drivers/net/e1000/igb_ethdev.c

   1 /* SPDX-License-Identifier: BSD-3-Clause
   2  * Copyright(c) 2010-2016 Intel Corporation
   3  */
   4
   5 #include <sys/queue.h>
   6 #include <stdio.h>
   7 #include <errno.h>
   8 #include <stdint.h>
   9 #include <stdarg.h>
  10
  11 #include <rte_string_fns.h>
  12 #include <rte_common.h>
  13 #include <rte_interrupts.h>
  14 #include <rte_byteorder.h>
  15 #include <rte_log.h>
  16 #include <rte_debug.h>
  17 #include <rte_pci.h>
  18 #include <rte_bus_pci.h>
  19 #include <rte_ether.h>
  20 #include <rte_ethdev_driver.h>
  21 #include <rte_ethdev_pci.h>
  22 #include <rte_memory.h>
  23 #include <rte_eal.h>
  24 #include <rte_malloc.h>
  25 #include <rte_dev.h>
  26
  27 #include "e1000_logs.h"
  28 #include "base/e1000_api.h"
  29 #include "e1000_ethdev.h"
  30 #include "igb_regs.h"
  31
  32 /*
  33  * Default values for port configuration
  34  */
  35 #define IGB_DEFAULT_RX_FREE_THRESH  32
  36
  37 #define IGB_DEFAULT_RX_PTHRESH      ((hw->mac.type == e1000_i354) ? 12 : 8)
  38 #define IGB_DEFAULT_RX_HTHRESH      8
  39 #define IGB_DEFAULT_RX_WTHRESH      ((hw->mac.type == e1000_82576) ? 1 : 4)
  40
  41 #define IGB_DEFAULT_TX_PTHRESH      ((hw->mac.type == e1000_i354) ? 20 : 8)
  42 #define IGB_DEFAULT_TX_HTHRESH      1
  43 #define IGB_DEFAULT_TX_WTHRESH      ((hw->mac.type == e1000_82576) ? 1 : 16)
  44
  45 /* Bit shift and mask */
  46 #define IGB_4_BIT_WIDTH  (CHAR_BIT / 2)
  47 #define IGB_4_BIT_MASK   RTE_LEN2MASK(IGB_4_BIT_WIDTH, uint8_t)
  48 #define IGB_8_BIT_WIDTH  CHAR_BIT
  49 #define IGB_8_BIT_MASK   UINT8_MAX
  50
  51 /* Additional timesync values. */
  52 #define E1000_CYCLECOUNTER_MASK      0xffffffffffffffffULL
  53 #define E1000_ETQF_FILTER_1588       3
  54 #define IGB_82576_TSYNC_SHIFT        16
  55 #define E1000_INCPERIOD_82576        (1 << E1000_TIMINCA_16NS_SHIFT)
  56 #define E1000_INCVALUE_82576         (16 << IGB_82576_TSYNC_SHIFT)
  57 #define E1000_TSAUXC_DISABLE_SYSTIME 0x80000000
  58
  59 #define E1000_VTIVAR_MISC                0x01740
  60 #define E1000_VTIVAR_MISC_MASK           0xFF
  61 #define E1000_VTIVAR_VALID               0x80
  62 #define E1000_VTIVAR_MISC_MAILBOX        0
  63 #define E1000_VTIVAR_MISC_INTR_MASK      0x3
  64
  65 /* External VLAN Enable bit mask */
  66 #define E1000_CTRL_EXT_EXT_VLAN      (1 << 26)
  67
  68 /* External VLAN Ether Type bit mask and shift */
  69 #define E1000_VET_VET_EXT            0xFFFF0000
  70 #define E1000_VET_VET_EXT_SHIFT      16
  71
  72 /* MSI-X other interrupt vector */
  73 #define IGB_MSIX_OTHER_INTR_VEC      0
  74
  75 static int  eth_igb_configure(struct rte_eth_dev *dev);
  76 static int  eth_igb_start(struct rte_eth_dev *dev);
  77 static void eth_igb_stop(struct rte_eth_dev *dev);
  78 static int  eth_igb_dev_set_link_up(struct rte_eth_dev *dev);
  79 static int  eth_igb_dev_set_link_down(struct rte_eth_dev *dev);
  80 static void eth_igb_close(struct rte_eth_dev *dev);
  81 static int eth_igb_reset(struct rte_eth_dev *dev);
  82 static int  eth_igb_promiscuous_enable(struct rte_eth_dev *dev);
  83 static int  eth_igb_promiscuous_disable(struct rte_eth_dev *dev);
  84 static int  eth_igb_allmulticast_enable(struct rte_eth_dev *dev);
  85 static int  eth_igb_allmulticast_disable(struct rte_eth_dev *dev);
  86 static int  eth_igb_link_update(struct rte_eth_dev *dev,
  87                                 int wait_to_complete);
  88 static int eth_igb_stats_get(struct rte_eth_dev *dev,
  89                                 struct rte_eth_stats *rte_stats);
  90 static int eth_igb_xstats_get(struct rte_eth_dev *dev,
  91                               struct rte_eth_xstat *xstats, unsigned n);
  92 static int eth_igb_xstats_get_by_id(struct rte_eth_dev *dev,
  93                 const uint64_t *ids,
  94                 uint64_t *values, unsigned int n);
  95 static int eth_igb_xstats_get_names(struct rte_eth_dev *dev,
  96                                     struct rte_eth_xstat_name *xstats_names,
  97                                     unsigned int size);
  98 static int eth_igb_xstats_get_names_by_id(struct rte_eth_dev *dev,
  99                 struct rte_eth_xstat_name *xstats_names, const uint64_t *ids,
 100                 unsigned int limit);
 101 static int eth_igb_stats_reset(struct rte_eth_dev *dev);
 102 static int eth_igb_xstats_reset(struct rte_eth_dev *dev);
 103 static int eth_igb_fw_version_get(struct rte_eth_dev *dev,
 104                                    char *fw_version, size_t fw_size);
 105 static int eth_igb_infos_get(struct rte_eth_dev *dev,
 106                               struct rte_eth_dev_info *dev_info);
 107 static const uint32_t *eth_igb_supported_ptypes_get(struct rte_eth_dev *dev);
 108 static int eth_igbvf_infos_get(struct rte_eth_dev *dev,
 109                                 struct rte_eth_dev_info *dev_info);
 110 static int  eth_igb_flow_ctrl_get(struct rte_eth_dev *dev,
 111                                 struct rte_eth_fc_conf *fc_conf);
 112 static int  eth_igb_flow_ctrl_set(struct rte_eth_dev *dev,
 113                                 struct rte_eth_fc_conf *fc_conf);
 114 static int eth_igb_lsc_interrupt_setup(struct rte_eth_dev *dev, uint8_t on);
 115 static int eth_igb_rxq_interrupt_setup(struct rte_eth_dev *dev);
 116 static int eth_igb_interrupt_get_status(struct rte_eth_dev *dev);
 117 static int eth_igb_interrupt_action(struct rte_eth_dev *dev,
 118                                     struct rte_intr_handle *handle);
 119 static void eth_igb_interrupt_handler(void *param);
 120 static int  igb_hardware_init(struct e1000_hw *hw);
 121 static void igb_hw_control_acquire(struct e1000_hw *hw);
 122 static void igb_hw_control_release(struct e1000_hw *hw);
 123 static void igb_init_manageability(struct e1000_hw *hw);
 124 static void igb_release_manageability(struct e1000_hw *hw);
 125
 126 static int  eth_igb_mtu_set(struct rte_eth_dev *dev, uint16_t mtu);
 127
 128 static int eth_igb_vlan_filter_set(struct rte_eth_dev *dev,
 129                 uint16_t vlan_id, int on);
 130 static int eth_igb_vlan_tpid_set(struct rte_eth_dev *dev,
 131                                  enum rte_vlan_type vlan_type,
 132                                  uint16_t tpid_id);
 133 static int eth_igb_vlan_offload_set(struct rte_eth_dev *dev, int mask);
 134
 135 static void igb_vlan_hw_filter_enable(struct rte_eth_dev *dev);
 136 static void igb_vlan_hw_filter_disable(struct rte_eth_dev *dev);
 137 static void igb_vlan_hw_strip_enable(struct rte_eth_dev *dev);
 138 static void igb_vlan_hw_strip_disable(struct rte_eth_dev *dev);
 139 static void igb_vlan_hw_extend_enable(struct rte_eth_dev *dev);
 140 static void igb_vlan_hw_extend_disable(struct rte_eth_dev *dev);
 141
 142 static int eth_igb_led_on(struct rte_eth_dev *dev);
 143 static int eth_igb_led_off(struct rte_eth_dev *dev);
 144
 145 static void igb_intr_disable(struct rte_eth_dev *dev);
 146 static int  igb_get_rx_buffer_size(struct e1000_hw *hw);
 147 static int eth_igb_rar_set(struct rte_eth_dev *dev,
 148                            struct rte_ether_addr *mac_addr,
 149                            uint32_t index, uint32_t pool);
 150 static void eth_igb_rar_clear(struct rte_eth_dev *dev, uint32_t index);
 151 static int eth_igb_default_mac_addr_set(struct rte_eth_dev *dev,
 152                 struct rte_ether_addr *addr);
 153
 154 static void igbvf_intr_disable(struct e1000_hw *hw);
 155 static int igbvf_dev_configure(struct rte_eth_dev *dev);
 156 static int igbvf_dev_start(struct rte_eth_dev *dev);
 157 static void igbvf_dev_stop(struct rte_eth_dev *dev);
 158 static void igbvf_dev_close(struct rte_eth_dev *dev);
 159 static int igbvf_promiscuous_enable(struct rte_eth_dev *dev);
 160 static int igbvf_promiscuous_disable(struct rte_eth_dev *dev);
 161 static int igbvf_allmulticast_enable(struct rte_eth_dev *dev);
 162 static int igbvf_allmulticast_disable(struct rte_eth_dev *dev);
 163 static int eth_igbvf_link_update(struct e1000_hw *hw);
 164 static int eth_igbvf_stats_get(struct rte_eth_dev *dev,
 165                                 struct rte_eth_stats *rte_stats);
 166 static int eth_igbvf_xstats_get(struct rte_eth_dev *dev,
 167                                 struct rte_eth_xstat *xstats, unsigned n);
 168 static int eth_igbvf_xstats_get_names(struct rte_eth_dev *dev,
 169                                       struct rte_eth_xstat_name *xstats_names,
 170                                       unsigned limit);
 171 static int eth_igbvf_stats_reset(struct rte_eth_dev *dev);
 172 static int igbvf_vlan_filter_set(struct rte_eth_dev *dev,
 173                 uint16_t vlan_id, int on);
 174 static int igbvf_set_vfta(struct e1000_hw *hw, uint16_t vid, bool on);
 175 static void igbvf_set_vfta_all(struct rte_eth_dev *dev, bool on);
 176 static int igbvf_default_mac_addr_set(struct rte_eth_dev *dev,
 177                 struct rte_ether_addr *addr);
 178 static int igbvf_get_reg_length(struct rte_eth_dev *dev);
 179 static int igbvf_get_regs(struct rte_eth_dev *dev,
 180                 struct rte_dev_reg_info *regs);
 181
 182 static int eth_igb_rss_reta_update(struct rte_eth_dev *dev,
 183                                    struct rte_eth_rss_reta_entry64 *reta_conf,
 184                                    uint16_t reta_size);
 185 static int eth_igb_rss_reta_query(struct rte_eth_dev *dev,
 186                                   struct rte_eth_rss_reta_entry64 *reta_conf,
 187                                   uint16_t reta_size);
 188
 189 static int eth_igb_syn_filter_get(struct rte_eth_dev *dev,
 190                         struct rte_eth_syn_filter *filter);
 191 static int eth_igb_syn_filter_handle(struct rte_eth_dev *dev,
 192                         enum rte_filter_op filter_op,
 193                         void *arg);
 194 static int igb_add_2tuple_filter(struct rte_eth_dev *dev,
 195                         struct rte_eth_ntuple_filter *ntuple_filter);
 196 static int igb_remove_2tuple_filter(struct rte_eth_dev *dev,
 197                         struct rte_eth_ntuple_filter *ntuple_filter);
 198 static int eth_igb_get_flex_filter(struct rte_eth_dev *dev,
 199                         struct rte_eth_flex_filter *filter);
 200 static int eth_igb_flex_filter_handle(struct rte_eth_dev *dev,
 201                         enum rte_filter_op filter_op,
 202                         void *arg);
 203 static int igb_add_5tuple_filter_82576(struct rte_eth_dev *dev,
 204                         struct rte_eth_ntuple_filter *ntuple_filter);
 205 static int igb_remove_5tuple_filter_82576(struct rte_eth_dev *dev,
 206                         struct rte_eth_ntuple_filter *ntuple_filter);
 207 static int igb_get_ntuple_filter(struct rte_eth_dev *dev,
 208                         struct rte_eth_ntuple_filter *filter);
 209 static int igb_ntuple_filter_handle(struct rte_eth_dev *dev,
 210                                 enum rte_filter_op filter_op,
 211                                 void *arg);
 212 static int igb_ethertype_filter_handle(struct rte_eth_dev *dev,
 213                                 enum rte_filter_op filter_op,
 214                                 void *arg);
 215 static int igb_get_ethertype_filter(struct rte_eth_dev *dev,
 216                         struct rte_eth_ethertype_filter *filter);
 217 static int eth_igb_filter_ctrl(struct rte_eth_dev *dev,
 218                      enum rte_filter_type filter_type,
 219                      enum rte_filter_op filter_op,
 220                      void *arg);
 221 static int eth_igb_get_reg_length(struct rte_eth_dev *dev);
 222 static int eth_igb_get_regs(struct rte_eth_dev *dev,
 223                 struct rte_dev_reg_info *regs);
 224 static int eth_igb_get_eeprom_length(struct rte_eth_dev *dev);
 225 static int eth_igb_get_eeprom(struct rte_eth_dev *dev,
 226                 struct rte_dev_eeprom_info *eeprom);
 227 static int eth_igb_set_eeprom(struct rte_eth_dev *dev,
 228                 struct rte_dev_eeprom_info *eeprom);
 229 static int eth_igb_get_module_info(struct rte_eth_dev *dev,
 230                                    struct rte_eth_dev_module_info *modinfo);
 231 static int eth_igb_get_module_eeprom(struct rte_eth_dev *dev,
 232                                      struct rte_dev_eeprom_info *info);
 233 static int eth_igb_set_mc_addr_list(struct rte_eth_dev *dev,
 234                                     struct rte_ether_addr *mc_addr_set,
 235                                     uint32_t nb_mc_addr);
 236 static int igb_timesync_enable(struct rte_eth_dev *dev);
 237 static int igb_timesync_disable(struct rte_eth_dev *dev);
 238 static int igb_timesync_read_rx_timestamp(struct rte_eth_dev *dev,
 239                                           struct timespec *timestamp,
 240                                           uint32_t flags);
 241 static int igb_timesync_read_tx_timestamp(struct rte_eth_dev *dev,
 242                                           struct timespec *timestamp);
 243 static int igb_timesync_adjust_time(struct rte_eth_dev *dev, int64_t delta);
 244 static int igb_timesync_read_time(struct rte_eth_dev *dev,
 245                                   struct timespec *timestamp);
 246 static int igb_timesync_write_time(struct rte_eth_dev *dev,
 247                                    const struct timespec *timestamp);
 248 static int eth_igb_rx_queue_intr_enable(struct rte_eth_dev *dev,
 249                                         uint16_t queue_id);
 250 static int eth_igb_rx_queue_intr_disable(struct rte_eth_dev *dev,
 251                                          uint16_t queue_id);
 252 static void eth_igb_assign_msix_vector(struct e1000_hw *hw, int8_t direction,
 253                                        uint8_t queue, uint8_t msix_vector);
 254 static void eth_igb_write_ivar(struct e1000_hw *hw, uint8_t msix_vector,
 255                                uint8_t index, uint8_t offset);
 256 static void eth_igb_configure_msix_intr(struct rte_eth_dev *dev);
 257 static void eth_igbvf_interrupt_handler(void *param);
 258 static void igbvf_mbx_process(struct rte_eth_dev *dev);
 259 static int igb_filter_restore(struct rte_eth_dev *dev);
 260
 261 /*
 262  * Define VF Stats MACRO for Non "cleared on read" register
 263  */
 264 #define UPDATE_VF_STAT(reg, last, cur)            \
 265 {                                                 \
 266         u32 latest = E1000_READ_REG(hw, reg);     \
 267         cur += (latest - last) & UINT_MAX;        \
 268         last = latest;                            \
 269 }
 270
 271 #define IGB_FC_PAUSE_TIME 0x0680
 272 #define IGB_LINK_UPDATE_CHECK_TIMEOUT  90  /* 9s */
 273 #define IGB_LINK_UPDATE_CHECK_INTERVAL 100 /* ms */
 274
 275 #define IGBVF_PMD_NAME "rte_igbvf_pmd"     /* PMD name */
 276
 277 static enum e1000_fc_mode igb_fc_setting = e1000_fc_full;
 278
 279 /*
 280  * The set of PCI devices this driver supports
 281  */
 282 static const struct rte_pci_id pci_id_igb_map[] = {
 283         { RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82576) },
 284         { RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82576_FIBER) },
 285         { RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82576_SERDES) },
 286         { RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82576_QUAD_COPPER) },
 287         { RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82576_QUAD_COPPER_ET2) },
 288         { RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82576_NS) },
 289         { RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82576_NS_SERDES) },
 290         { RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82576_SERDES_QUAD) },
 291
 292         { RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82575EB_COPPER) },
 293         { RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82575EB_FIBER_SERDES) },
 294         { RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82575GB_QUAD_COPPER) },
 295
 296         { RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82580_COPPER) },
 297         { RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82580_FIBER) },
 298         { RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82580_SERDES) },
 299         { RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82580_SGMII) },
 300         { RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82580_COPPER_DUAL) },
 301         { RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82580_QUAD_FIBER) },
 302
 303         { RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I350_COPPER) },
 304         { RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I350_FIBER) },
 305         { RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I350_SERDES) },
 306         { RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I350_SGMII) },
 307         { RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I350_DA4) },
 308         { RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I210_COPPER) },
 309         { RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I210_COPPER_OEM1) },
 310         { RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I210_COPPER_IT) },
 311         { RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I210_FIBER) },
 312         { RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I210_SERDES) },
 313         { RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I210_SGMII) },
 314         { RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I210_COPPER_FLASHLESS) },
 315         { RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I210_SERDES_FLASHLESS) },
 316         { RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I211_COPPER) },
 317         { RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I354_BACKPLANE_1GBPS) },
 318         { RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I354_SGMII) },
 319         { RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I354_BACKPLANE_2_5GBPS) },
 320         { RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_DH89XXCC_SGMII) },
 321         { RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_DH89XXCC_SERDES) },
 322         { RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_DH89XXCC_BACKPLANE) },
 323         { RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_DH89XXCC_SFP) },
 324         { .vendor_id = 0, /* sentinel */ },
 325 };
 326
 327 /*
 328  * The set of PCI devices this driver supports (for 82576&I350 VF)
 329  */
 330 static const struct rte_pci_id pci_id_igbvf_map[] = {
 331         { RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82576_VF) },
 332         { RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82576_VF_HV) },
 333         { RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I350_VF) },
 334         { RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_I350_VF_HV) },
 335         { .vendor_id = 0, /* sentinel */ },
 336 };
 337
 338 static const struct rte_eth_desc_lim rx_desc_lim = {
 339         .nb_max = E1000_MAX_RING_DESC,
 340         .nb_min = E1000_MIN_RING_DESC,
 341         .nb_align = IGB_RXD_ALIGN,
 342 };
 343
 344 static const struct rte_eth_desc_lim tx_desc_lim = {
 345         .nb_max = E1000_MAX_RING_DESC,
 346         .nb_min = E1000_MIN_RING_DESC,
 347         .nb_align = IGB_RXD_ALIGN,
 348         .nb_seg_max = IGB_TX_MAX_SEG,
 349         .nb_mtu_seg_max = IGB_TX_MAX_MTU_SEG,
 350 };
 351
 352 static const struct eth_dev_ops eth_igb_ops = {
 353         .dev_configure        = eth_igb_configure,
 354         .dev_start            = eth_igb_start,
 355         .dev_stop             = eth_igb_stop,
 356         .dev_set_link_up      = eth_igb_dev_set_link_up,
 357         .dev_set_link_down    = eth_igb_dev_set_link_down,
 358         .dev_close            = eth_igb_close,
 359         .dev_reset            = eth_igb_reset,
 360         .promiscuous_enable   = eth_igb_promiscuous_enable,
 361         .promiscuous_disable  = eth_igb_promiscuous_disable,
 362         .allmulticast_enable  = eth_igb_allmulticast_enable,
 363         .allmulticast_disable = eth_igb_allmulticast_disable,
 364         .link_update          = eth_igb_link_update,
 365         .stats_get            = eth_igb_stats_get,
 366         .xstats_get           = eth_igb_xstats_get,
 367         .xstats_get_by_id     = eth_igb_xstats_get_by_id,
 368         .xstats_get_names_by_id = eth_igb_xstats_get_names_by_id,
 369         .xstats_get_names     = eth_igb_xstats_get_names,
 370         .stats_reset          = eth_igb_stats_reset,
 371         .xstats_reset         = eth_igb_xstats_reset,
 372         .fw_version_get       = eth_igb_fw_version_get,
 373         .dev_infos_get        = eth_igb_infos_get,
 374         .dev_supported_ptypes_get = eth_igb_supported_ptypes_get,
 375         .mtu_set              = eth_igb_mtu_set,
 376         .vlan_filter_set      = eth_igb_vlan_filter_set,
 377         .vlan_tpid_set        = eth_igb_vlan_tpid_set,
 378         .vlan_offload_set     = eth_igb_vlan_offload_set,
 379         .rx_queue_setup       = eth_igb_rx_queue_setup,
 380         .rx_queue_intr_enable = eth_igb_rx_queue_intr_enable,
 381         .rx_queue_intr_disable = eth_igb_rx_queue_intr_disable,
 382         .rx_queue_release     = eth_igb_rx_queue_release,
 383         .rx_queue_count       = eth_igb_rx_queue_count,
 384         .rx_descriptor_done   = eth_igb_rx_descriptor_done,
 385         .rx_descriptor_status = eth_igb_rx_descriptor_status,
 386         .tx_descriptor_status = eth_igb_tx_descriptor_status,
 387         .tx_queue_setup       = eth_igb_tx_queue_setup,
 388         .tx_queue_release     = eth_igb_tx_queue_release,
 389         .tx_done_cleanup      = eth_igb_tx_done_cleanup,
 390         .dev_led_on           = eth_igb_led_on,
 391         .dev_led_off          = eth_igb_led_off,
 392         .flow_ctrl_get        = eth_igb_flow_ctrl_get,
 393         .flow_ctrl_set        = eth_igb_flow_ctrl_set,
 394         .mac_addr_add         = eth_igb_rar_set,
 395         .mac_addr_remove      = eth_igb_rar_clear,
 396         .mac_addr_set         = eth_igb_default_mac_addr_set,
 397         .reta_update          = eth_igb_rss_reta_update,
 398         .reta_query           = eth_igb_rss_reta_query,
 399         .rss_hash_update      = eth_igb_rss_hash_update,
 400         .rss_hash_conf_get    = eth_igb_rss_hash_conf_get,
 401         .filter_ctrl          = eth_igb_filter_ctrl,
 402         .set_mc_addr_list     = eth_igb_set_mc_addr_list,
 403         .rxq_info_get         = igb_rxq_info_get,
 404         .txq_info_get         = igb_txq_info_get,
 405         .timesync_enable      = igb_timesync_enable,
 406         .timesync_disable     = igb_timesync_disable,
 407         .timesync_read_rx_timestamp = igb_timesync_read_rx_timestamp,
 408         .timesync_read_tx_timestamp = igb_timesync_read_tx_timestamp,
 409         .get_reg              = eth_igb_get_regs,
 410         .get_eeprom_length    = eth_igb_get_eeprom_length,
 411         .get_eeprom           = eth_igb_get_eeprom,
 412         .set_eeprom           = eth_igb_set_eeprom,
 413         .get_module_info      = eth_igb_get_module_info,
 414         .get_module_eeprom    = eth_igb_get_module_eeprom,
 415         .timesync_adjust_time = igb_timesync_adjust_time,
 416         .timesync_read_time   = igb_timesync_read_time,
 417         .timesync_write_time  = igb_timesync_write_time,
 418 };
 419
 420 /*
 421  * dev_ops for virtual function, bare necessities for basic vf
 422  * operation have been implemented
 423  */
 424 static const struct eth_dev_ops igbvf_eth_dev_ops = {
 425         .dev_configure        = igbvf_dev_configure,
 426         .dev_start            = igbvf_dev_start,
 427         .dev_stop             = igbvf_dev_stop,
 428         .dev_close            = igbvf_dev_close,
 429         .promiscuous_enable   = igbvf_promiscuous_enable,
 430         .promiscuous_disable  = igbvf_promiscuous_disable,
 431         .allmulticast_enable  = igbvf_allmulticast_enable,
 432         .allmulticast_disable = igbvf_allmulticast_disable,
 433         .link_update          = eth_igb_link_update,
 434         .stats_get            = eth_igbvf_stats_get,
 435         .xstats_get           = eth_igbvf_xstats_get,
 436         .xstats_get_names     = eth_igbvf_xstats_get_names,
 437         .stats_reset          = eth_igbvf_stats_reset,
 438         .xstats_reset         = eth_igbvf_stats_reset,
 439         .vlan_filter_set      = igbvf_vlan_filter_set,
 440         .dev_infos_get        = eth_igbvf_infos_get,
 441         .dev_supported_ptypes_get = eth_igb_supported_ptypes_get,
 442         .rx_queue_setup       = eth_igb_rx_queue_setup,
 443         .rx_queue_release     = eth_igb_rx_queue_release,
 444         .rx_descriptor_done   = eth_igb_rx_descriptor_done,
 445         .rx_descriptor_status = eth_igb_rx_descriptor_status,
 446         .tx_descriptor_status = eth_igb_tx_descriptor_status,
 447         .tx_queue_setup       = eth_igb_tx_queue_setup,
 448         .tx_queue_release     = eth_igb_tx_queue_release,
 449         .set_mc_addr_list     = eth_igb_set_mc_addr_list,
 450         .rxq_info_get         = igb_rxq_info_get,
 451         .txq_info_get         = igb_txq_info_get,
 452         .mac_addr_set         = igbvf_default_mac_addr_set,
 453         .get_reg              = igbvf_get_regs,
 454 };
 455
 456 /* store statistics names and its offset in stats structure */
 457 struct rte_igb_xstats_name_off {
 458         char name[RTE_ETH_XSTATS_NAME_SIZE];
 459         unsigned offset;
 460 };
 461
 462 static const struct rte_igb_xstats_name_off rte_igb_stats_strings[] = {
 463         {"rx_crc_errors", offsetof(struct e1000_hw_stats, crcerrs)},
 464         {"rx_align_errors", offsetof(struct e1000_hw_stats, algnerrc)},
 465         {"rx_symbol_errors", offsetof(struct e1000_hw_stats, symerrs)},
 466         {"rx_missed_packets", offsetof(struct e1000_hw_stats, mpc)},
 467         {"tx_single_collision_packets", offsetof(struct e1000_hw_stats, scc)},
 468         {"tx_multiple_collision_packets", offsetof(struct e1000_hw_stats, mcc)},
 469         {"tx_excessive_collision_packets", offsetof(struct e1000_hw_stats,
 470                 ecol)},
 471         {"tx_late_collisions", offsetof(struct e1000_hw_stats, latecol)},
 472         {"tx_total_collisions", offsetof(struct e1000_hw_stats, colc)},
 473         {"tx_deferred_packets", offsetof(struct e1000_hw_stats, dc)},
 474         {"tx_no_carrier_sense_packets", offsetof(struct e1000_hw_stats, tncrs)},
 475         {"rx_carrier_ext_errors", offsetof(struct e1000_hw_stats, cexterr)},
 476         {"rx_length_errors", offsetof(struct e1000_hw_stats, rlec)},
 477         {"rx_xon_packets", offsetof(struct e1000_hw_stats, xonrxc)},
 478         {"tx_xon_packets", offsetof(struct e1000_hw_stats, xontxc)},
 479         {"rx_xoff_packets", offsetof(struct e1000_hw_stats, xoffrxc)},
 480         {"tx_xoff_packets", offsetof(struct e1000_hw_stats, xofftxc)},
 481         {"rx_flow_control_unsupported_packets", offsetof(struct e1000_hw_stats,
 482                 fcruc)},
 483         {"rx_size_64_packets", offsetof(struct e1000_hw_stats, prc64)},
 484         {"rx_size_65_to_127_packets", offsetof(struct e1000_hw_stats, prc127)},
 485         {"rx_size_128_to_255_packets", offsetof(struct e1000_hw_stats, prc255)},
 486         {"rx_size_256_to_511_packets", offsetof(struct e1000_hw_stats, prc511)},
 487         {"rx_size_512_to_1023_packets", offsetof(struct e1000_hw_stats,
 488                 prc1023)},
 489         {"rx_size_1024_to_max_packets", offsetof(struct e1000_hw_stats,
 490                 prc1522)},
 491         {"rx_broadcast_packets", offsetof(struct e1000_hw_stats, bprc)},
 492         {"rx_multicast_packets", offsetof(struct e1000_hw_stats, mprc)},
 493         {"rx_undersize_errors", offsetof(struct e1000_hw_stats, ruc)},
 494         {"rx_fragment_errors", offsetof(struct e1000_hw_stats, rfc)},
 495         {"rx_oversize_errors", offsetof(struct e1000_hw_stats, roc)},
 496         {"rx_jabber_errors", offsetof(struct e1000_hw_stats, rjc)},
 497         {"rx_management_packets", offsetof(struct e1000_hw_stats, mgprc)},
 498         {"rx_management_dropped", offsetof(struct e1000_hw_stats, mgpdc)},
 499         {"tx_management_packets", offsetof(struct e1000_hw_stats, mgptc)},
 500         {"rx_total_packets", offsetof(struct e1000_hw_stats, tpr)},
 501         {"tx_total_packets", offsetof(struct e1000_hw_stats, tpt)},
 502         {"rx_total_bytes", offsetof(struct e1000_hw_stats, tor)},
 503         {"tx_total_bytes", offsetof(struct e1000_hw_stats, tot)},
 504         {"tx_size_64_packets", offsetof(struct e1000_hw_stats, ptc64)},
 505         {"tx_size_65_to_127_packets", offsetof(struct e1000_hw_stats, ptc127)},
 506         {"tx_size_128_to_255_packets", offsetof(struct e1000_hw_stats, ptc255)},
 507         {"tx_size_256_to_511_packets", offsetof(struct e1000_hw_stats, ptc511)},
 508         {"tx_size_512_to_1023_packets", offsetof(struct e1000_hw_stats,
 509                 ptc1023)},
 510         {"tx_size_1023_to_max_packets", offsetof(struct e1000_hw_stats,
 511                 ptc1522)},
 512         {"tx_multicast_packets", offsetof(struct e1000_hw_stats, mptc)},
 513         {"tx_broadcast_packets", offsetof(struct e1000_hw_stats, bptc)},
 514         {"tx_tso_packets", offsetof(struct e1000_hw_stats, tsctc)},
 515         {"tx_tso_errors", offsetof(struct e1000_hw_stats, tsctfc)},
 516         {"rx_sent_to_host_packets", offsetof(struct e1000_hw_stats, rpthc)},
 517         {"tx_sent_by_host_packets", offsetof(struct e1000_hw_stats, hgptc)},
 518         {"rx_code_violation_packets", offsetof(struct e1000_hw_stats, scvpc)},
 519
 520         {"interrupt_assert_count", offsetof(struct e1000_hw_stats, iac)},
 521 };
 522
 523 #define IGB_NB_XSTATS (sizeof(rte_igb_stats_strings) / \
 524                 sizeof(rte_igb_stats_strings[0]))
 525
 526 static const struct rte_igb_xstats_name_off rte_igbvf_stats_strings[] = {
 527         {"rx_multicast_packets", offsetof(struct e1000_vf_stats, mprc)},
 528         {"rx_good_loopback_packets", offsetof(struct e1000_vf_stats, gprlbc)},
 529         {"tx_good_loopback_packets", offsetof(struct e1000_vf_stats, gptlbc)},
 530         {"rx_good_loopback_bytes", offsetof(struct e1000_vf_stats, gorlbc)},
 531         {"tx_good_loopback_bytes", offsetof(struct e1000_vf_stats, gotlbc)},
 532 };
 533
 534 #define IGBVF_NB_XSTATS (sizeof(rte_igbvf_stats_strings) / \
 535                 sizeof(rte_igbvf_stats_strings[0]))
 536
 537
 538 static inline void
 539 igb_intr_enable(struct rte_eth_dev *dev)
 540 {
 541         struct e1000_interrupt *intr =
 542                 E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
 543         struct e1000_hw *hw =
 544                 E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
 545         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
 546         struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
 547
 548         if (rte_intr_allow_others(intr_handle) &&
 549                 dev->data->dev_conf.intr_conf.lsc != 0) {
 550                 E1000_WRITE_REG(hw, E1000_EIMS, 1 << IGB_MSIX_OTHER_INTR_VEC);
 551         }
 552
 553         E1000_WRITE_REG(hw, E1000_IMS, intr->mask);
 554         E1000_WRITE_FLUSH(hw);
 555 }
 556
 557 static void
 558 igb_intr_disable(struct rte_eth_dev *dev)
 559 {
 560         struct e1000_hw *hw =
 561                 E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
 562         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
 563         struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
 564
 565         if (rte_intr_allow_others(intr_handle) &&
 566                 dev->data->dev_conf.intr_conf.lsc != 0) {
 567                 E1000_WRITE_REG(hw, E1000_EIMC, 1 << IGB_MSIX_OTHER_INTR_VEC);
 568         }
 569
 570         E1000_WRITE_REG(hw, E1000_IMC, ~0);
 571         E1000_WRITE_FLUSH(hw);
 572 }
 573
 574 static inline void
 575 igbvf_intr_enable(struct rte_eth_dev *dev)
 576 {
 577         struct e1000_hw *hw =
 578                 E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
 579
 580         /* only for mailbox */
 581         E1000_WRITE_REG(hw, E1000_EIAM, 1 << E1000_VTIVAR_MISC_MAILBOX);
 582         E1000_WRITE_REG(hw, E1000_EIAC, 1 << E1000_VTIVAR_MISC_MAILBOX);
 583         E1000_WRITE_REG(hw, E1000_EIMS, 1 << E1000_VTIVAR_MISC_MAILBOX);
 584         E1000_WRITE_FLUSH(hw);
 585 }
 586
 587 /* only for mailbox now. If RX/TX needed, should extend this function.  */
 588 static void
 589 igbvf_set_ivar_map(struct e1000_hw *hw, uint8_t msix_vector)
 590 {
 591         uint32_t tmp = 0;
 592
 593         /* mailbox */
 594         tmp |= (msix_vector & E1000_VTIVAR_MISC_INTR_MASK);
 595         tmp |= E1000_VTIVAR_VALID;
 596         E1000_WRITE_REG(hw, E1000_VTIVAR_MISC, tmp);
 597 }
 598
 599 static void
 600 eth_igbvf_configure_msix_intr(struct rte_eth_dev *dev)
 601 {
 602         struct e1000_hw *hw =
 603                 E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
 604
 605         /* Configure VF other cause ivar */
 606         igbvf_set_ivar_map(hw, E1000_VTIVAR_MISC_MAILBOX);
 607 }
 608
 609 static inline int32_t
 610 igb_pf_reset_hw(struct e1000_hw *hw)
 611 {
 612         uint32_t ctrl_ext;
 613         int32_t status;
 614
 615         status = e1000_reset_hw(hw);
 616
 617         ctrl_ext = E1000_READ_REG(hw, E1000_CTRL_EXT);
 618         /* Set PF Reset Done bit so PF/VF Mail Ops can work */
 619         ctrl_ext |= E1000_CTRL_EXT_PFRSTD;
 620         E1000_WRITE_REG(hw, E1000_CTRL_EXT, ctrl_ext);
 621         E1000_WRITE_FLUSH(hw);
 622
 623         return status;
 624 }
 625
 626 static void
 627 igb_identify_hardware(struct rte_eth_dev *dev, struct rte_pci_device *pci_dev)
 628 {
 629         struct e1000_hw *hw =
 630                 E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
 631
 632
 633         hw->vendor_id = pci_dev->id.vendor_id;
 634         hw->device_id = pci_dev->id.device_id;
 635         hw->subsystem_vendor_id = pci_dev->id.subsystem_vendor_id;
 636         hw->subsystem_device_id = pci_dev->id.subsystem_device_id;
 637
 638         e1000_set_mac_type(hw);
 639
 640         /* need to check if it is a vf device below */
 641 }
 642
 643 static int
 644 igb_reset_swfw_lock(struct e1000_hw *hw)
 645 {
 646         int ret_val;
 647
 648         /*
 649          * Do mac ops initialization manually here, since we will need
 650          * some function pointers set by this call.
 651          */
 652         ret_val = e1000_init_mac_params(hw);
 653         if (ret_val)
 654                 return ret_val;
 655
 656         /*
 657          * SMBI lock should not fail in this early stage. If this is the case,
 658          * it is due to an improper exit of the application.
 659          * So force the release of the faulty lock.
 660          */
 661         if (e1000_get_hw_semaphore_generic(hw) < 0) {
 662                 PMD_DRV_LOG(DEBUG, "SMBI lock released");
 663         }
 664         e1000_put_hw_semaphore_generic(hw);
 665
 666         if (hw->mac.ops.acquire_swfw_sync != NULL) {
 667                 uint16_t mask;
 668
 669                 /*
 670                  * Phy lock should not fail in this early stage. If this is the case,
 671                  * it is due to an improper exit of the application.
 672                  * So force the release of the faulty lock.
 673                  */
 674                 mask = E1000_SWFW_PHY0_SM << hw->bus.func;
 675                 if (hw->bus.func > E1000_FUNC_1)
 676                         mask <<= 2;
 677                 if (hw->mac.ops.acquire_swfw_sync(hw, mask) < 0) {
 678                         PMD_DRV_LOG(DEBUG, "SWFW phy%d lock released",
 679                                     hw->bus.func);
 680                 }
 681                 hw->mac.ops.release_swfw_sync(hw, mask);
 682
 683                 /*
 684                  * This one is more tricky since it is common to all ports; but
 685                  * swfw_sync retries last long enough (1s) to be almost sure that if
 686                  * lock can not be taken it is due to an improper lock of the
 687                  * semaphore.
 688                  */
 689                 mask = E1000_SWFW_EEP_SM;
 690                 if (hw->mac.ops.acquire_swfw_sync(hw, mask) < 0) {
 691                         PMD_DRV_LOG(DEBUG, "SWFW common locks released");
 692                 }
 693                 hw->mac.ops.release_swfw_sync(hw, mask);
 694         }
 695
 696         return E1000_SUCCESS;
 697 }
 698
 699 /* Remove all ntuple filters of the device */
 700 static int igb_ntuple_filter_uninit(struct rte_eth_dev *eth_dev)
 701 {
 702         struct e1000_filter_info *filter_info =
 703                 E1000_DEV_PRIVATE_TO_FILTER_INFO(eth_dev->data->dev_private);
 704         struct e1000_5tuple_filter *p_5tuple;
 705         struct e1000_2tuple_filter *p_2tuple;
 706
 707         while ((p_5tuple = TAILQ_FIRST(&filter_info->fivetuple_list))) {
 708                 TAILQ_REMOVE(&filter_info->fivetuple_list,
 709                         p_5tuple, entries);
 710                         rte_free(p_5tuple);
 711         }
 712         filter_info->fivetuple_mask = 0;
 713         while ((p_2tuple = TAILQ_FIRST(&filter_info->twotuple_list))) {
 714                 TAILQ_REMOVE(&filter_info->twotuple_list,
 715                         p_2tuple, entries);
 716                         rte_free(p_2tuple);
 717         }
 718         filter_info->twotuple_mask = 0;
 719
 720         return 0;
 721 }
 722
 723 /* Remove all flex filters of the device */
 724 static int igb_flex_filter_uninit(struct rte_eth_dev *eth_dev)
 725 {
 726         struct e1000_filter_info *filter_info =
 727                 E1000_DEV_PRIVATE_TO_FILTER_INFO(eth_dev->data->dev_private);
 728         struct e1000_flex_filter *p_flex;
 729
 730         while ((p_flex = TAILQ_FIRST(&filter_info->flex_list))) {
 731                 TAILQ_REMOVE(&filter_info->flex_list, p_flex, entries);
 732                 rte_free(p_flex);
 733         }
 734         filter_info->flex_mask = 0;
 735
 736         return 0;
 737 }
 738
 739 static int
 740 eth_igb_dev_init(struct rte_eth_dev *eth_dev)
 741 {
 742         int error = 0;
 743         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
 744         struct e1000_hw *hw =
 745                 E1000_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
 746         struct e1000_vfta * shadow_vfta =
 747                 E1000_DEV_PRIVATE_TO_VFTA(eth_dev->data->dev_private);
 748         struct e1000_filter_info *filter_info =
 749                 E1000_DEV_PRIVATE_TO_FILTER_INFO(eth_dev->data->dev_private);
 750         struct e1000_adapter *adapter =
 751                 E1000_DEV_PRIVATE(eth_dev->data->dev_private);
 752
 753         uint32_t ctrl_ext;
 754
 755         eth_dev->dev_ops = &eth_igb_ops;
 756         eth_dev->rx_pkt_burst = &eth_igb_recv_pkts;
 757         eth_dev->tx_pkt_burst = &eth_igb_xmit_pkts;
 758         eth_dev->tx_pkt_prepare = &eth_igb_prep_pkts;
 759
 760         /* for secondary processes, we don't initialise any further as primary
 761          * has already done this work. Only check we don't need a different
 762          * RX function */
 763         if (rte_eal_process_type() != RTE_PROC_PRIMARY){
 764                 if (eth_dev->data->scattered_rx)
 765                         eth_dev->rx_pkt_burst = &eth_igb_recv_scattered_pkts;
 766                 return 0;
 767         }
 768
 769         rte_eth_copy_pci_info(eth_dev, pci_dev);
 770
 771         hw->hw_addr= (void *)pci_dev->mem_resource[0].addr;
 772
 773         igb_identify_hardware(eth_dev, pci_dev);
 774         if (e1000_setup_init_funcs(hw, FALSE) != E1000_SUCCESS) {
 775                 error = -EIO;
 776                 goto err_late;
 777         }
 778
 779         e1000_get_bus_info(hw);
 780
 781         /* Reset any pending lock */
 782         if (igb_reset_swfw_lock(hw) != E1000_SUCCESS) {
 783                 error = -EIO;
 784                 goto err_late;
 785         }
 786
 787         /* Finish initialization */
 788         if (e1000_setup_init_funcs(hw, TRUE) != E1000_SUCCESS) {
 789                 error = -EIO;
 790                 goto err_late;
 791         }
 792
 793         hw->mac.autoneg = 1;
 794         hw->phy.autoneg_wait_to_complete = 0;
 795         hw->phy.autoneg_advertised = E1000_ALL_SPEED_DUPLEX;
 796
 797         /* Copper options */
 798         if (hw->phy.media_type == e1000_media_type_copper) {
 799                 hw->phy.mdix = 0; /* AUTO_ALL_MODES */
 800                 hw->phy.disable_polarity_correction = 0;
 801                 hw->phy.ms_type = e1000_ms_hw_default;
 802         }
 803
 804         /*
 805          * Start from a known state, this is important in reading the nvm
 806          * and mac from that.
 807          */
 808         igb_pf_reset_hw(hw);
 809
 810         /* Make sure we have a good EEPROM before we read from it */
 811         if (e1000_validate_nvm_checksum(hw) < 0) {
 812                 /*
 813                  * Some PCI-E parts fail the first check due to
 814                  * the link being in sleep state, call it again,
 815                  * if it fails a second time its a real issue.
 816                  */
 817                 if (e1000_validate_nvm_checksum(hw) < 0) {
 818                         PMD_INIT_LOG(ERR, "EEPROM checksum invalid");
 819                         error = -EIO;
 820                         goto err_late;
 821                 }
 822         }
 823
 824         /* Read the permanent MAC address out of the EEPROM */
 825         if (e1000_read_mac_addr(hw) != 0) {
 826                 PMD_INIT_LOG(ERR, "EEPROM error while reading MAC address");
 827                 error = -EIO;
 828                 goto err_late;
 829         }
 830
 831         /* Allocate memory for storing MAC addresses */
 832         eth_dev->data->mac_addrs = rte_zmalloc("e1000",
 833                 RTE_ETHER_ADDR_LEN * hw->mac.rar_entry_count, 0);
 834         if (eth_dev->data->mac_addrs == NULL) {
 835                 PMD_INIT_LOG(ERR, "Failed to allocate %d bytes needed to "
 836                                                 "store MAC addresses",
 837                                 RTE_ETHER_ADDR_LEN * hw->mac.rar_entry_count);
 838                 error = -ENOMEM;
 839                 goto err_late;
 840         }
 841
 842         /* Copy the permanent MAC address */
 843         rte_ether_addr_copy((struct rte_ether_addr *)hw->mac.addr,
 844                         &eth_dev->data->mac_addrs[0]);
 845
 846         /* initialize the vfta */
 847         memset(shadow_vfta, 0, sizeof(*shadow_vfta));
 848
 849         /* Now initialize the hardware */
 850         if (igb_hardware_init(hw) != 0) {
 851                 PMD_INIT_LOG(ERR, "Hardware initialization failed");
 852                 rte_free(eth_dev->data->mac_addrs);
 853                 eth_dev->data->mac_addrs = NULL;
 854                 error = -ENODEV;
 855                 goto err_late;
 856         }
 857         hw->mac.get_link_status = 1;
 858         adapter->stopped = 0;
 859
 860         /* Indicate SOL/IDER usage */
 861         if (e1000_check_reset_block(hw) < 0) {
 862                 PMD_INIT_LOG(ERR, "PHY reset is blocked due to"
 863                                         "SOL/IDER session");
 864         }
 865
 866         /* initialize PF if max_vfs not zero */
 867         igb_pf_host_init(eth_dev);
 868
 869         ctrl_ext = E1000_READ_REG(hw, E1000_CTRL_EXT);
 870         /* Set PF Reset Done bit so PF/VF Mail Ops can work */
 871         ctrl_ext |= E1000_CTRL_EXT_PFRSTD;
 872         E1000_WRITE_REG(hw, E1000_CTRL_EXT, ctrl_ext);
 873         E1000_WRITE_FLUSH(hw);
 874
 875         PMD_INIT_LOG(DEBUG, "port_id %d vendorID=0x%x deviceID=0x%x",
 876                      eth_dev->data->port_id, pci_dev->id.vendor_id,
 877                      pci_dev->id.device_id);
 878
 879         rte_intr_callback_register(&pci_dev->intr_handle,
 880                                    eth_igb_interrupt_handler,
 881                                    (void *)eth_dev);
 882
 883         /* enable uio/vfio intr/eventfd mapping */
 884         rte_intr_enable(&pci_dev->intr_handle);
 885
 886         /* enable support intr */
 887         igb_intr_enable(eth_dev);
 888
 889         /* initialize filter info */
 890         memset(filter_info, 0,
 891                sizeof(struct e1000_filter_info));
 892
 893         TAILQ_INIT(&filter_info->flex_list);
 894         TAILQ_INIT(&filter_info->twotuple_list);
 895         TAILQ_INIT(&filter_info->fivetuple_list);
 896
 897         TAILQ_INIT(&igb_filter_ntuple_list);
 898         TAILQ_INIT(&igb_filter_ethertype_list);
 899         TAILQ_INIT(&igb_filter_syn_list);
 900         TAILQ_INIT(&igb_filter_flex_list);
 901         TAILQ_INIT(&igb_filter_rss_list);
 902         TAILQ_INIT(&igb_flow_list);
 903
 904         return 0;
 905
 906 err_late:
 907         igb_hw_control_release(hw);
 908
 909         return error;
 910 }
 911
 912 static int
 913 eth_igb_dev_uninit(struct rte_eth_dev *eth_dev)
 914 {
 915         struct rte_pci_device *pci_dev;
 916         struct rte_intr_handle *intr_handle;
 917         struct e1000_hw *hw;
 918         struct e1000_adapter *adapter =
 919                 E1000_DEV_PRIVATE(eth_dev->data->dev_private);
 920         struct e1000_filter_info *filter_info =
 921                 E1000_DEV_PRIVATE_TO_FILTER_INFO(eth_dev->data->dev_private);
 922
 923         PMD_INIT_FUNC_TRACE();
 924
 925         if (rte_eal_process_type() != RTE_PROC_PRIMARY)
 926                 return -EPERM;
 927
 928         hw = E1000_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
 929         pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
 930         intr_handle = &pci_dev->intr_handle;
 931
 932         if (adapter->stopped == 0)
 933                 eth_igb_close(eth_dev);
 934
 935         eth_dev->dev_ops = NULL;
 936         eth_dev->rx_pkt_burst = NULL;
 937         eth_dev->tx_pkt_burst = NULL;
 938
 939         /* Reset any pending lock */
 940         igb_reset_swfw_lock(hw);
 941
 942         /* uninitialize PF if max_vfs not zero */
 943         igb_pf_host_uninit(eth_dev);
 944
 945         /* disable uio intr before callback unregister */
 946         rte_intr_disable(intr_handle);
 947         rte_intr_callback_unregister(intr_handle,
 948                                      eth_igb_interrupt_handler, eth_dev);
 949
 950         /* clear the SYN filter info */
 951         filter_info->syn_info = 0;
 952
 953         /* clear the ethertype filters info */
 954         filter_info->ethertype_mask = 0;
 955         memset(filter_info->ethertype_filters, 0,
 956                 E1000_MAX_ETQF_FILTERS * sizeof(struct igb_ethertype_filter));
 957
 958         /* clear the rss filter info */
 959         memset(&filter_info->rss_info, 0,
 960                 sizeof(struct igb_rte_flow_rss_conf));
 961
 962         /* remove all ntuple filters of the device */
 963         igb_ntuple_filter_uninit(eth_dev);
 964
 965         /* remove all flex filters of the device */
 966         igb_flex_filter_uninit(eth_dev);
 967
 968         /* clear all the filters list */
 969         igb_filterlist_flush(eth_dev);
 970
 971         return 0;
 972 }
 973
 974 /*
 975  * Virtual Function device init
 976  */
 977 static int
 978 eth_igbvf_dev_init(struct rte_eth_dev *eth_dev)
 979 {
 980         struct rte_pci_device *pci_dev;
 981         struct rte_intr_handle *intr_handle;
 982         struct e1000_adapter *adapter =
 983                 E1000_DEV_PRIVATE(eth_dev->data->dev_private);
 984         struct e1000_hw *hw =
 985                 E1000_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
 986         int diag;
 987         struct rte_ether_addr *perm_addr =
 988                 (struct rte_ether_addr *)hw->mac.perm_addr;
 989
 990         PMD_INIT_FUNC_TRACE();
 991
 992         eth_dev->dev_ops = &igbvf_eth_dev_ops;
 993         eth_dev->rx_pkt_burst = &eth_igb_recv_pkts;
 994         eth_dev->tx_pkt_burst = &eth_igb_xmit_pkts;
 995         eth_dev->tx_pkt_prepare = &eth_igb_prep_pkts;
 996
 997         /* for secondary processes, we don't initialise any further as primary
 998          * has already done this work. Only check we don't need a different
 999          * RX function */
1000         if (rte_eal_process_type() != RTE_PROC_PRIMARY){
1001                 if (eth_dev->data->scattered_rx)
1002                         eth_dev->rx_pkt_burst = &eth_igb_recv_scattered_pkts;
1003                 return 0;
1004         }
1005
1006         pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
1007         rte_eth_copy_pci_info(eth_dev, pci_dev);
1008
1009         hw->device_id = pci_dev->id.device_id;
1010         hw->vendor_id = pci_dev->id.vendor_id;
1011         hw->hw_addr = (void *)pci_dev->mem_resource[0].addr;
1012         adapter->stopped = 0;
1013
1014         /* Initialize the shared code (base driver) */
1015         diag = e1000_setup_init_funcs(hw, TRUE);
1016         if (diag != 0) {
1017                 PMD_INIT_LOG(ERR, "Shared code init failed for igbvf: %d",
1018                         diag);
1019                 return -EIO;
1020         }
1021
1022         /* init_mailbox_params */
1023         hw->mbx.ops.init_params(hw);
1024
1025         /* Disable the interrupts for VF */
1026         igbvf_intr_disable(hw);
1027
1028         diag = hw->mac.ops.reset_hw(hw);
1029
1030         /* Allocate memory for storing MAC addresses */
1031         eth_dev->data->mac_addrs = rte_zmalloc("igbvf", RTE_ETHER_ADDR_LEN *
1032                 hw->mac.rar_entry_count, 0);
1033         if (eth_dev->data->mac_addrs == NULL) {
1034                 PMD_INIT_LOG(ERR,
1035                         "Failed to allocate %d bytes needed to store MAC "
1036                         "addresses",
1037                         RTE_ETHER_ADDR_LEN * hw->mac.rar_entry_count);
1038                 return -ENOMEM;
1039         }
1040
1041         /* Generate a random MAC address, if none was assigned by PF. */
1042         if (rte_is_zero_ether_addr(perm_addr)) {
1043                 rte_eth_random_addr(perm_addr->addr_bytes);
1044                 PMD_INIT_LOG(INFO, "\tVF MAC address not assigned by Host PF");
1045                 PMD_INIT_LOG(INFO, "\tAssign randomly generated MAC address "
1046                              "%02x:%02x:%02x:%02x:%02x:%02x",
1047                              perm_addr->addr_bytes[0],
1048                              perm_addr->addr_bytes[1],
1049                              perm_addr->addr_bytes[2],
1050                              perm_addr->addr_bytes[3],
1051                              perm_addr->addr_bytes[4],
1052                              perm_addr->addr_bytes[5]);
1053         }
1054
1055         diag = e1000_rar_set(hw, perm_addr->addr_bytes, 0);
1056         if (diag) {
1057                 rte_free(eth_dev->data->mac_addrs);
1058                 eth_dev->data->mac_addrs = NULL;
1059                 return diag;
1060         }
1061         /* Copy the permanent MAC address */
1062         rte_ether_addr_copy((struct rte_ether_addr *)hw->mac.perm_addr,
1063                         &eth_dev->data->mac_addrs[0]);
1064
1065         PMD_INIT_LOG(DEBUG, "port %d vendorID=0x%x deviceID=0x%x "
1066                      "mac.type=%s",
1067                      eth_dev->data->port_id, pci_dev->id.vendor_id,
1068                      pci_dev->id.device_id, "igb_mac_82576_vf");
1069
1070         intr_handle = &pci_dev->intr_handle;
1071         rte_intr_callback_register(intr_handle,
1072                                    eth_igbvf_interrupt_handler, eth_dev);
1073
1074         return 0;
1075 }
1076
1077 static int
1078 eth_igbvf_dev_uninit(struct rte_eth_dev *eth_dev)
1079 {
1080         struct e1000_adapter *adapter =
1081                 E1000_DEV_PRIVATE(eth_dev->data->dev_private);
1082         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
1083
1084         PMD_INIT_FUNC_TRACE();
1085
1086         if (rte_eal_process_type() != RTE_PROC_PRIMARY)
1087                 return -EPERM;
1088
1089         if (adapter->stopped == 0)
1090                 igbvf_dev_close(eth_dev);
1091
1092         eth_dev->dev_ops = NULL;
1093         eth_dev->rx_pkt_burst = NULL;
1094         eth_dev->tx_pkt_burst = NULL;
1095
1096         /* disable uio intr before callback unregister */
1097         rte_intr_disable(&pci_dev->intr_handle);
1098         rte_intr_callback_unregister(&pci_dev->intr_handle,
1099                                      eth_igbvf_interrupt_handler,
1100                                      (void *)eth_dev);
1101
1102         return 0;
1103 }
1104
1105 static int eth_igb_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
1106         struct rte_pci_device *pci_dev)
1107 {
1108         return rte_eth_dev_pci_generic_probe(pci_dev,
1109                 sizeof(struct e1000_adapter), eth_igb_dev_init);
1110 }
1111
1112 static int eth_igb_pci_remove(struct rte_pci_device *pci_dev)
1113 {
1114         return rte_eth_dev_pci_generic_remove(pci_dev, eth_igb_dev_uninit);
1115 }
1116
1117 static struct rte_pci_driver rte_igb_pmd = {
1118         .id_table = pci_id_igb_map,
1119         .drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC,
1120         .probe = eth_igb_pci_probe,
1121         .remove = eth_igb_pci_remove,
1122 };
1123
1124
1125 static int eth_igbvf_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
1126         struct rte_pci_device *pci_dev)
1127 {
1128         return rte_eth_dev_pci_generic_probe(pci_dev,
1129                 sizeof(struct e1000_adapter), eth_igbvf_dev_init);
1130 }
1131
1132 static int eth_igbvf_pci_remove(struct rte_pci_device *pci_dev)
1133 {
1134         return rte_eth_dev_pci_generic_remove(pci_dev, eth_igbvf_dev_uninit);
1135 }
1136
1137 /*
1138  * virtual function driver struct
1139  */
1140 static struct rte_pci_driver rte_igbvf_pmd = {
1141         .id_table = pci_id_igbvf_map,
1142         .drv_flags = RTE_PCI_DRV_NEED_MAPPING,
1143         .probe = eth_igbvf_pci_probe,
1144         .remove = eth_igbvf_pci_remove,
1145 };
1146
1147 static void
1148 igb_vmdq_vlan_hw_filter_enable(struct rte_eth_dev *dev)
1149 {
1150         struct e1000_hw *hw =
1151                 E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1152         /* RCTL: enable VLAN filter since VMDq always use VLAN filter */
1153         uint32_t rctl = E1000_READ_REG(hw, E1000_RCTL);
1154         rctl |= E1000_RCTL_VFE;
1155         E1000_WRITE_REG(hw, E1000_RCTL, rctl);
1156 }
1157
1158 static int
1159 igb_check_mq_mode(struct rte_eth_dev *dev)
1160 {
1161         enum rte_eth_rx_mq_mode rx_mq_mode = dev->data->dev_conf.rxmode.mq_mode;
1162         enum rte_eth_tx_mq_mode tx_mq_mode = dev->data->dev_conf.txmode.mq_mode;
1163         uint16_t nb_rx_q = dev->data->nb_rx_queues;
1164         uint16_t nb_tx_q = dev->data->nb_tx_queues;
1165
1166         if ((rx_mq_mode & ETH_MQ_RX_DCB_FLAG) ||
1167             tx_mq_mode == ETH_MQ_TX_DCB ||
1168             tx_mq_mode == ETH_MQ_TX_VMDQ_DCB) {
1169                 PMD_INIT_LOG(ERR, "DCB mode is not supported.");
1170                 return -EINVAL;
1171         }
1172         if (RTE_ETH_DEV_SRIOV(dev).active != 0) {
1173                 /* Check multi-queue mode.
1174                  * To no break software we accept ETH_MQ_RX_NONE as this might
1175                  * be used to turn off VLAN filter.
1176                  */
1177
1178                 if (rx_mq_mode == ETH_MQ_RX_NONE ||
1179                     rx_mq_mode == ETH_MQ_RX_VMDQ_ONLY) {
1180                         dev->data->dev_conf.rxmode.mq_mode = ETH_MQ_RX_VMDQ_ONLY;
1181                         RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool = 1;
1182                 } else {
1183                         /* Only support one queue on VFs.
1184                          * RSS together with SRIOV is not supported.
1185                          */
1186                         PMD_INIT_LOG(ERR, "SRIOV is active,"
1187                                         " wrong mq_mode rx %d.",
1188                                         rx_mq_mode);
1189                         return -EINVAL;
1190                 }
1191                 /* TX mode is not used here, so mode might be ignored.*/
1192                 if (tx_mq_mode != ETH_MQ_TX_VMDQ_ONLY) {
1193                         /* SRIOV only works in VMDq enable mode */
1194                         PMD_INIT_LOG(WARNING, "SRIOV is active,"
1195                                         " TX mode %d is not supported. "
1196                                         " Driver will behave as %d mode.",
1197                                         tx_mq_mode, ETH_MQ_TX_VMDQ_ONLY);
1198                 }
1199
1200                 /* check valid queue number */
1201                 if ((nb_rx_q > 1) || (nb_tx_q > 1)) {
1202                         PMD_INIT_LOG(ERR, "SRIOV is active,"
1203                                         " only support one queue on VFs.");
1204                         return -EINVAL;
1205                 }
1206         } else {
1207                 /* To no break software that set invalid mode, only display
1208                  * warning if invalid mode is used.
1209                  */
1210                 if (rx_mq_mode != ETH_MQ_RX_NONE &&
1211                     rx_mq_mode != ETH_MQ_RX_VMDQ_ONLY &&
1212                     rx_mq_mode != ETH_MQ_RX_RSS) {
1213                         /* RSS together with VMDq not supported*/
1214                         PMD_INIT_LOG(ERR, "RX mode %d is not supported.",
1215                                      rx_mq_mode);
1216                         return -EINVAL;
1217                 }
1218
1219                 if (tx_mq_mode != ETH_MQ_TX_NONE &&
1220                     tx_mq_mode != ETH_MQ_TX_VMDQ_ONLY) {
1221                         PMD_INIT_LOG(WARNING, "TX mode %d is not supported."
1222                                         " Due to txmode is meaningless in this"
1223                                         " driver, just ignore.",
1224                                         tx_mq_mode);
1225                 }
1226         }
1227         return 0;
1228 }
1229
1230 static int
1231 eth_igb_configure(struct rte_eth_dev *dev)
1232 {
1233         struct e1000_interrupt *intr =
1234                 E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
1235         int ret;
1236
1237         PMD_INIT_FUNC_TRACE();
1238
1239         /* multipe queue mode checking */
1240         ret  = igb_check_mq_mode(dev);
1241         if (ret != 0) {
1242                 PMD_DRV_LOG(ERR, "igb_check_mq_mode fails with %d.",
1243                             ret);
1244                 return ret;
1245         }
1246
1247         intr->flags |= E1000_FLAG_NEED_LINK_UPDATE;
1248         PMD_INIT_FUNC_TRACE();
1249
1250         return 0;
1251 }
1252
1253 static void
1254 eth_igb_rxtx_control(struct rte_eth_dev *dev,
1255                      bool enable)
1256 {
1257         struct e1000_hw *hw =
1258                 E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1259         uint32_t tctl, rctl;
1260
1261         tctl = E1000_READ_REG(hw, E1000_TCTL);
1262         rctl = E1000_READ_REG(hw, E1000_RCTL);
1263
1264         if (enable) {
1265                 /* enable Tx/Rx */
1266                 tctl |= E1000_TCTL_EN;
1267                 rctl |= E1000_RCTL_EN;
1268         } else {
1269                 /* disable Tx/Rx */
1270                 tctl &= ~E1000_TCTL_EN;
1271                 rctl &= ~E1000_RCTL_EN;
1272         }
1273         E1000_WRITE_REG(hw, E1000_TCTL, tctl);
1274         E1000_WRITE_REG(hw, E1000_RCTL, rctl);
1275         E1000_WRITE_FLUSH(hw);
1276 }
1277
1278 static int
1279 eth_igb_start(struct rte_eth_dev *dev)
1280 {
1281         struct e1000_hw *hw =
1282                 E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1283         struct e1000_adapter *adapter =
1284                 E1000_DEV_PRIVATE(dev->data->dev_private);
1285         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1286         struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
1287         int ret, mask;
1288         uint32_t intr_vector = 0;
1289         uint32_t ctrl_ext;
1290         uint32_t *speeds;
1291         int num_speeds;
1292         bool autoneg;
1293
1294         PMD_INIT_FUNC_TRACE();
1295
1296         /* disable uio/vfio intr/eventfd mapping */
1297         rte_intr_disable(intr_handle);
1298
1299         /* Power up the phy. Needed to make the link go Up */
1300         eth_igb_dev_set_link_up(dev);
1301
1302         /*
1303          * Packet Buffer Allocation (PBA)
1304          * Writing PBA sets the receive portion of the buffer
1305          * the remainder is used for the transmit buffer.
1306          */
1307         if (hw->mac.type == e1000_82575) {
1308                 uint32_t pba;
1309
1310                 pba = E1000_PBA_32K; /* 32K for Rx, 16K for Tx */
1311                 E1000_WRITE_REG(hw, E1000_PBA, pba);
1312         }
1313
1314         /* Put the address into the Receive Address Array */
1315         e1000_rar_set(hw, hw->mac.addr, 0);
1316
1317         /* Initialize the hardware */
1318         if (igb_hardware_init(hw)) {
1319                 PMD_INIT_LOG(ERR, "Unable to initialize the hardware");
1320                 return -EIO;
1321         }
1322         adapter->stopped = 0;
1323
1324         E1000_WRITE_REG(hw, E1000_VET,
1325                         RTE_ETHER_TYPE_VLAN << 16 | RTE_ETHER_TYPE_VLAN);
1326
1327         ctrl_ext = E1000_READ_REG(hw, E1000_CTRL_EXT);
1328         /* Set PF Reset Done bit so PF/VF Mail Ops can work */
1329         ctrl_ext |= E1000_CTRL_EXT_PFRSTD;
1330         E1000_WRITE_REG(hw, E1000_CTRL_EXT, ctrl_ext);
1331         E1000_WRITE_FLUSH(hw);
1332
1333         /* configure PF module if SRIOV enabled */
1334         igb_pf_host_configure(dev);
1335
1336         /* check and configure queue intr-vector mapping */
1337         if ((rte_intr_cap_multiple(intr_handle) ||
1338              !RTE_ETH_DEV_SRIOV(dev).active) &&
1339             dev->data->dev_conf.intr_conf.rxq != 0) {
1340                 intr_vector = dev->data->nb_rx_queues;
1341                 if (rte_intr_efd_enable(intr_handle, intr_vector))
1342                         return -1;
1343         }
1344
1345         if (rte_intr_dp_is_en(intr_handle) && !intr_handle->intr_vec) {
1346                 intr_handle->intr_vec =
1347                         rte_zmalloc("intr_vec",
1348                                     dev->data->nb_rx_queues * sizeof(int), 0);
1349                 if (intr_handle->intr_vec == NULL) {
1350                         PMD_INIT_LOG(ERR, "Failed to allocate %d rx_queues"
1351                                      " intr_vec", dev->data->nb_rx_queues);
1352                         return -ENOMEM;
1353                 }
1354         }
1355
1356         /* confiugre msix for rx interrupt */
1357         eth_igb_configure_msix_intr(dev);
1358
1359         /* Configure for OS presence */
1360         igb_init_manageability(hw);
1361
1362         eth_igb_tx_init(dev);
1363
1364         /* This can fail when allocating mbufs for descriptor rings */
1365         ret = eth_igb_rx_init(dev);
1366         if (ret) {
1367                 PMD_INIT_LOG(ERR, "Unable to initialize RX hardware");
1368                 igb_dev_clear_queues(dev);
1369                 return ret;
1370         }
1371
1372         e1000_clear_hw_cntrs_base_generic(hw);
1373
1374         /*
1375          * VLAN Offload Settings
1376          */
1377         mask = ETH_VLAN_STRIP_MASK | ETH_VLAN_FILTER_MASK | \
1378                         ETH_VLAN_EXTEND_MASK;
1379         ret = eth_igb_vlan_offload_set(dev, mask);
1380         if (ret) {
1381                 PMD_INIT_LOG(ERR, "Unable to set vlan offload");
1382                 igb_dev_clear_queues(dev);
1383                 return ret;
1384         }
1385
1386         if (dev->data->dev_conf.rxmode.mq_mode == ETH_MQ_RX_VMDQ_ONLY) {
1387                 /* Enable VLAN filter since VMDq always use VLAN filter */
1388                 igb_vmdq_vlan_hw_filter_enable(dev);
1389         }
1390
1391         if ((hw->mac.type == e1000_82576) || (hw->mac.type == e1000_82580) ||
1392                 (hw->mac.type == e1000_i350) || (hw->mac.type == e1000_i210) ||
1393                 (hw->mac.type == e1000_i211)) {
1394                 /* Configure EITR with the maximum possible value (0xFFFF) */
1395                 E1000_WRITE_REG(hw, E1000_EITR(0), 0xFFFF);
1396         }
1397
1398         /* Setup link speed and duplex */
1399         speeds = &dev->data->dev_conf.link_speeds;
1400         if (*speeds == ETH_LINK_SPEED_AUTONEG) {
1401                 hw->phy.autoneg_advertised = E1000_ALL_SPEED_DUPLEX;
1402                 hw->mac.autoneg = 1;
1403         } else {
1404                 num_speeds = 0;
1405                 autoneg = (*speeds & ETH_LINK_SPEED_FIXED) == 0;
1406
1407                 /* Reset */
1408                 hw->phy.autoneg_advertised = 0;
1409
1410                 if (*speeds & ~(ETH_LINK_SPEED_10M_HD | ETH_LINK_SPEED_10M |
1411                                 ETH_LINK_SPEED_100M_HD | ETH_LINK_SPEED_100M |
1412                                 ETH_LINK_SPEED_1G | ETH_LINK_SPEED_FIXED)) {
1413                         num_speeds = -1;
1414                         goto error_invalid_config;
1415                 }
1416                 if (*speeds & ETH_LINK_SPEED_10M_HD) {
1417                         hw->phy.autoneg_advertised |= ADVERTISE_10_HALF;
1418                         num_speeds++;
1419                 }
1420                 if (*speeds & ETH_LINK_SPEED_10M) {
1421                         hw->phy.autoneg_advertised |= ADVERTISE_10_FULL;
1422                         num_speeds++;
1423                 }
1424                 if (*speeds & ETH_LINK_SPEED_100M_HD) {
1425                         hw->phy.autoneg_advertised |= ADVERTISE_100_HALF;
1426                         num_speeds++;
1427                 }
1428                 if (*speeds & ETH_LINK_SPEED_100M) {
1429                         hw->phy.autoneg_advertised |= ADVERTISE_100_FULL;
1430                         num_speeds++;
1431                 }
1432                 if (*speeds & ETH_LINK_SPEED_1G) {
1433                         hw->phy.autoneg_advertised |= ADVERTISE_1000_FULL;
1434                         num_speeds++;
1435                 }
1436                 if (num_speeds == 0 || (!autoneg && (num_speeds > 1)))
1437                         goto error_invalid_config;
1438
1439                 /* Set/reset the mac.autoneg based on the link speed,
1440                  * fixed or not
1441                  */
1442                 if (!autoneg) {
1443                         hw->mac.autoneg = 0;
1444                         hw->mac.forced_speed_duplex =
1445                                         hw->phy.autoneg_advertised;
1446                 } else {
1447                         hw->mac.autoneg = 1;
1448                 }
1449         }
1450
1451         e1000_setup_link(hw);
1452
1453         if (rte_intr_allow_others(intr_handle)) {
1454                 /* check if lsc interrupt is enabled */
1455                 if (dev->data->dev_conf.intr_conf.lsc != 0)
1456                         eth_igb_lsc_interrupt_setup(dev, TRUE);
1457                 else
1458                         eth_igb_lsc_interrupt_setup(dev, FALSE);
1459         } else {
1460                 rte_intr_callback_unregister(intr_handle,
1461                                              eth_igb_interrupt_handler,
1462                                              (void *)dev);
1463                 if (dev->data->dev_conf.intr_conf.lsc != 0)
1464                         PMD_INIT_LOG(INFO, "lsc won't enable because of"
1465                                      " no intr multiplex");
1466         }
1467
1468         /* check if rxq interrupt is enabled */
1469         if (dev->data->dev_conf.intr_conf.rxq != 0 &&
1470             rte_intr_dp_is_en(intr_handle))
1471                 eth_igb_rxq_interrupt_setup(dev);
1472
1473         /* enable uio/vfio intr/eventfd mapping */
1474         rte_intr_enable(intr_handle);
1475
1476         /* resume enabled intr since hw reset */
1477         igb_intr_enable(dev);
1478
1479         /* restore all types filter */
1480         igb_filter_restore(dev);
1481
1482         eth_igb_rxtx_control(dev, true);
1483         eth_igb_link_update(dev, 0);
1484
1485         PMD_INIT_LOG(DEBUG, "<<");
1486
1487         return 0;
1488
1489 error_invalid_config:
1490         PMD_INIT_LOG(ERR, "Invalid advertised speeds (%u) for port %u",
1491                      dev->data->dev_conf.link_speeds, dev->data->port_id);
1492         igb_dev_clear_queues(dev);
1493         return -EINVAL;
1494 }
1495
1496 /*********************************************************************
1497  *
1498  *  This routine disables all traffic on the adapter by issuing a
1499  *  global reset on the MAC.
1500  *
1501  **********************************************************************/
1502 static void
1503 eth_igb_stop(struct rte_eth_dev *dev)
1504 {
1505         struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1506         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1507         struct rte_eth_link link;
1508         struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
1509
1510         eth_igb_rxtx_control(dev, false);
1511
1512         igb_intr_disable(dev);
1513
1514         /* disable intr eventfd mapping */
1515         rte_intr_disable(intr_handle);
1516
1517         igb_pf_reset_hw(hw);
1518         E1000_WRITE_REG(hw, E1000_WUC, 0);
1519
1520         /* Set bit for Go Link disconnect */
1521         if (hw->mac.type >= e1000_82580) {
1522                 uint32_t phpm_reg;
1523
1524                 phpm_reg = E1000_READ_REG(hw, E1000_82580_PHY_POWER_MGMT);
1525                 phpm_reg |= E1000_82580_PM_GO_LINKD;
1526                 E1000_WRITE_REG(hw, E1000_82580_PHY_POWER_MGMT, phpm_reg);
1527         }
1528
1529         /* Power down the phy. Needed to make the link go Down */
1530         eth_igb_dev_set_link_down(dev);
1531
1532         igb_dev_clear_queues(dev);
1533
1534         /* clear the recorded link status */
1535         memset(&link, 0, sizeof(link));
1536         rte_eth_linkstatus_set(dev, &link);
1537
1538         if (!rte_intr_allow_others(intr_handle))
1539                 /* resume to the default handler */
1540                 rte_intr_callback_register(intr_handle,
1541                                            eth_igb_interrupt_handler,
1542                                            (void *)dev);
1543
1544         /* Clean datapath event and queue/vec mapping */
1545         rte_intr_efd_disable(intr_handle);
1546         if (intr_handle->intr_vec != NULL) {
1547                 rte_free(intr_handle->intr_vec);
1548                 intr_handle->intr_vec = NULL;
1549         }
1550 }
1551
1552 static int
1553 eth_igb_dev_set_link_up(struct rte_eth_dev *dev)
1554 {
1555         struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1556
1557         if (hw->phy.media_type == e1000_media_type_copper)
1558                 e1000_power_up_phy(hw);
1559         else
1560                 e1000_power_up_fiber_serdes_link(hw);
1561
1562         return 0;
1563 }
1564
1565 static int
1566 eth_igb_dev_set_link_down(struct rte_eth_dev *dev)
1567 {
1568         struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1569
1570         if (hw->phy.media_type == e1000_media_type_copper)
1571                 e1000_power_down_phy(hw);
1572         else
1573                 e1000_shutdown_fiber_serdes_link(hw);
1574
1575         return 0;
1576 }
1577
1578 static void
1579 eth_igb_close(struct rte_eth_dev *dev)
1580 {
1581         struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1582         struct e1000_adapter *adapter =
1583                 E1000_DEV_PRIVATE(dev->data->dev_private);
1584         struct rte_eth_link link;
1585         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1586         struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
1587
1588         eth_igb_stop(dev);
1589         adapter->stopped = 1;
1590
1591         e1000_phy_hw_reset(hw);
1592         igb_release_manageability(hw);
1593         igb_hw_control_release(hw);
1594
1595         /* Clear bit for Go Link disconnect */
1596         if (hw->mac.type >= e1000_82580) {
1597                 uint32_t phpm_reg;
1598
1599                 phpm_reg = E1000_READ_REG(hw, E1000_82580_PHY_POWER_MGMT);
1600                 phpm_reg &= ~E1000_82580_PM_GO_LINKD;
1601                 E1000_WRITE_REG(hw, E1000_82580_PHY_POWER_MGMT, phpm_reg);
1602         }
1603
1604         igb_dev_free_queues(dev);
1605
1606         if (intr_handle->intr_vec) {
1607                 rte_free(intr_handle->intr_vec);
1608                 intr_handle->intr_vec = NULL;
1609         }
1610
1611         memset(&link, 0, sizeof(link));
1612         rte_eth_linkstatus_set(dev, &link);
1613 }
1614
1615 /*
1616  * Reset PF device.
1617  */
1618 static int
1619 eth_igb_reset(struct rte_eth_dev *dev)
1620 {
1621         int ret;
1622
1623         /* When a DPDK PMD PF begin to reset PF port, it should notify all
1624          * its VF to make them align with it. The detailed notification
1625          * mechanism is PMD specific and is currently not implemented.
1626          * To avoid unexpected behavior in VF, currently reset of PF with
1627          * SR-IOV activation is not supported. It might be supported later.
1628          */
1629         if (dev->data->sriov.active)
1630                 return -ENOTSUP;
1631
1632         ret = eth_igb_dev_uninit(dev);
1633         if (ret)
1634                 return ret;
1635
1636         ret = eth_igb_dev_init(dev);
1637
1638         return ret;
1639 }
1640
1641
1642 static int
1643 igb_get_rx_buffer_size(struct e1000_hw *hw)
1644 {
1645         uint32_t rx_buf_size;
1646         if (hw->mac.type == e1000_82576) {
1647                 rx_buf_size = (E1000_READ_REG(hw, E1000_RXPBS) & 0xffff) << 10;
1648         } else if (hw->mac.type == e1000_82580 || hw->mac.type == e1000_i350) {
1649                 /* PBS needs to be translated according to a lookup table */
1650                 rx_buf_size = (E1000_READ_REG(hw, E1000_RXPBS) & 0xf);
1651                 rx_buf_size = (uint32_t) e1000_rxpbs_adjust_82580(rx_buf_size);
1652                 rx_buf_size = (rx_buf_size << 10);
1653         } else if (hw->mac.type == e1000_i210 || hw->mac.type == e1000_i211) {
1654                 rx_buf_size = (E1000_READ_REG(hw, E1000_RXPBS) & 0x3f) << 10;
1655         } else {
1656                 rx_buf_size = (E1000_READ_REG(hw, E1000_PBA) & 0xffff) << 10;
1657         }
1658
1659         return rx_buf_size;
1660 }
1661
1662 /*********************************************************************
1663  *
1664  *  Initialize the hardware
1665  *
1666  **********************************************************************/
1667 static int
1668 igb_hardware_init(struct e1000_hw *hw)
1669 {
1670         uint32_t rx_buf_size;
1671         int diag;
1672
1673         /* Let the firmware know the OS is in control */
1674         igb_hw_control_acquire(hw);
1675
1676         /*
1677          * These parameters control the automatic generation (Tx) and
1678          * response (Rx) to Ethernet PAUSE frames.
1679          * - High water mark should allow for at least two standard size (1518)
1680          *   frames to be received after sending an XOFF.
1681          * - Low water mark works best when it is very near the high water mark.
1682          *   This allows the receiver to restart by sending XON when it has
1683          *   drained a bit. Here we use an arbitrary value of 1500 which will
1684          *   restart after one full frame is pulled from the buffer. There
1685          *   could be several smaller frames in the buffer and if so they will
1686          *   not trigger the XON until their total number reduces the buffer
1687          *   by 1500.
1688          * - The pause time is fairly large at 1000 x 512ns = 512 usec.
1689          */
1690         rx_buf_size = igb_get_rx_buffer_size(hw);
1691
1692         hw->fc.high_water = rx_buf_size - (RTE_ETHER_MAX_LEN * 2);
1693         hw->fc.low_water = hw->fc.high_water - 1500;
1694         hw->fc.pause_time = IGB_FC_PAUSE_TIME;
1695         hw->fc.send_xon = 1;
1696
1697         /* Set Flow control, use the tunable location if sane */
1698         if ((igb_fc_setting != e1000_fc_none) && (igb_fc_setting < 4))
1699                 hw->fc.requested_mode = igb_fc_setting;
1700         else
1701                 hw->fc.requested_mode = e1000_fc_none;
1702
1703         /* Issue a global reset */
1704         igb_pf_reset_hw(hw);
1705         E1000_WRITE_REG(hw, E1000_WUC, 0);
1706
1707         diag = e1000_init_hw(hw);
1708         if (diag < 0)
1709                 return diag;
1710
1711         E1000_WRITE_REG(hw, E1000_VET,
1712                         RTE_ETHER_TYPE_VLAN << 16 | RTE_ETHER_TYPE_VLAN);
1713         e1000_get_phy_info(hw);
1714         e1000_check_for_link(hw);
1715
1716         return 0;
1717 }
1718
1719 /* This function is based on igb_update_stats_counters() in igb/if_igb.c */
1720 static void
1721 igb_read_stats_registers(struct e1000_hw *hw, struct e1000_hw_stats *stats)
1722 {
1723         int pause_frames;
1724
1725         uint64_t old_gprc  = stats->gprc;
1726         uint64_t old_gptc  = stats->gptc;
1727         uint64_t old_tpr   = stats->tpr;
1728         uint64_t old_tpt   = stats->tpt;
1729         uint64_t old_rpthc = stats->rpthc;
1730         uint64_t old_hgptc = stats->hgptc;
1731
1732         if(hw->phy.media_type == e1000_media_type_copper ||
1733             (E1000_READ_REG(hw, E1000_STATUS) & E1000_STATUS_LU)) {
1734                 stats->symerrs +=
1735                     E1000_READ_REG(hw,E1000_SYMERRS);
1736                 stats->sec += E1000_READ_REG(hw, E1000_SEC);
1737         }
1738
1739         stats->crcerrs += E1000_READ_REG(hw, E1000_CRCERRS);
1740         stats->mpc += E1000_READ_REG(hw, E1000_MPC);
1741         stats->scc += E1000_READ_REG(hw, E1000_SCC);
1742         stats->ecol += E1000_READ_REG(hw, E1000_ECOL);
1743
1744         stats->mcc += E1000_READ_REG(hw, E1000_MCC);
1745         stats->latecol += E1000_READ_REG(hw, E1000_LATECOL);
1746         stats->colc += E1000_READ_REG(hw, E1000_COLC);
1747         stats->dc += E1000_READ_REG(hw, E1000_DC);
1748         stats->rlec += E1000_READ_REG(hw, E1000_RLEC);
1749         stats->xonrxc += E1000_READ_REG(hw, E1000_XONRXC);
1750         stats->xontxc += E1000_READ_REG(hw, E1000_XONTXC);
1751         /*
1752         ** For watchdog management we need to know if we have been
1753         ** paused during the last interval, so capture that here.
1754         */
1755         pause_frames = E1000_READ_REG(hw, E1000_XOFFRXC);
1756         stats->xoffrxc += pause_frames;
1757         stats->xofftxc += E1000_READ_REG(hw, E1000_XOFFTXC);
1758         stats->fcruc += E1000_READ_REG(hw, E1000_FCRUC);
1759         stats->prc64 += E1000_READ_REG(hw, E1000_PRC64);
1760         stats->prc127 += E1000_READ_REG(hw, E1000_PRC127);
1761         stats->prc255 += E1000_READ_REG(hw, E1000_PRC255);
1762         stats->prc511 += E1000_READ_REG(hw, E1000_PRC511);
1763         stats->prc1023 += E1000_READ_REG(hw, E1000_PRC1023);
1764         stats->prc1522 += E1000_READ_REG(hw, E1000_PRC1522);
1765         stats->gprc += E1000_READ_REG(hw, E1000_GPRC);
1766         stats->bprc += E1000_READ_REG(hw, E1000_BPRC);
1767         stats->mprc += E1000_READ_REG(hw, E1000_MPRC);
1768         stats->gptc += E1000_READ_REG(hw, E1000_GPTC);
1769
1770         /* For the 64-bit byte counters the low dword must be read first. */
1771         /* Both registers clear on the read of the high dword */
1772
1773         /* Workaround CRC bytes included in size, take away 4 bytes/packet */
1774         stats->gorc += E1000_READ_REG(hw, E1000_GORCL);
1775         stats->gorc += ((uint64_t)E1000_READ_REG(hw, E1000_GORCH) << 32);
1776         stats->gorc -= (stats->gprc - old_gprc) * RTE_ETHER_CRC_LEN;
1777         stats->gotc += E1000_READ_REG(hw, E1000_GOTCL);
1778         stats->gotc += ((uint64_t)E1000_READ_REG(hw, E1000_GOTCH) << 32);
1779         stats->gotc -= (stats->gptc - old_gptc) * RTE_ETHER_CRC_LEN;
1780
1781         stats->rnbc += E1000_READ_REG(hw, E1000_RNBC);
1782         stats->ruc += E1000_READ_REG(hw, E1000_RUC);
1783         stats->rfc += E1000_READ_REG(hw, E1000_RFC);
1784         stats->roc += E1000_READ_REG(hw, E1000_ROC);
1785         stats->rjc += E1000_READ_REG(hw, E1000_RJC);
1786
1787         stats->tpr += E1000_READ_REG(hw, E1000_TPR);
1788         stats->tpt += E1000_READ_REG(hw, E1000_TPT);
1789
1790         stats->tor += E1000_READ_REG(hw, E1000_TORL);
1791         stats->tor += ((uint64_t)E1000_READ_REG(hw, E1000_TORH) << 32);
1792         stats->tor -= (stats->tpr - old_tpr) * RTE_ETHER_CRC_LEN;
1793         stats->tot += E1000_READ_REG(hw, E1000_TOTL);
1794         stats->tot += ((uint64_t)E1000_READ_REG(hw, E1000_TOTH) << 32);
1795         stats->tot -= (stats->tpt - old_tpt) * RTE_ETHER_CRC_LEN;
1796
1797         stats->ptc64 += E1000_READ_REG(hw, E1000_PTC64);
1798         stats->ptc127 += E1000_READ_REG(hw, E1000_PTC127);
1799         stats->ptc255 += E1000_READ_REG(hw, E1000_PTC255);
1800         stats->ptc511 += E1000_READ_REG(hw, E1000_PTC511);
1801         stats->ptc1023 += E1000_READ_REG(hw, E1000_PTC1023);
1802         stats->ptc1522 += E1000_READ_REG(hw, E1000_PTC1522);
1803         stats->mptc += E1000_READ_REG(hw, E1000_MPTC);
1804         stats->bptc += E1000_READ_REG(hw, E1000_BPTC);
1805
1806         /* Interrupt Counts */
1807
1808         stats->iac += E1000_READ_REG(hw, E1000_IAC);
1809         stats->icrxptc += E1000_READ_REG(hw, E1000_ICRXPTC);
1810         stats->icrxatc += E1000_READ_REG(hw, E1000_ICRXATC);
1811         stats->ictxptc += E1000_READ_REG(hw, E1000_ICTXPTC);
1812         stats->ictxatc += E1000_READ_REG(hw, E1000_ICTXATC);
1813         stats->ictxqec += E1000_READ_REG(hw, E1000_ICTXQEC);
1814         stats->ictxqmtc += E1000_READ_REG(hw, E1000_ICTXQMTC);
1815         stats->icrxdmtc += E1000_READ_REG(hw, E1000_ICRXDMTC);
1816         stats->icrxoc += E1000_READ_REG(hw, E1000_ICRXOC);
1817
1818         /* Host to Card Statistics */
1819
1820         stats->cbtmpc += E1000_READ_REG(hw, E1000_CBTMPC);
1821         stats->htdpmc += E1000_READ_REG(hw, E1000_HTDPMC);
1822         stats->cbrdpc += E1000_READ_REG(hw, E1000_CBRDPC);
1823         stats->cbrmpc += E1000_READ_REG(hw, E1000_CBRMPC);
1824         stats->rpthc += E1000_READ_REG(hw, E1000_RPTHC);
1825         stats->hgptc += E1000_READ_REG(hw, E1000_HGPTC);
1826         stats->htcbdpc += E1000_READ_REG(hw, E1000_HTCBDPC);
1827         stats->hgorc += E1000_READ_REG(hw, E1000_HGORCL);
1828         stats->hgorc += ((uint64_t)E1000_READ_REG(hw, E1000_HGORCH) << 32);
1829         stats->hgorc -= (stats->rpthc - old_rpthc) * RTE_ETHER_CRC_LEN;
1830         stats->hgotc += E1000_READ_REG(hw, E1000_HGOTCL);
1831         stats->hgotc += ((uint64_t)E1000_READ_REG(hw, E1000_HGOTCH) << 32);
1832         stats->hgotc -= (stats->hgptc - old_hgptc) * RTE_ETHER_CRC_LEN;
1833         stats->lenerrs += E1000_READ_REG(hw, E1000_LENERRS);
1834         stats->scvpc += E1000_READ_REG(hw, E1000_SCVPC);
1835         stats->hrmpc += E1000_READ_REG(hw, E1000_HRMPC);
1836
1837         stats->algnerrc += E1000_READ_REG(hw, E1000_ALGNERRC);
1838         stats->rxerrc += E1000_READ_REG(hw, E1000_RXERRC);
1839         stats->tncrs += E1000_READ_REG(hw, E1000_TNCRS);
1840         stats->cexterr += E1000_READ_REG(hw, E1000_CEXTERR);
1841         stats->tsctc += E1000_READ_REG(hw, E1000_TSCTC);
1842         stats->tsctfc += E1000_READ_REG(hw, E1000_TSCTFC);
1843 }
1844
1845 static int
1846 eth_igb_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *rte_stats)
1847 {
1848         struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1849         struct e1000_hw_stats *stats =
1850                         E1000_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
1851
1852         igb_read_stats_registers(hw, stats);
1853
1854         if (rte_stats == NULL)
1855                 return -EINVAL;
1856
1857         /* Rx Errors */
1858         rte_stats->imissed = stats->mpc;
1859         rte_stats->ierrors = stats->crcerrs +
1860                              stats->rlec + stats->ruc + stats->roc +
1861                              stats->rxerrc + stats->algnerrc + stats->cexterr;
1862
1863         /* Tx Errors */
1864         rte_stats->oerrors = stats->ecol + stats->latecol;
1865
1866         rte_stats->ipackets = stats->gprc;
1867         rte_stats->opackets = stats->gptc;
1868         rte_stats->ibytes   = stats->gorc;
1869         rte_stats->obytes   = stats->gotc;
1870         return 0;
1871 }
1872
1873 static int
1874 eth_igb_stats_reset(struct rte_eth_dev *dev)
1875 {
1876         struct e1000_hw_stats *hw_stats =
1877                         E1000_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
1878
1879         /* HW registers are cleared on read */
1880         eth_igb_stats_get(dev, NULL);
1881
1882         /* Reset software totals */
1883         memset(hw_stats, 0, sizeof(*hw_stats));
1884
1885         return 0;
1886 }
1887
1888 static int
1889 eth_igb_xstats_reset(struct rte_eth_dev *dev)
1890 {
1891         struct e1000_hw_stats *stats =
1892                         E1000_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
1893
1894         /* HW registers are cleared on read */
1895         eth_igb_xstats_get(dev, NULL, IGB_NB_XSTATS);
1896
1897         /* Reset software totals */
1898         memset(stats, 0, sizeof(*stats));
1899
1900         return 0;
1901 }
1902
1903 static int eth_igb_xstats_get_names(__rte_unused struct rte_eth_dev *dev,
1904         struct rte_eth_xstat_name *xstats_names,
1905         __rte_unused unsigned int size)
1906 {
1907         unsigned i;
1908
1909         if (xstats_names == NULL)
1910                 return IGB_NB_XSTATS;
1911
1912         /* Note: limit checked in rte_eth_xstats_names() */
1913
1914         for (i = 0; i < IGB_NB_XSTATS; i++) {
1915                 strlcpy(xstats_names[i].name, rte_igb_stats_strings[i].name,
1916                         sizeof(xstats_names[i].name));
1917         }
1918
1919         return IGB_NB_XSTATS;
1920 }
1921
1922 static int eth_igb_xstats_get_names_by_id(struct rte_eth_dev *dev,
1923                 struct rte_eth_xstat_name *xstats_names, const uint64_t *ids,
1924                 unsigned int limit)
1925 {
1926         unsigned int i;
1927
1928         if (!ids) {
1929                 if (xstats_names == NULL)
1930                         return IGB_NB_XSTATS;
1931
1932                 for (i = 0; i < IGB_NB_XSTATS; i++)
1933                         strlcpy(xstats_names[i].name,
1934                                 rte_igb_stats_strings[i].name,
1935                                 sizeof(xstats_names[i].name));
1936
1937                 return IGB_NB_XSTATS;
1938
1939         } else {
1940                 struct rte_eth_xstat_name xstats_names_copy[IGB_NB_XSTATS];
1941
1942                 eth_igb_xstats_get_names_by_id(dev, xstats_names_copy, NULL,
1943                                 IGB_NB_XSTATS);
1944
1945                 for (i = 0; i < limit; i++) {
1946                         if (ids[i] >= IGB_NB_XSTATS) {
1947                                 PMD_INIT_LOG(ERR, "id value isn't valid");
1948                                 return -1;
1949                         }
1950                         strcpy(xstats_names[i].name,
1951                                         xstats_names_copy[ids[i]].name);
1952                 }
1953                 return limit;
1954         }
1955 }
1956
1957 static int
1958 eth_igb_xstats_get(struct rte_eth_dev *dev, struct rte_eth_xstat *xstats,
1959                    unsigned n)
1960 {
1961         struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1962         struct e1000_hw_stats *hw_stats =
1963                         E1000_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
1964         unsigned i;
1965
1966         if (n < IGB_NB_XSTATS)
1967                 return IGB_NB_XSTATS;
1968
1969         igb_read_stats_registers(hw, hw_stats);
1970
1971         /* If this is a reset xstats is NULL, and we have cleared the
1972          * registers by reading them.
1973          */
1974         if (!xstats)
1975                 return 0;
1976
1977         /* Extended stats */
1978         for (i = 0; i < IGB_NB_XSTATS; i++) {
1979                 xstats[i].id = i;
1980                 xstats[i].value = *(uint64_t *)(((char *)hw_stats) +
1981                         rte_igb_stats_strings[i].offset);
1982         }
1983
1984         return IGB_NB_XSTATS;
1985 }
1986
1987 static int
1988 eth_igb_xstats_get_by_id(struct rte_eth_dev *dev, const uint64_t *ids,
1989                 uint64_t *values, unsigned int n)
1990 {
1991         unsigned int i;
1992
1993         if (!ids) {
1994                 struct e1000_hw *hw =
1995                         E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1996                 struct e1000_hw_stats *hw_stats =
1997                         E1000_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
1998
1999                 if (n < IGB_NB_XSTATS)
2000                         return IGB_NB_XSTATS;
2001
2002                 igb_read_stats_registers(hw, hw_stats);
2003
2004                 /* If this is a reset xstats is NULL, and we have cleared the
2005                  * registers by reading them.
2006                  */
2007                 if (!values)
2008                         return 0;
2009
2010                 /* Extended stats */
2011                 for (i = 0; i < IGB_NB_XSTATS; i++)
2012                         values[i] = *(uint64_t *)(((char *)hw_stats) +
2013                                         rte_igb_stats_strings[i].offset);
2014
2015                 return IGB_NB_XSTATS;
2016
2017         } else {
2018                 uint64_t values_copy[IGB_NB_XSTATS];
2019
2020                 eth_igb_xstats_get_by_id(dev, NULL, values_copy,
2021                                 IGB_NB_XSTATS);
2022
2023                 for (i = 0; i < n; i++) {
2024                         if (ids[i] >= IGB_NB_XSTATS) {
2025                                 PMD_INIT_LOG(ERR, "id value isn't valid");
2026                                 return -1;
2027                         }
2028                         values[i] = values_copy[ids[i]];
2029                 }
2030                 return n;
2031         }
2032 }
2033
2034 static void
2035 igbvf_read_stats_registers(struct e1000_hw *hw, struct e1000_vf_stats *hw_stats)
2036 {
2037         /* Good Rx packets, include VF loopback */
2038         UPDATE_VF_STAT(E1000_VFGPRC,
2039             hw_stats->last_gprc, hw_stats->gprc);
2040
2041         /* Good Rx octets, include VF loopback */
2042         UPDATE_VF_STAT(E1000_VFGORC,
2043             hw_stats->last_gorc, hw_stats->gorc);
2044
2045         /* Good Tx packets, include VF loopback */
2046         UPDATE_VF_STAT(E1000_VFGPTC,
2047             hw_stats->last_gptc, hw_stats->gptc);
2048
2049         /* Good Tx octets, include VF loopback */
2050         UPDATE_VF_STAT(E1000_VFGOTC,
2051             hw_stats->last_gotc, hw_stats->gotc);
2052
2053         /* Rx Multicst packets */
2054         UPDATE_VF_STAT(E1000_VFMPRC,
2055             hw_stats->last_mprc, hw_stats->mprc);
2056
2057         /* Good Rx loopback packets */
2058         UPDATE_VF_STAT(E1000_VFGPRLBC,
2059             hw_stats->last_gprlbc, hw_stats->gprlbc);
2060
2061         /* Good Rx loopback octets */
2062         UPDATE_VF_STAT(E1000_VFGORLBC,
2063             hw_stats->last_gorlbc, hw_stats->gorlbc);
2064
2065         /* Good Tx loopback packets */
2066         UPDATE_VF_STAT(E1000_VFGPTLBC,
2067             hw_stats->last_gptlbc, hw_stats->gptlbc);
2068
2069         /* Good Tx loopback octets */
2070         UPDATE_VF_STAT(E1000_VFGOTLBC,
2071             hw_stats->last_gotlbc, hw_stats->gotlbc);
2072 }
2073
2074 static int eth_igbvf_xstats_get_names(__rte_unused struct rte_eth_dev *dev,
2075                                      struct rte_eth_xstat_name *xstats_names,
2076                                      __rte_unused unsigned limit)
2077 {
2078         unsigned i;
2079
2080         if (xstats_names != NULL)
2081                 for (i = 0; i < IGBVF_NB_XSTATS; i++) {
2082                         strlcpy(xstats_names[i].name,
2083                                 rte_igbvf_stats_strings[i].name,
2084                                 sizeof(xstats_names[i].name));
2085                 }
2086         return IGBVF_NB_XSTATS;
2087 }
2088
2089 static int
2090 eth_igbvf_xstats_get(struct rte_eth_dev *dev, struct rte_eth_xstat *xstats,
2091                      unsigned n)
2092 {
2093         struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2094         struct e1000_vf_stats *hw_stats = (struct e1000_vf_stats *)
2095                         E1000_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
2096         unsigned i;
2097
2098         if (n < IGBVF_NB_XSTATS)
2099                 return IGBVF_NB_XSTATS;
2100
2101         igbvf_read_stats_registers(hw, hw_stats);
2102
2103         if (!xstats)
2104                 return 0;
2105
2106         for (i = 0; i < IGBVF_NB_XSTATS; i++) {
2107                 xstats[i].id = i;
2108                 xstats[i].value = *(uint64_t *)(((char *)hw_stats) +
2109                         rte_igbvf_stats_strings[i].offset);
2110         }
2111
2112         return IGBVF_NB_XSTATS;
2113 }
2114
2115 static int
2116 eth_igbvf_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *rte_stats)
2117 {
2118         struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2119         struct e1000_vf_stats *hw_stats = (struct e1000_vf_stats *)
2120                           E1000_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
2121
2122         igbvf_read_stats_registers(hw, hw_stats);
2123
2124         if (rte_stats == NULL)
2125                 return -EINVAL;
2126
2127         rte_stats->ipackets = hw_stats->gprc;
2128         rte_stats->ibytes = hw_stats->gorc;
2129         rte_stats->opackets = hw_stats->gptc;
2130         rte_stats->obytes = hw_stats->gotc;
2131         return 0;
2132 }
2133
2134 static int
2135 eth_igbvf_stats_reset(struct rte_eth_dev *dev)
2136 {
2137         struct e1000_vf_stats *hw_stats = (struct e1000_vf_stats*)
2138                         E1000_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
2139
2140         /* Sync HW register to the last stats */
2141         eth_igbvf_stats_get(dev, NULL);
2142
2143         /* reset HW current stats*/
2144         memset(&hw_stats->gprc, 0, sizeof(*hw_stats) -
2145                offsetof(struct e1000_vf_stats, gprc));
2146
2147         return 0;
2148 }
2149
2150 static int
2151 eth_igb_fw_version_get(struct rte_eth_dev *dev, char *fw_version,
2152                        size_t fw_size)
2153 {
2154         struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2155         struct e1000_fw_version fw;
2156         int ret;
2157
2158         e1000_get_fw_version(hw, &fw);
2159
2160         switch (hw->mac.type) {
2161         case e1000_i210:
2162         case e1000_i211:
2163                 if (!(e1000_get_flash_presence_i210(hw))) {
2164                         ret = snprintf(fw_version, fw_size,
2165                                  "%2d.%2d-%d",
2166                                  fw.invm_major, fw.invm_minor,
2167                                  fw.invm_img_type);
2168                         break;
2169                 }
2170                 /* fall through */
2171         default:
2172                 /* if option rom is valid, display its version too */
2173                 if (fw.or_valid) {
2174                         ret = snprintf(fw_version, fw_size,
2175                                  "%d.%d, 0x%08x, %d.%d.%d",
2176                                  fw.eep_major, fw.eep_minor, fw.etrack_id,
2177                                  fw.or_major, fw.or_build, fw.or_patch);
2178                 /* no option rom */
2179                 } else {
2180                         if (fw.etrack_id != 0X0000) {
2181                                 ret = snprintf(fw_version, fw_size,
2182                                          "%d.%d, 0x%08x",
2183                                          fw.eep_major, fw.eep_minor,
2184                                          fw.etrack_id);
2185                         } else {
2186                                 ret = snprintf(fw_version, fw_size,
2187                                          "%d.%d.%d",
2188                                          fw.eep_major, fw.eep_minor,
2189                                          fw.eep_build);
2190                         }
2191                 }
2192                 break;
2193         }
2194
2195         ret += 1; /* add the size of '\0' */
2196         if (fw_size < (u32)ret)
2197                 return ret;
2198         else
2199                 return 0;
2200 }
2201
2202 static int
2203 eth_igb_infos_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
2204 {
2205         struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2206
2207         dev_info->min_rx_bufsize = 256; /* See BSIZE field of RCTL register. */
2208         dev_info->max_rx_pktlen  = 0x3FFF; /* See RLPML register. */
2209         dev_info->max_mac_addrs = hw->mac.rar_entry_count;
2210         dev_info->rx_queue_offload_capa = igb_get_rx_queue_offloads_capa(dev);
2211         dev_info->rx_offload_capa = igb_get_rx_port_offloads_capa(dev) |
2212                                     dev_info->rx_queue_offload_capa;
2213         dev_info->tx_queue_offload_capa = igb_get_tx_queue_offloads_capa(dev);
2214         dev_info->tx_offload_capa = igb_get_tx_port_offloads_capa(dev) |
2215                                     dev_info->tx_queue_offload_capa;
2216
2217         switch (hw->mac.type) {
2218         case e1000_82575:
2219                 dev_info->max_rx_queues = 4;
2220                 dev_info->max_tx_queues = 4;
2221                 dev_info->max_vmdq_pools = 0;
2222                 break;
2223
2224         case e1000_82576:
2225                 dev_info->max_rx_queues = 16;
2226                 dev_info->max_tx_queues = 16;
2227                 dev_info->max_vmdq_pools = ETH_8_POOLS;
2228                 dev_info->vmdq_queue_num = 16;
2229                 break;
2230
2231         case e1000_82580:
2232                 dev_info->max_rx_queues = 8;
2233                 dev_info->max_tx_queues = 8;
2234                 dev_info->max_vmdq_pools = ETH_8_POOLS;
2235                 dev_info->vmdq_queue_num = 8;
2236                 break;
2237
2238         case e1000_i350:
2239                 dev_info->max_rx_queues = 8;
2240                 dev_info->max_tx_queues = 8;
2241                 dev_info->max_vmdq_pools = ETH_8_POOLS;
2242                 dev_info->vmdq_queue_num = 8;
2243                 break;
2244
2245         case e1000_i354:
2246                 dev_info->max_rx_queues = 8;
2247                 dev_info->max_tx_queues = 8;
2248                 break;
2249
2250         case e1000_i210:
2251                 dev_info->max_rx_queues = 4;
2252                 dev_info->max_tx_queues = 4;
2253                 dev_info->max_vmdq_pools = 0;
2254                 break;
2255
2256         case e1000_i211:
2257                 dev_info->max_rx_queues = 2;
2258                 dev_info->max_tx_queues = 2;
2259                 dev_info->max_vmdq_pools = 0;
2260                 break;
2261
2262         default:
2263                 /* Should not happen */
2264                 return -EINVAL;
2265         }
2266         dev_info->hash_key_size = IGB_HKEY_MAX_INDEX * sizeof(uint32_t);
2267         dev_info->reta_size = ETH_RSS_RETA_SIZE_128;
2268         dev_info->flow_type_rss_offloads = IGB_RSS_OFFLOAD_ALL;
2269
2270         dev_info->default_rxconf = (struct rte_eth_rxconf) {
2271                 .rx_thresh = {
2272                         .pthresh = IGB_DEFAULT_RX_PTHRESH,
2273                         .hthresh = IGB_DEFAULT_RX_HTHRESH,
2274                         .wthresh = IGB_DEFAULT_RX_WTHRESH,
2275                 },
2276                 .rx_free_thresh = IGB_DEFAULT_RX_FREE_THRESH,
2277                 .rx_drop_en = 0,
2278                 .offloads = 0,
2279         };
2280
2281         dev_info->default_txconf = (struct rte_eth_txconf) {
2282                 .tx_thresh = {
2283                         .pthresh = IGB_DEFAULT_TX_PTHRESH,
2284                         .hthresh = IGB_DEFAULT_TX_HTHRESH,
2285                         .wthresh = IGB_DEFAULT_TX_WTHRESH,
2286                 },
2287                 .offloads = 0,
2288         };
2289
2290         dev_info->rx_desc_lim = rx_desc_lim;
2291         dev_info->tx_desc_lim = tx_desc_lim;
2292
2293         dev_info->speed_capa = ETH_LINK_SPEED_10M_HD | ETH_LINK_SPEED_10M |
2294                         ETH_LINK_SPEED_100M_HD | ETH_LINK_SPEED_100M |
2295                         ETH_LINK_SPEED_1G;
2296
2297         dev_info->max_mtu = dev_info->max_rx_pktlen - E1000_ETH_OVERHEAD;
2298         dev_info->min_mtu = RTE_ETHER_MIN_MTU;
2299
2300         return 0;
2301 }
2302
2303 static const uint32_t *
2304 eth_igb_supported_ptypes_get(struct rte_eth_dev *dev)
2305 {
2306         static const uint32_t ptypes[] = {
2307                 /* refers to igb_rxd_pkt_info_to_pkt_type() */
2308                 RTE_PTYPE_L2_ETHER,
2309                 RTE_PTYPE_L3_IPV4,
2310                 RTE_PTYPE_L3_IPV4_EXT,
2311                 RTE_PTYPE_L3_IPV6,
2312                 RTE_PTYPE_L3_IPV6_EXT,
2313                 RTE_PTYPE_L4_TCP,
2314                 RTE_PTYPE_L4_UDP,
2315                 RTE_PTYPE_L4_SCTP,
2316                 RTE_PTYPE_TUNNEL_IP,
2317                 RTE_PTYPE_INNER_L3_IPV6,
2318                 RTE_PTYPE_INNER_L3_IPV6_EXT,
2319                 RTE_PTYPE_INNER_L4_TCP,
2320                 RTE_PTYPE_INNER_L4_UDP,
2321                 RTE_PTYPE_UNKNOWN
2322         };
2323
2324         if (dev->rx_pkt_burst == eth_igb_recv_pkts ||
2325             dev->rx_pkt_burst == eth_igb_recv_scattered_pkts)
2326                 return ptypes;
2327         return NULL;
2328 }
2329
2330 static int
2331 eth_igbvf_infos_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
2332 {
2333         struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2334
2335         dev_info->min_rx_bufsize = 256; /* See BSIZE field of RCTL register. */
2336         dev_info->max_rx_pktlen  = 0x3FFF; /* See RLPML register. */
2337         dev_info->max_mac_addrs = hw->mac.rar_entry_count;
2338         dev_info->tx_offload_capa = DEV_TX_OFFLOAD_VLAN_INSERT |
2339                                 DEV_TX_OFFLOAD_IPV4_CKSUM  |
2340                                 DEV_TX_OFFLOAD_UDP_CKSUM   |
2341                                 DEV_TX_OFFLOAD_TCP_CKSUM   |
2342                                 DEV_TX_OFFLOAD_SCTP_CKSUM  |
2343                                 DEV_TX_OFFLOAD_TCP_TSO;
2344         switch (hw->mac.type) {
2345         case e1000_vfadapt:
2346                 dev_info->max_rx_queues = 2;
2347                 dev_info->max_tx_queues = 2;
2348                 break;
2349         case e1000_vfadapt_i350:
2350                 dev_info->max_rx_queues = 1;
2351                 dev_info->max_tx_queues = 1;
2352                 break;
2353         default:
2354                 /* Should not happen */
2355                 return -EINVAL;
2356         }
2357
2358         dev_info->rx_queue_offload_capa = igb_get_rx_queue_offloads_capa(dev);
2359         dev_info->rx_offload_capa = igb_get_rx_port_offloads_capa(dev) |
2360                                     dev_info->rx_queue_offload_capa;
2361         dev_info->tx_queue_offload_capa = igb_get_tx_queue_offloads_capa(dev);
2362         dev_info->tx_offload_capa = igb_get_tx_port_offloads_capa(dev) |
2363                                     dev_info->tx_queue_offload_capa;
2364
2365         dev_info->default_rxconf = (struct rte_eth_rxconf) {
2366                 .rx_thresh = {
2367                         .pthresh = IGB_DEFAULT_RX_PTHRESH,
2368                         .hthresh = IGB_DEFAULT_RX_HTHRESH,
2369                         .wthresh = IGB_DEFAULT_RX_WTHRESH,
2370                 },
2371                 .rx_free_thresh = IGB_DEFAULT_RX_FREE_THRESH,
2372                 .rx_drop_en = 0,
2373                 .offloads = 0,
2374         };
2375
2376         dev_info->default_txconf = (struct rte_eth_txconf) {
2377                 .tx_thresh = {
2378                         .pthresh = IGB_DEFAULT_TX_PTHRESH,
2379                         .hthresh = IGB_DEFAULT_TX_HTHRESH,
2380                         .wthresh = IGB_DEFAULT_TX_WTHRESH,
2381                 },
2382                 .offloads = 0,
2383         };
2384
2385         dev_info->rx_desc_lim = rx_desc_lim;
2386         dev_info->tx_desc_lim = tx_desc_lim;
2387
2388         return 0;
2389 }
2390
2391 /* return 0 means link status changed, -1 means not changed */
2392 static int
2393 eth_igb_link_update(struct rte_eth_dev *dev, int wait_to_complete)
2394 {
2395         struct e1000_hw *hw =
2396                 E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2397         struct rte_eth_link link;
2398         int link_check, count;
2399
2400         link_check = 0;
2401         hw->mac.get_link_status = 1;
2402
2403         /* possible wait-to-complete in up to 9 seconds */
2404         for (count = 0; count < IGB_LINK_UPDATE_CHECK_TIMEOUT; count ++) {
2405                 /* Read the real link status */
2406                 switch (hw->phy.media_type) {
2407                 case e1000_media_type_copper:
2408                         /* Do the work to read phy */
2409                         e1000_check_for_link(hw);
2410                         link_check = !hw->mac.get_link_status;
2411                         break;
2412
2413                 case e1000_media_type_fiber:
2414                         e1000_check_for_link(hw);
2415                         link_check = (E1000_READ_REG(hw, E1000_STATUS) &
2416                                       E1000_STATUS_LU);
2417                         break;
2418
2419                 case e1000_media_type_internal_serdes:
2420                         e1000_check_for_link(hw);
2421                         link_check = hw->mac.serdes_has_link;
2422                         break;
2423
2424                 /* VF device is type_unknown */
2425                 case e1000_media_type_unknown:
2426                         eth_igbvf_link_update(hw);
2427                         link_check = !hw->mac.get_link_status;
2428                         break;
2429
2430                 default:
2431                         break;
2432                 }
2433                 if (link_check || wait_to_complete == 0)
2434                         break;
2435                 rte_delay_ms(IGB_LINK_UPDATE_CHECK_INTERVAL);
2436         }
2437         memset(&link, 0, sizeof(link));
2438
2439         /* Now we check if a transition has happened */
2440         if (link_check) {
2441                 uint16_t duplex, speed;
2442                 hw->mac.ops.get_link_up_info(hw, &speed, &duplex);
2443                 link.link_duplex = (duplex == FULL_DUPLEX) ?
2444                                 ETH_LINK_FULL_DUPLEX :
2445                                 ETH_LINK_HALF_DUPLEX;
2446                 link.link_speed = speed;
2447                 link.link_status = ETH_LINK_UP;
2448                 link.link_autoneg = !(dev->data->dev_conf.link_speeds &
2449                                 ETH_LINK_SPEED_FIXED);
2450         } else if (!link_check) {
2451                 link.link_speed = 0;
2452                 link.link_duplex = ETH_LINK_HALF_DUPLEX;
2453                 link.link_status = ETH_LINK_DOWN;
2454                 link.link_autoneg = ETH_LINK_FIXED;
2455         }
2456
2457         return rte_eth_linkstatus_set(dev, &link);
2458 }
2459
2460 /*
2461  * igb_hw_control_acquire sets CTRL_EXT:DRV_LOAD bit.
2462  * For ASF and Pass Through versions of f/w this means
2463  * that the driver is loaded.
2464  */
2465 static void
2466 igb_hw_control_acquire(struct e1000_hw *hw)
2467 {
2468         uint32_t ctrl_ext;
2469
2470         /* Let firmware know the driver has taken over */
2471         ctrl_ext = E1000_READ_REG(hw, E1000_CTRL_EXT);
2472         E1000_WRITE_REG(hw, E1000_CTRL_EXT, ctrl_ext | E1000_CTRL_EXT_DRV_LOAD);
2473 }
2474
2475 /*
2476  * igb_hw_control_release resets CTRL_EXT:DRV_LOAD bit.
2477  * For ASF and Pass Through versions of f/w this means that the
2478  * driver is no longer loaded.
2479  */
2480 static void
2481 igb_hw_control_release(struct e1000_hw *hw)
2482 {
2483         uint32_t ctrl_ext;
2484
2485         /* Let firmware taken over control of h/w */
2486         ctrl_ext = E1000_READ_REG(hw, E1000_CTRL_EXT);
2487         E1000_WRITE_REG(hw, E1000_CTRL_EXT,
2488                         ctrl_ext & ~E1000_CTRL_EXT_DRV_LOAD);
2489 }
2490
2491 /*
2492  * Bit of a misnomer, what this really means is
2493  * to enable OS management of the system... aka
2494  * to disable special hardware management features.
2495  */
2496 static void
2497 igb_init_manageability(struct e1000_hw *hw)
2498 {
2499         if (e1000_enable_mng_pass_thru(hw)) {
2500                 uint32_t manc2h = E1000_READ_REG(hw, E1000_MANC2H);
2501                 uint32_t manc = E1000_READ_REG(hw, E1000_MANC);
2502
2503                 /* disable hardware interception of ARP */
2504                 manc &= ~(E1000_MANC_ARP_EN);
2505
2506                 /* enable receiving management packets to the host */
2507                 manc |= E1000_MANC_EN_MNG2HOST;
2508                 manc2h |= 1 << 5;  /* Mng Port 623 */
2509                 manc2h |= 1 << 6;  /* Mng Port 664 */
2510                 E1000_WRITE_REG(hw, E1000_MANC2H, manc2h);
2511                 E1000_WRITE_REG(hw, E1000_MANC, manc);
2512         }
2513 }
2514
2515 static void
2516 igb_release_manageability(struct e1000_hw *hw)
2517 {
2518         if (e1000_enable_mng_pass_thru(hw)) {
2519                 uint32_t manc = E1000_READ_REG(hw, E1000_MANC);
2520
2521                 manc |= E1000_MANC_ARP_EN;
2522                 manc &= ~E1000_MANC_EN_MNG2HOST;
2523
2524                 E1000_WRITE_REG(hw, E1000_MANC, manc);
2525         }
2526 }
2527
2528 static int
2529 eth_igb_promiscuous_enable(struct rte_eth_dev *dev)
2530 {
2531         struct e1000_hw *hw =
2532                 E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2533         uint32_t rctl;
2534
2535         rctl = E1000_READ_REG(hw, E1000_RCTL);
2536         rctl |= (E1000_RCTL_UPE | E1000_RCTL_MPE);
2537         E1000_WRITE_REG(hw, E1000_RCTL, rctl);
2538
2539         return 0;
2540 }
2541
2542 static int
2543 eth_igb_promiscuous_disable(struct rte_eth_dev *dev)
2544 {
2545         struct e1000_hw *hw =
2546                 E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2547         uint32_t rctl;
2548
2549         rctl = E1000_READ_REG(hw, E1000_RCTL);
2550         rctl &= (~E1000_RCTL_UPE);
2551         if (dev->data->all_multicast == 1)
2552                 rctl |= E1000_RCTL_MPE;
2553         else
2554                 rctl &= (~E1000_RCTL_MPE);
2555         E1000_WRITE_REG(hw, E1000_RCTL, rctl);
2556
2557         return 0;
2558 }
2559
2560 static int
2561 eth_igb_allmulticast_enable(struct rte_eth_dev *dev)
2562 {
2563         struct e1000_hw *hw =
2564                 E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2565         uint32_t rctl;
2566
2567         rctl = E1000_READ_REG(hw, E1000_RCTL);
2568         rctl |= E1000_RCTL_MPE;
2569         E1000_WRITE_REG(hw, E1000_RCTL, rctl);
2570
2571         return 0;
2572 }
2573
2574 static int
2575 eth_igb_allmulticast_disable(struct rte_eth_dev *dev)
2576 {
2577         struct e1000_hw *hw =
2578                 E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2579         uint32_t rctl;
2580
2581         if (dev->data->promiscuous == 1)
2582                 return 0; /* must remain in all_multicast mode */
2583         rctl = E1000_READ_REG(hw, E1000_RCTL);
2584         rctl &= (~E1000_RCTL_MPE);
2585         E1000_WRITE_REG(hw, E1000_RCTL, rctl);
2586
2587         return 0;
2588 }
2589
2590 static int
2591 eth_igb_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
2592 {
2593         struct e1000_hw *hw =
2594                 E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2595         struct e1000_vfta * shadow_vfta =
2596                 E1000_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
2597         uint32_t vfta;
2598         uint32_t vid_idx;
2599         uint32_t vid_bit;
2600
2601         vid_idx = (uint32_t) ((vlan_id >> E1000_VFTA_ENTRY_SHIFT) &
2602                               E1000_VFTA_ENTRY_MASK);
2603         vid_bit = (uint32_t) (1 << (vlan_id & E1000_VFTA_ENTRY_BIT_SHIFT_MASK));
2604         vfta = E1000_READ_REG_ARRAY(hw, E1000_VFTA, vid_idx);
2605         if (on)
2606                 vfta |= vid_bit;
2607         else
2608                 vfta &= ~vid_bit;
2609         E1000_WRITE_REG_ARRAY(hw, E1000_VFTA, vid_idx, vfta);
2610
2611         /* update local VFTA copy */
2612         shadow_vfta->vfta[vid_idx] = vfta;
2613
2614         return 0;
2615 }
2616
2617 static int
2618 eth_igb_vlan_tpid_set(struct rte_eth_dev *dev,
2619                       enum rte_vlan_type vlan_type,
2620                       uint16_t tpid)
2621 {
2622         struct e1000_hw *hw =
2623                 E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2624         uint32_t reg, qinq;
2625
2626         qinq = E1000_READ_REG(hw, E1000_CTRL_EXT);
2627         qinq &= E1000_CTRL_EXT_EXT_VLAN;
2628
2629         /* only outer TPID of double VLAN can be configured*/
2630         if (qinq && vlan_type == ETH_VLAN_TYPE_OUTER) {
2631                 reg = E1000_READ_REG(hw, E1000_VET);
2632                 reg = (reg & (~E1000_VET_VET_EXT)) |
2633                         ((uint32_t)tpid << E1000_VET_VET_EXT_SHIFT);
2634                 E1000_WRITE_REG(hw, E1000_VET, reg);
2635
2636                 return 0;
2637         }
2638
2639         /* all other TPID values are read-only*/
2640         PMD_DRV_LOG(ERR, "Not supported");
2641
2642         return -ENOTSUP;
2643 }
2644
2645 static void
2646 igb_vlan_hw_filter_disable(struct rte_eth_dev *dev)
2647 {
2648         struct e1000_hw *hw =
2649                 E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2650         uint32_t reg;
2651
2652         /* Filter Table Disable */
2653         reg = E1000_READ_REG(hw, E1000_RCTL);
2654         reg &= ~E1000_RCTL_CFIEN;
2655         reg &= ~E1000_RCTL_VFE;
2656         E1000_WRITE_REG(hw, E1000_RCTL, reg);
2657 }
2658
2659 static void
2660 igb_vlan_hw_filter_enable(struct rte_eth_dev *dev)
2661 {
2662         struct e1000_hw *hw =
2663                 E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2664         struct e1000_vfta * shadow_vfta =
2665                 E1000_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
2666         uint32_t reg;
2667         int i;
2668
2669         /* Filter Table Enable, CFI not used for packet acceptance */
2670         reg = E1000_READ_REG(hw, E1000_RCTL);
2671         reg &= ~E1000_RCTL_CFIEN;
2672         reg |= E1000_RCTL_VFE;
2673         E1000_WRITE_REG(hw, E1000_RCTL, reg);
2674
2675         /* restore VFTA table */
2676         for (i = 0; i < IGB_VFTA_SIZE; i++)
2677                 E1000_WRITE_REG_ARRAY(hw, E1000_VFTA, i, shadow_vfta->vfta[i]);
2678 }
2679
2680 static void
2681 igb_vlan_hw_strip_disable(struct rte_eth_dev *dev)
2682 {
2683         struct e1000_hw *hw =
2684                 E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2685         uint32_t reg;
2686
2687         /* VLAN Mode Disable */
2688         reg = E1000_READ_REG(hw, E1000_CTRL);
2689         reg &= ~E1000_CTRL_VME;
2690         E1000_WRITE_REG(hw, E1000_CTRL, reg);
2691 }
2692
2693 static void
2694 igb_vlan_hw_strip_enable(struct rte_eth_dev *dev)
2695 {
2696         struct e1000_hw *hw =
2697                 E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2698         uint32_t reg;
2699
2700         /* VLAN Mode Enable */
2701         reg = E1000_READ_REG(hw, E1000_CTRL);
2702         reg |= E1000_CTRL_VME;
2703         E1000_WRITE_REG(hw, E1000_CTRL, reg);
2704 }
2705
2706 static void
2707 igb_vlan_hw_extend_disable(struct rte_eth_dev *dev)
2708 {
2709         struct e1000_hw *hw =
2710                 E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2711         uint32_t reg;
2712
2713         /* CTRL_EXT: Extended VLAN */
2714         reg = E1000_READ_REG(hw, E1000_CTRL_EXT);
2715         reg &= ~E1000_CTRL_EXT_EXTEND_VLAN;
2716         E1000_WRITE_REG(hw, E1000_CTRL_EXT, reg);
2717
2718         /* Update maximum packet length */
2719         if (dev->data->dev_conf.rxmode.offloads & DEV_RX_OFFLOAD_JUMBO_FRAME)
2720                 E1000_WRITE_REG(hw, E1000_RLPML,
2721                         dev->data->dev_conf.rxmode.max_rx_pkt_len +
2722                                                 VLAN_TAG_SIZE);
2723 }
2724
2725 static void
2726 igb_vlan_hw_extend_enable(struct rte_eth_dev *dev)
2727 {
2728         struct e1000_hw *hw =
2729                 E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2730         uint32_t reg;
2731
2732         /* CTRL_EXT: Extended VLAN */
2733         reg = E1000_READ_REG(hw, E1000_CTRL_EXT);
2734         reg |= E1000_CTRL_EXT_EXTEND_VLAN;
2735         E1000_WRITE_REG(hw, E1000_CTRL_EXT, reg);
2736
2737         /* Update maximum packet length */
2738         if (dev->data->dev_conf.rxmode.offloads & DEV_RX_OFFLOAD_JUMBO_FRAME)
2739                 E1000_WRITE_REG(hw, E1000_RLPML,
2740                         dev->data->dev_conf.rxmode.max_rx_pkt_len +
2741                                                 2 * VLAN_TAG_SIZE);
2742 }
2743
2744 static int
2745 eth_igb_vlan_offload_set(struct rte_eth_dev *dev, int mask)
2746 {
2747         struct rte_eth_rxmode *rxmode;
2748
2749         rxmode = &dev->data->dev_conf.rxmode;
2750         if(mask & ETH_VLAN_STRIP_MASK){
2751                 if (rxmode->offloads & DEV_RX_OFFLOAD_VLAN_STRIP)
2752                         igb_vlan_hw_strip_enable(dev);
2753                 else
2754                         igb_vlan_hw_strip_disable(dev);
2755         }
2756
2757         if(mask & ETH_VLAN_FILTER_MASK){
2758                 if (rxmode->offloads & DEV_RX_OFFLOAD_VLAN_FILTER)
2759                         igb_vlan_hw_filter_enable(dev);
2760                 else
2761                         igb_vlan_hw_filter_disable(dev);
2762         }
2763
2764         if(mask & ETH_VLAN_EXTEND_MASK){
2765                 if (rxmode->offloads & DEV_RX_OFFLOAD_VLAN_EXTEND)
2766                         igb_vlan_hw_extend_enable(dev);
2767                 else
2768                         igb_vlan_hw_extend_disable(dev);
2769         }
2770
2771         return 0;
2772 }
2773
2774
2775 /**
2776  * It enables the interrupt mask and then enable the interrupt.
2777  *
2778  * @param dev
2779  *  Pointer to struct rte_eth_dev.
2780  * @param on
2781  *  Enable or Disable
2782  *
2783  * @return
2784  *  - On success, zero.
2785  *  - On failure, a negative value.
2786  */
2787 static int
2788 eth_igb_lsc_interrupt_setup(struct rte_eth_dev *dev, uint8_t on)
2789 {
2790         struct e1000_interrupt *intr =
2791                 E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
2792
2793         if (on)
2794                 intr->mask |= E1000_ICR_LSC;
2795         else
2796                 intr->mask &= ~E1000_ICR_LSC;
2797
2798         return 0;
2799 }
2800
2801 /* It clears the interrupt causes and enables the interrupt.
2802  * It will be called once only during nic initialized.
2803  *
2804  * @param dev
2805  *  Pointer to struct rte_eth_dev.
2806  *
2807  * @return
2808  *  - On success, zero.
2809  *  - On failure, a negative value.
2810  */
2811 static int eth_igb_rxq_interrupt_setup(struct rte_eth_dev *dev)
2812 {
2813         uint32_t mask, regval;
2814         int ret;
2815         struct e1000_hw *hw =
2816                 E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2817         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
2818         struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
2819         int misc_shift = rte_intr_allow_others(intr_handle) ? 1 : 0;
2820         struct rte_eth_dev_info dev_info;
2821
2822         memset(&dev_info, 0, sizeof(dev_info));
2823         ret = eth_igb_infos_get(dev, &dev_info);
2824         if (ret != 0)
2825                 return ret;
2826
2827         mask = (0xFFFFFFFF >> (32 - dev_info.max_rx_queues)) << misc_shift;
2828         regval = E1000_READ_REG(hw, E1000_EIMS);
2829         E1000_WRITE_REG(hw, E1000_EIMS, regval | mask);
2830
2831         return 0;
2832 }
2833
2834 /*
2835  * It reads ICR and gets interrupt causes, check it and set a bit flag
2836  * to update link status.
2837  *
2838  * @param dev
2839  *  Pointer to struct rte_eth_dev.
2840  *
2841  * @return
2842  *  - On success, zero.
2843  *  - On failure, a negative value.
2844  */
2845 static int
2846 eth_igb_interrupt_get_status(struct rte_eth_dev *dev)
2847 {
2848         uint32_t icr;
2849         struct e1000_hw *hw =
2850                 E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2851         struct e1000_interrupt *intr =
2852                 E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
2853
2854         igb_intr_disable(dev);
2855
2856         /* read-on-clear nic registers here */
2857         icr = E1000_READ_REG(hw, E1000_ICR);
2858
2859         intr->flags = 0;
2860         if (icr & E1000_ICR_LSC) {
2861                 intr->flags |= E1000_FLAG_NEED_LINK_UPDATE;
2862         }
2863
2864         if (icr & E1000_ICR_VMMB)
2865                 intr->flags |= E1000_FLAG_MAILBOX;
2866
2867         return 0;
2868 }
2869
2870 /*
2871  * It executes link_update after knowing an interrupt is prsent.
2872  *
2873  * @param dev
2874  *  Pointer to struct rte_eth_dev.
2875  *
2876  * @return
2877  *  - On success, zero.
2878  *  - On failure, a negative value.
2879  */
2880 static int
2881 eth_igb_interrupt_action(struct rte_eth_dev *dev,
2882                          struct rte_intr_handle *intr_handle)
2883 {
2884         struct e1000_hw *hw =
2885                 E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2886         struct e1000_interrupt *intr =
2887                 E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
2888         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
2889         struct rte_eth_link link;
2890         int ret;
2891
2892         if (intr->flags & E1000_FLAG_MAILBOX) {
2893                 igb_pf_mbx_process(dev);
2894                 intr->flags &= ~E1000_FLAG_MAILBOX;
2895         }
2896
2897         igb_intr_enable(dev);
2898         rte_intr_ack(intr_handle);
2899
2900         if (intr->flags & E1000_FLAG_NEED_LINK_UPDATE) {
2901                 intr->flags &= ~E1000_FLAG_NEED_LINK_UPDATE;
2902
2903                 /* set get_link_status to check register later */
2904                 hw->mac.get_link_status = 1;
2905                 ret = eth_igb_link_update(dev, 0);
2906
2907                 /* check if link has changed */
2908                 if (ret < 0)
2909                         return 0;
2910
2911                 rte_eth_linkstatus_get(dev, &link);
2912                 if (link.link_status) {
2913                         PMD_INIT_LOG(INFO,
2914                                      " Port %d: Link Up - speed %u Mbps - %s",
2915                                      dev->data->port_id,
2916                                      (unsigned)link.link_speed,
2917                                      link.link_duplex == ETH_LINK_FULL_DUPLEX ?
2918                                      "full-duplex" : "half-duplex");
2919                 } else {
2920                         PMD_INIT_LOG(INFO, " Port %d: Link Down",
2921                                      dev->data->port_id);
2922                 }
2923
2924                 PMD_INIT_LOG(DEBUG, "PCI Address: %04d:%02d:%02d:%d",
2925                              pci_dev->addr.domain,
2926                              pci_dev->addr.bus,
2927                              pci_dev->addr.devid,
2928                              pci_dev->addr.function);
2929                 _rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_LSC,
2930                                               NULL);
2931         }
2932
2933         return 0;
2934 }
2935
2936 /**
2937  * Interrupt handler which shall be registered at first.
2938  *
2939  * @param handle
2940  *  Pointer to interrupt handle.
2941  * @param param
2942  *  The address of parameter (struct rte_eth_dev *) regsitered before.
2943  *
2944  * @return
2945  *  void
2946  */
2947 static void
2948 eth_igb_interrupt_handler(void *param)
2949 {
2950         struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
2951
2952         eth_igb_interrupt_get_status(dev);
2953         eth_igb_interrupt_action(dev, dev->intr_handle);
2954 }
2955
2956 static int
2957 eth_igbvf_interrupt_get_status(struct rte_eth_dev *dev)
2958 {
2959         uint32_t eicr;
2960         struct e1000_hw *hw =
2961                 E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2962         struct e1000_interrupt *intr =
2963                 E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
2964
2965         igbvf_intr_disable(hw);
2966
2967         /* read-on-clear nic registers here */
2968         eicr = E1000_READ_REG(hw, E1000_EICR);
2969         intr->flags = 0;
2970
2971         if (eicr == E1000_VTIVAR_MISC_MAILBOX)
2972                 intr->flags |= E1000_FLAG_MAILBOX;
2973
2974         return 0;
2975 }
2976
2977 void igbvf_mbx_process(struct rte_eth_dev *dev)
2978 {
2979         struct e1000_hw *hw =
2980                 E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2981         struct e1000_mbx_info *mbx = &hw->mbx;
2982         u32 in_msg = 0;
2983
2984         /* peek the message first */
2985         in_msg = E1000_READ_REG(hw, E1000_VMBMEM(0));
2986
2987         /* PF reset VF event */
2988         if (in_msg == E1000_PF_CONTROL_MSG) {
2989                 /* dummy mbx read to ack pf */
2990                 if (mbx->ops.read(hw, &in_msg, 1, 0))
2991                         return;
2992                 _rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_RESET,
2993                                               NULL);
2994         }
2995 }
2996
2997 static int
2998 eth_igbvf_interrupt_action(struct rte_eth_dev *dev, struct rte_intr_handle *intr_handle)
2999 {
3000         struct e1000_interrupt *intr =
3001                 E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
3002
3003         if (intr->flags & E1000_FLAG_MAILBOX) {
3004                 igbvf_mbx_process(dev);
3005                 intr->flags &= ~E1000_FLAG_MAILBOX;
3006         }
3007
3008         igbvf_intr_enable(dev);
3009         rte_intr_ack(intr_handle);
3010
3011         return 0;
3012 }
3013
3014 static void
3015 eth_igbvf_interrupt_handler(void *param)
3016 {
3017         struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
3018
3019         eth_igbvf_interrupt_get_status(dev);
3020         eth_igbvf_interrupt_action(dev, dev->intr_handle);
3021 }
3022
3023 static int
3024 eth_igb_led_on(struct rte_eth_dev *dev)
3025 {
3026         struct e1000_hw *hw;
3027
3028         hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3029         return e1000_led_on(hw) == E1000_SUCCESS ? 0 : -ENOTSUP;
3030 }
3031
3032 static int
3033 eth_igb_led_off(struct rte_eth_dev *dev)
3034 {
3035         struct e1000_hw *hw;
3036
3037         hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3038         return e1000_led_off(hw) == E1000_SUCCESS ? 0 : -ENOTSUP;
3039 }
3040
3041 static int
3042 eth_igb_flow_ctrl_get(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
3043 {
3044         struct e1000_hw *hw;
3045         uint32_t ctrl;
3046         int tx_pause;
3047         int rx_pause;
3048
3049         hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3050         fc_conf->pause_time = hw->fc.pause_time;
3051         fc_conf->high_water = hw->fc.high_water;
3052         fc_conf->low_water = hw->fc.low_water;
3053         fc_conf->send_xon = hw->fc.send_xon;
3054         fc_conf->autoneg = hw->mac.autoneg;
3055
3056         /*
3057          * Return rx_pause and tx_pause status according to actual setting of
3058          * the TFCE and RFCE bits in the CTRL register.
3059          */
3060         ctrl = E1000_READ_REG(hw, E1000_CTRL);
3061         if (ctrl & E1000_CTRL_TFCE)
3062                 tx_pause = 1;
3063         else
3064                 tx_pause = 0;
3065
3066         if (ctrl & E1000_CTRL_RFCE)
3067                 rx_pause = 1;
3068         else
3069                 rx_pause = 0;
3070
3071         if (rx_pause && tx_pause)
3072                 fc_conf->mode = RTE_FC_FULL;
3073         else if (rx_pause)
3074                 fc_conf->mode = RTE_FC_RX_PAUSE;
3075         else if (tx_pause)
3076                 fc_conf->mode = RTE_FC_TX_PAUSE;
3077         else
3078                 fc_conf->mode = RTE_FC_NONE;
3079
3080         return 0;
3081 }
3082
3083 static int
3084 eth_igb_flow_ctrl_set(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
3085 {
3086         struct e1000_hw *hw;
3087         int err;
3088         enum e1000_fc_mode rte_fcmode_2_e1000_fcmode[] = {
3089                 e1000_fc_none,
3090                 e1000_fc_rx_pause,
3091                 e1000_fc_tx_pause,
3092                 e1000_fc_full
3093         };
3094         uint32_t rx_buf_size;
3095         uint32_t max_high_water;
3096         uint32_t rctl;
3097
3098         hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3099         if (fc_conf->autoneg != hw->mac.autoneg)
3100                 return -ENOTSUP;
3101         rx_buf_size = igb_get_rx_buffer_size(hw);
3102         PMD_INIT_LOG(DEBUG, "Rx packet buffer size = 0x%x", rx_buf_size);
3103
3104         /* At least reserve one Ethernet frame for watermark */
3105         max_high_water = rx_buf_size - RTE_ETHER_MAX_LEN;
3106         if ((fc_conf->high_water > max_high_water) ||
3107             (fc_conf->high_water < fc_conf->low_water)) {
3108                 PMD_INIT_LOG(ERR, "e1000 incorrect high/low water value");
3109                 PMD_INIT_LOG(ERR, "high water must <=  0x%x", max_high_water);
3110                 return -EINVAL;
3111         }
3112
3113         hw->fc.requested_mode = rte_fcmode_2_e1000_fcmode[fc_conf->mode];
3114         hw->fc.pause_time     = fc_conf->pause_time;
3115         hw->fc.high_water     = fc_conf->high_water;
3116         hw->fc.low_water      = fc_conf->low_water;
3117         hw->fc.send_xon       = fc_conf->send_xon;
3118
3119         err = e1000_setup_link_generic(hw);
3120         if (err == E1000_SUCCESS) {
3121
3122                 /* check if we want to forward MAC frames - driver doesn't have native
3123                  * capability to do that, so we'll write the registers ourselves */
3124
3125                 rctl = E1000_READ_REG(hw, E1000_RCTL);
3126
3127                 /* set or clear MFLCN.PMCF bit depending on configuration */
3128                 if (fc_conf->mac_ctrl_frame_fwd != 0)
3129                         rctl |= E1000_RCTL_PMCF;
3130                 else
3131                         rctl &= ~E1000_RCTL_PMCF;
3132
3133                 E1000_WRITE_REG(hw, E1000_RCTL, rctl);
3134                 E1000_WRITE_FLUSH(hw);
3135
3136                 return 0;
3137         }
3138
3139         PMD_INIT_LOG(ERR, "e1000_setup_link_generic = 0x%x", err);
3140         return -EIO;
3141 }
3142
3143 #define E1000_RAH_POOLSEL_SHIFT      (18)
3144 static int
3145 eth_igb_rar_set(struct rte_eth_dev *dev, struct rte_ether_addr *mac_addr,
3146                 uint32_t index, uint32_t pool)
3147 {
3148         struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3149         uint32_t rah;
3150
3151         e1000_rar_set(hw, mac_addr->addr_bytes, index);
3152         rah = E1000_READ_REG(hw, E1000_RAH(index));
3153         rah |= (0x1 << (E1000_RAH_POOLSEL_SHIFT + pool));
3154         E1000_WRITE_REG(hw, E1000_RAH(index), rah);
3155         return 0;
3156 }
3157
3158 static void
3159 eth_igb_rar_clear(struct rte_eth_dev *dev, uint32_t index)
3160 {
3161         uint8_t addr[RTE_ETHER_ADDR_LEN];
3162         struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3163
3164         memset(addr, 0, sizeof(addr));
3165
3166         e1000_rar_set(hw, addr, index);
3167 }
3168
3169 static int
3170 eth_igb_default_mac_addr_set(struct rte_eth_dev *dev,
3171                                 struct rte_ether_addr *addr)
3172 {
3173         eth_igb_rar_clear(dev, 0);
3174         eth_igb_rar_set(dev, (void *)addr, 0, 0);
3175
3176         return 0;
3177 }
3178 /*
3179  * Virtual Function operations
3180  */
3181 static void
3182 igbvf_intr_disable(struct e1000_hw *hw)
3183 {
3184         PMD_INIT_FUNC_TRACE();
3185
3186         /* Clear interrupt mask to stop from interrupts being generated */
3187         E1000_WRITE_REG(hw, E1000_EIMC, 0xFFFF);
3188
3189         E1000_WRITE_FLUSH(hw);
3190 }
3191
3192 static void
3193 igbvf_stop_adapter(struct rte_eth_dev *dev)
3194 {
3195         u32 reg_val;
3196         u16 i;
3197         struct rte_eth_dev_info dev_info;
3198         struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3199         int ret;
3200
3201         memset(&dev_info, 0, sizeof(dev_info));
3202         ret = eth_igbvf_infos_get(dev, &dev_info);
3203         if (ret != 0)
3204                 return;
3205
3206         /* Clear interrupt mask to stop from interrupts being generated */
3207         igbvf_intr_disable(hw);
3208
3209         /* Clear any pending interrupts, flush previous writes */
3210         E1000_READ_REG(hw, E1000_EICR);
3211
3212         /* Disable the transmit unit.  Each queue must be disabled. */
3213         for (i = 0; i < dev_info.max_tx_queues; i++)
3214                 E1000_WRITE_REG(hw, E1000_TXDCTL(i), E1000_TXDCTL_SWFLSH);
3215
3216         /* Disable the receive unit by stopping each queue */
3217         for (i = 0; i < dev_info.max_rx_queues; i++) {
3218                 reg_val = E1000_READ_REG(hw, E1000_RXDCTL(i));
3219                 reg_val &= ~E1000_RXDCTL_QUEUE_ENABLE;
3220                 E1000_WRITE_REG(hw, E1000_RXDCTL(i), reg_val);
3221                 while (E1000_READ_REG(hw, E1000_RXDCTL(i)) & E1000_RXDCTL_QUEUE_ENABLE)
3222                         ;
3223         }
3224
3225         /* flush all queues disables */
3226         E1000_WRITE_FLUSH(hw);
3227         msec_delay(2);
3228 }
3229
3230 static int eth_igbvf_link_update(struct e1000_hw *hw)
3231 {
3232         struct e1000_mbx_info *mbx = &hw->mbx;
3233         struct e1000_mac_info *mac = &hw->mac;
3234         int ret_val = E1000_SUCCESS;
3235
3236         PMD_INIT_LOG(DEBUG, "e1000_check_for_link_vf");
3237
3238         /*
3239          * We only want to run this if there has been a rst asserted.
3240          * in this case that could mean a link change, device reset,
3241          * or a virtual function reset
3242          */
3243
3244         /* If we were hit with a reset or timeout drop the link */
3245         if (!e1000_check_for_rst(hw, 0) || !mbx->timeout)
3246                 mac->get_link_status = TRUE;
3247
3248         if (!mac->get_link_status)
3249                 goto out;
3250
3251         /* if link status is down no point in checking to see if pf is up */
3252         if (!(E1000_READ_REG(hw, E1000_STATUS) & E1000_STATUS_LU))
3253                 goto out;
3254
3255         /* if we passed all the tests above then the link is up and we no
3256          * longer need to check for link */
3257         mac->get_link_status = FALSE;
3258
3259 out:
3260         return ret_val;
3261 }
3262
3263
3264 static int
3265 igbvf_dev_configure(struct rte_eth_dev *dev)
3266 {
3267         struct rte_eth_conf* conf = &dev->data->dev_conf;
3268
3269         PMD_INIT_LOG(DEBUG, "Configured Virtual Function port id: %d",
3270                      dev->data->port_id);
3271
3272         /*
3273          * VF has no ability to enable/disable HW CRC
3274          * Keep the persistent behavior the same as Host PF
3275          */
3276 #ifndef RTE_LIBRTE_E1000_PF_DISABLE_STRIP_CRC
3277         if (conf->rxmode.offloads & DEV_RX_OFFLOAD_KEEP_CRC) {
3278                 PMD_INIT_LOG(NOTICE, "VF can't disable HW CRC Strip");
3279                 conf->rxmode.offloads &= ~DEV_RX_OFFLOAD_KEEP_CRC;
3280         }
3281 #else
3282         if (!(conf->rxmode.offloads & DEV_RX_OFFLOAD_KEEP_CRC)) {
3283                 PMD_INIT_LOG(NOTICE, "VF can't enable HW CRC Strip");
3284                 conf->rxmode.offloads |= DEV_RX_OFFLOAD_KEEP_CRC;
3285         }
3286 #endif
3287
3288         return 0;
3289 }
3290
3291 static int
3292 igbvf_dev_start(struct rte_eth_dev *dev)
3293 {
3294         struct e1000_hw *hw =
3295                 E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3296         struct e1000_adapter *adapter =
3297                 E1000_DEV_PRIVATE(dev->data->dev_private);
3298         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
3299         struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
3300         int ret;
3301         uint32_t intr_vector = 0;
3302
3303         PMD_INIT_FUNC_TRACE();
3304
3305         hw->mac.ops.reset_hw(hw);
3306         adapter->stopped = 0;
3307
3308         /* Set all vfta */
3309         igbvf_set_vfta_all(dev,1);
3310
3311         eth_igbvf_tx_init(dev);
3312
3313         /* This can fail when allocating mbufs for descriptor rings */
3314         ret = eth_igbvf_rx_init(dev);
3315         if (ret) {
3316                 PMD_INIT_LOG(ERR, "Unable to initialize RX hardware");
3317                 igb_dev_clear_queues(dev);
3318                 return ret;
3319         }
3320
3321         /* check and configure queue intr-vector mapping */
3322         if (rte_intr_cap_multiple(intr_handle) &&
3323             dev->data->dev_conf.intr_conf.rxq) {
3324                 intr_vector = dev->data->nb_rx_queues;
3325                 ret = rte_intr_efd_enable(intr_handle, intr_vector);
3326                 if (ret)
3327                         return ret;
3328         }
3329
3330         if (rte_intr_dp_is_en(intr_handle) && !intr_handle->intr_vec) {
3331                 intr_handle->intr_vec =
3332                         rte_zmalloc("intr_vec",
3333                                     dev->data->nb_rx_queues * sizeof(int), 0);
3334                 if (!intr_handle->intr_vec) {
3335                         PMD_INIT_LOG(ERR, "Failed to allocate %d rx_queues"
3336                                      " intr_vec", dev->data->nb_rx_queues);
3337                         return -ENOMEM;
3338                 }
3339         }
3340
3341         eth_igbvf_configure_msix_intr(dev);
3342
3343         /* enable uio/vfio intr/eventfd mapping */
3344         rte_intr_enable(intr_handle);
3345
3346         /* resume enabled intr since hw reset */
3347         igbvf_intr_enable(dev);
3348
3349         return 0;
3350 }
3351
3352 static void
3353 igbvf_dev_stop(struct rte_eth_dev *dev)
3354 {
3355         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
3356         struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
3357
3358         PMD_INIT_FUNC_TRACE();
3359
3360         igbvf_stop_adapter(dev);
3361
3362         /*
3363           * Clear what we set, but we still keep shadow_vfta to
3364           * restore after device starts
3365           */
3366         igbvf_set_vfta_all(dev,0);
3367
3368         igb_dev_clear_queues(dev);
3369
3370         /* disable intr eventfd mapping */
3371         rte_intr_disable(intr_handle);
3372
3373         /* Clean datapath event and queue/vec mapping */
3374         rte_intr_efd_disable(intr_handle);
3375         if (intr_handle->intr_vec) {
3376                 rte_free(intr_handle->intr_vec);
3377                 intr_handle->intr_vec = NULL;
3378         }
3379 }
3380
3381 static void
3382 igbvf_dev_close(struct rte_eth_dev *dev)
3383 {
3384         struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3385         struct e1000_adapter *adapter =
3386                 E1000_DEV_PRIVATE(dev->data->dev_private);
3387         struct rte_ether_addr addr;
3388
3389         PMD_INIT_FUNC_TRACE();
3390
3391         e1000_reset_hw(hw);
3392
3393         igbvf_dev_stop(dev);
3394         adapter->stopped = 1;
3395         igb_dev_free_queues(dev);
3396
3397         /**
3398          * reprogram the RAR with a zero mac address,
3399          * to ensure that the VF traffic goes to the PF
3400          * after stop, close and detach of the VF.
3401          **/
3402
3403         memset(&addr, 0, sizeof(addr));
3404         igbvf_default_mac_addr_set(dev, &addr);
3405 }
3406
3407 static int
3408 igbvf_promiscuous_enable(struct rte_eth_dev *dev)
3409 {
3410         struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3411
3412         /* Set both unicast and multicast promisc */
3413         e1000_promisc_set_vf(hw, e1000_promisc_enabled);
3414
3415         return 0;
3416 }
3417
3418 static int
3419 igbvf_promiscuous_disable(struct rte_eth_dev *dev)
3420 {
3421         struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3422
3423         /* If in allmulticast mode leave multicast promisc */
3424         if (dev->data->all_multicast == 1)
3425                 e1000_promisc_set_vf(hw, e1000_promisc_multicast);
3426         else
3427                 e1000_promisc_set_vf(hw, e1000_promisc_disabled);
3428
3429         return 0;
3430 }
3431
3432 static int
3433 igbvf_allmulticast_enable(struct rte_eth_dev *dev)
3434 {
3435         struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3436
3437         /* In promiscuous mode multicast promisc already set */
3438         if (dev->data->promiscuous == 0)
3439                 e1000_promisc_set_vf(hw, e1000_promisc_multicast);
3440
3441         return 0;
3442 }
3443
3444 static int
3445 igbvf_allmulticast_disable(struct rte_eth_dev *dev)
3446 {
3447         struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3448
3449         /* In promiscuous mode leave multicast promisc enabled */
3450         if (dev->data->promiscuous == 0)
3451                 e1000_promisc_set_vf(hw, e1000_promisc_disabled);
3452
3453         return 0;
3454 }
3455
3456 static int igbvf_set_vfta(struct e1000_hw *hw, uint16_t vid, bool on)
3457 {
3458         struct e1000_mbx_info *mbx = &hw->mbx;
3459         uint32_t msgbuf[2];
3460         s32 err;
3461
3462         /* After set vlan, vlan strip will also be enabled in igb driver*/
3463         msgbuf[0] = E1000_VF_SET_VLAN;
3464         msgbuf[1] = vid;
3465         /* Setting the 8 bit field MSG INFO to TRUE indicates "add" */
3466         if (on)
3467                 msgbuf[0] |= E1000_VF_SET_VLAN_ADD;
3468
3469         err = mbx->ops.write_posted(hw, msgbuf, 2, 0);
3470         if (err)
3471                 goto mbx_err;
3472
3473         err = mbx->ops.read_posted(hw, msgbuf, 2, 0);
3474         if (err)
3475                 goto mbx_err;
3476
3477         msgbuf[0] &= ~E1000_VT_MSGTYPE_CTS;
3478         if (msgbuf[0] == (E1000_VF_SET_VLAN | E1000_VT_MSGTYPE_NACK))
3479                 err = -EINVAL;
3480
3481 mbx_err:
3482         return err;
3483 }
3484
3485 static void igbvf_set_vfta_all(struct rte_eth_dev *dev, bool on)
3486 {
3487         struct e1000_hw *hw =
3488                 E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3489         struct e1000_vfta * shadow_vfta =
3490                 E1000_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
3491         int i = 0, j = 0, vfta = 0, mask = 1;
3492
3493         for (i = 0; i < IGB_VFTA_SIZE; i++){
3494                 vfta = shadow_vfta->vfta[i];
3495                 if(vfta){
3496                         mask = 1;
3497                         for (j = 0; j < 32; j++){
3498                                 if(vfta & mask)
3499                                         igbvf_set_vfta(hw,
3500                                                 (uint16_t)((i<<5)+j), on);
3501                                 mask<<=1;
3502                         }
3503                 }
3504         }
3505
3506 }
3507
3508 static int
3509 igbvf_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
3510 {
3511         struct e1000_hw *hw =
3512                 E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3513         struct e1000_vfta * shadow_vfta =
3514                 E1000_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
3515         uint32_t vid_idx = 0;
3516         uint32_t vid_bit = 0;
3517         int ret = 0;
3518
3519         PMD_INIT_FUNC_TRACE();
3520
3521         /*vind is not used in VF driver, set to 0, check ixgbe_set_vfta_vf*/
3522         ret = igbvf_set_vfta(hw, vlan_id, !!on);
3523         if(ret){
3524                 PMD_INIT_LOG(ERR, "Unable to set VF vlan");
3525                 return ret;
3526         }
3527         vid_idx = (uint32_t) ((vlan_id >> 5) & 0x7F);
3528         vid_bit = (uint32_t) (1 << (vlan_id & 0x1F));
3529
3530         /*Save what we set and retore it after device reset*/
3531         if (on)
3532                 shadow_vfta->vfta[vid_idx] |= vid_bit;
3533         else
3534                 shadow_vfta->vfta[vid_idx] &= ~vid_bit;
3535
3536         return 0;
3537 }
3538
3539 static int
3540 igbvf_default_mac_addr_set(struct rte_eth_dev *dev, struct rte_ether_addr *addr)
3541 {
3542         struct e1000_hw *hw =
3543                 E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3544
3545         /* index is not used by rar_set() */
3546         hw->mac.ops.rar_set(hw, (void *)addr, 0);
3547         return 0;
3548 }
3549
3550
3551 static int
3552 eth_igb_rss_reta_update(struct rte_eth_dev *dev,
3553                         struct rte_eth_rss_reta_entry64 *reta_conf,
3554                         uint16_t reta_size)
3555 {
3556         uint8_t i, j, mask;
3557         uint32_t reta, r;
3558         uint16_t idx, shift;
3559         struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3560
3561         if (reta_size != ETH_RSS_RETA_SIZE_128) {
3562                 PMD_DRV_LOG(ERR, "The size of hash lookup table configured "
3563                         "(%d) doesn't match the number hardware can supported "
3564                         "(%d)", reta_size, ETH_RSS_RETA_SIZE_128);
3565                 return -EINVAL;
3566         }
3567
3568         for (i = 0; i < reta_size; i += IGB_4_BIT_WIDTH) {
3569                 idx = i / RTE_RETA_GROUP_SIZE;
3570                 shift = i % RTE_RETA_GROUP_SIZE;
3571                 mask = (uint8_t)((reta_conf[idx].mask >> shift) &
3572                                                 IGB_4_BIT_MASK);
3573                 if (!mask)
3574                         continue;
3575                 if (mask == IGB_4_BIT_MASK)
3576                         r = 0;
3577                 else
3578                         r = E1000_READ_REG(hw, E1000_RETA(i >> 2));
3579                 for (j = 0, reta = 0; j < IGB_4_BIT_WIDTH; j++) {
3580                         if (mask & (0x1 << j))
3581                                 reta |= reta_conf[idx].reta[shift + j] <<
3582                                                         (CHAR_BIT * j);
3583                         else
3584                                 reta |= r & (IGB_8_BIT_MASK << (CHAR_BIT * j));
3585                 }
3586                 E1000_WRITE_REG(hw, E1000_RETA(i >> 2), reta);
3587         }
3588
3589         return 0;
3590 }
3591
3592 static int
3593 eth_igb_rss_reta_query(struct rte_eth_dev *dev,
3594                        struct rte_eth_rss_reta_entry64 *reta_conf,
3595                        uint16_t reta_size)
3596 {
3597         uint8_t i, j, mask;
3598         uint32_t reta;
3599         uint16_t idx, shift;
3600         struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3601
3602         if (reta_size != ETH_RSS_RETA_SIZE_128) {
3603                 PMD_DRV_LOG(ERR, "The size of hash lookup table configured "
3604                         "(%d) doesn't match the number hardware can supported "
3605                         "(%d)", reta_size, ETH_RSS_RETA_SIZE_128);
3606                 return -EINVAL;
3607         }
3608
3609         for (i = 0; i < reta_size; i += IGB_4_BIT_WIDTH) {
3610                 idx = i / RTE_RETA_GROUP_SIZE;
3611                 shift = i % RTE_RETA_GROUP_SIZE;
3612                 mask = (uint8_t)((reta_conf[idx].mask >> shift) &
3613                                                 IGB_4_BIT_MASK);
3614                 if (!mask)
3615                         continue;
3616                 reta = E1000_READ_REG(hw, E1000_RETA(i >> 2));
3617                 for (j = 0; j < IGB_4_BIT_WIDTH; j++) {
3618                         if (mask & (0x1 << j))
3619                                 reta_conf[idx].reta[shift + j] =
3620                                         ((reta >> (CHAR_BIT * j)) &
3621                                                 IGB_8_BIT_MASK);
3622                 }
3623         }
3624
3625         return 0;
3626 }
3627
3628 int
3629 eth_igb_syn_filter_set(struct rte_eth_dev *dev,
3630                         struct rte_eth_syn_filter *filter,
3631                         bool add)
3632 {
3633         struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3634         struct e1000_filter_info *filter_info =
3635                 E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
3636         uint32_t synqf, rfctl;
3637
3638         if (filter->queue >= IGB_MAX_RX_QUEUE_NUM)
3639                 return -EINVAL;
3640
3641         synqf = E1000_READ_REG(hw, E1000_SYNQF(0));
3642
3643         if (add) {
3644                 if (synqf & E1000_SYN_FILTER_ENABLE)
3645                         return -EINVAL;
3646
3647                 synqf = (uint32_t)(((filter->queue << E1000_SYN_FILTER_QUEUE_SHIFT) &
3648                         E1000_SYN_FILTER_QUEUE) | E1000_SYN_FILTER_ENABLE);
3649
3650                 rfctl = E1000_READ_REG(hw, E1000_RFCTL);
3651                 if (filter->hig_pri)
3652                         rfctl |= E1000_RFCTL_SYNQFP;
3653                 else
3654                         rfctl &= ~E1000_RFCTL_SYNQFP;
3655
3656                 E1000_WRITE_REG(hw, E1000_RFCTL, rfctl);
3657         } else {
3658                 if (!(synqf & E1000_SYN_FILTER_ENABLE))
3659                         return -ENOENT;
3660                 synqf = 0;
3661         }
3662
3663         filter_info->syn_info = synqf;
3664         E1000_WRITE_REG(hw, E1000_SYNQF(0), synqf);
3665         E1000_WRITE_FLUSH(hw);
3666         return 0;
3667 }
3668
3669 static int
3670 eth_igb_syn_filter_get(struct rte_eth_dev *dev,
3671                         struct rte_eth_syn_filter *filter)
3672 {
3673         struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3674         uint32_t synqf, rfctl;
3675
3676         synqf = E1000_READ_REG(hw, E1000_SYNQF(0));
3677         if (synqf & E1000_SYN_FILTER_ENABLE) {
3678                 rfctl = E1000_READ_REG(hw, E1000_RFCTL);
3679                 filter->hig_pri = (rfctl & E1000_RFCTL_SYNQFP) ? 1 : 0;
3680                 filter->queue = (uint8_t)((synqf & E1000_SYN_FILTER_QUEUE) >>
3681                                 E1000_SYN_FILTER_QUEUE_SHIFT);
3682                 return 0;
3683         }
3684
3685         return -ENOENT;
3686 }
3687
3688 static int
3689 eth_igb_syn_filter_handle(struct rte_eth_dev *dev,
3690                         enum rte_filter_op filter_op,
3691                         void *arg)
3692 {
3693         struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3694         int ret;
3695
3696         MAC_TYPE_FILTER_SUP(hw->mac.type);
3697
3698         if (filter_op == RTE_ETH_FILTER_NOP)
3699                 return 0;
3700
3701         if (arg == NULL) {
3702                 PMD_DRV_LOG(ERR, "arg shouldn't be NULL for operation %u",
3703                             filter_op);
3704                 return -EINVAL;
3705         }
3706
3707         switch (filter_op) {
3708         case RTE_ETH_FILTER_ADD:
3709                 ret = eth_igb_syn_filter_set(dev,
3710                                 (struct rte_eth_syn_filter *)arg,
3711                                 TRUE);
3712                 break;
3713         case RTE_ETH_FILTER_DELETE:
3714                 ret = eth_igb_syn_filter_set(dev,
3715                                 (struct rte_eth_syn_filter *)arg,
3716                                 FALSE);
3717                 break;
3718         case RTE_ETH_FILTER_GET:
3719                 ret = eth_igb_syn_filter_get(dev,
3720                                 (struct rte_eth_syn_filter *)arg);
3721                 break;
3722         default:
3723                 PMD_DRV_LOG(ERR, "unsupported operation %u", filter_op);
3724                 ret = -EINVAL;
3725                 break;
3726         }
3727
3728         return ret;
3729 }
3730
3731 /* translate elements in struct rte_eth_ntuple_filter to struct e1000_2tuple_filter_info*/
3732 static inline int
3733 ntuple_filter_to_2tuple(struct rte_eth_ntuple_filter *filter,
3734                         struct e1000_2tuple_filter_info *filter_info)
3735 {
3736         if (filter->queue >= IGB_MAX_RX_QUEUE_NUM)
3737                 return -EINVAL;
3738         if (filter->priority > E1000_2TUPLE_MAX_PRI)
3739                 return -EINVAL;  /* filter index is out of range. */
3740         if (filter->tcp_flags > RTE_NTUPLE_TCP_FLAGS_MASK)
3741                 return -EINVAL;  /* flags is invalid. */
3742
3743         switch (filter->dst_port_mask) {
3744         case UINT16_MAX:
3745                 filter_info->dst_port_mask = 0;
3746                 filter_info->dst_port = filter->dst_port;
3747                 break;
3748         case 0:
3749                 filter_info->dst_port_mask = 1;
3750                 break;
3751         default:
3752                 PMD_DRV_LOG(ERR, "invalid dst_port mask.");
3753                 return -EINVAL;
3754         }
3755
3756         switch (filter->proto_mask) {
3757         case UINT8_MAX:
3758                 filter_info->proto_mask = 0;
3759                 filter_info->proto = filter->proto;
3760                 break;
3761         case 0:
3762                 filter_info->proto_mask = 1;
3763                 break;
3764         default:
3765                 PMD_DRV_LOG(ERR, "invalid protocol mask.");
3766                 return -EINVAL;
3767         }
3768
3769         filter_info->priority = (uint8_t)filter->priority;
3770         if (filter->flags & RTE_NTUPLE_FLAGS_TCP_FLAG)
3771                 filter_info->tcp_flags = filter->tcp_flags;
3772         else
3773                 filter_info->tcp_flags = 0;
3774
3775         return 0;
3776 }
3777
3778 static inline struct e1000_2tuple_filter *
3779 igb_2tuple_filter_lookup(struct e1000_2tuple_filter_list *filter_list,
3780                         struct e1000_2tuple_filter_info *key)
3781 {
3782         struct e1000_2tuple_filter *it;
3783
3784         TAILQ_FOREACH(it, filter_list, entries) {
3785                 if (memcmp(key, &it->filter_info,
3786                         sizeof(struct e1000_2tuple_filter_info)) == 0) {
3787                         return it;
3788                 }
3789         }
3790         return NULL;
3791 }
3792
3793 /* inject a igb 2tuple filter to HW */
3794 static inline void
3795 igb_inject_2uple_filter(struct rte_eth_dev *dev,
3796                            struct e1000_2tuple_filter *filter)
3797 {
3798         struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3799         uint32_t ttqf = E1000_TTQF_DISABLE_MASK;
3800         uint32_t imir, imir_ext = E1000_IMIREXT_SIZE_BP;
3801         int i;
3802
3803         i = filter->index;
3804         imir = (uint32_t)(filter->filter_info.dst_port & E1000_IMIR_DSTPORT);
3805         if (filter->filter_info.dst_port_mask == 1) /* 1b means not compare. */
3806                 imir |= E1000_IMIR_PORT_BP;
3807         else
3808                 imir &= ~E1000_IMIR_PORT_BP;
3809
3810         imir |= filter->filter_info.priority << E1000_IMIR_PRIORITY_SHIFT;
3811
3812         ttqf |= E1000_TTQF_QUEUE_ENABLE;
3813         ttqf |= (uint32_t)(filter->queue << E1000_TTQF_QUEUE_SHIFT);
3814         ttqf |= (uint32_t)(filter->filter_info.proto &
3815                                                 E1000_TTQF_PROTOCOL_MASK);
3816         if (filter->filter_info.proto_mask == 0)
3817                 ttqf &= ~E1000_TTQF_MASK_ENABLE;
3818
3819         /* tcp flags bits setting. */
3820         if (filter->filter_info.tcp_flags & RTE_NTUPLE_TCP_FLAGS_MASK) {
3821                 if (filter->filter_info.tcp_flags & RTE_TCP_URG_FLAG)
3822                         imir_ext |= E1000_IMIREXT_CTRL_URG;
3823                 if (filter->filter_info.tcp_flags & RTE_TCP_ACK_FLAG)
3824                         imir_ext |= E1000_IMIREXT_CTRL_ACK;
3825                 if (filter->filter_info.tcp_flags & RTE_TCP_PSH_FLAG)
3826                         imir_ext |= E1000_IMIREXT_CTRL_PSH;
3827                 if (filter->filter_info.tcp_flags & RTE_TCP_RST_FLAG)
3828                         imir_ext |= E1000_IMIREXT_CTRL_RST;
3829                 if (filter->filter_info.tcp_flags & RTE_TCP_SYN_FLAG)
3830                         imir_ext |= E1000_IMIREXT_CTRL_SYN;
3831                 if (filter->filter_info.tcp_flags & RTE_TCP_FIN_FLAG)
3832                         imir_ext |= E1000_IMIREXT_CTRL_FIN;
3833         } else {
3834                 imir_ext |= E1000_IMIREXT_CTRL_BP;
3835         }
3836         E1000_WRITE_REG(hw, E1000_IMIR(i), imir);
3837         E1000_WRITE_REG(hw, E1000_TTQF(i), ttqf);
3838         E1000_WRITE_REG(hw, E1000_IMIREXT(i), imir_ext);
3839 }
3840
3841 /*
3842  * igb_add_2tuple_filter - add a 2tuple filter
3843  *
3844  * @param
3845  * dev: Pointer to struct rte_eth_dev.
3846  * ntuple_filter: ponter to the filter that will be added.
3847  *
3848  * @return
3849  *    - On success, zero.
3850  *    - On failure, a negative value.
3851  */
3852 static int
3853 igb_add_2tuple_filter(struct rte_eth_dev *dev,
3854                         struct rte_eth_ntuple_filter *ntuple_filter)
3855 {
3856         struct e1000_filter_info *filter_info =
3857                 E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
3858         struct e1000_2tuple_filter *filter;
3859         int i, ret;
3860
3861         filter = rte_zmalloc("e1000_2tuple_filter",
3862                         sizeof(struct e1000_2tuple_filter), 0);
3863         if (filter == NULL)
3864                 return -ENOMEM;
3865
3866         ret = ntuple_filter_to_2tuple(ntuple_filter,
3867                                       &filter->filter_info);
3868         if (ret < 0) {
3869                 rte_free(filter);
3870                 return ret;
3871         }
3872         if (igb_2tuple_filter_lookup(&filter_info->twotuple_list,
3873                                          &filter->filter_info) != NULL) {
3874                 PMD_DRV_LOG(ERR, "filter exists.");
3875                 rte_free(filter);
3876                 return -EEXIST;
3877         }
3878         filter->queue = ntuple_filter->queue;
3879
3880         /*
3881          * look for an unused 2tuple filter index,
3882          * and insert the filter to list.
3883          */
3884         for (i = 0; i < E1000_MAX_TTQF_FILTERS; i++) {
3885                 if (!(filter_info->twotuple_mask & (1 << i))) {
3886                         filter_info->twotuple_mask |= 1 << i;
3887                         filter->index = i;
3888                         TAILQ_INSERT_TAIL(&filter_info->twotuple_list,
3889                                           filter,
3890                                           entries);
3891                         break;
3892                 }
3893         }
3894         if (i >= E1000_MAX_TTQF_FILTERS) {
3895                 PMD_DRV_LOG(ERR, "2tuple filters are full.");
3896                 rte_free(filter);
3897                 return -ENOSYS;
3898         }
3899
3900         igb_inject_2uple_filter(dev, filter);
3901         return 0;
3902 }
3903
3904 int
3905 igb_delete_2tuple_filter(struct rte_eth_dev *dev,
3906                         struct e1000_2tuple_filter *filter)
3907 {
3908         struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3909         struct e1000_filter_info *filter_info =
3910                 E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
3911
3912         filter_info->twotuple_mask &= ~(1 << filter->index);
3913         TAILQ_REMOVE(&filter_info->twotuple_list, filter, entries);
3914         rte_free(filter);
3915
3916         E1000_WRITE_REG(hw, E1000_TTQF(filter->index), E1000_TTQF_DISABLE_MASK);
3917         E1000_WRITE_REG(hw, E1000_IMIR(filter->index), 0);
3918         E1000_WRITE_REG(hw, E1000_IMIREXT(filter->index), 0);
3919         return 0;
3920 }
3921
3922 /*
3923  * igb_remove_2tuple_filter - remove a 2tuple filter
3924  *
3925  * @param
3926  * dev: Pointer to struct rte_eth_dev.
3927  * ntuple_filter: ponter to the filter that will be removed.
3928  *
3929  * @return
3930  *    - On success, zero.
3931  *    - On failure, a negative value.
3932  */
3933 static int
3934 igb_remove_2tuple_filter(struct rte_eth_dev *dev,
3935                         struct rte_eth_ntuple_filter *ntuple_filter)
3936 {
3937         struct e1000_filter_info *filter_info =
3938                 E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
3939         struct e1000_2tuple_filter_info filter_2tuple;
3940         struct e1000_2tuple_filter *filter;
3941         int ret;
3942
3943         memset(&filter_2tuple, 0, sizeof(struct e1000_2tuple_filter_info));
3944         ret = ntuple_filter_to_2tuple(ntuple_filter,
3945                                       &filter_2tuple);
3946         if (ret < 0)
3947                 return ret;
3948
3949         filter = igb_2tuple_filter_lookup(&filter_info->twotuple_list,
3950                                          &filter_2tuple);
3951         if (filter == NULL) {
3952                 PMD_DRV_LOG(ERR, "filter doesn't exist.");
3953                 return -ENOENT;
3954         }
3955
3956         igb_delete_2tuple_filter(dev, filter);
3957
3958         return 0;
3959 }
3960
3961 /* inject a igb flex filter to HW */
3962 static inline void
3963 igb_inject_flex_filter(struct rte_eth_dev *dev,
3964                            struct e1000_flex_filter *filter)
3965 {
3966         struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3967         uint32_t wufc, queueing;
3968         uint32_t reg_off;
3969         uint8_t i, j = 0;
3970
3971         wufc = E1000_READ_REG(hw, E1000_WUFC);
3972         if (filter->index < E1000_MAX_FHFT)
3973                 reg_off = E1000_FHFT(filter->index);
3974         else
3975                 reg_off = E1000_FHFT_EXT(filter->index - E1000_MAX_FHFT);
3976
3977         E1000_WRITE_REG(hw, E1000_WUFC, wufc | E1000_WUFC_FLEX_HQ |
3978                         (E1000_WUFC_FLX0 << filter->index));
3979         queueing = filter->filter_info.len |
3980                 (filter->queue << E1000_FHFT_QUEUEING_QUEUE_SHIFT) |
3981                 (filter->filter_info.priority <<
3982                         E1000_FHFT_QUEUEING_PRIO_SHIFT);
3983         E1000_WRITE_REG(hw, reg_off + E1000_FHFT_QUEUEING_OFFSET,
3984                         queueing);
3985
3986         for (i = 0; i < E1000_FLEX_FILTERS_MASK_SIZE; i++) {
3987                 E1000_WRITE_REG(hw, reg_off,
3988                                 filter->filter_info.dwords[j]);
3989                 reg_off += sizeof(uint32_t);
3990                 E1000_WRITE_REG(hw, reg_off,
3991                                 filter->filter_info.dwords[++j]);
3992                 reg_off += sizeof(uint32_t);
3993                 E1000_WRITE_REG(hw, reg_off,
3994                         (uint32_t)filter->filter_info.mask[i]);
3995                 reg_off += sizeof(uint32_t) * 2;
3996                 ++j;
3997         }
3998 }
3999
4000 static inline struct e1000_flex_filter *
4001 eth_igb_flex_filter_lookup(struct e1000_flex_filter_list *filter_list,
4002                         struct e1000_flex_filter_info *key)
4003 {
4004         struct e1000_flex_filter *it;
4005
4006         TAILQ_FOREACH(it, filter_list, entries) {
4007                 if (memcmp(key, &it->filter_info,
4008                         sizeof(struct e1000_flex_filter_info)) == 0)
4009                         return it;
4010         }
4011
4012         return NULL;
4013 }
4014
4015 /* remove a flex byte filter
4016  * @param
4017  * dev: Pointer to struct rte_eth_dev.
4018  * filter: the pointer of the filter will be removed.
4019  */
4020 void
4021 igb_remove_flex_filter(struct rte_eth_dev *dev,
4022                         struct e1000_flex_filter *filter)
4023 {
4024         struct e1000_filter_info *filter_info =
4025                 E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
4026         struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4027         uint32_t wufc, i;
4028         uint32_t reg_off;
4029
4030         wufc = E1000_READ_REG(hw, E1000_WUFC);
4031         if (filter->index < E1000_MAX_FHFT)
4032                 reg_off = E1000_FHFT(filter->index);
4033         else
4034                 reg_off = E1000_FHFT_EXT(filter->index - E1000_MAX_FHFT);
4035
4036         for (i = 0; i < E1000_FHFT_SIZE_IN_DWD; i++)
4037                 E1000_WRITE_REG(hw, reg_off + i * sizeof(uint32_t), 0);
4038
4039         E1000_WRITE_REG(hw, E1000_WUFC, wufc &
4040                 (~(E1000_WUFC_FLX0 << filter->index)));
4041
4042         filter_info->flex_mask &= ~(1 << filter->index);
4043         TAILQ_REMOVE(&filter_info->flex_list, filter, entries);
4044         rte_free(filter);
4045 }
4046
4047 int
4048 eth_igb_add_del_flex_filter(struct rte_eth_dev *dev,
4049                         struct rte_eth_flex_filter *filter,
4050                         bool add)
4051 {
4052         struct e1000_filter_info *filter_info =
4053                 E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
4054         struct e1000_flex_filter *flex_filter, *it;
4055         uint32_t mask;
4056         uint8_t shift, i;
4057
4058         flex_filter = rte_zmalloc("e1000_flex_filter",
4059                         sizeof(struct e1000_flex_filter), 0);
4060         if (flex_filter == NULL)
4061                 return -ENOMEM;
4062
4063         flex_filter->filter_info.len = filter->len;
4064         flex_filter->filter_info.priority = filter->priority;
4065         memcpy(flex_filter->filter_info.dwords, filter->bytes, filter->len);
4066         for (i = 0; i < RTE_ALIGN(filter->len, CHAR_BIT) / CHAR_BIT; i++) {
4067                 mask = 0;
4068                 /* reverse bits in flex filter's mask*/
4069                 for (shift = 0; shift < CHAR_BIT; shift++) {
4070                         if (filter->mask[i] & (0x01 << shift))
4071                                 mask |= (0x80 >> shift);
4072                 }
4073                 flex_filter->filter_info.mask[i] = mask;
4074         }
4075
4076         it = eth_igb_flex_filter_lookup(&filter_info->flex_list,
4077                                 &flex_filter->filter_info);
4078         if (it == NULL && !add) {
4079                 PMD_DRV_LOG(ERR, "filter doesn't exist.");
4080                 rte_free(flex_filter);
4081                 return -ENOENT;
4082         }
4083         if (it != NULL && add) {
4084                 PMD_DRV_LOG(ERR, "filter exists.");
4085                 rte_free(flex_filter);
4086                 return -EEXIST;
4087         }
4088
4089         if (add) {
4090                 flex_filter->queue = filter->queue;
4091                 /*
4092                  * look for an unused flex filter index
4093                  * and insert the filter into the list.
4094                  */
4095                 for (i = 0; i < E1000_MAX_FLEX_FILTERS; i++) {
4096                         if (!(filter_info->flex_mask & (1 << i))) {
4097                                 filter_info->flex_mask |= 1 << i;
4098                                 flex_filter->index = i;
4099                                 TAILQ_INSERT_TAIL(&filter_info->flex_list,
4100                                         flex_filter,
4101                                         entries);
4102                                 break;
4103                         }
4104                 }
4105                 if (i >= E1000_MAX_FLEX_FILTERS) {
4106                         PMD_DRV_LOG(ERR, "flex filters are full.");
4107                         rte_free(flex_filter);
4108                         return -ENOSYS;
4109                 }
4110
4111                 igb_inject_flex_filter(dev, flex_filter);
4112
4113         } else {
4114                 igb_remove_flex_filter(dev, it);
4115                 rte_free(flex_filter);
4116         }
4117
4118         return 0;
4119 }
4120
4121 static int
4122 eth_igb_get_flex_filter(struct rte_eth_dev *dev,
4123                         struct rte_eth_flex_filter *filter)
4124 {
4125         struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4126         struct e1000_filter_info *filter_info =
4127                 E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
4128         struct e1000_flex_filter flex_filter, *it;
4129         uint32_t wufc, queueing, wufc_en = 0;
4130
4131         memset(&flex_filter, 0, sizeof(struct e1000_flex_filter));
4132         flex_filter.filter_info.len = filter->len;
4133         flex_filter.filter_info.priority = filter->priority;
4134         memcpy(flex_filter.filter_info.dwords, filter->bytes, filter->len);
4135         memcpy(flex_filter.filter_info.mask, filter->mask,
4136                         RTE_ALIGN(filter->len, CHAR_BIT) / CHAR_BIT);
4137
4138         it = eth_igb_flex_filter_lookup(&filter_info->flex_list,
4139                                 &flex_filter.filter_info);
4140         if (it == NULL) {
4141                 PMD_DRV_LOG(ERR, "filter doesn't exist.");
4142                 return -ENOENT;
4143         }
4144
4145         wufc = E1000_READ_REG(hw, E1000_WUFC);
4146         wufc_en = E1000_WUFC_FLEX_HQ | (E1000_WUFC_FLX0 << it->index);
4147
4148         if ((wufc & wufc_en) == wufc_en) {
4149                 uint32_t reg_off = 0;
4150                 if (it->index < E1000_MAX_FHFT)
4151                         reg_off = E1000_FHFT(it->index);
4152                 else
4153                         reg_off = E1000_FHFT_EXT(it->index - E1000_MAX_FHFT);
4154
4155                 queueing = E1000_READ_REG(hw,
4156                                 reg_off + E1000_FHFT_QUEUEING_OFFSET);
4157                 filter->len = queueing & E1000_FHFT_QUEUEING_LEN;
4158                 filter->priority = (queueing & E1000_FHFT_QUEUEING_PRIO) >>
4159                         E1000_FHFT_QUEUEING_PRIO_SHIFT;
4160                 filter->queue = (queueing & E1000_FHFT_QUEUEING_QUEUE) >>
4161                         E1000_FHFT_QUEUEING_QUEUE_SHIFT;
4162                 return 0;
4163         }
4164         return -ENOENT;
4165 }
4166
4167 static int
4168 eth_igb_flex_filter_handle(struct rte_eth_dev *dev,
4169                         enum rte_filter_op filter_op,
4170                         void *arg)
4171 {
4172         struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4173         struct rte_eth_flex_filter *filter;
4174         int ret = 0;
4175
4176         MAC_TYPE_FILTER_SUP_EXT(hw->mac.type);
4177
4178         if (filter_op == RTE_ETH_FILTER_NOP)
4179                 return ret;
4180
4181         if (arg == NULL) {
4182                 PMD_DRV_LOG(ERR, "arg shouldn't be NULL for operation %u",
4183                             filter_op);
4184                 return -EINVAL;
4185         }
4186
4187         filter = (struct rte_eth_flex_filter *)arg;
4188         if (filter->len == 0 || filter->len > E1000_MAX_FLEX_FILTER_LEN
4189             || filter->len % sizeof(uint64_t) != 0) {
4190                 PMD_DRV_LOG(ERR, "filter's length is out of range");
4191                 return -EINVAL;
4192         }
4193         if (filter->priority > E1000_MAX_FLEX_FILTER_PRI) {
4194                 PMD_DRV_LOG(ERR, "filter's priority is out of range");
4195                 return -EINVAL;
4196         }
4197
4198         switch (filter_op) {
4199         case RTE_ETH_FILTER_ADD:
4200                 ret = eth_igb_add_del_flex_filter(dev, filter, TRUE);
4201                 break;
4202         case RTE_ETH_FILTER_DELETE:
4203                 ret = eth_igb_add_del_flex_filter(dev, filter, FALSE);
4204                 break;
4205         case RTE_ETH_FILTER_GET:
4206                 ret = eth_igb_get_flex_filter(dev, filter);
4207                 break;
4208         default:
4209                 PMD_DRV_LOG(ERR, "unsupported operation %u", filter_op);
4210                 ret = -EINVAL;
4211                 break;
4212         }
4213
4214         return ret;
4215 }
4216
4217 /* translate elements in struct rte_eth_ntuple_filter to struct e1000_5tuple_filter_info*/
4218 static inline int
4219 ntuple_filter_to_5tuple_82576(struct rte_eth_ntuple_filter *filter,
4220                         struct e1000_5tuple_filter_info *filter_info)
4221 {
4222         if (filter->queue >= IGB_MAX_RX_QUEUE_NUM_82576)
4223                 return -EINVAL;
4224         if (filter->priority > E1000_2TUPLE_MAX_PRI)
4225                 return -EINVAL;  /* filter index is out of range. */
4226         if (filter->tcp_flags > RTE_NTUPLE_TCP_FLAGS_MASK)
4227                 return -EINVAL;  /* flags is invalid. */
4228
4229         switch (filter->dst_ip_mask) {
4230         case UINT32_MAX:
4231                 filter_info->dst_ip_mask = 0;
4232                 filter_info->dst_ip = filter->dst_ip;
4233                 break;
4234         case 0:
4235                 filter_info->dst_ip_mask = 1;
4236                 break;
4237         default:
4238                 PMD_DRV_LOG(ERR, "invalid dst_ip mask.");
4239                 return -EINVAL;
4240         }
4241
4242         switch (filter->src_ip_mask) {
4243         case UINT32_MAX:
4244                 filter_info->src_ip_mask = 0;
4245                 filter_info->src_ip = filter->src_ip;
4246                 break;
4247         case 0:
4248                 filter_info->src_ip_mask = 1;
4249                 break;
4250         default:
4251                 PMD_DRV_LOG(ERR, "invalid src_ip mask.");
4252                 return -EINVAL;
4253         }
4254
4255         switch (filter->dst_port_mask) {
4256         case UINT16_MAX:
4257                 filter_info->dst_port_mask = 0;
4258                 filter_info->dst_port = filter->dst_port;
4259                 break;
4260         case 0:
4261                 filter_info->dst_port_mask = 1;
4262                 break;
4263         default:
4264                 PMD_DRV_LOG(ERR, "invalid dst_port mask.");
4265                 return -EINVAL;
4266         }
4267
4268         switch (filter->src_port_mask) {
4269         case UINT16_MAX:
4270                 filter_info->src_port_mask = 0;
4271                 filter_info->src_port = filter->src_port;
4272                 break;
4273         case 0:
4274                 filter_info->src_port_mask = 1;
4275                 break;
4276         default:
4277                 PMD_DRV_LOG(ERR, "invalid src_port mask.");
4278                 return -EINVAL;
4279         }
4280
4281         switch (filter->proto_mask) {
4282         case UINT8_MAX:
4283                 filter_info->proto_mask = 0;
4284                 filter_info->proto = filter->proto;
4285                 break;
4286         case 0:
4287                 filter_info->proto_mask = 1;
4288                 break;
4289         default:
4290                 PMD_DRV_LOG(ERR, "invalid protocol mask.");
4291                 return -EINVAL;
4292         }
4293
4294         filter_info->priority = (uint8_t)filter->priority;
4295         if (filter->flags & RTE_NTUPLE_FLAGS_TCP_FLAG)
4296                 filter_info->tcp_flags = filter->tcp_flags;
4297         else
4298                 filter_info->tcp_flags = 0;
4299
4300         return 0;
4301 }
4302
4303 static inline struct e1000_5tuple_filter *
4304 igb_5tuple_filter_lookup_82576(struct e1000_5tuple_filter_list *filter_list,
4305                         struct e1000_5tuple_filter_info *key)
4306 {
4307         struct e1000_5tuple_filter *it;
4308
4309         TAILQ_FOREACH(it, filter_list, entries) {
4310                 if (memcmp(key, &it->filter_info,
4311                         sizeof(struct e1000_5tuple_filter_info)) == 0) {
4312                         return it;
4313                 }
4314         }
4315         return NULL;
4316 }
4317
4318 /* inject a igb 5-tuple filter to HW */
4319 static inline void
4320 igb_inject_5tuple_filter_82576(struct rte_eth_dev *dev,
4321                            struct e1000_5tuple_filter *filter)
4322 {
4323         struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4324         uint32_t ftqf = E1000_FTQF_VF_BP | E1000_FTQF_MASK;
4325         uint32_t spqf, imir, imir_ext = E1000_IMIREXT_SIZE_BP;
4326         uint8_t i;
4327
4328         i = filter->index;
4329         ftqf |= filter->filter_info.proto & E1000_FTQF_PROTOCOL_MASK;
4330         if (filter->filter_info.src_ip_mask == 0) /* 0b means compare. */
4331                 ftqf &= ~E1000_FTQF_MASK_SOURCE_ADDR_BP;
4332         if (filter->filter_info.dst_ip_mask == 0)
4333                 ftqf &= ~E1000_FTQF_MASK_DEST_ADDR_BP;
4334         if (filter->filter_info.src_port_mask == 0)
4335                 ftqf &= ~E1000_FTQF_MASK_SOURCE_PORT_BP;
4336         if (filter->filter_info.proto_mask == 0)
4337                 ftqf &= ~E1000_FTQF_MASK_PROTO_BP;
4338         ftqf |= (filter->queue << E1000_FTQF_QUEUE_SHIFT) &
4339                 E1000_FTQF_QUEUE_MASK;
4340         ftqf |= E1000_FTQF_QUEUE_ENABLE;
4341         E1000_WRITE_REG(hw, E1000_FTQF(i), ftqf);
4342         E1000_WRITE_REG(hw, E1000_DAQF(i), filter->filter_info.dst_ip);
4343         E1000_WRITE_REG(hw, E1000_SAQF(i), filter->filter_info.src_ip);
4344
4345         spqf = filter->filter_info.src_port & E1000_SPQF_SRCPORT;
4346         E1000_WRITE_REG(hw, E1000_SPQF(i), spqf);
4347
4348         imir = (uint32_t)(filter->filter_info.dst_port & E1000_IMIR_DSTPORT);
4349         if (filter->filter_info.dst_port_mask == 1) /* 1b means not compare. */
4350                 imir |= E1000_IMIR_PORT_BP;
4351         else
4352                 imir &= ~E1000_IMIR_PORT_BP;
4353         imir |= filter->filter_info.priority << E1000_IMIR_PRIORITY_SHIFT;
4354
4355         /* tcp flags bits setting. */
4356         if (filter->filter_info.tcp_flags & RTE_NTUPLE_TCP_FLAGS_MASK) {
4357                 if (filter->filter_info.tcp_flags & RTE_TCP_URG_FLAG)
4358                         imir_ext |= E1000_IMIREXT_CTRL_URG;
4359                 if (filter->filter_info.tcp_flags & RTE_TCP_ACK_FLAG)
4360                         imir_ext |= E1000_IMIREXT_CTRL_ACK;
4361                 if (filter->filter_info.tcp_flags & RTE_TCP_PSH_FLAG)
4362                         imir_ext |= E1000_IMIREXT_CTRL_PSH;
4363                 if (filter->filter_info.tcp_flags & RTE_TCP_RST_FLAG)
4364                         imir_ext |= E1000_IMIREXT_CTRL_RST;
4365                 if (filter->filter_info.tcp_flags & RTE_TCP_SYN_FLAG)
4366                         imir_ext |= E1000_IMIREXT_CTRL_SYN;
4367                 if (filter->filter_info.tcp_flags & RTE_TCP_FIN_FLAG)
4368                         imir_ext |= E1000_IMIREXT_CTRL_FIN;
4369         } else {
4370                 imir_ext |= E1000_IMIREXT_CTRL_BP;
4371         }
4372         E1000_WRITE_REG(hw, E1000_IMIR(i), imir);
4373         E1000_WRITE_REG(hw, E1000_IMIREXT(i), imir_ext);
4374 }
4375
4376 /*
4377  * igb_add_5tuple_filter_82576 - add a 5tuple filter
4378  *
4379  * @param
4380  * dev: Pointer to struct rte_eth_dev.
4381  * ntuple_filter: ponter to the filter that will be added.
4382  *
4383  * @return
4384  *    - On success, zero.
4385  *    - On failure, a negative value.
4386  */
4387 static int
4388 igb_add_5tuple_filter_82576(struct rte_eth_dev *dev,
4389                         struct rte_eth_ntuple_filter *ntuple_filter)
4390 {
4391         struct e1000_filter_info *filter_info =
4392                 E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
4393         struct e1000_5tuple_filter *filter;
4394         uint8_t i;
4395         int ret;
4396
4397         filter = rte_zmalloc("e1000_5tuple_filter",
4398                         sizeof(struct e1000_5tuple_filter), 0);
4399         if (filter == NULL)
4400                 return -ENOMEM;
4401
4402         ret = ntuple_filter_to_5tuple_82576(ntuple_filter,
4403                                             &filter->filter_info);
4404         if (ret < 0) {
4405                 rte_free(filter);
4406                 return ret;
4407         }
4408
4409         if (igb_5tuple_filter_lookup_82576(&filter_info->fivetuple_list,
4410                                          &filter->filter_info) != NULL) {
4411                 PMD_DRV_LOG(ERR, "filter exists.");
4412                 rte_free(filter);
4413                 return -EEXIST;
4414         }
4415         filter->queue = ntuple_filter->queue;
4416
4417         /*
4418          * look for an unused 5tuple filter index,
4419          * and insert the filter to list.
4420          */
4421         for (i = 0; i < E1000_MAX_FTQF_FILTERS; i++) {
4422                 if (!(filter_info->fivetuple_mask & (1 << i))) {
4423                         filter_info->fivetuple_mask |= 1 << i;
4424                         filter->index = i;
4425                         TAILQ_INSERT_TAIL(&filter_info->fivetuple_list,
4426                                           filter,
4427                                           entries);
4428                         break;
4429                 }
4430         }
4431         if (i >= E1000_MAX_FTQF_FILTERS) {
4432                 PMD_DRV_LOG(ERR, "5tuple filters are full.");
4433                 rte_free(filter);
4434                 return -ENOSYS;
4435         }
4436
4437         igb_inject_5tuple_filter_82576(dev, filter);
4438         return 0;
4439 }
4440
4441 int
4442 igb_delete_5tuple_filter_82576(struct rte_eth_dev *dev,
4443                                 struct e1000_5tuple_filter *filter)
4444 {
4445         struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4446         struct e1000_filter_info *filter_info =
4447                 E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
4448
4449         filter_info->fivetuple_mask &= ~(1 << filter->index);
4450         TAILQ_REMOVE(&filter_info->fivetuple_list, filter, entries);
4451         rte_free(filter);
4452
4453         E1000_WRITE_REG(hw, E1000_FTQF(filter->index),
4454                         E1000_FTQF_VF_BP | E1000_FTQF_MASK);
4455         E1000_WRITE_REG(hw, E1000_DAQF(filter->index), 0);
4456         E1000_WRITE_REG(hw, E1000_SAQF(filter->index), 0);
4457         E1000_WRITE_REG(hw, E1000_SPQF(filter->index), 0);
4458         E1000_WRITE_REG(hw, E1000_IMIR(filter->index), 0);
4459         E1000_WRITE_REG(hw, E1000_IMIREXT(filter->index), 0);
4460         return 0;
4461 }
4462
4463 /*
4464  * igb_remove_5tuple_filter_82576 - remove a 5tuple filter
4465  *
4466  * @param
4467  * dev: Pointer to struct rte_eth_dev.
4468  * ntuple_filter: ponter to the filter that will be removed.
4469  *
4470  * @return
4471  *    - On success, zero.
4472  *    - On failure, a negative value.
4473  */
4474 static int
4475 igb_remove_5tuple_filter_82576(struct rte_eth_dev *dev,
4476                                 struct rte_eth_ntuple_filter *ntuple_filter)
4477 {
4478         struct e1000_filter_info *filter_info =
4479                 E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
4480         struct e1000_5tuple_filter_info filter_5tuple;
4481         struct e1000_5tuple_filter *filter;
4482         int ret;
4483
4484         memset(&filter_5tuple, 0, sizeof(struct e1000_5tuple_filter_info));
4485         ret = ntuple_filter_to_5tuple_82576(ntuple_filter,
4486                                             &filter_5tuple);
4487         if (ret < 0)
4488                 return ret;
4489
4490         filter = igb_5tuple_filter_lookup_82576(&filter_info->fivetuple_list,
4491                                          &filter_5tuple);
4492         if (filter == NULL) {
4493                 PMD_DRV_LOG(ERR, "filter doesn't exist.");
4494                 return -ENOENT;
4495         }
4496
4497         igb_delete_5tuple_filter_82576(dev, filter);
4498
4499         return 0;
4500 }
4501
4502 static int
4503 eth_igb_mtu_set(struct rte_eth_dev *dev, uint16_t mtu)
4504 {
4505         uint32_t rctl;
4506         struct e1000_hw *hw;
4507         struct rte_eth_dev_info dev_info;
4508         uint32_t frame_size = mtu + E1000_ETH_OVERHEAD;
4509         int ret;
4510
4511         hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4512
4513 #ifdef RTE_LIBRTE_82571_SUPPORT
4514         /* XXX: not bigger than max_rx_pktlen */
4515         if (hw->mac.type == e1000_82571)
4516                 return -ENOTSUP;
4517 #endif
4518         ret = eth_igb_infos_get(dev, &dev_info);
4519         if (ret != 0)
4520                 return ret;
4521
4522         /* check that mtu is within the allowed range */
4523         if (mtu < RTE_ETHER_MIN_MTU ||
4524                         frame_size > dev_info.max_rx_pktlen)
4525                 return -EINVAL;
4526
4527         /* refuse mtu that requires the support of scattered packets when this
4528          * feature has not been enabled before. */
4529         if (!dev->data->scattered_rx &&
4530             frame_size > dev->data->min_rx_buf_size - RTE_PKTMBUF_HEADROOM)
4531                 return -EINVAL;
4532
4533         rctl = E1000_READ_REG(hw, E1000_RCTL);
4534
4535         /* switch to jumbo mode if needed */
4536         if (frame_size > RTE_ETHER_MAX_LEN) {
4537                 dev->data->dev_conf.rxmode.offloads |=
4538                         DEV_RX_OFFLOAD_JUMBO_FRAME;
4539                 rctl |= E1000_RCTL_LPE;
4540         } else {
4541                 dev->data->dev_conf.rxmode.offloads &=
4542                         ~DEV_RX_OFFLOAD_JUMBO_FRAME;
4543                 rctl &= ~E1000_RCTL_LPE;
4544         }
4545         E1000_WRITE_REG(hw, E1000_RCTL, rctl);
4546
4547         /* update max frame size */
4548         dev->data->dev_conf.rxmode.max_rx_pkt_len = frame_size;
4549
4550         E1000_WRITE_REG(hw, E1000_RLPML,
4551                         dev->data->dev_conf.rxmode.max_rx_pkt_len);
4552
4553         return 0;
4554 }
4555
4556 /*
4557  * igb_add_del_ntuple_filter - add or delete a ntuple filter
4558  *
4559  * @param
4560  * dev: Pointer to struct rte_eth_dev.
4561  * ntuple_filter: Pointer to struct rte_eth_ntuple_filter
4562  * add: if true, add filter, if false, remove filter
4563  *
4564  * @return
4565  *    - On success, zero.
4566  *    - On failure, a negative value.
4567  */
4568 int
4569 igb_add_del_ntuple_filter(struct rte_eth_dev *dev,
4570                         struct rte_eth_ntuple_filter *ntuple_filter,
4571                         bool add)
4572 {
4573         struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4574         int ret;
4575
4576         switch (ntuple_filter->flags) {
4577         case RTE_5TUPLE_FLAGS:
4578         case (RTE_5TUPLE_FLAGS | RTE_NTUPLE_FLAGS_TCP_FLAG):
4579                 if (hw->mac.type != e1000_82576)
4580                         return -ENOTSUP;
4581                 if (add)
4582                         ret = igb_add_5tuple_filter_82576(dev,
4583                                                           ntuple_filter);
4584                 else
4585                         ret = igb_remove_5tuple_filter_82576(dev,
4586                                                              ntuple_filter);
4587                 break;
4588         case RTE_2TUPLE_FLAGS:
4589         case (RTE_2TUPLE_FLAGS | RTE_NTUPLE_FLAGS_TCP_FLAG):
4590                 if (hw->mac.type != e1000_82580 && hw->mac.type != e1000_i350 &&
4591                         hw->mac.type != e1000_i210 &&
4592                         hw->mac.type != e1000_i211)
4593                         return -ENOTSUP;
4594                 if (add)
4595                         ret = igb_add_2tuple_filter(dev, ntuple_filter);
4596                 else
4597                         ret = igb_remove_2tuple_filter(dev, ntuple_filter);
4598                 break;
4599         default:
4600                 ret = -EINVAL;
4601                 break;
4602         }
4603
4604         return ret;
4605 }
4606
4607 /*
4608  * igb_get_ntuple_filter - get a ntuple filter
4609  *
4610  * @param
4611  * dev: Pointer to struct rte_eth_dev.
4612  * ntuple_filter: Pointer to struct rte_eth_ntuple_filter
4613  *
4614  * @return
4615  *    - On success, zero.
4616  *    - On failure, a negative value.
4617  */
4618 static int
4619 igb_get_ntuple_filter(struct rte_eth_dev *dev,
4620                         struct rte_eth_ntuple_filter *ntuple_filter)
4621 {
4622         struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4623         struct e1000_filter_info *filter_info =
4624                 E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
4625         struct e1000_5tuple_filter_info filter_5tuple;
4626         struct e1000_2tuple_filter_info filter_2tuple;
4627         struct e1000_5tuple_filter *p_5tuple_filter;
4628         struct e1000_2tuple_filter *p_2tuple_filter;
4629         int ret;
4630
4631         switch (ntuple_filter->flags) {
4632         case RTE_5TUPLE_FLAGS:
4633         case (RTE_5TUPLE_FLAGS | RTE_NTUPLE_FLAGS_TCP_FLAG):
4634                 if (hw->mac.type != e1000_82576)
4635                         return -ENOTSUP;
4636                 memset(&filter_5tuple,
4637                         0,
4638                         sizeof(struct e1000_5tuple_filter_info));
4639                 ret = ntuple_filter_to_5tuple_82576(ntuple_filter,
4640                                                     &filter_5tuple);
4641                 if (ret < 0)
4642                         return ret;
4643                 p_5tuple_filter = igb_5tuple_filter_lookup_82576(
4644                                         &filter_info->fivetuple_list,
4645                                         &filter_5tuple);
4646                 if (p_5tuple_filter == NULL) {
4647                         PMD_DRV_LOG(ERR, "filter doesn't exist.");
4648                         return -ENOENT;
4649                 }
4650                 ntuple_filter->queue = p_5tuple_filter->queue;
4651                 break;
4652         case RTE_2TUPLE_FLAGS:
4653         case (RTE_2TUPLE_FLAGS | RTE_NTUPLE_FLAGS_TCP_FLAG):
4654                 if (hw->mac.type != e1000_82580 && hw->mac.type != e1000_i350)
4655                         return -ENOTSUP;
4656                 memset(&filter_2tuple,
4657                         0,
4658                         sizeof(struct e1000_2tuple_filter_info));
4659                 ret = ntuple_filter_to_2tuple(ntuple_filter, &filter_2tuple);
4660                 if (ret < 0)
4661                         return ret;
4662                 p_2tuple_filter = igb_2tuple_filter_lookup(
4663                                         &filter_info->twotuple_list,
4664                                         &filter_2tuple);
4665                 if (p_2tuple_filter == NULL) {
4666                         PMD_DRV_LOG(ERR, "filter doesn't exist.");
4667                         return -ENOENT;
4668                 }
4669                 ntuple_filter->queue = p_2tuple_filter->queue;
4670                 break;
4671         default:
4672                 ret = -EINVAL;
4673                 break;
4674         }
4675
4676         return 0;
4677 }
4678
4679 /*
4680  * igb_ntuple_filter_handle - Handle operations for ntuple filter.
4681  * @dev: pointer to rte_eth_dev structure
4682  * @filter_op:operation will be taken.
4683  * @arg: a pointer to specific structure corresponding to the filter_op
4684  */
4685 static int
4686 igb_ntuple_filter_handle(struct rte_eth_dev *dev,
4687                                 enum rte_filter_op filter_op,
4688                                 void *arg)
4689 {
4690         struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4691         int ret;
4692
4693         MAC_TYPE_FILTER_SUP(hw->mac.type);
4694
4695         if (filter_op == RTE_ETH_FILTER_NOP)
4696                 return 0;
4697
4698         if (arg == NULL) {
4699                 PMD_DRV_LOG(ERR, "arg shouldn't be NULL for operation %u.",
4700                             filter_op);
4701                 return -EINVAL;
4702         }
4703
4704         switch (filter_op) {
4705         case RTE_ETH_FILTER_ADD:
4706                 ret = igb_add_del_ntuple_filter(dev,
4707                         (struct rte_eth_ntuple_filter *)arg,
4708                         TRUE);
4709                 break;
4710         case RTE_ETH_FILTER_DELETE:
4711                 ret = igb_add_del_ntuple_filter(dev,
4712                         (struct rte_eth_ntuple_filter *)arg,
4713                         FALSE);
4714                 break;
4715         case RTE_ETH_FILTER_GET:
4716                 ret = igb_get_ntuple_filter(dev,
4717                         (struct rte_eth_ntuple_filter *)arg);
4718                 break;
4719         default:
4720                 PMD_DRV_LOG(ERR, "unsupported operation %u.", filter_op);
4721                 ret = -EINVAL;
4722                 break;
4723         }
4724         return ret;
4725 }
4726
4727 static inline int
4728 igb_ethertype_filter_lookup(struct e1000_filter_info *filter_info,
4729                         uint16_t ethertype)
4730 {
4731         int i;
4732
4733         for (i = 0; i < E1000_MAX_ETQF_FILTERS; i++) {
4734                 if (filter_info->ethertype_filters[i].ethertype == ethertype &&
4735                     (filter_info->ethertype_mask & (1 << i)))
4736                         return i;
4737         }
4738         return -1;
4739 }
4740
4741 static inline int
4742 igb_ethertype_filter_insert(struct e1000_filter_info *filter_info,
4743                         uint16_t ethertype, uint32_t etqf)
4744 {
4745         int i;
4746
4747         for (i = 0; i < E1000_MAX_ETQF_FILTERS; i++) {
4748                 if (!(filter_info->ethertype_mask & (1 << i))) {
4749                         filter_info->ethertype_mask |= 1 << i;
4750                         filter_info->ethertype_filters[i].ethertype = ethertype;
4751                         filter_info->ethertype_filters[i].etqf = etqf;
4752                         return i;
4753                 }
4754         }
4755         return -1;
4756 }
4757
4758 int
4759 igb_ethertype_filter_remove(struct e1000_filter_info *filter_info,
4760                         uint8_t idx)
4761 {
4762         if (idx >= E1000_MAX_ETQF_FILTERS)
4763                 return -1;
4764         filter_info->ethertype_mask &= ~(1 << idx);
4765         filter_info->ethertype_filters[idx].ethertype = 0;
4766         filter_info->ethertype_filters[idx].etqf = 0;
4767         return idx;
4768 }
4769
4770
4771 int
4772 igb_add_del_ethertype_filter(struct rte_eth_dev *dev,
4773                         struct rte_eth_ethertype_filter *filter,
4774                         bool add)
4775 {
4776         struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4777         struct e1000_filter_info *filter_info =
4778                 E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
4779         uint32_t etqf = 0;
4780         int ret;
4781
4782         if (filter->ether_type == RTE_ETHER_TYPE_IPV4 ||
4783                 filter->ether_type == RTE_ETHER_TYPE_IPV6) {
4784                 PMD_DRV_LOG(ERR, "unsupported ether_type(0x%04x) in"
4785                         " ethertype filter.", filter->ether_type);
4786                 return -EINVAL;
4787         }
4788
4789         if (filter->flags & RTE_ETHTYPE_FLAGS_MAC) {
4790                 PMD_DRV_LOG(ERR, "mac compare is unsupported.");
4791                 return -EINVAL;
4792         }
4793         if (filter->flags & RTE_ETHTYPE_FLAGS_DROP) {
4794                 PMD_DRV_LOG(ERR, "drop option is unsupported.");
4795                 return -EINVAL;
4796         }
4797
4798         ret = igb_ethertype_filter_lookup(filter_info, filter->ether_type);
4799         if (ret >= 0 && add) {
4800                 PMD_DRV_LOG(ERR, "ethertype (0x%04x) filter exists.",
4801                             filter->ether_type);
4802                 return -EEXIST;
4803         }
4804         if (ret < 0 && !add) {
4805                 PMD_DRV_LOG(ERR, "ethertype (0x%04x) filter doesn't exist.",
4806                             filter->ether_type);
4807                 return -ENOENT;
4808         }
4809
4810         if (add) {
4811                 etqf |= E1000_ETQF_FILTER_ENABLE | E1000_ETQF_QUEUE_ENABLE;
4812                 etqf |= (uint32_t)(filter->ether_type & E1000_ETQF_ETHERTYPE);
4813                 etqf |= filter->queue << E1000_ETQF_QUEUE_SHIFT;
4814                 ret = igb_ethertype_filter_insert(filter_info,
4815                                 filter->ether_type, etqf);
4816                 if (ret < 0) {
4817                         PMD_DRV_LOG(ERR, "ethertype filters are full.");
4818                         return -ENOSYS;
4819                 }
4820         } else {
4821                 ret = igb_ethertype_filter_remove(filter_info, (uint8_t)ret);
4822                 if (ret < 0)
4823                         return -ENOSYS;
4824         }
4825         E1000_WRITE_REG(hw, E1000_ETQF(ret), etqf);
4826         E1000_WRITE_FLUSH(hw);
4827
4828         return 0;
4829 }
4830
4831 static int
4832 igb_get_ethertype_filter(struct rte_eth_dev *dev,
4833                         struct rte_eth_ethertype_filter *filter)
4834 {
4835         struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4836         struct e1000_filter_info *filter_info =
4837                 E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
4838         uint32_t etqf;
4839         int ret;
4840
4841         ret = igb_ethertype_filter_lookup(filter_info, filter->ether_type);
4842         if (ret < 0) {
4843                 PMD_DRV_LOG(ERR, "ethertype (0x%04x) filter doesn't exist.",
4844                             filter->ether_type);
4845                 return -ENOENT;
4846         }
4847
4848         etqf = E1000_READ_REG(hw, E1000_ETQF(ret));
4849         if (etqf & E1000_ETQF_FILTER_ENABLE) {
4850                 filter->ether_type = etqf & E1000_ETQF_ETHERTYPE;
4851                 filter->flags = 0;
4852                 filter->queue = (etqf & E1000_ETQF_QUEUE) >>
4853                                 E1000_ETQF_QUEUE_SHIFT;
4854                 return 0;
4855         }
4856
4857         return -ENOENT;
4858 }
4859
4860 /*
4861  * igb_ethertype_filter_handle - Handle operations for ethertype filter.
4862  * @dev: pointer to rte_eth_dev structure
4863  * @filter_op:operation will be taken.
4864  * @arg: a pointer to specific structure corresponding to the filter_op
4865  */
4866 static int
4867 igb_ethertype_filter_handle(struct rte_eth_dev *dev,
4868                                 enum rte_filter_op filter_op,
4869                                 void *arg)
4870 {
4871         struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4872         int ret;
4873
4874         MAC_TYPE_FILTER_SUP(hw->mac.type);
4875
4876         if (filter_op == RTE_ETH_FILTER_NOP)
4877                 return 0;
4878
4879         if (arg == NULL) {
4880                 PMD_DRV_LOG(ERR, "arg shouldn't be NULL for operation %u.",
4881                             filter_op);
4882                 return -EINVAL;
4883         }
4884
4885         switch (filter_op) {
4886         case RTE_ETH_FILTER_ADD:
4887                 ret = igb_add_del_ethertype_filter(dev,
4888                         (struct rte_eth_ethertype_filter *)arg,
4889                         TRUE);
4890                 break;
4891         case RTE_ETH_FILTER_DELETE:
4892                 ret = igb_add_del_ethertype_filter(dev,
4893                         (struct rte_eth_ethertype_filter *)arg,
4894                         FALSE);
4895                 break;
4896         case RTE_ETH_FILTER_GET:
4897                 ret = igb_get_ethertype_filter(dev,
4898                         (struct rte_eth_ethertype_filter *)arg);
4899                 break;
4900         default:
4901                 PMD_DRV_LOG(ERR, "unsupported operation %u.", filter_op);
4902                 ret = -EINVAL;
4903                 break;
4904         }
4905         return ret;
4906 }
4907
4908 static int
4909 eth_igb_filter_ctrl(struct rte_eth_dev *dev,
4910                      enum rte_filter_type filter_type,
4911                      enum rte_filter_op filter_op,
4912                      void *arg)
4913 {
4914         int ret = 0;
4915
4916         switch (filter_type) {
4917         case RTE_ETH_FILTER_NTUPLE:
4918                 ret = igb_ntuple_filter_handle(dev, filter_op, arg);
4919                 break;
4920         case RTE_ETH_FILTER_ETHERTYPE:
4921                 ret = igb_ethertype_filter_handle(dev, filter_op, arg);
4922                 break;
4923         case RTE_ETH_FILTER_SYN:
4924                 ret = eth_igb_syn_filter_handle(dev, filter_op, arg);
4925                 break;
4926         case RTE_ETH_FILTER_FLEXIBLE:
4927                 ret = eth_igb_flex_filter_handle(dev, filter_op, arg);
4928                 break;
4929         case RTE_ETH_FILTER_GENERIC:
4930                 if (filter_op != RTE_ETH_FILTER_GET)
4931                         return -EINVAL;
4932                 *(const void **)arg = &igb_flow_ops;
4933                 break;
4934         default:
4935                 PMD_DRV_LOG(WARNING, "Filter type (%d) not supported",
4936                                                         filter_type);
4937                 break;
4938         }
4939
4940         return ret;
4941 }
4942
4943 static int
4944 eth_igb_set_mc_addr_list(struct rte_eth_dev *dev,
4945                          struct rte_ether_addr *mc_addr_set,
4946                          uint32_t nb_mc_addr)
4947 {
4948         struct e1000_hw *hw;
4949
4950         hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4951         e1000_update_mc_addr_list(hw, (u8 *)mc_addr_set, nb_mc_addr);
4952         return 0;
4953 }
4954
4955 static uint64_t
4956 igb_read_systime_cyclecounter(struct rte_eth_dev *dev)
4957 {
4958         struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4959         uint64_t systime_cycles;
4960
4961         switch (hw->mac.type) {
4962         case e1000_i210:
4963         case e1000_i211:
4964                 /*
4965                  * Need to read System Time Residue Register to be able
4966                  * to read the other two registers.
4967                  */
4968                 E1000_READ_REG(hw, E1000_SYSTIMR);
4969                 /* SYSTIMEL stores ns and SYSTIMEH stores seconds. */
4970                 systime_cycles = (uint64_t)E1000_READ_REG(hw, E1000_SYSTIML);
4971                 systime_cycles += (uint64_t)E1000_READ_REG(hw, E1000_SYSTIMH)
4972                                 * NSEC_PER_SEC;
4973                 break;
4974         case e1000_82580:
4975         case e1000_i350:
4976         case e1000_i354:
4977                 /*
4978                  * Need to read System Time Residue Register to be able
4979                  * to read the other two registers.
4980                  */
4981                 E1000_READ_REG(hw, E1000_SYSTIMR);
4982                 systime_cycles = (uint64_t)E1000_READ_REG(hw, E1000_SYSTIML);
4983                 /* Only the 8 LSB are valid. */
4984                 systime_cycles |= (uint64_t)(E1000_READ_REG(hw, E1000_SYSTIMH)
4985                                 & 0xff) << 32;
4986                 break;
4987         default:
4988                 systime_cycles = (uint64_t)E1000_READ_REG(hw, E1000_SYSTIML);
4989                 systime_cycles |= (uint64_t)E1000_READ_REG(hw, E1000_SYSTIMH)
4990                                 << 32;
4991                 break;
4992         }
4993
4994         return systime_cycles;
4995 }
4996
4997 static uint64_t
4998 igb_read_rx_tstamp_cyclecounter(struct rte_eth_dev *dev)
4999 {
5000         struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5001         uint64_t rx_tstamp_cycles;
5002
5003         switch (hw->mac.type) {
5004         case e1000_i210:
5005         case e1000_i211:
5006                 /* RXSTMPL stores ns and RXSTMPH stores seconds. */
5007                 rx_tstamp_cycles = (uint64_t)E1000_READ_REG(hw, E1000_RXSTMPL);
5008                 rx_tstamp_cycles += (uint64_t)E1000_READ_REG(hw, E1000_RXSTMPH)
5009                                 * NSEC_PER_SEC;
5010                 break;
5011         case e1000_82580:
5012         case e1000_i350:
5013         case e1000_i354:
5014                 rx_tstamp_cycles = (uint64_t)E1000_READ_REG(hw, E1000_RXSTMPL);
5015                 /* Only the 8 LSB are valid. */
5016                 rx_tstamp_cycles |= (uint64_t)(E1000_READ_REG(hw, E1000_RXSTMPH)
5017                                 & 0xff) << 32;
5018                 break;
5019         default:
5020                 rx_tstamp_cycles = (uint64_t)E1000_READ_REG(hw, E1000_RXSTMPL);
5021                 rx_tstamp_cycles |= (uint64_t)E1000_READ_REG(hw, E1000_RXSTMPH)
5022                                 << 32;
5023                 break;
5024         }
5025
5026         return rx_tstamp_cycles;
5027 }
5028
5029 static uint64_t
5030 igb_read_tx_tstamp_cyclecounter(struct rte_eth_dev *dev)
5031 {
5032         struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5033         uint64_t tx_tstamp_cycles;
5034
5035         switch (hw->mac.type) {
5036         case e1000_i210:
5037         case e1000_i211:
5038                 /* RXSTMPL stores ns and RXSTMPH stores seconds. */
5039                 tx_tstamp_cycles = (uint64_t)E1000_READ_REG(hw, E1000_TXSTMPL);
5040                 tx_tstamp_cycles += (uint64_t)E1000_READ_REG(hw, E1000_TXSTMPH)
5041                                 * NSEC_PER_SEC;
5042                 break;
5043         case e1000_82580:
5044         case e1000_i350:
5045         case e1000_i354:
5046                 tx_tstamp_cycles = (uint64_t)E1000_READ_REG(hw, E1000_TXSTMPL);
5047                 /* Only the 8 LSB are valid. */
5048                 tx_tstamp_cycles |= (uint64_t)(E1000_READ_REG(hw, E1000_TXSTMPH)
5049                                 & 0xff) << 32;
5050                 break;
5051         default:
5052                 tx_tstamp_cycles = (uint64_t)E1000_READ_REG(hw, E1000_TXSTMPL);
5053                 tx_tstamp_cycles |= (uint64_t)E1000_READ_REG(hw, E1000_TXSTMPH)
5054                                 << 32;
5055                 break;
5056         }
5057
5058         return tx_tstamp_cycles;
5059 }
5060
5061 static void
5062 igb_start_timecounters(struct rte_eth_dev *dev)
5063 {
5064         struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5065         struct e1000_adapter *adapter = dev->data->dev_private;
5066         uint32_t incval = 1;
5067         uint32_t shift = 0;
5068         uint64_t mask = E1000_CYCLECOUNTER_MASK;
5069
5070         switch (hw->mac.type) {
5071         case e1000_82580:
5072         case e1000_i350:
5073         case e1000_i354:
5074                 /* 32 LSB bits + 8 MSB bits = 40 bits */
5075                 mask = (1ULL << 40) - 1;
5076                 /* fall-through */
5077         case e1000_i210:
5078         case e1000_i211:
5079                 /*
5080                  * Start incrementing the register
5081                  * used to timestamp PTP packets.
5082                  */
5083                 E1000_WRITE_REG(hw, E1000_TIMINCA, incval);
5084                 break;
5085         case e1000_82576:
5086                 incval = E1000_INCVALUE_82576;
5087                 shift = IGB_82576_TSYNC_SHIFT;
5088                 E1000_WRITE_REG(hw, E1000_TIMINCA,
5089                                 E1000_INCPERIOD_82576 | incval);
5090                 break;
5091         default:
5092                 /* Not supported */
5093                 return;
5094         }
5095
5096         memset(&adapter->systime_tc, 0, sizeof(struct rte_timecounter));
5097         memset(&adapter->rx_tstamp_tc, 0, sizeof(struct rte_timecounter));
5098         memset(&adapter->tx_tstamp_tc, 0, sizeof(struct rte_timecounter));
5099
5100         adapter->systime_tc.cc_mask = mask;
5101         adapter->systime_tc.cc_shift = shift;
5102         adapter->systime_tc.nsec_mask = (1ULL << shift) - 1;
5103
5104         adapter->rx_tstamp_tc.cc_mask = mask;
5105         adapter->rx_tstamp_tc.cc_shift = shift;
5106         adapter->rx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
5107
5108         adapter->tx_tstamp_tc.cc_mask = mask;
5109         adapter->tx_tstamp_tc.cc_shift = shift;
5110         adapter->tx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
5111 }
5112
5113 static int
5114 igb_timesync_adjust_time(struct rte_eth_dev *dev, int64_t delta)
5115 {
5116         struct e1000_adapter *adapter = dev->data->dev_private;
5117
5118         adapter->systime_tc.nsec += delta;
5119         adapter->rx_tstamp_tc.nsec += delta;
5120         adapter->tx_tstamp_tc.nsec += delta;
5121
5122         return 0;
5123 }
5124
5125 static int
5126 igb_timesync_write_time(struct rte_eth_dev *dev, const struct timespec *ts)
5127 {
5128         uint64_t ns;
5129         struct e1000_adapter *adapter = dev->data->dev_private;
5130
5131         ns = rte_timespec_to_ns(ts);
5132
5133         /* Set the timecounters to a new value. */
5134         adapter->systime_tc.nsec = ns;
5135         adapter->rx_tstamp_tc.nsec = ns;
5136         adapter->tx_tstamp_tc.nsec = ns;
5137
5138         return 0;
5139 }
5140
5141 static int
5142 igb_timesync_read_time(struct rte_eth_dev *dev, struct timespec *ts)
5143 {
5144         uint64_t ns, systime_cycles;
5145         struct e1000_adapter *adapter = dev->data->dev_private;
5146
5147         systime_cycles = igb_read_systime_cyclecounter(dev);
5148         ns = rte_timecounter_update(&adapter->systime_tc, systime_cycles);
5149         *ts = rte_ns_to_timespec(ns);
5150
5151         return 0;
5152 }
5153
5154 static int
5155 igb_timesync_enable(struct rte_eth_dev *dev)
5156 {
5157         struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5158         uint32_t tsync_ctl;
5159         uint32_t tsauxc;
5160
5161         /* Stop the timesync system time. */
5162         E1000_WRITE_REG(hw, E1000_TIMINCA, 0x0);
5163         /* Reset the timesync system time value. */
5164         switch (hw->mac.type) {
5165         case e1000_82580:
5166         case e1000_i350:
5167         case e1000_i354:
5168         case e1000_i210:
5169         case e1000_i211:
5170                 E1000_WRITE_REG(hw, E1000_SYSTIMR, 0x0);
5171                 /* fall-through */
5172         case e1000_82576:
5173                 E1000_WRITE_REG(hw, E1000_SYSTIML, 0x0);
5174                 E1000_WRITE_REG(hw, E1000_SYSTIMH, 0x0);
5175                 break;
5176         default:
5177                 /* Not supported. */
5178                 return -ENOTSUP;
5179         }
5180
5181         /* Enable system time for it isn't on by default. */
5182         tsauxc = E1000_READ_REG(hw, E1000_TSAUXC);
5183         tsauxc &= ~E1000_TSAUXC_DISABLE_SYSTIME;
5184         E1000_WRITE_REG(hw, E1000_TSAUXC, tsauxc);
5185
5186         igb_start_timecounters(dev);
5187
5188         /* Enable L2 filtering of IEEE1588/802.1AS Ethernet frame types. */
5189         E1000_WRITE_REG(hw, E1000_ETQF(E1000_ETQF_FILTER_1588),
5190                         (RTE_ETHER_TYPE_1588 |
5191                          E1000_ETQF_FILTER_ENABLE |
5192                          E1000_ETQF_1588));
5193
5194         /* Enable timestamping of received PTP packets. */
5195         tsync_ctl = E1000_READ_REG(hw, E1000_TSYNCRXCTL);
5196         tsync_ctl |= E1000_TSYNCRXCTL_ENABLED;
5197         E1000_WRITE_REG(hw, E1000_TSYNCRXCTL, tsync_ctl);
5198
5199         /* Enable Timestamping of transmitted PTP packets. */
5200         tsync_ctl = E1000_READ_REG(hw, E1000_TSYNCTXCTL);
5201         tsync_ctl |= E1000_TSYNCTXCTL_ENABLED;
5202         E1000_WRITE_REG(hw, E1000_TSYNCTXCTL, tsync_ctl);
5203
5204         return 0;
5205 }
5206
5207 static int
5208 igb_timesync_disable(struct rte_eth_dev *dev)
5209 {
5210         struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5211         uint32_t tsync_ctl;
5212
5213         /* Disable timestamping of transmitted PTP packets. */
5214         tsync_ctl = E1000_READ_REG(hw, E1000_TSYNCTXCTL);
5215         tsync_ctl &= ~E1000_TSYNCTXCTL_ENABLED;
5216         E1000_WRITE_REG(hw, E1000_TSYNCTXCTL, tsync_ctl);
5217
5218         /* Disable timestamping of received PTP packets. */
5219         tsync_ctl = E1000_READ_REG(hw, E1000_TSYNCRXCTL);
5220         tsync_ctl &= ~E1000_TSYNCRXCTL_ENABLED;
5221         E1000_WRITE_REG(hw, E1000_TSYNCRXCTL, tsync_ctl);
5222
5223         /* Disable L2 filtering of IEEE1588/802.1AS Ethernet frame types. */
5224         E1000_WRITE_REG(hw, E1000_ETQF(E1000_ETQF_FILTER_1588), 0);
5225
5226         /* Stop incrementating the System Time registers. */
5227         E1000_WRITE_REG(hw, E1000_TIMINCA, 0);
5228
5229         return 0;
5230 }
5231
5232 static int
5233 igb_timesync_read_rx_timestamp(struct rte_eth_dev *dev,
5234                                struct timespec *timestamp,
5235                                uint32_t flags __rte_unused)
5236 {
5237         struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5238         struct e1000_adapter *adapter = dev->data->dev_private;
5239         uint32_t tsync_rxctl;
5240         uint64_t rx_tstamp_cycles;
5241         uint64_t ns;
5242
5243         tsync_rxctl = E1000_READ_REG(hw, E1000_TSYNCRXCTL);
5244         if ((tsync_rxctl & E1000_TSYNCRXCTL_VALID) == 0)
5245                 return -EINVAL;
5246
5247         rx_tstamp_cycles = igb_read_rx_tstamp_cyclecounter(dev);
5248         ns = rte_timecounter_update(&adapter->rx_tstamp_tc, rx_tstamp_cycles);
5249         *timestamp = rte_ns_to_timespec(ns);
5250
5251         return  0;
5252 }
5253
5254 static int
5255 igb_timesync_read_tx_timestamp(struct rte_eth_dev *dev,
5256                                struct timespec *timestamp)
5257 {
5258         struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5259         struct e1000_adapter *adapter = dev->data->dev_private;
5260         uint32_t tsync_txctl;
5261         uint64_t tx_tstamp_cycles;
5262         uint64_t ns;
5263
5264         tsync_txctl = E1000_READ_REG(hw, E1000_TSYNCTXCTL);
5265         if ((tsync_txctl & E1000_TSYNCTXCTL_VALID) == 0)
5266                 return -EINVAL;
5267
5268         tx_tstamp_cycles = igb_read_tx_tstamp_cyclecounter(dev);
5269         ns = rte_timecounter_update(&adapter->tx_tstamp_tc, tx_tstamp_cycles);
5270         *timestamp = rte_ns_to_timespec(ns);
5271
5272         return  0;
5273 }
5274
5275 static int
5276 eth_igb_get_reg_length(struct rte_eth_dev *dev __rte_unused)
5277 {
5278         int count = 0;
5279         int g_ind = 0;
5280         const struct reg_info *reg_group;
5281
5282         while ((reg_group = igb_regs[g_ind++]))
5283                 count += igb_reg_group_count(reg_group);
5284
5285         return count;
5286 }
5287
5288 static int
5289 igbvf_get_reg_length(struct rte_eth_dev *dev __rte_unused)
5290 {
5291         int count = 0;
5292         int g_ind = 0;
5293         const struct reg_info *reg_group;
5294
5295         while ((reg_group = igbvf_regs[g_ind++]))
5296                 count += igb_reg_group_count(reg_group);
5297
5298         return count;
5299 }
5300
5301 static int
5302 eth_igb_get_regs(struct rte_eth_dev *dev,
5303         struct rte_dev_reg_info *regs)
5304 {
5305         struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5306         uint32_t *data = regs->data;
5307         int g_ind = 0;
5308         int count = 0;
5309         const struct reg_info *reg_group;
5310
5311         if (data == NULL) {
5312                 regs->length = eth_igb_get_reg_length(dev);
5313                 regs->width = sizeof(uint32_t);
5314                 return 0;
5315         }
5316
5317         /* Support only full register dump */
5318         if ((regs->length == 0) ||
5319             (regs->length == (uint32_t)eth_igb_get_reg_length(dev))) {
5320                 regs->version = hw->mac.type << 24 | hw->revision_id << 16 |
5321                         hw->device_id;
5322                 while ((reg_group = igb_regs[g_ind++]))
5323                         count += igb_read_regs_group(dev, &data[count],
5324                                                         reg_group);
5325                 return 0;
5326         }
5327
5328         return -ENOTSUP;
5329 }
5330
5331 static int
5332 igbvf_get_regs(struct rte_eth_dev *dev,
5333         struct rte_dev_reg_info *regs)
5334 {
5335         struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5336         uint32_t *data = regs->data;
5337         int g_ind = 0;
5338         int count = 0;
5339         const struct reg_info *reg_group;
5340
5341         if (data == NULL) {
5342                 regs->length = igbvf_get_reg_length(dev);
5343                 regs->width = sizeof(uint32_t);
5344                 return 0;
5345         }
5346
5347         /* Support only full register dump */
5348         if ((regs->length == 0) ||
5349             (regs->length == (uint32_t)igbvf_get_reg_length(dev))) {
5350                 regs->version = hw->mac.type << 24 | hw->revision_id << 16 |
5351                         hw->device_id;
5352                 while ((reg_group = igbvf_regs[g_ind++]))
5353                         count += igb_read_regs_group(dev, &data[count],
5354                                                         reg_group);
5355                 return 0;
5356         }
5357
5358         return -ENOTSUP;
5359 }
5360
5361 static int
5362 eth_igb_get_eeprom_length(struct rte_eth_dev *dev)
5363 {
5364         struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5365
5366         /* Return unit is byte count */
5367         return hw->nvm.word_size * 2;
5368 }
5369
5370 static int
5371 eth_igb_get_eeprom(struct rte_eth_dev *dev,
5372         struct rte_dev_eeprom_info *in_eeprom)
5373 {
5374         struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5375         struct e1000_nvm_info *nvm = &hw->nvm;
5376         uint16_t *data = in_eeprom->data;
5377         int first, length;
5378
5379         first = in_eeprom->offset >> 1;
5380         length = in_eeprom->length >> 1;
5381         if ((first >= hw->nvm.word_size) ||
5382             ((first + length) >= hw->nvm.word_size))
5383                 return -EINVAL;
5384
5385         in_eeprom->magic = hw->vendor_id |
5386                 ((uint32_t)hw->device_id << 16);
5387
5388         if ((nvm->ops.read) == NULL)
5389                 return -ENOTSUP;
5390
5391         return nvm->ops.read(hw, first, length, data);
5392 }
5393
5394 static int
5395 eth_igb_set_eeprom(struct rte_eth_dev *dev,
5396         struct rte_dev_eeprom_info *in_eeprom)
5397 {
5398         struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5399         struct e1000_nvm_info *nvm = &hw->nvm;
5400         uint16_t *data = in_eeprom->data;
5401         int first, length;
5402
5403         first = in_eeprom->offset >> 1;
5404         length = in_eeprom->length >> 1;
5405         if ((first >= hw->nvm.word_size) ||
5406             ((first + length) >= hw->nvm.word_size))
5407                 return -EINVAL;
5408
5409         in_eeprom->magic = (uint32_t)hw->vendor_id |
5410                 ((uint32_t)hw->device_id << 16);
5411
5412         if ((nvm->ops.write) == NULL)
5413                 return -ENOTSUP;
5414         return nvm->ops.write(hw,  first, length, data);
5415 }
5416
5417 static int
5418 eth_igb_get_module_info(struct rte_eth_dev *dev,
5419                         struct rte_eth_dev_module_info *modinfo)
5420 {
5421         struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5422
5423         uint32_t status = 0;
5424         uint16_t sff8472_rev, addr_mode;
5425         bool page_swap = false;
5426
5427         if (hw->phy.media_type == e1000_media_type_copper ||
5428             hw->phy.media_type == e1000_media_type_unknown)
5429                 return -EOPNOTSUPP;
5430
5431         /* Check whether we support SFF-8472 or not */
5432         status = e1000_read_phy_reg_i2c(hw, IGB_SFF_8472_COMP, &sff8472_rev);
5433         if (status)
5434                 return -EIO;
5435
5436         /* addressing mode is not supported */
5437         status = e1000_read_phy_reg_i2c(hw, IGB_SFF_8472_SWAP, &addr_mode);
5438         if (status)
5439                 return -EIO;
5440
5441         /* addressing mode is not supported */
5442         if ((addr_mode & 0xFF) & IGB_SFF_ADDRESSING_MODE) {
5443                 PMD_DRV_LOG(ERR,
5444                             "Address change required to access page 0xA2, "
5445                             "but not supported. Please report the module "
5446                             "type to the driver maintainers.\n");
5447                 page_swap = true;
5448         }
5449
5450         if ((sff8472_rev & 0xFF) == IGB_SFF_8472_UNSUP || page_swap) {
5451                 /* We have an SFP, but it does not support SFF-8472 */
5452                 modinfo->type = RTE_ETH_MODULE_SFF_8079;
5453                 modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8079_LEN;
5454         } else {
5455                 /* We have an SFP which supports a revision of SFF-8472 */
5456                 modinfo->type = RTE_ETH_MODULE_SFF_8472;
5457                 modinfo->eeprom_len = RTE_ETH_MODULE_SFF_8472_LEN;
5458         }
5459
5460         return 0;
5461 }
5462
5463 static int
5464 eth_igb_get_module_eeprom(struct rte_eth_dev *dev,
5465                           struct rte_dev_eeprom_info *info)
5466 {
5467         struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5468
5469         uint32_t status = 0;
5470         uint16_t dataword[RTE_ETH_MODULE_SFF_8472_LEN / 2 + 1];
5471         u16 first_word, last_word;
5472         int i = 0;
5473
5474         if (info->length == 0)
5475                 return -EINVAL;
5476
5477         first_word = info->offset >> 1;
5478         last_word = (info->offset + info->length - 1) >> 1;
5479
5480         /* Read EEPROM block, SFF-8079/SFF-8472, word at a time */
5481         for (i = 0; i < last_word - first_word + 1; i++) {
5482                 status = e1000_read_phy_reg_i2c(hw, (first_word + i) * 2,
5483                                                 &dataword[i]);
5484                 if (status) {
5485                         /* Error occurred while reading module */
5486                         return -EIO;
5487                 }
5488
5489                 dataword[i] = rte_be_to_cpu_16(dataword[i]);
5490         }
5491
5492         memcpy(info->data, (u8 *)dataword + (info->offset & 1), info->length);
5493
5494         return 0;
5495 }
5496
5497 static int
5498 eth_igb_rx_queue_intr_disable(struct rte_eth_dev *dev, uint16_t queue_id)
5499 {
5500         struct e1000_hw *hw =
5501                 E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5502         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5503         struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
5504         uint32_t vec = E1000_MISC_VEC_ID;
5505
5506         if (rte_intr_allow_others(intr_handle))
5507                 vec = E1000_RX_VEC_START;
5508
5509         uint32_t mask = 1 << (queue_id + vec);
5510
5511         E1000_WRITE_REG(hw, E1000_EIMC, mask);
5512         E1000_WRITE_FLUSH(hw);
5513
5514         return 0;
5515 }
5516
5517 static int
5518 eth_igb_rx_queue_intr_enable(struct rte_eth_dev *dev, uint16_t queue_id)
5519 {
5520         struct e1000_hw *hw =
5521                 E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5522         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5523         struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
5524         uint32_t vec = E1000_MISC_VEC_ID;
5525
5526         if (rte_intr_allow_others(intr_handle))
5527                 vec = E1000_RX_VEC_START;
5528
5529         uint32_t mask = 1 << (queue_id + vec);
5530         uint32_t regval;
5531
5532         regval = E1000_READ_REG(hw, E1000_EIMS);
5533         E1000_WRITE_REG(hw, E1000_EIMS, regval | mask);
5534         E1000_WRITE_FLUSH(hw);
5535
5536         rte_intr_ack(intr_handle);
5537
5538         return 0;
5539 }
5540
5541 static void
5542 eth_igb_write_ivar(struct e1000_hw *hw, uint8_t  msix_vector,
5543                    uint8_t index, uint8_t offset)
5544 {
5545         uint32_t val = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index);
5546
5547         /* clear bits */
5548         val &= ~((uint32_t)0xFF << offset);
5549
5550         /* write vector and valid bit */
5551         val |= (msix_vector | E1000_IVAR_VALID) << offset;
5552
5553         E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, val);
5554 }
5555
5556 static void
5557 eth_igb_assign_msix_vector(struct e1000_hw *hw, int8_t direction,
5558                            uint8_t queue, uint8_t msix_vector)
5559 {
5560         uint32_t tmp = 0;
5561
5562         if (hw->mac.type == e1000_82575) {
5563                 if (direction == 0)
5564                         tmp = E1000_EICR_RX_QUEUE0 << queue;
5565                 else if (direction == 1)
5566                         tmp = E1000_EICR_TX_QUEUE0 << queue;
5567                 E1000_WRITE_REG(hw, E1000_MSIXBM(msix_vector), tmp);
5568         } else if (hw->mac.type == e1000_82576) {
5569                 if ((direction == 0) || (direction == 1))
5570                         eth_igb_write_ivar(hw, msix_vector, queue & 0x7,
5571                                            ((queue & 0x8) << 1) +
5572                                            8 * direction);
5573         } else if ((hw->mac.type == e1000_82580) ||
5574                         (hw->mac.type == e1000_i350) ||
5575                         (hw->mac.type == e1000_i354) ||
5576                         (hw->mac.type == e1000_i210) ||
5577                         (hw->mac.type == e1000_i211)) {
5578                 if ((direction == 0) || (direction == 1))
5579                         eth_igb_write_ivar(hw, msix_vector,
5580                                            queue >> 1,
5581                                            ((queue & 0x1) << 4) +
5582                                            8 * direction);
5583         }
5584 }
5585
5586 /* Sets up the hardware to generate MSI-X interrupts properly
5587  * @hw
5588  *  board private structure
5589  */
5590 static void
5591 eth_igb_configure_msix_intr(struct rte_eth_dev *dev)
5592 {
5593         int queue_id;
5594         uint32_t tmpval, regval, intr_mask;
5595         struct e1000_hw *hw =
5596                 E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5597         uint32_t vec = E1000_MISC_VEC_ID;
5598         uint32_t base = E1000_MISC_VEC_ID;
5599         uint32_t misc_shift = 0;
5600         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
5601         struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
5602
5603         /* won't configure msix register if no mapping is done
5604          * between intr vector and event fd
5605          */
5606         if (!rte_intr_dp_is_en(intr_handle))
5607                 return;
5608
5609         if (rte_intr_allow_others(intr_handle)) {
5610                 vec = base = E1000_RX_VEC_START;
5611                 misc_shift = 1;
5612         }
5613
5614         /* set interrupt vector for other causes */
5615         if (hw->mac.type == e1000_82575) {
5616                 tmpval = E1000_READ_REG(hw, E1000_CTRL_EXT);
5617                 /* enable MSI-X PBA support */
5618                 tmpval |= E1000_CTRL_EXT_PBA_CLR;
5619
5620                 /* Auto-Mask interrupts upon ICR read */
5621                 tmpval |= E1000_CTRL_EXT_EIAME;
5622                 tmpval |= E1000_CTRL_EXT_IRCA;
5623
5624                 E1000_WRITE_REG(hw, E1000_CTRL_EXT, tmpval);
5625
5626                 /* enable msix_other interrupt */
5627                 E1000_WRITE_REG_ARRAY(hw, E1000_MSIXBM(0), 0, E1000_EIMS_OTHER);
5628                 regval = E1000_READ_REG(hw, E1000_EIAC);
5629                 E1000_WRITE_REG(hw, E1000_EIAC, regval | E1000_EIMS_OTHER);
5630                 regval = E1000_READ_REG(hw, E1000_EIAM);
5631                 E1000_WRITE_REG(hw, E1000_EIMS, regval | E1000_EIMS_OTHER);
5632         } else if ((hw->mac.type == e1000_82576) ||
5633                         (hw->mac.type == e1000_82580) ||
5634                         (hw->mac.type == e1000_i350) ||
5635                         (hw->mac.type == e1000_i354) ||
5636                         (hw->mac.type == e1000_i210) ||
5637                         (hw->mac.type == e1000_i211)) {
5638                 /* turn on MSI-X capability first */
5639                 E1000_WRITE_REG(hw, E1000_GPIE, E1000_GPIE_MSIX_MODE |
5640                                         E1000_GPIE_PBA | E1000_GPIE_EIAME |
5641                                         E1000_GPIE_NSICR);
5642                 intr_mask = RTE_LEN2MASK(intr_handle->nb_efd, uint32_t) <<
5643                         misc_shift;
5644
5645                 if (dev->data->dev_conf.intr_conf.lsc != 0)
5646                         intr_mask |= (1 << IGB_MSIX_OTHER_INTR_VEC);
5647
5648                 regval = E1000_READ_REG(hw, E1000_EIAC);
5649                 E1000_WRITE_REG(hw, E1000_EIAC, regval | intr_mask);
5650
5651                 /* enable msix_other interrupt */
5652                 regval = E1000_READ_REG(hw, E1000_EIMS);
5653                 E1000_WRITE_REG(hw, E1000_EIMS, regval | intr_mask);
5654                 tmpval = (IGB_MSIX_OTHER_INTR_VEC | E1000_IVAR_VALID) << 8;
5655                 E1000_WRITE_REG(hw, E1000_IVAR_MISC, tmpval);
5656         }
5657
5658         /* use EIAM to auto-mask when MSI-X interrupt
5659          * is asserted, this saves a register write for every interrupt
5660          */
5661         intr_mask = RTE_LEN2MASK(intr_handle->nb_efd, uint32_t) <<
5662                 misc_shift;
5663
5664         if (dev->data->dev_conf.intr_conf.lsc != 0)
5665                 intr_mask |= (1 << IGB_MSIX_OTHER_INTR_VEC);
5666
5667         regval = E1000_READ_REG(hw, E1000_EIAM);
5668         E1000_WRITE_REG(hw, E1000_EIAM, regval | intr_mask);
5669
5670         for (queue_id = 0; queue_id < dev->data->nb_rx_queues; queue_id++) {
5671                 eth_igb_assign_msix_vector(hw, 0, queue_id, vec);
5672                 intr_handle->intr_vec[queue_id] = vec;
5673                 if (vec < base + intr_handle->nb_efd - 1)
5674                         vec++;
5675         }
5676
5677         E1000_WRITE_FLUSH(hw);
5678 }
5679
5680 /* restore n-tuple filter */
5681 static inline void
5682 igb_ntuple_filter_restore(struct rte_eth_dev *dev)
5683 {
5684         struct e1000_filter_info *filter_info =
5685                 E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
5686         struct e1000_5tuple_filter *p_5tuple;
5687         struct e1000_2tuple_filter *p_2tuple;
5688
5689         TAILQ_FOREACH(p_5tuple, &filter_info->fivetuple_list, entries) {
5690                 igb_inject_5tuple_filter_82576(dev, p_5tuple);
5691         }
5692
5693         TAILQ_FOREACH(p_2tuple, &filter_info->twotuple_list, entries) {
5694                 igb_inject_2uple_filter(dev, p_2tuple);
5695         }
5696 }
5697
5698 /* restore SYN filter */
5699 static inline void
5700 igb_syn_filter_restore(struct rte_eth_dev *dev)
5701 {
5702         struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5703         struct e1000_filter_info *filter_info =
5704                 E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
5705         uint32_t synqf;
5706
5707         synqf = filter_info->syn_info;
5708
5709         if (synqf & E1000_SYN_FILTER_ENABLE) {
5710                 E1000_WRITE_REG(hw, E1000_SYNQF(0), synqf);
5711                 E1000_WRITE_FLUSH(hw);
5712         }
5713 }
5714
5715 /* restore ethernet type filter */
5716 static inline void
5717 igb_ethertype_filter_restore(struct rte_eth_dev *dev)
5718 {
5719         struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5720         struct e1000_filter_info *filter_info =
5721                 E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
5722         int i;
5723
5724         for (i = 0; i < E1000_MAX_ETQF_FILTERS; i++) {
5725                 if (filter_info->ethertype_mask & (1 << i)) {
5726                         E1000_WRITE_REG(hw, E1000_ETQF(i),
5727                                 filter_info->ethertype_filters[i].etqf);
5728                         E1000_WRITE_FLUSH(hw);
5729                 }
5730         }
5731 }
5732
5733 /* restore flex byte filter */
5734 static inline void
5735 igb_flex_filter_restore(struct rte_eth_dev *dev)
5736 {
5737         struct e1000_filter_info *filter_info =
5738                 E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
5739         struct e1000_flex_filter *flex_filter;
5740
5741         TAILQ_FOREACH(flex_filter, &filter_info->flex_list, entries) {
5742                 igb_inject_flex_filter(dev, flex_filter);
5743         }
5744 }
5745
5746 /* restore rss filter */
5747 static inline void
5748 igb_rss_filter_restore(struct rte_eth_dev *dev)
5749 {
5750         struct e1000_filter_info *filter_info =
5751                 E1000_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
5752
5753         if (filter_info->rss_info.conf.queue_num)
5754                 igb_config_rss_filter(dev, &filter_info->rss_info, TRUE);
5755 }
5756
5757 /* restore all types filter */
5758 static int
5759 igb_filter_restore(struct rte_eth_dev *dev)
5760 {
5761         igb_ntuple_filter_restore(dev);
5762         igb_ethertype_filter_restore(dev);
5763         igb_syn_filter_restore(dev);
5764         igb_flex_filter_restore(dev);
5765         igb_rss_filter_restore(dev);
5766
5767         return 0;
5768 }
5769
5770 RTE_PMD_REGISTER_PCI(net_e1000_igb, rte_igb_pmd);
5771 RTE_PMD_REGISTER_PCI_TABLE(net_e1000_igb, pci_id_igb_map);
5772 RTE_PMD_REGISTER_KMOD_DEP(net_e1000_igb, "* igb_uio | uio_pci_generic | vfio-pci");
5773 RTE_PMD_REGISTER_PCI(net_e1000_igb_vf, rte_igbvf_pmd);
5774 RTE_PMD_REGISTER_PCI_TABLE(net_e1000_igb_vf, pci_id_igbvf_map);
5775 RTE_PMD_REGISTER_KMOD_DEP(net_e1000_igb_vf, "* igb_uio | vfio-pci");
5776
5777 /* see e1000_logs.c */
5778 RTE_INIT(e1000_init_log)
5779 {
5780         e1000_igb_init_log();
5781 }