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 #define IAVF_QUANTA_SIZE_ARG       "quanta_size"
  38
  39 uint64_t iavf_timestamp_dynflag;
  40 int iavf_timestamp_dynfield_offset = -1;
  41
  42 static const char * const iavf_valid_args[] = {
  43         IAVF_PROTO_XTR_ARG,
  44         IAVF_QUANTA_SIZE_ARG,
  45         NULL
  46 };
  47
  48 static const struct rte_mbuf_dynfield iavf_proto_xtr_metadata_param = {
  49         .name = "intel_pmd_dynfield_proto_xtr_metadata",
  50         .size = sizeof(uint32_t),
  51         .align = __alignof__(uint32_t),
  52         .flags = 0,
  53 };
  54
  55 struct iavf_proto_xtr_ol {
  56         const struct rte_mbuf_dynflag param;
  57         uint64_t *ol_flag;
  58         bool required;
  59 };
  60
  61 static struct iavf_proto_xtr_ol iavf_proto_xtr_params[] = {
  62         [IAVF_PROTO_XTR_VLAN] = {
  63                 .param = { .name = "intel_pmd_dynflag_proto_xtr_vlan" },
  64                 .ol_flag = &rte_pmd_ifd_dynflag_proto_xtr_vlan_mask },
  65         [IAVF_PROTO_XTR_IPV4] = {
  66                 .param = { .name = "intel_pmd_dynflag_proto_xtr_ipv4" },
  67                 .ol_flag = &rte_pmd_ifd_dynflag_proto_xtr_ipv4_mask },
  68         [IAVF_PROTO_XTR_IPV6] = {
  69                 .param = { .name = "intel_pmd_dynflag_proto_xtr_ipv6" },
  70                 .ol_flag = &rte_pmd_ifd_dynflag_proto_xtr_ipv6_mask },
  71         [IAVF_PROTO_XTR_IPV6_FLOW] = {
  72                 .param = { .name = "intel_pmd_dynflag_proto_xtr_ipv6_flow" },
  73                 .ol_flag = &rte_pmd_ifd_dynflag_proto_xtr_ipv6_flow_mask },
  74         [IAVF_PROTO_XTR_TCP] = {
  75                 .param = { .name = "intel_pmd_dynflag_proto_xtr_tcp" },
  76                 .ol_flag = &rte_pmd_ifd_dynflag_proto_xtr_tcp_mask },
  77         [IAVF_PROTO_XTR_IP_OFFSET] = {
  78                 .param = { .name = "intel_pmd_dynflag_proto_xtr_ip_offset" },
  79                 .ol_flag = &rte_pmd_ifd_dynflag_proto_xtr_ip_offset_mask },
  80         [IAVF_PROTO_XTR_IPSEC_CRYPTO_SAID] = {
  81                 .param = {
  82                 .name = "intel_pmd_dynflag_proto_xtr_ipsec_crypto_said" },
  83                 .ol_flag =
  84                         &rte_pmd_ifd_dynflag_proto_xtr_ipsec_crypto_said_mask },
  85 };
  86
  87 static int iavf_dev_configure(struct rte_eth_dev *dev);
  88 static int iavf_dev_start(struct rte_eth_dev *dev);
  89 static int iavf_dev_stop(struct rte_eth_dev *dev);
  90 static int iavf_dev_close(struct rte_eth_dev *dev);
  91 static int iavf_dev_reset(struct rte_eth_dev *dev);
  92 static int iavf_dev_info_get(struct rte_eth_dev *dev,
  93                              struct rte_eth_dev_info *dev_info);
  94 static const uint32_t *iavf_dev_supported_ptypes_get(struct rte_eth_dev *dev);
  95 static int iavf_dev_stats_get(struct rte_eth_dev *dev,
  96                              struct rte_eth_stats *stats);
  97 static int iavf_dev_stats_reset(struct rte_eth_dev *dev);
  98 static int iavf_dev_xstats_reset(struct rte_eth_dev *dev);
  99 static int iavf_dev_xstats_get(struct rte_eth_dev *dev,
 100                                  struct rte_eth_xstat *xstats, unsigned int n);
 101 static int iavf_dev_xstats_get_names(struct rte_eth_dev *dev,
 102                                        struct rte_eth_xstat_name *xstats_names,
 103                                        unsigned int limit);
 104 static int iavf_dev_promiscuous_enable(struct rte_eth_dev *dev);
 105 static int iavf_dev_promiscuous_disable(struct rte_eth_dev *dev);
 106 static int iavf_dev_allmulticast_enable(struct rte_eth_dev *dev);
 107 static int iavf_dev_allmulticast_disable(struct rte_eth_dev *dev);
 108 static int iavf_dev_add_mac_addr(struct rte_eth_dev *dev,
 109                                 struct rte_ether_addr *addr,
 110                                 uint32_t index,
 111                                 uint32_t pool);
 112 static void iavf_dev_del_mac_addr(struct rte_eth_dev *dev, uint32_t index);
 113 static int iavf_dev_vlan_filter_set(struct rte_eth_dev *dev,
 114                                    uint16_t vlan_id, int on);
 115 static int iavf_dev_vlan_offload_set(struct rte_eth_dev *dev, int mask);
 116 static int iavf_dev_rss_reta_update(struct rte_eth_dev *dev,
 117                                    struct rte_eth_rss_reta_entry64 *reta_conf,
 118                                    uint16_t reta_size);
 119 static int iavf_dev_rss_reta_query(struct rte_eth_dev *dev,
 120                                   struct rte_eth_rss_reta_entry64 *reta_conf,
 121                                   uint16_t reta_size);
 122 static int iavf_dev_rss_hash_update(struct rte_eth_dev *dev,
 123                                    struct rte_eth_rss_conf *rss_conf);
 124 static int iavf_dev_rss_hash_conf_get(struct rte_eth_dev *dev,
 125                                      struct rte_eth_rss_conf *rss_conf);
 126 static int iavf_dev_mtu_set(struct rte_eth_dev *dev, uint16_t mtu);
 127 static int iavf_dev_set_default_mac_addr(struct rte_eth_dev *dev,
 128                                          struct rte_ether_addr *mac_addr);
 129 static int iavf_dev_rx_queue_intr_enable(struct rte_eth_dev *dev,
 130                                         uint16_t queue_id);
 131 static int iavf_dev_rx_queue_intr_disable(struct rte_eth_dev *dev,
 132                                          uint16_t queue_id);
 133 static int iavf_dev_flow_ops_get(struct rte_eth_dev *dev,
 134                                  const struct rte_flow_ops **ops);
 135 static int iavf_set_mc_addr_list(struct rte_eth_dev *dev,
 136                         struct rte_ether_addr *mc_addrs,
 137                         uint32_t mc_addrs_num);
 138 static int iavf_tm_ops_get(struct rte_eth_dev *dev __rte_unused, void *arg);
 139
 140 static const struct rte_pci_id pci_id_iavf_map[] = {
 141         { RTE_PCI_DEVICE(IAVF_INTEL_VENDOR_ID, IAVF_DEV_ID_ADAPTIVE_VF) },
 142         { RTE_PCI_DEVICE(IAVF_INTEL_VENDOR_ID, IAVF_DEV_ID_VF) },
 143         { RTE_PCI_DEVICE(IAVF_INTEL_VENDOR_ID, IAVF_DEV_ID_VF_HV) },
 144         { RTE_PCI_DEVICE(IAVF_INTEL_VENDOR_ID, IAVF_DEV_ID_X722_VF) },
 145         { RTE_PCI_DEVICE(IAVF_INTEL_VENDOR_ID, IAVF_DEV_ID_X722_A0_VF) },
 146         { .vendor_id = 0, /* sentinel */ },
 147 };
 148
 149 struct rte_iavf_xstats_name_off {
 150         char name[RTE_ETH_XSTATS_NAME_SIZE];
 151         unsigned int offset;
 152 };
 153
 154 #define _OFF_OF(a) offsetof(struct iavf_eth_xstats, a)
 155 static const struct rte_iavf_xstats_name_off rte_iavf_stats_strings[] = {
 156         {"rx_bytes", _OFF_OF(eth_stats.rx_bytes)},
 157         {"rx_unicast_packets", _OFF_OF(eth_stats.rx_unicast)},
 158         {"rx_multicast_packets", _OFF_OF(eth_stats.rx_multicast)},
 159         {"rx_broadcast_packets", _OFF_OF(eth_stats.rx_broadcast)},
 160         {"rx_dropped_packets", _OFF_OF(eth_stats.rx_discards)},
 161         {"rx_unknown_protocol_packets", offsetof(struct iavf_eth_stats,
 162                 rx_unknown_protocol)},
 163         {"tx_bytes", _OFF_OF(eth_stats.tx_bytes)},
 164         {"tx_unicast_packets", _OFF_OF(eth_stats.tx_unicast)},
 165         {"tx_multicast_packets", _OFF_OF(eth_stats.tx_multicast)},
 166         {"tx_broadcast_packets", _OFF_OF(eth_stats.tx_broadcast)},
 167         {"tx_dropped_packets", _OFF_OF(eth_stats.tx_discards)},
 168         {"tx_error_packets", _OFF_OF(eth_stats.tx_errors)},
 169
 170         {"inline_ipsec_crypto_ipackets", _OFF_OF(ips_stats.icount)},
 171         {"inline_ipsec_crypto_ibytes", _OFF_OF(ips_stats.ibytes)},
 172         {"inline_ipsec_crypto_ierrors", _OFF_OF(ips_stats.ierrors.count)},
 173         {"inline_ipsec_crypto_ierrors_sad_lookup",
 174                         _OFF_OF(ips_stats.ierrors.sad_miss)},
 175         {"inline_ipsec_crypto_ierrors_not_processed",
 176                         _OFF_OF(ips_stats.ierrors.not_processed)},
 177         {"inline_ipsec_crypto_ierrors_icv_fail",
 178                         _OFF_OF(ips_stats.ierrors.icv_check)},
 179         {"inline_ipsec_crypto_ierrors_length",
 180                         _OFF_OF(ips_stats.ierrors.ipsec_length)},
 181         {"inline_ipsec_crypto_ierrors_misc",
 182                         _OFF_OF(ips_stats.ierrors.misc)},
 183 };
 184 #undef _OFF_OF
 185
 186 #define IAVF_NB_XSTATS (sizeof(rte_iavf_stats_strings) / \
 187                 sizeof(rte_iavf_stats_strings[0]))
 188
 189 static const struct eth_dev_ops iavf_eth_dev_ops = {
 190         .dev_configure              = iavf_dev_configure,
 191         .dev_start                  = iavf_dev_start,
 192         .dev_stop                   = iavf_dev_stop,
 193         .dev_close                  = iavf_dev_close,
 194         .dev_reset                  = iavf_dev_reset,
 195         .dev_infos_get              = iavf_dev_info_get,
 196         .dev_supported_ptypes_get   = iavf_dev_supported_ptypes_get,
 197         .link_update                = iavf_dev_link_update,
 198         .stats_get                  = iavf_dev_stats_get,
 199         .stats_reset                = iavf_dev_stats_reset,
 200         .xstats_get                 = iavf_dev_xstats_get,
 201         .xstats_get_names           = iavf_dev_xstats_get_names,
 202         .xstats_reset               = iavf_dev_xstats_reset,
 203         .promiscuous_enable         = iavf_dev_promiscuous_enable,
 204         .promiscuous_disable        = iavf_dev_promiscuous_disable,
 205         .allmulticast_enable        = iavf_dev_allmulticast_enable,
 206         .allmulticast_disable       = iavf_dev_allmulticast_disable,
 207         .mac_addr_add               = iavf_dev_add_mac_addr,
 208         .mac_addr_remove            = iavf_dev_del_mac_addr,
 209         .set_mc_addr_list                       = iavf_set_mc_addr_list,
 210         .vlan_filter_set            = iavf_dev_vlan_filter_set,
 211         .vlan_offload_set           = iavf_dev_vlan_offload_set,
 212         .rx_queue_start             = iavf_dev_rx_queue_start,
 213         .rx_queue_stop              = iavf_dev_rx_queue_stop,
 214         .tx_queue_start             = iavf_dev_tx_queue_start,
 215         .tx_queue_stop              = iavf_dev_tx_queue_stop,
 216         .rx_queue_setup             = iavf_dev_rx_queue_setup,
 217         .rx_queue_release           = iavf_dev_rx_queue_release,
 218         .tx_queue_setup             = iavf_dev_tx_queue_setup,
 219         .tx_queue_release           = iavf_dev_tx_queue_release,
 220         .mac_addr_set               = iavf_dev_set_default_mac_addr,
 221         .reta_update                = iavf_dev_rss_reta_update,
 222         .reta_query                 = iavf_dev_rss_reta_query,
 223         .rss_hash_update            = iavf_dev_rss_hash_update,
 224         .rss_hash_conf_get          = iavf_dev_rss_hash_conf_get,
 225         .rxq_info_get               = iavf_dev_rxq_info_get,
 226         .txq_info_get               = iavf_dev_txq_info_get,
 227         .mtu_set                    = iavf_dev_mtu_set,
 228         .rx_queue_intr_enable       = iavf_dev_rx_queue_intr_enable,
 229         .rx_queue_intr_disable      = iavf_dev_rx_queue_intr_disable,
 230         .flow_ops_get               = iavf_dev_flow_ops_get,
 231         .tx_done_cleanup            = iavf_dev_tx_done_cleanup,
 232         .get_monitor_addr           = iavf_get_monitor_addr,
 233         .tm_ops_get                 = iavf_tm_ops_get,
 234 };
 235
 236 static int
 237 iavf_tm_ops_get(struct rte_eth_dev *dev,
 238                         void *arg)
 239 {
 240         struct iavf_adapter *adapter =
 241                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 242
 243         if (adapter->closed)
 244                 return -EIO;
 245
 246         if (!arg)
 247                 return -EINVAL;
 248
 249         *(const void **)arg = &iavf_tm_ops;
 250
 251         return 0;
 252 }
 253
 254 __rte_unused
 255 static int
 256 iavf_vfr_inprogress(struct iavf_hw *hw)
 257 {
 258         int inprogress = 0;
 259
 260         if ((IAVF_READ_REG(hw, IAVF_VFGEN_RSTAT) &
 261                 IAVF_VFGEN_RSTAT_VFR_STATE_MASK) ==
 262                 VIRTCHNL_VFR_INPROGRESS)
 263                 inprogress = 1;
 264
 265         if (inprogress)
 266                 PMD_DRV_LOG(INFO, "Watchdog detected VFR in progress");
 267
 268         return inprogress;
 269 }
 270
 271 __rte_unused
 272 static void
 273 iavf_dev_watchdog(void *cb_arg)
 274 {
 275         struct iavf_adapter *adapter = cb_arg;
 276         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(adapter);
 277         int vfr_inprogress = 0, rc = 0;
 278
 279         /* check if watchdog has been disabled since last call */
 280         if (!adapter->vf.watchdog_enabled)
 281                 return;
 282
 283         /* If in reset then poll vfr_inprogress register for completion */
 284         if (adapter->vf.vf_reset) {
 285                 vfr_inprogress = iavf_vfr_inprogress(hw);
 286
 287                 if (!vfr_inprogress) {
 288                         PMD_DRV_LOG(INFO, "VF \"%s\" reset has completed",
 289                                 adapter->vf.eth_dev->data->name);
 290                         adapter->vf.vf_reset = false;
 291                 }
 292         /* If not in reset then poll vfr_inprogress register for VFLR event */
 293         } else {
 294                 vfr_inprogress = iavf_vfr_inprogress(hw);
 295
 296                 if (vfr_inprogress) {
 297                         PMD_DRV_LOG(INFO,
 298                                 "VF \"%s\" reset event detected by watchdog",
 299                                 adapter->vf.eth_dev->data->name);
 300
 301                         /* enter reset state with VFLR event */
 302                         adapter->vf.vf_reset = true;
 303
 304                         rte_eth_dev_callback_process(adapter->vf.eth_dev,
 305                                 RTE_ETH_EVENT_INTR_RESET, NULL);
 306                 }
 307         }
 308
 309         /* re-alarm watchdog */
 310         rc = rte_eal_alarm_set(IAVF_DEV_WATCHDOG_PERIOD,
 311                         &iavf_dev_watchdog, cb_arg);
 312
 313         if (rc)
 314                 PMD_DRV_LOG(ERR, "Failed \"%s\" to reset device watchdog alarm",
 315                         adapter->vf.eth_dev->data->name);
 316 }
 317
 318 static void
 319 iavf_dev_watchdog_enable(struct iavf_adapter *adapter __rte_unused)
 320 {
 321 #if (IAVF_DEV_WATCHDOG_PERIOD > 0)
 322         PMD_DRV_LOG(INFO, "Enabling device watchdog");
 323         adapter->vf.watchdog_enabled = true;
 324         if (rte_eal_alarm_set(IAVF_DEV_WATCHDOG_PERIOD,
 325                         &iavf_dev_watchdog, (void *)adapter))
 326                 PMD_DRV_LOG(ERR, "Failed to enabled device watchdog");
 327 #endif
 328 }
 329
 330 static void
 331 iavf_dev_watchdog_disable(struct iavf_adapter *adapter __rte_unused)
 332 {
 333 #if (IAVF_DEV_WATCHDOG_PERIOD > 0)
 334         PMD_DRV_LOG(INFO, "Disabling device watchdog");
 335         adapter->vf.watchdog_enabled = false;
 336 #endif
 337 }
 338
 339 static int
 340 iavf_set_mc_addr_list(struct rte_eth_dev *dev,
 341                         struct rte_ether_addr *mc_addrs,
 342                         uint32_t mc_addrs_num)
 343 {
 344         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
 345         struct iavf_adapter *adapter =
 346                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 347         int err, ret;
 348
 349         if (mc_addrs_num > IAVF_NUM_MACADDR_MAX) {
 350                 PMD_DRV_LOG(ERR,
 351                             "can't add more than a limited number (%u) of addresses.",
 352                             (uint32_t)IAVF_NUM_MACADDR_MAX);
 353                 return -EINVAL;
 354         }
 355
 356         if (adapter->closed)
 357                 return -EIO;
 358
 359         /* flush previous addresses */
 360         err = iavf_add_del_mc_addr_list(adapter, vf->mc_addrs, vf->mc_addrs_num,
 361                                         false);
 362         if (err)
 363                 return err;
 364
 365         /* add new ones */
 366         err = iavf_add_del_mc_addr_list(adapter, mc_addrs, mc_addrs_num, true);
 367
 368         if (err) {
 369                 /* if adding mac address list fails, should add the previous
 370                  * addresses back.
 371                  */
 372                 ret = iavf_add_del_mc_addr_list(adapter, vf->mc_addrs,
 373                                                 vf->mc_addrs_num, true);
 374                 if (ret)
 375                         return ret;
 376         } else {
 377                 vf->mc_addrs_num = mc_addrs_num;
 378                 memcpy(vf->mc_addrs,
 379                        mc_addrs, mc_addrs_num * sizeof(*mc_addrs));
 380         }
 381
 382         return err;
 383 }
 384
 385 static void
 386 iavf_config_rss_hf(struct iavf_adapter *adapter, uint64_t rss_hf)
 387 {
 388         static const uint64_t map_hena_rss[] = {
 389                 /* IPv4 */
 390                 [IAVF_FILTER_PCTYPE_NONF_UNICAST_IPV4_UDP] =
 391                                 RTE_ETH_RSS_NONFRAG_IPV4_UDP,
 392                 [IAVF_FILTER_PCTYPE_NONF_MULTICAST_IPV4_UDP] =
 393                                 RTE_ETH_RSS_NONFRAG_IPV4_UDP,
 394                 [IAVF_FILTER_PCTYPE_NONF_IPV4_UDP] =
 395                                 RTE_ETH_RSS_NONFRAG_IPV4_UDP,
 396                 [IAVF_FILTER_PCTYPE_NONF_IPV4_TCP_SYN_NO_ACK] =
 397                                 RTE_ETH_RSS_NONFRAG_IPV4_TCP,
 398                 [IAVF_FILTER_PCTYPE_NONF_IPV4_TCP] =
 399                                 RTE_ETH_RSS_NONFRAG_IPV4_TCP,
 400                 [IAVF_FILTER_PCTYPE_NONF_IPV4_SCTP] =
 401                                 RTE_ETH_RSS_NONFRAG_IPV4_SCTP,
 402                 [IAVF_FILTER_PCTYPE_NONF_IPV4_OTHER] =
 403                                 RTE_ETH_RSS_NONFRAG_IPV4_OTHER,
 404                 [IAVF_FILTER_PCTYPE_FRAG_IPV4] = RTE_ETH_RSS_FRAG_IPV4,
 405
 406                 /* IPv6 */
 407                 [IAVF_FILTER_PCTYPE_NONF_UNICAST_IPV6_UDP] =
 408                                 RTE_ETH_RSS_NONFRAG_IPV6_UDP,
 409                 [IAVF_FILTER_PCTYPE_NONF_MULTICAST_IPV6_UDP] =
 410                                 RTE_ETH_RSS_NONFRAG_IPV6_UDP,
 411                 [IAVF_FILTER_PCTYPE_NONF_IPV6_UDP] =
 412                                 RTE_ETH_RSS_NONFRAG_IPV6_UDP,
 413                 [IAVF_FILTER_PCTYPE_NONF_IPV6_TCP_SYN_NO_ACK] =
 414                                 RTE_ETH_RSS_NONFRAG_IPV6_TCP,
 415                 [IAVF_FILTER_PCTYPE_NONF_IPV6_TCP] =
 416                                 RTE_ETH_RSS_NONFRAG_IPV6_TCP,
 417                 [IAVF_FILTER_PCTYPE_NONF_IPV6_SCTP] =
 418                                 RTE_ETH_RSS_NONFRAG_IPV6_SCTP,
 419                 [IAVF_FILTER_PCTYPE_NONF_IPV6_OTHER] =
 420                                 RTE_ETH_RSS_NONFRAG_IPV6_OTHER,
 421                 [IAVF_FILTER_PCTYPE_FRAG_IPV6] = RTE_ETH_RSS_FRAG_IPV6,
 422
 423                 /* L2 Payload */
 424                 [IAVF_FILTER_PCTYPE_L2_PAYLOAD] = RTE_ETH_RSS_L2_PAYLOAD
 425         };
 426
 427         const uint64_t ipv4_rss = RTE_ETH_RSS_NONFRAG_IPV4_UDP |
 428                                   RTE_ETH_RSS_NONFRAG_IPV4_TCP |
 429                                   RTE_ETH_RSS_NONFRAG_IPV4_SCTP |
 430                                   RTE_ETH_RSS_NONFRAG_IPV4_OTHER |
 431                                   RTE_ETH_RSS_FRAG_IPV4;
 432
 433         const uint64_t ipv6_rss = RTE_ETH_RSS_NONFRAG_IPV6_UDP |
 434                                   RTE_ETH_RSS_NONFRAG_IPV6_TCP |
 435                                   RTE_ETH_RSS_NONFRAG_IPV6_SCTP |
 436                                   RTE_ETH_RSS_NONFRAG_IPV6_OTHER |
 437                                   RTE_ETH_RSS_FRAG_IPV6;
 438
 439         struct iavf_info *vf =  IAVF_DEV_PRIVATE_TO_VF(adapter);
 440         uint64_t caps = 0, hena = 0, valid_rss_hf = 0;
 441         uint32_t i;
 442         int ret;
 443
 444         ret = iavf_get_hena_caps(adapter, &caps);
 445         if (ret) {
 446                 /**
 447                  * RSS offload type configuration is not a necessary feature
 448                  * for VF, so here just print a warning and return.
 449                  */
 450                 PMD_DRV_LOG(WARNING,
 451                             "fail to get RSS offload type caps, ret: %d", ret);
 452                 return;
 453         }
 454
 455         /**
 456          * RTE_ETH_RSS_IPV4 and RTE_ETH_RSS_IPV6 can be considered as 2
 457          * generalizations of all other IPv4 and IPv6 RSS types.
 458          */
 459         if (rss_hf & RTE_ETH_RSS_IPV4)
 460                 rss_hf |= ipv4_rss;
 461
 462         if (rss_hf & RTE_ETH_RSS_IPV6)
 463                 rss_hf |= ipv6_rss;
 464
 465         RTE_BUILD_BUG_ON(RTE_DIM(map_hena_rss) > sizeof(uint64_t) * CHAR_BIT);
 466
 467         for (i = 0; i < RTE_DIM(map_hena_rss); i++) {
 468                 uint64_t bit = BIT_ULL(i);
 469
 470                 if ((caps & bit) && (map_hena_rss[i] & rss_hf)) {
 471                         valid_rss_hf |= map_hena_rss[i];
 472                         hena |= bit;
 473                 }
 474         }
 475
 476         ret = iavf_set_hena(adapter, hena);
 477         if (ret) {
 478                 /**
 479                  * RSS offload type configuration is not a necessary feature
 480                  * for VF, so here just print a warning and return.
 481                  */
 482                 PMD_DRV_LOG(WARNING,
 483                             "fail to set RSS offload types, ret: %d", ret);
 484                 return;
 485         }
 486
 487         if (valid_rss_hf & ipv4_rss)
 488                 valid_rss_hf |= rss_hf & RTE_ETH_RSS_IPV4;
 489
 490         if (valid_rss_hf & ipv6_rss)
 491                 valid_rss_hf |= rss_hf & RTE_ETH_RSS_IPV6;
 492
 493         if (rss_hf & ~valid_rss_hf)
 494                 PMD_DRV_LOG(WARNING, "Unsupported rss_hf 0x%" PRIx64,
 495                             rss_hf & ~valid_rss_hf);
 496
 497         vf->rss_hf = valid_rss_hf;
 498 }
 499
 500 static int
 501 iavf_init_rss(struct iavf_adapter *adapter)
 502 {
 503         struct iavf_info *vf =  IAVF_DEV_PRIVATE_TO_VF(adapter);
 504         struct rte_eth_rss_conf *rss_conf;
 505         uint16_t i, j, nb_q;
 506         int ret;
 507
 508         rss_conf = &adapter->dev_data->dev_conf.rx_adv_conf.rss_conf;
 509         nb_q = RTE_MIN(adapter->dev_data->nb_rx_queues,
 510                        vf->max_rss_qregion);
 511
 512         if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF)) {
 513                 PMD_DRV_LOG(DEBUG, "RSS is not supported");
 514                 return -ENOTSUP;
 515         }
 516
 517         /* configure RSS key */
 518         if (!rss_conf->rss_key) {
 519                 /* Calculate the default hash key */
 520                 for (i = 0; i < vf->vf_res->rss_key_size; i++)
 521                         vf->rss_key[i] = (uint8_t)rte_rand();
 522         } else
 523                 rte_memcpy(vf->rss_key, rss_conf->rss_key,
 524                            RTE_MIN(rss_conf->rss_key_len,
 525                                    vf->vf_res->rss_key_size));
 526
 527         /* init RSS LUT table */
 528         for (i = 0, j = 0; i < vf->vf_res->rss_lut_size; i++, j++) {
 529                 if (j >= nb_q)
 530                         j = 0;
 531                 vf->rss_lut[i] = j;
 532         }
 533         /* send virtchnl ops to configure RSS */
 534         ret = iavf_configure_rss_lut(adapter);
 535         if (ret)
 536                 return ret;
 537         ret = iavf_configure_rss_key(adapter);
 538         if (ret)
 539                 return ret;
 540
 541         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_ADV_RSS_PF) {
 542                 /* Set RSS hash configuration based on rss_conf->rss_hf. */
 543                 ret = iavf_rss_hash_set(adapter, rss_conf->rss_hf, true);
 544                 if (ret) {
 545                         PMD_DRV_LOG(ERR, "fail to set default RSS");
 546                         return ret;
 547                 }
 548         } else {
 549                 iavf_config_rss_hf(adapter, rss_conf->rss_hf);
 550         }
 551
 552         return 0;
 553 }
 554
 555 static int
 556 iavf_queues_req_reset(struct rte_eth_dev *dev, uint16_t num)
 557 {
 558         struct iavf_adapter *ad =
 559                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 560         struct iavf_info *vf =  IAVF_DEV_PRIVATE_TO_VF(ad);
 561         int ret;
 562
 563         ret = iavf_request_queues(dev, num);
 564         if (ret) {
 565                 PMD_DRV_LOG(ERR, "request queues from PF failed");
 566                 return ret;
 567         }
 568         PMD_DRV_LOG(INFO, "change queue pairs from %u to %u",
 569                         vf->vsi_res->num_queue_pairs, num);
 570
 571         ret = iavf_dev_reset(dev);
 572         if (ret) {
 573                 PMD_DRV_LOG(ERR, "vf reset failed");
 574                 return ret;
 575         }
 576
 577         return 0;
 578 }
 579
 580 static int
 581 iavf_dev_vlan_insert_set(struct rte_eth_dev *dev)
 582 {
 583         struct iavf_adapter *adapter =
 584                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 585         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
 586         bool enable;
 587
 588         if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_VLAN_V2))
 589                 return 0;
 590
 591         enable = !!(dev->data->dev_conf.txmode.offloads &
 592                     RTE_ETH_TX_OFFLOAD_VLAN_INSERT);
 593         iavf_config_vlan_insert_v2(adapter, enable);
 594
 595         return 0;
 596 }
 597
 598 static int
 599 iavf_dev_init_vlan(struct rte_eth_dev *dev)
 600 {
 601         int err;
 602
 603         err = iavf_dev_vlan_offload_set(dev,
 604                                         RTE_ETH_VLAN_STRIP_MASK |
 605                                         RTE_ETH_QINQ_STRIP_MASK |
 606                                         RTE_ETH_VLAN_FILTER_MASK |
 607                                         RTE_ETH_VLAN_EXTEND_MASK);
 608         if (err) {
 609                 PMD_DRV_LOG(ERR, "Failed to update vlan offload");
 610                 return err;
 611         }
 612
 613         err = iavf_dev_vlan_insert_set(dev);
 614         if (err)
 615                 PMD_DRV_LOG(ERR, "Failed to update vlan insertion");
 616
 617         return err;
 618 }
 619
 620 static int
 621 iavf_dev_configure(struct rte_eth_dev *dev)
 622 {
 623         struct iavf_adapter *ad =
 624                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 625         struct iavf_info *vf =  IAVF_DEV_PRIVATE_TO_VF(ad);
 626         uint16_t num_queue_pairs = RTE_MAX(dev->data->nb_rx_queues,
 627                 dev->data->nb_tx_queues);
 628         int ret;
 629
 630         if (ad->closed)
 631                 return -EIO;
 632
 633         ad->rx_bulk_alloc_allowed = true;
 634         /* Initialize to TRUE. If any of Rx queues doesn't meet the
 635          * vector Rx/Tx preconditions, it will be reset.
 636          */
 637         ad->rx_vec_allowed = true;
 638         ad->tx_vec_allowed = true;
 639
 640         if (dev->data->dev_conf.rxmode.mq_mode & RTE_ETH_MQ_RX_RSS_FLAG)
 641                 dev->data->dev_conf.rxmode.offloads |= RTE_ETH_RX_OFFLOAD_RSS_HASH;
 642
 643         /* Large VF setting */
 644         if (num_queue_pairs > IAVF_MAX_NUM_QUEUES_DFLT) {
 645                 if (!(vf->vf_res->vf_cap_flags &
 646                                 VIRTCHNL_VF_LARGE_NUM_QPAIRS)) {
 647                         PMD_DRV_LOG(ERR, "large VF is not supported");
 648                         return -1;
 649                 }
 650
 651                 if (num_queue_pairs > IAVF_MAX_NUM_QUEUES_LV) {
 652                         PMD_DRV_LOG(ERR, "queue pairs number cannot be larger than %u",
 653                                 IAVF_MAX_NUM_QUEUES_LV);
 654                         return -1;
 655                 }
 656
 657                 ret = iavf_queues_req_reset(dev, num_queue_pairs);
 658                 if (ret)
 659                         return ret;
 660
 661                 ret = iavf_get_max_rss_queue_region(ad);
 662                 if (ret) {
 663                         PMD_INIT_LOG(ERR, "get max rss queue region failed");
 664                         return ret;
 665                 }
 666
 667                 vf->lv_enabled = true;
 668         } else {
 669                 /* Check if large VF is already enabled. If so, disable and
 670                  * release redundant queue resource.
 671                  * Or check if enough queue pairs. If not, request them from PF.
 672                  */
 673                 if (vf->lv_enabled ||
 674                     num_queue_pairs > vf->vsi_res->num_queue_pairs) {
 675                         ret = iavf_queues_req_reset(dev, num_queue_pairs);
 676                         if (ret)
 677                                 return ret;
 678
 679                         vf->lv_enabled = false;
 680                 }
 681                 /* if large VF is not required, use default rss queue region */
 682                 vf->max_rss_qregion = IAVF_MAX_NUM_QUEUES_DFLT;
 683         }
 684
 685         ret = iavf_dev_init_vlan(dev);
 686         if (ret)
 687                 PMD_DRV_LOG(ERR, "configure VLAN failed: %d", ret);
 688
 689         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF) {
 690                 if (iavf_init_rss(ad) != 0) {
 691                         PMD_DRV_LOG(ERR, "configure rss failed");
 692                         return -1;
 693                 }
 694         }
 695         return 0;
 696 }
 697
 698 static int
 699 iavf_init_rxq(struct rte_eth_dev *dev, struct iavf_rx_queue *rxq)
 700 {
 701         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(dev->data->dev_private);
 702         struct rte_eth_dev_data *dev_data = dev->data;
 703         uint16_t buf_size, max_pkt_len;
 704         uint32_t frame_size = dev->data->mtu + IAVF_ETH_OVERHEAD;
 705         enum iavf_status err;
 706
 707         buf_size = rte_pktmbuf_data_room_size(rxq->mp) - RTE_PKTMBUF_HEADROOM;
 708
 709         /* Calculate the maximum packet length allowed */
 710         max_pkt_len = RTE_MIN((uint32_t)
 711                         rxq->rx_buf_len * IAVF_MAX_CHAINED_RX_BUFFERS,
 712                         frame_size);
 713
 714         /* Check if maximum packet length is set correctly.  */
 715         if (max_pkt_len <= RTE_ETHER_MIN_LEN ||
 716             max_pkt_len > IAVF_FRAME_SIZE_MAX) {
 717                 PMD_DRV_LOG(ERR, "maximum packet length must be "
 718                             "larger than %u and smaller than %u",
 719                             (uint32_t)IAVF_ETH_MAX_LEN,
 720                             (uint32_t)IAVF_FRAME_SIZE_MAX);
 721                 return -EINVAL;
 722         }
 723
 724         if (rxq->offloads & RTE_ETH_RX_OFFLOAD_TIMESTAMP) {
 725                 /* Register mbuf field and flag for Rx timestamp */
 726                 err = rte_mbuf_dyn_rx_timestamp_register(
 727                         &iavf_timestamp_dynfield_offset,
 728                         &iavf_timestamp_dynflag);
 729                 if (err) {
 730                         PMD_DRV_LOG(ERR,
 731                                     "Cannot register mbuf field/flag for timestamp");
 732                         return -EINVAL;
 733                 }
 734         }
 735
 736         rxq->max_pkt_len = max_pkt_len;
 737         if ((dev_data->dev_conf.rxmode.offloads & RTE_ETH_RX_OFFLOAD_SCATTER) ||
 738             rxq->max_pkt_len > buf_size) {
 739                 dev_data->scattered_rx = 1;
 740         }
 741         IAVF_PCI_REG_WRITE(rxq->qrx_tail, rxq->nb_rx_desc - 1);
 742         IAVF_WRITE_FLUSH(hw);
 743
 744         return 0;
 745 }
 746
 747 static int
 748 iavf_init_queues(struct rte_eth_dev *dev)
 749 {
 750         struct iavf_rx_queue **rxq =
 751                 (struct iavf_rx_queue **)dev->data->rx_queues;
 752         int i, ret = IAVF_SUCCESS;
 753
 754         for (i = 0; i < dev->data->nb_rx_queues; i++) {
 755                 if (!rxq[i] || !rxq[i]->q_set)
 756                         continue;
 757                 ret = iavf_init_rxq(dev, rxq[i]);
 758                 if (ret != IAVF_SUCCESS)
 759                         break;
 760         }
 761         /* set rx/tx function to vector/scatter/single-segment
 762          * according to parameters
 763          */
 764         iavf_set_rx_function(dev);
 765         iavf_set_tx_function(dev);
 766
 767         return ret;
 768 }
 769
 770 static int iavf_config_rx_queues_irqs(struct rte_eth_dev *dev,
 771                                      struct rte_intr_handle *intr_handle)
 772 {
 773         struct iavf_adapter *adapter =
 774                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 775         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
 776         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(adapter);
 777         struct iavf_qv_map *qv_map;
 778         uint16_t interval, i;
 779         int vec;
 780
 781         if (rte_intr_cap_multiple(intr_handle) &&
 782             dev->data->dev_conf.intr_conf.rxq) {
 783                 if (rte_intr_efd_enable(intr_handle, dev->data->nb_rx_queues))
 784                         return -1;
 785         }
 786
 787         if (rte_intr_dp_is_en(intr_handle)) {
 788                 if (rte_intr_vec_list_alloc(intr_handle, "intr_vec",
 789                                                     dev->data->nb_rx_queues)) {
 790                         PMD_DRV_LOG(ERR, "Failed to allocate %d rx intr_vec",
 791                                     dev->data->nb_rx_queues);
 792                         return -1;
 793                 }
 794         }
 795
 796
 797         qv_map = rte_zmalloc("qv_map",
 798                 dev->data->nb_rx_queues * sizeof(struct iavf_qv_map), 0);
 799         if (!qv_map) {
 800                 PMD_DRV_LOG(ERR, "Failed to allocate %d queue-vector map",
 801                                 dev->data->nb_rx_queues);
 802                 goto qv_map_alloc_err;
 803         }
 804
 805         if (!dev->data->dev_conf.intr_conf.rxq ||
 806             !rte_intr_dp_is_en(intr_handle)) {
 807                 /* Rx interrupt disabled, Map interrupt only for writeback */
 808                 vf->nb_msix = 1;
 809                 if (vf->vf_res->vf_cap_flags &
 810                     VIRTCHNL_VF_OFFLOAD_WB_ON_ITR) {
 811                         /* If WB_ON_ITR supports, enable it */
 812                         vf->msix_base = IAVF_RX_VEC_START;
 813                         /* Set the ITR for index zero, to 2us to make sure that
 814                          * we leave time for aggregation to occur, but don't
 815                          * increase latency dramatically.
 816                          */
 817                         IAVF_WRITE_REG(hw,
 818                                        IAVF_VFINT_DYN_CTLN1(vf->msix_base - 1),
 819                                        (0 << IAVF_VFINT_DYN_CTLN1_ITR_INDX_SHIFT) |
 820                                        IAVF_VFINT_DYN_CTLN1_WB_ON_ITR_MASK |
 821                                        (2UL << IAVF_VFINT_DYN_CTLN1_INTERVAL_SHIFT));
 822                         /* debug - check for success! the return value
 823                          * should be 2, offset is 0x2800
 824                          */
 825                         /* IAVF_READ_REG(hw, IAVF_VFINT_ITRN1(0, 0)); */
 826                 } else {
 827                         /* If no WB_ON_ITR offload flags, need to set
 828                          * interrupt for descriptor write back.
 829                          */
 830                         vf->msix_base = IAVF_MISC_VEC_ID;
 831
 832                         /* set ITR to default */
 833                         interval = iavf_calc_itr_interval(
 834                                         IAVF_QUEUE_ITR_INTERVAL_DEFAULT);
 835                         IAVF_WRITE_REG(hw, IAVF_VFINT_DYN_CTL01,
 836                                        IAVF_VFINT_DYN_CTL01_INTENA_MASK |
 837                                        (IAVF_ITR_INDEX_DEFAULT <<
 838                                         IAVF_VFINT_DYN_CTL01_ITR_INDX_SHIFT) |
 839                                        (interval <<
 840                                         IAVF_VFINT_DYN_CTL01_INTERVAL_SHIFT));
 841                 }
 842                 IAVF_WRITE_FLUSH(hw);
 843                 /* map all queues to the same interrupt */
 844                 for (i = 0; i < dev->data->nb_rx_queues; i++) {
 845                         qv_map[i].queue_id = i;
 846                         qv_map[i].vector_id = vf->msix_base;
 847                 }
 848                 vf->qv_map = qv_map;
 849         } else {
 850                 if (!rte_intr_allow_others(intr_handle)) {
 851                         vf->nb_msix = 1;
 852                         vf->msix_base = IAVF_MISC_VEC_ID;
 853                         for (i = 0; i < dev->data->nb_rx_queues; i++) {
 854                                 qv_map[i].queue_id = i;
 855                                 qv_map[i].vector_id = vf->msix_base;
 856                                 rte_intr_vec_list_index_set(intr_handle,
 857                                                         i, IAVF_MISC_VEC_ID);
 858                         }
 859                         vf->qv_map = qv_map;
 860                         PMD_DRV_LOG(DEBUG,
 861                                     "vector %u are mapping to all Rx queues",
 862                                     vf->msix_base);
 863                 } else {
 864                         /* If Rx interrupt is required, and we can use
 865                          * multi interrupts, then the vec is from 1
 866                          */
 867                         vf->nb_msix =
 868                                 RTE_MIN(rte_intr_nb_efd_get(intr_handle),
 869                                 (uint16_t)(vf->vf_res->max_vectors - 1));
 870                         vf->msix_base = IAVF_RX_VEC_START;
 871                         vec = IAVF_RX_VEC_START;
 872                         for (i = 0; i < dev->data->nb_rx_queues; i++) {
 873                                 qv_map[i].queue_id = i;
 874                                 qv_map[i].vector_id = vec;
 875                                 rte_intr_vec_list_index_set(intr_handle,
 876                                                                    i, vec++);
 877                                 if (vec >= vf->nb_msix + IAVF_RX_VEC_START)
 878                                         vec = IAVF_RX_VEC_START;
 879                         }
 880                         vf->qv_map = qv_map;
 881                         PMD_DRV_LOG(DEBUG,
 882                                     "%u vectors are mapping to %u Rx queues",
 883                                     vf->nb_msix, dev->data->nb_rx_queues);
 884                 }
 885         }
 886
 887         if (!vf->lv_enabled) {
 888                 if (iavf_config_irq_map(adapter)) {
 889                         PMD_DRV_LOG(ERR, "config interrupt mapping failed");
 890                         goto config_irq_map_err;
 891                 }
 892         } else {
 893                 uint16_t num_qv_maps = dev->data->nb_rx_queues;
 894                 uint16_t index = 0;
 895
 896                 while (num_qv_maps > IAVF_IRQ_MAP_NUM_PER_BUF) {
 897                         if (iavf_config_irq_map_lv(adapter,
 898                                         IAVF_IRQ_MAP_NUM_PER_BUF, index)) {
 899                                 PMD_DRV_LOG(ERR, "config interrupt mapping for large VF failed");
 900                                 goto config_irq_map_err;
 901                         }
 902                         num_qv_maps -= IAVF_IRQ_MAP_NUM_PER_BUF;
 903                         index += IAVF_IRQ_MAP_NUM_PER_BUF;
 904                 }
 905
 906                 if (iavf_config_irq_map_lv(adapter, num_qv_maps, index)) {
 907                         PMD_DRV_LOG(ERR, "config interrupt mapping for large VF failed");
 908                         goto config_irq_map_err;
 909                 }
 910         }
 911         return 0;
 912
 913 config_irq_map_err:
 914         rte_free(vf->qv_map);
 915         vf->qv_map = NULL;
 916
 917 qv_map_alloc_err:
 918         rte_intr_vec_list_free(intr_handle);
 919
 920         return -1;
 921 }
 922
 923 static int
 924 iavf_start_queues(struct rte_eth_dev *dev)
 925 {
 926         struct iavf_rx_queue *rxq;
 927         struct iavf_tx_queue *txq;
 928         int i;
 929         uint16_t nb_txq, nb_rxq;
 930
 931         for (nb_txq = 0; nb_txq < dev->data->nb_tx_queues; nb_txq++) {
 932                 txq = dev->data->tx_queues[nb_txq];
 933                 if (txq->tx_deferred_start)
 934                         continue;
 935                 if (iavf_dev_tx_queue_start(dev, nb_txq) != 0) {
 936                         PMD_DRV_LOG(ERR, "Fail to start tx queue %u", nb_txq);
 937                         goto tx_err;
 938                 }
 939         }
 940
 941         for (nb_rxq = 0; nb_rxq < dev->data->nb_rx_queues; nb_rxq++) {
 942                 rxq = dev->data->rx_queues[nb_rxq];
 943                 if (rxq->rx_deferred_start)
 944                         continue;
 945                 if (iavf_dev_rx_queue_start(dev, nb_rxq) != 0) {
 946                         PMD_DRV_LOG(ERR, "Fail to start rx queue %u", nb_rxq);
 947                         goto rx_err;
 948                 }
 949         }
 950
 951         return 0;
 952
 953 rx_err:
 954         for (i = 0; i < nb_rxq; i++)
 955                 iavf_dev_rx_queue_stop(dev, i);
 956 tx_err:
 957         for (i = 0; i < nb_txq; i++)
 958                 iavf_dev_tx_queue_stop(dev, i);
 959
 960         return -1;
 961 }
 962
 963 static int
 964 iavf_dev_start(struct rte_eth_dev *dev)
 965 {
 966         struct iavf_adapter *adapter =
 967                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 968         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
 969         struct rte_intr_handle *intr_handle = dev->intr_handle;
 970         uint16_t num_queue_pairs;
 971         uint16_t index = 0;
 972
 973         PMD_INIT_FUNC_TRACE();
 974
 975         if (adapter->closed)
 976                 return -1;
 977
 978         adapter->stopped = 0;
 979
 980         vf->max_pkt_len = dev->data->mtu + IAVF_ETH_OVERHEAD;
 981         vf->num_queue_pairs = RTE_MAX(dev->data->nb_rx_queues,
 982                                       dev->data->nb_tx_queues);
 983         num_queue_pairs = vf->num_queue_pairs;
 984
 985         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_QOS)
 986                 if (iavf_get_qos_cap(adapter)) {
 987                         PMD_INIT_LOG(ERR, "Failed to get qos capability");
 988                         return -1;
 989                 }
 990
 991         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_CAP_PTP) {
 992                 if (iavf_get_ptp_cap(adapter)) {
 993                         PMD_INIT_LOG(ERR, "Failed to get ptp capability");
 994                         return -1;
 995                 }
 996         }
 997
 998         if (iavf_init_queues(dev) != 0) {
 999                 PMD_DRV_LOG(ERR, "failed to do Queue init");
1000                 return -1;
1001         }
1002
1003         if (iavf_set_vf_quanta_size(adapter, index, num_queue_pairs) != 0)
1004                 PMD_DRV_LOG(WARNING, "configure quanta size failed");
1005
1006         /* If needed, send configure queues msg multiple times to make the
1007          * adminq buffer length smaller than the 4K limitation.
1008          */
1009         while (num_queue_pairs > IAVF_CFG_Q_NUM_PER_BUF) {
1010                 if (iavf_configure_queues(adapter,
1011                                 IAVF_CFG_Q_NUM_PER_BUF, index) != 0) {
1012                         PMD_DRV_LOG(ERR, "configure queues failed");
1013                         goto err_queue;
1014                 }
1015                 num_queue_pairs -= IAVF_CFG_Q_NUM_PER_BUF;
1016                 index += IAVF_CFG_Q_NUM_PER_BUF;
1017         }
1018
1019         if (iavf_configure_queues(adapter, num_queue_pairs, index) != 0) {
1020                 PMD_DRV_LOG(ERR, "configure queues failed");
1021                 goto err_queue;
1022         }
1023
1024         if (iavf_config_rx_queues_irqs(dev, intr_handle) != 0) {
1025                 PMD_DRV_LOG(ERR, "configure irq failed");
1026                 goto err_queue;
1027         }
1028         /* re-enable intr again, because efd assign may change */
1029         if (dev->data->dev_conf.intr_conf.rxq != 0) {
1030                 if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_WB_ON_ITR)
1031                         rte_intr_disable(intr_handle);
1032                 rte_intr_enable(intr_handle);
1033         }
1034
1035         /* Set all mac addrs */
1036         iavf_add_del_all_mac_addr(adapter, true);
1037
1038         /* Set all multicast addresses */
1039         iavf_add_del_mc_addr_list(adapter, vf->mc_addrs, vf->mc_addrs_num,
1040                                   true);
1041
1042         rte_spinlock_init(&vf->phc_time_aq_lock);
1043
1044         if (iavf_start_queues(dev) != 0) {
1045                 PMD_DRV_LOG(ERR, "enable queues failed");
1046                 goto err_mac;
1047         }
1048
1049         return 0;
1050
1051 err_mac:
1052         iavf_add_del_all_mac_addr(adapter, false);
1053 err_queue:
1054         return -1;
1055 }
1056
1057 static int
1058 iavf_dev_stop(struct rte_eth_dev *dev)
1059 {
1060         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
1061         struct iavf_adapter *adapter =
1062                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1063         struct rte_intr_handle *intr_handle = dev->intr_handle;
1064
1065         PMD_INIT_FUNC_TRACE();
1066
1067         if (adapter->closed)
1068                 return -1;
1069
1070         if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_WB_ON_ITR) &&
1071             dev->data->dev_conf.intr_conf.rxq != 0)
1072                 rte_intr_disable(intr_handle);
1073
1074         if (adapter->stopped == 1)
1075                 return 0;
1076
1077         iavf_stop_queues(dev);
1078
1079         /* Disable the interrupt for Rx */
1080         rte_intr_efd_disable(intr_handle);
1081         /* Rx interrupt vector mapping free */
1082         rte_intr_vec_list_free(intr_handle);
1083
1084         /* remove all mac addrs */
1085         iavf_add_del_all_mac_addr(adapter, false);
1086
1087         /* remove all multicast addresses */
1088         iavf_add_del_mc_addr_list(adapter, vf->mc_addrs, vf->mc_addrs_num,
1089                                   false);
1090
1091         adapter->stopped = 1;
1092         dev->data->dev_started = 0;
1093
1094         return 0;
1095 }
1096
1097 static int
1098 iavf_dev_info_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
1099 {
1100         struct iavf_adapter *adapter =
1101                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1102         struct iavf_info *vf = &adapter->vf;
1103
1104         if (adapter->closed)
1105                 return -EIO;
1106
1107         dev_info->max_rx_queues = IAVF_MAX_NUM_QUEUES_LV;
1108         dev_info->max_tx_queues = IAVF_MAX_NUM_QUEUES_LV;
1109         dev_info->min_rx_bufsize = IAVF_BUF_SIZE_MIN;
1110         dev_info->max_rx_pktlen = IAVF_FRAME_SIZE_MAX;
1111         dev_info->max_mtu = dev_info->max_rx_pktlen - IAVF_ETH_OVERHEAD;
1112         dev_info->min_mtu = RTE_ETHER_MIN_MTU;
1113         dev_info->hash_key_size = vf->vf_res->rss_key_size;
1114         dev_info->reta_size = vf->vf_res->rss_lut_size;
1115         dev_info->flow_type_rss_offloads = IAVF_RSS_OFFLOAD_ALL;
1116         dev_info->max_mac_addrs = IAVF_NUM_MACADDR_MAX;
1117         dev_info->dev_capa &= ~RTE_ETH_DEV_CAPA_FLOW_RULE_KEEP;
1118         dev_info->rx_offload_capa =
1119                 RTE_ETH_RX_OFFLOAD_VLAN_STRIP |
1120                 RTE_ETH_RX_OFFLOAD_QINQ_STRIP |
1121                 RTE_ETH_RX_OFFLOAD_IPV4_CKSUM |
1122                 RTE_ETH_RX_OFFLOAD_UDP_CKSUM |
1123                 RTE_ETH_RX_OFFLOAD_TCP_CKSUM |
1124                 RTE_ETH_RX_OFFLOAD_OUTER_IPV4_CKSUM |
1125                 RTE_ETH_RX_OFFLOAD_SCATTER |
1126                 RTE_ETH_RX_OFFLOAD_VLAN_FILTER |
1127                 RTE_ETH_RX_OFFLOAD_RSS_HASH;
1128
1129         dev_info->tx_offload_capa =
1130                 RTE_ETH_TX_OFFLOAD_VLAN_INSERT |
1131                 RTE_ETH_TX_OFFLOAD_QINQ_INSERT |
1132                 RTE_ETH_TX_OFFLOAD_IPV4_CKSUM |
1133                 RTE_ETH_TX_OFFLOAD_UDP_CKSUM |
1134                 RTE_ETH_TX_OFFLOAD_TCP_CKSUM |
1135                 RTE_ETH_TX_OFFLOAD_SCTP_CKSUM |
1136                 RTE_ETH_TX_OFFLOAD_OUTER_IPV4_CKSUM |
1137                 RTE_ETH_TX_OFFLOAD_TCP_TSO |
1138                 RTE_ETH_TX_OFFLOAD_VXLAN_TNL_TSO |
1139                 RTE_ETH_TX_OFFLOAD_GRE_TNL_TSO |
1140                 RTE_ETH_TX_OFFLOAD_IPIP_TNL_TSO |
1141                 RTE_ETH_TX_OFFLOAD_GENEVE_TNL_TSO |
1142                 RTE_ETH_TX_OFFLOAD_MULTI_SEGS |
1143                 RTE_ETH_TX_OFFLOAD_MBUF_FAST_FREE;
1144
1145         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_CRC)
1146                 dev_info->rx_offload_capa |= RTE_ETH_RX_OFFLOAD_KEEP_CRC;
1147
1148         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_CAP_PTP)
1149                 dev_info->rx_offload_capa |= RTE_ETH_RX_OFFLOAD_TIMESTAMP;
1150
1151         if (iavf_ipsec_crypto_supported(adapter)) {
1152                 dev_info->rx_offload_capa |= RTE_ETH_RX_OFFLOAD_SECURITY;
1153                 dev_info->tx_offload_capa |= RTE_ETH_TX_OFFLOAD_SECURITY;
1154         }
1155
1156         dev_info->default_rxconf = (struct rte_eth_rxconf) {
1157                 .rx_free_thresh = IAVF_DEFAULT_RX_FREE_THRESH,
1158                 .rx_drop_en = 0,
1159                 .offloads = 0,
1160         };
1161
1162         dev_info->default_txconf = (struct rte_eth_txconf) {
1163                 .tx_free_thresh = IAVF_DEFAULT_TX_FREE_THRESH,
1164                 .tx_rs_thresh = IAVF_DEFAULT_TX_RS_THRESH,
1165                 .offloads = 0,
1166         };
1167
1168         dev_info->rx_desc_lim = (struct rte_eth_desc_lim) {
1169                 .nb_max = IAVF_MAX_RING_DESC,
1170                 .nb_min = IAVF_MIN_RING_DESC,
1171                 .nb_align = IAVF_ALIGN_RING_DESC,
1172         };
1173
1174         dev_info->tx_desc_lim = (struct rte_eth_desc_lim) {
1175                 .nb_max = IAVF_MAX_RING_DESC,
1176                 .nb_min = IAVF_MIN_RING_DESC,
1177                 .nb_align = IAVF_ALIGN_RING_DESC,
1178         };
1179
1180         return 0;
1181 }
1182
1183 static const uint32_t *
1184 iavf_dev_supported_ptypes_get(struct rte_eth_dev *dev __rte_unused)
1185 {
1186         static const uint32_t ptypes[] = {
1187                 RTE_PTYPE_L2_ETHER,
1188                 RTE_PTYPE_L3_IPV4_EXT_UNKNOWN,
1189                 RTE_PTYPE_L4_FRAG,
1190                 RTE_PTYPE_L4_ICMP,
1191                 RTE_PTYPE_L4_NONFRAG,
1192                 RTE_PTYPE_L4_SCTP,
1193                 RTE_PTYPE_L4_TCP,
1194                 RTE_PTYPE_L4_UDP,
1195                 RTE_PTYPE_UNKNOWN
1196         };
1197         return ptypes;
1198 }
1199
1200 int
1201 iavf_dev_link_update(struct rte_eth_dev *dev,
1202                     __rte_unused int wait_to_complete)
1203 {
1204         struct rte_eth_link new_link;
1205         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
1206
1207         memset(&new_link, 0, sizeof(new_link));
1208
1209         /* Only read status info stored in VF, and the info is updated
1210          *  when receive LINK_CHANGE evnet from PF by Virtchnnl.
1211          */
1212         switch (vf->link_speed) {
1213         case 10:
1214                 new_link.link_speed = RTE_ETH_SPEED_NUM_10M;
1215                 break;
1216         case 100:
1217                 new_link.link_speed = RTE_ETH_SPEED_NUM_100M;
1218                 break;
1219         case 1000:
1220                 new_link.link_speed = RTE_ETH_SPEED_NUM_1G;
1221                 break;
1222         case 10000:
1223                 new_link.link_speed = RTE_ETH_SPEED_NUM_10G;
1224                 break;
1225         case 20000:
1226                 new_link.link_speed = RTE_ETH_SPEED_NUM_20G;
1227                 break;
1228         case 25000:
1229                 new_link.link_speed = RTE_ETH_SPEED_NUM_25G;
1230                 break;
1231         case 40000:
1232                 new_link.link_speed = RTE_ETH_SPEED_NUM_40G;
1233                 break;
1234         case 50000:
1235                 new_link.link_speed = RTE_ETH_SPEED_NUM_50G;
1236                 break;
1237         case 100000:
1238                 new_link.link_speed = RTE_ETH_SPEED_NUM_100G;
1239                 break;
1240         default:
1241                 new_link.link_speed = RTE_ETH_SPEED_NUM_NONE;
1242                 break;
1243         }
1244
1245         new_link.link_duplex = RTE_ETH_LINK_FULL_DUPLEX;
1246         new_link.link_status = vf->link_up ? RTE_ETH_LINK_UP :
1247                                              RTE_ETH_LINK_DOWN;
1248         new_link.link_autoneg = !(dev->data->dev_conf.link_speeds &
1249                                 RTE_ETH_LINK_SPEED_FIXED);
1250
1251         return rte_eth_linkstatus_set(dev, &new_link);
1252 }
1253
1254 static int
1255 iavf_dev_promiscuous_enable(struct rte_eth_dev *dev)
1256 {
1257         struct iavf_adapter *adapter =
1258                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1259         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
1260
1261         return iavf_config_promisc(adapter,
1262                                   true, vf->promisc_multicast_enabled);
1263 }
1264
1265 static int
1266 iavf_dev_promiscuous_disable(struct rte_eth_dev *dev)
1267 {
1268         struct iavf_adapter *adapter =
1269                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1270         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
1271
1272         return iavf_config_promisc(adapter,
1273                                   false, vf->promisc_multicast_enabled);
1274 }
1275
1276 static int
1277 iavf_dev_allmulticast_enable(struct rte_eth_dev *dev)
1278 {
1279         struct iavf_adapter *adapter =
1280                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1281         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
1282
1283         return iavf_config_promisc(adapter,
1284                                   vf->promisc_unicast_enabled, true);
1285 }
1286
1287 static int
1288 iavf_dev_allmulticast_disable(struct rte_eth_dev *dev)
1289 {
1290         struct iavf_adapter *adapter =
1291                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1292         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
1293
1294         return iavf_config_promisc(adapter,
1295                                   vf->promisc_unicast_enabled, false);
1296 }
1297
1298 static int
1299 iavf_dev_add_mac_addr(struct rte_eth_dev *dev, struct rte_ether_addr *addr,
1300                      __rte_unused uint32_t index,
1301                      __rte_unused uint32_t pool)
1302 {
1303         struct iavf_adapter *adapter =
1304                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1305         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
1306         int err;
1307
1308         if (rte_is_zero_ether_addr(addr)) {
1309                 PMD_DRV_LOG(ERR, "Invalid Ethernet Address");
1310                 return -EINVAL;
1311         }
1312
1313         err = iavf_add_del_eth_addr(adapter, addr, true, VIRTCHNL_ETHER_ADDR_EXTRA);
1314         if (err) {
1315                 PMD_DRV_LOG(ERR, "fail to add MAC address");
1316                 return -EIO;
1317         }
1318
1319         vf->mac_num++;
1320
1321         return 0;
1322 }
1323
1324 static void
1325 iavf_dev_del_mac_addr(struct rte_eth_dev *dev, uint32_t index)
1326 {
1327         struct iavf_adapter *adapter =
1328                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1329         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
1330         struct rte_ether_addr *addr;
1331         int err;
1332
1333         addr = &dev->data->mac_addrs[index];
1334
1335         err = iavf_add_del_eth_addr(adapter, addr, false, VIRTCHNL_ETHER_ADDR_EXTRA);
1336         if (err)
1337                 PMD_DRV_LOG(ERR, "fail to delete MAC address");
1338
1339         vf->mac_num--;
1340 }
1341
1342 static int
1343 iavf_dev_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
1344 {
1345         struct iavf_adapter *adapter =
1346                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1347         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
1348         int err;
1349
1350         if (adapter->closed)
1351                 return -EIO;
1352
1353         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_VLAN_V2) {
1354                 err = iavf_add_del_vlan_v2(adapter, vlan_id, on);
1355                 if (err)
1356                         return -EIO;
1357                 return 0;
1358         }
1359
1360         if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_VLAN))
1361                 return -ENOTSUP;
1362
1363         err = iavf_add_del_vlan(adapter, vlan_id, on);
1364         if (err)
1365                 return -EIO;
1366         return 0;
1367 }
1368
1369 static void
1370 iavf_iterate_vlan_filters_v2(struct rte_eth_dev *dev, bool enable)
1371 {
1372         struct rte_vlan_filter_conf *vfc = &dev->data->vlan_filter_conf;
1373         struct iavf_adapter *adapter =
1374                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1375         uint32_t i, j;
1376         uint64_t ids;
1377
1378         for (i = 0; i < RTE_DIM(vfc->ids); i++) {
1379                 if (vfc->ids[i] == 0)
1380                         continue;
1381
1382                 ids = vfc->ids[i];
1383                 for (j = 0; ids != 0 && j < 64; j++, ids >>= 1) {
1384                         if (ids & 1)
1385                                 iavf_add_del_vlan_v2(adapter,
1386                                                      64 * i + j, enable);
1387                 }
1388         }
1389 }
1390
1391 static int
1392 iavf_dev_vlan_offload_set_v2(struct rte_eth_dev *dev, int mask)
1393 {
1394         struct rte_eth_rxmode *rxmode = &dev->data->dev_conf.rxmode;
1395         struct iavf_adapter *adapter =
1396                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1397         bool enable;
1398         int err;
1399
1400         if (mask & RTE_ETH_VLAN_FILTER_MASK) {
1401                 enable = !!(rxmode->offloads & RTE_ETH_RX_OFFLOAD_VLAN_FILTER);
1402
1403                 iavf_iterate_vlan_filters_v2(dev, enable);
1404         }
1405
1406         if (mask & RTE_ETH_VLAN_STRIP_MASK) {
1407                 enable = !!(rxmode->offloads & RTE_ETH_RX_OFFLOAD_VLAN_STRIP);
1408
1409                 err = iavf_config_vlan_strip_v2(adapter, enable);
1410                 /* If not support, the stripping is already disabled by PF */
1411                 if (err == -ENOTSUP && !enable)
1412                         err = 0;
1413                 if (err)
1414                         return -EIO;
1415         }
1416
1417         return 0;
1418 }
1419
1420 static int
1421 iavf_dev_vlan_offload_set(struct rte_eth_dev *dev, int mask)
1422 {
1423         struct iavf_adapter *adapter =
1424                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1425         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
1426         struct rte_eth_conf *dev_conf = &dev->data->dev_conf;
1427         int err;
1428
1429         if (adapter->closed)
1430                 return -EIO;
1431
1432         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_VLAN_V2)
1433                 return iavf_dev_vlan_offload_set_v2(dev, mask);
1434
1435         if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_VLAN))
1436                 return -ENOTSUP;
1437
1438         /* Vlan stripping setting */
1439         if (mask & RTE_ETH_VLAN_STRIP_MASK) {
1440                 /* Enable or disable VLAN stripping */
1441                 if (dev_conf->rxmode.offloads & RTE_ETH_RX_OFFLOAD_VLAN_STRIP)
1442                         err = iavf_enable_vlan_strip(adapter);
1443                 else
1444                         err = iavf_disable_vlan_strip(adapter);
1445
1446                 if (err)
1447                         return -EIO;
1448         }
1449         return 0;
1450 }
1451
1452 static int
1453 iavf_dev_rss_reta_update(struct rte_eth_dev *dev,
1454                         struct rte_eth_rss_reta_entry64 *reta_conf,
1455                         uint16_t reta_size)
1456 {
1457         struct iavf_adapter *adapter =
1458                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1459         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
1460         uint8_t *lut;
1461         uint16_t i, idx, shift;
1462         int ret;
1463
1464         if (adapter->closed)
1465                 return -EIO;
1466
1467         if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF))
1468                 return -ENOTSUP;
1469
1470         if (reta_size != vf->vf_res->rss_lut_size) {
1471                 PMD_DRV_LOG(ERR, "The size of hash lookup table configured "
1472                         "(%d) doesn't match the number of hardware can "
1473                         "support (%d)", reta_size, vf->vf_res->rss_lut_size);
1474                 return -EINVAL;
1475         }
1476
1477         lut = rte_zmalloc("rss_lut", reta_size, 0);
1478         if (!lut) {
1479                 PMD_DRV_LOG(ERR, "No memory can be allocated");
1480                 return -ENOMEM;
1481         }
1482         /* store the old lut table temporarily */
1483         rte_memcpy(lut, vf->rss_lut, reta_size);
1484
1485         for (i = 0; i < reta_size; i++) {
1486                 idx = i / RTE_ETH_RETA_GROUP_SIZE;
1487                 shift = i % RTE_ETH_RETA_GROUP_SIZE;
1488                 if (reta_conf[idx].mask & (1ULL << shift))
1489                         lut[i] = reta_conf[idx].reta[shift];
1490         }
1491
1492         rte_memcpy(vf->rss_lut, lut, reta_size);
1493         /* send virtchnl ops to configure RSS */
1494         ret = iavf_configure_rss_lut(adapter);
1495         if (ret) /* revert back */
1496                 rte_memcpy(vf->rss_lut, lut, reta_size);
1497         rte_free(lut);
1498
1499         return ret;
1500 }
1501
1502 static int
1503 iavf_dev_rss_reta_query(struct rte_eth_dev *dev,
1504                        struct rte_eth_rss_reta_entry64 *reta_conf,
1505                        uint16_t reta_size)
1506 {
1507         struct iavf_adapter *adapter =
1508                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1509         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
1510         uint16_t i, idx, shift;
1511
1512         if (adapter->closed)
1513                 return -EIO;
1514
1515         if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF))
1516                 return -ENOTSUP;
1517
1518         if (reta_size != vf->vf_res->rss_lut_size) {
1519                 PMD_DRV_LOG(ERR, "The size of hash lookup table configured "
1520                         "(%d) doesn't match the number of hardware can "
1521                         "support (%d)", reta_size, vf->vf_res->rss_lut_size);
1522                 return -EINVAL;
1523         }
1524
1525         for (i = 0; i < reta_size; i++) {
1526                 idx = i / RTE_ETH_RETA_GROUP_SIZE;
1527                 shift = i % RTE_ETH_RETA_GROUP_SIZE;
1528                 if (reta_conf[idx].mask & (1ULL << shift))
1529                         reta_conf[idx].reta[shift] = vf->rss_lut[i];
1530         }
1531
1532         return 0;
1533 }
1534
1535 static int
1536 iavf_set_rss_key(struct iavf_adapter *adapter, uint8_t *key, uint8_t key_len)
1537 {
1538         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
1539
1540         /* HENA setting, it is enabled by default, no change */
1541         if (!key || key_len == 0) {
1542                 PMD_DRV_LOG(DEBUG, "No key to be configured");
1543                 return 0;
1544         } else if (key_len != vf->vf_res->rss_key_size) {
1545                 PMD_DRV_LOG(ERR, "The size of hash key configured "
1546                         "(%d) doesn't match the size of hardware can "
1547                         "support (%d)", key_len,
1548                         vf->vf_res->rss_key_size);
1549                 return -EINVAL;
1550         }
1551
1552         rte_memcpy(vf->rss_key, key, key_len);
1553
1554         return iavf_configure_rss_key(adapter);
1555 }
1556
1557 static int
1558 iavf_dev_rss_hash_update(struct rte_eth_dev *dev,
1559                         struct rte_eth_rss_conf *rss_conf)
1560 {
1561         struct iavf_adapter *adapter =
1562                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1563         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
1564         int ret;
1565
1566         adapter->dev_data->dev_conf.rx_adv_conf.rss_conf = *rss_conf;
1567
1568         if (adapter->closed)
1569                 return -EIO;
1570
1571         if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF))
1572                 return -ENOTSUP;
1573
1574         /* Set hash key. */
1575         ret = iavf_set_rss_key(adapter, rss_conf->rss_key,
1576                                rss_conf->rss_key_len);
1577         if (ret)
1578                 return ret;
1579
1580         if (rss_conf->rss_hf == 0) {
1581                 vf->rss_hf = 0;
1582                 ret = iavf_set_hena(adapter, 0);
1583
1584                 /* It is a workaround, temporarily allow error to be returned
1585                  * due to possible lack of PF handling for hena = 0.
1586                  */
1587                 if (ret)
1588                         PMD_DRV_LOG(WARNING, "fail to clean existing RSS, lack PF support");
1589                 return 0;
1590         }
1591
1592         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_ADV_RSS_PF) {
1593                 /* Clear existing RSS. */
1594                 ret = iavf_set_hena(adapter, 0);
1595
1596                 /* It is a workaround, temporarily allow error to be returned
1597                  * due to possible lack of PF handling for hena = 0.
1598                  */
1599                 if (ret)
1600                         PMD_DRV_LOG(WARNING, "fail to clean existing RSS,"
1601                                     "lack PF support");
1602
1603                 /* Set new RSS configuration. */
1604                 ret = iavf_rss_hash_set(adapter, rss_conf->rss_hf, true);
1605                 if (ret) {
1606                         PMD_DRV_LOG(ERR, "fail to set new RSS");
1607                         return ret;
1608                 }
1609         } else {
1610                 iavf_config_rss_hf(adapter, rss_conf->rss_hf);
1611         }
1612
1613         return 0;
1614 }
1615
1616 static int
1617 iavf_dev_rss_hash_conf_get(struct rte_eth_dev *dev,
1618                           struct rte_eth_rss_conf *rss_conf)
1619 {
1620         struct iavf_adapter *adapter =
1621                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1622         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
1623
1624         if (adapter->closed)
1625                 return -EIO;
1626
1627         if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF))
1628                 return -ENOTSUP;
1629
1630         rss_conf->rss_hf = vf->rss_hf;
1631
1632         if (!rss_conf->rss_key)
1633                 return 0;
1634
1635         rss_conf->rss_key_len = vf->vf_res->rss_key_size;
1636         rte_memcpy(rss_conf->rss_key, vf->rss_key, rss_conf->rss_key_len);
1637
1638         return 0;
1639 }
1640
1641 static int
1642 iavf_dev_mtu_set(struct rte_eth_dev *dev, uint16_t mtu __rte_unused)
1643 {
1644         /* mtu setting is forbidden if port is start */
1645         if (dev->data->dev_started) {
1646                 PMD_DRV_LOG(ERR, "port must be stopped before configuration");
1647                 return -EBUSY;
1648         }
1649
1650         return 0;
1651 }
1652
1653 static int
1654 iavf_dev_set_default_mac_addr(struct rte_eth_dev *dev,
1655                              struct rte_ether_addr *mac_addr)
1656 {
1657         struct iavf_adapter *adapter =
1658                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1659         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(adapter);
1660         struct rte_ether_addr *old_addr;
1661         int ret;
1662
1663         old_addr = (struct rte_ether_addr *)hw->mac.addr;
1664
1665         if (rte_is_same_ether_addr(old_addr, mac_addr))
1666                 return 0;
1667
1668         ret = iavf_add_del_eth_addr(adapter, old_addr, false, VIRTCHNL_ETHER_ADDR_PRIMARY);
1669         if (ret)
1670                 PMD_DRV_LOG(ERR, "Fail to delete old MAC:"
1671                             RTE_ETHER_ADDR_PRT_FMT,
1672                                 RTE_ETHER_ADDR_BYTES(old_addr));
1673
1674         ret = iavf_add_del_eth_addr(adapter, mac_addr, true, VIRTCHNL_ETHER_ADDR_PRIMARY);
1675         if (ret)
1676                 PMD_DRV_LOG(ERR, "Fail to add new MAC:"
1677                             RTE_ETHER_ADDR_PRT_FMT,
1678                                 RTE_ETHER_ADDR_BYTES(mac_addr));
1679
1680         if (ret)
1681                 return -EIO;
1682
1683         rte_ether_addr_copy(mac_addr, (struct rte_ether_addr *)hw->mac.addr);
1684         return 0;
1685 }
1686
1687 static void
1688 iavf_stat_update_48(uint64_t *offset, uint64_t *stat)
1689 {
1690         if (*stat >= *offset)
1691                 *stat = *stat - *offset;
1692         else
1693                 *stat = (uint64_t)((*stat +
1694                         ((uint64_t)1 << IAVF_48_BIT_WIDTH)) - *offset);
1695
1696         *stat &= IAVF_48_BIT_MASK;
1697 }
1698
1699 static void
1700 iavf_stat_update_32(uint64_t *offset, uint64_t *stat)
1701 {
1702         if (*stat >= *offset)
1703                 *stat = (uint64_t)(*stat - *offset);
1704         else
1705                 *stat = (uint64_t)((*stat +
1706                         ((uint64_t)1 << IAVF_32_BIT_WIDTH)) - *offset);
1707 }
1708
1709 static void
1710 iavf_update_stats(struct iavf_vsi *vsi, struct virtchnl_eth_stats *nes)
1711 {
1712         struct virtchnl_eth_stats *oes = &vsi->eth_stats_offset.eth_stats;
1713
1714         iavf_stat_update_48(&oes->rx_bytes, &nes->rx_bytes);
1715         iavf_stat_update_48(&oes->rx_unicast, &nes->rx_unicast);
1716         iavf_stat_update_48(&oes->rx_multicast, &nes->rx_multicast);
1717         iavf_stat_update_48(&oes->rx_broadcast, &nes->rx_broadcast);
1718         iavf_stat_update_32(&oes->rx_discards, &nes->rx_discards);
1719         iavf_stat_update_48(&oes->tx_bytes, &nes->tx_bytes);
1720         iavf_stat_update_48(&oes->tx_unicast, &nes->tx_unicast);
1721         iavf_stat_update_48(&oes->tx_multicast, &nes->tx_multicast);
1722         iavf_stat_update_48(&oes->tx_broadcast, &nes->tx_broadcast);
1723         iavf_stat_update_32(&oes->tx_errors, &nes->tx_errors);
1724         iavf_stat_update_32(&oes->tx_discards, &nes->tx_discards);
1725 }
1726
1727 static int
1728 iavf_dev_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
1729 {
1730         struct iavf_adapter *adapter =
1731                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1732         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
1733         struct iavf_vsi *vsi = &vf->vsi;
1734         struct virtchnl_eth_stats *pstats = NULL;
1735         int ret;
1736
1737         ret = iavf_query_stats(adapter, &pstats);
1738         if (ret == 0) {
1739                 uint8_t crc_stats_len = (dev->data->dev_conf.rxmode.offloads &
1740                                          RTE_ETH_RX_OFFLOAD_KEEP_CRC) ? 0 :
1741                                          RTE_ETHER_CRC_LEN;
1742                 iavf_update_stats(vsi, pstats);
1743                 stats->ipackets = pstats->rx_unicast + pstats->rx_multicast +
1744                                 pstats->rx_broadcast - pstats->rx_discards;
1745                 stats->opackets = pstats->tx_broadcast + pstats->tx_multicast +
1746                                                 pstats->tx_unicast;
1747                 stats->imissed = pstats->rx_discards;
1748                 stats->oerrors = pstats->tx_errors + pstats->tx_discards;
1749                 stats->ibytes = pstats->rx_bytes;
1750                 stats->ibytes -= stats->ipackets * crc_stats_len;
1751                 stats->obytes = pstats->tx_bytes;
1752         } else {
1753                 PMD_DRV_LOG(ERR, "Get statistics failed");
1754         }
1755         return ret;
1756 }
1757
1758 static int
1759 iavf_dev_stats_reset(struct rte_eth_dev *dev)
1760 {
1761         int ret;
1762         struct iavf_adapter *adapter =
1763                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1764         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
1765         struct iavf_vsi *vsi = &vf->vsi;
1766         struct virtchnl_eth_stats *pstats = NULL;
1767
1768         /* read stat values to clear hardware registers */
1769         ret = iavf_query_stats(adapter, &pstats);
1770         if (ret != 0)
1771                 return ret;
1772
1773         /* set stats offset base on current values */
1774         vsi->eth_stats_offset.eth_stats = *pstats;
1775
1776         return 0;
1777 }
1778
1779 static int
1780 iavf_dev_xstats_reset(struct rte_eth_dev *dev)
1781 {
1782         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
1783         iavf_dev_stats_reset(dev);
1784         memset(&vf->vsi.eth_stats_offset.ips_stats, 0,
1785                         sizeof(struct iavf_ipsec_crypto_stats));
1786         return 0;
1787 }
1788
1789 static int iavf_dev_xstats_get_names(__rte_unused struct rte_eth_dev *dev,
1790                                       struct rte_eth_xstat_name *xstats_names,
1791                                       __rte_unused unsigned int limit)
1792 {
1793         unsigned int i;
1794
1795         if (xstats_names != NULL)
1796                 for (i = 0; i < IAVF_NB_XSTATS; i++) {
1797                         snprintf(xstats_names[i].name,
1798                                 sizeof(xstats_names[i].name),
1799                                 "%s", rte_iavf_stats_strings[i].name);
1800                 }
1801         return IAVF_NB_XSTATS;
1802 }
1803
1804 static void
1805 iavf_dev_update_ipsec_xstats(struct rte_eth_dev *ethdev,
1806                 struct iavf_ipsec_crypto_stats *ips)
1807 {
1808         uint16_t idx;
1809         for (idx = 0; idx < ethdev->data->nb_rx_queues; idx++) {
1810                 struct iavf_rx_queue *rxq;
1811                 struct iavf_ipsec_crypto_stats *stats;
1812                 rxq = (struct iavf_rx_queue *)ethdev->data->rx_queues[idx];
1813                 stats = &rxq->stats.ipsec_crypto;
1814                 ips->icount += stats->icount;
1815                 ips->ibytes += stats->ibytes;
1816                 ips->ierrors.count += stats->ierrors.count;
1817                 ips->ierrors.sad_miss += stats->ierrors.sad_miss;
1818                 ips->ierrors.not_processed += stats->ierrors.not_processed;
1819                 ips->ierrors.icv_check += stats->ierrors.icv_check;
1820                 ips->ierrors.ipsec_length += stats->ierrors.ipsec_length;
1821                 ips->ierrors.misc += stats->ierrors.misc;
1822         }
1823 }
1824
1825 static int iavf_dev_xstats_get(struct rte_eth_dev *dev,
1826                                  struct rte_eth_xstat *xstats, unsigned int n)
1827 {
1828         int ret;
1829         unsigned int i;
1830         struct iavf_adapter *adapter =
1831                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1832         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
1833         struct iavf_vsi *vsi = &vf->vsi;
1834         struct virtchnl_eth_stats *pstats = NULL;
1835         struct iavf_eth_xstats iavf_xtats = {{0}};
1836
1837         if (n < IAVF_NB_XSTATS)
1838                 return IAVF_NB_XSTATS;
1839
1840         ret = iavf_query_stats(adapter, &pstats);
1841         if (ret != 0)
1842                 return 0;
1843
1844         if (!xstats)
1845                 return 0;
1846
1847         iavf_update_stats(vsi, pstats);
1848         iavf_xtats.eth_stats = *pstats;
1849
1850         if (iavf_ipsec_crypto_supported(adapter))
1851                 iavf_dev_update_ipsec_xstats(dev, &iavf_xtats.ips_stats);
1852
1853         /* loop over xstats array and values from pstats */
1854         for (i = 0; i < IAVF_NB_XSTATS; i++) {
1855                 xstats[i].id = i;
1856                 xstats[i].value = *(uint64_t *)(((char *)&iavf_xtats) +
1857                         rte_iavf_stats_strings[i].offset);
1858         }
1859
1860         return IAVF_NB_XSTATS;
1861 }
1862
1863
1864 static int
1865 iavf_dev_rx_queue_intr_enable(struct rte_eth_dev *dev, uint16_t queue_id)
1866 {
1867         struct iavf_adapter *adapter =
1868                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1869         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1870         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(adapter);
1871         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
1872         uint16_t msix_intr;
1873
1874         if (adapter->closed)
1875                 return -EIO;
1876
1877         msix_intr = rte_intr_vec_list_index_get(pci_dev->intr_handle,
1878                                                        queue_id);
1879         if (msix_intr == IAVF_MISC_VEC_ID) {
1880                 PMD_DRV_LOG(INFO, "MISC is also enabled for control");
1881                 IAVF_WRITE_REG(hw, IAVF_VFINT_DYN_CTL01,
1882                                IAVF_VFINT_DYN_CTL01_INTENA_MASK |
1883                                IAVF_VFINT_DYN_CTL01_CLEARPBA_MASK |
1884                                IAVF_VFINT_DYN_CTL01_ITR_INDX_MASK);
1885         } else {
1886                 IAVF_WRITE_REG(hw,
1887                                IAVF_VFINT_DYN_CTLN1
1888                                 (msix_intr - IAVF_RX_VEC_START),
1889                                IAVF_VFINT_DYN_CTLN1_INTENA_MASK |
1890                                IAVF_VFINT_DYN_CTL01_CLEARPBA_MASK |
1891                                IAVF_VFINT_DYN_CTLN1_ITR_INDX_MASK);
1892         }
1893
1894         IAVF_WRITE_FLUSH(hw);
1895
1896         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_WB_ON_ITR)
1897                 rte_intr_ack(pci_dev->intr_handle);
1898
1899         return 0;
1900 }
1901
1902 static int
1903 iavf_dev_rx_queue_intr_disable(struct rte_eth_dev *dev, uint16_t queue_id)
1904 {
1905         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1906         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1907         uint16_t msix_intr;
1908
1909         msix_intr = rte_intr_vec_list_index_get(pci_dev->intr_handle,
1910                                                        queue_id);
1911         if (msix_intr == IAVF_MISC_VEC_ID) {
1912                 PMD_DRV_LOG(ERR, "MISC is used for control, cannot disable it");
1913                 return -EIO;
1914         }
1915
1916         IAVF_WRITE_REG(hw,
1917                       IAVF_VFINT_DYN_CTLN1(msix_intr - IAVF_RX_VEC_START),
1918                       IAVF_VFINT_DYN_CTLN1_WB_ON_ITR_MASK);
1919
1920         IAVF_WRITE_FLUSH(hw);
1921         return 0;
1922 }
1923
1924 static int
1925 iavf_check_vf_reset_done(struct iavf_hw *hw)
1926 {
1927         int i, reset;
1928
1929         for (i = 0; i < IAVF_RESET_WAIT_CNT; i++) {
1930                 reset = IAVF_READ_REG(hw, IAVF_VFGEN_RSTAT) &
1931                         IAVF_VFGEN_RSTAT_VFR_STATE_MASK;
1932                 reset = reset >> IAVF_VFGEN_RSTAT_VFR_STATE_SHIFT;
1933                 if (reset == VIRTCHNL_VFR_VFACTIVE ||
1934                     reset == VIRTCHNL_VFR_COMPLETED)
1935                         break;
1936                 rte_delay_ms(20);
1937         }
1938
1939         if (i >= IAVF_RESET_WAIT_CNT)
1940                 return -1;
1941
1942         return 0;
1943 }
1944
1945 static int
1946 iavf_lookup_proto_xtr_type(const char *flex_name)
1947 {
1948         static struct {
1949                 const char *name;
1950                 enum iavf_proto_xtr_type type;
1951         } xtr_type_map[] = {
1952                 { "vlan",      IAVF_PROTO_XTR_VLAN      },
1953                 { "ipv4",      IAVF_PROTO_XTR_IPV4      },
1954                 { "ipv6",      IAVF_PROTO_XTR_IPV6      },
1955                 { "ipv6_flow", IAVF_PROTO_XTR_IPV6_FLOW },
1956                 { "tcp",       IAVF_PROTO_XTR_TCP       },
1957                 { "ip_offset", IAVF_PROTO_XTR_IP_OFFSET },
1958                 { "ipsec_crypto_said", IAVF_PROTO_XTR_IPSEC_CRYPTO_SAID },
1959         };
1960         uint32_t i;
1961
1962         for (i = 0; i < RTE_DIM(xtr_type_map); i++) {
1963                 if (strcmp(flex_name, xtr_type_map[i].name) == 0)
1964                         return xtr_type_map[i].type;
1965         }
1966
1967         PMD_DRV_LOG(ERR, "wrong proto_xtr type, it should be: "
1968                         "vlan|ipv4|ipv6|ipv6_flow|tcp|ip_offset|ipsec_crypto_said");
1969
1970         return -1;
1971 }
1972
1973 /**
1974  * Parse elem, the elem could be single number/range or '(' ')' group
1975  * 1) A single number elem, it's just a simple digit. e.g. 9
1976  * 2) A single range elem, two digits with a '-' between. e.g. 2-6
1977  * 3) A group elem, combines multiple 1) or 2) with '( )'. e.g (0,2-4,6)
1978  *    Within group elem, '-' used for a range separator;
1979  *                       ',' used for a single number.
1980  */
1981 static int
1982 iavf_parse_queue_set(const char *input, int xtr_type,
1983                      struct iavf_devargs *devargs)
1984 {
1985         const char *str = input;
1986         char *end = NULL;
1987         uint32_t min, max;
1988         uint32_t idx;
1989
1990         while (isblank(*str))
1991                 str++;
1992
1993         if (!isdigit(*str) && *str != '(')
1994                 return -1;
1995
1996         /* process single number or single range of number */
1997         if (*str != '(') {
1998                 errno = 0;
1999                 idx = strtoul(str, &end, 10);
2000                 if (errno || !end || idx >= IAVF_MAX_QUEUE_NUM)
2001                         return -1;
2002
2003                 while (isblank(*end))
2004                         end++;
2005
2006                 min = idx;
2007                 max = idx;
2008
2009                 /* process single <number>-<number> */
2010                 if (*end == '-') {
2011                         end++;
2012                         while (isblank(*end))
2013                                 end++;
2014                         if (!isdigit(*end))
2015                                 return -1;
2016
2017                         errno = 0;
2018                         idx = strtoul(end, &end, 10);
2019                         if (errno || !end || idx >= IAVF_MAX_QUEUE_NUM)
2020                                 return -1;
2021
2022                         max = idx;
2023                         while (isblank(*end))
2024                                 end++;
2025                 }
2026
2027                 if (*end != ':')
2028                         return -1;
2029
2030                 for (idx = RTE_MIN(min, max);
2031                      idx <= RTE_MAX(min, max); idx++)
2032                         devargs->proto_xtr[idx] = xtr_type;
2033
2034                 return 0;
2035         }
2036
2037         /* process set within bracket */
2038         str++;
2039         while (isblank(*str))
2040                 str++;
2041         if (*str == '\0')
2042                 return -1;
2043
2044         min = IAVF_MAX_QUEUE_NUM;
2045         do {
2046                 /* go ahead to the first digit */
2047                 while (isblank(*str))
2048                         str++;
2049                 if (!isdigit(*str))
2050                         return -1;
2051
2052                 /* get the digit value */
2053                 errno = 0;
2054                 idx = strtoul(str, &end, 10);
2055                 if (errno || !end || idx >= IAVF_MAX_QUEUE_NUM)
2056                         return -1;
2057
2058                 /* go ahead to separator '-',',' and ')' */
2059                 while (isblank(*end))
2060                         end++;
2061                 if (*end == '-') {
2062                         if (min == IAVF_MAX_QUEUE_NUM)
2063                                 min = idx;
2064                         else /* avoid continuous '-' */
2065                                 return -1;
2066                 } else if (*end == ',' || *end == ')') {
2067                         max = idx;
2068                         if (min == IAVF_MAX_QUEUE_NUM)
2069                                 min = idx;
2070
2071                         for (idx = RTE_MIN(min, max);
2072                              idx <= RTE_MAX(min, max); idx++)
2073                                 devargs->proto_xtr[idx] = xtr_type;
2074
2075                         min = IAVF_MAX_QUEUE_NUM;
2076                 } else {
2077                         return -1;
2078                 }
2079
2080                 str = end + 1;
2081         } while (*end != ')' && *end != '\0');
2082
2083         return 0;
2084 }
2085
2086 static int
2087 iavf_parse_queue_proto_xtr(const char *queues, struct iavf_devargs *devargs)
2088 {
2089         const char *queue_start;
2090         uint32_t idx;
2091         int xtr_type;
2092         char flex_name[32];
2093
2094         while (isblank(*queues))
2095                 queues++;
2096
2097         if (*queues != '[') {
2098                 xtr_type = iavf_lookup_proto_xtr_type(queues);
2099                 if (xtr_type < 0)
2100                         return -1;
2101
2102                 devargs->proto_xtr_dflt = xtr_type;
2103
2104                 return 0;
2105         }
2106
2107         queues++;
2108         do {
2109                 while (isblank(*queues))
2110                         queues++;
2111                 if (*queues == '\0')
2112                         return -1;
2113
2114                 queue_start = queues;
2115
2116                 /* go across a complete bracket */
2117                 if (*queue_start == '(') {
2118                         queues += strcspn(queues, ")");
2119                         if (*queues != ')')
2120                                 return -1;
2121                 }
2122
2123                 /* scan the separator ':' */
2124                 queues += strcspn(queues, ":");
2125                 if (*queues++ != ':')
2126                         return -1;
2127                 while (isblank(*queues))
2128                         queues++;
2129
2130                 for (idx = 0; ; idx++) {
2131                         if (isblank(queues[idx]) ||
2132                             queues[idx] == ',' ||
2133                             queues[idx] == ']' ||
2134                             queues[idx] == '\0')
2135                                 break;
2136
2137                         if (idx > sizeof(flex_name) - 2)
2138                                 return -1;
2139
2140                         flex_name[idx] = queues[idx];
2141                 }
2142                 flex_name[idx] = '\0';
2143                 xtr_type = iavf_lookup_proto_xtr_type(flex_name);
2144                 if (xtr_type < 0)
2145                         return -1;
2146
2147                 queues += idx;
2148
2149                 while (isblank(*queues) || *queues == ',' || *queues == ']')
2150                         queues++;
2151
2152                 if (iavf_parse_queue_set(queue_start, xtr_type, devargs) < 0)
2153                         return -1;
2154         } while (*queues != '\0');
2155
2156         return 0;
2157 }
2158
2159 static int
2160 iavf_handle_proto_xtr_arg(__rte_unused const char *key, const char *value,
2161                           void *extra_args)
2162 {
2163         struct iavf_devargs *devargs = extra_args;
2164
2165         if (!value || !extra_args)
2166                 return -EINVAL;
2167
2168         if (iavf_parse_queue_proto_xtr(value, devargs) < 0) {
2169                 PMD_DRV_LOG(ERR, "the proto_xtr's parameter is wrong : '%s'",
2170                             value);
2171                 return -1;
2172         }
2173
2174         return 0;
2175 }
2176
2177 static int
2178 parse_u16(__rte_unused const char *key, const char *value, void *args)
2179 {
2180         u16 *num = (u16 *)args;
2181         u16 tmp;
2182
2183         errno = 0;
2184         tmp = strtoull(value, NULL, 10);
2185         if (errno || !tmp) {
2186                 PMD_DRV_LOG(WARNING, "%s: \"%s\" is not a valid u16",
2187                             key, value);
2188                 return -1;
2189         }
2190
2191         *num = tmp;
2192
2193         return 0;
2194 }
2195
2196 static int iavf_parse_devargs(struct rte_eth_dev *dev)
2197 {
2198         struct iavf_adapter *ad =
2199                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
2200         struct rte_devargs *devargs = dev->device->devargs;
2201         struct rte_kvargs *kvlist;
2202         int ret;
2203
2204         if (!devargs)
2205                 return 0;
2206
2207         kvlist = rte_kvargs_parse(devargs->args, iavf_valid_args);
2208         if (!kvlist) {
2209                 PMD_INIT_LOG(ERR, "invalid kvargs key\n");
2210                 return -EINVAL;
2211         }
2212
2213         ad->devargs.proto_xtr_dflt = IAVF_PROTO_XTR_NONE;
2214         memset(ad->devargs.proto_xtr, IAVF_PROTO_XTR_NONE,
2215                sizeof(ad->devargs.proto_xtr));
2216
2217         ret = rte_kvargs_process(kvlist, IAVF_PROTO_XTR_ARG,
2218                                  &iavf_handle_proto_xtr_arg, &ad->devargs);
2219         if (ret)
2220                 goto bail;
2221
2222         ret = rte_kvargs_process(kvlist, IAVF_QUANTA_SIZE_ARG,
2223                                  &parse_u16, &ad->devargs.quanta_size);
2224         if (ret)
2225                 goto bail;
2226
2227         if (ad->devargs.quanta_size != 0 &&
2228             (ad->devargs.quanta_size < 256 || ad->devargs.quanta_size > 4096 ||
2229              ad->devargs.quanta_size & 0x40)) {
2230                 PMD_INIT_LOG(ERR, "invalid quanta size\n");
2231                 ret = -EINVAL;
2232                 goto bail;
2233         }
2234
2235 bail:
2236         rte_kvargs_free(kvlist);
2237         return ret;
2238 }
2239
2240 static void
2241 iavf_init_proto_xtr(struct rte_eth_dev *dev)
2242 {
2243         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
2244         struct iavf_adapter *ad =
2245                         IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
2246         const struct iavf_proto_xtr_ol *xtr_ol;
2247         bool proto_xtr_enable = false;
2248         int offset;
2249         uint16_t i;
2250
2251         vf->proto_xtr = rte_zmalloc("vf proto xtr",
2252                                     vf->vsi_res->num_queue_pairs, 0);
2253         if (unlikely(!(vf->proto_xtr))) {
2254                 PMD_DRV_LOG(ERR, "no memory for setting up proto_xtr's table");
2255                 return;
2256         }
2257
2258         for (i = 0; i < vf->vsi_res->num_queue_pairs; i++) {
2259                 vf->proto_xtr[i] = ad->devargs.proto_xtr[i] !=
2260                                         IAVF_PROTO_XTR_NONE ?
2261                                         ad->devargs.proto_xtr[i] :
2262                                         ad->devargs.proto_xtr_dflt;
2263
2264                 if (vf->proto_xtr[i] != IAVF_PROTO_XTR_NONE) {
2265                         uint8_t type = vf->proto_xtr[i];
2266
2267                         iavf_proto_xtr_params[type].required = true;
2268                         proto_xtr_enable = true;
2269                 }
2270         }
2271
2272         if (likely(!proto_xtr_enable))
2273                 return;
2274
2275         offset = rte_mbuf_dynfield_register(&iavf_proto_xtr_metadata_param);
2276         if (unlikely(offset == -1)) {
2277                 PMD_DRV_LOG(ERR,
2278                             "failed to extract protocol metadata, error %d",
2279                             -rte_errno);
2280                 return;
2281         }
2282
2283         PMD_DRV_LOG(DEBUG,
2284                     "proto_xtr metadata offset in mbuf is : %d",
2285                     offset);
2286         rte_pmd_ifd_dynfield_proto_xtr_metadata_offs = offset;
2287
2288         for (i = 0; i < RTE_DIM(iavf_proto_xtr_params); i++) {
2289                 xtr_ol = &iavf_proto_xtr_params[i];
2290
2291                 uint8_t rxdid = iavf_proto_xtr_type_to_rxdid((uint8_t)i);
2292
2293                 if (!xtr_ol->required)
2294                         continue;
2295
2296                 if (!(vf->supported_rxdid & BIT(rxdid))) {
2297                         PMD_DRV_LOG(ERR,
2298                                     "rxdid[%u] is not supported in hardware",
2299                                     rxdid);
2300                         rte_pmd_ifd_dynfield_proto_xtr_metadata_offs = -1;
2301                         break;
2302                 }
2303
2304                 offset = rte_mbuf_dynflag_register(&xtr_ol->param);
2305                 if (unlikely(offset == -1)) {
2306                         PMD_DRV_LOG(ERR,
2307                                     "failed to register proto_xtr offload '%s', error %d",
2308                                     xtr_ol->param.name, -rte_errno);
2309
2310                         rte_pmd_ifd_dynfield_proto_xtr_metadata_offs = -1;
2311                         break;
2312                 }
2313
2314                 PMD_DRV_LOG(DEBUG,
2315                             "proto_xtr offload '%s' offset in mbuf is : %d",
2316                             xtr_ol->param.name, offset);
2317                 *xtr_ol->ol_flag = 1ULL << offset;
2318         }
2319 }
2320
2321 static int
2322 iavf_init_vf(struct rte_eth_dev *dev)
2323 {
2324         int err, bufsz;
2325         struct iavf_adapter *adapter =
2326                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
2327         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2328         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
2329
2330         vf->eth_dev = dev;
2331
2332         err = iavf_parse_devargs(dev);
2333         if (err) {
2334                 PMD_INIT_LOG(ERR, "Failed to parse devargs");
2335                 goto err;
2336         }
2337
2338         err = iavf_set_mac_type(hw);
2339         if (err) {
2340                 PMD_INIT_LOG(ERR, "set_mac_type failed: %d", err);
2341                 goto err;
2342         }
2343
2344         err = iavf_check_vf_reset_done(hw);
2345         if (err) {
2346                 PMD_INIT_LOG(ERR, "VF is still resetting");
2347                 goto err;
2348         }
2349
2350         iavf_init_adminq_parameter(hw);
2351         err = iavf_init_adminq(hw);
2352         if (err) {
2353                 PMD_INIT_LOG(ERR, "init_adminq failed: %d", err);
2354                 goto err;
2355         }
2356
2357         vf->aq_resp = rte_zmalloc("vf_aq_resp", IAVF_AQ_BUF_SZ, 0);
2358         if (!vf->aq_resp) {
2359                 PMD_INIT_LOG(ERR, "unable to allocate vf_aq_resp memory");
2360                 goto err_aq;
2361         }
2362         if (iavf_check_api_version(adapter) != 0) {
2363                 PMD_INIT_LOG(ERR, "check_api version failed");
2364                 goto err_api;
2365         }
2366
2367         bufsz = sizeof(struct virtchnl_vf_resource) +
2368                 (IAVF_MAX_VF_VSI * sizeof(struct virtchnl_vsi_resource));
2369         vf->vf_res = rte_zmalloc("vf_res", bufsz, 0);
2370         if (!vf->vf_res) {
2371                 PMD_INIT_LOG(ERR, "unable to allocate vf_res memory");
2372                 goto err_api;
2373         }
2374
2375         if (iavf_get_vf_resource(adapter) != 0) {
2376                 PMD_INIT_LOG(ERR, "iavf_get_vf_config failed");
2377                 goto err_alloc;
2378         }
2379         /* Allocate memort for RSS info */
2380         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF) {
2381                 vf->rss_key = rte_zmalloc("rss_key",
2382                                           vf->vf_res->rss_key_size, 0);
2383                 if (!vf->rss_key) {
2384                         PMD_INIT_LOG(ERR, "unable to allocate rss_key memory");
2385                         goto err_rss;
2386                 }
2387                 vf->rss_lut = rte_zmalloc("rss_lut",
2388                                           vf->vf_res->rss_lut_size, 0);
2389                 if (!vf->rss_lut) {
2390                         PMD_INIT_LOG(ERR, "unable to allocate rss_lut memory");
2391                         goto err_rss;
2392                 }
2393         }
2394
2395         if (vf->vsi_res->num_queue_pairs > IAVF_MAX_NUM_QUEUES_DFLT)
2396                 vf->lv_enabled = true;
2397
2398         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RX_FLEX_DESC) {
2399                 if (iavf_get_supported_rxdid(adapter) != 0) {
2400                         PMD_INIT_LOG(ERR, "failed to do get supported rxdid");
2401                         goto err_rss;
2402                 }
2403         }
2404
2405         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_VLAN_V2) {
2406                 if (iavf_get_vlan_offload_caps_v2(adapter) != 0) {
2407                         PMD_INIT_LOG(ERR, "failed to do get VLAN offload v2 capabilities");
2408                         goto err_rss;
2409                 }
2410         }
2411
2412         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_QOS) {
2413                 bufsz = sizeof(struct virtchnl_qos_cap_list) +
2414                         IAVF_MAX_TRAFFIC_CLASS *
2415                         sizeof(struct virtchnl_qos_cap_elem);
2416                 vf->qos_cap = rte_zmalloc("qos_cap", bufsz, 0);
2417                 if (!vf->qos_cap) {
2418                         PMD_INIT_LOG(ERR, "unable to allocate qos_cap memory");
2419                         goto err_rss;
2420                 }
2421                 iavf_tm_conf_init(dev);
2422         }
2423
2424         iavf_init_proto_xtr(dev);
2425
2426         return 0;
2427 err_rss:
2428         rte_free(vf->rss_key);
2429         rte_free(vf->rss_lut);
2430 err_alloc:
2431         rte_free(vf->qos_cap);
2432         rte_free(vf->vf_res);
2433         vf->vsi_res = NULL;
2434 err_api:
2435         rte_free(vf->aq_resp);
2436 err_aq:
2437         iavf_shutdown_adminq(hw);
2438 err:
2439         return -1;
2440 }
2441
2442 static void
2443 iavf_uninit_vf(struct rte_eth_dev *dev)
2444 {
2445         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2446         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
2447
2448         iavf_shutdown_adminq(hw);
2449
2450         rte_free(vf->vf_res);
2451         vf->vsi_res = NULL;
2452         vf->vf_res = NULL;
2453
2454         rte_free(vf->aq_resp);
2455         vf->aq_resp = NULL;
2456
2457         rte_free(vf->qos_cap);
2458         vf->qos_cap = NULL;
2459
2460         rte_free(vf->rss_lut);
2461         vf->rss_lut = NULL;
2462         rte_free(vf->rss_key);
2463         vf->rss_key = NULL;
2464 }
2465
2466 /* Enable default admin queue interrupt setting */
2467 static inline void
2468 iavf_enable_irq0(struct iavf_hw *hw)
2469 {
2470         /* Enable admin queue interrupt trigger */
2471         IAVF_WRITE_REG(hw, IAVF_VFINT_ICR0_ENA1,
2472                        IAVF_VFINT_ICR0_ENA1_ADMINQ_MASK);
2473
2474         IAVF_WRITE_REG(hw, IAVF_VFINT_DYN_CTL01,
2475                        IAVF_VFINT_DYN_CTL01_INTENA_MASK |
2476                        IAVF_VFINT_DYN_CTL01_CLEARPBA_MASK |
2477                        IAVF_VFINT_DYN_CTL01_ITR_INDX_MASK);
2478
2479         IAVF_WRITE_FLUSH(hw);
2480 }
2481
2482 static inline void
2483 iavf_disable_irq0(struct iavf_hw *hw)
2484 {
2485         /* Disable all interrupt types */
2486         IAVF_WRITE_REG(hw, IAVF_VFINT_ICR0_ENA1, 0);
2487         IAVF_WRITE_REG(hw, IAVF_VFINT_DYN_CTL01,
2488                        IAVF_VFINT_DYN_CTL01_ITR_INDX_MASK);
2489         IAVF_WRITE_FLUSH(hw);
2490 }
2491
2492 static void
2493 iavf_dev_interrupt_handler(void *param)
2494 {
2495         struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
2496         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2497
2498         iavf_disable_irq0(hw);
2499
2500         iavf_handle_virtchnl_msg(dev);
2501
2502         iavf_enable_irq0(hw);
2503 }
2504
2505 void
2506 iavf_dev_alarm_handler(void *param)
2507 {
2508         struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
2509         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2510         uint32_t icr0;
2511
2512         iavf_disable_irq0(hw);
2513
2514         /* read out interrupt causes */
2515         icr0 = IAVF_READ_REG(hw, IAVF_VFINT_ICR01);
2516
2517         if (icr0 & IAVF_VFINT_ICR01_ADMINQ_MASK) {
2518                 PMD_DRV_LOG(DEBUG, "ICR01_ADMINQ is reported");
2519                 iavf_handle_virtchnl_msg(dev);
2520         }
2521
2522         iavf_enable_irq0(hw);
2523
2524         rte_eal_alarm_set(IAVF_ALARM_INTERVAL,
2525                           iavf_dev_alarm_handler, dev);
2526 }
2527
2528 static int
2529 iavf_dev_flow_ops_get(struct rte_eth_dev *dev,
2530                       const struct rte_flow_ops **ops)
2531 {
2532         struct iavf_adapter *adapter =
2533                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
2534
2535         if (adapter->closed)
2536                 return -EIO;
2537
2538         *ops = &iavf_flow_ops;
2539         return 0;
2540 }
2541
2542 static void
2543 iavf_default_rss_disable(struct iavf_adapter *adapter)
2544 {
2545         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
2546         int ret = 0;
2547
2548         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF) {
2549                 /* Set hena = 0 to ask PF to cleanup all existing RSS. */
2550                 ret = iavf_set_hena(adapter, 0);
2551                 if (ret)
2552                         /* It is a workaround, temporarily allow error to be
2553                          * returned due to possible lack of PF handling for
2554                          * hena = 0.
2555                          */
2556                         PMD_INIT_LOG(WARNING, "fail to disable default RSS,"
2557                                     "lack PF support");
2558         }
2559 }
2560
2561 static int
2562 iavf_dev_init(struct rte_eth_dev *eth_dev)
2563 {
2564         struct iavf_adapter *adapter =
2565                 IAVF_DEV_PRIVATE_TO_ADAPTER(eth_dev->data->dev_private);
2566         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(adapter);
2567         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
2568         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
2569         int ret = 0;
2570
2571         PMD_INIT_FUNC_TRACE();
2572
2573         /* assign ops func pointer */
2574         eth_dev->dev_ops = &iavf_eth_dev_ops;
2575         eth_dev->rx_queue_count = iavf_dev_rxq_count;
2576         eth_dev->rx_descriptor_status = iavf_dev_rx_desc_status;
2577         eth_dev->tx_descriptor_status = iavf_dev_tx_desc_status;
2578         eth_dev->rx_pkt_burst = &iavf_recv_pkts;
2579         eth_dev->tx_pkt_burst = &iavf_xmit_pkts;
2580         eth_dev->tx_pkt_prepare = &iavf_prep_pkts;
2581
2582         /* For secondary processes, we don't initialise any further as primary
2583          * has already done this work. Only check if we need a different RX
2584          * and TX function.
2585          */
2586         if (rte_eal_process_type() != RTE_PROC_PRIMARY) {
2587                 iavf_set_rx_function(eth_dev);
2588                 iavf_set_tx_function(eth_dev);
2589                 return 0;
2590         }
2591         rte_eth_copy_pci_info(eth_dev, pci_dev);
2592
2593         hw->vendor_id = pci_dev->id.vendor_id;
2594         hw->device_id = pci_dev->id.device_id;
2595         hw->subsystem_vendor_id = pci_dev->id.subsystem_vendor_id;
2596         hw->subsystem_device_id = pci_dev->id.subsystem_device_id;
2597         hw->bus.bus_id = pci_dev->addr.bus;
2598         hw->bus.device = pci_dev->addr.devid;
2599         hw->bus.func = pci_dev->addr.function;
2600         hw->hw_addr = (void *)pci_dev->mem_resource[0].addr;
2601         hw->back = IAVF_DEV_PRIVATE_TO_ADAPTER(eth_dev->data->dev_private);
2602         adapter->dev_data = eth_dev->data;
2603         adapter->stopped = 1;
2604
2605         if (iavf_init_vf(eth_dev) != 0) {
2606                 PMD_INIT_LOG(ERR, "Init vf failed");
2607                 return -1;
2608         }
2609
2610         /* set default ptype table */
2611         iavf_set_default_ptype_table(eth_dev);
2612
2613         /* copy mac addr */
2614         eth_dev->data->mac_addrs = rte_zmalloc(
2615                 "iavf_mac", RTE_ETHER_ADDR_LEN * IAVF_NUM_MACADDR_MAX, 0);
2616         if (!eth_dev->data->mac_addrs) {
2617                 PMD_INIT_LOG(ERR, "Failed to allocate %d bytes needed to"
2618                              " store MAC addresses",
2619                              RTE_ETHER_ADDR_LEN * IAVF_NUM_MACADDR_MAX);
2620                 ret = -ENOMEM;
2621                 goto init_vf_err;
2622         }
2623         /* If the MAC address is not configured by host,
2624          * generate a random one.
2625          */
2626         if (!rte_is_valid_assigned_ether_addr(
2627                         (struct rte_ether_addr *)hw->mac.addr))
2628                 rte_eth_random_addr(hw->mac.addr);
2629         rte_ether_addr_copy((struct rte_ether_addr *)hw->mac.addr,
2630                         &eth_dev->data->mac_addrs[0]);
2631
2632         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_WB_ON_ITR) {
2633                 /* register callback func to eal lib */
2634                 rte_intr_callback_register(pci_dev->intr_handle,
2635                                            iavf_dev_interrupt_handler,
2636                                            (void *)eth_dev);
2637
2638                 /* enable uio intr after callback register */
2639                 rte_intr_enable(pci_dev->intr_handle);
2640         } else {
2641                 rte_eal_alarm_set(IAVF_ALARM_INTERVAL,
2642                                   iavf_dev_alarm_handler, eth_dev);
2643         }
2644
2645         /* configure and enable device interrupt */
2646         iavf_enable_irq0(hw);
2647
2648         ret = iavf_flow_init(adapter);
2649         if (ret) {
2650                 PMD_INIT_LOG(ERR, "Failed to initialize flow");
2651                 goto flow_init_err;
2652         }
2653
2654         /** Check if the IPsec Crypto offload is supported and create
2655          *  security_ctx if it is.
2656          */
2657         if (iavf_ipsec_crypto_supported(adapter)) {
2658                 /* Initialize security_ctx only for primary process*/
2659                 ret = iavf_security_ctx_create(adapter);
2660                 if (ret) {
2661                         PMD_INIT_LOG(ERR, "failed to create ipsec crypto security instance");
2662                         return ret;
2663                 }
2664
2665                 ret = iavf_security_init(adapter);
2666                 if (ret) {
2667                         PMD_INIT_LOG(ERR, "failed to initialized ipsec crypto resources");
2668                         return ret;
2669                 }
2670         }
2671
2672         iavf_default_rss_disable(adapter);
2673
2674
2675         /* Start device watchdog */
2676         iavf_dev_watchdog_enable(adapter);
2677         adapter->closed = false;
2678
2679         return 0;
2680
2681 flow_init_err:
2682         rte_free(eth_dev->data->mac_addrs);
2683         eth_dev->data->mac_addrs = NULL;
2684
2685 init_vf_err:
2686         iavf_uninit_vf(eth_dev);
2687
2688         return ret;
2689 }
2690
2691 static int
2692 iavf_dev_close(struct rte_eth_dev *dev)
2693 {
2694         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2695         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
2696         struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
2697         struct iavf_adapter *adapter =
2698                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
2699         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
2700         int ret;
2701
2702         if (rte_eal_process_type() != RTE_PROC_PRIMARY)
2703                 return 0;
2704
2705         if (adapter->closed) {
2706                 ret = 0;
2707                 goto out;
2708         }
2709
2710         ret = iavf_dev_stop(dev);
2711         adapter->closed = true;
2712
2713         /* free iAVF security device context all related resources */
2714         iavf_security_ctx_destroy(adapter);
2715
2716         iavf_flow_flush(dev, NULL);
2717         iavf_flow_uninit(adapter);
2718
2719         /*
2720          * disable promiscuous mode before reset vf
2721          * it is a workaround solution when work with kernel driver
2722          * and it is not the normal way
2723          */
2724         if (vf->promisc_unicast_enabled || vf->promisc_multicast_enabled)
2725                 iavf_config_promisc(adapter, false, false);
2726
2727         iavf_shutdown_adminq(hw);
2728         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_WB_ON_ITR) {
2729                 /* disable uio intr before callback unregister */
2730                 rte_intr_disable(intr_handle);
2731
2732                 /* unregister callback func from eal lib */
2733                 rte_intr_callback_unregister(intr_handle,
2734                                              iavf_dev_interrupt_handler, dev);
2735         } else {
2736                 rte_eal_alarm_cancel(iavf_dev_alarm_handler, dev);
2737         }
2738         iavf_disable_irq0(hw);
2739
2740         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_QOS)
2741                 iavf_tm_conf_uninit(dev);
2742
2743         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF) {
2744                 if (vf->rss_lut) {
2745                         rte_free(vf->rss_lut);
2746                         vf->rss_lut = NULL;
2747                 }
2748                 if (vf->rss_key) {
2749                         rte_free(vf->rss_key);
2750                         vf->rss_key = NULL;
2751                 }
2752         }
2753
2754         rte_free(vf->vf_res);
2755         vf->vsi_res = NULL;
2756         vf->vf_res = NULL;
2757
2758         rte_free(vf->aq_resp);
2759         vf->aq_resp = NULL;
2760
2761         /*
2762          * If the VF is reset via VFLR, the device will be knocked out of bus
2763          * master mode, and the driver will fail to recover from the reset. Fix
2764          * this by enabling bus mastering after every reset. In a non-VFLR case,
2765          * the bus master bit will not be disabled, and this call will have no
2766          * effect.
2767          */
2768 out:
2769         if (vf->vf_reset && !rte_pci_set_bus_master(pci_dev, true))
2770                 vf->vf_reset = false;
2771
2772         /* disable watchdog */
2773         iavf_dev_watchdog_disable(adapter);
2774
2775         return ret;
2776 }
2777
2778 static int
2779 iavf_dev_uninit(struct rte_eth_dev *dev)
2780 {
2781         if (rte_eal_process_type() != RTE_PROC_PRIMARY)
2782                 return -EPERM;
2783
2784         iavf_dev_close(dev);
2785
2786         return 0;
2787 }
2788
2789 /*
2790  * Reset VF device only to re-initialize resources in PMD layer
2791  */
2792 static int
2793 iavf_dev_reset(struct rte_eth_dev *dev)
2794 {
2795         int ret;
2796
2797         ret = iavf_dev_uninit(dev);
2798         if (ret)
2799                 return ret;
2800
2801         return iavf_dev_init(dev);
2802 }
2803
2804 static int
2805 iavf_dcf_cap_check_handler(__rte_unused const char *key,
2806                            const char *value, __rte_unused void *opaque)
2807 {
2808         if (strcmp(value, "dcf"))
2809                 return -1;
2810
2811         return 0;
2812 }
2813
2814 static int
2815 iavf_dcf_cap_selected(struct rte_devargs *devargs)
2816 {
2817         struct rte_kvargs *kvlist;
2818         const char *key = "cap";
2819         int ret = 0;
2820
2821         if (devargs == NULL)
2822                 return 0;
2823
2824         kvlist = rte_kvargs_parse(devargs->args, NULL);
2825         if (kvlist == NULL)
2826                 return 0;
2827
2828         if (!rte_kvargs_count(kvlist, key))
2829                 goto exit;
2830
2831         /* dcf capability selected when there's a key-value pair: cap=dcf */
2832         if (rte_kvargs_process(kvlist, key,
2833                                iavf_dcf_cap_check_handler, NULL) < 0)
2834                 goto exit;
2835
2836         ret = 1;
2837
2838 exit:
2839         rte_kvargs_free(kvlist);
2840         return ret;
2841 }
2842
2843 static int eth_iavf_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
2844                              struct rte_pci_device *pci_dev)
2845 {
2846         if (iavf_dcf_cap_selected(pci_dev->device.devargs))
2847                 return 1;
2848
2849         return rte_eth_dev_pci_generic_probe(pci_dev,
2850                 sizeof(struct iavf_adapter), iavf_dev_init);
2851 }
2852
2853 static int eth_iavf_pci_remove(struct rte_pci_device *pci_dev)
2854 {
2855         return rte_eth_dev_pci_generic_remove(pci_dev, iavf_dev_uninit);
2856 }
2857
2858 /* Adaptive virtual function driver struct */
2859 static struct rte_pci_driver rte_iavf_pmd = {
2860         .id_table = pci_id_iavf_map,
2861         .drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC,
2862         .probe = eth_iavf_pci_probe,
2863         .remove = eth_iavf_pci_remove,
2864 };
2865
2866 RTE_PMD_REGISTER_PCI(net_iavf, rte_iavf_pmd);
2867 RTE_PMD_REGISTER_PCI_TABLE(net_iavf, pci_id_iavf_map);
2868 RTE_PMD_REGISTER_KMOD_DEP(net_iavf, "* igb_uio | vfio-pci");
2869 RTE_PMD_REGISTER_PARAM_STRING(net_iavf, "cap=dcf");
2870 RTE_LOG_REGISTER_SUFFIX(iavf_logtype_init, init, NOTICE);
2871 RTE_LOG_REGISTER_SUFFIX(iavf_logtype_driver, driver, NOTICE);
2872 #ifdef RTE_ETHDEV_DEBUG_RX
2873 RTE_LOG_REGISTER_SUFFIX(iavf_logtype_rx, rx, DEBUG);
2874 #endif
2875 #ifdef RTE_ETHDEV_DEBUG_TX
2876 RTE_LOG_REGISTER_SUFFIX(iavf_logtype_tx, tx, DEBUG);
2877 #endif