drivers/net/e1000/em_ethdev.c

   1 /*-
   2  *   BSD LICENSE
   3  *
   4  *   Copyright(c) 2010-2016 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 <stdarg.h>
  39
  40 #include <rte_common.h>
  41 #include <rte_interrupts.h>
  42 #include <rte_byteorder.h>
  43 #include <rte_log.h>
  44 #include <rte_debug.h>
  45 #include <rte_pci.h>
  46 #include <rte_ether.h>
  47 #include <rte_ethdev.h>
  48 #include <rte_ethdev_pci.h>
  49 #include <rte_memory.h>
  50 #include <rte_memzone.h>
  51 #include <rte_eal.h>
  52 #include <rte_atomic.h>
  53 #include <rte_malloc.h>
  54 #include <rte_dev.h>
  55
  56 #include "e1000_logs.h"
  57 #include "base/e1000_api.h"
  58 #include "e1000_ethdev.h"
  59
  60 #define EM_EIAC                 0x000DC
  61
  62 #define PMD_ROUNDUP(x,y)        (((x) + (y) - 1)/(y) * (y))
  63
  64
  65 static int eth_em_configure(struct rte_eth_dev *dev);
  66 static int eth_em_start(struct rte_eth_dev *dev);
  67 static void eth_em_stop(struct rte_eth_dev *dev);
  68 static void eth_em_close(struct rte_eth_dev *dev);
  69 static void eth_em_promiscuous_enable(struct rte_eth_dev *dev);
  70 static void eth_em_promiscuous_disable(struct rte_eth_dev *dev);
  71 static void eth_em_allmulticast_enable(struct rte_eth_dev *dev);
  72 static void eth_em_allmulticast_disable(struct rte_eth_dev *dev);
  73 static int eth_em_link_update(struct rte_eth_dev *dev,
  74                                 int wait_to_complete);
  75 static int eth_em_stats_get(struct rte_eth_dev *dev,
  76                                 struct rte_eth_stats *rte_stats);
  77 static void eth_em_stats_reset(struct rte_eth_dev *dev);
  78 static void eth_em_infos_get(struct rte_eth_dev *dev,
  79                                 struct rte_eth_dev_info *dev_info);
  80 static int eth_em_flow_ctrl_get(struct rte_eth_dev *dev,
  81                                 struct rte_eth_fc_conf *fc_conf);
  82 static int eth_em_flow_ctrl_set(struct rte_eth_dev *dev,
  83                                 struct rte_eth_fc_conf *fc_conf);
  84 static int eth_em_interrupt_setup(struct rte_eth_dev *dev);
  85 static int eth_em_rxq_interrupt_setup(struct rte_eth_dev *dev);
  86 static int eth_em_interrupt_get_status(struct rte_eth_dev *dev);
  87 static int eth_em_interrupt_action(struct rte_eth_dev *dev,
  88                                    struct rte_intr_handle *handle);
  89 static void eth_em_interrupt_handler(void *param);
  90
  91 static int em_hw_init(struct e1000_hw *hw);
  92 static int em_hardware_init(struct e1000_hw *hw);
  93 static void em_hw_control_acquire(struct e1000_hw *hw);
  94 static void em_hw_control_release(struct e1000_hw *hw);
  95 static void em_init_manageability(struct e1000_hw *hw);
  96 static void em_release_manageability(struct e1000_hw *hw);
  97
  98 static int eth_em_mtu_set(struct rte_eth_dev *dev, uint16_t mtu);
  99
 100 static int eth_em_vlan_filter_set(struct rte_eth_dev *dev,
 101                 uint16_t vlan_id, int on);
 102 static void eth_em_vlan_offload_set(struct rte_eth_dev *dev, int mask);
 103 static void em_vlan_hw_filter_enable(struct rte_eth_dev *dev);
 104 static void em_vlan_hw_filter_disable(struct rte_eth_dev *dev);
 105 static void em_vlan_hw_strip_enable(struct rte_eth_dev *dev);
 106 static void em_vlan_hw_strip_disable(struct rte_eth_dev *dev);
 107
 108 /*
 109 static void eth_em_vlan_filter_set(struct rte_eth_dev *dev,
 110                                         uint16_t vlan_id, int on);
 111 */
 112
 113 static int eth_em_rx_queue_intr_enable(struct rte_eth_dev *dev, uint16_t queue_id);
 114 static int eth_em_rx_queue_intr_disable(struct rte_eth_dev *dev, uint16_t queue_id);
 115 static void em_lsc_intr_disable(struct e1000_hw *hw);
 116 static void em_rxq_intr_enable(struct e1000_hw *hw);
 117 static void em_rxq_intr_disable(struct e1000_hw *hw);
 118
 119 static int eth_em_led_on(struct rte_eth_dev *dev);
 120 static int eth_em_led_off(struct rte_eth_dev *dev);
 121
 122 static int em_get_rx_buffer_size(struct e1000_hw *hw);
 123 static int eth_em_rar_set(struct rte_eth_dev *dev, struct ether_addr *mac_addr,
 124                           uint32_t index, uint32_t pool);
 125 static void eth_em_rar_clear(struct rte_eth_dev *dev, uint32_t index);
 126
 127 static int eth_em_set_mc_addr_list(struct rte_eth_dev *dev,
 128                                    struct ether_addr *mc_addr_set,
 129                                    uint32_t nb_mc_addr);
 130
 131 #define EM_FC_PAUSE_TIME 0x0680
 132 #define EM_LINK_UPDATE_CHECK_TIMEOUT  90  /* 9s */
 133 #define EM_LINK_UPDATE_CHECK_INTERVAL 100 /* ms */
 134
 135 static enum e1000_fc_mode em_fc_setting = e1000_fc_full;
 136
 137 /*
 138  * The set of PCI devices this driver supports
 139  */
 140 static const struct rte_pci_id pci_id_em_map[] = {
 141         { RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82540EM) },
 142         { RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82545EM_COPPER) },
 143         { RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82545EM_FIBER) },
 144         { RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82546EB_COPPER) },
 145         { RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82546EB_FIBER) },
 146         { RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82546EB_QUAD_COPPER) },
 147         { RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82571EB_COPPER) },
 148         { RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82571EB_FIBER) },
 149         { RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82571EB_SERDES) },
 150         { RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82571EB_SERDES_DUAL) },
 151         { RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82571EB_SERDES_QUAD) },
 152         { RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82571EB_QUAD_COPPER) },
 153         { RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82571PT_QUAD_COPPER) },
 154         { RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82571EB_QUAD_FIBER) },
 155         { RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82571EB_QUAD_COPPER_LP) },
 156         { RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82572EI_COPPER) },
 157         { RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82572EI_FIBER) },
 158         { RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82572EI_SERDES) },
 159         { RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82572EI) },
 160         { RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82573L) },
 161         { RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82574L) },
 162         { RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82574LA) },
 163         { RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82583V) },
 164         { RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_LPT_I217_LM) },
 165         { RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_LPT_I217_V) },
 166         { RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_LPTLP_I218_LM) },
 167         { RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_LPTLP_I218_V) },
 168         { RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_I218_LM2) },
 169         { RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_I218_V2) },
 170         { RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_I218_LM3) },
 171         { RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_I218_V3) },
 172         { RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_SPT_I219_LM) },
 173         { RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_SPT_I219_V) },
 174         { RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_SPT_I219_LM2) },
 175         { RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_SPT_I219_V2) },
 176         { RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_LBG_I219_LM3) },
 177         { RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_SPT_I219_LM4) },
 178         { RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_SPT_I219_V4) },
 179         { RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_SPT_I219_LM5) },
 180         { RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_SPT_I219_V5) },
 181         { RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_CNP_I219_LM6) },
 182         { RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_CNP_I219_V6) },
 183         { RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_CNP_I219_LM7) },
 184         { RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_CNP_I219_V7) },
 185         { .vendor_id = 0, /* sentinel */ },
 186 };
 187
 188 static const struct eth_dev_ops eth_em_ops = {
 189         .dev_configure        = eth_em_configure,
 190         .dev_start            = eth_em_start,
 191         .dev_stop             = eth_em_stop,
 192         .dev_close            = eth_em_close,
 193         .promiscuous_enable   = eth_em_promiscuous_enable,
 194         .promiscuous_disable  = eth_em_promiscuous_disable,
 195         .allmulticast_enable  = eth_em_allmulticast_enable,
 196         .allmulticast_disable = eth_em_allmulticast_disable,
 197         .link_update          = eth_em_link_update,
 198         .stats_get            = eth_em_stats_get,
 199         .stats_reset          = eth_em_stats_reset,
 200         .dev_infos_get        = eth_em_infos_get,
 201         .mtu_set              = eth_em_mtu_set,
 202         .vlan_filter_set      = eth_em_vlan_filter_set,
 203         .vlan_offload_set     = eth_em_vlan_offload_set,
 204         .rx_queue_setup       = eth_em_rx_queue_setup,
 205         .rx_queue_release     = eth_em_rx_queue_release,
 206         .rx_queue_count       = eth_em_rx_queue_count,
 207         .rx_descriptor_done   = eth_em_rx_descriptor_done,
 208         .rx_descriptor_status = eth_em_rx_descriptor_status,
 209         .tx_descriptor_status = eth_em_tx_descriptor_status,
 210         .tx_queue_setup       = eth_em_tx_queue_setup,
 211         .tx_queue_release     = eth_em_tx_queue_release,
 212         .rx_queue_intr_enable = eth_em_rx_queue_intr_enable,
 213         .rx_queue_intr_disable = eth_em_rx_queue_intr_disable,
 214         .dev_led_on           = eth_em_led_on,
 215         .dev_led_off          = eth_em_led_off,
 216         .flow_ctrl_get        = eth_em_flow_ctrl_get,
 217         .flow_ctrl_set        = eth_em_flow_ctrl_set,
 218         .mac_addr_add         = eth_em_rar_set,
 219         .mac_addr_remove      = eth_em_rar_clear,
 220         .set_mc_addr_list     = eth_em_set_mc_addr_list,
 221         .rxq_info_get         = em_rxq_info_get,
 222         .txq_info_get         = em_txq_info_get,
 223 };
 224
 225 /**
 226  * Atomically reads the link status information from global
 227  * structure rte_eth_dev.
 228  *
 229  * @param dev
 230  *   - Pointer to the structure rte_eth_dev to read from.
 231  *   - Pointer to the buffer to be saved with the link status.
 232  *
 233  * @return
 234  *   - On success, zero.
 235  *   - On failure, negative value.
 236  */
 237 static inline int
 238 rte_em_dev_atomic_read_link_status(struct rte_eth_dev *dev,
 239                                 struct rte_eth_link *link)
 240 {
 241         struct rte_eth_link *dst = link;
 242         struct rte_eth_link *src = &(dev->data->dev_link);
 243
 244         if (rte_atomic64_cmpset((uint64_t *)dst, *(uint64_t *)dst,
 245                                         *(uint64_t *)src) == 0)
 246                 return -1;
 247
 248         return 0;
 249 }
 250
 251 /**
 252  * Atomically writes the link status information into global
 253  * structure rte_eth_dev.
 254  *
 255  * @param dev
 256  *   - Pointer to the structure rte_eth_dev to read from.
 257  *   - Pointer to the buffer to be saved with the link status.
 258  *
 259  * @return
 260  *   - On success, zero.
 261  *   - On failure, negative value.
 262  */
 263 static inline int
 264 rte_em_dev_atomic_write_link_status(struct rte_eth_dev *dev,
 265                                 struct rte_eth_link *link)
 266 {
 267         struct rte_eth_link *dst = &(dev->data->dev_link);
 268         struct rte_eth_link *src = link;
 269
 270         if (rte_atomic64_cmpset((uint64_t *)dst, *(uint64_t *)dst,
 271                                         *(uint64_t *)src) == 0)
 272                 return -1;
 273
 274         return 0;
 275 }
 276
 277 /**
 278  *  eth_em_dev_is_ich8 - Check for ICH8 device
 279  *  @hw: pointer to the HW structure
 280  *
 281  *  return TRUE for ICH8, otherwise FALSE
 282  **/
 283 static bool
 284 eth_em_dev_is_ich8(struct e1000_hw *hw)
 285 {
 286         DEBUGFUNC("eth_em_dev_is_ich8");
 287
 288         switch (hw->device_id) {
 289         case E1000_DEV_ID_PCH_LPT_I217_LM:
 290         case E1000_DEV_ID_PCH_LPT_I217_V:
 291         case E1000_DEV_ID_PCH_LPTLP_I218_LM:
 292         case E1000_DEV_ID_PCH_LPTLP_I218_V:
 293         case E1000_DEV_ID_PCH_I218_V2:
 294         case E1000_DEV_ID_PCH_I218_LM2:
 295         case E1000_DEV_ID_PCH_I218_V3:
 296         case E1000_DEV_ID_PCH_I218_LM3:
 297         case E1000_DEV_ID_PCH_SPT_I219_LM:
 298         case E1000_DEV_ID_PCH_SPT_I219_V:
 299         case E1000_DEV_ID_PCH_SPT_I219_LM2:
 300         case E1000_DEV_ID_PCH_SPT_I219_V2:
 301         case E1000_DEV_ID_PCH_LBG_I219_LM3:
 302         case E1000_DEV_ID_PCH_SPT_I219_LM4:
 303         case E1000_DEV_ID_PCH_SPT_I219_V4:
 304         case E1000_DEV_ID_PCH_SPT_I219_LM5:
 305         case E1000_DEV_ID_PCH_SPT_I219_V5:
 306         case E1000_DEV_ID_PCH_CNP_I219_LM6:
 307         case E1000_DEV_ID_PCH_CNP_I219_V6:
 308         case E1000_DEV_ID_PCH_CNP_I219_LM7:
 309         case E1000_DEV_ID_PCH_CNP_I219_V7:
 310                 return 1;
 311         default:
 312                 return 0;
 313         }
 314 }
 315
 316 static int
 317 eth_em_dev_init(struct rte_eth_dev *eth_dev)
 318 {
 319         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
 320         struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
 321         struct e1000_adapter *adapter =
 322                 E1000_DEV_PRIVATE(eth_dev->data->dev_private);
 323         struct e1000_hw *hw =
 324                 E1000_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
 325         struct e1000_vfta * shadow_vfta =
 326                 E1000_DEV_PRIVATE_TO_VFTA(eth_dev->data->dev_private);
 327
 328         eth_dev->dev_ops = &eth_em_ops;
 329         eth_dev->rx_pkt_burst = (eth_rx_burst_t)&eth_em_recv_pkts;
 330         eth_dev->tx_pkt_burst = (eth_tx_burst_t)&eth_em_xmit_pkts;
 331         eth_dev->tx_pkt_prepare = (eth_tx_prep_t)&eth_em_prep_pkts;
 332
 333         /* for secondary processes, we don't initialise any further as primary
 334          * has already done this work. Only check we don't need a different
 335          * RX function */
 336         if (rte_eal_process_type() != RTE_PROC_PRIMARY){
 337                 if (eth_dev->data->scattered_rx)
 338                         eth_dev->rx_pkt_burst =
 339                                 (eth_rx_burst_t)&eth_em_recv_scattered_pkts;
 340                 return 0;
 341         }
 342
 343         rte_eth_copy_pci_info(eth_dev, pci_dev);
 344         eth_dev->data->dev_flags |= RTE_ETH_DEV_DETACHABLE;
 345
 346         hw->hw_addr = (void *)pci_dev->mem_resource[0].addr;
 347         hw->device_id = pci_dev->id.device_id;
 348         adapter->stopped = 0;
 349
 350         /* For ICH8 support we'll need to map the flash memory BAR */
 351         if (eth_em_dev_is_ich8(hw))
 352                 hw->flash_address = (void *)pci_dev->mem_resource[1].addr;
 353
 354         if (e1000_setup_init_funcs(hw, TRUE) != E1000_SUCCESS ||
 355                         em_hw_init(hw) != 0) {
 356                 PMD_INIT_LOG(ERR, "port_id %d vendorID=0x%x deviceID=0x%x: "
 357                         "failed to init HW",
 358                         eth_dev->data->port_id, pci_dev->id.vendor_id,
 359                         pci_dev->id.device_id);
 360                 return -ENODEV;
 361         }
 362
 363         /* Allocate memory for storing MAC addresses */
 364         eth_dev->data->mac_addrs = rte_zmalloc("e1000", ETHER_ADDR_LEN *
 365                         hw->mac.rar_entry_count, 0);
 366         if (eth_dev->data->mac_addrs == NULL) {
 367                 PMD_INIT_LOG(ERR, "Failed to allocate %d bytes needed to "
 368                         "store MAC addresses",
 369                         ETHER_ADDR_LEN * hw->mac.rar_entry_count);
 370                 return -ENOMEM;
 371         }
 372
 373         /* Copy the permanent MAC address */
 374         ether_addr_copy((struct ether_addr *) hw->mac.addr,
 375                 eth_dev->data->mac_addrs);
 376
 377         /* initialize the vfta */
 378         memset(shadow_vfta, 0, sizeof(*shadow_vfta));
 379
 380         PMD_INIT_LOG(DEBUG, "port_id %d vendorID=0x%x deviceID=0x%x",
 381                      eth_dev->data->port_id, pci_dev->id.vendor_id,
 382                      pci_dev->id.device_id);
 383
 384         rte_intr_callback_register(intr_handle,
 385                                    eth_em_interrupt_handler, eth_dev);
 386
 387         return 0;
 388 }
 389
 390 static int
 391 eth_em_dev_uninit(struct rte_eth_dev *eth_dev)
 392 {
 393         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
 394         struct e1000_adapter *adapter =
 395                 E1000_DEV_PRIVATE(eth_dev->data->dev_private);
 396         struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
 397
 398         PMD_INIT_FUNC_TRACE();
 399
 400         if (rte_eal_process_type() != RTE_PROC_PRIMARY)
 401                 return -EPERM;
 402
 403         if (adapter->stopped == 0)
 404                 eth_em_close(eth_dev);
 405
 406         eth_dev->dev_ops = NULL;
 407         eth_dev->rx_pkt_burst = NULL;
 408         eth_dev->tx_pkt_burst = NULL;
 409
 410         rte_free(eth_dev->data->mac_addrs);
 411         eth_dev->data->mac_addrs = NULL;
 412
 413         /* disable uio intr before callback unregister */
 414         rte_intr_disable(intr_handle);
 415         rte_intr_callback_unregister(intr_handle,
 416                                      eth_em_interrupt_handler, eth_dev);
 417
 418         return 0;
 419 }
 420
 421 static int eth_em_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
 422         struct rte_pci_device *pci_dev)
 423 {
 424         return rte_eth_dev_pci_generic_probe(pci_dev,
 425                 sizeof(struct e1000_adapter), eth_em_dev_init);
 426 }
 427
 428 static int eth_em_pci_remove(struct rte_pci_device *pci_dev)
 429 {
 430         return rte_eth_dev_pci_generic_remove(pci_dev, eth_em_dev_uninit);
 431 }
 432
 433 static struct rte_pci_driver rte_em_pmd = {
 434         .id_table = pci_id_em_map,
 435         .drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC |
 436                      RTE_PCI_DRV_IOVA_AS_VA,
 437         .probe = eth_em_pci_probe,
 438         .remove = eth_em_pci_remove,
 439 };
 440
 441 static int
 442 em_hw_init(struct e1000_hw *hw)
 443 {
 444         int diag;
 445
 446         diag = hw->mac.ops.init_params(hw);
 447         if (diag != 0) {
 448                 PMD_INIT_LOG(ERR, "MAC Initialization Error");
 449                 return diag;
 450         }
 451         diag = hw->nvm.ops.init_params(hw);
 452         if (diag != 0) {
 453                 PMD_INIT_LOG(ERR, "NVM Initialization Error");
 454                 return diag;
 455         }
 456         diag = hw->phy.ops.init_params(hw);
 457         if (diag != 0) {
 458                 PMD_INIT_LOG(ERR, "PHY Initialization Error");
 459                 return diag;
 460         }
 461         (void) e1000_get_bus_info(hw);
 462
 463         hw->mac.autoneg = 1;
 464         hw->phy.autoneg_wait_to_complete = 0;
 465         hw->phy.autoneg_advertised = E1000_ALL_SPEED_DUPLEX;
 466
 467         e1000_init_script_state_82541(hw, TRUE);
 468         e1000_set_tbi_compatibility_82543(hw, TRUE);
 469
 470         /* Copper options */
 471         if (hw->phy.media_type == e1000_media_type_copper) {
 472                 hw->phy.mdix = 0; /* AUTO_ALL_MODES */
 473                 hw->phy.disable_polarity_correction = 0;
 474                 hw->phy.ms_type = e1000_ms_hw_default;
 475         }
 476
 477         /*
 478          * Start from a known state, this is important in reading the nvm
 479          * and mac from that.
 480          */
 481         e1000_reset_hw(hw);
 482
 483         /* Make sure we have a good EEPROM before we read from it */
 484         if (e1000_validate_nvm_checksum(hw) < 0) {
 485                 /*
 486                  * Some PCI-E parts fail the first check due to
 487                  * the link being in sleep state, call it again,
 488                  * if it fails a second time its a real issue.
 489                  */
 490                 diag = e1000_validate_nvm_checksum(hw);
 491                 if (diag < 0) {
 492                         PMD_INIT_LOG(ERR, "EEPROM checksum invalid");
 493                         goto error;
 494                 }
 495         }
 496
 497         /* Read the permanent MAC address out of the EEPROM */
 498         diag = e1000_read_mac_addr(hw);
 499         if (diag != 0) {
 500                 PMD_INIT_LOG(ERR, "EEPROM error while reading MAC address");
 501                 goto error;
 502         }
 503
 504         /* Now initialize the hardware */
 505         diag = em_hardware_init(hw);
 506         if (diag != 0) {
 507                 PMD_INIT_LOG(ERR, "Hardware initialization failed");
 508                 goto error;
 509         }
 510
 511         hw->mac.get_link_status = 1;
 512
 513         /* Indicate SOL/IDER usage */
 514         diag = e1000_check_reset_block(hw);
 515         if (diag < 0) {
 516                 PMD_INIT_LOG(ERR, "PHY reset is blocked due to "
 517                         "SOL/IDER session");
 518         }
 519         return 0;
 520
 521 error:
 522         em_hw_control_release(hw);
 523         return diag;
 524 }
 525
 526 static int
 527 eth_em_configure(struct rte_eth_dev *dev)
 528 {
 529         struct e1000_interrupt *intr =
 530                 E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
 531
 532         PMD_INIT_FUNC_TRACE();
 533         intr->flags |= E1000_FLAG_NEED_LINK_UPDATE;
 534         PMD_INIT_FUNC_TRACE();
 535
 536         return 0;
 537 }
 538
 539 static void
 540 em_set_pba(struct e1000_hw *hw)
 541 {
 542         uint32_t pba;
 543
 544         /*
 545          * Packet Buffer Allocation (PBA)
 546          * Writing PBA sets the receive portion of the buffer
 547          * the remainder is used for the transmit buffer.
 548          * Devices before the 82547 had a Packet Buffer of 64K.
 549          * After the 82547 the buffer was reduced to 40K.
 550          */
 551         switch (hw->mac.type) {
 552                 case e1000_82547:
 553                 case e1000_82547_rev_2:
 554                 /* 82547: Total Packet Buffer is 40K */
 555                         pba = E1000_PBA_22K; /* 22K for Rx, 18K for Tx */
 556                         break;
 557                 case e1000_82571:
 558                 case e1000_82572:
 559                 case e1000_80003es2lan:
 560                         pba = E1000_PBA_32K; /* 32K for Rx, 16K for Tx */
 561                         break;
 562                 case e1000_82573: /* 82573: Total Packet Buffer is 32K */
 563                         pba = E1000_PBA_12K; /* 12K for Rx, 20K for Tx */
 564                         break;
 565                 case e1000_82574:
 566                 case e1000_82583:
 567                         pba = E1000_PBA_20K; /* 20K for Rx, 20K for Tx */
 568                         break;
 569                 case e1000_ich8lan:
 570                         pba = E1000_PBA_8K;
 571                         break;
 572                 case e1000_ich9lan:
 573                 case e1000_ich10lan:
 574                         pba = E1000_PBA_10K;
 575                         break;
 576                 case e1000_pchlan:
 577                 case e1000_pch2lan:
 578                 case e1000_pch_lpt:
 579                 case e1000_pch_spt:
 580                 case e1000_pch_cnp:
 581                         pba = E1000_PBA_26K;
 582                         break;
 583                 default:
 584                         pba = E1000_PBA_40K; /* 40K for Rx, 24K for Tx */
 585         }
 586
 587         E1000_WRITE_REG(hw, E1000_PBA, pba);
 588 }
 589
 590 static int
 591 eth_em_start(struct rte_eth_dev *dev)
 592 {
 593         struct e1000_adapter *adapter =
 594                 E1000_DEV_PRIVATE(dev->data->dev_private);
 595         struct e1000_hw *hw =
 596                 E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
 597         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
 598         struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
 599         int ret, mask;
 600         uint32_t intr_vector = 0;
 601         uint32_t *speeds;
 602         int num_speeds;
 603         bool autoneg;
 604
 605         PMD_INIT_FUNC_TRACE();
 606
 607         eth_em_stop(dev);
 608
 609         e1000_power_up_phy(hw);
 610
 611         /* Set default PBA value */
 612         em_set_pba(hw);
 613
 614         /* Put the address into the Receive Address Array */
 615         e1000_rar_set(hw, hw->mac.addr, 0);
 616
 617         /*
 618          * With the 82571 adapter, RAR[0] may be overwritten
 619          * when the other port is reset, we make a duplicate
 620          * in RAR[14] for that eventuality, this assures
 621          * the interface continues to function.
 622          */
 623         if (hw->mac.type == e1000_82571) {
 624                 e1000_set_laa_state_82571(hw, TRUE);
 625                 e1000_rar_set(hw, hw->mac.addr, E1000_RAR_ENTRIES - 1);
 626         }
 627
 628         /* Initialize the hardware */
 629         if (em_hardware_init(hw)) {
 630                 PMD_INIT_LOG(ERR, "Unable to initialize the hardware");
 631                 return -EIO;
 632         }
 633
 634         E1000_WRITE_REG(hw, E1000_VET, ETHER_TYPE_VLAN);
 635
 636         /* Configure for OS presence */
 637         em_init_manageability(hw);
 638
 639         if (dev->data->dev_conf.intr_conf.rxq != 0) {
 640                 intr_vector = dev->data->nb_rx_queues;
 641                 if (rte_intr_efd_enable(intr_handle, intr_vector))
 642                         return -1;
 643         }
 644
 645         if (rte_intr_dp_is_en(intr_handle)) {
 646                 intr_handle->intr_vec =
 647                         rte_zmalloc("intr_vec",
 648                                         dev->data->nb_rx_queues * sizeof(int), 0);
 649                 if (intr_handle->intr_vec == NULL) {
 650                         PMD_INIT_LOG(ERR, "Failed to allocate %d rx_queues"
 651                                                 " intr_vec", dev->data->nb_rx_queues);
 652                         return -ENOMEM;
 653                 }
 654
 655                 /* enable rx interrupt */
 656                 em_rxq_intr_enable(hw);
 657         }
 658
 659         eth_em_tx_init(dev);
 660
 661         ret = eth_em_rx_init(dev);
 662         if (ret) {
 663                 PMD_INIT_LOG(ERR, "Unable to initialize RX hardware");
 664                 em_dev_clear_queues(dev);
 665                 return ret;
 666         }
 667
 668         e1000_clear_hw_cntrs_base_generic(hw);
 669
 670         mask = ETH_VLAN_STRIP_MASK | ETH_VLAN_FILTER_MASK | \
 671                         ETH_VLAN_EXTEND_MASK;
 672         eth_em_vlan_offload_set(dev, mask);
 673
 674         /* Set Interrupt Throttling Rate to maximum allowed value. */
 675         E1000_WRITE_REG(hw, E1000_ITR, UINT16_MAX);
 676
 677         /* Setup link speed and duplex */
 678         speeds = &dev->data->dev_conf.link_speeds;
 679         if (*speeds == ETH_LINK_SPEED_AUTONEG) {
 680                 hw->phy.autoneg_advertised = E1000_ALL_SPEED_DUPLEX;
 681                 hw->mac.autoneg = 1;
 682         } else {
 683                 num_speeds = 0;
 684                 autoneg = (*speeds & ETH_LINK_SPEED_FIXED) == 0;
 685
 686                 /* Reset */
 687                 hw->phy.autoneg_advertised = 0;
 688
 689                 if (*speeds & ~(ETH_LINK_SPEED_10M_HD | ETH_LINK_SPEED_10M |
 690                                 ETH_LINK_SPEED_100M_HD | ETH_LINK_SPEED_100M |
 691                                 ETH_LINK_SPEED_1G | ETH_LINK_SPEED_FIXED)) {
 692                         num_speeds = -1;
 693                         goto error_invalid_config;
 694                 }
 695                 if (*speeds & ETH_LINK_SPEED_10M_HD) {
 696                         hw->phy.autoneg_advertised |= ADVERTISE_10_HALF;
 697                         num_speeds++;
 698                 }
 699                 if (*speeds & ETH_LINK_SPEED_10M) {
 700                         hw->phy.autoneg_advertised |= ADVERTISE_10_FULL;
 701                         num_speeds++;
 702                 }
 703                 if (*speeds & ETH_LINK_SPEED_100M_HD) {
 704                         hw->phy.autoneg_advertised |= ADVERTISE_100_HALF;
 705                         num_speeds++;
 706                 }
 707                 if (*speeds & ETH_LINK_SPEED_100M) {
 708                         hw->phy.autoneg_advertised |= ADVERTISE_100_FULL;
 709                         num_speeds++;
 710                 }
 711                 if (*speeds & ETH_LINK_SPEED_1G) {
 712                         hw->phy.autoneg_advertised |= ADVERTISE_1000_FULL;
 713                         num_speeds++;
 714                 }
 715                 if (num_speeds == 0 || (!autoneg && (num_speeds > 1)))
 716                         goto error_invalid_config;
 717
 718                 /* Set/reset the mac.autoneg based on the link speed,
 719                  * fixed or not
 720                  */
 721                 if (!autoneg) {
 722                         hw->mac.autoneg = 0;
 723                         hw->mac.forced_speed_duplex =
 724                                         hw->phy.autoneg_advertised;
 725                 } else {
 726                         hw->mac.autoneg = 1;
 727                 }
 728         }
 729
 730         e1000_setup_link(hw);
 731
 732         if (rte_intr_allow_others(intr_handle)) {
 733                 /* check if lsc interrupt is enabled */
 734                 if (dev->data->dev_conf.intr_conf.lsc != 0) {
 735                         ret = eth_em_interrupt_setup(dev);
 736                         if (ret) {
 737                                 PMD_INIT_LOG(ERR, "Unable to setup interrupts");
 738                                 em_dev_clear_queues(dev);
 739                                 return ret;
 740                         }
 741                 }
 742         } else {
 743                 rte_intr_callback_unregister(intr_handle,
 744                                                 eth_em_interrupt_handler,
 745                                                 (void *)dev);
 746                 if (dev->data->dev_conf.intr_conf.lsc != 0)
 747                         PMD_INIT_LOG(INFO, "lsc won't enable because of"
 748                                      " no intr multiplexn");
 749         }
 750         /* check if rxq interrupt is enabled */
 751         if (dev->data->dev_conf.intr_conf.rxq != 0)
 752                 eth_em_rxq_interrupt_setup(dev);
 753
 754         rte_intr_enable(intr_handle);
 755
 756         adapter->stopped = 0;
 757
 758         PMD_INIT_LOG(DEBUG, "<<");
 759
 760         return 0;
 761
 762 error_invalid_config:
 763         PMD_INIT_LOG(ERR, "Invalid advertised speeds (%u) for port %u",
 764                      dev->data->dev_conf.link_speeds, dev->data->port_id);
 765         em_dev_clear_queues(dev);
 766         return -EINVAL;
 767 }
 768
 769 /*********************************************************************
 770  *
 771  *  This routine disables all traffic on the adapter by issuing a
 772  *  global reset on the MAC.
 773  *
 774  **********************************************************************/
 775 static void
 776 eth_em_stop(struct rte_eth_dev *dev)
 777 {
 778         struct rte_eth_link link;
 779         struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
 780         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
 781         struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
 782
 783         em_rxq_intr_disable(hw);
 784         em_lsc_intr_disable(hw);
 785
 786         e1000_reset_hw(hw);
 787         if (hw->mac.type >= e1000_82544)
 788                 E1000_WRITE_REG(hw, E1000_WUC, 0);
 789
 790         /* Power down the phy. Needed to make the link go down */
 791         e1000_power_down_phy(hw);
 792
 793         em_dev_clear_queues(dev);
 794
 795         /* clear the recorded link status */
 796         memset(&link, 0, sizeof(link));
 797         rte_em_dev_atomic_write_link_status(dev, &link);
 798
 799         if (!rte_intr_allow_others(intr_handle))
 800                 /* resume to the default handler */
 801                 rte_intr_callback_register(intr_handle,
 802                                            eth_em_interrupt_handler,
 803                                            (void *)dev);
 804
 805         /* Clean datapath event and queue/vec mapping */
 806         rte_intr_efd_disable(intr_handle);
 807         if (intr_handle->intr_vec != NULL) {
 808                 rte_free(intr_handle->intr_vec);
 809                 intr_handle->intr_vec = NULL;
 810         }
 811 }
 812
 813 static void
 814 eth_em_close(struct rte_eth_dev *dev)
 815 {
 816         struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
 817         struct e1000_adapter *adapter =
 818                 E1000_DEV_PRIVATE(dev->data->dev_private);
 819
 820         eth_em_stop(dev);
 821         adapter->stopped = 1;
 822         em_dev_free_queues(dev);
 823         e1000_phy_hw_reset(hw);
 824         em_release_manageability(hw);
 825         em_hw_control_release(hw);
 826 }
 827
 828 static int
 829 em_get_rx_buffer_size(struct e1000_hw *hw)
 830 {
 831         uint32_t rx_buf_size;
 832
 833         rx_buf_size = ((E1000_READ_REG(hw, E1000_PBA) & UINT16_MAX) << 10);
 834         return rx_buf_size;
 835 }
 836
 837 /*********************************************************************
 838  *
 839  *  Initialize the hardware
 840  *
 841  **********************************************************************/
 842 static int
 843 em_hardware_init(struct e1000_hw *hw)
 844 {
 845         uint32_t rx_buf_size;
 846         int diag;
 847
 848         /* Issue a global reset */
 849         e1000_reset_hw(hw);
 850
 851         /* Let the firmware know the OS is in control */
 852         em_hw_control_acquire(hw);
 853
 854         /*
 855          * These parameters control the automatic generation (Tx) and
 856          * response (Rx) to Ethernet PAUSE frames.
 857          * - High water mark should allow for at least two standard size (1518)
 858          *   frames to be received after sending an XOFF.
 859          * - Low water mark works best when it is very near the high water mark.
 860          *   This allows the receiver to restart by sending XON when it has
 861          *   drained a bit. Here we use an arbitrary value of 1500 which will
 862          *   restart after one full frame is pulled from the buffer. There
 863          *   could be several smaller frames in the buffer and if so they will
 864          *   not trigger the XON until their total number reduces the buffer
 865          *   by 1500.
 866          * - The pause time is fairly large at 1000 x 512ns = 512 usec.
 867          */
 868         rx_buf_size = em_get_rx_buffer_size(hw);
 869
 870         hw->fc.high_water = rx_buf_size - PMD_ROUNDUP(ETHER_MAX_LEN * 2, 1024);
 871         hw->fc.low_water = hw->fc.high_water - 1500;
 872
 873         if (hw->mac.type == e1000_80003es2lan)
 874                 hw->fc.pause_time = UINT16_MAX;
 875         else
 876                 hw->fc.pause_time = EM_FC_PAUSE_TIME;
 877
 878         hw->fc.send_xon = 1;
 879
 880         /* Set Flow control, use the tunable location if sane */
 881         if (em_fc_setting <= e1000_fc_full)
 882                 hw->fc.requested_mode = em_fc_setting;
 883         else
 884                 hw->fc.requested_mode = e1000_fc_none;
 885
 886         /* Workaround: no TX flow ctrl for PCH */
 887         if (hw->mac.type == e1000_pchlan)
 888                 hw->fc.requested_mode = e1000_fc_rx_pause;
 889
 890         /* Override - settings for PCH2LAN, ya its magic :) */
 891         if (hw->mac.type == e1000_pch2lan) {
 892                 hw->fc.high_water = 0x5C20;
 893                 hw->fc.low_water = 0x5048;
 894                 hw->fc.pause_time = 0x0650;
 895                 hw->fc.refresh_time = 0x0400;
 896         } else if (hw->mac.type == e1000_pch_lpt ||
 897                    hw->mac.type == e1000_pch_spt ||
 898                    hw->mac.type == e1000_pch_cnp) {
 899                 hw->fc.requested_mode = e1000_fc_full;
 900         }
 901
 902         diag = e1000_init_hw(hw);
 903         if (diag < 0)
 904                 return diag;
 905         e1000_check_for_link(hw);
 906         return 0;
 907 }
 908
 909 /* This function is based on em_update_stats_counters() in e1000/if_em.c */
 910 static int
 911 eth_em_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *rte_stats)
 912 {
 913         struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
 914         struct e1000_hw_stats *stats =
 915                         E1000_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
 916         int pause_frames;
 917
 918         if(hw->phy.media_type == e1000_media_type_copper ||
 919                         (E1000_READ_REG(hw, E1000_STATUS) & E1000_STATUS_LU)) {
 920                 stats->symerrs += E1000_READ_REG(hw,E1000_SYMERRS);
 921                 stats->sec += E1000_READ_REG(hw, E1000_SEC);
 922         }
 923
 924         stats->crcerrs += E1000_READ_REG(hw, E1000_CRCERRS);
 925         stats->mpc += E1000_READ_REG(hw, E1000_MPC);
 926         stats->scc += E1000_READ_REG(hw, E1000_SCC);
 927         stats->ecol += E1000_READ_REG(hw, E1000_ECOL);
 928
 929         stats->mcc += E1000_READ_REG(hw, E1000_MCC);
 930         stats->latecol += E1000_READ_REG(hw, E1000_LATECOL);
 931         stats->colc += E1000_READ_REG(hw, E1000_COLC);
 932         stats->dc += E1000_READ_REG(hw, E1000_DC);
 933         stats->rlec += E1000_READ_REG(hw, E1000_RLEC);
 934         stats->xonrxc += E1000_READ_REG(hw, E1000_XONRXC);
 935         stats->xontxc += E1000_READ_REG(hw, E1000_XONTXC);
 936
 937         /*
 938          * For watchdog management we need to know if we have been
 939          * paused during the last interval, so capture that here.
 940          */
 941         pause_frames = E1000_READ_REG(hw, E1000_XOFFRXC);
 942         stats->xoffrxc += pause_frames;
 943         stats->xofftxc += E1000_READ_REG(hw, E1000_XOFFTXC);
 944         stats->fcruc += E1000_READ_REG(hw, E1000_FCRUC);
 945         stats->prc64 += E1000_READ_REG(hw, E1000_PRC64);
 946         stats->prc127 += E1000_READ_REG(hw, E1000_PRC127);
 947         stats->prc255 += E1000_READ_REG(hw, E1000_PRC255);
 948         stats->prc511 += E1000_READ_REG(hw, E1000_PRC511);
 949         stats->prc1023 += E1000_READ_REG(hw, E1000_PRC1023);
 950         stats->prc1522 += E1000_READ_REG(hw, E1000_PRC1522);
 951         stats->gprc += E1000_READ_REG(hw, E1000_GPRC);
 952         stats->bprc += E1000_READ_REG(hw, E1000_BPRC);
 953         stats->mprc += E1000_READ_REG(hw, E1000_MPRC);
 954         stats->gptc += E1000_READ_REG(hw, E1000_GPTC);
 955
 956         /*
 957          * For the 64-bit byte counters the low dword must be read first.
 958          * Both registers clear on the read of the high dword.
 959          */
 960
 961         stats->gorc += E1000_READ_REG(hw, E1000_GORCL);
 962         stats->gorc += ((uint64_t)E1000_READ_REG(hw, E1000_GORCH) << 32);
 963         stats->gotc += E1000_READ_REG(hw, E1000_GOTCL);
 964         stats->gotc += ((uint64_t)E1000_READ_REG(hw, E1000_GOTCH) << 32);
 965
 966         stats->rnbc += E1000_READ_REG(hw, E1000_RNBC);
 967         stats->ruc += E1000_READ_REG(hw, E1000_RUC);
 968         stats->rfc += E1000_READ_REG(hw, E1000_RFC);
 969         stats->roc += E1000_READ_REG(hw, E1000_ROC);
 970         stats->rjc += E1000_READ_REG(hw, E1000_RJC);
 971
 972         stats->tor += E1000_READ_REG(hw, E1000_TORH);
 973         stats->tot += E1000_READ_REG(hw, E1000_TOTH);
 974
 975         stats->tpr += E1000_READ_REG(hw, E1000_TPR);
 976         stats->tpt += E1000_READ_REG(hw, E1000_TPT);
 977         stats->ptc64 += E1000_READ_REG(hw, E1000_PTC64);
 978         stats->ptc127 += E1000_READ_REG(hw, E1000_PTC127);
 979         stats->ptc255 += E1000_READ_REG(hw, E1000_PTC255);
 980         stats->ptc511 += E1000_READ_REG(hw, E1000_PTC511);
 981         stats->ptc1023 += E1000_READ_REG(hw, E1000_PTC1023);
 982         stats->ptc1522 += E1000_READ_REG(hw, E1000_PTC1522);
 983         stats->mptc += E1000_READ_REG(hw, E1000_MPTC);
 984         stats->bptc += E1000_READ_REG(hw, E1000_BPTC);
 985
 986         /* Interrupt Counts */
 987
 988         if (hw->mac.type >= e1000_82571) {
 989                 stats->iac += E1000_READ_REG(hw, E1000_IAC);
 990                 stats->icrxptc += E1000_READ_REG(hw, E1000_ICRXPTC);
 991                 stats->icrxatc += E1000_READ_REG(hw, E1000_ICRXATC);
 992                 stats->ictxptc += E1000_READ_REG(hw, E1000_ICTXPTC);
 993                 stats->ictxatc += E1000_READ_REG(hw, E1000_ICTXATC);
 994                 stats->ictxqec += E1000_READ_REG(hw, E1000_ICTXQEC);
 995                 stats->ictxqmtc += E1000_READ_REG(hw, E1000_ICTXQMTC);
 996                 stats->icrxdmtc += E1000_READ_REG(hw, E1000_ICRXDMTC);
 997                 stats->icrxoc += E1000_READ_REG(hw, E1000_ICRXOC);
 998         }
 999
1000         if (hw->mac.type >= e1000_82543) {
1001                 stats->algnerrc += E1000_READ_REG(hw, E1000_ALGNERRC);
1002                 stats->rxerrc += E1000_READ_REG(hw, E1000_RXERRC);
1003                 stats->tncrs += E1000_READ_REG(hw, E1000_TNCRS);
1004                 stats->cexterr += E1000_READ_REG(hw, E1000_CEXTERR);
1005                 stats->tsctc += E1000_READ_REG(hw, E1000_TSCTC);
1006                 stats->tsctfc += E1000_READ_REG(hw, E1000_TSCTFC);
1007         }
1008
1009         if (rte_stats == NULL)
1010                 return -EINVAL;
1011
1012         /* Rx Errors */
1013         rte_stats->imissed = stats->mpc;
1014         rte_stats->ierrors = stats->crcerrs +
1015                              stats->rlec + stats->ruc + stats->roc +
1016                              stats->rxerrc + stats->algnerrc + stats->cexterr;
1017
1018         /* Tx Errors */
1019         rte_stats->oerrors = stats->ecol + stats->latecol;
1020
1021         rte_stats->ipackets = stats->gprc;
1022         rte_stats->opackets = stats->gptc;
1023         rte_stats->ibytes   = stats->gorc;
1024         rte_stats->obytes   = stats->gotc;
1025         return 0;
1026 }
1027
1028 static void
1029 eth_em_stats_reset(struct rte_eth_dev *dev)
1030 {
1031         struct e1000_hw_stats *hw_stats =
1032                         E1000_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
1033
1034         /* HW registers are cleared on read */
1035         eth_em_stats_get(dev, NULL);
1036
1037         /* Reset software totals */
1038         memset(hw_stats, 0, sizeof(*hw_stats));
1039 }
1040
1041 static int
1042 eth_em_rx_queue_intr_enable(struct rte_eth_dev *dev, __rte_unused uint16_t queue_id)
1043 {
1044         struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1045         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1046         struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
1047
1048         em_rxq_intr_enable(hw);
1049         rte_intr_enable(intr_handle);
1050
1051         return 0;
1052 }
1053
1054 static int
1055 eth_em_rx_queue_intr_disable(struct rte_eth_dev *dev, __rte_unused uint16_t queue_id)
1056 {
1057         struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1058
1059         em_rxq_intr_disable(hw);
1060
1061         return 0;
1062 }
1063
1064 static uint32_t
1065 em_get_max_pktlen(const struct e1000_hw *hw)
1066 {
1067         switch (hw->mac.type) {
1068         case e1000_82571:
1069         case e1000_82572:
1070         case e1000_ich9lan:
1071         case e1000_ich10lan:
1072         case e1000_pch2lan:
1073         case e1000_pch_lpt:
1074         case e1000_pch_spt:
1075         case e1000_pch_cnp:
1076         case e1000_82574:
1077         case e1000_80003es2lan: /* 9K Jumbo Frame size */
1078         case e1000_82583:
1079                 return 0x2412;
1080         case e1000_pchlan:
1081                 return 0x1000;
1082         /* Adapters that do not support jumbo frames */
1083         case e1000_ich8lan:
1084                 return ETHER_MAX_LEN;
1085         default:
1086                 return MAX_JUMBO_FRAME_SIZE;
1087         }
1088 }
1089
1090 static void
1091 eth_em_infos_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
1092 {
1093         struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1094
1095         dev_info->pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1096         dev_info->min_rx_bufsize = 256; /* See BSIZE field of RCTL register. */
1097         dev_info->max_rx_pktlen = em_get_max_pktlen(hw);
1098         dev_info->max_mac_addrs = hw->mac.rar_entry_count;
1099         dev_info->rx_offload_capa =
1100                 DEV_RX_OFFLOAD_VLAN_STRIP |
1101                 DEV_RX_OFFLOAD_IPV4_CKSUM |
1102                 DEV_RX_OFFLOAD_UDP_CKSUM  |
1103                 DEV_RX_OFFLOAD_TCP_CKSUM;
1104         dev_info->tx_offload_capa =
1105                 DEV_TX_OFFLOAD_VLAN_INSERT |
1106                 DEV_TX_OFFLOAD_IPV4_CKSUM  |
1107                 DEV_TX_OFFLOAD_UDP_CKSUM   |
1108                 DEV_TX_OFFLOAD_TCP_CKSUM;
1109
1110         /*
1111          * Starting with 631xESB hw supports 2 TX/RX queues per port.
1112          * Unfortunatelly, all these nics have just one TX context.
1113          * So we have few choises for TX:
1114          * - Use just one TX queue.
1115          * - Allow cksum offload only for one TX queue.
1116          * - Don't allow TX cksum offload at all.
1117          * For now, option #1 was chosen.
1118          * To use second RX queue we have to use extended RX descriptor
1119          * (Multiple Receive Queues are mutually exclusive with UDP
1120          * fragmentation and are not supported when a legacy receive
1121          * descriptor format is used).
1122          * Which means separate RX routinies - as legacy nics (82540, 82545)
1123          * don't support extended RXD.
1124          * To avoid it we support just one RX queue for now (no RSS).
1125          */
1126
1127         dev_info->max_rx_queues = 1;
1128         dev_info->max_tx_queues = 1;
1129
1130         dev_info->rx_desc_lim = (struct rte_eth_desc_lim) {
1131                 .nb_max = E1000_MAX_RING_DESC,
1132                 .nb_min = E1000_MIN_RING_DESC,
1133                 .nb_align = EM_RXD_ALIGN,
1134         };
1135
1136         dev_info->tx_desc_lim = (struct rte_eth_desc_lim) {
1137                 .nb_max = E1000_MAX_RING_DESC,
1138                 .nb_min = E1000_MIN_RING_DESC,
1139                 .nb_align = EM_TXD_ALIGN,
1140                 .nb_seg_max = EM_TX_MAX_SEG,
1141                 .nb_mtu_seg_max = EM_TX_MAX_MTU_SEG,
1142         };
1143
1144         dev_info->speed_capa = ETH_LINK_SPEED_10M_HD | ETH_LINK_SPEED_10M |
1145                         ETH_LINK_SPEED_100M_HD | ETH_LINK_SPEED_100M |
1146                         ETH_LINK_SPEED_1G;
1147 }
1148
1149 /* return 0 means link status changed, -1 means not changed */
1150 static int
1151 eth_em_link_update(struct rte_eth_dev *dev, int wait_to_complete)
1152 {
1153         struct e1000_hw *hw =
1154                 E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1155         struct rte_eth_link link, old;
1156         int link_check, count;
1157
1158         link_check = 0;
1159         hw->mac.get_link_status = 1;
1160
1161         /* possible wait-to-complete in up to 9 seconds */
1162         for (count = 0; count < EM_LINK_UPDATE_CHECK_TIMEOUT; count ++) {
1163                 /* Read the real link status */
1164                 switch (hw->phy.media_type) {
1165                 case e1000_media_type_copper:
1166                         /* Do the work to read phy */
1167                         e1000_check_for_link(hw);
1168                         link_check = !hw->mac.get_link_status;
1169                         break;
1170
1171                 case e1000_media_type_fiber:
1172                         e1000_check_for_link(hw);
1173                         link_check = (E1000_READ_REG(hw, E1000_STATUS) &
1174                                         E1000_STATUS_LU);
1175                         break;
1176
1177                 case e1000_media_type_internal_serdes:
1178                         e1000_check_for_link(hw);
1179                         link_check = hw->mac.serdes_has_link;
1180                         break;
1181
1182                 default:
1183                         break;
1184                 }
1185                 if (link_check || wait_to_complete == 0)
1186                         break;
1187                 rte_delay_ms(EM_LINK_UPDATE_CHECK_INTERVAL);
1188         }
1189         memset(&link, 0, sizeof(link));
1190         rte_em_dev_atomic_read_link_status(dev, &link);
1191         old = link;
1192
1193         /* Now we check if a transition has happened */
1194         if (link_check && (link.link_status == ETH_LINK_DOWN)) {
1195                 uint16_t duplex, speed;
1196                 hw->mac.ops.get_link_up_info(hw, &speed, &duplex);
1197                 link.link_duplex = (duplex == FULL_DUPLEX) ?
1198                                 ETH_LINK_FULL_DUPLEX :
1199                                 ETH_LINK_HALF_DUPLEX;
1200                 link.link_speed = speed;
1201                 link.link_status = ETH_LINK_UP;
1202                 link.link_autoneg = !(dev->data->dev_conf.link_speeds &
1203                                 ETH_LINK_SPEED_FIXED);
1204         } else if (!link_check && (link.link_status == ETH_LINK_UP)) {
1205                 link.link_speed = 0;
1206                 link.link_duplex = ETH_LINK_HALF_DUPLEX;
1207                 link.link_status = ETH_LINK_DOWN;
1208                 link.link_autoneg = ETH_LINK_SPEED_FIXED;
1209         }
1210         rte_em_dev_atomic_write_link_status(dev, &link);
1211
1212         /* not changed */
1213         if (old.link_status == link.link_status)
1214                 return -1;
1215
1216         /* changed */
1217         return 0;
1218 }
1219
1220 /*
1221  * em_hw_control_acquire sets {CTRL_EXT|FWSM}:DRV_LOAD bit.
1222  * For ASF and Pass Through versions of f/w this means
1223  * that the driver is loaded. For AMT version type f/w
1224  * this means that the network i/f is open.
1225  */
1226 static void
1227 em_hw_control_acquire(struct e1000_hw *hw)
1228 {
1229         uint32_t ctrl_ext, swsm;
1230
1231         /* Let firmware know the driver has taken over */
1232         if (hw->mac.type == e1000_82573) {
1233                 swsm = E1000_READ_REG(hw, E1000_SWSM);
1234                 E1000_WRITE_REG(hw, E1000_SWSM, swsm | E1000_SWSM_DRV_LOAD);
1235
1236         } else {
1237                 ctrl_ext = E1000_READ_REG(hw, E1000_CTRL_EXT);
1238                 E1000_WRITE_REG(hw, E1000_CTRL_EXT,
1239                         ctrl_ext | E1000_CTRL_EXT_DRV_LOAD);
1240         }
1241 }
1242
1243 /*
1244  * em_hw_control_release resets {CTRL_EXTT|FWSM}:DRV_LOAD bit.
1245  * For ASF and Pass Through versions of f/w this means that the
1246  * driver is no longer loaded. For AMT versions of the
1247  * f/w this means that the network i/f is closed.
1248  */
1249 static void
1250 em_hw_control_release(struct e1000_hw *hw)
1251 {
1252         uint32_t ctrl_ext, swsm;
1253
1254         /* Let firmware taken over control of h/w */
1255         if (hw->mac.type == e1000_82573) {
1256                 swsm = E1000_READ_REG(hw, E1000_SWSM);
1257                 E1000_WRITE_REG(hw, E1000_SWSM, swsm & ~E1000_SWSM_DRV_LOAD);
1258         } else {
1259                 ctrl_ext = E1000_READ_REG(hw, E1000_CTRL_EXT);
1260                 E1000_WRITE_REG(hw, E1000_CTRL_EXT,
1261                         ctrl_ext & ~E1000_CTRL_EXT_DRV_LOAD);
1262         }
1263 }
1264
1265 /*
1266  * Bit of a misnomer, what this really means is
1267  * to enable OS management of the system... aka
1268  * to disable special hardware management features.
1269  */
1270 static void
1271 em_init_manageability(struct e1000_hw *hw)
1272 {
1273         if (e1000_enable_mng_pass_thru(hw)) {
1274                 uint32_t manc2h = E1000_READ_REG(hw, E1000_MANC2H);
1275                 uint32_t manc = E1000_READ_REG(hw, E1000_MANC);
1276
1277                 /* disable hardware interception of ARP */
1278                 manc &= ~(E1000_MANC_ARP_EN);
1279
1280                 /* enable receiving management packets to the host */
1281                 manc |= E1000_MANC_EN_MNG2HOST;
1282                 manc2h |= 1 << 5;  /* Mng Port 623 */
1283                 manc2h |= 1 << 6;  /* Mng Port 664 */
1284                 E1000_WRITE_REG(hw, E1000_MANC2H, manc2h);
1285                 E1000_WRITE_REG(hw, E1000_MANC, manc);
1286         }
1287 }
1288
1289 /*
1290  * Give control back to hardware management
1291  * controller if there is one.
1292  */
1293 static void
1294 em_release_manageability(struct e1000_hw *hw)
1295 {
1296         uint32_t manc;
1297
1298         if (e1000_enable_mng_pass_thru(hw)) {
1299                 manc = E1000_READ_REG(hw, E1000_MANC);
1300
1301                 /* re-enable hardware interception of ARP */
1302                 manc |= E1000_MANC_ARP_EN;
1303                 manc &= ~E1000_MANC_EN_MNG2HOST;
1304
1305                 E1000_WRITE_REG(hw, E1000_MANC, manc);
1306         }
1307 }
1308
1309 static void
1310 eth_em_promiscuous_enable(struct rte_eth_dev *dev)
1311 {
1312         struct e1000_hw *hw =
1313                 E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1314         uint32_t rctl;
1315
1316         rctl = E1000_READ_REG(hw, E1000_RCTL);
1317         rctl |= (E1000_RCTL_UPE | E1000_RCTL_MPE);
1318         E1000_WRITE_REG(hw, E1000_RCTL, rctl);
1319 }
1320
1321 static void
1322 eth_em_promiscuous_disable(struct rte_eth_dev *dev)
1323 {
1324         struct e1000_hw *hw =
1325                 E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1326         uint32_t rctl;
1327
1328         rctl = E1000_READ_REG(hw, E1000_RCTL);
1329         rctl &= ~(E1000_RCTL_UPE | E1000_RCTL_SBP);
1330         if (dev->data->all_multicast == 1)
1331                 rctl |= E1000_RCTL_MPE;
1332         else
1333                 rctl &= (~E1000_RCTL_MPE);
1334         E1000_WRITE_REG(hw, E1000_RCTL, rctl);
1335 }
1336
1337 static void
1338 eth_em_allmulticast_enable(struct rte_eth_dev *dev)
1339 {
1340         struct e1000_hw *hw =
1341                 E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1342         uint32_t rctl;
1343
1344         rctl = E1000_READ_REG(hw, E1000_RCTL);
1345         rctl |= E1000_RCTL_MPE;
1346         E1000_WRITE_REG(hw, E1000_RCTL, rctl);
1347 }
1348
1349 static void
1350 eth_em_allmulticast_disable(struct rte_eth_dev *dev)
1351 {
1352         struct e1000_hw *hw =
1353                 E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1354         uint32_t rctl;
1355
1356         if (dev->data->promiscuous == 1)
1357                 return; /* must remain in all_multicast mode */
1358         rctl = E1000_READ_REG(hw, E1000_RCTL);
1359         rctl &= (~E1000_RCTL_MPE);
1360         E1000_WRITE_REG(hw, E1000_RCTL, rctl);
1361 }
1362
1363 static int
1364 eth_em_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
1365 {
1366         struct e1000_hw *hw =
1367                 E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1368         struct e1000_vfta * shadow_vfta =
1369                 E1000_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
1370         uint32_t vfta;
1371         uint32_t vid_idx;
1372         uint32_t vid_bit;
1373
1374         vid_idx = (uint32_t) ((vlan_id >> E1000_VFTA_ENTRY_SHIFT) &
1375                               E1000_VFTA_ENTRY_MASK);
1376         vid_bit = (uint32_t) (1 << (vlan_id & E1000_VFTA_ENTRY_BIT_SHIFT_MASK));
1377         vfta = E1000_READ_REG_ARRAY(hw, E1000_VFTA, vid_idx);
1378         if (on)
1379                 vfta |= vid_bit;
1380         else
1381                 vfta &= ~vid_bit;
1382         E1000_WRITE_REG_ARRAY(hw, E1000_VFTA, vid_idx, vfta);
1383
1384         /* update local VFTA copy */
1385         shadow_vfta->vfta[vid_idx] = vfta;
1386
1387         return 0;
1388 }
1389
1390 static void
1391 em_vlan_hw_filter_disable(struct rte_eth_dev *dev)
1392 {
1393         struct e1000_hw *hw =
1394                 E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1395         uint32_t reg;
1396
1397         /* Filter Table Disable */
1398         reg = E1000_READ_REG(hw, E1000_RCTL);
1399         reg &= ~E1000_RCTL_CFIEN;
1400         reg &= ~E1000_RCTL_VFE;
1401         E1000_WRITE_REG(hw, E1000_RCTL, reg);
1402 }
1403
1404 static void
1405 em_vlan_hw_filter_enable(struct rte_eth_dev *dev)
1406 {
1407         struct e1000_hw *hw =
1408                 E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1409         struct e1000_vfta * shadow_vfta =
1410                 E1000_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
1411         uint32_t reg;
1412         int i;
1413
1414         /* Filter Table Enable, CFI not used for packet acceptance */
1415         reg = E1000_READ_REG(hw, E1000_RCTL);
1416         reg &= ~E1000_RCTL_CFIEN;
1417         reg |= E1000_RCTL_VFE;
1418         E1000_WRITE_REG(hw, E1000_RCTL, reg);
1419
1420         /* restore vfta from local copy */
1421         for (i = 0; i < IGB_VFTA_SIZE; i++)
1422                 E1000_WRITE_REG_ARRAY(hw, E1000_VFTA, i, shadow_vfta->vfta[i]);
1423 }
1424
1425 static void
1426 em_vlan_hw_strip_disable(struct rte_eth_dev *dev)
1427 {
1428         struct e1000_hw *hw =
1429                 E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1430         uint32_t reg;
1431
1432         /* VLAN Mode Disable */
1433         reg = E1000_READ_REG(hw, E1000_CTRL);
1434         reg &= ~E1000_CTRL_VME;
1435         E1000_WRITE_REG(hw, E1000_CTRL, reg);
1436
1437 }
1438
1439 static void
1440 em_vlan_hw_strip_enable(struct rte_eth_dev *dev)
1441 {
1442         struct e1000_hw *hw =
1443                 E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1444         uint32_t reg;
1445
1446         /* VLAN Mode Enable */
1447         reg = E1000_READ_REG(hw, E1000_CTRL);
1448         reg |= E1000_CTRL_VME;
1449         E1000_WRITE_REG(hw, E1000_CTRL, reg);
1450 }
1451
1452 static void
1453 eth_em_vlan_offload_set(struct rte_eth_dev *dev, int mask)
1454 {
1455         if(mask & ETH_VLAN_STRIP_MASK){
1456                 if (dev->data->dev_conf.rxmode.hw_vlan_strip)
1457                         em_vlan_hw_strip_enable(dev);
1458                 else
1459                         em_vlan_hw_strip_disable(dev);
1460         }
1461
1462         if(mask & ETH_VLAN_FILTER_MASK){
1463                 if (dev->data->dev_conf.rxmode.hw_vlan_filter)
1464                         em_vlan_hw_filter_enable(dev);
1465                 else
1466                         em_vlan_hw_filter_disable(dev);
1467         }
1468 }
1469
1470 /*
1471  * It enables the interrupt mask and then enable the interrupt.
1472  *
1473  * @param dev
1474  *  Pointer to struct rte_eth_dev.
1475  *
1476  * @return
1477  *  - On success, zero.
1478  *  - On failure, a negative value.
1479  */
1480 static int
1481 eth_em_interrupt_setup(struct rte_eth_dev *dev)
1482 {
1483         uint32_t regval;
1484         struct e1000_hw *hw =
1485                 E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1486
1487         /* clear interrupt */
1488         E1000_READ_REG(hw, E1000_ICR);
1489         regval = E1000_READ_REG(hw, E1000_IMS);
1490         E1000_WRITE_REG(hw, E1000_IMS, regval | E1000_ICR_LSC);
1491         return 0;
1492 }
1493
1494 /*
1495  * It clears the interrupt causes and enables the interrupt.
1496  * It will be called once only during nic initialized.
1497  *
1498  * @param dev
1499  *  Pointer to struct rte_eth_dev.
1500  *
1501  * @return
1502  *  - On success, zero.
1503  *  - On failure, a negative value.
1504  */
1505 static int
1506 eth_em_rxq_interrupt_setup(struct rte_eth_dev *dev)
1507 {
1508         struct e1000_hw *hw =
1509         E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1510
1511         E1000_READ_REG(hw, E1000_ICR);
1512         em_rxq_intr_enable(hw);
1513         return 0;
1514 }
1515
1516 /*
1517  * It enable receive packet interrupt.
1518  * @param hw
1519  * Pointer to struct e1000_hw
1520  *
1521  * @return
1522  */
1523 static void
1524 em_rxq_intr_enable(struct e1000_hw *hw)
1525 {
1526         E1000_WRITE_REG(hw, E1000_IMS, E1000_IMS_RXT0);
1527         E1000_WRITE_FLUSH(hw);
1528 }
1529
1530 /*
1531  * It disabled lsc interrupt.
1532  * @param hw
1533  * Pointer to struct e1000_hw
1534  *
1535  * @return
1536  */
1537 static void
1538 em_lsc_intr_disable(struct e1000_hw *hw)
1539 {
1540         E1000_WRITE_REG(hw, E1000_IMC, E1000_IMS_LSC);
1541         E1000_WRITE_FLUSH(hw);
1542 }
1543
1544 /*
1545  * It disabled receive packet interrupt.
1546  * @param hw
1547  * Pointer to struct e1000_hw
1548  *
1549  * @return
1550  */
1551 static void
1552 em_rxq_intr_disable(struct e1000_hw *hw)
1553 {
1554         E1000_READ_REG(hw, E1000_ICR);
1555         E1000_WRITE_REG(hw, E1000_IMC, E1000_IMS_RXT0);
1556         E1000_WRITE_FLUSH(hw);
1557 }
1558
1559 /*
1560  * It reads ICR and gets interrupt causes, check it and set a bit flag
1561  * to update link status.
1562  *
1563  * @param dev
1564  *  Pointer to struct rte_eth_dev.
1565  *
1566  * @return
1567  *  - On success, zero.
1568  *  - On failure, a negative value.
1569  */
1570 static int
1571 eth_em_interrupt_get_status(struct rte_eth_dev *dev)
1572 {
1573         uint32_t icr;
1574         struct e1000_hw *hw =
1575                 E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1576         struct e1000_interrupt *intr =
1577                 E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
1578
1579         /* read-on-clear nic registers here */
1580         icr = E1000_READ_REG(hw, E1000_ICR);
1581         if (icr & E1000_ICR_LSC) {
1582                 intr->flags |= E1000_FLAG_NEED_LINK_UPDATE;
1583         }
1584
1585         return 0;
1586 }
1587
1588 /*
1589  * It executes link_update after knowing an interrupt is prsent.
1590  *
1591  * @param dev
1592  *  Pointer to struct rte_eth_dev.
1593  *
1594  * @return
1595  *  - On success, zero.
1596  *  - On failure, a negative value.
1597  */
1598 static int
1599 eth_em_interrupt_action(struct rte_eth_dev *dev,
1600                         struct rte_intr_handle *intr_handle)
1601 {
1602         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1603         struct e1000_hw *hw =
1604                 E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1605         struct e1000_interrupt *intr =
1606                 E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
1607         uint32_t tctl, rctl;
1608         struct rte_eth_link link;
1609         int ret;
1610
1611         if (!(intr->flags & E1000_FLAG_NEED_LINK_UPDATE))
1612                 return -1;
1613
1614         intr->flags &= ~E1000_FLAG_NEED_LINK_UPDATE;
1615         rte_intr_enable(intr_handle);
1616
1617         /* set get_link_status to check register later */
1618         hw->mac.get_link_status = 1;
1619         ret = eth_em_link_update(dev, 0);
1620
1621         /* check if link has changed */
1622         if (ret < 0)
1623                 return 0;
1624
1625         memset(&link, 0, sizeof(link));
1626         rte_em_dev_atomic_read_link_status(dev, &link);
1627         if (link.link_status) {
1628                 PMD_INIT_LOG(INFO, " Port %d: Link Up - speed %u Mbps - %s",
1629                              dev->data->port_id, link.link_speed,
1630                              link.link_duplex == ETH_LINK_FULL_DUPLEX ?
1631                              "full-duplex" : "half-duplex");
1632         } else {
1633                 PMD_INIT_LOG(INFO, " Port %d: Link Down", dev->data->port_id);
1634         }
1635         PMD_INIT_LOG(DEBUG, "PCI Address: %04d:%02d:%02d:%d",
1636                      pci_dev->addr.domain, pci_dev->addr.bus,
1637                      pci_dev->addr.devid, pci_dev->addr.function);
1638
1639         tctl = E1000_READ_REG(hw, E1000_TCTL);
1640         rctl = E1000_READ_REG(hw, E1000_RCTL);
1641         if (link.link_status) {
1642                 /* enable Tx/Rx */
1643                 tctl |= E1000_TCTL_EN;
1644                 rctl |= E1000_RCTL_EN;
1645         } else {
1646                 /* disable Tx/Rx */
1647                 tctl &= ~E1000_TCTL_EN;
1648                 rctl &= ~E1000_RCTL_EN;
1649         }
1650         E1000_WRITE_REG(hw, E1000_TCTL, tctl);
1651         E1000_WRITE_REG(hw, E1000_RCTL, rctl);
1652         E1000_WRITE_FLUSH(hw);
1653
1654         return 0;
1655 }
1656
1657 /**
1658  * Interrupt handler which shall be registered at first.
1659  *
1660  * @param handle
1661  *  Pointer to interrupt handle.
1662  * @param param
1663  *  The address of parameter (struct rte_eth_dev *) regsitered before.
1664  *
1665  * @return
1666  *  void
1667  */
1668 static void
1669 eth_em_interrupt_handler(void *param)
1670 {
1671         struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
1672
1673         eth_em_interrupt_get_status(dev);
1674         eth_em_interrupt_action(dev, dev->intr_handle);
1675         _rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_LSC, NULL, NULL);
1676 }
1677
1678 static int
1679 eth_em_led_on(struct rte_eth_dev *dev)
1680 {
1681         struct e1000_hw *hw;
1682
1683         hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1684         return e1000_led_on(hw) == E1000_SUCCESS ? 0 : -ENOTSUP;
1685 }
1686
1687 static int
1688 eth_em_led_off(struct rte_eth_dev *dev)
1689 {
1690         struct e1000_hw *hw;
1691
1692         hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1693         return e1000_led_off(hw) == E1000_SUCCESS ? 0 : -ENOTSUP;
1694 }
1695
1696 static int
1697 eth_em_flow_ctrl_get(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
1698 {
1699         struct e1000_hw *hw;
1700         uint32_t ctrl;
1701         int tx_pause;
1702         int rx_pause;
1703
1704         hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1705         fc_conf->pause_time = hw->fc.pause_time;
1706         fc_conf->high_water = hw->fc.high_water;
1707         fc_conf->low_water = hw->fc.low_water;
1708         fc_conf->send_xon = hw->fc.send_xon;
1709         fc_conf->autoneg = hw->mac.autoneg;
1710
1711         /*
1712          * Return rx_pause and tx_pause status according to actual setting of
1713          * the TFCE and RFCE bits in the CTRL register.
1714          */
1715         ctrl = E1000_READ_REG(hw, E1000_CTRL);
1716         if (ctrl & E1000_CTRL_TFCE)
1717                 tx_pause = 1;
1718         else
1719                 tx_pause = 0;
1720
1721         if (ctrl & E1000_CTRL_RFCE)
1722                 rx_pause = 1;
1723         else
1724                 rx_pause = 0;
1725
1726         if (rx_pause && tx_pause)
1727                 fc_conf->mode = RTE_FC_FULL;
1728         else if (rx_pause)
1729                 fc_conf->mode = RTE_FC_RX_PAUSE;
1730         else if (tx_pause)
1731                 fc_conf->mode = RTE_FC_TX_PAUSE;
1732         else
1733                 fc_conf->mode = RTE_FC_NONE;
1734
1735         return 0;
1736 }
1737
1738 static int
1739 eth_em_flow_ctrl_set(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
1740 {
1741         struct e1000_hw *hw;
1742         int err;
1743         enum e1000_fc_mode rte_fcmode_2_e1000_fcmode[] = {
1744                 e1000_fc_none,
1745                 e1000_fc_rx_pause,
1746                 e1000_fc_tx_pause,
1747                 e1000_fc_full
1748         };
1749         uint32_t rx_buf_size;
1750         uint32_t max_high_water;
1751         uint32_t rctl;
1752
1753         hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1754         if (fc_conf->autoneg != hw->mac.autoneg)
1755                 return -ENOTSUP;
1756         rx_buf_size = em_get_rx_buffer_size(hw);
1757         PMD_INIT_LOG(DEBUG, "Rx packet buffer size = 0x%x", rx_buf_size);
1758
1759         /* At least reserve one Ethernet frame for watermark */
1760         max_high_water = rx_buf_size - ETHER_MAX_LEN;
1761         if ((fc_conf->high_water > max_high_water) ||
1762             (fc_conf->high_water < fc_conf->low_water)) {
1763                 PMD_INIT_LOG(ERR, "e1000 incorrect high/low water value");
1764                 PMD_INIT_LOG(ERR, "high water must <= 0x%x", max_high_water);
1765                 return -EINVAL;
1766         }
1767
1768         hw->fc.requested_mode = rte_fcmode_2_e1000_fcmode[fc_conf->mode];
1769         hw->fc.pause_time     = fc_conf->pause_time;
1770         hw->fc.high_water     = fc_conf->high_water;
1771         hw->fc.low_water      = fc_conf->low_water;
1772         hw->fc.send_xon       = fc_conf->send_xon;
1773
1774         err = e1000_setup_link_generic(hw);
1775         if (err == E1000_SUCCESS) {
1776
1777                 /* check if we want to forward MAC frames - driver doesn't have native
1778                  * capability to do that, so we'll write the registers ourselves */
1779
1780                 rctl = E1000_READ_REG(hw, E1000_RCTL);
1781
1782                 /* set or clear MFLCN.PMCF bit depending on configuration */
1783                 if (fc_conf->mac_ctrl_frame_fwd != 0)
1784                         rctl |= E1000_RCTL_PMCF;
1785                 else
1786                         rctl &= ~E1000_RCTL_PMCF;
1787
1788                 E1000_WRITE_REG(hw, E1000_RCTL, rctl);
1789                 E1000_WRITE_FLUSH(hw);
1790
1791                 return 0;
1792         }
1793
1794         PMD_INIT_LOG(ERR, "e1000_setup_link_generic = 0x%x", err);
1795         return -EIO;
1796 }
1797
1798 static int
1799 eth_em_rar_set(struct rte_eth_dev *dev, struct ether_addr *mac_addr,
1800                 uint32_t index, __rte_unused uint32_t pool)
1801 {
1802         struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1803
1804         return e1000_rar_set(hw, mac_addr->addr_bytes, index);
1805 }
1806
1807 static void
1808 eth_em_rar_clear(struct rte_eth_dev *dev, uint32_t index)
1809 {
1810         uint8_t addr[ETHER_ADDR_LEN];
1811         struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1812
1813         memset(addr, 0, sizeof(addr));
1814
1815         e1000_rar_set(hw, addr, index);
1816 }
1817
1818 static int
1819 eth_em_mtu_set(struct rte_eth_dev *dev, uint16_t mtu)
1820 {
1821         struct rte_eth_dev_info dev_info;
1822         struct e1000_hw *hw;
1823         uint32_t frame_size;
1824         uint32_t rctl;
1825
1826         eth_em_infos_get(dev, &dev_info);
1827         frame_size = mtu + ETHER_HDR_LEN + ETHER_CRC_LEN + VLAN_TAG_SIZE;
1828
1829         /* check that mtu is within the allowed range */
1830         if ((mtu < ETHER_MIN_MTU) || (frame_size > dev_info.max_rx_pktlen))
1831                 return -EINVAL;
1832
1833         /* refuse mtu that requires the support of scattered packets when this
1834          * feature has not been enabled before. */
1835         if (!dev->data->scattered_rx &&
1836             frame_size > dev->data->min_rx_buf_size - RTE_PKTMBUF_HEADROOM)
1837                 return -EINVAL;
1838
1839         hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1840         rctl = E1000_READ_REG(hw, E1000_RCTL);
1841
1842         /* switch to jumbo mode if needed */
1843         if (frame_size > ETHER_MAX_LEN) {
1844                 dev->data->dev_conf.rxmode.jumbo_frame = 1;
1845                 rctl |= E1000_RCTL_LPE;
1846         } else {
1847                 dev->data->dev_conf.rxmode.jumbo_frame = 0;
1848                 rctl &= ~E1000_RCTL_LPE;
1849         }
1850         E1000_WRITE_REG(hw, E1000_RCTL, rctl);
1851
1852         /* update max frame size */
1853         dev->data->dev_conf.rxmode.max_rx_pkt_len = frame_size;
1854         return 0;
1855 }
1856
1857 static int
1858 eth_em_set_mc_addr_list(struct rte_eth_dev *dev,
1859                         struct ether_addr *mc_addr_set,
1860                         uint32_t nb_mc_addr)
1861 {
1862         struct e1000_hw *hw;
1863
1864         hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1865         e1000_update_mc_addr_list(hw, (u8 *)mc_addr_set, nb_mc_addr);
1866         return 0;
1867 }
1868
1869 RTE_PMD_REGISTER_PCI(net_e1000_em, rte_em_pmd);
1870 RTE_PMD_REGISTER_PCI_TABLE(net_e1000_em, pci_id_em_map);
1871 RTE_PMD_REGISTER_KMOD_DEP(net_e1000_em, "* igb_uio | uio_pci_generic | vfio-pci");