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