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                                 ETH_RSS_NONFRAG_IPV4_UDP,
 270                 [IAVF_FILTER_PCTYPE_NONF_MULTICAST_IPV4_UDP] =
 271                                 ETH_RSS_NONFRAG_IPV4_UDP,
 272                 [IAVF_FILTER_PCTYPE_NONF_IPV4_UDP] =
 273                                 ETH_RSS_NONFRAG_IPV4_UDP,
 274                 [IAVF_FILTER_PCTYPE_NONF_IPV4_TCP_SYN_NO_ACK] =
 275                                 ETH_RSS_NONFRAG_IPV4_TCP,
 276                 [IAVF_FILTER_PCTYPE_NONF_IPV4_TCP] =
 277                                 ETH_RSS_NONFRAG_IPV4_TCP,
 278                 [IAVF_FILTER_PCTYPE_NONF_IPV4_SCTP] =
 279                                 ETH_RSS_NONFRAG_IPV4_SCTP,
 280                 [IAVF_FILTER_PCTYPE_NONF_IPV4_OTHER] =
 281                                 ETH_RSS_NONFRAG_IPV4_OTHER,
 282                 [IAVF_FILTER_PCTYPE_FRAG_IPV4] = ETH_RSS_FRAG_IPV4,
 283
 284                 /* IPv6 */
 285                 [IAVF_FILTER_PCTYPE_NONF_UNICAST_IPV6_UDP] =
 286                                 ETH_RSS_NONFRAG_IPV6_UDP,
 287                 [IAVF_FILTER_PCTYPE_NONF_MULTICAST_IPV6_UDP] =
 288                                 ETH_RSS_NONFRAG_IPV6_UDP,
 289                 [IAVF_FILTER_PCTYPE_NONF_IPV6_UDP] =
 290                                 ETH_RSS_NONFRAG_IPV6_UDP,
 291                 [IAVF_FILTER_PCTYPE_NONF_IPV6_TCP_SYN_NO_ACK] =
 292                                 ETH_RSS_NONFRAG_IPV6_TCP,
 293                 [IAVF_FILTER_PCTYPE_NONF_IPV6_TCP] =
 294                                 ETH_RSS_NONFRAG_IPV6_TCP,
 295                 [IAVF_FILTER_PCTYPE_NONF_IPV6_SCTP] =
 296                                 ETH_RSS_NONFRAG_IPV6_SCTP,
 297                 [IAVF_FILTER_PCTYPE_NONF_IPV6_OTHER] =
 298                                 ETH_RSS_NONFRAG_IPV6_OTHER,
 299                 [IAVF_FILTER_PCTYPE_FRAG_IPV6] = ETH_RSS_FRAG_IPV6,
 300
 301                 /* L2 Payload */
 302                 [IAVF_FILTER_PCTYPE_L2_PAYLOAD] = ETH_RSS_L2_PAYLOAD
 303         };
 304
 305         const uint64_t ipv4_rss = ETH_RSS_NONFRAG_IPV4_UDP |
 306                                   ETH_RSS_NONFRAG_IPV4_TCP |
 307                                   ETH_RSS_NONFRAG_IPV4_SCTP |
 308                                   ETH_RSS_NONFRAG_IPV4_OTHER |
 309                                   ETH_RSS_FRAG_IPV4;
 310
 311         const uint64_t ipv6_rss = ETH_RSS_NONFRAG_IPV6_UDP |
 312                                   ETH_RSS_NONFRAG_IPV6_TCP |
 313                                   ETH_RSS_NONFRAG_IPV6_SCTP |
 314                                   ETH_RSS_NONFRAG_IPV6_OTHER |
 315                                   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          * ETH_RSS_IPV4 and ETH_RSS_IPV6 can be considered as 2
 335          * generalizations of all other IPv4 and IPv6 RSS types.
 336          */
 337         if (rss_hf & ETH_RSS_IPV4)
 338                 rss_hf |= ipv4_rss;
 339
 340         if (rss_hf & 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 & ETH_RSS_IPV4;
 367
 368         if (valid_rss_hf & ipv6_rss)
 369                 valid_rss_hf |= rss_hf & 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->eth_dev->data->dev_conf.rx_adv_conf.rss_conf;
 387         nb_q = RTE_MIN(adapter->eth_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(ad, 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                     DEV_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                                         ETH_VLAN_STRIP_MASK |
 483                                         ETH_QINQ_STRIP_MASK |
 484                                         ETH_VLAN_FILTER_MASK |
 485                                         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 & ETH_MQ_RX_RSS_FLAG)
 516                 dev->data->dev_conf.rxmode.offloads |= DEV_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
 580         buf_size = rte_pktmbuf_data_room_size(rxq->mp) - RTE_PKTMBUF_HEADROOM;
 581
 582         /* Calculate the maximum packet length allowed */
 583         max_pkt_len = RTE_MIN((uint32_t)
 584                         rxq->rx_buf_len * IAVF_MAX_CHAINED_RX_BUFFERS,
 585                         dev->data->dev_conf.rxmode.max_rx_pkt_len);
 586
 587         /* Check if the jumbo frame and maximum packet length are set
 588          * correctly.
 589          */
 590         if (dev->data->dev_conf.rxmode.offloads & DEV_RX_OFFLOAD_JUMBO_FRAME) {
 591                 if (max_pkt_len <= IAVF_ETH_MAX_LEN ||
 592                     max_pkt_len > IAVF_FRAME_SIZE_MAX) {
 593                         PMD_DRV_LOG(ERR, "maximum packet length must be "
 594                                     "larger than %u and smaller than %u, "
 595                                     "as jumbo frame is enabled",
 596                                     (uint32_t)IAVF_ETH_MAX_LEN,
 597                                     (uint32_t)IAVF_FRAME_SIZE_MAX);
 598                         return -EINVAL;
 599                 }
 600         } else {
 601                 if (max_pkt_len < RTE_ETHER_MIN_LEN ||
 602                     max_pkt_len > IAVF_ETH_MAX_LEN) {
 603                         PMD_DRV_LOG(ERR, "maximum packet length must be "
 604                                     "larger than %u and smaller than %u, "
 605                                     "as jumbo frame is disabled",
 606                                     (uint32_t)RTE_ETHER_MIN_LEN,
 607                                     (uint32_t)IAVF_ETH_MAX_LEN);
 608                         return -EINVAL;
 609                 }
 610         }
 611
 612         rxq->max_pkt_len = max_pkt_len;
 613         if ((dev_data->dev_conf.rxmode.offloads & DEV_RX_OFFLOAD_SCATTER) ||
 614             rxq->max_pkt_len > buf_size) {
 615                 dev_data->scattered_rx = 1;
 616         }
 617         IAVF_PCI_REG_WRITE(rxq->qrx_tail, rxq->nb_rx_desc - 1);
 618         IAVF_WRITE_FLUSH(hw);
 619
 620         return 0;
 621 }
 622
 623 static int
 624 iavf_init_queues(struct rte_eth_dev *dev)
 625 {
 626         struct iavf_rx_queue **rxq =
 627                 (struct iavf_rx_queue **)dev->data->rx_queues;
 628         int i, ret = IAVF_SUCCESS;
 629
 630         for (i = 0; i < dev->data->nb_rx_queues; i++) {
 631                 if (!rxq[i] || !rxq[i]->q_set)
 632                         continue;
 633                 ret = iavf_init_rxq(dev, rxq[i]);
 634                 if (ret != IAVF_SUCCESS)
 635                         break;
 636         }
 637         /* set rx/tx function to vector/scatter/single-segment
 638          * according to parameters
 639          */
 640         iavf_set_rx_function(dev);
 641         iavf_set_tx_function(dev);
 642
 643         return ret;
 644 }
 645
 646 static int iavf_config_rx_queues_irqs(struct rte_eth_dev *dev,
 647                                      struct rte_intr_handle *intr_handle)
 648 {
 649         struct iavf_adapter *adapter =
 650                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 651         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
 652         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(adapter);
 653         struct iavf_qv_map *qv_map;
 654         uint16_t interval, i;
 655         int vec;
 656
 657         if (rte_intr_cap_multiple(intr_handle) &&
 658             dev->data->dev_conf.intr_conf.rxq) {
 659                 if (rte_intr_efd_enable(intr_handle, dev->data->nb_rx_queues))
 660                         return -1;
 661         }
 662
 663         if (rte_intr_dp_is_en(intr_handle) && !intr_handle->intr_vec) {
 664                 intr_handle->intr_vec =
 665                         rte_zmalloc("intr_vec",
 666                                     dev->data->nb_rx_queues * sizeof(int), 0);
 667                 if (!intr_handle->intr_vec) {
 668                         PMD_DRV_LOG(ERR, "Failed to allocate %d rx intr_vec",
 669                                     dev->data->nb_rx_queues);
 670                         return -1;
 671                 }
 672         }
 673
 674         qv_map = rte_zmalloc("qv_map",
 675                 dev->data->nb_rx_queues * sizeof(struct iavf_qv_map), 0);
 676         if (!qv_map) {
 677                 PMD_DRV_LOG(ERR, "Failed to allocate %d queue-vector map",
 678                                 dev->data->nb_rx_queues);
 679                 return -1;
 680         }
 681
 682         if (!dev->data->dev_conf.intr_conf.rxq ||
 683             !rte_intr_dp_is_en(intr_handle)) {
 684                 /* Rx interrupt disabled, Map interrupt only for writeback */
 685                 vf->nb_msix = 1;
 686                 if (vf->vf_res->vf_cap_flags &
 687                     VIRTCHNL_VF_OFFLOAD_WB_ON_ITR) {
 688                         /* If WB_ON_ITR supports, enable it */
 689                         vf->msix_base = IAVF_RX_VEC_START;
 690                         /* Set the ITR for index zero, to 2us to make sure that
 691                          * we leave time for aggregation to occur, but don't
 692                          * increase latency dramatically.
 693                          */
 694                         IAVF_WRITE_REG(hw,
 695                                        IAVF_VFINT_DYN_CTLN1(vf->msix_base - 1),
 696                                        (0 << IAVF_VFINT_DYN_CTLN1_ITR_INDX_SHIFT) |
 697                                        IAVF_VFINT_DYN_CTLN1_WB_ON_ITR_MASK |
 698                                        (2UL << IAVF_VFINT_DYN_CTLN1_INTERVAL_SHIFT));
 699                         /* debug - check for success! the return value
 700                          * should be 2, offset is 0x2800
 701                          */
 702                         /* IAVF_READ_REG(hw, IAVF_VFINT_ITRN1(0, 0)); */
 703                 } else {
 704                         /* If no WB_ON_ITR offload flags, need to set
 705                          * interrupt for descriptor write back.
 706                          */
 707                         vf->msix_base = IAVF_MISC_VEC_ID;
 708
 709                         /* set ITR to default */
 710                         interval = iavf_calc_itr_interval(
 711                                         IAVF_QUEUE_ITR_INTERVAL_DEFAULT);
 712                         IAVF_WRITE_REG(hw, IAVF_VFINT_DYN_CTL01,
 713                                        IAVF_VFINT_DYN_CTL01_INTENA_MASK |
 714                                        (IAVF_ITR_INDEX_DEFAULT <<
 715                                         IAVF_VFINT_DYN_CTL01_ITR_INDX_SHIFT) |
 716                                        (interval <<
 717                                         IAVF_VFINT_DYN_CTL01_INTERVAL_SHIFT));
 718                 }
 719                 IAVF_WRITE_FLUSH(hw);
 720                 /* map all queues to the same interrupt */
 721                 for (i = 0; i < dev->data->nb_rx_queues; i++) {
 722                         qv_map[i].queue_id = i;
 723                         qv_map[i].vector_id = vf->msix_base;
 724                 }
 725                 vf->qv_map = qv_map;
 726         } else {
 727                 if (!rte_intr_allow_others(intr_handle)) {
 728                         vf->nb_msix = 1;
 729                         vf->msix_base = IAVF_MISC_VEC_ID;
 730                         for (i = 0; i < dev->data->nb_rx_queues; i++) {
 731                                 qv_map[i].queue_id = i;
 732                                 qv_map[i].vector_id = vf->msix_base;
 733                                 intr_handle->intr_vec[i] = IAVF_MISC_VEC_ID;
 734                         }
 735                         vf->qv_map = qv_map;
 736                         PMD_DRV_LOG(DEBUG,
 737                                     "vector %u are mapping to all Rx queues",
 738                                     vf->msix_base);
 739                 } else {
 740                         /* If Rx interrupt is reuquired, and we can use
 741                          * multi interrupts, then the vec is from 1
 742                          */
 743                         vf->nb_msix = RTE_MIN(intr_handle->nb_efd,
 744                                  (uint16_t)(vf->vf_res->max_vectors - 1));
 745                         vf->msix_base = IAVF_RX_VEC_START;
 746                         vec = IAVF_RX_VEC_START;
 747                         for (i = 0; i < dev->data->nb_rx_queues; i++) {
 748                                 qv_map[i].queue_id = i;
 749                                 qv_map[i].vector_id = vec;
 750                                 intr_handle->intr_vec[i] = vec++;
 751                                 if (vec >= vf->nb_msix + IAVF_RX_VEC_START)
 752                                         vec = IAVF_RX_VEC_START;
 753                         }
 754                         vf->qv_map = qv_map;
 755                         PMD_DRV_LOG(DEBUG,
 756                                     "%u vectors are mapping to %u Rx queues",
 757                                     vf->nb_msix, dev->data->nb_rx_queues);
 758                 }
 759         }
 760
 761         if (!vf->lv_enabled) {
 762                 if (iavf_config_irq_map(adapter)) {
 763                         PMD_DRV_LOG(ERR, "config interrupt mapping failed");
 764                         return -1;
 765                 }
 766         } else {
 767                 uint16_t num_qv_maps = dev->data->nb_rx_queues;
 768                 uint16_t index = 0;
 769
 770                 while (num_qv_maps > IAVF_IRQ_MAP_NUM_PER_BUF) {
 771                         if (iavf_config_irq_map_lv(adapter,
 772                                         IAVF_IRQ_MAP_NUM_PER_BUF, index)) {
 773                                 PMD_DRV_LOG(ERR, "config interrupt mapping for large VF failed");
 774                                 return -1;
 775                         }
 776                         num_qv_maps -= IAVF_IRQ_MAP_NUM_PER_BUF;
 777                         index += IAVF_IRQ_MAP_NUM_PER_BUF;
 778                 }
 779
 780                 if (iavf_config_irq_map_lv(adapter, num_qv_maps, index)) {
 781                         PMD_DRV_LOG(ERR, "config interrupt mapping for large VF failed");
 782                         return -1;
 783                 }
 784         }
 785         return 0;
 786 }
 787
 788 static int
 789 iavf_start_queues(struct rte_eth_dev *dev)
 790 {
 791         struct iavf_rx_queue *rxq;
 792         struct iavf_tx_queue *txq;
 793         int i;
 794
 795         for (i = 0; i < dev->data->nb_tx_queues; i++) {
 796                 txq = dev->data->tx_queues[i];
 797                 if (txq->tx_deferred_start)
 798                         continue;
 799                 if (iavf_dev_tx_queue_start(dev, i) != 0) {
 800                         PMD_DRV_LOG(ERR, "Fail to start queue %u", i);
 801                         return -1;
 802                 }
 803         }
 804
 805         for (i = 0; i < dev->data->nb_rx_queues; i++) {
 806                 rxq = dev->data->rx_queues[i];
 807                 if (rxq->rx_deferred_start)
 808                         continue;
 809                 if (iavf_dev_rx_queue_start(dev, i) != 0) {
 810                         PMD_DRV_LOG(ERR, "Fail to start queue %u", i);
 811                         return -1;
 812                 }
 813         }
 814
 815         return 0;
 816 }
 817
 818 static int
 819 iavf_dev_start(struct rte_eth_dev *dev)
 820 {
 821         struct iavf_adapter *adapter =
 822                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 823         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
 824         struct rte_intr_handle *intr_handle = dev->intr_handle;
 825         uint16_t num_queue_pairs;
 826         uint16_t index = 0;
 827
 828         PMD_INIT_FUNC_TRACE();
 829
 830         adapter->stopped = 0;
 831
 832         vf->max_pkt_len = dev->data->dev_conf.rxmode.max_rx_pkt_len;
 833         vf->num_queue_pairs = RTE_MAX(dev->data->nb_rx_queues,
 834                                       dev->data->nb_tx_queues);
 835         num_queue_pairs = vf->num_queue_pairs;
 836
 837         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_QOS)
 838                 if (iavf_get_qos_cap(adapter)) {
 839                         PMD_INIT_LOG(ERR, "Failed to get qos capability");
 840                         return -1;
 841                 }
 842
 843         if (iavf_init_queues(dev) != 0) {
 844                 PMD_DRV_LOG(ERR, "failed to do Queue init");
 845                 return -1;
 846         }
 847
 848         /* If needed, send configure queues msg multiple times to make the
 849          * adminq buffer length smaller than the 4K limitation.
 850          */
 851         while (num_queue_pairs > IAVF_CFG_Q_NUM_PER_BUF) {
 852                 if (iavf_configure_queues(adapter,
 853                                 IAVF_CFG_Q_NUM_PER_BUF, index) != 0) {
 854                         PMD_DRV_LOG(ERR, "configure queues failed");
 855                         goto err_queue;
 856                 }
 857                 num_queue_pairs -= IAVF_CFG_Q_NUM_PER_BUF;
 858                 index += IAVF_CFG_Q_NUM_PER_BUF;
 859         }
 860
 861         if (iavf_configure_queues(adapter, num_queue_pairs, index) != 0) {
 862                 PMD_DRV_LOG(ERR, "configure queues failed");
 863                 goto err_queue;
 864         }
 865
 866         if (iavf_config_rx_queues_irqs(dev, intr_handle) != 0) {
 867                 PMD_DRV_LOG(ERR, "configure irq failed");
 868                 goto err_queue;
 869         }
 870         /* re-enable intr again, because efd assign may change */
 871         if (dev->data->dev_conf.intr_conf.rxq != 0) {
 872                 if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_WB_ON_ITR)
 873                         rte_intr_disable(intr_handle);
 874                 rte_intr_enable(intr_handle);
 875         }
 876
 877         /* Set all mac addrs */
 878         iavf_add_del_all_mac_addr(adapter, true);
 879
 880         /* Set all multicast addresses */
 881         iavf_add_del_mc_addr_list(adapter, vf->mc_addrs, vf->mc_addrs_num,
 882                                   true);
 883
 884         if (iavf_start_queues(dev) != 0) {
 885                 PMD_DRV_LOG(ERR, "enable queues failed");
 886                 goto err_mac;
 887         }
 888
 889         return 0;
 890
 891 err_mac:
 892         iavf_add_del_all_mac_addr(adapter, false);
 893 err_queue:
 894         return -1;
 895 }
 896
 897 static int
 898 iavf_dev_stop(struct rte_eth_dev *dev)
 899 {
 900         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
 901         struct iavf_adapter *adapter =
 902                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
 903         struct rte_intr_handle *intr_handle = dev->intr_handle;
 904
 905         PMD_INIT_FUNC_TRACE();
 906
 907         if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_WB_ON_ITR) &&
 908             dev->data->dev_conf.intr_conf.rxq != 0)
 909                 rte_intr_disable(intr_handle);
 910
 911         if (adapter->stopped == 1)
 912                 return 0;
 913
 914         iavf_stop_queues(dev);
 915
 916         /* Disable the interrupt for Rx */
 917         rte_intr_efd_disable(intr_handle);
 918         /* Rx interrupt vector mapping free */
 919         if (intr_handle->intr_vec) {
 920                 rte_free(intr_handle->intr_vec);
 921                 intr_handle->intr_vec = NULL;
 922         }
 923
 924         /* remove all mac addrs */
 925         iavf_add_del_all_mac_addr(adapter, false);
 926
 927         /* remove all multicast addresses */
 928         iavf_add_del_mc_addr_list(adapter, vf->mc_addrs, vf->mc_addrs_num,
 929                                   false);
 930
 931         adapter->stopped = 1;
 932         dev->data->dev_started = 0;
 933
 934         return 0;
 935 }
 936
 937 static int
 938 iavf_dev_info_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
 939 {
 940         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
 941
 942         dev_info->max_rx_queues = IAVF_MAX_NUM_QUEUES_LV;
 943         dev_info->max_tx_queues = IAVF_MAX_NUM_QUEUES_LV;
 944         dev_info->min_rx_bufsize = IAVF_BUF_SIZE_MIN;
 945         dev_info->max_rx_pktlen = IAVF_FRAME_SIZE_MAX;
 946         dev_info->max_mtu = dev_info->max_rx_pktlen - IAVF_ETH_OVERHEAD;
 947         dev_info->min_mtu = RTE_ETHER_MIN_MTU;
 948         dev_info->hash_key_size = vf->vf_res->rss_key_size;
 949         dev_info->reta_size = vf->vf_res->rss_lut_size;
 950         dev_info->flow_type_rss_offloads = IAVF_RSS_OFFLOAD_ALL;
 951         dev_info->max_mac_addrs = IAVF_NUM_MACADDR_MAX;
 952         dev_info->rx_offload_capa =
 953                 DEV_RX_OFFLOAD_VLAN_STRIP |
 954                 DEV_RX_OFFLOAD_QINQ_STRIP |
 955                 DEV_RX_OFFLOAD_IPV4_CKSUM |
 956                 DEV_RX_OFFLOAD_UDP_CKSUM |
 957                 DEV_RX_OFFLOAD_TCP_CKSUM |
 958                 DEV_RX_OFFLOAD_OUTER_IPV4_CKSUM |
 959                 DEV_RX_OFFLOAD_SCATTER |
 960                 DEV_RX_OFFLOAD_JUMBO_FRAME |
 961                 DEV_RX_OFFLOAD_VLAN_FILTER |
 962                 DEV_RX_OFFLOAD_RSS_HASH;
 963
 964         dev_info->tx_offload_capa =
 965                 DEV_TX_OFFLOAD_VLAN_INSERT |
 966                 DEV_TX_OFFLOAD_QINQ_INSERT |
 967                 DEV_TX_OFFLOAD_IPV4_CKSUM |
 968                 DEV_TX_OFFLOAD_UDP_CKSUM |
 969                 DEV_TX_OFFLOAD_TCP_CKSUM |
 970                 DEV_TX_OFFLOAD_SCTP_CKSUM |
 971                 DEV_TX_OFFLOAD_OUTER_IPV4_CKSUM |
 972                 DEV_TX_OFFLOAD_TCP_TSO |
 973                 DEV_TX_OFFLOAD_VXLAN_TNL_TSO |
 974                 DEV_TX_OFFLOAD_GRE_TNL_TSO |
 975                 DEV_TX_OFFLOAD_IPIP_TNL_TSO |
 976                 DEV_TX_OFFLOAD_GENEVE_TNL_TSO |
 977                 DEV_TX_OFFLOAD_MULTI_SEGS |
 978                 DEV_TX_OFFLOAD_MBUF_FAST_FREE;
 979
 980         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_CRC)
 981                 dev_info->rx_offload_capa |= DEV_RX_OFFLOAD_KEEP_CRC;
 982
 983         dev_info->default_rxconf = (struct rte_eth_rxconf) {
 984                 .rx_free_thresh = IAVF_DEFAULT_RX_FREE_THRESH,
 985                 .rx_drop_en = 0,
 986                 .offloads = 0,
 987         };
 988
 989         dev_info->default_txconf = (struct rte_eth_txconf) {
 990                 .tx_free_thresh = IAVF_DEFAULT_TX_FREE_THRESH,
 991                 .tx_rs_thresh = IAVF_DEFAULT_TX_RS_THRESH,
 992                 .offloads = 0,
 993         };
 994
 995         dev_info->rx_desc_lim = (struct rte_eth_desc_lim) {
 996                 .nb_max = IAVF_MAX_RING_DESC,
 997                 .nb_min = IAVF_MIN_RING_DESC,
 998                 .nb_align = IAVF_ALIGN_RING_DESC,
 999         };
1000
1001         dev_info->tx_desc_lim = (struct rte_eth_desc_lim) {
1002                 .nb_max = IAVF_MAX_RING_DESC,
1003                 .nb_min = IAVF_MIN_RING_DESC,
1004                 .nb_align = IAVF_ALIGN_RING_DESC,
1005         };
1006
1007         return 0;
1008 }
1009
1010 static const uint32_t *
1011 iavf_dev_supported_ptypes_get(struct rte_eth_dev *dev __rte_unused)
1012 {
1013         static const uint32_t ptypes[] = {
1014                 RTE_PTYPE_L2_ETHER,
1015                 RTE_PTYPE_L3_IPV4_EXT_UNKNOWN,
1016                 RTE_PTYPE_L4_FRAG,
1017                 RTE_PTYPE_L4_ICMP,
1018                 RTE_PTYPE_L4_NONFRAG,
1019                 RTE_PTYPE_L4_SCTP,
1020                 RTE_PTYPE_L4_TCP,
1021                 RTE_PTYPE_L4_UDP,
1022                 RTE_PTYPE_UNKNOWN
1023         };
1024         return ptypes;
1025 }
1026
1027 int
1028 iavf_dev_link_update(struct rte_eth_dev *dev,
1029                     __rte_unused int wait_to_complete)
1030 {
1031         struct rte_eth_link new_link;
1032         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
1033
1034         memset(&new_link, 0, sizeof(new_link));
1035
1036         /* Only read status info stored in VF, and the info is updated
1037          *  when receive LINK_CHANGE evnet from PF by Virtchnnl.
1038          */
1039         switch (vf->link_speed) {
1040         case 10:
1041                 new_link.link_speed = ETH_SPEED_NUM_10M;
1042                 break;
1043         case 100:
1044                 new_link.link_speed = ETH_SPEED_NUM_100M;
1045                 break;
1046         case 1000:
1047                 new_link.link_speed = ETH_SPEED_NUM_1G;
1048                 break;
1049         case 10000:
1050                 new_link.link_speed = ETH_SPEED_NUM_10G;
1051                 break;
1052         case 20000:
1053                 new_link.link_speed = ETH_SPEED_NUM_20G;
1054                 break;
1055         case 25000:
1056                 new_link.link_speed = ETH_SPEED_NUM_25G;
1057                 break;
1058         case 40000:
1059                 new_link.link_speed = ETH_SPEED_NUM_40G;
1060                 break;
1061         case 50000:
1062                 new_link.link_speed = ETH_SPEED_NUM_50G;
1063                 break;
1064         case 100000:
1065                 new_link.link_speed = ETH_SPEED_NUM_100G;
1066                 break;
1067         default:
1068                 new_link.link_speed = ETH_SPEED_NUM_NONE;
1069                 break;
1070         }
1071
1072         new_link.link_duplex = ETH_LINK_FULL_DUPLEX;
1073         new_link.link_status = vf->link_up ? ETH_LINK_UP :
1074                                              ETH_LINK_DOWN;
1075         new_link.link_autoneg = !(dev->data->dev_conf.link_speeds &
1076                                 ETH_LINK_SPEED_FIXED);
1077
1078         return rte_eth_linkstatus_set(dev, &new_link);
1079 }
1080
1081 static int
1082 iavf_dev_promiscuous_enable(struct rte_eth_dev *dev)
1083 {
1084         struct iavf_adapter *adapter =
1085                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1086         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
1087
1088         return iavf_config_promisc(adapter,
1089                                   true, vf->promisc_multicast_enabled);
1090 }
1091
1092 static int
1093 iavf_dev_promiscuous_disable(struct rte_eth_dev *dev)
1094 {
1095         struct iavf_adapter *adapter =
1096                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1097         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
1098
1099         return iavf_config_promisc(adapter,
1100                                   false, vf->promisc_multicast_enabled);
1101 }
1102
1103 static int
1104 iavf_dev_allmulticast_enable(struct rte_eth_dev *dev)
1105 {
1106         struct iavf_adapter *adapter =
1107                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1108         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
1109
1110         return iavf_config_promisc(adapter,
1111                                   vf->promisc_unicast_enabled, true);
1112 }
1113
1114 static int
1115 iavf_dev_allmulticast_disable(struct rte_eth_dev *dev)
1116 {
1117         struct iavf_adapter *adapter =
1118                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1119         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
1120
1121         return iavf_config_promisc(adapter,
1122                                   vf->promisc_unicast_enabled, false);
1123 }
1124
1125 static int
1126 iavf_dev_add_mac_addr(struct rte_eth_dev *dev, struct rte_ether_addr *addr,
1127                      __rte_unused uint32_t index,
1128                      __rte_unused uint32_t pool)
1129 {
1130         struct iavf_adapter *adapter =
1131                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1132         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
1133         int err;
1134
1135         if (rte_is_zero_ether_addr(addr)) {
1136                 PMD_DRV_LOG(ERR, "Invalid Ethernet Address");
1137                 return -EINVAL;
1138         }
1139
1140         err = iavf_add_del_eth_addr(adapter, addr, true, VIRTCHNL_ETHER_ADDR_EXTRA);
1141         if (err) {
1142                 PMD_DRV_LOG(ERR, "fail to add MAC address");
1143                 return -EIO;
1144         }
1145
1146         vf->mac_num++;
1147
1148         return 0;
1149 }
1150
1151 static void
1152 iavf_dev_del_mac_addr(struct rte_eth_dev *dev, uint32_t index)
1153 {
1154         struct iavf_adapter *adapter =
1155                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1156         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
1157         struct rte_ether_addr *addr;
1158         int err;
1159
1160         addr = &dev->data->mac_addrs[index];
1161
1162         err = iavf_add_del_eth_addr(adapter, addr, false, VIRTCHNL_ETHER_ADDR_EXTRA);
1163         if (err)
1164                 PMD_DRV_LOG(ERR, "fail to delete MAC address");
1165
1166         vf->mac_num--;
1167 }
1168
1169 static int
1170 iavf_dev_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
1171 {
1172         struct iavf_adapter *adapter =
1173                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1174         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
1175         int err;
1176
1177         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_VLAN_V2) {
1178                 err = iavf_add_del_vlan_v2(adapter, vlan_id, on);
1179                 if (err)
1180                         return -EIO;
1181                 return 0;
1182         }
1183
1184         if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_VLAN))
1185                 return -ENOTSUP;
1186
1187         err = iavf_add_del_vlan(adapter, vlan_id, on);
1188         if (err)
1189                 return -EIO;
1190         return 0;
1191 }
1192
1193 static void
1194 iavf_iterate_vlan_filters_v2(struct rte_eth_dev *dev, bool enable)
1195 {
1196         struct rte_vlan_filter_conf *vfc = &dev->data->vlan_filter_conf;
1197         struct iavf_adapter *adapter =
1198                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1199         uint32_t i, j;
1200         uint64_t ids;
1201
1202         for (i = 0; i < RTE_DIM(vfc->ids); i++) {
1203                 if (vfc->ids[i] == 0)
1204                         continue;
1205
1206                 ids = vfc->ids[i];
1207                 for (j = 0; ids != 0 && j < 64; j++, ids >>= 1) {
1208                         if (ids & 1)
1209                                 iavf_add_del_vlan_v2(adapter,
1210                                                      64 * i + j, enable);
1211                 }
1212         }
1213 }
1214
1215 static int
1216 iavf_dev_vlan_offload_set_v2(struct rte_eth_dev *dev, int mask)
1217 {
1218         struct rte_eth_rxmode *rxmode = &dev->data->dev_conf.rxmode;
1219         struct iavf_adapter *adapter =
1220                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1221         bool enable;
1222         int err;
1223
1224         if (mask & ETH_VLAN_FILTER_MASK) {
1225                 enable = !!(rxmode->offloads & DEV_RX_OFFLOAD_VLAN_FILTER);
1226
1227                 iavf_iterate_vlan_filters_v2(dev, enable);
1228         }
1229
1230         if (mask & ETH_VLAN_STRIP_MASK) {
1231                 enable = !!(rxmode->offloads & DEV_RX_OFFLOAD_VLAN_STRIP);
1232
1233                 err = iavf_config_vlan_strip_v2(adapter, enable);
1234                 /* If not support, the stripping is already disabled by PF */
1235                 if (err == -ENOTSUP && !enable)
1236                         err = 0;
1237                 if (err)
1238                         return -EIO;
1239         }
1240
1241         return 0;
1242 }
1243
1244 static int
1245 iavf_dev_vlan_offload_set(struct rte_eth_dev *dev, int mask)
1246 {
1247         struct iavf_adapter *adapter =
1248                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1249         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
1250         struct rte_eth_conf *dev_conf = &dev->data->dev_conf;
1251         int err;
1252
1253         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_VLAN_V2)
1254                 return iavf_dev_vlan_offload_set_v2(dev, mask);
1255
1256         if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_VLAN))
1257                 return -ENOTSUP;
1258
1259         /* Vlan stripping setting */
1260         if (mask & ETH_VLAN_STRIP_MASK) {
1261                 /* Enable or disable VLAN stripping */
1262                 if (dev_conf->rxmode.offloads & DEV_RX_OFFLOAD_VLAN_STRIP)
1263                         err = iavf_enable_vlan_strip(adapter);
1264                 else
1265                         err = iavf_disable_vlan_strip(adapter);
1266
1267                 if (err)
1268                         return -EIO;
1269         }
1270         return 0;
1271 }
1272
1273 static int
1274 iavf_dev_rss_reta_update(struct rte_eth_dev *dev,
1275                         struct rte_eth_rss_reta_entry64 *reta_conf,
1276                         uint16_t reta_size)
1277 {
1278         struct iavf_adapter *adapter =
1279                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1280         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
1281         uint8_t *lut;
1282         uint16_t i, idx, shift;
1283         int ret;
1284
1285         if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF))
1286                 return -ENOTSUP;
1287
1288         if (reta_size != vf->vf_res->rss_lut_size) {
1289                 PMD_DRV_LOG(ERR, "The size of hash lookup table configured "
1290                         "(%d) doesn't match the number of hardware can "
1291                         "support (%d)", reta_size, vf->vf_res->rss_lut_size);
1292                 return -EINVAL;
1293         }
1294
1295         lut = rte_zmalloc("rss_lut", reta_size, 0);
1296         if (!lut) {
1297                 PMD_DRV_LOG(ERR, "No memory can be allocated");
1298                 return -ENOMEM;
1299         }
1300         /* store the old lut table temporarily */
1301         rte_memcpy(lut, vf->rss_lut, reta_size);
1302
1303         for (i = 0; i < reta_size; i++) {
1304                 idx = i / RTE_RETA_GROUP_SIZE;
1305                 shift = i % RTE_RETA_GROUP_SIZE;
1306                 if (reta_conf[idx].mask & (1ULL << shift))
1307                         lut[i] = reta_conf[idx].reta[shift];
1308         }
1309
1310         rte_memcpy(vf->rss_lut, lut, reta_size);
1311         /* send virtchnnl ops to configure rss*/
1312         ret = iavf_configure_rss_lut(adapter);
1313         if (ret) /* revert back */
1314                 rte_memcpy(vf->rss_lut, lut, reta_size);
1315         rte_free(lut);
1316
1317         return ret;
1318 }
1319
1320 static int
1321 iavf_dev_rss_reta_query(struct rte_eth_dev *dev,
1322                        struct rte_eth_rss_reta_entry64 *reta_conf,
1323                        uint16_t reta_size)
1324 {
1325         struct iavf_adapter *adapter =
1326                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1327         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
1328         uint16_t i, idx, shift;
1329
1330         if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF))
1331                 return -ENOTSUP;
1332
1333         if (reta_size != vf->vf_res->rss_lut_size) {
1334                 PMD_DRV_LOG(ERR, "The size of hash lookup table configured "
1335                         "(%d) doesn't match the number of hardware can "
1336                         "support (%d)", reta_size, vf->vf_res->rss_lut_size);
1337                 return -EINVAL;
1338         }
1339
1340         for (i = 0; i < reta_size; i++) {
1341                 idx = i / RTE_RETA_GROUP_SIZE;
1342                 shift = i % RTE_RETA_GROUP_SIZE;
1343                 if (reta_conf[idx].mask & (1ULL << shift))
1344                         reta_conf[idx].reta[shift] = vf->rss_lut[i];
1345         }
1346
1347         return 0;
1348 }
1349
1350 static int
1351 iavf_set_rss_key(struct iavf_adapter *adapter, uint8_t *key, uint8_t key_len)
1352 {
1353         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
1354
1355         /* HENA setting, it is enabled by default, no change */
1356         if (!key || key_len == 0) {
1357                 PMD_DRV_LOG(DEBUG, "No key to be configured");
1358                 return 0;
1359         } else if (key_len != vf->vf_res->rss_key_size) {
1360                 PMD_DRV_LOG(ERR, "The size of hash key configured "
1361                         "(%d) doesn't match the size of hardware can "
1362                         "support (%d)", key_len,
1363                         vf->vf_res->rss_key_size);
1364                 return -EINVAL;
1365         }
1366
1367         rte_memcpy(vf->rss_key, key, key_len);
1368
1369         return iavf_configure_rss_key(adapter);
1370 }
1371
1372 static int
1373 iavf_dev_rss_hash_update(struct rte_eth_dev *dev,
1374                         struct rte_eth_rss_conf *rss_conf)
1375 {
1376         struct iavf_adapter *adapter =
1377                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1378         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
1379         int ret;
1380
1381         adapter->eth_dev->data->dev_conf.rx_adv_conf.rss_conf = *rss_conf;
1382
1383         if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF))
1384                 return -ENOTSUP;
1385
1386         /* Set hash key. */
1387         ret = iavf_set_rss_key(adapter, rss_conf->rss_key,
1388                                rss_conf->rss_key_len);
1389         if (ret)
1390                 return ret;
1391
1392         if (rss_conf->rss_hf == 0) {
1393                 vf->rss_hf = 0;
1394                 ret = iavf_set_hena(adapter, 0);
1395
1396                 /* It is a workaround, temporarily allow error to be returned
1397                  * due to possible lack of PF handling for hena = 0.
1398                  */
1399                 if (ret)
1400                         PMD_DRV_LOG(WARNING, "fail to clean existing RSS, lack PF support");
1401                 return 0;
1402         }
1403
1404         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_ADV_RSS_PF) {
1405                 /* Clear existing RSS. */
1406                 ret = iavf_set_hena(adapter, 0);
1407
1408                 /* It is a workaround, temporarily allow error to be returned
1409                  * due to possible lack of PF handling for hena = 0.
1410                  */
1411                 if (ret)
1412                         PMD_DRV_LOG(WARNING, "fail to clean existing RSS,"
1413                                     "lack PF support");
1414
1415                 /* Set new RSS configuration. */
1416                 ret = iavf_rss_hash_set(adapter, rss_conf->rss_hf, true);
1417                 if (ret) {
1418                         PMD_DRV_LOG(ERR, "fail to set new RSS");
1419                         return ret;
1420                 }
1421         } else {
1422                 iavf_config_rss_hf(adapter, rss_conf->rss_hf);
1423         }
1424
1425         return 0;
1426 }
1427
1428 static int
1429 iavf_dev_rss_hash_conf_get(struct rte_eth_dev *dev,
1430                           struct rte_eth_rss_conf *rss_conf)
1431 {
1432         struct iavf_adapter *adapter =
1433                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1434         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
1435
1436         if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF))
1437                 return -ENOTSUP;
1438
1439         rss_conf->rss_hf = vf->rss_hf;
1440
1441         if (!rss_conf->rss_key)
1442                 return 0;
1443
1444         rss_conf->rss_key_len = vf->vf_res->rss_key_size;
1445         rte_memcpy(rss_conf->rss_key, vf->rss_key, rss_conf->rss_key_len);
1446
1447         return 0;
1448 }
1449
1450 static int
1451 iavf_dev_mtu_set(struct rte_eth_dev *dev, uint16_t mtu)
1452 {
1453         uint32_t frame_size = mtu + IAVF_ETH_OVERHEAD;
1454         int ret = 0;
1455
1456         if (mtu < RTE_ETHER_MIN_MTU || frame_size > IAVF_FRAME_SIZE_MAX)
1457                 return -EINVAL;
1458
1459         /* mtu setting is forbidden if port is start */
1460         if (dev->data->dev_started) {
1461                 PMD_DRV_LOG(ERR, "port must be stopped before configuration");
1462                 return -EBUSY;
1463         }
1464
1465         if (frame_size > IAVF_ETH_MAX_LEN)
1466                 dev->data->dev_conf.rxmode.offloads |=
1467                                 DEV_RX_OFFLOAD_JUMBO_FRAME;
1468         else
1469                 dev->data->dev_conf.rxmode.offloads &=
1470                                 ~DEV_RX_OFFLOAD_JUMBO_FRAME;
1471
1472         dev->data->dev_conf.rxmode.max_rx_pkt_len = frame_size;
1473
1474         return ret;
1475 }
1476
1477 static int
1478 iavf_dev_set_default_mac_addr(struct rte_eth_dev *dev,
1479                              struct rte_ether_addr *mac_addr)
1480 {
1481         struct iavf_adapter *adapter =
1482                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1483         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(adapter);
1484         struct rte_ether_addr *old_addr;
1485         int ret;
1486
1487         old_addr = (struct rte_ether_addr *)hw->mac.addr;
1488
1489         if (rte_is_same_ether_addr(old_addr, mac_addr))
1490                 return 0;
1491
1492         ret = iavf_add_del_eth_addr(adapter, old_addr, false, VIRTCHNL_ETHER_ADDR_PRIMARY);
1493         if (ret)
1494                 PMD_DRV_LOG(ERR, "Fail to delete old MAC:"
1495                             RTE_ETHER_ADDR_PRT_FMT,
1496                                 RTE_ETHER_ADDR_BYTES(old_addr));
1497
1498         ret = iavf_add_del_eth_addr(adapter, mac_addr, true, VIRTCHNL_ETHER_ADDR_PRIMARY);
1499         if (ret)
1500                 PMD_DRV_LOG(ERR, "Fail to add new MAC:"
1501                             RTE_ETHER_ADDR_PRT_FMT,
1502                                 RTE_ETHER_ADDR_BYTES(mac_addr));
1503
1504         if (ret)
1505                 return -EIO;
1506
1507         rte_ether_addr_copy(mac_addr, (struct rte_ether_addr *)hw->mac.addr);
1508         return 0;
1509 }
1510
1511 static void
1512 iavf_stat_update_48(uint64_t *offset, uint64_t *stat)
1513 {
1514         if (*stat >= *offset)
1515                 *stat = *stat - *offset;
1516         else
1517                 *stat = (uint64_t)((*stat +
1518                         ((uint64_t)1 << IAVF_48_BIT_WIDTH)) - *offset);
1519
1520         *stat &= IAVF_48_BIT_MASK;
1521 }
1522
1523 static void
1524 iavf_stat_update_32(uint64_t *offset, uint64_t *stat)
1525 {
1526         if (*stat >= *offset)
1527                 *stat = (uint64_t)(*stat - *offset);
1528         else
1529                 *stat = (uint64_t)((*stat +
1530                         ((uint64_t)1 << IAVF_32_BIT_WIDTH)) - *offset);
1531 }
1532
1533 static void
1534 iavf_update_stats(struct iavf_vsi *vsi, struct virtchnl_eth_stats *nes)
1535 {
1536         struct virtchnl_eth_stats *oes = &vsi->eth_stats_offset;
1537
1538         iavf_stat_update_48(&oes->rx_bytes, &nes->rx_bytes);
1539         iavf_stat_update_48(&oes->rx_unicast, &nes->rx_unicast);
1540         iavf_stat_update_48(&oes->rx_multicast, &nes->rx_multicast);
1541         iavf_stat_update_48(&oes->rx_broadcast, &nes->rx_broadcast);
1542         iavf_stat_update_32(&oes->rx_discards, &nes->rx_discards);
1543         iavf_stat_update_48(&oes->tx_bytes, &nes->tx_bytes);
1544         iavf_stat_update_48(&oes->tx_unicast, &nes->tx_unicast);
1545         iavf_stat_update_48(&oes->tx_multicast, &nes->tx_multicast);
1546         iavf_stat_update_48(&oes->tx_broadcast, &nes->tx_broadcast);
1547         iavf_stat_update_32(&oes->tx_errors, &nes->tx_errors);
1548         iavf_stat_update_32(&oes->tx_discards, &nes->tx_discards);
1549 }
1550
1551 static int
1552 iavf_dev_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
1553 {
1554         struct iavf_adapter *adapter =
1555                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1556         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
1557         struct iavf_vsi *vsi = &vf->vsi;
1558         struct virtchnl_eth_stats *pstats = NULL;
1559         int ret;
1560
1561         ret = iavf_query_stats(adapter, &pstats);
1562         if (ret == 0) {
1563                 uint8_t crc_stats_len = (dev->data->dev_conf.rxmode.offloads &
1564                                          DEV_RX_OFFLOAD_KEEP_CRC) ? 0 :
1565                                          RTE_ETHER_CRC_LEN;
1566                 iavf_update_stats(vsi, pstats);
1567                 stats->ipackets = pstats->rx_unicast + pstats->rx_multicast +
1568                                 pstats->rx_broadcast - pstats->rx_discards;
1569                 stats->opackets = pstats->tx_broadcast + pstats->tx_multicast +
1570                                                 pstats->tx_unicast;
1571                 stats->imissed = pstats->rx_discards;
1572                 stats->oerrors = pstats->tx_errors + pstats->tx_discards;
1573                 stats->ibytes = pstats->rx_bytes;
1574                 stats->ibytes -= stats->ipackets * crc_stats_len;
1575                 stats->obytes = pstats->tx_bytes;
1576         } else {
1577                 PMD_DRV_LOG(ERR, "Get statistics failed");
1578         }
1579         return ret;
1580 }
1581
1582 static int
1583 iavf_dev_stats_reset(struct rte_eth_dev *dev)
1584 {
1585         int ret;
1586         struct iavf_adapter *adapter =
1587                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1588         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
1589         struct iavf_vsi *vsi = &vf->vsi;
1590         struct virtchnl_eth_stats *pstats = NULL;
1591
1592         /* read stat values to clear hardware registers */
1593         ret = iavf_query_stats(adapter, &pstats);
1594         if (ret != 0)
1595                 return ret;
1596
1597         /* set stats offset base on current values */
1598         vsi->eth_stats_offset = *pstats;
1599
1600         return 0;
1601 }
1602
1603 static int iavf_dev_xstats_get_names(__rte_unused struct rte_eth_dev *dev,
1604                                       struct rte_eth_xstat_name *xstats_names,
1605                                       __rte_unused unsigned int limit)
1606 {
1607         unsigned int i;
1608
1609         if (xstats_names != NULL)
1610                 for (i = 0; i < IAVF_NB_XSTATS; i++) {
1611                         snprintf(xstats_names[i].name,
1612                                 sizeof(xstats_names[i].name),
1613                                 "%s", rte_iavf_stats_strings[i].name);
1614                 }
1615         return IAVF_NB_XSTATS;
1616 }
1617
1618 static int iavf_dev_xstats_get(struct rte_eth_dev *dev,
1619                                  struct rte_eth_xstat *xstats, unsigned int n)
1620 {
1621         int ret;
1622         unsigned int i;
1623         struct iavf_adapter *adapter =
1624                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1625         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
1626         struct iavf_vsi *vsi = &vf->vsi;
1627         struct virtchnl_eth_stats *pstats = NULL;
1628
1629         if (n < IAVF_NB_XSTATS)
1630                 return IAVF_NB_XSTATS;
1631
1632         ret = iavf_query_stats(adapter, &pstats);
1633         if (ret != 0)
1634                 return 0;
1635
1636         if (!xstats)
1637                 return 0;
1638
1639         iavf_update_stats(vsi, pstats);
1640
1641         /* loop over xstats array and values from pstats */
1642         for (i = 0; i < IAVF_NB_XSTATS; i++) {
1643                 xstats[i].id = i;
1644                 xstats[i].value = *(uint64_t *)(((char *)pstats) +
1645                         rte_iavf_stats_strings[i].offset);
1646         }
1647
1648         return IAVF_NB_XSTATS;
1649 }
1650
1651
1652 static int
1653 iavf_dev_rx_queue_intr_enable(struct rte_eth_dev *dev, uint16_t queue_id)
1654 {
1655         struct iavf_adapter *adapter =
1656                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1657         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1658         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(adapter);
1659         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
1660         uint16_t msix_intr;
1661
1662         msix_intr = pci_dev->intr_handle.intr_vec[queue_id];
1663         if (msix_intr == IAVF_MISC_VEC_ID) {
1664                 PMD_DRV_LOG(INFO, "MISC is also enabled for control");
1665                 IAVF_WRITE_REG(hw, IAVF_VFINT_DYN_CTL01,
1666                                IAVF_VFINT_DYN_CTL01_INTENA_MASK |
1667                                IAVF_VFINT_DYN_CTL01_CLEARPBA_MASK |
1668                                IAVF_VFINT_DYN_CTL01_ITR_INDX_MASK);
1669         } else {
1670                 IAVF_WRITE_REG(hw,
1671                                IAVF_VFINT_DYN_CTLN1
1672                                 (msix_intr - IAVF_RX_VEC_START),
1673                                IAVF_VFINT_DYN_CTLN1_INTENA_MASK |
1674                                IAVF_VFINT_DYN_CTL01_CLEARPBA_MASK |
1675                                IAVF_VFINT_DYN_CTLN1_ITR_INDX_MASK);
1676         }
1677
1678         IAVF_WRITE_FLUSH(hw);
1679
1680         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_WB_ON_ITR)
1681                 rte_intr_ack(&pci_dev->intr_handle);
1682
1683         return 0;
1684 }
1685
1686 static int
1687 iavf_dev_rx_queue_intr_disable(struct rte_eth_dev *dev, uint16_t queue_id)
1688 {
1689         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1690         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1691         uint16_t msix_intr;
1692
1693         msix_intr = pci_dev->intr_handle.intr_vec[queue_id];
1694         if (msix_intr == IAVF_MISC_VEC_ID) {
1695                 PMD_DRV_LOG(ERR, "MISC is used for control, cannot disable it");
1696                 return -EIO;
1697         }
1698
1699         IAVF_WRITE_REG(hw,
1700                       IAVF_VFINT_DYN_CTLN1(msix_intr - IAVF_RX_VEC_START),
1701                       0);
1702
1703         IAVF_WRITE_FLUSH(hw);
1704         return 0;
1705 }
1706
1707 static int
1708 iavf_check_vf_reset_done(struct iavf_hw *hw)
1709 {
1710         int i, reset;
1711
1712         for (i = 0; i < IAVF_RESET_WAIT_CNT; i++) {
1713                 reset = IAVF_READ_REG(hw, IAVF_VFGEN_RSTAT) &
1714                         IAVF_VFGEN_RSTAT_VFR_STATE_MASK;
1715                 reset = reset >> IAVF_VFGEN_RSTAT_VFR_STATE_SHIFT;
1716                 if (reset == VIRTCHNL_VFR_VFACTIVE ||
1717                     reset == VIRTCHNL_VFR_COMPLETED)
1718                         break;
1719                 rte_delay_ms(20);
1720         }
1721
1722         if (i >= IAVF_RESET_WAIT_CNT)
1723                 return -1;
1724
1725         return 0;
1726 }
1727
1728 static int
1729 iavf_lookup_proto_xtr_type(const char *flex_name)
1730 {
1731         static struct {
1732                 const char *name;
1733                 enum iavf_proto_xtr_type type;
1734         } xtr_type_map[] = {
1735                 { "vlan",      IAVF_PROTO_XTR_VLAN      },
1736                 { "ipv4",      IAVF_PROTO_XTR_IPV4      },
1737                 { "ipv6",      IAVF_PROTO_XTR_IPV6      },
1738                 { "ipv6_flow", IAVF_PROTO_XTR_IPV6_FLOW },
1739                 { "tcp",       IAVF_PROTO_XTR_TCP       },
1740                 { "ip_offset", IAVF_PROTO_XTR_IP_OFFSET },
1741         };
1742         uint32_t i;
1743
1744         for (i = 0; i < RTE_DIM(xtr_type_map); i++) {
1745                 if (strcmp(flex_name, xtr_type_map[i].name) == 0)
1746                         return xtr_type_map[i].type;
1747         }
1748
1749         PMD_DRV_LOG(ERR, "wrong proto_xtr type, "
1750                     "it should be: vlan|ipv4|ipv6|ipv6_flow|tcp|ip_offset");
1751
1752         return -1;
1753 }
1754
1755 /**
1756  * Parse elem, the elem could be single number/range or '(' ')' group
1757  * 1) A single number elem, it's just a simple digit. e.g. 9
1758  * 2) A single range elem, two digits with a '-' between. e.g. 2-6
1759  * 3) A group elem, combines multiple 1) or 2) with '( )'. e.g (0,2-4,6)
1760  *    Within group elem, '-' used for a range separator;
1761  *                       ',' used for a single number.
1762  */
1763 static int
1764 iavf_parse_queue_set(const char *input, int xtr_type,
1765                      struct iavf_devargs *devargs)
1766 {
1767         const char *str = input;
1768         char *end = NULL;
1769         uint32_t min, max;
1770         uint32_t idx;
1771
1772         while (isblank(*str))
1773                 str++;
1774
1775         if (!isdigit(*str) && *str != '(')
1776                 return -1;
1777
1778         /* process single number or single range of number */
1779         if (*str != '(') {
1780                 errno = 0;
1781                 idx = strtoul(str, &end, 10);
1782                 if (errno || !end || idx >= IAVF_MAX_QUEUE_NUM)
1783                         return -1;
1784
1785                 while (isblank(*end))
1786                         end++;
1787
1788                 min = idx;
1789                 max = idx;
1790
1791                 /* process single <number>-<number> */
1792                 if (*end == '-') {
1793                         end++;
1794                         while (isblank(*end))
1795                                 end++;
1796                         if (!isdigit(*end))
1797                                 return -1;
1798
1799                         errno = 0;
1800                         idx = strtoul(end, &end, 10);
1801                         if (errno || !end || idx >= IAVF_MAX_QUEUE_NUM)
1802                                 return -1;
1803
1804                         max = idx;
1805                         while (isblank(*end))
1806                                 end++;
1807                 }
1808
1809                 if (*end != ':')
1810                         return -1;
1811
1812                 for (idx = RTE_MIN(min, max);
1813                      idx <= RTE_MAX(min, max); idx++)
1814                         devargs->proto_xtr[idx] = xtr_type;
1815
1816                 return 0;
1817         }
1818
1819         /* process set within bracket */
1820         str++;
1821         while (isblank(*str))
1822                 str++;
1823         if (*str == '\0')
1824                 return -1;
1825
1826         min = IAVF_MAX_QUEUE_NUM;
1827         do {
1828                 /* go ahead to the first digit */
1829                 while (isblank(*str))
1830                         str++;
1831                 if (!isdigit(*str))
1832                         return -1;
1833
1834                 /* get the digit value */
1835                 errno = 0;
1836                 idx = strtoul(str, &end, 10);
1837                 if (errno || !end || idx >= IAVF_MAX_QUEUE_NUM)
1838                         return -1;
1839
1840                 /* go ahead to separator '-',',' and ')' */
1841                 while (isblank(*end))
1842                         end++;
1843                 if (*end == '-') {
1844                         if (min == IAVF_MAX_QUEUE_NUM)
1845                                 min = idx;
1846                         else /* avoid continuous '-' */
1847                                 return -1;
1848                 } else if (*end == ',' || *end == ')') {
1849                         max = idx;
1850                         if (min == IAVF_MAX_QUEUE_NUM)
1851                                 min = idx;
1852
1853                         for (idx = RTE_MIN(min, max);
1854                              idx <= RTE_MAX(min, max); idx++)
1855                                 devargs->proto_xtr[idx] = xtr_type;
1856
1857                         min = IAVF_MAX_QUEUE_NUM;
1858                 } else {
1859                         return -1;
1860                 }
1861
1862                 str = end + 1;
1863         } while (*end != ')' && *end != '\0');
1864
1865         return 0;
1866 }
1867
1868 static int
1869 iavf_parse_queue_proto_xtr(const char *queues, struct iavf_devargs *devargs)
1870 {
1871         const char *queue_start;
1872         uint32_t idx;
1873         int xtr_type;
1874         char flex_name[32];
1875
1876         while (isblank(*queues))
1877                 queues++;
1878
1879         if (*queues != '[') {
1880                 xtr_type = iavf_lookup_proto_xtr_type(queues);
1881                 if (xtr_type < 0)
1882                         return -1;
1883
1884                 devargs->proto_xtr_dflt = xtr_type;
1885
1886                 return 0;
1887         }
1888
1889         queues++;
1890         do {
1891                 while (isblank(*queues))
1892                         queues++;
1893                 if (*queues == '\0')
1894                         return -1;
1895
1896                 queue_start = queues;
1897
1898                 /* go across a complete bracket */
1899                 if (*queue_start == '(') {
1900                         queues += strcspn(queues, ")");
1901                         if (*queues != ')')
1902                                 return -1;
1903                 }
1904
1905                 /* scan the separator ':' */
1906                 queues += strcspn(queues, ":");
1907                 if (*queues++ != ':')
1908                         return -1;
1909                 while (isblank(*queues))
1910                         queues++;
1911
1912                 for (idx = 0; ; idx++) {
1913                         if (isblank(queues[idx]) ||
1914                             queues[idx] == ',' ||
1915                             queues[idx] == ']' ||
1916                             queues[idx] == '\0')
1917                                 break;
1918
1919                         if (idx > sizeof(flex_name) - 2)
1920                                 return -1;
1921
1922                         flex_name[idx] = queues[idx];
1923                 }
1924                 flex_name[idx] = '\0';
1925                 xtr_type = iavf_lookup_proto_xtr_type(flex_name);
1926                 if (xtr_type < 0)
1927                         return -1;
1928
1929                 queues += idx;
1930
1931                 while (isblank(*queues) || *queues == ',' || *queues == ']')
1932                         queues++;
1933
1934                 if (iavf_parse_queue_set(queue_start, xtr_type, devargs) < 0)
1935                         return -1;
1936         } while (*queues != '\0');
1937
1938         return 0;
1939 }
1940
1941 static int
1942 iavf_handle_proto_xtr_arg(__rte_unused const char *key, const char *value,
1943                           void *extra_args)
1944 {
1945         struct iavf_devargs *devargs = extra_args;
1946
1947         if (!value || !extra_args)
1948                 return -EINVAL;
1949
1950         if (iavf_parse_queue_proto_xtr(value, devargs) < 0) {
1951                 PMD_DRV_LOG(ERR, "the proto_xtr's parameter is wrong : '%s'",
1952                             value);
1953                 return -1;
1954         }
1955
1956         return 0;
1957 }
1958
1959 static int iavf_parse_devargs(struct rte_eth_dev *dev)
1960 {
1961         struct iavf_adapter *ad =
1962                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1963         struct rte_devargs *devargs = dev->device->devargs;
1964         struct rte_kvargs *kvlist;
1965         int ret;
1966
1967         if (!devargs)
1968                 return 0;
1969
1970         kvlist = rte_kvargs_parse(devargs->args, iavf_valid_args);
1971         if (!kvlist) {
1972                 PMD_INIT_LOG(ERR, "invalid kvargs key\n");
1973                 return -EINVAL;
1974         }
1975
1976         ad->devargs.proto_xtr_dflt = IAVF_PROTO_XTR_NONE;
1977         memset(ad->devargs.proto_xtr, IAVF_PROTO_XTR_NONE,
1978                sizeof(ad->devargs.proto_xtr));
1979
1980         ret = rte_kvargs_process(kvlist, IAVF_PROTO_XTR_ARG,
1981                                  &iavf_handle_proto_xtr_arg, &ad->devargs);
1982         if (ret)
1983                 goto bail;
1984
1985 bail:
1986         rte_kvargs_free(kvlist);
1987         return ret;
1988 }
1989
1990 static void
1991 iavf_init_proto_xtr(struct rte_eth_dev *dev)
1992 {
1993         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
1994         struct iavf_adapter *ad =
1995                         IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1996         const struct iavf_proto_xtr_ol *xtr_ol;
1997         bool proto_xtr_enable = false;
1998         int offset;
1999         uint16_t i;
2000
2001         vf->proto_xtr = rte_zmalloc("vf proto xtr",
2002                                     vf->vsi_res->num_queue_pairs, 0);
2003         if (unlikely(!(vf->proto_xtr))) {
2004                 PMD_DRV_LOG(ERR, "no memory for setting up proto_xtr's table");
2005                 return;
2006         }
2007
2008         for (i = 0; i < vf->vsi_res->num_queue_pairs; i++) {
2009                 vf->proto_xtr[i] = ad->devargs.proto_xtr[i] !=
2010                                         IAVF_PROTO_XTR_NONE ?
2011                                         ad->devargs.proto_xtr[i] :
2012                                         ad->devargs.proto_xtr_dflt;
2013
2014                 if (vf->proto_xtr[i] != IAVF_PROTO_XTR_NONE) {
2015                         uint8_t type = vf->proto_xtr[i];
2016
2017                         iavf_proto_xtr_params[type].required = true;
2018                         proto_xtr_enable = true;
2019                 }
2020         }
2021
2022         if (likely(!proto_xtr_enable))
2023                 return;
2024
2025         offset = rte_mbuf_dynfield_register(&iavf_proto_xtr_metadata_param);
2026         if (unlikely(offset == -1)) {
2027                 PMD_DRV_LOG(ERR,
2028                             "failed to extract protocol metadata, error %d",
2029                             -rte_errno);
2030                 return;
2031         }
2032
2033         PMD_DRV_LOG(DEBUG,
2034                     "proto_xtr metadata offset in mbuf is : %d",
2035                     offset);
2036         rte_pmd_ifd_dynfield_proto_xtr_metadata_offs = offset;
2037
2038         for (i = 0; i < RTE_DIM(iavf_proto_xtr_params); i++) {
2039                 xtr_ol = &iavf_proto_xtr_params[i];
2040
2041                 uint8_t rxdid = iavf_proto_xtr_type_to_rxdid((uint8_t)i);
2042
2043                 if (!xtr_ol->required)
2044                         continue;
2045
2046                 if (!(vf->supported_rxdid & BIT(rxdid))) {
2047                         PMD_DRV_LOG(ERR,
2048                                     "rxdid[%u] is not supported in hardware",
2049                                     rxdid);
2050                         rte_pmd_ifd_dynfield_proto_xtr_metadata_offs = -1;
2051                         break;
2052                 }
2053
2054                 offset = rte_mbuf_dynflag_register(&xtr_ol->param);
2055                 if (unlikely(offset == -1)) {
2056                         PMD_DRV_LOG(ERR,
2057                                     "failed to register proto_xtr offload '%s', error %d",
2058                                     xtr_ol->param.name, -rte_errno);
2059
2060                         rte_pmd_ifd_dynfield_proto_xtr_metadata_offs = -1;
2061                         break;
2062                 }
2063
2064                 PMD_DRV_LOG(DEBUG,
2065                             "proto_xtr offload '%s' offset in mbuf is : %d",
2066                             xtr_ol->param.name, offset);
2067                 *xtr_ol->ol_flag = 1ULL << offset;
2068         }
2069 }
2070
2071 static int
2072 iavf_init_vf(struct rte_eth_dev *dev)
2073 {
2074         int err, bufsz;
2075         struct iavf_adapter *adapter =
2076                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
2077         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2078         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
2079
2080         err = iavf_parse_devargs(dev);
2081         if (err) {
2082                 PMD_INIT_LOG(ERR, "Failed to parse devargs");
2083                 goto err;
2084         }
2085
2086         err = iavf_set_mac_type(hw);
2087         if (err) {
2088                 PMD_INIT_LOG(ERR, "set_mac_type failed: %d", err);
2089                 goto err;
2090         }
2091
2092         err = iavf_check_vf_reset_done(hw);
2093         if (err) {
2094                 PMD_INIT_LOG(ERR, "VF is still resetting");
2095                 goto err;
2096         }
2097
2098         iavf_init_adminq_parameter(hw);
2099         err = iavf_init_adminq(hw);
2100         if (err) {
2101                 PMD_INIT_LOG(ERR, "init_adminq failed: %d", err);
2102                 goto err;
2103         }
2104
2105         vf->aq_resp = rte_zmalloc("vf_aq_resp", IAVF_AQ_BUF_SZ, 0);
2106         if (!vf->aq_resp) {
2107                 PMD_INIT_LOG(ERR, "unable to allocate vf_aq_resp memory");
2108                 goto err_aq;
2109         }
2110         if (iavf_check_api_version(adapter) != 0) {
2111                 PMD_INIT_LOG(ERR, "check_api version failed");
2112                 goto err_api;
2113         }
2114
2115         bufsz = sizeof(struct virtchnl_vf_resource) +
2116                 (IAVF_MAX_VF_VSI * sizeof(struct virtchnl_vsi_resource));
2117         vf->vf_res = rte_zmalloc("vf_res", bufsz, 0);
2118         if (!vf->vf_res) {
2119                 PMD_INIT_LOG(ERR, "unable to allocate vf_res memory");
2120                 goto err_api;
2121         }
2122
2123         if (iavf_get_vf_resource(adapter) != 0) {
2124                 PMD_INIT_LOG(ERR, "iavf_get_vf_config failed");
2125                 goto err_alloc;
2126         }
2127         /* Allocate memort for RSS info */
2128         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF) {
2129                 vf->rss_key = rte_zmalloc("rss_key",
2130                                           vf->vf_res->rss_key_size, 0);
2131                 if (!vf->rss_key) {
2132                         PMD_INIT_LOG(ERR, "unable to allocate rss_key memory");
2133                         goto err_rss;
2134                 }
2135                 vf->rss_lut = rte_zmalloc("rss_lut",
2136                                           vf->vf_res->rss_lut_size, 0);
2137                 if (!vf->rss_lut) {
2138                         PMD_INIT_LOG(ERR, "unable to allocate rss_lut memory");
2139                         goto err_rss;
2140                 }
2141         }
2142
2143         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RX_FLEX_DESC) {
2144                 if (iavf_get_supported_rxdid(adapter) != 0) {
2145                         PMD_INIT_LOG(ERR, "failed to do get supported rxdid");
2146                         goto err_rss;
2147                 }
2148         }
2149
2150         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_VLAN_V2) {
2151                 if (iavf_get_vlan_offload_caps_v2(adapter) != 0) {
2152                         PMD_INIT_LOG(ERR, "failed to do get VLAN offload v2 capabilities");
2153                         goto err_rss;
2154                 }
2155         }
2156
2157         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_QOS) {
2158                 bufsz = sizeof(struct virtchnl_qos_cap_list) +
2159                         IAVF_MAX_TRAFFIC_CLASS *
2160                         sizeof(struct virtchnl_qos_cap_elem);
2161                 vf->qos_cap = rte_zmalloc("qos_cap", bufsz, 0);
2162                 if (!vf->qos_cap) {
2163                         PMD_INIT_LOG(ERR, "unable to allocate qos_cap memory");
2164                         goto err_rss;
2165                 }
2166                 iavf_tm_conf_init(dev);
2167         }
2168
2169         iavf_init_proto_xtr(dev);
2170
2171         return 0;
2172 err_rss:
2173         rte_free(vf->rss_key);
2174         rte_free(vf->rss_lut);
2175 err_alloc:
2176         rte_free(vf->qos_cap);
2177         rte_free(vf->vf_res);
2178         vf->vsi_res = NULL;
2179 err_api:
2180         rte_free(vf->aq_resp);
2181 err_aq:
2182         iavf_shutdown_adminq(hw);
2183 err:
2184         return -1;
2185 }
2186
2187 static void
2188 iavf_uninit_vf(struct rte_eth_dev *dev)
2189 {
2190         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2191         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
2192
2193         iavf_shutdown_adminq(hw);
2194
2195         rte_free(vf->vf_res);
2196         vf->vsi_res = NULL;
2197         vf->vf_res = NULL;
2198
2199         rte_free(vf->aq_resp);
2200         vf->aq_resp = NULL;
2201
2202         rte_free(vf->qos_cap);
2203         vf->qos_cap = NULL;
2204
2205         rte_free(vf->rss_lut);
2206         vf->rss_lut = NULL;
2207         rte_free(vf->rss_key);
2208         vf->rss_key = NULL;
2209 }
2210
2211 /* Enable default admin queue interrupt setting */
2212 static inline void
2213 iavf_enable_irq0(struct iavf_hw *hw)
2214 {
2215         /* Enable admin queue interrupt trigger */
2216         IAVF_WRITE_REG(hw, IAVF_VFINT_ICR0_ENA1,
2217                        IAVF_VFINT_ICR0_ENA1_ADMINQ_MASK);
2218
2219         IAVF_WRITE_REG(hw, IAVF_VFINT_DYN_CTL01,
2220                        IAVF_VFINT_DYN_CTL01_INTENA_MASK |
2221                        IAVF_VFINT_DYN_CTL01_CLEARPBA_MASK |
2222                        IAVF_VFINT_DYN_CTL01_ITR_INDX_MASK);
2223
2224         IAVF_WRITE_FLUSH(hw);
2225 }
2226
2227 static inline void
2228 iavf_disable_irq0(struct iavf_hw *hw)
2229 {
2230         /* Disable all interrupt types */
2231         IAVF_WRITE_REG(hw, IAVF_VFINT_ICR0_ENA1, 0);
2232         IAVF_WRITE_REG(hw, IAVF_VFINT_DYN_CTL01,
2233                        IAVF_VFINT_DYN_CTL01_ITR_INDX_MASK);
2234         IAVF_WRITE_FLUSH(hw);
2235 }
2236
2237 static void
2238 iavf_dev_interrupt_handler(void *param)
2239 {
2240         struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
2241         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2242
2243         iavf_disable_irq0(hw);
2244
2245         iavf_handle_virtchnl_msg(dev);
2246
2247         iavf_enable_irq0(hw);
2248 }
2249
2250 void
2251 iavf_dev_alarm_handler(void *param)
2252 {
2253         struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
2254         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2255         uint32_t icr0;
2256
2257         iavf_disable_irq0(hw);
2258
2259         /* read out interrupt causes */
2260         icr0 = IAVF_READ_REG(hw, IAVF_VFINT_ICR01);
2261
2262         if (icr0 & IAVF_VFINT_ICR01_ADMINQ_MASK) {
2263                 PMD_DRV_LOG(DEBUG, "ICR01_ADMINQ is reported");
2264                 iavf_handle_virtchnl_msg(dev);
2265         }
2266
2267         iavf_enable_irq0(hw);
2268
2269         rte_eal_alarm_set(IAVF_ALARM_INTERVAL,
2270                           iavf_dev_alarm_handler, dev);
2271 }
2272
2273 static int
2274 iavf_dev_flow_ops_get(struct rte_eth_dev *dev,
2275                       const struct rte_flow_ops **ops)
2276 {
2277         if (!dev)
2278                 return -EINVAL;
2279
2280         *ops = &iavf_flow_ops;
2281         return 0;
2282 }
2283
2284 static void
2285 iavf_default_rss_disable(struct iavf_adapter *adapter)
2286 {
2287         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
2288         int ret = 0;
2289
2290         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF) {
2291                 /* Set hena = 0 to ask PF to cleanup all existing RSS. */
2292                 ret = iavf_set_hena(adapter, 0);
2293                 if (ret)
2294                         /* It is a workaround, temporarily allow error to be
2295                          * returned due to possible lack of PF handling for
2296                          * hena = 0.
2297                          */
2298                         PMD_INIT_LOG(WARNING, "fail to disable default RSS,"
2299                                     "lack PF support");
2300         }
2301 }
2302
2303 static int
2304 iavf_dev_init(struct rte_eth_dev *eth_dev)
2305 {
2306         struct iavf_adapter *adapter =
2307                 IAVF_DEV_PRIVATE_TO_ADAPTER(eth_dev->data->dev_private);
2308         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(adapter);
2309         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
2310         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
2311         int ret = 0;
2312
2313         PMD_INIT_FUNC_TRACE();
2314
2315         /* assign ops func pointer */
2316         eth_dev->dev_ops = &iavf_eth_dev_ops;
2317         eth_dev->rx_queue_count = iavf_dev_rxq_count;
2318         eth_dev->rx_descriptor_status = iavf_dev_rx_desc_status;
2319         eth_dev->tx_descriptor_status = iavf_dev_tx_desc_status;
2320         eth_dev->rx_pkt_burst = &iavf_recv_pkts;
2321         eth_dev->tx_pkt_burst = &iavf_xmit_pkts;
2322         eth_dev->tx_pkt_prepare = &iavf_prep_pkts;
2323
2324         /* For secondary processes, we don't initialise any further as primary
2325          * has already done this work. Only check if we need a different RX
2326          * and TX function.
2327          */
2328         if (rte_eal_process_type() != RTE_PROC_PRIMARY) {
2329                 iavf_set_rx_function(eth_dev);
2330                 iavf_set_tx_function(eth_dev);
2331                 return 0;
2332         }
2333         rte_eth_copy_pci_info(eth_dev, pci_dev);
2334         eth_dev->data->dev_flags |= RTE_ETH_DEV_AUTOFILL_QUEUE_XSTATS;
2335
2336         hw->vendor_id = pci_dev->id.vendor_id;
2337         hw->device_id = pci_dev->id.device_id;
2338         hw->subsystem_vendor_id = pci_dev->id.subsystem_vendor_id;
2339         hw->subsystem_device_id = pci_dev->id.subsystem_device_id;
2340         hw->bus.bus_id = pci_dev->addr.bus;
2341         hw->bus.device = pci_dev->addr.devid;
2342         hw->bus.func = pci_dev->addr.function;
2343         hw->hw_addr = (void *)pci_dev->mem_resource[0].addr;
2344         hw->back = IAVF_DEV_PRIVATE_TO_ADAPTER(eth_dev->data->dev_private);
2345         adapter->eth_dev = eth_dev;
2346         adapter->stopped = 1;
2347
2348         if (iavf_init_vf(eth_dev) != 0) {
2349                 PMD_INIT_LOG(ERR, "Init vf failed");
2350                 return -1;
2351         }
2352
2353         /* set default ptype table */
2354         adapter->ptype_tbl = iavf_get_default_ptype_table();
2355
2356         /* copy mac addr */
2357         eth_dev->data->mac_addrs = rte_zmalloc(
2358                 "iavf_mac", RTE_ETHER_ADDR_LEN * IAVF_NUM_MACADDR_MAX, 0);
2359         if (!eth_dev->data->mac_addrs) {
2360                 PMD_INIT_LOG(ERR, "Failed to allocate %d bytes needed to"
2361                              " store MAC addresses",
2362                              RTE_ETHER_ADDR_LEN * IAVF_NUM_MACADDR_MAX);
2363                 ret = -ENOMEM;
2364                 goto init_vf_err;
2365         }
2366         /* If the MAC address is not configured by host,
2367          * generate a random one.
2368          */
2369         if (!rte_is_valid_assigned_ether_addr(
2370                         (struct rte_ether_addr *)hw->mac.addr))
2371                 rte_eth_random_addr(hw->mac.addr);
2372         rte_ether_addr_copy((struct rte_ether_addr *)hw->mac.addr,
2373                         &eth_dev->data->mac_addrs[0]);
2374
2375         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_WB_ON_ITR) {
2376                 /* register callback func to eal lib */
2377                 rte_intr_callback_register(&pci_dev->intr_handle,
2378                                            iavf_dev_interrupt_handler,
2379                                            (void *)eth_dev);
2380
2381                 /* enable uio intr after callback register */
2382                 rte_intr_enable(&pci_dev->intr_handle);
2383         } else {
2384                 rte_eal_alarm_set(IAVF_ALARM_INTERVAL,
2385                                   iavf_dev_alarm_handler, eth_dev);
2386         }
2387
2388         /* configure and enable device interrupt */
2389         iavf_enable_irq0(hw);
2390
2391         ret = iavf_flow_init(adapter);
2392         if (ret) {
2393                 PMD_INIT_LOG(ERR, "Failed to initialize flow");
2394                 goto flow_init_err;
2395         }
2396
2397         iavf_default_rss_disable(adapter);
2398
2399         return 0;
2400
2401 flow_init_err:
2402         rte_free(eth_dev->data->mac_addrs);
2403         eth_dev->data->mac_addrs = NULL;
2404
2405 init_vf_err:
2406         iavf_uninit_vf(eth_dev);
2407
2408         return ret;
2409 }
2410
2411 static int
2412 iavf_dev_close(struct rte_eth_dev *dev)
2413 {
2414         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2415         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
2416         struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
2417         struct iavf_adapter *adapter =
2418                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
2419         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
2420         int ret;
2421
2422         if (rte_eal_process_type() != RTE_PROC_PRIMARY)
2423                 return 0;
2424
2425         ret = iavf_dev_stop(dev);
2426
2427         iavf_flow_flush(dev, NULL);
2428         iavf_flow_uninit(adapter);
2429
2430         /*
2431          * disable promiscuous mode before reset vf
2432          * it is a workaround solution when work with kernel driver
2433          * and it is not the normal way
2434          */
2435         if (vf->promisc_unicast_enabled || vf->promisc_multicast_enabled)
2436                 iavf_config_promisc(adapter, false, false);
2437
2438         iavf_shutdown_adminq(hw);
2439         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_WB_ON_ITR) {
2440                 /* disable uio intr before callback unregister */
2441                 rte_intr_disable(intr_handle);
2442
2443                 /* unregister callback func from eal lib */
2444                 rte_intr_callback_unregister(intr_handle,
2445                                              iavf_dev_interrupt_handler, dev);
2446         } else {
2447                 rte_eal_alarm_cancel(iavf_dev_alarm_handler, dev);
2448         }
2449         iavf_disable_irq0(hw);
2450
2451         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_QOS)
2452                 iavf_tm_conf_uninit(dev);
2453
2454         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF) {
2455                 if (vf->rss_lut) {
2456                         rte_free(vf->rss_lut);
2457                         vf->rss_lut = NULL;
2458                 }
2459                 if (vf->rss_key) {
2460                         rte_free(vf->rss_key);
2461                         vf->rss_key = NULL;
2462                 }
2463         }
2464
2465         rte_free(vf->vf_res);
2466         vf->vsi_res = NULL;
2467         vf->vf_res = NULL;
2468
2469         rte_free(vf->aq_resp);
2470         vf->aq_resp = NULL;
2471
2472         /*
2473          * If the VF is reset via VFLR, the device will be knocked out of bus
2474          * master mode, and the driver will fail to recover from the reset. Fix
2475          * this by enabling bus mastering after every reset. In a non-VFLR case,
2476          * the bus master bit will not be disabled, and this call will have no
2477          * effect.
2478          */
2479         if (vf->vf_reset && !rte_pci_set_bus_master(pci_dev, true))
2480                 vf->vf_reset = false;
2481
2482         return ret;
2483 }
2484
2485 static int
2486 iavf_dev_uninit(struct rte_eth_dev *dev)
2487 {
2488         if (rte_eal_process_type() != RTE_PROC_PRIMARY)
2489                 return -EPERM;
2490
2491         iavf_dev_close(dev);
2492
2493         return 0;
2494 }
2495
2496 /*
2497  * Reset VF device only to re-initialize resources in PMD layer
2498  */
2499 static int
2500 iavf_dev_reset(struct rte_eth_dev *dev)
2501 {
2502         int ret;
2503
2504         ret = iavf_dev_uninit(dev);
2505         if (ret)
2506                 return ret;
2507
2508         return iavf_dev_init(dev);
2509 }
2510
2511 static int
2512 iavf_dcf_cap_check_handler(__rte_unused const char *key,
2513                            const char *value, __rte_unused void *opaque)
2514 {
2515         if (strcmp(value, "dcf"))
2516                 return -1;
2517
2518         return 0;
2519 }
2520
2521 static int
2522 iavf_dcf_cap_selected(struct rte_devargs *devargs)
2523 {
2524         struct rte_kvargs *kvlist;
2525         const char *key = "cap";
2526         int ret = 0;
2527
2528         if (devargs == NULL)
2529                 return 0;
2530
2531         kvlist = rte_kvargs_parse(devargs->args, NULL);
2532         if (kvlist == NULL)
2533                 return 0;
2534
2535         if (!rte_kvargs_count(kvlist, key))
2536                 goto exit;
2537
2538         /* dcf capability selected when there's a key-value pair: cap=dcf */
2539         if (rte_kvargs_process(kvlist, key,
2540                                iavf_dcf_cap_check_handler, NULL) < 0)
2541                 goto exit;
2542
2543         ret = 1;
2544
2545 exit:
2546         rte_kvargs_free(kvlist);
2547         return ret;
2548 }
2549
2550 static int eth_iavf_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
2551                              struct rte_pci_device *pci_dev)
2552 {
2553         if (iavf_dcf_cap_selected(pci_dev->device.devargs))
2554                 return 1;
2555
2556         return rte_eth_dev_pci_generic_probe(pci_dev,
2557                 sizeof(struct iavf_adapter), iavf_dev_init);
2558 }
2559
2560 static int eth_iavf_pci_remove(struct rte_pci_device *pci_dev)
2561 {
2562         return rte_eth_dev_pci_generic_remove(pci_dev, iavf_dev_uninit);
2563 }
2564
2565 /* Adaptive virtual function driver struct */
2566 static struct rte_pci_driver rte_iavf_pmd = {
2567         .id_table = pci_id_iavf_map,
2568         .drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC,
2569         .probe = eth_iavf_pci_probe,
2570         .remove = eth_iavf_pci_remove,
2571 };
2572
2573 RTE_PMD_REGISTER_PCI(net_iavf, rte_iavf_pmd);
2574 RTE_PMD_REGISTER_PCI_TABLE(net_iavf, pci_id_iavf_map);
2575 RTE_PMD_REGISTER_KMOD_DEP(net_iavf, "* igb_uio | vfio-pci");
2576 RTE_PMD_REGISTER_PARAM_STRING(net_iavf, "cap=dcf");
2577 RTE_LOG_REGISTER_SUFFIX(iavf_logtype_init, init, NOTICE);
2578 RTE_LOG_REGISTER_SUFFIX(iavf_logtype_driver, driver, NOTICE);
2579 #ifdef RTE_ETHDEV_DEBUG_RX
2580 RTE_LOG_REGISTER_SUFFIX(iavf_logtype_rx, rx, DEBUG);
2581 #endif
2582 #ifdef RTE_ETHDEV_DEBUG_TX
2583 RTE_LOG_REGISTER_SUFFIX(iavf_logtype_tx, tx, DEBUG);
2584 #endif