drivers/net/ixgbe/ixgbe_ethdev.c

   1 /*-
   2  *   BSD LICENSE
   3  *
   4  *   Copyright(c) 2010-2015 Intel Corporation. All rights reserved.
   5  *   All rights reserved.
   6  *
   7  *   Redistribution and use in source and binary forms, with or without
   8  *   modification, are permitted provided that the following conditions
   9  *   are met:
  10  *
  11  *     * Redistributions of source code must retain the above copyright
  12  *       notice, this list of conditions and the following disclaimer.
  13  *     * Redistributions in binary form must reproduce the above copyright
  14  *       notice, this list of conditions and the following disclaimer in
  15  *       the documentation and/or other materials provided with the
  16  *       distribution.
  17  *     * Neither the name of Intel Corporation nor the names of its
  18  *       contributors may be used to endorse or promote products derived
  19  *       from this software without specific prior written permission.
  20  *
  21  *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
  22  *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
  23  *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
  24  *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
  25  *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
  26  *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
  27  *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
  28  *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
  29  *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
  30  *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
  31  *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  32  */
  33
  34 #include <sys/queue.h>
  35 #include <stdio.h>
  36 #include <errno.h>
  37 #include <stdint.h>
  38 #include <string.h>
  39 #include <unistd.h>
  40 #include <stdarg.h>
  41 #include <inttypes.h>
  42 #include <netinet/in.h>
  43 #include <rte_byteorder.h>
  44 #include <rte_common.h>
  45 #include <rte_cycles.h>
  46
  47 #include <rte_interrupts.h>
  48 #include <rte_log.h>
  49 #include <rte_debug.h>
  50 #include <rte_pci.h>
  51 #include <rte_atomic.h>
  52 #include <rte_branch_prediction.h>
  53 #include <rte_memory.h>
  54 #include <rte_memzone.h>
  55 #include <rte_eal.h>
  56 #include <rte_alarm.h>
  57 #include <rte_ether.h>
  58 #include <rte_ethdev.h>
  59 #include <rte_atomic.h>
  60 #include <rte_malloc.h>
  61 #include <rte_random.h>
  62 #include <rte_dev.h>
  63
  64 #include "ixgbe_logs.h"
  65 #include "base/ixgbe_api.h"
  66 #include "base/ixgbe_vf.h"
  67 #include "base/ixgbe_common.h"
  68 #include "ixgbe_ethdev.h"
  69 #include "ixgbe_bypass.h"
  70 #include "ixgbe_rxtx.h"
  71 #include "base/ixgbe_type.h"
  72 #include "base/ixgbe_phy.h"
  73 #include "ixgbe_regs.h"
  74
  75 /*
  76  * High threshold controlling when to start sending XOFF frames. Must be at
  77  * least 8 bytes less than receive packet buffer size. This value is in units
  78  * of 1024 bytes.
  79  */
  80 #define IXGBE_FC_HI    0x80
  81
  82 /*
  83  * Low threshold controlling when to start sending XON frames. This value is
  84  * in units of 1024 bytes.
  85  */
  86 #define IXGBE_FC_LO    0x40
  87
  88 /* Default minimum inter-interrupt interval for EITR configuration */
  89 #define IXGBE_MIN_INTER_INTERRUPT_INTERVAL_DEFAULT    0x79E
  90
  91 /* Timer value included in XOFF frames. */
  92 #define IXGBE_FC_PAUSE 0x680
  93
  94 #define IXGBE_LINK_DOWN_CHECK_TIMEOUT 4000 /* ms */
  95 #define IXGBE_LINK_UP_CHECK_TIMEOUT   1000 /* ms */
  96 #define IXGBE_VMDQ_NUM_UC_MAC         4096 /* Maximum nb. of UC MAC addr. */
  97
  98 #define IXGBE_MMW_SIZE_DEFAULT        0x4
  99 #define IXGBE_MMW_SIZE_JUMBO_FRAME    0x14
 100 #define IXGBE_MAX_RING_DESC           4096 /* replicate define from rxtx */
 101
 102 /*
 103  *  Default values for RX/TX configuration
 104  */
 105 #define IXGBE_DEFAULT_RX_FREE_THRESH  32
 106 #define IXGBE_DEFAULT_RX_PTHRESH      8
 107 #define IXGBE_DEFAULT_RX_HTHRESH      8
 108 #define IXGBE_DEFAULT_RX_WTHRESH      0
 109
 110 #define IXGBE_DEFAULT_TX_FREE_THRESH  32
 111 #define IXGBE_DEFAULT_TX_PTHRESH      32
 112 #define IXGBE_DEFAULT_TX_HTHRESH      0
 113 #define IXGBE_DEFAULT_TX_WTHRESH      0
 114 #define IXGBE_DEFAULT_TX_RSBIT_THRESH 32
 115
 116 /* Bit shift and mask */
 117 #define IXGBE_4_BIT_WIDTH  (CHAR_BIT / 2)
 118 #define IXGBE_4_BIT_MASK   RTE_LEN2MASK(IXGBE_4_BIT_WIDTH, uint8_t)
 119 #define IXGBE_8_BIT_WIDTH  CHAR_BIT
 120 #define IXGBE_8_BIT_MASK   UINT8_MAX
 121
 122 #define IXGBEVF_PMD_NAME "rte_ixgbevf_pmd" /* PMD name */
 123
 124 #define IXGBE_QUEUE_STAT_COUNTERS (sizeof(hw_stats->qprc) / sizeof(hw_stats->qprc[0]))
 125
 126 #define IXGBE_HKEY_MAX_INDEX 10
 127
 128 /* Additional timesync values. */
 129 #define IXGBE_TIMINCA_16NS_SHIFT 24
 130 #define IXGBE_TIMINCA_INCVALUE   16000000
 131 #define IXGBE_TIMINCA_INIT       ((0x02 << IXGBE_TIMINCA_16NS_SHIFT) \
 132                                   | IXGBE_TIMINCA_INCVALUE)
 133
 134 static int eth_ixgbe_dev_init(struct rte_eth_dev *eth_dev);
 135 static int eth_ixgbe_dev_uninit(struct rte_eth_dev *eth_dev);
 136 static int  ixgbe_dev_configure(struct rte_eth_dev *dev);
 137 static int  ixgbe_dev_start(struct rte_eth_dev *dev);
 138 static void ixgbe_dev_stop(struct rte_eth_dev *dev);
 139 static int  ixgbe_dev_set_link_up(struct rte_eth_dev *dev);
 140 static int  ixgbe_dev_set_link_down(struct rte_eth_dev *dev);
 141 static void ixgbe_dev_close(struct rte_eth_dev *dev);
 142 static void ixgbe_dev_promiscuous_enable(struct rte_eth_dev *dev);
 143 static void ixgbe_dev_promiscuous_disable(struct rte_eth_dev *dev);
 144 static void ixgbe_dev_allmulticast_enable(struct rte_eth_dev *dev);
 145 static void ixgbe_dev_allmulticast_disable(struct rte_eth_dev *dev);
 146 static int ixgbe_dev_link_update(struct rte_eth_dev *dev,
 147                                 int wait_to_complete);
 148 static void ixgbe_dev_stats_get(struct rte_eth_dev *dev,
 149                                 struct rte_eth_stats *stats);
 150 static int ixgbe_dev_xstats_get(struct rte_eth_dev *dev,
 151                                 struct rte_eth_xstats *xstats, unsigned n);
 152 static void ixgbe_dev_stats_reset(struct rte_eth_dev *dev);
 153 static void ixgbe_dev_xstats_reset(struct rte_eth_dev *dev);
 154 static int ixgbe_dev_queue_stats_mapping_set(struct rte_eth_dev *eth_dev,
 155                                              uint16_t queue_id,
 156                                              uint8_t stat_idx,
 157                                              uint8_t is_rx);
 158 static void ixgbe_dev_info_get(struct rte_eth_dev *dev,
 159                                struct rte_eth_dev_info *dev_info);
 160 static void ixgbevf_dev_info_get(struct rte_eth_dev *dev,
 161                                  struct rte_eth_dev_info *dev_info);
 162 static int ixgbe_dev_mtu_set(struct rte_eth_dev *dev, uint16_t mtu);
 163
 164 static int ixgbe_vlan_filter_set(struct rte_eth_dev *dev,
 165                 uint16_t vlan_id, int on);
 166 static void ixgbe_vlan_tpid_set(struct rte_eth_dev *dev, uint16_t tpid_id);
 167 static void ixgbe_vlan_hw_strip_bitmap_set(struct rte_eth_dev *dev,
 168                 uint16_t queue, bool on);
 169 static void ixgbe_vlan_strip_queue_set(struct rte_eth_dev *dev, uint16_t queue,
 170                 int on);
 171 static void ixgbe_vlan_offload_set(struct rte_eth_dev *dev, int mask);
 172 static void ixgbe_vlan_hw_strip_enable(struct rte_eth_dev *dev, uint16_t queue);
 173 static void ixgbe_vlan_hw_strip_disable(struct rte_eth_dev *dev, uint16_t queue);
 174 static void ixgbe_vlan_hw_extend_enable(struct rte_eth_dev *dev);
 175 static void ixgbe_vlan_hw_extend_disable(struct rte_eth_dev *dev);
 176
 177 static int ixgbe_dev_led_on(struct rte_eth_dev *dev);
 178 static int ixgbe_dev_led_off(struct rte_eth_dev *dev);
 179 static int ixgbe_flow_ctrl_get(struct rte_eth_dev *dev,
 180                                struct rte_eth_fc_conf *fc_conf);
 181 static int ixgbe_flow_ctrl_set(struct rte_eth_dev *dev,
 182                                struct rte_eth_fc_conf *fc_conf);
 183 static int ixgbe_priority_flow_ctrl_set(struct rte_eth_dev *dev,
 184                 struct rte_eth_pfc_conf *pfc_conf);
 185 static int ixgbe_dev_rss_reta_update(struct rte_eth_dev *dev,
 186                         struct rte_eth_rss_reta_entry64 *reta_conf,
 187                         uint16_t reta_size);
 188 static int ixgbe_dev_rss_reta_query(struct rte_eth_dev *dev,
 189                         struct rte_eth_rss_reta_entry64 *reta_conf,
 190                         uint16_t reta_size);
 191 static void ixgbe_dev_link_status_print(struct rte_eth_dev *dev);
 192 static int ixgbe_dev_lsc_interrupt_setup(struct rte_eth_dev *dev);
 193 static int ixgbe_dev_rxq_interrupt_setup(struct rte_eth_dev *dev);
 194 static int ixgbe_dev_interrupt_get_status(struct rte_eth_dev *dev);
 195 static int ixgbe_dev_interrupt_action(struct rte_eth_dev *dev);
 196 static void ixgbe_dev_interrupt_handler(struct rte_intr_handle *handle,
 197                 void *param);
 198 static void ixgbe_dev_interrupt_delayed_handler(void *param);
 199 static void ixgbe_add_rar(struct rte_eth_dev *dev, struct ether_addr *mac_addr,
 200                 uint32_t index, uint32_t pool);
 201 static void ixgbe_remove_rar(struct rte_eth_dev *dev, uint32_t index);
 202 static void ixgbe_set_default_mac_addr(struct rte_eth_dev *dev,
 203                                            struct ether_addr *mac_addr);
 204 static void ixgbe_dcb_init(struct ixgbe_hw *hw,struct ixgbe_dcb_config *dcb_config);
 205
 206 /* For Virtual Function support */
 207 static int eth_ixgbevf_dev_init(struct rte_eth_dev *eth_dev);
 208 static int eth_ixgbevf_dev_uninit(struct rte_eth_dev *eth_dev);
 209 static int ixgbevf_dev_interrupt_get_status(struct rte_eth_dev *dev);
 210 static int ixgbevf_dev_interrupt_action(struct rte_eth_dev *dev);
 211 static int  ixgbevf_dev_configure(struct rte_eth_dev *dev);
 212 static int  ixgbevf_dev_start(struct rte_eth_dev *dev);
 213 static void ixgbevf_dev_stop(struct rte_eth_dev *dev);
 214 static void ixgbevf_dev_close(struct rte_eth_dev *dev);
 215 static void ixgbevf_intr_disable(struct ixgbe_hw *hw);
 216 static void ixgbevf_intr_enable(struct ixgbe_hw *hw);
 217 static void ixgbevf_dev_stats_get(struct rte_eth_dev *dev,
 218                 struct rte_eth_stats *stats);
 219 static void ixgbevf_dev_stats_reset(struct rte_eth_dev *dev);
 220 static int ixgbevf_vlan_filter_set(struct rte_eth_dev *dev,
 221                 uint16_t vlan_id, int on);
 222 static void ixgbevf_vlan_strip_queue_set(struct rte_eth_dev *dev,
 223                 uint16_t queue, int on);
 224 static void ixgbevf_vlan_offload_set(struct rte_eth_dev *dev, int mask);
 225 static void ixgbevf_set_vfta_all(struct rte_eth_dev *dev, bool on);
 226 static void ixgbevf_dev_interrupt_handler(struct rte_intr_handle *handle,
 227                                           void *param);
 228 static int ixgbevf_dev_rx_queue_intr_enable(struct rte_eth_dev *dev,
 229                                             uint16_t queue_id);
 230 static int ixgbevf_dev_rx_queue_intr_disable(struct rte_eth_dev *dev,
 231                                              uint16_t queue_id);
 232 static void ixgbevf_set_ivar_map(struct ixgbe_hw *hw, int8_t direction,
 233                                  uint8_t queue, uint8_t msix_vector);
 234 static void ixgbevf_configure_msix(struct rte_eth_dev *dev);
 235
 236 /* For Eth VMDQ APIs support */
 237 static int ixgbe_uc_hash_table_set(struct rte_eth_dev *dev, struct
 238                 ether_addr* mac_addr,uint8_t on);
 239 static int ixgbe_uc_all_hash_table_set(struct rte_eth_dev *dev,uint8_t on);
 240 static int  ixgbe_set_pool_rx_mode(struct rte_eth_dev *dev,  uint16_t pool,
 241                 uint16_t rx_mask, uint8_t on);
 242 static int ixgbe_set_pool_rx(struct rte_eth_dev *dev,uint16_t pool,uint8_t on);
 243 static int ixgbe_set_pool_tx(struct rte_eth_dev *dev,uint16_t pool,uint8_t on);
 244 static int ixgbe_set_pool_vlan_filter(struct rte_eth_dev *dev, uint16_t vlan,
 245                 uint64_t pool_mask,uint8_t vlan_on);
 246 static int ixgbe_mirror_rule_set(struct rte_eth_dev *dev,
 247                 struct rte_eth_mirror_conf *mirror_conf,
 248                 uint8_t rule_id, uint8_t on);
 249 static int ixgbe_mirror_rule_reset(struct rte_eth_dev *dev,
 250                 uint8_t rule_id);
 251 static int ixgbe_dev_rx_queue_intr_enable(struct rte_eth_dev *dev,
 252                                           uint16_t queue_id);
 253 static int ixgbe_dev_rx_queue_intr_disable(struct rte_eth_dev *dev,
 254                                            uint16_t queue_id);
 255 static void ixgbe_set_ivar_map(struct ixgbe_hw *hw, int8_t direction,
 256                                uint8_t queue, uint8_t msix_vector);
 257 static void ixgbe_configure_msix(struct rte_eth_dev *dev);
 258
 259 static int ixgbe_set_queue_rate_limit(struct rte_eth_dev *dev,
 260                 uint16_t queue_idx, uint16_t tx_rate);
 261 static int ixgbe_set_vf_rate_limit(struct rte_eth_dev *dev, uint16_t vf,
 262                 uint16_t tx_rate, uint64_t q_msk);
 263
 264 static void ixgbevf_add_mac_addr(struct rte_eth_dev *dev,
 265                                  struct ether_addr *mac_addr,
 266                                  uint32_t index, uint32_t pool);
 267 static void ixgbevf_remove_mac_addr(struct rte_eth_dev *dev, uint32_t index);
 268 static void ixgbevf_set_default_mac_addr(struct rte_eth_dev *dev,
 269                                              struct ether_addr *mac_addr);
 270 static int ixgbe_syn_filter_set(struct rte_eth_dev *dev,
 271                         struct rte_eth_syn_filter *filter,
 272                         bool add);
 273 static int ixgbe_syn_filter_get(struct rte_eth_dev *dev,
 274                         struct rte_eth_syn_filter *filter);
 275 static int ixgbe_syn_filter_handle(struct rte_eth_dev *dev,
 276                         enum rte_filter_op filter_op,
 277                         void *arg);
 278 static int ixgbe_add_5tuple_filter(struct rte_eth_dev *dev,
 279                         struct ixgbe_5tuple_filter *filter);
 280 static void ixgbe_remove_5tuple_filter(struct rte_eth_dev *dev,
 281                         struct ixgbe_5tuple_filter *filter);
 282 static int ixgbe_add_del_ntuple_filter(struct rte_eth_dev *dev,
 283                         struct rte_eth_ntuple_filter *filter,
 284                         bool add);
 285 static int ixgbe_ntuple_filter_handle(struct rte_eth_dev *dev,
 286                                 enum rte_filter_op filter_op,
 287                                 void *arg);
 288 static int ixgbe_get_ntuple_filter(struct rte_eth_dev *dev,
 289                         struct rte_eth_ntuple_filter *filter);
 290 static int ixgbe_add_del_ethertype_filter(struct rte_eth_dev *dev,
 291                         struct rte_eth_ethertype_filter *filter,
 292                         bool add);
 293 static int ixgbe_ethertype_filter_handle(struct rte_eth_dev *dev,
 294                                 enum rte_filter_op filter_op,
 295                                 void *arg);
 296 static int ixgbe_get_ethertype_filter(struct rte_eth_dev *dev,
 297                         struct rte_eth_ethertype_filter *filter);
 298 static int ixgbe_dev_filter_ctrl(struct rte_eth_dev *dev,
 299                      enum rte_filter_type filter_type,
 300                      enum rte_filter_op filter_op,
 301                      void *arg);
 302 static int ixgbevf_dev_set_mtu(struct rte_eth_dev *dev, uint16_t mtu);
 303
 304 static int ixgbe_dev_set_mc_addr_list(struct rte_eth_dev *dev,
 305                                       struct ether_addr *mc_addr_set,
 306                                       uint32_t nb_mc_addr);
 307
 308 static int ixgbe_get_reg_length(struct rte_eth_dev *dev);
 309 static int ixgbe_get_regs(struct rte_eth_dev *dev,
 310                             struct rte_dev_reg_info *regs);
 311 static int ixgbe_get_eeprom_length(struct rte_eth_dev *dev);
 312 static int ixgbe_get_eeprom(struct rte_eth_dev *dev,
 313                                 struct rte_dev_eeprom_info *eeprom);
 314 static int ixgbe_set_eeprom(struct rte_eth_dev *dev,
 315                                 struct rte_dev_eeprom_info *eeprom);
 316
 317 static int ixgbevf_get_reg_length(struct rte_eth_dev *dev);
 318 static int ixgbevf_get_regs(struct rte_eth_dev *dev,
 319                                 struct rte_dev_reg_info *regs);
 320
 321 static int ixgbe_timesync_enable(struct rte_eth_dev *dev);
 322 static int ixgbe_timesync_disable(struct rte_eth_dev *dev);
 323 static int ixgbe_timesync_read_rx_timestamp(struct rte_eth_dev *dev,
 324                                             struct timespec *timestamp,
 325                                             uint32_t flags);
 326 static int ixgbe_timesync_read_tx_timestamp(struct rte_eth_dev *dev,
 327                                             struct timespec *timestamp);
 328
 329 /*
 330  * Define VF Stats MACRO for Non "cleared on read" register
 331  */
 332 #define UPDATE_VF_STAT(reg, last, cur)                          \
 333 {                                                               \
 334         uint32_t latest = IXGBE_READ_REG(hw, reg);              \
 335         cur += (latest - last) & UINT_MAX;                      \
 336         last = latest;                                          \
 337 }
 338
 339 #define UPDATE_VF_STAT_36BIT(lsb, msb, last, cur)                \
 340 {                                                                \
 341         u64 new_lsb = IXGBE_READ_REG(hw, lsb);                   \
 342         u64 new_msb = IXGBE_READ_REG(hw, msb);                   \
 343         u64 latest = ((new_msb << 32) | new_lsb);                \
 344         cur += (0x1000000000LL + latest - last) & 0xFFFFFFFFFLL; \
 345         last = latest;                                           \
 346 }
 347
 348 #define IXGBE_SET_HWSTRIP(h, q) do{\
 349                 uint32_t idx = (q) / (sizeof ((h)->bitmap[0]) * NBBY); \
 350                 uint32_t bit = (q) % (sizeof ((h)->bitmap[0]) * NBBY); \
 351                 (h)->bitmap[idx] |= 1 << bit;\
 352         }while(0)
 353
 354 #define IXGBE_CLEAR_HWSTRIP(h, q) do{\
 355                 uint32_t idx = (q) / (sizeof ((h)->bitmap[0]) * NBBY); \
 356                 uint32_t bit = (q) % (sizeof ((h)->bitmap[0]) * NBBY); \
 357                 (h)->bitmap[idx] &= ~(1 << bit);\
 358         }while(0)
 359
 360 #define IXGBE_GET_HWSTRIP(h, q, r) do{\
 361                 uint32_t idx = (q) / (sizeof ((h)->bitmap[0]) * NBBY); \
 362                 uint32_t bit = (q) % (sizeof ((h)->bitmap[0]) * NBBY); \
 363                 (r) = (h)->bitmap[idx] >> bit & 1;\
 364         }while(0)
 365
 366 /*
 367  * The set of PCI devices this driver supports
 368  */
 369 static const struct rte_pci_id pci_id_ixgbe_map[] = {
 370
 371 #define RTE_PCI_DEV_ID_DECL_IXGBE(vend, dev) {RTE_PCI_DEVICE(vend, dev)},
 372 #include "rte_pci_dev_ids.h"
 373
 374 { .vendor_id = 0, /* sentinel */ },
 375 };
 376
 377
 378 /*
 379  * The set of PCI devices this driver supports (for 82599 VF)
 380  */
 381 static const struct rte_pci_id pci_id_ixgbevf_map[] = {
 382
 383 #define RTE_PCI_DEV_ID_DECL_IXGBEVF(vend, dev) {RTE_PCI_DEVICE(vend, dev)},
 384 #include "rte_pci_dev_ids.h"
 385 { .vendor_id = 0, /* sentinel */ },
 386
 387 };
 388
 389 static const struct eth_dev_ops ixgbe_eth_dev_ops = {
 390         .dev_configure        = ixgbe_dev_configure,
 391         .dev_start            = ixgbe_dev_start,
 392         .dev_stop             = ixgbe_dev_stop,
 393         .dev_set_link_up    = ixgbe_dev_set_link_up,
 394         .dev_set_link_down  = ixgbe_dev_set_link_down,
 395         .dev_close            = ixgbe_dev_close,
 396         .promiscuous_enable   = ixgbe_dev_promiscuous_enable,
 397         .promiscuous_disable  = ixgbe_dev_promiscuous_disable,
 398         .allmulticast_enable  = ixgbe_dev_allmulticast_enable,
 399         .allmulticast_disable = ixgbe_dev_allmulticast_disable,
 400         .link_update          = ixgbe_dev_link_update,
 401         .stats_get            = ixgbe_dev_stats_get,
 402         .xstats_get           = ixgbe_dev_xstats_get,
 403         .stats_reset          = ixgbe_dev_stats_reset,
 404         .xstats_reset         = ixgbe_dev_xstats_reset,
 405         .queue_stats_mapping_set = ixgbe_dev_queue_stats_mapping_set,
 406         .dev_infos_get        = ixgbe_dev_info_get,
 407         .mtu_set              = ixgbe_dev_mtu_set,
 408         .vlan_filter_set      = ixgbe_vlan_filter_set,
 409         .vlan_tpid_set        = ixgbe_vlan_tpid_set,
 410         .vlan_offload_set     = ixgbe_vlan_offload_set,
 411         .vlan_strip_queue_set = ixgbe_vlan_strip_queue_set,
 412         .rx_queue_start       = ixgbe_dev_rx_queue_start,
 413         .rx_queue_stop        = ixgbe_dev_rx_queue_stop,
 414         .tx_queue_start       = ixgbe_dev_tx_queue_start,
 415         .tx_queue_stop        = ixgbe_dev_tx_queue_stop,
 416         .rx_queue_setup       = ixgbe_dev_rx_queue_setup,
 417         .rx_queue_intr_enable = ixgbe_dev_rx_queue_intr_enable,
 418         .rx_queue_intr_disable = ixgbe_dev_rx_queue_intr_disable,
 419         .rx_queue_release     = ixgbe_dev_rx_queue_release,
 420         .rx_queue_count       = ixgbe_dev_rx_queue_count,
 421         .rx_descriptor_done   = ixgbe_dev_rx_descriptor_done,
 422         .tx_queue_setup       = ixgbe_dev_tx_queue_setup,
 423         .tx_queue_release     = ixgbe_dev_tx_queue_release,
 424         .dev_led_on           = ixgbe_dev_led_on,
 425         .dev_led_off          = ixgbe_dev_led_off,
 426         .flow_ctrl_get        = ixgbe_flow_ctrl_get,
 427         .flow_ctrl_set        = ixgbe_flow_ctrl_set,
 428         .priority_flow_ctrl_set = ixgbe_priority_flow_ctrl_set,
 429         .mac_addr_add         = ixgbe_add_rar,
 430         .mac_addr_remove      = ixgbe_remove_rar,
 431         .mac_addr_set         = ixgbe_set_default_mac_addr,
 432         .uc_hash_table_set    = ixgbe_uc_hash_table_set,
 433         .uc_all_hash_table_set  = ixgbe_uc_all_hash_table_set,
 434         .mirror_rule_set      = ixgbe_mirror_rule_set,
 435         .mirror_rule_reset    = ixgbe_mirror_rule_reset,
 436         .set_vf_rx_mode       = ixgbe_set_pool_rx_mode,
 437         .set_vf_rx            = ixgbe_set_pool_rx,
 438         .set_vf_tx            = ixgbe_set_pool_tx,
 439         .set_vf_vlan_filter   = ixgbe_set_pool_vlan_filter,
 440         .set_queue_rate_limit = ixgbe_set_queue_rate_limit,
 441         .set_vf_rate_limit    = ixgbe_set_vf_rate_limit,
 442         .reta_update          = ixgbe_dev_rss_reta_update,
 443         .reta_query           = ixgbe_dev_rss_reta_query,
 444 #ifdef RTE_NIC_BYPASS
 445         .bypass_init          = ixgbe_bypass_init,
 446         .bypass_state_set     = ixgbe_bypass_state_store,
 447         .bypass_state_show    = ixgbe_bypass_state_show,
 448         .bypass_event_set     = ixgbe_bypass_event_store,
 449         .bypass_event_show    = ixgbe_bypass_event_show,
 450         .bypass_wd_timeout_set  = ixgbe_bypass_wd_timeout_store,
 451         .bypass_wd_timeout_show = ixgbe_bypass_wd_timeout_show,
 452         .bypass_ver_show      = ixgbe_bypass_ver_show,
 453         .bypass_wd_reset      = ixgbe_bypass_wd_reset,
 454 #endif /* RTE_NIC_BYPASS */
 455         .rss_hash_update      = ixgbe_dev_rss_hash_update,
 456         .rss_hash_conf_get    = ixgbe_dev_rss_hash_conf_get,
 457         .filter_ctrl          = ixgbe_dev_filter_ctrl,
 458         .set_mc_addr_list     = ixgbe_dev_set_mc_addr_list,
 459         .timesync_enable      = ixgbe_timesync_enable,
 460         .timesync_disable     = ixgbe_timesync_disable,
 461         .timesync_read_rx_timestamp = ixgbe_timesync_read_rx_timestamp,
 462         .timesync_read_tx_timestamp = ixgbe_timesync_read_tx_timestamp,
 463         .get_reg_length       = ixgbe_get_reg_length,
 464         .get_reg              = ixgbe_get_regs,
 465         .get_eeprom_length    = ixgbe_get_eeprom_length,
 466         .get_eeprom           = ixgbe_get_eeprom,
 467         .set_eeprom           = ixgbe_set_eeprom,
 468 };
 469
 470 /*
 471  * dev_ops for virtual function, bare necessities for basic vf
 472  * operation have been implemented
 473  */
 474 static const struct eth_dev_ops ixgbevf_eth_dev_ops = {
 475         .dev_configure        = ixgbevf_dev_configure,
 476         .dev_start            = ixgbevf_dev_start,
 477         .dev_stop             = ixgbevf_dev_stop,
 478         .link_update          = ixgbe_dev_link_update,
 479         .stats_get            = ixgbevf_dev_stats_get,
 480         .stats_reset          = ixgbevf_dev_stats_reset,
 481         .dev_close            = ixgbevf_dev_close,
 482         .dev_infos_get        = ixgbevf_dev_info_get,
 483         .mtu_set              = ixgbevf_dev_set_mtu,
 484         .vlan_filter_set      = ixgbevf_vlan_filter_set,
 485         .vlan_strip_queue_set = ixgbevf_vlan_strip_queue_set,
 486         .vlan_offload_set     = ixgbevf_vlan_offload_set,
 487         .rx_queue_setup       = ixgbe_dev_rx_queue_setup,
 488         .rx_queue_release     = ixgbe_dev_rx_queue_release,
 489         .rx_descriptor_done   = ixgbe_dev_rx_descriptor_done,
 490         .tx_queue_setup       = ixgbe_dev_tx_queue_setup,
 491         .tx_queue_release     = ixgbe_dev_tx_queue_release,
 492         .rx_queue_intr_enable = ixgbevf_dev_rx_queue_intr_enable,
 493         .rx_queue_intr_disable = ixgbevf_dev_rx_queue_intr_disable,
 494         .mac_addr_add         = ixgbevf_add_mac_addr,
 495         .mac_addr_remove      = ixgbevf_remove_mac_addr,
 496         .set_mc_addr_list     = ixgbe_dev_set_mc_addr_list,
 497         .mac_addr_set         = ixgbevf_set_default_mac_addr,
 498         .get_reg_length       = ixgbevf_get_reg_length,
 499         .get_reg              = ixgbevf_get_regs,
 500 };
 501
 502 /* store statistics names and its offset in stats structure */
 503 struct rte_ixgbe_xstats_name_off {
 504         char name[RTE_ETH_XSTATS_NAME_SIZE];
 505         unsigned offset;
 506 };
 507
 508 static const struct rte_ixgbe_xstats_name_off rte_ixgbe_stats_strings[] = {
 509         {"rx_illegal_byte_err", offsetof(struct ixgbe_hw_stats, errbc)},
 510         {"rx_len_err", offsetof(struct ixgbe_hw_stats, rlec)},
 511         {"rx_undersize_count", offsetof(struct ixgbe_hw_stats, ruc)},
 512         {"rx_oversize_count", offsetof(struct ixgbe_hw_stats, roc)},
 513         {"rx_fragment_count", offsetof(struct ixgbe_hw_stats, rfc)},
 514         {"rx_jabber_count", offsetof(struct ixgbe_hw_stats, rjc)},
 515         {"l3_l4_xsum_error", offsetof(struct ixgbe_hw_stats, xec)},
 516         {"mac_local_fault", offsetof(struct ixgbe_hw_stats, mlfc)},
 517         {"mac_remote_fault", offsetof(struct ixgbe_hw_stats, mrfc)},
 518         {"mac_short_pkt_discard", offsetof(struct ixgbe_hw_stats, mspdc)},
 519         {"fccrc_error", offsetof(struct ixgbe_hw_stats, fccrc)},
 520         {"fcoe_drop", offsetof(struct ixgbe_hw_stats, fcoerpdc)},
 521         {"fc_last_error", offsetof(struct ixgbe_hw_stats, fclast)},
 522         {"rx_broadcast_packets", offsetof(struct ixgbe_hw_stats, bprc)},
 523         {"rx_phy_multicast_packets", offsetof(struct ixgbe_hw_stats, mprc)},
 524         {"mgmt_pkts_dropped", offsetof(struct ixgbe_hw_stats, mngpdc)},
 525         {"rx_crc_errors", offsetof(struct ixgbe_hw_stats, crcerrs)},
 526         {"fdir_match", offsetof(struct ixgbe_hw_stats, fdirmatch)},
 527         {"fdir_miss", offsetof(struct ixgbe_hw_stats, fdirmiss)},
 528         {"tx_flow_control_xon", offsetof(struct ixgbe_hw_stats, lxontxc)},
 529         {"rx_flow_control_xon", offsetof(struct ixgbe_hw_stats, lxonrxc)},
 530         {"tx_flow_control_xoff", offsetof(struct ixgbe_hw_stats, lxofftxc)},
 531         {"rx_flow_control_xoff", offsetof(struct ixgbe_hw_stats, lxoffrxc)},
 532 };
 533
 534 #define IXGBE_NB_XSTATS (sizeof(rte_ixgbe_stats_strings) /      \
 535                 sizeof(rte_ixgbe_stats_strings[0]))
 536
 537 /**
 538  * Atomically reads the link status information from global
 539  * structure rte_eth_dev.
 540  *
 541  * @param dev
 542  *   - Pointer to the structure rte_eth_dev to read from.
 543  *   - Pointer to the buffer to be saved with the link status.
 544  *
 545  * @return
 546  *   - On success, zero.
 547  *   - On failure, negative value.
 548  */
 549 static inline int
 550 rte_ixgbe_dev_atomic_read_link_status(struct rte_eth_dev *dev,
 551                                 struct rte_eth_link *link)
 552 {
 553         struct rte_eth_link *dst = link;
 554         struct rte_eth_link *src = &(dev->data->dev_link);
 555
 556         if (rte_atomic64_cmpset((uint64_t *)dst, *(uint64_t *)dst,
 557                                         *(uint64_t *)src) == 0)
 558                 return -1;
 559
 560         return 0;
 561 }
 562
 563 /**
 564  * Atomically writes the link status information into global
 565  * structure rte_eth_dev.
 566  *
 567  * @param dev
 568  *   - Pointer to the structure rte_eth_dev to read from.
 569  *   - Pointer to the buffer to be saved with the link status.
 570  *
 571  * @return
 572  *   - On success, zero.
 573  *   - On failure, negative value.
 574  */
 575 static inline int
 576 rte_ixgbe_dev_atomic_write_link_status(struct rte_eth_dev *dev,
 577                                 struct rte_eth_link *link)
 578 {
 579         struct rte_eth_link *dst = &(dev->data->dev_link);
 580         struct rte_eth_link *src = link;
 581
 582         if (rte_atomic64_cmpset((uint64_t *)dst, *(uint64_t *)dst,
 583                                         *(uint64_t *)src) == 0)
 584                 return -1;
 585
 586         return 0;
 587 }
 588
 589 /*
 590  * This function is the same as ixgbe_is_sfp() in base/ixgbe.h.
 591  */
 592 static inline int
 593 ixgbe_is_sfp(struct ixgbe_hw *hw)
 594 {
 595         switch (hw->phy.type) {
 596         case ixgbe_phy_sfp_avago:
 597         case ixgbe_phy_sfp_ftl:
 598         case ixgbe_phy_sfp_intel:
 599         case ixgbe_phy_sfp_unknown:
 600         case ixgbe_phy_sfp_passive_tyco:
 601         case ixgbe_phy_sfp_passive_unknown:
 602                 return 1;
 603         default:
 604                 return 0;
 605         }
 606 }
 607
 608 static inline int32_t
 609 ixgbe_pf_reset_hw(struct ixgbe_hw *hw)
 610 {
 611         uint32_t ctrl_ext;
 612         int32_t status;
 613
 614         status = ixgbe_reset_hw(hw);
 615
 616         ctrl_ext = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
 617         /* Set PF Reset Done bit so PF/VF Mail Ops can work */
 618         ctrl_ext |= IXGBE_CTRL_EXT_PFRSTD;
 619         IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl_ext);
 620         IXGBE_WRITE_FLUSH(hw);
 621
 622         return status;
 623 }
 624
 625 static inline void
 626 ixgbe_enable_intr(struct rte_eth_dev *dev)
 627 {
 628         struct ixgbe_interrupt *intr =
 629                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
 630         struct ixgbe_hw *hw =
 631                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
 632
 633         IXGBE_WRITE_REG(hw, IXGBE_EIMS, intr->mask);
 634         IXGBE_WRITE_FLUSH(hw);
 635 }
 636
 637 /*
 638  * This function is based on ixgbe_disable_intr() in base/ixgbe.h.
 639  */
 640 static void
 641 ixgbe_disable_intr(struct ixgbe_hw *hw)
 642 {
 643         PMD_INIT_FUNC_TRACE();
 644
 645         if (hw->mac.type == ixgbe_mac_82598EB) {
 646                 IXGBE_WRITE_REG(hw, IXGBE_EIMC, ~0);
 647         } else {
 648                 IXGBE_WRITE_REG(hw, IXGBE_EIMC, 0xFFFF0000);
 649                 IXGBE_WRITE_REG(hw, IXGBE_EIMC_EX(0), ~0);
 650                 IXGBE_WRITE_REG(hw, IXGBE_EIMC_EX(1), ~0);
 651         }
 652         IXGBE_WRITE_FLUSH(hw);
 653 }
 654
 655 /*
 656  * This function resets queue statistics mapping registers.
 657  * From Niantic datasheet, Initialization of Statistics section:
 658  * "...if software requires the queue counters, the RQSMR and TQSM registers
 659  * must be re-programmed following a device reset.
 660  */
 661 static void
 662 ixgbe_reset_qstat_mappings(struct ixgbe_hw *hw)
 663 {
 664         uint32_t i;
 665
 666         for(i = 0; i != IXGBE_NB_STAT_MAPPING_REGS; i++) {
 667                 IXGBE_WRITE_REG(hw, IXGBE_RQSMR(i), 0);
 668                 IXGBE_WRITE_REG(hw, IXGBE_TQSM(i), 0);
 669         }
 670 }
 671
 672
 673 static int
 674 ixgbe_dev_queue_stats_mapping_set(struct rte_eth_dev *eth_dev,
 675                                   uint16_t queue_id,
 676                                   uint8_t stat_idx,
 677                                   uint8_t is_rx)
 678 {
 679 #define QSM_REG_NB_BITS_PER_QMAP_FIELD 8
 680 #define NB_QMAP_FIELDS_PER_QSM_REG 4
 681 #define QMAP_FIELD_RESERVED_BITS_MASK 0x0f
 682
 683         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
 684         struct ixgbe_stat_mapping_registers *stat_mappings =
 685                 IXGBE_DEV_PRIVATE_TO_STAT_MAPPINGS(eth_dev->data->dev_private);
 686         uint32_t qsmr_mask = 0;
 687         uint32_t clearing_mask = QMAP_FIELD_RESERVED_BITS_MASK;
 688         uint32_t q_map;
 689         uint8_t n, offset;
 690
 691         if ((hw->mac.type != ixgbe_mac_82599EB) &&
 692                 (hw->mac.type != ixgbe_mac_X540) &&
 693                 (hw->mac.type != ixgbe_mac_X550) &&
 694                 (hw->mac.type != ixgbe_mac_X550EM_x))
 695                 return -ENOSYS;
 696
 697         PMD_INIT_LOG(DEBUG, "Setting port %d, %s queue_id %d to stat index %d",
 698                      (int)(eth_dev->data->port_id), is_rx ? "RX" : "TX",
 699                      queue_id, stat_idx);
 700
 701         n = (uint8_t)(queue_id / NB_QMAP_FIELDS_PER_QSM_REG);
 702         if (n >= IXGBE_NB_STAT_MAPPING_REGS) {
 703                 PMD_INIT_LOG(ERR, "Nb of stat mapping registers exceeded");
 704                 return -EIO;
 705         }
 706         offset = (uint8_t)(queue_id % NB_QMAP_FIELDS_PER_QSM_REG);
 707
 708         /* Now clear any previous stat_idx set */
 709         clearing_mask <<= (QSM_REG_NB_BITS_PER_QMAP_FIELD * offset);
 710         if (!is_rx)
 711                 stat_mappings->tqsm[n] &= ~clearing_mask;
 712         else
 713                 stat_mappings->rqsmr[n] &= ~clearing_mask;
 714
 715         q_map = (uint32_t)stat_idx;
 716         q_map &= QMAP_FIELD_RESERVED_BITS_MASK;
 717         qsmr_mask = q_map << (QSM_REG_NB_BITS_PER_QMAP_FIELD * offset);
 718         if (!is_rx)
 719                 stat_mappings->tqsm[n] |= qsmr_mask;
 720         else
 721                 stat_mappings->rqsmr[n] |= qsmr_mask;
 722
 723         PMD_INIT_LOG(DEBUG, "Set port %d, %s queue_id %d to stat index %d",
 724                      (int)(eth_dev->data->port_id), is_rx ? "RX" : "TX",
 725                      queue_id, stat_idx);
 726         PMD_INIT_LOG(DEBUG, "%s[%d] = 0x%08x", is_rx ? "RQSMR" : "TQSM", n,
 727                      is_rx ? stat_mappings->rqsmr[n] : stat_mappings->tqsm[n]);
 728
 729         /* Now write the mapping in the appropriate register */
 730         if (is_rx) {
 731                 PMD_INIT_LOG(DEBUG, "Write 0x%x to RX IXGBE stat mapping reg:%d",
 732                              stat_mappings->rqsmr[n], n);
 733                 IXGBE_WRITE_REG(hw, IXGBE_RQSMR(n), stat_mappings->rqsmr[n]);
 734         }
 735         else {
 736                 PMD_INIT_LOG(DEBUG, "Write 0x%x to TX IXGBE stat mapping reg:%d",
 737                              stat_mappings->tqsm[n], n);
 738                 IXGBE_WRITE_REG(hw, IXGBE_TQSM(n), stat_mappings->tqsm[n]);
 739         }
 740         return 0;
 741 }
 742
 743 static void
 744 ixgbe_restore_statistics_mapping(struct rte_eth_dev * dev)
 745 {
 746         struct ixgbe_stat_mapping_registers *stat_mappings =
 747                 IXGBE_DEV_PRIVATE_TO_STAT_MAPPINGS(dev->data->dev_private);
 748         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
 749         int i;
 750
 751         /* write whatever was in stat mapping table to the NIC */
 752         for (i = 0; i < IXGBE_NB_STAT_MAPPING_REGS; i++) {
 753                 /* rx */
 754                 IXGBE_WRITE_REG(hw, IXGBE_RQSMR(i), stat_mappings->rqsmr[i]);
 755
 756                 /* tx */
 757                 IXGBE_WRITE_REG(hw, IXGBE_TQSM(i), stat_mappings->tqsm[i]);
 758         }
 759 }
 760
 761 static void
 762 ixgbe_dcb_init(struct ixgbe_hw *hw,struct ixgbe_dcb_config *dcb_config)
 763 {
 764         uint8_t i;
 765         struct ixgbe_dcb_tc_config *tc;
 766         uint8_t dcb_max_tc = IXGBE_DCB_MAX_TRAFFIC_CLASS;
 767
 768         dcb_config->num_tcs.pg_tcs = dcb_max_tc;
 769         dcb_config->num_tcs.pfc_tcs = dcb_max_tc;
 770         for (i = 0; i < dcb_max_tc; i++) {
 771                 tc = &dcb_config->tc_config[i];
 772                 tc->path[IXGBE_DCB_TX_CONFIG].bwg_id = i;
 773                 tc->path[IXGBE_DCB_TX_CONFIG].bwg_percent =
 774                                  (uint8_t)(100/dcb_max_tc + (i & 1));
 775                 tc->path[IXGBE_DCB_RX_CONFIG].bwg_id = i;
 776                 tc->path[IXGBE_DCB_RX_CONFIG].bwg_percent =
 777                                  (uint8_t)(100/dcb_max_tc + (i & 1));
 778                 tc->pfc = ixgbe_dcb_pfc_disabled;
 779         }
 780
 781         /* Initialize default user to priority mapping, UPx->TC0 */
 782         tc = &dcb_config->tc_config[0];
 783         tc->path[IXGBE_DCB_TX_CONFIG].up_to_tc_bitmap = 0xFF;
 784         tc->path[IXGBE_DCB_RX_CONFIG].up_to_tc_bitmap = 0xFF;
 785         for (i = 0; i< IXGBE_DCB_MAX_BW_GROUP; i++) {
 786                 dcb_config->bw_percentage[IXGBE_DCB_TX_CONFIG][i] = 100;
 787                 dcb_config->bw_percentage[IXGBE_DCB_RX_CONFIG][i] = 100;
 788         }
 789         dcb_config->rx_pba_cfg = ixgbe_dcb_pba_equal;
 790         dcb_config->pfc_mode_enable = false;
 791         dcb_config->vt_mode = true;
 792         dcb_config->round_robin_enable = false;
 793         /* support all DCB capabilities in 82599 */
 794         dcb_config->support.capabilities = 0xFF;
 795
 796         /*we only support 4 Tcs for X540, X550 */
 797         if (hw->mac.type == ixgbe_mac_X540 ||
 798                 hw->mac.type == ixgbe_mac_X550 ||
 799                 hw->mac.type == ixgbe_mac_X550EM_x) {
 800                 dcb_config->num_tcs.pg_tcs = 4;
 801                 dcb_config->num_tcs.pfc_tcs = 4;
 802         }
 803 }
 804
 805 /*
 806  * Ensure that all locks are released before first NVM or PHY access
 807  */
 808 static void
 809 ixgbe_swfw_lock_reset(struct ixgbe_hw *hw)
 810 {
 811         uint16_t mask;
 812
 813         /*
 814          * Phy lock should not fail in this early stage. If this is the case,
 815          * it is due to an improper exit of the application.
 816          * So force the release of the faulty lock. Release of common lock
 817          * is done automatically by swfw_sync function.
 818          */
 819         mask = IXGBE_GSSR_PHY0_SM << hw->bus.func;
 820         if (ixgbe_acquire_swfw_semaphore(hw, mask) < 0) {
 821                 PMD_DRV_LOG(DEBUG, "SWFW phy%d lock released", hw->bus.func);
 822         }
 823         ixgbe_release_swfw_semaphore(hw, mask);
 824
 825         /*
 826          * These ones are more tricky since they are common to all ports; but
 827          * swfw_sync retries last long enough (1s) to be almost sure that if
 828          * lock can not be taken it is due to an improper lock of the
 829          * semaphore.
 830          */
 831         mask = IXGBE_GSSR_EEP_SM | IXGBE_GSSR_MAC_CSR_SM | IXGBE_GSSR_SW_MNG_SM;
 832         if (ixgbe_acquire_swfw_semaphore(hw, mask) < 0) {
 833                 PMD_DRV_LOG(DEBUG, "SWFW common locks released");
 834         }
 835         ixgbe_release_swfw_semaphore(hw, mask);
 836 }
 837
 838 /*
 839  * This function is based on code in ixgbe_attach() in base/ixgbe.c.
 840  * It returns 0 on success.
 841  */
 842 static int
 843 eth_ixgbe_dev_init(struct rte_eth_dev *eth_dev)
 844 {
 845         struct rte_pci_device *pci_dev;
 846         struct ixgbe_hw *hw =
 847                 IXGBE_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
 848         struct ixgbe_vfta * shadow_vfta =
 849                 IXGBE_DEV_PRIVATE_TO_VFTA(eth_dev->data->dev_private);
 850         struct ixgbe_hwstrip *hwstrip =
 851                 IXGBE_DEV_PRIVATE_TO_HWSTRIP_BITMAP(eth_dev->data->dev_private);
 852         struct ixgbe_dcb_config *dcb_config =
 853                 IXGBE_DEV_PRIVATE_TO_DCB_CFG(eth_dev->data->dev_private);
 854         struct ixgbe_filter_info *filter_info =
 855                 IXGBE_DEV_PRIVATE_TO_FILTER_INFO(eth_dev->data->dev_private);
 856         uint32_t ctrl_ext;
 857         uint16_t csum;
 858         int diag, i;
 859
 860         PMD_INIT_FUNC_TRACE();
 861
 862         eth_dev->dev_ops = &ixgbe_eth_dev_ops;
 863         eth_dev->rx_pkt_burst = &ixgbe_recv_pkts;
 864         eth_dev->tx_pkt_burst = &ixgbe_xmit_pkts;
 865
 866         /*
 867          * For secondary processes, we don't initialise any further as primary
 868          * has already done this work. Only check we don't need a different
 869          * RX and TX function.
 870          */
 871         if (rte_eal_process_type() != RTE_PROC_PRIMARY){
 872                 struct ixgbe_tx_queue *txq;
 873                 /* TX queue function in primary, set by last queue initialized
 874                  * Tx queue may not initialized by primary process */
 875                 if (eth_dev->data->tx_queues) {
 876                         txq = eth_dev->data->tx_queues[eth_dev->data->nb_tx_queues-1];
 877                         ixgbe_set_tx_function(eth_dev, txq);
 878                 } else {
 879                         /* Use default TX function if we get here */
 880                         PMD_INIT_LOG(NOTICE, "No TX queues configured yet. "
 881                                              "Using default TX function.");
 882                 }
 883
 884                 ixgbe_set_rx_function(eth_dev);
 885
 886                 return 0;
 887         }
 888         pci_dev = eth_dev->pci_dev;
 889
 890         /* Vendor and Device ID need to be set before init of shared code */
 891         hw->device_id = pci_dev->id.device_id;
 892         hw->vendor_id = pci_dev->id.vendor_id;
 893         hw->hw_addr = (void *)pci_dev->mem_resource[0].addr;
 894         hw->allow_unsupported_sfp = 1;
 895
 896         /* Initialize the shared code (base driver) */
 897 #ifdef RTE_NIC_BYPASS
 898         diag = ixgbe_bypass_init_shared_code(hw);
 899 #else
 900         diag = ixgbe_init_shared_code(hw);
 901 #endif /* RTE_NIC_BYPASS */
 902
 903         if (diag != IXGBE_SUCCESS) {
 904                 PMD_INIT_LOG(ERR, "Shared code init failed: %d", diag);
 905                 return -EIO;
 906         }
 907
 908         /* pick up the PCI bus settings for reporting later */
 909         ixgbe_get_bus_info(hw);
 910
 911         /* Unlock any pending hardware semaphore */
 912         ixgbe_swfw_lock_reset(hw);
 913
 914         /* Initialize DCB configuration*/
 915         memset(dcb_config, 0, sizeof(struct ixgbe_dcb_config));
 916         ixgbe_dcb_init(hw,dcb_config);
 917         /* Get Hardware Flow Control setting */
 918         hw->fc.requested_mode = ixgbe_fc_full;
 919         hw->fc.current_mode = ixgbe_fc_full;
 920         hw->fc.pause_time = IXGBE_FC_PAUSE;
 921         for (i = 0; i < IXGBE_DCB_MAX_TRAFFIC_CLASS; i++) {
 922                 hw->fc.low_water[i] = IXGBE_FC_LO;
 923                 hw->fc.high_water[i] = IXGBE_FC_HI;
 924         }
 925         hw->fc.send_xon = 1;
 926
 927         /* Make sure we have a good EEPROM before we read from it */
 928         diag = ixgbe_validate_eeprom_checksum(hw, &csum);
 929         if (diag != IXGBE_SUCCESS) {
 930                 PMD_INIT_LOG(ERR, "The EEPROM checksum is not valid: %d", diag);
 931                 return -EIO;
 932         }
 933
 934 #ifdef RTE_NIC_BYPASS
 935         diag = ixgbe_bypass_init_hw(hw);
 936 #else
 937         diag = ixgbe_init_hw(hw);
 938 #endif /* RTE_NIC_BYPASS */
 939
 940         /*
 941          * Devices with copper phys will fail to initialise if ixgbe_init_hw()
 942          * is called too soon after the kernel driver unbinding/binding occurs.
 943          * The failure occurs in ixgbe_identify_phy_generic() for all devices,
 944          * but for non-copper devies, ixgbe_identify_sfp_module_generic() is
 945          * also called. See ixgbe_identify_phy_82599(). The reason for the
 946          * failure is not known, and only occuts when virtualisation features
 947          * are disabled in the bios. A delay of 100ms  was found to be enough by
 948          * trial-and-error, and is doubled to be safe.
 949          */
 950         if (diag && (hw->mac.ops.get_media_type(hw) == ixgbe_media_type_copper)) {
 951                 rte_delay_ms(200);
 952                 diag = ixgbe_init_hw(hw);
 953         }
 954
 955         if (diag == IXGBE_ERR_EEPROM_VERSION) {
 956                 PMD_INIT_LOG(ERR, "This device is a pre-production adapter/"
 957                     "LOM.  Please be aware there may be issues associated "
 958                     "with your hardware.");
 959                 PMD_INIT_LOG(ERR, "If you are experiencing problems "
 960                     "please contact your Intel or hardware representative "
 961                     "who provided you with this hardware.");
 962         } else if (diag == IXGBE_ERR_SFP_NOT_SUPPORTED)
 963                 PMD_INIT_LOG(ERR, "Unsupported SFP+ Module");
 964         if (diag) {
 965                 PMD_INIT_LOG(ERR, "Hardware Initialization Failure: %d", diag);
 966                 return -EIO;
 967         }
 968
 969         /* Reset the hw statistics */
 970         ixgbe_dev_stats_reset(eth_dev);
 971
 972         /* disable interrupt */
 973         ixgbe_disable_intr(hw);
 974
 975         /* reset mappings for queue statistics hw counters*/
 976         ixgbe_reset_qstat_mappings(hw);
 977
 978         /* Allocate memory for storing MAC addresses */
 979         eth_dev->data->mac_addrs = rte_zmalloc("ixgbe", ETHER_ADDR_LEN *
 980                         hw->mac.num_rar_entries, 0);
 981         if (eth_dev->data->mac_addrs == NULL) {
 982                 PMD_INIT_LOG(ERR,
 983                         "Failed to allocate %u bytes needed to store "
 984                         "MAC addresses",
 985                         ETHER_ADDR_LEN * hw->mac.num_rar_entries);
 986                 return -ENOMEM;
 987         }
 988         /* Copy the permanent MAC address */
 989         ether_addr_copy((struct ether_addr *) hw->mac.perm_addr,
 990                         &eth_dev->data->mac_addrs[0]);
 991
 992         /* Allocate memory for storing hash filter MAC addresses */
 993         eth_dev->data->hash_mac_addrs = rte_zmalloc("ixgbe", ETHER_ADDR_LEN *
 994                         IXGBE_VMDQ_NUM_UC_MAC, 0);
 995         if (eth_dev->data->hash_mac_addrs == NULL) {
 996                 PMD_INIT_LOG(ERR,
 997                         "Failed to allocate %d bytes needed to store MAC addresses",
 998                         ETHER_ADDR_LEN * IXGBE_VMDQ_NUM_UC_MAC);
 999                 return -ENOMEM;
1000         }
1001
1002         /* initialize the vfta */
1003         memset(shadow_vfta, 0, sizeof(*shadow_vfta));
1004
1005         /* initialize the hw strip bitmap*/
1006         memset(hwstrip, 0, sizeof(*hwstrip));
1007
1008         /* initialize PF if max_vfs not zero */
1009         ixgbe_pf_host_init(eth_dev);
1010
1011         ctrl_ext = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
1012         /* let hardware know driver is loaded */
1013         ctrl_ext |= IXGBE_CTRL_EXT_DRV_LOAD;
1014         /* Set PF Reset Done bit so PF/VF Mail Ops can work */
1015         ctrl_ext |= IXGBE_CTRL_EXT_PFRSTD;
1016         IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl_ext);
1017         IXGBE_WRITE_FLUSH(hw);
1018
1019         if (ixgbe_is_sfp(hw) && hw->phy.sfp_type != ixgbe_sfp_type_not_present)
1020                 PMD_INIT_LOG(DEBUG, "MAC: %d, PHY: %d, SFP+: %d",
1021                              (int) hw->mac.type, (int) hw->phy.type,
1022                              (int) hw->phy.sfp_type);
1023         else
1024                 PMD_INIT_LOG(DEBUG, "MAC: %d, PHY: %d",
1025                              (int) hw->mac.type, (int) hw->phy.type);
1026
1027         PMD_INIT_LOG(DEBUG, "port %d vendorID=0x%x deviceID=0x%x",
1028                         eth_dev->data->port_id, pci_dev->id.vendor_id,
1029                         pci_dev->id.device_id);
1030
1031         /* enable support intr */
1032         ixgbe_enable_intr(eth_dev);
1033
1034         /* initialize 5tuple filter list */
1035         TAILQ_INIT(&filter_info->fivetuple_list);
1036         memset(filter_info->fivetuple_mask, 0,
1037                 sizeof(uint32_t) * IXGBE_5TUPLE_ARRAY_SIZE);
1038
1039         return 0;
1040 }
1041
1042 static int
1043 eth_ixgbe_dev_uninit(struct rte_eth_dev *eth_dev)
1044 {
1045         struct rte_pci_device *pci_dev;
1046         struct ixgbe_hw *hw;
1047
1048         PMD_INIT_FUNC_TRACE();
1049
1050         if (rte_eal_process_type() != RTE_PROC_PRIMARY)
1051                 return -EPERM;
1052
1053         hw = IXGBE_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
1054         pci_dev = eth_dev->pci_dev;
1055
1056         if (hw->adapter_stopped == 0)
1057                 ixgbe_dev_close(eth_dev);
1058
1059         eth_dev->dev_ops = NULL;
1060         eth_dev->rx_pkt_burst = NULL;
1061         eth_dev->tx_pkt_burst = NULL;
1062
1063         /* Unlock any pending hardware semaphore */
1064         ixgbe_swfw_lock_reset(hw);
1065
1066         /* disable uio intr before callback unregister */
1067         rte_intr_disable(&(pci_dev->intr_handle));
1068         rte_intr_callback_unregister(&(pci_dev->intr_handle),
1069                 ixgbe_dev_interrupt_handler, (void *)eth_dev);
1070
1071         /* uninitialize PF if max_vfs not zero */
1072         ixgbe_pf_host_uninit(eth_dev);
1073
1074         rte_free(eth_dev->data->mac_addrs);
1075         eth_dev->data->mac_addrs = NULL;
1076
1077         rte_free(eth_dev->data->hash_mac_addrs);
1078         eth_dev->data->hash_mac_addrs = NULL;
1079
1080         return 0;
1081 }
1082
1083 /*
1084  * Negotiate mailbox API version with the PF.
1085  * After reset API version is always set to the basic one (ixgbe_mbox_api_10).
1086  * Then we try to negotiate starting with the most recent one.
1087  * If all negotiation attempts fail, then we will proceed with
1088  * the default one (ixgbe_mbox_api_10).
1089  */
1090 static void
1091 ixgbevf_negotiate_api(struct ixgbe_hw *hw)
1092 {
1093         int32_t i;
1094
1095         /* start with highest supported, proceed down */
1096         static const enum ixgbe_pfvf_api_rev sup_ver[] = {
1097                 ixgbe_mbox_api_11,
1098                 ixgbe_mbox_api_10,
1099         };
1100
1101         for (i = 0;
1102                         i != RTE_DIM(sup_ver) &&
1103                         ixgbevf_negotiate_api_version(hw, sup_ver[i]) != 0;
1104                         i++)
1105                 ;
1106 }
1107
1108 static void
1109 generate_random_mac_addr(struct ether_addr *mac_addr)
1110 {
1111         uint64_t random;
1112
1113         /* Set Organizationally Unique Identifier (OUI) prefix. */
1114         mac_addr->addr_bytes[0] = 0x00;
1115         mac_addr->addr_bytes[1] = 0x09;
1116         mac_addr->addr_bytes[2] = 0xC0;
1117         /* Force indication of locally assigned MAC address. */
1118         mac_addr->addr_bytes[0] |= ETHER_LOCAL_ADMIN_ADDR;
1119         /* Generate the last 3 bytes of the MAC address with a random number. */
1120         random = rte_rand();
1121         memcpy(&mac_addr->addr_bytes[3], &random, 3);
1122 }
1123
1124 /*
1125  * Virtual Function device init
1126  */
1127 static int
1128 eth_ixgbevf_dev_init(struct rte_eth_dev *eth_dev)
1129 {
1130         int diag;
1131         uint32_t tc, tcs;
1132         struct rte_pci_device *pci_dev;
1133         struct ixgbe_hw *hw =
1134                 IXGBE_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
1135         struct ixgbe_vfta * shadow_vfta =
1136                 IXGBE_DEV_PRIVATE_TO_VFTA(eth_dev->data->dev_private);
1137         struct ixgbe_hwstrip *hwstrip =
1138                 IXGBE_DEV_PRIVATE_TO_HWSTRIP_BITMAP(eth_dev->data->dev_private);
1139         struct ether_addr *perm_addr = (struct ether_addr *) hw->mac.perm_addr;
1140
1141         PMD_INIT_FUNC_TRACE();
1142
1143         eth_dev->dev_ops = &ixgbevf_eth_dev_ops;
1144         eth_dev->rx_pkt_burst = &ixgbe_recv_pkts;
1145         eth_dev->tx_pkt_burst = &ixgbe_xmit_pkts;
1146
1147         /* for secondary processes, we don't initialise any further as primary
1148          * has already done this work. Only check we don't need a different
1149          * RX function */
1150         if (rte_eal_process_type() != RTE_PROC_PRIMARY){
1151                 if (eth_dev->data->scattered_rx)
1152                         eth_dev->rx_pkt_burst = ixgbe_recv_pkts_lro_single_alloc;
1153                 return 0;
1154         }
1155
1156         pci_dev = eth_dev->pci_dev;
1157
1158         hw->device_id = pci_dev->id.device_id;
1159         hw->vendor_id = pci_dev->id.vendor_id;
1160         hw->hw_addr = (void *)pci_dev->mem_resource[0].addr;
1161
1162         /* initialize the vfta */
1163         memset(shadow_vfta, 0, sizeof(*shadow_vfta));
1164
1165         /* initialize the hw strip bitmap*/
1166         memset(hwstrip, 0, sizeof(*hwstrip));
1167
1168         /* Initialize the shared code (base driver) */
1169         diag = ixgbe_init_shared_code(hw);
1170         if (diag != IXGBE_SUCCESS) {
1171                 PMD_INIT_LOG(ERR, "Shared code init failed for ixgbevf: %d", diag);
1172                 return -EIO;
1173         }
1174
1175         /* init_mailbox_params */
1176         hw->mbx.ops.init_params(hw);
1177
1178         /* Reset the hw statistics */
1179         ixgbevf_dev_stats_reset(eth_dev);
1180
1181         /* Disable the interrupts for VF */
1182         ixgbevf_intr_disable(hw);
1183
1184         hw->mac.num_rar_entries = 128; /* The MAX of the underlying PF */
1185         diag = hw->mac.ops.reset_hw(hw);
1186
1187         /*
1188          * The VF reset operation returns the IXGBE_ERR_INVALID_MAC_ADDR when
1189          * the underlying PF driver has not assigned a MAC address to the VF.
1190          * In this case, assign a random MAC address.
1191          */
1192         if ((diag != IXGBE_SUCCESS) && (diag != IXGBE_ERR_INVALID_MAC_ADDR)) {
1193                 PMD_INIT_LOG(ERR, "VF Initialization Failure: %d", diag);
1194                 return (diag);
1195         }
1196
1197         /* negotiate mailbox API version to use with the PF. */
1198         ixgbevf_negotiate_api(hw);
1199
1200         /* Get Rx/Tx queue count via mailbox, which is ready after reset_hw */
1201         ixgbevf_get_queues(hw, &tcs, &tc);
1202
1203         /* Allocate memory for storing MAC addresses */
1204         eth_dev->data->mac_addrs = rte_zmalloc("ixgbevf", ETHER_ADDR_LEN *
1205                         hw->mac.num_rar_entries, 0);
1206         if (eth_dev->data->mac_addrs == NULL) {
1207                 PMD_INIT_LOG(ERR,
1208                         "Failed to allocate %u bytes needed to store "
1209                         "MAC addresses",
1210                         ETHER_ADDR_LEN * hw->mac.num_rar_entries);
1211                 return -ENOMEM;
1212         }
1213
1214         /* Generate a random MAC address, if none was assigned by PF. */
1215         if (is_zero_ether_addr(perm_addr)) {
1216                 generate_random_mac_addr(perm_addr);
1217                 diag = ixgbe_set_rar_vf(hw, 1, perm_addr->addr_bytes, 0, 1);
1218                 if (diag) {
1219                         rte_free(eth_dev->data->mac_addrs);
1220                         eth_dev->data->mac_addrs = NULL;
1221                         return diag;
1222                 }
1223                 PMD_INIT_LOG(INFO, "\tVF MAC address not assigned by Host PF");
1224                 PMD_INIT_LOG(INFO, "\tAssign randomly generated MAC address "
1225                              "%02x:%02x:%02x:%02x:%02x:%02x",
1226                              perm_addr->addr_bytes[0],
1227                              perm_addr->addr_bytes[1],
1228                              perm_addr->addr_bytes[2],
1229                              perm_addr->addr_bytes[3],
1230                              perm_addr->addr_bytes[4],
1231                              perm_addr->addr_bytes[5]);
1232         }
1233
1234         /* Copy the permanent MAC address */
1235         ether_addr_copy(perm_addr, &eth_dev->data->mac_addrs[0]);
1236
1237         /* reset the hardware with the new settings */
1238         diag = hw->mac.ops.start_hw(hw);
1239         switch (diag) {
1240                 case  0:
1241                         break;
1242
1243                 default:
1244                         PMD_INIT_LOG(ERR, "VF Initialization Failure: %d", diag);
1245                         return (-EIO);
1246         }
1247
1248         PMD_INIT_LOG(DEBUG, "port %d vendorID=0x%x deviceID=0x%x mac.type=%s",
1249                      eth_dev->data->port_id, pci_dev->id.vendor_id,
1250                      pci_dev->id.device_id, "ixgbe_mac_82599_vf");
1251
1252         return 0;
1253 }
1254
1255 /* Virtual Function device uninit */
1256
1257 static int
1258 eth_ixgbevf_dev_uninit(struct rte_eth_dev *eth_dev)
1259 {
1260         struct ixgbe_hw *hw;
1261         unsigned i;
1262
1263         PMD_INIT_FUNC_TRACE();
1264
1265         if (rte_eal_process_type() != RTE_PROC_PRIMARY)
1266                 return -EPERM;
1267
1268         hw = IXGBE_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
1269
1270         if (hw->adapter_stopped == 0)
1271                 ixgbevf_dev_close(eth_dev);
1272
1273         eth_dev->dev_ops = NULL;
1274         eth_dev->rx_pkt_burst = NULL;
1275         eth_dev->tx_pkt_burst = NULL;
1276
1277         /* Disable the interrupts for VF */
1278         ixgbevf_intr_disable(hw);
1279
1280         for (i = 0; i < eth_dev->data->nb_rx_queues; i++) {
1281                 ixgbe_dev_rx_queue_release(eth_dev->data->rx_queues[i]);
1282                 eth_dev->data->rx_queues[i] = NULL;
1283         }
1284         eth_dev->data->nb_rx_queues = 0;
1285
1286         for (i = 0; i < eth_dev->data->nb_tx_queues; i++) {
1287                 ixgbe_dev_tx_queue_release(eth_dev->data->tx_queues[i]);
1288                 eth_dev->data->tx_queues[i] = NULL;
1289         }
1290         eth_dev->data->nb_tx_queues = 0;
1291
1292         rte_free(eth_dev->data->mac_addrs);
1293         eth_dev->data->mac_addrs = NULL;
1294
1295         return 0;
1296 }
1297
1298 static struct eth_driver rte_ixgbe_pmd = {
1299         .pci_drv = {
1300                 .name = "rte_ixgbe_pmd",
1301                 .id_table = pci_id_ixgbe_map,
1302                 .drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC |
1303                         RTE_PCI_DRV_DETACHABLE,
1304         },
1305         .eth_dev_init = eth_ixgbe_dev_init,
1306         .eth_dev_uninit = eth_ixgbe_dev_uninit,
1307         .dev_private_size = sizeof(struct ixgbe_adapter),
1308 };
1309
1310 /*
1311  * virtual function driver struct
1312  */
1313 static struct eth_driver rte_ixgbevf_pmd = {
1314         .pci_drv = {
1315                 .name = "rte_ixgbevf_pmd",
1316                 .id_table = pci_id_ixgbevf_map,
1317                 .drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_DETACHABLE,
1318         },
1319         .eth_dev_init = eth_ixgbevf_dev_init,
1320         .eth_dev_uninit = eth_ixgbevf_dev_uninit,
1321         .dev_private_size = sizeof(struct ixgbe_adapter),
1322 };
1323
1324 /*
1325  * Driver initialization routine.
1326  * Invoked once at EAL init time.
1327  * Register itself as the [Poll Mode] Driver of PCI IXGBE devices.
1328  */
1329 static int
1330 rte_ixgbe_pmd_init(const char *name __rte_unused, const char *params __rte_unused)
1331 {
1332         PMD_INIT_FUNC_TRACE();
1333
1334         rte_eth_driver_register(&rte_ixgbe_pmd);
1335         return 0;
1336 }
1337
1338 /*
1339  * VF Driver initialization routine.
1340  * Invoked one at EAL init time.
1341  * Register itself as the [Virtual Poll Mode] Driver of PCI niantic devices.
1342  */
1343 static int
1344 rte_ixgbevf_pmd_init(const char *name __rte_unused, const char *param __rte_unused)
1345 {
1346         PMD_INIT_FUNC_TRACE();
1347
1348         rte_eth_driver_register(&rte_ixgbevf_pmd);
1349         return (0);
1350 }
1351
1352 static int
1353 ixgbe_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
1354 {
1355         struct ixgbe_hw *hw =
1356                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1357         struct ixgbe_vfta * shadow_vfta =
1358                 IXGBE_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
1359         uint32_t vfta;
1360         uint32_t vid_idx;
1361         uint32_t vid_bit;
1362
1363         vid_idx = (uint32_t) ((vlan_id >> 5) & 0x7F);
1364         vid_bit = (uint32_t) (1 << (vlan_id & 0x1F));
1365         vfta = IXGBE_READ_REG(hw, IXGBE_VFTA(vid_idx));
1366         if (on)
1367                 vfta |= vid_bit;
1368         else
1369                 vfta &= ~vid_bit;
1370         IXGBE_WRITE_REG(hw, IXGBE_VFTA(vid_idx), vfta);
1371
1372         /* update local VFTA copy */
1373         shadow_vfta->vfta[vid_idx] = vfta;
1374
1375         return 0;
1376 }
1377
1378 static void
1379 ixgbe_vlan_strip_queue_set(struct rte_eth_dev *dev, uint16_t queue, int on)
1380 {
1381         if (on)
1382                 ixgbe_vlan_hw_strip_enable(dev, queue);
1383         else
1384                 ixgbe_vlan_hw_strip_disable(dev, queue);
1385 }
1386
1387 static void
1388 ixgbe_vlan_tpid_set(struct rte_eth_dev *dev, uint16_t tpid)
1389 {
1390         struct ixgbe_hw *hw =
1391                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1392
1393         /* Only the high 16-bits is valid */
1394         IXGBE_WRITE_REG(hw, IXGBE_EXVET, tpid << 16);
1395 }
1396
1397 void
1398 ixgbe_vlan_hw_filter_disable(struct rte_eth_dev *dev)
1399 {
1400         struct ixgbe_hw *hw =
1401                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1402         uint32_t vlnctrl;
1403
1404         PMD_INIT_FUNC_TRACE();
1405
1406         /* Filter Table Disable */
1407         vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
1408         vlnctrl &= ~IXGBE_VLNCTRL_VFE;
1409
1410         IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
1411 }
1412
1413 void
1414 ixgbe_vlan_hw_filter_enable(struct rte_eth_dev *dev)
1415 {
1416         struct ixgbe_hw *hw =
1417                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1418         struct ixgbe_vfta * shadow_vfta =
1419                 IXGBE_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
1420         uint32_t vlnctrl;
1421         uint16_t i;
1422
1423         PMD_INIT_FUNC_TRACE();
1424
1425         /* Filter Table Enable */
1426         vlnctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
1427         vlnctrl &= ~IXGBE_VLNCTRL_CFIEN;
1428         vlnctrl |= IXGBE_VLNCTRL_VFE;
1429
1430         IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlnctrl);
1431
1432         /* write whatever is in local vfta copy */
1433         for (i = 0; i < IXGBE_VFTA_SIZE; i++)
1434                 IXGBE_WRITE_REG(hw, IXGBE_VFTA(i), shadow_vfta->vfta[i]);
1435 }
1436
1437 static void
1438 ixgbe_vlan_hw_strip_bitmap_set(struct rte_eth_dev *dev, uint16_t queue, bool on)
1439 {
1440         struct ixgbe_hwstrip *hwstrip =
1441                 IXGBE_DEV_PRIVATE_TO_HWSTRIP_BITMAP(dev->data->dev_private);
1442
1443         if(queue >= IXGBE_MAX_RX_QUEUE_NUM)
1444                 return;
1445
1446         if (on)
1447                 IXGBE_SET_HWSTRIP(hwstrip, queue);
1448         else
1449                 IXGBE_CLEAR_HWSTRIP(hwstrip, queue);
1450 }
1451
1452 static void
1453 ixgbe_vlan_hw_strip_disable(struct rte_eth_dev *dev, uint16_t queue)
1454 {
1455         struct ixgbe_hw *hw =
1456                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1457         uint32_t ctrl;
1458
1459         PMD_INIT_FUNC_TRACE();
1460
1461         if (hw->mac.type == ixgbe_mac_82598EB) {
1462                 /* No queue level support */
1463                 PMD_INIT_LOG(NOTICE, "82598EB not support queue level hw strip");
1464                 return;
1465         }
1466         else {
1467                 /* Other 10G NIC, the VLAN strip can be setup per queue in RXDCTL */
1468                 ctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(queue));
1469                 ctrl &= ~IXGBE_RXDCTL_VME;
1470                 IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(queue), ctrl);
1471         }
1472         /* record those setting for HW strip per queue */
1473         ixgbe_vlan_hw_strip_bitmap_set(dev, queue, 0);
1474 }
1475
1476 static void
1477 ixgbe_vlan_hw_strip_enable(struct rte_eth_dev *dev, uint16_t queue)
1478 {
1479         struct ixgbe_hw *hw =
1480                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1481         uint32_t ctrl;
1482
1483         PMD_INIT_FUNC_TRACE();
1484
1485         if (hw->mac.type == ixgbe_mac_82598EB) {
1486                 /* No queue level supported */
1487                 PMD_INIT_LOG(NOTICE, "82598EB not support queue level hw strip");
1488                 return;
1489         }
1490         else {
1491                 /* Other 10G NIC, the VLAN strip can be setup per queue in RXDCTL */
1492                 ctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(queue));
1493                 ctrl |= IXGBE_RXDCTL_VME;
1494                 IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(queue), ctrl);
1495         }
1496         /* record those setting for HW strip per queue */
1497         ixgbe_vlan_hw_strip_bitmap_set(dev, queue, 1);
1498 }
1499
1500 void
1501 ixgbe_vlan_hw_strip_disable_all(struct rte_eth_dev *dev)
1502 {
1503         struct ixgbe_hw *hw =
1504                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1505         uint32_t ctrl;
1506         uint16_t i;
1507
1508         PMD_INIT_FUNC_TRACE();
1509
1510         if (hw->mac.type == ixgbe_mac_82598EB) {
1511                 ctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
1512                 ctrl &= ~IXGBE_VLNCTRL_VME;
1513                 IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, ctrl);
1514         }
1515         else {
1516                 /* Other 10G NIC, the VLAN strip can be setup per queue in RXDCTL */
1517                 for (i = 0; i < dev->data->nb_rx_queues; i++) {
1518                         ctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(i));
1519                         ctrl &= ~IXGBE_RXDCTL_VME;
1520                         IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(i), ctrl);
1521
1522                         /* record those setting for HW strip per queue */
1523                         ixgbe_vlan_hw_strip_bitmap_set(dev, i, 0);
1524                 }
1525         }
1526 }
1527
1528 void
1529 ixgbe_vlan_hw_strip_enable_all(struct rte_eth_dev *dev)
1530 {
1531         struct ixgbe_hw *hw =
1532                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1533         uint32_t ctrl;
1534         uint16_t i;
1535
1536         PMD_INIT_FUNC_TRACE();
1537
1538         if (hw->mac.type == ixgbe_mac_82598EB) {
1539                 ctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
1540                 ctrl |= IXGBE_VLNCTRL_VME;
1541                 IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, ctrl);
1542         }
1543         else {
1544                 /* Other 10G NIC, the VLAN strip can be setup per queue in RXDCTL */
1545                 for (i = 0; i < dev->data->nb_rx_queues; i++) {
1546                         ctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(i));
1547                         ctrl |= IXGBE_RXDCTL_VME;
1548                         IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(i), ctrl);
1549
1550                         /* record those setting for HW strip per queue */
1551                         ixgbe_vlan_hw_strip_bitmap_set(dev, i, 1);
1552                 }
1553         }
1554 }
1555
1556 static void
1557 ixgbe_vlan_hw_extend_disable(struct rte_eth_dev *dev)
1558 {
1559         struct ixgbe_hw *hw =
1560                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1561         uint32_t ctrl;
1562
1563         PMD_INIT_FUNC_TRACE();
1564
1565         /* DMATXCTRL: Geric Double VLAN Disable */
1566         ctrl = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
1567         ctrl &= ~IXGBE_DMATXCTL_GDV;
1568         IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL, ctrl);
1569
1570         /* CTRL_EXT: Global Double VLAN Disable */
1571         ctrl = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
1572         ctrl &= ~IXGBE_EXTENDED_VLAN;
1573         IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl);
1574
1575 }
1576
1577 static void
1578 ixgbe_vlan_hw_extend_enable(struct rte_eth_dev *dev)
1579 {
1580         struct ixgbe_hw *hw =
1581                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1582         uint32_t ctrl;
1583
1584         PMD_INIT_FUNC_TRACE();
1585
1586         /* DMATXCTRL: Geric Double VLAN Enable */
1587         ctrl  = IXGBE_READ_REG(hw, IXGBE_DMATXCTL);
1588         ctrl |= IXGBE_DMATXCTL_GDV;
1589         IXGBE_WRITE_REG(hw, IXGBE_DMATXCTL, ctrl);
1590
1591         /* CTRL_EXT: Global Double VLAN Enable */
1592         ctrl  = IXGBE_READ_REG(hw, IXGBE_CTRL_EXT);
1593         ctrl |= IXGBE_EXTENDED_VLAN;
1594         IXGBE_WRITE_REG(hw, IXGBE_CTRL_EXT, ctrl);
1595
1596         /*
1597          * VET EXT field in the EXVET register = 0x8100 by default
1598          * So no need to change. Same to VT field of DMATXCTL register
1599          */
1600 }
1601
1602 static void
1603 ixgbe_vlan_offload_set(struct rte_eth_dev *dev, int mask)
1604 {
1605         if(mask & ETH_VLAN_STRIP_MASK){
1606                 if (dev->data->dev_conf.rxmode.hw_vlan_strip)
1607                         ixgbe_vlan_hw_strip_enable_all(dev);
1608                 else
1609                         ixgbe_vlan_hw_strip_disable_all(dev);
1610         }
1611
1612         if(mask & ETH_VLAN_FILTER_MASK){
1613                 if (dev->data->dev_conf.rxmode.hw_vlan_filter)
1614                         ixgbe_vlan_hw_filter_enable(dev);
1615                 else
1616                         ixgbe_vlan_hw_filter_disable(dev);
1617         }
1618
1619         if(mask & ETH_VLAN_EXTEND_MASK){
1620                 if (dev->data->dev_conf.rxmode.hw_vlan_extend)
1621                         ixgbe_vlan_hw_extend_enable(dev);
1622                 else
1623                         ixgbe_vlan_hw_extend_disable(dev);
1624         }
1625 }
1626
1627 static void
1628 ixgbe_vmdq_vlan_hw_filter_enable(struct rte_eth_dev *dev)
1629 {
1630         struct ixgbe_hw *hw =
1631                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1632         /* VLNCTRL: enable vlan filtering and allow all vlan tags through */
1633         uint32_t vlanctrl = IXGBE_READ_REG(hw, IXGBE_VLNCTRL);
1634         vlanctrl |= IXGBE_VLNCTRL_VFE ; /* enable vlan filters */
1635         IXGBE_WRITE_REG(hw, IXGBE_VLNCTRL, vlanctrl);
1636 }
1637
1638 static int
1639 ixgbe_dev_configure(struct rte_eth_dev *dev)
1640 {
1641         struct ixgbe_interrupt *intr =
1642                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
1643         struct ixgbe_adapter *adapter =
1644                 (struct ixgbe_adapter *)dev->data->dev_private;
1645
1646         PMD_INIT_FUNC_TRACE();
1647
1648         /* set flag to update link status after init */
1649         intr->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
1650
1651         /*
1652          * Initialize to TRUE. If any of Rx queues doesn't meet the bulk
1653          * allocation or vector Rx preconditions we will reset it.
1654          */
1655         adapter->rx_bulk_alloc_allowed = true;
1656         adapter->rx_vec_allowed = true;
1657
1658         return 0;
1659 }
1660
1661 /*
1662  * Configure device link speed and setup link.
1663  * It returns 0 on success.
1664  */
1665 static int
1666 ixgbe_dev_start(struct rte_eth_dev *dev)
1667 {
1668         struct ixgbe_hw *hw =
1669                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1670         struct ixgbe_vf_info *vfinfo =
1671                 *IXGBE_DEV_PRIVATE_TO_P_VFDATA(dev->data->dev_private);
1672         struct rte_intr_handle *intr_handle = &dev->pci_dev->intr_handle;
1673         uint32_t intr_vector = 0;
1674         int err, link_up = 0, negotiate = 0;
1675         uint32_t speed = 0;
1676         int mask = 0;
1677         int status;
1678         uint16_t vf, idx;
1679
1680         PMD_INIT_FUNC_TRACE();
1681
1682         /* IXGBE devices don't support half duplex */
1683         if ((dev->data->dev_conf.link_duplex != ETH_LINK_AUTONEG_DUPLEX) &&
1684                         (dev->data->dev_conf.link_duplex != ETH_LINK_FULL_DUPLEX)) {
1685                 PMD_INIT_LOG(ERR, "Invalid link_duplex (%hu) for port %hhu",
1686                              dev->data->dev_conf.link_duplex,
1687                              dev->data->port_id);
1688                 return -EINVAL;
1689         }
1690
1691         /* stop adapter */
1692         hw->adapter_stopped = 0;
1693         ixgbe_stop_adapter(hw);
1694
1695         /* reinitialize adapter
1696          * this calls reset and start */
1697         status = ixgbe_pf_reset_hw(hw);
1698         if (status != 0)
1699                 return -1;
1700         hw->mac.ops.start_hw(hw);
1701         hw->mac.get_link_status = true;
1702
1703         /* configure PF module if SRIOV enabled */
1704         ixgbe_pf_host_configure(dev);
1705
1706         /* check and configure queue intr-vector mapping */
1707         if (dev->data->dev_conf.intr_conf.rxq != 0)
1708                 intr_vector = dev->data->nb_rx_queues;
1709
1710         if (rte_intr_efd_enable(intr_handle, intr_vector))
1711                 return -1;
1712
1713         if (rte_intr_dp_is_en(intr_handle) && !intr_handle->intr_vec) {
1714                 intr_handle->intr_vec =
1715                         rte_zmalloc("intr_vec",
1716                                     dev->data->nb_rx_queues * sizeof(int),
1717                                     0);
1718                 if (intr_handle->intr_vec == NULL) {
1719                         PMD_INIT_LOG(ERR, "Failed to allocate %d rx_queues"
1720                                      " intr_vec\n", dev->data->nb_rx_queues);
1721                         return -ENOMEM;
1722                 }
1723         }
1724
1725         /* confiugre msix for sleep until rx interrupt */
1726         ixgbe_configure_msix(dev);
1727
1728         /* initialize transmission unit */
1729         ixgbe_dev_tx_init(dev);
1730
1731         /* This can fail when allocating mbufs for descriptor rings */
1732         err = ixgbe_dev_rx_init(dev);
1733         if (err) {
1734                 PMD_INIT_LOG(ERR, "Unable to initialize RX hardware");
1735                 goto error;
1736         }
1737
1738         err = ixgbe_dev_rxtx_start(dev);
1739         if (err < 0) {
1740                 PMD_INIT_LOG(ERR, "Unable to start rxtx queues");
1741                 goto error;
1742         }
1743
1744         /* Skip link setup if loopback mode is enabled for 82599. */
1745         if (hw->mac.type == ixgbe_mac_82599EB &&
1746                         dev->data->dev_conf.lpbk_mode == IXGBE_LPBK_82599_TX_RX)
1747                 goto skip_link_setup;
1748
1749         if (ixgbe_is_sfp(hw) && hw->phy.multispeed_fiber) {
1750                 err = hw->mac.ops.setup_sfp(hw);
1751                 if (err)
1752                         goto error;
1753         }
1754
1755         if (hw->mac.ops.get_media_type(hw) == ixgbe_media_type_copper) {
1756                 /* Turn on the copper */
1757                 ixgbe_set_phy_power(hw, true);
1758         } else {
1759                 /* Turn on the laser */
1760                 ixgbe_enable_tx_laser(hw);
1761         }
1762
1763         err = ixgbe_check_link(hw, &speed, &link_up, 0);
1764         if (err)
1765                 goto error;
1766         dev->data->dev_link.link_status = link_up;
1767
1768         err = ixgbe_get_link_capabilities(hw, &speed, &negotiate);
1769         if (err)
1770                 goto error;
1771
1772         switch(dev->data->dev_conf.link_speed) {
1773         case ETH_LINK_SPEED_AUTONEG:
1774                 speed = (hw->mac.type != ixgbe_mac_82598EB) ?
1775                                 IXGBE_LINK_SPEED_82599_AUTONEG :
1776                                 IXGBE_LINK_SPEED_82598_AUTONEG;
1777                 break;
1778         case ETH_LINK_SPEED_100:
1779                 /*
1780                  * Invalid for 82598 but error will be detected by
1781                  * ixgbe_setup_link()
1782                  */
1783                 speed = IXGBE_LINK_SPEED_100_FULL;
1784                 break;
1785         case ETH_LINK_SPEED_1000:
1786                 speed = IXGBE_LINK_SPEED_1GB_FULL;
1787                 break;
1788         case ETH_LINK_SPEED_10000:
1789                 speed = IXGBE_LINK_SPEED_10GB_FULL;
1790                 break;
1791         default:
1792                 PMD_INIT_LOG(ERR, "Invalid link_speed (%hu) for port %hhu",
1793                              dev->data->dev_conf.link_speed,
1794                              dev->data->port_id);
1795                 goto error;
1796         }
1797
1798         err = ixgbe_setup_link(hw, speed, link_up);
1799         if (err)
1800                 goto error;
1801
1802 skip_link_setup:
1803
1804         /* check if lsc interrupt is enabled */
1805         if (dev->data->dev_conf.intr_conf.lsc != 0) {
1806                 if (rte_intr_allow_others(intr_handle)) {
1807                         rte_intr_callback_register(intr_handle,
1808                                                    ixgbe_dev_interrupt_handler,
1809                                                    (void *)dev);
1810                         ixgbe_dev_lsc_interrupt_setup(dev);
1811                 } else
1812                         PMD_INIT_LOG(INFO, "lsc won't enable because of"
1813                                      " no intr multiplex\n");
1814         }
1815
1816         /* check if rxq interrupt is enabled */
1817         if (dev->data->dev_conf.intr_conf.rxq != 0)
1818                 ixgbe_dev_rxq_interrupt_setup(dev);
1819
1820         /* enable uio/vfio intr/eventfd mapping */
1821         rte_intr_enable(intr_handle);
1822
1823         /* resume enabled intr since hw reset */
1824         ixgbe_enable_intr(dev);
1825
1826         mask = ETH_VLAN_STRIP_MASK | ETH_VLAN_FILTER_MASK | \
1827                 ETH_VLAN_EXTEND_MASK;
1828         ixgbe_vlan_offload_set(dev, mask);
1829
1830         if (dev->data->dev_conf.rxmode.mq_mode == ETH_MQ_RX_VMDQ_ONLY) {
1831                 /* Enable vlan filtering for VMDq */
1832                 ixgbe_vmdq_vlan_hw_filter_enable(dev);
1833         }
1834
1835         /* Configure DCB hw */
1836         ixgbe_configure_dcb(dev);
1837
1838         if (dev->data->dev_conf.fdir_conf.mode != RTE_FDIR_MODE_NONE) {
1839                 err = ixgbe_fdir_configure(dev);
1840                 if (err)
1841                         goto error;
1842         }
1843
1844         /* Restore vf rate limit */
1845         if (vfinfo != NULL) {
1846                 for (vf = 0; vf < dev->pci_dev->max_vfs; vf++)
1847                         for (idx = 0; idx < IXGBE_MAX_QUEUE_NUM_PER_VF; idx++)
1848                                 if (vfinfo[vf].tx_rate[idx] != 0)
1849                                         ixgbe_set_vf_rate_limit(dev, vf,
1850                                                 vfinfo[vf].tx_rate[idx],
1851                                                 1 << idx);
1852         }
1853
1854         ixgbe_restore_statistics_mapping(dev);
1855
1856         return (0);
1857
1858 error:
1859         PMD_INIT_LOG(ERR, "failure in ixgbe_dev_start(): %d", err);
1860         ixgbe_dev_clear_queues(dev);
1861         return -EIO;
1862 }
1863
1864 /*
1865  * Stop device: disable rx and tx functions to allow for reconfiguring.
1866  */
1867 static void
1868 ixgbe_dev_stop(struct rte_eth_dev *dev)
1869 {
1870         struct rte_eth_link link;
1871         struct ixgbe_hw *hw =
1872                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1873         struct ixgbe_vf_info *vfinfo =
1874                 *IXGBE_DEV_PRIVATE_TO_P_VFDATA(dev->data->dev_private);
1875         struct ixgbe_filter_info *filter_info =
1876                 IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
1877         struct ixgbe_5tuple_filter *p_5tuple, *p_5tuple_next;
1878         struct rte_intr_handle *intr_handle = &dev->pci_dev->intr_handle;
1879         int vf;
1880
1881         PMD_INIT_FUNC_TRACE();
1882
1883         /* disable interrupts */
1884         ixgbe_disable_intr(hw);
1885
1886         /* disable intr eventfd mapping */
1887         rte_intr_disable(intr_handle);
1888
1889         /* reset the NIC */
1890         ixgbe_pf_reset_hw(hw);
1891         hw->adapter_stopped = 0;
1892
1893         /* stop adapter */
1894         ixgbe_stop_adapter(hw);
1895
1896         for (vf = 0; vfinfo != NULL &&
1897                      vf < dev->pci_dev->max_vfs; vf++)
1898                 vfinfo[vf].clear_to_send = false;
1899
1900         if (hw->mac.ops.get_media_type(hw) == ixgbe_media_type_copper) {
1901                 /* Turn off the copper */
1902                 ixgbe_set_phy_power(hw, false);
1903         } else {
1904                 /* Turn off the laser */
1905                 ixgbe_disable_tx_laser(hw);
1906         }
1907
1908         ixgbe_dev_clear_queues(dev);
1909
1910         /* Clear stored conf */
1911         dev->data->scattered_rx = 0;
1912         dev->data->lro = 0;
1913
1914         /* Clear recorded link status */
1915         memset(&link, 0, sizeof(link));
1916         rte_ixgbe_dev_atomic_write_link_status(dev, &link);
1917
1918         /* Remove all ntuple filters of the device */
1919         for (p_5tuple = TAILQ_FIRST(&filter_info->fivetuple_list);
1920              p_5tuple != NULL; p_5tuple = p_5tuple_next) {
1921                 p_5tuple_next = TAILQ_NEXT(p_5tuple, entries);
1922                 TAILQ_REMOVE(&filter_info->fivetuple_list,
1923                              p_5tuple, entries);
1924                 rte_free(p_5tuple);
1925         }
1926         memset(filter_info->fivetuple_mask, 0,
1927                 sizeof(uint32_t) * IXGBE_5TUPLE_ARRAY_SIZE);
1928
1929         /* Clean datapath event and queue/vec mapping */
1930         rte_intr_efd_disable(intr_handle);
1931         if (intr_handle->intr_vec != NULL) {
1932                 rte_free(intr_handle->intr_vec);
1933                 intr_handle->intr_vec = NULL;
1934         }
1935 }
1936
1937 /*
1938  * Set device link up: enable tx.
1939  */
1940 static int
1941 ixgbe_dev_set_link_up(struct rte_eth_dev *dev)
1942 {
1943         struct ixgbe_hw *hw =
1944                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1945         if (hw->mac.type == ixgbe_mac_82599EB) {
1946 #ifdef RTE_NIC_BYPASS
1947                 if (hw->device_id == IXGBE_DEV_ID_82599_BYPASS) {
1948                         /* Not suported in bypass mode */
1949                         PMD_INIT_LOG(ERR, "Set link up is not supported "
1950                                      "by device id 0x%x", hw->device_id);
1951                         return -ENOTSUP;
1952                 }
1953 #endif
1954         }
1955
1956         if (hw->mac.ops.get_media_type(hw) == ixgbe_media_type_copper) {
1957                 /* Turn on the copper */
1958                 ixgbe_set_phy_power(hw, true);
1959         } else {
1960                 /* Turn on the laser */
1961                 ixgbe_enable_tx_laser(hw);
1962         }
1963
1964         return 0;
1965 }
1966
1967 /*
1968  * Set device link down: disable tx.
1969  */
1970 static int
1971 ixgbe_dev_set_link_down(struct rte_eth_dev *dev)
1972 {
1973         struct ixgbe_hw *hw =
1974                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1975         if (hw->mac.type == ixgbe_mac_82599EB) {
1976 #ifdef RTE_NIC_BYPASS
1977                 if (hw->device_id == IXGBE_DEV_ID_82599_BYPASS) {
1978                         /* Not suported in bypass mode */
1979                         PMD_INIT_LOG(ERR, "Set link down is not supported "
1980                                      "by device id 0x%x", hw->device_id);
1981                         return -ENOTSUP;
1982                 }
1983 #endif
1984         }
1985
1986         if (hw->mac.ops.get_media_type(hw) == ixgbe_media_type_copper) {
1987                 /* Turn off the copper */
1988                 ixgbe_set_phy_power(hw, false);
1989         } else {
1990                 /* Turn off the laser */
1991                 ixgbe_disable_tx_laser(hw);
1992         }
1993
1994         return 0;
1995 }
1996
1997 /*
1998  * Reest and stop device.
1999  */
2000 static void
2001 ixgbe_dev_close(struct rte_eth_dev *dev)
2002 {
2003         struct ixgbe_hw *hw =
2004                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2005
2006         PMD_INIT_FUNC_TRACE();
2007
2008         ixgbe_pf_reset_hw(hw);
2009
2010         ixgbe_dev_stop(dev);
2011         hw->adapter_stopped = 1;
2012
2013         ixgbe_dev_free_queues(dev);
2014
2015         ixgbe_disable_pcie_master(hw);
2016
2017         /* reprogram the RAR[0] in case user changed it. */
2018         ixgbe_set_rar(hw, 0, hw->mac.addr, 0, IXGBE_RAH_AV);
2019 }
2020
2021 static void
2022 ixgbe_read_stats_registers(struct ixgbe_hw *hw, struct ixgbe_hw_stats
2023                                                    *hw_stats, uint64_t *total_missed_rx,
2024                                                    uint64_t *total_qbrc, uint64_t *total_qprc,
2025                                                    uint64_t *total_qprdc)
2026 {
2027         uint32_t bprc, lxon, lxoff, total;
2028         unsigned i;
2029
2030         hw_stats->crcerrs += IXGBE_READ_REG(hw, IXGBE_CRCERRS);
2031         hw_stats->illerrc += IXGBE_READ_REG(hw, IXGBE_ILLERRC);
2032         hw_stats->errbc += IXGBE_READ_REG(hw, IXGBE_ERRBC);
2033         hw_stats->mspdc += IXGBE_READ_REG(hw, IXGBE_MSPDC);
2034
2035         for (i = 0; i < 8; i++) {
2036                 uint32_t mp;
2037                 mp = IXGBE_READ_REG(hw, IXGBE_MPC(i));
2038                 /* global total per queue */
2039                 hw_stats->mpc[i] += mp;
2040                 /* Running comprehensive total for stats display */
2041                 *total_missed_rx += hw_stats->mpc[i];
2042                 if (hw->mac.type == ixgbe_mac_82598EB)
2043                         hw_stats->rnbc[i] +=
2044                             IXGBE_READ_REG(hw, IXGBE_RNBC(i));
2045                 hw_stats->pxontxc[i] +=
2046                     IXGBE_READ_REG(hw, IXGBE_PXONTXC(i));
2047                 hw_stats->pxonrxc[i] +=
2048                     IXGBE_READ_REG(hw, IXGBE_PXONRXC(i));
2049                 hw_stats->pxofftxc[i] +=
2050                     IXGBE_READ_REG(hw, IXGBE_PXOFFTXC(i));
2051                 hw_stats->pxoffrxc[i] +=
2052                     IXGBE_READ_REG(hw, IXGBE_PXOFFRXC(i));
2053                 hw_stats->pxon2offc[i] +=
2054                     IXGBE_READ_REG(hw, IXGBE_PXON2OFFCNT(i));
2055         }
2056         for (i = 0; i < IXGBE_QUEUE_STAT_COUNTERS; i++) {
2057                 hw_stats->qprc[i] += IXGBE_READ_REG(hw, IXGBE_QPRC(i));
2058                 hw_stats->qptc[i] += IXGBE_READ_REG(hw, IXGBE_QPTC(i));
2059                 hw_stats->qbrc[i] += IXGBE_READ_REG(hw, IXGBE_QBRC_L(i));
2060                 hw_stats->qbrc[i] +=
2061                     ((uint64_t)IXGBE_READ_REG(hw, IXGBE_QBRC_H(i)) << 32);
2062                 hw_stats->qbtc[i] += IXGBE_READ_REG(hw, IXGBE_QBTC_L(i));
2063                 hw_stats->qbtc[i] +=
2064                     ((uint64_t)IXGBE_READ_REG(hw, IXGBE_QBTC_H(i)) << 32);
2065                 *total_qprdc += hw_stats->qprdc[i] +=
2066                                 IXGBE_READ_REG(hw, IXGBE_QPRDC(i));
2067
2068                 *total_qprc += hw_stats->qprc[i];
2069                 *total_qbrc += hw_stats->qbrc[i];
2070         }
2071         hw_stats->mlfc += IXGBE_READ_REG(hw, IXGBE_MLFC);
2072         hw_stats->mrfc += IXGBE_READ_REG(hw, IXGBE_MRFC);
2073         hw_stats->rlec += IXGBE_READ_REG(hw, IXGBE_RLEC);
2074
2075         /* Note that gprc counts missed packets */
2076         hw_stats->gprc += IXGBE_READ_REG(hw, IXGBE_GPRC);
2077
2078         if (hw->mac.type != ixgbe_mac_82598EB) {
2079                 hw_stats->gorc += IXGBE_READ_REG(hw, IXGBE_GORCL);
2080                 hw_stats->gorc += ((u64)IXGBE_READ_REG(hw, IXGBE_GORCH) << 32);
2081                 hw_stats->gotc += IXGBE_READ_REG(hw, IXGBE_GOTCL);
2082                 hw_stats->gotc += ((u64)IXGBE_READ_REG(hw, IXGBE_GOTCH) << 32);
2083                 hw_stats->tor += IXGBE_READ_REG(hw, IXGBE_TORL);
2084                 hw_stats->tor += ((u64)IXGBE_READ_REG(hw, IXGBE_TORH) << 32);
2085                 hw_stats->lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXCNT);
2086                 hw_stats->lxoffrxc += IXGBE_READ_REG(hw, IXGBE_LXOFFRXCNT);
2087         } else {
2088                 hw_stats->lxonrxc += IXGBE_READ_REG(hw, IXGBE_LXONRXC);
2089                 hw_stats->lxoffrxc += IXGBE_READ_REG(hw, IXGBE_LXOFFRXC);
2090                 /* 82598 only has a counter in the high register */
2091                 hw_stats->gorc += IXGBE_READ_REG(hw, IXGBE_GORCH);
2092                 hw_stats->gotc += IXGBE_READ_REG(hw, IXGBE_GOTCH);
2093                 hw_stats->tor += IXGBE_READ_REG(hw, IXGBE_TORH);
2094         }
2095
2096         /*
2097          * Workaround: mprc hardware is incorrectly counting
2098          * broadcasts, so for now we subtract those.
2099          */
2100         bprc = IXGBE_READ_REG(hw, IXGBE_BPRC);
2101         hw_stats->bprc += bprc;
2102         hw_stats->mprc += IXGBE_READ_REG(hw, IXGBE_MPRC);
2103         if (hw->mac.type == ixgbe_mac_82598EB)
2104                 hw_stats->mprc -= bprc;
2105
2106         hw_stats->prc64 += IXGBE_READ_REG(hw, IXGBE_PRC64);
2107         hw_stats->prc127 += IXGBE_READ_REG(hw, IXGBE_PRC127);
2108         hw_stats->prc255 += IXGBE_READ_REG(hw, IXGBE_PRC255);
2109         hw_stats->prc511 += IXGBE_READ_REG(hw, IXGBE_PRC511);
2110         hw_stats->prc1023 += IXGBE_READ_REG(hw, IXGBE_PRC1023);
2111         hw_stats->prc1522 += IXGBE_READ_REG(hw, IXGBE_PRC1522);
2112
2113         lxon = IXGBE_READ_REG(hw, IXGBE_LXONTXC);
2114         hw_stats->lxontxc += lxon;
2115         lxoff = IXGBE_READ_REG(hw, IXGBE_LXOFFTXC);
2116         hw_stats->lxofftxc += lxoff;
2117         total = lxon + lxoff;
2118
2119         hw_stats->gptc += IXGBE_READ_REG(hw, IXGBE_GPTC);
2120         hw_stats->mptc += IXGBE_READ_REG(hw, IXGBE_MPTC);
2121         hw_stats->ptc64 += IXGBE_READ_REG(hw, IXGBE_PTC64);
2122         hw_stats->gptc -= total;
2123         hw_stats->mptc -= total;
2124         hw_stats->ptc64 -= total;
2125         hw_stats->gotc -= total * ETHER_MIN_LEN;
2126
2127         hw_stats->ruc += IXGBE_READ_REG(hw, IXGBE_RUC);
2128         hw_stats->rfc += IXGBE_READ_REG(hw, IXGBE_RFC);
2129         hw_stats->roc += IXGBE_READ_REG(hw, IXGBE_ROC);
2130         hw_stats->rjc += IXGBE_READ_REG(hw, IXGBE_RJC);
2131         hw_stats->mngprc += IXGBE_READ_REG(hw, IXGBE_MNGPRC);
2132         hw_stats->mngpdc += IXGBE_READ_REG(hw, IXGBE_MNGPDC);
2133         hw_stats->mngptc += IXGBE_READ_REG(hw, IXGBE_MNGPTC);
2134         hw_stats->tpr += IXGBE_READ_REG(hw, IXGBE_TPR);
2135         hw_stats->tpt += IXGBE_READ_REG(hw, IXGBE_TPT);
2136         hw_stats->ptc127 += IXGBE_READ_REG(hw, IXGBE_PTC127);
2137         hw_stats->ptc255 += IXGBE_READ_REG(hw, IXGBE_PTC255);
2138         hw_stats->ptc511 += IXGBE_READ_REG(hw, IXGBE_PTC511);
2139         hw_stats->ptc1023 += IXGBE_READ_REG(hw, IXGBE_PTC1023);
2140         hw_stats->ptc1522 += IXGBE_READ_REG(hw, IXGBE_PTC1522);
2141         hw_stats->bptc += IXGBE_READ_REG(hw, IXGBE_BPTC);
2142         hw_stats->xec += IXGBE_READ_REG(hw, IXGBE_XEC);
2143         hw_stats->fccrc += IXGBE_READ_REG(hw, IXGBE_FCCRC);
2144         hw_stats->fclast += IXGBE_READ_REG(hw, IXGBE_FCLAST);
2145         /* Only read FCOE on 82599 */
2146         if (hw->mac.type != ixgbe_mac_82598EB) {
2147                 hw_stats->fcoerpdc += IXGBE_READ_REG(hw, IXGBE_FCOERPDC);
2148                 hw_stats->fcoeprc += IXGBE_READ_REG(hw, IXGBE_FCOEPRC);
2149                 hw_stats->fcoeptc += IXGBE_READ_REG(hw, IXGBE_FCOEPTC);
2150                 hw_stats->fcoedwrc += IXGBE_READ_REG(hw, IXGBE_FCOEDWRC);
2151                 hw_stats->fcoedwtc += IXGBE_READ_REG(hw, IXGBE_FCOEDWTC);
2152         }
2153
2154         /* Flow Director Stats registers */
2155         hw_stats->fdirmatch += IXGBE_READ_REG(hw, IXGBE_FDIRMATCH);
2156         hw_stats->fdirmiss += IXGBE_READ_REG(hw, IXGBE_FDIRMISS);
2157 }
2158
2159 /*
2160  * This function is based on ixgbe_update_stats_counters() in ixgbe/ixgbe.c
2161  */
2162 static void
2163 ixgbe_dev_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
2164 {
2165         struct ixgbe_hw *hw =
2166                         IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2167         struct ixgbe_hw_stats *hw_stats =
2168                         IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
2169         uint64_t total_missed_rx, total_qbrc, total_qprc, total_qprdc;
2170         unsigned i;
2171
2172         total_missed_rx = 0;
2173         total_qbrc = 0;
2174         total_qprc = 0;
2175         total_qprdc = 0;
2176
2177         ixgbe_read_stats_registers(hw, hw_stats, &total_missed_rx, &total_qbrc,
2178                         &total_qprc, &total_qprdc);
2179
2180         if (stats == NULL)
2181                 return;
2182
2183         /* Fill out the rte_eth_stats statistics structure */
2184         stats->ipackets = total_qprc;
2185         stats->ibytes = total_qbrc;
2186         stats->opackets = hw_stats->gptc;
2187         stats->obytes = hw_stats->gotc;
2188
2189         for (i = 0; i < IXGBE_QUEUE_STAT_COUNTERS; i++) {
2190                 stats->q_ipackets[i] = hw_stats->qprc[i];
2191                 stats->q_opackets[i] = hw_stats->qptc[i];
2192                 stats->q_ibytes[i] = hw_stats->qbrc[i];
2193                 stats->q_obytes[i] = hw_stats->qbtc[i];
2194                 stats->q_errors[i] = hw_stats->qprdc[i];
2195         }
2196
2197         /* Rx Errors */
2198         stats->ierrors  = hw_stats->crcerrs +
2199                           hw_stats->rlec +
2200                           hw_stats->ruc +
2201                           hw_stats->roc +
2202                           total_missed_rx +
2203                           hw_stats->illerrc +
2204                           hw_stats->errbc +
2205                           hw_stats->xec +
2206                           hw_stats->mlfc +
2207                           hw_stats->mrfc +
2208                           hw_stats->rfc +
2209                           hw_stats->rjc +
2210                           hw_stats->fccrc +
2211                           hw_stats->fclast;
2212
2213         /* Tx Errors */
2214         stats->oerrors  = 0;
2215 }
2216
2217 static void
2218 ixgbe_dev_stats_reset(struct rte_eth_dev *dev)
2219 {
2220         struct ixgbe_hw_stats *stats =
2221                         IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
2222
2223         /* HW registers are cleared on read */
2224         ixgbe_dev_stats_get(dev, NULL);
2225
2226         /* Reset software totals */
2227         memset(stats, 0, sizeof(*stats));
2228 }
2229
2230 static int
2231 ixgbe_dev_xstats_get(struct rte_eth_dev *dev, struct rte_eth_xstats *xstats,
2232                                          unsigned n)
2233 {
2234         struct ixgbe_hw *hw =
2235                         IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2236         struct ixgbe_hw_stats *hw_stats =
2237                         IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
2238         uint64_t total_missed_rx, total_qbrc, total_qprc, total_qprdc;
2239         unsigned i, count = IXGBE_NB_XSTATS;
2240
2241         if (n < count)
2242                 return count;
2243
2244         total_missed_rx = 0;
2245         total_qbrc = 0;
2246         total_qprc = 0;
2247         total_qprdc = 0;
2248
2249         ixgbe_read_stats_registers(hw, hw_stats, &total_missed_rx, &total_qbrc,
2250                                                            &total_qprc, &total_qprdc);
2251
2252         /* If this is a reset xstats is NULL, and we have cleared the
2253          * registers by reading them.
2254          */
2255         if (!xstats)
2256                 return 0;
2257
2258         /* Extended stats */
2259         for (i = 0; i < IXGBE_NB_XSTATS; i++) {
2260                 snprintf(xstats[i].name, sizeof(xstats[i].name),
2261                                 "%s", rte_ixgbe_stats_strings[i].name);
2262                 xstats[i].value = *(uint64_t *)(((char *)hw_stats) +
2263                                                         rte_ixgbe_stats_strings[i].offset);
2264         }
2265
2266         return count;
2267 }
2268
2269 static void
2270 ixgbe_dev_xstats_reset(struct rte_eth_dev *dev)
2271 {
2272         struct ixgbe_hw_stats *stats =
2273                         IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
2274
2275         /* HW registers are cleared on read */
2276         ixgbe_dev_xstats_get(dev, NULL, IXGBE_NB_XSTATS);
2277
2278         /* Reset software totals */
2279         memset(stats, 0, sizeof(*stats));
2280 }
2281
2282 static void
2283 ixgbevf_dev_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
2284 {
2285         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2286         struct ixgbevf_hw_stats *hw_stats = (struct ixgbevf_hw_stats*)
2287                           IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
2288
2289         /* Good Rx packet, include VF loopback */
2290         UPDATE_VF_STAT(IXGBE_VFGPRC,
2291             hw_stats->last_vfgprc, hw_stats->vfgprc);
2292
2293         /* Good Rx octets, include VF loopback */
2294         UPDATE_VF_STAT_36BIT(IXGBE_VFGORC_LSB, IXGBE_VFGORC_MSB,
2295             hw_stats->last_vfgorc, hw_stats->vfgorc);
2296
2297         /* Good Tx packet, include VF loopback */
2298         UPDATE_VF_STAT(IXGBE_VFGPTC,
2299             hw_stats->last_vfgptc, hw_stats->vfgptc);
2300
2301         /* Good Tx octets, include VF loopback */
2302         UPDATE_VF_STAT_36BIT(IXGBE_VFGOTC_LSB, IXGBE_VFGOTC_MSB,
2303             hw_stats->last_vfgotc, hw_stats->vfgotc);
2304
2305         /* Rx Multicst Packet */
2306         UPDATE_VF_STAT(IXGBE_VFMPRC,
2307             hw_stats->last_vfmprc, hw_stats->vfmprc);
2308
2309         if (stats == NULL)
2310                 return;
2311
2312         stats->ipackets = hw_stats->vfgprc;
2313         stats->ibytes = hw_stats->vfgorc;
2314         stats->opackets = hw_stats->vfgptc;
2315         stats->obytes = hw_stats->vfgotc;
2316         stats->imcasts = hw_stats->vfmprc;
2317         /* stats->imcasts should be removed as imcasts is deprecated */
2318 }
2319
2320 static void
2321 ixgbevf_dev_stats_reset(struct rte_eth_dev *dev)
2322 {
2323         struct ixgbevf_hw_stats *hw_stats = (struct ixgbevf_hw_stats*)
2324                         IXGBE_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
2325
2326         /* Sync HW register to the last stats */
2327         ixgbevf_dev_stats_get(dev, NULL);
2328
2329         /* reset HW current stats*/
2330         hw_stats->vfgprc = 0;
2331         hw_stats->vfgorc = 0;
2332         hw_stats->vfgptc = 0;
2333         hw_stats->vfgotc = 0;
2334         hw_stats->vfmprc = 0;
2335
2336 }
2337
2338 static void
2339 ixgbe_dev_info_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
2340 {
2341         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2342
2343         dev_info->max_rx_queues = (uint16_t)hw->mac.max_rx_queues;
2344         dev_info->max_tx_queues = (uint16_t)hw->mac.max_tx_queues;
2345         dev_info->min_rx_bufsize = 1024; /* cf BSIZEPACKET in SRRCTL register */
2346         dev_info->max_rx_pktlen = 15872; /* includes CRC, cf MAXFRS register */
2347         dev_info->max_mac_addrs = hw->mac.num_rar_entries;
2348         dev_info->max_hash_mac_addrs = IXGBE_VMDQ_NUM_UC_MAC;
2349         dev_info->max_vfs = dev->pci_dev->max_vfs;
2350         if (hw->mac.type == ixgbe_mac_82598EB)
2351                 dev_info->max_vmdq_pools = ETH_16_POOLS;
2352         else
2353                 dev_info->max_vmdq_pools = ETH_64_POOLS;
2354         dev_info->vmdq_queue_num = dev_info->max_rx_queues;
2355         dev_info->rx_offload_capa =
2356                 DEV_RX_OFFLOAD_VLAN_STRIP |
2357                 DEV_RX_OFFLOAD_IPV4_CKSUM |
2358                 DEV_RX_OFFLOAD_UDP_CKSUM  |
2359                 DEV_RX_OFFLOAD_TCP_CKSUM;
2360
2361         /*
2362          * RSC is only supported by 82599 and x540 PF devices in a non-SR-IOV
2363          * mode.
2364          */
2365         if ((hw->mac.type == ixgbe_mac_82599EB ||
2366              hw->mac.type == ixgbe_mac_X540) &&
2367             !RTE_ETH_DEV_SRIOV(dev).active)
2368                 dev_info->rx_offload_capa |= DEV_RX_OFFLOAD_TCP_LRO;
2369
2370         dev_info->tx_offload_capa =
2371                 DEV_TX_OFFLOAD_VLAN_INSERT |
2372                 DEV_TX_OFFLOAD_IPV4_CKSUM  |
2373                 DEV_TX_OFFLOAD_UDP_CKSUM   |
2374                 DEV_TX_OFFLOAD_TCP_CKSUM   |
2375                 DEV_TX_OFFLOAD_SCTP_CKSUM  |
2376                 DEV_TX_OFFLOAD_TCP_TSO;
2377
2378         dev_info->default_rxconf = (struct rte_eth_rxconf) {
2379                 .rx_thresh = {
2380                         .pthresh = IXGBE_DEFAULT_RX_PTHRESH,
2381                         .hthresh = IXGBE_DEFAULT_RX_HTHRESH,
2382                         .wthresh = IXGBE_DEFAULT_RX_WTHRESH,
2383                 },
2384                 .rx_free_thresh = IXGBE_DEFAULT_RX_FREE_THRESH,
2385                 .rx_drop_en = 0,
2386         };
2387
2388         dev_info->default_txconf = (struct rte_eth_txconf) {
2389                 .tx_thresh = {
2390                         .pthresh = IXGBE_DEFAULT_TX_PTHRESH,
2391                         .hthresh = IXGBE_DEFAULT_TX_HTHRESH,
2392                         .wthresh = IXGBE_DEFAULT_TX_WTHRESH,
2393                 },
2394                 .tx_free_thresh = IXGBE_DEFAULT_TX_FREE_THRESH,
2395                 .tx_rs_thresh = IXGBE_DEFAULT_TX_RSBIT_THRESH,
2396                 .txq_flags = ETH_TXQ_FLAGS_NOMULTSEGS |
2397                                 ETH_TXQ_FLAGS_NOOFFLOADS,
2398         };
2399         dev_info->hash_key_size = IXGBE_HKEY_MAX_INDEX * sizeof(uint32_t);
2400         dev_info->reta_size = ETH_RSS_RETA_SIZE_128;
2401         dev_info->flow_type_rss_offloads = IXGBE_RSS_OFFLOAD_ALL;
2402 }
2403
2404 static void
2405 ixgbevf_dev_info_get(struct rte_eth_dev *dev,
2406                      struct rte_eth_dev_info *dev_info)
2407 {
2408         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2409
2410         dev_info->max_rx_queues = (uint16_t)hw->mac.max_rx_queues;
2411         dev_info->max_tx_queues = (uint16_t)hw->mac.max_tx_queues;
2412         dev_info->min_rx_bufsize = 1024; /* cf BSIZEPACKET in SRRCTL reg */
2413         dev_info->max_rx_pktlen = 15872; /* includes CRC, cf MAXFRS reg */
2414         dev_info->max_mac_addrs = hw->mac.num_rar_entries;
2415         dev_info->max_hash_mac_addrs = IXGBE_VMDQ_NUM_UC_MAC;
2416         dev_info->max_vfs = dev->pci_dev->max_vfs;
2417         if (hw->mac.type == ixgbe_mac_82598EB)
2418                 dev_info->max_vmdq_pools = ETH_16_POOLS;
2419         else
2420                 dev_info->max_vmdq_pools = ETH_64_POOLS;
2421         dev_info->rx_offload_capa = DEV_RX_OFFLOAD_VLAN_STRIP |
2422                                 DEV_RX_OFFLOAD_IPV4_CKSUM |
2423                                 DEV_RX_OFFLOAD_UDP_CKSUM  |
2424                                 DEV_RX_OFFLOAD_TCP_CKSUM;
2425         dev_info->tx_offload_capa = DEV_TX_OFFLOAD_VLAN_INSERT |
2426                                 DEV_TX_OFFLOAD_IPV4_CKSUM  |
2427                                 DEV_TX_OFFLOAD_UDP_CKSUM   |
2428                                 DEV_TX_OFFLOAD_TCP_CKSUM   |
2429                                 DEV_TX_OFFLOAD_SCTP_CKSUM;
2430
2431         dev_info->default_rxconf = (struct rte_eth_rxconf) {
2432                 .rx_thresh = {
2433                         .pthresh = IXGBE_DEFAULT_RX_PTHRESH,
2434                         .hthresh = IXGBE_DEFAULT_RX_HTHRESH,
2435                         .wthresh = IXGBE_DEFAULT_RX_WTHRESH,
2436                 },
2437                 .rx_free_thresh = IXGBE_DEFAULT_RX_FREE_THRESH,
2438                 .rx_drop_en = 0,
2439         };
2440
2441         dev_info->default_txconf = (struct rte_eth_txconf) {
2442                 .tx_thresh = {
2443                         .pthresh = IXGBE_DEFAULT_TX_PTHRESH,
2444                         .hthresh = IXGBE_DEFAULT_TX_HTHRESH,
2445                         .wthresh = IXGBE_DEFAULT_TX_WTHRESH,
2446                 },
2447                 .tx_free_thresh = IXGBE_DEFAULT_TX_FREE_THRESH,
2448                 .tx_rs_thresh = IXGBE_DEFAULT_TX_RSBIT_THRESH,
2449                 .txq_flags = ETH_TXQ_FLAGS_NOMULTSEGS |
2450                                 ETH_TXQ_FLAGS_NOOFFLOADS,
2451         };
2452 }
2453
2454 /* return 0 means link status changed, -1 means not changed */
2455 static int
2456 ixgbe_dev_link_update(struct rte_eth_dev *dev, int wait_to_complete)
2457 {
2458         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2459         struct rte_eth_link link, old;
2460         ixgbe_link_speed link_speed = IXGBE_LINK_SPEED_UNKNOWN;
2461         int link_up;
2462         int diag;
2463
2464         link.link_status = 0;
2465         link.link_speed = 0;
2466         link.link_duplex = 0;
2467         memset(&old, 0, sizeof(old));
2468         rte_ixgbe_dev_atomic_read_link_status(dev, &old);
2469
2470         hw->mac.get_link_status = true;
2471
2472         /* check if it needs to wait to complete, if lsc interrupt is enabled */
2473         if (wait_to_complete == 0 || dev->data->dev_conf.intr_conf.lsc != 0)
2474                 diag = ixgbe_check_link(hw, &link_speed, &link_up, 0);
2475         else
2476                 diag = ixgbe_check_link(hw, &link_speed, &link_up, 1);
2477
2478         if (diag != 0) {
2479                 link.link_speed = ETH_LINK_SPEED_100;
2480                 link.link_duplex = ETH_LINK_HALF_DUPLEX;
2481                 rte_ixgbe_dev_atomic_write_link_status(dev, &link);
2482                 if (link.link_status == old.link_status)
2483                         return -1;
2484                 return 0;
2485         }
2486
2487         if (link_up == 0) {
2488                 rte_ixgbe_dev_atomic_write_link_status(dev, &link);
2489                 if (link.link_status == old.link_status)
2490                         return -1;
2491                 return 0;
2492         }
2493         link.link_status = 1;
2494         link.link_duplex = ETH_LINK_FULL_DUPLEX;
2495
2496         switch (link_speed) {
2497         default:
2498         case IXGBE_LINK_SPEED_UNKNOWN:
2499                 link.link_duplex = ETH_LINK_HALF_DUPLEX;
2500                 link.link_speed = ETH_LINK_SPEED_100;
2501                 break;
2502
2503         case IXGBE_LINK_SPEED_100_FULL:
2504                 link.link_speed = ETH_LINK_SPEED_100;
2505                 break;
2506
2507         case IXGBE_LINK_SPEED_1GB_FULL:
2508                 link.link_speed = ETH_LINK_SPEED_1000;
2509                 break;
2510
2511         case IXGBE_LINK_SPEED_10GB_FULL:
2512                 link.link_speed = ETH_LINK_SPEED_10000;
2513                 break;
2514         }
2515         rte_ixgbe_dev_atomic_write_link_status(dev, &link);
2516
2517         if (link.link_status == old.link_status)
2518                 return -1;
2519
2520         return 0;
2521 }
2522
2523 static void
2524 ixgbe_dev_promiscuous_enable(struct rte_eth_dev *dev)
2525 {
2526         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2527         uint32_t fctrl;
2528
2529         fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
2530         fctrl |= (IXGBE_FCTRL_UPE | IXGBE_FCTRL_MPE);
2531         IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
2532 }
2533
2534 static void
2535 ixgbe_dev_promiscuous_disable(struct rte_eth_dev *dev)
2536 {
2537         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2538         uint32_t fctrl;
2539
2540         fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
2541         fctrl &= (~IXGBE_FCTRL_UPE);
2542         if (dev->data->all_multicast == 1)
2543                 fctrl |= IXGBE_FCTRL_MPE;
2544         else
2545                 fctrl &= (~IXGBE_FCTRL_MPE);
2546         IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
2547 }
2548
2549 static void
2550 ixgbe_dev_allmulticast_enable(struct rte_eth_dev *dev)
2551 {
2552         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2553         uint32_t fctrl;
2554
2555         fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
2556         fctrl |= IXGBE_FCTRL_MPE;
2557         IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
2558 }
2559
2560 static void
2561 ixgbe_dev_allmulticast_disable(struct rte_eth_dev *dev)
2562 {
2563         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2564         uint32_t fctrl;
2565
2566         if (dev->data->promiscuous == 1)
2567                 return; /* must remain in all_multicast mode */
2568
2569         fctrl = IXGBE_READ_REG(hw, IXGBE_FCTRL);
2570         fctrl &= (~IXGBE_FCTRL_MPE);
2571         IXGBE_WRITE_REG(hw, IXGBE_FCTRL, fctrl);
2572 }
2573
2574 /**
2575  * It clears the interrupt causes and enables the interrupt.
2576  * It will be called once only during nic initialized.
2577  *
2578  * @param dev
2579  *  Pointer to struct rte_eth_dev.
2580  *
2581  * @return
2582  *  - On success, zero.
2583  *  - On failure, a negative value.
2584  */
2585 static int
2586 ixgbe_dev_lsc_interrupt_setup(struct rte_eth_dev *dev)
2587 {
2588         struct ixgbe_interrupt *intr =
2589                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
2590
2591         ixgbe_dev_link_status_print(dev);
2592         intr->mask |= IXGBE_EICR_LSC;
2593
2594         return 0;
2595 }
2596
2597 /**
2598  * It clears the interrupt causes and enables the interrupt.
2599  * It will be called once only during nic initialized.
2600  *
2601  * @param dev
2602  *  Pointer to struct rte_eth_dev.
2603  *
2604  * @return
2605  *  - On success, zero.
2606  *  - On failure, a negative value.
2607  */
2608 static int
2609 ixgbe_dev_rxq_interrupt_setup(struct rte_eth_dev *dev)
2610 {
2611         struct ixgbe_interrupt *intr =
2612                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
2613
2614         intr->mask |= IXGBE_EICR_RTX_QUEUE;
2615
2616         return 0;
2617 }
2618
2619 /*
2620  * It reads ICR and sets flag (IXGBE_EICR_LSC) for the link_update.
2621  *
2622  * @param dev
2623  *  Pointer to struct rte_eth_dev.
2624  *
2625  * @return
2626  *  - On success, zero.
2627  *  - On failure, a negative value.
2628  */
2629 static int
2630 ixgbe_dev_interrupt_get_status(struct rte_eth_dev *dev)
2631 {
2632         uint32_t eicr;
2633         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2634         struct ixgbe_interrupt *intr =
2635                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
2636
2637         /* clear all cause mask */
2638         ixgbe_disable_intr(hw);
2639
2640         /* read-on-clear nic registers here */
2641         eicr = IXGBE_READ_REG(hw, IXGBE_EICR);
2642         PMD_DRV_LOG(DEBUG, "eicr %x", eicr);
2643
2644         intr->flags = 0;
2645
2646         /* set flag for async link update */
2647         if (eicr & IXGBE_EICR_LSC)
2648                 intr->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
2649
2650         if (eicr & IXGBE_EICR_MAILBOX)
2651                 intr->flags |= IXGBE_FLAG_MAILBOX;
2652
2653         return 0;
2654 }
2655
2656 static int
2657 ixgbevf_dev_interrupt_get_status(struct rte_eth_dev *dev)
2658 {
2659         uint32_t eicr;
2660         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2661         struct ixgbe_interrupt *intr =
2662                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
2663
2664         /* clear all cause mask */
2665         ixgbevf_intr_disable(hw);
2666
2667         /* read-on-clear nic registers here */
2668         eicr = IXGBE_READ_REG(hw, IXGBE_VTEICR);
2669         PMD_DRV_LOG(INFO, "eicr %x", eicr);
2670
2671         intr->flags = 0;
2672
2673         /* set flag for async link update */
2674         if (eicr & IXGBE_EICR_LSC)
2675                 intr->flags |= IXGBE_FLAG_NEED_LINK_UPDATE;
2676
2677         return 0;
2678 }
2679
2680 /**
2681  * It gets and then prints the link status.
2682  *
2683  * @param dev
2684  *  Pointer to struct rte_eth_dev.
2685  *
2686  * @return
2687  *  - On success, zero.
2688  *  - On failure, a negative value.
2689  */
2690 static void
2691 ixgbe_dev_link_status_print(struct rte_eth_dev *dev)
2692 {
2693         struct rte_eth_link link;
2694
2695         memset(&link, 0, sizeof(link));
2696         rte_ixgbe_dev_atomic_read_link_status(dev, &link);
2697         if (link.link_status) {
2698                 PMD_INIT_LOG(INFO, "Port %d: Link Up - speed %u Mbps - %s",
2699                                         (int)(dev->data->port_id),
2700                                         (unsigned)link.link_speed,
2701                         link.link_duplex == ETH_LINK_FULL_DUPLEX ?
2702                                         "full-duplex" : "half-duplex");
2703         } else {
2704                 PMD_INIT_LOG(INFO, " Port %d: Link Down",
2705                                 (int)(dev->data->port_id));
2706         }
2707         PMD_INIT_LOG(DEBUG, "PCI Address: %04d:%02d:%02d:%d",
2708                                 dev->pci_dev->addr.domain,
2709                                 dev->pci_dev->addr.bus,
2710                                 dev->pci_dev->addr.devid,
2711                                 dev->pci_dev->addr.function);
2712 }
2713
2714 /*
2715  * It executes link_update after knowing an interrupt occurred.
2716  *
2717  * @param dev
2718  *  Pointer to struct rte_eth_dev.
2719  *
2720  * @return
2721  *  - On success, zero.
2722  *  - On failure, a negative value.
2723  */
2724 static int
2725 ixgbe_dev_interrupt_action(struct rte_eth_dev *dev)
2726 {
2727         struct ixgbe_interrupt *intr =
2728                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
2729         int64_t timeout;
2730         struct rte_eth_link link;
2731         int intr_enable_delay = false;
2732
2733         PMD_DRV_LOG(DEBUG, "intr action type %d", intr->flags);
2734
2735         if (intr->flags & IXGBE_FLAG_MAILBOX) {
2736                 ixgbe_pf_mbx_process(dev);
2737                 intr->flags &= ~IXGBE_FLAG_MAILBOX;
2738         }
2739
2740         if (intr->flags & IXGBE_FLAG_NEED_LINK_UPDATE) {
2741                 /* get the link status before link update, for predicting later */
2742                 memset(&link, 0, sizeof(link));
2743                 rte_ixgbe_dev_atomic_read_link_status(dev, &link);
2744
2745                 ixgbe_dev_link_update(dev, 0);
2746
2747                 /* likely to up */
2748                 if (!link.link_status)
2749                         /* handle it 1 sec later, wait it being stable */
2750                         timeout = IXGBE_LINK_UP_CHECK_TIMEOUT;
2751                 /* likely to down */
2752                 else
2753                         /* handle it 4 sec later, wait it being stable */
2754                         timeout = IXGBE_LINK_DOWN_CHECK_TIMEOUT;
2755
2756                 ixgbe_dev_link_status_print(dev);
2757
2758                 intr_enable_delay = true;
2759         }
2760
2761         if (intr_enable_delay) {
2762                 if (rte_eal_alarm_set(timeout * 1000,
2763                                       ixgbe_dev_interrupt_delayed_handler, (void*)dev) < 0)
2764                         PMD_DRV_LOG(ERR, "Error setting alarm");
2765         } else {
2766                 PMD_DRV_LOG(DEBUG, "enable intr immediately");
2767                 ixgbe_enable_intr(dev);
2768                 rte_intr_enable(&(dev->pci_dev->intr_handle));
2769         }
2770
2771
2772         return 0;
2773 }
2774
2775 static int
2776 ixgbevf_dev_interrupt_action(struct rte_eth_dev *dev)
2777 {
2778         struct ixgbe_hw *hw =
2779                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2780
2781         PMD_DRV_LOG(DEBUG, "enable intr immediately");
2782         ixgbevf_intr_enable(hw);
2783         rte_intr_enable(&dev->pci_dev->intr_handle);
2784         return 0;
2785 }
2786
2787 /**
2788  * Interrupt handler which shall be registered for alarm callback for delayed
2789  * handling specific interrupt to wait for the stable nic state. As the
2790  * NIC interrupt state is not stable for ixgbe after link is just down,
2791  * it needs to wait 4 seconds to get the stable status.
2792  *
2793  * @param handle
2794  *  Pointer to interrupt handle.
2795  * @param param
2796  *  The address of parameter (struct rte_eth_dev *) regsitered before.
2797  *
2798  * @return
2799  *  void
2800  */
2801 static void
2802 ixgbe_dev_interrupt_delayed_handler(void *param)
2803 {
2804         struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
2805         struct ixgbe_interrupt *intr =
2806                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
2807         struct ixgbe_hw *hw =
2808                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2809         uint32_t eicr;
2810
2811         eicr = IXGBE_READ_REG(hw, IXGBE_EICR);
2812         if (eicr & IXGBE_EICR_MAILBOX)
2813                 ixgbe_pf_mbx_process(dev);
2814
2815         if (intr->flags & IXGBE_FLAG_NEED_LINK_UPDATE) {
2816                 ixgbe_dev_link_update(dev, 0);
2817                 intr->flags &= ~IXGBE_FLAG_NEED_LINK_UPDATE;
2818                 ixgbe_dev_link_status_print(dev);
2819                 _rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_LSC);
2820         }
2821
2822         PMD_DRV_LOG(DEBUG, "enable intr in delayed handler S[%08x]", eicr);
2823         ixgbe_enable_intr(dev);
2824         rte_intr_enable(&(dev->pci_dev->intr_handle));
2825 }
2826
2827 /**
2828  * Interrupt handler triggered by NIC  for handling
2829  * specific interrupt.
2830  *
2831  * @param handle
2832  *  Pointer to interrupt handle.
2833  * @param param
2834  *  The address of parameter (struct rte_eth_dev *) regsitered before.
2835  *
2836  * @return
2837  *  void
2838  */
2839 static void
2840 ixgbe_dev_interrupt_handler(__rte_unused struct rte_intr_handle *handle,
2841                             void *param)
2842 {
2843         struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
2844
2845         ixgbe_dev_interrupt_get_status(dev);
2846         ixgbe_dev_interrupt_action(dev);
2847 }
2848
2849 static void
2850 ixgbevf_dev_interrupt_handler(__rte_unused struct rte_intr_handle *handle,
2851                               void *param)
2852 {
2853         struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
2854
2855         ixgbevf_dev_interrupt_get_status(dev);
2856         ixgbevf_dev_interrupt_action(dev);
2857 }
2858
2859 static int
2860 ixgbe_dev_led_on(struct rte_eth_dev *dev)
2861 {
2862         struct ixgbe_hw *hw;
2863
2864         hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2865         return (ixgbe_led_on(hw, 0) == IXGBE_SUCCESS ? 0 : -ENOTSUP);
2866 }
2867
2868 static int
2869 ixgbe_dev_led_off(struct rte_eth_dev *dev)
2870 {
2871         struct ixgbe_hw *hw;
2872
2873         hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2874         return (ixgbe_led_off(hw, 0) == IXGBE_SUCCESS ? 0 : -ENOTSUP);
2875 }
2876
2877 static int
2878 ixgbe_flow_ctrl_get(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
2879 {
2880         struct ixgbe_hw *hw;
2881         uint32_t mflcn_reg;
2882         uint32_t fccfg_reg;
2883         int rx_pause;
2884         int tx_pause;
2885
2886         hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2887
2888         fc_conf->pause_time = hw->fc.pause_time;
2889         fc_conf->high_water = hw->fc.high_water[0];
2890         fc_conf->low_water = hw->fc.low_water[0];
2891         fc_conf->send_xon = hw->fc.send_xon;
2892         fc_conf->autoneg = !hw->fc.disable_fc_autoneg;
2893
2894         /*
2895          * Return rx_pause status according to actual setting of
2896          * MFLCN register.
2897          */
2898         mflcn_reg = IXGBE_READ_REG(hw, IXGBE_MFLCN);
2899         if (mflcn_reg & (IXGBE_MFLCN_RPFCE | IXGBE_MFLCN_RFCE))
2900                 rx_pause = 1;
2901         else
2902                 rx_pause = 0;
2903
2904         /*
2905          * Return tx_pause status according to actual setting of
2906          * FCCFG register.
2907          */
2908         fccfg_reg = IXGBE_READ_REG(hw, IXGBE_FCCFG);
2909         if (fccfg_reg & (IXGBE_FCCFG_TFCE_802_3X | IXGBE_FCCFG_TFCE_PRIORITY))
2910                 tx_pause = 1;
2911         else
2912                 tx_pause = 0;
2913
2914         if (rx_pause && tx_pause)
2915                 fc_conf->mode = RTE_FC_FULL;
2916         else if (rx_pause)
2917                 fc_conf->mode = RTE_FC_RX_PAUSE;
2918         else if (tx_pause)
2919                 fc_conf->mode = RTE_FC_TX_PAUSE;
2920         else
2921                 fc_conf->mode = RTE_FC_NONE;
2922
2923         return 0;
2924 }
2925
2926 static int
2927 ixgbe_flow_ctrl_set(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
2928 {
2929         struct ixgbe_hw *hw;
2930         int err;
2931         uint32_t rx_buf_size;
2932         uint32_t max_high_water;
2933         uint32_t mflcn;
2934         enum ixgbe_fc_mode rte_fcmode_2_ixgbe_fcmode[] = {
2935                 ixgbe_fc_none,
2936                 ixgbe_fc_rx_pause,
2937                 ixgbe_fc_tx_pause,
2938                 ixgbe_fc_full
2939         };
2940
2941         PMD_INIT_FUNC_TRACE();
2942
2943         hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2944         rx_buf_size = IXGBE_READ_REG(hw, IXGBE_RXPBSIZE(0));
2945         PMD_INIT_LOG(DEBUG, "Rx packet buffer size = 0x%x", rx_buf_size);
2946
2947         /*
2948          * At least reserve one Ethernet frame for watermark
2949          * high_water/low_water in kilo bytes for ixgbe
2950          */
2951         max_high_water = (rx_buf_size - ETHER_MAX_LEN) >> IXGBE_RXPBSIZE_SHIFT;
2952         if ((fc_conf->high_water > max_high_water) ||
2953                 (fc_conf->high_water < fc_conf->low_water)) {
2954                 PMD_INIT_LOG(ERR, "Invalid high/low water setup value in KB");
2955                 PMD_INIT_LOG(ERR, "High_water must <= 0x%x", max_high_water);
2956                 return (-EINVAL);
2957         }
2958
2959         hw->fc.requested_mode = rte_fcmode_2_ixgbe_fcmode[fc_conf->mode];
2960         hw->fc.pause_time     = fc_conf->pause_time;
2961         hw->fc.high_water[0]  = fc_conf->high_water;
2962         hw->fc.low_water[0]   = fc_conf->low_water;
2963         hw->fc.send_xon       = fc_conf->send_xon;
2964         hw->fc.disable_fc_autoneg = !fc_conf->autoneg;
2965
2966         err = ixgbe_fc_enable(hw);
2967
2968         /* Not negotiated is not an error case */
2969         if ((err == IXGBE_SUCCESS) || (err == IXGBE_ERR_FC_NOT_NEGOTIATED)) {
2970
2971                 /* check if we want to forward MAC frames - driver doesn't have native
2972                  * capability to do that, so we'll write the registers ourselves */
2973
2974                 mflcn = IXGBE_READ_REG(hw, IXGBE_MFLCN);
2975
2976                 /* set or clear MFLCN.PMCF bit depending on configuration */
2977                 if (fc_conf->mac_ctrl_frame_fwd != 0)
2978                         mflcn |= IXGBE_MFLCN_PMCF;
2979                 else
2980                         mflcn &= ~IXGBE_MFLCN_PMCF;
2981
2982                 IXGBE_WRITE_REG(hw, IXGBE_MFLCN, mflcn);
2983                 IXGBE_WRITE_FLUSH(hw);
2984
2985                 return 0;
2986         }
2987
2988         PMD_INIT_LOG(ERR, "ixgbe_fc_enable = 0x%x", err);
2989         return -EIO;
2990 }
2991
2992 /**
2993  *  ixgbe_pfc_enable_generic - Enable flow control
2994  *  @hw: pointer to hardware structure
2995  *  @tc_num: traffic class number
2996  *  Enable flow control according to the current settings.
2997  */
2998 static int
2999 ixgbe_dcb_pfc_enable_generic(struct ixgbe_hw *hw,uint8_t tc_num)
3000 {
3001         int ret_val = 0;
3002         uint32_t mflcn_reg, fccfg_reg;
3003         uint32_t reg;
3004         uint32_t fcrtl, fcrth;
3005         uint8_t i;
3006         uint8_t nb_rx_en;
3007
3008         /* Validate the water mark configuration */
3009         if (!hw->fc.pause_time) {
3010                 ret_val = IXGBE_ERR_INVALID_LINK_SETTINGS;
3011                 goto out;
3012         }
3013
3014         /* Low water mark of zero causes XOFF floods */
3015         if (hw->fc.current_mode & ixgbe_fc_tx_pause) {
3016                  /* High/Low water can not be 0 */
3017                 if( (!hw->fc.high_water[tc_num])|| (!hw->fc.low_water[tc_num])) {
3018                         PMD_INIT_LOG(ERR, "Invalid water mark configuration");
3019                         ret_val = IXGBE_ERR_INVALID_LINK_SETTINGS;
3020                         goto out;
3021                 }
3022
3023                 if(hw->fc.low_water[tc_num] >= hw->fc.high_water[tc_num]) {
3024                         PMD_INIT_LOG(ERR, "Invalid water mark configuration");
3025                         ret_val = IXGBE_ERR_INVALID_LINK_SETTINGS;
3026                         goto out;
3027                 }
3028         }
3029         /* Negotiate the fc mode to use */
3030         ixgbe_fc_autoneg(hw);
3031
3032         /* Disable any previous flow control settings */
3033         mflcn_reg = IXGBE_READ_REG(hw, IXGBE_MFLCN);
3034         mflcn_reg &= ~(IXGBE_MFLCN_RPFCE_SHIFT | IXGBE_MFLCN_RFCE|IXGBE_MFLCN_RPFCE);
3035
3036         fccfg_reg = IXGBE_READ_REG(hw, IXGBE_FCCFG);
3037         fccfg_reg &= ~(IXGBE_FCCFG_TFCE_802_3X | IXGBE_FCCFG_TFCE_PRIORITY);
3038
3039         switch (hw->fc.current_mode) {
3040         case ixgbe_fc_none:
3041                 /*
3042                  * If the count of enabled RX Priority Flow control >1,
3043                  * and the TX pause can not be disabled
3044                  */
3045                 nb_rx_en = 0;
3046                 for (i =0; i < IXGBE_DCB_MAX_TRAFFIC_CLASS; i++) {
3047                         reg = IXGBE_READ_REG(hw, IXGBE_FCRTH_82599(i));
3048                         if (reg & IXGBE_FCRTH_FCEN)
3049                                 nb_rx_en++;
3050                 }
3051                 if (nb_rx_en > 1)
3052                         fccfg_reg |=IXGBE_FCCFG_TFCE_PRIORITY;
3053                 break;
3054         case ixgbe_fc_rx_pause:
3055                 /*
3056                  * Rx Flow control is enabled and Tx Flow control is
3057                  * disabled by software override. Since there really
3058                  * isn't a way to advertise that we are capable of RX
3059                  * Pause ONLY, we will advertise that we support both
3060                  * symmetric and asymmetric Rx PAUSE.  Later, we will
3061                  * disable the adapter's ability to send PAUSE frames.
3062                  */
3063                 mflcn_reg |= IXGBE_MFLCN_RPFCE;
3064                 /*
3065                  * If the count of enabled RX Priority Flow control >1,
3066                  * and the TX pause can not be disabled
3067                  */
3068                 nb_rx_en = 0;
3069                 for (i =0; i < IXGBE_DCB_MAX_TRAFFIC_CLASS; i++) {
3070                         reg = IXGBE_READ_REG(hw, IXGBE_FCRTH_82599(i));
3071                         if (reg & IXGBE_FCRTH_FCEN)
3072                                 nb_rx_en++;
3073                 }
3074                 if (nb_rx_en > 1)
3075                         fccfg_reg |=IXGBE_FCCFG_TFCE_PRIORITY;
3076                 break;
3077         case ixgbe_fc_tx_pause:
3078                 /*
3079                  * Tx Flow control is enabled, and Rx Flow control is
3080                  * disabled by software override.
3081                  */
3082                 fccfg_reg |=IXGBE_FCCFG_TFCE_PRIORITY;
3083                 break;
3084         case ixgbe_fc_full:
3085                 /* Flow control (both Rx and Tx) is enabled by SW override. */
3086                 mflcn_reg |= IXGBE_MFLCN_RPFCE;
3087                 fccfg_reg |= IXGBE_FCCFG_TFCE_PRIORITY;
3088                 break;
3089         default:
3090                 PMD_DRV_LOG(DEBUG, "Flow control param set incorrectly");
3091                 ret_val = IXGBE_ERR_CONFIG;
3092                 goto out;
3093                 break;
3094         }
3095
3096         /* Set 802.3x based flow control settings. */
3097         mflcn_reg |= IXGBE_MFLCN_DPF;
3098         IXGBE_WRITE_REG(hw, IXGBE_MFLCN, mflcn_reg);
3099         IXGBE_WRITE_REG(hw, IXGBE_FCCFG, fccfg_reg);
3100
3101         /* Set up and enable Rx high/low water mark thresholds, enable XON. */
3102         if ((hw->fc.current_mode & ixgbe_fc_tx_pause) &&
3103                 hw->fc.high_water[tc_num]) {
3104                 fcrtl = (hw->fc.low_water[tc_num] << 10) | IXGBE_FCRTL_XONE;
3105                 IXGBE_WRITE_REG(hw, IXGBE_FCRTL_82599(tc_num), fcrtl);
3106                 fcrth = (hw->fc.high_water[tc_num] << 10) | IXGBE_FCRTH_FCEN;
3107         } else {
3108                 IXGBE_WRITE_REG(hw, IXGBE_FCRTL_82599(tc_num), 0);
3109                 /*
3110                  * In order to prevent Tx hangs when the internal Tx
3111                  * switch is enabled we must set the high water mark
3112                  * to the maximum FCRTH value.  This allows the Tx
3113                  * switch to function even under heavy Rx workloads.
3114                  */
3115                 fcrth = IXGBE_READ_REG(hw, IXGBE_RXPBSIZE(tc_num)) - 32;
3116         }
3117         IXGBE_WRITE_REG(hw, IXGBE_FCRTH_82599(tc_num), fcrth);
3118
3119         /* Configure pause time (2 TCs per register) */
3120         reg = hw->fc.pause_time * 0x00010001;
3121         for (i = 0; i < (IXGBE_DCB_MAX_TRAFFIC_CLASS / 2); i++)
3122                 IXGBE_WRITE_REG(hw, IXGBE_FCTTV(i), reg);
3123
3124         /* Configure flow control refresh threshold value */
3125         IXGBE_WRITE_REG(hw, IXGBE_FCRTV, hw->fc.pause_time / 2);
3126
3127 out:
3128         return ret_val;
3129 }
3130
3131 static int
3132 ixgbe_dcb_pfc_enable(struct rte_eth_dev *dev,uint8_t tc_num)
3133 {
3134         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3135         int32_t ret_val = IXGBE_NOT_IMPLEMENTED;
3136
3137         if(hw->mac.type != ixgbe_mac_82598EB) {
3138                 ret_val = ixgbe_dcb_pfc_enable_generic(hw,tc_num);
3139         }
3140         return ret_val;
3141 }
3142
3143 static int
3144 ixgbe_priority_flow_ctrl_set(struct rte_eth_dev *dev, struct rte_eth_pfc_conf *pfc_conf)
3145 {
3146         int err;
3147         uint32_t rx_buf_size;
3148         uint32_t max_high_water;
3149         uint8_t tc_num;
3150         uint8_t  map[IXGBE_DCB_MAX_USER_PRIORITY] = { 0 };
3151         struct ixgbe_hw *hw =
3152                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3153         struct ixgbe_dcb_config *dcb_config =
3154                 IXGBE_DEV_PRIVATE_TO_DCB_CFG(dev->data->dev_private);
3155
3156         enum ixgbe_fc_mode rte_fcmode_2_ixgbe_fcmode[] = {
3157                 ixgbe_fc_none,
3158                 ixgbe_fc_rx_pause,
3159                 ixgbe_fc_tx_pause,
3160                 ixgbe_fc_full
3161         };
3162
3163         PMD_INIT_FUNC_TRACE();
3164
3165         ixgbe_dcb_unpack_map_cee(dcb_config, IXGBE_DCB_RX_CONFIG, map);
3166         tc_num = map[pfc_conf->priority];
3167         rx_buf_size = IXGBE_READ_REG(hw, IXGBE_RXPBSIZE(tc_num));
3168         PMD_INIT_LOG(DEBUG, "Rx packet buffer size = 0x%x", rx_buf_size);
3169         /*
3170          * At least reserve one Ethernet frame for watermark
3171          * high_water/low_water in kilo bytes for ixgbe
3172          */
3173         max_high_water = (rx_buf_size - ETHER_MAX_LEN) >> IXGBE_RXPBSIZE_SHIFT;
3174         if ((pfc_conf->fc.high_water > max_high_water) ||
3175             (pfc_conf->fc.high_water <= pfc_conf->fc.low_water)) {
3176                 PMD_INIT_LOG(ERR, "Invalid high/low water setup value in KB");
3177                 PMD_INIT_LOG(ERR, "High_water must <= 0x%x", max_high_water);
3178                 return (-EINVAL);
3179         }
3180
3181         hw->fc.requested_mode = rte_fcmode_2_ixgbe_fcmode[pfc_conf->fc.mode];
3182         hw->fc.pause_time = pfc_conf->fc.pause_time;
3183         hw->fc.send_xon = pfc_conf->fc.send_xon;
3184         hw->fc.low_water[tc_num] =  pfc_conf->fc.low_water;
3185         hw->fc.high_water[tc_num] = pfc_conf->fc.high_water;
3186
3187         err = ixgbe_dcb_pfc_enable(dev,tc_num);
3188
3189         /* Not negotiated is not an error case */
3190         if ((err == IXGBE_SUCCESS) || (err == IXGBE_ERR_FC_NOT_NEGOTIATED))
3191                 return 0;
3192
3193         PMD_INIT_LOG(ERR, "ixgbe_dcb_pfc_enable = 0x%x", err);
3194         return -EIO;
3195 }
3196
3197 static int
3198 ixgbe_dev_rss_reta_update(struct rte_eth_dev *dev,
3199                           struct rte_eth_rss_reta_entry64 *reta_conf,
3200                           uint16_t reta_size)
3201 {
3202         uint8_t i, j, mask;
3203         uint32_t reta, r;
3204         uint16_t idx, shift;
3205         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3206
3207         PMD_INIT_FUNC_TRACE();
3208         if (reta_size != ETH_RSS_RETA_SIZE_128) {
3209                 PMD_DRV_LOG(ERR, "The size of hash lookup table configured "
3210                         "(%d) doesn't match the number hardware can supported "
3211                         "(%d)\n", reta_size, ETH_RSS_RETA_SIZE_128);
3212                 return -EINVAL;
3213         }
3214
3215         for (i = 0; i < reta_size; i += IXGBE_4_BIT_WIDTH) {
3216                 idx = i / RTE_RETA_GROUP_SIZE;
3217                 shift = i % RTE_RETA_GROUP_SIZE;
3218                 mask = (uint8_t)((reta_conf[idx].mask >> shift) &
3219                                                 IXGBE_4_BIT_MASK);
3220                 if (!mask)
3221                         continue;
3222                 if (mask == IXGBE_4_BIT_MASK)
3223                         r = 0;
3224                 else
3225                         r = IXGBE_READ_REG(hw, IXGBE_RETA(i >> 2));
3226                 for (j = 0, reta = 0; j < IXGBE_4_BIT_WIDTH; j++) {
3227                         if (mask & (0x1 << j))
3228                                 reta |= reta_conf[idx].reta[shift + j] <<
3229                                                         (CHAR_BIT * j);
3230                         else
3231                                 reta |= r & (IXGBE_8_BIT_MASK <<
3232                                                 (CHAR_BIT * j));
3233                 }
3234                 IXGBE_WRITE_REG(hw, IXGBE_RETA(i >> 2), reta);
3235         }
3236
3237         return 0;
3238 }
3239
3240 static int
3241 ixgbe_dev_rss_reta_query(struct rte_eth_dev *dev,
3242                          struct rte_eth_rss_reta_entry64 *reta_conf,
3243                          uint16_t reta_size)
3244 {
3245         uint8_t i, j, mask;
3246         uint32_t reta;
3247         uint16_t idx, shift;
3248         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3249
3250         PMD_INIT_FUNC_TRACE();
3251         if (reta_size != ETH_RSS_RETA_SIZE_128) {
3252                 PMD_DRV_LOG(ERR, "The size of hash lookup table configured "
3253                         "(%d) doesn't match the number hardware can supported "
3254                                 "(%d)\n", reta_size, ETH_RSS_RETA_SIZE_128);
3255                 return -EINVAL;
3256         }
3257
3258         for (i = 0; i < ETH_RSS_RETA_SIZE_128; i += IXGBE_4_BIT_WIDTH) {
3259                 idx = i / RTE_RETA_GROUP_SIZE;
3260                 shift = i % RTE_RETA_GROUP_SIZE;
3261                 mask = (uint8_t)((reta_conf[idx].mask >> shift) &
3262                                                 IXGBE_4_BIT_MASK);
3263                 if (!mask)
3264                         continue;
3265
3266                 reta = IXGBE_READ_REG(hw, IXGBE_RETA(i >> 2));
3267                 for (j = 0; j < IXGBE_4_BIT_WIDTH; j++) {
3268                         if (mask & (0x1 << j))
3269                                 reta_conf[idx].reta[shift + j] =
3270                                         ((reta >> (CHAR_BIT * j)) &
3271                                                 IXGBE_8_BIT_MASK);
3272                 }
3273         }
3274
3275         return 0;
3276 }
3277
3278 static void
3279 ixgbe_add_rar(struct rte_eth_dev *dev, struct ether_addr *mac_addr,
3280                                 uint32_t index, uint32_t pool)
3281 {
3282         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3283         uint32_t enable_addr = 1;
3284
3285         ixgbe_set_rar(hw, index, mac_addr->addr_bytes, pool, enable_addr);
3286 }
3287
3288 static void
3289 ixgbe_remove_rar(struct rte_eth_dev *dev, uint32_t index)
3290 {
3291         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3292
3293         ixgbe_clear_rar(hw, index);
3294 }
3295
3296 static void
3297 ixgbe_set_default_mac_addr(struct rte_eth_dev *dev, struct ether_addr *addr)
3298 {
3299         ixgbe_remove_rar(dev, 0);
3300
3301         ixgbe_add_rar(dev, addr, 0, 0);
3302 }
3303
3304 static int
3305 ixgbe_dev_mtu_set(struct rte_eth_dev *dev, uint16_t mtu)
3306 {
3307         uint32_t hlreg0;
3308         uint32_t maxfrs;
3309         struct ixgbe_hw *hw;
3310         struct rte_eth_dev_info dev_info;
3311         uint32_t frame_size = mtu + ETHER_HDR_LEN + ETHER_CRC_LEN;
3312
3313         ixgbe_dev_info_get(dev, &dev_info);
3314
3315         /* check that mtu is within the allowed range */
3316         if ((mtu < ETHER_MIN_MTU) || (frame_size > dev_info.max_rx_pktlen))
3317                 return -EINVAL;
3318
3319         /* refuse mtu that requires the support of scattered packets when this
3320          * feature has not been enabled before. */
3321         if (!dev->data->scattered_rx &&
3322             (frame_size + 2 * IXGBE_VLAN_TAG_SIZE >
3323              dev->data->min_rx_buf_size - RTE_PKTMBUF_HEADROOM))
3324                 return -EINVAL;
3325
3326         hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3327         hlreg0 = IXGBE_READ_REG(hw, IXGBE_HLREG0);
3328
3329         /* switch to jumbo mode if needed */
3330         if (frame_size > ETHER_MAX_LEN) {
3331                 dev->data->dev_conf.rxmode.jumbo_frame = 1;
3332                 hlreg0 |= IXGBE_HLREG0_JUMBOEN;
3333         } else {
3334                 dev->data->dev_conf.rxmode.jumbo_frame = 0;
3335                 hlreg0 &= ~IXGBE_HLREG0_JUMBOEN;
3336         }
3337         IXGBE_WRITE_REG(hw, IXGBE_HLREG0, hlreg0);
3338
3339         /* update max frame size */
3340         dev->data->dev_conf.rxmode.max_rx_pkt_len = frame_size;
3341
3342         maxfrs = IXGBE_READ_REG(hw, IXGBE_MAXFRS);
3343         maxfrs &= 0x0000FFFF;
3344         maxfrs |= (dev->data->dev_conf.rxmode.max_rx_pkt_len << 16);
3345         IXGBE_WRITE_REG(hw, IXGBE_MAXFRS, maxfrs);
3346
3347         return 0;
3348 }
3349
3350 /*
3351  * Virtual Function operations
3352  */
3353 static void
3354 ixgbevf_intr_disable(struct ixgbe_hw *hw)
3355 {
3356         PMD_INIT_FUNC_TRACE();
3357
3358         /* Clear interrupt mask to stop from interrupts being generated */
3359         IXGBE_WRITE_REG(hw, IXGBE_VTEIMC, IXGBE_VF_IRQ_CLEAR_MASK);
3360
3361         IXGBE_WRITE_FLUSH(hw);
3362 }
3363
3364 static void
3365 ixgbevf_intr_enable(struct ixgbe_hw *hw)
3366 {
3367         PMD_INIT_FUNC_TRACE();
3368
3369         /* VF enable interrupt autoclean */
3370         IXGBE_WRITE_REG(hw, IXGBE_VTEIAM, IXGBE_VF_IRQ_ENABLE_MASK);
3371         IXGBE_WRITE_REG(hw, IXGBE_VTEIAC, IXGBE_VF_IRQ_ENABLE_MASK);
3372         IXGBE_WRITE_REG(hw, IXGBE_VTEIMS, IXGBE_VF_IRQ_ENABLE_MASK);
3373
3374         IXGBE_WRITE_FLUSH(hw);
3375 }
3376
3377 static int
3378 ixgbevf_dev_configure(struct rte_eth_dev *dev)
3379 {
3380         struct rte_eth_conf* conf = &dev->data->dev_conf;
3381         struct ixgbe_adapter *adapter =
3382                         (struct ixgbe_adapter *)dev->data->dev_private;
3383
3384         PMD_INIT_LOG(DEBUG, "Configured Virtual Function port id: %d",
3385                      dev->data->port_id);
3386
3387         /*
3388          * VF has no ability to enable/disable HW CRC
3389          * Keep the persistent behavior the same as Host PF
3390          */
3391 #ifndef RTE_LIBRTE_IXGBE_PF_DISABLE_STRIP_CRC
3392         if (!conf->rxmode.hw_strip_crc) {
3393                 PMD_INIT_LOG(NOTICE, "VF can't disable HW CRC Strip");
3394                 conf->rxmode.hw_strip_crc = 1;
3395         }
3396 #else
3397         if (conf->rxmode.hw_strip_crc) {
3398                 PMD_INIT_LOG(NOTICE, "VF can't enable HW CRC Strip");
3399                 conf->rxmode.hw_strip_crc = 0;
3400         }
3401 #endif
3402
3403         /*
3404          * Initialize to TRUE. If any of Rx queues doesn't meet the bulk
3405          * allocation or vector Rx preconditions we will reset it.
3406          */
3407         adapter->rx_bulk_alloc_allowed = true;
3408         adapter->rx_vec_allowed = true;
3409
3410         return 0;
3411 }
3412
3413 static int
3414 ixgbevf_dev_start(struct rte_eth_dev *dev)
3415 {
3416         struct ixgbe_hw *hw =
3417                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3418         uint32_t intr_vector = 0;
3419         struct rte_intr_handle *intr_handle = &dev->pci_dev->intr_handle;
3420
3421         int err, mask = 0;
3422
3423         PMD_INIT_FUNC_TRACE();
3424
3425         hw->mac.ops.reset_hw(hw);
3426         hw->mac.get_link_status = true;
3427
3428         /* negotiate mailbox API version to use with the PF. */
3429         ixgbevf_negotiate_api(hw);
3430
3431         ixgbevf_dev_tx_init(dev);
3432
3433         /* This can fail when allocating mbufs for descriptor rings */
3434         err = ixgbevf_dev_rx_init(dev);
3435         if (err) {
3436                 PMD_INIT_LOG(ERR, "Unable to initialize RX hardware (%d)", err);
3437                 ixgbe_dev_clear_queues(dev);
3438                 return err;
3439         }
3440
3441         /* Set vfta */
3442         ixgbevf_set_vfta_all(dev,1);
3443
3444         /* Set HW strip */
3445         mask = ETH_VLAN_STRIP_MASK | ETH_VLAN_FILTER_MASK | \
3446                 ETH_VLAN_EXTEND_MASK;
3447         ixgbevf_vlan_offload_set(dev, mask);
3448
3449         ixgbevf_dev_rxtx_start(dev);
3450
3451         /* check and configure queue intr-vector mapping */
3452         if (dev->data->dev_conf.intr_conf.rxq != 0)
3453                 intr_vector = dev->data->nb_rx_queues;
3454
3455         if (rte_intr_efd_enable(intr_handle, intr_vector))
3456                 return -1;
3457
3458         if (rte_intr_dp_is_en(intr_handle) && !intr_handle->intr_vec) {
3459                 intr_handle->intr_vec =
3460                         rte_zmalloc("intr_vec",
3461                                     dev->data->nb_rx_queues * sizeof(int), 0);
3462                 if (intr_handle->intr_vec == NULL) {
3463                         PMD_INIT_LOG(ERR, "Failed to allocate %d rx_queues"
3464                                      " intr_vec\n", dev->data->nb_rx_queues);
3465                         return -ENOMEM;
3466                 }
3467         }
3468         ixgbevf_configure_msix(dev);
3469
3470         if (dev->data->dev_conf.intr_conf.lsc != 0) {
3471                 if (rte_intr_allow_others(intr_handle))
3472                         rte_intr_callback_register(intr_handle,
3473                                         ixgbevf_dev_interrupt_handler,
3474                                         (void *)dev);
3475                 else
3476                         PMD_INIT_LOG(INFO, "lsc won't enable because of"
3477                                      " no intr multiplex\n");
3478         }
3479
3480         rte_intr_enable(intr_handle);
3481
3482         /* Re-enable interrupt for VF */
3483         ixgbevf_intr_enable(hw);
3484
3485         return 0;
3486 }
3487
3488 static void
3489 ixgbevf_dev_stop(struct rte_eth_dev *dev)
3490 {
3491         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3492         struct rte_intr_handle *intr_handle = &dev->pci_dev->intr_handle;
3493
3494         PMD_INIT_FUNC_TRACE();
3495
3496         hw->adapter_stopped = 1;
3497         ixgbe_stop_adapter(hw);
3498
3499         /*
3500           * Clear what we set, but we still keep shadow_vfta to
3501           * restore after device starts
3502           */
3503         ixgbevf_set_vfta_all(dev,0);
3504
3505         /* Clear stored conf */
3506         dev->data->scattered_rx = 0;
3507
3508         ixgbe_dev_clear_queues(dev);
3509
3510         /* disable intr eventfd mapping */
3511         rte_intr_disable(intr_handle);
3512
3513         /* Clean datapath event and queue/vec mapping */
3514         rte_intr_efd_disable(intr_handle);
3515         if (intr_handle->intr_vec != NULL) {
3516                 rte_free(intr_handle->intr_vec);
3517                 intr_handle->intr_vec = NULL;
3518         }
3519 }
3520
3521 static void
3522 ixgbevf_dev_close(struct rte_eth_dev *dev)
3523 {
3524         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3525         struct rte_pci_device *pci_dev;
3526
3527         PMD_INIT_FUNC_TRACE();
3528
3529         ixgbe_reset_hw(hw);
3530
3531         ixgbevf_dev_stop(dev);
3532
3533         ixgbe_dev_free_queues(dev);
3534
3535         /* reprogram the RAR[0] in case user changed it. */
3536         ixgbe_set_rar(hw, 0, hw->mac.addr, 0, IXGBE_RAH_AV);
3537
3538         pci_dev = dev->pci_dev;
3539         if (pci_dev->intr_handle.intr_vec) {
3540                 rte_free(pci_dev->intr_handle.intr_vec);
3541                 pci_dev->intr_handle.intr_vec = NULL;
3542         }
3543 }
3544
3545 static void ixgbevf_set_vfta_all(struct rte_eth_dev *dev, bool on)
3546 {
3547         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3548         struct ixgbe_vfta * shadow_vfta =
3549                 IXGBE_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
3550         int i = 0, j = 0, vfta = 0, mask = 1;
3551
3552         for (i = 0; i < IXGBE_VFTA_SIZE; i++){
3553                 vfta = shadow_vfta->vfta[i];
3554                 if(vfta){
3555                         mask = 1;
3556                         for (j = 0; j < 32; j++){
3557                                 if(vfta & mask)
3558                                         ixgbe_set_vfta(hw, (i<<5)+j, 0, on);
3559                                 mask<<=1;
3560                         }
3561                 }
3562         }
3563
3564 }
3565
3566 static int
3567 ixgbevf_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
3568 {
3569         struct ixgbe_hw *hw =
3570                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3571         struct ixgbe_vfta * shadow_vfta =
3572                 IXGBE_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
3573         uint32_t vid_idx = 0;
3574         uint32_t vid_bit = 0;
3575         int ret = 0;
3576
3577         PMD_INIT_FUNC_TRACE();
3578
3579         /* vind is not used in VF driver, set to 0, check ixgbe_set_vfta_vf */
3580         ret = ixgbe_set_vfta(hw, vlan_id, 0, !!on);
3581         if(ret){
3582                 PMD_INIT_LOG(ERR, "Unable to set VF vlan");
3583                 return ret;
3584         }
3585         vid_idx = (uint32_t) ((vlan_id >> 5) & 0x7F);
3586         vid_bit = (uint32_t) (1 << (vlan_id & 0x1F));
3587
3588         /* Save what we set and retore it after device reset */
3589         if (on)
3590                 shadow_vfta->vfta[vid_idx] |= vid_bit;
3591         else
3592                 shadow_vfta->vfta[vid_idx] &= ~vid_bit;
3593
3594         return 0;
3595 }
3596
3597 static void
3598 ixgbevf_vlan_strip_queue_set(struct rte_eth_dev *dev, uint16_t queue, int on)
3599 {
3600         struct ixgbe_hw *hw =
3601                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3602         uint32_t ctrl;
3603
3604         PMD_INIT_FUNC_TRACE();
3605
3606         if(queue >= hw->mac.max_rx_queues)
3607                 return;
3608
3609         ctrl = IXGBE_READ_REG(hw, IXGBE_RXDCTL(queue));
3610         if(on)
3611                 ctrl |= IXGBE_RXDCTL_VME;
3612         else
3613                 ctrl &= ~IXGBE_RXDCTL_VME;
3614         IXGBE_WRITE_REG(hw, IXGBE_RXDCTL(queue), ctrl);
3615
3616         ixgbe_vlan_hw_strip_bitmap_set( dev, queue, on);
3617 }
3618
3619 static void
3620 ixgbevf_vlan_offload_set(struct rte_eth_dev *dev, int mask)
3621 {
3622         struct ixgbe_hw *hw =
3623                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3624         uint16_t i;
3625         int on = 0;
3626
3627         /* VF function only support hw strip feature, others are not support */
3628         if(mask & ETH_VLAN_STRIP_MASK){
3629                 on = !!(dev->data->dev_conf.rxmode.hw_vlan_strip);
3630
3631                 for(i=0; i < hw->mac.max_rx_queues; i++)
3632                         ixgbevf_vlan_strip_queue_set(dev,i,on);
3633         }
3634 }
3635
3636 static int
3637 ixgbe_vmdq_mode_check(struct ixgbe_hw *hw)
3638 {
3639         uint32_t reg_val;
3640
3641         /* we only need to do this if VMDq is enabled */
3642         reg_val = IXGBE_READ_REG(hw, IXGBE_VT_CTL);
3643         if (!(reg_val & IXGBE_VT_CTL_VT_ENABLE)) {
3644                 PMD_INIT_LOG(ERR, "VMDq must be enabled for this setting");
3645                 return (-1);
3646         }
3647
3648         return 0;
3649 }
3650
3651 static uint32_t
3652 ixgbe_uta_vector(struct ixgbe_hw *hw, struct ether_addr* uc_addr)
3653 {
3654         uint32_t vector = 0;
3655         switch (hw->mac.mc_filter_type) {
3656         case 0:   /* use bits [47:36] of the address */
3657                 vector = ((uc_addr->addr_bytes[4] >> 4) |
3658                         (((uint16_t)uc_addr->addr_bytes[5]) << 4));
3659                 break;
3660         case 1:   /* use bits [46:35] of the address */
3661                 vector = ((uc_addr->addr_bytes[4] >> 3) |
3662                         (((uint16_t)uc_addr->addr_bytes[5]) << 5));
3663                 break;
3664         case 2:   /* use bits [45:34] of the address */
3665                 vector = ((uc_addr->addr_bytes[4] >> 2) |
3666                         (((uint16_t)uc_addr->addr_bytes[5]) << 6));
3667                 break;
3668         case 3:   /* use bits [43:32] of the address */
3669                 vector = ((uc_addr->addr_bytes[4]) |
3670                         (((uint16_t)uc_addr->addr_bytes[5]) << 8));
3671                 break;
3672         default:  /* Invalid mc_filter_type */
3673                 break;
3674         }
3675
3676         /* vector can only be 12-bits or boundary will be exceeded */
3677         vector &= 0xFFF;
3678         return vector;
3679 }
3680
3681 static int
3682 ixgbe_uc_hash_table_set(struct rte_eth_dev *dev,struct ether_addr* mac_addr,
3683                                uint8_t on)
3684 {
3685         uint32_t vector;
3686         uint32_t uta_idx;
3687         uint32_t reg_val;
3688         uint32_t uta_shift;
3689         uint32_t rc;
3690         const uint32_t ixgbe_uta_idx_mask = 0x7F;
3691         const uint32_t ixgbe_uta_bit_shift = 5;
3692         const uint32_t ixgbe_uta_bit_mask = (0x1 << ixgbe_uta_bit_shift) - 1;
3693         const uint32_t bit1 = 0x1;
3694
3695         struct ixgbe_hw *hw =
3696                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3697         struct ixgbe_uta_info *uta_info =
3698                 IXGBE_DEV_PRIVATE_TO_UTA(dev->data->dev_private);
3699
3700         /* The UTA table only exists on 82599 hardware and newer */
3701         if (hw->mac.type < ixgbe_mac_82599EB)
3702                 return (-ENOTSUP);
3703
3704         vector = ixgbe_uta_vector(hw,mac_addr);
3705         uta_idx = (vector >> ixgbe_uta_bit_shift) & ixgbe_uta_idx_mask;
3706         uta_shift = vector & ixgbe_uta_bit_mask;
3707
3708         rc = ((uta_info->uta_shadow[uta_idx] >> uta_shift & bit1) != 0);
3709         if(rc == on)
3710                 return 0;
3711
3712         reg_val = IXGBE_READ_REG(hw, IXGBE_UTA(uta_idx));
3713         if (on) {
3714                 uta_info->uta_in_use++;
3715                 reg_val |= (bit1 << uta_shift);
3716                 uta_info->uta_shadow[uta_idx] |= (bit1 << uta_shift);
3717         } else {
3718                 uta_info->uta_in_use--;
3719                 reg_val &= ~(bit1 << uta_shift);
3720                 uta_info->uta_shadow[uta_idx] &= ~(bit1 << uta_shift);
3721         }
3722
3723         IXGBE_WRITE_REG(hw, IXGBE_UTA(uta_idx), reg_val);
3724
3725         if (uta_info->uta_in_use > 0)
3726                 IXGBE_WRITE_REG(hw, IXGBE_MCSTCTRL,
3727                                 IXGBE_MCSTCTRL_MFE | hw->mac.mc_filter_type);
3728         else
3729                 IXGBE_WRITE_REG(hw, IXGBE_MCSTCTRL,hw->mac.mc_filter_type);
3730
3731         return 0;
3732 }
3733
3734 static int
3735 ixgbe_uc_all_hash_table_set(struct rte_eth_dev *dev, uint8_t on)
3736 {
3737         int i;
3738         struct ixgbe_hw *hw =
3739                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3740         struct ixgbe_uta_info *uta_info =
3741                 IXGBE_DEV_PRIVATE_TO_UTA(dev->data->dev_private);
3742
3743         /* The UTA table only exists on 82599 hardware and newer */
3744         if (hw->mac.type < ixgbe_mac_82599EB)
3745                 return (-ENOTSUP);
3746
3747         if(on) {
3748                 for (i = 0; i < ETH_VMDQ_NUM_UC_HASH_ARRAY; i++) {
3749                         uta_info->uta_shadow[i] = ~0;
3750                         IXGBE_WRITE_REG(hw, IXGBE_UTA(i), ~0);
3751                 }
3752         } else {
3753                 for (i = 0; i < ETH_VMDQ_NUM_UC_HASH_ARRAY; i++) {
3754                         uta_info->uta_shadow[i] = 0;
3755                         IXGBE_WRITE_REG(hw, IXGBE_UTA(i), 0);
3756                 }
3757         }
3758         return 0;
3759
3760 }
3761
3762 uint32_t
3763 ixgbe_convert_vm_rx_mask_to_val(uint16_t rx_mask, uint32_t orig_val)
3764 {
3765         uint32_t new_val = orig_val;
3766
3767         if (rx_mask & ETH_VMDQ_ACCEPT_UNTAG)
3768                 new_val |= IXGBE_VMOLR_AUPE;
3769         if (rx_mask & ETH_VMDQ_ACCEPT_HASH_MC)
3770                 new_val |= IXGBE_VMOLR_ROMPE;
3771         if (rx_mask & ETH_VMDQ_ACCEPT_HASH_UC)
3772                 new_val |= IXGBE_VMOLR_ROPE;
3773         if (rx_mask & ETH_VMDQ_ACCEPT_BROADCAST)
3774                 new_val |= IXGBE_VMOLR_BAM;
3775         if (rx_mask & ETH_VMDQ_ACCEPT_MULTICAST)
3776                 new_val |= IXGBE_VMOLR_MPE;
3777
3778         return new_val;
3779 }
3780
3781 static int
3782 ixgbe_set_pool_rx_mode(struct rte_eth_dev *dev, uint16_t pool,
3783                                uint16_t rx_mask, uint8_t on)
3784 {
3785         int val = 0;
3786
3787         struct ixgbe_hw *hw =
3788                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3789         uint32_t vmolr = IXGBE_READ_REG(hw, IXGBE_VMOLR(pool));
3790
3791         if (hw->mac.type == ixgbe_mac_82598EB) {
3792                 PMD_INIT_LOG(ERR, "setting VF receive mode set should be done"
3793                              " on 82599 hardware and newer");
3794                 return (-ENOTSUP);
3795         }
3796         if (ixgbe_vmdq_mode_check(hw) < 0)
3797                 return (-ENOTSUP);
3798
3799         val = ixgbe_convert_vm_rx_mask_to_val(rx_mask, val);
3800
3801         if (on)
3802                 vmolr |= val;
3803         else
3804                 vmolr &= ~val;
3805
3806         IXGBE_WRITE_REG(hw, IXGBE_VMOLR(pool), vmolr);
3807
3808         return 0;
3809 }
3810
3811 static int
3812 ixgbe_set_pool_rx(struct rte_eth_dev *dev, uint16_t pool, uint8_t on)
3813 {
3814         uint32_t reg,addr;
3815         uint32_t val;
3816         const uint8_t bit1 = 0x1;
3817
3818         struct ixgbe_hw *hw =
3819                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3820
3821         if (ixgbe_vmdq_mode_check(hw) < 0)
3822                 return (-ENOTSUP);
3823
3824         addr = IXGBE_VFRE(pool >= ETH_64_POOLS/2);
3825         reg = IXGBE_READ_REG(hw, addr);
3826         val = bit1 << pool;
3827
3828         if (on)
3829                 reg |= val;
3830         else
3831                 reg &= ~val;
3832
3833         IXGBE_WRITE_REG(hw, addr,reg);
3834
3835         return 0;
3836 }
3837
3838 static int
3839 ixgbe_set_pool_tx(struct rte_eth_dev *dev, uint16_t pool, uint8_t on)
3840 {
3841         uint32_t reg,addr;
3842         uint32_t val;
3843         const uint8_t bit1 = 0x1;
3844
3845         struct ixgbe_hw *hw =
3846                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3847
3848         if (ixgbe_vmdq_mode_check(hw) < 0)
3849                 return (-ENOTSUP);
3850
3851         addr = IXGBE_VFTE(pool >= ETH_64_POOLS/2);
3852         reg = IXGBE_READ_REG(hw, addr);
3853         val = bit1 << pool;
3854
3855         if (on)
3856                 reg |= val;
3857         else
3858                 reg &= ~val;
3859
3860         IXGBE_WRITE_REG(hw, addr,reg);
3861
3862         return 0;
3863 }
3864
3865 static int
3866 ixgbe_set_pool_vlan_filter(struct rte_eth_dev *dev, uint16_t vlan,
3867                         uint64_t pool_mask, uint8_t vlan_on)
3868 {
3869         int ret = 0;
3870         uint16_t pool_idx;
3871         struct ixgbe_hw *hw =
3872                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3873
3874         if (ixgbe_vmdq_mode_check(hw) < 0)
3875                 return (-ENOTSUP);
3876         for (pool_idx = 0; pool_idx < ETH_64_POOLS; pool_idx++) {
3877                 if (pool_mask & ((uint64_t)(1ULL << pool_idx)))
3878                         ret = hw->mac.ops.set_vfta(hw,vlan,pool_idx,vlan_on);
3879                         if (ret < 0)
3880                                 return ret;
3881         }
3882
3883         return ret;
3884 }
3885
3886 #define IXGBE_MRCTL_VPME  0x01 /* Virtual Pool Mirroring. */
3887 #define IXGBE_MRCTL_UPME  0x02 /* Uplink Port Mirroring. */
3888 #define IXGBE_MRCTL_DPME  0x04 /* Downlink Port Mirroring. */
3889 #define IXGBE_MRCTL_VLME  0x08 /* VLAN Mirroring. */
3890 #define IXGBE_INVALID_MIRROR_TYPE(mirror_type) \
3891         ((mirror_type) & ~(uint8_t)(ETH_MIRROR_VIRTUAL_POOL_UP | \
3892         ETH_MIRROR_UPLINK_PORT | ETH_MIRROR_DOWNLINK_PORT | ETH_MIRROR_VLAN))
3893
3894 static int
3895 ixgbe_mirror_rule_set(struct rte_eth_dev *dev,
3896                         struct rte_eth_mirror_conf *mirror_conf,
3897                         uint8_t rule_id, uint8_t on)
3898 {
3899         uint32_t mr_ctl,vlvf;
3900         uint32_t mp_lsb = 0;
3901         uint32_t mv_msb = 0;
3902         uint32_t mv_lsb = 0;
3903         uint32_t mp_msb = 0;
3904         uint8_t i = 0;
3905         int reg_index = 0;
3906         uint64_t vlan_mask = 0;
3907
3908         const uint8_t pool_mask_offset = 32;
3909         const uint8_t vlan_mask_offset = 32;
3910         const uint8_t dst_pool_offset = 8;
3911         const uint8_t rule_mr_offset  = 4;
3912         const uint8_t mirror_rule_mask= 0x0F;
3913
3914         struct ixgbe_mirror_info *mr_info =
3915                         (IXGBE_DEV_PRIVATE_TO_PFDATA(dev->data->dev_private));
3916         struct ixgbe_hw *hw =
3917                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
3918         uint8_t mirror_type = 0;
3919
3920         if (ixgbe_vmdq_mode_check(hw) < 0)
3921                 return -ENOTSUP;
3922
3923         if (rule_id >= IXGBE_MAX_MIRROR_RULES)
3924                 return -EINVAL;
3925
3926         if (IXGBE_INVALID_MIRROR_TYPE(mirror_conf->rule_type)) {
3927                 PMD_DRV_LOG(ERR, "unsupported mirror type 0x%x.",
3928                         mirror_conf->rule_type);
3929                 return -EINVAL;
3930         }
3931
3932         if (mirror_conf->rule_type & ETH_MIRROR_VLAN) {
3933                 mirror_type |= IXGBE_MRCTL_VLME;
3934                 /* Check if vlan id is valid and find conresponding VLAN ID index in VLVF */
3935                 for (i = 0;i < IXGBE_VLVF_ENTRIES; i++) {
3936                         if (mirror_conf->vlan.vlan_mask & (1ULL << i)) {
3937                                 /* search vlan id related pool vlan filter index */
3938                                 reg_index = ixgbe_find_vlvf_slot(hw,
3939                                                 mirror_conf->vlan.vlan_id[i]);
3940                                 if(reg_index < 0)
3941                                         return -EINVAL;
3942                                 vlvf = IXGBE_READ_REG(hw, IXGBE_VLVF(reg_index));
3943                                 if ((vlvf & IXGBE_VLVF_VIEN) &&
3944                                     ((vlvf & IXGBE_VLVF_VLANID_MASK) ==
3945                                       mirror_conf->vlan.vlan_id[i]))
3946                                         vlan_mask |= (1ULL << reg_index);
3947                                 else
3948                                         return -EINVAL;
3949                         }
3950                 }
3951
3952                 if (on) {
3953                         mv_lsb = vlan_mask & 0xFFFFFFFF;
3954                         mv_msb = vlan_mask >> vlan_mask_offset;
3955
3956                         mr_info->mr_conf[rule_id].vlan.vlan_mask =
3957                                                 mirror_conf->vlan.vlan_mask;
3958                         for(i = 0 ;i < ETH_VMDQ_MAX_VLAN_FILTERS; i++) {
3959                                 if(mirror_conf->vlan.vlan_mask & (1ULL << i))
3960                                         mr_info->mr_conf[rule_id].vlan.vlan_id[i] =
3961                                                 mirror_conf->vlan.vlan_id[i];
3962                         }
3963                 } else {
3964                         mv_lsb = 0;
3965                         mv_msb = 0;
3966                         mr_info->mr_conf[rule_id].vlan.vlan_mask = 0;
3967                         for(i = 0 ;i < ETH_VMDQ_MAX_VLAN_FILTERS; i++)
3968                                 mr_info->mr_conf[rule_id].vlan.vlan_id[i] = 0;
3969                 }
3970         }
3971
3972         /*
3973          * if enable pool mirror, write related pool mask register,if disable
3974          * pool mirror, clear PFMRVM register
3975          */
3976         if (mirror_conf->rule_type & ETH_MIRROR_VIRTUAL_POOL_UP) {
3977                 mirror_type |= IXGBE_MRCTL_VPME;
3978                 if (on) {
3979                         mp_lsb = mirror_conf->pool_mask & 0xFFFFFFFF;
3980                         mp_msb = mirror_conf->pool_mask >> pool_mask_offset;
3981                         mr_info->mr_conf[rule_id].pool_mask =
3982                                         mirror_conf->pool_mask;
3983
3984                 } else {
3985                         mp_lsb = 0;
3986                         mp_msb = 0;
3987                         mr_info->mr_conf[rule_id].pool_mask = 0;
3988                 }
3989         }
3990         if (mirror_conf->rule_type & ETH_MIRROR_UPLINK_PORT)
3991                 mirror_type |= IXGBE_MRCTL_UPME;
3992         if (mirror_conf->rule_type & ETH_MIRROR_DOWNLINK_PORT)
3993                 mirror_type |= IXGBE_MRCTL_DPME;
3994
3995         /* read  mirror control register and recalculate it */
3996         mr_ctl = IXGBE_READ_REG(hw, IXGBE_MRCTL(rule_id));
3997
3998         if (on) {
3999                 mr_ctl |= mirror_type;
4000                 mr_ctl &= mirror_rule_mask;
4001                 mr_ctl |= mirror_conf->dst_pool << dst_pool_offset;
4002         } else
4003                 mr_ctl &= ~(mirror_conf->rule_type & mirror_rule_mask);
4004
4005         mr_info->mr_conf[rule_id].rule_type = mirror_conf->rule_type;
4006         mr_info->mr_conf[rule_id].dst_pool = mirror_conf->dst_pool;
4007
4008         /* write mirrror control  register */
4009         IXGBE_WRITE_REG(hw, IXGBE_MRCTL(rule_id), mr_ctl);
4010
4011         /* write pool mirrror control  register */
4012         if (mirror_conf->rule_type == ETH_MIRROR_VIRTUAL_POOL_UP) {
4013                 IXGBE_WRITE_REG(hw, IXGBE_VMRVM(rule_id), mp_lsb);
4014                 IXGBE_WRITE_REG(hw, IXGBE_VMRVM(rule_id + rule_mr_offset),
4015                                 mp_msb);
4016         }
4017         /* write VLAN mirrror control  register */
4018         if (mirror_conf->rule_type == ETH_MIRROR_VLAN) {
4019                 IXGBE_WRITE_REG(hw, IXGBE_VMRVLAN(rule_id), mv_lsb);
4020                 IXGBE_WRITE_REG(hw, IXGBE_VMRVLAN(rule_id + rule_mr_offset),
4021                                 mv_msb);
4022         }
4023
4024         return 0;
4025 }
4026
4027 static int
4028 ixgbe_mirror_rule_reset(struct rte_eth_dev *dev, uint8_t rule_id)
4029 {
4030         int mr_ctl = 0;
4031         uint32_t lsb_val = 0;
4032         uint32_t msb_val = 0;
4033         const uint8_t rule_mr_offset = 4;
4034
4035         struct ixgbe_hw *hw =
4036                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4037         struct ixgbe_mirror_info *mr_info =
4038                 (IXGBE_DEV_PRIVATE_TO_PFDATA(dev->data->dev_private));
4039
4040         if (ixgbe_vmdq_mode_check(hw) < 0)
4041                 return (-ENOTSUP);
4042
4043         memset(&mr_info->mr_conf[rule_id], 0,
4044                 sizeof(struct rte_eth_mirror_conf));
4045
4046         /* clear PFVMCTL register */
4047         IXGBE_WRITE_REG(hw, IXGBE_MRCTL(rule_id), mr_ctl);
4048
4049         /* clear pool mask register */
4050         IXGBE_WRITE_REG(hw, IXGBE_VMRVM(rule_id), lsb_val);
4051         IXGBE_WRITE_REG(hw, IXGBE_VMRVM(rule_id + rule_mr_offset), msb_val);
4052
4053         /* clear vlan mask register */
4054         IXGBE_WRITE_REG(hw, IXGBE_VMRVLAN(rule_id), lsb_val);
4055         IXGBE_WRITE_REG(hw, IXGBE_VMRVLAN(rule_id + rule_mr_offset), msb_val);
4056
4057         return 0;
4058 }
4059
4060 static int
4061 ixgbevf_dev_rx_queue_intr_enable(struct rte_eth_dev *dev, uint16_t queue_id)
4062 {
4063         uint32_t mask;
4064         struct ixgbe_hw *hw =
4065                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4066
4067         mask = IXGBE_READ_REG(hw, IXGBE_VTEIMS);
4068         mask |= (1 << queue_id);
4069         IXGBE_WRITE_REG(hw, IXGBE_VTEIMS, mask);
4070
4071         rte_intr_enable(&dev->pci_dev->intr_handle);
4072
4073         return 0;
4074 }
4075
4076 static int
4077 ixgbevf_dev_rx_queue_intr_disable(struct rte_eth_dev *dev, uint16_t queue_id)
4078 {
4079         uint32_t mask;
4080         struct ixgbe_hw *hw =
4081                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4082
4083         mask = IXGBE_READ_REG(hw, IXGBE_VTEIMS);
4084         mask &= ~(1 << queue_id);
4085         IXGBE_WRITE_REG(hw, IXGBE_VTEIMS, mask);
4086
4087         return 0;
4088 }
4089
4090 static int
4091 ixgbe_dev_rx_queue_intr_enable(struct rte_eth_dev *dev, uint16_t queue_id)
4092 {
4093         uint32_t mask;
4094         struct ixgbe_hw *hw =
4095                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4096         struct ixgbe_interrupt *intr =
4097                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4098
4099         if (queue_id < 16) {
4100                 ixgbe_disable_intr(hw);
4101                 intr->mask |= (1 << queue_id);
4102                 ixgbe_enable_intr(dev);
4103         } else if (queue_id < 32) {
4104                 mask = IXGBE_READ_REG(hw, IXGBE_EIMS_EX(0));
4105                 mask &= (1 << queue_id);
4106                 IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(0), mask);
4107         } else if (queue_id < 64) {
4108                 mask = IXGBE_READ_REG(hw, IXGBE_EIMS_EX(1));
4109                 mask &= (1 << (queue_id - 32));
4110                 IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(1), mask);
4111         }
4112         rte_intr_enable(&dev->pci_dev->intr_handle);
4113
4114         return 0;
4115 }
4116
4117 static int
4118 ixgbe_dev_rx_queue_intr_disable(struct rte_eth_dev *dev, uint16_t queue_id)
4119 {
4120         uint32_t mask;
4121         struct ixgbe_hw *hw =
4122                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4123         struct ixgbe_interrupt *intr =
4124                 IXGBE_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
4125
4126         if (queue_id < 16) {
4127                 ixgbe_disable_intr(hw);
4128                 intr->mask &= ~(1 << queue_id);
4129                 ixgbe_enable_intr(dev);
4130         } else if (queue_id < 32) {
4131                 mask = IXGBE_READ_REG(hw, IXGBE_EIMS_EX(0));
4132                 mask &= ~(1 << queue_id);
4133                 IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(0), mask);
4134         } else if (queue_id < 64) {
4135                 mask = IXGBE_READ_REG(hw, IXGBE_EIMS_EX(1));
4136                 mask &= ~(1 << (queue_id - 32));
4137                 IXGBE_WRITE_REG(hw, IXGBE_EIMS_EX(1), mask);
4138         }
4139
4140         return 0;
4141 }
4142
4143 static void
4144 ixgbevf_set_ivar_map(struct ixgbe_hw *hw, int8_t direction,
4145                      uint8_t queue, uint8_t msix_vector)
4146 {
4147         uint32_t tmp, idx;
4148
4149         if (direction == -1) {
4150                 /* other causes */
4151                 msix_vector |= IXGBE_IVAR_ALLOC_VAL;
4152                 tmp = IXGBE_READ_REG(hw, IXGBE_VTIVAR_MISC);
4153                 tmp &= ~0xFF;
4154                 tmp |= msix_vector;
4155                 IXGBE_WRITE_REG(hw, IXGBE_VTIVAR_MISC, tmp);
4156         } else {
4157                 /* rx or tx cause */
4158                 msix_vector |= IXGBE_IVAR_ALLOC_VAL;
4159                 idx = ((16 * (queue & 1)) + (8 * direction));
4160                 tmp = IXGBE_READ_REG(hw, IXGBE_VTIVAR(queue >> 1));
4161                 tmp &= ~(0xFF << idx);
4162                 tmp |= (msix_vector << idx);
4163                 IXGBE_WRITE_REG(hw, IXGBE_VTIVAR(queue >> 1), tmp);
4164         }
4165 }
4166
4167 /**
4168  * set the IVAR registers, mapping interrupt causes to vectors
4169  * @param hw
4170  *  pointer to ixgbe_hw struct
4171  * @direction
4172  *  0 for Rx, 1 for Tx, -1 for other causes
4173  * @queue
4174  *  queue to map the corresponding interrupt to
4175  * @msix_vector
4176  *  the vector to map to the corresponding queue
4177  */
4178 static void
4179 ixgbe_set_ivar_map(struct ixgbe_hw *hw, int8_t direction,
4180                    uint8_t queue, uint8_t msix_vector)
4181 {
4182         uint32_t tmp, idx;
4183
4184         msix_vector |= IXGBE_IVAR_ALLOC_VAL;
4185         if (hw->mac.type == ixgbe_mac_82598EB) {
4186                 if (direction == -1)
4187                         direction = 0;
4188                 idx = (((direction * 64) + queue) >> 2) & 0x1F;
4189                 tmp = IXGBE_READ_REG(hw, IXGBE_IVAR(idx));
4190                 tmp &= ~(0xFF << (8 * (queue & 0x3)));
4191                 tmp |= (msix_vector << (8 * (queue & 0x3)));
4192                 IXGBE_WRITE_REG(hw, IXGBE_IVAR(idx), tmp);
4193         } else if ((hw->mac.type == ixgbe_mac_82599EB) ||
4194                         (hw->mac.type == ixgbe_mac_X540)) {
4195                 if (direction == -1) {
4196                         /* other causes */
4197                         idx = ((queue & 1) * 8);
4198                         tmp = IXGBE_READ_REG(hw, IXGBE_IVAR_MISC);
4199                         tmp &= ~(0xFF << idx);
4200                         tmp |= (msix_vector << idx);
4201                         IXGBE_WRITE_REG(hw, IXGBE_IVAR_MISC, tmp);
4202                 } else {
4203                         /* rx or tx causes */
4204                         idx = ((16 * (queue & 1)) + (8 * direction));
4205                         tmp = IXGBE_READ_REG(hw, IXGBE_IVAR(queue >> 1));
4206                         tmp &= ~(0xFF << idx);
4207                         tmp |= (msix_vector << idx);
4208                         IXGBE_WRITE_REG(hw, IXGBE_IVAR(queue >> 1), tmp);
4209                 }
4210         }
4211 }
4212
4213 static void
4214 ixgbevf_configure_msix(struct rte_eth_dev *dev)
4215 {
4216         struct rte_intr_handle *intr_handle = &dev->pci_dev->intr_handle;
4217         struct ixgbe_hw *hw =
4218                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4219         uint32_t q_idx;
4220         uint32_t vector_idx = 0;
4221
4222         /* won't configure msix register if no mapping is done
4223          * between intr vector and event fd.
4224          */
4225         if (!rte_intr_dp_is_en(intr_handle))
4226                 return;
4227
4228         /* Configure all RX queues of VF */
4229         for (q_idx = 0; q_idx < dev->data->nb_rx_queues; q_idx++) {
4230                 /* Force all queue use vector 0,
4231                  * as IXGBE_VF_MAXMSIVECOTR = 1
4232                  */
4233                 ixgbevf_set_ivar_map(hw, 0, q_idx, vector_idx);
4234                 intr_handle->intr_vec[q_idx] = vector_idx;
4235         }
4236
4237         /* Configure VF Rx queue ivar */
4238         ixgbevf_set_ivar_map(hw, -1, 1, vector_idx);
4239 }
4240
4241 /**
4242  * Sets up the hardware to properly generate MSI-X interrupts
4243  * @hw
4244  *  board private structure
4245  */
4246 static void
4247 ixgbe_configure_msix(struct rte_eth_dev *dev)
4248 {
4249         struct rte_intr_handle *intr_handle = &dev->pci_dev->intr_handle;
4250         struct ixgbe_hw *hw =
4251                 IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4252         uint32_t queue_id, vec = 0;
4253         uint32_t mask;
4254         uint32_t gpie;
4255
4256         /* won't configure msix register if no mapping is done
4257          * between intr vector and event fd
4258          */
4259         if (!rte_intr_dp_is_en(intr_handle))
4260                 return;
4261
4262         /* setup GPIE for MSI-x mode */
4263         gpie = IXGBE_READ_REG(hw, IXGBE_GPIE);
4264         gpie |= IXGBE_GPIE_MSIX_MODE | IXGBE_GPIE_PBA_SUPPORT |
4265                 IXGBE_GPIE_OCD | IXGBE_GPIE_EIAME;
4266         /* auto clearing and auto setting corresponding bits in EIMS
4267          * when MSI-X interrupt is triggered
4268          */
4269         if (hw->mac.type == ixgbe_mac_82598EB) {
4270                 IXGBE_WRITE_REG(hw, IXGBE_EIAM, IXGBE_EICS_RTX_QUEUE);
4271         } else {
4272                 IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(0), 0xFFFFFFFF);
4273                 IXGBE_WRITE_REG(hw, IXGBE_EIAM_EX(1), 0xFFFFFFFF);
4274         }
4275         IXGBE_WRITE_REG(hw, IXGBE_GPIE, gpie);
4276
4277         /* Populate the IVAR table and set the ITR values to the
4278          * corresponding register.
4279          */
4280         for (queue_id = 0; queue_id < dev->data->nb_rx_queues;
4281              queue_id++) {
4282                 /* by default, 1:1 mapping */
4283                 ixgbe_set_ivar_map(hw, 0, queue_id, vec);
4284                 intr_handle->intr_vec[queue_id] = vec;
4285                 if (vec < intr_handle->nb_efd - 1)
4286                         vec++;
4287         }
4288
4289         switch (hw->mac.type) {
4290         case ixgbe_mac_82598EB:
4291                 ixgbe_set_ivar_map(hw, -1, IXGBE_IVAR_OTHER_CAUSES_INDEX,
4292                                    intr_handle->max_intr - 1);
4293                 break;
4294         case ixgbe_mac_82599EB:
4295         case ixgbe_mac_X540:
4296                 ixgbe_set_ivar_map(hw, -1, 1, intr_handle->max_intr - 1);
4297                 break;
4298         default:
4299                 break;
4300         }
4301         IXGBE_WRITE_REG(hw, IXGBE_EITR(queue_id),
4302                         IXGBE_MIN_INTER_INTERRUPT_INTERVAL_DEFAULT & 0xFFF);
4303
4304         /* set up to autoclear timer, and the vectors */
4305         mask = IXGBE_EIMS_ENABLE_MASK;
4306         mask &= ~(IXGBE_EIMS_OTHER |
4307                   IXGBE_EIMS_MAILBOX |
4308                   IXGBE_EIMS_LSC);
4309
4310         IXGBE_WRITE_REG(hw, IXGBE_EIAC, mask);
4311 }
4312
4313 static int ixgbe_set_queue_rate_limit(struct rte_eth_dev *dev,
4314         uint16_t queue_idx, uint16_t tx_rate)
4315 {
4316         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4317         uint32_t rf_dec, rf_int;
4318         uint32_t bcnrc_val;
4319         uint16_t link_speed = dev->data->dev_link.link_speed;
4320
4321         if (queue_idx >= hw->mac.max_tx_queues)
4322                 return -EINVAL;
4323
4324         if (tx_rate != 0) {
4325                 /* Calculate the rate factor values to set */
4326                 rf_int = (uint32_t)link_speed / (uint32_t)tx_rate;
4327                 rf_dec = (uint32_t)link_speed % (uint32_t)tx_rate;
4328                 rf_dec = (rf_dec << IXGBE_RTTBCNRC_RF_INT_SHIFT) / tx_rate;
4329
4330                 bcnrc_val = IXGBE_RTTBCNRC_RS_ENA;
4331                 bcnrc_val |= ((rf_int << IXGBE_RTTBCNRC_RF_INT_SHIFT) &
4332                                 IXGBE_RTTBCNRC_RF_INT_MASK_M);
4333                 bcnrc_val |= (rf_dec & IXGBE_RTTBCNRC_RF_DEC_MASK);
4334         } else {
4335                 bcnrc_val = 0;
4336         }
4337
4338         /*
4339          * Set global transmit compensation time to the MMW_SIZE in RTTBCNRM
4340          * register. MMW_SIZE=0x014 if 9728-byte jumbo is supported, otherwise
4341          * set as 0x4.
4342          */
4343         if ((dev->data->dev_conf.rxmode.jumbo_frame == 1) &&
4344                 (dev->data->dev_conf.rxmode.max_rx_pkt_len >=
4345                                 IXGBE_MAX_JUMBO_FRAME_SIZE))
4346                 IXGBE_WRITE_REG(hw, IXGBE_RTTBCNRM,
4347                         IXGBE_MMW_SIZE_JUMBO_FRAME);
4348         else
4349                 IXGBE_WRITE_REG(hw, IXGBE_RTTBCNRM,
4350                         IXGBE_MMW_SIZE_DEFAULT);
4351
4352         /* Set RTTBCNRC of queue X */
4353         IXGBE_WRITE_REG(hw, IXGBE_RTTDQSEL, queue_idx);
4354         IXGBE_WRITE_REG(hw, IXGBE_RTTBCNRC, bcnrc_val);
4355         IXGBE_WRITE_FLUSH(hw);
4356
4357         return 0;
4358 }
4359
4360 static int ixgbe_set_vf_rate_limit(struct rte_eth_dev *dev, uint16_t vf,
4361         uint16_t tx_rate, uint64_t q_msk)
4362 {
4363         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4364         struct ixgbe_vf_info *vfinfo =
4365                 *(IXGBE_DEV_PRIVATE_TO_P_VFDATA(dev->data->dev_private));
4366         uint8_t  nb_q_per_pool = RTE_ETH_DEV_SRIOV(dev).nb_q_per_pool;
4367         uint32_t queue_stride =
4368                 IXGBE_MAX_RX_QUEUE_NUM / RTE_ETH_DEV_SRIOV(dev).active;
4369         uint32_t queue_idx = vf * queue_stride, idx = 0, vf_idx;
4370         uint32_t queue_end = queue_idx + nb_q_per_pool - 1;
4371         uint16_t total_rate = 0;
4372
4373         if (queue_end >= hw->mac.max_tx_queues)
4374                 return -EINVAL;
4375
4376         if (vfinfo != NULL) {
4377                 for (vf_idx = 0; vf_idx < dev->pci_dev->max_vfs; vf_idx++) {
4378                         if (vf_idx == vf)
4379                                 continue;
4380                         for (idx = 0; idx < RTE_DIM(vfinfo[vf_idx].tx_rate);
4381                                 idx++)
4382                                 total_rate += vfinfo[vf_idx].tx_rate[idx];
4383                 }
4384         } else
4385                 return -EINVAL;
4386
4387         /* Store tx_rate for this vf. */
4388         for (idx = 0; idx < nb_q_per_pool; idx++) {
4389                 if (((uint64_t)0x1 << idx) & q_msk) {
4390                         if (vfinfo[vf].tx_rate[idx] != tx_rate)
4391                                 vfinfo[vf].tx_rate[idx] = tx_rate;
4392                         total_rate += tx_rate;
4393                 }
4394         }
4395
4396         if (total_rate > dev->data->dev_link.link_speed) {
4397                 /*
4398                  * Reset stored TX rate of the VF if it causes exceed
4399                  * link speed.
4400                  */
4401                 memset(vfinfo[vf].tx_rate, 0, sizeof(vfinfo[vf].tx_rate));
4402                 return -EINVAL;
4403         }
4404
4405         /* Set RTTBCNRC of each queue/pool for vf X  */
4406         for (; queue_idx <= queue_end; queue_idx++) {
4407                 if (0x1 & q_msk)
4408                         ixgbe_set_queue_rate_limit(dev, queue_idx, tx_rate);
4409                 q_msk = q_msk >> 1;
4410         }
4411
4412         return 0;
4413 }
4414
4415 static void
4416 ixgbevf_add_mac_addr(struct rte_eth_dev *dev, struct ether_addr *mac_addr,
4417                      __attribute__((unused)) uint32_t index,
4418                      __attribute__((unused)) uint32_t pool)
4419 {
4420         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4421         int diag;
4422
4423         /*
4424          * On a 82599 VF, adding again the same MAC addr is not an idempotent
4425          * operation. Trap this case to avoid exhausting the [very limited]
4426          * set of PF resources used to store VF MAC addresses.
4427          */
4428         if (memcmp(hw->mac.perm_addr, mac_addr, sizeof(struct ether_addr)) == 0)
4429                 return;
4430         diag = ixgbevf_set_uc_addr_vf(hw, 2, mac_addr->addr_bytes);
4431         if (diag == 0)
4432                 return;
4433         PMD_DRV_LOG(ERR, "Unable to add MAC address - diag=%d", diag);
4434 }
4435
4436 static void
4437 ixgbevf_remove_mac_addr(struct rte_eth_dev *dev, uint32_t index)
4438 {
4439         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4440         struct ether_addr *perm_addr = (struct ether_addr *) hw->mac.perm_addr;
4441         struct ether_addr *mac_addr;
4442         uint32_t i;
4443         int diag;
4444
4445         /*
4446          * The IXGBE_VF_SET_MACVLAN command of the ixgbe-pf driver does
4447          * not support the deletion of a given MAC address.
4448          * Instead, it imposes to delete all MAC addresses, then to add again
4449          * all MAC addresses with the exception of the one to be deleted.
4450          */
4451         (void) ixgbevf_set_uc_addr_vf(hw, 0, NULL);
4452
4453         /*
4454          * Add again all MAC addresses, with the exception of the deleted one
4455          * and of the permanent MAC address.
4456          */
4457         for (i = 0, mac_addr = dev->data->mac_addrs;
4458              i < hw->mac.num_rar_entries; i++, mac_addr++) {
4459                 /* Skip the deleted MAC address */
4460                 if (i == index)
4461                         continue;
4462                 /* Skip NULL MAC addresses */
4463                 if (is_zero_ether_addr(mac_addr))
4464                         continue;
4465                 /* Skip the permanent MAC address */
4466                 if (memcmp(perm_addr, mac_addr, sizeof(struct ether_addr)) == 0)
4467                         continue;
4468                 diag = ixgbevf_set_uc_addr_vf(hw, 2, mac_addr->addr_bytes);
4469                 if (diag != 0)
4470                         PMD_DRV_LOG(ERR,
4471                                     "Adding again MAC address "
4472                                     "%02x:%02x:%02x:%02x:%02x:%02x failed "
4473                                     "diag=%d",
4474                                     mac_addr->addr_bytes[0],
4475                                     mac_addr->addr_bytes[1],
4476                                     mac_addr->addr_bytes[2],
4477                                     mac_addr->addr_bytes[3],
4478                                     mac_addr->addr_bytes[4],
4479                                     mac_addr->addr_bytes[5],
4480                                     diag);
4481         }
4482 }
4483
4484 static void
4485 ixgbevf_set_default_mac_addr(struct rte_eth_dev *dev, struct ether_addr *addr)
4486 {
4487         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4488
4489         hw->mac.ops.set_rar(hw, 0, (void *)addr, 0, 0);
4490 }
4491
4492 #define MAC_TYPE_FILTER_SUP(type)    do {\
4493         if ((type) != ixgbe_mac_82599EB && (type) != ixgbe_mac_X540 &&\
4494                 (type) != ixgbe_mac_X550)\
4495                 return -ENOTSUP;\
4496 } while (0)
4497
4498 static int
4499 ixgbe_syn_filter_set(struct rte_eth_dev *dev,
4500                         struct rte_eth_syn_filter *filter,
4501                         bool add)
4502 {
4503         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4504         uint32_t synqf;
4505
4506         if (filter->queue >= IXGBE_MAX_RX_QUEUE_NUM)
4507                 return -EINVAL;
4508
4509         synqf = IXGBE_READ_REG(hw, IXGBE_SYNQF);
4510
4511         if (add) {
4512                 if (synqf & IXGBE_SYN_FILTER_ENABLE)
4513                         return -EINVAL;
4514                 synqf = (uint32_t)(((filter->queue << IXGBE_SYN_FILTER_QUEUE_SHIFT) &
4515                         IXGBE_SYN_FILTER_QUEUE) | IXGBE_SYN_FILTER_ENABLE);
4516
4517                 if (filter->hig_pri)
4518                         synqf |= IXGBE_SYN_FILTER_SYNQFP;
4519                 else
4520                         synqf &= ~IXGBE_SYN_FILTER_SYNQFP;
4521         } else {
4522                 if (!(synqf & IXGBE_SYN_FILTER_ENABLE))
4523                         return -ENOENT;
4524                 synqf &= ~(IXGBE_SYN_FILTER_QUEUE | IXGBE_SYN_FILTER_ENABLE);
4525         }
4526         IXGBE_WRITE_REG(hw, IXGBE_SYNQF, synqf);
4527         IXGBE_WRITE_FLUSH(hw);
4528         return 0;
4529 }
4530
4531 static int
4532 ixgbe_syn_filter_get(struct rte_eth_dev *dev,
4533                         struct rte_eth_syn_filter *filter)
4534 {
4535         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4536         uint32_t synqf = IXGBE_READ_REG(hw, IXGBE_SYNQF);
4537
4538         if (synqf & IXGBE_SYN_FILTER_ENABLE) {
4539                 filter->hig_pri = (synqf & IXGBE_SYN_FILTER_SYNQFP) ? 1 : 0;
4540                 filter->queue = (uint16_t)((synqf & IXGBE_SYN_FILTER_QUEUE) >> 1);
4541                 return 0;
4542         }
4543         return -ENOENT;
4544 }
4545
4546 static int
4547 ixgbe_syn_filter_handle(struct rte_eth_dev *dev,
4548                         enum rte_filter_op filter_op,
4549                         void *arg)
4550 {
4551         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4552         int ret;
4553
4554         MAC_TYPE_FILTER_SUP(hw->mac.type);
4555
4556         if (filter_op == RTE_ETH_FILTER_NOP)
4557                 return 0;
4558
4559         if (arg == NULL) {
4560                 PMD_DRV_LOG(ERR, "arg shouldn't be NULL for operation %u",
4561                             filter_op);
4562                 return -EINVAL;
4563         }
4564
4565         switch (filter_op) {
4566         case RTE_ETH_FILTER_ADD:
4567                 ret = ixgbe_syn_filter_set(dev,
4568                                 (struct rte_eth_syn_filter *)arg,
4569                                 TRUE);
4570                 break;
4571         case RTE_ETH_FILTER_DELETE:
4572                 ret = ixgbe_syn_filter_set(dev,
4573                                 (struct rte_eth_syn_filter *)arg,
4574                                 FALSE);
4575                 break;
4576         case RTE_ETH_FILTER_GET:
4577                 ret = ixgbe_syn_filter_get(dev,
4578                                 (struct rte_eth_syn_filter *)arg);
4579                 break;
4580         default:
4581                 PMD_DRV_LOG(ERR, "unsupported operation %u\n", filter_op);
4582                 ret = -EINVAL;
4583                 break;
4584         }
4585
4586         return ret;
4587 }
4588
4589
4590 static inline enum ixgbe_5tuple_protocol
4591 convert_protocol_type(uint8_t protocol_value)
4592 {
4593         if (protocol_value == IPPROTO_TCP)
4594                 return IXGBE_FILTER_PROTOCOL_TCP;
4595         else if (protocol_value == IPPROTO_UDP)
4596                 return IXGBE_FILTER_PROTOCOL_UDP;
4597         else if (protocol_value == IPPROTO_SCTP)
4598                 return IXGBE_FILTER_PROTOCOL_SCTP;
4599         else
4600                 return IXGBE_FILTER_PROTOCOL_NONE;
4601 }
4602
4603 /*
4604  * add a 5tuple filter
4605  *
4606  * @param
4607  * dev: Pointer to struct rte_eth_dev.
4608  * index: the index the filter allocates.
4609  * filter: ponter to the filter that will be added.
4610  * rx_queue: the queue id the filter assigned to.
4611  *
4612  * @return
4613  *    - On success, zero.
4614  *    - On failure, a negative value.
4615  */
4616 static int
4617 ixgbe_add_5tuple_filter(struct rte_eth_dev *dev,
4618                         struct ixgbe_5tuple_filter *filter)
4619 {
4620         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4621         struct ixgbe_filter_info *filter_info =
4622                 IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
4623         int i, idx, shift;
4624         uint32_t ftqf, sdpqf;
4625         uint32_t l34timir = 0;
4626         uint8_t mask = 0xff;
4627
4628         /*
4629          * look for an unused 5tuple filter index,
4630          * and insert the filter to list.
4631          */
4632         for (i = 0; i < IXGBE_MAX_FTQF_FILTERS; i++) {
4633                 idx = i / (sizeof(uint32_t) * NBBY);
4634                 shift = i % (sizeof(uint32_t) * NBBY);
4635                 if (!(filter_info->fivetuple_mask[idx] & (1 << shift))) {
4636                         filter_info->fivetuple_mask[idx] |= 1 << shift;
4637                         filter->index = i;
4638                         TAILQ_INSERT_TAIL(&filter_info->fivetuple_list,
4639                                           filter,
4640                                           entries);
4641                         break;
4642                 }
4643         }
4644         if (i >= IXGBE_MAX_FTQF_FILTERS) {
4645                 PMD_DRV_LOG(ERR, "5tuple filters are full.");
4646                 return -ENOSYS;
4647         }
4648
4649         sdpqf = (uint32_t)(filter->filter_info.dst_port <<
4650                                 IXGBE_SDPQF_DSTPORT_SHIFT);
4651         sdpqf = sdpqf | (filter->filter_info.src_port & IXGBE_SDPQF_SRCPORT);
4652
4653         ftqf = (uint32_t)(filter->filter_info.proto &
4654                 IXGBE_FTQF_PROTOCOL_MASK);
4655         ftqf |= (uint32_t)((filter->filter_info.priority &
4656                 IXGBE_FTQF_PRIORITY_MASK) << IXGBE_FTQF_PRIORITY_SHIFT);
4657         if (filter->filter_info.src_ip_mask == 0) /* 0 means compare. */
4658                 mask &= IXGBE_FTQF_SOURCE_ADDR_MASK;
4659         if (filter->filter_info.dst_ip_mask == 0)
4660                 mask &= IXGBE_FTQF_DEST_ADDR_MASK;
4661         if (filter->filter_info.src_port_mask == 0)
4662                 mask &= IXGBE_FTQF_SOURCE_PORT_MASK;
4663         if (filter->filter_info.dst_port_mask == 0)
4664                 mask &= IXGBE_FTQF_DEST_PORT_MASK;
4665         if (filter->filter_info.proto_mask == 0)
4666                 mask &= IXGBE_FTQF_PROTOCOL_COMP_MASK;
4667         ftqf |= mask << IXGBE_FTQF_5TUPLE_MASK_SHIFT;
4668         ftqf |= IXGBE_FTQF_POOL_MASK_EN;
4669         ftqf |= IXGBE_FTQF_QUEUE_ENABLE;
4670
4671         IXGBE_WRITE_REG(hw, IXGBE_DAQF(i), filter->filter_info.dst_ip);
4672         IXGBE_WRITE_REG(hw, IXGBE_SAQF(i), filter->filter_info.src_ip);
4673         IXGBE_WRITE_REG(hw, IXGBE_SDPQF(i), sdpqf);
4674         IXGBE_WRITE_REG(hw, IXGBE_FTQF(i), ftqf);
4675
4676         l34timir |= IXGBE_L34T_IMIR_RESERVE;
4677         l34timir |= (uint32_t)(filter->queue <<
4678                                 IXGBE_L34T_IMIR_QUEUE_SHIFT);
4679         IXGBE_WRITE_REG(hw, IXGBE_L34T_IMIR(i), l34timir);
4680         return 0;
4681 }
4682
4683 /*
4684  * remove a 5tuple filter
4685  *
4686  * @param
4687  * dev: Pointer to struct rte_eth_dev.
4688  * filter: the pointer of the filter will be removed.
4689  */
4690 static void
4691 ixgbe_remove_5tuple_filter(struct rte_eth_dev *dev,
4692                         struct ixgbe_5tuple_filter *filter)
4693 {
4694         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4695         struct ixgbe_filter_info *filter_info =
4696                 IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
4697         uint16_t index = filter->index;
4698
4699         filter_info->fivetuple_mask[index / (sizeof(uint32_t) * NBBY)] &=
4700                                 ~(1 << (index % (sizeof(uint32_t) * NBBY)));
4701         TAILQ_REMOVE(&filter_info->fivetuple_list, filter, entries);
4702         rte_free(filter);
4703
4704         IXGBE_WRITE_REG(hw, IXGBE_DAQF(index), 0);
4705         IXGBE_WRITE_REG(hw, IXGBE_SAQF(index), 0);
4706         IXGBE_WRITE_REG(hw, IXGBE_SDPQF(index), 0);
4707         IXGBE_WRITE_REG(hw, IXGBE_FTQF(index), 0);
4708         IXGBE_WRITE_REG(hw, IXGBE_L34T_IMIR(index), 0);
4709 }
4710
4711 static int
4712 ixgbevf_dev_set_mtu(struct rte_eth_dev *dev, uint16_t mtu)
4713 {
4714         struct ixgbe_hw *hw;
4715         uint32_t max_frame = mtu + ETHER_HDR_LEN + ETHER_CRC_LEN;
4716
4717         hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4718
4719         if ((mtu < ETHER_MIN_MTU) || (max_frame > ETHER_MAX_JUMBO_FRAME_LEN))
4720                 return -EINVAL;
4721
4722         /* refuse mtu that requires the support of scattered packets when this
4723          * feature has not been enabled before. */
4724         if (!dev->data->scattered_rx &&
4725             (max_frame + 2 * IXGBE_VLAN_TAG_SIZE >
4726              dev->data->min_rx_buf_size - RTE_PKTMBUF_HEADROOM))
4727                 return -EINVAL;
4728
4729         /*
4730          * When supported by the underlying PF driver, use the IXGBE_VF_SET_MTU
4731          * request of the version 2.0 of the mailbox API.
4732          * For now, use the IXGBE_VF_SET_LPE request of the version 1.0
4733          * of the mailbox API.
4734          * This call to IXGBE_SET_LPE action won't work with ixgbe pf drivers
4735          * prior to 3.11.33 which contains the following change:
4736          * "ixgbe: Enable jumbo frames support w/ SR-IOV"
4737          */
4738         ixgbevf_rlpml_set_vf(hw, max_frame);
4739
4740         /* update max frame size */
4741         dev->data->dev_conf.rxmode.max_rx_pkt_len = max_frame;
4742         return 0;
4743 }
4744
4745 #define MAC_TYPE_FILTER_SUP_EXT(type)    do {\
4746         if ((type) != ixgbe_mac_82599EB && (type) != ixgbe_mac_X540)\
4747                 return -ENOTSUP;\
4748 } while (0)
4749
4750 static inline struct ixgbe_5tuple_filter *
4751 ixgbe_5tuple_filter_lookup(struct ixgbe_5tuple_filter_list *filter_list,
4752                         struct ixgbe_5tuple_filter_info *key)
4753 {
4754         struct ixgbe_5tuple_filter *it;
4755
4756         TAILQ_FOREACH(it, filter_list, entries) {
4757                 if (memcmp(key, &it->filter_info,
4758                         sizeof(struct ixgbe_5tuple_filter_info)) == 0) {
4759                         return it;
4760                 }
4761         }
4762         return NULL;
4763 }
4764
4765 /* translate elements in struct rte_eth_ntuple_filter to struct ixgbe_5tuple_filter_info*/
4766 static inline int
4767 ntuple_filter_to_5tuple(struct rte_eth_ntuple_filter *filter,
4768                         struct ixgbe_5tuple_filter_info *filter_info)
4769 {
4770         if (filter->queue >= IXGBE_MAX_RX_QUEUE_NUM ||
4771                 filter->priority > IXGBE_5TUPLE_MAX_PRI ||
4772                 filter->priority < IXGBE_5TUPLE_MIN_PRI)
4773                 return -EINVAL;
4774
4775         switch (filter->dst_ip_mask) {
4776         case UINT32_MAX:
4777                 filter_info->dst_ip_mask = 0;
4778                 filter_info->dst_ip = filter->dst_ip;
4779                 break;
4780         case 0:
4781                 filter_info->dst_ip_mask = 1;
4782                 break;
4783         default:
4784                 PMD_DRV_LOG(ERR, "invalid dst_ip mask.");
4785                 return -EINVAL;
4786         }
4787
4788         switch (filter->src_ip_mask) {
4789         case UINT32_MAX:
4790                 filter_info->src_ip_mask = 0;
4791                 filter_info->src_ip = filter->src_ip;
4792                 break;
4793         case 0:
4794                 filter_info->src_ip_mask = 1;
4795                 break;
4796         default:
4797                 PMD_DRV_LOG(ERR, "invalid src_ip mask.");
4798                 return -EINVAL;
4799         }
4800
4801         switch (filter->dst_port_mask) {
4802         case UINT16_MAX:
4803                 filter_info->dst_port_mask = 0;
4804                 filter_info->dst_port = filter->dst_port;
4805                 break;
4806         case 0:
4807                 filter_info->dst_port_mask = 1;
4808                 break;
4809         default:
4810                 PMD_DRV_LOG(ERR, "invalid dst_port mask.");
4811                 return -EINVAL;
4812         }
4813
4814         switch (filter->src_port_mask) {
4815         case UINT16_MAX:
4816                 filter_info->src_port_mask = 0;
4817                 filter_info->src_port = filter->src_port;
4818                 break;
4819         case 0:
4820                 filter_info->src_port_mask = 1;
4821                 break;
4822         default:
4823                 PMD_DRV_LOG(ERR, "invalid src_port mask.");
4824                 return -EINVAL;
4825         }
4826
4827         switch (filter->proto_mask) {
4828         case UINT8_MAX:
4829                 filter_info->proto_mask = 0;
4830                 filter_info->proto =
4831                         convert_protocol_type(filter->proto);
4832                 break;
4833         case 0:
4834                 filter_info->proto_mask = 1;
4835                 break;
4836         default:
4837                 PMD_DRV_LOG(ERR, "invalid protocol mask.");
4838                 return -EINVAL;
4839         }
4840
4841         filter_info->priority = (uint8_t)filter->priority;
4842         return 0;
4843 }
4844
4845 /*
4846  * add or delete a ntuple filter
4847  *
4848  * @param
4849  * dev: Pointer to struct rte_eth_dev.
4850  * ntuple_filter: Pointer to struct rte_eth_ntuple_filter
4851  * add: if true, add filter, if false, remove filter
4852  *
4853  * @return
4854  *    - On success, zero.
4855  *    - On failure, a negative value.
4856  */
4857 static int
4858 ixgbe_add_del_ntuple_filter(struct rte_eth_dev *dev,
4859                         struct rte_eth_ntuple_filter *ntuple_filter,
4860                         bool add)
4861 {
4862         struct ixgbe_filter_info *filter_info =
4863                 IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
4864         struct ixgbe_5tuple_filter_info filter_5tuple;
4865         struct ixgbe_5tuple_filter *filter;
4866         int ret;
4867
4868         if (ntuple_filter->flags != RTE_5TUPLE_FLAGS) {
4869                 PMD_DRV_LOG(ERR, "only 5tuple is supported.");
4870                 return -EINVAL;
4871         }
4872
4873         memset(&filter_5tuple, 0, sizeof(struct ixgbe_5tuple_filter_info));
4874         ret = ntuple_filter_to_5tuple(ntuple_filter, &filter_5tuple);
4875         if (ret < 0)
4876                 return ret;
4877
4878         filter = ixgbe_5tuple_filter_lookup(&filter_info->fivetuple_list,
4879                                          &filter_5tuple);
4880         if (filter != NULL && add) {
4881                 PMD_DRV_LOG(ERR, "filter exists.");
4882                 return -EEXIST;
4883         }
4884         if (filter == NULL && !add) {
4885                 PMD_DRV_LOG(ERR, "filter doesn't exist.");
4886                 return -ENOENT;
4887         }
4888
4889         if (add) {
4890                 filter = rte_zmalloc("ixgbe_5tuple_filter",
4891                                 sizeof(struct ixgbe_5tuple_filter), 0);
4892                 if (filter == NULL)
4893                         return -ENOMEM;
4894                 (void)rte_memcpy(&filter->filter_info,
4895                                  &filter_5tuple,
4896                                  sizeof(struct ixgbe_5tuple_filter_info));
4897                 filter->queue = ntuple_filter->queue;
4898                 ret = ixgbe_add_5tuple_filter(dev, filter);
4899                 if (ret < 0) {
4900                         rte_free(filter);
4901                         return ret;
4902                 }
4903         } else
4904                 ixgbe_remove_5tuple_filter(dev, filter);
4905
4906         return 0;
4907 }
4908
4909 /*
4910  * get a ntuple filter
4911  *
4912  * @param
4913  * dev: Pointer to struct rte_eth_dev.
4914  * ntuple_filter: Pointer to struct rte_eth_ntuple_filter
4915  *
4916  * @return
4917  *    - On success, zero.
4918  *    - On failure, a negative value.
4919  */
4920 static int
4921 ixgbe_get_ntuple_filter(struct rte_eth_dev *dev,
4922                         struct rte_eth_ntuple_filter *ntuple_filter)
4923 {
4924         struct ixgbe_filter_info *filter_info =
4925                 IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
4926         struct ixgbe_5tuple_filter_info filter_5tuple;
4927         struct ixgbe_5tuple_filter *filter;
4928         int ret;
4929
4930         if (ntuple_filter->flags != RTE_5TUPLE_FLAGS) {
4931                 PMD_DRV_LOG(ERR, "only 5tuple is supported.");
4932                 return -EINVAL;
4933         }
4934
4935         memset(&filter_5tuple, 0, sizeof(struct ixgbe_5tuple_filter_info));
4936         ret = ntuple_filter_to_5tuple(ntuple_filter, &filter_5tuple);
4937         if (ret < 0)
4938                 return ret;
4939
4940         filter = ixgbe_5tuple_filter_lookup(&filter_info->fivetuple_list,
4941                                          &filter_5tuple);
4942         if (filter == NULL) {
4943                 PMD_DRV_LOG(ERR, "filter doesn't exist.");
4944                 return -ENOENT;
4945         }
4946         ntuple_filter->queue = filter->queue;
4947         return 0;
4948 }
4949
4950 /*
4951  * ixgbe_ntuple_filter_handle - Handle operations for ntuple filter.
4952  * @dev: pointer to rte_eth_dev structure
4953  * @filter_op:operation will be taken.
4954  * @arg: a pointer to specific structure corresponding to the filter_op
4955  *
4956  * @return
4957  *    - On success, zero.
4958  *    - On failure, a negative value.
4959  */
4960 static int
4961 ixgbe_ntuple_filter_handle(struct rte_eth_dev *dev,
4962                                 enum rte_filter_op filter_op,
4963                                 void *arg)
4964 {
4965         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
4966         int ret;
4967
4968         MAC_TYPE_FILTER_SUP_EXT(hw->mac.type);
4969
4970         if (filter_op == RTE_ETH_FILTER_NOP)
4971                 return 0;
4972
4973         if (arg == NULL) {
4974                 PMD_DRV_LOG(ERR, "arg shouldn't be NULL for operation %u.",
4975                             filter_op);
4976                 return -EINVAL;
4977         }
4978
4979         switch (filter_op) {
4980         case RTE_ETH_FILTER_ADD:
4981                 ret = ixgbe_add_del_ntuple_filter(dev,
4982                         (struct rte_eth_ntuple_filter *)arg,
4983                         TRUE);
4984                 break;
4985         case RTE_ETH_FILTER_DELETE:
4986                 ret = ixgbe_add_del_ntuple_filter(dev,
4987                         (struct rte_eth_ntuple_filter *)arg,
4988                         FALSE);
4989                 break;
4990         case RTE_ETH_FILTER_GET:
4991                 ret = ixgbe_get_ntuple_filter(dev,
4992                         (struct rte_eth_ntuple_filter *)arg);
4993                 break;
4994         default:
4995                 PMD_DRV_LOG(ERR, "unsupported operation %u.", filter_op);
4996                 ret = -EINVAL;
4997                 break;
4998         }
4999         return ret;
5000 }
5001
5002 static inline int
5003 ixgbe_ethertype_filter_lookup(struct ixgbe_filter_info *filter_info,
5004                         uint16_t ethertype)
5005 {
5006         int i;
5007
5008         for (i = 0; i < IXGBE_MAX_ETQF_FILTERS; i++) {
5009                 if (filter_info->ethertype_filters[i] == ethertype &&
5010                     (filter_info->ethertype_mask & (1 << i)))
5011                         return i;
5012         }
5013         return -1;
5014 }
5015
5016 static inline int
5017 ixgbe_ethertype_filter_insert(struct ixgbe_filter_info *filter_info,
5018                         uint16_t ethertype)
5019 {
5020         int i;
5021
5022         for (i = 0; i < IXGBE_MAX_ETQF_FILTERS; i++) {
5023                 if (!(filter_info->ethertype_mask & (1 << i))) {
5024                         filter_info->ethertype_mask |= 1 << i;
5025                         filter_info->ethertype_filters[i] = ethertype;
5026                         return i;
5027                 }
5028         }
5029         return -1;
5030 }
5031
5032 static inline int
5033 ixgbe_ethertype_filter_remove(struct ixgbe_filter_info *filter_info,
5034                         uint8_t idx)
5035 {
5036         if (idx >= IXGBE_MAX_ETQF_FILTERS)
5037                 return -1;
5038         filter_info->ethertype_mask &= ~(1 << idx);
5039         filter_info->ethertype_filters[idx] = 0;
5040         return idx;
5041 }
5042
5043 static int
5044 ixgbe_add_del_ethertype_filter(struct rte_eth_dev *dev,
5045                         struct rte_eth_ethertype_filter *filter,
5046                         bool add)
5047 {
5048         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5049         struct ixgbe_filter_info *filter_info =
5050                 IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
5051         uint32_t etqf = 0;
5052         uint32_t etqs = 0;
5053         int ret;
5054
5055         if (filter->queue >= IXGBE_MAX_RX_QUEUE_NUM)
5056                 return -EINVAL;
5057
5058         if (filter->ether_type == ETHER_TYPE_IPv4 ||
5059                 filter->ether_type == ETHER_TYPE_IPv6) {
5060                 PMD_DRV_LOG(ERR, "unsupported ether_type(0x%04x) in"
5061                         " ethertype filter.", filter->ether_type);
5062                 return -EINVAL;
5063         }
5064
5065         if (filter->flags & RTE_ETHTYPE_FLAGS_MAC) {
5066                 PMD_DRV_LOG(ERR, "mac compare is unsupported.");
5067                 return -EINVAL;
5068         }
5069         if (filter->flags & RTE_ETHTYPE_FLAGS_DROP) {
5070                 PMD_DRV_LOG(ERR, "drop option is unsupported.");
5071                 return -EINVAL;
5072         }
5073
5074         ret = ixgbe_ethertype_filter_lookup(filter_info, filter->ether_type);
5075         if (ret >= 0 && add) {
5076                 PMD_DRV_LOG(ERR, "ethertype (0x%04x) filter exists.",
5077                             filter->ether_type);
5078                 return -EEXIST;
5079         }
5080         if (ret < 0 && !add) {
5081                 PMD_DRV_LOG(ERR, "ethertype (0x%04x) filter doesn't exist.",
5082                             filter->ether_type);
5083                 return -ENOENT;
5084         }
5085
5086         if (add) {
5087                 ret = ixgbe_ethertype_filter_insert(filter_info,
5088                         filter->ether_type);
5089                 if (ret < 0) {
5090                         PMD_DRV_LOG(ERR, "ethertype filters are full.");
5091                         return -ENOSYS;
5092                 }
5093                 etqf = IXGBE_ETQF_FILTER_EN;
5094                 etqf |= (uint32_t)filter->ether_type;
5095                 etqs |= (uint32_t)((filter->queue <<
5096                                     IXGBE_ETQS_RX_QUEUE_SHIFT) &
5097                                     IXGBE_ETQS_RX_QUEUE);
5098                 etqs |= IXGBE_ETQS_QUEUE_EN;
5099         } else {
5100                 ret = ixgbe_ethertype_filter_remove(filter_info, (uint8_t)ret);
5101                 if (ret < 0)
5102                         return -ENOSYS;
5103         }
5104         IXGBE_WRITE_REG(hw, IXGBE_ETQF(ret), etqf);
5105         IXGBE_WRITE_REG(hw, IXGBE_ETQS(ret), etqs);
5106         IXGBE_WRITE_FLUSH(hw);
5107
5108         return 0;
5109 }
5110
5111 static int
5112 ixgbe_get_ethertype_filter(struct rte_eth_dev *dev,
5113                         struct rte_eth_ethertype_filter *filter)
5114 {
5115         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5116         struct ixgbe_filter_info *filter_info =
5117                 IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
5118         uint32_t etqf, etqs;
5119         int ret;
5120
5121         ret = ixgbe_ethertype_filter_lookup(filter_info, filter->ether_type);
5122         if (ret < 0) {
5123                 PMD_DRV_LOG(ERR, "ethertype (0x%04x) filter doesn't exist.",
5124                             filter->ether_type);
5125                 return -ENOENT;
5126         }
5127
5128         etqf = IXGBE_READ_REG(hw, IXGBE_ETQF(ret));
5129         if (etqf & IXGBE_ETQF_FILTER_EN) {
5130                 etqs = IXGBE_READ_REG(hw, IXGBE_ETQS(ret));
5131                 filter->ether_type = etqf & IXGBE_ETQF_ETHERTYPE;
5132                 filter->flags = 0;
5133                 filter->queue = (etqs & IXGBE_ETQS_RX_QUEUE) >>
5134                                IXGBE_ETQS_RX_QUEUE_SHIFT;
5135                 return 0;
5136         }
5137         return -ENOENT;
5138 }
5139
5140 /*
5141  * ixgbe_ethertype_filter_handle - Handle operations for ethertype filter.
5142  * @dev: pointer to rte_eth_dev structure
5143  * @filter_op:operation will be taken.
5144  * @arg: a pointer to specific structure corresponding to the filter_op
5145  */
5146 static int
5147 ixgbe_ethertype_filter_handle(struct rte_eth_dev *dev,
5148                                 enum rte_filter_op filter_op,
5149                                 void *arg)
5150 {
5151         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5152         int ret;
5153
5154         MAC_TYPE_FILTER_SUP(hw->mac.type);
5155
5156         if (filter_op == RTE_ETH_FILTER_NOP)
5157                 return 0;
5158
5159         if (arg == NULL) {
5160                 PMD_DRV_LOG(ERR, "arg shouldn't be NULL for operation %u.",
5161                             filter_op);
5162                 return -EINVAL;
5163         }
5164
5165         switch (filter_op) {
5166         case RTE_ETH_FILTER_ADD:
5167                 ret = ixgbe_add_del_ethertype_filter(dev,
5168                         (struct rte_eth_ethertype_filter *)arg,
5169                         TRUE);
5170                 break;
5171         case RTE_ETH_FILTER_DELETE:
5172                 ret = ixgbe_add_del_ethertype_filter(dev,
5173                         (struct rte_eth_ethertype_filter *)arg,
5174                         FALSE);
5175                 break;
5176         case RTE_ETH_FILTER_GET:
5177                 ret = ixgbe_get_ethertype_filter(dev,
5178                         (struct rte_eth_ethertype_filter *)arg);
5179                 break;
5180         default:
5181                 PMD_DRV_LOG(ERR, "unsupported operation %u.", filter_op);
5182                 ret = -EINVAL;
5183                 break;
5184         }
5185         return ret;
5186 }
5187
5188 static int
5189 ixgbe_dev_filter_ctrl(struct rte_eth_dev *dev,
5190                      enum rte_filter_type filter_type,
5191                      enum rte_filter_op filter_op,
5192                      void *arg)
5193 {
5194         int ret = -EINVAL;
5195
5196         switch (filter_type) {
5197         case RTE_ETH_FILTER_NTUPLE:
5198                 ret = ixgbe_ntuple_filter_handle(dev, filter_op, arg);
5199                 break;
5200         case RTE_ETH_FILTER_ETHERTYPE:
5201                 ret = ixgbe_ethertype_filter_handle(dev, filter_op, arg);
5202                 break;
5203         case RTE_ETH_FILTER_SYN:
5204                 ret = ixgbe_syn_filter_handle(dev, filter_op, arg);
5205                 break;
5206         case RTE_ETH_FILTER_FDIR:
5207                 ret = ixgbe_fdir_ctrl_func(dev, filter_op, arg);
5208                 break;
5209         default:
5210                 PMD_DRV_LOG(WARNING, "Filter type (%d) not supported",
5211                                                         filter_type);
5212                 break;
5213         }
5214
5215         return ret;
5216 }
5217
5218 static u8 *
5219 ixgbe_dev_addr_list_itr(__attribute__((unused)) struct ixgbe_hw *hw,
5220                         u8 **mc_addr_ptr, u32 *vmdq)
5221 {
5222         u8 *mc_addr;
5223
5224         *vmdq = 0;
5225         mc_addr = *mc_addr_ptr;
5226         *mc_addr_ptr = (mc_addr + sizeof(struct ether_addr));
5227         return mc_addr;
5228 }
5229
5230 static int
5231 ixgbe_dev_set_mc_addr_list(struct rte_eth_dev *dev,
5232                           struct ether_addr *mc_addr_set,
5233                           uint32_t nb_mc_addr)
5234 {
5235         struct ixgbe_hw *hw;
5236         u8 *mc_addr_list;
5237
5238         hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5239         mc_addr_list = (u8 *)mc_addr_set;
5240         return ixgbe_update_mc_addr_list(hw, mc_addr_list, nb_mc_addr,
5241                                          ixgbe_dev_addr_list_itr, TRUE);
5242 }
5243
5244 static int
5245 ixgbe_timesync_enable(struct rte_eth_dev *dev)
5246 {
5247         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5248         uint32_t tsync_ctl;
5249         uint32_t tsauxc;
5250
5251         /* Enable system time for platforms where it isn't on by default. */
5252         tsauxc = IXGBE_READ_REG(hw, IXGBE_TSAUXC);
5253         tsauxc &= ~IXGBE_TSAUXC_DISABLE_SYSTIME;
5254         IXGBE_WRITE_REG(hw, IXGBE_TSAUXC, tsauxc);
5255
5256         /* Start incrementing the register used to timestamp PTP packets. */
5257         IXGBE_WRITE_REG(hw, IXGBE_TIMINCA, IXGBE_TIMINCA_INIT);
5258
5259         /* Enable L2 filtering of IEEE1588/802.1AS Ethernet frame types. */
5260         IXGBE_WRITE_REG(hw, IXGBE_ETQF(IXGBE_ETQF_FILTER_1588),
5261                         (ETHER_TYPE_1588 |
5262                          IXGBE_ETQF_FILTER_EN |
5263                          IXGBE_ETQF_1588));
5264
5265         /* Enable timestamping of received PTP packets. */
5266         tsync_ctl = IXGBE_READ_REG(hw, IXGBE_TSYNCRXCTL);
5267         tsync_ctl |= IXGBE_TSYNCRXCTL_ENABLED;
5268         IXGBE_WRITE_REG(hw, IXGBE_TSYNCRXCTL, tsync_ctl);
5269
5270         /* Enable timestamping of transmitted PTP packets. */
5271         tsync_ctl = IXGBE_READ_REG(hw, IXGBE_TSYNCTXCTL);
5272         tsync_ctl |= IXGBE_TSYNCTXCTL_ENABLED;
5273         IXGBE_WRITE_REG(hw, IXGBE_TSYNCTXCTL, tsync_ctl);
5274
5275         return 0;
5276 }
5277
5278 static int
5279 ixgbe_timesync_disable(struct rte_eth_dev *dev)
5280 {
5281         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5282         uint32_t tsync_ctl;
5283
5284         /* Disable timestamping of transmitted PTP packets. */
5285         tsync_ctl = IXGBE_READ_REG(hw, IXGBE_TSYNCTXCTL);
5286         tsync_ctl &= ~IXGBE_TSYNCTXCTL_ENABLED;
5287         IXGBE_WRITE_REG(hw, IXGBE_TSYNCTXCTL, tsync_ctl);
5288
5289         /* Disable timestamping of received PTP packets. */
5290         tsync_ctl = IXGBE_READ_REG(hw, IXGBE_TSYNCRXCTL);
5291         tsync_ctl &= ~IXGBE_TSYNCRXCTL_ENABLED;
5292         IXGBE_WRITE_REG(hw, IXGBE_TSYNCRXCTL, tsync_ctl);
5293
5294         /* Disable L2 filtering of IEEE1588/802.1AS Ethernet frame types. */
5295         IXGBE_WRITE_REG(hw, IXGBE_ETQF(IXGBE_ETQF_FILTER_1588), 0);
5296
5297         /* Stop incrementating the System Time registers. */
5298         IXGBE_WRITE_REG(hw, IXGBE_TIMINCA, 0);
5299
5300         return 0;
5301 }
5302
5303 static int
5304 ixgbe_timesync_read_rx_timestamp(struct rte_eth_dev *dev,
5305                                  struct timespec *timestamp,
5306                                  uint32_t flags __rte_unused)
5307 {
5308         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5309         uint32_t tsync_rxctl;
5310         uint32_t rx_stmpl;
5311         uint32_t rx_stmph;
5312
5313         tsync_rxctl = IXGBE_READ_REG(hw, IXGBE_TSYNCRXCTL);
5314         if ((tsync_rxctl & IXGBE_TSYNCRXCTL_VALID) == 0)
5315                 return -EINVAL;
5316
5317         rx_stmpl = IXGBE_READ_REG(hw, IXGBE_RXSTMPL);
5318         rx_stmph = IXGBE_READ_REG(hw, IXGBE_RXSTMPH);
5319
5320         timestamp->tv_sec = (uint64_t)(((uint64_t)rx_stmph << 32) | rx_stmpl);
5321         timestamp->tv_nsec = 0;
5322
5323         return  0;
5324 }
5325
5326 static int
5327 ixgbe_timesync_read_tx_timestamp(struct rte_eth_dev *dev,
5328                                  struct timespec *timestamp)
5329 {
5330         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5331         uint32_t tsync_txctl;
5332         uint32_t tx_stmpl;
5333         uint32_t tx_stmph;
5334
5335         tsync_txctl = IXGBE_READ_REG(hw, IXGBE_TSYNCTXCTL);
5336         if ((tsync_txctl & IXGBE_TSYNCTXCTL_VALID) == 0)
5337                 return -EINVAL;
5338
5339         tx_stmpl = IXGBE_READ_REG(hw, IXGBE_TXSTMPL);
5340         tx_stmph = IXGBE_READ_REG(hw, IXGBE_TXSTMPH);
5341
5342         timestamp->tv_sec = (uint64_t)(((uint64_t)tx_stmph << 32) | tx_stmpl);
5343         timestamp->tv_nsec = 0;
5344
5345         return  0;
5346 }
5347
5348 static int
5349 ixgbe_get_reg_length(struct rte_eth_dev *dev)
5350 {
5351         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5352         int count = 0;
5353         int g_ind = 0;
5354         const struct reg_info *reg_group;
5355         const struct reg_info **reg_set = (hw->mac.type == ixgbe_mac_82598EB) ?
5356                                     ixgbe_regs_mac_82598EB : ixgbe_regs_others;
5357
5358         while ((reg_group = reg_set[g_ind++]))
5359                 count += ixgbe_regs_group_count(reg_group);
5360
5361         return count;
5362 }
5363
5364 static int
5365 ixgbevf_get_reg_length(struct rte_eth_dev *dev __rte_unused)
5366 {
5367         int count = 0;
5368         int g_ind = 0;
5369         const struct reg_info *reg_group;
5370
5371         while ((reg_group = ixgbevf_regs[g_ind++]))
5372                 count += ixgbe_regs_group_count(reg_group);
5373
5374         return count;
5375 }
5376
5377 static int
5378 ixgbe_get_regs(struct rte_eth_dev *dev,
5379               struct rte_dev_reg_info *regs)
5380 {
5381         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5382         uint32_t *data = regs->data;
5383         int g_ind = 0;
5384         int count = 0;
5385         const struct reg_info *reg_group;
5386         const struct reg_info **reg_set = (hw->mac.type == ixgbe_mac_82598EB) ?
5387                                     ixgbe_regs_mac_82598EB : ixgbe_regs_others;
5388
5389         /* Support only full register dump */
5390         if ((regs->length == 0) ||
5391             (regs->length == (uint32_t)ixgbe_get_reg_length(dev))) {
5392                 regs->version = hw->mac.type << 24 | hw->revision_id << 16 |
5393                         hw->device_id;
5394                 while ((reg_group = reg_set[g_ind++]))
5395                         count += ixgbe_read_regs_group(dev, &data[count],
5396                                 reg_group);
5397                 return 0;
5398         }
5399
5400         return -ENOTSUP;
5401 }
5402
5403 static int
5404 ixgbevf_get_regs(struct rte_eth_dev *dev,
5405                 struct rte_dev_reg_info *regs)
5406 {
5407         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5408         uint32_t *data = regs->data;
5409         int g_ind = 0;
5410         int count = 0;
5411         const struct reg_info *reg_group;
5412
5413         /* Support only full register dump */
5414         if ((regs->length == 0) ||
5415             (regs->length == (uint32_t)ixgbevf_get_reg_length(dev))) {
5416                 regs->version = hw->mac.type << 24 | hw->revision_id << 16 |
5417                         hw->device_id;
5418                 while ((reg_group = ixgbevf_regs[g_ind++]))
5419                         count += ixgbe_read_regs_group(dev, &data[count],
5420                                                       reg_group);
5421                 return 0;
5422         }
5423
5424         return -ENOTSUP;
5425 }
5426
5427 static int
5428 ixgbe_get_eeprom_length(struct rte_eth_dev *dev)
5429 {
5430         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5431
5432         /* Return unit is byte count */
5433         return hw->eeprom.word_size * 2;
5434 }
5435
5436 static int
5437 ixgbe_get_eeprom(struct rte_eth_dev *dev,
5438                 struct rte_dev_eeprom_info *in_eeprom)
5439 {
5440         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5441         struct ixgbe_eeprom_info *eeprom = &hw->eeprom;
5442         uint16_t *data = in_eeprom->data;
5443         int first, length;
5444
5445         first = in_eeprom->offset >> 1;
5446         length = in_eeprom->length >> 1;
5447         if ((first >= hw->eeprom.word_size) ||
5448             ((first + length) >= hw->eeprom.word_size))
5449                 return -EINVAL;
5450
5451         in_eeprom->magic = hw->vendor_id | (hw->device_id << 16);
5452
5453         return eeprom->ops.read_buffer(hw, first, length, data);
5454 }
5455
5456 static int
5457 ixgbe_set_eeprom(struct rte_eth_dev *dev,
5458                 struct rte_dev_eeprom_info *in_eeprom)
5459 {
5460         struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
5461         struct ixgbe_eeprom_info *eeprom = &hw->eeprom;
5462         uint16_t *data = in_eeprom->data;
5463         int first, length;
5464
5465         first = in_eeprom->offset >> 1;
5466         length = in_eeprom->length >> 1;
5467         if ((first >= hw->eeprom.word_size) ||
5468             ((first + length) >= hw->eeprom.word_size))
5469                 return -EINVAL;
5470
5471         in_eeprom->magic = hw->vendor_id | (hw->device_id << 16);
5472
5473         return eeprom->ops.write_buffer(hw,  first, length, data);
5474 }
5475
5476 static struct rte_driver rte_ixgbe_driver = {
5477         .type = PMD_PDEV,
5478         .init = rte_ixgbe_pmd_init,
5479 };
5480
5481 static struct rte_driver rte_ixgbevf_driver = {
5482         .type = PMD_PDEV,
5483         .init = rte_ixgbevf_pmd_init,
5484 };
5485
5486 PMD_REGISTER_DRIVER(rte_ixgbe_driver);
5487 PMD_REGISTER_DRIVER(rte_ixgbevf_driver);