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_alarm.h>
  20 #include <rte_atomic.h>
  21 #include <rte_eal.h>
  22 #include <rte_ether.h>
  23 #include <ethdev_driver.h>
  24 #include <ethdev_pci.h>
  25 #include <rte_malloc.h>
  26 #include <rte_memzone.h>
  27 #include <rte_dev.h>
  28
  29 #include "iavf.h"
  30 #include "iavf_rxtx.h"
  31 #include "iavf_generic_flow.h"
  32 #include "rte_pmd_iavf.h"
  33 #include "iavf_ipsec_crypto.h"
  34
  35 /* devargs */
  36 #define IAVF_PROTO_XTR_ARG         "proto_xtr"
  37
  38 static const char * const iavf_valid_args[] = {
  39         IAVF_PROTO_XTR_ARG,
  40         NULL
  41 };
  42
  43 static const struct rte_mbuf_dynfield iavf_proto_xtr_metadata_param = {
  44         .name = "intel_pmd_dynfield_proto_xtr_metadata",
  45         .size = sizeof(uint32_t),
  46         .align = __alignof__(uint32_t),
  47         .flags = 0,
  48 };
  49
  50 struct iavf_proto_xtr_ol {
  51         const struct rte_mbuf_dynflag param;
  52         uint64_t *ol_flag;
  53         bool required;
  54 };
  55
  56 static struct iavf_proto_xtr_ol iavf_proto_xtr_params[] = {
  57         [IAVF_PROTO_XTR_VLAN] = {
  58                 .param = { .name = "intel_pmd_dynflag_proto_xtr_vlan" },
  59                 .ol_flag = &rte_pmd_ifd_dynflag_proto_xtr_vlan_mask },
  60         [IAVF_PROTO_XTR_IPV4] = {
  61                 .param = { .name = "intel_pmd_dynflag_proto_xtr_ipv4" },
  62                 .ol_flag = &rte_pmd_ifd_dynflag_proto_xtr_ipv4_mask },
  63         [IAVF_PROTO_XTR_IPV6] = {
  64                 .param = { .name = "intel_pmd_dynflag_proto_xtr_ipv6" },
  65                 .ol_flag = &rte_pmd_ifd_dynflag_proto_xtr_ipv6_mask },
  66         [IAVF_PROTO_XTR_IPV6_FLOW] = {
  67                 .param = { .name = "intel_pmd_dynflag_proto_xtr_ipv6_flow" },
  68                 .ol_flag = &rte_pmd_ifd_dynflag_proto_xtr_ipv6_flow_mask },
  69         [IAVF_PROTO_XTR_TCP] = {
  70                 .param = { .name = "intel_pmd_dynflag_proto_xtr_tcp" },
  71                 .ol_flag = &rte_pmd_ifd_dynflag_proto_xtr_tcp_mask },
  72         [IAVF_PROTO_XTR_IP_OFFSET] = {
  73                 .param = { .name = "intel_pmd_dynflag_proto_xtr_ip_offset" },
  74                 .ol_flag = &rte_pmd_ifd_dynflag_proto_xtr_ip_offset_mask },
  75         [IAVF_PROTO_XTR_IPSEC_CRYPTO_SAID] = {
  76                 .param = {
  77                 .name = "intel_pmd_dynflag_proto_xtr_ipsec_crypto_said" },
  78                 .ol_flag =
  79                         &rte_pmd_ifd_dynflag_proto_xtr_ipsec_crypto_said_mask },
  80 };
  81
  82 static int iavf_dev_configure(struct rte_eth_dev *dev);
  83 static int iavf_dev_start(struct rte_eth_dev *dev);
  84 static int iavf_dev_stop(struct rte_eth_dev *dev);
  85 static int iavf_dev_close(struct rte_eth_dev *dev);
  86 static int iavf_dev_reset(struct rte_eth_dev *dev);
  87 static int iavf_dev_info_get(struct rte_eth_dev *dev,
  88                              struct rte_eth_dev_info *dev_info);
  89 static const uint32_t *iavf_dev_supported_ptypes_get(struct rte_eth_dev *dev);
  90 static int iavf_dev_stats_get(struct rte_eth_dev *dev,
  91                              struct rte_eth_stats *stats);
  92 static int iavf_dev_stats_reset(struct rte_eth_dev *dev);
  93 static int iavf_dev_xstats_reset(struct rte_eth_dev *dev);
  94 static int iavf_dev_xstats_get(struct rte_eth_dev *dev,
  95                                  struct rte_eth_xstat *xstats, unsigned int n);
  96 static int iavf_dev_xstats_get_names(struct rte_eth_dev *dev,
  97                                        struct rte_eth_xstat_name *xstats_names,
  98                                        unsigned int limit);
  99 static int iavf_dev_promiscuous_enable(struct rte_eth_dev *dev);
 100 static int iavf_dev_promiscuous_disable(struct rte_eth_dev *dev);
 101 static int iavf_dev_allmulticast_enable(struct rte_eth_dev *dev);
 102 static int iavf_dev_allmulticast_disable(struct rte_eth_dev *dev);
 103 static int iavf_dev_add_mac_addr(struct rte_eth_dev *dev,
 104                                 struct rte_ether_addr *addr,
 105                                 uint32_t index,
 106                                 uint32_t pool);
 107 static void iavf_dev_del_mac_addr(struct rte_eth_dev *dev, uint32_t index);
 108 static int iavf_dev_vlan_filter_set(struct rte_eth_dev *dev,
 109                                    uint16_t vlan_id, int on);
 110 static int iavf_dev_vlan_offload_set(struct rte_eth_dev *dev, int mask);
 111 static int iavf_dev_rss_reta_update(struct rte_eth_dev *dev,
 112                                    struct rte_eth_rss_reta_entry64 *reta_conf,
 113                                    uint16_t reta_size);
 114 static int iavf_dev_rss_reta_query(struct rte_eth_dev *dev,
 115                                   struct rte_eth_rss_reta_entry64 *reta_conf,
 116                                   uint16_t reta_size);
 117 static int iavf_dev_rss_hash_update(struct rte_eth_dev *dev,
 118                                    struct rte_eth_rss_conf *rss_conf);
 119 static int iavf_dev_rss_hash_conf_get(struct rte_eth_dev *dev,
 120                                      struct rte_eth_rss_conf *rss_conf);
 121 static int iavf_dev_mtu_set(struct rte_eth_dev *dev, uint16_t mtu);
 122 static int iavf_dev_set_default_mac_addr(struct rte_eth_dev *dev,
 123                                          struct rte_ether_addr *mac_addr);
 124 static int iavf_dev_rx_queue_intr_enable(struct rte_eth_dev *dev,
 125                                         uint16_t queue_id);
 126 static int iavf_dev_rx_queue_intr_disable(struct rte_eth_dev *dev,
 127                                          uint16_t queue_id);
 128 static int iavf_dev_flow_ops_get(struct rte_eth_dev *dev,
 129                                  const struct rte_flow_ops **ops);
 130 static int iavf_set_mc_addr_list(struct rte_eth_dev *dev,
 131                         struct rte_ether_addr *mc_addrs,
 132                         uint32_t mc_addrs_num);
 133 static int iavf_tm_ops_get(struct rte_eth_dev *dev __rte_unused, void *arg);
 134
 135 static const struct rte_pci_id pci_id_iavf_map[] = {
 136         { RTE_PCI_DEVICE(IAVF_INTEL_VENDOR_ID, IAVF_DEV_ID_ADAPTIVE_VF) },
 137         { RTE_PCI_DEVICE(IAVF_INTEL_VENDOR_ID, IAVF_DEV_ID_VF) },
 138         { RTE_PCI_DEVICE(IAVF_INTEL_VENDOR_ID, IAVF_DEV_ID_VF_HV) },
 139         { RTE_PCI_DEVICE(IAVF_INTEL_VENDOR_ID, IAVF_DEV_ID_X722_VF) },
 140         { RTE_PCI_DEVICE(IAVF_INTEL_VENDOR_ID, IAVF_DEV_ID_X722_A0_VF) },
 141         { .vendor_id = 0, /* sentinel */ },
 142 };
 143
 144 struct rte_iavf_xstats_name_off {
 145         char name[RTE_ETH_XSTATS_NAME_SIZE];
 146         unsigned int offset;
 147 };
 148
 149 #define _OFF_OF(a) offsetof(struct iavf_eth_xstats, a)
 150 static const struct rte_iavf_xstats_name_off rte_iavf_stats_strings[] = {
 151         {"rx_bytes", _OFF_OF(eth_stats.rx_bytes)},
 152         {"rx_unicast_packets", _OFF_OF(eth_stats.rx_unicast)},
 153         {"rx_multicast_packets", _OFF_OF(eth_stats.rx_multicast)},
 154         {"rx_broadcast_packets", _OFF_OF(eth_stats.rx_broadcast)},
 155         {"rx_dropped_packets", _OFF_OF(eth_stats.rx_discards)},
 156         {"rx_unknown_protocol_packets", offsetof(struct iavf_eth_stats,
 157                 rx_unknown_protocol)},
 158         {"tx_bytes", _OFF_OF(eth_stats.tx_bytes)},
 159         {"tx_unicast_packets", _OFF_OF(eth_stats.tx_unicast)},
 160         {"tx_multicast_packets", _OFF_OF(eth_stats.tx_multicast)},
 161         {"tx_broadcast_packets", _OFF_OF(eth_stats.tx_broadcast)},
 162         {"tx_dropped_packets", _OFF_OF(eth_stats.tx_discards)},
 163         {"tx_error_packets", _OFF_OF(eth_stats.tx_errors)},
 164
 165         {"inline_ipsec_crypto_ipackets", _OFF_OF(ips_stats.icount)},
 166         {"inline_ipsec_crypto_ibytes", _OFF_OF(ips_stats.ibytes)},
 167         {"inline_ipsec_crypto_ierrors", _OFF_OF(ips_stats.ierrors.count)},
 168         {"inline_ipsec_crypto_ierrors_sad_lookup",
 169                         _OFF_OF(ips_stats.ierrors.sad_miss)},
 170         {"inline_ipsec_crypto_ierrors_not_processed",
 171                         _OFF_OF(ips_stats.ierrors.not_processed)},
 172         {"inline_ipsec_crypto_ierrors_icv_fail",
 173                         _OFF_OF(ips_stats.ierrors.icv_check)},
 174         {"inline_ipsec_crypto_ierrors_length",
 175                         _OFF_OF(ips_stats.ierrors.ipsec_length)},
 176         {"inline_ipsec_crypto_ierrors_misc",
 177                         _OFF_OF(ips_stats.ierrors.misc)},
 178 };
 179 #undef _OFF_OF
 180
 181 #define IAVF_NB_XSTATS (sizeof(rte_iavf_stats_strings) / \
 182                 sizeof(rte_iavf_stats_strings[0]))
 183
 184 static const struct eth_dev_ops iavf_eth_dev_ops = {
 185         .dev_configure              = iavf_dev_configure,
 186         .dev_start                  = iavf_dev_start,
 187         .dev_stop                   = iavf_dev_stop,
 188         .dev_close                  = iavf_dev_close,
 189         .dev_reset                  = iavf_dev_reset,
 190         .dev_infos_get              = iavf_dev_info_get,
 191         .dev_supported_ptypes_get   = iavf_dev_supported_ptypes_get,
 192         .link_update                = iavf_dev_link_update,
 193         .stats_get                  = iavf_dev_stats_get,
 194         .stats_reset                = iavf_dev_stats_reset,
 195         .xstats_get                 = iavf_dev_xstats_get,
 196         .xstats_get_names           = iavf_dev_xstats_get_names,
 197         .xstats_reset               = iavf_dev_xstats_reset,
 198         .promiscuous_enable         = iavf_dev_promiscuous_enable,
 199         .promiscuous_disable        = iavf_dev_promiscuous_disable,
 200         .allmulticast_enable        = iavf_dev_allmulticast_enable,
 201         .allmulticast_disable       = iavf_dev_allmulticast_disable,
 202         .mac_addr_add               = iavf_dev_add_mac_addr,
 203         .mac_addr_remove            = iavf_dev_del_mac_addr,
 204         .set_mc_addr_list                       = iavf_set_mc_addr_list,
 205         .vlan_filter_set            = iavf_dev_vlan_filter_set,
 206         .vlan_offload_set           = iavf_dev_vlan_offload_set,
 207         .rx_queue_start             = iavf_dev_rx_queue_start,
 208         .rx_queue_stop              = iavf_dev_rx_queue_stop,
 209         .tx_queue_start             = iavf_dev_tx_queue_start,
 210         .tx_queue_stop              = iavf_dev_tx_queue_stop,
 211         .rx_queue_setup             = iavf_dev_rx_queue_setup,
 212         .rx_queue_release           = iavf_dev_rx_queue_release,
 213         .tx_queue_setup             = iavf_dev_tx_queue_setup,
 214         .tx_queue_release           = iavf_dev_tx_queue_release,
 215         .mac_addr_set               = iavf_dev_set_default_mac_addr,
 216         .reta_update                = iavf_dev_rss_reta_update,
 217         .reta_query                 = iavf_dev_rss_reta_query,
 218         .rss_hash_update            = iavf_dev_rss_hash_update,
 219         .rss_hash_conf_get          = iavf_dev_rss_hash_conf_get,
 220         .rxq_info_get               = iavf_dev_rxq_info_get,
 221         .txq_info_get               = iavf_dev_txq_info_get,
 222         .mtu_set                    = iavf_dev_mtu_set,
 223         .rx_queue_intr_enable       = iavf_dev_rx_queue_intr_enable,
 224         .rx_queue_intr_disable      = iavf_dev_rx_queue_intr_disable,
 225         .flow_ops_get               = iavf_dev_flow_ops_get,
 226         .tx_done_cleanup            = iavf_dev_tx_done_cleanup,
 227         .get_monitor_addr           = iavf_get_monitor_addr,
 228         .tm_ops_get                 = iavf_tm_ops_get,
 229 };
 230
 231 static int
 232 iavf_tm_ops_get(struct rte_eth_dev *dev __rte_unused,
 233                         void *arg)
 234 {
 235         if (!arg)
 236                 return -EINVAL;
 237
 238         *(const void **)arg = &iavf_tm_ops;
 239
 240         return 0;
 241 }
 242
 243 static int
 244 iavf_set_mc_addr_list(struct rte_eth_dev *dev,
 245                         struct rte_ether_addr *mc_addrs,
 246                         uint32_t mc_addrs_num)
 247 {
 248         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
 249         struct iavf_adapter *adapter =
 250                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 251         int err, ret;
 252
 253         if (mc_addrs_num > IAVF_NUM_MACADDR_MAX) {
 254                 PMD_DRV_LOG(ERR,
 255                             "can't add more than a limited number (%u) of addresses.",
 256                             (uint32_t)IAVF_NUM_MACADDR_MAX);
 257                 return -EINVAL;
 258         }
 259
 260         /* flush previous addresses */
 261         err = iavf_add_del_mc_addr_list(adapter, vf->mc_addrs, vf->mc_addrs_num,
 262                                         false);
 263         if (err)
 264                 return err;
 265
 266         /* add new ones */
 267         err = iavf_add_del_mc_addr_list(adapter, mc_addrs, mc_addrs_num, true);
 268
 269         if (err) {
 270                 /* if adding mac address list fails, should add the previous
 271                  * addresses back.
 272                  */
 273                 ret = iavf_add_del_mc_addr_list(adapter, vf->mc_addrs,
 274                                                 vf->mc_addrs_num, true);
 275                 if (ret)
 276                         return ret;
 277         } else {
 278                 vf->mc_addrs_num = mc_addrs_num;
 279                 memcpy(vf->mc_addrs,
 280                        mc_addrs, mc_addrs_num * sizeof(*mc_addrs));
 281         }
 282
 283         return err;
 284 }
 285
 286 static void
 287 iavf_config_rss_hf(struct iavf_adapter *adapter, uint64_t rss_hf)
 288 {
 289         static const uint64_t map_hena_rss[] = {
 290                 /* IPv4 */
 291                 [IAVF_FILTER_PCTYPE_NONF_UNICAST_IPV4_UDP] =
 292                                 RTE_ETH_RSS_NONFRAG_IPV4_UDP,
 293                 [IAVF_FILTER_PCTYPE_NONF_MULTICAST_IPV4_UDP] =
 294                                 RTE_ETH_RSS_NONFRAG_IPV4_UDP,
 295                 [IAVF_FILTER_PCTYPE_NONF_IPV4_UDP] =
 296                                 RTE_ETH_RSS_NONFRAG_IPV4_UDP,
 297                 [IAVF_FILTER_PCTYPE_NONF_IPV4_TCP_SYN_NO_ACK] =
 298                                 RTE_ETH_RSS_NONFRAG_IPV4_TCP,
 299                 [IAVF_FILTER_PCTYPE_NONF_IPV4_TCP] =
 300                                 RTE_ETH_RSS_NONFRAG_IPV4_TCP,
 301                 [IAVF_FILTER_PCTYPE_NONF_IPV4_SCTP] =
 302                                 RTE_ETH_RSS_NONFRAG_IPV4_SCTP,
 303                 [IAVF_FILTER_PCTYPE_NONF_IPV4_OTHER] =
 304                                 RTE_ETH_RSS_NONFRAG_IPV4_OTHER,
 305                 [IAVF_FILTER_PCTYPE_FRAG_IPV4] = RTE_ETH_RSS_FRAG_IPV4,
 306
 307                 /* IPv6 */
 308                 [IAVF_FILTER_PCTYPE_NONF_UNICAST_IPV6_UDP] =
 309                                 RTE_ETH_RSS_NONFRAG_IPV6_UDP,
 310                 [IAVF_FILTER_PCTYPE_NONF_MULTICAST_IPV6_UDP] =
 311                                 RTE_ETH_RSS_NONFRAG_IPV6_UDP,
 312                 [IAVF_FILTER_PCTYPE_NONF_IPV6_UDP] =
 313                                 RTE_ETH_RSS_NONFRAG_IPV6_UDP,
 314                 [IAVF_FILTER_PCTYPE_NONF_IPV6_TCP_SYN_NO_ACK] =
 315                                 RTE_ETH_RSS_NONFRAG_IPV6_TCP,
 316                 [IAVF_FILTER_PCTYPE_NONF_IPV6_TCP] =
 317                                 RTE_ETH_RSS_NONFRAG_IPV6_TCP,
 318                 [IAVF_FILTER_PCTYPE_NONF_IPV6_SCTP] =
 319                                 RTE_ETH_RSS_NONFRAG_IPV6_SCTP,
 320                 [IAVF_FILTER_PCTYPE_NONF_IPV6_OTHER] =
 321                                 RTE_ETH_RSS_NONFRAG_IPV6_OTHER,
 322                 [IAVF_FILTER_PCTYPE_FRAG_IPV6] = RTE_ETH_RSS_FRAG_IPV6,
 323
 324                 /* L2 Payload */
 325                 [IAVF_FILTER_PCTYPE_L2_PAYLOAD] = RTE_ETH_RSS_L2_PAYLOAD
 326         };
 327
 328         const uint64_t ipv4_rss = RTE_ETH_RSS_NONFRAG_IPV4_UDP |
 329                                   RTE_ETH_RSS_NONFRAG_IPV4_TCP |
 330                                   RTE_ETH_RSS_NONFRAG_IPV4_SCTP |
 331                                   RTE_ETH_RSS_NONFRAG_IPV4_OTHER |
 332                                   RTE_ETH_RSS_FRAG_IPV4;
 333
 334         const uint64_t ipv6_rss = RTE_ETH_RSS_NONFRAG_IPV6_UDP |
 335                                   RTE_ETH_RSS_NONFRAG_IPV6_TCP |
 336                                   RTE_ETH_RSS_NONFRAG_IPV6_SCTP |
 337                                   RTE_ETH_RSS_NONFRAG_IPV6_OTHER |
 338                                   RTE_ETH_RSS_FRAG_IPV6;
 339
 340         struct iavf_info *vf =  IAVF_DEV_PRIVATE_TO_VF(adapter);
 341         uint64_t caps = 0, hena = 0, valid_rss_hf = 0;
 342         uint32_t i;
 343         int ret;
 344
 345         ret = iavf_get_hena_caps(adapter, &caps);
 346         if (ret) {
 347                 /**
 348                  * RSS offload type configuration is not a necessary feature
 349                  * for VF, so here just print a warning and return.
 350                  */
 351                 PMD_DRV_LOG(WARNING,
 352                             "fail to get RSS offload type caps, ret: %d", ret);
 353                 return;
 354         }
 355
 356         /**
 357          * RTE_ETH_RSS_IPV4 and RTE_ETH_RSS_IPV6 can be considered as 2
 358          * generalizations of all other IPv4 and IPv6 RSS types.
 359          */
 360         if (rss_hf & RTE_ETH_RSS_IPV4)
 361                 rss_hf |= ipv4_rss;
 362
 363         if (rss_hf & RTE_ETH_RSS_IPV6)
 364                 rss_hf |= ipv6_rss;
 365
 366         RTE_BUILD_BUG_ON(RTE_DIM(map_hena_rss) > sizeof(uint64_t) * CHAR_BIT);
 367
 368         for (i = 0; i < RTE_DIM(map_hena_rss); i++) {
 369                 uint64_t bit = BIT_ULL(i);
 370
 371                 if ((caps & bit) && (map_hena_rss[i] & rss_hf)) {
 372                         valid_rss_hf |= map_hena_rss[i];
 373                         hena |= bit;
 374                 }
 375         }
 376
 377         ret = iavf_set_hena(adapter, hena);
 378         if (ret) {
 379                 /**
 380                  * RSS offload type configuration is not a necessary feature
 381                  * for VF, so here just print a warning and return.
 382                  */
 383                 PMD_DRV_LOG(WARNING,
 384                             "fail to set RSS offload types, ret: %d", ret);
 385                 return;
 386         }
 387
 388         if (valid_rss_hf & ipv4_rss)
 389                 valid_rss_hf |= rss_hf & RTE_ETH_RSS_IPV4;
 390
 391         if (valid_rss_hf & ipv6_rss)
 392                 valid_rss_hf |= rss_hf & RTE_ETH_RSS_IPV6;
 393
 394         if (rss_hf & ~valid_rss_hf)
 395                 PMD_DRV_LOG(WARNING, "Unsupported rss_hf 0x%" PRIx64,
 396                             rss_hf & ~valid_rss_hf);
 397
 398         vf->rss_hf = valid_rss_hf;
 399 }
 400
 401 static int
 402 iavf_init_rss(struct iavf_adapter *adapter)
 403 {
 404         struct iavf_info *vf =  IAVF_DEV_PRIVATE_TO_VF(adapter);
 405         struct rte_eth_rss_conf *rss_conf;
 406         uint16_t i, j, nb_q;
 407         int ret;
 408
 409         rss_conf = &adapter->dev_data->dev_conf.rx_adv_conf.rss_conf;
 410         nb_q = RTE_MIN(adapter->dev_data->nb_rx_queues,
 411                        vf->max_rss_qregion);
 412
 413         if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF)) {
 414                 PMD_DRV_LOG(DEBUG, "RSS is not supported");
 415                 return -ENOTSUP;
 416         }
 417
 418         /* configure RSS key */
 419         if (!rss_conf->rss_key) {
 420                 /* Calculate the default hash key */
 421                 for (i = 0; i < vf->vf_res->rss_key_size; i++)
 422                         vf->rss_key[i] = (uint8_t)rte_rand();
 423         } else
 424                 rte_memcpy(vf->rss_key, rss_conf->rss_key,
 425                            RTE_MIN(rss_conf->rss_key_len,
 426                                    vf->vf_res->rss_key_size));
 427
 428         /* init RSS LUT table */
 429         for (i = 0, j = 0; i < vf->vf_res->rss_lut_size; i++, j++) {
 430                 if (j >= nb_q)
 431                         j = 0;
 432                 vf->rss_lut[i] = j;
 433         }
 434         /* send virtchnnl ops to configure rss*/
 435         ret = iavf_configure_rss_lut(adapter);
 436         if (ret)
 437                 return ret;
 438         ret = iavf_configure_rss_key(adapter);
 439         if (ret)
 440                 return ret;
 441
 442         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_ADV_RSS_PF) {
 443                 /* Set RSS hash configuration based on rss_conf->rss_hf. */
 444                 ret = iavf_rss_hash_set(adapter, rss_conf->rss_hf, true);
 445                 if (ret) {
 446                         PMD_DRV_LOG(ERR, "fail to set default RSS");
 447                         return ret;
 448                 }
 449         } else {
 450                 iavf_config_rss_hf(adapter, rss_conf->rss_hf);
 451         }
 452
 453         return 0;
 454 }
 455
 456 static int
 457 iavf_queues_req_reset(struct rte_eth_dev *dev, uint16_t num)
 458 {
 459         struct iavf_adapter *ad =
 460                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 461         struct iavf_info *vf =  IAVF_DEV_PRIVATE_TO_VF(ad);
 462         int ret;
 463
 464         ret = iavf_request_queues(dev, num);
 465         if (ret) {
 466                 PMD_DRV_LOG(ERR, "request queues from PF failed");
 467                 return ret;
 468         }
 469         PMD_DRV_LOG(INFO, "change queue pairs from %u to %u",
 470                         vf->vsi_res->num_queue_pairs, num);
 471
 472         ret = iavf_dev_reset(dev);
 473         if (ret) {
 474                 PMD_DRV_LOG(ERR, "vf reset failed");
 475                 return ret;
 476         }
 477
 478         return 0;
 479 }
 480
 481 static int
 482 iavf_dev_vlan_insert_set(struct rte_eth_dev *dev)
 483 {
 484         struct iavf_adapter *adapter =
 485                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 486         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
 487         bool enable;
 488
 489         if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_VLAN_V2))
 490                 return 0;
 491
 492         enable = !!(dev->data->dev_conf.txmode.offloads &
 493                     RTE_ETH_TX_OFFLOAD_VLAN_INSERT);
 494         iavf_config_vlan_insert_v2(adapter, enable);
 495
 496         return 0;
 497 }
 498
 499 static int
 500 iavf_dev_init_vlan(struct rte_eth_dev *dev)
 501 {
 502         int err;
 503
 504         err = iavf_dev_vlan_offload_set(dev,
 505                                         RTE_ETH_VLAN_STRIP_MASK |
 506                                         RTE_ETH_QINQ_STRIP_MASK |
 507                                         RTE_ETH_VLAN_FILTER_MASK |
 508                                         RTE_ETH_VLAN_EXTEND_MASK);
 509         if (err) {
 510                 PMD_DRV_LOG(ERR, "Failed to update vlan offload");
 511                 return err;
 512         }
 513
 514         err = iavf_dev_vlan_insert_set(dev);
 515         if (err)
 516                 PMD_DRV_LOG(ERR, "Failed to update vlan insertion");
 517
 518         return err;
 519 }
 520
 521 static int
 522 iavf_dev_configure(struct rte_eth_dev *dev)
 523 {
 524         struct iavf_adapter *ad =
 525                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 526         struct iavf_info *vf =  IAVF_DEV_PRIVATE_TO_VF(ad);
 527         uint16_t num_queue_pairs = RTE_MAX(dev->data->nb_rx_queues,
 528                 dev->data->nb_tx_queues);
 529         int ret;
 530
 531         ad->rx_bulk_alloc_allowed = true;
 532         /* Initialize to TRUE. If any of Rx queues doesn't meet the
 533          * vector Rx/Tx preconditions, it will be reset.
 534          */
 535         ad->rx_vec_allowed = true;
 536         ad->tx_vec_allowed = true;
 537
 538         if (dev->data->dev_conf.rxmode.mq_mode & RTE_ETH_MQ_RX_RSS_FLAG)
 539                 dev->data->dev_conf.rxmode.offloads |= RTE_ETH_RX_OFFLOAD_RSS_HASH;
 540
 541         /* Large VF setting */
 542         if (num_queue_pairs > IAVF_MAX_NUM_QUEUES_DFLT) {
 543                 if (!(vf->vf_res->vf_cap_flags &
 544                                 VIRTCHNL_VF_LARGE_NUM_QPAIRS)) {
 545                         PMD_DRV_LOG(ERR, "large VF is not supported");
 546                         return -1;
 547                 }
 548
 549                 if (num_queue_pairs > IAVF_MAX_NUM_QUEUES_LV) {
 550                         PMD_DRV_LOG(ERR, "queue pairs number cannot be larger than %u",
 551                                 IAVF_MAX_NUM_QUEUES_LV);
 552                         return -1;
 553                 }
 554
 555                 ret = iavf_queues_req_reset(dev, num_queue_pairs);
 556                 if (ret)
 557                         return ret;
 558
 559                 ret = iavf_get_max_rss_queue_region(ad);
 560                 if (ret) {
 561                         PMD_INIT_LOG(ERR, "get max rss queue region failed");
 562                         return ret;
 563                 }
 564
 565                 vf->lv_enabled = true;
 566         } else {
 567                 /* Check if large VF is already enabled. If so, disable and
 568                  * release redundant queue resource.
 569                  * Or check if enough queue pairs. If not, request them from PF.
 570                  */
 571                 if (vf->lv_enabled ||
 572                     num_queue_pairs > vf->vsi_res->num_queue_pairs) {
 573                         ret = iavf_queues_req_reset(dev, num_queue_pairs);
 574                         if (ret)
 575                                 return ret;
 576
 577                         vf->lv_enabled = false;
 578                 }
 579                 /* if large VF is not required, use default rss queue region */
 580                 vf->max_rss_qregion = IAVF_MAX_NUM_QUEUES_DFLT;
 581         }
 582
 583         ret = iavf_dev_init_vlan(dev);
 584         if (ret)
 585                 PMD_DRV_LOG(ERR, "configure VLAN failed: %d", ret);
 586
 587         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF) {
 588                 if (iavf_init_rss(ad) != 0) {
 589                         PMD_DRV_LOG(ERR, "configure rss failed");
 590                         return -1;
 591                 }
 592         }
 593         return 0;
 594 }
 595
 596 static int
 597 iavf_init_rxq(struct rte_eth_dev *dev, struct iavf_rx_queue *rxq)
 598 {
 599         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(dev->data->dev_private);
 600         struct rte_eth_dev_data *dev_data = dev->data;
 601         uint16_t buf_size, max_pkt_len;
 602         uint32_t frame_size = dev->data->mtu + IAVF_ETH_OVERHEAD;
 603
 604         buf_size = rte_pktmbuf_data_room_size(rxq->mp) - RTE_PKTMBUF_HEADROOM;
 605
 606         /* Calculate the maximum packet length allowed */
 607         max_pkt_len = RTE_MIN((uint32_t)
 608                         rxq->rx_buf_len * IAVF_MAX_CHAINED_RX_BUFFERS,
 609                         frame_size);
 610
 611         /* Check if maximum packet length is set correctly.  */
 612         if (max_pkt_len <= RTE_ETHER_MIN_LEN ||
 613             max_pkt_len > IAVF_FRAME_SIZE_MAX) {
 614                 PMD_DRV_LOG(ERR, "maximum packet length must be "
 615                             "larger than %u and smaller than %u",
 616                             (uint32_t)IAVF_ETH_MAX_LEN,
 617                             (uint32_t)IAVF_FRAME_SIZE_MAX);
 618                 return -EINVAL;
 619         }
 620
 621         rxq->max_pkt_len = max_pkt_len;
 622         if ((dev_data->dev_conf.rxmode.offloads & RTE_ETH_RX_OFFLOAD_SCATTER) ||
 623             rxq->max_pkt_len > buf_size) {
 624                 dev_data->scattered_rx = 1;
 625         }
 626         IAVF_PCI_REG_WRITE(rxq->qrx_tail, rxq->nb_rx_desc - 1);
 627         IAVF_WRITE_FLUSH(hw);
 628
 629         return 0;
 630 }
 631
 632 static int
 633 iavf_init_queues(struct rte_eth_dev *dev)
 634 {
 635         struct iavf_rx_queue **rxq =
 636                 (struct iavf_rx_queue **)dev->data->rx_queues;
 637         int i, ret = IAVF_SUCCESS;
 638
 639         for (i = 0; i < dev->data->nb_rx_queues; i++) {
 640                 if (!rxq[i] || !rxq[i]->q_set)
 641                         continue;
 642                 ret = iavf_init_rxq(dev, rxq[i]);
 643                 if (ret != IAVF_SUCCESS)
 644                         break;
 645         }
 646         /* set rx/tx function to vector/scatter/single-segment
 647          * according to parameters
 648          */
 649         iavf_set_rx_function(dev);
 650         iavf_set_tx_function(dev);
 651
 652         return ret;
 653 }
 654
 655 static int iavf_config_rx_queues_irqs(struct rte_eth_dev *dev,
 656                                      struct rte_intr_handle *intr_handle)
 657 {
 658         struct iavf_adapter *adapter =
 659                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 660         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
 661         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(adapter);
 662         struct iavf_qv_map *qv_map;
 663         uint16_t interval, i;
 664         int vec;
 665
 666         if (rte_intr_cap_multiple(intr_handle) &&
 667             dev->data->dev_conf.intr_conf.rxq) {
 668                 if (rte_intr_efd_enable(intr_handle, dev->data->nb_rx_queues))
 669                         return -1;
 670         }
 671
 672         if (rte_intr_dp_is_en(intr_handle)) {
 673                 if (rte_intr_vec_list_alloc(intr_handle, "intr_vec",
 674                                                     dev->data->nb_rx_queues)) {
 675                         PMD_DRV_LOG(ERR, "Failed to allocate %d rx intr_vec",
 676                                     dev->data->nb_rx_queues);
 677                         return -1;
 678                 }
 679         }
 680
 681
 682         qv_map = rte_zmalloc("qv_map",
 683                 dev->data->nb_rx_queues * sizeof(struct iavf_qv_map), 0);
 684         if (!qv_map) {
 685                 PMD_DRV_LOG(ERR, "Failed to allocate %d queue-vector map",
 686                                 dev->data->nb_rx_queues);
 687                 goto qv_map_alloc_err;
 688         }
 689
 690         if (!dev->data->dev_conf.intr_conf.rxq ||
 691             !rte_intr_dp_is_en(intr_handle)) {
 692                 /* Rx interrupt disabled, Map interrupt only for writeback */
 693                 vf->nb_msix = 1;
 694                 if (vf->vf_res->vf_cap_flags &
 695                     VIRTCHNL_VF_OFFLOAD_WB_ON_ITR) {
 696                         /* If WB_ON_ITR supports, enable it */
 697                         vf->msix_base = IAVF_RX_VEC_START;
 698                         /* Set the ITR for index zero, to 2us to make sure that
 699                          * we leave time for aggregation to occur, but don't
 700                          * increase latency dramatically.
 701                          */
 702                         IAVF_WRITE_REG(hw,
 703                                        IAVF_VFINT_DYN_CTLN1(vf->msix_base - 1),
 704                                        (0 << IAVF_VFINT_DYN_CTLN1_ITR_INDX_SHIFT) |
 705                                        IAVF_VFINT_DYN_CTLN1_WB_ON_ITR_MASK |
 706                                        (2UL << IAVF_VFINT_DYN_CTLN1_INTERVAL_SHIFT));
 707                         /* debug - check for success! the return value
 708                          * should be 2, offset is 0x2800
 709                          */
 710                         /* IAVF_READ_REG(hw, IAVF_VFINT_ITRN1(0, 0)); */
 711                 } else {
 712                         /* If no WB_ON_ITR offload flags, need to set
 713                          * interrupt for descriptor write back.
 714                          */
 715                         vf->msix_base = IAVF_MISC_VEC_ID;
 716
 717                         /* set ITR to default */
 718                         interval = iavf_calc_itr_interval(
 719                                         IAVF_QUEUE_ITR_INTERVAL_DEFAULT);
 720                         IAVF_WRITE_REG(hw, IAVF_VFINT_DYN_CTL01,
 721                                        IAVF_VFINT_DYN_CTL01_INTENA_MASK |
 722                                        (IAVF_ITR_INDEX_DEFAULT <<
 723                                         IAVF_VFINT_DYN_CTL01_ITR_INDX_SHIFT) |
 724                                        (interval <<
 725                                         IAVF_VFINT_DYN_CTL01_INTERVAL_SHIFT));
 726                 }
 727                 IAVF_WRITE_FLUSH(hw);
 728                 /* map all queues to the same interrupt */
 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                 }
 733                 vf->qv_map = qv_map;
 734         } else {
 735                 if (!rte_intr_allow_others(intr_handle)) {
 736                         vf->nb_msix = 1;
 737                         vf->msix_base = IAVF_MISC_VEC_ID;
 738                         for (i = 0; i < dev->data->nb_rx_queues; i++) {
 739                                 qv_map[i].queue_id = i;
 740                                 qv_map[i].vector_id = vf->msix_base;
 741                                 rte_intr_vec_list_index_set(intr_handle,
 742                                                         i, IAVF_MISC_VEC_ID);
 743                         }
 744                         vf->qv_map = qv_map;
 745                         PMD_DRV_LOG(DEBUG,
 746                                     "vector %u are mapping to all Rx queues",
 747                                     vf->msix_base);
 748                 } else {
 749                         /* If Rx interrupt is reuquired, and we can use
 750                          * multi interrupts, then the vec is from 1
 751                          */
 752                         vf->nb_msix =
 753                                 RTE_MIN(rte_intr_nb_efd_get(intr_handle),
 754                                 (uint16_t)(vf->vf_res->max_vectors - 1));
 755                         vf->msix_base = IAVF_RX_VEC_START;
 756                         vec = IAVF_RX_VEC_START;
 757                         for (i = 0; i < dev->data->nb_rx_queues; i++) {
 758                                 qv_map[i].queue_id = i;
 759                                 qv_map[i].vector_id = vec;
 760                                 rte_intr_vec_list_index_set(intr_handle,
 761                                                                    i, vec++);
 762                                 if (vec >= vf->nb_msix + IAVF_RX_VEC_START)
 763                                         vec = IAVF_RX_VEC_START;
 764                         }
 765                         vf->qv_map = qv_map;
 766                         PMD_DRV_LOG(DEBUG,
 767                                     "%u vectors are mapping to %u Rx queues",
 768                                     vf->nb_msix, dev->data->nb_rx_queues);
 769                 }
 770         }
 771
 772         if (!vf->lv_enabled) {
 773                 if (iavf_config_irq_map(adapter)) {
 774                         PMD_DRV_LOG(ERR, "config interrupt mapping failed");
 775                         goto config_irq_map_err;
 776                 }
 777         } else {
 778                 uint16_t num_qv_maps = dev->data->nb_rx_queues;
 779                 uint16_t index = 0;
 780
 781                 while (num_qv_maps > IAVF_IRQ_MAP_NUM_PER_BUF) {
 782                         if (iavf_config_irq_map_lv(adapter,
 783                                         IAVF_IRQ_MAP_NUM_PER_BUF, index)) {
 784                                 PMD_DRV_LOG(ERR, "config interrupt mapping for large VF failed");
 785                                 goto config_irq_map_err;
 786                         }
 787                         num_qv_maps -= IAVF_IRQ_MAP_NUM_PER_BUF;
 788                         index += IAVF_IRQ_MAP_NUM_PER_BUF;
 789                 }
 790
 791                 if (iavf_config_irq_map_lv(adapter, num_qv_maps, index)) {
 792                         PMD_DRV_LOG(ERR, "config interrupt mapping for large VF failed");
 793                         goto config_irq_map_err;
 794                 }
 795         }
 796         return 0;
 797
 798 config_irq_map_err:
 799         rte_free(vf->qv_map);
 800         vf->qv_map = NULL;
 801
 802 qv_map_alloc_err:
 803         rte_intr_vec_list_free(intr_handle);
 804
 805         return -1;
 806 }
 807
 808 static int
 809 iavf_start_queues(struct rte_eth_dev *dev)
 810 {
 811         struct iavf_rx_queue *rxq;
 812         struct iavf_tx_queue *txq;
 813         int i;
 814
 815         for (i = 0; i < dev->data->nb_tx_queues; i++) {
 816                 txq = dev->data->tx_queues[i];
 817                 if (txq->tx_deferred_start)
 818                         continue;
 819                 if (iavf_dev_tx_queue_start(dev, i) != 0) {
 820                         PMD_DRV_LOG(ERR, "Fail to start queue %u", i);
 821                         return -1;
 822                 }
 823         }
 824
 825         for (i = 0; i < dev->data->nb_rx_queues; i++) {
 826                 rxq = dev->data->rx_queues[i];
 827                 if (rxq->rx_deferred_start)
 828                         continue;
 829                 if (iavf_dev_rx_queue_start(dev, i) != 0) {
 830                         PMD_DRV_LOG(ERR, "Fail to start queue %u", i);
 831                         return -1;
 832                 }
 833         }
 834
 835         return 0;
 836 }
 837
 838 static int
 839 iavf_dev_start(struct rte_eth_dev *dev)
 840 {
 841         struct iavf_adapter *adapter =
 842                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 843         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
 844         struct rte_intr_handle *intr_handle = dev->intr_handle;
 845         uint16_t num_queue_pairs;
 846         uint16_t index = 0;
 847
 848         PMD_INIT_FUNC_TRACE();
 849
 850         adapter->stopped = 0;
 851
 852         vf->max_pkt_len = dev->data->mtu + IAVF_ETH_OVERHEAD;
 853         vf->num_queue_pairs = RTE_MAX(dev->data->nb_rx_queues,
 854                                       dev->data->nb_tx_queues);
 855         num_queue_pairs = vf->num_queue_pairs;
 856
 857         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_QOS)
 858                 if (iavf_get_qos_cap(adapter)) {
 859                         PMD_INIT_LOG(ERR, "Failed to get qos capability");
 860                         return -1;
 861                 }
 862
 863         if (iavf_init_queues(dev) != 0) {
 864                 PMD_DRV_LOG(ERR, "failed to do Queue init");
 865                 return -1;
 866         }
 867
 868         /* If needed, send configure queues msg multiple times to make the
 869          * adminq buffer length smaller than the 4K limitation.
 870          */
 871         while (num_queue_pairs > IAVF_CFG_Q_NUM_PER_BUF) {
 872                 if (iavf_configure_queues(adapter,
 873                                 IAVF_CFG_Q_NUM_PER_BUF, index) != 0) {
 874                         PMD_DRV_LOG(ERR, "configure queues failed");
 875                         goto err_queue;
 876                 }
 877                 num_queue_pairs -= IAVF_CFG_Q_NUM_PER_BUF;
 878                 index += IAVF_CFG_Q_NUM_PER_BUF;
 879         }
 880
 881         if (iavf_configure_queues(adapter, num_queue_pairs, index) != 0) {
 882                 PMD_DRV_LOG(ERR, "configure queues failed");
 883                 goto err_queue;
 884         }
 885
 886         if (iavf_config_rx_queues_irqs(dev, intr_handle) != 0) {
 887                 PMD_DRV_LOG(ERR, "configure irq failed");
 888                 goto err_queue;
 889         }
 890         /* re-enable intr again, because efd assign may change */
 891         if (dev->data->dev_conf.intr_conf.rxq != 0) {
 892                 if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_WB_ON_ITR)
 893                         rte_intr_disable(intr_handle);
 894                 rte_intr_enable(intr_handle);
 895         }
 896
 897         /* Set all mac addrs */
 898         iavf_add_del_all_mac_addr(adapter, true);
 899
 900         /* Set all multicast addresses */
 901         iavf_add_del_mc_addr_list(adapter, vf->mc_addrs, vf->mc_addrs_num,
 902                                   true);
 903
 904         if (iavf_start_queues(dev) != 0) {
 905                 PMD_DRV_LOG(ERR, "enable queues failed");
 906                 goto err_mac;
 907         }
 908
 909         return 0;
 910
 911 err_mac:
 912         iavf_add_del_all_mac_addr(adapter, false);
 913 err_queue:
 914         return -1;
 915 }
 916
 917 static int
 918 iavf_dev_stop(struct rte_eth_dev *dev)
 919 {
 920         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
 921         struct iavf_adapter *adapter =
 922                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 923         struct rte_intr_handle *intr_handle = dev->intr_handle;
 924
 925         PMD_INIT_FUNC_TRACE();
 926
 927         if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_WB_ON_ITR) &&
 928             dev->data->dev_conf.intr_conf.rxq != 0)
 929                 rte_intr_disable(intr_handle);
 930
 931         if (adapter->stopped == 1)
 932                 return 0;
 933
 934         iavf_stop_queues(dev);
 935
 936         /* Disable the interrupt for Rx */
 937         rte_intr_efd_disable(intr_handle);
 938         /* Rx interrupt vector mapping free */
 939         rte_intr_vec_list_free(intr_handle);
 940
 941         /* remove all mac addrs */
 942         iavf_add_del_all_mac_addr(adapter, false);
 943
 944         /* remove all multicast addresses */
 945         iavf_add_del_mc_addr_list(adapter, vf->mc_addrs, vf->mc_addrs_num,
 946                                   false);
 947
 948         /* free iAVF security device context all related resources */
 949         iavf_security_ctx_destroy(adapter);
 950
 951         adapter->stopped = 1;
 952         dev->data->dev_started = 0;
 953
 954         return 0;
 955 }
 956
 957 static int
 958 iavf_dev_info_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
 959 {
 960         struct iavf_adapter *adapter =
 961                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 962         struct iavf_info *vf = &adapter->vf;
 963
 964         dev_info->max_rx_queues = IAVF_MAX_NUM_QUEUES_LV;
 965         dev_info->max_tx_queues = IAVF_MAX_NUM_QUEUES_LV;
 966         dev_info->min_rx_bufsize = IAVF_BUF_SIZE_MIN;
 967         dev_info->max_rx_pktlen = IAVF_FRAME_SIZE_MAX;
 968         dev_info->max_mtu = dev_info->max_rx_pktlen - IAVF_ETH_OVERHEAD;
 969         dev_info->min_mtu = RTE_ETHER_MIN_MTU;
 970         dev_info->hash_key_size = vf->vf_res->rss_key_size;
 971         dev_info->reta_size = vf->vf_res->rss_lut_size;
 972         dev_info->flow_type_rss_offloads = IAVF_RSS_OFFLOAD_ALL;
 973         dev_info->max_mac_addrs = IAVF_NUM_MACADDR_MAX;
 974         dev_info->rx_offload_capa =
 975                 RTE_ETH_RX_OFFLOAD_VLAN_STRIP |
 976                 RTE_ETH_RX_OFFLOAD_QINQ_STRIP |
 977                 RTE_ETH_RX_OFFLOAD_IPV4_CKSUM |
 978                 RTE_ETH_RX_OFFLOAD_UDP_CKSUM |
 979                 RTE_ETH_RX_OFFLOAD_TCP_CKSUM |
 980                 RTE_ETH_RX_OFFLOAD_OUTER_IPV4_CKSUM |
 981                 RTE_ETH_RX_OFFLOAD_SCATTER |
 982                 RTE_ETH_RX_OFFLOAD_VLAN_FILTER |
 983                 RTE_ETH_RX_OFFLOAD_RSS_HASH;
 984
 985         dev_info->tx_offload_capa =
 986                 RTE_ETH_TX_OFFLOAD_VLAN_INSERT |
 987                 RTE_ETH_TX_OFFLOAD_QINQ_INSERT |
 988                 RTE_ETH_TX_OFFLOAD_IPV4_CKSUM |
 989                 RTE_ETH_TX_OFFLOAD_UDP_CKSUM |
 990                 RTE_ETH_TX_OFFLOAD_TCP_CKSUM |
 991                 RTE_ETH_TX_OFFLOAD_SCTP_CKSUM |
 992                 RTE_ETH_TX_OFFLOAD_OUTER_IPV4_CKSUM |
 993                 RTE_ETH_TX_OFFLOAD_TCP_TSO |
 994                 RTE_ETH_TX_OFFLOAD_VXLAN_TNL_TSO |
 995                 RTE_ETH_TX_OFFLOAD_GRE_TNL_TSO |
 996                 RTE_ETH_TX_OFFLOAD_IPIP_TNL_TSO |
 997                 RTE_ETH_TX_OFFLOAD_GENEVE_TNL_TSO |
 998                 RTE_ETH_TX_OFFLOAD_MULTI_SEGS |
 999                 RTE_ETH_TX_OFFLOAD_MBUF_FAST_FREE;
1000
1001         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_CRC)
1002                 dev_info->rx_offload_capa |= RTE_ETH_RX_OFFLOAD_KEEP_CRC;
1003
1004         if (iavf_ipsec_crypto_supported(adapter)) {
1005                 dev_info->rx_offload_capa |= RTE_ETH_RX_OFFLOAD_SECURITY;
1006                 dev_info->tx_offload_capa |= RTE_ETH_TX_OFFLOAD_SECURITY;
1007         }
1008
1009         dev_info->default_rxconf = (struct rte_eth_rxconf) {
1010                 .rx_free_thresh = IAVF_DEFAULT_RX_FREE_THRESH,
1011                 .rx_drop_en = 0,
1012                 .offloads = 0,
1013         };
1014
1015         dev_info->default_txconf = (struct rte_eth_txconf) {
1016                 .tx_free_thresh = IAVF_DEFAULT_TX_FREE_THRESH,
1017                 .tx_rs_thresh = IAVF_DEFAULT_TX_RS_THRESH,
1018                 .offloads = 0,
1019         };
1020
1021         dev_info->rx_desc_lim = (struct rte_eth_desc_lim) {
1022                 .nb_max = IAVF_MAX_RING_DESC,
1023                 .nb_min = IAVF_MIN_RING_DESC,
1024                 .nb_align = IAVF_ALIGN_RING_DESC,
1025         };
1026
1027         dev_info->tx_desc_lim = (struct rte_eth_desc_lim) {
1028                 .nb_max = IAVF_MAX_RING_DESC,
1029                 .nb_min = IAVF_MIN_RING_DESC,
1030                 .nb_align = IAVF_ALIGN_RING_DESC,
1031         };
1032
1033         return 0;
1034 }
1035
1036 static const uint32_t *
1037 iavf_dev_supported_ptypes_get(struct rte_eth_dev *dev __rte_unused)
1038 {
1039         static const uint32_t ptypes[] = {
1040                 RTE_PTYPE_L2_ETHER,
1041                 RTE_PTYPE_L3_IPV4_EXT_UNKNOWN,
1042                 RTE_PTYPE_L4_FRAG,
1043                 RTE_PTYPE_L4_ICMP,
1044                 RTE_PTYPE_L4_NONFRAG,
1045                 RTE_PTYPE_L4_SCTP,
1046                 RTE_PTYPE_L4_TCP,
1047                 RTE_PTYPE_L4_UDP,
1048                 RTE_PTYPE_UNKNOWN
1049         };
1050         return ptypes;
1051 }
1052
1053 int
1054 iavf_dev_link_update(struct rte_eth_dev *dev,
1055                     __rte_unused int wait_to_complete)
1056 {
1057         struct rte_eth_link new_link;
1058         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
1059
1060         memset(&new_link, 0, sizeof(new_link));
1061
1062         /* Only read status info stored in VF, and the info is updated
1063          *  when receive LINK_CHANGE evnet from PF by Virtchnnl.
1064          */
1065         switch (vf->link_speed) {
1066         case 10:
1067                 new_link.link_speed = RTE_ETH_SPEED_NUM_10M;
1068                 break;
1069         case 100:
1070                 new_link.link_speed = RTE_ETH_SPEED_NUM_100M;
1071                 break;
1072         case 1000:
1073                 new_link.link_speed = RTE_ETH_SPEED_NUM_1G;
1074                 break;
1075         case 10000:
1076                 new_link.link_speed = RTE_ETH_SPEED_NUM_10G;
1077                 break;
1078         case 20000:
1079                 new_link.link_speed = RTE_ETH_SPEED_NUM_20G;
1080                 break;
1081         case 25000:
1082                 new_link.link_speed = RTE_ETH_SPEED_NUM_25G;
1083                 break;
1084         case 40000:
1085                 new_link.link_speed = RTE_ETH_SPEED_NUM_40G;
1086                 break;
1087         case 50000:
1088                 new_link.link_speed = RTE_ETH_SPEED_NUM_50G;
1089                 break;
1090         case 100000:
1091                 new_link.link_speed = RTE_ETH_SPEED_NUM_100G;
1092                 break;
1093         default:
1094                 new_link.link_speed = RTE_ETH_SPEED_NUM_NONE;
1095                 break;
1096         }
1097
1098         new_link.link_duplex = RTE_ETH_LINK_FULL_DUPLEX;
1099         new_link.link_status = vf->link_up ? RTE_ETH_LINK_UP :
1100                                              RTE_ETH_LINK_DOWN;
1101         new_link.link_autoneg = !(dev->data->dev_conf.link_speeds &
1102                                 RTE_ETH_LINK_SPEED_FIXED);
1103
1104         return rte_eth_linkstatus_set(dev, &new_link);
1105 }
1106
1107 static int
1108 iavf_dev_promiscuous_enable(struct rte_eth_dev *dev)
1109 {
1110         struct iavf_adapter *adapter =
1111                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1112         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
1113
1114         return iavf_config_promisc(adapter,
1115                                   true, vf->promisc_multicast_enabled);
1116 }
1117
1118 static int
1119 iavf_dev_promiscuous_disable(struct rte_eth_dev *dev)
1120 {
1121         struct iavf_adapter *adapter =
1122                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1123         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
1124
1125         return iavf_config_promisc(adapter,
1126                                   false, vf->promisc_multicast_enabled);
1127 }
1128
1129 static int
1130 iavf_dev_allmulticast_enable(struct rte_eth_dev *dev)
1131 {
1132         struct iavf_adapter *adapter =
1133                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1134         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
1135
1136         return iavf_config_promisc(adapter,
1137                                   vf->promisc_unicast_enabled, true);
1138 }
1139
1140 static int
1141 iavf_dev_allmulticast_disable(struct rte_eth_dev *dev)
1142 {
1143         struct iavf_adapter *adapter =
1144                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1145         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
1146
1147         return iavf_config_promisc(adapter,
1148                                   vf->promisc_unicast_enabled, false);
1149 }
1150
1151 static int
1152 iavf_dev_add_mac_addr(struct rte_eth_dev *dev, struct rte_ether_addr *addr,
1153                      __rte_unused uint32_t index,
1154                      __rte_unused uint32_t pool)
1155 {
1156         struct iavf_adapter *adapter =
1157                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1158         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
1159         int err;
1160
1161         if (rte_is_zero_ether_addr(addr)) {
1162                 PMD_DRV_LOG(ERR, "Invalid Ethernet Address");
1163                 return -EINVAL;
1164         }
1165
1166         err = iavf_add_del_eth_addr(adapter, addr, true, VIRTCHNL_ETHER_ADDR_EXTRA);
1167         if (err) {
1168                 PMD_DRV_LOG(ERR, "fail to add MAC address");
1169                 return -EIO;
1170         }
1171
1172         vf->mac_num++;
1173
1174         return 0;
1175 }
1176
1177 static void
1178 iavf_dev_del_mac_addr(struct rte_eth_dev *dev, uint32_t index)
1179 {
1180         struct iavf_adapter *adapter =
1181                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1182         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
1183         struct rte_ether_addr *addr;
1184         int err;
1185
1186         addr = &dev->data->mac_addrs[index];
1187
1188         err = iavf_add_del_eth_addr(adapter, addr, false, VIRTCHNL_ETHER_ADDR_EXTRA);
1189         if (err)
1190                 PMD_DRV_LOG(ERR, "fail to delete MAC address");
1191
1192         vf->mac_num--;
1193 }
1194
1195 static int
1196 iavf_dev_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
1197 {
1198         struct iavf_adapter *adapter =
1199                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1200         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
1201         int err;
1202
1203         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_VLAN_V2) {
1204                 err = iavf_add_del_vlan_v2(adapter, vlan_id, on);
1205                 if (err)
1206                         return -EIO;
1207                 return 0;
1208         }
1209
1210         if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_VLAN))
1211                 return -ENOTSUP;
1212
1213         err = iavf_add_del_vlan(adapter, vlan_id, on);
1214         if (err)
1215                 return -EIO;
1216         return 0;
1217 }
1218
1219 static void
1220 iavf_iterate_vlan_filters_v2(struct rte_eth_dev *dev, bool enable)
1221 {
1222         struct rte_vlan_filter_conf *vfc = &dev->data->vlan_filter_conf;
1223         struct iavf_adapter *adapter =
1224                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1225         uint32_t i, j;
1226         uint64_t ids;
1227
1228         for (i = 0; i < RTE_DIM(vfc->ids); i++) {
1229                 if (vfc->ids[i] == 0)
1230                         continue;
1231
1232                 ids = vfc->ids[i];
1233                 for (j = 0; ids != 0 && j < 64; j++, ids >>= 1) {
1234                         if (ids & 1)
1235                                 iavf_add_del_vlan_v2(adapter,
1236                                                      64 * i + j, enable);
1237                 }
1238         }
1239 }
1240
1241 static int
1242 iavf_dev_vlan_offload_set_v2(struct rte_eth_dev *dev, int mask)
1243 {
1244         struct rte_eth_rxmode *rxmode = &dev->data->dev_conf.rxmode;
1245         struct iavf_adapter *adapter =
1246                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1247         bool enable;
1248         int err;
1249
1250         if (mask & RTE_ETH_VLAN_FILTER_MASK) {
1251                 enable = !!(rxmode->offloads & RTE_ETH_RX_OFFLOAD_VLAN_FILTER);
1252
1253                 iavf_iterate_vlan_filters_v2(dev, enable);
1254         }
1255
1256         if (mask & RTE_ETH_VLAN_STRIP_MASK) {
1257                 enable = !!(rxmode->offloads & RTE_ETH_RX_OFFLOAD_VLAN_STRIP);
1258
1259                 err = iavf_config_vlan_strip_v2(adapter, enable);
1260                 /* If not support, the stripping is already disabled by PF */
1261                 if (err == -ENOTSUP && !enable)
1262                         err = 0;
1263                 if (err)
1264                         return -EIO;
1265         }
1266
1267         return 0;
1268 }
1269
1270 static int
1271 iavf_dev_vlan_offload_set(struct rte_eth_dev *dev, int mask)
1272 {
1273         struct iavf_adapter *adapter =
1274                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1275         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
1276         struct rte_eth_conf *dev_conf = &dev->data->dev_conf;
1277         int err;
1278
1279         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_VLAN_V2)
1280                 return iavf_dev_vlan_offload_set_v2(dev, mask);
1281
1282         if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_VLAN))
1283                 return -ENOTSUP;
1284
1285         /* Vlan stripping setting */
1286         if (mask & RTE_ETH_VLAN_STRIP_MASK) {
1287                 /* Enable or disable VLAN stripping */
1288                 if (dev_conf->rxmode.offloads & RTE_ETH_RX_OFFLOAD_VLAN_STRIP)
1289                         err = iavf_enable_vlan_strip(adapter);
1290                 else
1291                         err = iavf_disable_vlan_strip(adapter);
1292
1293                 if (err)
1294                         return -EIO;
1295         }
1296         return 0;
1297 }
1298
1299 static int
1300 iavf_dev_rss_reta_update(struct rte_eth_dev *dev,
1301                         struct rte_eth_rss_reta_entry64 *reta_conf,
1302                         uint16_t reta_size)
1303 {
1304         struct iavf_adapter *adapter =
1305                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1306         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
1307         uint8_t *lut;
1308         uint16_t i, idx, shift;
1309         int ret;
1310
1311         if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF))
1312                 return -ENOTSUP;
1313
1314         if (reta_size != vf->vf_res->rss_lut_size) {
1315                 PMD_DRV_LOG(ERR, "The size of hash lookup table configured "
1316                         "(%d) doesn't match the number of hardware can "
1317                         "support (%d)", reta_size, vf->vf_res->rss_lut_size);
1318                 return -EINVAL;
1319         }
1320
1321         lut = rte_zmalloc("rss_lut", reta_size, 0);
1322         if (!lut) {
1323                 PMD_DRV_LOG(ERR, "No memory can be allocated");
1324                 return -ENOMEM;
1325         }
1326         /* store the old lut table temporarily */
1327         rte_memcpy(lut, vf->rss_lut, reta_size);
1328
1329         for (i = 0; i < reta_size; i++) {
1330                 idx = i / RTE_ETH_RETA_GROUP_SIZE;
1331                 shift = i % RTE_ETH_RETA_GROUP_SIZE;
1332                 if (reta_conf[idx].mask & (1ULL << shift))
1333                         lut[i] = reta_conf[idx].reta[shift];
1334         }
1335
1336         rte_memcpy(vf->rss_lut, lut, reta_size);
1337         /* send virtchnnl ops to configure rss*/
1338         ret = iavf_configure_rss_lut(adapter);
1339         if (ret) /* revert back */
1340                 rte_memcpy(vf->rss_lut, lut, reta_size);
1341         rte_free(lut);
1342
1343         return ret;
1344 }
1345
1346 static int
1347 iavf_dev_rss_reta_query(struct rte_eth_dev *dev,
1348                        struct rte_eth_rss_reta_entry64 *reta_conf,
1349                        uint16_t reta_size)
1350 {
1351         struct iavf_adapter *adapter =
1352                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1353         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
1354         uint16_t i, idx, shift;
1355
1356         if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF))
1357                 return -ENOTSUP;
1358
1359         if (reta_size != vf->vf_res->rss_lut_size) {
1360                 PMD_DRV_LOG(ERR, "The size of hash lookup table configured "
1361                         "(%d) doesn't match the number of hardware can "
1362                         "support (%d)", reta_size, vf->vf_res->rss_lut_size);
1363                 return -EINVAL;
1364         }
1365
1366         for (i = 0; i < reta_size; i++) {
1367                 idx = i / RTE_ETH_RETA_GROUP_SIZE;
1368                 shift = i % RTE_ETH_RETA_GROUP_SIZE;
1369                 if (reta_conf[idx].mask & (1ULL << shift))
1370                         reta_conf[idx].reta[shift] = vf->rss_lut[i];
1371         }
1372
1373         return 0;
1374 }
1375
1376 static int
1377 iavf_set_rss_key(struct iavf_adapter *adapter, uint8_t *key, uint8_t key_len)
1378 {
1379         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
1380
1381         /* HENA setting, it is enabled by default, no change */
1382         if (!key || key_len == 0) {
1383                 PMD_DRV_LOG(DEBUG, "No key to be configured");
1384                 return 0;
1385         } else if (key_len != vf->vf_res->rss_key_size) {
1386                 PMD_DRV_LOG(ERR, "The size of hash key configured "
1387                         "(%d) doesn't match the size of hardware can "
1388                         "support (%d)", key_len,
1389                         vf->vf_res->rss_key_size);
1390                 return -EINVAL;
1391         }
1392
1393         rte_memcpy(vf->rss_key, key, key_len);
1394
1395         return iavf_configure_rss_key(adapter);
1396 }
1397
1398 static int
1399 iavf_dev_rss_hash_update(struct rte_eth_dev *dev,
1400                         struct rte_eth_rss_conf *rss_conf)
1401 {
1402         struct iavf_adapter *adapter =
1403                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1404         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
1405         int ret;
1406
1407         adapter->dev_data->dev_conf.rx_adv_conf.rss_conf = *rss_conf;
1408
1409         if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF))
1410                 return -ENOTSUP;
1411
1412         /* Set hash key. */
1413         ret = iavf_set_rss_key(adapter, rss_conf->rss_key,
1414                                rss_conf->rss_key_len);
1415         if (ret)
1416                 return ret;
1417
1418         if (rss_conf->rss_hf == 0) {
1419                 vf->rss_hf = 0;
1420                 ret = iavf_set_hena(adapter, 0);
1421
1422                 /* It is a workaround, temporarily allow error to be returned
1423                  * due to possible lack of PF handling for hena = 0.
1424                  */
1425                 if (ret)
1426                         PMD_DRV_LOG(WARNING, "fail to clean existing RSS, lack PF support");
1427                 return 0;
1428         }
1429
1430         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_ADV_RSS_PF) {
1431                 /* Clear existing RSS. */
1432                 ret = iavf_set_hena(adapter, 0);
1433
1434                 /* It is a workaround, temporarily allow error to be returned
1435                  * due to possible lack of PF handling for hena = 0.
1436                  */
1437                 if (ret)
1438                         PMD_DRV_LOG(WARNING, "fail to clean existing RSS,"
1439                                     "lack PF support");
1440
1441                 /* Set new RSS configuration. */
1442                 ret = iavf_rss_hash_set(adapter, rss_conf->rss_hf, true);
1443                 if (ret) {
1444                         PMD_DRV_LOG(ERR, "fail to set new RSS");
1445                         return ret;
1446                 }
1447         } else {
1448                 iavf_config_rss_hf(adapter, rss_conf->rss_hf);
1449         }
1450
1451         return 0;
1452 }
1453
1454 static int
1455 iavf_dev_rss_hash_conf_get(struct rte_eth_dev *dev,
1456                           struct rte_eth_rss_conf *rss_conf)
1457 {
1458         struct iavf_adapter *adapter =
1459                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1460         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
1461
1462         if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF))
1463                 return -ENOTSUP;
1464
1465         rss_conf->rss_hf = vf->rss_hf;
1466
1467         if (!rss_conf->rss_key)
1468                 return 0;
1469
1470         rss_conf->rss_key_len = vf->vf_res->rss_key_size;
1471         rte_memcpy(rss_conf->rss_key, vf->rss_key, rss_conf->rss_key_len);
1472
1473         return 0;
1474 }
1475
1476 static int
1477 iavf_dev_mtu_set(struct rte_eth_dev *dev, uint16_t mtu __rte_unused)
1478 {
1479         /* mtu setting is forbidden if port is start */
1480         if (dev->data->dev_started) {
1481                 PMD_DRV_LOG(ERR, "port must be stopped before configuration");
1482                 return -EBUSY;
1483         }
1484
1485         return 0;
1486 }
1487
1488 static int
1489 iavf_dev_set_default_mac_addr(struct rte_eth_dev *dev,
1490                              struct rte_ether_addr *mac_addr)
1491 {
1492         struct iavf_adapter *adapter =
1493                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1494         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(adapter);
1495         struct rte_ether_addr *old_addr;
1496         int ret;
1497
1498         old_addr = (struct rte_ether_addr *)hw->mac.addr;
1499
1500         if (rte_is_same_ether_addr(old_addr, mac_addr))
1501                 return 0;
1502
1503         ret = iavf_add_del_eth_addr(adapter, old_addr, false, VIRTCHNL_ETHER_ADDR_PRIMARY);
1504         if (ret)
1505                 PMD_DRV_LOG(ERR, "Fail to delete old MAC:"
1506                             RTE_ETHER_ADDR_PRT_FMT,
1507                                 RTE_ETHER_ADDR_BYTES(old_addr));
1508
1509         ret = iavf_add_del_eth_addr(adapter, mac_addr, true, VIRTCHNL_ETHER_ADDR_PRIMARY);
1510         if (ret)
1511                 PMD_DRV_LOG(ERR, "Fail to add new MAC:"
1512                             RTE_ETHER_ADDR_PRT_FMT,
1513                                 RTE_ETHER_ADDR_BYTES(mac_addr));
1514
1515         if (ret)
1516                 return -EIO;
1517
1518         rte_ether_addr_copy(mac_addr, (struct rte_ether_addr *)hw->mac.addr);
1519         return 0;
1520 }
1521
1522 static void
1523 iavf_stat_update_48(uint64_t *offset, uint64_t *stat)
1524 {
1525         if (*stat >= *offset)
1526                 *stat = *stat - *offset;
1527         else
1528                 *stat = (uint64_t)((*stat +
1529                         ((uint64_t)1 << IAVF_48_BIT_WIDTH)) - *offset);
1530
1531         *stat &= IAVF_48_BIT_MASK;
1532 }
1533
1534 static void
1535 iavf_stat_update_32(uint64_t *offset, uint64_t *stat)
1536 {
1537         if (*stat >= *offset)
1538                 *stat = (uint64_t)(*stat - *offset);
1539         else
1540                 *stat = (uint64_t)((*stat +
1541                         ((uint64_t)1 << IAVF_32_BIT_WIDTH)) - *offset);
1542 }
1543
1544 static void
1545 iavf_update_stats(struct iavf_vsi *vsi, struct virtchnl_eth_stats *nes)
1546 {
1547         struct virtchnl_eth_stats *oes = &vsi->eth_stats_offset.eth_stats;
1548
1549         iavf_stat_update_48(&oes->rx_bytes, &nes->rx_bytes);
1550         iavf_stat_update_48(&oes->rx_unicast, &nes->rx_unicast);
1551         iavf_stat_update_48(&oes->rx_multicast, &nes->rx_multicast);
1552         iavf_stat_update_48(&oes->rx_broadcast, &nes->rx_broadcast);
1553         iavf_stat_update_32(&oes->rx_discards, &nes->rx_discards);
1554         iavf_stat_update_48(&oes->tx_bytes, &nes->tx_bytes);
1555         iavf_stat_update_48(&oes->tx_unicast, &nes->tx_unicast);
1556         iavf_stat_update_48(&oes->tx_multicast, &nes->tx_multicast);
1557         iavf_stat_update_48(&oes->tx_broadcast, &nes->tx_broadcast);
1558         iavf_stat_update_32(&oes->tx_errors, &nes->tx_errors);
1559         iavf_stat_update_32(&oes->tx_discards, &nes->tx_discards);
1560 }
1561
1562 static int
1563 iavf_dev_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
1564 {
1565         struct iavf_adapter *adapter =
1566                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1567         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
1568         struct iavf_vsi *vsi = &vf->vsi;
1569         struct virtchnl_eth_stats *pstats = NULL;
1570         int ret;
1571
1572         ret = iavf_query_stats(adapter, &pstats);
1573         if (ret == 0) {
1574                 uint8_t crc_stats_len = (dev->data->dev_conf.rxmode.offloads &
1575                                          RTE_ETH_RX_OFFLOAD_KEEP_CRC) ? 0 :
1576                                          RTE_ETHER_CRC_LEN;
1577                 iavf_update_stats(vsi, pstats);
1578                 stats->ipackets = pstats->rx_unicast + pstats->rx_multicast +
1579                                 pstats->rx_broadcast - pstats->rx_discards;
1580                 stats->opackets = pstats->tx_broadcast + pstats->tx_multicast +
1581                                                 pstats->tx_unicast;
1582                 stats->imissed = pstats->rx_discards;
1583                 stats->oerrors = pstats->tx_errors + pstats->tx_discards;
1584                 stats->ibytes = pstats->rx_bytes;
1585                 stats->ibytes -= stats->ipackets * crc_stats_len;
1586                 stats->obytes = pstats->tx_bytes;
1587         } else {
1588                 PMD_DRV_LOG(ERR, "Get statistics failed");
1589         }
1590         return ret;
1591 }
1592
1593 static int
1594 iavf_dev_stats_reset(struct rte_eth_dev *dev)
1595 {
1596         int ret;
1597         struct iavf_adapter *adapter =
1598                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1599         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
1600         struct iavf_vsi *vsi = &vf->vsi;
1601         struct virtchnl_eth_stats *pstats = NULL;
1602
1603         /* read stat values to clear hardware registers */
1604         ret = iavf_query_stats(adapter, &pstats);
1605         if (ret != 0)
1606                 return ret;
1607
1608         /* set stats offset base on current values */
1609         vsi->eth_stats_offset.eth_stats = *pstats;
1610
1611         return 0;
1612 }
1613
1614 static int
1615 iavf_dev_xstats_reset(struct rte_eth_dev *dev)
1616 {
1617         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
1618
1619         iavf_dev_stats_reset(dev);
1620         memset(&vf->vsi.eth_stats_offset, 0, sizeof(struct iavf_eth_xstats));
1621
1622         return 0;
1623 }
1624
1625 static int iavf_dev_xstats_get_names(__rte_unused struct rte_eth_dev *dev,
1626                                       struct rte_eth_xstat_name *xstats_names,
1627                                       __rte_unused unsigned int limit)
1628 {
1629         unsigned int i;
1630
1631         if (xstats_names != NULL)
1632                 for (i = 0; i < IAVF_NB_XSTATS; i++) {
1633                         snprintf(xstats_names[i].name,
1634                                 sizeof(xstats_names[i].name),
1635                                 "%s", rte_iavf_stats_strings[i].name);
1636                 }
1637         return IAVF_NB_XSTATS;
1638 }
1639
1640 static void
1641 iavf_dev_update_ipsec_xstats(struct rte_eth_dev *ethdev,
1642                 struct iavf_ipsec_crypto_stats *ips)
1643 {
1644         uint16_t idx;
1645         for (idx = 0; idx < ethdev->data->nb_rx_queues; idx++) {
1646                 struct iavf_rx_queue *rxq;
1647                 struct iavf_ipsec_crypto_stats *stats;
1648                 rxq = (struct iavf_rx_queue *)ethdev->data->rx_queues[idx];
1649                 stats = &rxq->stats.ipsec_crypto;
1650                 ips->icount += stats->icount;
1651                 ips->ibytes += stats->ibytes;
1652                 ips->ierrors.count += stats->ierrors.count;
1653                 ips->ierrors.sad_miss += stats->ierrors.sad_miss;
1654                 ips->ierrors.not_processed += stats->ierrors.not_processed;
1655                 ips->ierrors.icv_check += stats->ierrors.icv_check;
1656                 ips->ierrors.ipsec_length += stats->ierrors.ipsec_length;
1657                 ips->ierrors.misc += stats->ierrors.misc;
1658         }
1659 }
1660
1661 static int iavf_dev_xstats_get(struct rte_eth_dev *dev,
1662                                  struct rte_eth_xstat *xstats, unsigned int n)
1663 {
1664         int ret;
1665         unsigned int i;
1666         struct iavf_adapter *adapter =
1667                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1668         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
1669         struct iavf_vsi *vsi = &vf->vsi;
1670         struct virtchnl_eth_stats *pstats = NULL;
1671         struct iavf_eth_xstats iavf_xtats = {{0}};
1672
1673         if (n < IAVF_NB_XSTATS)
1674                 return IAVF_NB_XSTATS;
1675
1676         ret = iavf_query_stats(adapter, &pstats);
1677         if (ret != 0)
1678                 return 0;
1679
1680         if (!xstats)
1681                 return 0;
1682
1683         iavf_update_stats(vsi, pstats);
1684         iavf_xtats.eth_stats = *pstats;
1685
1686         if (iavf_ipsec_crypto_supported(adapter))
1687                 iavf_dev_update_ipsec_xstats(dev, &iavf_xtats.ips_stats);
1688
1689         /* loop over xstats array and values from pstats */
1690         for (i = 0; i < IAVF_NB_XSTATS; i++) {
1691                 xstats[i].id = i;
1692                 xstats[i].value = *(uint64_t *)(((char *)&iavf_xtats) +
1693                         rte_iavf_stats_strings[i].offset);
1694         }
1695
1696         return IAVF_NB_XSTATS;
1697 }
1698
1699
1700 static int
1701 iavf_dev_rx_queue_intr_enable(struct rte_eth_dev *dev, uint16_t queue_id)
1702 {
1703         struct iavf_adapter *adapter =
1704                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1705         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1706         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(adapter);
1707         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
1708         uint16_t msix_intr;
1709
1710         msix_intr = rte_intr_vec_list_index_get(pci_dev->intr_handle,
1711                                                        queue_id);
1712         if (msix_intr == IAVF_MISC_VEC_ID) {
1713                 PMD_DRV_LOG(INFO, "MISC is also enabled for control");
1714                 IAVF_WRITE_REG(hw, IAVF_VFINT_DYN_CTL01,
1715                                IAVF_VFINT_DYN_CTL01_INTENA_MASK |
1716                                IAVF_VFINT_DYN_CTL01_CLEARPBA_MASK |
1717                                IAVF_VFINT_DYN_CTL01_ITR_INDX_MASK);
1718         } else {
1719                 IAVF_WRITE_REG(hw,
1720                                IAVF_VFINT_DYN_CTLN1
1721                                 (msix_intr - IAVF_RX_VEC_START),
1722                                IAVF_VFINT_DYN_CTLN1_INTENA_MASK |
1723                                IAVF_VFINT_DYN_CTL01_CLEARPBA_MASK |
1724                                IAVF_VFINT_DYN_CTLN1_ITR_INDX_MASK);
1725         }
1726
1727         IAVF_WRITE_FLUSH(hw);
1728
1729         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_WB_ON_ITR)
1730                 rte_intr_ack(pci_dev->intr_handle);
1731
1732         return 0;
1733 }
1734
1735 static int
1736 iavf_dev_rx_queue_intr_disable(struct rte_eth_dev *dev, uint16_t queue_id)
1737 {
1738         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1739         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1740         uint16_t msix_intr;
1741
1742         msix_intr = rte_intr_vec_list_index_get(pci_dev->intr_handle,
1743                                                        queue_id);
1744         if (msix_intr == IAVF_MISC_VEC_ID) {
1745                 PMD_DRV_LOG(ERR, "MISC is used for control, cannot disable it");
1746                 return -EIO;
1747         }
1748
1749         IAVF_WRITE_REG(hw,
1750                       IAVF_VFINT_DYN_CTLN1(msix_intr - IAVF_RX_VEC_START),
1751                       0);
1752
1753         IAVF_WRITE_FLUSH(hw);
1754         return 0;
1755 }
1756
1757 static int
1758 iavf_check_vf_reset_done(struct iavf_hw *hw)
1759 {
1760         int i, reset;
1761
1762         for (i = 0; i < IAVF_RESET_WAIT_CNT; i++) {
1763                 reset = IAVF_READ_REG(hw, IAVF_VFGEN_RSTAT) &
1764                         IAVF_VFGEN_RSTAT_VFR_STATE_MASK;
1765                 reset = reset >> IAVF_VFGEN_RSTAT_VFR_STATE_SHIFT;
1766                 if (reset == VIRTCHNL_VFR_VFACTIVE ||
1767                     reset == VIRTCHNL_VFR_COMPLETED)
1768                         break;
1769                 rte_delay_ms(20);
1770         }
1771
1772         if (i >= IAVF_RESET_WAIT_CNT)
1773                 return -1;
1774
1775         return 0;
1776 }
1777
1778 static int
1779 iavf_lookup_proto_xtr_type(const char *flex_name)
1780 {
1781         static struct {
1782                 const char *name;
1783                 enum iavf_proto_xtr_type type;
1784         } xtr_type_map[] = {
1785                 { "vlan",      IAVF_PROTO_XTR_VLAN      },
1786                 { "ipv4",      IAVF_PROTO_XTR_IPV4      },
1787                 { "ipv6",      IAVF_PROTO_XTR_IPV6      },
1788                 { "ipv6_flow", IAVF_PROTO_XTR_IPV6_FLOW },
1789                 { "tcp",       IAVF_PROTO_XTR_TCP       },
1790                 { "ip_offset", IAVF_PROTO_XTR_IP_OFFSET },
1791                 { "ipsec_crypto_said", IAVF_PROTO_XTR_IPSEC_CRYPTO_SAID },
1792         };
1793         uint32_t i;
1794
1795         for (i = 0; i < RTE_DIM(xtr_type_map); i++) {
1796                 if (strcmp(flex_name, xtr_type_map[i].name) == 0)
1797                         return xtr_type_map[i].type;
1798         }
1799
1800         PMD_DRV_LOG(ERR, "wrong proto_xtr type, it should be: "
1801                         "vlan|ipv4|ipv6|ipv6_flow|tcp|ip_offset|ipsec_crypto_said");
1802
1803         return -1;
1804 }
1805
1806 /**
1807  * Parse elem, the elem could be single number/range or '(' ')' group
1808  * 1) A single number elem, it's just a simple digit. e.g. 9
1809  * 2) A single range elem, two digits with a '-' between. e.g. 2-6
1810  * 3) A group elem, combines multiple 1) or 2) with '( )'. e.g (0,2-4,6)
1811  *    Within group elem, '-' used for a range separator;
1812  *                       ',' used for a single number.
1813  */
1814 static int
1815 iavf_parse_queue_set(const char *input, int xtr_type,
1816                      struct iavf_devargs *devargs)
1817 {
1818         const char *str = input;
1819         char *end = NULL;
1820         uint32_t min, max;
1821         uint32_t idx;
1822
1823         while (isblank(*str))
1824                 str++;
1825
1826         if (!isdigit(*str) && *str != '(')
1827                 return -1;
1828
1829         /* process single number or single range of number */
1830         if (*str != '(') {
1831                 errno = 0;
1832                 idx = strtoul(str, &end, 10);
1833                 if (errno || !end || idx >= IAVF_MAX_QUEUE_NUM)
1834                         return -1;
1835
1836                 while (isblank(*end))
1837                         end++;
1838
1839                 min = idx;
1840                 max = idx;
1841
1842                 /* process single <number>-<number> */
1843                 if (*end == '-') {
1844                         end++;
1845                         while (isblank(*end))
1846                                 end++;
1847                         if (!isdigit(*end))
1848                                 return -1;
1849
1850                         errno = 0;
1851                         idx = strtoul(end, &end, 10);
1852                         if (errno || !end || idx >= IAVF_MAX_QUEUE_NUM)
1853                                 return -1;
1854
1855                         max = idx;
1856                         while (isblank(*end))
1857                                 end++;
1858                 }
1859
1860                 if (*end != ':')
1861                         return -1;
1862
1863                 for (idx = RTE_MIN(min, max);
1864                      idx <= RTE_MAX(min, max); idx++)
1865                         devargs->proto_xtr[idx] = xtr_type;
1866
1867                 return 0;
1868         }
1869
1870         /* process set within bracket */
1871         str++;
1872         while (isblank(*str))
1873                 str++;
1874         if (*str == '\0')
1875                 return -1;
1876
1877         min = IAVF_MAX_QUEUE_NUM;
1878         do {
1879                 /* go ahead to the first digit */
1880                 while (isblank(*str))
1881                         str++;
1882                 if (!isdigit(*str))
1883                         return -1;
1884
1885                 /* get the digit value */
1886                 errno = 0;
1887                 idx = strtoul(str, &end, 10);
1888                 if (errno || !end || idx >= IAVF_MAX_QUEUE_NUM)
1889                         return -1;
1890
1891                 /* go ahead to separator '-',',' and ')' */
1892                 while (isblank(*end))
1893                         end++;
1894                 if (*end == '-') {
1895                         if (min == IAVF_MAX_QUEUE_NUM)
1896                                 min = idx;
1897                         else /* avoid continuous '-' */
1898                                 return -1;
1899                 } else if (*end == ',' || *end == ')') {
1900                         max = idx;
1901                         if (min == IAVF_MAX_QUEUE_NUM)
1902                                 min = idx;
1903
1904                         for (idx = RTE_MIN(min, max);
1905                              idx <= RTE_MAX(min, max); idx++)
1906                                 devargs->proto_xtr[idx] = xtr_type;
1907
1908                         min = IAVF_MAX_QUEUE_NUM;
1909                 } else {
1910                         return -1;
1911                 }
1912
1913                 str = end + 1;
1914         } while (*end != ')' && *end != '\0');
1915
1916         return 0;
1917 }
1918
1919 static int
1920 iavf_parse_queue_proto_xtr(const char *queues, struct iavf_devargs *devargs)
1921 {
1922         const char *queue_start;
1923         uint32_t idx;
1924         int xtr_type;
1925         char flex_name[32];
1926
1927         while (isblank(*queues))
1928                 queues++;
1929
1930         if (*queues != '[') {
1931                 xtr_type = iavf_lookup_proto_xtr_type(queues);
1932                 if (xtr_type < 0)
1933                         return -1;
1934
1935                 devargs->proto_xtr_dflt = xtr_type;
1936
1937                 return 0;
1938         }
1939
1940         queues++;
1941         do {
1942                 while (isblank(*queues))
1943                         queues++;
1944                 if (*queues == '\0')
1945                         return -1;
1946
1947                 queue_start = queues;
1948
1949                 /* go across a complete bracket */
1950                 if (*queue_start == '(') {
1951                         queues += strcspn(queues, ")");
1952                         if (*queues != ')')
1953                                 return -1;
1954                 }
1955
1956                 /* scan the separator ':' */
1957                 queues += strcspn(queues, ":");
1958                 if (*queues++ != ':')
1959                         return -1;
1960                 while (isblank(*queues))
1961                         queues++;
1962
1963                 for (idx = 0; ; idx++) {
1964                         if (isblank(queues[idx]) ||
1965                             queues[idx] == ',' ||
1966                             queues[idx] == ']' ||
1967                             queues[idx] == '\0')
1968                                 break;
1969
1970                         if (idx > sizeof(flex_name) - 2)
1971                                 return -1;
1972
1973                         flex_name[idx] = queues[idx];
1974                 }
1975                 flex_name[idx] = '\0';
1976                 xtr_type = iavf_lookup_proto_xtr_type(flex_name);
1977                 if (xtr_type < 0)
1978                         return -1;
1979
1980                 queues += idx;
1981
1982                 while (isblank(*queues) || *queues == ',' || *queues == ']')
1983                         queues++;
1984
1985                 if (iavf_parse_queue_set(queue_start, xtr_type, devargs) < 0)
1986                         return -1;
1987         } while (*queues != '\0');
1988
1989         return 0;
1990 }
1991
1992 static int
1993 iavf_handle_proto_xtr_arg(__rte_unused const char *key, const char *value,
1994                           void *extra_args)
1995 {
1996         struct iavf_devargs *devargs = extra_args;
1997
1998         if (!value || !extra_args)
1999                 return -EINVAL;
2000
2001         if (iavf_parse_queue_proto_xtr(value, devargs) < 0) {
2002                 PMD_DRV_LOG(ERR, "the proto_xtr's parameter is wrong : '%s'",
2003                             value);
2004                 return -1;
2005         }
2006
2007         return 0;
2008 }
2009
2010 static int iavf_parse_devargs(struct rte_eth_dev *dev)
2011 {
2012         struct iavf_adapter *ad =
2013                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
2014         struct rte_devargs *devargs = dev->device->devargs;
2015         struct rte_kvargs *kvlist;
2016         int ret;
2017
2018         if (!devargs)
2019                 return 0;
2020
2021         kvlist = rte_kvargs_parse(devargs->args, iavf_valid_args);
2022         if (!kvlist) {
2023                 PMD_INIT_LOG(ERR, "invalid kvargs key\n");
2024                 return -EINVAL;
2025         }
2026
2027         ad->devargs.proto_xtr_dflt = IAVF_PROTO_XTR_NONE;
2028         memset(ad->devargs.proto_xtr, IAVF_PROTO_XTR_NONE,
2029                sizeof(ad->devargs.proto_xtr));
2030
2031         ret = rte_kvargs_process(kvlist, IAVF_PROTO_XTR_ARG,
2032                                  &iavf_handle_proto_xtr_arg, &ad->devargs);
2033         if (ret)
2034                 goto bail;
2035
2036 bail:
2037         rte_kvargs_free(kvlist);
2038         return ret;
2039 }
2040
2041 static void
2042 iavf_init_proto_xtr(struct rte_eth_dev *dev)
2043 {
2044         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
2045         struct iavf_adapter *ad =
2046                         IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
2047         const struct iavf_proto_xtr_ol *xtr_ol;
2048         bool proto_xtr_enable = false;
2049         int offset;
2050         uint16_t i;
2051
2052         vf->proto_xtr = rte_zmalloc("vf proto xtr",
2053                                     vf->vsi_res->num_queue_pairs, 0);
2054         if (unlikely(!(vf->proto_xtr))) {
2055                 PMD_DRV_LOG(ERR, "no memory for setting up proto_xtr's table");
2056                 return;
2057         }
2058
2059         for (i = 0; i < vf->vsi_res->num_queue_pairs; i++) {
2060                 vf->proto_xtr[i] = ad->devargs.proto_xtr[i] !=
2061                                         IAVF_PROTO_XTR_NONE ?
2062                                         ad->devargs.proto_xtr[i] :
2063                                         ad->devargs.proto_xtr_dflt;
2064
2065                 if (vf->proto_xtr[i] != IAVF_PROTO_XTR_NONE) {
2066                         uint8_t type = vf->proto_xtr[i];
2067
2068                         iavf_proto_xtr_params[type].required = true;
2069                         proto_xtr_enable = true;
2070                 }
2071         }
2072
2073         if (likely(!proto_xtr_enable))
2074                 return;
2075
2076         offset = rte_mbuf_dynfield_register(&iavf_proto_xtr_metadata_param);
2077         if (unlikely(offset == -1)) {
2078                 PMD_DRV_LOG(ERR,
2079                             "failed to extract protocol metadata, error %d",
2080                             -rte_errno);
2081                 return;
2082         }
2083
2084         PMD_DRV_LOG(DEBUG,
2085                     "proto_xtr metadata offset in mbuf is : %d",
2086                     offset);
2087         rte_pmd_ifd_dynfield_proto_xtr_metadata_offs = offset;
2088
2089         for (i = 0; i < RTE_DIM(iavf_proto_xtr_params); i++) {
2090                 xtr_ol = &iavf_proto_xtr_params[i];
2091
2092                 uint8_t rxdid = iavf_proto_xtr_type_to_rxdid((uint8_t)i);
2093
2094                 if (!xtr_ol->required)
2095                         continue;
2096
2097                 if (!(vf->supported_rxdid & BIT(rxdid))) {
2098                         PMD_DRV_LOG(ERR,
2099                                     "rxdid[%u] is not supported in hardware",
2100                                     rxdid);
2101                         rte_pmd_ifd_dynfield_proto_xtr_metadata_offs = -1;
2102                         break;
2103                 }
2104
2105                 offset = rte_mbuf_dynflag_register(&xtr_ol->param);
2106                 if (unlikely(offset == -1)) {
2107                         PMD_DRV_LOG(ERR,
2108                                     "failed to register proto_xtr offload '%s', error %d",
2109                                     xtr_ol->param.name, -rte_errno);
2110
2111                         rte_pmd_ifd_dynfield_proto_xtr_metadata_offs = -1;
2112                         break;
2113                 }
2114
2115                 PMD_DRV_LOG(DEBUG,
2116                             "proto_xtr offload '%s' offset in mbuf is : %d",
2117                             xtr_ol->param.name, offset);
2118                 *xtr_ol->ol_flag = 1ULL << offset;
2119         }
2120 }
2121
2122 static int
2123 iavf_init_vf(struct rte_eth_dev *dev)
2124 {
2125         int err, bufsz;
2126         struct iavf_adapter *adapter =
2127                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
2128         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2129         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
2130
2131         vf->eth_dev = dev;
2132
2133         err = iavf_parse_devargs(dev);
2134         if (err) {
2135                 PMD_INIT_LOG(ERR, "Failed to parse devargs");
2136                 goto err;
2137         }
2138
2139         err = iavf_set_mac_type(hw);
2140         if (err) {
2141                 PMD_INIT_LOG(ERR, "set_mac_type failed: %d", err);
2142                 goto err;
2143         }
2144
2145         err = iavf_check_vf_reset_done(hw);
2146         if (err) {
2147                 PMD_INIT_LOG(ERR, "VF is still resetting");
2148                 goto err;
2149         }
2150
2151         iavf_init_adminq_parameter(hw);
2152         err = iavf_init_adminq(hw);
2153         if (err) {
2154                 PMD_INIT_LOG(ERR, "init_adminq failed: %d", err);
2155                 goto err;
2156         }
2157
2158         vf->aq_resp = rte_zmalloc("vf_aq_resp", IAVF_AQ_BUF_SZ, 0);
2159         if (!vf->aq_resp) {
2160                 PMD_INIT_LOG(ERR, "unable to allocate vf_aq_resp memory");
2161                 goto err_aq;
2162         }
2163         if (iavf_check_api_version(adapter) != 0) {
2164                 PMD_INIT_LOG(ERR, "check_api version failed");
2165                 goto err_api;
2166         }
2167
2168         bufsz = sizeof(struct virtchnl_vf_resource) +
2169                 (IAVF_MAX_VF_VSI * sizeof(struct virtchnl_vsi_resource));
2170         vf->vf_res = rte_zmalloc("vf_res", bufsz, 0);
2171         if (!vf->vf_res) {
2172                 PMD_INIT_LOG(ERR, "unable to allocate vf_res memory");
2173                 goto err_api;
2174         }
2175
2176         if (iavf_get_vf_resource(adapter) != 0) {
2177                 PMD_INIT_LOG(ERR, "iavf_get_vf_config failed");
2178                 goto err_alloc;
2179         }
2180         /* Allocate memort for RSS info */
2181         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF) {
2182                 vf->rss_key = rte_zmalloc("rss_key",
2183                                           vf->vf_res->rss_key_size, 0);
2184                 if (!vf->rss_key) {
2185                         PMD_INIT_LOG(ERR, "unable to allocate rss_key memory");
2186                         goto err_rss;
2187                 }
2188                 vf->rss_lut = rte_zmalloc("rss_lut",
2189                                           vf->vf_res->rss_lut_size, 0);
2190                 if (!vf->rss_lut) {
2191                         PMD_INIT_LOG(ERR, "unable to allocate rss_lut memory");
2192                         goto err_rss;
2193                 }
2194         }
2195
2196         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RX_FLEX_DESC) {
2197                 if (iavf_get_supported_rxdid(adapter) != 0) {
2198                         PMD_INIT_LOG(ERR, "failed to do get supported rxdid");
2199                         goto err_rss;
2200                 }
2201         }
2202
2203         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_VLAN_V2) {
2204                 if (iavf_get_vlan_offload_caps_v2(adapter) != 0) {
2205                         PMD_INIT_LOG(ERR, "failed to do get VLAN offload v2 capabilities");
2206                         goto err_rss;
2207                 }
2208         }
2209
2210         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_QOS) {
2211                 bufsz = sizeof(struct virtchnl_qos_cap_list) +
2212                         IAVF_MAX_TRAFFIC_CLASS *
2213                         sizeof(struct virtchnl_qos_cap_elem);
2214                 vf->qos_cap = rte_zmalloc("qos_cap", bufsz, 0);
2215                 if (!vf->qos_cap) {
2216                         PMD_INIT_LOG(ERR, "unable to allocate qos_cap memory");
2217                         goto err_rss;
2218                 }
2219                 iavf_tm_conf_init(dev);
2220         }
2221
2222         iavf_init_proto_xtr(dev);
2223
2224         return 0;
2225 err_rss:
2226         rte_free(vf->rss_key);
2227         rte_free(vf->rss_lut);
2228 err_alloc:
2229         rte_free(vf->qos_cap);
2230         rte_free(vf->vf_res);
2231         vf->vsi_res = NULL;
2232 err_api:
2233         rte_free(vf->aq_resp);
2234 err_aq:
2235         iavf_shutdown_adminq(hw);
2236 err:
2237         return -1;
2238 }
2239
2240 static void
2241 iavf_uninit_vf(struct rte_eth_dev *dev)
2242 {
2243         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2244         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
2245
2246         iavf_shutdown_adminq(hw);
2247
2248         rte_free(vf->vf_res);
2249         vf->vsi_res = NULL;
2250         vf->vf_res = NULL;
2251
2252         rte_free(vf->aq_resp);
2253         vf->aq_resp = NULL;
2254
2255         rte_free(vf->qos_cap);
2256         vf->qos_cap = NULL;
2257
2258         rte_free(vf->rss_lut);
2259         vf->rss_lut = NULL;
2260         rte_free(vf->rss_key);
2261         vf->rss_key = NULL;
2262 }
2263
2264 /* Enable default admin queue interrupt setting */
2265 static inline void
2266 iavf_enable_irq0(struct iavf_hw *hw)
2267 {
2268         /* Enable admin queue interrupt trigger */
2269         IAVF_WRITE_REG(hw, IAVF_VFINT_ICR0_ENA1,
2270                        IAVF_VFINT_ICR0_ENA1_ADMINQ_MASK);
2271
2272         IAVF_WRITE_REG(hw, IAVF_VFINT_DYN_CTL01,
2273                        IAVF_VFINT_DYN_CTL01_INTENA_MASK |
2274                        IAVF_VFINT_DYN_CTL01_CLEARPBA_MASK |
2275                        IAVF_VFINT_DYN_CTL01_ITR_INDX_MASK);
2276
2277         IAVF_WRITE_FLUSH(hw);
2278 }
2279
2280 static inline void
2281 iavf_disable_irq0(struct iavf_hw *hw)
2282 {
2283         /* Disable all interrupt types */
2284         IAVF_WRITE_REG(hw, IAVF_VFINT_ICR0_ENA1, 0);
2285         IAVF_WRITE_REG(hw, IAVF_VFINT_DYN_CTL01,
2286                        IAVF_VFINT_DYN_CTL01_ITR_INDX_MASK);
2287         IAVF_WRITE_FLUSH(hw);
2288 }
2289
2290 static void
2291 iavf_dev_interrupt_handler(void *param)
2292 {
2293         struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
2294         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2295
2296         iavf_disable_irq0(hw);
2297
2298         iavf_handle_virtchnl_msg(dev);
2299
2300         iavf_enable_irq0(hw);
2301 }
2302
2303 void
2304 iavf_dev_alarm_handler(void *param)
2305 {
2306         struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
2307         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2308         uint32_t icr0;
2309
2310         iavf_disable_irq0(hw);
2311
2312         /* read out interrupt causes */
2313         icr0 = IAVF_READ_REG(hw, IAVF_VFINT_ICR01);
2314
2315         if (icr0 & IAVF_VFINT_ICR01_ADMINQ_MASK) {
2316                 PMD_DRV_LOG(DEBUG, "ICR01_ADMINQ is reported");
2317                 iavf_handle_virtchnl_msg(dev);
2318         }
2319
2320         iavf_enable_irq0(hw);
2321
2322         rte_eal_alarm_set(IAVF_ALARM_INTERVAL,
2323                           iavf_dev_alarm_handler, dev);
2324 }
2325
2326 static int
2327 iavf_dev_flow_ops_get(struct rte_eth_dev *dev,
2328                       const struct rte_flow_ops **ops)
2329 {
2330         if (!dev)
2331                 return -EINVAL;
2332
2333         *ops = &iavf_flow_ops;
2334         return 0;
2335 }
2336
2337 static void
2338 iavf_default_rss_disable(struct iavf_adapter *adapter)
2339 {
2340         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
2341         int ret = 0;
2342
2343         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF) {
2344                 /* Set hena = 0 to ask PF to cleanup all existing RSS. */
2345                 ret = iavf_set_hena(adapter, 0);
2346                 if (ret)
2347                         /* It is a workaround, temporarily allow error to be
2348                          * returned due to possible lack of PF handling for
2349                          * hena = 0.
2350                          */
2351                         PMD_INIT_LOG(WARNING, "fail to disable default RSS,"
2352                                     "lack PF support");
2353         }
2354 }
2355
2356 static int
2357 iavf_dev_init(struct rte_eth_dev *eth_dev)
2358 {
2359         struct iavf_adapter *adapter =
2360                 IAVF_DEV_PRIVATE_TO_ADAPTER(eth_dev->data->dev_private);
2361         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(adapter);
2362         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
2363         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
2364         int ret = 0;
2365
2366         PMD_INIT_FUNC_TRACE();
2367
2368         /* assign ops func pointer */
2369         eth_dev->dev_ops = &iavf_eth_dev_ops;
2370         eth_dev->rx_queue_count = iavf_dev_rxq_count;
2371         eth_dev->rx_descriptor_status = iavf_dev_rx_desc_status;
2372         eth_dev->tx_descriptor_status = iavf_dev_tx_desc_status;
2373         eth_dev->rx_pkt_burst = &iavf_recv_pkts;
2374         eth_dev->tx_pkt_burst = &iavf_xmit_pkts;
2375         eth_dev->tx_pkt_prepare = &iavf_prep_pkts;
2376
2377         /* For secondary processes, we don't initialise any further as primary
2378          * has already done this work. Only check if we need a different RX
2379          * and TX function.
2380          */
2381         if (rte_eal_process_type() != RTE_PROC_PRIMARY) {
2382                 iavf_set_rx_function(eth_dev);
2383                 iavf_set_tx_function(eth_dev);
2384                 return 0;
2385         }
2386         rte_eth_copy_pci_info(eth_dev, pci_dev);
2387
2388         hw->vendor_id = pci_dev->id.vendor_id;
2389         hw->device_id = pci_dev->id.device_id;
2390         hw->subsystem_vendor_id = pci_dev->id.subsystem_vendor_id;
2391         hw->subsystem_device_id = pci_dev->id.subsystem_device_id;
2392         hw->bus.bus_id = pci_dev->addr.bus;
2393         hw->bus.device = pci_dev->addr.devid;
2394         hw->bus.func = pci_dev->addr.function;
2395         hw->hw_addr = (void *)pci_dev->mem_resource[0].addr;
2396         hw->back = IAVF_DEV_PRIVATE_TO_ADAPTER(eth_dev->data->dev_private);
2397         adapter->dev_data = eth_dev->data;
2398         adapter->stopped = 1;
2399
2400         if (iavf_init_vf(eth_dev) != 0) {
2401                 PMD_INIT_LOG(ERR, "Init vf failed");
2402                 return -1;
2403         }
2404
2405         /* set default ptype table */
2406         iavf_set_default_ptype_table(eth_dev);
2407
2408         /* copy mac addr */
2409         eth_dev->data->mac_addrs = rte_zmalloc(
2410                 "iavf_mac", RTE_ETHER_ADDR_LEN * IAVF_NUM_MACADDR_MAX, 0);
2411         if (!eth_dev->data->mac_addrs) {
2412                 PMD_INIT_LOG(ERR, "Failed to allocate %d bytes needed to"
2413                              " store MAC addresses",
2414                              RTE_ETHER_ADDR_LEN * IAVF_NUM_MACADDR_MAX);
2415                 ret = -ENOMEM;
2416                 goto init_vf_err;
2417         }
2418         /* If the MAC address is not configured by host,
2419          * generate a random one.
2420          */
2421         if (!rte_is_valid_assigned_ether_addr(
2422                         (struct rte_ether_addr *)hw->mac.addr))
2423                 rte_eth_random_addr(hw->mac.addr);
2424         rte_ether_addr_copy((struct rte_ether_addr *)hw->mac.addr,
2425                         &eth_dev->data->mac_addrs[0]);
2426
2427         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_WB_ON_ITR) {
2428                 /* register callback func to eal lib */
2429                 rte_intr_callback_register(pci_dev->intr_handle,
2430                                            iavf_dev_interrupt_handler,
2431                                            (void *)eth_dev);
2432
2433                 /* enable uio intr after callback register */
2434                 rte_intr_enable(pci_dev->intr_handle);
2435         } else {
2436                 rte_eal_alarm_set(IAVF_ALARM_INTERVAL,
2437                                   iavf_dev_alarm_handler, eth_dev);
2438         }
2439
2440         /* configure and enable device interrupt */
2441         iavf_enable_irq0(hw);
2442
2443         ret = iavf_flow_init(adapter);
2444         if (ret) {
2445                 PMD_INIT_LOG(ERR, "Failed to initialize flow");
2446                 goto flow_init_err;
2447         }
2448
2449         /** Check if the IPsec Crypto offload is supported and create
2450          *  security_ctx if it is.
2451          */
2452         if (iavf_ipsec_crypto_supported(adapter)) {
2453                 /* Initialize security_ctx only for primary process*/
2454                 ret = iavf_security_ctx_create(adapter);
2455                 if (ret) {
2456                         PMD_INIT_LOG(ERR, "failed to create ipsec crypto security instance");
2457                         return ret;
2458                 }
2459
2460                 ret = iavf_security_init(adapter);
2461                 if (ret) {
2462                         PMD_INIT_LOG(ERR, "failed to initialized ipsec crypto resources");
2463                         return ret;
2464                 }
2465         }
2466
2467         iavf_default_rss_disable(adapter);
2468
2469         return 0;
2470
2471 flow_init_err:
2472         rte_free(eth_dev->data->mac_addrs);
2473         eth_dev->data->mac_addrs = NULL;
2474
2475 init_vf_err:
2476         iavf_uninit_vf(eth_dev);
2477
2478         return ret;
2479 }
2480
2481 static int
2482 iavf_dev_close(struct rte_eth_dev *dev)
2483 {
2484         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2485         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
2486         struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
2487         struct iavf_adapter *adapter =
2488                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
2489         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
2490         int ret;
2491
2492         if (rte_eal_process_type() != RTE_PROC_PRIMARY)
2493                 return 0;
2494
2495         ret = iavf_dev_stop(dev);
2496
2497         iavf_flow_flush(dev, NULL);
2498         iavf_flow_uninit(adapter);
2499
2500         /*
2501          * disable promiscuous mode before reset vf
2502          * it is a workaround solution when work with kernel driver
2503          * and it is not the normal way
2504          */
2505         if (vf->promisc_unicast_enabled || vf->promisc_multicast_enabled)
2506                 iavf_config_promisc(adapter, false, false);
2507
2508         iavf_shutdown_adminq(hw);
2509         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_WB_ON_ITR) {
2510                 /* disable uio intr before callback unregister */
2511                 rte_intr_disable(intr_handle);
2512
2513                 /* unregister callback func from eal lib */
2514                 rte_intr_callback_unregister(intr_handle,
2515                                              iavf_dev_interrupt_handler, dev);
2516         } else {
2517                 rte_eal_alarm_cancel(iavf_dev_alarm_handler, dev);
2518         }
2519         iavf_disable_irq0(hw);
2520
2521         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_QOS)
2522                 iavf_tm_conf_uninit(dev);
2523
2524         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF) {
2525                 if (vf->rss_lut) {
2526                         rte_free(vf->rss_lut);
2527                         vf->rss_lut = NULL;
2528                 }
2529                 if (vf->rss_key) {
2530                         rte_free(vf->rss_key);
2531                         vf->rss_key = NULL;
2532                 }
2533         }
2534
2535         rte_free(vf->vf_res);
2536         vf->vsi_res = NULL;
2537         vf->vf_res = NULL;
2538
2539         rte_free(vf->aq_resp);
2540         vf->aq_resp = NULL;
2541
2542         /*
2543          * If the VF is reset via VFLR, the device will be knocked out of bus
2544          * master mode, and the driver will fail to recover from the reset. Fix
2545          * this by enabling bus mastering after every reset. In a non-VFLR case,
2546          * the bus master bit will not be disabled, and this call will have no
2547          * effect.
2548          */
2549         if (vf->vf_reset && !rte_pci_set_bus_master(pci_dev, true))
2550                 vf->vf_reset = false;
2551
2552         return ret;
2553 }
2554
2555 static int
2556 iavf_dev_uninit(struct rte_eth_dev *dev)
2557 {
2558         if (rte_eal_process_type() != RTE_PROC_PRIMARY)
2559                 return -EPERM;
2560
2561         iavf_dev_close(dev);
2562
2563         return 0;
2564 }
2565
2566 /*
2567  * Reset VF device only to re-initialize resources in PMD layer
2568  */
2569 static int
2570 iavf_dev_reset(struct rte_eth_dev *dev)
2571 {
2572         int ret;
2573
2574         ret = iavf_dev_uninit(dev);
2575         if (ret)
2576                 return ret;
2577
2578         return iavf_dev_init(dev);
2579 }
2580
2581 static int
2582 iavf_dcf_cap_check_handler(__rte_unused const char *key,
2583                            const char *value, __rte_unused void *opaque)
2584 {
2585         if (strcmp(value, "dcf"))
2586                 return -1;
2587
2588         return 0;
2589 }
2590
2591 static int
2592 iavf_dcf_cap_selected(struct rte_devargs *devargs)
2593 {
2594         struct rte_kvargs *kvlist;
2595         const char *key = "cap";
2596         int ret = 0;
2597
2598         if (devargs == NULL)
2599                 return 0;
2600
2601         kvlist = rte_kvargs_parse(devargs->args, NULL);
2602         if (kvlist == NULL)
2603                 return 0;
2604
2605         if (!rte_kvargs_count(kvlist, key))
2606                 goto exit;
2607
2608         /* dcf capability selected when there's a key-value pair: cap=dcf */
2609         if (rte_kvargs_process(kvlist, key,
2610                                iavf_dcf_cap_check_handler, NULL) < 0)
2611                 goto exit;
2612
2613         ret = 1;
2614
2615 exit:
2616         rte_kvargs_free(kvlist);
2617         return ret;
2618 }
2619
2620 static int eth_iavf_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
2621                              struct rte_pci_device *pci_dev)
2622 {
2623         if (iavf_dcf_cap_selected(pci_dev->device.devargs))
2624                 return 1;
2625
2626         return rte_eth_dev_pci_generic_probe(pci_dev,
2627                 sizeof(struct iavf_adapter), iavf_dev_init);
2628 }
2629
2630 static int eth_iavf_pci_remove(struct rte_pci_device *pci_dev)
2631 {
2632         return rte_eth_dev_pci_generic_remove(pci_dev, iavf_dev_uninit);
2633 }
2634
2635 /* Adaptive virtual function driver struct */
2636 static struct rte_pci_driver rte_iavf_pmd = {
2637         .id_table = pci_id_iavf_map,
2638         .drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC,
2639         .probe = eth_iavf_pci_probe,
2640         .remove = eth_iavf_pci_remove,
2641 };
2642
2643 RTE_PMD_REGISTER_PCI(net_iavf, rte_iavf_pmd);
2644 RTE_PMD_REGISTER_PCI_TABLE(net_iavf, pci_id_iavf_map);
2645 RTE_PMD_REGISTER_KMOD_DEP(net_iavf, "* igb_uio | vfio-pci");
2646 RTE_PMD_REGISTER_PARAM_STRING(net_iavf, "cap=dcf");
2647 RTE_LOG_REGISTER_SUFFIX(iavf_logtype_init, init, NOTICE);
2648 RTE_LOG_REGISTER_SUFFIX(iavf_logtype_driver, driver, NOTICE);
2649 #ifdef RTE_ETHDEV_DEBUG_RX
2650 RTE_LOG_REGISTER_SUFFIX(iavf_logtype_rx, rx, DEBUG);
2651 #endif
2652 #ifdef RTE_ETHDEV_DEBUG_TX
2653 RTE_LOG_REGISTER_SUFFIX(iavf_logtype_tx, tx, DEBUG);
2654 #endif