte_unused void *opaque)
2449 {
2450         if (strcmp(value, "dcf"))
2451                 return -1;
2452
2453         return 0;
2454 }
2455
2456 static int
2457 iavf_dcf_cap_selected(struct rte_devargs *devargs)
2458 {
2459         struct rte_kvargs *kvlist;
2460         const char *key = "cap";
2461         int ret = 0;
2462
2463         if (devargs == NULL)
2464                 return 0;
2465
2466         kvlist = rte_kvargs_parse(devargs->args, NULL);
2467         if (kvlist == NULL)
2468                 return 0;
2469
2470         if (!rte_kvargs_count(kvlist, key))
2471                 goto exit;
2472
2473         /* dcf capability selected when there's a key-value pair: cap=dcf */
2474         if (rte_kvargs_process(kvlist, key,
2475                                iavf_dcf_cap_check_handler, NULL) < 0)
2476                 goto exit;
2477
2478         ret = 1;
2479
2480 exit:
2481         rte_kvargs_free(kvlist);
2482         return ret;
2483 }
2484
2485 static int
2486 iavf_drv_i40evf_check_handler(__rte_unused const char *key,
2487                               const char *value, __rte_unused void *opaque)
2488 {
2489         if (strcmp(value, "i40evf"))
2490                 return -1;
2491
2492         return 0;
2493 }
2494
2495 static int
2496 iavf_drv_i40evf_selected(struct rte_devargs *devargs, uint16_t device_id)
2497 {
2498         struct rte_kvargs *kvlist;
2499         int ret = 0;
2500
2501         if (device_id != IAVF_DEV_ID_VF &&
2502             device_id != IAVF_DEV_ID_VF_HV &&
2503             device_id != IAVF_DEV_ID_X722_VF &&
2504             device_id != IAVF_DEV_ID_X722_A0_VF)
2505                 return 0;
2506
2507         if (devargs == NULL)
2508                 return 0;
2509
2510         kvlist = rte_kvargs_parse(devargs->args, NULL);
2511         if (kvlist == NULL)
2512                 return 0;
2513
2514         if (!rte_kvargs_count(kvlist, RTE_DEVARGS_KEY_DRIVER))
2515                 goto exit;
2516
2517         /* i40evf driver selected when there's a key-value pair:
2518          * driver=i40evf
2519          */
2520         if (rte_kvargs_process(kvlist, RTE_DEVARGS_KEY_DRIVER,
2521                                iavf_drv_i40evf_check_handler, NULL) < 0)
2522                 goto exit;
2523
2524         ret = 1;
2525
2526 exit:
2527         rte_kvargs_free(kvlist);
2528         return ret;
2529 }
2530
2531 static int eth_iavf_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
2532                              struct rte_pci_device *pci_dev)
2533 {
2534         if (iavf_dcf_cap_selected(pci_dev->device.devargs) ||
2535             iavf_drv_i40evf_selected(pci_dev->device.devargs,
2536                                      pci_dev->id.device_id))
2537                 return 1;
2538
2539         return rte_eth_dev_pci_generic_probe(pci_dev,
2540                 sizeof(struct iavf_adapter), iavf_dev_init);
2541 }
2542
2543 static int eth_iavf_pci_remove(struct rte_pci_device *pci_dev)
2544 {
2545         return rte_eth_dev_pci_generic_remove(pci_dev, iavf_dev_uninit);
2546 }
2547
2548 /* Adaptive virtual function driver struct */
2549 static struct rte_pci_driver rte_iavf_pmd = {
2550         .id_table = pci_id_iavf_map,
2551         .drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC,
2552         .probe = eth_iavf_pci_probe,
2553         .remove = eth_iavf_pci_remove,
2554 };
2555
2556 RTE_PMD_REGISTER_PCI(net_iavf, rte_iavf_pmd);
2557 RTE_PMD_REGISTER_PCI_TABLE(net_iavf, pci_id_iavf_map);
2558 RTE_PMD_REGISTER_KMOD_DEP(net_iavf, "* igb_uio | vfio-pci");
2559 RTE_PMD_REGISTER_PARAM_STRING(net_iavf, "cap=dcf driver=i40evf");
2560 RTE_LOG_REGISTER_SUFFIX(iavf_logtype_init, init, NOTICE);
2561 RTE_LOG_REGISTER_SUFFIX(iavf_logtype_driver, driver, NOTICE);
2562 #ifdef RTE_ETHDEV_DEBUG_RX
2563 RTE_LOG_REGISTER_SUFFIX(iavf_logtype_rx, rx, DEBUG);
2564 #endif
2565 #ifdef RTE_ETHDEV_DEBUG_TX
2566 RTE_LOG_REGISTER_SUFFIX(iavf_logtype_tx, tx, DEBUG);
2567 #endif