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