drivers/net/ixgbe/ixgbe_ethdev.c

   1 /* SPDX-License-Identifier: BSD-3-Clause
   2  * Copyright(c) 2010-2017 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 <netinet/in.h>
  14 #include <rte_string_fns.h>
  15 #include <rte_byteorder.h>
  16 #include <rte_common.h>
  17 #include <rte_cycles.h>
  18
  19 #include <rte_interrupts.h>
  20 #include <rte_log.h>
  21 #include <rte_debug.h>
  22 #include <rte_pci.h>
  23 #include <rte_bus_pci.h>
  24 #include <rte_branch_prediction.h>
  25 #include <rte_memory.h>
  26 #include <rte_kvargs.h>
  27 #include <rte_eal.h>
  28 #include <rte_alarm.h>
  29 #include <rte_ether.h>
  30 #include <rte_ethdev_driver.h>
  31 #include <rte_ethdev_pci.h>
  32 #include <rte_malloc.h>
  33 #include <rte_random.h>
  34 #include <rte_dev.h>
  35 #include <rte_hash_crc.h>
  36 #ifdef RTE_LIBRTE_SECURITY
  37 #include <rte_security_driver.h>
  38 #endif
  39
  40 #include "ixgbe_logs.h"
  41 #include "base/ixgbe_api.h"
  42 #include "base/ixgbe_vf.h"
  43 #include "base/ixgbe_common.h"
  44 #include "ixgbe_ethdev.h"
  45 #include "ixgbe_bypass.h"
  46 #include "ixgbe_rxtx.h"
  47 #include "base/ixgbe_type.h"
  48 #include "base/ixgbe_phy.h"
  49 #include "ixgbe_regs.h"
  50
  51 /*
  52  * High threshold controlling when to start sending XOFF frames. Must be at
  53  * least 8 bytes less than receive packet buffer size. This value is in units
  54  * of 1024 bytes.
  55  */
  56 #define IXGBE_FC_HI    0x80
  57
  58 /*
  59  * Low threshold controlling when to start sending XON frames. This value is
  60  * in units of 1024 bytes.
  61  */
  62 #define IXGBE_FC_LO    0x40
  63
  64 /* Timer value included in XOFF frames. */
  65 #define IXGBE_FC_PAUSE 0x680
  66
  67 /*Default value of Max Rx Queue*/
  68 #define IXGBE_MAX_RX_QUEUE_NUM 128
  69
  70 #define IXGBE_LINK_DOWN_CHECK_TIMEOUT 4000 /* ms */
  71 #define IXGBE_LINK_UP_CHECK_TIMEOUT   1000 /* ms */
  72 #define IXGBE_VMDQ_NUM_UC_MAC         4096 /* Maximum nb. of UC MAC addr. */
  73
  74 #define IXGBE_MMW_SIZE_DEFAULT        0x4
  75 #define IXGBE_MMW_SIZE_JUMBO_FRAME    0x14
  76 #define IXGBE_MAX_RING_DESC           4096 /* replicate define from rxtx */
  77
  78 /*
  79  *  Default values for RX/TX configuration
  80  */
  81 #define IXGBE_DEFAULT_RX_FREE_THRESH  32
  82 #define IXGBE_DEFAULT_RX_PTHRESH      8
  83 #define IXGBE_DEFAULT_RX_HTHRESH      8
  84 #define IXGBE_DEFAULT_RX_WTHRESH      0
  85
  86 #define IXGBE_DEFAULT_TX_FREE_THRESH  32
  87 #define IXGBE_DEFAULT_TX_PTHRESH      32
  88 #define IXGBE_DEFAULT_TX_HTHRESH      0
  89 #define IXGBE_DEFAULT_TX_WTHRESH      0
  90 #define IXGBE_DEFAULT_TX_RSBIT_THRESH 32
  91
  92 /* Bit shift and mask */
  93 #define IXGBE_4_BIT_WIDTH  (CHAR_BIT / 2)
  94 #define IXGBE_4_BIT_MASK   RTE_LEN2MASK(IXGBE_4_BIT_WIDTH, uint8_t)
  95 #define IXGBE_8_BIT_WIDTH  CHAR_BIT
  96 #define IXGBE_8_BIT_MASK   UINT8_MAX
  97
  98 #define IXGBEVF_PMD_NAME "rte_ixgbevf_pmd" /* PMD name */
  99
 100 #define IXGBE_QUEUE_STAT_COUNTERS (sizeof(hw_stats->qprc) / sizeof(hw_stats->qprc[0]))
 101
 102 /* Additional timesync values. */
 103 #define NSEC_PER_SEC             1000000000L
 104 #define IXGBE_INCVAL_10GB        0x66666666
 105 #define IXGBE_INCVAL_1GB         0x40000000
 106 #define IXGBE_INCVAL_100         0x50000000
 107 #define IXGBE_INCVAL_SHIFT_10GB  28
 108 #define IXGBE_INCVAL_SHIFT_1GB   24
 109 #define IXGBE_INCVAL_SHIFT_100   21
 110 #define IXGBE_INCVAL_SHIFT_82599 7
 111 #define IXGBE_INCPER_SHIFT_82599 24
 112
 113 #define IXGBE_CYCLECOUNTER_MASK   0xffffffffffffffffULL
 114
 115 #define IXGBE_VT_CTL_POOLING_MODE_MASK         0x00030000
 116 #define IXGBE_VT_CTL_POOLING_MODE_ETAG         0x00010000
 117 #define IXGBE_ETAG_ETYPE                       0x00005084
 118 #define IXGBE_ETAG_ETYPE_MASK                  0x0000ffff
 119 #define IXGBE_ETAG_ETYPE_VALID                 0x80000000
 120 #define IXGBE_RAH_ADTYPE                       0x40000000
 121 #define IXGBE_RAL_ETAG_FILTER_MASK             0x00003fff
 122 #define IXGBE_VMVIR_TAGA_MASK                  0x18000000
 123 #define IXGBE_VMVIR_TAGA_ETAG_INSERT           0x08000000
 124 #define IXGBE_VMTIR(_i) (0x00017000 + ((_i) * 4)) /* 64 of these (0-63) */
 125 #define IXGBE_QDE_STRIP_TAG                    0x00000004
 126 #define IXGBE_VTEICR_MASK                      0x07
 127
 128 #define IXGBE_EXVET_VET_EXT_SHIFT              16
 129 #define IXGBE_DMATXCTL_VT_MASK                 0xFFFF0000
 130
 131 #define IXGBEVF_DEVARG_PFLINK_FULLCHK           "pflink_fullchk"
 132
 133 static const char * const ixgbevf_valid_arguments[] = {
 134         IXGBEVF_DEVARG_PFLINK_FULLCHK,
 135         NULL
 136 };
 137
 138 static int eth_ixgbe_dev_init(struct rte_eth_dev *eth_dev, void *init_params);
 139 static int eth_ixgbe_dev_uninit(struct rte_eth_dev *eth_dev);
 140 static int ixgbe_fdir_filter_init(struct rte_eth_dev *eth_dev);
 141 static int ixgbe_fdir_filter_uninit(struct rte_eth_dev *eth_dev);
 142 static int ixgbe_l2_tn_filter_init(struct rte_eth_dev *eth_dev);
 143 static int ixgbe_l2_tn_filter_uninit(struct rte_eth_dev *eth_dev);
 144 static int ixgbe_ntuple_filter_uninit(struct rte_eth_dev *eth_dev);
 145 static int  ixgbe_dev_configure(struct rte_eth_dev *dev);
 146 static int  ixgbe_dev_start(struct rte_eth_dev *dev);
 147 static void ixgbe_dev_stop(struct rte_eth_dev *dev);
 148 static int  ixgbe_dev_set_link_up(struct rte_eth_dev *dev);
 149 static int  ixgbe_dev_set_link_down(struct rte_eth_dev *dev);
 150 static int  ixgbe_dev_close(struct rte_eth_dev *dev);
 151 static int  ixgbe_dev_reset(struct rte_eth_dev *dev);
 152 static int ixgbe_dev_promiscuous_enable(struct rte_eth_dev *dev);
 153 static int ixgbe_dev_promiscuous_disable(struct rte_eth_dev *dev);
 154 static int ixgbe_dev_allmulticast_enable(struct rte_eth_dev *dev);
 155 static int ixgbe_dev_allmulticast_disable(struct rte_eth_dev *dev);
 156 static int ixgbe_dev_link_update(struct rte_eth_dev *dev,
 157                                 int wait_to_complete);
 158 static int ixgbe_dev_stats_get(struct rte_eth_dev *dev,
 159                                 struct rte_eth_stats *stats);
 160 static int ixgbe_dev_xstats_get(struct rte_eth_dev *dev,
 161                                 struct rte_eth_xstat *xstats, unsigned n);
 162 static int ixgbevf_dev_xstats_get(struct rte_eth_dev *dev,
 163                                   struct rte_eth_xstat *xstats, unsigned n);
 164 static int
 165 ixgbe_dev_xstats_get_by_id(struct rte_eth_dev *dev, const uint64_t *ids,
 166                 uint64_t *values, unsigned int n);
 167 static int ixgbe_dev_stats_reset(struct rte_eth_dev *dev);
 168 static int ixgbe_dev_xstats_reset(struct rte_eth_dev *dev);
 169 static int ixgbe_dev_xstats_get_names(struct rte_eth_dev *dev,
 170         struct rte_eth_xstat_name *xstats_names,
 171         unsigned int size);
 172 static int ixgbevf_dev_xstats_get_names(struct rte_eth_dev *dev,
 173         struct rte_eth_xstat_name *xstats_names, unsigned limit);
 174 static int ixgbe_dev_xstats_get_names_by_id(
 175         struct rte_eth_dev *dev,
 176         struct rte_eth_xstat_name *xstats_names,
 177         const uint64_t *ids,
 178         unsigned int limit);
 179 static int ixgbe_dev_queue_stats_mapping_set(struct rte_eth_dev *eth_dev,
 180                                              uint16_t queue_id,
 181                                              uint8_t stat_idx,
 182                                              uint8_t is_rx);
 183 static int ixgbe_fw_version_get(struct rte_eth_dev *dev, char *fw_version,
 184                                  size_t fw_size);
 185 static int ixgbe_dev_info_get(struct rte_eth_dev *dev,
 186                               struct rte_eth_dev_info *dev_info);
 187 static const uint32_t *ixgbe_dev_supported_ptypes_get(struct rte_eth_dev *dev);
 188 static int ixgbevf_dev_info_get(struct rte_eth_dev *dev,
 189                                 struct rte_eth_dev_info *dev_info);
 190 static int ixgbe_dev_mtu_set(struct rte_eth_dev *dev, uint16_t mtu);
 191
 192 static int ixgbe_vlan_filter_set(struct rte_eth_dev *dev,
 193                 uint16_t vlan_id, int on);
 194 static int ixgbe_vlan_tpid_set(struct rte_eth_dev *dev,
 195                                enum rte_vlan_type vlan_type,
 196                                uint16_t tpid_id);
 197 static void ixgbe_vlan_hw_strip_bitmap_set(struct rte_eth_dev *dev,
 198                 uint16_t queue, bool on);
 199 static void ixgbe_vlan_strip_queue_set(struct rte_eth_dev *dev, uint16_t queue,
 200                 int on);
 201 static void ixgbe_config_vlan_strip_on_all_queues(struct rte_eth_dev *dev,
 202                                                   int mask);
 203 static int ixgbe_vlan_offload_config(struct rte_eth_dev *dev, int mask);
 204 static int ixgbe_vlan_offload_set(struct rte_eth_dev *dev, int mask);
 205 static void ixgbe_vlan_hw_strip_enable(struct rte_eth_dev *dev, uint16_t queue);
 206 static void ixgbe_vlan_hw_strip_disable(struct rte_eth_dev *dev, uint16_t queue);
 207 static void ixgbe_vlan_hw_extend_enable(struct rte_eth_dev *dev);
 208 static void ixgbe_vlan_hw_extend_disable(struct rte_eth_dev *dev);
 209
 210 static int ixgbe_dev_led_on(struct rte_eth_dev *dev);
 211 static int ixgbe_dev_led_off(struct rte_eth_dev *dev);
 212 static int ixgbe_flow_ctrl_get(struct rte_eth_dev *dev,
 213                                struct rte_eth_fc_conf *fc_conf);
 214 static int ixgbe_flow_ctrl_set(struct rte_eth_dev *dev,
 215                                struct rte_eth_fc_conf *fc_conf);
 216 static int ixgbe_priority_flow_ctrl_set(struct rte_eth_dev *dev,
 217                 struct rte_eth_pfc_conf *pfc_conf);
 218 static int ixgbe_dev_rss_reta_update(struct rte_eth_dev *dev,
 219                         struct rte_eth_rss_reta_entry64 *reta_conf,
 220                         uint16_t reta_size);
 221 static int ixgbe_dev_rss_reta_query(struct rte_eth_dev *dev,
 222                         struct rte_eth_rss_reta_entry64 *reta_conf,
 223                         uint16_t reta_size);
 224 static void ixgbe_dev_link_status_print(struct rte_eth_dev *dev);
 225 static int ixgbe_dev_lsc_interrupt_setup(struct rte_eth_dev *dev, uint8_t on);
 226 static int ixgbe_dev_macsec_interrupt_setup(struct rte_eth_dev *dev);
 227 static int ixgbe_dev_rxq_interrupt_setup(struct rte_eth_dev *dev);
 228 static int ixgbe_dev_interrupt_get_status(struct rte_eth_dev *dev);
 229 static int ixgbe_dev_interrupt_action(struct rte_eth_dev *dev);
 230 static void ixgbe_dev_interrupt_handler(void *param);
 231 static void ixgbe_dev_interrupt_delayed_handler(void *param);
 232 static void *ixgbe_dev_setup_link_thread_handler(void *param);
 233 static int ixgbe_dev_wait_setup_link_complete(struct rte_eth_dev *dev,
 234                                               uint32_t timeout_ms);
 235
 236 static int ixgbe_add_rar(struct rte_eth_dev *dev,
 237                         struct rte_ether_addr *mac_addr,
 238                         uint32_t index, uint32_t pool);
 239 static void ixgbe_remove_rar(struct rte_eth_dev *dev, uint32_t index);
 240 static int ixgbe_set_default_mac_addr(struct rte_eth_dev *dev,
 241                                            struct rte_ether_addr *mac_addr);
 242 static void ixgbe_dcb_init(struct ixgbe_hw *hw, struct ixgbe_dcb_config *dcb_config);
 243 static bool is_device_supported(struct rte_eth_dev *dev,
 244                                 struct rte_pci_driver *drv);
 245
 246 /* For Virtual Function support */
 247 static int eth_ixgbevf_dev_init(struct rte_eth_dev *eth_dev);
 248 static int eth_ixgbevf_dev_uninit(struct rte_eth_dev *eth_dev);
 249 static int  ixgbevf_dev_configure(struct rte_eth_dev *dev);
 250 static int  ixgbevf_dev_start(struct rte_eth_dev *dev);
 251 static int ixgbevf_dev_link_update(struct rte_eth_dev *dev,
 252                                    int wait_to_complete);
 253 static void ixgbevf_dev_stop(struct rte_eth_dev *dev);
 254 static int ixgbevf_dev_close(struct rte_eth_dev *dev);
 255 static int  ixgbevf_dev_reset(struct rte_eth_dev *dev);
 256 static void ixgbevf_intr_disable(struct rte_eth_dev *dev);
 257 static void ixgbevf_intr_enable(struct rte_eth_dev *dev);
 258 static int ixgbevf_dev_stats_get(struct rte_eth_dev *dev,
 259                 struct rte_eth_stats *stats);
 260 static int ixgbevf_dev_stats_reset(struct rte_eth_dev *dev);
 261 static int ixgbevf_vlan_filter_set(struct rte_eth_dev *dev,
 262                 uint16_t vlan_id, int on);
 263 static void ixgbevf_vlan_strip_queue_set(struct rte_eth_dev *dev,
 264                 uint16_t queue, int on);
 265 static int ixgbevf_vlan_offload_config(struct rte_eth_dev *dev, int mask);
 266 static int ixgbevf_vlan_offload_set(struct rte_eth_dev *dev, int mask);
 267 static void ixgbevf_set_vfta_all(struct rte_eth_dev *dev, bool on);
 268 static int ixgbevf_dev_rx_queue_intr_enable(struct rte_eth_dev *dev,
 269                                             uint16_t queue_id);
 270 static int ixgbevf_dev_rx_queue_intr_disable(struct rte_eth_dev *dev,
 271                                              uint16_t queue_id);
 272 static void ixgbevf_set_ivar_map(struct ixgbe_hw *hw, int8_t direction,
 273                                  uint8_t queue, uint8_t msix_vector);
 274 static void ixgbevf_configure_msix(struct rte_eth_dev *dev);
 275 static int ixgbevf_dev_promiscuous_enable(struct rte_eth_dev *dev);
 276 static int ixgbevf_dev_promiscuous_disable(struct rte_eth_dev *dev);
 277 static int ixgbevf_dev_allmulticast_enable(struct rte_eth_dev *dev);
 278 static int ixgbevf_dev_allmulticast_disable(struct rte_eth_dev *dev);
 279
 280 /* For Eth VMDQ APIs support */
 281 static int ixgbe_uc_hash_table_set(struct rte_eth_dev *dev, struct
 282                 rte_ether_addr * mac_addr, uint8_t on);
 283 static int ixgbe_uc_all_hash_table_set(struct rte_eth_dev *dev, uint8_t on);
 284 static int ixgbe_mirror_rule_set(struct rte_eth_dev *dev,
 285                 struct rte_eth_mirror_conf *mirror_conf,
 286                 uint8_t rule_id, uint8_t on);
 287 static int ixgbe_mirror_rule_reset(struct rte_eth_dev *dev,
 288                 uint8_t rule_id);
 289 static int ixgbe_dev_rx_queue_intr_enable(struct rte_eth_dev *dev,
 290                                           uint16_t queue_id);
 291 static int ixgbe_dev_rx_queue_intr_disable(struct rte_eth_dev *dev,
 292                                            uint16_t queue_id);
 293 static void ixgbe_set_ivar_map(struct ixgbe_hw *hw, int8_t direction,
 294                                uint8_t queue, uint8_t msix_vector);
 295 static void ixgbe_configure_msix(struct rte_eth_dev *dev);
 296
 297 static int ixgbevf_add_mac_addr(struct rte_eth_dev *dev,
 298                                 struct rte_ether_addr *mac_addr,
 299                                 uint32_t index, uint32_t pool);
 300 static void ixgbevf_remove_mac_addr(struct rte_eth_dev *dev, uint32_t index);
 301 static int ixgbevf_set_default_mac_addr(struct rte_eth_dev *dev,
 302                                              struct rte_ether_addr *mac_addr);
 303 static int ixgbe_syn_filter_get(struct rte_eth_dev *dev,
 304                         struct rte_eth_syn_filter *filter);
 305 static int ixgbe_syn_filter_handle(struct rte_eth_dev *dev,
 306                         enum rte_filter_op filter_op,
 307                         void *arg);
 308 static int ixgbe_add_5tuple_filter(struct rte_eth_dev *dev,
 309                         struct ixgbe_5tuple_filter *filter);
 310 static void ixgbe_remove_5tuple_filter(struct rte_eth_dev *dev,
 311                         struct ixgbe_5tuple_filter *filter);
 312 static int ixgbe_ntuple_filter_handle(struct rte_eth_dev *dev,
 313                                 enum rte_filter_op filter_op,
 314                                 void *arg);
 315 static int ixgbe_get_ntuple_filter(struct rte_eth_dev *dev,
 316                         struct rte_eth_ntuple_filter *filter);
 317 static int ixgbe_ethertype_filter_handle(struct rte_eth_dev *dev,
 318                                 enum rte_filter_op filter_op,
 319                                 void *arg);
 320 static int ixgbe_get_ethertype_filter(struct rte_eth_dev *dev,
 321                         struct rte_eth_ethertype_filter *filter);
 322 static int ixgbe_dev_filter_ctrl(struct rte_eth_dev *dev,
 323                      enum rte_filter_type filter_type,
 324                      enum rte_filter_op filter_op,
 325                      void *arg);
 326 static int ixgbevf_dev_set_mtu(struct rte_eth_dev *dev, uint16_t mtu);
 327
 328 static int ixgbe_dev_set_mc_addr_list(struct rte_eth_dev *dev,
 329                                       struct rte_ether_addr *mc_addr_set,
 330                                       uint32_t nb_mc_addr);
 331 static int ixgbe_dev_get_dcb_info(struct rte_eth_dev *dev,
 332                                    struct rte_eth_dcb_info *dcb_info);
 333
 334 static int ixgbe_get_reg_length(struct rte_eth_dev *dev);
 335 static int ixgbe_get_regs(struct rte_eth_dev *dev,
 336                             struct rte_dev_reg_info *regs);
 337 static int ixgbe_get_eeprom_length(struct rte_eth_dev *dev);
 338 static int ixgbe_get_eeprom(struct rte_eth_dev *dev,
 339                                 struct rte_dev_eeprom_info *eeprom);
 340 static int ixgbe_set_eeprom(struct rte_eth_dev *dev,
 341                                 struct rte_dev_eeprom_info *eeprom);
 342
 343 static int ixgbe_get_module_info(struct rte_eth_dev *dev,
 344                                  struct rte_eth_dev_module_info *modinfo);
 345 static int ixgbe_get_module_eeprom(struct rte_eth_dev *dev,
 346                                    struct rte_dev_eeprom_info *info);
 347
 348 static int ixgbevf_get_reg_length(struct rte_eth_dev *dev);
 349 static int ixgbevf_get_regs(struct rte_eth_dev *dev,
 350                                 struct rte_dev_reg_info *regs);
 351
 352 static int ixgbe_timesync_enable(struct rte_eth_dev *dev);
 353 static int ixgbe_timesync_disable(struct rte_eth_dev *dev);
 354 static int ixgbe_timesync_read_rx_timestamp(struct rte_eth_dev *dev,
 355                                             struct timespec *timestamp,
 356                                             uint32_t flags);
 357 static int ixgbe_timesync_read_tx_timestamp(struct rte_eth_dev *dev,
 358                                             struct timespec *timestamp);
 359 static int ixgbe_timesync_adjust_time(struct rte_eth_dev *dev, int64_t delta);
 360 static int ixgbe_timesync_read_time(struct rte_eth_dev *dev,
 361                                    struct timespec *timestamp);
 362 static int ixgbe_timesync_write_time(struct rte_eth_dev *dev,
 363                                    const struct timespec *timestamp);
 364 static void ixgbevf_dev_interrupt_handler(void *param);
 365
 366 static int ixgbe_dev_l2_tunnel_eth_type_conf
 367         (struct rte_eth_dev *dev, struct rte_eth_l2_tunnel_conf *l2_tunnel);
 368 static int ixgbe_dev_l2_tunnel_offload_set
 369         (struct rte_eth_dev *dev,
 370          struct rte_eth_l2_tunnel_conf *l2_tunnel,
 371          uint32_t mask,
 372          uint8_t en);
 373 static int ixgbe_dev_l2_tunnel_filter_handle(struct rte_eth_dev *dev,
 374                                              enum rte_filter_op filter_op,
 375                                              void *arg);
 376
 377 static int ixgbe_dev_udp_tunnel_port_add(struct rte_eth_dev *dev,
 378                                          struct rte_eth_udp_tunnel *udp_tunnel);
 379 static int ixgbe_dev_udp_tunnel_port_del(struct rte_eth_dev *dev,
 380                                          struct rte_eth_udp_tunnel *udp_tunnel);
 381 static int ixgbe_filter_restore(struct rte_eth_dev *dev);
 382 static void ixgbe_l2_tunnel_conf(struct rte_eth_dev *dev);
 383 static int ixgbe_wait_for_link_up(struct ixgbe_hw *hw);
 384
 385 /*
 386  * Define VF Stats MACRO for Non "cleared on read" register
 387  */
 388 #define UPDATE_VF_STAT(reg, last, cur)                          \
 389 {                                                               \
 390         uint32_t latest = IXGBE_READ_REG(hw, reg);              \
 391         cur += (latest - last) & UINT_MAX;                      \
 392         last = latest;                                          \
 393 }
 394
 395 #define UPDATE_VF_STAT_36BIT(lsb, msb, last, cur)                \
 396 {                                                                \
 397         u64 new_lsb = IXGBE_READ_REG(hw, lsb);                   \
 398         u64 new_msb = IXGBE_READ_REG(hw, msb);                   \
 399         u64 latest = ((new_msb << 32) | new_lsb);                \
 400         cur += (0x1000000000LL + latest - last) & 0xFFFFFFFFFLL; \
 401         last = latest;                                           \
 402 }
 403
 404 #define IXGBE_SET_HWSTRIP(h, q) do {\
 405                 uint32_t idx = (q) / (sizeof((h)->bitmap[0]) * NBBY); \
 406                 uint32_t bit = (q) % (sizeof((h)->bitmap[0]) * NBBY); \
 407                 (h)->bitmap[idx] |= 1 << bit;\
 408         } while (0)
 409
 410 #define IXGBE_CLEAR_HWSTRIP(h, q) do {\
 411                 uint32_t idx = (q) / (sizeof((h)->bitmap[0]) * NBBY); \
 412                 uint32_t bit = (q) % (sizeof((h)->bitmap[0]) * NBBY); \
 413                 (h)->bitmap[idx] &= ~(1 << bit);\
 414         } while (0)
 415
 416 #define IXGBE_GET_HWSTRIP(h, q, r) do {\
 417                 uint32_t idx = (q) / (sizeof((h)->bitmap[0]) * NBBY); \
 418                 uint32_t bit = (q) % (sizeof((h)->bitmap[0]) * NBBY); \
 419                 (r) = (h)->bitmap[idx] >> bit & 1;\
 420         } while (0)
 421
 422 /*
 423  * The set of PCI devices this driver supports
 424  */
 425 static const struct rte_pci_id pci_id_ixgbe_map[] = {
 426         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598) },
 427         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598_BX) },
 428         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598AF_DUAL_PORT) },
 429         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598AF_SINGLE_PORT) },
 430         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598AT) },
 431         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598AT2) },
 432         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598EB_SFP_LOM) },
 433         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598EB_CX4) },
 434         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598_CX4_DUAL_PORT) },
 435         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598_DA_DUAL_PORT) },
 436         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598_SR_DUAL_PORT_EM) },
 437         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82598EB_XF_LR) },
 438         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_KX4) },
 439         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_KX4_MEZZ) },
 440         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_KR) },
 441         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_COMBO_BACKPLANE) },
 442         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_CX4) },
 443         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_SFP) },
 444         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_BACKPLANE_FCOE) },
 445         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_SFP_FCOE) },
 446         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_SFP_EM) },
 447         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_SFP_SF2) },
 448         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_SFP_SF_QP) },
 449         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_QSFP_SF_QP) },
 450         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599EN_SFP) },
 451         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_XAUI_LOM) },
 452         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_T3_LOM) },
 453         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X540T) },
 454         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X540T1) },
 455         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_SFP) },
 456         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_10G_T) },
 457         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_1G_T) },
 458         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550T) },
 459         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550T1) },
 460         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_KR) },
 461         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_KR_L) },
 462         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_SFP_N) },
 463         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_SGMII) },
 464         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_SGMII_L) },
 465         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_10G_T) },
 466         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_QSFP) },
 467         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_QSFP_N) },
 468         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_SFP) },
 469         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_1G_T) },
 470         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_1G_T_L) },
 471         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_KX4) },
 472         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_KR) },
 473         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_XFI) },
 474 #ifdef RTE_LIBRTE_IXGBE_BYPASS
 475         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_BYPASS) },
 476 #endif
 477         { .vendor_id = 0, /* sentinel */ },
 478 };
 479
 480 /*
 481  * The set of PCI devices this driver supports (for 82599 VF)
 482  */
 483 static const struct rte_pci_id pci_id_ixgbevf_map[] = {
 484         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_VF) },
 485         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_82599_VF_HV) },
 486         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X540_VF) },
 487         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X540_VF_HV) },
 488         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550_VF_HV) },
 489         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550_VF) },
 490         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_VF) },
 491         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_A_VF_HV) },
 492         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_VF) },
 493         { RTE_PCI_DEVICE(IXGBE_INTEL_VENDOR_ID, IXGBE_DEV_ID_X550EM_X_VF_HV) },
 494         { .vendor_id = 0, /* sentinel */ },
 495 };
 496
 497 static const struct rte_eth_desc_lim rx_desc_lim = {
 498         .nb_max = IXGBE_MAX_RING_DESC,
 499         .nb_min = IXGBE_MIN_RING_DESC,
 500         .nb_align = IXGBE_RXD_ALIGN,
 501 };
 502
 503 static const struct rte_eth_desc_lim tx_desc_lim = {
 504         .nb_max = IXGBE_MAX_RING_DESC,
 505         .nb_min = IXGBE_MIN_RING_DESC,
 506         .nb_align = IXGBE_TXD_ALIGN,
 507         .nb_seg_max = IXGBE_TX_MAX_SEG,
 508         .nb_mtu_seg_max = IXGBE_TX_MAX_SEG,
 509 };
 510
 511 static const struct eth_dev_ops ixgbe_eth_dev_ops = {
 512         .dev_configure        = ixgbe_dev_configure,
 513         .dev_start            = ixgbe_dev_start,
 514         .dev_stop             = ixgbe_dev_stop,
 515         .dev_set_link_up    = ixgbe_dev_set_link_up,
 516         .dev_set_link_down  = ixgbe_dev_set_link_down,
 517         .dev_close            = ixgbe_dev_close,
 518         .dev_reset            = ixgbe_dev_reset,
 519         .promiscuous_enable   = ixgbe_dev_promiscuous_enable,
 520         .promiscuous_disable  = ixgbe_dev_promiscuous_disable,
 521         .allmulticast_enable  = ixgbe_dev_allmulticast_enable,
 522         .allmulticast_disable = ixgbe_dev_allmulticast_disable,
 523         .link_update          = ixgbe_dev_link_update,
 524         .stats_get            = ixgbe_dev_stats_get,
 525         .xstats_get           = ixgbe_dev_xstats_get,
 526         .xstats_get_by_id     = ixgbe_dev_xstats_get_by_id,
 527         .stats_reset          = ixgbe_dev_stats_reset,
 528         .xstats_reset         = ixgbe_dev_xstats_reset,
 529         .xstats_get_names     = ixgbe_dev_xstats_get_names,
 530         .xstats_get_names_by_id = ixgbe_dev_xstats_get_names_by_id,
 531         .queue_stats_mapping_set = ixgbe_dev_queue_stats_mapping_set,
 532         .fw_version_get       = ixgbe_fw_version_get,
 533         .dev_infos_get        = ixgbe_dev_info_get,
 534         .dev_supported_ptypes_get = ixgbe_dev_supported_ptypes_get,
 535         .mtu_set              = ixgbe_dev_mtu_set,
 536         .vlan_filter_set      = ixgbe_vlan_filter_set,
 537         .vlan_tpid_set        = ixgbe_vlan_tpid_set,
 538         .vlan_offload_set     = ixgbe_vlan_offload_set,
 539         .vlan_strip_queue_set = ixgbe_vlan_strip_queue_set,
 540         .rx_queue_start       = ixgbe_dev_rx_queue_start,
 541         .rx_queue_stop        = ixgbe_dev_rx_queue_stop,
 542         .tx_queue_start       = ixgbe_dev_tx_queue_start,
 543         .tx_queue_stop        = ixgbe_dev_tx_queue_stop,
 544         .rx_queue_setup       = ixgbe_dev_rx_queue_setup,
 545         .rx_queue_intr_enable = ixgbe_dev_rx_queue_intr_enable,
 546         .rx_queue_intr_disable = ixgbe_dev_rx_queue_intr_disable,
 547         .rx_queue_release     = ixgbe_dev_rx_queue_release,
 548         .tx_queue_setup       = ixgbe_dev_tx_queue_setup,
 549         .tx_queue_release     = ixgbe_dev_tx_queue_release,
 550         .dev_led_on           = ixgbe_dev_led_on,
 551         .dev_led_off          = ixgbe_dev_led_off,
 552         .flow_ctrl_get        = ixgbe_flow_ctrl_get,
 553         .flow_ctrl_set        = ixgbe_flow_ctrl_set,
 554         .priority_flow_ctrl_set = ixgbe_priority_flow_ctrl_set,
 555         .mac_addr_add         = ixgbe_add_rar,
 556         .mac_addr_remove      = ixgbe_remove_rar,
 557         .mac_addr_set         = ixgbe_set_default_mac_addr,
 558         .uc_hash_table_set    = ixgbe_uc_hash_table_set,
 559         .uc_all_hash_table_set  = ixgbe_uc_all_hash_table_set,
 560         .mirror_rule_set      = ixgbe_mirror_rule_set,
 561         .mirror_rule_reset    = ixgbe_mirror_rule_reset,
 562         .set_queue_rate_limit = ixgbe_set_queue_rate_limit,
 563         .reta_update          = ixgbe_dev_rss_reta_update,
 564         .reta_query           = ixgbe_dev_rss_reta_query,
 565         .rss_hash_update      = ixgbe_dev_rss_hash_update,
 566         .rss_hash_conf_get    = ixgbe_dev_rss_hash_conf_get,
 567         .filter_ctrl          = ixgbe_dev_filter_ctrl,
 568         .set_mc_addr_list     = ixgbe_dev_set_mc_addr_list,
 569         .rxq_info_get         = ixgbe_rxq_info_get,
 570         .txq_info_get         = ixgbe_txq_info_get,
 571         .timesync_enable      = ixgbe_timesync_enable,
 572         .timesync_disable     = ixgbe_timesync_disable,
 573         .timesync_read_rx_timestamp = ixgbe_timesync_read_rx_timestamp,
 574         .timesync_read_tx_timestamp = ixgbe_timesync_read_tx_timestamp,
 575         .get_reg              = ixgbe_get_regs,
 576         .get_eeprom_length    = ixgbe_get_eeprom_length,
 577         .get_eeprom           = ixgbe_get_eeprom,
 578         .set_eeprom           = ixgbe_set_eeprom,
 579         .get_module_info      = ixgbe_get_module_info,
 580         .get_module_eeprom    = ixgbe_get_module_eeprom,
 581         .get_dcb_info         = ixgbe_dev_get_dcb_info,
 582         .timesync_adjust_time = ixgbe_timesync_adjust_time,
 583         .timesync_read_time   = ixgbe_timesync_read_time,
 584         .timesync_write_time  = ixgbe_timesync_write_time,
 585         .l2_tunnel_eth_type_conf = ixgbe_dev_l2_tunnel_eth_type_conf,
 586         .l2_tunnel_offload_set   = ixgbe_dev_l2_tunnel_offload_set,
 587         .udp_tunnel_port_add  = ixgbe_dev_udp_tunnel_port_add,
 588         .udp_tunnel_port_del  = ixgbe_dev_udp_tunnel_port_del,
 589         .tm_ops_get           = ixgbe_tm_ops_get,
 590         .tx_done_cleanup      = ixgbe_dev_tx_done_cleanup,
 591 };
 592
 593 /*
 594  * dev_ops for virtual function, bare necessities for basic vf
 595  * operation have been implemented
 596  */
 597 static const struct eth_dev_ops ixgbevf_eth_dev_ops = {
 598         .dev_configure        = ixgbevf_dev_configure,
 599         .dev_start            = ixgbevf_dev_start,
 600         .dev_stop             = ixgbevf_dev_stop,
 601         .link_update          = ixgbevf_dev_link_update,
 602         .stats_get            = ixgbevf_dev_stats_get,
 603         .xstats_get           = ixgbevf_dev_xstats_get,
 604         .stats_reset          = ixgbevf_dev_stats_reset,
 605         .xstats_reset         = ixgbevf_dev_stats_reset,
 606         .xstats_get_names     = ixgbevf_dev_xstats_get_names,
 607         .dev_close            = ixgbevf_dev_close,
 608         .dev_reset            = ixgbevf_dev_reset,
 609         .promiscuous_enable   = ixgbevf_dev_promiscuous_enable,
 610         .promiscuous_disable  = ixgbevf_dev_promiscuous_disable,
 611         .allmulticast_enable  = ixgbevf_dev_allmulticast_enable,
 612         .allmulticast_disable = ixgbevf_dev_allmulticast_disable,
 613         .dev_infos_get        = ixgbevf_dev_info_get,
 614         .dev_supported_ptypes_get = ixgbe_dev_supported_ptypes_get,
 615         .mtu_set              = ixgbevf_dev_set_mtu,
 616         .vlan_filter_set      = ixgbevf_vlan_filter_set,
 617         .vlan_strip_queue_set = ixgbevf_vlan_strip_queue_set,
 618         .vlan_offload_set     = ixgbevf_vlan_offload_set,
 619         .rx_queue_setup       = ixgbe_dev_rx_queue_setup,
 620         .rx_queue_release     = ixgbe_dev_rx_queue_release,
 621         .tx_queue_setup       = ixgbe_dev_tx_queue_setup,
 622         .tx_queue_release     = ixgbe_dev_tx_queue_release,
 623         .rx_queue_intr_enable = ixgbevf_dev_rx_queue_intr_enable,
 624         .rx_queue_intr_disable = ixgbevf_dev_rx_queue_intr_disable,
 625         .mac_addr_add         = ixgbevf_add_mac_addr,
 626         .mac_addr_remove      = ixgbevf_remove_mac_addr,
 627         .set_mc_addr_list     = ixgbe_dev_set_mc_addr_list,
 628         .rxq_info_get         = ixgbe_rxq_info_get,
 629         .txq_info_get         = ixgbe_txq_info_get,
 630         .mac_addr_set         = ixgbevf_set_default_mac_addr,
 631         .get_reg              = ixgbevf_get_regs,
 632         .reta_update          = ixgbe_dev_rss_reta_update,
 633         .reta_query           = ixgbe_dev_rss_reta_query,
 634         .rss_hash_update      = ixgbe_dev_rss_hash_update,
 635         .rss_hash_conf_get    = ixgbe_dev_rss_hash_conf_get,
 636         .tx_done_cleanup      = ixgbe_dev_tx_done_cleanup,
 637 };
 638
 639 /* store statistics names and its offset in stats structure */
 640 struct rte_ixgbe_xstats_name_off {
 641         char name[RTE_ETH_XSTATS_NAME_SIZE];
 642         unsigned offset;
 643 };
 644
 645 static const struct rte_ixgbe_xstats_name_off rte_ixgbe_stats_strings[] = {
 646         {"rx_crc_errors", offsetof(struct ixgbe_hw_stats, crcerrs)},
 647         {"rx_illegal_byte_errors", offsetof(struct ixgbe_hw_stats, illerrc)},
 648         {"rx_error_bytes", offsetof(struct ixgbe_hw_stats, errbc)},
 649         {"mac_local_errors", offsetof(struct ixgbe_hw_stats, mlfc)},
 650         {"mac_remote_errors", offsetof(struct ixgbe_hw_stats, mrfc)},
 651         {"rx_length_errors", offsetof(struct ixgbe_hw_stats, rlec)},
 652         {"tx_xon_packets", offsetof(struct ixgbe_hw_stats, lxontxc)},
 653         {"rx_xon_packets", offsetof(struct ixgbe_hw_stats, lxonrxc)},
 654         {"tx_xoff_packets", offsetof(struct ixgbe_hw_stats, lxofftxc)},
 655         {"rx_xoff_packets", offsetof(struct ixgbe_hw_stats, lxoffrxc)},
 656         {"rx_size_64_packets", offsetof(struct ixgbe_hw_stats, prc64)},
 657         {"rx_size_65_to_127_packets", offsetof(struct ixgbe_hw_stats, prc127)},
 658         {"rx_size_128_to_255_packets", offsetof(struct ixgbe_hw_stats, prc255)},
 659         {"rx_size_256_to_511_packets", offsetof(struct ixgbe_hw_stats, prc511)},
 660         {"rx_size_512_to_1023_packets", offsetof(struct ixgbe_hw_stats,
 661                 prc1023)},
 662         {"rx_size_1024_to_max_packets", offsetof(struct ixgbe_hw_stats,
 663                 prc1522)},
 664         {"rx_broadcast_packets", offsetof(struct ixgbe_hw_stats, bprc)},
 665         {"rx_multicast_packets", offsetof(struct ixgbe_hw_stats, mprc)},
 666         {"rx_fragment_errors", offsetof(struct ixgbe_hw_stats, rfc)},
 667         {"rx_undersize_errors", offsetof(struct ixgbe_hw_stats, ruc)},
 668         {"rx_oversize_errors", offsetof(struct ixgbe_hw_stats, roc)},
 669         {"rx_jabber_errors", offsetof(struct ixgbe_hw_stats, rjc)},
 670         {"rx_management_packets", offsetof(struct ixgbe_hw_stats, mngprc)},
 671         {"rx_management_dropped", offsetof(struct ixgbe_hw_stats, mngpdc)},
 672         {"tx_management_packets", offsetof(struct ixgbe_hw_stats, mngptc)},
 673         {"rx_total_packets", offsetof(struct ixgbe_hw_stats, tpr)},
 674         {"rx_total_bytes", offsetof(struct ixgbe_hw_stats, tor)},
 675         {"tx_total_packets", offsetof(struct ixgbe_hw_stats, tpt)},
 676         {"tx_size_64_packets", offsetof(struct ixgbe_hw_stats, ptc64)},
 677         {"tx_size_65_to_127_packets", offsetof(struct ixgbe_hw_stats, ptc127)},
 678         {"tx_size_128_to_255_packets", offsetof(struct ixgbe_hw_stats, ptc255)},
 679         {"tx_size_256_to_511_packets", offsetof(struct ixgbe_hw_stats, ptc511)},
 680         {"tx_size_512_to_1023_packets", offsetof(struct ixgbe_hw_stats,
 681                 ptc1023)},
 682         {"tx_size_1024_to_max_packets", offsetof(struct ixgbe_hw_stats,
 683                 ptc1522)},
 684         {"tx_multicast_packets", offsetof(struct ixgbe_hw_stats, mptc)},
 685         {"tx_broadcast_packets", offsetof(struct ixgbe_hw_stats, bptc)},
 686         {"rx_mac_short_packet_dropped", offsetof(struct ixgbe_hw_stats, mspdc)},
 687         {"rx_l3_l4_xsum_error", offsetof(struct ixgbe_hw_stats, xec)},
 688
 689         {"flow_director_added_filters", offsetof(struct ixgbe_hw_stats,
 690                 fdirustat_add)},
 691         {"flow_director_removed_filters", offsetof(struct ixgbe_hw_stats,
 692                 fdirustat_remove)},
 693         {"flow_director_filter_add_errors", offsetof(struct ixgbe_hw_stats,
 694                 fdirfstat_fadd)},
 695         {"flow_director_filter_remove_errors", offsetof(struct ixgbe_hw_stats,
 696                 fdirfstat_fremove)},
 697         {"flow_director_matched_filters", offsetof(struct ixgbe_hw_stats,
 698                 fdirmatch)},
 699         {"flow_director_missed_filters", offsetof(struct ixgbe_hw_stats,
 700                 fdirmiss)},
 701
 702         {"rx_fcoe_crc_errors", offsetof(struct ixgbe_hw_stats, fccrc)},
 703         {"rx_fcoe_dropped", offsetof(struct ixgbe_hw_stats, fcoerpdc)},
 704         {"rx_fcoe_mbuf_allocation_errors", offsetof(struct ixgbe_hw_stats,
 705                 fclast)},
 706         {"rx_fcoe_packets", offsetof(struct ixgbe_hw_stats, fcoeprc)},
 707         {"tx_fcoe_packets", offsetof(struct ixgbe_hw_stats, fcoeptc)},
 708         {"rx_fcoe_bytes", offsetof(struct ixgbe_hw_stats, fcoedwrc)},
 709         {"tx_fcoe_bytes", offsetof(struct ixgbe_hw_stats, fcoedwtc)},
 710         {"rx_fcoe_no_direct_data_placement", offsetof(struct ixgbe_hw_stats,
 711                 fcoe_noddp)},
 712         {"rx_fcoe_no_direct_data_placement_ext_buff",
 713                 offsetof(struct ixgbe_hw_stats, fcoe_noddp_ext_buff)},
 714
 715         {"tx_flow_control_xon_packets", offsetof(struct ixgbe_hw_stats,
 716                 lxontxc)},
 717         {"rx_flow_control_xon_packets", offsetof(struct ixgbe_hw_stats,
 718                 lxonrxc)},
 719         {"tx_flow_control_xoff_packets", offsetof(struct ixgbe_hw_stats,
 720                 lxofftxc)},
 721         {"rx_flow_control_xoff_packets", offsetof(struct ixgbe_hw_stats,
 722                 lxoffrxc)},
 723         {"rx_total_missed_packets", offsetof(struct ixgbe_hw_stats, mpctotal)},
 724 };
 725
 726 #define IXGBE_NB_HW_STATS (sizeof(rte_ixgbe_stats_strings) / \
 727                            sizeof(rte_ixgbe_stats_strings[0]))
 728
 729 /* MACsec statistics */
 730 static const struct rte_ixgbe_xstats_name_off rte_ixgbe_macsec_strings[] = {
 731         {"out_pkts_untagged", offsetof(struct ixgbe_macsec_stats,
 732                 out_pkts_untagged)},
 733         {"out_pkts_encrypted", offsetof(struct ixgbe_macsec_stats,
 734                 out_pkts_encrypted)},
 735         {"out_pkts_protected", offsetof(struct ixgbe_macsec_stats,
 736                 out_pkts_protected)},
 737         {"out_octets_encrypted", offsetof(struct ixgbe_macsec_stats,
 738                 out_octets_encrypted)},
 739         {"out_octets_protected", offsetof(struct ixgbe_macsec_stats,
 740                 out_octets_protected)},
 741         {"in_pkts_untagged", offsetof(struct ixgbe_macsec_stats,
 742                 in_pkts_untagged)},
 743         {"in_pkts_badtag", offsetof(struct ixgbe_macsec_stats,
 744                 in_pkts_badtag)},
 745         {"in_pkts_nosci", offsetof(struct ixgbe_macsec_stats,
 746                 in_pkts_nosci)},
 747         {"in_pkts_unknownsci", offsetof(struct ixgbe_macsec_stats,
 748                 in_pkts_unknownsci)},
 749         {"in_octets_decrypted", offsetof(struct ixgbe_macsec_stats,
 750                 in_octets_decrypted)},
 751         {"in_octets_validated", offsetof(struct ixgbe_macsec_stats,
 752                 in_octets_validated)},
 753         {"in_pkts_unchecked", offsetof(struct ixgbe_macsec_stats,
 754                 in_pkts_unchecked)},
 755         {"in_pkts_delayed", offsetof(struct ixgbe_macsec_stats,
 756                 in_pkts_delayed)},
 757         {"in_pkts_late", offsetof(struct ixgbe_macsec_stats,
 758                 in_pkts_late)},
 759         {"in_pkts_ok", offsetof(struct ixgbe_macsec_stats,
 760                 in_pkts_ok)},
 761         {"in_pkts_invalid", offsetof(struct ixgbe_macsec_stats,
 762                 in_pkts_invalid)},
 763         {"in_pkts_notvalid", offsetof(struct ixgbe_macsec_stats,
 764                 in_pkts_notvalid)},
 765         {"in_pkts_unusedsa", offsetof(struct ixgbe_macsec_stats,
 766                 in_pkts_unusedsa)},
 767         {"in_pkts_notusingsa", offsetof(struct ixgbe_macsec_stats,
 768                 in_pkts_notusingsa)},
 769 };
 770
 771 #define IXGBE_NB_MACSEC_STATS (sizeof(rte_ixgbe_macsec_strings) / \
 772                            sizeof(rte_ixgbe_macsec_strings[0]))
 773
 774 /* Per-queue statistics */
 775 static const struct rte_ixgbe_xstats_name_off rte_ixgbe_rxq_strings[] = {
 776         {"mbuf_allocation_errors", offsetof(struct ixgbe_hw_stats, rnbc)},
 777         {"dropped", offsetof(struct ixgbe_hw_stats, mpc)},
 778         {"xon_packets", offsetof(struct ixgbe_hw_stats, pxonrxc)},
 779         {"xoff_packets", offsetof(struct ixgbe_hw_stats, pxoffrxc)},
 780 };
 781
 782 #define IXGBE_NB_RXQ_PRIO_STATS (sizeof(rte_ixgbe_rxq_strings) / \
 783                            sizeof(rte_ixgbe_rxq_strings[0]))
 784 #define IXGBE_NB_RXQ_PRIO_VALUES 8
 785
 786 static const struct rte_ixgbe_xstats_name_off rte_ixgbe_txq_strings[] = {
 787         {"xon_packets", offsetof(struct ixgbe_hw_stats, pxontxc)},
 788         {"xoff_packets", offsetof(struct ixgbe_hw_stats, pxofftxc)},
 789         {"xon_to_xoff_packets", offsetof(struct ixgbe_hw_stats,
 790                 pxon2offc)},
 791 };
 792
 793 #define IXGBE_NB_TXQ_PRIO_STATS (sizeof(rte_ixgbe_txq_strings) / \
 794                            sizeof(rte_ixgbe_txq_strings[0]))
 795 #define IXGBE_NB_TXQ_PRIO_VALUES 8
 796
 797 static const struct rte_ixgbe_xstats_name_off rte_ixgbevf_stats_strings[] = {
 798         {"rx_multicast_packets", offsetof(struct ixgbevf_hw_stats, vfmprc)},
 799 };
 800
 801 #define IXGBEVF_NB_XSTATS (sizeof(rte_ixgbevf_stats_strings) /  \
 802                 sizeof(rte_ixgbevf_stats_strings[0]))
 803
 804 /*
 805  * This function is the same as ixgbe_is_sfp() in base/ixgbe.h.
 806  */
 807 static inline int
 808 ixgbe_is_sfp(struct ixgbe_hw *hw)
 809 {
 810         switch (hw->phy.type) {
 811         case ixgbe_phy_sfp_avago:
 812         case ixgbe_phy_sfp_ftl:
 813         case ixgbe_phy_sfp_intel:
 814         case ixgbe_phy_sfp_unknown:
 815         case ixgbe_phy_sfp_passive_tyco:
 816         case ixgbe_phy_sfp_passive_unknown:
 817                 return 1;
 818         default:
 819                 return 0;
 820         }
 821 }
 822
 823 static inline int32_t
 824 ixgbe_pf_reset_hw(struct ixgbe_hw *hw)
 825 {
 826         uint32_t ctrl_ext;
 827         int32_t status;
 828
 829         status = ixgbe_reset_hw(hw);
 830
 831         ctrl_ext = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
 832         /* Set PF Reset Done bit so PF/VF Mail Ops can work */
 833         ctrl_ext |= IXGBE_CTRL_EXT_PFRSTD;
 834         IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl_ext);
 835         IXGBE_WRITE_FLUSH(hw);
 836
 837         if (status == IXGBE_ERR_SFP_NOT_PRESENT)
 838                 status = IXGBE_SUCCESS;
 839         return status;
 840 }
 841
 842 static inline void
 843 ixgbe_enable_intr(struct rte_eth_dev *dev)
 844 {
 845         struct ixgbe_interrupt *intr =
 846                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
 847         struct ixgbe_hw *hw =
 848                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
 849
 850         IXGBE_WRITE_REG(hw, IXGBE_EIMS, intr->mask);
 851         IXGBE_WRITE_FLUSH(hw);
 852 }
 853
 854 /*
 855  * This function is based on ixgbe_disable_intr() in base/ixgbe.h.
 856  */
 857 static void
 858 ixgbe_disable_intr(struct ixgbe_hw *hw)
 859 {
 860         PMD_INIT_FUNC_TRACE();
 861
 862         if (hw->mac.type == ixgbe_mac_82598EB) {
 863                 IXGBE_WRITE_REG(hw, IXGBE_EIMC, ~0);
 864         } else {
 865                 IXGBE_WRITE_REG(hw, IXGBE_EIMC, 0xFFFF0000);
 866                 IXGBE_WRITE_REG(hw, IXGBE_EIMC_EX(0), ~0);
 867                 IXGBE_WRITE_REG(hw, IXGBE_EIMC_EX(1), ~0);
 868         }
 869         IXGBE_WRITE_FLUSH(hw);
 870 }
 871
 872 /*
 873  * This function resets queue statistics mapping registers.
 874  * From Niantic datasheet, Initialization of Statistics section:
 875  * "...if software requires the queue counters, the RQSMR and TQSM registers
 876  * must be re-programmed following a device reset.
 877  */
 878 static void
 879 ixgbe_reset_qstat_mappings(struct ixgbe_hw *hw)
 880 {
 881         uint32_t i;
 882
 883         for (i = 0; i != IXGBE_NB_STAT_MAPPING_REGS; i++) {
 884                 IXGBE_WRITE_REG(hw, IXGBE_RQSMR(i), 0);
 885                 IXGBE_WRITE_REG(hw, IXGBE_TQSM(i), 0);
 886         }
 887 }
 888
 889
 890 static int
 891 ixgbe_dev_queue_stats_mapping_set(struct rte_eth_dev *eth_dev,
 892                                   uint16_t queue_id,
 893                                   uint8_t stat_idx,
 894                                   uint8_t is_rx)
 895 {
 896 #define QSM_REG_NB_BITS_PER_QMAP_FIELD 8
 897 #define NB_QMAP_FIELDS_PER_QSM_REG 4
 898 #define QMAP_FIELD_RESERVED_BITS_MASK 0x0f
 899
 900         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
 901         struct ixgbe_stat_mapping_registers *stat_mappings =
 902                 IXGBE_DEV_PRIVATE_TO_STAT_MAPPINGS(eth_dev->data->dev_private);
 903         uint32_t qsmr_mask = 0;
 904         uint32_t clearing_mask = QMAP_FIELD_RESERVED_BITS_MASK;
 905         uint32_t q_map;
 906         uint8_t n, offset;
 907
 908         if ((hw->mac.type != ixgbe_mac_82599EB) &&
 909                 (hw->mac.type != ixgbe_mac_X540) &&
 910                 (hw->mac.type != ixgbe_mac_X550) &&
 911                 (hw->mac.type != ixgbe_mac_X550EM_x) &&
 912                 (hw->mac.type != ixgbe_mac_X550EM_a))
 913                 return -ENOSYS;
 914
 915         PMD_INIT_LOG(DEBUG, "Setting port %d, %s queue_id %d to stat index %d",
 916                      (int)(eth_dev->data->port_id), is_rx ? "RX" : "TX",
 917                      queue_id, stat_idx);
 918
 919         n = (uint8_t)(queue_id / NB_QMAP_FIELDS_PER_QSM_REG);
 920         if (n >= IXGBE_NB_STAT_MAPPING_REGS) {
 921                 PMD_INIT_LOG(ERR, "Nb of stat mapping registers exceeded");
 922                 return -EIO;
 923         }
 924         offset = (uint8_t)(queue_id % NB_QMAP_FIELDS_PER_QSM_REG);
 925
 926         /* Now clear any previous stat_idx set */
 927         clearing_mask <<= (QSM_REG_NB_BITS_PER_QMAP_FIELD * offset);
 928         if (!is_rx)
 929                 stat_mappings->tqsm[n] &= ~clearing_mask;
 930         else
 931                 stat_mappings->rqsmr[n] &= ~clearing_mask;
 932
 933         q_map = (uint32_t)stat_idx;
 934         q_map &= QMAP_FIELD_RESERVED_BITS_MASK;
 935         qsmr_mask = q_map << (QSM_REG_NB_BITS_PER_QMAP_FIELD * offset);
 936         if (!is_rx)
 937                 stat_mappings->tqsm[n] |= qsmr_mask;
 938         else
 939                 stat_mappings->rqsmr[n] |= qsmr_mask;
 940
 941         PMD_INIT_LOG(DEBUG, "Set port %d, %s queue_id %d to stat index %d",
 942                      (int)(eth_dev->data->port_id), is_rx ? "RX" : "TX",
 943                      queue_id, stat_idx);
 944         PMD_INIT_LOG(DEBUG, "%s[%d] = 0x%08x", is_rx ? "RQSMR" : "TQSM", n,
 945                      is_rx ? stat_mappings->rqsmr[n] : stat_mappings->tqsm[n]);
 946
 947         /* Now write the mapping in the appropriate register */
 948         if (is_rx) {
 949                 PMD_INIT_LOG(DEBUG, "Write 0x%x to RX IXGBE stat mapping reg:%d",
 950                              stat_mappings->rqsmr[n], n);
 951                 IXGBE_WRITE_REG(hw, IXGBE_RQSMR(n), stat_mappings->rqsmr[n]);
 952         } else {
 953                 PMD_INIT_LOG(DEBUG, "Write 0x%x to TX IXGBE stat mapping reg:%d",
 954                              stat_mappings->tqsm[n], n);
 955                 IXGBE_WRITE_REG(hw, IXGBE_TQSM(n), stat_mappings->tqsm[n]);
 956         }
 957         return 0;
 958 }
 959
 960 static void
 961 ixgbe_restore_statistics_mapping(struct rte_eth_dev *dev)
 962 {
 963         struct ixgbe_stat_mapping_registers *stat_mappings =
 964                 IXGBE_DEV_PRIVATE_TO_STAT_MAPPINGS(dev->data->dev_private);
 965         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
 966         int i;
 967
 968         /* write whatever was in stat mapping table to the NIC */
 969         for (i = 0; i < IXGBE_NB_STAT_MAPPING_REGS; i++) {
 970                 /* rx */
 971                 IXGBE_WRITE_REG(hw, IXGBE_RQSMR(i), stat_mappings->rqsmr[i]);
 972
 973                 /* tx */
 974                 IXGBE_WRITE_REG(hw, IXGBE_TQSM(i), stat_mappings->tqsm[i]);
 975         }
 976 }
 977
 978 static void
 979 ixgbe_dcb_init(struct ixgbe_hw *hw, struct ixgbe_dcb_config *dcb_config)
 980 {
 981         uint8_t i;
 982         struct ixgbe_dcb_tc_config *tc;
 983         uint8_t dcb_max_tc = IXGBE_DCB_MAX_TRAFFIC_CLASS;
 984
 985         dcb_config->num_tcs.pg_tcs = dcb_max_tc;
 986         dcb_config->num_tcs.pfc_tcs = dcb_max_tc;
 987         for (i = 0; i < dcb_max_tc; i++) {
 988                 tc = &dcb_config->tc_config[i];
 989                 tc->path[IXGBE_DCB_TX_CONFIG].bwg_id = i;
 990                 tc->path[IXGBE_DCB_TX_CONFIG].bwg_percent =
 991                                  (uint8_t)(100/dcb_max_tc + (i & 1));
 992                 tc->path[IXGBE_DCB_RX_CONFIG].bwg_id = i;
 993                 tc->path[IXGBE_DCB_RX_CONFIG].bwg_percent =
 994                                  (uint8_t)(100/dcb_max_tc + (i & 1));
 995                 tc->pfc = ixgbe_dcb_pfc_disabled;
 996         }
 997
 998         /* Initialize default user to priority mapping, UPx->TC0 */
 999         tc = &dcb_config->tc_config[0];
1000         tc->path[IXGBE_DCB_TX_CONFIG].up_to_tc_bitmap = 0xFF;
1001         tc->path[IXGBE_DCB_RX_CONFIG].up_to_tc_bitmap = 0xFF;
1002         for (i = 0; i < IXGBE_DCB_MAX_BW_GROUP; i++) {
1003                 dcb_config->bw_percentage[IXGBE_DCB_TX_CONFIG][i] = 100;
1004                 dcb_config->bw_percentage[IXGBE_DCB_RX_CONFIG][i] = 100;
1005         }
1006         dcb_config->rx_pba_cfg = ixgbe_dcb_pba_equal;
1007         dcb_config->pfc_mode_enable = false;
1008         dcb_config->vt_mode = true;
1009         dcb_config->round_robin_enable = false;
1010         /* support all DCB capabilities in 82599 */
1011         dcb_config->support.capabilities = 0xFF;
1012
1013         /*we only support 4 Tcs for X540, X550 */
1014         if (hw->mac.type == ixgbe_mac_X540 ||
1015                 hw->mac.type == ixgbe_mac_X550 ||
1016                 hw->mac.type == ixgbe_mac_X550EM_x ||
1017                 hw->mac.type == ixgbe_mac_X550EM_a) {
1018                 dcb_config->num_tcs.pg_tcs = 4;
1019                 dcb_config->num_tcs.pfc_tcs = 4;
1020         }
1021 }
1022
1023 /*
1024  * Ensure that all locks are released before first NVM or PHY access
1025  */
1026 static void
1027 ixgbe_swfw_lock_reset(struct ixgbe_hw *hw)
1028 {
1029         uint16_t mask;
1030
1031         /*
1032          * Phy lock should not fail in this early stage. If this is the case,
1033          * it is due to an improper exit of the application.
1034          * So force the release of the faulty lock. Release of common lock
1035          * is done automatically by swfw_sync function.
1036          */
1037         mask = IXGBE_GSSR_PHY0_SM << hw->bus.func;
1038         if (ixgbe_acquire_swfw_semaphore(hw, mask) < 0) {
1039                 PMD_DRV_LOG(DEBUG, "SWFW phy%d lock released", hw->bus.func);
1040         }
1041         ixgbe_release_swfw_semaphore(hw, mask);
1042
1043         /*
1044          * These ones are more tricky since they are common to all ports; but
1045          * swfw_sync retries last long enough (1s) to be almost sure that if
1046          * lock can not be taken it is due to an improper lock of the
1047          * semaphore.
1048          */
1049         mask = IXGBE_GSSR_EEP_SM | IXGBE_GSSR_MAC_CSR_SM | IXGBE_GSSR_SW_MNG_SM;
1050         if (ixgbe_acquire_swfw_semaphore(hw, mask) < 0) {
1051                 PMD_DRV_LOG(DEBUG, "SWFW common locks released");
1052         }
1053         ixgbe_release_swfw_semaphore(hw, mask);
1054 }
1055
1056 /*
1057  * This function is based on code in ixgbe_attach() in base/ixgbe.c.
1058  * It returns 0 on success.
1059  */
1060 static int
1061 eth_ixgbe_dev_init(struct rte_eth_dev *eth_dev, void *init_params __rte_unused)
1062 {
1063         struct ixgbe_adapter *ad = eth_dev->data->dev_private;
1064         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
1065         struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
1066         struct ixgbe_hw *hw =
1067                 IXGBE_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
1068         struct ixgbe_vfta *shadow_vfta =
1069                 IXGBE_DEV_PRIVATE_TO_VFTA(eth_dev->data->dev_private);
1070         struct ixgbe_hwstrip *hwstrip =
1071                 IXGBE_DEV_PRIVATE_TO_HWSTRIP_BITMAP(eth_dev->data->dev_private);
1072         struct ixgbe_dcb_config *dcb_config =
1073                 IXGBE_DEV_PRIVATE_TO_DCB_CFG(eth_dev->data->dev_private);
1074         struct ixgbe_filter_info *filter_info =
1075                 IXGBE_DEV_PRIVATE_TO_FILTER_INFO(eth_dev->data->dev_private);
1076         struct ixgbe_bw_conf *bw_conf =
1077                 IXGBE_DEV_PRIVATE_TO_BW_CONF(eth_dev->data->dev_private);
1078         uint32_t ctrl_ext;
1079         uint16_t csum;
1080         int diag, i;
1081
1082         PMD_INIT_FUNC_TRACE();
1083
1084         ixgbe_dev_macsec_setting_reset(eth_dev);
1085
1086         eth_dev->dev_ops = &ixgbe_eth_dev_ops;
1087         eth_dev->rx_queue_count       = ixgbe_dev_rx_queue_count;
1088         eth_dev->rx_descriptor_done   = ixgbe_dev_rx_descriptor_done;
1089         eth_dev->rx_descriptor_status = ixgbe_dev_rx_descriptor_status;
1090         eth_dev->tx_descriptor_status = ixgbe_dev_tx_descriptor_status;
1091         eth_dev->rx_pkt_burst = &ixgbe_recv_pkts;
1092         eth_dev->tx_pkt_burst = &ixgbe_xmit_pkts;
1093         eth_dev->tx_pkt_prepare = &ixgbe_prep_pkts;
1094
1095         /*
1096          * For secondary processes, we don't initialise any further as primary
1097          * has already done this work. Only check we don't need a different
1098          * RX and TX function.
1099          */
1100         if (rte_eal_process_type() != RTE_PROC_PRIMARY) {
1101                 struct ixgbe_tx_queue *txq;
1102                 /* TX queue function in primary, set by last queue initialized
1103                  * Tx queue may not initialized by primary process
1104                  */
1105                 if (eth_dev->data->tx_queues) {
1106                         txq = eth_dev->data->tx_queues[eth_dev->data->nb_tx_queues-1];
1107                         ixgbe_set_tx_function(eth_dev, txq);
1108                 } else {
1109                         /* Use default TX function if we get here */
1110                         PMD_INIT_LOG(NOTICE, "No TX queues configured yet. "
1111                                      "Using default TX function.");
1112                 }
1113
1114                 ixgbe_set_rx_function(eth_dev);
1115
1116                 return 0;
1117         }
1118
1119         rte_atomic32_clear(&ad->link_thread_running);
1120         rte_eth_copy_pci_info(eth_dev, pci_dev);
1121
1122         /* Vendor and Device ID need to be set before init of shared code */
1123         hw->device_id = pci_dev->id.device_id;
1124         hw->vendor_id = pci_dev->id.vendor_id;
1125         hw->hw_addr = (void *)pci_dev->mem_resource[0].addr;
1126         hw->allow_unsupported_sfp = 1;
1127
1128         /* Initialize the shared code (base driver) */
1129 #ifdef RTE_LIBRTE_IXGBE_BYPASS
1130         diag = ixgbe_bypass_init_shared_code(hw);
1131 #else
1132         diag = ixgbe_init_shared_code(hw);
1133 #endif /* RTE_LIBRTE_IXGBE_BYPASS */
1134
1135         if (diag != IXGBE_SUCCESS) {
1136                 PMD_INIT_LOG(ERR, "Shared code init failed: %d", diag);
1137                 return -EIO;
1138         }
1139
1140         if (hw->mac.ops.fw_recovery_mode && hw->mac.ops.fw_recovery_mode(hw)) {
1141                 PMD_INIT_LOG(ERR, "\nERROR: "
1142                         "Firmware recovery mode detected. Limiting functionality.\n"
1143                         "Refer to the Intel(R) Ethernet Adapters and Devices "
1144                         "User Guide for details on firmware recovery mode.");
1145                 return -EIO;
1146         }
1147
1148         /* pick up the PCI bus settings for reporting later */
1149         ixgbe_get_bus_info(hw);
1150
1151         /* Unlock any pending hardware semaphore */
1152         ixgbe_swfw_lock_reset(hw);
1153
1154 #ifdef RTE_LIBRTE_SECURITY
1155         /* Initialize security_ctx only for primary process*/
1156         if (ixgbe_ipsec_ctx_create(eth_dev))
1157                 return -ENOMEM;
1158 #endif
1159
1160         /* Initialize DCB configuration*/
1161         memset(dcb_config, 0, sizeof(struct ixgbe_dcb_config));
1162         ixgbe_dcb_init(hw, dcb_config);
1163         /* Get Hardware Flow Control setting */
1164         hw->fc.requested_mode = ixgbe_fc_none;
1165         hw->fc.current_mode = ixgbe_fc_none;
1166         hw->fc.pause_time = IXGBE_FC_PAUSE;
1167         for (i = 0; i < IXGBE_DCB_MAX_TRAFFIC_CLASS; i++) {
1168                 hw->fc.low_water[i] = IXGBE_FC_LO;
1169                 hw->fc.high_water[i] = IXGBE_FC_HI;
1170         }
1171         hw->fc.send_xon = 1;
1172
1173         /* Make sure we have a good EEPROM before we read from it */
1174         diag = ixgbe_validate_eeprom_checksum(hw, &csum);
1175         if (diag != IXGBE_SUCCESS) {
1176                 PMD_INIT_LOG(ERR, "The EEPROM checksum is not valid: %d", diag);
1177                 return -EIO;
1178         }
1179
1180 #ifdef RTE_LIBRTE_IXGBE_BYPASS
1181         diag = ixgbe_bypass_init_hw(hw);
1182 #else
1183         diag = ixgbe_init_hw(hw);
1184 #endif /* RTE_LIBRTE_IXGBE_BYPASS */
1185
1186         /*
1187          * Devices with copper phys will fail to initialise if ixgbe_init_hw()
1188          * is called too soon after the kernel driver unbinding/binding occurs.
1189          * The failure occurs in ixgbe_identify_phy_generic() for all devices,
1190          * but for non-copper devies, ixgbe_identify_sfp_module_generic() is
1191          * also called. See ixgbe_identify_phy_82599(). The reason for the
1192          * failure is not known, and only occuts when virtualisation features
1193          * are disabled in the bios. A delay of 100ms  was found to be enough by
1194          * trial-and-error, and is doubled to be safe.
1195          */
1196         if (diag && (hw->mac.ops.get_media_type(hw) == ixgbe_media_type_copper)) {
1197                 rte_delay_ms(200);
1198                 diag = ixgbe_init_hw(hw);
1199         }
1200
1201         if (diag == IXGBE_ERR_SFP_NOT_PRESENT)
1202                 diag = IXGBE_SUCCESS;
1203
1204         if (diag == IXGBE_ERR_EEPROM_VERSION) {
1205                 PMD_INIT_LOG(ERR, "This device is a pre-production adapter/"
1206                              "LOM.  Please be aware there may be issues associated "
1207                              "with your hardware.");
1208                 PMD_INIT_LOG(ERR, "If you are experiencing problems "
1209                              "please contact your Intel or hardware representative "
1210                              "who provided you with this hardware.");
1211         } else if (diag == IXGBE_ERR_SFP_NOT_SUPPORTED)
1212                 PMD_INIT_LOG(ERR, "Unsupported SFP+ Module");
1213         if (diag) {
1214                 PMD_INIT_LOG(ERR, "Hardware Initialization Failure: %d", diag);
1215                 return -EIO;
1216         }
1217
1218         /* Reset the hw statistics */
1219         ixgbe_dev_stats_reset(eth_dev);
1220
1221         /* disable interrupt */
1222         ixgbe_disable_intr(hw);
1223
1224         /* reset mappings for queue statistics hw counters*/
1225         ixgbe_reset_qstat_mappings(hw);
1226
1227         /* Allocate memory for storing MAC addresses */
1228         eth_dev->data->mac_addrs = rte_zmalloc("ixgbe", RTE_ETHER_ADDR_LEN *
1229                                                hw->mac.num_rar_entries, 0);
1230         if (eth_dev->data->mac_addrs == NULL) {
1231                 PMD_INIT_LOG(ERR,
1232                              "Failed to allocate %u bytes needed to store "
1233                              "MAC addresses",
1234                              RTE_ETHER_ADDR_LEN * hw->mac.num_rar_entries);
1235                 return -ENOMEM;
1236         }
1237         /* Copy the permanent MAC address */
1238         rte_ether_addr_copy((struct rte_ether_addr *)hw->mac.perm_addr,
1239                         &eth_dev->data->mac_addrs[0]);
1240
1241         /* Allocate memory for storing hash filter MAC addresses */
1242         eth_dev->data->hash_mac_addrs = rte_zmalloc(
1243                 "ixgbe", RTE_ETHER_ADDR_LEN * IXGBE_VMDQ_NUM_UC_MAC, 0);
1244         if (eth_dev->data->hash_mac_addrs == NULL) {
1245                 PMD_INIT_LOG(ERR,
1246                              "Failed to allocate %d bytes needed to store MAC addresses",
1247                              RTE_ETHER_ADDR_LEN * IXGBE_VMDQ_NUM_UC_MAC);
1248                 return -ENOMEM;
1249         }
1250
1251         /* Pass the information to the rte_eth_dev_close() that it should also
1252          * release the private port resources.
1253          */
1254         eth_dev->data->dev_flags |= RTE_ETH_DEV_CLOSE_REMOVE;
1255
1256         /* initialize the vfta */
1257         memset(shadow_vfta, 0, sizeof(*shadow_vfta));
1258
1259         /* initialize the hw strip bitmap*/
1260         memset(hwstrip, 0, sizeof(*hwstrip));
1261
1262         /* initialize PF if max_vfs not zero */
1263         ixgbe_pf_host_init(eth_dev);
1264
1265         ctrl_ext = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
1266         /* let hardware know driver is loaded */
1267         ctrl_ext |= IXGBE_CTRL_EXT_DRV_LOAD;
1268         /* Set PF Reset Done bit so PF/VF Mail Ops can work */
1269         ctrl_ext |= IXGBE_CTRL_EXT_PFRSTD;
1270         IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl_ext);
1271         IXGBE_WRITE_FLUSH(hw);
1272
1273         if (ixgbe_is_sfp(hw) && hw->phy.sfp_type != ixgbe_sfp_type_not_present)
1274                 PMD_INIT_LOG(DEBUG, "MAC: %d, PHY: %d, SFP+: %d",
1275                              (int) hw->mac.type, (int) hw->phy.type,
1276                              (int) hw->phy.sfp_type);
1277         else
1278                 PMD_INIT_LOG(DEBUG, "MAC: %d, PHY: %d",
1279                              (int) hw->mac.type, (int) hw->phy.type);
1280
1281         PMD_INIT_LOG(DEBUG, "port %d vendorID=0x%x deviceID=0x%x",
1282                      eth_dev->data->port_id, pci_dev->id.vendor_id,
1283                      pci_dev->id.device_id);
1284
1285         rte_intr_callback_register(intr_handle,
1286                                    ixgbe_dev_interrupt_handler, eth_dev);
1287
1288         /* enable uio/vfio intr/eventfd mapping */
1289         rte_intr_enable(intr_handle);
1290
1291         /* enable support intr */
1292         ixgbe_enable_intr(eth_dev);
1293
1294         /* initialize filter info */
1295         memset(filter_info, 0,
1296                sizeof(struct ixgbe_filter_info));
1297
1298         /* initialize 5tuple filter list */
1299         TAILQ_INIT(&filter_info->fivetuple_list);
1300
1301         /* initialize flow director filter list & hash */
1302         ixgbe_fdir_filter_init(eth_dev);
1303
1304         /* initialize l2 tunnel filter list & hash */
1305         ixgbe_l2_tn_filter_init(eth_dev);
1306
1307         /* initialize flow filter lists */
1308         ixgbe_filterlist_init();
1309
1310         /* initialize bandwidth configuration info */
1311         memset(bw_conf, 0, sizeof(struct ixgbe_bw_conf));
1312
1313         /* initialize Traffic Manager configuration */
1314         ixgbe_tm_conf_init(eth_dev);
1315
1316         return 0;
1317 }
1318
1319 static int
1320 eth_ixgbe_dev_uninit(struct rte_eth_dev *eth_dev)
1321 {
1322         PMD_INIT_FUNC_TRACE();
1323
1324         if (rte_eal_process_type() != RTE_PROC_PRIMARY)
1325                 return 0;
1326
1327         ixgbe_dev_close(eth_dev);
1328
1329         return 0;
1330 }
1331
1332 static int ixgbe_ntuple_filter_uninit(struct rte_eth_dev *eth_dev)
1333 {
1334         struct ixgbe_filter_info *filter_info =
1335                 IXGBE_DEV_PRIVATE_TO_FILTER_INFO(eth_dev->data->dev_private);
1336         struct ixgbe_5tuple_filter *p_5tuple;
1337
1338         while ((p_5tuple = TAILQ_FIRST(&filter_info->fivetuple_list))) {
1339                 TAILQ_REMOVE(&filter_info->fivetuple_list,
1340                              p_5tuple,
1341                              entries);
1342                 rte_free(p_5tuple);
1343         }
1344         memset(filter_info->fivetuple_mask, 0,
1345                sizeof(uint32_t) * IXGBE_5TUPLE_ARRAY_SIZE);
1346
1347         return 0;
1348 }
1349
1350 static int ixgbe_fdir_filter_uninit(struct rte_eth_dev *eth_dev)
1351 {
1352         struct ixgbe_hw_fdir_info *fdir_info =
1353                 IXGBE_DEV_PRIVATE_TO_FDIR_INFO(eth_dev->data->dev_private);
1354         struct ixgbe_fdir_filter *fdir_filter;
1355
1356                 if (fdir_info->hash_map)
1357                 rte_free(fdir_info->hash_map);
1358         if (fdir_info->hash_handle)
1359                 rte_hash_free(fdir_info->hash_handle);
1360
1361         while ((fdir_filter = TAILQ_FIRST(&fdir_info->fdir_list))) {
1362                 TAILQ_REMOVE(&fdir_info->fdir_list,
1363                              fdir_filter,
1364                              entries);
1365                 rte_free(fdir_filter);
1366         }
1367
1368         return 0;
1369 }
1370
1371 static int ixgbe_l2_tn_filter_uninit(struct rte_eth_dev *eth_dev)
1372 {
1373         struct ixgbe_l2_tn_info *l2_tn_info =
1374                 IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(eth_dev->data->dev_private);
1375         struct ixgbe_l2_tn_filter *l2_tn_filter;
1376
1377         if (l2_tn_info->hash_map)
1378                 rte_free(l2_tn_info->hash_map);
1379         if (l2_tn_info->hash_handle)
1380                 rte_hash_free(l2_tn_info->hash_handle);
1381
1382         while ((l2_tn_filter = TAILQ_FIRST(&l2_tn_info->l2_tn_list))) {
1383                 TAILQ_REMOVE(&l2_tn_info->l2_tn_list,
1384                              l2_tn_filter,
1385                              entries);
1386                 rte_free(l2_tn_filter);
1387         }
1388
1389         return 0;
1390 }
1391
1392 static int ixgbe_fdir_filter_init(struct rte_eth_dev *eth_dev)
1393 {
1394         struct ixgbe_hw_fdir_info *fdir_info =
1395                 IXGBE_DEV_PRIVATE_TO_FDIR_INFO(eth_dev->data->dev_private);
1396         char fdir_hash_name[RTE_HASH_NAMESIZE];
1397         struct rte_hash_parameters fdir_hash_params = {
1398                 .name = fdir_hash_name,
1399                 .entries = IXGBE_MAX_FDIR_FILTER_NUM,
1400                 .key_len = sizeof(union ixgbe_atr_input),
1401                 .hash_func = rte_hash_crc,
1402                 .hash_func_init_val = 0,
1403                 .socket_id = rte_socket_id(),
1404         };
1405
1406         TAILQ_INIT(&fdir_info->fdir_list);
1407         snprintf(fdir_hash_name, RTE_HASH_NAMESIZE,
1408                  "fdir_%s", eth_dev->device->name);
1409         fdir_info->hash_handle = rte_hash_create(&fdir_hash_params);
1410         if (!fdir_info->hash_handle) {
1411                 PMD_INIT_LOG(ERR, "Failed to create fdir hash table!");
1412                 return -EINVAL;
1413         }
1414         fdir_info->hash_map = rte_zmalloc("ixgbe",
1415                                           sizeof(struct ixgbe_fdir_filter *) *
1416                                           IXGBE_MAX_FDIR_FILTER_NUM,
1417                                           0);
1418         if (!fdir_info->hash_map) {
1419                 PMD_INIT_LOG(ERR,
1420                              "Failed to allocate memory for fdir hash map!");
1421                 return -ENOMEM;
1422         }
1423         fdir_info->mask_added = FALSE;
1424
1425         return 0;
1426 }
1427
1428 static int ixgbe_l2_tn_filter_init(struct rte_eth_dev *eth_dev)
1429 {
1430         struct ixgbe_l2_tn_info *l2_tn_info =
1431                 IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(eth_dev->data->dev_private);
1432         char l2_tn_hash_name[RTE_HASH_NAMESIZE];
1433         struct rte_hash_parameters l2_tn_hash_params = {
1434                 .name = l2_tn_hash_name,
1435                 .entries = IXGBE_MAX_L2_TN_FILTER_NUM,
1436                 .key_len = sizeof(struct ixgbe_l2_tn_key),
1437                 .hash_func = rte_hash_crc,
1438                 .hash_func_init_val = 0,
1439                 .socket_id = rte_socket_id(),
1440         };
1441
1442         TAILQ_INIT(&l2_tn_info->l2_tn_list);
1443         snprintf(l2_tn_hash_name, RTE_HASH_NAMESIZE,
1444                  "l2_tn_%s", eth_dev->device->name);
1445         l2_tn_info->hash_handle = rte_hash_create(&l2_tn_hash_params);
1446         if (!l2_tn_info->hash_handle) {
1447                 PMD_INIT_LOG(ERR, "Failed to create L2 TN hash table!");
1448                 return -EINVAL;
1449         }
1450         l2_tn_info->hash_map = rte_zmalloc("ixgbe",
1451                                    sizeof(struct ixgbe_l2_tn_filter *) *
1452                                    IXGBE_MAX_L2_TN_FILTER_NUM,
1453                                    0);
1454         if (!l2_tn_info->hash_map) {
1455                 PMD_INIT_LOG(ERR,
1456                         "Failed to allocate memory for L2 TN hash map!");
1457                 return -ENOMEM;
1458         }
1459         l2_tn_info->e_tag_en = FALSE;
1460         l2_tn_info->e_tag_fwd_en = FALSE;
1461         l2_tn_info->e_tag_ether_type = RTE_ETHER_TYPE_ETAG;
1462
1463         return 0;
1464 }
1465 /*
1466  * Negotiate mailbox API version with the PF.
1467  * After reset API version is always set to the basic one (ixgbe_mbox_api_10).
1468  * Then we try to negotiate starting with the most recent one.
1469  * If all negotiation attempts fail, then we will proceed with
1470  * the default one (ixgbe_mbox_api_10).
1471  */
1472 static void
1473 ixgbevf_negotiate_api(struct ixgbe_hw *hw)
1474 {
1475         int32_t i;
1476
1477         /* start with highest supported, proceed down */
1478         static const enum ixgbe_pfvf_api_rev sup_ver[] = {
1479                 ixgbe_mbox_api_13,
1480                 ixgbe_mbox_api_12,
1481                 ixgbe_mbox_api_11,
1482                 ixgbe_mbox_api_10,
1483         };
1484
1485         for (i = 0;
1486                         i != RTE_DIM(sup_ver) &&
1487                         ixgbevf_negotiate_api_version(hw, sup_ver[i]) != 0;
1488                         i++)
1489                 ;
1490 }
1491
1492 static void
1493 generate_random_mac_addr(struct rte_ether_addr *mac_addr)
1494 {
1495         uint64_t random;
1496
1497         /* Set Organizationally Unique Identifier (OUI) prefix. */
1498         mac_addr->addr_bytes[0] = 0x00;
1499         mac_addr->addr_bytes[1] = 0x09;
1500         mac_addr->addr_bytes[2] = 0xC0;
1501         /* Force indication of locally assigned MAC address. */
1502         mac_addr->addr_bytes[0] |= RTE_ETHER_LOCAL_ADMIN_ADDR;
1503         /* Generate the last 3 bytes of the MAC address with a random number. */
1504         random = rte_rand();
1505         memcpy(&mac_addr->addr_bytes[3], &random, 3);
1506 }
1507
1508 static int
1509 devarg_handle_int(__rte_unused const char *key, const char *value,
1510                   void *extra_args)
1511 {
1512         uint16_t *n = extra_args;
1513
1514         if (value == NULL || extra_args == NULL)
1515                 return -EINVAL;
1516
1517         *n = (uint16_t)strtoul(value, NULL, 0);
1518         if (*n == USHRT_MAX && errno == ERANGE)
1519                 return -1;
1520
1521         return 0;
1522 }
1523
1524 static void
1525 ixgbevf_parse_devargs(struct ixgbe_adapter *adapter,
1526                       struct rte_devargs *devargs)
1527 {
1528         struct rte_kvargs *kvlist;
1529         uint16_t pflink_fullchk;
1530
1531         if (devargs == NULL)
1532                 return;
1533
1534         kvlist = rte_kvargs_parse(devargs->args, ixgbevf_valid_arguments);
1535         if (kvlist == NULL)
1536                 return;
1537
1538         if (rte_kvargs_count(kvlist, IXGBEVF_DEVARG_PFLINK_FULLCHK) == 1 &&
1539             rte_kvargs_process(kvlist, IXGBEVF_DEVARG_PFLINK_FULLCHK,
1540                                devarg_handle_int, &pflink_fullchk) == 0 &&
1541             pflink_fullchk == 1)
1542                 adapter->pflink_fullchk = 1;
1543
1544         rte_kvargs_free(kvlist);
1545 }
1546
1547 /*
1548  * Virtual Function device init
1549  */
1550 static int
1551 eth_ixgbevf_dev_init(struct rte_eth_dev *eth_dev)
1552 {
1553         int diag;
1554         uint32_t tc, tcs;
1555         struct ixgbe_adapter *ad = eth_dev->data->dev_private;
1556         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
1557         struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
1558         struct ixgbe_hw *hw =
1559                 IXGBE_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
1560         struct ixgbe_vfta *shadow_vfta =
1561                 IXGBE_DEV_PRIVATE_TO_VFTA(eth_dev->data->dev_private);
1562         struct ixgbe_hwstrip *hwstrip =
1563                 IXGBE_DEV_PRIVATE_TO_HWSTRIP_BITMAP(eth_dev->data->dev_private);
1564         struct rte_ether_addr *perm_addr =
1565                 (struct rte_ether_addr *)hw->mac.perm_addr;
1566
1567         PMD_INIT_FUNC_TRACE();
1568
1569         eth_dev->dev_ops = &ixgbevf_eth_dev_ops;
1570         eth_dev->rx_descriptor_done   = ixgbe_dev_rx_descriptor_done;
1571         eth_dev->rx_descriptor_status = ixgbe_dev_rx_descriptor_status;
1572         eth_dev->tx_descriptor_status = ixgbe_dev_tx_descriptor_status;
1573         eth_dev->rx_pkt_burst = &ixgbe_recv_pkts;
1574         eth_dev->tx_pkt_burst = &ixgbe_xmit_pkts;
1575
1576         /* for secondary processes, we don't initialise any further as primary
1577          * has already done this work. Only check we don't need a different
1578          * RX function
1579          */
1580         if (rte_eal_process_type() != RTE_PROC_PRIMARY) {
1581                 struct ixgbe_tx_queue *txq;
1582                 /* TX queue function in primary, set by last queue initialized
1583                  * Tx queue may not initialized by primary process
1584                  */
1585                 if (eth_dev->data->tx_queues) {
1586                         txq = eth_dev->data->tx_queues[eth_dev->data->nb_tx_queues - 1];
1587                         ixgbe_set_tx_function(eth_dev, txq);
1588                 } else {
1589                         /* Use default TX function if we get here */
1590                         PMD_INIT_LOG(NOTICE,
1591                                      "No TX queues configured yet. Using default TX function.");
1592                 }
1593
1594                 ixgbe_set_rx_function(eth_dev);
1595
1596                 return 0;
1597         }
1598
1599         rte_atomic32_clear(&ad->link_thread_running);
1600         ixgbevf_parse_devargs(eth_dev->data->dev_private,
1601                               pci_dev->device.devargs);
1602
1603         rte_eth_copy_pci_info(eth_dev, pci_dev);
1604
1605         hw->device_id = pci_dev->id.device_id;
1606         hw->vendor_id = pci_dev->id.vendor_id;
1607         hw->hw_addr = (void *)pci_dev->mem_resource[0].addr;
1608
1609         /* initialize the vfta */
1610         memset(shadow_vfta, 0, sizeof(*shadow_vfta));
1611
1612         /* initialize the hw strip bitmap*/
1613         memset(hwstrip, 0, sizeof(*hwstrip));
1614
1615         /* Initialize the shared code (base driver) */
1616         diag = ixgbe_init_shared_code(hw);
1617         if (diag != IXGBE_SUCCESS) {
1618                 PMD_INIT_LOG(ERR, "Shared code init failed for ixgbevf: %d", diag);
1619                 return -EIO;
1620         }
1621
1622         /* init_mailbox_params */
1623         hw->mbx.ops.init_params(hw);
1624
1625         /* Reset the hw statistics */
1626         ixgbevf_dev_stats_reset(eth_dev);
1627
1628         /* Disable the interrupts for VF */
1629         ixgbevf_intr_disable(eth_dev);
1630
1631         hw->mac.num_rar_entries = 128; /* The MAX of the underlying PF */
1632         diag = hw->mac.ops.reset_hw(hw);
1633
1634         /*
1635          * The VF reset operation returns the IXGBE_ERR_INVALID_MAC_ADDR when
1636          * the underlying PF driver has not assigned a MAC address to the VF.
1637          * In this case, assign a random MAC address.
1638          */
1639         if ((diag != IXGBE_SUCCESS) && (diag != IXGBE_ERR_INVALID_MAC_ADDR)) {
1640                 PMD_INIT_LOG(ERR, "VF Initialization Failure: %d", diag);
1641                 /*
1642                  * This error code will be propagated to the app by
1643                  * rte_eth_dev_reset, so use a public error code rather than
1644                  * the internal-only IXGBE_ERR_RESET_FAILED
1645                  */
1646                 return -EAGAIN;
1647         }
1648
1649         /* negotiate mailbox API version to use with the PF. */
1650         ixgbevf_negotiate_api(hw);
1651
1652         /* Get Rx/Tx queue count via mailbox, which is ready after reset_hw */
1653         ixgbevf_get_queues(hw, &tcs, &tc);
1654
1655         /* Allocate memory for storing MAC addresses */
1656         eth_dev->data->mac_addrs = rte_zmalloc("ixgbevf", RTE_ETHER_ADDR_LEN *
1657                                                hw->mac.num_rar_entries, 0);
1658         if (eth_dev->data->mac_addrs == NULL) {
1659                 PMD_INIT_LOG(ERR,
1660                              "Failed to allocate %u bytes needed to store "
1661                              "MAC addresses",
1662                              RTE_ETHER_ADDR_LEN * hw->mac.num_rar_entries);
1663                 return -ENOMEM;
1664         }
1665
1666         /* Pass the information to the rte_eth_dev_close() that it should also
1667          * release the private port resources.
1668          */
1669         eth_dev->data->dev_flags |= RTE_ETH_DEV_CLOSE_REMOVE;
1670
1671         /* Generate a random MAC address, if none was assigned by PF. */
1672         if (rte_is_zero_ether_addr(perm_addr)) {
1673                 generate_random_mac_addr(perm_addr);
1674                 diag = ixgbe_set_rar_vf(hw, 1, perm_addr->addr_bytes, 0, 1);
1675                 if (diag) {
1676                         rte_free(eth_dev->data->mac_addrs);
1677                         eth_dev->data->mac_addrs = NULL;
1678                         return diag;
1679                 }
1680                 PMD_INIT_LOG(INFO, "\tVF MAC address not assigned by Host PF");
1681                 PMD_INIT_LOG(INFO, "\tAssign randomly generated MAC address "
1682                              "%02x:%02x:%02x:%02x:%02x:%02x",
1683                              perm_addr->addr_bytes[0],
1684                              perm_addr->addr_bytes[1],
1685                              perm_addr->addr_bytes[2],
1686                              perm_addr->addr_bytes[3],
1687                              perm_addr->addr_bytes[4],
1688                              perm_addr->addr_bytes[5]);
1689         }
1690
1691         /* Copy the permanent MAC address */
1692         rte_ether_addr_copy(perm_addr, &eth_dev->data->mac_addrs[0]);
1693
1694         /* reset the hardware with the new settings */
1695         diag = hw->mac.ops.start_hw(hw);
1696         switch (diag) {
1697         case  0:
1698                 break;
1699
1700         default:
1701                 PMD_INIT_LOG(ERR, "VF Initialization Failure: %d", diag);
1702                 return -EIO;
1703         }
1704
1705         rte_intr_callback_register(intr_handle,
1706                                    ixgbevf_dev_interrupt_handler, eth_dev);
1707         rte_intr_enable(intr_handle);
1708         ixgbevf_intr_enable(eth_dev);
1709
1710         PMD_INIT_LOG(DEBUG, "port %d vendorID=0x%x deviceID=0x%x mac.type=%s",
1711                      eth_dev->data->port_id, pci_dev->id.vendor_id,
1712                      pci_dev->id.device_id, "ixgbe_mac_82599_vf");
1713
1714         return 0;
1715 }
1716
1717 /* Virtual Function device uninit */
1718
1719 static int
1720 eth_ixgbevf_dev_uninit(struct rte_eth_dev *eth_dev)
1721 {
1722         PMD_INIT_FUNC_TRACE();
1723
1724         if (rte_eal_process_type() != RTE_PROC_PRIMARY)
1725                 return 0;
1726
1727         ixgbevf_dev_close(eth_dev);
1728
1729         return 0;
1730 }
1731
1732 static int
1733 eth_ixgbe_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
1734                 struct rte_pci_device *pci_dev)
1735 {
1736         char name[RTE_ETH_NAME_MAX_LEN];
1737         struct rte_eth_dev *pf_ethdev;
1738         struct rte_eth_devargs eth_da;
1739         int i, retval;
1740
1741         if (pci_dev->device.devargs) {
1742                 retval = rte_eth_devargs_parse(pci_dev->device.devargs->args,
1743                                 &eth_da);
1744                 if (retval)
1745                         return retval;
1746         } else
1747                 memset(&eth_da, 0, sizeof(eth_da));
1748
1749         retval = rte_eth_dev_create(&pci_dev->device, pci_dev->device.name,
1750                 sizeof(struct ixgbe_adapter),
1751                 eth_dev_pci_specific_init, pci_dev,
1752                 eth_ixgbe_dev_init, NULL);
1753
1754         if (retval || eth_da.nb_representor_ports < 1)
1755                 return retval;
1756
1757         pf_ethdev = rte_eth_dev_allocated(pci_dev->device.name);
1758         if (pf_ethdev == NULL)
1759                 return -ENODEV;
1760
1761         /* probe VF representor ports */
1762         for (i = 0; i < eth_da.nb_representor_ports; i++) {
1763                 struct ixgbe_vf_info *vfinfo;
1764                 struct ixgbe_vf_representor representor;
1765
1766                 vfinfo = *IXGBE_DEV_PRIVATE_TO_P_VFDATA(
1767                         pf_ethdev->data->dev_private);
1768                 if (vfinfo == NULL) {
1769                         PMD_DRV_LOG(ERR,
1770                                 "no virtual functions supported by PF");
1771                         break;
1772                 }
1773
1774                 representor.vf_id = eth_da.representor_ports[i];
1775                 representor.switch_domain_id = vfinfo->switch_domain_id;
1776                 representor.pf_ethdev = pf_ethdev;
1777
1778                 /* representor port net_bdf_port */
1779                 snprintf(name, sizeof(name), "net_%s_representor_%d",
1780                         pci_dev->device.name,
1781                         eth_da.representor_ports[i]);
1782
1783                 retval = rte_eth_dev_create(&pci_dev->device, name,
1784                         sizeof(struct ixgbe_vf_representor), NULL, NULL,
1785                         ixgbe_vf_representor_init, &representor);
1786
1787                 if (retval)
1788                         PMD_DRV_LOG(ERR, "failed to create ixgbe vf "
1789                                 "representor %s.", name);
1790         }
1791
1792         return 0;
1793 }
1794
1795 static int eth_ixgbe_pci_remove(struct rte_pci_device *pci_dev)
1796 {
1797         struct rte_eth_dev *ethdev;
1798
1799         ethdev = rte_eth_dev_allocated(pci_dev->device.name);
1800         if (!ethdev)
1801                 return 0;
1802
1803         if (ethdev->data->dev_flags & RTE_ETH_DEV_REPRESENTOR)
1804                 return rte_eth_dev_pci_generic_remove(pci_dev,
1805                                         ixgbe_vf_representor_uninit);
1806         else
1807                 return rte_eth_dev_pci_generic_remove(pci_dev,
1808                                                 eth_ixgbe_dev_uninit);
1809 }
1810
1811 static struct rte_pci_driver rte_ixgbe_pmd = {
1812         .id_table = pci_id_ixgbe_map,
1813         .drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC,
1814         .probe = eth_ixgbe_pci_probe,
1815         .remove = eth_ixgbe_pci_remove,
1816 };
1817
1818 static int eth_ixgbevf_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
1819         struct rte_pci_device *pci_dev)
1820 {
1821         return rte_eth_dev_pci_generic_probe(pci_dev,
1822                 sizeof(struct ixgbe_adapter), eth_ixgbevf_dev_init);
1823 }
1824
1825 static int eth_ixgbevf_pci_remove(struct rte_pci_device *pci_dev)
1826 {
1827         return rte_eth_dev_pci_generic_remove(pci_dev, eth_ixgbevf_dev_uninit);
1828 }
1829
1830 /*
1831  * virtual function driver struct
1832  */
1833 static struct rte_pci_driver rte_ixgbevf_pmd = {
1834         .id_table = pci_id_ixgbevf_map,
1835         .drv_flags = RTE_PCI_DRV_NEED_MAPPING,
1836         .probe = eth_ixgbevf_pci_probe,
1837         .remove = eth_ixgbevf_pci_remove,
1838 };
1839
1840 static int
1841 ixgbe_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
1842 {
1843         struct ixgbe_hw *hw =
1844                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1845         struct ixgbe_vfta *shadow_vfta =
1846                 IXGBE_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
1847         uint32_t vfta;
1848         uint32_t vid_idx;
1849         uint32_t vid_bit;
1850
1851         vid_idx = (uint32_t) ((vlan_id >> 5) & 0x7F);
1852         vid_bit = (uint32_t) (1 << (vlan_id & 0x1F));
1853         vfta = IXGBE_READ_REG(hw, IXGBE_VFTA(vid_idx));
1854         if (on)
1855                 vfta |= vid_bit;
1856         else
1857                 vfta &= ~vid_bit;
1858         IXGBE_WRITE_REG(hw, IXGBE_VFTA(vid_idx), vfta);
1859
1860         /* update local VFTA copy */
1861         shadow_vfta->vfta[vid_idx] = vfta;
1862
1863         return 0;
1864 }
1865
1866 static void
1867 ixgbe_vlan_strip_queue_set(struct rte_eth_dev *dev, uint16_t queue, int on)
1868 {
1869         if (on)
1870                 ixgbe_vlan_hw_strip_enable(dev, queue);
1871         else
1872                 ixgbe_vlan_hw_strip_disable(dev, queue);
1873 }
1874
1875 static int
1876 ixgbe_vlan_tpid_set(struct rte_eth_dev *dev,
1877                     enum rte_vlan_type vlan_type,
1878                     uint16_t tpid)
1879 {
1880         struct ixgbe_hw *hw =
1881                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1882         int ret = 0;
1883         uint32_t reg;
1884         uint32_t qinq;
1885
1886         qinq = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
1887         qinq &= IXGBE_DMATXCTL_GDV;
1888
1889         switch (vlan_type) {
1890         case ETH_VLAN_TYPE_INNER:
1891                 if (qinq) {
1892                         reg = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
1893                         reg = (reg & (~IXGBE_VLNCTRL_VET)) | (uint32_t)tpid;
1894                         IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, reg);
1895                         reg = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
1896                         reg = (reg & (~IXGBE_DMATXCTL_VT_MASK))
1897                                 | ((uint32_t)tpid << IXGBE_DMATXCTL_VT_SHIFT);
1898                         IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL, reg);
1899                 } else {
1900                         ret = -ENOTSUP;
1901                         PMD_DRV_LOG(ERR, "Inner type is not supported"
1902                                     " by single VLAN");
1903                 }
1904                 break;
1905         case ETH_VLAN_TYPE_OUTER:
1906                 if (qinq) {
1907                         /* Only the high 16-bits is valid */
1908                         IXGBE_WRITE_REG(hw, IXGBE_EXVET, (uint32_t)tpid <<
1909                                         IXGBE_EXVET_VET_EXT_SHIFT);
1910                 } else {
1911                         reg = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
1912                         reg = (reg & (~IXGBE_VLNCTRL_VET)) | (uint32_t)tpid;
1913                         IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, reg);
1914                         reg = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
1915                         reg = (reg & (~IXGBE_DMATXCTL_VT_MASK))
1916                                 | ((uint32_t)tpid << IXGBE_DMATXCTL_VT_SHIFT);
1917                         IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL, reg);
1918                 }
1919
1920                 break;
1921         default:
1922                 ret = -EINVAL;
1923                 PMD_DRV_LOG(ERR, "Unsupported VLAN type %d", vlan_type);
1924                 break;
1925         }
1926
1927         return ret;
1928 }
1929
1930 void
1931 ixgbe_vlan_hw_filter_disable(struct rte_eth_dev *dev)
1932 {
1933         struct ixgbe_hw *hw =
1934                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1935         uint32_t vlnctrl;
1936
1937         PMD_INIT_FUNC_TRACE();
1938
1939         /* Filter Table Disable */
1940         vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
1941         vlnctrl &= ~IXGBE_VLNCTRL_VFE;
1942
1943         IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
1944 }
1945
1946 void
1947 ixgbe_vlan_hw_filter_enable(struct rte_eth_dev *dev)
1948 {
1949         struct ixgbe_hw *hw =
1950                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1951         struct ixgbe_vfta *shadow_vfta =
1952                 IXGBE_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
1953         uint32_t vlnctrl;
1954         uint16_t i;
1955
1956         PMD_INIT_FUNC_TRACE();
1957
1958         /* Filter Table Enable */
1959         vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
1960         vlnctrl &= ~IXGBE_VLNCTRL_CFIEN;
1961         vlnctrl |= IXGBE_VLNCTRL_VFE;
1962
1963         IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
1964
1965         /* write whatever is in local vfta copy */
1966         for (i = 0; i < IXGBE_VFTA_SIZE; i++)
1967                 IXGBE_WRITE_REG(hw, IXGBE_VFTA(i), shadow_vfta->vfta[i]);
1968 }
1969
1970 static void
1971 ixgbe_vlan_hw_strip_bitmap_set(struct rte_eth_dev *dev, uint16_t queue, bool on)
1972 {
1973         struct ixgbe_hwstrip *hwstrip =
1974                 IXGBE_DEV_PRIVATE_TO_HWSTRIP_BITMAP(dev->data->dev_private);
1975         struct ixgbe_rx_queue *rxq;
1976
1977         if (queue >= IXGBE_MAX_RX_QUEUE_NUM)
1978                 return;
1979
1980         if (on)
1981                 IXGBE_SET_HWSTRIP(hwstrip, queue);
1982         else
1983                 IXGBE_CLEAR_HWSTRIP(hwstrip, queue);
1984
1985         if (queue >= dev->data->nb_rx_queues)
1986                 return;
1987
1988         rxq = dev->data->rx_queues[queue];
1989
1990         if (on) {
1991                 rxq->vlan_flags = PKT_RX_VLAN | PKT_RX_VLAN_STRIPPED;
1992                 rxq->offloads |= DEV_RX_OFFLOAD_VLAN_STRIP;
1993         } else {
1994                 rxq->vlan_flags = PKT_RX_VLAN;
1995                 rxq->offloads &= ~DEV_RX_OFFLOAD_VLAN_STRIP;
1996         }
1997 }
1998
1999 static void
2000 ixgbe_vlan_hw_strip_disable(struct rte_eth_dev *dev, uint16_t queue)
2001 {
2002         struct ixgbe_hw *hw =
2003                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2004         uint32_t ctrl;
2005
2006         PMD_INIT_FUNC_TRACE();
2007
2008         if (hw->mac.type == ixgbe_mac_82598EB) {
2009                 /* No queue level support */
2010                 PMD_INIT_LOG(NOTICE, "82598EB not support queue level hw strip");
2011                 return;
2012         }
2013
2014         /* Other 10G NIC, the VLAN strip can be setup per queue in RXDCTL */
2015         ctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(queue));
2016         ctrl &= ~IXGBE_RXDCTL_VME;
2017         IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(queue), ctrl);
2018
2019         /* record those setting for HW strip per queue */
2020         ixgbe_vlan_hw_strip_bitmap_set(dev, queue, 0);
2021 }
2022
2023 static void
2024 ixgbe_vlan_hw_strip_enable(struct rte_eth_dev *dev, uint16_t queue)
2025 {
2026         struct ixgbe_hw *hw =
2027                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2028         uint32_t ctrl;
2029
2030         PMD_INIT_FUNC_TRACE();
2031
2032         if (hw->mac.type == ixgbe_mac_82598EB) {
2033                 /* No queue level supported */
2034                 PMD_INIT_LOG(NOTICE, "82598EB not support queue level hw strip");
2035                 return;
2036         }
2037
2038         /* Other 10G NIC, the VLAN strip can be setup per queue in RXDCTL */
2039         ctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(queue));
2040         ctrl |= IXGBE_RXDCTL_VME;
2041         IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(queue), ctrl);
2042
2043         /* record those setting for HW strip per queue */
2044         ixgbe_vlan_hw_strip_bitmap_set(dev, queue, 1);
2045 }
2046
2047 static void
2048 ixgbe_vlan_hw_extend_disable(struct rte_eth_dev *dev)
2049 {
2050         struct ixgbe_hw *hw =
2051                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2052         uint32_t ctrl;
2053
2054         PMD_INIT_FUNC_TRACE();
2055
2056         /* DMATXCTRL: Geric Double VLAN Disable */
2057         ctrl = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
2058         ctrl &= ~IXGBE_DMATXCTL_GDV;
2059         IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL, ctrl);
2060
2061         /* CTRL_EXT: Global Double VLAN Disable */
2062         ctrl = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
2063         ctrl &= ~IXGBE_EXTENDED_VLAN;
2064         IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl);
2065
2066 }
2067
2068 static void
2069 ixgbe_vlan_hw_extend_enable(struct rte_eth_dev *dev)
2070 {
2071         struct ixgbe_hw *hw =
2072                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2073         uint32_t ctrl;
2074
2075         PMD_INIT_FUNC_TRACE();
2076
2077         /* DMATXCTRL: Geric Double VLAN Enable */
2078         ctrl  = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
2079         ctrl |= IXGBE_DMATXCTL_GDV;
2080         IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL, ctrl);
2081
2082         /* CTRL_EXT: Global Double VLAN Enable */
2083         ctrl  = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
2084         ctrl |= IXGBE_EXTENDED_VLAN;
2085         IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl);
2086
2087         /* Clear pooling mode of PFVTCTL. It's required by X550. */
2088         if (hw->mac.type == ixgbe_mac_X550 ||
2089             hw->mac.type == ixgbe_mac_X550EM_x ||
2090             hw->mac.type == ixgbe_mac_X550EM_a) {
2091                 ctrl = IXGBE_READ_REG(hw, IXGBE_VT_CTL);
2092                 ctrl &= ~IXGBE_VT_CTL_POOLING_MODE_MASK;
2093                 IXGBE_WRITE_REG(hw, IXGBE_VT_CTL, ctrl);
2094         }
2095
2096         /*
2097          * VET EXT field in the EXVET register = 0x8100 by default
2098          * So no need to change. Same to VT field of DMATXCTL register
2099          */
2100 }
2101
2102 void
2103 ixgbe_vlan_hw_strip_config(struct rte_eth_dev *dev)
2104 {
2105         struct ixgbe_hw *hw =
2106                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2107         struct rte_eth_rxmode *rxmode = &dev->data->dev_conf.rxmode;
2108         uint32_t ctrl;
2109         uint16_t i;
2110         struct ixgbe_rx_queue *rxq;
2111         bool on;
2112
2113         PMD_INIT_FUNC_TRACE();
2114
2115         if (hw->mac.type == ixgbe_mac_82598EB) {
2116                 if (rxmode->offloads & DEV_RX_OFFLOAD_VLAN_STRIP) {
2117                         ctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
2118                         ctrl |= IXGBE_VLNCTRL_VME;
2119                         IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, ctrl);
2120                 } else {
2121                         ctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
2122                         ctrl &= ~IXGBE_VLNCTRL_VME;
2123                         IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, ctrl);
2124                 }
2125         } else {
2126                 /*
2127                  * Other 10G NIC, the VLAN strip can be setup
2128                  * per queue in RXDCTL
2129                  */
2130                 for (i = 0; i < dev->data->nb_rx_queues; i++) {
2131                         rxq = dev->data->rx_queues[i];
2132                         ctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(rxq->reg_idx));
2133                         if (rxq->offloads & DEV_RX_OFFLOAD_VLAN_STRIP) {
2134                                 ctrl |= IXGBE_RXDCTL_VME;
2135                                 on = TRUE;
2136                         } else {
2137                                 ctrl &= ~IXGBE_RXDCTL_VME;
2138                                 on = FALSE;
2139                         }
2140                         IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(rxq->reg_idx), ctrl);
2141
2142                         /* record those setting for HW strip per queue */
2143                         ixgbe_vlan_hw_strip_bitmap_set(dev, i, on);
2144                 }
2145         }
2146 }
2147
2148 static void
2149 ixgbe_config_vlan_strip_on_all_queues(struct rte_eth_dev *dev, int mask)
2150 {
2151         uint16_t i;
2152         struct rte_eth_rxmode *rxmode;
2153         struct ixgbe_rx_queue *rxq;
2154
2155         if (mask & ETH_VLAN_STRIP_MASK) {
2156                 rxmode = &dev->data->dev_conf.rxmode;
2157                 if (rxmode->offloads & DEV_RX_OFFLOAD_VLAN_STRIP)
2158                         for (i = 0; i < dev->data->nb_rx_queues; i++) {
2159                                 rxq = dev->data->rx_queues[i];
2160                                 rxq->offloads |= DEV_RX_OFFLOAD_VLAN_STRIP;
2161                         }
2162                 else
2163                         for (i = 0; i < dev->data->nb_rx_queues; i++) {
2164                                 rxq = dev->data->rx_queues[i];
2165                                 rxq->offloads &= ~DEV_RX_OFFLOAD_VLAN_STRIP;
2166                         }
2167         }
2168 }
2169
2170 static int
2171 ixgbe_vlan_offload_config(struct rte_eth_dev *dev, int mask)
2172 {
2173         struct rte_eth_rxmode *rxmode;
2174         rxmode = &dev->data->dev_conf.rxmode;
2175
2176         if (mask & ETH_VLAN_STRIP_MASK) {
2177                 ixgbe_vlan_hw_strip_config(dev);
2178         }
2179
2180         if (mask & ETH_VLAN_FILTER_MASK) {
2181                 if (rxmode->offloads & DEV_RX_OFFLOAD_VLAN_FILTER)
2182                         ixgbe_vlan_hw_filter_enable(dev);
2183                 else
2184                         ixgbe_vlan_hw_filter_disable(dev);
2185         }
2186
2187         if (mask & ETH_VLAN_EXTEND_MASK) {
2188                 if (rxmode->offloads & DEV_RX_OFFLOAD_VLAN_EXTEND)
2189                         ixgbe_vlan_hw_extend_enable(dev);
2190                 else
2191                         ixgbe_vlan_hw_extend_disable(dev);
2192         }
2193
2194         return 0;
2195 }
2196
2197 static int
2198 ixgbe_vlan_offload_set(struct rte_eth_dev *dev, int mask)
2199 {
2200         ixgbe_config_vlan_strip_on_all_queues(dev, mask);
2201
2202         ixgbe_vlan_offload_config(dev, mask);
2203
2204         return 0;
2205 }
2206
2207 static void
2208 ixgbe_vmdq_vlan_hw_filter_enable(struct rte_eth_dev *dev)
2209 {
2210         struct ixgbe_hw *hw =
2211                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2212         /* VLNCTRL: enable vlan filtering and allow all vlan tags through */
2213         uint32_t vlanctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
2214
2215         vlanctrl |= IXGBE_VLNCTRL_VFE; /* enable vlan filters */
2216         IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlanctrl);
2217 }
2218
2219 static int
2220 ixgbe_check_vf_rss_rxq_num(struct rte_eth_dev *dev, uint16_t nb_rx_q)
2221 {
2222         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
2223
2224         switch (nb_rx_q) {
2225         case 1:
2226         case 2:
2227                 RTE_ETH_DEV_SRIOV(dev).active = ETH_64_POOLS;
2228                 break;
2229         case 4:
2230                 RTE_ETH_DEV_SRIOV(dev).active = ETH_32_POOLS;
2231                 break;
2232         default:
2233                 return -EINVAL;
2234         }
2235
2236         RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool =
2237                 IXGBE_MAX_RX_QUEUE_NUM / RTE_ETH_DEV_SRIOV(dev).active;
2238         RTE_ETH_DEV_SRIOV(dev).def_pool_q_idx =
2239                 pci_dev->max_vfs * RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool;
2240         return 0;
2241 }
2242
2243 static int
2244 ixgbe_check_mq_mode(struct rte_eth_dev *dev)
2245 {
2246         struct rte_eth_conf *dev_conf = &dev->data->dev_conf;
2247         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2248         uint16_t nb_rx_q = dev->data->nb_rx_queues;
2249         uint16_t nb_tx_q = dev->data->nb_tx_queues;
2250
2251         if (RTE_ETH_DEV_SRIOV(dev).active != 0) {
2252                 /* check multi-queue mode */
2253                 switch (dev_conf->rxmode.mq_mode) {
2254                 case ETH_MQ_RX_VMDQ_DCB:
2255                         PMD_INIT_LOG(INFO, "ETH_MQ_RX_VMDQ_DCB mode supported in SRIOV");
2256                         break;
2257                 case ETH_MQ_RX_VMDQ_DCB_RSS:
2258                         /* DCB/RSS VMDQ in SRIOV mode, not implement yet */
2259                         PMD_INIT_LOG(ERR, "SRIOV active,"
2260                                         " unsupported mq_mode rx %d.",
2261                                         dev_conf->rxmode.mq_mode);
2262                         return -EINVAL;
2263                 case ETH_MQ_RX_RSS:
2264                 case ETH_MQ_RX_VMDQ_RSS:
2265                         dev->data->dev_conf.rxmode.mq_mode = ETH_MQ_RX_VMDQ_RSS;
2266                         if (nb_rx_q <= RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool)
2267                                 if (ixgbe_check_vf_rss_rxq_num(dev, nb_rx_q)) {
2268                                         PMD_INIT_LOG(ERR, "SRIOV is active,"
2269                                                 " invalid queue number"
2270                                                 " for VMDQ RSS, allowed"
2271                                                 " value are 1, 2 or 4.");
2272                                         return -EINVAL;
2273                                 }
2274                         break;
2275                 case ETH_MQ_RX_VMDQ_ONLY:
2276                 case ETH_MQ_RX_NONE:
2277                         /* if nothing mq mode configure, use default scheme */
2278                         dev->data->dev_conf.rxmode.mq_mode = ETH_MQ_RX_VMDQ_ONLY;
2279                         break;
2280                 default: /* ETH_MQ_RX_DCB, ETH_MQ_RX_DCB_RSS or ETH_MQ_TX_DCB*/
2281                         /* SRIOV only works in VMDq enable mode */
2282                         PMD_INIT_LOG(ERR, "SRIOV is active,"
2283                                         " wrong mq_mode rx %d.",
2284                                         dev_conf->rxmode.mq_mode);
2285                         return -EINVAL;
2286                 }
2287
2288                 switch (dev_conf->txmode.mq_mode) {
2289                 case ETH_MQ_TX_VMDQ_DCB:
2290                         PMD_INIT_LOG(INFO, "ETH_MQ_TX_VMDQ_DCB mode supported in SRIOV");
2291                         dev->data->dev_conf.txmode.mq_mode = ETH_MQ_TX_VMDQ_DCB;
2292                         break;
2293                 default: /* ETH_MQ_TX_VMDQ_ONLY or ETH_MQ_TX_NONE */
2294                         dev->data->dev_conf.txmode.mq_mode = ETH_MQ_TX_VMDQ_ONLY;
2295                         break;
2296                 }
2297
2298                 /* check valid queue number */
2299                 if ((nb_rx_q > RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool) ||
2300                     (nb_tx_q > RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool)) {
2301                         PMD_INIT_LOG(ERR, "SRIOV is active,"
2302                                         " nb_rx_q=%d nb_tx_q=%d queue number"
2303                                         " must be less than or equal to %d.",
2304                                         nb_rx_q, nb_tx_q,
2305                                         RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool);
2306                         return -EINVAL;
2307                 }
2308         } else {
2309                 if (dev_conf->rxmode.mq_mode == ETH_MQ_RX_VMDQ_DCB_RSS) {
2310                         PMD_INIT_LOG(ERR, "VMDQ+DCB+RSS mq_mode is"
2311                                           " not supported.");
2312                         return -EINVAL;
2313                 }
2314                 /* check configuration for vmdb+dcb mode */
2315                 if (dev_conf->rxmode.mq_mode == ETH_MQ_RX_VMDQ_DCB) {
2316                         const struct rte_eth_vmdq_dcb_conf *conf;
2317
2318                         if (nb_rx_q != IXGBE_VMDQ_DCB_NB_QUEUES) {
2319                                 PMD_INIT_LOG(ERR, "VMDQ+DCB, nb_rx_q != %d.",
2320                                                 IXGBE_VMDQ_DCB_NB_QUEUES);
2321                                 return -EINVAL;
2322                         }
2323                         conf = &dev_conf->rx_adv_conf.vmdq_dcb_conf;
2324                         if (!(conf->nb_queue_pools == ETH_16_POOLS ||
2325                                conf->nb_queue_pools == ETH_32_POOLS)) {
2326                                 PMD_INIT_LOG(ERR, "VMDQ+DCB selected,"
2327                                                 " nb_queue_pools must be %d or %d.",
2328                                                 ETH_16_POOLS, ETH_32_POOLS);
2329                                 return -EINVAL;
2330                         }
2331                 }
2332                 if (dev_conf->txmode.mq_mode == ETH_MQ_TX_VMDQ_DCB) {
2333                         const struct rte_eth_vmdq_dcb_tx_conf *conf;
2334
2335                         if (nb_tx_q != IXGBE_VMDQ_DCB_NB_QUEUES) {
2336                                 PMD_INIT_LOG(ERR, "VMDQ+DCB, nb_tx_q != %d",
2337                                                  IXGBE_VMDQ_DCB_NB_QUEUES);
2338                                 return -EINVAL;
2339                         }
2340                         conf = &dev_conf->tx_adv_conf.vmdq_dcb_tx_conf;
2341                         if (!(conf->nb_queue_pools == ETH_16_POOLS ||
2342                                conf->nb_queue_pools == ETH_32_POOLS)) {
2343                                 PMD_INIT_LOG(ERR, "VMDQ+DCB selected,"
2344                                                 " nb_queue_pools != %d and"
2345                                                 " nb_queue_pools != %d.",
2346                                                 ETH_16_POOLS, ETH_32_POOLS);
2347                                 return -EINVAL;
2348                         }
2349                 }
2350
2351                 /* For DCB mode check our configuration before we go further */
2352                 if (dev_conf->rxmode.mq_mode == ETH_MQ_RX_DCB) {
2353                         const struct rte_eth_dcb_rx_conf *conf;
2354
2355                         conf = &dev_conf->rx_adv_conf.dcb_rx_conf;
2356                         if (!(conf->nb_tcs == ETH_4_TCS ||
2357                                conf->nb_tcs == ETH_8_TCS)) {
2358                                 PMD_INIT_LOG(ERR, "DCB selected, nb_tcs != %d"
2359                                                 " and nb_tcs != %d.",
2360                                                 ETH_4_TCS, ETH_8_TCS);
2361                                 return -EINVAL;
2362                         }
2363                 }
2364
2365                 if (dev_conf->txmode.mq_mode == ETH_MQ_TX_DCB) {
2366                         const struct rte_eth_dcb_tx_conf *conf;
2367
2368                         conf = &dev_conf->tx_adv_conf.dcb_tx_conf;
2369                         if (!(conf->nb_tcs == ETH_4_TCS ||
2370                                conf->nb_tcs == ETH_8_TCS)) {
2371                                 PMD_INIT_LOG(ERR, "DCB selected, nb_tcs != %d"
2372                                                 " and nb_tcs != %d.",
2373                                                 ETH_4_TCS, ETH_8_TCS);
2374                                 return -EINVAL;
2375                         }
2376                 }
2377
2378                 /*
2379                  * When DCB/VT is off, maximum number of queues changes,
2380                  * except for 82598EB, which remains constant.
2381                  */
2382                 if (dev_conf->txmode.mq_mode == ETH_MQ_TX_NONE &&
2383                                 hw->mac.type != ixgbe_mac_82598EB) {
2384                         if (nb_tx_q > IXGBE_NONE_MODE_TX_NB_QUEUES) {
2385                                 PMD_INIT_LOG(ERR,
2386                                              "Neither VT nor DCB are enabled, "
2387                                              "nb_tx_q > %d.",
2388                                              IXGBE_NONE_MODE_TX_NB_QUEUES);
2389                                 return -EINVAL;
2390                         }
2391                 }
2392         }
2393         return 0;
2394 }
2395
2396 static int
2397 ixgbe_dev_configure(struct rte_eth_dev *dev)
2398 {
2399         struct ixgbe_interrupt *intr =
2400                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
2401         struct ixgbe_adapter *adapter = dev->data->dev_private;
2402         int ret;
2403
2404         PMD_INIT_FUNC_TRACE();
2405
2406         if (dev->data->dev_conf.rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG)
2407                 dev->data->dev_conf.rxmode.offloads |= DEV_RX_OFFLOAD_RSS_HASH;
2408
2409         /* multipe queue mode checking */
2410         ret  = ixgbe_check_mq_mode(dev);
2411         if (ret != 0) {
2412                 PMD_DRV_LOG(ERR, "ixgbe_check_mq_mode fails with %d.",
2413                             ret);
2414                 return ret;
2415         }
2416
2417         /* set flag to update link status after init */
2418         intr->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
2419
2420         /*
2421          * Initialize to TRUE. If any of Rx queues doesn't meet the bulk
2422          * allocation or vector Rx preconditions we will reset it.
2423          */
2424         adapter->rx_bulk_alloc_allowed = true;
2425         adapter->rx_vec_allowed = true;
2426
2427         return 0;
2428 }
2429
2430 static void
2431 ixgbe_dev_phy_intr_setup(struct rte_eth_dev *dev)
2432 {
2433         struct ixgbe_hw *hw =
2434                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2435         struct ixgbe_interrupt *intr =
2436                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
2437         uint32_t gpie;
2438
2439         /* only set up it on X550EM_X */
2440         if (hw->mac.type == ixgbe_mac_X550EM_x) {
2441                 gpie = IXGBE_READ_REG(hw, IXGBE_GPIE);
2442                 gpie |= IXGBE_SDP0_GPIEN_X550EM_x;
2443                 IXGBE_WRITE_REG(hw, IXGBE_GPIE, gpie);
2444                 if (hw->phy.type == ixgbe_phy_x550em_ext_t)
2445                         intr->mask |= IXGBE_EICR_GPI_SDP0_X550EM_x;
2446         }
2447 }
2448
2449 int
2450 ixgbe_set_vf_rate_limit(struct rte_eth_dev *dev, uint16_t vf,
2451                         uint16_t tx_rate, uint64_t q_msk)
2452 {
2453         struct ixgbe_hw *hw;
2454         struct ixgbe_vf_info *vfinfo;
2455         struct rte_eth_link link;
2456         uint8_t  nb_q_per_pool;
2457         uint32_t queue_stride;
2458         uint32_t queue_idx, idx = 0, vf_idx;
2459         uint32_t queue_end;
2460         uint16_t total_rate = 0;
2461         struct rte_pci_device *pci_dev;
2462         int ret;
2463
2464         pci_dev = RTE_ETH_DEV_TO_PCI(dev);
2465         ret = rte_eth_link_get_nowait(dev->data->port_id, &link);
2466         if (ret < 0)
2467                 return ret;
2468
2469         if (vf >= pci_dev->max_vfs)
2470                 return -EINVAL;
2471
2472         if (tx_rate > link.link_speed)
2473                 return -EINVAL;
2474
2475         if (q_msk == 0)
2476                 return 0;
2477
2478         hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2479         vfinfo = *(IXGBE_DEV_PRIVATE_TO_P_VFDATA(dev->data->dev_private));
2480         nb_q_per_pool = RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool;
2481         queue_stride = IXGBE_MAX_RX_QUEUE_NUM / RTE_ETH_DEV_SRIOV(dev).active;
2482         queue_idx = vf * queue_stride;
2483         queue_end = queue_idx + nb_q_per_pool - 1;
2484         if (queue_end >= hw->mac.max_tx_queues)
2485                 return -EINVAL;
2486
2487         if (vfinfo) {
2488                 for (vf_idx = 0; vf_idx < pci_dev->max_vfs; vf_idx++) {
2489                         if (vf_idx == vf)
2490                                 continue;
2491                         for (idx = 0; idx < RTE_DIM(vfinfo[vf_idx].tx_rate);
2492                                 idx++)
2493                                 total_rate += vfinfo[vf_idx].tx_rate[idx];
2494                 }
2495         } else {
2496                 return -EINVAL;
2497         }
2498
2499         /* Store tx_rate for this vf. */
2500         for (idx = 0; idx < nb_q_per_pool; idx++) {
2501                 if (((uint64_t)0x1 << idx) & q_msk) {
2502                         if (vfinfo[vf].tx_rate[idx] != tx_rate)
2503                                 vfinfo[vf].tx_rate[idx] = tx_rate;
2504                         total_rate += tx_rate;
2505                 }
2506         }
2507
2508         if (total_rate > dev->data->dev_link.link_speed) {
2509                 /* Reset stored TX rate of the VF if it causes exceed
2510                  * link speed.
2511                  */
2512                 memset(vfinfo[vf].tx_rate, 0, sizeof(vfinfo[vf].tx_rate));
2513                 return -EINVAL;
2514         }
2515
2516         /* Set RTTBCNRC of each queue/pool for vf X  */
2517         for (; queue_idx <= queue_end; queue_idx++) {
2518                 if (0x1 & q_msk)
2519                         ixgbe_set_queue_rate_limit(dev, queue_idx, tx_rate);
2520                 q_msk = q_msk >> 1;
2521         }
2522
2523         return 0;
2524 }
2525
2526 static int
2527 ixgbe_flow_ctrl_enable(struct rte_eth_dev *dev, struct ixgbe_hw *hw)
2528 {
2529         struct ixgbe_adapter *adapter = dev->data->dev_private;
2530         int err;
2531         uint32_t mflcn;
2532
2533         ixgbe_setup_fc(hw);
2534
2535         err = ixgbe_fc_enable(hw);
2536
2537         /* Not negotiated is not an error case */
2538         if (err == IXGBE_SUCCESS || err == IXGBE_ERR_FC_NOT_NEGOTIATED) {
2539                 /*
2540                  *check if we want to forward MAC frames - driver doesn't
2541                  *have native capability to do that,
2542                  *so we'll write the registers ourselves
2543                  */
2544
2545                 mflcn = IXGBE_READ_REG(hw, IXGBE_MFLCN);
2546
2547                 /* set or clear MFLCN.PMCF bit depending on configuration */
2548                 if (adapter->mac_ctrl_frame_fwd != 0)
2549                         mflcn |= IXGBE_MFLCN_PMCF;
2550                 else
2551                         mflcn &= ~IXGBE_MFLCN_PMCF;
2552
2553                 IXGBE_WRITE_REG(hw, IXGBE_MFLCN, mflcn);
2554                 IXGBE_WRITE_FLUSH(hw);
2555
2556                 return 0;
2557         }
2558         return err;
2559 }
2560
2561 /*
2562  * Configure device link speed and setup link.
2563  * It returns 0 on success.
2564  */
2565 static int
2566 ixgbe_dev_start(struct rte_eth_dev *dev)
2567 {
2568         struct ixgbe_hw *hw =
2569                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2570         struct ixgbe_vf_info *vfinfo =
2571                 *IXGBE_DEV_PRIVATE_TO_P_VFDATA(dev->data->dev_private);
2572         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
2573         struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
2574         uint32_t intr_vector = 0;
2575         int err;
2576         bool link_up = false, negotiate = 0;
2577         uint32_t speed = 0;
2578         uint32_t allowed_speeds = 0;
2579         int mask = 0;
2580         int status;
2581         uint16_t vf, idx;
2582         uint32_t *link_speeds;
2583         struct ixgbe_tm_conf *tm_conf =
2584                 IXGBE_DEV_PRIVATE_TO_TM_CONF(dev->data->dev_private);
2585         struct ixgbe_macsec_setting *macsec_setting =
2586                 IXGBE_DEV_PRIVATE_TO_MACSEC_SETTING(dev->data->dev_private);
2587
2588         PMD_INIT_FUNC_TRACE();
2589
2590         /* Stop the link setup handler before resetting the HW. */
2591         ixgbe_dev_wait_setup_link_complete(dev, 0);
2592
2593         /* disable uio/vfio intr/eventfd mapping */
2594         rte_intr_disable(intr_handle);
2595
2596         /* stop adapter */
2597         hw->adapter_stopped = 0;
2598         ixgbe_stop_adapter(hw);
2599
2600         /* reinitialize adapter
2601          * this calls reset and start
2602          */
2603         status = ixgbe_pf_reset_hw(hw);
2604         if (status != 0)
2605                 return -1;
2606         hw->mac.ops.start_hw(hw);
2607         hw->mac.get_link_status = true;
2608
2609         /* configure PF module if SRIOV enabled */
2610         ixgbe_pf_host_configure(dev);
2611
2612         ixgbe_dev_phy_intr_setup(dev);
2613
2614         /* check and configure queue intr-vector mapping */
2615         if ((rte_intr_cap_multiple(intr_handle) ||
2616              !RTE_ETH_DEV_SRIOV(dev).active) &&
2617             dev->data->dev_conf.intr_conf.rxq != 0) {
2618                 intr_vector = dev->data->nb_rx_queues;
2619                 if (intr_vector > IXGBE_MAX_INTR_QUEUE_NUM) {
2620                         PMD_INIT_LOG(ERR, "At most %d intr queues supported",
2621                                         IXGBE_MAX_INTR_QUEUE_NUM);
2622                         return -ENOTSUP;
2623                 }
2624                 if (rte_intr_efd_enable(intr_handle, intr_vector))
2625                         return -1;
2626         }
2627
2628         if (rte_intr_dp_is_en(intr_handle) && !intr_handle->intr_vec) {
2629                 intr_handle->intr_vec =
2630                         rte_zmalloc("intr_vec",
2631                                     dev->data->nb_rx_queues * sizeof(int), 0);
2632                 if (intr_handle->intr_vec == NULL) {
2633                         PMD_INIT_LOG(ERR, "Failed to allocate %d rx_queues"
2634                                      " intr_vec", dev->data->nb_rx_queues);
2635                         return -ENOMEM;
2636                 }
2637         }
2638
2639         /* confiugre msix for sleep until rx interrupt */
2640         ixgbe_configure_msix(dev);
2641
2642         /* initialize transmission unit */
2643         ixgbe_dev_tx_init(dev);
2644
2645         /* This can fail when allocating mbufs for descriptor rings */
2646         err = ixgbe_dev_rx_init(dev);
2647         if (err) {
2648                 PMD_INIT_LOG(ERR, "Unable to initialize RX hardware");
2649                 goto error;
2650         }
2651
2652         mask = ETH_VLAN_STRIP_MASK | ETH_VLAN_FILTER_MASK |
2653                 ETH_VLAN_EXTEND_MASK;
2654         err = ixgbe_vlan_offload_config(dev, mask);
2655         if (err) {
2656                 PMD_INIT_LOG(ERR, "Unable to set VLAN offload");
2657                 goto error;
2658         }
2659
2660         if (dev->data->dev_conf.rxmode.mq_mode == ETH_MQ_RX_VMDQ_ONLY) {
2661                 /* Enable vlan filtering for VMDq */
2662                 ixgbe_vmdq_vlan_hw_filter_enable(dev);
2663         }
2664
2665         /* Configure DCB hw */
2666         ixgbe_configure_dcb(dev);
2667
2668         if (dev->data->dev_conf.fdir_conf.mode != RTE_FDIR_MODE_NONE) {
2669                 err = ixgbe_fdir_configure(dev);
2670                 if (err)
2671                         goto error;
2672         }
2673
2674         /* Restore vf rate limit */
2675         if (vfinfo != NULL) {
2676                 for (vf = 0; vf < pci_dev->max_vfs; vf++)
2677                         for (idx = 0; idx < IXGBE_MAX_QUEUE_NUM_PER_VF; idx++)
2678                                 if (vfinfo[vf].tx_rate[idx] != 0)
2679                                         ixgbe_set_vf_rate_limit(
2680                                                 dev, vf,
2681                                                 vfinfo[vf].tx_rate[idx],
2682                                                 1 << idx);
2683         }
2684
2685         ixgbe_restore_statistics_mapping(dev);
2686
2687         err = ixgbe_flow_ctrl_enable(dev, hw);
2688         if (err < 0) {
2689                 PMD_INIT_LOG(ERR, "enable flow ctrl err");
2690                 goto error;
2691         }
2692
2693         err = ixgbe_dev_rxtx_start(dev);
2694         if (err < 0) {
2695                 PMD_INIT_LOG(ERR, "Unable to start rxtx queues");
2696                 goto error;
2697         }
2698
2699         /* Skip link setup if loopback mode is enabled. */
2700         if (dev->data->dev_conf.lpbk_mode != 0) {
2701                 err = ixgbe_check_supported_loopback_mode(dev);
2702                 if (err < 0) {
2703                         PMD_INIT_LOG(ERR, "Unsupported loopback mode");
2704                         goto error;
2705                 } else {
2706                         goto skip_link_setup;
2707                 }
2708         }
2709
2710         if (ixgbe_is_sfp(hw) && hw->phy.multispeed_fiber) {
2711                 err = hw->mac.ops.setup_sfp(hw);
2712                 if (err)
2713                         goto error;
2714         }
2715
2716         if (hw->mac.ops.get_media_type(hw) == ixgbe_media_type_copper) {
2717                 /* Turn on the copper */
2718                 ixgbe_set_phy_power(hw, true);
2719         } else {
2720                 /* Turn on the laser */
2721                 ixgbe_enable_tx_laser(hw);
2722         }
2723
2724         err = ixgbe_check_link(hw, &speed, &link_up, 0);
2725         if (err)
2726                 goto error;
2727         dev->data->dev_link.link_status = link_up;
2728
2729         err = ixgbe_get_link_capabilities(hw, &speed, &negotiate);
2730         if (err)
2731                 goto error;
2732
2733         switch (hw->mac.type) {
2734         case ixgbe_mac_X550:
2735         case ixgbe_mac_X550EM_x:
2736         case ixgbe_mac_X550EM_a:
2737                 allowed_speeds = ETH_LINK_SPEED_100M | ETH_LINK_SPEED_1G |
2738                         ETH_LINK_SPEED_2_5G |  ETH_LINK_SPEED_5G |
2739                         ETH_LINK_SPEED_10G;
2740                 if (hw->device_id == IXGBE_DEV_ID_X550EM_A_1G_T ||
2741                                 hw->device_id == IXGBE_DEV_ID_X550EM_A_1G_T_L)
2742                         allowed_speeds = ETH_LINK_SPEED_10M |
2743                                 ETH_LINK_SPEED_100M | ETH_LINK_SPEED_1G;
2744                 break;
2745         default:
2746                 allowed_speeds = ETH_LINK_SPEED_100M | ETH_LINK_SPEED_1G |
2747                         ETH_LINK_SPEED_10G;
2748         }
2749
2750         link_speeds = &dev->data->dev_conf.link_speeds;
2751
2752         /* Ignore autoneg flag bit and check the validity of
2753          * link_speed
2754          */
2755         if (((*link_speeds) >> 1) & ~(allowed_speeds >> 1)) {
2756                 PMD_INIT_LOG(ERR, "Invalid link setting");
2757                 goto error;
2758         }
2759
2760         speed = 0x0;
2761         if (*link_speeds == ETH_LINK_SPEED_AUTONEG) {
2762                 switch (hw->mac.type) {
2763                 case ixgbe_mac_82598EB:
2764                         speed = IXGBE_LINK_SPEED_82598_AUTONEG;
2765                         break;
2766                 case ixgbe_mac_82599EB:
2767                 case ixgbe_mac_X540:
2768                         speed = IXGBE_LINK_SPEED_82599_AUTONEG;
2769                         break;
2770                 case ixgbe_mac_X550:
2771                 case ixgbe_mac_X550EM_x:
2772                 case ixgbe_mac_X550EM_a:
2773                         speed = IXGBE_LINK_SPEED_X550_AUTONEG;
2774                         break;
2775                 default:
2776                         speed = IXGBE_LINK_SPEED_82599_AUTONEG;
2777                 }
2778         } else {
2779                 if (*link_speeds & ETH_LINK_SPEED_10G)
2780                         speed |= IXGBE_LINK_SPEED_10GB_FULL;
2781                 if (*link_speeds & ETH_LINK_SPEED_5G)
2782                         speed |= IXGBE_LINK_SPEED_5GB_FULL;
2783                 if (*link_speeds & ETH_LINK_SPEED_2_5G)
2784                         speed |= IXGBE_LINK_SPEED_2_5GB_FULL;
2785                 if (*link_speeds & ETH_LINK_SPEED_1G)
2786                         speed |= IXGBE_LINK_SPEED_1GB_FULL;
2787                 if (*link_speeds & ETH_LINK_SPEED_100M)
2788                         speed |= IXGBE_LINK_SPEED_100_FULL;
2789                 if (*link_speeds & ETH_LINK_SPEED_10M)
2790                         speed |= IXGBE_LINK_SPEED_10_FULL;
2791         }
2792
2793         err = ixgbe_setup_link(hw, speed, link_up);
2794         if (err)
2795                 goto error;
2796
2797 skip_link_setup:
2798
2799         if (rte_intr_allow_others(intr_handle)) {
2800                 /* check if lsc interrupt is enabled */
2801                 if (dev->data->dev_conf.intr_conf.lsc != 0)
2802                         ixgbe_dev_lsc_interrupt_setup(dev, TRUE);
2803                 else
2804                         ixgbe_dev_lsc_interrupt_setup(dev, FALSE);
2805                 ixgbe_dev_macsec_interrupt_setup(dev);
2806         } else {
2807                 rte_intr_callback_unregister(intr_handle,
2808                                              ixgbe_dev_interrupt_handler, dev);
2809                 if (dev->data->dev_conf.intr_conf.lsc != 0)
2810                         PMD_INIT_LOG(INFO, "lsc won't enable because of"
2811                                      " no intr multiplex");
2812         }
2813
2814         /* check if rxq interrupt is enabled */
2815         if (dev->data->dev_conf.intr_conf.rxq != 0 &&
2816             rte_intr_dp_is_en(intr_handle))
2817                 ixgbe_dev_rxq_interrupt_setup(dev);
2818
2819         /* enable uio/vfio intr/eventfd mapping */
2820         rte_intr_enable(intr_handle);
2821
2822         /* resume enabled intr since hw reset */
2823         ixgbe_enable_intr(dev);
2824         ixgbe_l2_tunnel_conf(dev);
2825         ixgbe_filter_restore(dev);
2826
2827         if (tm_conf->root && !tm_conf->committed)
2828                 PMD_DRV_LOG(WARNING,
2829                             "please call hierarchy_commit() "
2830                             "before starting the port");
2831
2832         /* wait for the controller to acquire link */
2833         err = ixgbe_wait_for_link_up(hw);
2834         if (err)
2835                 goto error;
2836
2837         /*
2838          * Update link status right before return, because it may
2839          * start link configuration process in a separate thread.
2840          */
2841         ixgbe_dev_link_update(dev, 0);
2842
2843         /* setup the macsec setting register */
2844         if (macsec_setting->offload_en)
2845                 ixgbe_dev_macsec_register_enable(dev, macsec_setting);
2846
2847         return 0;
2848
2849 error:
2850         PMD_INIT_LOG(ERR, "failure in ixgbe_dev_start(): %d", err);
2851         ixgbe_dev_clear_queues(dev);
2852         return -EIO;
2853 }
2854
2855 /*
2856  * Stop device: disable rx and tx functions to allow for reconfiguring.
2857  */
2858 static void
2859 ixgbe_dev_stop(struct rte_eth_dev *dev)
2860 {
2861         struct rte_eth_link link;
2862         struct ixgbe_adapter *adapter = dev->data->dev_private;
2863         struct ixgbe_hw *hw =
2864                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2865         struct ixgbe_vf_info *vfinfo =
2866                 *IXGBE_DEV_PRIVATE_TO_P_VFDATA(dev->data->dev_private);
2867         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
2868         struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
2869         int vf;
2870         struct ixgbe_tm_conf *tm_conf =
2871                 IXGBE_DEV_PRIVATE_TO_TM_CONF(dev->data->dev_private);
2872
2873         if (hw->adapter_stopped)
2874                 return;
2875
2876         PMD_INIT_FUNC_TRACE();
2877
2878         ixgbe_dev_wait_setup_link_complete(dev, 0);
2879
2880         /* disable interrupts */
2881         ixgbe_disable_intr(hw);
2882
2883         /* reset the NIC */
2884         ixgbe_pf_reset_hw(hw);
2885         hw->adapter_stopped = 0;
2886
2887         /* stop adapter */
2888         ixgbe_stop_adapter(hw);
2889
2890         for (vf = 0; vfinfo != NULL && vf < pci_dev->max_vfs; vf++)
2891                 vfinfo[vf].clear_to_send = false;
2892
2893         if (hw->mac.ops.get_media_type(hw) == ixgbe_media_type_copper) {
2894                 /* Turn off the copper */
2895                 ixgbe_set_phy_power(hw, false);
2896         } else {
2897                 /* Turn off the laser */
2898                 ixgbe_disable_tx_laser(hw);
2899         }
2900
2901         ixgbe_dev_clear_queues(dev);
2902
2903         /* Clear stored conf */
2904         dev->data->scattered_rx = 0;
2905         dev->data->lro = 0;
2906
2907         /* Clear recorded link status */
2908         memset(&link, 0, sizeof(link));
2909         rte_eth_linkstatus_set(dev, &link);
2910
2911         if (!rte_intr_allow_others(intr_handle))
2912                 /* resume to the default handler */
2913                 rte_intr_callback_register(intr_handle,
2914                                            ixgbe_dev_interrupt_handler,
2915                                            (void *)dev);
2916
2917         /* Clean datapath event and queue/vec mapping */
2918         rte_intr_efd_disable(intr_handle);
2919         if (intr_handle->intr_vec != NULL) {
2920                 rte_free(intr_handle->intr_vec);
2921                 intr_handle->intr_vec = NULL;
2922         }
2923
2924         /* reset hierarchy commit */
2925         tm_conf->committed = false;
2926
2927         adapter->rss_reta_updated = 0;
2928
2929         hw->adapter_stopped = true;
2930 }
2931
2932 /*
2933  * Set device link up: enable tx.
2934  */
2935 static int
2936 ixgbe_dev_set_link_up(struct rte_eth_dev *dev)
2937 {
2938         struct ixgbe_hw *hw =
2939                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2940         if (hw->mac.type == ixgbe_mac_82599EB) {
2941 #ifdef RTE_LIBRTE_IXGBE_BYPASS
2942                 if (hw->device_id == IXGBE_DEV_ID_82599_BYPASS) {
2943                         /* Not suported in bypass mode */
2944                         PMD_INIT_LOG(ERR, "Set link up is not supported "
2945                                      "by device id 0x%x", hw->device_id);
2946                         return -ENOTSUP;
2947                 }
2948 #endif
2949         }
2950
2951         if (hw->mac.ops.get_media_type(hw) == ixgbe_media_type_copper) {
2952                 /* Turn on the copper */
2953                 ixgbe_set_phy_power(hw, true);
2954         } else {
2955                 /* Turn on the laser */
2956                 ixgbe_enable_tx_laser(hw);
2957                 ixgbe_dev_link_update(dev, 0);
2958         }
2959
2960         return 0;
2961 }
2962
2963 /*
2964  * Set device link down: disable tx.
2965  */
2966 static int
2967 ixgbe_dev_set_link_down(struct rte_eth_dev *dev)
2968 {
2969         struct ixgbe_hw *hw =
2970                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2971         if (hw->mac.type == ixgbe_mac_82599EB) {
2972 #ifdef RTE_LIBRTE_IXGBE_BYPASS
2973                 if (hw->device_id == IXGBE_DEV_ID_82599_BYPASS) {
2974                         /* Not suported in bypass mode */
2975                         PMD_INIT_LOG(ERR, "Set link down is not supported "
2976                                      "by device id 0x%x", hw->device_id);
2977                         return -ENOTSUP;
2978                 }
2979 #endif
2980         }
2981
2982         if (hw->mac.ops.get_media_type(hw) == ixgbe_media_type_copper) {
2983                 /* Turn off the copper */
2984                 ixgbe_set_phy_power(hw, false);
2985         } else {
2986                 /* Turn off the laser */
2987                 ixgbe_disable_tx_laser(hw);
2988                 ixgbe_dev_link_update(dev, 0);
2989         }
2990
2991         return 0;
2992 }
2993
2994 /*
2995  * Reset and stop device.
2996  */
2997 static int
2998 ixgbe_dev_close(struct rte_eth_dev *dev)
2999 {
3000         struct ixgbe_hw *hw =
3001                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3002         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
3003         struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
3004         int retries = 0;
3005         int ret;
3006
3007         PMD_INIT_FUNC_TRACE();
3008
3009         ixgbe_pf_reset_hw(hw);
3010
3011         ixgbe_dev_stop(dev);
3012
3013         ixgbe_dev_free_queues(dev);
3014
3015         ixgbe_disable_pcie_master(hw);
3016
3017         /* reprogram the RAR[0] in case user changed it. */
3018         ixgbe_set_rar(hw, 0, hw->mac.addr, 0, IXGBE_RAH_AV);
3019
3020         dev->dev_ops = NULL;
3021         dev->rx_pkt_burst = NULL;
3022         dev->tx_pkt_burst = NULL;
3023
3024         /* Unlock any pending hardware semaphore */
3025         ixgbe_swfw_lock_reset(hw);
3026
3027         /* disable uio intr before callback unregister */
3028         rte_intr_disable(intr_handle);
3029
3030         do {
3031                 ret = rte_intr_callback_unregister(intr_handle,
3032                                 ixgbe_dev_interrupt_handler, dev);
3033                 if (ret >= 0 || ret == -ENOENT) {
3034                         break;
3035                 } else if (ret != -EAGAIN) {
3036                         PMD_INIT_LOG(ERR,
3037                                 "intr callback unregister failed: %d",
3038                                 ret);
3039                 }
3040                 rte_delay_ms(100);
3041         } while (retries++ < (10 + IXGBE_LINK_UP_TIME));
3042
3043         /* cancel the delay handler before remove dev */
3044         rte_eal_alarm_cancel(ixgbe_dev_interrupt_delayed_handler, dev);
3045
3046         /* uninitialize PF if max_vfs not zero */
3047         ixgbe_pf_host_uninit(dev);
3048
3049         /* remove all the fdir filters & hash */
3050         ixgbe_fdir_filter_uninit(dev);
3051
3052         /* remove all the L2 tunnel filters & hash */
3053         ixgbe_l2_tn_filter_uninit(dev);
3054
3055         /* Remove all ntuple filters of the device */
3056         ixgbe_ntuple_filter_uninit(dev);
3057
3058         /* clear all the filters list */
3059         ixgbe_filterlist_flush();
3060
3061         /* Remove all Traffic Manager configuration */
3062         ixgbe_tm_conf_uninit(dev);
3063
3064 #ifdef RTE_LIBRTE_SECURITY
3065         rte_free(dev->security_ctx);
3066 #endif
3067
3068         return 0;
3069 }
3070
3071 /*
3072  * Reset PF device.
3073  */
3074 static int
3075 ixgbe_dev_reset(struct rte_eth_dev *dev)
3076 {
3077         int ret;
3078
3079         /* When a DPDK PMD PF begin to reset PF port, it should notify all
3080          * its VF to make them align with it. The detailed notification
3081          * mechanism is PMD specific. As to ixgbe PF, it is rather complex.
3082          * To avoid unexpected behavior in VF, currently reset of PF with
3083          * SR-IOV activation is not supported. It might be supported later.
3084          */
3085         if (dev->data->sriov.active)
3086                 return -ENOTSUP;
3087
3088         ret = eth_ixgbe_dev_uninit(dev);
3089         if (ret)
3090                 return ret;
3091
3092         ret = eth_ixgbe_dev_init(dev, NULL);
3093
3094         return ret;
3095 }
3096
3097 static void
3098 ixgbe_read_stats_registers(struct ixgbe_hw *hw,
3099                            struct ixgbe_hw_stats *hw_stats,
3100                            struct ixgbe_macsec_stats *macsec_stats,
3101                            uint64_t *total_missed_rx, uint64_t *total_qbrc,
3102                            uint64_t *total_qprc, uint64_t *total_qprdc)
3103 {
3104         uint32_t bprc, lxon, lxoff, total;
3105         uint32_t delta_gprc = 0;
3106         unsigned i;
3107         /* Workaround for RX byte count not including CRC bytes when CRC
3108          * strip is enabled. CRC bytes are removed from counters when crc_strip
3109          * is disabled.
3110          */
3111         int crc_strip = (IXGBE_READ_REG(hw, IXGBE_HLREG0) &
3112                         IXGBE_HLREG0_RXCRCSTRP);
3113
3114         hw_stats->crcerrs += IXGBE_READ_REG(hw, IXGBE_CRCERRS);
3115         hw_stats->illerrc += IXGBE_READ_REG(hw, IXGBE_ILLERRC);
3116         hw_stats->errbc += IXGBE_READ_REG(hw, IXGBE_ERRBC);
3117         hw_stats->mspdc += IXGBE_READ_REG(hw, IXGBE_MSPDC);
3118
3119         for (i = 0; i < 8; i++) {
3120                 uint32_t mp = IXGBE_READ_REG(hw, IXGBE_MPC(i));
3121
3122                 /* global total per queue */
3123                 hw_stats->mpc[i] += mp;
3124                 /* Running comprehensive total for stats display */
3125                 *total_missed_rx += hw_stats->mpc[i];
3126                 if (hw->mac.type == ixgbe_mac_82598EB) {
3127                         hw_stats->rnbc[i] +=
3128                             IXGBE_READ_REG(hw, IXGBE_RNBC(i));
3129                         hw_stats->pxonrxc[i] +=
3130                                 IXGBE_READ_REG(hw, IXGBE_PXONRXC(i));
3131                         hw_stats->pxoffrxc[i] +=
3132                                 IXGBE_READ_REG(hw, IXGBE_PXOFFRXC(i));
3133                 } else {
3134                         hw_stats->pxonrxc[i] +=
3135                                 IXGBE_READ_REG(hw, IXGBE_PXONRXCNT(i));
3136                         hw_stats->pxoffrxc[i] +=
3137                                 IXGBE_READ_REG(hw, IXGBE_PXOFFRXCNT(i));
3138                         hw_stats->pxon2offc[i] +=
3139                                 IXGBE_READ_REG(hw, IXGBE_PXON2OFFCNT(i));
3140                 }
3141                 hw_stats->pxontxc[i] +=
3142                     IXGBE_READ_REG(hw, IXGBE_PXONTXC(i));
3143                 hw_stats->pxofftxc[i] +=
3144                     IXGBE_READ_REG(hw, IXGBE_PXOFFTXC(i));
3145         }
3146         for (i = 0; i < IXGBE_QUEUE_STAT_COUNTERS; i++) {
3147                 uint32_t delta_qprc = IXGBE_READ_REG(hw, IXGBE_QPRC(i));
3148                 uint32_t delta_qptc = IXGBE_READ_REG(hw, IXGBE_QPTC(i));
3149                 uint32_t delta_qprdc = IXGBE_READ_REG(hw, IXGBE_QPRDC(i));
3150
3151                 delta_gprc += delta_qprc;
3152
3153                 hw_stats->qprc[i] += delta_qprc;
3154                 hw_stats->qptc[i] += delta_qptc;
3155
3156                 hw_stats->qbrc[i] += IXGBE_READ_REG(hw, IXGBE_QBRC_L(i));
3157                 hw_stats->qbrc[i] +=
3158                     ((uint64_t)IXGBE_READ_REG(hw, IXGBE_QBRC_H(i)) << 32);
3159                 if (crc_strip == 0)
3160                         hw_stats->qbrc[i] -= delta_qprc * RTE_ETHER_CRC_LEN;
3161
3162                 hw_stats->qbtc[i] += IXGBE_READ_REG(hw, IXGBE_QBTC_L(i));
3163                 hw_stats->qbtc[i] +=
3164                     ((uint64_t)IXGBE_READ_REG(hw, IXGBE_QBTC_H(i)) << 32);
3165
3166                 hw_stats->qprdc[i] += delta_qprdc;
3167                 *total_qprdc += hw_stats->qprdc[i];
3168
3169                 *total_qprc += hw_stats->qprc[i];
3170                 *total_qbrc += hw_stats->qbrc[i];
3171         }
3172         hw_stats->mlfc += IXGBE_READ_REG(hw, IXGBE_MLFC);
3173         hw_stats->mrfc += IXGBE_READ_REG(hw, IXGBE_MRFC);
3174         hw_stats->rlec += IXGBE_READ_REG(hw, IXGBE_RLEC);
3175
3176         /*
3177          * An errata states that gprc actually counts good + missed packets:
3178          * Workaround to set gprc to summated queue packet receives
3179          */
3180         hw_stats->gprc = *total_qprc;
3181
3182         if (hw->mac.type != ixgbe_mac_82598EB) {
3183                 hw_stats->gorc += IXGBE_READ_REG(hw, IXGBE_GORCL);
3184                 hw_stats->gorc += ((u64)IXGBE_READ_REG(hw, IXGBE_GORCH) << 32);
3185                 hw_stats->gotc += IXGBE_READ_REG(hw, IXGBE_GOTCL);
3186                 hw_stats->gotc += ((u64)IXGBE_READ_REG(hw, IXGBE_GOTCH) << 32);
3187                 hw_stats->tor += IXGBE_READ_REG(hw, IXGBE_TORL);
3188                 hw_stats->tor += ((u64)IXGBE_READ_REG(hw, IXGBE_TORH) << 32);
3189                 hw_stats->lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXCNT);
3190                 hw_stats->lxoffrxc += IXGBE_READ_REG(hw, IXGBE_LXOFFRXCNT);
3191         } else {
3192                 hw_stats->lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXC);
3193                 hw_stats->lxoffrxc += IXGBE_READ_REG(hw, IXGBE_LXOFFRXC);
3194                 /* 82598 only has a counter in the high register */
3195                 hw_stats->gorc += IXGBE_READ_REG(hw, IXGBE_GORCH);
3196                 hw_stats->gotc += IXGBE_READ_REG(hw, IXGBE_GOTCH);
3197                 hw_stats->tor += IXGBE_READ_REG(hw, IXGBE_TORH);
3198         }
3199         uint64_t old_tpr = hw_stats->tpr;
3200
3201         hw_stats->tpr += IXGBE_READ_REG(hw, IXGBE_TPR);
3202         hw_stats->tpt += IXGBE_READ_REG(hw, IXGBE_TPT);
3203
3204         if (crc_strip == 0)
3205                 hw_stats->gorc -= delta_gprc * RTE_ETHER_CRC_LEN;
3206
3207         uint64_t delta_gptc = IXGBE_READ_REG(hw, IXGBE_GPTC);
3208         hw_stats->gptc += delta_gptc;
3209         hw_stats->gotc -= delta_gptc * RTE_ETHER_CRC_LEN;
3210         hw_stats->tor -= (hw_stats->tpr - old_tpr) * RTE_ETHER_CRC_LEN;
3211
3212         /*
3213          * Workaround: mprc hardware is incorrectly counting
3214          * broadcasts, so for now we subtract those.
3215          */
3216         bprc = IXGBE_READ_REG(hw, IXGBE_BPRC);
3217         hw_stats->bprc += bprc;
3218         hw_stats->mprc += IXGBE_READ_REG(hw, IXGBE_MPRC);
3219         if (hw->mac.type == ixgbe_mac_82598EB)
3220                 hw_stats->mprc -= bprc;
3221
3222         hw_stats->prc64 += IXGBE_READ_REG(hw, IXGBE_PRC64);
3223         hw_stats->prc127 += IXGBE_READ_REG(hw, IXGBE_PRC127);
3224         hw_stats->prc255 += IXGBE_READ_REG(hw, IXGBE_PRC255);
3225         hw_stats->prc511 += IXGBE_READ_REG(hw, IXGBE_PRC511);
3226         hw_stats->prc1023 += IXGBE_READ_REG(hw, IXGBE_PRC1023);
3227         hw_stats->prc1522 += IXGBE_READ_REG(hw, IXGBE_PRC1522);
3228
3229         lxon = IXGBE_READ_REG(hw, IXGBE_LXONTXC);
3230         hw_stats->lxontxc += lxon;
3231         lxoff = IXGBE_READ_REG(hw, IXGBE_LXOFFTXC);
3232         hw_stats->lxofftxc += lxoff;
3233         total = lxon + lxoff;
3234
3235         hw_stats->mptc += IXGBE_READ_REG(hw, IXGBE_MPTC);
3236         hw_stats->ptc64 += IXGBE_READ_REG(hw, IXGBE_PTC64);
3237         hw_stats->gptc -= total;
3238         hw_stats->mptc -= total;
3239         hw_stats->ptc64 -= total;
3240         hw_stats->gotc -= total * RTE_ETHER_MIN_LEN;
3241
3242         hw_stats->ruc += IXGBE_READ_REG(hw, IXGBE_RUC);
3243         hw_stats->rfc += IXGBE_READ_REG(hw, IXGBE_RFC);
3244         hw_stats->roc += IXGBE_READ_REG(hw, IXGBE_ROC);
3245         hw_stats->rjc += IXGBE_READ_REG(hw, IXGBE_RJC);
3246         hw_stats->mngprc += IXGBE_READ_REG(hw, IXGBE_MNGPRC);
3247         hw_stats->mngpdc += IXGBE_READ_REG(hw, IXGBE_MNGPDC);
3248         hw_stats->mngptc += IXGBE_READ_REG(hw, IXGBE_MNGPTC);
3249         hw_stats->ptc127 += IXGBE_READ_REG(hw, IXGBE_PTC127);
3250         hw_stats->ptc255 += IXGBE_READ_REG(hw, IXGBE_PTC255);
3251         hw_stats->ptc511 += IXGBE_READ_REG(hw, IXGBE_PTC511);
3252         hw_stats->ptc1023 += IXGBE_READ_REG(hw, IXGBE_PTC1023);
3253         hw_stats->ptc1522 += IXGBE_READ_REG(hw, IXGBE_PTC1522);
3254         hw_stats->bptc += IXGBE_READ_REG(hw, IXGBE_BPTC);
3255         hw_stats->xec += IXGBE_READ_REG(hw, IXGBE_XEC);
3256         hw_stats->fccrc += IXGBE_READ_REG(hw, IXGBE_FCCRC);
3257         hw_stats->fclast += IXGBE_READ_REG(hw, IXGBE_FCLAST);
3258         /* Only read FCOE on 82599 */
3259         if (hw->mac.type != ixgbe_mac_82598EB) {
3260                 hw_stats->fcoerpdc += IXGBE_READ_REG(hw, IXGBE_FCOERPDC);
3261                 hw_stats->fcoeprc += IXGBE_READ_REG(hw, IXGBE_FCOEPRC);
3262                 hw_stats->fcoeptc += IXGBE_READ_REG(hw, IXGBE_FCOEPTC);
3263                 hw_stats->fcoedwrc += IXGBE_READ_REG(hw, IXGBE_FCOEDWRC);
3264                 hw_stats->fcoedwtc += IXGBE_READ_REG(hw, IXGBE_FCOEDWTC);
3265         }
3266
3267         /* Flow Director Stats registers */
3268         if (hw->mac.type != ixgbe_mac_82598EB) {
3269                 hw_stats->fdirmatch += IXGBE_READ_REG(hw, IXGBE_FDIRMATCH);
3270                 hw_stats->fdirmiss += IXGBE_READ_REG(hw, IXGBE_FDIRMISS);
3271                 hw_stats->fdirustat_add += IXGBE_READ_REG(hw,
3272                                         IXGBE_FDIRUSTAT) & 0xFFFF;
3273                 hw_stats->fdirustat_remove += (IXGBE_READ_REG(hw,
3274                                         IXGBE_FDIRUSTAT) >> 16) & 0xFFFF;
3275                 hw_stats->fdirfstat_fadd += IXGBE_READ_REG(hw,
3276                                         IXGBE_FDIRFSTAT) & 0xFFFF;
3277                 hw_stats->fdirfstat_fremove += (IXGBE_READ_REG(hw,
3278                                         IXGBE_FDIRFSTAT) >> 16) & 0xFFFF;
3279         }
3280         /* MACsec Stats registers */
3281         macsec_stats->out_pkts_untagged += IXGBE_READ_REG(hw, IXGBE_LSECTXUT);
3282         macsec_stats->out_pkts_encrypted +=
3283                 IXGBE_READ_REG(hw, IXGBE_LSECTXPKTE);
3284         macsec_stats->out_pkts_protected +=
3285                 IXGBE_READ_REG(hw, IXGBE_LSECTXPKTP);
3286         macsec_stats->out_octets_encrypted +=
3287                 IXGBE_READ_REG(hw, IXGBE_LSECTXOCTE);
3288         macsec_stats->out_octets_protected +=
3289                 IXGBE_READ_REG(hw, IXGBE_LSECTXOCTP);
3290         macsec_stats->in_pkts_untagged += IXGBE_READ_REG(hw, IXGBE_LSECRXUT);
3291         macsec_stats->in_pkts_badtag += IXGBE_READ_REG(hw, IXGBE_LSECRXBAD);
3292         macsec_stats->in_pkts_nosci += IXGBE_READ_REG(hw, IXGBE_LSECRXNOSCI);
3293         macsec_stats->in_pkts_unknownsci +=
3294                 IXGBE_READ_REG(hw, IXGBE_LSECRXUNSCI);
3295         macsec_stats->in_octets_decrypted +=
3296                 IXGBE_READ_REG(hw, IXGBE_LSECRXOCTD);
3297         macsec_stats->in_octets_validated +=
3298                 IXGBE_READ_REG(hw, IXGBE_LSECRXOCTV);
3299         macsec_stats->in_pkts_unchecked += IXGBE_READ_REG(hw, IXGBE_LSECRXUNCH);
3300         macsec_stats->in_pkts_delayed += IXGBE_READ_REG(hw, IXGBE_LSECRXDELAY);
3301         macsec_stats->in_pkts_late += IXGBE_READ_REG(hw, IXGBE_LSECRXLATE);
3302         for (i = 0; i < 2; i++) {
3303                 macsec_stats->in_pkts_ok +=
3304                         IXGBE_READ_REG(hw, IXGBE_LSECRXOK(i));
3305                 macsec_stats->in_pkts_invalid +=
3306                         IXGBE_READ_REG(hw, IXGBE_LSECRXINV(i));
3307                 macsec_stats->in_pkts_notvalid +=
3308                         IXGBE_READ_REG(hw, IXGBE_LSECRXNV(i));
3309         }
3310         macsec_stats->in_pkts_unusedsa += IXGBE_READ_REG(hw, IXGBE_LSECRXUNSA);
3311         macsec_stats->in_pkts_notusingsa +=
3312                 IXGBE_READ_REG(hw, IXGBE_LSECRXNUSA);
3313 }
3314
3315 /*
3316  * This function is based on ixgbe_update_stats_counters() in ixgbe/ixgbe.c
3317  */
3318 static int
3319 ixgbe_dev_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
3320 {
3321         struct ixgbe_hw *hw =
3322                         IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3323         struct ixgbe_hw_stats *hw_stats =
3324                         IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3325         struct ixgbe_macsec_stats *macsec_stats =
3326                         IXGBE_DEV_PRIVATE_TO_MACSEC_STATS(
3327                                 dev->data->dev_private);
3328         uint64_t total_missed_rx, total_qbrc, total_qprc, total_qprdc;
3329         unsigned i;
3330
3331         total_missed_rx = 0;
3332         total_qbrc = 0;
3333         total_qprc = 0;
3334         total_qprdc = 0;
3335
3336         ixgbe_read_stats_registers(hw, hw_stats, macsec_stats, &total_missed_rx,
3337                         &total_qbrc, &total_qprc, &total_qprdc);
3338
3339         if (stats == NULL)
3340                 return -EINVAL;
3341
3342         /* Fill out the rte_eth_stats statistics structure */
3343         stats->ipackets = total_qprc;
3344         stats->ibytes = total_qbrc;
3345         stats->opackets = hw_stats->gptc;
3346         stats->obytes = hw_stats->gotc;
3347
3348         for (i = 0; i < IXGBE_QUEUE_STAT_COUNTERS; i++) {
3349                 stats->q_ipackets[i] = hw_stats->qprc[i];
3350                 stats->q_opackets[i] = hw_stats->qptc[i];
3351                 stats->q_ibytes[i] = hw_stats->qbrc[i];
3352                 stats->q_obytes[i] = hw_stats->qbtc[i];
3353                 stats->q_errors[i] = hw_stats->qprdc[i];
3354         }
3355
3356         /* Rx Errors */
3357         stats->imissed  = total_missed_rx;
3358         stats->ierrors  = hw_stats->crcerrs +
3359                           hw_stats->mspdc +
3360                           hw_stats->rlec +
3361                           hw_stats->ruc +
3362                           hw_stats->roc +
3363                           hw_stats->illerrc +
3364                           hw_stats->errbc +
3365                           hw_stats->rfc +
3366                           hw_stats->fccrc +
3367                           hw_stats->fclast;
3368
3369         /* Tx Errors */
3370         stats->oerrors  = 0;
3371         return 0;
3372 }
3373
3374 static int
3375 ixgbe_dev_stats_reset(struct rte_eth_dev *dev)
3376 {
3377         struct ixgbe_hw_stats *stats =
3378                         IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3379
3380         /* HW registers are cleared on read */
3381         ixgbe_dev_stats_get(dev, NULL);
3382
3383         /* Reset software totals */
3384         memset(stats, 0, sizeof(*stats));
3385
3386         return 0;
3387 }
3388
3389 /* This function calculates the number of xstats based on the current config */
3390 static unsigned
3391 ixgbe_xstats_calc_num(void) {
3392         return IXGBE_NB_HW_STATS + IXGBE_NB_MACSEC_STATS +
3393                 (IXGBE_NB_RXQ_PRIO_STATS * IXGBE_NB_RXQ_PRIO_VALUES) +
3394                 (IXGBE_NB_TXQ_PRIO_STATS * IXGBE_NB_TXQ_PRIO_VALUES);
3395 }
3396
3397 static int ixgbe_dev_xstats_get_names(__rte_unused struct rte_eth_dev *dev,
3398         struct rte_eth_xstat_name *xstats_names, __rte_unused unsigned int size)
3399 {
3400         const unsigned cnt_stats = ixgbe_xstats_calc_num();
3401         unsigned stat, i, count;
3402
3403         if (xstats_names != NULL) {
3404                 count = 0;
3405
3406                 /* Note: limit >= cnt_stats checked upstream
3407                  * in rte_eth_xstats_names()
3408                  */
3409
3410                 /* Extended stats from ixgbe_hw_stats */
3411                 for (i = 0; i < IXGBE_NB_HW_STATS; i++) {
3412                         strlcpy(xstats_names[count].name,
3413                                 rte_ixgbe_stats_strings[i].name,
3414                                 sizeof(xstats_names[count].name));
3415                         count++;
3416                 }
3417
3418                 /* MACsec Stats */
3419                 for (i = 0; i < IXGBE_NB_MACSEC_STATS; i++) {
3420                         strlcpy(xstats_names[count].name,
3421                                 rte_ixgbe_macsec_strings[i].name,
3422                                 sizeof(xstats_names[count].name));
3423                         count++;
3424                 }
3425
3426                 /* RX Priority Stats */
3427                 for (stat = 0; stat < IXGBE_NB_RXQ_PRIO_STATS; stat++) {
3428                         for (i = 0; i < IXGBE_NB_RXQ_PRIO_VALUES; i++) {
3429                                 snprintf(xstats_names[count].name,
3430                                         sizeof(xstats_names[count].name),
3431                                         "rx_priority%u_%s", i,
3432                                         rte_ixgbe_rxq_strings[stat].name);
3433                                 count++;
3434                         }
3435                 }
3436
3437                 /* TX Priority Stats */
3438                 for (stat = 0; stat < IXGBE_NB_TXQ_PRIO_STATS; stat++) {
3439                         for (i = 0; i < IXGBE_NB_TXQ_PRIO_VALUES; i++) {
3440                                 snprintf(xstats_names[count].name,
3441                                         sizeof(xstats_names[count].name),
3442                                         "tx_priority%u_%s", i,
3443                                         rte_ixgbe_txq_strings[stat].name);
3444                                 count++;
3445                         }
3446                 }
3447         }
3448         return cnt_stats;
3449 }
3450
3451 static int ixgbe_dev_xstats_get_names_by_id(
3452         struct rte_eth_dev *dev,
3453         struct rte_eth_xstat_name *xstats_names,
3454         const uint64_t *ids,
3455         unsigned int limit)
3456 {
3457         if (!ids) {
3458                 const unsigned int cnt_stats = ixgbe_xstats_calc_num();
3459                 unsigned int stat, i, count;
3460
3461                 if (xstats_names != NULL) {
3462                         count = 0;
3463
3464                         /* Note: limit >= cnt_stats checked upstream
3465                          * in rte_eth_xstats_names()
3466                          */
3467
3468                         /* Extended stats from ixgbe_hw_stats */
3469                         for (i = 0; i < IXGBE_NB_HW_STATS; i++) {
3470                                 strlcpy(xstats_names[count].name,
3471                                         rte_ixgbe_stats_strings[i].name,
3472                                         sizeof(xstats_names[count].name));
3473                                 count++;
3474                         }
3475
3476                         /* MACsec Stats */
3477                         for (i = 0; i < IXGBE_NB_MACSEC_STATS; i++) {
3478                                 strlcpy(xstats_names[count].name,
3479                                         rte_ixgbe_macsec_strings[i].name,
3480                                         sizeof(xstats_names[count].name));
3481                                 count++;
3482                         }
3483
3484                         /* RX Priority Stats */
3485                         for (stat = 0; stat < IXGBE_NB_RXQ_PRIO_STATS; stat++) {
3486                                 for (i = 0; i < IXGBE_NB_RXQ_PRIO_VALUES; i++) {
3487                                         snprintf(xstats_names[count].name,
3488                                             sizeof(xstats_names[count].name),
3489                                             "rx_priority%u_%s", i,
3490                                             rte_ixgbe_rxq_strings[stat].name);
3491                                         count++;
3492                                 }
3493                         }
3494
3495                         /* TX Priority Stats */
3496                         for (stat = 0; stat < IXGBE_NB_TXQ_PRIO_STATS; stat++) {
3497                                 for (i = 0; i < IXGBE_NB_TXQ_PRIO_VALUES; i++) {
3498                                         snprintf(xstats_names[count].name,
3499                                             sizeof(xstats_names[count].name),
3500                                             "tx_priority%u_%s", i,
3501                                             rte_ixgbe_txq_strings[stat].name);
3502                                         count++;
3503                                 }
3504                         }
3505                 }
3506                 return cnt_stats;
3507         }
3508
3509         uint16_t i;
3510         uint16_t size = ixgbe_xstats_calc_num();
3511         struct rte_eth_xstat_name xstats_names_copy[size];
3512
3513         ixgbe_dev_xstats_get_names_by_id(dev, xstats_names_copy, NULL,
3514                         size);
3515
3516         for (i = 0; i < limit; i++) {
3517                 if (ids[i] >= size) {
3518                         PMD_INIT_LOG(ERR, "id value isn't valid");
3519                         return -1;
3520                 }
3521                 strcpy(xstats_names[i].name,
3522                                 xstats_names_copy[ids[i]].name);
3523         }
3524         return limit;
3525 }
3526
3527 static int ixgbevf_dev_xstats_get_names(__rte_unused struct rte_eth_dev *dev,
3528         struct rte_eth_xstat_name *xstats_names, unsigned limit)
3529 {
3530         unsigned i;
3531
3532         if (limit < IXGBEVF_NB_XSTATS && xstats_names != NULL)
3533                 return -ENOMEM;
3534
3535         if (xstats_names != NULL)
3536                 for (i = 0; i < IXGBEVF_NB_XSTATS; i++)
3537                         strlcpy(xstats_names[i].name,
3538                                 rte_ixgbevf_stats_strings[i].name,
3539                                 sizeof(xstats_names[i].name));
3540         return IXGBEVF_NB_XSTATS;
3541 }
3542
3543 static int
3544 ixgbe_dev_xstats_get(struct rte_eth_dev *dev, struct rte_eth_xstat *xstats,
3545                                          unsigned n)
3546 {
3547         struct ixgbe_hw *hw =
3548                         IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3549         struct ixgbe_hw_stats *hw_stats =
3550                         IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3551         struct ixgbe_macsec_stats *macsec_stats =
3552                         IXGBE_DEV_PRIVATE_TO_MACSEC_STATS(
3553                                 dev->data->dev_private);
3554         uint64_t total_missed_rx, total_qbrc, total_qprc, total_qprdc;
3555         unsigned i, stat, count = 0;
3556
3557         count = ixgbe_xstats_calc_num();
3558
3559         if (n < count)
3560                 return count;
3561
3562         total_missed_rx = 0;
3563         total_qbrc = 0;
3564         total_qprc = 0;
3565         total_qprdc = 0;
3566
3567         ixgbe_read_stats_registers(hw, hw_stats, macsec_stats, &total_missed_rx,
3568                         &total_qbrc, &total_qprc, &total_qprdc);
3569
3570         /* If this is a reset xstats is NULL, and we have cleared the
3571          * registers by reading them.
3572          */
3573         if (!xstats)
3574                 return 0;
3575
3576         /* Extended stats from ixgbe_hw_stats */
3577         count = 0;
3578         for (i = 0; i < IXGBE_NB_HW_STATS; i++) {
3579                 xstats[count].value = *(uint64_t *)(((char *)hw_stats) +
3580                                 rte_ixgbe_stats_strings[i].offset);
3581                 xstats[count].id = count;
3582                 count++;
3583         }
3584
3585         /* MACsec Stats */
3586         for (i = 0; i < IXGBE_NB_MACSEC_STATS; i++) {
3587                 xstats[count].value = *(uint64_t *)(((char *)macsec_stats) +
3588                                 rte_ixgbe_macsec_strings[i].offset);
3589                 xstats[count].id = count;
3590                 count++;
3591         }
3592
3593         /* RX Priority Stats */
3594         for (stat = 0; stat < IXGBE_NB_RXQ_PRIO_STATS; stat++) {
3595                 for (i = 0; i < IXGBE_NB_RXQ_PRIO_VALUES; i++) {
3596                         xstats[count].value = *(uint64_t *)(((char *)hw_stats) +
3597                                         rte_ixgbe_rxq_strings[stat].offset +
3598                                         (sizeof(uint64_t) * i));
3599                         xstats[count].id = count;
3600                         count++;
3601                 }
3602         }
3603
3604         /* TX Priority Stats */
3605         for (stat = 0; stat < IXGBE_NB_TXQ_PRIO_STATS; stat++) {
3606                 for (i = 0; i < IXGBE_NB_TXQ_PRIO_VALUES; i++) {
3607                         xstats[count].value = *(uint64_t *)(((char *)hw_stats) +
3608                                         rte_ixgbe_txq_strings[stat].offset +
3609                                         (sizeof(uint64_t) * i));
3610                         xstats[count].id = count;
3611                         count++;
3612                 }
3613         }
3614         return count;
3615 }
3616
3617 static int
3618 ixgbe_dev_xstats_get_by_id(struct rte_eth_dev *dev, const uint64_t *ids,
3619                 uint64_t *values, unsigned int n)
3620 {
3621         if (!ids) {
3622                 struct ixgbe_hw *hw =
3623                                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3624                 struct ixgbe_hw_stats *hw_stats =
3625                                 IXGBE_DEV_PRIVATE_TO_STATS(
3626                                                 dev->data->dev_private);
3627                 struct ixgbe_macsec_stats *macsec_stats =
3628                                 IXGBE_DEV_PRIVATE_TO_MACSEC_STATS(
3629                                         dev->data->dev_private);
3630                 uint64_t total_missed_rx, total_qbrc, total_qprc, total_qprdc;
3631                 unsigned int i, stat, count = 0;
3632
3633                 count = ixgbe_xstats_calc_num();
3634
3635                 if (!ids && n < count)
3636                         return count;
3637
3638                 total_missed_rx = 0;
3639                 total_qbrc = 0;
3640                 total_qprc = 0;
3641                 total_qprdc = 0;
3642
3643                 ixgbe_read_stats_registers(hw, hw_stats, macsec_stats,
3644                                 &total_missed_rx, &total_qbrc, &total_qprc,
3645                                 &total_qprdc);
3646
3647                 /* If this is a reset xstats is NULL, and we have cleared the
3648                  * registers by reading them.
3649                  */
3650                 if (!ids && !values)
3651                         return 0;
3652
3653                 /* Extended stats from ixgbe_hw_stats */
3654                 count = 0;
3655                 for (i = 0; i < IXGBE_NB_HW_STATS; i++) {
3656                         values[count] = *(uint64_t *)(((char *)hw_stats) +
3657                                         rte_ixgbe_stats_strings[i].offset);
3658                         count++;
3659                 }
3660
3661                 /* MACsec Stats */
3662                 for (i = 0; i < IXGBE_NB_MACSEC_STATS; i++) {
3663                         values[count] = *(uint64_t *)(((char *)macsec_stats) +
3664                                         rte_ixgbe_macsec_strings[i].offset);
3665                         count++;
3666                 }
3667
3668                 /* RX Priority Stats */
3669                 for (stat = 0; stat < IXGBE_NB_RXQ_PRIO_STATS; stat++) {
3670                         for (i = 0; i < IXGBE_NB_RXQ_PRIO_VALUES; i++) {
3671                                 values[count] =
3672                                         *(uint64_t *)(((char *)hw_stats) +
3673                                         rte_ixgbe_rxq_strings[stat].offset +
3674                                         (sizeof(uint64_t) * i));
3675                                 count++;
3676                         }
3677                 }
3678
3679                 /* TX Priority Stats */
3680                 for (stat = 0; stat < IXGBE_NB_TXQ_PRIO_STATS; stat++) {
3681                         for (i = 0; i < IXGBE_NB_TXQ_PRIO_VALUES; i++) {
3682                                 values[count] =
3683                                         *(uint64_t *)(((char *)hw_stats) +
3684                                         rte_ixgbe_txq_strings[stat].offset +
3685                                         (sizeof(uint64_t) * i));
3686                                 count++;
3687                         }
3688                 }
3689                 return count;
3690         }
3691
3692         uint16_t i;
3693         uint16_t size = ixgbe_xstats_calc_num();
3694         uint64_t values_copy[size];
3695
3696         ixgbe_dev_xstats_get_by_id(dev, NULL, values_copy, size);
3697
3698         for (i = 0; i < n; i++) {
3699                 if (ids[i] >= size) {
3700                         PMD_INIT_LOG(ERR, "id value isn't valid");
3701                         return -1;
3702                 }
3703                 values[i] = values_copy[ids[i]];
3704         }
3705         return n;
3706 }
3707
3708 static int
3709 ixgbe_dev_xstats_reset(struct rte_eth_dev *dev)
3710 {
3711         struct ixgbe_hw_stats *stats =
3712                         IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3713         struct ixgbe_macsec_stats *macsec_stats =
3714                         IXGBE_DEV_PRIVATE_TO_MACSEC_STATS(
3715                                 dev->data->dev_private);
3716
3717         unsigned count = ixgbe_xstats_calc_num();
3718
3719         /* HW registers are cleared on read */
3720         ixgbe_dev_xstats_get(dev, NULL, count);
3721
3722         /* Reset software totals */
3723         memset(stats, 0, sizeof(*stats));
3724         memset(macsec_stats, 0, sizeof(*macsec_stats));
3725
3726         return 0;
3727 }
3728
3729 static void
3730 ixgbevf_update_stats(struct rte_eth_dev *dev)
3731 {
3732         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3733         struct ixgbevf_hw_stats *hw_stats = (struct ixgbevf_hw_stats *)
3734                           IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3735
3736         /* Good Rx packet, include VF loopback */
3737         UPDATE_VF_STAT(IXGBE_VFGPRC,
3738             hw_stats->last_vfgprc, hw_stats->vfgprc);
3739
3740         /* Good Rx octets, include VF loopback */
3741         UPDATE_VF_STAT_36BIT(IXGBE_VFGORC_LSB, IXGBE_VFGORC_MSB,
3742             hw_stats->last_vfgorc, hw_stats->vfgorc);
3743
3744         /* Good Tx packet, include VF loopback */
3745         UPDATE_VF_STAT(IXGBE_VFGPTC,
3746             hw_stats->last_vfgptc, hw_stats->vfgptc);
3747
3748         /* Good Tx octets, include VF loopback */
3749         UPDATE_VF_STAT_36BIT(IXGBE_VFGOTC_LSB, IXGBE_VFGOTC_MSB,
3750             hw_stats->last_vfgotc, hw_stats->vfgotc);
3751
3752         /* Rx Multicst Packet */
3753         UPDATE_VF_STAT(IXGBE_VFMPRC,
3754             hw_stats->last_vfmprc, hw_stats->vfmprc);
3755 }
3756
3757 static int
3758 ixgbevf_dev_xstats_get(struct rte_eth_dev *dev, struct rte_eth_xstat *xstats,
3759                        unsigned n)
3760 {
3761         struct ixgbevf_hw_stats *hw_stats = (struct ixgbevf_hw_stats *)
3762                         IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3763         unsigned i;
3764
3765         if (n < IXGBEVF_NB_XSTATS)
3766                 return IXGBEVF_NB_XSTATS;
3767
3768         ixgbevf_update_stats(dev);
3769
3770         if (!xstats)
3771                 return 0;
3772
3773         /* Extended stats */
3774         for (i = 0; i < IXGBEVF_NB_XSTATS; i++) {
3775                 xstats[i].id = i;
3776                 xstats[i].value = *(uint64_t *)(((char *)hw_stats) +
3777                         rte_ixgbevf_stats_strings[i].offset);
3778         }
3779
3780         return IXGBEVF_NB_XSTATS;
3781 }
3782
3783 static int
3784 ixgbevf_dev_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
3785 {
3786         struct ixgbevf_hw_stats *hw_stats = (struct ixgbevf_hw_stats *)
3787                           IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3788
3789         ixgbevf_update_stats(dev);
3790
3791         if (stats == NULL)
3792                 return -EINVAL;
3793
3794         stats->ipackets = hw_stats->vfgprc;
3795         stats->ibytes = hw_stats->vfgorc;
3796         stats->opackets = hw_stats->vfgptc;
3797         stats->obytes = hw_stats->vfgotc;
3798         return 0;
3799 }
3800
3801 static int
3802 ixgbevf_dev_stats_reset(struct rte_eth_dev *dev)
3803 {
3804         struct ixgbevf_hw_stats *hw_stats = (struct ixgbevf_hw_stats *)
3805                         IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
3806
3807         /* Sync HW register to the last stats */
3808         ixgbevf_dev_stats_get(dev, NULL);
3809
3810         /* reset HW current stats*/
3811         hw_stats->vfgprc = 0;
3812         hw_stats->vfgorc = 0;
3813         hw_stats->vfgptc = 0;
3814         hw_stats->vfgotc = 0;
3815
3816         return 0;
3817 }
3818
3819 static int
3820 ixgbe_fw_version_get(struct rte_eth_dev *dev, char *fw_version, size_t fw_size)
3821 {
3822         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3823         u16 eeprom_verh, eeprom_verl;
3824         u32 etrack_id;
3825         int ret;
3826
3827         ixgbe_read_eeprom(hw, 0x2e, &eeprom_verh);
3828         ixgbe_read_eeprom(hw, 0x2d, &eeprom_verl);
3829
3830         etrack_id = (eeprom_verh << 16) | eeprom_verl;
3831         ret = snprintf(fw_version, fw_size, "0x%08x", etrack_id);
3832
3833         ret += 1; /* add the size of '\0' */
3834         if (fw_size < (u32)ret)
3835                 return ret;
3836         else
3837                 return 0;
3838 }
3839
3840 static int
3841 ixgbe_dev_info_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
3842 {
3843         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
3844         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3845         struct rte_eth_conf *dev_conf = &dev->data->dev_conf;
3846
3847         dev_info->max_rx_queues = (uint16_t)hw->mac.max_rx_queues;
3848         dev_info->max_tx_queues = (uint16_t)hw->mac.max_tx_queues;
3849         if (RTE_ETH_DEV_SRIOV(dev).active == 0) {
3850                 /*
3851                  * When DCB/VT is off, maximum number of queues changes,
3852                  * except for 82598EB, which remains constant.
3853                  */
3854                 if (dev_conf->txmode.mq_mode == ETH_MQ_TX_NONE &&
3855                                 hw->mac.type != ixgbe_mac_82598EB)
3856                         dev_info->max_tx_queues = IXGBE_NONE_MODE_TX_NB_QUEUES;
3857         }
3858         dev_info->min_rx_bufsize = 1024; /* cf BSIZEPACKET in SRRCTL register */
3859         dev_info->max_rx_pktlen = 15872; /* includes CRC, cf MAXFRS register */
3860         dev_info->max_mac_addrs = hw->mac.num_rar_entries;
3861         dev_info->max_hash_mac_addrs = IXGBE_VMDQ_NUM_UC_MAC;
3862         dev_info->max_vfs = pci_dev->max_vfs;
3863         if (hw->mac.type == ixgbe_mac_82598EB)
3864                 dev_info->max_vmdq_pools = ETH_16_POOLS;
3865         else
3866                 dev_info->max_vmdq_pools = ETH_64_POOLS;
3867         dev_info->max_mtu =  dev_info->max_rx_pktlen - IXGBE_ETH_OVERHEAD;
3868         dev_info->min_mtu = RTE_ETHER_MIN_MTU;
3869         dev_info->vmdq_queue_num = dev_info->max_rx_queues;
3870         dev_info->rx_queue_offload_capa = ixgbe_get_rx_queue_offloads(dev);
3871         dev_info->rx_offload_capa = (ixgbe_get_rx_port_offloads(dev) |
3872                                      dev_info->rx_queue_offload_capa);
3873         dev_info->tx_queue_offload_capa = ixgbe_get_tx_queue_offloads(dev);
3874         dev_info->tx_offload_capa = ixgbe_get_tx_port_offloads(dev);
3875
3876         dev_info->default_rxconf = (struct rte_eth_rxconf) {
3877                 .rx_thresh = {
3878                         .pthresh = IXGBE_DEFAULT_RX_PTHRESH,
3879                         .hthresh = IXGBE_DEFAULT_RX_HTHRESH,
3880                         .wthresh = IXGBE_DEFAULT_RX_WTHRESH,
3881                 },
3882                 .rx_free_thresh = IXGBE_DEFAULT_RX_FREE_THRESH,
3883                 .rx_drop_en = 0,
3884                 .offloads = 0,
3885         };
3886
3887         dev_info->default_txconf = (struct rte_eth_txconf) {
3888                 .tx_thresh = {
3889                         .pthresh = IXGBE_DEFAULT_TX_PTHRESH,
3890                         .hthresh = IXGBE_DEFAULT_TX_HTHRESH,
3891                         .wthresh = IXGBE_DEFAULT_TX_WTHRESH,
3892                 },
3893                 .tx_free_thresh = IXGBE_DEFAULT_TX_FREE_THRESH,
3894                 .tx_rs_thresh = IXGBE_DEFAULT_TX_RSBIT_THRESH,
3895                 .offloads = 0,
3896         };
3897
3898         dev_info->rx_desc_lim = rx_desc_lim;
3899         dev_info->tx_desc_lim = tx_desc_lim;
3900
3901         dev_info->hash_key_size = IXGBE_HKEY_MAX_INDEX * sizeof(uint32_t);
3902         dev_info->reta_size = ixgbe_reta_size_get(hw->mac.type);
3903         dev_info->flow_type_rss_offloads = IXGBE_RSS_OFFLOAD_ALL;
3904
3905         dev_info->speed_capa = ETH_LINK_SPEED_1G | ETH_LINK_SPEED_10G;
3906         if (hw->device_id == IXGBE_DEV_ID_X550EM_A_1G_T ||
3907                         hw->device_id == IXGBE_DEV_ID_X550EM_A_1G_T_L)
3908                 dev_info->speed_capa = ETH_LINK_SPEED_10M |
3909                         ETH_LINK_SPEED_100M | ETH_LINK_SPEED_1G;
3910
3911         if (hw->mac.type == ixgbe_mac_X540 ||
3912             hw->mac.type == ixgbe_mac_X540_vf ||
3913             hw->mac.type == ixgbe_mac_X550 ||
3914             hw->mac.type == ixgbe_mac_X550_vf) {
3915                 dev_info->speed_capa |= ETH_LINK_SPEED_100M;
3916         }
3917         if (hw->mac.type == ixgbe_mac_X550) {
3918                 dev_info->speed_capa |= ETH_LINK_SPEED_2_5G;
3919                 dev_info->speed_capa |= ETH_LINK_SPEED_5G;
3920         }
3921
3922         /* Driver-preferred Rx/Tx parameters */
3923         dev_info->default_rxportconf.burst_size = 32;
3924         dev_info->default_txportconf.burst_size = 32;
3925         dev_info->default_rxportconf.nb_queues = 1;
3926         dev_info->default_txportconf.nb_queues = 1;
3927         dev_info->default_rxportconf.ring_size = 256;
3928         dev_info->default_txportconf.ring_size = 256;
3929
3930         return 0;
3931 }
3932
3933 static const uint32_t *
3934 ixgbe_dev_supported_ptypes_get(struct rte_eth_dev *dev)
3935 {
3936         static const uint32_t ptypes[] = {
3937                 /* For non-vec functions,
3938                  * refers to ixgbe_rxd_pkt_info_to_pkt_type();
3939                  * for vec functions,
3940                  * refers to _recv_raw_pkts_vec().
3941                  */
3942                 RTE_PTYPE_L2_ETHER,
3943                 RTE_PTYPE_L3_IPV4,
3944                 RTE_PTYPE_L3_IPV4_EXT,
3945                 RTE_PTYPE_L3_IPV6,
3946                 RTE_PTYPE_L3_IPV6_EXT,
3947                 RTE_PTYPE_L4_SCTP,
3948                 RTE_PTYPE_L4_TCP,
3949                 RTE_PTYPE_L4_UDP,
3950                 RTE_PTYPE_TUNNEL_IP,
3951                 RTE_PTYPE_INNER_L3_IPV6,
3952                 RTE_PTYPE_INNER_L3_IPV6_EXT,
3953                 RTE_PTYPE_INNER_L4_TCP,
3954                 RTE_PTYPE_INNER_L4_UDP,
3955                 RTE_PTYPE_UNKNOWN
3956         };
3957
3958         if (dev->rx_pkt_burst == ixgbe_recv_pkts ||
3959             dev->rx_pkt_burst == ixgbe_recv_pkts_lro_single_alloc ||
3960             dev->rx_pkt_burst == ixgbe_recv_pkts_lro_bulk_alloc ||
3961             dev->rx_pkt_burst == ixgbe_recv_pkts_bulk_alloc)
3962                 return ptypes;
3963
3964 #if defined(RTE_ARCH_X86) || defined(__ARM_NEON)
3965         if (dev->rx_pkt_burst == ixgbe_recv_pkts_vec ||
3966             dev->rx_pkt_burst == ixgbe_recv_scattered_pkts_vec)
3967                 return ptypes;
3968 #endif
3969         return NULL;
3970 }
3971
3972 static int
3973 ixgbevf_dev_info_get(struct rte_eth_dev *dev,
3974                      struct rte_eth_dev_info *dev_info)
3975 {
3976         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
3977         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3978
3979         dev_info->max_rx_queues = (uint16_t)hw->mac.max_rx_queues;
3980         dev_info->max_tx_queues = (uint16_t)hw->mac.max_tx_queues;
3981         dev_info->min_rx_bufsize = 1024; /* cf BSIZEPACKET in SRRCTL reg */
3982         dev_info->max_rx_pktlen = 9728; /* includes CRC, cf MAXFRS reg */
3983         dev_info->max_mtu = dev_info->max_rx_pktlen - IXGBE_ETH_OVERHEAD;
3984         dev_info->max_mac_addrs = hw->mac.num_rar_entries;
3985         dev_info->max_hash_mac_addrs = IXGBE_VMDQ_NUM_UC_MAC;
3986         dev_info->max_vfs = pci_dev->max_vfs;
3987         if (hw->mac.type == ixgbe_mac_82598EB)
3988                 dev_info->max_vmdq_pools = ETH_16_POOLS;
3989         else
3990                 dev_info->max_vmdq_pools = ETH_64_POOLS;
3991         dev_info->rx_queue_offload_capa = ixgbe_get_rx_queue_offloads(dev);
3992         dev_info->rx_offload_capa = (ixgbe_get_rx_port_offloads(dev) |
3993                                      dev_info->rx_queue_offload_capa);
3994         dev_info->tx_queue_offload_capa = ixgbe_get_tx_queue_offloads(dev);
3995         dev_info->tx_offload_capa = ixgbe_get_tx_port_offloads(dev);
3996         dev_info->hash_key_size = IXGBE_HKEY_MAX_INDEX * sizeof(uint32_t);
3997         dev_info->reta_size = ixgbe_reta_size_get(hw->mac.type);
3998         dev_info->flow_type_rss_offloads = IXGBE_RSS_OFFLOAD_ALL;
3999
4000         dev_info->default_rxconf = (struct rte_eth_rxconf) {
4001                 .rx_thresh = {
4002                         .pthresh = IXGBE_DEFAULT_RX_PTHRESH,
4003                         .hthresh = IXGBE_DEFAULT_RX_HTHRESH,
4004                         .wthresh = IXGBE_DEFAULT_RX_WTHRESH,
4005                 },
4006                 .rx_free_thresh = IXGBE_DEFAULT_RX_FREE_THRESH,
4007                 .rx_drop_en = 0,
4008                 .offloads = 0,
4009         };
4010
4011         dev_info->default_txconf = (struct rte_eth_txconf) {
4012                 .tx_thresh = {
4013                         .pthresh = IXGBE_DEFAULT_TX_PTHRESH,
4014                         .hthresh = IXGBE_DEFAULT_TX_HTHRESH,
4015                         .wthresh = IXGBE_DEFAULT_TX_WTHRESH,
4016                 },
4017                 .tx_free_thresh = IXGBE_DEFAULT_TX_FREE_THRESH,
4018                 .tx_rs_thresh = IXGBE_DEFAULT_TX_RSBIT_THRESH,
4019                 .offloads = 0,
4020         };
4021
4022         dev_info->rx_desc_lim = rx_desc_lim;
4023         dev_info->tx_desc_lim = tx_desc_lim;
4024
4025         return 0;
4026 }
4027
4028 static int
4029 ixgbevf_check_link(struct ixgbe_hw *hw, ixgbe_link_speed *speed,
4030                    bool *link_up, int wait_to_complete)
4031 {
4032         struct ixgbe_adapter *adapter = container_of(hw,
4033                                                      struct ixgbe_adapter, hw);
4034         struct ixgbe_mbx_info *mbx = &hw->mbx;
4035         struct ixgbe_mac_info *mac = &hw->mac;
4036         uint32_t links_reg, in_msg;
4037         int ret_val = 0;
4038
4039         /* If we were hit with a reset drop the link */
4040         if (!mbx->ops.check_for_rst(hw, 0) || !mbx->timeout)
4041                 mac->get_link_status = true;
4042
4043         if (!mac->get_link_status)
4044                 goto out;
4045
4046         /* if link status is down no point in checking to see if pf is up */
4047         links_reg = IXGBE_READ_REG(hw, IXGBE_VFLINKS);
4048         if (!(links_reg & IXGBE_LINKS_UP))
4049                 goto out;
4050
4051         /* for SFP+ modules and DA cables on 82599 it can take up to 500usecs
4052          * before the link status is correct
4053          */
4054         if (mac->type == ixgbe_mac_82599_vf && wait_to_complete) {
4055                 int i;
4056
4057                 for (i = 0; i < 5; i++) {
4058                         rte_delay_us(100);
4059                         links_reg = IXGBE_READ_REG(hw, IXGBE_VFLINKS);
4060
4061                         if (!(links_reg & IXGBE_LINKS_UP))
4062                                 goto out;
4063                 }
4064         }
4065
4066         switch (links_reg & IXGBE_LINKS_SPEED_82599) {
4067         case IXGBE_LINKS_SPEED_10G_82599:
4068                 *speed = IXGBE_LINK_SPEED_10GB_FULL;
4069                 if (hw->mac.type >= ixgbe_mac_X550) {
4070                         if (links_reg & IXGBE_LINKS_SPEED_NON_STD)
4071                                 *speed = IXGBE_LINK_SPEED_2_5GB_FULL;
4072                 }
4073                 break;
4074         case IXGBE_LINKS_SPEED_1G_82599:
4075                 *speed = IXGBE_LINK_SPEED_1GB_FULL;
4076                 break;
4077         case IXGBE_LINKS_SPEED_100_82599:
4078                 *speed = IXGBE_LINK_SPEED_100_FULL;
4079                 if (hw->mac.type == ixgbe_mac_X550) {
4080                         if (links_reg & IXGBE_LINKS_SPEED_NON_STD)
4081                                 *speed = IXGBE_LINK_SPEED_5GB_FULL;
4082                 }
4083                 break;
4084         case IXGBE_LINKS_SPEED_10_X550EM_A:
4085                 *speed = IXGBE_LINK_SPEED_UNKNOWN;
4086                 /* Since Reserved in older MAC's */
4087                 if (hw->mac.type >= ixgbe_mac_X550)
4088                         *speed = IXGBE_LINK_SPEED_10_FULL;
4089                 break;
4090         default:
4091                 *speed = IXGBE_LINK_SPEED_UNKNOWN;
4092         }
4093
4094         if (wait_to_complete == 0 && adapter->pflink_fullchk == 0) {
4095                 if (*speed == IXGBE_LINK_SPEED_UNKNOWN)
4096                         mac->get_link_status = true;
4097                 else
4098                         mac->get_link_status = false;
4099
4100                 goto out;
4101         }
4102
4103         /* if the read failed it could just be a mailbox collision, best wait
4104          * until we are called again and don't report an error
4105          */
4106         if (mbx->ops.read(hw, &in_msg, 1, 0))
4107                 goto out;
4108
4109         if (!(in_msg & IXGBE_VT_MSGTYPE_CTS)) {
4110                 /* msg is not CTS and is NACK we must have lost CTS status */
4111                 if (in_msg & IXGBE_VT_MSGTYPE_NACK)
4112                         mac->get_link_status = false;
4113                 goto out;
4114         }
4115
4116         /* the pf is talking, if we timed out in the past we reinit */
4117         if (!mbx->timeout) {
4118                 ret_val = -1;
4119                 goto out;
4120         }
4121
4122         /* if we passed all the tests above then the link is up and we no
4123          * longer need to check for link
4124          */
4125         mac->get_link_status = false;
4126
4127 out:
4128         *link_up = !mac->get_link_status;
4129         return ret_val;
4130 }
4131
4132 /*
4133  * If @timeout_ms was 0, it means that it will not return until link complete.
4134  * It returns 1 on complete, return 0 on timeout.
4135  */
4136 static int
4137 ixgbe_dev_wait_setup_link_complete(struct rte_eth_dev *dev, uint32_t timeout_ms)
4138 {
4139 #define WARNING_TIMEOUT    9000 /* 9s  in total */
4140         struct ixgbe_adapter *ad = dev->data->dev_private;
4141         uint32_t timeout = timeout_ms ? timeout_ms : WARNING_TIMEOUT;
4142
4143         while (rte_atomic32_read(&ad->link_thread_running)) {
4144                 msec_delay(1);
4145                 timeout--;
4146
4147                 if (timeout_ms) {
4148                         if (!timeout)
4149                                 return 0;
4150                 } else if (!timeout) {
4151                         /* It will not return until link complete */
4152                         timeout = WARNING_TIMEOUT;
4153                         PMD_DRV_LOG(ERR, "IXGBE link thread not complete too long time!");
4154                 }
4155         }
4156
4157         return 1;
4158 }
4159
4160 static void *
4161 ixgbe_dev_setup_link_thread_handler(void *param)
4162 {
4163         struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
4164         struct ixgbe_adapter *ad = dev->data->dev_private;
4165         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4166         struct ixgbe_interrupt *intr =
4167                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4168         u32 speed;
4169         bool autoneg = false;
4170
4171         pthread_detach(pthread_self());
4172         speed = hw->phy.autoneg_advertised;
4173         if (!speed)
4174                 ixgbe_get_link_capabilities(hw, &speed, &autoneg);
4175
4176         ixgbe_setup_link(hw, speed, true);
4177
4178         intr->flags &= ~IXGBE_FLAG_NEED_LINK_CONFIG;
4179         rte_atomic32_clear(&ad->link_thread_running);
4180         return NULL;
4181 }
4182
4183 /*
4184  * In freebsd environment, nic_uio drivers do not support interrupts,
4185  * rte_intr_callback_register() will fail to register interrupts.
4186  * We can not make link status to change from down to up by interrupt
4187  * callback. So we need to wait for the controller to acquire link
4188  * when ports start.
4189  * It returns 0 on link up.
4190  */
4191 static int
4192 ixgbe_wait_for_link_up(struct ixgbe_hw *hw)
4193 {
4194 #ifdef RTE_EXEC_ENV_FREEBSD
4195         int err, i;
4196         bool link_up = false;
4197         uint32_t speed = 0;
4198         const int nb_iter = 25;
4199
4200         for (i = 0; i < nb_iter; i++) {
4201                 err = ixgbe_check_link(hw, &speed, &link_up, 0);
4202                 if (err)
4203                         return err;
4204                 if (link_up)
4205                         return 0;
4206                 msec_delay(200);
4207         }
4208
4209         return 0;
4210 #else
4211         RTE_SET_USED(hw);
4212         return 0;
4213 #endif
4214 }
4215
4216 /* return 0 means link status changed, -1 means not changed */
4217 int
4218 ixgbe_dev_link_update_share(struct rte_eth_dev *dev,
4219                             int wait_to_complete, int vf)
4220 {
4221         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4222         struct ixgbe_adapter *ad = dev->data->dev_private;
4223         struct rte_eth_link link;
4224         ixgbe_link_speed link_speed = IXGBE_LINK_SPEED_UNKNOWN;
4225         struct ixgbe_interrupt *intr =
4226                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4227         bool link_up;
4228         int diag;
4229         int wait = 1;
4230         u32 esdp_reg;
4231
4232         memset(&link, 0, sizeof(link));
4233         link.link_status = ETH_LINK_DOWN;
4234         link.link_speed = ETH_SPEED_NUM_NONE;
4235         link.link_duplex = ETH_LINK_HALF_DUPLEX;
4236         link.link_autoneg = !(dev->data->dev_conf.link_speeds &
4237                         ETH_LINK_SPEED_FIXED);
4238
4239         hw->mac.get_link_status = true;
4240
4241         if (intr->flags & IXGBE_FLAG_NEED_LINK_CONFIG)
4242                 return rte_eth_linkstatus_set(dev, &link);
4243
4244         /* check if it needs to wait to complete, if lsc interrupt is enabled */
4245         if (wait_to_complete == 0 || dev->data->dev_conf.intr_conf.lsc != 0)
4246                 wait = 0;
4247
4248 /* BSD has no interrupt mechanism, so force NIC status synchronization. */
4249 #ifdef RTE_EXEC_ENV_FREEBSD
4250         wait = 1;
4251 #endif
4252
4253         if (vf)
4254                 diag = ixgbevf_check_link(hw, &link_speed, &link_up, wait);
4255         else
4256                 diag = ixgbe_check_link(hw, &link_speed, &link_up, wait);
4257
4258         if (diag != 0) {
4259                 link.link_speed = ETH_SPEED_NUM_100M;
4260                 link.link_duplex = ETH_LINK_FULL_DUPLEX;
4261                 return rte_eth_linkstatus_set(dev, &link);
4262         }
4263
4264         if (ixgbe_get_media_type(hw) == ixgbe_media_type_fiber) {
4265                 esdp_reg = IXGBE_READ_REG(hw, IXGBE_ESDP);
4266                 if ((esdp_reg & IXGBE_ESDP_SDP3))
4267                         link_up = 0;
4268         }
4269
4270         if (link_up == 0) {
4271                 if (ixgbe_get_media_type(hw) == ixgbe_media_type_fiber) {
4272                         ixgbe_dev_wait_setup_link_complete(dev, 0);
4273                         if (rte_atomic32_test_and_set(&ad->link_thread_running)) {
4274                                 /* To avoid race condition between threads, set
4275                                  * the IXGBE_FLAG_NEED_LINK_CONFIG flag only
4276                                  * when there is no link thread running.
4277                                  */
4278                                 intr->flags |= IXGBE_FLAG_NEED_LINK_CONFIG;
4279                                 if (rte_ctrl_thread_create(&ad->link_thread_tid,
4280                                         "ixgbe-link-handler",
4281                                         NULL,
4282                                         ixgbe_dev_setup_link_thread_handler,
4283                                         dev) < 0) {
4284                                         PMD_DRV_LOG(ERR,
4285                                                 "Create link thread failed!");
4286                                         rte_atomic32_clear(&ad->link_thread_running);
4287                                 }
4288                         } else {
4289                                 PMD_DRV_LOG(ERR,
4290                                         "Other link thread is running now!");
4291                         }
4292                 }
4293                 return rte_eth_linkstatus_set(dev, &link);
4294         }
4295
4296         link.link_status = ETH_LINK_UP;
4297         link.link_duplex = ETH_LINK_FULL_DUPLEX;
4298
4299         switch (link_speed) {
4300         default:
4301         case IXGBE_LINK_SPEED_UNKNOWN:
4302                 link.link_speed = ETH_SPEED_NUM_UNKNOWN;
4303                 break;
4304
4305         case IXGBE_LINK_SPEED_10_FULL:
4306                 link.link_speed = ETH_SPEED_NUM_10M;
4307                 break;
4308
4309         case IXGBE_LINK_SPEED_100_FULL:
4310                 link.link_speed = ETH_SPEED_NUM_100M;
4311                 break;
4312
4313         case IXGBE_LINK_SPEED_1GB_FULL:
4314                 link.link_speed = ETH_SPEED_NUM_1G;
4315                 break;
4316
4317         case IXGBE_LINK_SPEED_2_5GB_FULL:
4318                 link.link_speed = ETH_SPEED_NUM_2_5G;
4319                 break;
4320
4321         case IXGBE_LINK_SPEED_5GB_FULL:
4322                 link.link_speed = ETH_SPEED_NUM_5G;
4323                 break;
4324
4325         case IXGBE_LINK_SPEED_10GB_FULL:
4326                 link.link_speed = ETH_SPEED_NUM_10G;
4327                 break;
4328         }
4329
4330         return rte_eth_linkstatus_set(dev, &link);
4331 }
4332
4333 static int
4334 ixgbe_dev_link_update(struct rte_eth_dev *dev, int wait_to_complete)
4335 {
4336         return ixgbe_dev_link_update_share(dev, wait_to_complete, 0);
4337 }
4338
4339 static int
4340 ixgbevf_dev_link_update(struct rte_eth_dev *dev, int wait_to_complete)
4341 {
4342         return ixgbe_dev_link_update_share(dev, wait_to_complete, 1);
4343 }
4344
4345 static int
4346 ixgbe_dev_promiscuous_enable(struct rte_eth_dev *dev)
4347 {
4348         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4349         uint32_t fctrl;
4350
4351         fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
4352         fctrl |= (IXGBE_FCTRL_UPE | IXGBE_FCTRL_MPE);
4353         IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
4354
4355         return 0;
4356 }
4357
4358 static int
4359 ixgbe_dev_promiscuous_disable(struct rte_eth_dev *dev)
4360 {
4361         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4362         uint32_t fctrl;
4363
4364         fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
4365         fctrl &= (~IXGBE_FCTRL_UPE);
4366         if (dev->data->all_multicast == 1)
4367                 fctrl |= IXGBE_FCTRL_MPE;
4368         else
4369                 fctrl &= (~IXGBE_FCTRL_MPE);
4370         IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
4371
4372         return 0;
4373 }
4374
4375 static int
4376 ixgbe_dev_allmulticast_enable(struct rte_eth_dev *dev)
4377 {
4378         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4379         uint32_t fctrl;
4380
4381         fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
4382         fctrl |= IXGBE_FCTRL_MPE;
4383         IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
4384
4385         return 0;
4386 }
4387
4388 static int
4389 ixgbe_dev_allmulticast_disable(struct rte_eth_dev *dev)
4390 {
4391         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4392         uint32_t fctrl;
4393
4394         if (dev->data->promiscuous == 1)
4395                 return 0; /* must remain in all_multicast mode */
4396
4397         fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
4398         fctrl &= (~IXGBE_FCTRL_MPE);
4399         IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
4400
4401         return 0;
4402 }
4403
4404 /**
4405  * It clears the interrupt causes and enables the interrupt.
4406  * It will be called once only during nic initialized.
4407  *
4408  * @param dev
4409  *  Pointer to struct rte_eth_dev.
4410  * @param on
4411  *  Enable or Disable.
4412  *
4413  * @return
4414  *  - On success, zero.
4415  *  - On failure, a negative value.
4416  */
4417 static int
4418 ixgbe_dev_lsc_interrupt_setup(struct rte_eth_dev *dev, uint8_t on)
4419 {
4420         struct ixgbe_interrupt *intr =
4421                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4422
4423         ixgbe_dev_link_status_print(dev);
4424         if (on)
4425                 intr->mask |= IXGBE_EICR_LSC;
4426         else
4427                 intr->mask &= ~IXGBE_EICR_LSC;
4428
4429         return 0;
4430 }
4431
4432 /**
4433  * It clears the interrupt causes and enables the interrupt.
4434  * It will be called once only during nic initialized.
4435  *
4436  * @param dev
4437  *  Pointer to struct rte_eth_dev.
4438  *
4439  * @return
4440  *  - On success, zero.
4441  *  - On failure, a negative value.
4442  */
4443 static int
4444 ixgbe_dev_rxq_interrupt_setup(struct rte_eth_dev *dev)
4445 {
4446         struct ixgbe_interrupt *intr =
4447                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4448
4449         intr->mask |= IXGBE_EICR_RTX_QUEUE;
4450
4451         return 0;
4452 }
4453
4454 /**
4455  * It clears the interrupt causes and enables the interrupt.
4456  * It will be called once only during nic initialized.
4457  *
4458  * @param dev
4459  *  Pointer to struct rte_eth_dev.
4460  *
4461  * @return
4462  *  - On success, zero.
4463  *  - On failure, a negative value.
4464  */
4465 static int
4466 ixgbe_dev_macsec_interrupt_setup(struct rte_eth_dev *dev)
4467 {
4468         struct ixgbe_interrupt *intr =
4469                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4470
4471         intr->mask |= IXGBE_EICR_LINKSEC;
4472
4473         return 0;
4474 }
4475
4476 /*
4477  * It reads ICR and sets flag (IXGBE_EICR_LSC) for the link_update.
4478  *
4479  * @param dev
4480  *  Pointer to struct rte_eth_dev.
4481  *
4482  * @return
4483  *  - On success, zero.
4484  *  - On failure, a negative value.
4485  */
4486 static int
4487 ixgbe_dev_interrupt_get_status(struct rte_eth_dev *dev)
4488 {
4489         uint32_t eicr;
4490         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4491         struct ixgbe_interrupt *intr =
4492                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4493
4494         /* clear all cause mask */
4495         ixgbe_disable_intr(hw);
4496
4497         /* read-on-clear nic registers here */
4498         eicr = IXGBE_READ_REG(hw, IXGBE_EICR);
4499         PMD_DRV_LOG(DEBUG, "eicr %x", eicr);
4500
4501         intr->flags = 0;
4502
4503         /* set flag for async link update */
4504         if (eicr & IXGBE_EICR_LSC)
4505                 intr->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
4506
4507         if (eicr & IXGBE_EICR_MAILBOX)
4508                 intr->flags |= IXGBE_FLAG_MAILBOX;
4509
4510         if (eicr & IXGBE_EICR_LINKSEC)
4511                 intr->flags |= IXGBE_FLAG_MACSEC;
4512
4513         if (hw->mac.type ==  ixgbe_mac_X550EM_x &&
4514             hw->phy.type == ixgbe_phy_x550em_ext_t &&
4515             (eicr & IXGBE_EICR_GPI_SDP0_X550EM_x))
4516                 intr->flags |= IXGBE_FLAG_PHY_INTERRUPT;
4517
4518         return 0;
4519 }
4520
4521 /**
4522  * It gets and then prints the link status.
4523  *
4524  * @param dev
4525  *  Pointer to struct rte_eth_dev.
4526  *
4527  * @return
4528  *  - On success, zero.
4529  *  - On failure, a negative value.
4530  */
4531 static void
4532 ixgbe_dev_link_status_print(struct rte_eth_dev *dev)
4533 {
4534         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
4535         struct rte_eth_link link;
4536
4537         rte_eth_linkstatus_get(dev, &link);
4538
4539         if (link.link_status) {
4540                 PMD_INIT_LOG(INFO, "Port %d: Link Up - speed %u Mbps - %s",
4541                                         (int)(dev->data->port_id),
4542                                         (unsigned)link.link_speed,
4543                         link.link_duplex == ETH_LINK_FULL_DUPLEX ?
4544                                         "full-duplex" : "half-duplex");
4545         } else {
4546                 PMD_INIT_LOG(INFO, " Port %d: Link Down",
4547                                 (int)(dev->data->port_id));
4548         }
4549         PMD_INIT_LOG(DEBUG, "PCI Address: " PCI_PRI_FMT,
4550                                 pci_dev->addr.domain,
4551                                 pci_dev->addr.bus,
4552                                 pci_dev->addr.devid,
4553                                 pci_dev->addr.function);
4554 }
4555
4556 /*
4557  * It executes link_update after knowing an interrupt occurred.
4558  *
4559  * @param dev
4560  *  Pointer to struct rte_eth_dev.
4561  *
4562  * @return
4563  *  - On success, zero.
4564  *  - On failure, a negative value.
4565  */
4566 static int
4567 ixgbe_dev_interrupt_action(struct rte_eth_dev *dev)
4568 {
4569         struct ixgbe_interrupt *intr =
4570                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4571         int64_t timeout;
4572         struct ixgbe_hw *hw =
4573                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4574
4575         PMD_DRV_LOG(DEBUG, "intr action type %d", intr->flags);
4576
4577         if (intr->flags & IXGBE_FLAG_MAILBOX) {
4578                 ixgbe_pf_mbx_process(dev);
4579                 intr->flags &= ~IXGBE_FLAG_MAILBOX;
4580         }
4581
4582         if (intr->flags & IXGBE_FLAG_PHY_INTERRUPT) {
4583                 ixgbe_handle_lasi(hw);
4584                 intr->flags &= ~IXGBE_FLAG_PHY_INTERRUPT;
4585         }
4586
4587         if (intr->flags & IXGBE_FLAG_NEED_LINK_UPDATE) {
4588                 struct rte_eth_link link;
4589
4590                 /* get the link status before link update, for predicting later */
4591                 rte_eth_linkstatus_get(dev, &link);
4592
4593                 ixgbe_dev_link_update(dev, 0);
4594
4595                 /* likely to up */
4596                 if (!link.link_status)
4597                         /* handle it 1 sec later, wait it being stable */
4598                         timeout = IXGBE_LINK_UP_CHECK_TIMEOUT;
4599                 /* likely to down */
4600                 else
4601                         /* handle it 4 sec later, wait it being stable */
4602                         timeout = IXGBE_LINK_DOWN_CHECK_TIMEOUT;
4603
4604                 ixgbe_dev_link_status_print(dev);
4605                 if (rte_eal_alarm_set(timeout * 1000,
4606                                       ixgbe_dev_interrupt_delayed_handler, (void *)dev) < 0)
4607                         PMD_DRV_LOG(ERR, "Error setting alarm");
4608                 else {
4609                         /* remember original mask */
4610                         intr->mask_original = intr->mask;
4611                         /* only disable lsc interrupt */
4612                         intr->mask &= ~IXGBE_EIMS_LSC;
4613                 }
4614         }
4615
4616         PMD_DRV_LOG(DEBUG, "enable intr immediately");
4617         ixgbe_enable_intr(dev);
4618
4619         return 0;
4620 }
4621
4622 /**
4623  * Interrupt handler which shall be registered for alarm callback for delayed
4624  * handling specific interrupt to wait for the stable nic state. As the
4625  * NIC interrupt state is not stable for ixgbe after link is just down,
4626  * it needs to wait 4 seconds to get the stable status.
4627  *
4628  * @param handle
4629  *  Pointer to interrupt handle.
4630  * @param param
4631  *  The address of parameter (struct rte_eth_dev *) regsitered before.
4632  *
4633  * @return
4634  *  void
4635  */
4636 static void
4637 ixgbe_dev_interrupt_delayed_handler(void *param)
4638 {
4639         struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
4640         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
4641         struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
4642         struct ixgbe_interrupt *intr =
4643                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4644         struct ixgbe_hw *hw =
4645                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4646         uint32_t eicr;
4647
4648         ixgbe_disable_intr(hw);
4649
4650         eicr = IXGBE_READ_REG(hw, IXGBE_EICR);
4651         if (eicr & IXGBE_EICR_MAILBOX)
4652                 ixgbe_pf_mbx_process(dev);
4653
4654         if (intr->flags & IXGBE_FLAG_PHY_INTERRUPT) {
4655                 ixgbe_handle_lasi(hw);
4656                 intr->flags &= ~IXGBE_FLAG_PHY_INTERRUPT;
4657         }
4658
4659         if (intr->flags & IXGBE_FLAG_NEED_LINK_UPDATE) {
4660                 ixgbe_dev_link_update(dev, 0);
4661                 intr->flags &= ~IXGBE_FLAG_NEED_LINK_UPDATE;
4662                 ixgbe_dev_link_status_print(dev);
4663                 rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_LSC, NULL);
4664         }
4665
4666         if (intr->flags & IXGBE_FLAG_MACSEC) {
4667                 rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_MACSEC, NULL);
4668                 intr->flags &= ~IXGBE_FLAG_MACSEC;
4669         }
4670
4671         /* restore original mask */
4672         intr->mask = intr->mask_original;
4673         intr->mask_original = 0;
4674
4675         PMD_DRV_LOG(DEBUG, "enable intr in delayed handler S[%08x]", eicr);
4676         ixgbe_enable_intr(dev);
4677         rte_intr_ack(intr_handle);
4678 }
4679
4680 /**
4681  * Interrupt handler triggered by NIC  for handling
4682  * specific interrupt.
4683  *
4684  * @param handle
4685  *  Pointer to interrupt handle.
4686  * @param param
4687  *  The address of parameter (struct rte_eth_dev *) regsitered before.
4688  *
4689  * @return
4690  *  void
4691  */
4692 static void
4693 ixgbe_dev_interrupt_handler(void *param)
4694 {
4695         struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
4696
4697         ixgbe_dev_interrupt_get_status(dev);
4698         ixgbe_dev_interrupt_action(dev);
4699 }
4700
4701 static int
4702 ixgbe_dev_led_on(struct rte_eth_dev *dev)
4703 {
4704         struct ixgbe_hw *hw;
4705
4706         hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4707         return ixgbe_led_on(hw, 0) == IXGBE_SUCCESS ? 0 : -ENOTSUP;
4708 }
4709
4710 static int
4711 ixgbe_dev_led_off(struct rte_eth_dev *dev)
4712 {
4713         struct ixgbe_hw *hw;
4714
4715         hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4716         return ixgbe_led_off(hw, 0) == IXGBE_SUCCESS ? 0 : -ENOTSUP;
4717 }
4718
4719 static int
4720 ixgbe_flow_ctrl_get(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
4721 {
4722         struct ixgbe_hw *hw;
4723         uint32_t mflcn_reg;
4724         uint32_t fccfg_reg;
4725         int rx_pause;
4726         int tx_pause;
4727
4728         hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4729
4730         fc_conf->pause_time = hw->fc.pause_time;
4731         fc_conf->high_water = hw->fc.high_water[0];
4732         fc_conf->low_water = hw->fc.low_water[0];
4733         fc_conf->send_xon = hw->fc.send_xon;
4734         fc_conf->autoneg = !hw->fc.disable_fc_autoneg;
4735
4736         /*
4737          * Return rx_pause status according to actual setting of
4738          * MFLCN register.
4739          */
4740         mflcn_reg = IXGBE_READ_REG(hw, IXGBE_MFLCN);
4741         if (mflcn_reg & IXGBE_MFLCN_PMCF)
4742                 fc_conf->mac_ctrl_frame_fwd = 1;
4743         else
4744                 fc_conf->mac_ctrl_frame_fwd = 0;
4745
4746         if (mflcn_reg & (IXGBE_MFLCN_RPFCE | IXGBE_MFLCN_RFCE))
4747                 rx_pause = 1;
4748         else
4749                 rx_pause = 0;
4750
4751         /*
4752          * Return tx_pause status according to actual setting of
4753          * FCCFG register.
4754          */
4755         fccfg_reg = IXGBE_READ_REG(hw, IXGBE_FCCFG);
4756         if (fccfg_reg & (IXGBE_FCCFG_TFCE_802_3X | IXGBE_FCCFG_TFCE_PRIORITY))
4757                 tx_pause = 1;
4758         else
4759                 tx_pause = 0;
4760
4761         if (rx_pause && tx_pause)
4762                 fc_conf->mode = RTE_FC_FULL;
4763         else if (rx_pause)
4764                 fc_conf->mode = RTE_FC_RX_PAUSE;
4765         else if (tx_pause)
4766                 fc_conf->mode = RTE_FC_TX_PAUSE;
4767         else
4768                 fc_conf->mode = RTE_FC_NONE;
4769
4770         return 0;
4771 }
4772
4773 static int
4774 ixgbe_flow_ctrl_set(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
4775 {
4776         struct ixgbe_hw *hw;
4777         struct ixgbe_adapter *adapter = dev->data->dev_private;
4778         int err;
4779         uint32_t rx_buf_size;
4780         uint32_t max_high_water;
4781         enum ixgbe_fc_mode rte_fcmode_2_ixgbe_fcmode[] = {
4782                 ixgbe_fc_none,
4783                 ixgbe_fc_rx_pause,
4784                 ixgbe_fc_tx_pause,
4785                 ixgbe_fc_full
4786         };
4787
4788         PMD_INIT_FUNC_TRACE();
4789
4790         hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4791         rx_buf_size = IXGBE_READ_REG(hw, IXGBE_RXPBSIZE(0));
4792         PMD_INIT_LOG(DEBUG, "Rx packet buffer size = 0x%x", rx_buf_size);
4793
4794         /*
4795          * At least reserve one Ethernet frame for watermark
4796          * high_water/low_water in kilo bytes for ixgbe
4797          */
4798         max_high_water = (rx_buf_size -
4799                         RTE_ETHER_MAX_LEN) >> IXGBE_RXPBSIZE_SHIFT;
4800         if ((fc_conf->high_water > max_high_water) ||
4801                 (fc_conf->high_water < fc_conf->low_water)) {
4802                 PMD_INIT_LOG(ERR, "Invalid high/low water setup value in KB");
4803                 PMD_INIT_LOG(ERR, "High_water must <= 0x%x", max_high_water);
4804                 return -EINVAL;
4805         }
4806
4807         hw->fc.requested_mode = rte_fcmode_2_ixgbe_fcmode[fc_conf->mode];
4808         hw->fc.pause_time     = fc_conf->pause_time;
4809         hw->fc.high_water[0]  = fc_conf->high_water;
4810         hw->fc.low_water[0]   = fc_conf->low_water;
4811         hw->fc.send_xon       = fc_conf->send_xon;
4812         hw->fc.disable_fc_autoneg = !fc_conf->autoneg;
4813         adapter->mac_ctrl_frame_fwd = fc_conf->mac_ctrl_frame_fwd;
4814
4815         err = ixgbe_flow_ctrl_enable(dev, hw);
4816         if (err < 0) {
4817                 PMD_INIT_LOG(ERR, "ixgbe_flow_ctrl_enable = 0x%x", err);
4818                 return -EIO;
4819         }
4820         return err;
4821 }
4822
4823 /**
4824  *  ixgbe_pfc_enable_generic - Enable flow control
4825  *  @hw: pointer to hardware structure
4826  *  @tc_num: traffic class number
4827  *  Enable flow control according to the current settings.
4828  */
4829 static int
4830 ixgbe_dcb_pfc_enable_generic(struct ixgbe_hw *hw, uint8_t tc_num)
4831 {
4832         int ret_val = 0;
4833         uint32_t mflcn_reg, fccfg_reg;
4834         uint32_t reg;
4835         uint32_t fcrtl, fcrth;
4836         uint8_t i;
4837         uint8_t nb_rx_en;
4838
4839         /* Validate the water mark configuration */
4840         if (!hw->fc.pause_time) {
4841                 ret_val = IXGBE_ERR_INVALID_LINK_SETTINGS;
4842                 goto out;
4843         }
4844
4845         /* Low water mark of zero causes XOFF floods */
4846         if (hw->fc.current_mode & ixgbe_fc_tx_pause) {
4847                  /* High/Low water can not be 0 */
4848                 if ((!hw->fc.high_water[tc_num]) || (!hw->fc.low_water[tc_num])) {
4849                         PMD_INIT_LOG(ERR, "Invalid water mark configuration");
4850                         ret_val = IXGBE_ERR_INVALID_LINK_SETTINGS;
4851                         goto out;
4852                 }
4853
4854                 if (hw->fc.low_water[tc_num] >= hw->fc.high_water[tc_num]) {
4855                         PMD_INIT_LOG(ERR, "Invalid water mark configuration");
4856                         ret_val = IXGBE_ERR_INVALID_LINK_SETTINGS;
4857                         goto out;
4858                 }
4859         }
4860         /* Negotiate the fc mode to use */
4861         ixgbe_fc_autoneg(hw);
4862
4863         /* Disable any previous flow control settings */
4864         mflcn_reg = IXGBE_READ_REG(hw, IXGBE_MFLCN);
4865         mflcn_reg &= ~(IXGBE_MFLCN_RPFCE_SHIFT | IXGBE_MFLCN_RFCE|IXGBE_MFLCN_RPFCE);
4866
4867         fccfg_reg = IXGBE_READ_REG(hw, IXGBE_FCCFG);
4868         fccfg_reg &= ~(IXGBE_FCCFG_TFCE_802_3X | IXGBE_FCCFG_TFCE_PRIORITY);
4869
4870         switch (hw->fc.current_mode) {
4871         case ixgbe_fc_none:
4872                 /*
4873                  * If the count of enabled RX Priority Flow control >1,
4874                  * and the TX pause can not be disabled
4875                  */
4876                 nb_rx_en = 0;
4877                 for (i = 0; i < IXGBE_DCB_MAX_TRAFFIC_CLASS; i++) {
4878                         reg = IXGBE_READ_REG(hw, IXGBE_FCRTH_82599(i));
4879                         if (reg & IXGBE_FCRTH_FCEN)
4880                                 nb_rx_en++;
4881                 }
4882                 if (nb_rx_en > 1)
4883                         fccfg_reg |= IXGBE_FCCFG_TFCE_PRIORITY;
4884                 break;
4885         case ixgbe_fc_rx_pause:
4886                 /*
4887                  * Rx Flow control is enabled and Tx Flow control is
4888                  * disabled by software override. Since there really
4889                  * isn't a way to advertise that we are capable of RX
4890                  * Pause ONLY, we will advertise that we support both
4891                  * symmetric and asymmetric Rx PAUSE.  Later, we will
4892                  * disable the adapter's ability to send PAUSE frames.
4893                  */
4894                 mflcn_reg |= IXGBE_MFLCN_RPFCE;
4895                 /*
4896                  * If the count of enabled RX Priority Flow control >1,
4897                  * and the TX pause can not be disabled
4898                  */
4899                 nb_rx_en = 0;
4900                 for (i = 0; i < IXGBE_DCB_MAX_TRAFFIC_CLASS; i++) {
4901                         reg = IXGBE_READ_REG(hw, IXGBE_FCRTH_82599(i));
4902                         if (reg & IXGBE_FCRTH_FCEN)
4903                                 nb_rx_en++;
4904                 }
4905                 if (nb_rx_en > 1)
4906                         fccfg_reg |= IXGBE_FCCFG_TFCE_PRIORITY;
4907                 break;
4908         case ixgbe_fc_tx_pause:
4909                 /*
4910                  * Tx Flow control is enabled, and Rx Flow control is
4911                  * disabled by software override.
4912                  */
4913                 fccfg_reg |= IXGBE_FCCFG_TFCE_PRIORITY;
4914                 break;
4915         case ixgbe_fc_full:
4916                 /* Flow control (both Rx and Tx) is enabled by SW override. */
4917                 mflcn_reg |= IXGBE_MFLCN_RPFCE;
4918                 fccfg_reg |= IXGBE_FCCFG_TFCE_PRIORITY;
4919                 break;
4920         default:
4921                 PMD_DRV_LOG(DEBUG, "Flow control param set incorrectly");
4922                 ret_val = IXGBE_ERR_CONFIG;
4923                 goto out;
4924         }
4925
4926         /* Set 802.3x based flow control settings. */
4927         mflcn_reg |= IXGBE_MFLCN_DPF;
4928         IXGBE_WRITE_REG(hw, IXGBE_MFLCN, mflcn_reg);
4929         IXGBE_WRITE_REG(hw, IXGBE_FCCFG, fccfg_reg);
4930
4931         /* Set up and enable Rx high/low water mark thresholds, enable XON. */
4932         if ((hw->fc.current_mode & ixgbe_fc_tx_pause) &&
4933                 hw->fc.high_water[tc_num]) {
4934                 fcrtl = (hw->fc.low_water[tc_num] << 10) | IXGBE_FCRTL_XONE;
4935                 IXGBE_WRITE_REG(hw, IXGBE_FCRTL_82599(tc_num), fcrtl);
4936                 fcrth = (hw->fc.high_water[tc_num] << 10) | IXGBE_FCRTH_FCEN;
4937         } else {
4938                 IXGBE_WRITE_REG(hw, IXGBE_FCRTL_82599(tc_num), 0);
4939                 /*
4940                  * In order to prevent Tx hangs when the internal Tx
4941                  * switch is enabled we must set the high water mark
4942                  * to the maximum FCRTH value.  This allows the Tx
4943                  * switch to function even under heavy Rx workloads.
4944                  */
4945                 fcrth = IXGBE_READ_REG(hw, IXGBE_RXPBSIZE(tc_num)) - 32;
4946         }
4947         IXGBE_WRITE_REG(hw, IXGBE_FCRTH_82599(tc_num), fcrth);
4948
4949         /* Configure pause time (2 TCs per register) */
4950         reg = hw->fc.pause_time * 0x00010001;
4951         for (i = 0; i < (IXGBE_DCB_MAX_TRAFFIC_CLASS / 2); i++)
4952                 IXGBE_WRITE_REG(hw, IXGBE_FCTTV(i), reg);
4953
4954         /* Configure flow control refresh threshold value */
4955         IXGBE_WRITE_REG(hw, IXGBE_FCRTV, hw->fc.pause_time / 2);
4956
4957 out:
4958         return ret_val;
4959 }
4960
4961 static int
4962 ixgbe_dcb_pfc_enable(struct rte_eth_dev *dev, uint8_t tc_num)
4963 {
4964         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4965         int32_t ret_val = IXGBE_NOT_IMPLEMENTED;
4966
4967         if (hw->mac.type != ixgbe_mac_82598EB) {
4968                 ret_val = ixgbe_dcb_pfc_enable_generic(hw, tc_num);
4969         }
4970         return ret_val;
4971 }
4972
4973 static int
4974 ixgbe_priority_flow_ctrl_set(struct rte_eth_dev *dev, struct rte_eth_pfc_conf *pfc_conf)
4975 {
4976         int err;
4977         uint32_t rx_buf_size;
4978         uint32_t max_high_water;
4979         uint8_t tc_num;
4980         uint8_t  map[IXGBE_DCB_MAX_USER_PRIORITY] = { 0 };
4981         struct ixgbe_hw *hw =
4982                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4983         struct ixgbe_dcb_config *dcb_config =
4984                 IXGBE_DEV_PRIVATE_TO_DCB_CFG(dev->data->dev_private);
4985
4986         enum ixgbe_fc_mode rte_fcmode_2_ixgbe_fcmode[] = {
4987                 ixgbe_fc_none,
4988                 ixgbe_fc_rx_pause,
4989                 ixgbe_fc_tx_pause,
4990                 ixgbe_fc_full
4991         };
4992
4993         PMD_INIT_FUNC_TRACE();
4994
4995         ixgbe_dcb_unpack_map_cee(dcb_config, IXGBE_DCB_RX_CONFIG, map);
4996         tc_num = map[pfc_conf->priority];
4997         rx_buf_size = IXGBE_READ_REG(hw, IXGBE_RXPBSIZE(tc_num));
4998         PMD_INIT_LOG(DEBUG, "Rx packet buffer size = 0x%x", rx_buf_size);
4999         /*
5000          * At least reserve one Ethernet frame for watermark
5001          * high_water/low_water in kilo bytes for ixgbe
5002          */
5003         max_high_water = (rx_buf_size -
5004                         RTE_ETHER_MAX_LEN) >> IXGBE_RXPBSIZE_SHIFT;
5005         if ((pfc_conf->fc.high_water > max_high_water) ||
5006             (pfc_conf->fc.high_water <= pfc_conf->fc.low_water)) {
5007                 PMD_INIT_LOG(ERR, "Invalid high/low water setup value in KB");
5008                 PMD_INIT_LOG(ERR, "High_water must <= 0x%x", max_high_water);
5009                 return -EINVAL;
5010         }
5011
5012         hw->fc.requested_mode = rte_fcmode_2_ixgbe_fcmode[pfc_conf->fc.mode];
5013         hw->fc.pause_time = pfc_conf->fc.pause_time;
5014         hw->fc.send_xon = pfc_conf->fc.send_xon;
5015         hw->fc.low_water[tc_num] =  pfc_conf->fc.low_water;
5016         hw->fc.high_water[tc_num] = pfc_conf->fc.high_water;
5017
5018         err = ixgbe_dcb_pfc_enable(dev, tc_num);
5019
5020         /* Not negotiated is not an error case */
5021         if ((err == IXGBE_SUCCESS) || (err == IXGBE_ERR_FC_NOT_NEGOTIATED))
5022                 return 0;
5023
5024         PMD_INIT_LOG(ERR, "ixgbe_dcb_pfc_enable = 0x%x", err);
5025         return -EIO;
5026 }
5027
5028 static int
5029 ixgbe_dev_rss_reta_update(struct rte_eth_dev *dev,
5030                           struct rte_eth_rss_reta_entry64 *reta_conf,
5031                           uint16_t reta_size)
5032 {
5033         uint16_t i, sp_reta_size;
5034         uint8_t j, mask;
5035         uint32_t reta, r;
5036         uint16_t idx, shift;
5037         struct ixgbe_adapter *adapter = dev->data->dev_private;
5038         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5039         uint32_t reta_reg;
5040
5041         PMD_INIT_FUNC_TRACE();
5042
5043         if (!ixgbe_rss_update_sp(hw->mac.type)) {
5044                 PMD_DRV_LOG(ERR, "RSS reta update is not supported on this "
5045                         "NIC.");
5046                 return -ENOTSUP;
5047         }
5048
5049         sp_reta_size = ixgbe_reta_size_get(hw->mac.type);
5050         if (reta_size != sp_reta_size) {
5051                 PMD_DRV_LOG(ERR, "The size of hash lookup table configured "
5052                         "(%d) doesn't match the number hardware can supported "
5053                         "(%d)", reta_size, sp_reta_size);
5054                 return -EINVAL;
5055         }
5056
5057         for (i = 0; i < reta_size; i += IXGBE_4_BIT_WIDTH) {
5058                 idx = i / RTE_RETA_GROUP_SIZE;
5059                 shift = i % RTE_RETA_GROUP_SIZE;
5060                 mask = (uint8_t)((reta_conf[idx].mask >> shift) &
5061                                                 IXGBE_4_BIT_MASK);
5062                 if (!mask)
5063                         continue;
5064                 reta_reg = ixgbe_reta_reg_get(hw->mac.type, i);
5065                 if (mask == IXGBE_4_BIT_MASK)
5066                         r = 0;
5067                 else
5068                         r = IXGBE_READ_REG(hw, reta_reg);
5069                 for (j = 0, reta = 0; j < IXGBE_4_BIT_WIDTH; j++) {
5070                         if (mask & (0x1 << j))
5071                                 reta |= reta_conf[idx].reta[shift + j] <<
5072                                                         (CHAR_BIT * j);
5073                         else
5074                                 reta |= r & (IXGBE_8_BIT_MASK <<
5075                                                 (CHAR_BIT * j));
5076                 }
5077                 IXGBE_WRITE_REG(hw, reta_reg, reta);
5078         }
5079         adapter->rss_reta_updated = 1;
5080
5081         return 0;
5082 }
5083
5084 static int
5085 ixgbe_dev_rss_reta_query(struct rte_eth_dev *dev,
5086                          struct rte_eth_rss_reta_entry64 *reta_conf,
5087                          uint16_t reta_size)
5088 {
5089         uint16_t i, sp_reta_size;
5090         uint8_t j, mask;
5091         uint32_t reta;
5092         uint16_t idx, shift;
5093         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5094         uint32_t reta_reg;
5095
5096         PMD_INIT_FUNC_TRACE();
5097         sp_reta_size = ixgbe_reta_size_get(hw->mac.type);
5098         if (reta_size != sp_reta_size) {
5099                 PMD_DRV_LOG(ERR, "The size of hash lookup table configured "
5100                         "(%d) doesn't match the number hardware can supported "
5101                         "(%d)", reta_size, sp_reta_size);
5102                 return -EINVAL;
5103         }
5104
5105         for (i = 0; i < reta_size; i += IXGBE_4_BIT_WIDTH) {
5106                 idx = i / RTE_RETA_GROUP_SIZE;
5107                 shift = i % RTE_RETA_GROUP_SIZE;
5108                 mask = (uint8_t)((reta_conf[idx].mask >> shift) &
5109                                                 IXGBE_4_BIT_MASK);
5110                 if (!mask)
5111                         continue;
5112
5113                 reta_reg = ixgbe_reta_reg_get(hw->mac.type, i);
5114                 reta = IXGBE_READ_REG(hw, reta_reg);
5115                 for (j = 0; j < IXGBE_4_BIT_WIDTH; j++) {
5116                         if (mask & (0x1 << j))
5117                                 reta_conf[idx].reta[shift + j] =
5118                                         ((reta >> (CHAR_BIT * j)) &
5119                                                 IXGBE_8_BIT_MASK);
5120                 }
5121         }
5122
5123         return 0;
5124 }
5125
5126 static int
5127 ixgbe_add_rar(struct rte_eth_dev *dev, struct rte_ether_addr *mac_addr,
5128                                 uint32_t index, uint32_t pool)
5129 {
5130         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5131         uint32_t enable_addr = 1;
5132
5133         return ixgbe_set_rar(hw, index, mac_addr->addr_bytes,
5134                              pool, enable_addr);
5135 }
5136
5137 static void
5138 ixgbe_remove_rar(struct rte_eth_dev *dev, uint32_t index)
5139 {
5140         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5141
5142         ixgbe_clear_rar(hw, index);
5143 }
5144
5145 static int
5146 ixgbe_set_default_mac_addr(struct rte_eth_dev *dev, struct rte_ether_addr *addr)
5147 {
5148         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5149
5150         ixgbe_remove_rar(dev, 0);
5151         ixgbe_add_rar(dev, addr, 0, pci_dev->max_vfs);
5152
5153         return 0;
5154 }
5155
5156 static bool
5157 is_device_supported(struct rte_eth_dev *dev, struct rte_pci_driver *drv)
5158 {
5159         if (strcmp(dev->device->driver->name, drv->driver.name))
5160                 return false;
5161
5162         return true;
5163 }
5164
5165 bool
5166 is_ixgbe_supported(struct rte_eth_dev *dev)
5167 {
5168         return is_device_supported(dev, &rte_ixgbe_pmd);
5169 }
5170
5171 static int
5172 ixgbe_dev_mtu_set(struct rte_eth_dev *dev, uint16_t mtu)
5173 {
5174         uint32_t hlreg0;
5175         uint32_t maxfrs;
5176         struct ixgbe_hw *hw;
5177         struct rte_eth_dev_info dev_info;
5178         uint32_t frame_size = mtu + IXGBE_ETH_OVERHEAD;
5179         struct rte_eth_dev_data *dev_data = dev->data;
5180         int ret;
5181
5182         ret = ixgbe_dev_info_get(dev, &dev_info);
5183         if (ret != 0)
5184                 return ret;
5185
5186         /* check that mtu is within the allowed range */
5187         if (mtu < RTE_ETHER_MIN_MTU || frame_size > dev_info.max_rx_pktlen)
5188                 return -EINVAL;
5189
5190         /* If device is started, refuse mtu that requires the support of
5191          * scattered packets when this feature has not been enabled before.
5192          */
5193         if (dev_data->dev_started && !dev_data->scattered_rx &&
5194             (frame_size + 2 * IXGBE_VLAN_TAG_SIZE >
5195              dev->data->min_rx_buf_size - RTE_PKTMBUF_HEADROOM)) {
5196                 PMD_INIT_LOG(ERR, "Stop port first.");
5197                 return -EINVAL;
5198         }
5199
5200         hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5201         hlreg0 = IXGBE_READ_REG(hw, IXGBE_HLREG0);
5202
5203         /* switch to jumbo mode if needed */
5204         if (frame_size > RTE_ETHER_MAX_LEN) {
5205                 dev->data->dev_conf.rxmode.offloads |=
5206                         DEV_RX_OFFLOAD_JUMBO_FRAME;
5207                 hlreg0 |= IXGBE_HLREG0_JUMBOEN;
5208         } else {
5209                 dev->data->dev_conf.rxmode.offloads &=
5210                         ~DEV_RX_OFFLOAD_JUMBO_FRAME;
5211                 hlreg0 &= ~IXGBE_HLREG0_JUMBOEN;
5212         }
5213         IXGBE_WRITE_REG(hw, IXGBE_HLREG0, hlreg0);
5214
5215         /* update max frame size */
5216         dev->data->dev_conf.rxmode.max_rx_pkt_len = frame_size;
5217
5218         maxfrs = IXGBE_READ_REG(hw, IXGBE_MAXFRS);
5219         maxfrs &= 0x0000FFFF;
5220         maxfrs |= (dev->data->dev_conf.rxmode.max_rx_pkt_len << 16);
5221         IXGBE_WRITE_REG(hw, IXGBE_MAXFRS, maxfrs);
5222
5223         return 0;
5224 }
5225
5226 /*
5227  * Virtual Function operations
5228  */
5229 static void
5230 ixgbevf_intr_disable(struct rte_eth_dev *dev)
5231 {
5232         struct ixgbe_interrupt *intr =
5233                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
5234         struct ixgbe_hw *hw =
5235                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5236
5237         PMD_INIT_FUNC_TRACE();
5238
5239         /* Clear interrupt mask to stop from interrupts being generated */
5240         IXGBE_WRITE_REG(hw, IXGBE_VTEIMC, IXGBE_VF_IRQ_CLEAR_MASK);
5241
5242         IXGBE_WRITE_FLUSH(hw);
5243
5244         /* Clear mask value. */
5245         intr->mask = 0;
5246 }
5247
5248 static void
5249 ixgbevf_intr_enable(struct rte_eth_dev *dev)
5250 {
5251         struct ixgbe_interrupt *intr =
5252                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
5253         struct ixgbe_hw *hw =
5254                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5255
5256         PMD_INIT_FUNC_TRACE();
5257
5258         /* VF enable interrupt autoclean */
5259         IXGBE_WRITE_REG(hw, IXGBE_VTEIAM, IXGBE_VF_IRQ_ENABLE_MASK);
5260         IXGBE_WRITE_REG(hw, IXGBE_VTEIAC, IXGBE_VF_IRQ_ENABLE_MASK);
5261         IXGBE_WRITE_REG(hw, IXGBE_VTEIMS, IXGBE_VF_IRQ_ENABLE_MASK);
5262
5263         IXGBE_WRITE_FLUSH(hw);
5264
5265         /* Save IXGBE_VTEIMS value to mask. */
5266         intr->mask = IXGBE_VF_IRQ_ENABLE_MASK;
5267 }
5268
5269 static int
5270 ixgbevf_dev_configure(struct rte_eth_dev *dev)
5271 {
5272         struct rte_eth_conf *conf = &dev->data->dev_conf;
5273         struct ixgbe_adapter *adapter = dev->data->dev_private;
5274
5275         PMD_INIT_LOG(DEBUG, "Configured Virtual Function port id: %d",
5276                      dev->data->port_id);
5277
5278         if (dev->data->dev_conf.rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG)
5279                 dev->data->dev_conf.rxmode.offloads |= DEV_RX_OFFLOAD_RSS_HASH;
5280
5281         /*
5282          * VF has no ability to enable/disable HW CRC
5283          * Keep the persistent behavior the same as Host PF
5284          */
5285 #ifndef RTE_LIBRTE_IXGBE_PF_DISABLE_STRIP_CRC
5286         if (conf->rxmode.offloads & DEV_RX_OFFLOAD_KEEP_CRC) {
5287                 PMD_INIT_LOG(NOTICE, "VF can't disable HW CRC Strip");
5288                 conf->rxmode.offloads &= ~DEV_RX_OFFLOAD_KEEP_CRC;
5289         }
5290 #else
5291         if (!(conf->rxmode.offloads & DEV_RX_OFFLOAD_KEEP_CRC)) {
5292                 PMD_INIT_LOG(NOTICE, "VF can't enable HW CRC Strip");
5293                 conf->rxmode.offloads |= DEV_RX_OFFLOAD_KEEP_CRC;
5294         }
5295 #endif
5296
5297         /*
5298          * Initialize to TRUE. If any of Rx queues doesn't meet the bulk
5299          * allocation or vector Rx preconditions we will reset it.
5300          */
5301         adapter->rx_bulk_alloc_allowed = true;
5302         adapter->rx_vec_allowed = true;
5303
5304         return 0;
5305 }
5306
5307 static int
5308 ixgbevf_dev_start(struct rte_eth_dev *dev)
5309 {
5310         struct ixgbe_hw *hw =
5311                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5312         uint32_t intr_vector = 0;
5313         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5314         struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
5315
5316         int err, mask = 0;
5317
5318         PMD_INIT_FUNC_TRACE();
5319
5320         /* Stop the link setup handler before resetting the HW. */
5321         ixgbe_dev_wait_setup_link_complete(dev, 0);
5322
5323         err = hw->mac.ops.reset_hw(hw);
5324
5325         /**
5326          * In this case, reuses the MAC address assigned by VF
5327          * initialization.
5328          */
5329         if (err != IXGBE_SUCCESS && err != IXGBE_ERR_INVALID_MAC_ADDR) {
5330                 PMD_INIT_LOG(ERR, "Unable to reset vf hardware (%d)", err);
5331                 return err;
5332         }
5333
5334         hw->mac.get_link_status = true;
5335
5336         /* negotiate mailbox API version to use with the PF. */
5337         ixgbevf_negotiate_api(hw);
5338
5339         ixgbevf_dev_tx_init(dev);
5340
5341         /* This can fail when allocating mbufs for descriptor rings */
5342         err = ixgbevf_dev_rx_init(dev);
5343         if (err) {
5344                 PMD_INIT_LOG(ERR, "Unable to initialize RX hardware (%d)", err);
5345                 ixgbe_dev_clear_queues(dev);
5346                 return err;
5347         }
5348
5349         /* Set vfta */
5350         ixgbevf_set_vfta_all(dev, 1);
5351
5352         /* Set HW strip */
5353         mask = ETH_VLAN_STRIP_MASK | ETH_VLAN_FILTER_MASK |
5354                 ETH_VLAN_EXTEND_MASK;
5355         err = ixgbevf_vlan_offload_config(dev, mask);
5356         if (err) {
5357                 PMD_INIT_LOG(ERR, "Unable to set VLAN offload (%d)", err);
5358                 ixgbe_dev_clear_queues(dev);
5359                 return err;
5360         }
5361
5362         ixgbevf_dev_rxtx_start(dev);
5363
5364         /* check and configure queue intr-vector mapping */
5365         if (rte_intr_cap_multiple(intr_handle) &&
5366             dev->data->dev_conf.intr_conf.rxq) {
5367                 /* According to datasheet, only vector 0/1/2 can be used,
5368                  * now only one vector is used for Rx queue
5369                  */
5370                 intr_vector = 1;
5371                 if (rte_intr_efd_enable(intr_handle, intr_vector))
5372                         return -1;
5373         }
5374
5375         if (rte_intr_dp_is_en(intr_handle) && !intr_handle->intr_vec) {
5376                 intr_handle->intr_vec =
5377                         rte_zmalloc("intr_vec",
5378                                     dev->data->nb_rx_queues * sizeof(int), 0);
5379                 if (intr_handle->intr_vec == NULL) {
5380                         PMD_INIT_LOG(ERR, "Failed to allocate %d rx_queues"
5381                                      " intr_vec", dev->data->nb_rx_queues);
5382                         return -ENOMEM;
5383                 }
5384         }
5385         ixgbevf_configure_msix(dev);
5386
5387         /* When a VF port is bound to VFIO-PCI, only miscellaneous interrupt
5388          * is mapped to VFIO vector 0 in eth_ixgbevf_dev_init( ).
5389          * If previous VFIO interrupt mapping setting in eth_ixgbevf_dev_init( )
5390          * is not cleared, it will fail when following rte_intr_enable( ) tries
5391          * to map Rx queue interrupt to other VFIO vectors.
5392          * So clear uio/vfio intr/evevnfd first to avoid failure.
5393          */
5394         rte_intr_disable(intr_handle);
5395
5396         rte_intr_enable(intr_handle);
5397
5398         /* Re-enable interrupt for VF */
5399         ixgbevf_intr_enable(dev);
5400
5401         /*
5402          * Update link status right before return, because it may
5403          * start link configuration process in a separate thread.
5404          */
5405         ixgbevf_dev_link_update(dev, 0);
5406
5407         hw->adapter_stopped = false;
5408
5409         return 0;
5410 }
5411
5412 static void
5413 ixgbevf_dev_stop(struct rte_eth_dev *dev)
5414 {
5415         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5416         struct ixgbe_adapter *adapter = dev->data->dev_private;
5417         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5418         struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
5419
5420         if (hw->adapter_stopped)
5421                 return;
5422
5423         PMD_INIT_FUNC_TRACE();
5424
5425         ixgbe_dev_wait_setup_link_complete(dev, 0);
5426
5427         ixgbevf_intr_disable(dev);
5428
5429         hw->adapter_stopped = 1;
5430         ixgbe_stop_adapter(hw);
5431
5432         /*
5433           * Clear what we set, but we still keep shadow_vfta to
5434           * restore after device starts
5435           */
5436         ixgbevf_set_vfta_all(dev, 0);
5437
5438         /* Clear stored conf */
5439         dev->data->scattered_rx = 0;
5440
5441         ixgbe_dev_clear_queues(dev);
5442
5443         /* Clean datapath event and queue/vec mapping */
5444         rte_intr_efd_disable(intr_handle);
5445         if (intr_handle->intr_vec != NULL) {
5446                 rte_free(intr_handle->intr_vec);
5447                 intr_handle->intr_vec = NULL;
5448         }
5449
5450         adapter->rss_reta_updated = 0;
5451 }
5452
5453 static int
5454 ixgbevf_dev_close(struct rte_eth_dev *dev)
5455 {
5456         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5457         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5458         struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
5459
5460         PMD_INIT_FUNC_TRACE();
5461
5462         ixgbe_reset_hw(hw);
5463
5464         ixgbevf_dev_stop(dev);
5465
5466         ixgbe_dev_free_queues(dev);
5467
5468         /**
5469          * Remove the VF MAC address ro ensure
5470          * that the VF traffic goes to the PF
5471          * after stop, close and detach of the VF
5472          **/
5473         ixgbevf_remove_mac_addr(dev, 0);
5474
5475         dev->dev_ops = NULL;
5476         dev->rx_pkt_burst = NULL;
5477         dev->tx_pkt_burst = NULL;
5478
5479         rte_intr_disable(intr_handle);
5480         rte_intr_callback_unregister(intr_handle,
5481                                      ixgbevf_dev_interrupt_handler, dev);
5482
5483         return 0;
5484 }
5485
5486 /*
5487  * Reset VF device
5488  */
5489 static int
5490 ixgbevf_dev_reset(struct rte_eth_dev *dev)
5491 {
5492         int ret;
5493
5494         ret = eth_ixgbevf_dev_uninit(dev);
5495         if (ret)
5496                 return ret;
5497
5498         ret = eth_ixgbevf_dev_init(dev);
5499
5500         return ret;
5501 }
5502
5503 static void ixgbevf_set_vfta_all(struct rte_eth_dev *dev, bool on)
5504 {
5505         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5506         struct ixgbe_vfta *shadow_vfta =
5507                 IXGBE_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
5508         int i = 0, j = 0, vfta = 0, mask = 1;
5509
5510         for (i = 0; i < IXGBE_VFTA_SIZE; i++) {
5511                 vfta = shadow_vfta->vfta[i];
5512                 if (vfta) {
5513                         mask = 1;
5514                         for (j = 0; j < 32; j++) {
5515                                 if (vfta & mask)
5516                                         ixgbe_set_vfta(hw, (i<<5)+j, 0,
5517                                                        on, false);
5518                                 mask <<= 1;
5519                         }
5520                 }
5521         }
5522
5523 }
5524
5525 static int
5526 ixgbevf_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
5527 {
5528         struct ixgbe_hw *hw =
5529                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5530         struct ixgbe_vfta *shadow_vfta =
5531                 IXGBE_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
5532         uint32_t vid_idx = 0;
5533         uint32_t vid_bit = 0;
5534         int ret = 0;
5535
5536         PMD_INIT_FUNC_TRACE();
5537
5538         /* vind is not used in VF driver, set to 0, check ixgbe_set_vfta_vf */
5539         ret = ixgbe_set_vfta(hw, vlan_id, 0, !!on, false);
5540         if (ret) {
5541                 PMD_INIT_LOG(ERR, "Unable to set VF vlan");
5542                 return ret;
5543         }
5544         vid_idx = (uint32_t) ((vlan_id >> 5) & 0x7F);
5545         vid_bit = (uint32_t) (1 << (vlan_id & 0x1F));
5546
5547         /* Save what we set and retore it after device reset */
5548         if (on)
5549                 shadow_vfta->vfta[vid_idx] |= vid_bit;
5550         else
5551                 shadow_vfta->vfta[vid_idx] &= ~vid_bit;
5552
5553         return 0;
5554 }
5555
5556 static void
5557 ixgbevf_vlan_strip_queue_set(struct rte_eth_dev *dev, uint16_t queue, int on)
5558 {
5559         struct ixgbe_hw *hw =
5560                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5561         uint32_t ctrl;
5562
5563         PMD_INIT_FUNC_TRACE();
5564
5565         if (queue >= hw->mac.max_rx_queues)
5566                 return;
5567
5568         ctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(queue));
5569         if (on)
5570                 ctrl |= IXGBE_RXDCTL_VME;
5571         else
5572                 ctrl &= ~IXGBE_RXDCTL_VME;
5573         IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(queue), ctrl);
5574
5575         ixgbe_vlan_hw_strip_bitmap_set(dev, queue, on);
5576 }
5577
5578 static int
5579 ixgbevf_vlan_offload_config(struct rte_eth_dev *dev, int mask)
5580 {
5581         struct ixgbe_rx_queue *rxq;
5582         uint16_t i;
5583         int on = 0;
5584
5585         /* VF function only support hw strip feature, others are not support */
5586         if (mask & ETH_VLAN_STRIP_MASK) {
5587                 for (i = 0; i < dev->data->nb_rx_queues; i++) {
5588                         rxq = dev->data->rx_queues[i];
5589                         on = !!(rxq->offloads & DEV_RX_OFFLOAD_VLAN_STRIP);
5590                         ixgbevf_vlan_strip_queue_set(dev, i, on);
5591                 }
5592         }
5593
5594         return 0;
5595 }
5596
5597 static int
5598 ixgbevf_vlan_offload_set(struct rte_eth_dev *dev, int mask)
5599 {
5600         ixgbe_config_vlan_strip_on_all_queues(dev, mask);
5601
5602         ixgbevf_vlan_offload_config(dev, mask);
5603
5604         return 0;
5605 }
5606
5607 int
5608 ixgbe_vt_check(struct ixgbe_hw *hw)
5609 {
5610         uint32_t reg_val;
5611
5612         /* if Virtualization Technology is enabled */
5613         reg_val = IXGBE_READ_REG(hw, IXGBE_VT_CTL);
5614         if (!(reg_val & IXGBE_VT_CTL_VT_ENABLE)) {
5615                 PMD_INIT_LOG(ERR, "VT must be enabled for this setting");
5616                 return -1;
5617         }
5618
5619         return 0;
5620 }
5621
5622 static uint32_t
5623 ixgbe_uta_vector(struct ixgbe_hw *hw, struct rte_ether_addr *uc_addr)
5624 {
5625         uint32_t vector = 0;
5626
5627         switch (hw->mac.mc_filter_type) {
5628         case 0:   /* use bits [47:36] of the address */
5629                 vector = ((uc_addr->addr_bytes[4] >> 4) |
5630                         (((uint16_t)uc_addr->addr_bytes[5]) << 4));
5631                 break;
5632         case 1:   /* use bits [46:35] of the address */
5633                 vector = ((uc_addr->addr_bytes[4] >> 3) |
5634                         (((uint16_t)uc_addr->addr_bytes[5]) << 5));
5635                 break;
5636         case 2:   /* use bits [45:34] of the address */
5637                 vector = ((uc_addr->addr_bytes[4] >> 2) |
5638                         (((uint16_t)uc_addr->addr_bytes[5]) << 6));
5639                 break;
5640         case 3:   /* use bits [43:32] of the address */
5641                 vector = ((uc_addr->addr_bytes[4]) |
5642                         (((uint16_t)uc_addr->addr_bytes[5]) << 8));
5643                 break;
5644         default:  /* Invalid mc_filter_type */
5645                 break;
5646         }
5647
5648         /* vector can only be 12-bits or boundary will be exceeded */
5649         vector &= 0xFFF;
5650         return vector;
5651 }
5652
5653 static int
5654 ixgbe_uc_hash_table_set(struct rte_eth_dev *dev,
5655                         struct rte_ether_addr *mac_addr, uint8_t on)
5656 {
5657         uint32_t vector;
5658         uint32_t uta_idx;
5659         uint32_t reg_val;
5660         uint32_t uta_shift;
5661         uint32_t rc;
5662         const uint32_t ixgbe_uta_idx_mask = 0x7F;
5663         const uint32_t ixgbe_uta_bit_shift = 5;
5664         const uint32_t ixgbe_uta_bit_mask = (0x1 << ixgbe_uta_bit_shift) - 1;
5665         const uint32_t bit1 = 0x1;
5666
5667         struct ixgbe_hw *hw =
5668                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5669         struct ixgbe_uta_info *uta_info =
5670                 IXGBE_DEV_PRIVATE_TO_UTA(dev->data->dev_private);
5671
5672         /* The UTA table only exists on 82599 hardware and newer */
5673         if (hw->mac.type < ixgbe_mac_82599EB)
5674                 return -ENOTSUP;
5675
5676         vector = ixgbe_uta_vector(hw, mac_addr);
5677         uta_idx = (vector >> ixgbe_uta_bit_shift) & ixgbe_uta_idx_mask;
5678         uta_shift = vector & ixgbe_uta_bit_mask;
5679
5680         rc = ((uta_info->uta_shadow[uta_idx] >> uta_shift & bit1) != 0);
5681         if (rc == on)
5682                 return 0;
5683
5684         reg_val = IXGBE_READ_REG(hw, IXGBE_UTA(uta_idx));
5685         if (on) {
5686                 uta_info->uta_in_use++;
5687                 reg_val |= (bit1 << uta_shift);
5688                 uta_info->uta_shadow[uta_idx] |= (bit1 << uta_shift);
5689         } else {
5690                 uta_info->uta_in_use--;
5691                 reg_val &= ~(bit1 << uta_shift);
5692                 uta_info->uta_shadow[uta_idx] &= ~(bit1 << uta_shift);
5693         }
5694
5695         IXGBE_WRITE_REG(hw, IXGBE_UTA(uta_idx), reg_val);
5696
5697         if (uta_info->uta_in_use > 0)
5698                 IXGBE_WRITE_REG(hw, IXGBE_MCSTCTRL,
5699                                 IXGBE_MCSTCTRL_MFE | hw->mac.mc_filter_type);
5700         else
5701                 IXGBE_WRITE_REG(hw, IXGBE_MCSTCTRL, hw->mac.mc_filter_type);
5702
5703         return 0;
5704 }
5705
5706 static int
5707 ixgbe_uc_all_hash_table_set(struct rte_eth_dev *dev, uint8_t on)
5708 {
5709         int i;
5710         struct ixgbe_hw *hw =
5711                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5712         struct ixgbe_uta_info *uta_info =
5713                 IXGBE_DEV_PRIVATE_TO_UTA(dev->data->dev_private);
5714
5715         /* The UTA table only exists on 82599 hardware and newer */
5716         if (hw->mac.type < ixgbe_mac_82599EB)
5717                 return -ENOTSUP;
5718
5719         if (on) {
5720                 for (i = 0; i < ETH_VMDQ_NUM_UC_HASH_ARRAY; i++) {
5721                         uta_info->uta_shadow[i] = ~0;
5722                         IXGBE_WRITE_REG(hw, IXGBE_UTA(i), ~0);
5723                 }
5724         } else {
5725                 for (i = 0; i < ETH_VMDQ_NUM_UC_HASH_ARRAY; i++) {
5726                         uta_info->uta_shadow[i] = 0;
5727                         IXGBE_WRITE_REG(hw, IXGBE_UTA(i), 0);
5728                 }
5729         }
5730         return 0;
5731
5732 }
5733
5734 uint32_t
5735 ixgbe_convert_vm_rx_mask_to_val(uint16_t rx_mask, uint32_t orig_val)
5736 {
5737         uint32_t new_val = orig_val;
5738
5739         if (rx_mask & ETH_VMDQ_ACCEPT_UNTAG)
5740                 new_val |= IXGBE_VMOLR_AUPE;
5741         if (rx_mask & ETH_VMDQ_ACCEPT_HASH_MC)
5742                 new_val |= IXGBE_VMOLR_ROMPE;
5743         if (rx_mask & ETH_VMDQ_ACCEPT_HASH_UC)
5744                 new_val |= IXGBE_VMOLR_ROPE;
5745         if (rx_mask & ETH_VMDQ_ACCEPT_BROADCAST)
5746                 new_val |= IXGBE_VMOLR_BAM;
5747         if (rx_mask & ETH_VMDQ_ACCEPT_MULTICAST)
5748                 new_val |= IXGBE_VMOLR_MPE;
5749
5750         return new_val;
5751 }
5752
5753 #define IXGBE_MRCTL_VPME  0x01 /* Virtual Pool Mirroring. */
5754 #define IXGBE_MRCTL_UPME  0x02 /* Uplink Port Mirroring. */
5755 #define IXGBE_MRCTL_DPME  0x04 /* Downlink Port Mirroring. */
5756 #define IXGBE_MRCTL_VLME  0x08 /* VLAN Mirroring. */
5757 #define IXGBE_INVALID_MIRROR_TYPE(mirror_type) \
5758         ((mirror_type) & ~(uint8_t)(ETH_MIRROR_VIRTUAL_POOL_UP | \
5759         ETH_MIRROR_UPLINK_PORT | ETH_MIRROR_DOWNLINK_PORT | ETH_MIRROR_VLAN))
5760
5761 static int
5762 ixgbe_mirror_rule_set(struct rte_eth_dev *dev,
5763                       struct rte_eth_mirror_conf *mirror_conf,
5764                       uint8_t rule_id, uint8_t on)
5765 {
5766         uint32_t mr_ctl, vlvf;
5767         uint32_t mp_lsb = 0;
5768         uint32_t mv_msb = 0;
5769         uint32_t mv_lsb = 0;
5770         uint32_t mp_msb = 0;
5771         uint8_t i = 0;
5772         int reg_index = 0;
5773         uint64_t vlan_mask = 0;
5774
5775         const uint8_t pool_mask_offset = 32;
5776         const uint8_t vlan_mask_offset = 32;
5777         const uint8_t dst_pool_offset = 8;
5778         const uint8_t rule_mr_offset  = 4;
5779         const uint8_t mirror_rule_mask = 0x0F;
5780
5781         struct ixgbe_mirror_info *mr_info =
5782                         (IXGBE_DEV_PRIVATE_TO_PFDATA(dev->data->dev_private));
5783         struct ixgbe_hw *hw =
5784                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5785         uint8_t mirror_type = 0;
5786
5787         if (ixgbe_vt_check(hw) < 0)
5788                 return -ENOTSUP;
5789
5790         if (rule_id >= IXGBE_MAX_MIRROR_RULES)
5791                 return -EINVAL;
5792
5793         if (IXGBE_INVALID_MIRROR_TYPE(mirror_conf->rule_type)) {
5794                 PMD_DRV_LOG(ERR, "unsupported mirror type 0x%x.",
5795                             mirror_conf->rule_type);
5796                 return -EINVAL;
5797         }
5798
5799         if (mirror_conf->rule_type & ETH_MIRROR_VLAN) {
5800                 mirror_type |= IXGBE_MRCTL_VLME;
5801                 /* Check if vlan id is valid and find conresponding VLAN ID
5802                  * index in VLVF
5803                  */
5804                 for (i = 0; i < IXGBE_VLVF_ENTRIES; i++) {
5805                         if (mirror_conf->vlan.vlan_mask & (1ULL << i)) {
5806                                 /* search vlan id related pool vlan filter
5807                                  * index
5808                                  */
5809                                 reg_index = ixgbe_find_vlvf_slot(
5810                                                 hw,
5811                                                 mirror_conf->vlan.vlan_id[i],
5812                                                 false);
5813                                 if (reg_index < 0)
5814                                         return -EINVAL;
5815                                 vlvf = IXGBE_READ_REG(hw,
5816                                                       IXGBE_VLVF(reg_index));
5817                                 if ((vlvf & IXGBE_VLVF_VIEN) &&
5818                                     ((vlvf & IXGBE_VLVF_VLANID_MASK) ==
5819                                       mirror_conf->vlan.vlan_id[i]))
5820                                         vlan_mask |= (1ULL << reg_index);
5821                                 else
5822                                         return -EINVAL;
5823                         }
5824                 }
5825
5826                 if (on) {
5827                         mv_lsb = vlan_mask & 0xFFFFFFFF;
5828                         mv_msb = vlan_mask >> vlan_mask_offset;
5829
5830                         mr_info->mr_conf[rule_id].vlan.vlan_mask =
5831                                                 mirror_conf->vlan.vlan_mask;
5832                         for (i = 0; i < ETH_VMDQ_MAX_VLAN_FILTERS; i++) {
5833                                 if (mirror_conf->vlan.vlan_mask & (1ULL << i))
5834                                         mr_info->mr_conf[rule_id].vlan.vlan_id[i] =
5835                                                 mirror_conf->vlan.vlan_id[i];
5836                         }
5837                 } else {
5838                         mv_lsb = 0;
5839                         mv_msb = 0;
5840                         mr_info->mr_conf[rule_id].vlan.vlan_mask = 0;
5841                         for (i = 0; i < ETH_VMDQ_MAX_VLAN_FILTERS; i++)
5842                                 mr_info->mr_conf[rule_id].vlan.vlan_id[i] = 0;
5843                 }
5844         }
5845
5846         /**
5847          * if enable pool mirror, write related pool mask register,if disable
5848          * pool mirror, clear PFMRVM register
5849          */
5850         if (mirror_conf->rule_type & ETH_MIRROR_VIRTUAL_POOL_UP) {
5851                 mirror_type |= IXGBE_MRCTL_VPME;
5852                 if (on) {
5853                         mp_lsb = mirror_conf->pool_mask & 0xFFFFFFFF;
5854                         mp_msb = mirror_conf->pool_mask >> pool_mask_offset;
5855                         mr_info->mr_conf[rule_id].pool_mask =
5856                                         mirror_conf->pool_mask;
5857
5858                 } else {
5859                         mp_lsb = 0;
5860                         mp_msb = 0;
5861                         mr_info->mr_conf[rule_id].pool_mask = 0;
5862                 }
5863         }
5864         if (mirror_conf->rule_type & ETH_MIRROR_UPLINK_PORT)
5865                 mirror_type |= IXGBE_MRCTL_UPME;
5866         if (mirror_conf->rule_type & ETH_MIRROR_DOWNLINK_PORT)
5867                 mirror_type |= IXGBE_MRCTL_DPME;
5868
5869         /* read  mirror control register and recalculate it */
5870         mr_ctl = IXGBE_READ_REG(hw, IXGBE_MRCTL(rule_id));
5871
5872         if (on) {
5873                 mr_ctl |= mirror_type;
5874                 mr_ctl &= mirror_rule_mask;
5875                 mr_ctl |= mirror_conf->dst_pool << dst_pool_offset;
5876         } else {
5877                 mr_ctl &= ~(mirror_conf->rule_type & mirror_rule_mask);
5878         }
5879
5880         mr_info->mr_conf[rule_id].rule_type = mirror_conf->rule_type;
5881         mr_info->mr_conf[rule_id].dst_pool = mirror_conf->dst_pool;
5882
5883         /* write mirrror control  register */
5884         IXGBE_WRITE_REG(hw, IXGBE_MRCTL(rule_id), mr_ctl);
5885
5886         /* write pool mirrror control  register */
5887         if (mirror_conf->rule_type & ETH_MIRROR_VIRTUAL_POOL_UP) {
5888                 IXGBE_WRITE_REG(hw, IXGBE_VMRVM(rule_id), mp_lsb);
5889                 IXGBE_WRITE_REG(hw, IXGBE_VMRVM(rule_id + rule_mr_offset),
5890                                 mp_msb);
5891         }
5892         /* write VLAN mirrror control  register */
5893         if (mirror_conf->rule_type & ETH_MIRROR_VLAN) {
5894                 IXGBE_WRITE_REG(hw, IXGBE_VMRVLAN(rule_id), mv_lsb);
5895                 IXGBE_WRITE_REG(hw, IXGBE_VMRVLAN(rule_id + rule_mr_offset),
5896                                 mv_msb);
5897         }
5898
5899         return 0;
5900 }
5901
5902 static int
5903 ixgbe_mirror_rule_reset(struct rte_eth_dev *dev, uint8_t rule_id)
5904 {
5905         int mr_ctl = 0;
5906         uint32_t lsb_val = 0;
5907         uint32_t msb_val = 0;
5908         const uint8_t rule_mr_offset = 4;
5909
5910         struct ixgbe_hw *hw =
5911                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5912         struct ixgbe_mirror_info *mr_info =
5913                 (IXGBE_DEV_PRIVATE_TO_PFDATA(dev->data->dev_private));
5914
5915         if (ixgbe_vt_check(hw) < 0)
5916                 return -ENOTSUP;
5917
5918         if (rule_id >= IXGBE_MAX_MIRROR_RULES)
5919                 return -EINVAL;
5920
5921         memset(&mr_info->mr_conf[rule_id], 0,
5922                sizeof(struct rte_eth_mirror_conf));
5923
5924         /* clear PFVMCTL register */
5925         IXGBE_WRITE_REG(hw, IXGBE_MRCTL(rule_id), mr_ctl);
5926
5927         /* clear pool mask register */
5928         IXGBE_WRITE_REG(hw, IXGBE_VMRVM(rule_id), lsb_val);
5929         IXGBE_WRITE_REG(hw, IXGBE_VMRVM(rule_id + rule_mr_offset), msb_val);
5930
5931         /* clear vlan mask register */
5932         IXGBE_WRITE_REG(hw, IXGBE_VMRVLAN(rule_id), lsb_val);
5933         IXGBE_WRITE_REG(hw, IXGBE_VMRVLAN(rule_id + rule_mr_offset), msb_val);
5934
5935         return 0;
5936 }
5937
5938 static int
5939 ixgbevf_dev_rx_queue_intr_enable(struct rte_eth_dev *dev, uint16_t queue_id)
5940 {
5941         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5942         struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
5943         struct ixgbe_interrupt *intr =
5944                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
5945         struct ixgbe_hw *hw =
5946                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5947         uint32_t vec = IXGBE_MISC_VEC_ID;
5948
5949         if (rte_intr_allow_others(intr_handle))
5950                 vec = IXGBE_RX_VEC_START;
5951         intr->mask |= (1 << vec);
5952         RTE_SET_USED(queue_id);
5953         IXGBE_WRITE_REG(hw, IXGBE_VTEIMS, intr->mask);
5954
5955         rte_intr_ack(intr_handle);
5956
5957         return 0;
5958 }
5959
5960 static int
5961 ixgbevf_dev_rx_queue_intr_disable(struct rte_eth_dev *dev, uint16_t queue_id)
5962 {
5963         struct ixgbe_interrupt *intr =
5964                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
5965         struct ixgbe_hw *hw =
5966                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5967         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5968         struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
5969         uint32_t vec = IXGBE_MISC_VEC_ID;
5970
5971         if (rte_intr_allow_others(intr_handle))
5972                 vec = IXGBE_RX_VEC_START;
5973         intr->mask &= ~(1 << vec);
5974         RTE_SET_USED(queue_id);
5975         IXGBE_WRITE_REG(hw, IXGBE_VTEIMS, intr->mask);
5976
5977         return 0;
5978 }
5979
5980 static int
5981 ixgbe_dev_rx_queue_intr_enable(struct rte_eth_dev *dev, uint16_t queue_id)
5982 {
5983         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5984         struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
5985         uint32_t mask;
5986         struct ixgbe_hw *hw =
5987                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5988         struct ixgbe_interrupt *intr =
5989                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
5990
5991         if (queue_id < 16) {
5992                 ixgbe_disable_intr(hw);
5993                 intr->mask |= (1 << queue_id);
5994                 ixgbe_enable_intr(dev);
5995         } else if (queue_id < 32) {
5996                 mask = IXGBE_READ_REG(hw, IXGBE_EIMS_EX(0));
5997                 mask &= (1 << queue_id);
5998                 IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(0), mask);
5999         } else if (queue_id < 64) {
6000                 mask = IXGBE_READ_REG(hw, IXGBE_EIMS_EX(1));
6001                 mask &= (1 << (queue_id - 32));
6002                 IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(1), mask);
6003         }
6004         rte_intr_ack(intr_handle);
6005
6006         return 0;
6007 }
6008
6009 static int
6010 ixgbe_dev_rx_queue_intr_disable(struct rte_eth_dev *dev, uint16_t queue_id)
6011 {
6012         uint32_t mask;
6013         struct ixgbe_hw *hw =
6014                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6015         struct ixgbe_interrupt *intr =
6016                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
6017
6018         if (queue_id < 16) {
6019                 ixgbe_disable_intr(hw);
6020                 intr->mask &= ~(1 << queue_id);
6021                 ixgbe_enable_intr(dev);
6022         } else if (queue_id < 32) {
6023                 mask = IXGBE_READ_REG(hw, IXGBE_EIMS_EX(0));
6024                 mask &= ~(1 << queue_id);
6025                 IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(0), mask);
6026         } else if (queue_id < 64) {
6027                 mask = IXGBE_READ_REG(hw, IXGBE_EIMS_EX(1));
6028                 mask &= ~(1 << (queue_id - 32));
6029                 IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(1), mask);
6030         }
6031
6032         return 0;
6033 }
6034
6035 static void
6036 ixgbevf_set_ivar_map(struct ixgbe_hw *hw, int8_t direction,
6037                      uint8_t queue, uint8_t msix_vector)
6038 {
6039         uint32_t tmp, idx;
6040
6041         if (direction == -1) {
6042                 /* other causes */
6043                 msix_vector |= IXGBE_IVAR_ALLOC_VAL;
6044                 tmp = IXGBE_READ_REG(hw, IXGBE_VTIVAR_MISC);
6045                 tmp &= ~0xFF;
6046                 tmp |= msix_vector;
6047                 IXGBE_WRITE_REG(hw, IXGBE_VTIVAR_MISC, tmp);
6048         } else {
6049                 /* rx or tx cause */
6050                 msix_vector |= IXGBE_IVAR_ALLOC_VAL;
6051                 idx = ((16 * (queue & 1)) + (8 * direction));
6052                 tmp = IXGBE_READ_REG(hw, IXGBE_VTIVAR(queue >> 1));
6053                 tmp &= ~(0xFF << idx);
6054                 tmp |= (msix_vector << idx);
6055                 IXGBE_WRITE_REG(hw, IXGBE_VTIVAR(queue >> 1), tmp);
6056         }
6057 }
6058
6059 /**
6060  * set the IVAR registers, mapping interrupt causes to vectors
6061  * @param hw
6062  *  pointer to ixgbe_hw struct
6063  * @direction
6064  *  0 for Rx, 1 for Tx, -1 for other causes
6065  * @queue
6066  *  queue to map the corresponding interrupt to
6067  * @msix_vector
6068  *  the vector to map to the corresponding queue
6069  */
6070 static void
6071 ixgbe_set_ivar_map(struct ixgbe_hw *hw, int8_t direction,
6072                    uint8_t queue, uint8_t msix_vector)
6073 {
6074         uint32_t tmp, idx;
6075
6076         msix_vector |= IXGBE_IVAR_ALLOC_VAL;
6077         if (hw->mac.type == ixgbe_mac_82598EB) {
6078                 if (direction == -1)
6079                         direction = 0;
6080                 idx = (((direction * 64) + queue) >> 2) & 0x1F;
6081                 tmp = IXGBE_READ_REG(hw, IXGBE_IVAR(idx));
6082                 tmp &= ~(0xFF << (8 * (queue & 0x3)));
6083                 tmp |= (msix_vector << (8 * (queue & 0x3)));
6084                 IXGBE_WRITE_REG(hw, IXGBE_IVAR(idx), tmp);
6085         } else if ((hw->mac.type == ixgbe_mac_82599EB) ||
6086                         (hw->mac.type == ixgbe_mac_X540) ||
6087                         (hw->mac.type == ixgbe_mac_X550) ||
6088                         (hw->mac.type == ixgbe_mac_X550EM_x)) {
6089                 if (direction == -1) {
6090                         /* other causes */
6091                         idx = ((queue & 1) * 8);
6092                         tmp = IXGBE_READ_REG(hw, IXGBE_IVAR_MISC);
6093                         tmp &= ~(0xFF << idx);
6094                         tmp |= (msix_vector << idx);
6095                         IXGBE_WRITE_REG(hw, IXGBE_IVAR_MISC, tmp);
6096                 } else {
6097                         /* rx or tx causes */
6098                         idx = ((16 * (queue & 1)) + (8 * direction));
6099                         tmp = IXGBE_READ_REG(hw, IXGBE_IVAR(queue >> 1));
6100                         tmp &= ~(0xFF << idx);
6101                         tmp |= (msix_vector << idx);
6102                         IXGBE_WRITE_REG(hw, IXGBE_IVAR(queue >> 1), tmp);
6103                 }
6104         }
6105 }
6106
6107 static void
6108 ixgbevf_configure_msix(struct rte_eth_dev *dev)
6109 {
6110         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
6111         struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
6112         struct ixgbe_hw *hw =
6113                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6114         uint32_t q_idx;
6115         uint32_t vector_idx = IXGBE_MISC_VEC_ID;
6116         uint32_t base = IXGBE_MISC_VEC_ID;
6117
6118         /* Configure VF other cause ivar */
6119         ixgbevf_set_ivar_map(hw, -1, 1, vector_idx);
6120
6121         /* won't configure msix register if no mapping is done
6122          * between intr vector and event fd.
6123          */
6124         if (!rte_intr_dp_is_en(intr_handle))
6125                 return;
6126
6127         if (rte_intr_allow_others(intr_handle)) {
6128                 base = IXGBE_RX_VEC_START;
6129                 vector_idx = IXGBE_RX_VEC_START;
6130         }
6131
6132         /* Configure all RX queues of VF */
6133         for (q_idx = 0; q_idx < dev->data->nb_rx_queues; q_idx++) {
6134                 /* Force all queue use vector 0,
6135                  * as IXGBE_VF_MAXMSIVECOTR = 1
6136                  */
6137                 ixgbevf_set_ivar_map(hw, 0, q_idx, vector_idx);
6138                 intr_handle->intr_vec[q_idx] = vector_idx;
6139                 if (vector_idx < base + intr_handle->nb_efd - 1)
6140                         vector_idx++;
6141         }
6142
6143         /* As RX queue setting above show, all queues use the vector 0.
6144          * Set only the ITR value of IXGBE_MISC_VEC_ID.
6145          */
6146         IXGBE_WRITE_REG(hw, IXGBE_VTEITR(IXGBE_MISC_VEC_ID),
6147                         IXGBE_EITR_INTERVAL_US(IXGBE_QUEUE_ITR_INTERVAL_DEFAULT)
6148                         | IXGBE_EITR_CNT_WDIS);
6149 }
6150
6151 /**
6152  * Sets up the hardware to properly generate MSI-X interrupts
6153  * @hw
6154  *  board private structure
6155  */
6156 static void
6157 ixgbe_configure_msix(struct rte_eth_dev *dev)
6158 {
6159         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
6160         struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
6161         struct ixgbe_hw *hw =
6162                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6163         uint32_t queue_id, base = IXGBE_MISC_VEC_ID;
6164         uint32_t vec = IXGBE_MISC_VEC_ID;
6165         uint32_t mask;
6166         uint32_t gpie;
6167
6168         /* won't configure msix register if no mapping is done
6169          * between intr vector and event fd
6170          * but if misx has been enabled already, need to configure
6171          * auto clean, auto mask and throttling.
6172          */
6173         gpie = IXGBE_READ_REG(hw, IXGBE_GPIE);
6174         if (!rte_intr_dp_is_en(intr_handle) &&
6175             !(gpie & (IXGBE_GPIE_MSIX_MODE | IXGBE_GPIE_PBA_SUPPORT)))
6176                 return;
6177
6178         if (rte_intr_allow_others(intr_handle))
6179                 vec = base = IXGBE_RX_VEC_START;
6180
6181         /* setup GPIE for MSI-x mode */
6182         gpie = IXGBE_READ_REG(hw, IXGBE_GPIE);
6183         gpie |= IXGBE_GPIE_MSIX_MODE | IXGBE_GPIE_PBA_SUPPORT |
6184                 IXGBE_GPIE_OCD | IXGBE_GPIE_EIAME;
6185         /* auto clearing and auto setting corresponding bits in EIMS
6186          * when MSI-X interrupt is triggered
6187          */
6188         if (hw->mac.type == ixgbe_mac_82598EB) {
6189                 IXGBE_WRITE_REG(hw, IXGBE_EIAM, IXGBE_EICS_RTX_QUEUE);
6190         } else {
6191                 IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(0), 0xFFFFFFFF);
6192                 IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(1), 0xFFFFFFFF);
6193         }
6194         IXGBE_WRITE_REG(hw, IXGBE_GPIE, gpie);
6195
6196         /* Populate the IVAR table and set the ITR values to the
6197          * corresponding register.
6198          */
6199         if (rte_intr_dp_is_en(intr_handle)) {
6200                 for (queue_id = 0; queue_id < dev->data->nb_rx_queues;
6201                         queue_id++) {
6202                         /* by default, 1:1 mapping */
6203                         ixgbe_set_ivar_map(hw, 0, queue_id, vec);
6204                         intr_handle->intr_vec[queue_id] = vec;
6205                         if (vec < base + intr_handle->nb_efd - 1)
6206                                 vec++;
6207                 }
6208
6209                 switch (hw->mac.type) {
6210                 case ixgbe_mac_82598EB:
6211                         ixgbe_set_ivar_map(hw, -1,
6212                                            IXGBE_IVAR_OTHER_CAUSES_INDEX,
6213                                            IXGBE_MISC_VEC_ID);
6214                         break;
6215                 case ixgbe_mac_82599EB:
6216                 case ixgbe_mac_X540:
6217                 case ixgbe_mac_X550:
6218                 case ixgbe_mac_X550EM_x:
6219                         ixgbe_set_ivar_map(hw, -1, 1, IXGBE_MISC_VEC_ID);
6220                         break;
6221                 default:
6222                         break;
6223                 }
6224         }
6225         IXGBE_WRITE_REG(hw, IXGBE_EITR(IXGBE_MISC_VEC_ID),
6226                         IXGBE_EITR_INTERVAL_US(IXGBE_QUEUE_ITR_INTERVAL_DEFAULT)
6227                         | IXGBE_EITR_CNT_WDIS);
6228
6229         /* set up to autoclear timer, and the vectors */
6230         mask = IXGBE_EIMS_ENABLE_MASK;
6231         mask &= ~(IXGBE_EIMS_OTHER |
6232                   IXGBE_EIMS_MAILBOX |
6233                   IXGBE_EIMS_LSC);
6234
6235         IXGBE_WRITE_REG(hw, IXGBE_EIAC, mask);
6236 }
6237
6238 int
6239 ixgbe_set_queue_rate_limit(struct rte_eth_dev *dev,
6240                            uint16_t queue_idx, uint16_t tx_rate)
6241 {
6242         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6243         struct rte_eth_rxmode *rxmode;
6244         uint32_t rf_dec, rf_int;
6245         uint32_t bcnrc_val;
6246         uint16_t link_speed = dev->data->dev_link.link_speed;
6247
6248         if (queue_idx >= hw->mac.max_tx_queues)
6249                 return -EINVAL;
6250
6251         if (tx_rate != 0) {
6252                 /* Calculate the rate factor values to set */
6253                 rf_int = (uint32_t)link_speed / (uint32_t)tx_rate;
6254                 rf_dec = (uint32_t)link_speed % (uint32_t)tx_rate;
6255                 rf_dec = (rf_dec << IXGBE_RTTBCNRC_RF_INT_SHIFT) / tx_rate;
6256
6257                 bcnrc_val = IXGBE_RTTBCNRC_RS_ENA;
6258                 bcnrc_val |= ((rf_int << IXGBE_RTTBCNRC_RF_INT_SHIFT) &
6259                                 IXGBE_RTTBCNRC_RF_INT_MASK_M);
6260                 bcnrc_val |= (rf_dec & IXGBE_RTTBCNRC_RF_DEC_MASK);
6261         } else {
6262                 bcnrc_val = 0;
6263         }
6264
6265         rxmode = &dev->data->dev_conf.rxmode;
6266         /*
6267          * Set global transmit compensation time to the MMW_SIZE in RTTBCNRM
6268          * register. MMW_SIZE=0x014 if 9728-byte jumbo is supported, otherwise
6269          * set as 0x4.
6270          */
6271         if ((rxmode->offloads & DEV_RX_OFFLOAD_JUMBO_FRAME) &&
6272             (rxmode->max_rx_pkt_len >= IXGBE_MAX_JUMBO_FRAME_SIZE))
6273                 IXGBE_WRITE_REG(hw, IXGBE_RTTBCNRM,
6274                         IXGBE_MMW_SIZE_JUMBO_FRAME);
6275         else
6276                 IXGBE_WRITE_REG(hw, IXGBE_RTTBCNRM,
6277                         IXGBE_MMW_SIZE_DEFAULT);
6278
6279         /* Set RTTBCNRC of queue X */
6280         IXGBE_WRITE_REG(hw, IXGBE_RTTDQSEL, queue_idx);
6281         IXGBE_WRITE_REG(hw, IXGBE_RTTBCNRC, bcnrc_val);
6282         IXGBE_WRITE_FLUSH(hw);
6283
6284         return 0;
6285 }
6286
6287 static int
6288 ixgbevf_add_mac_addr(struct rte_eth_dev *dev, struct rte_ether_addr *mac_addr,
6289                      __rte_unused uint32_t index,
6290                      __rte_unused uint32_t pool)
6291 {
6292         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6293         int diag;
6294
6295         /*
6296          * On a 82599 VF, adding again the same MAC addr is not an idempotent
6297          * operation. Trap this case to avoid exhausting the [very limited]
6298          * set of PF resources used to store VF MAC addresses.
6299          */
6300         if (memcmp(hw->mac.perm_addr, mac_addr,
6301                         sizeof(struct rte_ether_addr)) == 0)
6302                 return -1;
6303         diag = ixgbevf_set_uc_addr_vf(hw, 2, mac_addr->addr_bytes);
6304         if (diag != 0)
6305                 PMD_DRV_LOG(ERR, "Unable to add MAC address "
6306                             "%02x:%02x:%02x:%02x:%02x:%02x - diag=%d",
6307                             mac_addr->addr_bytes[0],
6308                             mac_addr->addr_bytes[1],
6309                             mac_addr->addr_bytes[2],
6310                             mac_addr->addr_bytes[3],
6311                             mac_addr->addr_bytes[4],
6312                             mac_addr->addr_bytes[5],
6313                             diag);
6314         return diag;
6315 }
6316
6317 static void
6318 ixgbevf_remove_mac_addr(struct rte_eth_dev *dev, uint32_t index)
6319 {
6320         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6321         struct rte_ether_addr *perm_addr =
6322                 (struct rte_ether_addr *)hw->mac.perm_addr;
6323         struct rte_ether_addr *mac_addr;
6324         uint32_t i;
6325         int diag;
6326
6327         /*
6328          * The IXGBE_VF_SET_MACVLAN command of the ixgbe-pf driver does
6329          * not support the deletion of a given MAC address.
6330          * Instead, it imposes to delete all MAC addresses, then to add again
6331          * all MAC addresses with the exception of the one to be deleted.
6332          */
6333         (void) ixgbevf_set_uc_addr_vf(hw, 0, NULL);
6334
6335         /*
6336          * Add again all MAC addresses, with the exception of the deleted one
6337          * and of the permanent MAC address.
6338          */
6339         for (i = 0, mac_addr = dev->data->mac_addrs;
6340              i < hw->mac.num_rar_entries; i++, mac_addr++) {
6341                 /* Skip the deleted MAC address */
6342                 if (i == index)
6343                         continue;
6344                 /* Skip NULL MAC addresses */
6345                 if (rte_is_zero_ether_addr(mac_addr))
6346                         continue;
6347                 /* Skip the permanent MAC address */
6348                 if (memcmp(perm_addr, mac_addr,
6349                                 sizeof(struct rte_ether_addr)) == 0)
6350                         continue;
6351                 diag = ixgbevf_set_uc_addr_vf(hw, 2, mac_addr->addr_bytes);
6352                 if (diag != 0)
6353                         PMD_DRV_LOG(ERR,
6354                                     "Adding again MAC address "
6355                                     "%02x:%02x:%02x:%02x:%02x:%02x failed "
6356                                     "diag=%d",
6357                                     mac_addr->addr_bytes[0],
6358                                     mac_addr->addr_bytes[1],
6359                                     mac_addr->addr_bytes[2],
6360                                     mac_addr->addr_bytes[3],
6361                                     mac_addr->addr_bytes[4],
6362                                     mac_addr->addr_bytes[5],
6363                                     diag);
6364         }
6365 }
6366
6367 static int
6368 ixgbevf_set_default_mac_addr(struct rte_eth_dev *dev,
6369                         struct rte_ether_addr *addr)
6370 {
6371         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6372
6373         hw->mac.ops.set_rar(hw, 0, (void *)addr, 0, 0);
6374
6375         return 0;
6376 }
6377
6378 int
6379 ixgbe_syn_filter_set(struct rte_eth_dev *dev,
6380                         struct rte_eth_syn_filter *filter,
6381                         bool add)
6382 {
6383         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6384         struct ixgbe_filter_info *filter_info =
6385                 IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
6386         uint32_t syn_info;
6387         uint32_t synqf;
6388
6389         if (filter->queue >= IXGBE_MAX_RX_QUEUE_NUM)
6390                 return -EINVAL;
6391
6392         syn_info = filter_info->syn_info;
6393
6394         if (add) {
6395                 if (syn_info & IXGBE_SYN_FILTER_ENABLE)
6396                         return -EINVAL;
6397                 synqf = (uint32_t)(((filter->queue << IXGBE_SYN_FILTER_QUEUE_SHIFT) &
6398                         IXGBE_SYN_FILTER_QUEUE) | IXGBE_SYN_FILTER_ENABLE);
6399
6400                 if (filter->hig_pri)
6401                         synqf |= IXGBE_SYN_FILTER_SYNQFP;
6402                 else
6403                         synqf &= ~IXGBE_SYN_FILTER_SYNQFP;
6404         } else {
6405                 synqf = IXGBE_READ_REG(hw, IXGBE_SYNQF);
6406                 if (!(syn_info & IXGBE_SYN_FILTER_ENABLE))
6407                         return -ENOENT;
6408                 synqf &= ~(IXGBE_SYN_FILTER_QUEUE | IXGBE_SYN_FILTER_ENABLE);
6409         }
6410
6411         filter_info->syn_info = synqf;
6412         IXGBE_WRITE_REG(hw, IXGBE_SYNQF, synqf);
6413         IXGBE_WRITE_FLUSH(hw);
6414         return 0;
6415 }
6416
6417 static int
6418 ixgbe_syn_filter_get(struct rte_eth_dev *dev,
6419                         struct rte_eth_syn_filter *filter)
6420 {
6421         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6422         uint32_t synqf = IXGBE_READ_REG(hw, IXGBE_SYNQF);
6423
6424         if (synqf & IXGBE_SYN_FILTER_ENABLE) {
6425                 filter->hig_pri = (synqf & IXGBE_SYN_FILTER_SYNQFP) ? 1 : 0;
6426                 filter->queue = (uint16_t)((synqf & IXGBE_SYN_FILTER_QUEUE) >> 1);
6427                 return 0;
6428         }
6429         return -ENOENT;
6430 }
6431
6432 static int
6433 ixgbe_syn_filter_handle(struct rte_eth_dev *dev,
6434                         enum rte_filter_op filter_op,
6435                         void *arg)
6436 {
6437         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6438         int ret;
6439
6440         MAC_TYPE_FILTER_SUP(hw->mac.type);
6441
6442         if (filter_op == RTE_ETH_FILTER_NOP)
6443                 return 0;
6444
6445         if (arg == NULL) {
6446                 PMD_DRV_LOG(ERR, "arg shouldn't be NULL for operation %u",
6447                             filter_op);
6448                 return -EINVAL;
6449         }
6450
6451         switch (filter_op) {
6452         case RTE_ETH_FILTER_ADD:
6453                 ret = ixgbe_syn_filter_set(dev,
6454                                 (struct rte_eth_syn_filter *)arg,
6455                                 TRUE);
6456                 break;
6457         case RTE_ETH_FILTER_DELETE:
6458                 ret = ixgbe_syn_filter_set(dev,
6459                                 (struct rte_eth_syn_filter *)arg,
6460                                 FALSE);
6461                 break;
6462         case RTE_ETH_FILTER_GET:
6463                 ret = ixgbe_syn_filter_get(dev,
6464                                 (struct rte_eth_syn_filter *)arg);
6465                 break;
6466         default:
6467                 PMD_DRV_LOG(ERR, "unsupported operation %u", filter_op);
6468                 ret = -EINVAL;
6469                 break;
6470         }
6471
6472         return ret;
6473 }
6474
6475
6476 static inline enum ixgbe_5tuple_protocol
6477 convert_protocol_type(uint8_t protocol_value)
6478 {
6479         if (protocol_value == IPPROTO_TCP)
6480                 return IXGBE_FILTER_PROTOCOL_TCP;
6481         else if (protocol_value == IPPROTO_UDP)
6482                 return IXGBE_FILTER_PROTOCOL_UDP;
6483         else if (protocol_value == IPPROTO_SCTP)
6484                 return IXGBE_FILTER_PROTOCOL_SCTP;
6485         else
6486                 return IXGBE_FILTER_PROTOCOL_NONE;
6487 }
6488
6489 /* inject a 5-tuple filter to HW */
6490 static inline void
6491 ixgbe_inject_5tuple_filter(struct rte_eth_dev *dev,
6492                            struct ixgbe_5tuple_filter *filter)
6493 {
6494         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6495         int i;
6496         uint32_t ftqf, sdpqf;
6497         uint32_t l34timir = 0;
6498         uint8_t mask = 0xff;
6499
6500         i = filter->index;
6501
6502         sdpqf = (uint32_t)(filter->filter_info.dst_port <<
6503                                 IXGBE_SDPQF_DSTPORT_SHIFT);
6504         sdpqf = sdpqf | (filter->filter_info.src_port & IXGBE_SDPQF_SRCPORT);
6505
6506         ftqf = (uint32_t)(filter->filter_info.proto &
6507                 IXGBE_FTQF_PROTOCOL_MASK);
6508         ftqf |= (uint32_t)((filter->filter_info.priority &
6509                 IXGBE_FTQF_PRIORITY_MASK) << IXGBE_FTQF_PRIORITY_SHIFT);
6510         if (filter->filter_info.src_ip_mask == 0) /* 0 means compare. */
6511                 mask &= IXGBE_FTQF_SOURCE_ADDR_MASK;
6512         if (filter->filter_info.dst_ip_mask == 0)
6513                 mask &= IXGBE_FTQF_DEST_ADDR_MASK;
6514         if (filter->filter_info.src_port_mask == 0)
6515                 mask &= IXGBE_FTQF_SOURCE_PORT_MASK;
6516         if (filter->filter_info.dst_port_mask == 0)
6517                 mask &= IXGBE_FTQF_DEST_PORT_MASK;
6518         if (filter->filter_info.proto_mask == 0)
6519                 mask &= IXGBE_FTQF_PROTOCOL_COMP_MASK;
6520         ftqf |= mask << IXGBE_FTQF_5TUPLE_MASK_SHIFT;
6521         ftqf |= IXGBE_FTQF_POOL_MASK_EN;
6522         ftqf |= IXGBE_FTQF_QUEUE_ENABLE;
6523
6524         IXGBE_WRITE_REG(hw, IXGBE_DAQF(i), filter->filter_info.dst_ip);
6525         IXGBE_WRITE_REG(hw, IXGBE_SAQF(i), filter->filter_info.src_ip);
6526         IXGBE_WRITE_REG(hw, IXGBE_SDPQF(i), sdpqf);
6527         IXGBE_WRITE_REG(hw, IXGBE_FTQF(i), ftqf);
6528
6529         l34timir |= IXGBE_L34T_IMIR_RESERVE;
6530         l34timir |= (uint32_t)(filter->queue <<
6531                                 IXGBE_L34T_IMIR_QUEUE_SHIFT);
6532         IXGBE_WRITE_REG(hw, IXGBE_L34T_IMIR(i), l34timir);
6533 }
6534
6535 /*
6536  * add a 5tuple filter
6537  *
6538  * @param
6539  * dev: Pointer to struct rte_eth_dev.
6540  * index: the index the filter allocates.
6541  * filter: ponter to the filter that will be added.
6542  * rx_queue: the queue id the filter assigned to.
6543  *
6544  * @return
6545  *    - On success, zero.
6546  *    - On failure, a negative value.
6547  */
6548 static int
6549 ixgbe_add_5tuple_filter(struct rte_eth_dev *dev,
6550                         struct ixgbe_5tuple_filter *filter)
6551 {
6552         struct ixgbe_filter_info *filter_info =
6553                 IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
6554         int i, idx, shift;
6555
6556         /*
6557          * look for an unused 5tuple filter index,
6558          * and insert the filter to list.
6559          */
6560         for (i = 0; i < IXGBE_MAX_FTQF_FILTERS; i++) {
6561                 idx = i / (sizeof(uint32_t) * NBBY);
6562                 shift = i % (sizeof(uint32_t) * NBBY);
6563                 if (!(filter_info->fivetuple_mask[idx] & (1 << shift))) {
6564                         filter_info->fivetuple_mask[idx] |= 1 << shift;
6565                         filter->index = i;
6566                         TAILQ_INSERT_TAIL(&filter_info->fivetuple_list,
6567                                           filter,
6568                                           entries);
6569                         break;
6570                 }
6571         }
6572         if (i >= IXGBE_MAX_FTQF_FILTERS) {
6573                 PMD_DRV_LOG(ERR, "5tuple filters are full.");
6574                 return -ENOSYS;
6575         }
6576
6577         ixgbe_inject_5tuple_filter(dev, filter);
6578
6579         return 0;
6580 }
6581
6582 /*
6583  * remove a 5tuple filter
6584  *
6585  * @param
6586  * dev: Pointer to struct rte_eth_dev.
6587  * filter: the pointer of the filter will be removed.
6588  */
6589 static void
6590 ixgbe_remove_5tuple_filter(struct rte_eth_dev *dev,
6591                         struct ixgbe_5tuple_filter *filter)
6592 {
6593         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6594         struct ixgbe_filter_info *filter_info =
6595                 IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
6596         uint16_t index = filter->index;
6597
6598         filter_info->fivetuple_mask[index / (sizeof(uint32_t) * NBBY)] &=
6599                                 ~(1 << (index % (sizeof(uint32_t) * NBBY)));
6600         TAILQ_REMOVE(&filter_info->fivetuple_list, filter, entries);
6601         rte_free(filter);
6602
6603         IXGBE_WRITE_REG(hw, IXGBE_DAQF(index), 0);
6604         IXGBE_WRITE_REG(hw, IXGBE_SAQF(index), 0);
6605         IXGBE_WRITE_REG(hw, IXGBE_SDPQF(index), 0);
6606         IXGBE_WRITE_REG(hw, IXGBE_FTQF(index), 0);
6607         IXGBE_WRITE_REG(hw, IXGBE_L34T_IMIR(index), 0);
6608 }
6609
6610 static int
6611 ixgbevf_dev_set_mtu(struct rte_eth_dev *dev, uint16_t mtu)
6612 {
6613         struct ixgbe_hw *hw;
6614         uint32_t max_frame = mtu + IXGBE_ETH_OVERHEAD;
6615         struct rte_eth_dev_data *dev_data = dev->data;
6616
6617         hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6618
6619         if (mtu < RTE_ETHER_MIN_MTU ||
6620                         max_frame > RTE_ETHER_MAX_JUMBO_FRAME_LEN)
6621                 return -EINVAL;
6622
6623         /* If device is started, refuse mtu that requires the support of
6624          * scattered packets when this feature has not been enabled before.
6625          */
6626         if (dev_data->dev_started && !dev_data->scattered_rx &&
6627             (max_frame + 2 * IXGBE_VLAN_TAG_SIZE >
6628              dev->data->min_rx_buf_size - RTE_PKTMBUF_HEADROOM)) {
6629                 PMD_INIT_LOG(ERR, "Stop port first.");
6630                 return -EINVAL;
6631         }
6632
6633         /*
6634          * When supported by the underlying PF driver, use the IXGBE_VF_SET_MTU
6635          * request of the version 2.0 of the mailbox API.
6636          * For now, use the IXGBE_VF_SET_LPE request of the version 1.0
6637          * of the mailbox API.
6638          * This call to IXGBE_SET_LPE action won't work with ixgbe pf drivers
6639          * prior to 3.11.33 which contains the following change:
6640          * "ixgbe: Enable jumbo frames support w/ SR-IOV"
6641          */
6642         ixgbevf_rlpml_set_vf(hw, max_frame);
6643
6644         /* update max frame size */
6645         dev->data->dev_conf.rxmode.max_rx_pkt_len = max_frame;
6646         return 0;
6647 }
6648
6649 static inline struct ixgbe_5tuple_filter *
6650 ixgbe_5tuple_filter_lookup(struct ixgbe_5tuple_filter_list *filter_list,
6651                         struct ixgbe_5tuple_filter_info *key)
6652 {
6653         struct ixgbe_5tuple_filter *it;
6654
6655         TAILQ_FOREACH(it, filter_list, entries) {
6656                 if (memcmp(key, &it->filter_info,
6657                         sizeof(struct ixgbe_5tuple_filter_info)) == 0) {
6658                         return it;
6659                 }
6660         }
6661         return NULL;
6662 }
6663
6664 /* translate elements in struct rte_eth_ntuple_filter to struct ixgbe_5tuple_filter_info*/
6665 static inline int
6666 ntuple_filter_to_5tuple(struct rte_eth_ntuple_filter *filter,
6667                         struct ixgbe_5tuple_filter_info *filter_info)
6668 {
6669         if (filter->queue >= IXGBE_MAX_RX_QUEUE_NUM ||
6670                 filter->priority > IXGBE_5TUPLE_MAX_PRI ||
6671                 filter->priority < IXGBE_5TUPLE_MIN_PRI)
6672                 return -EINVAL;
6673
6674         switch (filter->dst_ip_mask) {
6675         case UINT32_MAX:
6676                 filter_info->dst_ip_mask = 0;
6677                 filter_info->dst_ip = filter->dst_ip;
6678                 break;
6679         case 0:
6680                 filter_info->dst_ip_mask = 1;
6681                 break;
6682         default:
6683                 PMD_DRV_LOG(ERR, "invalid dst_ip mask.");
6684                 return -EINVAL;
6685         }
6686
6687         switch (filter->src_ip_mask) {
6688         case UINT32_MAX:
6689                 filter_info->src_ip_mask = 0;
6690                 filter_info->src_ip = filter->src_ip;
6691                 break;
6692         case 0:
6693                 filter_info->src_ip_mask = 1;
6694                 break;
6695         default:
6696                 PMD_DRV_LOG(ERR, "invalid src_ip mask.");
6697                 return -EINVAL;
6698         }
6699
6700         switch (filter->dst_port_mask) {
6701         case UINT16_MAX:
6702                 filter_info->dst_port_mask = 0;
6703                 filter_info->dst_port = filter->dst_port;
6704                 break;
6705         case 0:
6706                 filter_info->dst_port_mask = 1;
6707                 break;
6708         default:
6709                 PMD_DRV_LOG(ERR, "invalid dst_port mask.");
6710                 return -EINVAL;
6711         }
6712
6713         switch (filter->src_port_mask) {
6714         case UINT16_MAX:
6715                 filter_info->src_port_mask = 0;
6716                 filter_info->src_port = filter->src_port;
6717                 break;
6718         case 0:
6719                 filter_info->src_port_mask = 1;
6720                 break;
6721         default:
6722                 PMD_DRV_LOG(ERR, "invalid src_port mask.");
6723                 return -EINVAL;
6724         }
6725
6726         switch (filter->proto_mask) {
6727         case UINT8_MAX:
6728                 filter_info->proto_mask = 0;
6729                 filter_info->proto =
6730                         convert_protocol_type(filter->proto);
6731                 break;
6732         case 0:
6733                 filter_info->proto_mask = 1;
6734                 break;
6735         default:
6736                 PMD_DRV_LOG(ERR, "invalid protocol mask.");
6737                 return -EINVAL;
6738         }
6739
6740         filter_info->priority = (uint8_t)filter->priority;
6741         return 0;
6742 }
6743
6744 /*
6745  * add or delete a ntuple filter
6746  *
6747  * @param
6748  * dev: Pointer to struct rte_eth_dev.
6749  * ntuple_filter: Pointer to struct rte_eth_ntuple_filter
6750  * add: if true, add filter, if false, remove filter
6751  *
6752  * @return
6753  *    - On success, zero.
6754  *    - On failure, a negative value.
6755  */
6756 int
6757 ixgbe_add_del_ntuple_filter(struct rte_eth_dev *dev,
6758                         struct rte_eth_ntuple_filter *ntuple_filter,
6759                         bool add)
6760 {
6761         struct ixgbe_filter_info *filter_info =
6762                 IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
6763         struct ixgbe_5tuple_filter_info filter_5tuple;
6764         struct ixgbe_5tuple_filter *filter;
6765         int ret;
6766
6767         if (ntuple_filter->flags != RTE_5TUPLE_FLAGS) {
6768                 PMD_DRV_LOG(ERR, "only 5tuple is supported.");
6769                 return -EINVAL;
6770         }
6771
6772         memset(&filter_5tuple, 0, sizeof(struct ixgbe_5tuple_filter_info));
6773         ret = ntuple_filter_to_5tuple(ntuple_filter, &filter_5tuple);
6774         if (ret < 0)
6775                 return ret;
6776
6777         filter = ixgbe_5tuple_filter_lookup(&filter_info->fivetuple_list,
6778                                          &filter_5tuple);
6779         if (filter != NULL && add) {
6780                 PMD_DRV_LOG(ERR, "filter exists.");
6781                 return -EEXIST;
6782         }
6783         if (filter == NULL && !add) {
6784                 PMD_DRV_LOG(ERR, "filter doesn't exist.");
6785                 return -ENOENT;
6786         }
6787
6788         if (add) {
6789                 filter = rte_zmalloc("ixgbe_5tuple_filter",
6790                                 sizeof(struct ixgbe_5tuple_filter), 0);
6791                 if (filter == NULL)
6792                         return -ENOMEM;
6793                 rte_memcpy(&filter->filter_info,
6794                                  &filter_5tuple,
6795                                  sizeof(struct ixgbe_5tuple_filter_info));
6796                 filter->queue = ntuple_filter->queue;
6797                 ret = ixgbe_add_5tuple_filter(dev, filter);
6798                 if (ret < 0) {
6799                         rte_free(filter);
6800                         return ret;
6801                 }
6802         } else
6803                 ixgbe_remove_5tuple_filter(dev, filter);
6804
6805         return 0;
6806 }
6807
6808 /*
6809  * get a ntuple filter
6810  *
6811  * @param
6812  * dev: Pointer to struct rte_eth_dev.
6813  * ntuple_filter: Pointer to struct rte_eth_ntuple_filter
6814  *
6815  * @return
6816  *    - On success, zero.
6817  *    - On failure, a negative value.
6818  */
6819 static int
6820 ixgbe_get_ntuple_filter(struct rte_eth_dev *dev,
6821                         struct rte_eth_ntuple_filter *ntuple_filter)
6822 {
6823         struct ixgbe_filter_info *filter_info =
6824                 IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
6825         struct ixgbe_5tuple_filter_info filter_5tuple;
6826         struct ixgbe_5tuple_filter *filter;
6827         int ret;
6828
6829         if (ntuple_filter->flags != RTE_5TUPLE_FLAGS) {
6830                 PMD_DRV_LOG(ERR, "only 5tuple is supported.");
6831                 return -EINVAL;
6832         }
6833
6834         memset(&filter_5tuple, 0, sizeof(struct ixgbe_5tuple_filter_info));
6835         ret = ntuple_filter_to_5tuple(ntuple_filter, &filter_5tuple);
6836         if (ret < 0)
6837                 return ret;
6838
6839         filter = ixgbe_5tuple_filter_lookup(&filter_info->fivetuple_list,
6840                                          &filter_5tuple);
6841         if (filter == NULL) {
6842                 PMD_DRV_LOG(ERR, "filter doesn't exist.");
6843                 return -ENOENT;
6844         }
6845         ntuple_filter->queue = filter->queue;
6846         return 0;
6847 }
6848
6849 /*
6850  * ixgbe_ntuple_filter_handle - Handle operations for ntuple filter.
6851  * @dev: pointer to rte_eth_dev structure
6852  * @filter_op:operation will be taken.
6853  * @arg: a pointer to specific structure corresponding to the filter_op
6854  *
6855  * @return
6856  *    - On success, zero.
6857  *    - On failure, a negative value.
6858  */
6859 static int
6860 ixgbe_ntuple_filter_handle(struct rte_eth_dev *dev,
6861                                 enum rte_filter_op filter_op,
6862                                 void *arg)
6863 {
6864         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6865         int ret;
6866
6867         MAC_TYPE_FILTER_SUP_EXT(hw->mac.type);
6868
6869         if (filter_op == RTE_ETH_FILTER_NOP)
6870                 return 0;
6871
6872         if (arg == NULL) {
6873                 PMD_DRV_LOG(ERR, "arg shouldn't be NULL for operation %u.",
6874                             filter_op);
6875                 return -EINVAL;
6876         }
6877
6878         switch (filter_op) {
6879         case RTE_ETH_FILTER_ADD:
6880                 ret = ixgbe_add_del_ntuple_filter(dev,
6881                         (struct rte_eth_ntuple_filter *)arg,
6882                         TRUE);
6883                 break;
6884         case RTE_ETH_FILTER_DELETE:
6885                 ret = ixgbe_add_del_ntuple_filter(dev,
6886                         (struct rte_eth_ntuple_filter *)arg,
6887                         FALSE);
6888                 break;
6889         case RTE_ETH_FILTER_GET:
6890                 ret = ixgbe_get_ntuple_filter(dev,
6891                         (struct rte_eth_ntuple_filter *)arg);
6892                 break;
6893         default:
6894                 PMD_DRV_LOG(ERR, "unsupported operation %u.", filter_op);
6895                 ret = -EINVAL;
6896                 break;
6897         }
6898         return ret;
6899 }
6900
6901 int
6902 ixgbe_add_del_ethertype_filter(struct rte_eth_dev *dev,
6903                         struct rte_eth_ethertype_filter *filter,
6904                         bool add)
6905 {
6906         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6907         struct ixgbe_filter_info *filter_info =
6908                 IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
6909         uint32_t etqf = 0;
6910         uint32_t etqs = 0;
6911         int ret;
6912         struct ixgbe_ethertype_filter ethertype_filter;
6913
6914         if (filter->queue >= IXGBE_MAX_RX_QUEUE_NUM)
6915                 return -EINVAL;
6916
6917         if (filter->ether_type == RTE_ETHER_TYPE_IPV4 ||
6918                 filter->ether_type == RTE_ETHER_TYPE_IPV6) {
6919                 PMD_DRV_LOG(ERR, "unsupported ether_type(0x%04x) in"
6920                         " ethertype filter.", filter->ether_type);
6921                 return -EINVAL;
6922         }
6923
6924         if (filter->flags & RTE_ETHTYPE_FLAGS_MAC) {
6925                 PMD_DRV_LOG(ERR, "mac compare is unsupported.");
6926                 return -EINVAL;
6927         }
6928         if (filter->flags & RTE_ETHTYPE_FLAGS_DROP) {
6929                 PMD_DRV_LOG(ERR, "drop option is unsupported.");
6930                 return -EINVAL;
6931         }
6932
6933         ret = ixgbe_ethertype_filter_lookup(filter_info, filter->ether_type);
6934         if (ret >= 0 && add) {
6935                 PMD_DRV_LOG(ERR, "ethertype (0x%04x) filter exists.",
6936                             filter->ether_type);
6937                 return -EEXIST;
6938         }
6939         if (ret < 0 && !add) {
6940                 PMD_DRV_LOG(ERR, "ethertype (0x%04x) filter doesn't exist.",
6941                             filter->ether_type);
6942                 return -ENOENT;
6943         }
6944
6945         if (add) {
6946                 etqf = IXGBE_ETQF_FILTER_EN;
6947                 etqf |= (uint32_t)filter->ether_type;
6948                 etqs |= (uint32_t)((filter->queue <<
6949                                     IXGBE_ETQS_RX_QUEUE_SHIFT) &
6950                                     IXGBE_ETQS_RX_QUEUE);
6951                 etqs |= IXGBE_ETQS_QUEUE_EN;
6952
6953                 ethertype_filter.ethertype = filter->ether_type;
6954                 ethertype_filter.etqf = etqf;
6955                 ethertype_filter.etqs = etqs;
6956                 ethertype_filter.conf = FALSE;
6957                 ret = ixgbe_ethertype_filter_insert(filter_info,
6958                                                     &ethertype_filter);
6959                 if (ret < 0) {
6960                         PMD_DRV_LOG(ERR, "ethertype filters are full.");
6961                         return -ENOSPC;
6962                 }
6963         } else {
6964                 ret = ixgbe_ethertype_filter_remove(filter_info, (uint8_t)ret);
6965                 if (ret < 0)
6966                         return -ENOSYS;
6967         }
6968         IXGBE_WRITE_REG(hw, IXGBE_ETQF(ret), etqf);
6969         IXGBE_WRITE_REG(hw, IXGBE_ETQS(ret), etqs);
6970         IXGBE_WRITE_FLUSH(hw);
6971
6972         return 0;
6973 }
6974
6975 static int
6976 ixgbe_get_ethertype_filter(struct rte_eth_dev *dev,
6977                         struct rte_eth_ethertype_filter *filter)
6978 {
6979         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
6980         struct ixgbe_filter_info *filter_info =
6981                 IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
6982         uint32_t etqf, etqs;
6983         int ret;
6984
6985         ret = ixgbe_ethertype_filter_lookup(filter_info, filter->ether_type);
6986         if (ret < 0) {
6987                 PMD_DRV_LOG(ERR, "ethertype (0x%04x) filter doesn't exist.",
6988                             filter->ether_type);
6989                 return -ENOENT;
6990         }
6991
6992         etqf = IXGBE_READ_REG(hw, IXGBE_ETQF(ret));
6993         if (etqf & IXGBE_ETQF_FILTER_EN) {
6994                 etqs = IXGBE_READ_REG(hw, IXGBE_ETQS(ret));
6995                 filter->ether_type = etqf & IXGBE_ETQF_ETHERTYPE;
6996                 filter->flags = 0;
6997                 filter->queue = (etqs & IXGBE_ETQS_RX_QUEUE) >>
6998                                IXGBE_ETQS_RX_QUEUE_SHIFT;
6999                 return 0;
7000         }
7001         return -ENOENT;
7002 }
7003
7004 /*
7005  * ixgbe_ethertype_filter_handle - Handle operations for ethertype filter.
7006  * @dev: pointer to rte_eth_dev structure
7007  * @filter_op:operation will be taken.
7008  * @arg: a pointer to specific structure corresponding to the filter_op
7009  */
7010 static int
7011 ixgbe_ethertype_filter_handle(struct rte_eth_dev *dev,
7012                                 enum rte_filter_op filter_op,
7013                                 void *arg)
7014 {
7015         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7016         int ret;
7017
7018         MAC_TYPE_FILTER_SUP(hw->mac.type);
7019
7020         if (filter_op == RTE_ETH_FILTER_NOP)
7021                 return 0;
7022
7023         if (arg == NULL) {
7024                 PMD_DRV_LOG(ERR, "arg shouldn't be NULL for operation %u.",
7025                             filter_op);
7026                 return -EINVAL;
7027         }
7028
7029         switch (filter_op) {
7030         case RTE_ETH_FILTER_ADD:
7031                 ret = ixgbe_add_del_ethertype_filter(dev,
7032                         (struct rte_eth_ethertype_filter *)arg,
7033                         TRUE);
7034                 break;
7035         case RTE_ETH_FILTER_DELETE:
7036                 ret = ixgbe_add_del_ethertype_filter(dev,
7037                         (struct rte_eth_ethertype_filter *)arg,
7038                         FALSE);
7039                 break;
7040         case RTE_ETH_FILTER_GET:
7041                 ret = ixgbe_get_ethertype_filter(dev,
7042                         (struct rte_eth_ethertype_filter *)arg);
7043                 break;
7044         default:
7045                 PMD_DRV_LOG(ERR, "unsupported operation %u.", filter_op);
7046                 ret = -EINVAL;
7047                 break;
7048         }
7049         return ret;
7050 }
7051
7052 static int
7053 ixgbe_dev_filter_ctrl(struct rte_eth_dev *dev,
7054                      enum rte_filter_type filter_type,
7055                      enum rte_filter_op filter_op,
7056                      void *arg)
7057 {
7058         int ret = 0;
7059
7060         switch (filter_type) {
7061         case RTE_ETH_FILTER_NTUPLE:
7062                 ret = ixgbe_ntuple_filter_handle(dev, filter_op, arg);
7063                 break;
7064         case RTE_ETH_FILTER_ETHERTYPE:
7065                 ret = ixgbe_ethertype_filter_handle(dev, filter_op, arg);
7066                 break;
7067         case RTE_ETH_FILTER_SYN:
7068                 ret = ixgbe_syn_filter_handle(dev, filter_op, arg);
7069                 break;
7070         case RTE_ETH_FILTER_FDIR:
7071                 ret = ixgbe_fdir_ctrl_func(dev, filter_op, arg);
7072                 break;
7073         case RTE_ETH_FILTER_L2_TUNNEL:
7074                 ret = ixgbe_dev_l2_tunnel_filter_handle(dev, filter_op, arg);
7075                 break;
7076         case RTE_ETH_FILTER_GENERIC:
7077                 if (filter_op != RTE_ETH_FILTER_GET)
7078                         return -EINVAL;
7079                 *(const void **)arg = &ixgbe_flow_ops;
7080                 break;
7081         default:
7082                 PMD_DRV_LOG(WARNING, "Filter type (%d) not supported",
7083                                                         filter_type);
7084                 ret = -EINVAL;
7085                 break;
7086         }
7087
7088         return ret;
7089 }
7090
7091 static u8 *
7092 ixgbe_dev_addr_list_itr(__rte_unused struct ixgbe_hw *hw,
7093                         u8 **mc_addr_ptr, u32 *vmdq)
7094 {
7095         u8 *mc_addr;
7096
7097         *vmdq = 0;
7098         mc_addr = *mc_addr_ptr;
7099         *mc_addr_ptr = (mc_addr + sizeof(struct rte_ether_addr));
7100         return mc_addr;
7101 }
7102
7103 static int
7104 ixgbe_dev_set_mc_addr_list(struct rte_eth_dev *dev,
7105                           struct rte_ether_addr *mc_addr_set,
7106                           uint32_t nb_mc_addr)
7107 {
7108         struct ixgbe_hw *hw;
7109         u8 *mc_addr_list;
7110
7111         hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7112         mc_addr_list = (u8 *)mc_addr_set;
7113         return ixgbe_update_mc_addr_list(hw, mc_addr_list, nb_mc_addr,
7114                                          ixgbe_dev_addr_list_itr, TRUE);
7115 }
7116
7117 static uint64_t
7118 ixgbe_read_systime_cyclecounter(struct rte_eth_dev *dev)
7119 {
7120         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7121         uint64_t systime_cycles;
7122
7123         switch (hw->mac.type) {
7124         case ixgbe_mac_X550:
7125         case ixgbe_mac_X550EM_x:
7126         case ixgbe_mac_X550EM_a:
7127                 /* SYSTIMEL stores ns and SYSTIMEH stores seconds. */
7128                 systime_cycles = (uint64_t)IXGBE_READ_REG(hw, IXGBE_SYSTIML);
7129                 systime_cycles += (uint64_t)IXGBE_READ_REG(hw, IXGBE_SYSTIMH)
7130                                 * NSEC_PER_SEC;
7131                 break;
7132         default:
7133                 systime_cycles = (uint64_t)IXGBE_READ_REG(hw, IXGBE_SYSTIML);
7134                 systime_cycles |= (uint64_t)IXGBE_READ_REG(hw, IXGBE_SYSTIMH)
7135                                 << 32;
7136         }
7137
7138         return systime_cycles;
7139 }
7140
7141 static uint64_t
7142 ixgbe_read_rx_tstamp_cyclecounter(struct rte_eth_dev *dev)
7143 {
7144         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7145         uint64_t rx_tstamp_cycles;
7146
7147         switch (hw->mac.type) {
7148         case ixgbe_mac_X550:
7149         case ixgbe_mac_X550EM_x:
7150         case ixgbe_mac_X550EM_a:
7151                 /* RXSTMPL stores ns and RXSTMPH stores seconds. */
7152                 rx_tstamp_cycles = (uint64_t)IXGBE_READ_REG(hw, IXGBE_RXSTMPL);
7153                 rx_tstamp_cycles += (uint64_t)IXGBE_READ_REG(hw, IXGBE_RXSTMPH)
7154                                 * NSEC_PER_SEC;
7155                 break;
7156         default:
7157                 /* RXSTMPL stores ns and RXSTMPH stores seconds. */
7158                 rx_tstamp_cycles = (uint64_t)IXGBE_READ_REG(hw, IXGBE_RXSTMPL);
7159                 rx_tstamp_cycles |= (uint64_t)IXGBE_READ_REG(hw, IXGBE_RXSTMPH)
7160                                 << 32;
7161         }
7162
7163         return rx_tstamp_cycles;
7164 }
7165
7166 static uint64_t
7167 ixgbe_read_tx_tstamp_cyclecounter(struct rte_eth_dev *dev)
7168 {
7169         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7170         uint64_t tx_tstamp_cycles;
7171
7172         switch (hw->mac.type) {
7173         case ixgbe_mac_X550:
7174         case ixgbe_mac_X550EM_x:
7175         case ixgbe_mac_X550EM_a:
7176                 /* TXSTMPL stores ns and TXSTMPH stores seconds. */
7177                 tx_tstamp_cycles = (uint64_t)IXGBE_READ_REG(hw, IXGBE_TXSTMPL);
7178                 tx_tstamp_cycles += (uint64_t)IXGBE_READ_REG(hw, IXGBE_TXSTMPH)
7179                                 * NSEC_PER_SEC;
7180                 break;
7181         default:
7182                 /* TXSTMPL stores ns and TXSTMPH stores seconds. */
7183                 tx_tstamp_cycles = (uint64_t)IXGBE_READ_REG(hw, IXGBE_TXSTMPL);
7184                 tx_tstamp_cycles |= (uint64_t)IXGBE_READ_REG(hw, IXGBE_TXSTMPH)
7185                                 << 32;
7186         }
7187
7188         return tx_tstamp_cycles;
7189 }
7190
7191 static void
7192 ixgbe_start_timecounters(struct rte_eth_dev *dev)
7193 {
7194         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7195         struct ixgbe_adapter *adapter = dev->data->dev_private;
7196         struct rte_eth_link link;
7197         uint32_t incval = 0;
7198         uint32_t shift = 0;
7199
7200         /* Get current link speed. */
7201         ixgbe_dev_link_update(dev, 1);
7202         rte_eth_linkstatus_get(dev, &link);
7203
7204         switch (link.link_speed) {
7205         case ETH_SPEED_NUM_100M:
7206                 incval = IXGBE_INCVAL_100;
7207                 shift = IXGBE_INCVAL_SHIFT_100;
7208                 break;
7209         case ETH_SPEED_NUM_1G:
7210                 incval = IXGBE_INCVAL_1GB;
7211                 shift = IXGBE_INCVAL_SHIFT_1GB;
7212                 break;
7213         case ETH_SPEED_NUM_10G:
7214         default:
7215                 incval = IXGBE_INCVAL_10GB;
7216                 shift = IXGBE_INCVAL_SHIFT_10GB;
7217                 break;
7218         }
7219
7220         switch (hw->mac.type) {
7221         case ixgbe_mac_X550:
7222         case ixgbe_mac_X550EM_x:
7223         case ixgbe_mac_X550EM_a:
7224                 /* Independent of link speed. */
7225                 incval = 1;
7226                 /* Cycles read will be interpreted as ns. */
7227                 shift = 0;
7228                 /* Fall-through */
7229         case ixgbe_mac_X540:
7230                 IXGBE_WRITE_REG(hw, IXGBE_TIMINCA, incval);
7231                 break;
7232         case ixgbe_mac_82599EB:
7233                 incval >>= IXGBE_INCVAL_SHIFT_82599;
7234                 shift -= IXGBE_INCVAL_SHIFT_82599;
7235                 IXGBE_WRITE_REG(hw, IXGBE_TIMINCA,
7236                                 (1 << IXGBE_INCPER_SHIFT_82599) | incval);
7237                 break;
7238         default:
7239                 /* Not supported. */
7240                 return;
7241         }
7242
7243         memset(&adapter->systime_tc, 0, sizeof(struct rte_timecounter));
7244         memset(&adapter->rx_tstamp_tc, 0, sizeof(struct rte_timecounter));
7245         memset(&adapter->tx_tstamp_tc, 0, sizeof(struct rte_timecounter));
7246
7247         adapter->systime_tc.cc_mask = IXGBE_CYCLECOUNTER_MASK;
7248         adapter->systime_tc.cc_shift = shift;
7249         adapter->systime_tc.nsec_mask = (1ULL << shift) - 1;
7250
7251         adapter->rx_tstamp_tc.cc_mask = IXGBE_CYCLECOUNTER_MASK;
7252         adapter->rx_tstamp_tc.cc_shift = shift;
7253         adapter->rx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
7254
7255         adapter->tx_tstamp_tc.cc_mask = IXGBE_CYCLECOUNTER_MASK;
7256         adapter->tx_tstamp_tc.cc_shift = shift;
7257         adapter->tx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
7258 }
7259
7260 static int
7261 ixgbe_timesync_adjust_time(struct rte_eth_dev *dev, int64_t delta)
7262 {
7263         struct ixgbe_adapter *adapter = dev->data->dev_private;
7264
7265         adapter->systime_tc.nsec += delta;
7266         adapter->rx_tstamp_tc.nsec += delta;
7267         adapter->tx_tstamp_tc.nsec += delta;
7268
7269         return 0;
7270 }
7271
7272 static int
7273 ixgbe_timesync_write_time(struct rte_eth_dev *dev, const struct timespec *ts)
7274 {
7275         uint64_t ns;
7276         struct ixgbe_adapter *adapter = dev->data->dev_private;
7277
7278         ns = rte_timespec_to_ns(ts);
7279         /* Set the timecounters to a new value. */
7280         adapter->systime_tc.nsec = ns;
7281         adapter->rx_tstamp_tc.nsec = ns;
7282         adapter->tx_tstamp_tc.nsec = ns;
7283
7284         return 0;
7285 }
7286
7287 static int
7288 ixgbe_timesync_read_time(struct rte_eth_dev *dev, struct timespec *ts)
7289 {
7290         uint64_t ns, systime_cycles;
7291         struct ixgbe_adapter *adapter = dev->data->dev_private;
7292
7293         systime_cycles = ixgbe_read_systime_cyclecounter(dev);
7294         ns = rte_timecounter_update(&adapter->systime_tc, systime_cycles);
7295         *ts = rte_ns_to_timespec(ns);
7296
7297         return 0;
7298 }
7299
7300 static int
7301 ixgbe_timesync_enable(struct rte_eth_dev *dev)
7302 {
7303         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7304         uint32_t tsync_ctl;
7305         uint32_t tsauxc;
7306
7307         /* Stop the timesync system time. */
7308         IXGBE_WRITE_REG(hw, IXGBE_TIMINCA, 0x0);
7309         /* Reset the timesync system time value. */
7310         IXGBE_WRITE_REG(hw, IXGBE_SYSTIML, 0x0);
7311         IXGBE_WRITE_REG(hw, IXGBE_SYSTIMH, 0x0);
7312
7313         /* Enable system time for platforms where it isn't on by default. */
7314         tsauxc = IXGBE_READ_REG(hw, IXGBE_TSAUXC);
7315         tsauxc &= ~IXGBE_TSAUXC_DISABLE_SYSTIME;
7316         IXGBE_WRITE_REG(hw, IXGBE_TSAUXC, tsauxc);
7317
7318         ixgbe_start_timecounters(dev);
7319
7320         /* Enable L2 filtering of IEEE1588/802.1AS Ethernet frame types. */
7321         IXGBE_WRITE_REG(hw, IXGBE_ETQF(IXGBE_ETQF_FILTER_1588),
7322                         (RTE_ETHER_TYPE_1588 |
7323                          IXGBE_ETQF_FILTER_EN |
7324                          IXGBE_ETQF_1588));
7325
7326         /* Enable timestamping of received PTP packets. */
7327         tsync_ctl = IXGBE_READ_REG(hw, IXGBE_TSYNCRXCTL);
7328         tsync_ctl |= IXGBE_TSYNCRXCTL_ENABLED;
7329         IXGBE_WRITE_REG(hw, IXGBE_TSYNCRXCTL, tsync_ctl);
7330
7331         /* Enable timestamping of transmitted PTP packets. */
7332         tsync_ctl = IXGBE_READ_REG(hw, IXGBE_TSYNCTXCTL);
7333         tsync_ctl |= IXGBE_TSYNCTXCTL_ENABLED;
7334         IXGBE_WRITE_REG(hw, IXGBE_TSYNCTXCTL, tsync_ctl);
7335
7336         IXGBE_WRITE_FLUSH(hw);
7337
7338         return 0;
7339 }
7340
7341 static int
7342 ixgbe_timesync_disable(struct rte_eth_dev *dev)
7343 {
7344         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7345         uint32_t tsync_ctl;
7346
7347         /* Disable timestamping of transmitted PTP packets. */
7348         tsync_ctl = IXGBE_READ_REG(hw, IXGBE_TSYNCTXCTL);
7349         tsync_ctl &= ~IXGBE_TSYNCTXCTL_ENABLED;
7350         IXGBE_WRITE_REG(hw, IXGBE_TSYNCTXCTL, tsync_ctl);
7351
7352         /* Disable timestamping of received PTP packets. */
7353         tsync_ctl = IXGBE_READ_REG(hw, IXGBE_TSYNCRXCTL);
7354         tsync_ctl &= ~IXGBE_TSYNCRXCTL_ENABLED;
7355         IXGBE_WRITE_REG(hw, IXGBE_TSYNCRXCTL, tsync_ctl);
7356
7357         /* Disable L2 filtering of IEEE1588/802.1AS Ethernet frame types. */
7358         IXGBE_WRITE_REG(hw, IXGBE_ETQF(IXGBE_ETQF_FILTER_1588), 0);
7359
7360         /* Stop incrementating the System Time registers. */
7361         IXGBE_WRITE_REG(hw, IXGBE_TIMINCA, 0);
7362
7363         return 0;
7364 }
7365
7366 static int
7367 ixgbe_timesync_read_rx_timestamp(struct rte_eth_dev *dev,
7368                                  struct timespec *timestamp,
7369                                  uint32_t flags __rte_unused)
7370 {
7371         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7372         struct ixgbe_adapter *adapter = dev->data->dev_private;
7373         uint32_t tsync_rxctl;
7374         uint64_t rx_tstamp_cycles;
7375         uint64_t ns;
7376
7377         tsync_rxctl = IXGBE_READ_REG(hw, IXGBE_TSYNCRXCTL);
7378         if ((tsync_rxctl & IXGBE_TSYNCRXCTL_VALID) == 0)
7379                 return -EINVAL;
7380
7381         rx_tstamp_cycles = ixgbe_read_rx_tstamp_cyclecounter(dev);
7382         ns = rte_timecounter_update(&adapter->rx_tstamp_tc, rx_tstamp_cycles);
7383         *timestamp = rte_ns_to_timespec(ns);
7384
7385         return  0;
7386 }
7387
7388 static int
7389 ixgbe_timesync_read_tx_timestamp(struct rte_eth_dev *dev,
7390                                  struct timespec *timestamp)
7391 {
7392         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7393         struct ixgbe_adapter *adapter = dev->data->dev_private;
7394         uint32_t tsync_txctl;
7395         uint64_t tx_tstamp_cycles;
7396         uint64_t ns;
7397
7398         tsync_txctl = IXGBE_READ_REG(hw, IXGBE_TSYNCTXCTL);
7399         if ((tsync_txctl & IXGBE_TSYNCTXCTL_VALID) == 0)
7400                 return -EINVAL;
7401
7402         tx_tstamp_cycles = ixgbe_read_tx_tstamp_cyclecounter(dev);
7403         ns = rte_timecounter_update(&adapter->tx_tstamp_tc, tx_tstamp_cycles);
7404         *timestamp = rte_ns_to_timespec(ns);
7405
7406         return 0;
7407 }
7408
7409 static int
7410 ixgbe_get_reg_length(struct rte_eth_dev *dev)
7411 {
7412         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7413         int count = 0;
7414         int g_ind = 0;
7415         const struct reg_info *reg_group;
7416         const struct reg_info **reg_set = (hw->mac.type == ixgbe_mac_82598EB) ?
7417                                     ixgbe_regs_mac_82598EB : ixgbe_regs_others;
7418
7419         while ((reg_group = reg_set[g_ind++]))
7420                 count += ixgbe_regs_group_count(reg_group);
7421
7422         return count;
7423 }
7424
7425 static int
7426 ixgbevf_get_reg_length(struct rte_eth_dev *dev __rte_unused)
7427 {
7428         int count = 0;
7429         int g_ind = 0;
7430         const struct reg_info *reg_group;
7431
7432         while ((reg_group = ixgbevf_regs[g_ind++]))
7433                 count += ixgbe_regs_group_count(reg_group);
7434
7435         return count;
7436 }
7437
7438 static int
7439 ixgbe_get_regs(struct rte_eth_dev *dev,
7440               struct rte_dev_reg_info *regs)
7441 {
7442         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7443         uint32_t *data = regs->data;
7444         int g_ind = 0;
7445         int count = 0;
7446         const struct reg_info *reg_group;
7447         const struct reg_info **reg_set = (hw->mac.type == ixgbe_mac_82598EB) ?
7448                                     ixgbe_regs_mac_82598EB : ixgbe_regs_others;
7449
7450         if (data == NULL) {
7451                 regs->length = ixgbe_get_reg_length(dev);
7452                 regs->width = sizeof(uint32_t);
7453                 return 0;
7454         }
7455
7456         /* Support only full register dump */
7457         if ((regs->length == 0) ||
7458             (regs->length == (uint32_t)ixgbe_get_reg_length(dev))) {
7459                 regs->version = hw->mac.type << 24 | hw->revision_id << 16 |
7460                         hw->device_id;
7461                 while ((reg_group = reg_set[g_ind++]))
7462                         count += ixgbe_read_regs_group(dev, &data[count],
7463                                 reg_group);
7464                 return 0;
7465         }
7466
7467         return -ENOTSUP;
7468 }
7469
7470 static int
7471 ixgbevf_get_regs(struct rte_eth_dev *dev,
7472                 struct rte_dev_reg_info *regs)
7473 {
7474         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7475         uint32_t *data = regs->data;
7476         int g_ind = 0;
7477         int count = 0;
7478         const struct reg_info *reg_group;
7479
7480         if (data == NULL) {
7481                 regs->length = ixgbevf_get_reg_length(dev);
7482                 regs->width = sizeof(uint32_t);
7483                 return 0;
7484         }
7485
7486         /* Support only full register dump */
7487         if ((regs->length == 0) ||
7488             (regs->length == (uint32_t)ixgbevf_get_reg_length(dev))) {
7489                 regs->version = hw->mac.type << 24 | hw->revision_id << 16 |
7490                         hw->device_id;
7491                 while ((reg_group = ixgbevf_regs[g_ind++]))
7492                         count += ixgbe_read_regs_group(dev, &data[count],
7493                                                       reg_group);
7494                 return 0;
7495         }
7496
7497         return -ENOTSUP;
7498 }
7499
7500 static int
7501 ixgbe_get_eeprom_length(struct rte_eth_dev *dev)
7502 {
7503         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7504
7505         /* Return unit is byte count */
7506         return hw->eeprom.word_size * 2;
7507 }
7508
7509 static int
7510 ixgbe_get_eeprom(struct rte_eth_dev *dev,
7511                 struct rte_dev_eeprom_info *in_eeprom)
7512 {
7513         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7514         struct ixgbe_eeprom_info *eeprom = &hw->eeprom;
7515         uint16_t *data = in_eeprom->data;
7516         int first, length;
7517
7518         first = in_eeprom->offset >> 1;
7519         length = in_eeprom->length >> 1;
7520         if ((first > hw->eeprom.word_size) ||
7521             ((first + length) > hw->eeprom.word_size))
7522                 return -EINVAL;
7523
7524         in_eeprom->magic = hw->vendor_id | (hw->device_id << 16);
7525
7526         return eeprom->ops.read_buffer(hw, first, length, data);
7527 }
7528
7529 static int
7530 ixgbe_set_eeprom(struct rte_eth_dev *dev,
7531                 struct rte_dev_eeprom_info *in_eeprom)
7532 {
7533         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7534         struct ixgbe_eeprom_info *eeprom = &hw->eeprom;
7535         uint16_t *data = in_eeprom->data;
7536         int first, length;
7537
7538         first = in_eeprom->offset >> 1;
7539         length = in_eeprom->length >> 1;
7540         if ((first > hw->eeprom.word_size) ||
7541             ((first + length) > hw->eeprom.word_size))
7542                 return -EINVAL;
7543
7544         in_eeprom->magic = hw->vendor_id | (hw->device_id << 16);
7545
7546         return eeprom->ops.write_buffer(hw,  first, length, data);
7547 }
7548
7549 static int
7550 ixgbe_get_module_info(struct rte_eth_dev *dev,
7551                       struct rte_eth_dev_module_info *modinfo)
7552 {
7553         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7554         uint32_t status;
7555         uint8_t sff8472_rev, addr_mode;
7556         bool page_swap = false;
7557
7558         /* Check whether we support SFF-8472 or not */
7559         status = hw->phy.ops.read_i2c_eeprom(hw,
7560                                              IXGBE_SFF_SFF_8472_COMP,
7561                                              &sff8472_rev);
7562         if (status != 0)
7563                 return -EIO;
7564
7565         /* addressing mode is not supported */
7566         status = hw->phy.ops.read_i2c_eeprom(hw,
7567                                              IXGBE_SFF_SFF_8472_SWAP,
7568                                              &addr_mode);
7569         if (status != 0)
7570                 return -EIO;
7571
7572         if (addr_mode & IXGBE_SFF_ADDRESSING_MODE) {
7573                 PMD_DRV_LOG(ERR,
7574                             "Address change required to access page 0xA2, "
7575                             "but not supported. Please report the module "
7576                             "type to the driver maintainers.");
7577                 page_swap = true;
7578         }
7579
7580         if (sff8472_rev == IXGBE_SFF_SFF_8472_UNSUP || page_swap) {
7581                 /* We have a SFP, but it does not support SFF-8472 */
7582                 modinfo->type = RTE_ETH_MODULE_SFF_8079;
7583                 modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8079_LEN;
7584         } else {
7585                 /* We have a SFP which supports a revision of SFF-8472. */
7586                 modinfo->type = RTE_ETH_MODULE_SFF_8472;
7587                 modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8472_LEN;
7588         }
7589
7590         return 0;
7591 }
7592
7593 static int
7594 ixgbe_get_module_eeprom(struct rte_eth_dev *dev,
7595                         struct rte_dev_eeprom_info *info)
7596 {
7597         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7598         uint32_t status = IXGBE_ERR_PHY_ADDR_INVALID;
7599         uint8_t databyte = 0xFF;
7600         uint8_t *data = info->data;
7601         uint32_t i = 0;
7602
7603         if (info->length == 0)
7604                 return -EINVAL;
7605
7606         for (i = info->offset; i < info->offset + info->length; i++) {
7607                 if (i < RTE_ETH_MODULE_SFF_8079_LEN)
7608                         status = hw->phy.ops.read_i2c_eeprom(hw, i, &databyte);
7609                 else
7610                         status = hw->phy.ops.read_i2c_sff8472(hw, i, &databyte);
7611
7612                 if (status != 0)
7613                         return -EIO;
7614
7615                 data[i - info->offset] = databyte;
7616         }
7617
7618         return 0;
7619 }
7620
7621 uint16_t
7622 ixgbe_reta_size_get(enum ixgbe_mac_type mac_type) {
7623         switch (mac_type) {
7624         case ixgbe_mac_X550:
7625         case ixgbe_mac_X550EM_x:
7626         case ixgbe_mac_X550EM_a:
7627                 return ETH_RSS_RETA_SIZE_512;
7628         case ixgbe_mac_X550_vf:
7629         case ixgbe_mac_X550EM_x_vf:
7630         case ixgbe_mac_X550EM_a_vf:
7631                 return ETH_RSS_RETA_SIZE_64;
7632         case ixgbe_mac_X540_vf:
7633         case ixgbe_mac_82599_vf:
7634                 return 0;
7635         default:
7636                 return ETH_RSS_RETA_SIZE_128;
7637         }
7638 }
7639
7640 uint32_t
7641 ixgbe_reta_reg_get(enum ixgbe_mac_type mac_type, uint16_t reta_idx) {
7642         switch (mac_type) {
7643         case ixgbe_mac_X550:
7644         case ixgbe_mac_X550EM_x:
7645         case ixgbe_mac_X550EM_a:
7646                 if (reta_idx < ETH_RSS_RETA_SIZE_128)
7647                         return IXGBE_RETA(reta_idx >> 2);
7648                 else
7649                         return IXGBE_ERETA((reta_idx - ETH_RSS_RETA_SIZE_128) >> 2);
7650         case ixgbe_mac_X550_vf:
7651         case ixgbe_mac_X550EM_x_vf:
7652         case ixgbe_mac_X550EM_a_vf:
7653                 return IXGBE_VFRETA(reta_idx >> 2);
7654         default:
7655                 return IXGBE_RETA(reta_idx >> 2);
7656         }
7657 }
7658
7659 uint32_t
7660 ixgbe_mrqc_reg_get(enum ixgbe_mac_type mac_type) {
7661         switch (mac_type) {
7662         case ixgbe_mac_X550_vf:
7663         case ixgbe_mac_X550EM_x_vf:
7664         case ixgbe_mac_X550EM_a_vf:
7665                 return IXGBE_VFMRQC;
7666         default:
7667                 return IXGBE_MRQC;
7668         }
7669 }
7670
7671 uint32_t
7672 ixgbe_rssrk_reg_get(enum ixgbe_mac_type mac_type, uint8_t i) {
7673         switch (mac_type) {
7674         case ixgbe_mac_X550_vf:
7675         case ixgbe_mac_X550EM_x_vf:
7676         case ixgbe_mac_X550EM_a_vf:
7677                 return IXGBE_VFRSSRK(i);
7678         default:
7679                 return IXGBE_RSSRK(i);
7680         }
7681 }
7682
7683 bool
7684 ixgbe_rss_update_sp(enum ixgbe_mac_type mac_type) {
7685         switch (mac_type) {
7686         case ixgbe_mac_82599_vf:
7687         case ixgbe_mac_X540_vf:
7688                 return 0;
7689         default:
7690                 return 1;
7691         }
7692 }
7693
7694 static int
7695 ixgbe_dev_get_dcb_info(struct rte_eth_dev *dev,
7696                         struct rte_eth_dcb_info *dcb_info)
7697 {
7698         struct ixgbe_dcb_config *dcb_config =
7699                         IXGBE_DEV_PRIVATE_TO_DCB_CFG(dev->data->dev_private);
7700         struct ixgbe_dcb_tc_config *tc;
7701         struct rte_eth_dcb_tc_queue_mapping *tc_queue;
7702         uint8_t nb_tcs;
7703         uint8_t i, j;
7704
7705         if (dev->data->dev_conf.rxmode.mq_mode & ETH_MQ_RX_DCB_FLAG)
7706                 dcb_info->nb_tcs = dcb_config->num_tcs.pg_tcs;
7707         else
7708                 dcb_info->nb_tcs = 1;
7709
7710         tc_queue = &dcb_info->tc_queue;
7711         nb_tcs = dcb_info->nb_tcs;
7712
7713         if (dcb_config->vt_mode) { /* vt is enabled*/
7714                 struct rte_eth_vmdq_dcb_conf *vmdq_rx_conf =
7715                                 &dev->data->dev_conf.rx_adv_conf.vmdq_dcb_conf;
7716                 for (i = 0; i < ETH_DCB_NUM_USER_PRIORITIES; i++)
7717                         dcb_info->prio_tc[i] = vmdq_rx_conf->dcb_tc[i];
7718                 if (RTE_ETH_DEV_SRIOV(dev).active > 0) {
7719                         for (j = 0; j < nb_tcs; j++) {
7720                                 tc_queue->tc_rxq[0][j].base = j;
7721                                 tc_queue->tc_rxq[0][j].nb_queue = 1;
7722                                 tc_queue->tc_txq[0][j].base = j;
7723                                 tc_queue->tc_txq[0][j].nb_queue = 1;
7724                         }
7725                 } else {
7726                         for (i = 0; i < vmdq_rx_conf->nb_queue_pools; i++) {
7727                                 for (j = 0; j < nb_tcs; j++) {
7728                                         tc_queue->tc_rxq[i][j].base =
7729                                                 i * nb_tcs + j;
7730                                         tc_queue->tc_rxq[i][j].nb_queue = 1;
7731                                         tc_queue->tc_txq[i][j].base =
7732                                                 i * nb_tcs + j;
7733                                         tc_queue->tc_txq[i][j].nb_queue = 1;
7734                                 }
7735                         }
7736                 }
7737         } else { /* vt is disabled*/
7738                 struct rte_eth_dcb_rx_conf *rx_conf =
7739                                 &dev->data->dev_conf.rx_adv_conf.dcb_rx_conf;
7740                 for (i = 0; i < ETH_DCB_NUM_USER_PRIORITIES; i++)
7741                         dcb_info->prio_tc[i] = rx_conf->dcb_tc[i];
7742                 if (dcb_info->nb_tcs == ETH_4_TCS) {
7743                         for (i = 0; i < dcb_info->nb_tcs; i++) {
7744                                 dcb_info->tc_queue.tc_rxq[0][i].base = i * 32;
7745                                 dcb_info->tc_queue.tc_rxq[0][i].nb_queue = 16;
7746                         }
7747                         dcb_info->tc_queue.tc_txq[0][0].base = 0;
7748                         dcb_info->tc_queue.tc_txq[0][1].base = 64;
7749                         dcb_info->tc_queue.tc_txq[0][2].base = 96;
7750                         dcb_info->tc_queue.tc_txq[0][3].base = 112;
7751                         dcb_info->tc_queue.tc_txq[0][0].nb_queue = 64;
7752                         dcb_info->tc_queue.tc_txq[0][1].nb_queue = 32;
7753                         dcb_info->tc_queue.tc_txq[0][2].nb_queue = 16;
7754                         dcb_info->tc_queue.tc_txq[0][3].nb_queue = 16;
7755                 } else if (dcb_info->nb_tcs == ETH_8_TCS) {
7756                         for (i = 0; i < dcb_info->nb_tcs; i++) {
7757                                 dcb_info->tc_queue.tc_rxq[0][i].base = i * 16;
7758                                 dcb_info->tc_queue.tc_rxq[0][i].nb_queue = 16;
7759                         }
7760                         dcb_info->tc_queue.tc_txq[0][0].base = 0;
7761                         dcb_info->tc_queue.tc_txq[0][1].base = 32;
7762                         dcb_info->tc_queue.tc_txq[0][2].base = 64;
7763                         dcb_info->tc_queue.tc_txq[0][3].base = 80;
7764                         dcb_info->tc_queue.tc_txq[0][4].base = 96;
7765                         dcb_info->tc_queue.tc_txq[0][5].base = 104;
7766                         dcb_info->tc_queue.tc_txq[0][6].base = 112;
7767                         dcb_info->tc_queue.tc_txq[0][7].base = 120;
7768                         dcb_info->tc_queue.tc_txq[0][0].nb_queue = 32;
7769                         dcb_info->tc_queue.tc_txq[0][1].nb_queue = 32;
7770                         dcb_info->tc_queue.tc_txq[0][2].nb_queue = 16;
7771                         dcb_info->tc_queue.tc_txq[0][3].nb_queue = 16;
7772                         dcb_info->tc_queue.tc_txq[0][4].nb_queue = 8;
7773                         dcb_info->tc_queue.tc_txq[0][5].nb_queue = 8;
7774                         dcb_info->tc_queue.tc_txq[0][6].nb_queue = 8;
7775                         dcb_info->tc_queue.tc_txq[0][7].nb_queue = 8;
7776                 }
7777         }
7778         for (i = 0; i < dcb_info->nb_tcs; i++) {
7779                 tc = &dcb_config->tc_config[i];
7780                 dcb_info->tc_bws[i] = tc->path[IXGBE_DCB_TX_CONFIG].bwg_percent;
7781         }
7782         return 0;
7783 }
7784
7785 /* Update e-tag ether type */
7786 static int
7787 ixgbe_update_e_tag_eth_type(struct ixgbe_hw *hw,
7788                             uint16_t ether_type)
7789 {
7790         uint32_t etag_etype;
7791
7792         if (hw->mac.type != ixgbe_mac_X550 &&
7793             hw->mac.type != ixgbe_mac_X550EM_x &&
7794             hw->mac.type != ixgbe_mac_X550EM_a) {
7795                 return -ENOTSUP;
7796         }
7797
7798         etag_etype = IXGBE_READ_REG(hw, IXGBE_ETAG_ETYPE);
7799         etag_etype &= ~IXGBE_ETAG_ETYPE_MASK;
7800         etag_etype |= ether_type;
7801         IXGBE_WRITE_REG(hw, IXGBE_ETAG_ETYPE, etag_etype);
7802         IXGBE_WRITE_FLUSH(hw);
7803
7804         return 0;
7805 }
7806
7807 /* Config l2 tunnel ether type */
7808 static int
7809 ixgbe_dev_l2_tunnel_eth_type_conf(struct rte_eth_dev *dev,
7810                                   struct rte_eth_l2_tunnel_conf *l2_tunnel)
7811 {
7812         int ret = 0;
7813         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7814         struct ixgbe_l2_tn_info *l2_tn_info =
7815                 IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(dev->data->dev_private);
7816
7817         if (l2_tunnel == NULL)
7818                 return -EINVAL;
7819
7820         switch (l2_tunnel->l2_tunnel_type) {
7821         case RTE_L2_TUNNEL_TYPE_E_TAG:
7822                 l2_tn_info->e_tag_ether_type = l2_tunnel->ether_type;
7823                 ret = ixgbe_update_e_tag_eth_type(hw, l2_tunnel->ether_type);
7824                 break;
7825         default:
7826                 PMD_DRV_LOG(ERR, "Invalid tunnel type");
7827                 ret = -EINVAL;
7828                 break;
7829         }
7830
7831         return ret;
7832 }
7833
7834 /* Enable e-tag tunnel */
7835 static int
7836 ixgbe_e_tag_enable(struct ixgbe_hw *hw)
7837 {
7838         uint32_t etag_etype;
7839
7840         if (hw->mac.type != ixgbe_mac_X550 &&
7841             hw->mac.type != ixgbe_mac_X550EM_x &&
7842             hw->mac.type != ixgbe_mac_X550EM_a) {
7843                 return -ENOTSUP;
7844         }
7845
7846         etag_etype = IXGBE_READ_REG(hw, IXGBE_ETAG_ETYPE);
7847         etag_etype |= IXGBE_ETAG_ETYPE_VALID;
7848         IXGBE_WRITE_REG(hw, IXGBE_ETAG_ETYPE, etag_etype);
7849         IXGBE_WRITE_FLUSH(hw);
7850
7851         return 0;
7852 }
7853
7854 /* Enable l2 tunnel */
7855 static int
7856 ixgbe_dev_l2_tunnel_enable(struct rte_eth_dev *dev,
7857                            enum rte_eth_tunnel_type l2_tunnel_type)
7858 {
7859         int ret = 0;
7860         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7861         struct ixgbe_l2_tn_info *l2_tn_info =
7862                 IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(dev->data->dev_private);
7863
7864         switch (l2_tunnel_type) {
7865         case RTE_L2_TUNNEL_TYPE_E_TAG:
7866                 l2_tn_info->e_tag_en = TRUE;
7867                 ret = ixgbe_e_tag_enable(hw);
7868                 break;
7869         default:
7870                 PMD_DRV_LOG(ERR, "Invalid tunnel type");
7871                 ret = -EINVAL;
7872                 break;
7873         }
7874
7875         return ret;
7876 }
7877
7878 /* Disable e-tag tunnel */
7879 static int
7880 ixgbe_e_tag_disable(struct ixgbe_hw *hw)
7881 {
7882         uint32_t etag_etype;
7883
7884         if (hw->mac.type != ixgbe_mac_X550 &&
7885             hw->mac.type != ixgbe_mac_X550EM_x &&
7886             hw->mac.type != ixgbe_mac_X550EM_a) {
7887                 return -ENOTSUP;
7888         }
7889
7890         etag_etype = IXGBE_READ_REG(hw, IXGBE_ETAG_ETYPE);
7891         etag_etype &= ~IXGBE_ETAG_ETYPE_VALID;
7892         IXGBE_WRITE_REG(hw, IXGBE_ETAG_ETYPE, etag_etype);
7893         IXGBE_WRITE_FLUSH(hw);
7894
7895         return 0;
7896 }
7897
7898 /* Disable l2 tunnel */
7899 static int
7900 ixgbe_dev_l2_tunnel_disable(struct rte_eth_dev *dev,
7901                             enum rte_eth_tunnel_type l2_tunnel_type)
7902 {
7903         int ret = 0;
7904         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7905         struct ixgbe_l2_tn_info *l2_tn_info =
7906                 IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(dev->data->dev_private);
7907
7908         switch (l2_tunnel_type) {
7909         case RTE_L2_TUNNEL_TYPE_E_TAG:
7910                 l2_tn_info->e_tag_en = FALSE;
7911                 ret = ixgbe_e_tag_disable(hw);
7912                 break;
7913         default:
7914                 PMD_DRV_LOG(ERR, "Invalid tunnel type");
7915                 ret = -EINVAL;
7916                 break;
7917         }
7918
7919         return ret;
7920 }
7921
7922 static int
7923 ixgbe_e_tag_filter_del(struct rte_eth_dev *dev,
7924                        struct rte_eth_l2_tunnel_conf *l2_tunnel)
7925 {
7926         int ret = 0;
7927         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7928         uint32_t i, rar_entries;
7929         uint32_t rar_low, rar_high;
7930
7931         if (hw->mac.type != ixgbe_mac_X550 &&
7932             hw->mac.type != ixgbe_mac_X550EM_x &&
7933             hw->mac.type != ixgbe_mac_X550EM_a) {
7934                 return -ENOTSUP;
7935         }
7936
7937         rar_entries = ixgbe_get_num_rx_addrs(hw);
7938
7939         for (i = 1; i < rar_entries; i++) {
7940                 rar_high = IXGBE_READ_REG(hw, IXGBE_RAH(i));
7941                 rar_low  = IXGBE_READ_REG(hw, IXGBE_RAL(i));
7942                 if ((rar_high & IXGBE_RAH_AV) &&
7943                     (rar_high & IXGBE_RAH_ADTYPE) &&
7944                     ((rar_low & IXGBE_RAL_ETAG_FILTER_MASK) ==
7945                      l2_tunnel->tunnel_id)) {
7946                         IXGBE_WRITE_REG(hw, IXGBE_RAL(i), 0);
7947                         IXGBE_WRITE_REG(hw, IXGBE_RAH(i), 0);
7948
7949                         ixgbe_clear_vmdq(hw, i, IXGBE_CLEAR_VMDQ_ALL);
7950
7951                         return ret;
7952                 }
7953         }
7954
7955         return ret;
7956 }
7957
7958 static int
7959 ixgbe_e_tag_filter_add(struct rte_eth_dev *dev,
7960                        struct rte_eth_l2_tunnel_conf *l2_tunnel)
7961 {
7962         int ret = 0;
7963         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
7964         uint32_t i, rar_entries;
7965         uint32_t rar_low, rar_high;
7966
7967         if (hw->mac.type != ixgbe_mac_X550 &&
7968             hw->mac.type != ixgbe_mac_X550EM_x &&
7969             hw->mac.type != ixgbe_mac_X550EM_a) {
7970                 return -ENOTSUP;
7971         }
7972
7973         /* One entry for one tunnel. Try to remove potential existing entry. */
7974         ixgbe_e_tag_filter_del(dev, l2_tunnel);
7975
7976         rar_entries = ixgbe_get_num_rx_addrs(hw);
7977
7978         for (i = 1; i < rar_entries; i++) {
7979                 rar_high = IXGBE_READ_REG(hw, IXGBE_RAH(i));
7980                 if (rar_high & IXGBE_RAH_AV) {
7981                         continue;
7982                 } else {
7983                         ixgbe_set_vmdq(hw, i, l2_tunnel->pool);
7984                         rar_high = IXGBE_RAH_AV | IXGBE_RAH_ADTYPE;
7985                         rar_low = l2_tunnel->tunnel_id;
7986
7987                         IXGBE_WRITE_REG(hw, IXGBE_RAL(i), rar_low);
7988                         IXGBE_WRITE_REG(hw, IXGBE_RAH(i), rar_high);
7989
7990                         return ret;
7991                 }
7992         }
7993
7994         PMD_INIT_LOG(NOTICE, "The table of E-tag forwarding rule is full."
7995                      " Please remove a rule before adding a new one.");
7996         return -EINVAL;
7997 }
7998
7999 static inline struct ixgbe_l2_tn_filter *
8000 ixgbe_l2_tn_filter_lookup(struct ixgbe_l2_tn_info *l2_tn_info,
8001                           struct ixgbe_l2_tn_key *key)
8002 {
8003         int ret;
8004
8005         ret = rte_hash_lookup(l2_tn_info->hash_handle, (const void *)key);
8006         if (ret < 0)
8007                 return NULL;
8008
8009         return l2_tn_info->hash_map[ret];
8010 }
8011
8012 static inline int
8013 ixgbe_insert_l2_tn_filter(struct ixgbe_l2_tn_info *l2_tn_info,
8014                           struct ixgbe_l2_tn_filter *l2_tn_filter)
8015 {
8016         int ret;
8017
8018         ret = rte_hash_add_key(l2_tn_info->hash_handle,
8019                                &l2_tn_filter->key);
8020
8021         if (ret < 0) {
8022                 PMD_DRV_LOG(ERR,
8023                             "Failed to insert L2 tunnel filter"
8024                             " to hash table %d!",
8025                             ret);
8026                 return ret;
8027         }
8028
8029         l2_tn_info->hash_map[ret] = l2_tn_filter;
8030
8031         TAILQ_INSERT_TAIL(&l2_tn_info->l2_tn_list, l2_tn_filter, entries);
8032
8033         return 0;
8034 }
8035
8036 static inline int
8037 ixgbe_remove_l2_tn_filter(struct ixgbe_l2_tn_info *l2_tn_info,
8038                           struct ixgbe_l2_tn_key *key)
8039 {
8040         int ret;
8041         struct ixgbe_l2_tn_filter *l2_tn_filter;
8042
8043         ret = rte_hash_del_key(l2_tn_info->hash_handle, key);
8044
8045         if (ret < 0) {
8046                 PMD_DRV_LOG(ERR,
8047                             "No such L2 tunnel filter to delete %d!",
8048                             ret);
8049                 return ret;
8050         }
8051
8052         l2_tn_filter = l2_tn_info->hash_map[ret];
8053         l2_tn_info->hash_map[ret] = NULL;
8054
8055         TAILQ_REMOVE(&l2_tn_info->l2_tn_list, l2_tn_filter, entries);
8056         rte_free(l2_tn_filter);
8057
8058         return 0;
8059 }
8060
8061 /* Add l2 tunnel filter */
8062 int
8063 ixgbe_dev_l2_tunnel_filter_add(struct rte_eth_dev *dev,
8064                                struct rte_eth_l2_tunnel_conf *l2_tunnel,
8065                                bool restore)
8066 {
8067         int ret;
8068         struct ixgbe_l2_tn_info *l2_tn_info =
8069                 IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(dev->data->dev_private);
8070         struct ixgbe_l2_tn_key key;
8071         struct ixgbe_l2_tn_filter *node;
8072
8073         if (!restore) {
8074                 key.l2_tn_type = l2_tunnel->l2_tunnel_type;
8075                 key.tn_id = l2_tunnel->tunnel_id;
8076
8077                 node = ixgbe_l2_tn_filter_lookup(l2_tn_info, &key);
8078
8079                 if (node) {
8080                         PMD_DRV_LOG(ERR,
8081                                     "The L2 tunnel filter already exists!");
8082                         return -EINVAL;
8083                 }
8084
8085                 node = rte_zmalloc("ixgbe_l2_tn",
8086                                    sizeof(struct ixgbe_l2_tn_filter),
8087                                    0);
8088                 if (!node)
8089                         return -ENOMEM;
8090
8091                 rte_memcpy(&node->key,
8092                                  &key,
8093                                  sizeof(struct ixgbe_l2_tn_key));
8094                 node->pool = l2_tunnel->pool;
8095                 ret = ixgbe_insert_l2_tn_filter(l2_tn_info, node);
8096                 if (ret < 0) {
8097                         rte_free(node);
8098                         return ret;
8099                 }
8100         }
8101
8102         switch (l2_tunnel->l2_tunnel_type) {
8103         case RTE_L2_TUNNEL_TYPE_E_TAG:
8104                 ret = ixgbe_e_tag_filter_add(dev, l2_tunnel);
8105                 break;
8106         default:
8107                 PMD_DRV_LOG(ERR, "Invalid tunnel type");
8108                 ret = -EINVAL;
8109                 break;
8110         }
8111
8112         if ((!restore) && (ret < 0))
8113                 (void)ixgbe_remove_l2_tn_filter(l2_tn_info, &key);
8114
8115         return ret;
8116 }
8117
8118 /* Delete l2 tunnel filter */
8119 int
8120 ixgbe_dev_l2_tunnel_filter_del(struct rte_eth_dev *dev,
8121                                struct rte_eth_l2_tunnel_conf *l2_tunnel)
8122 {
8123         int ret;
8124         struct ixgbe_l2_tn_info *l2_tn_info =
8125                 IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(dev->data->dev_private);
8126         struct ixgbe_l2_tn_key key;
8127
8128         key.l2_tn_type = l2_tunnel->l2_tunnel_type;
8129         key.tn_id = l2_tunnel->tunnel_id;
8130         ret = ixgbe_remove_l2_tn_filter(l2_tn_info, &key);
8131         if (ret < 0)
8132                 return ret;
8133
8134         switch (l2_tunnel->l2_tunnel_type) {
8135         case RTE_L2_TUNNEL_TYPE_E_TAG:
8136                 ret = ixgbe_e_tag_filter_del(dev, l2_tunnel);
8137                 break;
8138         default:
8139                 PMD_DRV_LOG(ERR, "Invalid tunnel type");
8140                 ret = -EINVAL;
8141                 break;
8142         }
8143
8144         return ret;
8145 }
8146
8147 /**
8148  * ixgbe_dev_l2_tunnel_filter_handle - Handle operations for l2 tunnel filter.
8149  * @dev: pointer to rte_eth_dev structure
8150  * @filter_op:operation will be taken.
8151  * @arg: a pointer to specific structure corresponding to the filter_op
8152  */
8153 static int
8154 ixgbe_dev_l2_tunnel_filter_handle(struct rte_eth_dev *dev,
8155                                   enum rte_filter_op filter_op,
8156                                   void *arg)
8157 {
8158         int ret;
8159
8160         if (filter_op == RTE_ETH_FILTER_NOP)
8161                 return 0;
8162
8163         if (arg == NULL) {
8164                 PMD_DRV_LOG(ERR, "arg shouldn't be NULL for operation %u.",
8165                             filter_op);
8166                 return -EINVAL;
8167         }
8168
8169         switch (filter_op) {
8170         case RTE_ETH_FILTER_ADD:
8171                 ret = ixgbe_dev_l2_tunnel_filter_add
8172                         (dev,
8173                          (struct rte_eth_l2_tunnel_conf *)arg,
8174                          FALSE);
8175                 break;
8176         case RTE_ETH_FILTER_DELETE:
8177                 ret = ixgbe_dev_l2_tunnel_filter_del
8178                         (dev,
8179                          (struct rte_eth_l2_tunnel_conf *)arg);
8180                 break;
8181         default:
8182                 PMD_DRV_LOG(ERR, "unsupported operation %u.", filter_op);
8183                 ret = -EINVAL;
8184                 break;
8185         }
8186         return ret;
8187 }
8188
8189 static int
8190 ixgbe_e_tag_forwarding_en_dis(struct rte_eth_dev *dev, bool en)
8191 {
8192         int ret = 0;
8193         uint32_t ctrl;
8194         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8195
8196         if (hw->mac.type != ixgbe_mac_X550 &&
8197             hw->mac.type != ixgbe_mac_X550EM_x &&
8198             hw->mac.type != ixgbe_mac_X550EM_a) {
8199                 return -ENOTSUP;
8200         }
8201
8202         ctrl = IXGBE_READ_REG(hw, IXGBE_VT_CTL);
8203         ctrl &= ~IXGBE_VT_CTL_POOLING_MODE_MASK;
8204         if (en)
8205                 ctrl |= IXGBE_VT_CTL_POOLING_MODE_ETAG;
8206         IXGBE_WRITE_REG(hw, IXGBE_VT_CTL, ctrl);
8207
8208         return ret;
8209 }
8210
8211 /* Enable l2 tunnel forwarding */
8212 static int
8213 ixgbe_dev_l2_tunnel_forwarding_enable
8214         (struct rte_eth_dev *dev,
8215          enum rte_eth_tunnel_type l2_tunnel_type)
8216 {
8217         struct ixgbe_l2_tn_info *l2_tn_info =
8218                 IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(dev->data->dev_private);
8219         int ret = 0;
8220
8221         switch (l2_tunnel_type) {
8222         case RTE_L2_TUNNEL_TYPE_E_TAG:
8223                 l2_tn_info->e_tag_fwd_en = TRUE;
8224                 ret = ixgbe_e_tag_forwarding_en_dis(dev, 1);
8225                 break;
8226         default:
8227                 PMD_DRV_LOG(ERR, "Invalid tunnel type");
8228                 ret = -EINVAL;
8229                 break;
8230         }
8231
8232         return ret;
8233 }
8234
8235 /* Disable l2 tunnel forwarding */
8236 static int
8237 ixgbe_dev_l2_tunnel_forwarding_disable
8238         (struct rte_eth_dev *dev,
8239          enum rte_eth_tunnel_type l2_tunnel_type)
8240 {
8241         struct ixgbe_l2_tn_info *l2_tn_info =
8242                 IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(dev->data->dev_private);
8243         int ret = 0;
8244
8245         switch (l2_tunnel_type) {
8246         case RTE_L2_TUNNEL_TYPE_E_TAG:
8247                 l2_tn_info->e_tag_fwd_en = FALSE;
8248                 ret = ixgbe_e_tag_forwarding_en_dis(dev, 0);
8249                 break;
8250         default:
8251                 PMD_DRV_LOG(ERR, "Invalid tunnel type");
8252                 ret = -EINVAL;
8253                 break;
8254         }
8255
8256         return ret;
8257 }
8258
8259 static int
8260 ixgbe_e_tag_insertion_en_dis(struct rte_eth_dev *dev,
8261                              struct rte_eth_l2_tunnel_conf *l2_tunnel,
8262                              bool en)
8263 {
8264         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
8265         int ret = 0;
8266         uint32_t vmtir, vmvir;
8267         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8268
8269         if (l2_tunnel->vf_id >= pci_dev->max_vfs) {
8270                 PMD_DRV_LOG(ERR,
8271                             "VF id %u should be less than %u",
8272                             l2_tunnel->vf_id,
8273                             pci_dev->max_vfs);
8274                 return -EINVAL;
8275         }
8276
8277         if (hw->mac.type != ixgbe_mac_X550 &&
8278             hw->mac.type != ixgbe_mac_X550EM_x &&
8279             hw->mac.type != ixgbe_mac_X550EM_a) {
8280                 return -ENOTSUP;
8281         }
8282
8283         if (en)
8284                 vmtir = l2_tunnel->tunnel_id;
8285         else
8286                 vmtir = 0;
8287
8288         IXGBE_WRITE_REG(hw, IXGBE_VMTIR(l2_tunnel->vf_id), vmtir);
8289
8290         vmvir = IXGBE_READ_REG(hw, IXGBE_VMVIR(l2_tunnel->vf_id));
8291         vmvir &= ~IXGBE_VMVIR_TAGA_MASK;
8292         if (en)
8293                 vmvir |= IXGBE_VMVIR_TAGA_ETAG_INSERT;
8294         IXGBE_WRITE_REG(hw, IXGBE_VMVIR(l2_tunnel->vf_id), vmvir);
8295
8296         return ret;
8297 }
8298
8299 /* Enable l2 tunnel tag insertion */
8300 static int
8301 ixgbe_dev_l2_tunnel_insertion_enable(struct rte_eth_dev *dev,
8302                                      struct rte_eth_l2_tunnel_conf *l2_tunnel)
8303 {
8304         int ret = 0;
8305
8306         switch (l2_tunnel->l2_tunnel_type) {
8307         case RTE_L2_TUNNEL_TYPE_E_TAG:
8308                 ret = ixgbe_e_tag_insertion_en_dis(dev, l2_tunnel, 1);
8309                 break;
8310         default:
8311                 PMD_DRV_LOG(ERR, "Invalid tunnel type");
8312                 ret = -EINVAL;
8313                 break;
8314         }
8315
8316         return ret;
8317 }
8318
8319 /* Disable l2 tunnel tag insertion */
8320 static int
8321 ixgbe_dev_l2_tunnel_insertion_disable
8322         (struct rte_eth_dev *dev,
8323          struct rte_eth_l2_tunnel_conf *l2_tunnel)
8324 {
8325         int ret = 0;
8326
8327         switch (l2_tunnel->l2_tunnel_type) {
8328         case RTE_L2_TUNNEL_TYPE_E_TAG:
8329                 ret = ixgbe_e_tag_insertion_en_dis(dev, l2_tunnel, 0);
8330                 break;
8331         default:
8332                 PMD_DRV_LOG(ERR, "Invalid tunnel type");
8333                 ret = -EINVAL;
8334                 break;
8335         }
8336
8337         return ret;
8338 }
8339
8340 static int
8341 ixgbe_e_tag_stripping_en_dis(struct rte_eth_dev *dev,
8342                              bool en)
8343 {
8344         int ret = 0;
8345         uint32_t qde;
8346         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8347
8348         if (hw->mac.type != ixgbe_mac_X550 &&
8349             hw->mac.type != ixgbe_mac_X550EM_x &&
8350             hw->mac.type != ixgbe_mac_X550EM_a) {
8351                 return -ENOTSUP;
8352         }
8353
8354         qde = IXGBE_READ_REG(hw, IXGBE_QDE);
8355         if (en)
8356                 qde |= IXGBE_QDE_STRIP_TAG;
8357         else
8358                 qde &= ~IXGBE_QDE_STRIP_TAG;
8359         qde &= ~IXGBE_QDE_READ;
8360         qde |= IXGBE_QDE_WRITE;
8361         IXGBE_WRITE_REG(hw, IXGBE_QDE, qde);
8362
8363         return ret;
8364 }
8365
8366 /* Enable l2 tunnel tag stripping */
8367 static int
8368 ixgbe_dev_l2_tunnel_stripping_enable
8369         (struct rte_eth_dev *dev,
8370          enum rte_eth_tunnel_type l2_tunnel_type)
8371 {
8372         int ret = 0;
8373
8374         switch (l2_tunnel_type) {
8375         case RTE_L2_TUNNEL_TYPE_E_TAG:
8376                 ret = ixgbe_e_tag_stripping_en_dis(dev, 1);
8377                 break;
8378         default:
8379                 PMD_DRV_LOG(ERR, "Invalid tunnel type");
8380                 ret = -EINVAL;
8381                 break;
8382         }
8383
8384         return ret;
8385 }
8386
8387 /* Disable l2 tunnel tag stripping */
8388 static int
8389 ixgbe_dev_l2_tunnel_stripping_disable
8390         (struct rte_eth_dev *dev,
8391          enum rte_eth_tunnel_type l2_tunnel_type)
8392 {
8393         int ret = 0;
8394
8395         switch (l2_tunnel_type) {
8396         case RTE_L2_TUNNEL_TYPE_E_TAG:
8397                 ret = ixgbe_e_tag_stripping_en_dis(dev, 0);
8398                 break;
8399         default:
8400                 PMD_DRV_LOG(ERR, "Invalid tunnel type");
8401                 ret = -EINVAL;
8402                 break;
8403         }
8404
8405         return ret;
8406 }
8407
8408 /* Enable/disable l2 tunnel offload functions */
8409 static int
8410 ixgbe_dev_l2_tunnel_offload_set
8411         (struct rte_eth_dev *dev,
8412          struct rte_eth_l2_tunnel_conf *l2_tunnel,
8413          uint32_t mask,
8414          uint8_t en)
8415 {
8416         int ret = 0;
8417
8418         if (l2_tunnel == NULL)
8419                 return -EINVAL;
8420
8421         ret = -EINVAL;
8422         if (mask & ETH_L2_TUNNEL_ENABLE_MASK) {
8423                 if (en)
8424                         ret = ixgbe_dev_l2_tunnel_enable(
8425                                 dev,
8426                                 l2_tunnel->l2_tunnel_type);
8427                 else
8428                         ret = ixgbe_dev_l2_tunnel_disable(
8429                                 dev,
8430                                 l2_tunnel->l2_tunnel_type);
8431         }
8432
8433         if (mask & ETH_L2_TUNNEL_INSERTION_MASK) {
8434                 if (en)
8435                         ret = ixgbe_dev_l2_tunnel_insertion_enable(
8436                                 dev,
8437                                 l2_tunnel);
8438                 else
8439                         ret = ixgbe_dev_l2_tunnel_insertion_disable(
8440                                 dev,
8441                                 l2_tunnel);
8442         }
8443
8444         if (mask & ETH_L2_TUNNEL_STRIPPING_MASK) {
8445                 if (en)
8446                         ret = ixgbe_dev_l2_tunnel_stripping_enable(
8447                                 dev,
8448                                 l2_tunnel->l2_tunnel_type);
8449                 else
8450                         ret = ixgbe_dev_l2_tunnel_stripping_disable(
8451                                 dev,
8452                                 l2_tunnel->l2_tunnel_type);
8453         }
8454
8455         if (mask & ETH_L2_TUNNEL_FORWARDING_MASK) {
8456                 if (en)
8457                         ret = ixgbe_dev_l2_tunnel_forwarding_enable(
8458                                 dev,
8459                                 l2_tunnel->l2_tunnel_type);
8460                 else
8461                         ret = ixgbe_dev_l2_tunnel_forwarding_disable(
8462                                 dev,
8463                                 l2_tunnel->l2_tunnel_type);
8464         }
8465
8466         return ret;
8467 }
8468
8469 static int
8470 ixgbe_update_vxlan_port(struct ixgbe_hw *hw,
8471                         uint16_t port)
8472 {
8473         IXGBE_WRITE_REG(hw, IXGBE_VXLANCTRL, port);
8474         IXGBE_WRITE_FLUSH(hw);
8475
8476         return 0;
8477 }
8478
8479 /* There's only one register for VxLAN UDP port.
8480  * So, we cannot add several ports. Will update it.
8481  */
8482 static int
8483 ixgbe_add_vxlan_port(struct ixgbe_hw *hw,
8484                      uint16_t port)
8485 {
8486         if (port == 0) {
8487                 PMD_DRV_LOG(ERR, "Add VxLAN port 0 is not allowed.");
8488                 return -EINVAL;
8489         }
8490
8491         return ixgbe_update_vxlan_port(hw, port);
8492 }
8493
8494 /* We cannot delete the VxLAN port. For there's a register for VxLAN
8495  * UDP port, it must have a value.
8496  * So, will reset it to the original value 0.
8497  */
8498 static int
8499 ixgbe_del_vxlan_port(struct ixgbe_hw *hw,
8500                      uint16_t port)
8501 {
8502         uint16_t cur_port;
8503
8504         cur_port = (uint16_t)IXGBE_READ_REG(hw, IXGBE_VXLANCTRL);
8505
8506         if (cur_port != port) {
8507                 PMD_DRV_LOG(ERR, "Port %u does not exist.", port);
8508                 return -EINVAL;
8509         }
8510
8511         return ixgbe_update_vxlan_port(hw, 0);
8512 }
8513
8514 /* Add UDP tunneling port */
8515 static int
8516 ixgbe_dev_udp_tunnel_port_add(struct rte_eth_dev *dev,
8517                               struct rte_eth_udp_tunnel *udp_tunnel)
8518 {
8519         int ret = 0;
8520         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8521
8522         if (hw->mac.type != ixgbe_mac_X550 &&
8523             hw->mac.type != ixgbe_mac_X550EM_x &&
8524             hw->mac.type != ixgbe_mac_X550EM_a) {
8525                 return -ENOTSUP;
8526         }
8527
8528         if (udp_tunnel == NULL)
8529                 return -EINVAL;
8530
8531         switch (udp_tunnel->prot_type) {
8532         case RTE_TUNNEL_TYPE_VXLAN:
8533                 ret = ixgbe_add_vxlan_port(hw, udp_tunnel->udp_port);
8534                 break;
8535
8536         case RTE_TUNNEL_TYPE_GENEVE:
8537         case RTE_TUNNEL_TYPE_TEREDO:
8538                 PMD_DRV_LOG(ERR, "Tunnel type is not supported now.");
8539                 ret = -EINVAL;
8540                 break;
8541
8542         default:
8543                 PMD_DRV_LOG(ERR, "Invalid tunnel type");
8544                 ret = -EINVAL;
8545                 break;
8546         }
8547
8548         return ret;
8549 }
8550
8551 /* Remove UDP tunneling port */
8552 static int
8553 ixgbe_dev_udp_tunnel_port_del(struct rte_eth_dev *dev,
8554                               struct rte_eth_udp_tunnel *udp_tunnel)
8555 {
8556         int ret = 0;
8557         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8558
8559         if (hw->mac.type != ixgbe_mac_X550 &&
8560             hw->mac.type != ixgbe_mac_X550EM_x &&
8561             hw->mac.type != ixgbe_mac_X550EM_a) {
8562                 return -ENOTSUP;
8563         }
8564
8565         if (udp_tunnel == NULL)
8566                 return -EINVAL;
8567
8568         switch (udp_tunnel->prot_type) {
8569         case RTE_TUNNEL_TYPE_VXLAN:
8570                 ret = ixgbe_del_vxlan_port(hw, udp_tunnel->udp_port);
8571                 break;
8572         case RTE_TUNNEL_TYPE_GENEVE:
8573         case RTE_TUNNEL_TYPE_TEREDO:
8574                 PMD_DRV_LOG(ERR, "Tunnel type is not supported now.");
8575                 ret = -EINVAL;
8576                 break;
8577         default:
8578                 PMD_DRV_LOG(ERR, "Invalid tunnel type");
8579                 ret = -EINVAL;
8580                 break;
8581         }
8582
8583         return ret;
8584 }
8585
8586 static int
8587 ixgbevf_dev_promiscuous_enable(struct rte_eth_dev *dev)
8588 {
8589         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8590         int ret;
8591
8592         switch (hw->mac.ops.update_xcast_mode(hw, IXGBEVF_XCAST_MODE_PROMISC)) {
8593         case IXGBE_SUCCESS:
8594                 ret = 0;
8595                 break;
8596         case IXGBE_ERR_FEATURE_NOT_SUPPORTED:
8597                 ret = -ENOTSUP;
8598                 break;
8599         default:
8600                 ret = -EAGAIN;
8601                 break;
8602         }
8603
8604         return ret;
8605 }
8606
8607 static int
8608 ixgbevf_dev_promiscuous_disable(struct rte_eth_dev *dev)
8609 {
8610         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8611         int ret;
8612
8613         switch (hw->mac.ops.update_xcast_mode(hw, IXGBEVF_XCAST_MODE_NONE)) {
8614         case IXGBE_SUCCESS:
8615                 ret = 0;
8616                 break;
8617         case IXGBE_ERR_FEATURE_NOT_SUPPORTED:
8618                 ret = -ENOTSUP;
8619                 break;
8620         default:
8621                 ret = -EAGAIN;
8622                 break;
8623         }
8624
8625         return ret;
8626 }
8627
8628 static int
8629 ixgbevf_dev_allmulticast_enable(struct rte_eth_dev *dev)
8630 {
8631         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8632         int ret;
8633         int mode = IXGBEVF_XCAST_MODE_ALLMULTI;
8634
8635         switch (hw->mac.ops.update_xcast_mode(hw, mode)) {
8636         case IXGBE_SUCCESS:
8637                 ret = 0;
8638                 break;
8639         case IXGBE_ERR_FEATURE_NOT_SUPPORTED:
8640                 ret = -ENOTSUP;
8641                 break;
8642         default:
8643                 ret = -EAGAIN;
8644                 break;
8645         }
8646
8647         return ret;
8648 }
8649
8650 static int
8651 ixgbevf_dev_allmulticast_disable(struct rte_eth_dev *dev)
8652 {
8653         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8654         int ret;
8655
8656         switch (hw->mac.ops.update_xcast_mode(hw, IXGBEVF_XCAST_MODE_MULTI)) {
8657         case IXGBE_SUCCESS:
8658                 ret = 0;
8659                 break;
8660         case IXGBE_ERR_FEATURE_NOT_SUPPORTED:
8661                 ret = -ENOTSUP;
8662                 break;
8663         default:
8664                 ret = -EAGAIN;
8665                 break;
8666         }
8667
8668         return ret;
8669 }
8670
8671 static void ixgbevf_mbx_process(struct rte_eth_dev *dev)
8672 {
8673         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8674         u32 in_msg = 0;
8675
8676         /* peek the message first */
8677         in_msg = IXGBE_READ_REG(hw, IXGBE_VFMBMEM);
8678
8679         /* PF reset VF event */
8680         if (in_msg == IXGBE_PF_CONTROL_MSG) {
8681                 /* dummy mbx read to ack pf */
8682                 if (ixgbe_read_mbx(hw, &in_msg, 1, 0))
8683                         return;
8684                 rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_RESET,
8685                                              NULL);
8686         }
8687 }
8688
8689 static int
8690 ixgbevf_dev_interrupt_get_status(struct rte_eth_dev *dev)
8691 {
8692         uint32_t eicr;
8693         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8694         struct ixgbe_interrupt *intr =
8695                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
8696         ixgbevf_intr_disable(dev);
8697
8698         /* read-on-clear nic registers here */
8699         eicr = IXGBE_READ_REG(hw, IXGBE_VTEICR);
8700         intr->flags = 0;
8701
8702         /* only one misc vector supported - mailbox */
8703         eicr &= IXGBE_VTEICR_MASK;
8704         if (eicr == IXGBE_MISC_VEC_ID)
8705                 intr->flags |= IXGBE_FLAG_MAILBOX;
8706
8707         return 0;
8708 }
8709
8710 static int
8711 ixgbevf_dev_interrupt_action(struct rte_eth_dev *dev)
8712 {
8713         struct ixgbe_interrupt *intr =
8714                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
8715
8716         if (intr->flags & IXGBE_FLAG_MAILBOX) {
8717                 ixgbevf_mbx_process(dev);
8718                 intr->flags &= ~IXGBE_FLAG_MAILBOX;
8719         }
8720
8721         ixgbevf_intr_enable(dev);
8722
8723         return 0;
8724 }
8725
8726 static void
8727 ixgbevf_dev_interrupt_handler(void *param)
8728 {
8729         struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
8730
8731         ixgbevf_dev_interrupt_get_status(dev);
8732         ixgbevf_dev_interrupt_action(dev);
8733 }
8734
8735 /**
8736  *  ixgbe_disable_sec_tx_path_generic - Stops the transmit data path
8737  *  @hw: pointer to hardware structure
8738  *
8739  *  Stops the transmit data path and waits for the HW to internally empty
8740  *  the Tx security block
8741  **/
8742 int ixgbe_disable_sec_tx_path_generic(struct ixgbe_hw *hw)
8743 {
8744 #define IXGBE_MAX_SECTX_POLL 40
8745
8746         int i;
8747         int sectxreg;
8748
8749         sectxreg = IXGBE_READ_REG(hw, IXGBE_SECTXCTRL);
8750         sectxreg |= IXGBE_SECTXCTRL_TX_DIS;
8751         IXGBE_WRITE_REG(hw, IXGBE_SECTXCTRL, sectxreg);
8752         for (i = 0; i < IXGBE_MAX_SECTX_POLL; i++) {
8753                 sectxreg = IXGBE_READ_REG(hw, IXGBE_SECTXSTAT);
8754                 if (sectxreg & IXGBE_SECTXSTAT_SECTX_RDY)
8755                         break;
8756                 /* Use interrupt-safe sleep just in case */
8757                 usec_delay(1000);
8758         }
8759
8760         /* For informational purposes only */
8761         if (i >= IXGBE_MAX_SECTX_POLL)
8762                 PMD_DRV_LOG(DEBUG, "Tx unit being enabled before security "
8763                          "path fully disabled.  Continuing with init.");
8764
8765         return IXGBE_SUCCESS;
8766 }
8767
8768 /**
8769  *  ixgbe_enable_sec_tx_path_generic - Enables the transmit data path
8770  *  @hw: pointer to hardware structure
8771  *
8772  *  Enables the transmit data path.
8773  **/
8774 int ixgbe_enable_sec_tx_path_generic(struct ixgbe_hw *hw)
8775 {
8776         uint32_t sectxreg;
8777
8778         sectxreg = IXGBE_READ_REG(hw, IXGBE_SECTXCTRL);
8779         sectxreg &= ~IXGBE_SECTXCTRL_TX_DIS;
8780         IXGBE_WRITE_REG(hw, IXGBE_SECTXCTRL, sectxreg);
8781         IXGBE_WRITE_FLUSH(hw);
8782
8783         return IXGBE_SUCCESS;
8784 }
8785
8786 /* restore n-tuple filter */
8787 static inline void
8788 ixgbe_ntuple_filter_restore(struct rte_eth_dev *dev)
8789 {
8790         struct ixgbe_filter_info *filter_info =
8791                 IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
8792         struct ixgbe_5tuple_filter *node;
8793
8794         TAILQ_FOREACH(node, &filter_info->fivetuple_list, entries) {
8795                 ixgbe_inject_5tuple_filter(dev, node);
8796         }
8797 }
8798
8799 /* restore ethernet type filter */
8800 static inline void
8801 ixgbe_ethertype_filter_restore(struct rte_eth_dev *dev)
8802 {
8803         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8804         struct ixgbe_filter_info *filter_info =
8805                 IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
8806         int i;
8807
8808         for (i = 0; i < IXGBE_MAX_ETQF_FILTERS; i++) {
8809                 if (filter_info->ethertype_mask & (1 << i)) {
8810                         IXGBE_WRITE_REG(hw, IXGBE_ETQF(i),
8811                                         filter_info->ethertype_filters[i].etqf);
8812                         IXGBE_WRITE_REG(hw, IXGBE_ETQS(i),
8813                                         filter_info->ethertype_filters[i].etqs);
8814                         IXGBE_WRITE_FLUSH(hw);
8815                 }
8816         }
8817 }
8818
8819 /* restore SYN filter */
8820 static inline void
8821 ixgbe_syn_filter_restore(struct rte_eth_dev *dev)
8822 {
8823         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8824         struct ixgbe_filter_info *filter_info =
8825                 IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
8826         uint32_t synqf;
8827
8828         synqf = filter_info->syn_info;
8829
8830         if (synqf & IXGBE_SYN_FILTER_ENABLE) {
8831                 IXGBE_WRITE_REG(hw, IXGBE_SYNQF, synqf);
8832                 IXGBE_WRITE_FLUSH(hw);
8833         }
8834 }
8835
8836 /* restore L2 tunnel filter */
8837 static inline void
8838 ixgbe_l2_tn_filter_restore(struct rte_eth_dev *dev)
8839 {
8840         struct ixgbe_l2_tn_info *l2_tn_info =
8841                 IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(dev->data->dev_private);
8842         struct ixgbe_l2_tn_filter *node;
8843         struct rte_eth_l2_tunnel_conf l2_tn_conf;
8844
8845         TAILQ_FOREACH(node, &l2_tn_info->l2_tn_list, entries) {
8846                 l2_tn_conf.l2_tunnel_type = node->key.l2_tn_type;
8847                 l2_tn_conf.tunnel_id      = node->key.tn_id;
8848                 l2_tn_conf.pool           = node->pool;
8849                 (void)ixgbe_dev_l2_tunnel_filter_add(dev, &l2_tn_conf, TRUE);
8850         }
8851 }
8852
8853 /* restore rss filter */
8854 static inline void
8855 ixgbe_rss_filter_restore(struct rte_eth_dev *dev)
8856 {
8857         struct ixgbe_filter_info *filter_info =
8858                 IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
8859
8860         if (filter_info->rss_info.conf.queue_num)
8861                 ixgbe_config_rss_filter(dev,
8862                         &filter_info->rss_info, TRUE);
8863 }
8864
8865 static int
8866 ixgbe_filter_restore(struct rte_eth_dev *dev)
8867 {
8868         ixgbe_ntuple_filter_restore(dev);
8869         ixgbe_ethertype_filter_restore(dev);
8870         ixgbe_syn_filter_restore(dev);
8871         ixgbe_fdir_filter_restore(dev);
8872         ixgbe_l2_tn_filter_restore(dev);
8873         ixgbe_rss_filter_restore(dev);
8874
8875         return 0;
8876 }
8877
8878 static void
8879 ixgbe_l2_tunnel_conf(struct rte_eth_dev *dev)
8880 {
8881         struct ixgbe_l2_tn_info *l2_tn_info =
8882                 IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(dev->data->dev_private);
8883         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8884
8885         if (l2_tn_info->e_tag_en)
8886                 (void)ixgbe_e_tag_enable(hw);
8887
8888         if (l2_tn_info->e_tag_fwd_en)
8889                 (void)ixgbe_e_tag_forwarding_en_dis(dev, 1);
8890
8891         (void)ixgbe_update_e_tag_eth_type(hw, l2_tn_info->e_tag_ether_type);
8892 }
8893
8894 /* remove all the n-tuple filters */
8895 void
8896 ixgbe_clear_all_ntuple_filter(struct rte_eth_dev *dev)
8897 {
8898         struct ixgbe_filter_info *filter_info =
8899                 IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
8900         struct ixgbe_5tuple_filter *p_5tuple;
8901
8902         while ((p_5tuple = TAILQ_FIRST(&filter_info->fivetuple_list)))
8903                 ixgbe_remove_5tuple_filter(dev, p_5tuple);
8904 }
8905
8906 /* remove all the ether type filters */
8907 void
8908 ixgbe_clear_all_ethertype_filter(struct rte_eth_dev *dev)
8909 {
8910         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8911         struct ixgbe_filter_info *filter_info =
8912                 IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
8913         int i;
8914
8915         for (i = 0; i < IXGBE_MAX_ETQF_FILTERS; i++) {
8916                 if (filter_info->ethertype_mask & (1 << i) &&
8917                     !filter_info->ethertype_filters[i].conf) {
8918                         (void)ixgbe_ethertype_filter_remove(filter_info,
8919                                                             (uint8_t)i);
8920                         IXGBE_WRITE_REG(hw, IXGBE_ETQF(i), 0);
8921                         IXGBE_WRITE_REG(hw, IXGBE_ETQS(i), 0);
8922                         IXGBE_WRITE_FLUSH(hw);
8923                 }
8924         }
8925 }
8926
8927 /* remove the SYN filter */
8928 void
8929 ixgbe_clear_syn_filter(struct rte_eth_dev *dev)
8930 {
8931         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8932         struct ixgbe_filter_info *filter_info =
8933                 IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
8934
8935         if (filter_info->syn_info & IXGBE_SYN_FILTER_ENABLE) {
8936                 filter_info->syn_info = 0;
8937
8938                 IXGBE_WRITE_REG(hw, IXGBE_SYNQF, 0);
8939                 IXGBE_WRITE_FLUSH(hw);
8940         }
8941 }
8942
8943 /* remove all the L2 tunnel filters */
8944 int
8945 ixgbe_clear_all_l2_tn_filter(struct rte_eth_dev *dev)
8946 {
8947         struct ixgbe_l2_tn_info *l2_tn_info =
8948                 IXGBE_DEV_PRIVATE_TO_L2_TN_INFO(dev->data->dev_private);
8949         struct ixgbe_l2_tn_filter *l2_tn_filter;
8950         struct rte_eth_l2_tunnel_conf l2_tn_conf;
8951         int ret = 0;
8952
8953         while ((l2_tn_filter = TAILQ_FIRST(&l2_tn_info->l2_tn_list))) {
8954                 l2_tn_conf.l2_tunnel_type = l2_tn_filter->key.l2_tn_type;
8955                 l2_tn_conf.tunnel_id      = l2_tn_filter->key.tn_id;
8956                 l2_tn_conf.pool           = l2_tn_filter->pool;
8957                 ret = ixgbe_dev_l2_tunnel_filter_del(dev, &l2_tn_conf);
8958                 if (ret < 0)
8959                         return ret;
8960         }
8961
8962         return 0;
8963 }
8964
8965 void
8966 ixgbe_dev_macsec_setting_save(struct rte_eth_dev *dev,
8967                                 struct ixgbe_macsec_setting *macsec_setting)
8968 {
8969         struct ixgbe_macsec_setting *macsec =
8970                 IXGBE_DEV_PRIVATE_TO_MACSEC_SETTING(dev->data->dev_private);
8971
8972         macsec->offload_en = macsec_setting->offload_en;
8973         macsec->encrypt_en = macsec_setting->encrypt_en;
8974         macsec->replayprotect_en = macsec_setting->replayprotect_en;
8975 }
8976
8977 void
8978 ixgbe_dev_macsec_setting_reset(struct rte_eth_dev *dev)
8979 {
8980         struct ixgbe_macsec_setting *macsec =
8981                 IXGBE_DEV_PRIVATE_TO_MACSEC_SETTING(dev->data->dev_private);
8982
8983         macsec->offload_en = 0;
8984         macsec->encrypt_en = 0;
8985         macsec->replayprotect_en = 0;
8986 }
8987
8988 void
8989 ixgbe_dev_macsec_register_enable(struct rte_eth_dev *dev,
8990                                 struct ixgbe_macsec_setting *macsec_setting)
8991 {
8992         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
8993         uint32_t ctrl;
8994         uint8_t en = macsec_setting->encrypt_en;
8995         uint8_t rp = macsec_setting->replayprotect_en;
8996
8997         /**
8998          * Workaround:
8999          * As no ixgbe_disable_sec_rx_path equivalent is
9000          * implemented for tx in the base code, and we are
9001          * not allowed to modify the base code in DPDK, so
9002          * just call the hand-written one directly for now.
9003          * The hardware support has been checked by
9004          * ixgbe_disable_sec_rx_path().
9005          */
9006         ixgbe_disable_sec_tx_path_generic(hw);
9007
9008         /* Enable Ethernet CRC (required by MACsec offload) */
9009         ctrl = IXGBE_READ_REG(hw, IXGBE_HLREG0);
9010         ctrl |= IXGBE_HLREG0_TXCRCEN | IXGBE_HLREG0_RXCRCSTRP;
9011         IXGBE_WRITE_REG(hw, IXGBE_HLREG0, ctrl);
9012
9013         /* Enable the TX and RX crypto engines */
9014         ctrl = IXGBE_READ_REG(hw, IXGBE_SECTXCTRL);
9015         ctrl &= ~IXGBE_SECTXCTRL_SECTX_DIS;
9016         IXGBE_WRITE_REG(hw, IXGBE_SECTXCTRL, ctrl);
9017
9018         ctrl = IXGBE_READ_REG(hw, IXGBE_SECRXCTRL);
9019         ctrl &= ~IXGBE_SECRXCTRL_SECRX_DIS;
9020         IXGBE_WRITE_REG(hw, IXGBE_SECRXCTRL, ctrl);
9021
9022         ctrl = IXGBE_READ_REG(hw, IXGBE_SECTXMINIFG);
9023         ctrl &= ~IXGBE_SECTX_MINSECIFG_MASK;
9024         ctrl |= 0x3;
9025         IXGBE_WRITE_REG(hw, IXGBE_SECTXMINIFG, ctrl);
9026
9027         /* Enable SA lookup */
9028         ctrl = IXGBE_READ_REG(hw, IXGBE_LSECTXCTRL);
9029         ctrl &= ~IXGBE_LSECTXCTRL_EN_MASK;
9030         ctrl |= en ? IXGBE_LSECTXCTRL_AUTH_ENCRYPT :
9031                      IXGBE_LSECTXCTRL_AUTH;
9032         ctrl |= IXGBE_LSECTXCTRL_AISCI;
9033         ctrl &= ~IXGBE_LSECTXCTRL_PNTHRSH_MASK;
9034         ctrl |= IXGBE_MACSEC_PNTHRSH & IXGBE_LSECTXCTRL_PNTHRSH_MASK;
9035         IXGBE_WRITE_REG(hw, IXGBE_LSECTXCTRL, ctrl);
9036
9037         ctrl = IXGBE_READ_REG(hw, IXGBE_LSECRXCTRL);
9038         ctrl &= ~IXGBE_LSECRXCTRL_EN_MASK;
9039         ctrl |= IXGBE_LSECRXCTRL_STRICT << IXGBE_LSECRXCTRL_EN_SHIFT;
9040         ctrl &= ~IXGBE_LSECRXCTRL_PLSH;
9041         if (rp)
9042                 ctrl |= IXGBE_LSECRXCTRL_RP;
9043         else
9044                 ctrl &= ~IXGBE_LSECRXCTRL_RP;
9045         IXGBE_WRITE_REG(hw, IXGBE_LSECRXCTRL, ctrl);
9046
9047         /* Start the data paths */
9048         ixgbe_enable_sec_rx_path(hw);
9049         /**
9050          * Workaround:
9051          * As no ixgbe_enable_sec_rx_path equivalent is
9052          * implemented for tx in the base code, and we are
9053          * not allowed to modify the base code in DPDK, so
9054          * just call the hand-written one directly for now.
9055          */
9056         ixgbe_enable_sec_tx_path_generic(hw);
9057 }
9058
9059 void
9060 ixgbe_dev_macsec_register_disable(struct rte_eth_dev *dev)
9061 {
9062         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
9063         uint32_t ctrl;
9064
9065         /**
9066          * Workaround:
9067          * As no ixgbe_disable_sec_rx_path equivalent is
9068          * implemented for tx in the base code, and we are
9069          * not allowed to modify the base code in DPDK, so
9070          * just call the hand-written one directly for now.
9071          * The hardware support has been checked by
9072          * ixgbe_disable_sec_rx_path().
9073          */
9074         ixgbe_disable_sec_tx_path_generic(hw);
9075
9076         /* Disable the TX and RX crypto engines */
9077         ctrl = IXGBE_READ_REG(hw, IXGBE_SECTXCTRL);
9078         ctrl |= IXGBE_SECTXCTRL_SECTX_DIS;
9079         IXGBE_WRITE_REG(hw, IXGBE_SECTXCTRL, ctrl);
9080
9081         ctrl = IXGBE_READ_REG(hw, IXGBE_SECRXCTRL);
9082         ctrl |= IXGBE_SECRXCTRL_SECRX_DIS;
9083         IXGBE_WRITE_REG(hw, IXGBE_SECRXCTRL, ctrl);
9084
9085         /* Disable SA lookup */
9086         ctrl = IXGBE_READ_REG(hw, IXGBE_LSECTXCTRL);
9087         ctrl &= ~IXGBE_LSECTXCTRL_EN_MASK;
9088         ctrl |= IXGBE_LSECTXCTRL_DISABLE;
9089         IXGBE_WRITE_REG(hw, IXGBE_LSECTXCTRL, ctrl);
9090
9091         ctrl = IXGBE_READ_REG(hw, IXGBE_LSECRXCTRL);
9092         ctrl &= ~IXGBE_LSECRXCTRL_EN_MASK;
9093         ctrl |= IXGBE_LSECRXCTRL_DISABLE << IXGBE_LSECRXCTRL_EN_SHIFT;
9094         IXGBE_WRITE_REG(hw, IXGBE_LSECRXCTRL, ctrl);
9095
9096         /* Start the data paths */
9097         ixgbe_enable_sec_rx_path(hw);
9098         /**
9099          * Workaround:
9100          * As no ixgbe_enable_sec_rx_path equivalent is
9101          * implemented for tx in the base code, and we are
9102          * not allowed to modify the base code in DPDK, so
9103          * just call the hand-written one directly for now.
9104          */
9105         ixgbe_enable_sec_tx_path_generic(hw);
9106 }
9107
9108 RTE_PMD_REGISTER_PCI(net_ixgbe, rte_ixgbe_pmd);
9109 RTE_PMD_REGISTER_PCI_TABLE(net_ixgbe, pci_id_ixgbe_map);
9110 RTE_PMD_REGISTER_KMOD_DEP(net_ixgbe, "* igb_uio | uio_pci_generic | vfio-pci");
9111 RTE_PMD_REGISTER_PCI(net_ixgbe_vf, rte_ixgbevf_pmd);
9112 RTE_PMD_REGISTER_PCI_TABLE(net_ixgbe_vf, pci_id_ixgbevf_map);
9113 RTE_PMD_REGISTER_KMOD_DEP(net_ixgbe_vf, "* igb_uio | vfio-pci");
9114 RTE_PMD_REGISTER_PARAM_STRING(net_ixgbe_vf,
9115                               IXGBEVF_DEVARG_PFLINK_FULLCHK "=<0|1>");
9116
9117 RTE_LOG_REGISTER(ixgbe_logtype_init, pmd.net.ixgbe.init, NOTICE);
9118 RTE_LOG_REGISTER(ixgbe_logtype_driver, pmd.net.ixgbe.driver, NOTICE);
9119
9120 #ifdef RTE_LIBRTE_IXGBE_DEBUG_RX
9121 RTE_LOG_REGISTER(ixgbe_logtype_rx, pmd.net.ixgbe.rx, DEBUG);
9122 #endif
9123 #ifdef RTE_LIBRTE_IXGBE_DEBUG_TX
9124 RTE_LOG_REGISTER(ixgbe_logtype_tx, pmd.net.ixgbe.tx, DEBUG);
9125 #endif
9126 #ifdef RTE_LIBRTE_IXGBE_DEBUG_TX_FREE
9127 RTE_LOG_REGISTER(ixgbe_logtype_tx_free, pmd.net.ixgbe.tx_free, DEBUG);
9128 #endif