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