sys/dev/e1000/if_em.c

   1 /*-
   2  * SPDX-License-Identifier: BSD-2-Clause
   3  *
   4  * Copyright (c) 2016 Nicole Graziano <nicole@nextbsd.org>
   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  * 1. Redistributions of source code must retain the above copyright
  11  *    notice, this list of conditions and the following disclaimer.
  12  * 2. Redistributions in binary form must reproduce the above copyright
  13  *    notice, this list of conditions and the following disclaimer in the
  14  *    documentation and/or other materials provided with the distribution.
  15  *
  16  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
  17  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  18  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  19  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
  20  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  21  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  22  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  23  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  24  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  25  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  26  * SUCH DAMAGE.
  27  */
  28
  29 /* $FreeBSD$ */
  30 #include "if_em.h"
  31 #include <sys/sbuf.h>
  32 #include <machine/_inttypes.h>
  33
  34 #define em_mac_min e1000_82547
  35 #define igb_mac_min e1000_82575
  36
  37 /*********************************************************************
  38  *  Driver version:
  39  *********************************************************************/
  40 char em_driver_version[] = "7.6.1-k";
  41
  42 /*********************************************************************
  43  *  PCI Device ID Table
  44  *
  45  *  Used by probe to select devices to load on
  46  *  Last field stores an index into e1000_strings
  47  *  Last entry must be all 0s
  48  *
  49  *  { Vendor ID, Device ID, SubVendor ID, SubDevice ID, String Index }
  50  *********************************************************************/
  51
  52 static pci_vendor_info_t em_vendor_info_array[] =
  53 {
  54         /* Intel(R) PRO/1000 Network Connection - Legacy em*/
  55         PVID(0x8086, E1000_DEV_ID_82540EM, "Intel(R) PRO/1000 Network Connection"),
  56         PVID(0x8086, E1000_DEV_ID_82540EM_LOM, "Intel(R) PRO/1000 Network Connection"),
  57         PVID(0x8086, E1000_DEV_ID_82540EP, "Intel(R) PRO/1000 Network Connection"),
  58         PVID(0x8086, E1000_DEV_ID_82540EP_LOM, "Intel(R) PRO/1000 Network Connection"),
  59         PVID(0x8086, E1000_DEV_ID_82540EP_LP, "Intel(R) PRO/1000 Network Connection"),
  60
  61         PVID(0x8086, E1000_DEV_ID_82541EI, "Intel(R) PRO/1000 Network Connection"),
  62         PVID(0x8086, E1000_DEV_ID_82541ER, "Intel(R) PRO/1000 Network Connection"),
  63         PVID(0x8086, E1000_DEV_ID_82541ER_LOM, "Intel(R) PRO/1000 Network Connection"),
  64         PVID(0x8086, E1000_DEV_ID_82541EI_MOBILE, "Intel(R) PRO/1000 Network Connection"),
  65         PVID(0x8086, E1000_DEV_ID_82541GI, "Intel(R) PRO/1000 Network Connection"),
  66         PVID(0x8086, E1000_DEV_ID_82541GI_LF, "Intel(R) PRO/1000 Network Connection"),
  67         PVID(0x8086, E1000_DEV_ID_82541GI_MOBILE, "Intel(R) PRO/1000 Network Connection"),
  68
  69         PVID(0x8086, E1000_DEV_ID_82542, "Intel(R) PRO/1000 Network Connection"),
  70
  71         PVID(0x8086, E1000_DEV_ID_82543GC_FIBER, "Intel(R) PRO/1000 Network Connection"),
  72         PVID(0x8086, E1000_DEV_ID_82543GC_COPPER, "Intel(R) PRO/1000 Network Connection"),
  73
  74         PVID(0x8086, E1000_DEV_ID_82544EI_COPPER, "Intel(R) PRO/1000 Network Connection"),
  75         PVID(0x8086, E1000_DEV_ID_82544EI_FIBER, "Intel(R) PRO/1000 Network Connection"),
  76         PVID(0x8086, E1000_DEV_ID_82544GC_COPPER, "Intel(R) PRO/1000 Network Connection"),
  77         PVID(0x8086, E1000_DEV_ID_82544GC_LOM, "Intel(R) PRO/1000 Network Connection"),
  78
  79         PVID(0x8086, E1000_DEV_ID_82545EM_COPPER, "Intel(R) PRO/1000 Network Connection"),
  80         PVID(0x8086, E1000_DEV_ID_82545EM_FIBER, "Intel(R) PRO/1000 Network Connection"),
  81         PVID(0x8086, E1000_DEV_ID_82545GM_COPPER, "Intel(R) PRO/1000 Network Connection"),
  82         PVID(0x8086, E1000_DEV_ID_82545GM_FIBER, "Intel(R) PRO/1000 Network Connection"),
  83         PVID(0x8086, E1000_DEV_ID_82545GM_SERDES, "Intel(R) PRO/1000 Network Connection"),
  84
  85         PVID(0x8086, E1000_DEV_ID_82546EB_COPPER, "Intel(R) PRO/1000 Network Connection"),
  86         PVID(0x8086, E1000_DEV_ID_82546EB_FIBER, "Intel(R) PRO/1000 Network Connection"),
  87         PVID(0x8086, E1000_DEV_ID_82546EB_QUAD_COPPER, "Intel(R) PRO/1000 Network Connection"),
  88         PVID(0x8086, E1000_DEV_ID_82546GB_COPPER, "Intel(R) PRO/1000 Network Connection"),
  89         PVID(0x8086, E1000_DEV_ID_82546GB_FIBER, "Intel(R) PRO/1000 Network Connection"),
  90         PVID(0x8086, E1000_DEV_ID_82546GB_SERDES, "Intel(R) PRO/1000 Network Connection"),
  91         PVID(0x8086, E1000_DEV_ID_82546GB_PCIE, "Intel(R) PRO/1000 Network Connection"),
  92         PVID(0x8086, E1000_DEV_ID_82546GB_QUAD_COPPER, "Intel(R) PRO/1000 Network Connection"),
  93         PVID(0x8086, E1000_DEV_ID_82546GB_QUAD_COPPER_KSP3, "Intel(R) PRO/1000 Network Connection"),
  94
  95         PVID(0x8086, E1000_DEV_ID_82547EI, "Intel(R) PRO/1000 Network Connection"),
  96         PVID(0x8086, E1000_DEV_ID_82547EI_MOBILE, "Intel(R) PRO/1000 Network Connection"),
  97         PVID(0x8086, E1000_DEV_ID_82547GI, "Intel(R) PRO/1000 Network Connection"),
  98
  99         /* Intel(R) PRO/1000 Network Connection - em */
 100         PVID(0x8086, E1000_DEV_ID_82571EB_COPPER, "Intel(R) PRO/1000 Network Connection"),
 101         PVID(0x8086, E1000_DEV_ID_82571EB_FIBER, "Intel(R) PRO/1000 Network Connection"),
 102         PVID(0x8086, E1000_DEV_ID_82571EB_SERDES, "Intel(R) PRO/1000 Network Connection"),
 103         PVID(0x8086, E1000_DEV_ID_82571EB_SERDES_DUAL, "Intel(R) PRO/1000 Network Connection"),
 104         PVID(0x8086, E1000_DEV_ID_82571EB_SERDES_QUAD, "Intel(R) PRO/1000 Network Connection"),
 105         PVID(0x8086, E1000_DEV_ID_82571EB_QUAD_COPPER, "Intel(R) PRO/1000 Network Connection"),
 106         PVID(0x8086, E1000_DEV_ID_82571EB_QUAD_COPPER_LP, "Intel(R) PRO/1000 Network Connection"),
 107         PVID(0x8086, E1000_DEV_ID_82571EB_QUAD_FIBER, "Intel(R) PRO/1000 Network Connection"),
 108         PVID(0x8086, E1000_DEV_ID_82571PT_QUAD_COPPER, "Intel(R) PRO/1000 Network Connection"),
 109         PVID(0x8086, E1000_DEV_ID_82572EI, "Intel(R) PRO/1000 Network Connection"),
 110         PVID(0x8086, E1000_DEV_ID_82572EI_COPPER, "Intel(R) PRO/1000 Network Connection"),
 111         PVID(0x8086, E1000_DEV_ID_82572EI_FIBER, "Intel(R) PRO/1000 Network Connection"),
 112         PVID(0x8086, E1000_DEV_ID_82572EI_SERDES, "Intel(R) PRO/1000 Network Connection"),
 113         PVID(0x8086, E1000_DEV_ID_82573E, "Intel(R) PRO/1000 Network Connection"),
 114         PVID(0x8086, E1000_DEV_ID_82573E_IAMT, "Intel(R) PRO/1000 Network Connection"),
 115         PVID(0x8086, E1000_DEV_ID_82573L, "Intel(R) PRO/1000 Network Connection"),
 116         PVID(0x8086, E1000_DEV_ID_82583V, "Intel(R) PRO/1000 Network Connection"),
 117         PVID(0x8086, E1000_DEV_ID_80003ES2LAN_COPPER_SPT, "Intel(R) PRO/1000 Network Connection"),
 118         PVID(0x8086, E1000_DEV_ID_80003ES2LAN_SERDES_SPT, "Intel(R) PRO/1000 Network Connection"),
 119         PVID(0x8086, E1000_DEV_ID_80003ES2LAN_COPPER_DPT, "Intel(R) PRO/1000 Network Connection"),
 120         PVID(0x8086, E1000_DEV_ID_80003ES2LAN_SERDES_DPT, "Intel(R) PRO/1000 Network Connection"),
 121         PVID(0x8086, E1000_DEV_ID_ICH8_IGP_M_AMT, "Intel(R) PRO/1000 Network Connection"),
 122         PVID(0x8086, E1000_DEV_ID_ICH8_IGP_AMT, "Intel(R) PRO/1000 Network Connection"),
 123         PVID(0x8086, E1000_DEV_ID_ICH8_IGP_C, "Intel(R) PRO/1000 Network Connection"),
 124         PVID(0x8086, E1000_DEV_ID_ICH8_IFE, "Intel(R) PRO/1000 Network Connection"),
 125         PVID(0x8086, E1000_DEV_ID_ICH8_IFE_GT, "Intel(R) PRO/1000 Network Connection"),
 126         PVID(0x8086, E1000_DEV_ID_ICH8_IFE_G, "Intel(R) PRO/1000 Network Connection"),
 127         PVID(0x8086, E1000_DEV_ID_ICH8_IGP_M, "Intel(R) PRO/1000 Network Connection"),
 128         PVID(0x8086, E1000_DEV_ID_ICH8_82567V_3, "Intel(R) PRO/1000 Network Connection"),
 129         PVID(0x8086, E1000_DEV_ID_ICH9_IGP_M_AMT, "Intel(R) PRO/1000 Network Connection"),
 130         PVID(0x8086, E1000_DEV_ID_ICH9_IGP_AMT, "Intel(R) PRO/1000 Network Connection"),
 131         PVID(0x8086, E1000_DEV_ID_ICH9_IGP_C, "Intel(R) PRO/1000 Network Connection"),
 132         PVID(0x8086, E1000_DEV_ID_ICH9_IGP_M, "Intel(R) PRO/1000 Network Connection"),
 133         PVID(0x8086, E1000_DEV_ID_ICH9_IGP_M_V, "Intel(R) PRO/1000 Network Connection"),
 134         PVID(0x8086, E1000_DEV_ID_ICH9_IFE, "Intel(R) PRO/1000 Network Connection"),
 135         PVID(0x8086, E1000_DEV_ID_ICH9_IFE_GT, "Intel(R) PRO/1000 Network Connection"),
 136         PVID(0x8086, E1000_DEV_ID_ICH9_IFE_G, "Intel(R) PRO/1000 Network Connection"),
 137         PVID(0x8086, E1000_DEV_ID_ICH9_BM, "Intel(R) PRO/1000 Network Connection"),
 138         PVID(0x8086, E1000_DEV_ID_82574L, "Intel(R) PRO/1000 Network Connection"),
 139         PVID(0x8086, E1000_DEV_ID_82574LA, "Intel(R) PRO/1000 Network Connection"),
 140         PVID(0x8086, E1000_DEV_ID_ICH10_R_BM_LM, "Intel(R) PRO/1000 Network Connection"),
 141         PVID(0x8086, E1000_DEV_ID_ICH10_R_BM_LF, "Intel(R) PRO/1000 Network Connection"),
 142         PVID(0x8086, E1000_DEV_ID_ICH10_R_BM_V, "Intel(R) PRO/1000 Network Connection"),
 143         PVID(0x8086, E1000_DEV_ID_ICH10_D_BM_LM, "Intel(R) PRO/1000 Network Connection"),
 144         PVID(0x8086, E1000_DEV_ID_ICH10_D_BM_LF, "Intel(R) PRO/1000 Network Connection"),
 145         PVID(0x8086, E1000_DEV_ID_ICH10_D_BM_V, "Intel(R) PRO/1000 Network Connection"),
 146         PVID(0x8086, E1000_DEV_ID_PCH_M_HV_LM, "Intel(R) PRO/1000 Network Connection"),
 147         PVID(0x8086, E1000_DEV_ID_PCH_M_HV_LC, "Intel(R) PRO/1000 Network Connection"),
 148         PVID(0x8086, E1000_DEV_ID_PCH_D_HV_DM, "Intel(R) PRO/1000 Network Connection"),
 149         PVID(0x8086, E1000_DEV_ID_PCH_D_HV_DC, "Intel(R) PRO/1000 Network Connection"),
 150         PVID(0x8086, E1000_DEV_ID_PCH2_LV_LM, "Intel(R) PRO/1000 Network Connection"),
 151         PVID(0x8086, E1000_DEV_ID_PCH2_LV_V, "Intel(R) PRO/1000 Network Connection"),
 152         PVID(0x8086, E1000_DEV_ID_PCH_LPT_I217_LM, "Intel(R) PRO/1000 Network Connection"),
 153         PVID(0x8086, E1000_DEV_ID_PCH_LPT_I217_V, "Intel(R) PRO/1000 Network Connection"),
 154         PVID(0x8086, E1000_DEV_ID_PCH_LPTLP_I218_LM, "Intel(R) PRO/1000 Network Connection"),
 155         PVID(0x8086, E1000_DEV_ID_PCH_LPTLP_I218_V, "Intel(R) PRO/1000 Network Connection"),
 156         PVID(0x8086, E1000_DEV_ID_PCH_I218_LM2, "Intel(R) PRO/1000 Network Connection"),
 157         PVID(0x8086, E1000_DEV_ID_PCH_I218_V2, "Intel(R) PRO/1000 Network Connection"),
 158         PVID(0x8086, E1000_DEV_ID_PCH_I218_LM3, "Intel(R) PRO/1000 Network Connection"),
 159         PVID(0x8086, E1000_DEV_ID_PCH_I218_V3, "Intel(R) PRO/1000 Network Connection"),
 160         PVID(0x8086, E1000_DEV_ID_PCH_SPT_I219_LM, "Intel(R) PRO/1000 Network Connection"),
 161         PVID(0x8086, E1000_DEV_ID_PCH_SPT_I219_V, "Intel(R) PRO/1000 Network Connection"),
 162         PVID(0x8086, E1000_DEV_ID_PCH_SPT_I219_LM2, "Intel(R) PRO/1000 Network Connection"),
 163         PVID(0x8086, E1000_DEV_ID_PCH_SPT_I219_V2, "Intel(R) PRO/1000 Network Connection"),
 164         PVID(0x8086, E1000_DEV_ID_PCH_LBG_I219_LM3, "Intel(R) PRO/1000 Network Connection"),
 165         PVID(0x8086, E1000_DEV_ID_PCH_SPT_I219_LM4, "Intel(R) PRO/1000 Network Connection"),
 166         PVID(0x8086, E1000_DEV_ID_PCH_SPT_I219_V4, "Intel(R) PRO/1000 Network Connection"),
 167         PVID(0x8086, E1000_DEV_ID_PCH_SPT_I219_LM5, "Intel(R) PRO/1000 Network Connection"),
 168         PVID(0x8086, E1000_DEV_ID_PCH_SPT_I219_V5, "Intel(R) PRO/1000 Network Connection"),
 169         PVID(0x8086, E1000_DEV_ID_PCH_CNP_I219_LM6, "Intel(R) PRO/1000 Network Connection"),
 170         PVID(0x8086, E1000_DEV_ID_PCH_CNP_I219_V6, "Intel(R) PRO/1000 Network Connection"),
 171         PVID(0x8086, E1000_DEV_ID_PCH_CNP_I219_LM7, "Intel(R) PRO/1000 Network Connection"),
 172         PVID(0x8086, E1000_DEV_ID_PCH_CNP_I219_V7, "Intel(R) PRO/1000 Network Connection"),
 173         PVID(0x8086, E1000_DEV_ID_PCH_ICP_I219_LM8, "Intel(R) PRO/1000 Network Connection"),
 174         PVID(0x8086, E1000_DEV_ID_PCH_ICP_I219_V8, "Intel(R) PRO/1000 Network Connection"),
 175         PVID(0x8086, E1000_DEV_ID_PCH_ICP_I219_LM9, "Intel(R) PRO/1000 Network Connection"),
 176         PVID(0x8086, E1000_DEV_ID_PCH_ICP_I219_V9, "Intel(R) PRO/1000 Network Connection"),
 177         PVID(0x8086, E1000_DEV_ID_PCH_CMP_I219_LM10, "Intel(R) PRO/1000 Network Connection"),
 178         PVID(0x8086, E1000_DEV_ID_PCH_CMP_I219_V10, "Intel(R) PRO/1000 Network Connection"),
 179         PVID(0x8086, E1000_DEV_ID_PCH_CMP_I219_LM11, "Intel(R) PRO/1000 Network Connection"),
 180         PVID(0x8086, E1000_DEV_ID_PCH_CMP_I219_V11, "Intel(R) PRO/1000 Network Connection"),
 181         PVID(0x8086, E1000_DEV_ID_PCH_CMP_I219_LM12, "Intel(R) PRO/1000 Network Connection"),
 182         PVID(0x8086, E1000_DEV_ID_PCH_CMP_I219_V12, "Intel(R) PRO/1000 Network Connection"),
 183         /* required last entry */
 184         PVID_END
 185 };
 186
 187 static pci_vendor_info_t igb_vendor_info_array[] =
 188 {
 189         /* Intel(R) PRO/1000 Network Connection - igb */
 190         PVID(0x8086, E1000_DEV_ID_82575EB_COPPER, "Intel(R) PRO/1000 PCI-Express Network Driver"),
 191         PVID(0x8086, E1000_DEV_ID_82575EB_FIBER_SERDES, "Intel(R) PRO/1000 PCI-Express Network Driver"),
 192         PVID(0x8086, E1000_DEV_ID_82575GB_QUAD_COPPER, "Intel(R) PRO/1000 PCI-Express Network Driver"),
 193         PVID(0x8086, E1000_DEV_ID_82576, "Intel(R) PRO/1000 PCI-Express Network Driver"),
 194         PVID(0x8086, E1000_DEV_ID_82576_NS, "Intel(R) PRO/1000 PCI-Express Network Driver"),
 195         PVID(0x8086, E1000_DEV_ID_82576_NS_SERDES, "Intel(R) PRO/1000 PCI-Express Network Driver"),
 196         PVID(0x8086, E1000_DEV_ID_82576_FIBER, "Intel(R) PRO/1000 PCI-Express Network Driver"),
 197         PVID(0x8086, E1000_DEV_ID_82576_SERDES, "Intel(R) PRO/1000 PCI-Express Network Driver"),
 198         PVID(0x8086, E1000_DEV_ID_82576_SERDES_QUAD, "Intel(R) PRO/1000 PCI-Express Network Driver"),
 199         PVID(0x8086, E1000_DEV_ID_82576_QUAD_COPPER, "Intel(R) PRO/1000 PCI-Express Network Driver"),
 200         PVID(0x8086, E1000_DEV_ID_82576_QUAD_COPPER_ET2, "Intel(R) PRO/1000 PCI-Express Network Driver"),
 201         PVID(0x8086, E1000_DEV_ID_82576_VF, "Intel(R) PRO/1000 PCI-Express Network Driver"),
 202         PVID(0x8086, E1000_DEV_ID_82580_COPPER, "Intel(R) PRO/1000 PCI-Express Network Driver"),
 203         PVID(0x8086, E1000_DEV_ID_82580_FIBER, "Intel(R) PRO/1000 PCI-Express Network Driver"),
 204         PVID(0x8086, E1000_DEV_ID_82580_SERDES, "Intel(R) PRO/1000 PCI-Express Network Driver"),
 205         PVID(0x8086, E1000_DEV_ID_82580_SGMII, "Intel(R) PRO/1000 PCI-Express Network Driver"),
 206         PVID(0x8086, E1000_DEV_ID_82580_COPPER_DUAL, "Intel(R) PRO/1000 PCI-Express Network Driver"),
 207         PVID(0x8086, E1000_DEV_ID_82580_QUAD_FIBER, "Intel(R) PRO/1000 PCI-Express Network Driver"),
 208         PVID(0x8086, E1000_DEV_ID_DH89XXCC_SERDES, "Intel(R) PRO/1000 PCI-Express Network Driver"),
 209         PVID(0x8086, E1000_DEV_ID_DH89XXCC_SGMII, "Intel(R) PRO/1000 PCI-Express Network Driver"),
 210         PVID(0x8086, E1000_DEV_ID_DH89XXCC_SFP, "Intel(R) PRO/1000 PCI-Express Network Driver"),
 211         PVID(0x8086, E1000_DEV_ID_DH89XXCC_BACKPLANE, "Intel(R) PRO/1000 PCI-Express Network Driver"),
 212         PVID(0x8086, E1000_DEV_ID_I350_COPPER, "Intel(R) PRO/1000 PCI-Express Network Driver"),
 213         PVID(0x8086, E1000_DEV_ID_I350_FIBER, "Intel(R) PRO/1000 PCI-Express Network Driver"),
 214         PVID(0x8086, E1000_DEV_ID_I350_SERDES, "Intel(R) PRO/1000 PCI-Express Network Driver"),
 215         PVID(0x8086, E1000_DEV_ID_I350_SGMII, "Intel(R) PRO/1000 PCI-Express Network Driver"),
 216         PVID(0x8086, E1000_DEV_ID_I350_VF, "Intel(R) PRO/1000 PCI-Express Network Driver"),
 217         PVID(0x8086, E1000_DEV_ID_I210_COPPER, "Intel(R) PRO/1000 PCI-Express Network Driver"),
 218         PVID(0x8086, E1000_DEV_ID_I210_COPPER_IT, "Intel(R) PRO/1000 PCI-Express Network Driver"),
 219         PVID(0x8086, E1000_DEV_ID_I210_COPPER_OEM1, "Intel(R) PRO/1000 PCI-Express Network Driver"),
 220         PVID(0x8086, E1000_DEV_ID_I210_COPPER_FLASHLESS, "Intel(R) PRO/1000 PCI-Express Network Driver"),
 221         PVID(0x8086, E1000_DEV_ID_I210_SERDES_FLASHLESS, "Intel(R) PRO/1000 PCI-Express Network Driver"),
 222         PVID(0x8086, E1000_DEV_ID_I210_FIBER, "Intel(R) PRO/1000 PCI-Express Network Driver"),
 223         PVID(0x8086, E1000_DEV_ID_I210_SERDES, "Intel(R) PRO/1000 PCI-Express Network Driver"),
 224         PVID(0x8086, E1000_DEV_ID_I210_SGMII, "Intel(R) PRO/1000 PCI-Express Network Driver"),
 225         PVID(0x8086, E1000_DEV_ID_I211_COPPER, "Intel(R) PRO/1000 PCI-Express Network Driver"),
 226         PVID(0x8086, E1000_DEV_ID_I354_BACKPLANE_1GBPS, "Intel(R) PRO/1000 PCI-Express Network Driver"),
 227         PVID(0x8086, E1000_DEV_ID_I354_BACKPLANE_2_5GBPS, "Intel(R) PRO/1000 PCI-Express Network Driver"),
 228         PVID(0x8086, E1000_DEV_ID_I354_SGMII, "Intel(R) PRO/1000 PCI-Express Network Driver"),
 229         /* required last entry */
 230         PVID_END
 231 };
 232
 233 /*********************************************************************
 234  *  Function prototypes
 235  *********************************************************************/
 236 static void     *em_register(device_t dev);
 237 static void     *igb_register(device_t dev);
 238 static int      em_if_attach_pre(if_ctx_t ctx);
 239 static int      em_if_attach_post(if_ctx_t ctx);
 240 static int      em_if_detach(if_ctx_t ctx);
 241 static int      em_if_shutdown(if_ctx_t ctx);
 242 static int      em_if_suspend(if_ctx_t ctx);
 243 static int      em_if_resume(if_ctx_t ctx);
 244
 245 static int      em_if_tx_queues_alloc(if_ctx_t ctx, caddr_t *vaddrs, uint64_t *paddrs, int ntxqs, int ntxqsets);
 246 static int      em_if_rx_queues_alloc(if_ctx_t ctx, caddr_t *vaddrs, uint64_t *paddrs, int nrxqs, int nrxqsets);
 247 static void     em_if_queues_free(if_ctx_t ctx);
 248
 249 static uint64_t em_if_get_counter(if_ctx_t, ift_counter);
 250 static void     em_if_init(if_ctx_t ctx);
 251 static void     em_if_stop(if_ctx_t ctx);
 252 static void     em_if_media_status(if_ctx_t, struct ifmediareq *);
 253 static int      em_if_media_change(if_ctx_t ctx);
 254 static int      em_if_mtu_set(if_ctx_t ctx, uint32_t mtu);
 255 static void     em_if_timer(if_ctx_t ctx, uint16_t qid);
 256 static void     em_if_vlan_register(if_ctx_t ctx, u16 vtag);
 257 static void     em_if_vlan_unregister(if_ctx_t ctx, u16 vtag);
 258 static void     em_if_watchdog_reset(if_ctx_t ctx);
 259 static bool     em_if_needs_restart(if_ctx_t ctx, enum iflib_restart_event event);
 260
 261 static void     em_identify_hardware(if_ctx_t ctx);
 262 static int      em_allocate_pci_resources(if_ctx_t ctx);
 263 static void     em_free_pci_resources(if_ctx_t ctx);
 264 static void     em_reset(if_ctx_t ctx);
 265 static int      em_setup_interface(if_ctx_t ctx);
 266 static int      em_setup_msix(if_ctx_t ctx);
 267
 268 static void     em_initialize_transmit_unit(if_ctx_t ctx);
 269 static void     em_initialize_receive_unit(if_ctx_t ctx);
 270
 271 static void     em_if_intr_enable(if_ctx_t ctx);
 272 static void     em_if_intr_disable(if_ctx_t ctx);
 273 static void     igb_if_intr_enable(if_ctx_t ctx);
 274 static void     igb_if_intr_disable(if_ctx_t ctx);
 275 static int      em_if_rx_queue_intr_enable(if_ctx_t ctx, uint16_t rxqid);
 276 static int      em_if_tx_queue_intr_enable(if_ctx_t ctx, uint16_t txqid);
 277 static int      igb_if_rx_queue_intr_enable(if_ctx_t ctx, uint16_t rxqid);
 278 static int      igb_if_tx_queue_intr_enable(if_ctx_t ctx, uint16_t txqid);
 279 static void     em_if_multi_set(if_ctx_t ctx);
 280 static void     em_if_update_admin_status(if_ctx_t ctx);
 281 static void     em_if_debug(if_ctx_t ctx);
 282 static void     em_update_stats_counters(struct adapter *);
 283 static void     em_add_hw_stats(struct adapter *adapter);
 284 static int      em_if_set_promisc(if_ctx_t ctx, int flags);
 285 static void     em_setup_vlan_hw_support(struct adapter *);
 286 static int      em_sysctl_nvm_info(SYSCTL_HANDLER_ARGS);
 287 static void     em_print_nvm_info(struct adapter *);
 288 static int      em_sysctl_debug_info(SYSCTL_HANDLER_ARGS);
 289 static int      em_get_rs(SYSCTL_HANDLER_ARGS);
 290 static void     em_print_debug_info(struct adapter *);
 291 static int      em_is_valid_ether_addr(u8 *);
 292 static int      em_sysctl_int_delay(SYSCTL_HANDLER_ARGS);
 293 static void     em_add_int_delay_sysctl(struct adapter *, const char *,
 294                     const char *, struct em_int_delay_info *, int, int);
 295 /* Management and WOL Support */
 296 static void     em_init_manageability(struct adapter *);
 297 static void     em_release_manageability(struct adapter *);
 298 static void     em_get_hw_control(struct adapter *);
 299 static void     em_release_hw_control(struct adapter *);
 300 static void     em_get_wakeup(if_ctx_t ctx);
 301 static void     em_enable_wakeup(if_ctx_t ctx);
 302 static int      em_enable_phy_wakeup(struct adapter *);
 303 static void     em_disable_aspm(struct adapter *);
 304
 305 int             em_intr(void *arg);
 306 static void     em_disable_promisc(if_ctx_t ctx);
 307
 308 /* MSI-X handlers */
 309 static int      em_if_msix_intr_assign(if_ctx_t, int);
 310 static int      em_msix_link(void *);
 311 static void     em_handle_link(void *context);
 312
 313 static void     em_enable_vectors_82574(if_ctx_t);
 314
 315 static int      em_set_flowcntl(SYSCTL_HANDLER_ARGS);
 316 static int      em_sysctl_eee(SYSCTL_HANDLER_ARGS);
 317 static void     em_if_led_func(if_ctx_t ctx, int onoff);
 318
 319 static int      em_get_regs(SYSCTL_HANDLER_ARGS);
 320
 321 static void     lem_smartspeed(struct adapter *adapter);
 322 static void     igb_configure_queues(struct adapter *adapter);
 323
 324
 325 /*********************************************************************
 326  *  FreeBSD Device Interface Entry Points
 327  *********************************************************************/
 328 static device_method_t em_methods[] = {
 329         /* Device interface */
 330         DEVMETHOD(device_register, em_register),
 331         DEVMETHOD(device_probe, iflib_device_probe),
 332         DEVMETHOD(device_attach, iflib_device_attach),
 333         DEVMETHOD(device_detach, iflib_device_detach),
 334         DEVMETHOD(device_shutdown, iflib_device_shutdown),
 335         DEVMETHOD(device_suspend, iflib_device_suspend),
 336         DEVMETHOD(device_resume, iflib_device_resume),
 337         DEVMETHOD_END
 338 };
 339
 340 static device_method_t igb_methods[] = {
 341         /* Device interface */
 342         DEVMETHOD(device_register, igb_register),
 343         DEVMETHOD(device_probe, iflib_device_probe),
 344         DEVMETHOD(device_attach, iflib_device_attach),
 345         DEVMETHOD(device_detach, iflib_device_detach),
 346         DEVMETHOD(device_shutdown, iflib_device_shutdown),
 347         DEVMETHOD(device_suspend, iflib_device_suspend),
 348         DEVMETHOD(device_resume, iflib_device_resume),
 349         DEVMETHOD_END
 350 };
 351
 352
 353 static driver_t em_driver = {
 354         "em", em_methods, sizeof(struct adapter),
 355 };
 356
 357 static devclass_t em_devclass;
 358 DRIVER_MODULE(em, pci, em_driver, em_devclass, 0, 0);
 359
 360 MODULE_DEPEND(em, pci, 1, 1, 1);
 361 MODULE_DEPEND(em, ether, 1, 1, 1);
 362 MODULE_DEPEND(em, iflib, 1, 1, 1);
 363
 364 IFLIB_PNP_INFO(pci, em, em_vendor_info_array);
 365
 366 static driver_t igb_driver = {
 367         "igb", igb_methods, sizeof(struct adapter),
 368 };
 369
 370 static devclass_t igb_devclass;
 371 DRIVER_MODULE(igb, pci, igb_driver, igb_devclass, 0, 0);
 372
 373 MODULE_DEPEND(igb, pci, 1, 1, 1);
 374 MODULE_DEPEND(igb, ether, 1, 1, 1);
 375 MODULE_DEPEND(igb, iflib, 1, 1, 1);
 376
 377 IFLIB_PNP_INFO(pci, igb, igb_vendor_info_array);
 378
 379 static device_method_t em_if_methods[] = {
 380         DEVMETHOD(ifdi_attach_pre, em_if_attach_pre),
 381         DEVMETHOD(ifdi_attach_post, em_if_attach_post),
 382         DEVMETHOD(ifdi_detach, em_if_detach),
 383         DEVMETHOD(ifdi_shutdown, em_if_shutdown),
 384         DEVMETHOD(ifdi_suspend, em_if_suspend),
 385         DEVMETHOD(ifdi_resume, em_if_resume),
 386         DEVMETHOD(ifdi_init, em_if_init),
 387         DEVMETHOD(ifdi_stop, em_if_stop),
 388         DEVMETHOD(ifdi_msix_intr_assign, em_if_msix_intr_assign),
 389         DEVMETHOD(ifdi_intr_enable, em_if_intr_enable),
 390         DEVMETHOD(ifdi_intr_disable, em_if_intr_disable),
 391         DEVMETHOD(ifdi_tx_queues_alloc, em_if_tx_queues_alloc),
 392         DEVMETHOD(ifdi_rx_queues_alloc, em_if_rx_queues_alloc),
 393         DEVMETHOD(ifdi_queues_free, em_if_queues_free),
 394         DEVMETHOD(ifdi_update_admin_status, em_if_update_admin_status),
 395         DEVMETHOD(ifdi_multi_set, em_if_multi_set),
 396         DEVMETHOD(ifdi_media_status, em_if_media_status),
 397         DEVMETHOD(ifdi_media_change, em_if_media_change),
 398         DEVMETHOD(ifdi_mtu_set, em_if_mtu_set),
 399         DEVMETHOD(ifdi_promisc_set, em_if_set_promisc),
 400         DEVMETHOD(ifdi_timer, em_if_timer),
 401         DEVMETHOD(ifdi_watchdog_reset, em_if_watchdog_reset),
 402         DEVMETHOD(ifdi_vlan_register, em_if_vlan_register),
 403         DEVMETHOD(ifdi_vlan_unregister, em_if_vlan_unregister),
 404         DEVMETHOD(ifdi_get_counter, em_if_get_counter),
 405         DEVMETHOD(ifdi_led_func, em_if_led_func),
 406         DEVMETHOD(ifdi_rx_queue_intr_enable, em_if_rx_queue_intr_enable),
 407         DEVMETHOD(ifdi_tx_queue_intr_enable, em_if_tx_queue_intr_enable),
 408         DEVMETHOD(ifdi_debug, em_if_debug),
 409         DEVMETHOD(ifdi_needs_restart, em_if_needs_restart),
 410         DEVMETHOD_END
 411 };
 412
 413 static driver_t em_if_driver = {
 414         "em_if", em_if_methods, sizeof(struct adapter)
 415 };
 416
 417 static device_method_t igb_if_methods[] = {
 418         DEVMETHOD(ifdi_attach_pre, em_if_attach_pre),
 419         DEVMETHOD(ifdi_attach_post, em_if_attach_post),
 420         DEVMETHOD(ifdi_detach, em_if_detach),
 421         DEVMETHOD(ifdi_shutdown, em_if_shutdown),
 422         DEVMETHOD(ifdi_suspend, em_if_suspend),
 423         DEVMETHOD(ifdi_resume, em_if_resume),
 424         DEVMETHOD(ifdi_init, em_if_init),
 425         DEVMETHOD(ifdi_stop, em_if_stop),
 426         DEVMETHOD(ifdi_msix_intr_assign, em_if_msix_intr_assign),
 427         DEVMETHOD(ifdi_intr_enable, igb_if_intr_enable),
 428         DEVMETHOD(ifdi_intr_disable, igb_if_intr_disable),
 429         DEVMETHOD(ifdi_tx_queues_alloc, em_if_tx_queues_alloc),
 430         DEVMETHOD(ifdi_rx_queues_alloc, em_if_rx_queues_alloc),
 431         DEVMETHOD(ifdi_queues_free, em_if_queues_free),
 432         DEVMETHOD(ifdi_update_admin_status, em_if_update_admin_status),
 433         DEVMETHOD(ifdi_multi_set, em_if_multi_set),
 434         DEVMETHOD(ifdi_media_status, em_if_media_status),
 435         DEVMETHOD(ifdi_media_change, em_if_media_change),
 436         DEVMETHOD(ifdi_mtu_set, em_if_mtu_set),
 437         DEVMETHOD(ifdi_promisc_set, em_if_set_promisc),
 438         DEVMETHOD(ifdi_timer, em_if_timer),
 439         DEVMETHOD(ifdi_watchdog_reset, em_if_watchdog_reset),
 440         DEVMETHOD(ifdi_vlan_register, em_if_vlan_register),
 441         DEVMETHOD(ifdi_vlan_unregister, em_if_vlan_unregister),
 442         DEVMETHOD(ifdi_get_counter, em_if_get_counter),
 443         DEVMETHOD(ifdi_led_func, em_if_led_func),
 444         DEVMETHOD(ifdi_rx_queue_intr_enable, igb_if_rx_queue_intr_enable),
 445         DEVMETHOD(ifdi_tx_queue_intr_enable, igb_if_tx_queue_intr_enable),
 446         DEVMETHOD(ifdi_debug, em_if_debug),
 447         DEVMETHOD(ifdi_needs_restart, em_if_needs_restart),
 448         DEVMETHOD_END
 449 };
 450
 451 static driver_t igb_if_driver = {
 452         "igb_if", igb_if_methods, sizeof(struct adapter)
 453 };
 454
 455 /*********************************************************************
 456  *  Tunable default values.
 457  *********************************************************************/
 458
 459 #define EM_TICKS_TO_USECS(ticks)        ((1024 * (ticks) + 500) / 1000)
 460 #define EM_USECS_TO_TICKS(usecs)        ((1000 * (usecs) + 512) / 1024)
 461
 462 #define MAX_INTS_PER_SEC        8000
 463 #define DEFAULT_ITR             (1000000000/(MAX_INTS_PER_SEC * 256))
 464
 465 /* Allow common code without TSO */
 466 #ifndef CSUM_TSO
 467 #define CSUM_TSO        0
 468 #endif
 469
 470 static SYSCTL_NODE(_hw, OID_AUTO, em, CTLFLAG_RD | CTLFLAG_MPSAFE, 0,
 471     "EM driver parameters");
 472
 473 static int em_disable_crc_stripping = 0;
 474 SYSCTL_INT(_hw_em, OID_AUTO, disable_crc_stripping, CTLFLAG_RDTUN,
 475     &em_disable_crc_stripping, 0, "Disable CRC Stripping");
 476
 477 static int em_tx_int_delay_dflt = EM_TICKS_TO_USECS(EM_TIDV);
 478 static int em_rx_int_delay_dflt = EM_TICKS_TO_USECS(EM_RDTR);
 479 SYSCTL_INT(_hw_em, OID_AUTO, tx_int_delay, CTLFLAG_RDTUN, &em_tx_int_delay_dflt,
 480     0, "Default transmit interrupt delay in usecs");
 481 SYSCTL_INT(_hw_em, OID_AUTO, rx_int_delay, CTLFLAG_RDTUN, &em_rx_int_delay_dflt,
 482     0, "Default receive interrupt delay in usecs");
 483
 484 static int em_tx_abs_int_delay_dflt = EM_TICKS_TO_USECS(EM_TADV);
 485 static int em_rx_abs_int_delay_dflt = EM_TICKS_TO_USECS(EM_RADV);
 486 SYSCTL_INT(_hw_em, OID_AUTO, tx_abs_int_delay, CTLFLAG_RDTUN,
 487     &em_tx_abs_int_delay_dflt, 0,
 488     "Default transmit interrupt delay limit in usecs");
 489 SYSCTL_INT(_hw_em, OID_AUTO, rx_abs_int_delay, CTLFLAG_RDTUN,
 490     &em_rx_abs_int_delay_dflt, 0,
 491     "Default receive interrupt delay limit in usecs");
 492
 493 static int em_smart_pwr_down = FALSE;
 494 SYSCTL_INT(_hw_em, OID_AUTO, smart_pwr_down, CTLFLAG_RDTUN, &em_smart_pwr_down,
 495     0, "Set to true to leave smart power down enabled on newer adapters");
 496
 497 /* Controls whether promiscuous also shows bad packets */
 498 static int em_debug_sbp = TRUE;
 499 SYSCTL_INT(_hw_em, OID_AUTO, sbp, CTLFLAG_RDTUN, &em_debug_sbp, 0,
 500     "Show bad packets in promiscuous mode");
 501
 502 /* How many packets rxeof tries to clean at a time */
 503 static int em_rx_process_limit = 100;
 504 SYSCTL_INT(_hw_em, OID_AUTO, rx_process_limit, CTLFLAG_RDTUN,
 505     &em_rx_process_limit, 0,
 506     "Maximum number of received packets to process "
 507     "at a time, -1 means unlimited");
 508
 509 /* Energy efficient ethernet - default to OFF */
 510 static int eee_setting = 1;
 511 SYSCTL_INT(_hw_em, OID_AUTO, eee_setting, CTLFLAG_RDTUN, &eee_setting, 0,
 512     "Enable Energy Efficient Ethernet");
 513
 514 /*
 515 ** Tuneable Interrupt rate
 516 */
 517 static int em_max_interrupt_rate = 8000;
 518 SYSCTL_INT(_hw_em, OID_AUTO, max_interrupt_rate, CTLFLAG_RDTUN,
 519     &em_max_interrupt_rate, 0, "Maximum interrupts per second");
 520
 521
 522
 523 /* Global used in WOL setup with multiport cards */
 524 static int global_quad_port_a = 0;
 525
 526 extern struct if_txrx igb_txrx;
 527 extern struct if_txrx em_txrx;
 528 extern struct if_txrx lem_txrx;
 529
 530 static struct if_shared_ctx em_sctx_init = {
 531         .isc_magic = IFLIB_MAGIC,
 532         .isc_q_align = PAGE_SIZE,
 533         .isc_tx_maxsize = EM_TSO_SIZE + sizeof(struct ether_vlan_header),
 534         .isc_tx_maxsegsize = PAGE_SIZE,
 535         .isc_tso_maxsize = EM_TSO_SIZE + sizeof(struct ether_vlan_header),
 536         .isc_tso_maxsegsize = EM_TSO_SEG_SIZE,
 537         .isc_rx_maxsize = MJUM9BYTES,
 538         .isc_rx_nsegments = 1,
 539         .isc_rx_maxsegsize = MJUM9BYTES,
 540         .isc_nfl = 1,
 541         .isc_nrxqs = 1,
 542         .isc_ntxqs = 1,
 543         .isc_admin_intrcnt = 1,
 544         .isc_vendor_info = em_vendor_info_array,
 545         .isc_driver_version = em_driver_version,
 546         .isc_driver = &em_if_driver,
 547         .isc_flags = IFLIB_NEED_SCRATCH | IFLIB_TSO_INIT_IP | IFLIB_NEED_ZERO_CSUM,
 548
 549         .isc_nrxd_min = {EM_MIN_RXD},
 550         .isc_ntxd_min = {EM_MIN_TXD},
 551         .isc_nrxd_max = {EM_MAX_RXD},
 552         .isc_ntxd_max = {EM_MAX_TXD},
 553         .isc_nrxd_default = {EM_DEFAULT_RXD},
 554         .isc_ntxd_default = {EM_DEFAULT_TXD},
 555 };
 556
 557 static struct if_shared_ctx igb_sctx_init = {
 558         .isc_magic = IFLIB_MAGIC,
 559         .isc_q_align = PAGE_SIZE,
 560         .isc_tx_maxsize = EM_TSO_SIZE + sizeof(struct ether_vlan_header),
 561         .isc_tx_maxsegsize = PAGE_SIZE,
 562         .isc_tso_maxsize = EM_TSO_SIZE + sizeof(struct ether_vlan_header),
 563         .isc_tso_maxsegsize = EM_TSO_SEG_SIZE,
 564         .isc_rx_maxsize = MJUM9BYTES,
 565         .isc_rx_nsegments = 1,
 566         .isc_rx_maxsegsize = MJUM9BYTES,
 567         .isc_nfl = 1,
 568         .isc_nrxqs = 1,
 569         .isc_ntxqs = 1,
 570         .isc_admin_intrcnt = 1,
 571         .isc_vendor_info = igb_vendor_info_array,
 572         .isc_driver_version = em_driver_version,
 573         .isc_driver = &igb_if_driver,
 574         .isc_flags = IFLIB_NEED_SCRATCH | IFLIB_TSO_INIT_IP | IFLIB_NEED_ZERO_CSUM,
 575
 576         .isc_nrxd_min = {EM_MIN_RXD},
 577         .isc_ntxd_min = {EM_MIN_TXD},
 578         .isc_nrxd_max = {IGB_MAX_RXD},
 579         .isc_ntxd_max = {IGB_MAX_TXD},
 580         .isc_nrxd_default = {EM_DEFAULT_RXD},
 581         .isc_ntxd_default = {EM_DEFAULT_TXD},
 582 };
 583
 584 /*****************************************************************
 585  *
 586  * Dump Registers
 587  *
 588  ****************************************************************/
 589 #define IGB_REGS_LEN 739
 590
 591 static int em_get_regs(SYSCTL_HANDLER_ARGS)
 592 {
 593         struct adapter *adapter = (struct adapter *)arg1;
 594         struct e1000_hw *hw = &adapter->hw;
 595         struct sbuf *sb;
 596         u32 *regs_buff;
 597         int rc;
 598
 599         regs_buff = malloc(sizeof(u32) * IGB_REGS_LEN, M_DEVBUF, M_WAITOK);
 600         memset(regs_buff, 0, IGB_REGS_LEN * sizeof(u32));
 601
 602         rc = sysctl_wire_old_buffer(req, 0);
 603         MPASS(rc == 0);
 604         if (rc != 0) {
 605                 free(regs_buff, M_DEVBUF);
 606                 return (rc);
 607         }
 608
 609         sb = sbuf_new_for_sysctl(NULL, NULL, 32*400, req);
 610         MPASS(sb != NULL);
 611         if (sb == NULL) {
 612                 free(regs_buff, M_DEVBUF);
 613                 return (ENOMEM);
 614         }
 615
 616         /* General Registers */
 617         regs_buff[0] = E1000_READ_REG(hw, E1000_CTRL);
 618         regs_buff[1] = E1000_READ_REG(hw, E1000_STATUS);
 619         regs_buff[2] = E1000_READ_REG(hw, E1000_CTRL_EXT);
 620         regs_buff[3] = E1000_READ_REG(hw, E1000_ICR);
 621         regs_buff[4] = E1000_READ_REG(hw, E1000_RCTL);
 622         regs_buff[5] = E1000_READ_REG(hw, E1000_RDLEN(0));
 623         regs_buff[6] = E1000_READ_REG(hw, E1000_RDH(0));
 624         regs_buff[7] = E1000_READ_REG(hw, E1000_RDT(0));
 625         regs_buff[8] = E1000_READ_REG(hw, E1000_RXDCTL(0));
 626         regs_buff[9] = E1000_READ_REG(hw, E1000_RDBAL(0));
 627         regs_buff[10] = E1000_READ_REG(hw, E1000_RDBAH(0));
 628         regs_buff[11] = E1000_READ_REG(hw, E1000_TCTL);
 629         regs_buff[12] = E1000_READ_REG(hw, E1000_TDBAL(0));
 630         regs_buff[13] = E1000_READ_REG(hw, E1000_TDBAH(0));
 631         regs_buff[14] = E1000_READ_REG(hw, E1000_TDLEN(0));
 632         regs_buff[15] = E1000_READ_REG(hw, E1000_TDH(0));
 633         regs_buff[16] = E1000_READ_REG(hw, E1000_TDT(0));
 634         regs_buff[17] = E1000_READ_REG(hw, E1000_TXDCTL(0));
 635         regs_buff[18] = E1000_READ_REG(hw, E1000_TDFH);
 636         regs_buff[19] = E1000_READ_REG(hw, E1000_TDFT);
 637         regs_buff[20] = E1000_READ_REG(hw, E1000_TDFHS);
 638         regs_buff[21] = E1000_READ_REG(hw, E1000_TDFPC);
 639
 640         sbuf_printf(sb, "General Registers\n");
 641         sbuf_printf(sb, "\tCTRL\t %08x\n", regs_buff[0]);
 642         sbuf_printf(sb, "\tSTATUS\t %08x\n", regs_buff[1]);
 643         sbuf_printf(sb, "\tCTRL_EXIT\t %08x\n\n", regs_buff[2]);
 644
 645         sbuf_printf(sb, "Interrupt Registers\n");
 646         sbuf_printf(sb, "\tICR\t %08x\n\n", regs_buff[3]);
 647
 648         sbuf_printf(sb, "RX Registers\n");
 649         sbuf_printf(sb, "\tRCTL\t %08x\n", regs_buff[4]);
 650         sbuf_printf(sb, "\tRDLEN\t %08x\n", regs_buff[5]);
 651         sbuf_printf(sb, "\tRDH\t %08x\n", regs_buff[6]);
 652         sbuf_printf(sb, "\tRDT\t %08x\n", regs_buff[7]);
 653         sbuf_printf(sb, "\tRXDCTL\t %08x\n", regs_buff[8]);
 654         sbuf_printf(sb, "\tRDBAL\t %08x\n", regs_buff[9]);
 655         sbuf_printf(sb, "\tRDBAH\t %08x\n\n", regs_buff[10]);
 656
 657         sbuf_printf(sb, "TX Registers\n");
 658         sbuf_printf(sb, "\tTCTL\t %08x\n", regs_buff[11]);
 659         sbuf_printf(sb, "\tTDBAL\t %08x\n", regs_buff[12]);
 660         sbuf_printf(sb, "\tTDBAH\t %08x\n", regs_buff[13]);
 661         sbuf_printf(sb, "\tTDLEN\t %08x\n", regs_buff[14]);
 662         sbuf_printf(sb, "\tTDH\t %08x\n", regs_buff[15]);
 663         sbuf_printf(sb, "\tTDT\t %08x\n", regs_buff[16]);
 664         sbuf_printf(sb, "\tTXDCTL\t %08x\n", regs_buff[17]);
 665         sbuf_printf(sb, "\tTDFH\t %08x\n", regs_buff[18]);
 666         sbuf_printf(sb, "\tTDFT\t %08x\n", regs_buff[19]);
 667         sbuf_printf(sb, "\tTDFHS\t %08x\n", regs_buff[20]);
 668         sbuf_printf(sb, "\tTDFPC\t %08x\n\n", regs_buff[21]);
 669
 670         free(regs_buff, M_DEVBUF);
 671
 672 #ifdef DUMP_DESCS
 673         {
 674                 if_softc_ctx_t scctx = adapter->shared;
 675                 struct rx_ring *rxr = &rx_que->rxr;
 676                 struct tx_ring *txr = &tx_que->txr;
 677                 int ntxd = scctx->isc_ntxd[0];
 678                 int nrxd = scctx->isc_nrxd[0];
 679                 int j;
 680
 681         for (j = 0; j < nrxd; j++) {
 682                 u32 staterr = le32toh(rxr->rx_base[j].wb.upper.status_error);
 683                 u32 length =  le32toh(rxr->rx_base[j].wb.upper.length);
 684                 sbuf_printf(sb, "\tReceive Descriptor Address %d: %08" PRIx64 "  Error:%d  Length:%d\n", j, rxr->rx_base[j].read.buffer_addr, staterr, length);
 685         }
 686
 687         for (j = 0; j < min(ntxd, 256); j++) {
 688                 unsigned int *ptr = (unsigned int *)&txr->tx_base[j];
 689
 690                 sbuf_printf(sb, "\tTXD[%03d] [0]: %08x [1]: %08x [2]: %08x [3]: %08x  eop: %d DD=%d\n",
 691                             j, ptr[0], ptr[1], ptr[2], ptr[3], buf->eop,
 692                             buf->eop != -1 ? txr->tx_base[buf->eop].upper.fields.status & E1000_TXD_STAT_DD : 0);
 693
 694         }
 695         }
 696 #endif
 697
 698         rc = sbuf_finish(sb);
 699         sbuf_delete(sb);
 700         return(rc);
 701 }
 702
 703 static void *
 704 em_register(device_t dev)
 705 {
 706         return (&em_sctx_init);
 707 }
 708
 709 static void *
 710 igb_register(device_t dev)
 711 {
 712         return (&igb_sctx_init);
 713 }
 714
 715 static int
 716 em_set_num_queues(if_ctx_t ctx)
 717 {
 718         struct adapter *adapter = iflib_get_softc(ctx);
 719         int maxqueues;
 720
 721         /* Sanity check based on HW */
 722         switch (adapter->hw.mac.type) {
 723         case e1000_82576:
 724         case e1000_82580:
 725         case e1000_i350:
 726         case e1000_i354:
 727                 maxqueues = 8;
 728                 break;
 729         case e1000_i210:
 730         case e1000_82575:
 731                 maxqueues = 4;
 732                 break;
 733         case e1000_i211:
 734         case e1000_82574:
 735                 maxqueues = 2;
 736                 break;
 737         default:
 738                 maxqueues = 1;
 739                 break;
 740         }
 741
 742         return (maxqueues);
 743 }
 744
 745 #define LEM_CAPS                                                        \
 746     IFCAP_HWCSUM | IFCAP_VLAN_MTU | IFCAP_VLAN_HWTAGGING |              \
 747     IFCAP_VLAN_HWCSUM | IFCAP_WOL | IFCAP_VLAN_HWFILTER
 748
 749 #define EM_CAPS                                                         \
 750     IFCAP_HWCSUM | IFCAP_VLAN_MTU | IFCAP_VLAN_HWTAGGING |              \
 751     IFCAP_VLAN_HWCSUM | IFCAP_WOL | IFCAP_VLAN_HWFILTER | IFCAP_TSO4 |  \
 752     IFCAP_LRO | IFCAP_VLAN_HWTSO
 753
 754 #define IGB_CAPS                                                        \
 755     IFCAP_HWCSUM | IFCAP_VLAN_MTU | IFCAP_VLAN_HWTAGGING |              \
 756     IFCAP_VLAN_HWCSUM | IFCAP_WOL | IFCAP_VLAN_HWFILTER | IFCAP_TSO4 |  \
 757     IFCAP_LRO | IFCAP_VLAN_HWTSO | IFCAP_JUMBO_MTU | IFCAP_HWCSUM_IPV6 |\
 758     IFCAP_TSO6
 759
 760 /*********************************************************************
 761  *  Device initialization routine
 762  *
 763  *  The attach entry point is called when the driver is being loaded.
 764  *  This routine identifies the type of hardware, allocates all resources
 765  *  and initializes the hardware.
 766  *
 767  *  return 0 on success, positive on failure
 768  *********************************************************************/
 769 static int
 770 em_if_attach_pre(if_ctx_t ctx)
 771 {
 772         struct adapter *adapter;
 773         if_softc_ctx_t scctx;
 774         device_t dev;
 775         struct e1000_hw *hw;
 776         int error = 0;
 777
 778         INIT_DEBUGOUT("em_if_attach_pre: begin");
 779         dev = iflib_get_dev(ctx);
 780         adapter = iflib_get_softc(ctx);
 781
 782         adapter->ctx = adapter->osdep.ctx = ctx;
 783         adapter->dev = adapter->osdep.dev = dev;
 784         scctx = adapter->shared = iflib_get_softc_ctx(ctx);
 785         adapter->media = iflib_get_media(ctx);
 786         hw = &adapter->hw;
 787
 788         adapter->tx_process_limit = scctx->isc_ntxd[0];
 789
 790         /* SYSCTL stuff */
 791         SYSCTL_ADD_PROC(device_get_sysctl_ctx(dev),
 792             SYSCTL_CHILDREN(device_get_sysctl_tree(dev)),
 793             OID_AUTO, "nvm", CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_NEEDGIANT,
 794             adapter, 0, em_sysctl_nvm_info, "I", "NVM Information");
 795
 796         SYSCTL_ADD_PROC(device_get_sysctl_ctx(dev),
 797             SYSCTL_CHILDREN(device_get_sysctl_tree(dev)),
 798             OID_AUTO, "debug", CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_NEEDGIANT,
 799             adapter, 0, em_sysctl_debug_info, "I", "Debug Information");
 800
 801         SYSCTL_ADD_PROC(device_get_sysctl_ctx(dev),
 802             SYSCTL_CHILDREN(device_get_sysctl_tree(dev)),
 803             OID_AUTO, "fc", CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_NEEDGIANT,
 804             adapter, 0, em_set_flowcntl, "I", "Flow Control");
 805
 806         SYSCTL_ADD_PROC(device_get_sysctl_ctx(dev),
 807             SYSCTL_CHILDREN(device_get_sysctl_tree(dev)),
 808             OID_AUTO, "reg_dump",
 809             CTLTYPE_STRING | CTLFLAG_RD | CTLFLAG_NEEDGIANT, adapter, 0,
 810             em_get_regs, "A", "Dump Registers");
 811
 812         SYSCTL_ADD_PROC(device_get_sysctl_ctx(dev),
 813             SYSCTL_CHILDREN(device_get_sysctl_tree(dev)),
 814             OID_AUTO, "rs_dump",
 815             CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_NEEDGIANT, adapter, 0,
 816             em_get_rs, "I", "Dump RS indexes");
 817
 818         /* Determine hardware and mac info */
 819         em_identify_hardware(ctx);
 820
 821         scctx->isc_tx_nsegments = EM_MAX_SCATTER;
 822         scctx->isc_nrxqsets_max = scctx->isc_ntxqsets_max = em_set_num_queues(ctx);
 823         if (bootverbose)
 824                 device_printf(dev, "attach_pre capping queues at %d\n",
 825                     scctx->isc_ntxqsets_max);
 826
 827         if (hw->mac.type >= igb_mac_min) {
 828                 scctx->isc_txqsizes[0] = roundup2(scctx->isc_ntxd[0] * sizeof(union e1000_adv_tx_desc), EM_DBA_ALIGN);
 829                 scctx->isc_rxqsizes[0] = roundup2(scctx->isc_nrxd[0] * sizeof(union e1000_adv_rx_desc), EM_DBA_ALIGN);
 830                 scctx->isc_txd_size[0] = sizeof(union e1000_adv_tx_desc);
 831                 scctx->isc_rxd_size[0] = sizeof(union e1000_adv_rx_desc);
 832                 scctx->isc_txrx = &igb_txrx;
 833                 scctx->isc_tx_tso_segments_max = EM_MAX_SCATTER;
 834                 scctx->isc_tx_tso_size_max = EM_TSO_SIZE;
 835                 scctx->isc_tx_tso_segsize_max = EM_TSO_SEG_SIZE;
 836                 scctx->isc_capabilities = scctx->isc_capenable = IGB_CAPS;
 837                 scctx->isc_tx_csum_flags = CSUM_TCP | CSUM_UDP | CSUM_TSO |
 838                      CSUM_IP6_TCP | CSUM_IP6_UDP;
 839                 if (hw->mac.type != e1000_82575)
 840                         scctx->isc_tx_csum_flags |= CSUM_SCTP | CSUM_IP6_SCTP;
 841                 /*
 842                 ** Some new devices, as with ixgbe, now may
 843                 ** use a different BAR, so we need to keep
 844                 ** track of which is used.
 845                 */
 846                 scctx->isc_msix_bar = pci_msix_table_bar(dev);
 847         } else if (hw->mac.type >= em_mac_min) {
 848                 scctx->isc_txqsizes[0] = roundup2(scctx->isc_ntxd[0]* sizeof(struct e1000_tx_desc), EM_DBA_ALIGN);
 849                 scctx->isc_rxqsizes[0] = roundup2(scctx->isc_nrxd[0] * sizeof(union e1000_rx_desc_extended), EM_DBA_ALIGN);
 850                 scctx->isc_txd_size[0] = sizeof(struct e1000_tx_desc);
 851                 scctx->isc_rxd_size[0] = sizeof(union e1000_rx_desc_extended);
 852                 scctx->isc_txrx = &em_txrx;
 853                 scctx->isc_tx_tso_segments_max = EM_MAX_SCATTER;
 854                 scctx->isc_tx_tso_size_max = EM_TSO_SIZE;
 855                 scctx->isc_tx_tso_segsize_max = EM_TSO_SEG_SIZE;
 856                 scctx->isc_capabilities = scctx->isc_capenable = EM_CAPS;
 857                 /*
 858                  * For EM-class devices, don't enable IFCAP_{TSO4,VLAN_HWTSO}
 859                  * by default as we don't have workarounds for all associated
 860                  * silicon errata.  E. g., with several MACs such as 82573E,
 861                  * TSO only works at Gigabit speed and otherwise can cause the
 862                  * hardware to hang (which also would be next to impossible to
 863                  * work around given that already queued TSO-using descriptors
 864                  * would need to be flushed and vlan(4) reconfigured at runtime
 865                  * in case of a link speed change).  Moreover, MACs like 82579
 866                  * still can hang at Gigabit even with all publicly documented
 867                  * TSO workarounds implemented.  Generally, the penality of
 868                  * these workarounds is rather high and may involve copying
 869                  * mbuf data around so advantages of TSO lapse.  Still, TSO may
 870                  * work for a few MACs of this class - at least when sticking
 871                  * with Gigabit - in which case users may enable TSO manually.
 872                  */
 873                 scctx->isc_capenable &= ~(IFCAP_TSO4 | IFCAP_VLAN_HWTSO);
 874                 scctx->isc_tx_csum_flags = CSUM_TCP | CSUM_UDP | CSUM_IP_TSO;
 875                 /*
 876                  * We support MSI-X with 82574 only, but indicate to iflib(4)
 877                  * that it shall give MSI at least a try with other devices.
 878                  */
 879                 if (hw->mac.type == e1000_82574) {
 880                         scctx->isc_msix_bar = pci_msix_table_bar(dev);;
 881                 } else {
 882                         scctx->isc_msix_bar = -1;
 883                         scctx->isc_disable_msix = 1;
 884                 }
 885         } else {
 886                 scctx->isc_txqsizes[0] = roundup2((scctx->isc_ntxd[0] + 1) * sizeof(struct e1000_tx_desc), EM_DBA_ALIGN);
 887                 scctx->isc_rxqsizes[0] = roundup2((scctx->isc_nrxd[0] + 1) * sizeof(struct e1000_rx_desc), EM_DBA_ALIGN);
 888                 scctx->isc_txd_size[0] = sizeof(struct e1000_tx_desc);
 889                 scctx->isc_rxd_size[0] = sizeof(struct e1000_rx_desc);
 890                 scctx->isc_tx_csum_flags = CSUM_TCP | CSUM_UDP;
 891                 scctx->isc_txrx = &lem_txrx;
 892                 scctx->isc_capabilities = scctx->isc_capenable = LEM_CAPS;
 893                 if (hw->mac.type < e1000_82543)
 894                         scctx->isc_capenable &= ~(IFCAP_HWCSUM|IFCAP_VLAN_HWCSUM);
 895                 /* INTx only */
 896                 scctx->isc_msix_bar = 0;
 897         }
 898
 899         /* Setup PCI resources */
 900         if (em_allocate_pci_resources(ctx)) {
 901                 device_printf(dev, "Allocation of PCI resources failed\n");
 902                 error = ENXIO;
 903                 goto err_pci;
 904         }
 905
 906         /*
 907         ** For ICH8 and family we need to
 908         ** map the flash memory, and this
 909         ** must happen after the MAC is
 910         ** identified
 911         */
 912         if ((hw->mac.type == e1000_ich8lan) ||
 913             (hw->mac.type == e1000_ich9lan) ||
 914             (hw->mac.type == e1000_ich10lan) ||
 915             (hw->mac.type == e1000_pchlan) ||
 916             (hw->mac.type == e1000_pch2lan) ||
 917             (hw->mac.type == e1000_pch_lpt)) {
 918                 int rid = EM_BAR_TYPE_FLASH;
 919                 adapter->flash = bus_alloc_resource_any(dev,
 920                     SYS_RES_MEMORY, &rid, RF_ACTIVE);
 921                 if (adapter->flash == NULL) {
 922                         device_printf(dev, "Mapping of Flash failed\n");
 923                         error = ENXIO;
 924                         goto err_pci;
 925                 }
 926                 /* This is used in the shared code */
 927                 hw->flash_address = (u8 *)adapter->flash;
 928                 adapter->osdep.flash_bus_space_tag =
 929                     rman_get_bustag(adapter->flash);
 930                 adapter->osdep.flash_bus_space_handle =
 931                     rman_get_bushandle(adapter->flash);
 932         }
 933         /*
 934         ** In the new SPT device flash is not  a
 935         ** separate BAR, rather it is also in BAR0,
 936         ** so use the same tag and an offset handle for the
 937         ** FLASH read/write macros in the shared code.
 938         */
 939         else if (hw->mac.type >= e1000_pch_spt) {
 940                 adapter->osdep.flash_bus_space_tag =
 941                     adapter->osdep.mem_bus_space_tag;
 942                 adapter->osdep.flash_bus_space_handle =
 943                     adapter->osdep.mem_bus_space_handle
 944                     + E1000_FLASH_BASE_ADDR;
 945         }
 946
 947         /* Do Shared Code initialization */
 948         error = e1000_setup_init_funcs(hw, TRUE);
 949         if (error) {
 950                 device_printf(dev, "Setup of Shared code failed, error %d\n",
 951                     error);
 952                 error = ENXIO;
 953                 goto err_pci;
 954         }
 955
 956         em_setup_msix(ctx);
 957         e1000_get_bus_info(hw);
 958
 959         /* Set up some sysctls for the tunable interrupt delays */
 960         em_add_int_delay_sysctl(adapter, "rx_int_delay",
 961             "receive interrupt delay in usecs", &adapter->rx_int_delay,
 962             E1000_REGISTER(hw, E1000_RDTR), em_rx_int_delay_dflt);
 963         em_add_int_delay_sysctl(adapter, "tx_int_delay",
 964             "transmit interrupt delay in usecs", &adapter->tx_int_delay,
 965             E1000_REGISTER(hw, E1000_TIDV), em_tx_int_delay_dflt);
 966         em_add_int_delay_sysctl(adapter, "rx_abs_int_delay",
 967             "receive interrupt delay limit in usecs",
 968             &adapter->rx_abs_int_delay,
 969             E1000_REGISTER(hw, E1000_RADV),
 970             em_rx_abs_int_delay_dflt);
 971         em_add_int_delay_sysctl(adapter, "tx_abs_int_delay",
 972             "transmit interrupt delay limit in usecs",
 973             &adapter->tx_abs_int_delay,
 974             E1000_REGISTER(hw, E1000_TADV),
 975             em_tx_abs_int_delay_dflt);
 976         em_add_int_delay_sysctl(adapter, "itr",
 977             "interrupt delay limit in usecs/4",
 978             &adapter->tx_itr,
 979             E1000_REGISTER(hw, E1000_ITR),
 980             DEFAULT_ITR);
 981
 982         hw->mac.autoneg = DO_AUTO_NEG;
 983         hw->phy.autoneg_wait_to_complete = FALSE;
 984         hw->phy.autoneg_advertised = AUTONEG_ADV_DEFAULT;
 985
 986         if (hw->mac.type < em_mac_min) {
 987                 e1000_init_script_state_82541(hw, TRUE);
 988                 e1000_set_tbi_compatibility_82543(hw, TRUE);
 989         }
 990         /* Copper options */
 991         if (hw->phy.media_type == e1000_media_type_copper) {
 992                 hw->phy.mdix = AUTO_ALL_MODES;
 993                 hw->phy.disable_polarity_correction = FALSE;
 994                 hw->phy.ms_type = EM_MASTER_SLAVE;
 995         }
 996
 997         /*
 998          * Set the frame limits assuming
 999          * standard ethernet sized frames.
1000          */
1001         scctx->isc_max_frame_size = hw->mac.max_frame_size =
1002             ETHERMTU + ETHER_HDR_LEN + ETHERNET_FCS_SIZE;
1003
1004         /*
1005          * This controls when hardware reports transmit completion
1006          * status.
1007          */
1008         hw->mac.report_tx_early = 1;
1009
1010         /* Allocate multicast array memory. */
1011         adapter->mta = malloc(sizeof(u8) * ETHER_ADDR_LEN *
1012             MAX_NUM_MULTICAST_ADDRESSES, M_DEVBUF, M_NOWAIT);
1013         if (adapter->mta == NULL) {
1014                 device_printf(dev, "Can not allocate multicast setup array\n");
1015                 error = ENOMEM;
1016                 goto err_late;
1017         }
1018
1019         /* Check SOL/IDER usage */
1020         if (e1000_check_reset_block(hw))
1021                 device_printf(dev, "PHY reset is blocked"
1022                               " due to SOL/IDER session.\n");
1023
1024         /* Sysctl for setting Energy Efficient Ethernet */
1025         hw->dev_spec.ich8lan.eee_disable = eee_setting;
1026         SYSCTL_ADD_PROC(device_get_sysctl_ctx(dev),
1027             SYSCTL_CHILDREN(device_get_sysctl_tree(dev)),
1028             OID_AUTO, "eee_control",
1029             CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_NEEDGIANT,
1030             adapter, 0, em_sysctl_eee, "I",
1031             "Disable Energy Efficient Ethernet");
1032
1033         /*
1034         ** Start from a known state, this is
1035         ** important in reading the nvm and
1036         ** mac from that.
1037         */
1038         e1000_reset_hw(hw);
1039
1040         /* Make sure we have a good EEPROM before we read from it */
1041         if (e1000_validate_nvm_checksum(hw) < 0) {
1042                 /*
1043                 ** Some PCI-E parts fail the first check due to
1044                 ** the link being in sleep state, call it again,
1045                 ** if it fails a second time its a real issue.
1046                 */
1047                 if (e1000_validate_nvm_checksum(hw) < 0) {
1048                         device_printf(dev,
1049                             "The EEPROM Checksum Is Not Valid\n");
1050                         error = EIO;
1051                         goto err_late;
1052                 }
1053         }
1054
1055         /* Copy the permanent MAC address out of the EEPROM */
1056         if (e1000_read_mac_addr(hw) < 0) {
1057                 device_printf(dev, "EEPROM read error while reading MAC"
1058                               " address\n");
1059                 error = EIO;
1060                 goto err_late;
1061         }
1062
1063         if (!em_is_valid_ether_addr(hw->mac.addr)) {
1064                 device_printf(dev, "Invalid MAC address\n");
1065                 error = EIO;
1066                 goto err_late;
1067         }
1068
1069         /* Disable ULP support */
1070         e1000_disable_ulp_lpt_lp(hw, TRUE);
1071
1072         /*
1073          * Get Wake-on-Lan and Management info for later use
1074          */
1075         em_get_wakeup(ctx);
1076
1077         /* Enable only WOL MAGIC by default */
1078         scctx->isc_capenable &= ~IFCAP_WOL;
1079         if (adapter->wol != 0)
1080                 scctx->isc_capenable |= IFCAP_WOL_MAGIC;
1081
1082         iflib_set_mac(ctx, hw->mac.addr);
1083
1084         return (0);
1085
1086 err_late:
1087         em_release_hw_control(adapter);
1088 err_pci:
1089         em_free_pci_resources(ctx);
1090         free(adapter->mta, M_DEVBUF);
1091
1092         return (error);
1093 }
1094
1095 static int
1096 em_if_attach_post(if_ctx_t ctx)
1097 {
1098         struct adapter *adapter = iflib_get_softc(ctx);
1099         struct e1000_hw *hw = &adapter->hw;
1100         int error = 0;
1101
1102         /* Setup OS specific network interface */
1103         error = em_setup_interface(ctx);
1104         if (error != 0) {
1105                 device_printf(adapter->dev, "Interface setup failed: %d\n", error);
1106                 goto err_late;
1107         }
1108
1109         em_reset(ctx);
1110
1111         /* Initialize statistics */
1112         em_update_stats_counters(adapter);
1113         hw->mac.get_link_status = 1;
1114         em_if_update_admin_status(ctx);
1115         em_add_hw_stats(adapter);
1116
1117         /* Non-AMT based hardware can now take control from firmware */
1118         if (adapter->has_manage && !adapter->has_amt)
1119                 em_get_hw_control(adapter);
1120
1121         INIT_DEBUGOUT("em_if_attach_post: end");
1122
1123         return (0);
1124
1125 err_late:
1126         /* upon attach_post() error, iflib calls _if_detach() to free resources. */
1127         return (error);
1128 }
1129
1130 /*********************************************************************
1131  *  Device removal routine
1132  *
1133  *  The detach entry point is called when the driver is being removed.
1134  *  This routine stops the adapter and deallocates all the resources
1135  *  that were allocated for driver operation.
1136  *
1137  *  return 0 on success, positive on failure
1138  *********************************************************************/
1139 static int
1140 em_if_detach(if_ctx_t ctx)
1141 {
1142         struct adapter  *adapter = iflib_get_softc(ctx);
1143
1144         INIT_DEBUGOUT("em_if_detach: begin");
1145
1146         e1000_phy_hw_reset(&adapter->hw);
1147
1148         em_release_manageability(adapter);
1149         em_release_hw_control(adapter);
1150         em_free_pci_resources(ctx);
1151         free(adapter->mta, M_DEVBUF);
1152         adapter->mta = NULL;
1153
1154         return (0);
1155 }
1156
1157 /*********************************************************************
1158  *
1159  *  Shutdown entry point
1160  *
1161  **********************************************************************/
1162
1163 static int
1164 em_if_shutdown(if_ctx_t ctx)
1165 {
1166         return em_if_suspend(ctx);
1167 }
1168
1169 /*
1170  * Suspend/resume device methods.
1171  */
1172 static int
1173 em_if_suspend(if_ctx_t ctx)
1174 {
1175         struct adapter *adapter = iflib_get_softc(ctx);
1176
1177         em_release_manageability(adapter);
1178         em_release_hw_control(adapter);
1179         em_enable_wakeup(ctx);
1180         return (0);
1181 }
1182
1183 static int
1184 em_if_resume(if_ctx_t ctx)
1185 {
1186         struct adapter *adapter = iflib_get_softc(ctx);
1187
1188         if (adapter->hw.mac.type == e1000_pch2lan)
1189                 e1000_resume_workarounds_pchlan(&adapter->hw);
1190         em_if_init(ctx);
1191         em_init_manageability(adapter);
1192
1193         return(0);
1194 }
1195
1196 static int
1197 em_if_mtu_set(if_ctx_t ctx, uint32_t mtu)
1198 {
1199         int max_frame_size;
1200         struct adapter *adapter = iflib_get_softc(ctx);
1201         if_softc_ctx_t scctx = iflib_get_softc_ctx(ctx);
1202
1203         IOCTL_DEBUGOUT("ioctl rcv'd: SIOCSIFMTU (Set Interface MTU)");
1204
1205         switch (adapter->hw.mac.type) {
1206         case e1000_82571:
1207         case e1000_82572:
1208         case e1000_ich9lan:
1209         case e1000_ich10lan:
1210         case e1000_pch2lan:
1211         case e1000_pch_lpt:
1212         case e1000_pch_spt:
1213         case e1000_pch_cnp:
1214         case e1000_82574:
1215         case e1000_82583:
1216         case e1000_80003es2lan:
1217                 /* 9K Jumbo Frame size */
1218                 max_frame_size = 9234;
1219                 break;
1220         case e1000_pchlan:
1221                 max_frame_size = 4096;
1222                 break;
1223         case e1000_82542:
1224         case e1000_ich8lan:
1225                 /* Adapters that do not support jumbo frames */
1226                 max_frame_size = ETHER_MAX_LEN;
1227                 break;
1228         default:
1229                 if (adapter->hw.mac.type >= igb_mac_min)
1230                         max_frame_size = 9234;
1231                 else /* lem */
1232                         max_frame_size = MAX_JUMBO_FRAME_SIZE;
1233         }
1234         if (mtu > max_frame_size - ETHER_HDR_LEN - ETHER_CRC_LEN) {
1235                 return (EINVAL);
1236         }
1237
1238         scctx->isc_max_frame_size = adapter->hw.mac.max_frame_size =
1239             mtu + ETHER_HDR_LEN + ETHER_CRC_LEN;
1240         return (0);
1241 }
1242
1243 /*********************************************************************
1244  *  Init entry point
1245  *
1246  *  This routine is used in two ways. It is used by the stack as
1247  *  init entry point in network interface structure. It is also used
1248  *  by the driver as a hw/sw initialization routine to get to a
1249  *  consistent state.
1250  *
1251  **********************************************************************/
1252 static void
1253 em_if_init(if_ctx_t ctx)
1254 {
1255         struct adapter *adapter = iflib_get_softc(ctx);
1256         if_softc_ctx_t scctx = adapter->shared;
1257         struct ifnet *ifp = iflib_get_ifp(ctx);
1258         struct em_tx_queue *tx_que;
1259         int i;
1260
1261         INIT_DEBUGOUT("em_if_init: begin");
1262
1263         /* Get the latest mac address, User can use a LAA */
1264         bcopy(if_getlladdr(ifp), adapter->hw.mac.addr,
1265             ETHER_ADDR_LEN);
1266
1267         /* Put the address into the Receive Address Array */
1268         e1000_rar_set(&adapter->hw, adapter->hw.mac.addr, 0);
1269
1270         /*
1271          * With the 82571 adapter, RAR[0] may be overwritten
1272          * when the other port is reset, we make a duplicate
1273          * in RAR[14] for that eventuality, this assures
1274          * the interface continues to function.
1275          */
1276         if (adapter->hw.mac.type == e1000_82571) {
1277                 e1000_set_laa_state_82571(&adapter->hw, TRUE);
1278                 e1000_rar_set(&adapter->hw, adapter->hw.mac.addr,
1279                     E1000_RAR_ENTRIES - 1);
1280         }
1281
1282
1283         /* Initialize the hardware */
1284         em_reset(ctx);
1285         em_if_update_admin_status(ctx);
1286
1287         for (i = 0, tx_que = adapter->tx_queues; i < adapter->tx_num_queues; i++, tx_que++) {
1288                 struct tx_ring *txr = &tx_que->txr;
1289
1290                 txr->tx_rs_cidx = txr->tx_rs_pidx;
1291
1292                 /* Initialize the last processed descriptor to be the end of
1293                  * the ring, rather than the start, so that we avoid an
1294                  * off-by-one error when calculating how many descriptors are
1295                  * done in the credits_update function.
1296                  */
1297                 txr->tx_cidx_processed = scctx->isc_ntxd[0] - 1;
1298         }
1299
1300         /* Setup VLAN support, basic and offload if available */
1301         E1000_WRITE_REG(&adapter->hw, E1000_VET, ETHERTYPE_VLAN);
1302
1303         /* Clear bad data from Rx FIFOs */
1304         if (adapter->hw.mac.type >= igb_mac_min)
1305                 e1000_rx_fifo_flush_82575(&adapter->hw);
1306
1307         /* Configure for OS presence */
1308         em_init_manageability(adapter);
1309
1310         /* Prepare transmit descriptors and buffers */
1311         em_initialize_transmit_unit(ctx);
1312
1313         /* Setup Multicast table */
1314         em_if_multi_set(ctx);
1315
1316         adapter->rx_mbuf_sz = iflib_get_rx_mbuf_sz(ctx);
1317         em_initialize_receive_unit(ctx);
1318
1319         /* Use real VLAN Filter support? */
1320         if (if_getcapenable(ifp) & IFCAP_VLAN_HWTAGGING) {
1321                 if (if_getcapenable(ifp) & IFCAP_VLAN_HWFILTER)
1322                         /* Use real VLAN Filter support */
1323                         em_setup_vlan_hw_support(adapter);
1324                 else {
1325                         u32 ctrl;
1326                         ctrl = E1000_READ_REG(&adapter->hw, E1000_CTRL);
1327                         ctrl |= E1000_CTRL_VME;
1328                         E1000_WRITE_REG(&adapter->hw, E1000_CTRL, ctrl);
1329                 }
1330         } else {
1331                 u32 ctrl;
1332                 ctrl = E1000_READ_REG(&adapter->hw, E1000_CTRL);
1333                 ctrl &= ~E1000_CTRL_VME;
1334                 E1000_WRITE_REG(&adapter->hw, E1000_CTRL, ctrl);
1335         }
1336
1337         /* Don't lose promiscuous settings */
1338         em_if_set_promisc(ctx, if_getflags(ifp));
1339         e1000_clear_hw_cntrs_base_generic(&adapter->hw);
1340
1341         /* MSI-X configuration for 82574 */
1342         if (adapter->hw.mac.type == e1000_82574) {
1343                 int tmp = E1000_READ_REG(&adapter->hw, E1000_CTRL_EXT);
1344
1345                 tmp |= E1000_CTRL_EXT_PBA_CLR;
1346                 E1000_WRITE_REG(&adapter->hw, E1000_CTRL_EXT, tmp);
1347                 /* Set the IVAR - interrupt vector routing. */
1348                 E1000_WRITE_REG(&adapter->hw, E1000_IVAR, adapter->ivars);
1349         } else if (adapter->intr_type == IFLIB_INTR_MSIX) /* Set up queue routing */
1350                 igb_configure_queues(adapter);
1351
1352         /* this clears any pending interrupts */
1353         E1000_READ_REG(&adapter->hw, E1000_ICR);
1354         E1000_WRITE_REG(&adapter->hw, E1000_ICS, E1000_ICS_LSC);
1355
1356         /* AMT based hardware can now take control from firmware */
1357         if (adapter->has_manage && adapter->has_amt)
1358                 em_get_hw_control(adapter);
1359
1360         /* Set Energy Efficient Ethernet */
1361         if (adapter->hw.mac.type >= igb_mac_min &&
1362             adapter->hw.phy.media_type == e1000_media_type_copper) {
1363                 if (adapter->hw.mac.type == e1000_i354)
1364                         e1000_set_eee_i354(&adapter->hw, TRUE, TRUE);
1365                 else
1366                         e1000_set_eee_i350(&adapter->hw, TRUE, TRUE);
1367         }
1368 }
1369
1370 /*********************************************************************
1371  *
1372  *  Fast Legacy/MSI Combined Interrupt Service routine
1373  *
1374  *********************************************************************/
1375 int
1376 em_intr(void *arg)
1377 {
1378         struct adapter *adapter = arg;
1379         if_ctx_t ctx = adapter->ctx;
1380         u32 reg_icr;
1381
1382         reg_icr = E1000_READ_REG(&adapter->hw, E1000_ICR);
1383
1384         /* Hot eject? */
1385         if (reg_icr == 0xffffffff)
1386                 return FILTER_STRAY;
1387
1388         /* Definitely not our interrupt. */
1389         if (reg_icr == 0x0)
1390                 return FILTER_STRAY;
1391
1392         /*
1393          * Starting with the 82571 chip, bit 31 should be used to
1394          * determine whether the interrupt belongs to us.
1395          */
1396         if (adapter->hw.mac.type >= e1000_82571 &&
1397             (reg_icr & E1000_ICR_INT_ASSERTED) == 0)
1398                 return FILTER_STRAY;
1399
1400         /*
1401          * Only MSI-X interrupts have one-shot behavior by taking advantage
1402          * of the EIAC register.  Thus, explicitly disable interrupts.  This
1403          * also works around the MSI message reordering errata on certain
1404          * systems.
1405          */
1406         IFDI_INTR_DISABLE(ctx);
1407
1408         /* Link status change */
1409         if (reg_icr & (E1000_ICR_RXSEQ | E1000_ICR_LSC))
1410                 em_handle_link(ctx);
1411
1412         if (reg_icr & E1000_ICR_RXO)
1413                 adapter->rx_overruns++;
1414
1415         return (FILTER_SCHEDULE_THREAD);
1416 }
1417
1418 static int
1419 em_if_rx_queue_intr_enable(if_ctx_t ctx, uint16_t rxqid)
1420 {
1421         struct adapter *adapter = iflib_get_softc(ctx);
1422         struct em_rx_queue *rxq = &adapter->rx_queues[rxqid];
1423
1424         E1000_WRITE_REG(&adapter->hw, E1000_IMS, rxq->eims);
1425         return (0);
1426 }
1427
1428 static int
1429 em_if_tx_queue_intr_enable(if_ctx_t ctx, uint16_t txqid)
1430 {
1431         struct adapter *adapter = iflib_get_softc(ctx);
1432         struct em_tx_queue *txq = &adapter->tx_queues[txqid];
1433
1434         E1000_WRITE_REG(&adapter->hw, E1000_IMS, txq->eims);
1435         return (0);
1436 }
1437
1438 static int
1439 igb_if_rx_queue_intr_enable(if_ctx_t ctx, uint16_t rxqid)
1440 {
1441         struct adapter *adapter = iflib_get_softc(ctx);
1442         struct em_rx_queue *rxq = &adapter->rx_queues[rxqid];
1443
1444         E1000_WRITE_REG(&adapter->hw, E1000_EIMS, rxq->eims);
1445         return (0);
1446 }
1447
1448 static int
1449 igb_if_tx_queue_intr_enable(if_ctx_t ctx, uint16_t txqid)
1450 {
1451         struct adapter *adapter = iflib_get_softc(ctx);
1452         struct em_tx_queue *txq = &adapter->tx_queues[txqid];
1453
1454         E1000_WRITE_REG(&adapter->hw, E1000_EIMS, txq->eims);
1455         return (0);
1456 }
1457
1458 /*********************************************************************
1459  *
1460  *  MSI-X RX Interrupt Service routine
1461  *
1462  **********************************************************************/
1463 static int
1464 em_msix_que(void *arg)
1465 {
1466         struct em_rx_queue *que = arg;
1467
1468         ++que->irqs;
1469
1470         return (FILTER_SCHEDULE_THREAD);
1471 }
1472
1473 /*********************************************************************
1474  *
1475  *  MSI-X Link Fast Interrupt Service routine
1476  *
1477  **********************************************************************/
1478 static int
1479 em_msix_link(void *arg)
1480 {
1481         struct adapter *adapter = arg;
1482         u32 reg_icr;
1483
1484         ++adapter->link_irq;
1485         MPASS(adapter->hw.back != NULL);
1486         reg_icr = E1000_READ_REG(&adapter->hw, E1000_ICR);
1487
1488         if (reg_icr & E1000_ICR_RXO)
1489                 adapter->rx_overruns++;
1490
1491         if (reg_icr & (E1000_ICR_RXSEQ | E1000_ICR_LSC)) {
1492                 em_handle_link(adapter->ctx);
1493         } else if (adapter->hw.mac.type == e1000_82574) {
1494                 /* Only re-arm 82574 if em_if_update_admin_status() won't. */
1495                 E1000_WRITE_REG(&adapter->hw, E1000_IMS, EM_MSIX_LINK |
1496                     E1000_IMS_LSC);
1497         }
1498
1499         if (adapter->hw.mac.type == e1000_82574) {
1500                 /*
1501                  * Because we must read the ICR for this interrupt it may
1502                  * clear other causes using autoclear, for this reason we
1503                  * simply create a soft interrupt for all these vectors.
1504                  */
1505                 if (reg_icr)
1506                         E1000_WRITE_REG(&adapter->hw, E1000_ICS, adapter->ims);
1507         } else {
1508                 /* Re-arm unconditionally */
1509                 E1000_WRITE_REG(&adapter->hw, E1000_IMS, E1000_IMS_LSC);
1510                 E1000_WRITE_REG(&adapter->hw, E1000_EIMS, adapter->link_mask);
1511         }
1512
1513         return (FILTER_HANDLED);
1514 }
1515
1516 static void
1517 em_handle_link(void *context)
1518 {
1519         if_ctx_t ctx = context;
1520         struct adapter *adapter = iflib_get_softc(ctx);
1521
1522         adapter->hw.mac.get_link_status = 1;
1523         iflib_admin_intr_deferred(ctx);
1524 }
1525
1526 /*********************************************************************
1527  *
1528  *  Media Ioctl callback
1529  *
1530  *  This routine is called whenever the user queries the status of
1531  *  the interface using ifconfig.
1532  *
1533  **********************************************************************/
1534 static void
1535 em_if_media_status(if_ctx_t ctx, struct ifmediareq *ifmr)
1536 {
1537         struct adapter *adapter = iflib_get_softc(ctx);
1538         u_char fiber_type = IFM_1000_SX;
1539
1540         INIT_DEBUGOUT("em_if_media_status: begin");
1541
1542         iflib_admin_intr_deferred(ctx);
1543
1544         ifmr->ifm_status = IFM_AVALID;
1545         ifmr->ifm_active = IFM_ETHER;
1546
1547         if (!adapter->link_active) {
1548                 return;
1549         }
1550
1551         ifmr->ifm_status |= IFM_ACTIVE;
1552
1553         if ((adapter->hw.phy.media_type == e1000_media_type_fiber) ||
1554             (adapter->hw.phy.media_type == e1000_media_type_internal_serdes)) {
1555                 if (adapter->hw.mac.type == e1000_82545)
1556                         fiber_type = IFM_1000_LX;
1557                 ifmr->ifm_active |= fiber_type | IFM_FDX;
1558         } else {
1559                 switch (adapter->link_speed) {
1560                 case 10:
1561                         ifmr->ifm_active |= IFM_10_T;
1562                         break;
1563                 case 100:
1564                         ifmr->ifm_active |= IFM_100_TX;
1565                         break;
1566                 case 1000:
1567                         ifmr->ifm_active |= IFM_1000_T;
1568                         break;
1569                 }
1570                 if (adapter->link_duplex == FULL_DUPLEX)
1571                         ifmr->ifm_active |= IFM_FDX;
1572                 else
1573                         ifmr->ifm_active |= IFM_HDX;
1574         }
1575 }
1576
1577 /*********************************************************************
1578  *
1579  *  Media Ioctl callback
1580  *
1581  *  This routine is called when the user changes speed/duplex using
1582  *  media/mediopt option with ifconfig.
1583  *
1584  **********************************************************************/
1585 static int
1586 em_if_media_change(if_ctx_t ctx)
1587 {
1588         struct adapter *adapter = iflib_get_softc(ctx);
1589         struct ifmedia *ifm = iflib_get_media(ctx);
1590
1591         INIT_DEBUGOUT("em_if_media_change: begin");
1592
1593         if (IFM_TYPE(ifm->ifm_media) != IFM_ETHER)
1594                 return (EINVAL);
1595
1596         switch (IFM_SUBTYPE(ifm->ifm_media)) {
1597         case IFM_AUTO:
1598                 adapter->hw.mac.autoneg = DO_AUTO_NEG;
1599                 adapter->hw.phy.autoneg_advertised = AUTONEG_ADV_DEFAULT;
1600                 break;
1601         case IFM_1000_LX:
1602         case IFM_1000_SX:
1603         case IFM_1000_T:
1604                 adapter->hw.mac.autoneg = DO_AUTO_NEG;
1605                 adapter->hw.phy.autoneg_advertised = ADVERTISE_1000_FULL;
1606                 break;
1607         case IFM_100_TX:
1608                 adapter->hw.mac.autoneg = FALSE;
1609                 adapter->hw.phy.autoneg_advertised = 0;
1610                 if ((ifm->ifm_media & IFM_GMASK) == IFM_FDX)
1611                         adapter->hw.mac.forced_speed_duplex = ADVERTISE_100_FULL;
1612                 else
1613                         adapter->hw.mac.forced_speed_duplex = ADVERTISE_100_HALF;
1614                 break;
1615         case IFM_10_T:
1616                 adapter->hw.mac.autoneg = FALSE;
1617                 adapter->hw.phy.autoneg_advertised = 0;
1618                 if ((ifm->ifm_media & IFM_GMASK) == IFM_FDX)
1619                         adapter->hw.mac.forced_speed_duplex = ADVERTISE_10_FULL;
1620                 else
1621                         adapter->hw.mac.forced_speed_duplex = ADVERTISE_10_HALF;
1622                 break;
1623         default:
1624                 device_printf(adapter->dev, "Unsupported media type\n");
1625         }
1626
1627         em_if_init(ctx);
1628
1629         return (0);
1630 }
1631
1632 static int
1633 em_if_set_promisc(if_ctx_t ctx, int flags)
1634 {
1635         struct adapter *adapter = iflib_get_softc(ctx);
1636         u32 reg_rctl;
1637
1638         em_disable_promisc(ctx);
1639
1640         reg_rctl = E1000_READ_REG(&adapter->hw, E1000_RCTL);
1641
1642         if (flags & IFF_PROMISC) {
1643                 reg_rctl |= (E1000_RCTL_UPE | E1000_RCTL_MPE);
1644                 /* Turn this on if you want to see bad packets */
1645                 if (em_debug_sbp)
1646                         reg_rctl |= E1000_RCTL_SBP;
1647                 E1000_WRITE_REG(&adapter->hw, E1000_RCTL, reg_rctl);
1648         } else if (flags & IFF_ALLMULTI) {
1649                 reg_rctl |= E1000_RCTL_MPE;
1650                 reg_rctl &= ~E1000_RCTL_UPE;
1651                 E1000_WRITE_REG(&adapter->hw, E1000_RCTL, reg_rctl);
1652         }
1653         return (0);
1654 }
1655
1656 static void
1657 em_disable_promisc(if_ctx_t ctx)
1658 {
1659         struct adapter *adapter = iflib_get_softc(ctx);
1660         struct ifnet *ifp = iflib_get_ifp(ctx);
1661         u32 reg_rctl;
1662         int mcnt = 0;
1663
1664         reg_rctl = E1000_READ_REG(&adapter->hw, E1000_RCTL);
1665         reg_rctl &= (~E1000_RCTL_UPE);
1666         if (if_getflags(ifp) & IFF_ALLMULTI)
1667                 mcnt = MAX_NUM_MULTICAST_ADDRESSES;
1668         else
1669                 mcnt = if_llmaddr_count(ifp);
1670         /* Don't disable if in MAX groups */
1671         if (mcnt < MAX_NUM_MULTICAST_ADDRESSES)
1672                 reg_rctl &=  (~E1000_RCTL_MPE);
1673         reg_rctl &=  (~E1000_RCTL_SBP);
1674         E1000_WRITE_REG(&adapter->hw, E1000_RCTL, reg_rctl);
1675 }
1676
1677
1678 static u_int
1679 em_copy_maddr(void *arg, struct sockaddr_dl *sdl, u_int cnt)
1680 {
1681         u8 *mta = arg;
1682
1683         if (cnt == MAX_NUM_MULTICAST_ADDRESSES)
1684                 return (1);
1685
1686         bcopy(LLADDR(sdl), &mta[cnt * ETHER_ADDR_LEN], ETHER_ADDR_LEN);
1687
1688         return (1);
1689 }
1690
1691 /*********************************************************************
1692  *  Multicast Update
1693  *
1694  *  This routine is called whenever multicast address list is updated.
1695  *
1696  **********************************************************************/
1697
1698 static void
1699 em_if_multi_set(if_ctx_t ctx)
1700 {
1701         struct adapter *adapter = iflib_get_softc(ctx);
1702         struct ifnet *ifp = iflib_get_ifp(ctx);
1703         u32 reg_rctl = 0;
1704         u8  *mta; /* Multicast array memory */
1705         int mcnt = 0;
1706
1707         IOCTL_DEBUGOUT("em_set_multi: begin");
1708
1709         mta = adapter->mta;
1710         bzero(mta, sizeof(u8) * ETHER_ADDR_LEN * MAX_NUM_MULTICAST_ADDRESSES);
1711
1712         if (adapter->hw.mac.type == e1000_82542 &&
1713             adapter->hw.revision_id == E1000_REVISION_2) {
1714                 reg_rctl = E1000_READ_REG(&adapter->hw, E1000_RCTL);
1715                 if (adapter->hw.bus.pci_cmd_word & CMD_MEM_WRT_INVALIDATE)
1716                         e1000_pci_clear_mwi(&adapter->hw);
1717                 reg_rctl |= E1000_RCTL_RST;
1718                 E1000_WRITE_REG(&adapter->hw, E1000_RCTL, reg_rctl);
1719                 msec_delay(5);
1720         }
1721
1722         mcnt = if_foreach_llmaddr(ifp, em_copy_maddr, mta);
1723
1724         if (mcnt >= MAX_NUM_MULTICAST_ADDRESSES) {
1725                 reg_rctl = E1000_READ_REG(&adapter->hw, E1000_RCTL);
1726                 reg_rctl |= E1000_RCTL_MPE;
1727                 E1000_WRITE_REG(&adapter->hw, E1000_RCTL, reg_rctl);
1728         } else
1729                 e1000_update_mc_addr_list(&adapter->hw, mta, mcnt);
1730
1731         if (adapter->hw.mac.type == e1000_82542 &&
1732             adapter->hw.revision_id == E1000_REVISION_2) {
1733                 reg_rctl = E1000_READ_REG(&adapter->hw, E1000_RCTL);
1734                 reg_rctl &= ~E1000_RCTL_RST;
1735                 E1000_WRITE_REG(&adapter->hw, E1000_RCTL, reg_rctl);
1736                 msec_delay(5);
1737                 if (adapter->hw.bus.pci_cmd_word & CMD_MEM_WRT_INVALIDATE)
1738                         e1000_pci_set_mwi(&adapter->hw);
1739         }
1740 }
1741
1742 /*********************************************************************
1743  *  Timer routine
1744  *
1745  *  This routine schedules em_if_update_admin_status() to check for
1746  *  link status and to gather statistics as well as to perform some
1747  *  controller-specific hardware patting.
1748  *
1749  **********************************************************************/
1750 static void
1751 em_if_timer(if_ctx_t ctx, uint16_t qid)
1752 {
1753
1754         if (qid != 0)
1755                 return;
1756
1757         iflib_admin_intr_deferred(ctx);
1758 }
1759
1760 static void
1761 em_if_update_admin_status(if_ctx_t ctx)
1762 {
1763         struct adapter *adapter = iflib_get_softc(ctx);
1764         struct e1000_hw *hw = &adapter->hw;
1765         device_t dev = iflib_get_dev(ctx);
1766         u32 link_check, thstat, ctrl;
1767
1768         link_check = thstat = ctrl = 0;
1769         /* Get the cached link value or read phy for real */
1770         switch (hw->phy.media_type) {
1771         case e1000_media_type_copper:
1772                 if (hw->mac.get_link_status) {
1773                         if (hw->mac.type == e1000_pch_spt)
1774                                 msec_delay(50);
1775                         /* Do the work to read phy */
1776                         e1000_check_for_link(hw);
1777                         link_check = !hw->mac.get_link_status;
1778                         if (link_check) /* ESB2 fix */
1779                                 e1000_cfg_on_link_up(hw);
1780                 } else {
1781                         link_check = TRUE;
1782                 }
1783                 break;
1784         case e1000_media_type_fiber:
1785                 e1000_check_for_link(hw);
1786                 link_check = (E1000_READ_REG(hw, E1000_STATUS) &
1787                             E1000_STATUS_LU);
1788                 break;
1789         case e1000_media_type_internal_serdes:
1790                 e1000_check_for_link(hw);
1791                 link_check = hw->mac.serdes_has_link;
1792                 break;
1793         /* VF device is type_unknown */
1794         case e1000_media_type_unknown:
1795                 e1000_check_for_link(hw);
1796                 link_check = !hw->mac.get_link_status;
1797                 /* FALLTHROUGH */
1798         default:
1799                 break;
1800         }
1801
1802         /* Check for thermal downshift or shutdown */
1803         if (hw->mac.type == e1000_i350) {
1804                 thstat = E1000_READ_REG(hw, E1000_THSTAT);
1805                 ctrl = E1000_READ_REG(hw, E1000_CTRL_EXT);
1806         }
1807
1808         /* Now check for a transition */
1809         if (link_check && (adapter->link_active == 0)) {
1810                 e1000_get_speed_and_duplex(hw, &adapter->link_speed,
1811                     &adapter->link_duplex);
1812                 /* Check if we must disable SPEED_MODE bit on PCI-E */
1813                 if ((adapter->link_speed != SPEED_1000) &&
1814                     ((hw->mac.type == e1000_82571) ||
1815                     (hw->mac.type == e1000_82572))) {
1816                         int tarc0;
1817                         tarc0 = E1000_READ_REG(hw, E1000_TARC(0));
1818                         tarc0 &= ~TARC_SPEED_MODE_BIT;
1819                         E1000_WRITE_REG(hw, E1000_TARC(0), tarc0);
1820                 }
1821                 if (bootverbose)
1822                         device_printf(dev, "Link is up %d Mbps %s\n",
1823                             adapter->link_speed,
1824                             ((adapter->link_duplex == FULL_DUPLEX) ?
1825                             "Full Duplex" : "Half Duplex"));
1826                 adapter->link_active = 1;
1827                 adapter->smartspeed = 0;
1828                 if ((ctrl & E1000_CTRL_EXT_LINK_MODE_MASK) ==
1829                     E1000_CTRL_EXT_LINK_MODE_GMII &&
1830                     (thstat & E1000_THSTAT_LINK_THROTTLE))
1831                         device_printf(dev, "Link: thermal downshift\n");
1832                 /* Delay Link Up for Phy update */
1833                 if (((hw->mac.type == e1000_i210) ||
1834                     (hw->mac.type == e1000_i211)) &&
1835                     (hw->phy.id == I210_I_PHY_ID))
1836                         msec_delay(I210_LINK_DELAY);
1837                 /* Reset if the media type changed. */
1838                 if (hw->dev_spec._82575.media_changed &&
1839                     hw->mac.type >= igb_mac_min) {
1840                         hw->dev_spec._82575.media_changed = false;
1841                         adapter->flags |= IGB_MEDIA_RESET;
1842                         em_reset(ctx);
1843                 }
1844                 iflib_link_state_change(ctx, LINK_STATE_UP,
1845                     IF_Mbps(adapter->link_speed));
1846         } else if (!link_check && (adapter->link_active == 1)) {
1847                 adapter->link_speed = 0;
1848                 adapter->link_duplex = 0;
1849                 adapter->link_active = 0;
1850                 iflib_link_state_change(ctx, LINK_STATE_DOWN, 0);
1851         }
1852         em_update_stats_counters(adapter);
1853
1854         /* Reset LAA into RAR[0] on 82571 */
1855         if (hw->mac.type == e1000_82571 && e1000_get_laa_state_82571(hw))
1856                 e1000_rar_set(hw, hw->mac.addr, 0);
1857
1858         if (hw->mac.type < em_mac_min)
1859                 lem_smartspeed(adapter);
1860         else if (hw->mac.type == e1000_82574 &&
1861             adapter->intr_type == IFLIB_INTR_MSIX)
1862                 E1000_WRITE_REG(hw, E1000_IMS, EM_MSIX_LINK | E1000_IMS_LSC);
1863 }
1864
1865 static void
1866 em_if_watchdog_reset(if_ctx_t ctx)
1867 {
1868         struct adapter *adapter = iflib_get_softc(ctx);
1869
1870         /*
1871          * Just count the event; iflib(4) will already trigger a
1872          * sufficient reset of the controller.
1873          */
1874         adapter->watchdog_events++;
1875 }
1876
1877 /*********************************************************************
1878  *
1879  *  This routine disables all traffic on the adapter by issuing a
1880  *  global reset on the MAC.
1881  *
1882  **********************************************************************/
1883 static void
1884 em_if_stop(if_ctx_t ctx)
1885 {
1886         struct adapter *adapter = iflib_get_softc(ctx);
1887
1888         INIT_DEBUGOUT("em_if_stop: begin");
1889
1890         e1000_reset_hw(&adapter->hw);
1891         if (adapter->hw.mac.type >= e1000_82544)
1892                 E1000_WRITE_REG(&adapter->hw, E1000_WUFC, 0);
1893
1894         e1000_led_off(&adapter->hw);
1895         e1000_cleanup_led(&adapter->hw);
1896 }
1897
1898 /*********************************************************************
1899  *
1900  *  Determine hardware revision.
1901  *
1902  **********************************************************************/
1903 static void
1904 em_identify_hardware(if_ctx_t ctx)
1905 {
1906         device_t dev = iflib_get_dev(ctx);
1907         struct adapter *adapter = iflib_get_softc(ctx);
1908
1909         /* Make sure our PCI config space has the necessary stuff set */
1910         adapter->hw.bus.pci_cmd_word = pci_read_config(dev, PCIR_COMMAND, 2);
1911
1912         /* Save off the information about this board */
1913         adapter->hw.vendor_id = pci_get_vendor(dev);
1914         adapter->hw.device_id = pci_get_device(dev);
1915         adapter->hw.revision_id = pci_read_config(dev, PCIR_REVID, 1);
1916         adapter->hw.subsystem_vendor_id =
1917             pci_read_config(dev, PCIR_SUBVEND_0, 2);
1918         adapter->hw.subsystem_device_id =
1919             pci_read_config(dev, PCIR_SUBDEV_0, 2);
1920
1921         /* Do Shared Code Init and Setup */
1922         if (e1000_set_mac_type(&adapter->hw)) {
1923                 device_printf(dev, "Setup init failure\n");
1924                 return;
1925         }
1926 }
1927
1928 static int
1929 em_allocate_pci_resources(if_ctx_t ctx)
1930 {
1931         struct adapter *adapter = iflib_get_softc(ctx);
1932         device_t dev = iflib_get_dev(ctx);
1933         int rid, val;
1934
1935         rid = PCIR_BAR(0);
1936         adapter->memory = bus_alloc_resource_any(dev, SYS_RES_MEMORY,
1937             &rid, RF_ACTIVE);
1938         if (adapter->memory == NULL) {
1939                 device_printf(dev, "Unable to allocate bus resource: memory\n");
1940                 return (ENXIO);
1941         }
1942         adapter->osdep.mem_bus_space_tag = rman_get_bustag(adapter->memory);
1943         adapter->osdep.mem_bus_space_handle =
1944             rman_get_bushandle(adapter->memory);
1945         adapter->hw.hw_addr = (u8 *)&adapter->osdep.mem_bus_space_handle;
1946
1947         /* Only older adapters use IO mapping */
1948         if (adapter->hw.mac.type < em_mac_min &&
1949             adapter->hw.mac.type > e1000_82543) {
1950                 /* Figure our where our IO BAR is ? */
1951                 for (rid = PCIR_BAR(0); rid < PCIR_CIS;) {
1952                         val = pci_read_config(dev, rid, 4);
1953                         if (EM_BAR_TYPE(val) == EM_BAR_TYPE_IO) {
1954                                 break;
1955                         }
1956                         rid += 4;
1957                         /* check for 64bit BAR */
1958                         if (EM_BAR_MEM_TYPE(val) == EM_BAR_MEM_TYPE_64BIT)
1959                                 rid += 4;
1960                 }
1961                 if (rid >= PCIR_CIS) {
1962                         device_printf(dev, "Unable to locate IO BAR\n");
1963                         return (ENXIO);
1964                 }
1965                 adapter->ioport = bus_alloc_resource_any(dev, SYS_RES_IOPORT,
1966                     &rid, RF_ACTIVE);
1967                 if (adapter->ioport == NULL) {
1968                         device_printf(dev, "Unable to allocate bus resource: "
1969                             "ioport\n");
1970                         return (ENXIO);
1971                 }
1972                 adapter->hw.io_base = 0;
1973                 adapter->osdep.io_bus_space_tag =
1974                     rman_get_bustag(adapter->ioport);
1975                 adapter->osdep.io_bus_space_handle =
1976                     rman_get_bushandle(adapter->ioport);
1977         }
1978
1979         adapter->hw.back = &adapter->osdep;
1980
1981         return (0);
1982 }
1983
1984 /*********************************************************************
1985  *
1986  *  Set up the MSI-X Interrupt handlers
1987  *
1988  **********************************************************************/
1989 static int
1990 em_if_msix_intr_assign(if_ctx_t ctx, int msix)
1991 {
1992         struct adapter *adapter = iflib_get_softc(ctx);
1993         struct em_rx_queue *rx_que = adapter->rx_queues;
1994         struct em_tx_queue *tx_que = adapter->tx_queues;
1995         int error, rid, i, vector = 0, rx_vectors;
1996         char buf[16];
1997
1998         /* First set up ring resources */
1999         for (i = 0; i < adapter->rx_num_queues; i++, rx_que++, vector++) {
2000                 rid = vector + 1;
2001                 snprintf(buf, sizeof(buf), "rxq%d", i);
2002                 error = iflib_irq_alloc_generic(ctx, &rx_que->que_irq, rid, IFLIB_INTR_RXTX, em_msix_que, rx_que, rx_que->me, buf);
2003                 if (error) {
2004                         device_printf(iflib_get_dev(ctx), "Failed to allocate que int %d err: %d", i, error);
2005                         adapter->rx_num_queues = i + 1;
2006                         goto fail;
2007                 }
2008
2009                 rx_que->msix =  vector;
2010
2011                 /*
2012                  * Set the bit to enable interrupt
2013                  * in E1000_IMS -- bits 20 and 21
2014                  * are for RX0 and RX1, note this has
2015                  * NOTHING to do with the MSI-X vector
2016                  */
2017                 if (adapter->hw.mac.type == e1000_82574) {
2018                         rx_que->eims = 1 << (20 + i);
2019                         adapter->ims |= rx_que->eims;
2020                         adapter->ivars |= (8 | rx_que->msix) << (i * 4);
2021                 } else if (adapter->hw.mac.type == e1000_82575)
2022                         rx_que->eims = E1000_EICR_TX_QUEUE0 << vector;
2023                 else
2024                         rx_que->eims = 1 << vector;
2025         }
2026         rx_vectors = vector;
2027
2028         vector = 0;
2029         for (i = 0; i < adapter->tx_num_queues; i++, tx_que++, vector++) {
2030                 snprintf(buf, sizeof(buf), "txq%d", i);
2031                 tx_que = &adapter->tx_queues[i];
2032                 iflib_softirq_alloc_generic(ctx,
2033                     &adapter->rx_queues[i % adapter->rx_num_queues].que_irq,
2034                     IFLIB_INTR_TX, tx_que, tx_que->me, buf);
2035
2036                 tx_que->msix = (vector % adapter->rx_num_queues);
2037
2038                 /*
2039                  * Set the bit to enable interrupt
2040                  * in E1000_IMS -- bits 22 and 23
2041                  * are for TX0 and TX1, note this has
2042                  * NOTHING to do with the MSI-X vector
2043                  */
2044                 if (adapter->hw.mac.type == e1000_82574) {
2045                         tx_que->eims = 1 << (22 + i);
2046                         adapter->ims |= tx_que->eims;
2047                         adapter->ivars |= (8 | tx_que->msix) << (8 + (i * 4));
2048                 } else if (adapter->hw.mac.type == e1000_82575) {
2049                         tx_que->eims = E1000_EICR_TX_QUEUE0 << i;
2050                 } else {
2051                         tx_que->eims = 1 << i;
2052                 }
2053         }
2054
2055         /* Link interrupt */
2056         rid = rx_vectors + 1;
2057         error = iflib_irq_alloc_generic(ctx, &adapter->irq, rid, IFLIB_INTR_ADMIN, em_msix_link, adapter, 0, "aq");
2058
2059         if (error) {
2060                 device_printf(iflib_get_dev(ctx), "Failed to register admin handler");
2061                 goto fail;
2062         }
2063         adapter->linkvec = rx_vectors;
2064         if (adapter->hw.mac.type < igb_mac_min) {
2065                 adapter->ivars |=  (8 | rx_vectors) << 16;
2066                 adapter->ivars |= 0x80000000;
2067         }
2068         return (0);
2069 fail:
2070         iflib_irq_free(ctx, &adapter->irq);
2071         rx_que = adapter->rx_queues;
2072         for (int i = 0; i < adapter->rx_num_queues; i++, rx_que++)
2073                 iflib_irq_free(ctx, &rx_que->que_irq);
2074         return (error);
2075 }
2076
2077 static void
2078 igb_configure_queues(struct adapter *adapter)
2079 {
2080         struct e1000_hw *hw = &adapter->hw;
2081         struct em_rx_queue *rx_que;
2082         struct em_tx_queue *tx_que;
2083         u32 tmp, ivar = 0, newitr = 0;
2084
2085         /* First turn on RSS capability */
2086         if (hw->mac.type != e1000_82575)
2087                 E1000_WRITE_REG(hw, E1000_GPIE,
2088                     E1000_GPIE_MSIX_MODE | E1000_GPIE_EIAME |
2089                     E1000_GPIE_PBA | E1000_GPIE_NSICR);
2090
2091         /* Turn on MSI-X */
2092         switch (hw->mac.type) {
2093         case e1000_82580:
2094         case e1000_i350:
2095         case e1000_i354:
2096         case e1000_i210:
2097         case e1000_i211:
2098         case e1000_vfadapt:
2099         case e1000_vfadapt_i350:
2100                 /* RX entries */
2101                 for (int i = 0; i < adapter->rx_num_queues; i++) {
2102                         u32 index = i >> 1;
2103                         ivar = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index);
2104                         rx_que = &adapter->rx_queues[i];
2105                         if (i & 1) {
2106                                 ivar &= 0xFF00FFFF;
2107                                 ivar |= (rx_que->msix | E1000_IVAR_VALID) << 16;
2108                         } else {
2109                                 ivar &= 0xFFFFFF00;
2110                                 ivar |= rx_que->msix | E1000_IVAR_VALID;
2111                         }
2112                         E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, ivar);
2113                 }
2114                 /* TX entries */
2115                 for (int i = 0; i < adapter->tx_num_queues; i++) {
2116                         u32 index = i >> 1;
2117                         ivar = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index);
2118                         tx_que = &adapter->tx_queues[i];
2119                         if (i & 1) {
2120                                 ivar &= 0x00FFFFFF;
2121                                 ivar |= (tx_que->msix | E1000_IVAR_VALID) << 24;
2122                         } else {
2123                                 ivar &= 0xFFFF00FF;
2124                                 ivar |= (tx_que->msix | E1000_IVAR_VALID) << 8;
2125                         }
2126                         E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, ivar);
2127                         adapter->que_mask |= tx_que->eims;
2128                 }
2129
2130                 /* And for the link interrupt */
2131                 ivar = (adapter->linkvec | E1000_IVAR_VALID) << 8;
2132                 adapter->link_mask = 1 << adapter->linkvec;
2133                 E1000_WRITE_REG(hw, E1000_IVAR_MISC, ivar);
2134                 break;
2135         case e1000_82576:
2136                 /* RX entries */
2137                 for (int i = 0; i < adapter->rx_num_queues; i++) {
2138                         u32 index = i & 0x7; /* Each IVAR has two entries */
2139                         ivar = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index);
2140                         rx_que = &adapter->rx_queues[i];
2141                         if (i < 8) {
2142                                 ivar &= 0xFFFFFF00;
2143                                 ivar |= rx_que->msix | E1000_IVAR_VALID;
2144                         } else {
2145                                 ivar &= 0xFF00FFFF;
2146                                 ivar |= (rx_que->msix | E1000_IVAR_VALID) << 16;
2147                         }
2148                         E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, ivar);
2149                         adapter->que_mask |= rx_que->eims;
2150                 }
2151                 /* TX entries */
2152                 for (int i = 0; i < adapter->tx_num_queues; i++) {
2153                         u32 index = i & 0x7; /* Each IVAR has two entries */
2154                         ivar = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index);
2155                         tx_que = &adapter->tx_queues[i];
2156                         if (i < 8) {
2157                                 ivar &= 0xFFFF00FF;
2158                                 ivar |= (tx_que->msix | E1000_IVAR_VALID) << 8;
2159                         } else {
2160                                 ivar &= 0x00FFFFFF;
2161                                 ivar |= (tx_que->msix | E1000_IVAR_VALID) << 24;
2162                         }
2163                         E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, ivar);
2164                         adapter->que_mask |= tx_que->eims;
2165                 }
2166
2167                 /* And for the link interrupt */
2168                 ivar = (adapter->linkvec | E1000_IVAR_VALID) << 8;
2169                 adapter->link_mask = 1 << adapter->linkvec;
2170                 E1000_WRITE_REG(hw, E1000_IVAR_MISC, ivar);
2171                 break;
2172
2173         case e1000_82575:
2174                 /* enable MSI-X support*/
2175                 tmp = E1000_READ_REG(hw, E1000_CTRL_EXT);
2176                 tmp |= E1000_CTRL_EXT_PBA_CLR;
2177                 /* Auto-Mask interrupts upon ICR read. */
2178                 tmp |= E1000_CTRL_EXT_EIAME;
2179                 tmp |= E1000_CTRL_EXT_IRCA;
2180                 E1000_WRITE_REG(hw, E1000_CTRL_EXT, tmp);
2181
2182                 /* Queues */
2183                 for (int i = 0; i < adapter->rx_num_queues; i++) {
2184                         rx_que = &adapter->rx_queues[i];
2185                         tmp = E1000_EICR_RX_QUEUE0 << i;
2186                         tmp |= E1000_EICR_TX_QUEUE0 << i;
2187                         rx_que->eims = tmp;
2188                         E1000_WRITE_REG_ARRAY(hw, E1000_MSIXBM(0),
2189                             i, rx_que->eims);
2190                         adapter->que_mask |= rx_que->eims;
2191                 }
2192
2193                 /* Link */
2194                 E1000_WRITE_REG(hw, E1000_MSIXBM(adapter->linkvec),
2195                     E1000_EIMS_OTHER);
2196                 adapter->link_mask |= E1000_EIMS_OTHER;
2197         default:
2198                 break;
2199         }
2200
2201         /* Set the starting interrupt rate */
2202         if (em_max_interrupt_rate > 0)
2203                 newitr = (4000000 / em_max_interrupt_rate) & 0x7FFC;
2204
2205         if (hw->mac.type == e1000_82575)
2206                 newitr |= newitr << 16;
2207         else
2208                 newitr |= E1000_EITR_CNT_IGNR;
2209
2210         for (int i = 0; i < adapter->rx_num_queues; i++) {
2211                 rx_que = &adapter->rx_queues[i];
2212                 E1000_WRITE_REG(hw, E1000_EITR(rx_que->msix), newitr);
2213         }
2214
2215         return;
2216 }
2217
2218 static void
2219 em_free_pci_resources(if_ctx_t ctx)
2220 {
2221         struct adapter *adapter = iflib_get_softc(ctx);
2222         struct em_rx_queue *que = adapter->rx_queues;
2223         device_t dev = iflib_get_dev(ctx);
2224
2225         /* Release all MSI-X queue resources */
2226         if (adapter->intr_type == IFLIB_INTR_MSIX)
2227                 iflib_irq_free(ctx, &adapter->irq);
2228
2229         if (que != NULL) {
2230                 for (int i = 0; i < adapter->rx_num_queues; i++, que++) {
2231                         iflib_irq_free(ctx, &que->que_irq);
2232                 }
2233         }
2234
2235         if (adapter->memory != NULL) {
2236                 bus_release_resource(dev, SYS_RES_MEMORY,
2237                     rman_get_rid(adapter->memory), adapter->memory);
2238                 adapter->memory = NULL;
2239         }
2240
2241         if (adapter->flash != NULL) {
2242                 bus_release_resource(dev, SYS_RES_MEMORY,
2243                     rman_get_rid(adapter->flash), adapter->flash);
2244                 adapter->flash = NULL;
2245         }
2246
2247         if (adapter->ioport != NULL) {
2248                 bus_release_resource(dev, SYS_RES_IOPORT,
2249                     rman_get_rid(adapter->ioport), adapter->ioport);
2250                 adapter->ioport = NULL;
2251         }
2252 }
2253
2254 /* Set up MSI or MSI-X */
2255 static int
2256 em_setup_msix(if_ctx_t ctx)
2257 {
2258         struct adapter *adapter = iflib_get_softc(ctx);
2259
2260         if (adapter->hw.mac.type == e1000_82574) {
2261                 em_enable_vectors_82574(ctx);
2262         }
2263         return (0);
2264 }
2265
2266 /*********************************************************************
2267  *
2268  *  Workaround for SmartSpeed on 82541 and 82547 controllers
2269  *
2270  **********************************************************************/
2271 static void
2272 lem_smartspeed(struct adapter *adapter)
2273 {
2274         u16 phy_tmp;
2275
2276         if (adapter->link_active || (adapter->hw.phy.type != e1000_phy_igp) ||
2277             adapter->hw.mac.autoneg == 0 ||
2278             (adapter->hw.phy.autoneg_advertised & ADVERTISE_1000_FULL) == 0)
2279                 return;
2280
2281         if (adapter->smartspeed == 0) {
2282                 /* If Master/Slave config fault is asserted twice,
2283                  * we assume back-to-back */
2284                 e1000_read_phy_reg(&adapter->hw, PHY_1000T_STATUS, &phy_tmp);
2285                 if (!(phy_tmp & SR_1000T_MS_CONFIG_FAULT))
2286                         return;
2287                 e1000_read_phy_reg(&adapter->hw, PHY_1000T_STATUS, &phy_tmp);
2288                 if (phy_tmp & SR_1000T_MS_CONFIG_FAULT) {
2289                         e1000_read_phy_reg(&adapter->hw,
2290                             PHY_1000T_CTRL, &phy_tmp);
2291                         if(phy_tmp & CR_1000T_MS_ENABLE) {
2292                                 phy_tmp &= ~CR_1000T_MS_ENABLE;
2293                                 e1000_write_phy_reg(&adapter->hw,
2294                                     PHY_1000T_CTRL, phy_tmp);
2295                                 adapter->smartspeed++;
2296                                 if(adapter->hw.mac.autoneg &&
2297                                    !e1000_copper_link_autoneg(&adapter->hw) &&
2298                                    !e1000_read_phy_reg(&adapter->hw,
2299                                     PHY_CONTROL, &phy_tmp)) {
2300                                         phy_tmp |= (MII_CR_AUTO_NEG_EN |
2301                                                     MII_CR_RESTART_AUTO_NEG);
2302                                         e1000_write_phy_reg(&adapter->hw,
2303                                             PHY_CONTROL, phy_tmp);
2304                                 }
2305                         }
2306                 }
2307                 return;
2308         } else if(adapter->smartspeed == EM_SMARTSPEED_DOWNSHIFT) {
2309                 /* If still no link, perhaps using 2/3 pair cable */
2310                 e1000_read_phy_reg(&adapter->hw, PHY_1000T_CTRL, &phy_tmp);
2311                 phy_tmp |= CR_1000T_MS_ENABLE;
2312                 e1000_write_phy_reg(&adapter->hw, PHY_1000T_CTRL, phy_tmp);
2313                 if(adapter->hw.mac.autoneg &&
2314                    !e1000_copper_link_autoneg(&adapter->hw) &&
2315                    !e1000_read_phy_reg(&adapter->hw, PHY_CONTROL, &phy_tmp)) {
2316                         phy_tmp |= (MII_CR_AUTO_NEG_EN |
2317                                     MII_CR_RESTART_AUTO_NEG);
2318                         e1000_write_phy_reg(&adapter->hw, PHY_CONTROL, phy_tmp);
2319                 }
2320         }
2321         /* Restart process after EM_SMARTSPEED_MAX iterations */
2322         if(adapter->smartspeed++ == EM_SMARTSPEED_MAX)
2323                 adapter->smartspeed = 0;
2324 }
2325
2326 /*********************************************************************
2327  *
2328  *  Initialize the DMA Coalescing feature
2329  *
2330  **********************************************************************/
2331 static void
2332 igb_init_dmac(struct adapter *adapter, u32 pba)
2333 {
2334         device_t        dev = adapter->dev;
2335         struct e1000_hw *hw = &adapter->hw;
2336         u32             dmac, reg = ~E1000_DMACR_DMAC_EN;
2337         u16             hwm;
2338         u16             max_frame_size;
2339
2340         if (hw->mac.type == e1000_i211)
2341                 return;
2342
2343         max_frame_size = adapter->shared->isc_max_frame_size;
2344         if (hw->mac.type > e1000_82580) {
2345
2346                 if (adapter->dmac == 0) { /* Disabling it */
2347                         E1000_WRITE_REG(hw, E1000_DMACR, reg);
2348                         return;
2349                 } else
2350                         device_printf(dev, "DMA Coalescing enabled\n");
2351
2352                 /* Set starting threshold */
2353                 E1000_WRITE_REG(hw, E1000_DMCTXTH, 0);
2354
2355                 hwm = 64 * pba - max_frame_size / 16;
2356                 if (hwm < 64 * (pba - 6))
2357                         hwm = 64 * (pba - 6);
2358                 reg = E1000_READ_REG(hw, E1000_FCRTC);
2359                 reg &= ~E1000_FCRTC_RTH_COAL_MASK;
2360                 reg |= ((hwm << E1000_FCRTC_RTH_COAL_SHIFT)
2361                     & E1000_FCRTC_RTH_COAL_MASK);
2362                 E1000_WRITE_REG(hw, E1000_FCRTC, reg);
2363
2364
2365                 dmac = pba - max_frame_size / 512;
2366                 if (dmac < pba - 10)
2367                         dmac = pba - 10;
2368                 reg = E1000_READ_REG(hw, E1000_DMACR);
2369                 reg &= ~E1000_DMACR_DMACTHR_MASK;
2370                 reg |= ((dmac << E1000_DMACR_DMACTHR_SHIFT)
2371                     & E1000_DMACR_DMACTHR_MASK);
2372
2373                 /* transition to L0x or L1 if available..*/
2374                 reg |= (E1000_DMACR_DMAC_EN | E1000_DMACR_DMAC_LX_MASK);
2375
2376                 /* Check if status is 2.5Gb backplane connection
2377                 * before configuration of watchdog timer, which is
2378                 * in msec values in 12.8usec intervals
2379                 * watchdog timer= msec values in 32usec intervals
2380                 * for non 2.5Gb connection
2381                 */
2382                 if (hw->mac.type == e1000_i354) {
2383                         int status = E1000_READ_REG(hw, E1000_STATUS);
2384                         if ((status & E1000_STATUS_2P5_SKU) &&
2385                             (!(status & E1000_STATUS_2P5_SKU_OVER)))
2386                                 reg |= ((adapter->dmac * 5) >> 6);
2387                         else
2388                                 reg |= (adapter->dmac >> 5);
2389                 } else {
2390                         reg |= (adapter->dmac >> 5);
2391                 }
2392
2393                 E1000_WRITE_REG(hw, E1000_DMACR, reg);
2394
2395                 E1000_WRITE_REG(hw, E1000_DMCRTRH, 0);
2396
2397                 /* Set the interval before transition */
2398                 reg = E1000_READ_REG(hw, E1000_DMCTLX);
2399                 if (hw->mac.type == e1000_i350)
2400                         reg |= IGB_DMCTLX_DCFLUSH_DIS;
2401                 /*
2402                 ** in 2.5Gb connection, TTLX unit is 0.4 usec
2403                 ** which is 0x4*2 = 0xA. But delay is still 4 usec
2404                 */
2405                 if (hw->mac.type == e1000_i354) {
2406                         int status = E1000_READ_REG(hw, E1000_STATUS);
2407                         if ((status & E1000_STATUS_2P5_SKU) &&
2408                             (!(status & E1000_STATUS_2P5_SKU_OVER)))
2409                                 reg |= 0xA;
2410                         else
2411                                 reg |= 0x4;
2412                 } else {
2413                         reg |= 0x4;
2414                 }
2415
2416                 E1000_WRITE_REG(hw, E1000_DMCTLX, reg);
2417
2418                 /* free space in tx packet buffer to wake from DMA coal */
2419                 E1000_WRITE_REG(hw, E1000_DMCTXTH, (IGB_TXPBSIZE -
2420                     (2 * max_frame_size)) >> 6);
2421
2422                 /* make low power state decision controlled by DMA coal */
2423                 reg = E1000_READ_REG(hw, E1000_PCIEMISC);
2424                 reg &= ~E1000_PCIEMISC_LX_DECISION;
2425                 E1000_WRITE_REG(hw, E1000_PCIEMISC, reg);
2426
2427         } else if (hw->mac.type == e1000_82580) {
2428                 u32 reg = E1000_READ_REG(hw, E1000_PCIEMISC);
2429                 E1000_WRITE_REG(hw, E1000_PCIEMISC,
2430                     reg & ~E1000_PCIEMISC_LX_DECISION);
2431                 E1000_WRITE_REG(hw, E1000_DMACR, 0);
2432         }
2433 }
2434
2435 /*********************************************************************
2436  *
2437  *  Initialize the hardware to a configuration as specified by the
2438  *  adapter structure.
2439  *
2440  **********************************************************************/
2441 static void
2442 em_reset(if_ctx_t ctx)
2443 {
2444         device_t dev = iflib_get_dev(ctx);
2445         struct adapter *adapter = iflib_get_softc(ctx);
2446         struct ifnet *ifp = iflib_get_ifp(ctx);
2447         struct e1000_hw *hw = &adapter->hw;
2448         u16 rx_buffer_size;
2449         u32 pba;
2450
2451         INIT_DEBUGOUT("em_reset: begin");
2452         /* Let the firmware know the OS is in control */
2453         em_get_hw_control(adapter);
2454
2455         /* Set up smart power down as default off on newer adapters. */
2456         if (!em_smart_pwr_down && (hw->mac.type == e1000_82571 ||
2457             hw->mac.type == e1000_82572)) {
2458                 u16 phy_tmp = 0;
2459
2460                 /* Speed up time to link by disabling smart power down. */
2461                 e1000_read_phy_reg(hw, IGP02E1000_PHY_POWER_MGMT, &phy_tmp);
2462                 phy_tmp &= ~IGP02E1000_PM_SPD;
2463                 e1000_write_phy_reg(hw, IGP02E1000_PHY_POWER_MGMT, phy_tmp);
2464         }
2465
2466         /*
2467          * Packet Buffer Allocation (PBA)
2468          * Writing PBA sets the receive portion of the buffer
2469          * the remainder is used for the transmit buffer.
2470          */
2471         switch (hw->mac.type) {
2472         /* Total Packet Buffer on these is 48K */
2473         case e1000_82571:
2474         case e1000_82572:
2475         case e1000_80003es2lan:
2476                         pba = E1000_PBA_32K; /* 32K for Rx, 16K for Tx */
2477                 break;
2478         case e1000_82573: /* 82573: Total Packet Buffer is 32K */
2479                         pba = E1000_PBA_12K; /* 12K for Rx, 20K for Tx */
2480                 break;
2481         case e1000_82574:
2482         case e1000_82583:
2483                         pba = E1000_PBA_20K; /* 20K for Rx, 20K for Tx */
2484                 break;
2485         case e1000_ich8lan:
2486                 pba = E1000_PBA_8K;
2487                 break;
2488         case e1000_ich9lan:
2489         case e1000_ich10lan:
2490                 /* Boost Receive side for jumbo frames */
2491                 if (hw->mac.max_frame_size > 4096)
2492                         pba = E1000_PBA_14K;
2493                 else
2494                         pba = E1000_PBA_10K;
2495                 break;
2496         case e1000_pchlan:
2497         case e1000_pch2lan:
2498         case e1000_pch_lpt:
2499         case e1000_pch_spt:
2500         case e1000_pch_cnp:
2501                 pba = E1000_PBA_26K;
2502                 break;
2503         case e1000_82575:
2504                 pba = E1000_PBA_32K;
2505                 break;
2506         case e1000_82576:
2507         case e1000_vfadapt:
2508                 pba = E1000_READ_REG(hw, E1000_RXPBS);
2509                 pba &= E1000_RXPBS_SIZE_MASK_82576;
2510                 break;
2511         case e1000_82580:
2512         case e1000_i350:
2513         case e1000_i354:
2514         case e1000_vfadapt_i350:
2515                 pba = E1000_READ_REG(hw, E1000_RXPBS);
2516                 pba = e1000_rxpbs_adjust_82580(pba);
2517                 break;
2518         case e1000_i210:
2519         case e1000_i211:
2520                 pba = E1000_PBA_34K;
2521                 break;
2522         default:
2523                 if (hw->mac.max_frame_size > 8192)
2524                         pba = E1000_PBA_40K; /* 40K for Rx, 24K for Tx */
2525                 else
2526                         pba = E1000_PBA_48K; /* 48K for Rx, 16K for Tx */
2527         }
2528
2529         /* Special needs in case of Jumbo frames */
2530         if ((hw->mac.type == e1000_82575) && (ifp->if_mtu > ETHERMTU)) {
2531                 u32 tx_space, min_tx, min_rx;
2532                 pba = E1000_READ_REG(hw, E1000_PBA);
2533                 tx_space = pba >> 16;
2534                 pba &= 0xffff;
2535                 min_tx = (hw->mac.max_frame_size +
2536                     sizeof(struct e1000_tx_desc) - ETHERNET_FCS_SIZE) * 2;
2537                 min_tx = roundup2(min_tx, 1024);
2538                 min_tx >>= 10;
2539                 min_rx = hw->mac.max_frame_size;
2540                 min_rx = roundup2(min_rx, 1024);
2541                 min_rx >>= 10;
2542                 if (tx_space < min_tx &&
2543                     ((min_tx - tx_space) < pba)) {
2544                         pba = pba - (min_tx - tx_space);
2545                         /*
2546                          * if short on rx space, rx wins
2547                          * and must trump tx adjustment
2548                          */
2549                         if (pba < min_rx)
2550                                 pba = min_rx;
2551                 }
2552                 E1000_WRITE_REG(hw, E1000_PBA, pba);
2553         }
2554
2555         if (hw->mac.type < igb_mac_min)
2556                 E1000_WRITE_REG(hw, E1000_PBA, pba);
2557
2558         INIT_DEBUGOUT1("em_reset: pba=%dK",pba);
2559
2560         /*
2561          * These parameters control the automatic generation (Tx) and
2562          * response (Rx) to Ethernet PAUSE frames.
2563          * - High water mark should allow for at least two frames to be
2564          *   received after sending an XOFF.
2565          * - Low water mark works best when it is very near the high water mark.
2566          *   This allows the receiver to restart by sending XON when it has
2567          *   drained a bit. Here we use an arbitrary value of 1500 which will
2568          *   restart after one full frame is pulled from the buffer. There
2569          *   could be several smaller frames in the buffer and if so they will
2570          *   not trigger the XON until their total number reduces the buffer
2571          *   by 1500.
2572          * - The pause time is fairly large at 1000 x 512ns = 512 usec.
2573          */
2574         rx_buffer_size = (pba & 0xffff) << 10;
2575         hw->fc.high_water = rx_buffer_size -
2576             roundup2(hw->mac.max_frame_size, 1024);
2577         hw->fc.low_water = hw->fc.high_water - 1500;
2578
2579         if (adapter->fc) /* locally set flow control value? */
2580                 hw->fc.requested_mode = adapter->fc;
2581         else
2582                 hw->fc.requested_mode = e1000_fc_full;
2583
2584         if (hw->mac.type == e1000_80003es2lan)
2585                 hw->fc.pause_time = 0xFFFF;
2586         else
2587                 hw->fc.pause_time = EM_FC_PAUSE_TIME;
2588
2589         hw->fc.send_xon = TRUE;
2590
2591         /* Device specific overrides/settings */
2592         switch (hw->mac.type) {
2593         case e1000_pchlan:
2594                 /* Workaround: no TX flow ctrl for PCH */
2595                 hw->fc.requested_mode = e1000_fc_rx_pause;
2596                 hw->fc.pause_time = 0xFFFF; /* override */
2597                 if (if_getmtu(ifp) > ETHERMTU) {
2598                         hw->fc.high_water = 0x3500;
2599                         hw->fc.low_water = 0x1500;
2600                 } else {
2601                         hw->fc.high_water = 0x5000;
2602                         hw->fc.low_water = 0x3000;
2603                 }
2604                 hw->fc.refresh_time = 0x1000;
2605                 break;
2606         case e1000_pch2lan:
2607         case e1000_pch_lpt:
2608         case e1000_pch_spt:
2609         case e1000_pch_cnp:
2610                 hw->fc.high_water = 0x5C20;
2611                 hw->fc.low_water = 0x5048;
2612                 hw->fc.pause_time = 0x0650;
2613                 hw->fc.refresh_time = 0x0400;
2614                 /* Jumbos need adjusted PBA */
2615                 if (if_getmtu(ifp) > ETHERMTU)
2616                         E1000_WRITE_REG(hw, E1000_PBA, 12);
2617                 else
2618                         E1000_WRITE_REG(hw, E1000_PBA, 26);
2619                 break;
2620         case e1000_82575:
2621         case e1000_82576:
2622                 /* 8-byte granularity */
2623                 hw->fc.low_water = hw->fc.high_water - 8;
2624                 break;
2625         case e1000_82580:
2626         case e1000_i350:
2627         case e1000_i354:
2628         case e1000_i210:
2629         case e1000_i211:
2630         case e1000_vfadapt:
2631         case e1000_vfadapt_i350:
2632                 /* 16-byte granularity */
2633                 hw->fc.low_water = hw->fc.high_water - 16;
2634                 break;
2635         case e1000_ich9lan:
2636         case e1000_ich10lan:
2637                 if (if_getmtu(ifp) > ETHERMTU) {
2638                         hw->fc.high_water = 0x2800;
2639                         hw->fc.low_water = hw->fc.high_water - 8;
2640                         break;
2641                 }
2642                 /* FALLTHROUGH */
2643         default:
2644                 if (hw->mac.type == e1000_80003es2lan)
2645                         hw->fc.pause_time = 0xFFFF;
2646                 break;
2647         }
2648
2649         /* Issue a global reset */
2650         e1000_reset_hw(hw);
2651         if (hw->mac.type >= igb_mac_min) {
2652                 E1000_WRITE_REG(hw, E1000_WUC, 0);
2653         } else {
2654                 E1000_WRITE_REG(hw, E1000_WUFC, 0);
2655                 em_disable_aspm(adapter);
2656         }
2657         if (adapter->flags & IGB_MEDIA_RESET) {
2658                 e1000_setup_init_funcs(hw, TRUE);
2659                 e1000_get_bus_info(hw);
2660                 adapter->flags &= ~IGB_MEDIA_RESET;
2661         }
2662         /* and a re-init */
2663         if (e1000_init_hw(hw) < 0) {
2664                 device_printf(dev, "Hardware Initialization Failed\n");
2665                 return;
2666         }
2667         if (hw->mac.type >= igb_mac_min)
2668                 igb_init_dmac(adapter, pba);
2669
2670         E1000_WRITE_REG(hw, E1000_VET, ETHERTYPE_VLAN);
2671         e1000_get_phy_info(hw);
2672         e1000_check_for_link(hw);
2673 }
2674
2675 /*
2676  * Initialise the RSS mapping for NICs that support multiple transmit/
2677  * receive rings.
2678  */
2679
2680 #define RSSKEYLEN 10
2681 static void
2682 em_initialize_rss_mapping(struct adapter *adapter)
2683 {
2684         uint8_t  rss_key[4 * RSSKEYLEN];
2685         uint32_t reta = 0;
2686         struct e1000_hw *hw = &adapter->hw;
2687         int i;
2688
2689         /*
2690          * Configure RSS key
2691          */
2692         arc4rand(rss_key, sizeof(rss_key), 0);
2693         for (i = 0; i < RSSKEYLEN; ++i) {
2694                 uint32_t rssrk = 0;
2695
2696                 rssrk = EM_RSSRK_VAL(rss_key, i);
2697                 E1000_WRITE_REG(hw,E1000_RSSRK(i), rssrk);
2698         }
2699
2700         /*
2701          * Configure RSS redirect table in following fashion:
2702          * (hash & ring_cnt_mask) == rdr_table[(hash & rdr_table_mask)]
2703          */
2704         for (i = 0; i < sizeof(reta); ++i) {
2705                 uint32_t q;
2706
2707                 q = (i % adapter->rx_num_queues) << 7;
2708                 reta |= q << (8 * i);
2709         }
2710
2711         for (i = 0; i < 32; ++i)
2712                 E1000_WRITE_REG(hw, E1000_RETA(i), reta);
2713
2714         E1000_WRITE_REG(hw, E1000_MRQC, E1000_MRQC_RSS_ENABLE_2Q |
2715                         E1000_MRQC_RSS_FIELD_IPV4_TCP |
2716                         E1000_MRQC_RSS_FIELD_IPV4 |
2717                         E1000_MRQC_RSS_FIELD_IPV6_TCP_EX |
2718                         E1000_MRQC_RSS_FIELD_IPV6_EX |
2719                         E1000_MRQC_RSS_FIELD_IPV6);
2720 }
2721
2722 static void
2723 igb_initialize_rss_mapping(struct adapter *adapter)
2724 {
2725         struct e1000_hw *hw = &adapter->hw;
2726         int i;
2727         int queue_id;
2728         u32 reta;
2729         u32 rss_key[10], mrqc, shift = 0;
2730
2731         /* XXX? */
2732         if (hw->mac.type == e1000_82575)
2733                 shift = 6;
2734
2735         /*
2736          * The redirection table controls which destination
2737          * queue each bucket redirects traffic to.
2738          * Each DWORD represents four queues, with the LSB
2739          * being the first queue in the DWORD.
2740          *
2741          * This just allocates buckets to queues using round-robin
2742          * allocation.
2743          *
2744          * NOTE: It Just Happens to line up with the default
2745          * RSS allocation method.
2746          */
2747
2748         /* Warning FM follows */
2749         reta = 0;
2750         for (i = 0; i < 128; i++) {
2751 #ifdef RSS
2752                 queue_id = rss_get_indirection_to_bucket(i);
2753                 /*
2754                  * If we have more queues than buckets, we'll
2755                  * end up mapping buckets to a subset of the
2756                  * queues.
2757                  *
2758                  * If we have more buckets than queues, we'll
2759                  * end up instead assigning multiple buckets
2760                  * to queues.
2761                  *
2762                  * Both are suboptimal, but we need to handle
2763                  * the case so we don't go out of bounds
2764                  * indexing arrays and such.
2765                  */
2766                 queue_id = queue_id % adapter->rx_num_queues;
2767 #else
2768                 queue_id = (i % adapter->rx_num_queues);
2769 #endif
2770                 /* Adjust if required */
2771                 queue_id = queue_id << shift;
2772
2773                 /*
2774                  * The low 8 bits are for hash value (n+0);
2775                  * The next 8 bits are for hash value (n+1), etc.
2776                  */
2777                 reta = reta >> 8;
2778                 reta = reta | ( ((uint32_t) queue_id) << 24);
2779                 if ((i & 3) == 3) {
2780                         E1000_WRITE_REG(hw, E1000_RETA(i >> 2), reta);
2781                         reta = 0;
2782                 }
2783         }
2784
2785         /* Now fill in hash table */
2786
2787         /*
2788          * MRQC: Multiple Receive Queues Command
2789          * Set queuing to RSS control, number depends on the device.
2790          */
2791         mrqc = E1000_MRQC_ENABLE_RSS_8Q;
2792
2793 #ifdef RSS
2794         /* XXX ew typecasting */
2795         rss_getkey((uint8_t *) &rss_key);
2796 #else
2797         arc4rand(&rss_key, sizeof(rss_key), 0);
2798 #endif
2799         for (i = 0; i < 10; i++)
2800                 E1000_WRITE_REG_ARRAY(hw, E1000_RSSRK(0), i, rss_key[i]);
2801
2802         /*
2803          * Configure the RSS fields to hash upon.
2804          */
2805         mrqc |= (E1000_MRQC_RSS_FIELD_IPV4 |
2806             E1000_MRQC_RSS_FIELD_IPV4_TCP);
2807         mrqc |= (E1000_MRQC_RSS_FIELD_IPV6 |
2808             E1000_MRQC_RSS_FIELD_IPV6_TCP);
2809         mrqc |=( E1000_MRQC_RSS_FIELD_IPV4_UDP |
2810             E1000_MRQC_RSS_FIELD_IPV6_UDP);
2811         mrqc |=( E1000_MRQC_RSS_FIELD_IPV6_UDP_EX |
2812             E1000_MRQC_RSS_FIELD_IPV6_TCP_EX);
2813
2814         E1000_WRITE_REG(hw, E1000_MRQC, mrqc);
2815 }
2816
2817 /*********************************************************************
2818  *
2819  *  Setup networking device structure and register interface media.
2820  *
2821  **********************************************************************/
2822 static int
2823 em_setup_interface(if_ctx_t ctx)
2824 {
2825         struct ifnet *ifp = iflib_get_ifp(ctx);
2826         struct adapter *adapter = iflib_get_softc(ctx);
2827         if_softc_ctx_t scctx = adapter->shared;
2828
2829         INIT_DEBUGOUT("em_setup_interface: begin");
2830
2831         /* Single Queue */
2832         if (adapter->tx_num_queues == 1) {
2833                 if_setsendqlen(ifp, scctx->isc_ntxd[0] - 1);
2834                 if_setsendqready(ifp);
2835         }
2836
2837         /*
2838          * Specify the media types supported by this adapter and register
2839          * callbacks to update media and link information
2840          */
2841         if (adapter->hw.phy.media_type == e1000_media_type_fiber ||
2842             adapter->hw.phy.media_type == e1000_media_type_internal_serdes) {
2843                 u_char fiber_type = IFM_1000_SX;        /* default type */
2844
2845                 if (adapter->hw.mac.type == e1000_82545)
2846                         fiber_type = IFM_1000_LX;
2847                 ifmedia_add(adapter->media, IFM_ETHER | fiber_type | IFM_FDX, 0, NULL);
2848                 ifmedia_add(adapter->media, IFM_ETHER | fiber_type, 0, NULL);
2849         } else {
2850                 ifmedia_add(adapter->media, IFM_ETHER | IFM_10_T, 0, NULL);
2851                 ifmedia_add(adapter->media, IFM_ETHER | IFM_10_T | IFM_FDX, 0, NULL);
2852                 ifmedia_add(adapter->media, IFM_ETHER | IFM_100_TX, 0, NULL);
2853                 ifmedia_add(adapter->media, IFM_ETHER | IFM_100_TX | IFM_FDX, 0, NULL);
2854                 if (adapter->hw.phy.type != e1000_phy_ife) {
2855                         ifmedia_add(adapter->media, IFM_ETHER | IFM_1000_T | IFM_FDX, 0, NULL);
2856                         ifmedia_add(adapter->media, IFM_ETHER | IFM_1000_T, 0, NULL);
2857                 }
2858         }
2859         ifmedia_add(adapter->media, IFM_ETHER | IFM_AUTO, 0, NULL);
2860         ifmedia_set(adapter->media, IFM_ETHER | IFM_AUTO);
2861         return (0);
2862 }
2863
2864 static int
2865 em_if_tx_queues_alloc(if_ctx_t ctx, caddr_t *vaddrs, uint64_t *paddrs, int ntxqs, int ntxqsets)
2866 {
2867         struct adapter *adapter = iflib_get_softc(ctx);
2868         if_softc_ctx_t scctx = adapter->shared;
2869         int error = E1000_SUCCESS;
2870         struct em_tx_queue *que;
2871         int i, j;
2872
2873         MPASS(adapter->tx_num_queues > 0);
2874         MPASS(adapter->tx_num_queues == ntxqsets);
2875
2876         /* First allocate the top level queue structs */
2877         if (!(adapter->tx_queues =
2878             (struct em_tx_queue *) malloc(sizeof(struct em_tx_queue) *
2879             adapter->tx_num_queues, M_DEVBUF, M_NOWAIT | M_ZERO))) {
2880                 device_printf(iflib_get_dev(ctx), "Unable to allocate queue memory\n");
2881                 return(ENOMEM);
2882         }
2883
2884         for (i = 0, que = adapter->tx_queues; i < adapter->tx_num_queues; i++, que++) {
2885                 /* Set up some basics */
2886
2887                 struct tx_ring *txr = &que->txr;
2888                 txr->adapter = que->adapter = adapter;
2889                 que->me = txr->me =  i;
2890
2891                 /* Allocate report status array */
2892                 if (!(txr->tx_rsq = (qidx_t *) malloc(sizeof(qidx_t) * scctx->isc_ntxd[0], M_DEVBUF, M_NOWAIT | M_ZERO))) {
2893                         device_printf(iflib_get_dev(ctx), "failed to allocate rs_idxs memory\n");
2894                         error = ENOMEM;
2895                         goto fail;
2896                 }
2897                 for (j = 0; j < scctx->isc_ntxd[0]; j++)
2898                         txr->tx_rsq[j] = QIDX_INVALID;
2899                 /* get the virtual and physical address of the hardware queues */
2900                 txr->tx_base = (struct e1000_tx_desc *)vaddrs[i*ntxqs];
2901                 txr->tx_paddr = paddrs[i*ntxqs];
2902         }
2903
2904         if (bootverbose)
2905                 device_printf(iflib_get_dev(ctx),
2906                     "allocated for %d tx_queues\n", adapter->tx_num_queues);
2907         return (0);
2908 fail:
2909         em_if_queues_free(ctx);
2910         return (error);
2911 }
2912
2913 static int
2914 em_if_rx_queues_alloc(if_ctx_t ctx, caddr_t *vaddrs, uint64_t *paddrs, int nrxqs, int nrxqsets)
2915 {
2916         struct adapter *adapter = iflib_get_softc(ctx);
2917         int error = E1000_SUCCESS;
2918         struct em_rx_queue *que;
2919         int i;
2920
2921         MPASS(adapter->rx_num_queues > 0);
2922         MPASS(adapter->rx_num_queues == nrxqsets);
2923
2924         /* First allocate the top level queue structs */
2925         if (!(adapter->rx_queues =
2926             (struct em_rx_queue *) malloc(sizeof(struct em_rx_queue) *
2927             adapter->rx_num_queues, M_DEVBUF, M_NOWAIT | M_ZERO))) {
2928                 device_printf(iflib_get_dev(ctx), "Unable to allocate queue memory\n");
2929                 error = ENOMEM;
2930                 goto fail;
2931         }
2932
2933         for (i = 0, que = adapter->rx_queues; i < nrxqsets; i++, que++) {
2934                 /* Set up some basics */
2935                 struct rx_ring *rxr = &que->rxr;
2936                 rxr->adapter = que->adapter = adapter;
2937                 rxr->que = que;
2938                 que->me = rxr->me =  i;
2939
2940                 /* get the virtual and physical address of the hardware queues */
2941                 rxr->rx_base = (union e1000_rx_desc_extended *)vaddrs[i*nrxqs];
2942                 rxr->rx_paddr = paddrs[i*nrxqs];
2943         }
2944
2945         if (bootverbose)
2946                 device_printf(iflib_get_dev(ctx),
2947                     "allocated for %d rx_queues\n", adapter->rx_num_queues);
2948
2949         return (0);
2950 fail:
2951         em_if_queues_free(ctx);
2952         return (error);
2953 }
2954
2955 static void
2956 em_if_queues_free(if_ctx_t ctx)
2957 {
2958         struct adapter *adapter = iflib_get_softc(ctx);
2959         struct em_tx_queue *tx_que = adapter->tx_queues;
2960         struct em_rx_queue *rx_que = adapter->rx_queues;
2961
2962         if (tx_que != NULL) {
2963                 for (int i = 0; i < adapter->tx_num_queues; i++, tx_que++) {
2964                         struct tx_ring *txr = &tx_que->txr;
2965                         if (txr->tx_rsq == NULL)
2966                                 break;
2967
2968                         free(txr->tx_rsq, M_DEVBUF);
2969                         txr->tx_rsq = NULL;
2970                 }
2971                 free(adapter->tx_queues, M_DEVBUF);
2972                 adapter->tx_queues = NULL;
2973         }
2974
2975         if (rx_que != NULL) {
2976                 free(adapter->rx_queues, M_DEVBUF);
2977                 adapter->rx_queues = NULL;
2978         }
2979 }
2980
2981 /*********************************************************************
2982  *
2983  *  Enable transmit unit.
2984  *
2985  **********************************************************************/
2986 static void
2987 em_initialize_transmit_unit(if_ctx_t ctx)
2988 {
2989         struct adapter *adapter = iflib_get_softc(ctx);
2990         if_softc_ctx_t scctx = adapter->shared;
2991         struct em_tx_queue *que;
2992         struct tx_ring  *txr;
2993         struct e1000_hw *hw = &adapter->hw;
2994         u32 tctl, txdctl = 0, tarc, tipg = 0;
2995
2996         INIT_DEBUGOUT("em_initialize_transmit_unit: begin");
2997
2998         for (int i = 0; i < adapter->tx_num_queues; i++, txr++) {
2999                 u64 bus_addr;
3000                 caddr_t offp, endp;
3001
3002                 que = &adapter->tx_queues[i];
3003                 txr = &que->txr;
3004                 bus_addr = txr->tx_paddr;
3005
3006                 /* Clear checksum offload context. */
3007                 offp = (caddr_t)&txr->csum_flags;
3008                 endp = (caddr_t)(txr + 1);
3009                 bzero(offp, endp - offp);
3010
3011                 /* Base and Len of TX Ring */
3012                 E1000_WRITE_REG(hw, E1000_TDLEN(i),
3013                     scctx->isc_ntxd[0] * sizeof(struct e1000_tx_desc));
3014                 E1000_WRITE_REG(hw, E1000_TDBAH(i),
3015                     (u32)(bus_addr >> 32));
3016                 E1000_WRITE_REG(hw, E1000_TDBAL(i),
3017                     (u32)bus_addr);
3018                 /* Init the HEAD/TAIL indices */
3019                 E1000_WRITE_REG(hw, E1000_TDT(i), 0);
3020                 E1000_WRITE_REG(hw, E1000_TDH(i), 0);
3021
3022                 HW_DEBUGOUT2("Base = %x, Length = %x\n",
3023                     E1000_READ_REG(hw, E1000_TDBAL(i)),
3024                     E1000_READ_REG(hw, E1000_TDLEN(i)));
3025
3026                 txdctl = 0; /* clear txdctl */
3027                 txdctl |= 0x1f; /* PTHRESH */
3028                 txdctl |= 1 << 8; /* HTHRESH */
3029                 txdctl |= 1 << 16;/* WTHRESH */
3030                 txdctl |= 1 << 22; /* Reserved bit 22 must always be 1 */
3031                 txdctl |= E1000_TXDCTL_GRAN;
3032                 txdctl |= 1 << 25; /* LWTHRESH */
3033
3034                 E1000_WRITE_REG(hw, E1000_TXDCTL(i), txdctl);
3035         }
3036
3037         /* Set the default values for the Tx Inter Packet Gap timer */
3038         switch (hw->mac.type) {
3039         case e1000_80003es2lan:
3040                 tipg = DEFAULT_82543_TIPG_IPGR1;
3041                 tipg |= DEFAULT_80003ES2LAN_TIPG_IPGR2 <<
3042                     E1000_TIPG_IPGR2_SHIFT;
3043                 break;
3044         case e1000_82542:
3045                 tipg = DEFAULT_82542_TIPG_IPGT;
3046                 tipg |= DEFAULT_82542_TIPG_IPGR1 << E1000_TIPG_IPGR1_SHIFT;
3047                 tipg |= DEFAULT_82542_TIPG_IPGR2 << E1000_TIPG_IPGR2_SHIFT;
3048                 break;
3049         default:
3050                 if (hw->phy.media_type == e1000_media_type_fiber ||
3051                     hw->phy.media_type == e1000_media_type_internal_serdes)
3052                         tipg = DEFAULT_82543_TIPG_IPGT_FIBER;
3053                 else
3054                         tipg = DEFAULT_82543_TIPG_IPGT_COPPER;
3055                 tipg |= DEFAULT_82543_TIPG_IPGR1 << E1000_TIPG_IPGR1_SHIFT;
3056                 tipg |= DEFAULT_82543_TIPG_IPGR2 << E1000_TIPG_IPGR2_SHIFT;
3057         }
3058
3059         E1000_WRITE_REG(hw, E1000_TIPG, tipg);
3060         E1000_WRITE_REG(hw, E1000_TIDV, adapter->tx_int_delay.value);
3061
3062         if(hw->mac.type >= e1000_82540)
3063                 E1000_WRITE_REG(hw, E1000_TADV,
3064                     adapter->tx_abs_int_delay.value);
3065
3066         if (hw->mac.type == e1000_82571 || hw->mac.type == e1000_82572) {
3067                 tarc = E1000_READ_REG(hw, E1000_TARC(0));
3068                 tarc |= TARC_SPEED_MODE_BIT;
3069                 E1000_WRITE_REG(hw, E1000_TARC(0), tarc);
3070         } else if (hw->mac.type == e1000_80003es2lan) {
3071                 /* errata: program both queues to unweighted RR */
3072                 tarc = E1000_READ_REG(hw, E1000_TARC(0));
3073                 tarc |= 1;
3074                 E1000_WRITE_REG(hw, E1000_TARC(0), tarc);
3075                 tarc = E1000_READ_REG(hw, E1000_TARC(1));
3076                 tarc |= 1;
3077                 E1000_WRITE_REG(hw, E1000_TARC(1), tarc);
3078         } else if (hw->mac.type == e1000_82574) {
3079                 tarc = E1000_READ_REG(hw, E1000_TARC(0));
3080                 tarc |= TARC_ERRATA_BIT;
3081                 if ( adapter->tx_num_queues > 1) {
3082                         tarc |= (TARC_COMPENSATION_MODE | TARC_MQ_FIX);
3083                         E1000_WRITE_REG(hw, E1000_TARC(0), tarc);
3084                         E1000_WRITE_REG(hw, E1000_TARC(1), tarc);
3085                 } else
3086                         E1000_WRITE_REG(hw, E1000_TARC(0), tarc);
3087         }
3088
3089         if (adapter->tx_int_delay.value > 0)
3090                 adapter->txd_cmd |= E1000_TXD_CMD_IDE;
3091
3092         /* Program the Transmit Control Register */
3093         tctl = E1000_READ_REG(hw, E1000_TCTL);
3094         tctl &= ~E1000_TCTL_CT;
3095         tctl |= (E1000_TCTL_PSP | E1000_TCTL_RTLC | E1000_TCTL_EN |
3096                    (E1000_COLLISION_THRESHOLD << E1000_CT_SHIFT));
3097
3098         if (hw->mac.type >= e1000_82571)
3099                 tctl |= E1000_TCTL_MULR;
3100
3101         /* This write will effectively turn on the transmit unit. */
3102         E1000_WRITE_REG(hw, E1000_TCTL, tctl);
3103
3104         /* SPT and KBL errata workarounds */
3105         if (hw->mac.type == e1000_pch_spt) {
3106                 u32 reg;
3107                 reg = E1000_READ_REG(hw, E1000_IOSFPC);
3108                 reg |= E1000_RCTL_RDMTS_HEX;
3109                 E1000_WRITE_REG(hw, E1000_IOSFPC, reg);
3110                 /* i218-i219 Specification Update 1.5.4.5 */
3111                 reg = E1000_READ_REG(hw, E1000_TARC(0));
3112                 reg &= ~E1000_TARC0_CB_MULTIQ_3_REQ;
3113                 reg |= E1000_TARC0_CB_MULTIQ_2_REQ;
3114                 E1000_WRITE_REG(hw, E1000_TARC(0), reg);
3115         }
3116 }
3117
3118 /*********************************************************************
3119  *
3120  *  Enable receive unit.
3121  *
3122  **********************************************************************/
3123
3124 static void
3125 em_initialize_receive_unit(if_ctx_t ctx)
3126 {
3127         struct adapter *adapter = iflib_get_softc(ctx);
3128         if_softc_ctx_t scctx = adapter->shared;
3129         struct ifnet *ifp = iflib_get_ifp(ctx);
3130         struct e1000_hw *hw = &adapter->hw;
3131         struct em_rx_queue *que;
3132         int i;
3133         u32 rctl, rxcsum, rfctl;
3134
3135         INIT_DEBUGOUT("em_initialize_receive_units: begin");
3136
3137         /*
3138          * Make sure receives are disabled while setting
3139          * up the descriptor ring
3140          */
3141         rctl = E1000_READ_REG(hw, E1000_RCTL);
3142         /* Do not disable if ever enabled on this hardware */
3143         if ((hw->mac.type != e1000_82574) && (hw->mac.type != e1000_82583))
3144                 E1000_WRITE_REG(hw, E1000_RCTL, rctl & ~E1000_RCTL_EN);
3145
3146         /* Setup the Receive Control Register */
3147         rctl &= ~(3 << E1000_RCTL_MO_SHIFT);
3148         rctl |= E1000_RCTL_EN | E1000_RCTL_BAM |
3149             E1000_RCTL_LBM_NO | E1000_RCTL_RDMTS_HALF |
3150             (hw->mac.mc_filter_type << E1000_RCTL_MO_SHIFT);
3151
3152         /* Do not store bad packets */
3153         rctl &= ~E1000_RCTL_SBP;
3154
3155         /* Enable Long Packet receive */
3156         if (if_getmtu(ifp) > ETHERMTU)
3157                 rctl |= E1000_RCTL_LPE;
3158         else
3159                 rctl &= ~E1000_RCTL_LPE;
3160
3161         /* Strip the CRC */
3162         if (!em_disable_crc_stripping)
3163                 rctl |= E1000_RCTL_SECRC;
3164
3165         if (hw->mac.type >= e1000_82540) {
3166                 E1000_WRITE_REG(hw, E1000_RADV,
3167                     adapter->rx_abs_int_delay.value);
3168
3169                 /*
3170                  * Set the interrupt throttling rate. Value is calculated
3171                  * as DEFAULT_ITR = 1/(MAX_INTS_PER_SEC * 256ns)
3172                  */
3173                 E1000_WRITE_REG(hw, E1000_ITR, DEFAULT_ITR);
3174         }
3175         E1000_WRITE_REG(hw, E1000_RDTR, adapter->rx_int_delay.value);
3176
3177         /* Use extended rx descriptor formats */
3178         rfctl = E1000_READ_REG(hw, E1000_RFCTL);
3179         rfctl |= E1000_RFCTL_EXTEN;
3180         /*
3181          * When using MSI-X interrupts we need to throttle
3182          * using the EITR register (82574 only)
3183          */
3184         if (hw->mac.type == e1000_82574) {
3185                 for (int i = 0; i < 4; i++)
3186                         E1000_WRITE_REG(hw, E1000_EITR_82574(i),
3187                             DEFAULT_ITR);
3188                 /* Disable accelerated acknowledge */
3189                 rfctl |= E1000_RFCTL_ACK_DIS;
3190         }
3191         E1000_WRITE_REG(hw, E1000_RFCTL, rfctl);
3192
3193         rxcsum = E1000_READ_REG(hw, E1000_RXCSUM);
3194         if (if_getcapenable(ifp) & IFCAP_RXCSUM &&
3195             hw->mac.type >= e1000_82543) {
3196                 if (adapter->tx_num_queues > 1) {
3197                         if (hw->mac.type >= igb_mac_min) {
3198                                 rxcsum |= E1000_RXCSUM_PCSD;
3199                                 if (hw->mac.type != e1000_82575)
3200                                         rxcsum |= E1000_RXCSUM_CRCOFL;
3201                         } else
3202                                 rxcsum |= E1000_RXCSUM_TUOFL |
3203                                         E1000_RXCSUM_IPOFL |
3204                                         E1000_RXCSUM_PCSD;
3205                 } else {
3206                         if (hw->mac.type >= igb_mac_min)
3207                                 rxcsum |= E1000_RXCSUM_IPPCSE;
3208                         else
3209                                 rxcsum |= E1000_RXCSUM_TUOFL | E1000_RXCSUM_IPOFL;
3210                         if (hw->mac.type > e1000_82575)
3211                                 rxcsum |= E1000_RXCSUM_CRCOFL;
3212                 }
3213         } else
3214                 rxcsum &= ~E1000_RXCSUM_TUOFL;
3215
3216         E1000_WRITE_REG(hw, E1000_RXCSUM, rxcsum);
3217
3218         if (adapter->rx_num_queues > 1) {
3219                 if (hw->mac.type >= igb_mac_min)
3220                         igb_initialize_rss_mapping(adapter);
3221                 else
3222                         em_initialize_rss_mapping(adapter);
3223         }
3224
3225         /*
3226          * XXX TEMPORARY WORKAROUND: on some systems with 82573
3227          * long latencies are observed, like Lenovo X60. This
3228          * change eliminates the problem, but since having positive
3229          * values in RDTR is a known source of problems on other
3230          * platforms another solution is being sought.
3231          */
3232         if (hw->mac.type == e1000_82573)
3233                 E1000_WRITE_REG(hw, E1000_RDTR, 0x20);
3234
3235         for (i = 0, que = adapter->rx_queues; i < adapter->rx_num_queues; i++, que++) {
3236                 struct rx_ring *rxr = &que->rxr;
3237                 /* Setup the Base and Length of the Rx Descriptor Ring */
3238                 u64 bus_addr = rxr->rx_paddr;
3239 #if 0
3240                 u32 rdt = adapter->rx_num_queues -1;  /* default */
3241 #endif
3242
3243                 E1000_WRITE_REG(hw, E1000_RDLEN(i),
3244                     scctx->isc_nrxd[0] * sizeof(union e1000_rx_desc_extended));
3245                 E1000_WRITE_REG(hw, E1000_RDBAH(i), (u32)(bus_addr >> 32));
3246                 E1000_WRITE_REG(hw, E1000_RDBAL(i), (u32)bus_addr);
3247                 /* Setup the Head and Tail Descriptor Pointers */
3248                 E1000_WRITE_REG(hw, E1000_RDH(i), 0);
3249                 E1000_WRITE_REG(hw, E1000_RDT(i), 0);
3250         }
3251
3252         /*
3253          * Set PTHRESH for improved jumbo performance
3254          * According to 10.2.5.11 of Intel 82574 Datasheet,
3255          * RXDCTL(1) is written whenever RXDCTL(0) is written.
3256          * Only write to RXDCTL(1) if there is a need for different
3257          * settings.
3258          */
3259         if ((hw->mac.type == e1000_ich9lan || hw->mac.type == e1000_pch2lan ||
3260             hw->mac.type == e1000_ich10lan) && if_getmtu(ifp) > ETHERMTU) {
3261                 u32 rxdctl = E1000_READ_REG(hw, E1000_RXDCTL(0));
3262                 E1000_WRITE_REG(hw, E1000_RXDCTL(0), rxdctl | 3);
3263         } else if (hw->mac.type == e1000_82574) {
3264                 for (int i = 0; i < adapter->rx_num_queues; i++) {
3265                         u32 rxdctl = E1000_READ_REG(hw, E1000_RXDCTL(i));
3266                         rxdctl |= 0x20; /* PTHRESH */
3267                         rxdctl |= 4 << 8; /* HTHRESH */
3268                         rxdctl |= 4 << 16;/* WTHRESH */
3269                         rxdctl |= 1 << 24; /* Switch to granularity */
3270                         E1000_WRITE_REG(hw, E1000_RXDCTL(i), rxdctl);
3271                 }
3272         } else if (hw->mac.type >= igb_mac_min) {
3273                 u32 psize, srrctl = 0;
3274
3275                 if (if_getmtu(ifp) > ETHERMTU) {
3276                         /* Set maximum packet len */
3277                         if (adapter->rx_mbuf_sz <= 4096) {
3278                                 srrctl |= 4096 >> E1000_SRRCTL_BSIZEPKT_SHIFT;
3279                                 rctl |= E1000_RCTL_SZ_4096 | E1000_RCTL_BSEX;
3280                         } else if (adapter->rx_mbuf_sz > 4096) {
3281                                 srrctl |= 8192 >> E1000_SRRCTL_BSIZEPKT_SHIFT;
3282                                 rctl |= E1000_RCTL_SZ_8192 | E1000_RCTL_BSEX;
3283                         }
3284                         psize = scctx->isc_max_frame_size;
3285                         /* are we on a vlan? */
3286                         if (ifp->if_vlantrunk != NULL)
3287                                 psize += VLAN_TAG_SIZE;
3288                         E1000_WRITE_REG(hw, E1000_RLPML, psize);
3289                 } else {
3290                         srrctl |= 2048 >> E1000_SRRCTL_BSIZEPKT_SHIFT;
3291                         rctl |= E1000_RCTL_SZ_2048;
3292                 }
3293
3294                 /*
3295                  * If TX flow control is disabled and there's >1 queue defined,
3296                  * enable DROP.
3297                  *
3298                  * This drops frames rather than hanging the RX MAC for all queues.
3299                  */
3300                 if ((adapter->rx_num_queues > 1) &&
3301                     (adapter->fc == e1000_fc_none ||
3302                      adapter->fc == e1000_fc_rx_pause)) {
3303                         srrctl |= E1000_SRRCTL_DROP_EN;
3304                 }
3305                         /* Setup the Base and Length of the Rx Descriptor Rings */
3306                 for (i = 0, que = adapter->rx_queues; i < adapter->rx_num_queues; i++, que++) {
3307                         struct rx_ring *rxr = &que->rxr;
3308                         u64 bus_addr = rxr->rx_paddr;
3309                         u32 rxdctl;
3310
3311 #ifdef notyet
3312                         /* Configure for header split? -- ignore for now */
3313                         rxr->hdr_split = igb_header_split;
3314 #else
3315                         srrctl |= E1000_SRRCTL_DESCTYPE_ADV_ONEBUF;
3316 #endif
3317
3318                         E1000_WRITE_REG(hw, E1000_RDLEN(i),
3319                                         scctx->isc_nrxd[0] * sizeof(struct e1000_rx_desc));
3320                         E1000_WRITE_REG(hw, E1000_RDBAH(i),
3321                                         (uint32_t)(bus_addr >> 32));
3322                         E1000_WRITE_REG(hw, E1000_RDBAL(i),
3323                                         (uint32_t)bus_addr);
3324                         E1000_WRITE_REG(hw, E1000_SRRCTL(i), srrctl);
3325                         /* Enable this Queue */
3326                         rxdctl = E1000_READ_REG(hw, E1000_RXDCTL(i));
3327                         rxdctl |= E1000_RXDCTL_QUEUE_ENABLE;
3328                         rxdctl &= 0xFFF00000;
3329                         rxdctl |= IGB_RX_PTHRESH;
3330                         rxdctl |= IGB_RX_HTHRESH << 8;
3331                         rxdctl |= IGB_RX_WTHRESH << 16;
3332                         E1000_WRITE_REG(hw, E1000_RXDCTL(i), rxdctl);
3333                 }
3334         } else if (hw->mac.type >= e1000_pch2lan) {
3335                 if (if_getmtu(ifp) > ETHERMTU)
3336                         e1000_lv_jumbo_workaround_ich8lan(hw, TRUE);
3337                 else
3338                         e1000_lv_jumbo_workaround_ich8lan(hw, FALSE);
3339         }
3340
3341         /* Make sure VLAN Filters are off */
3342         rctl &= ~E1000_RCTL_VFE;
3343
3344         if (hw->mac.type < igb_mac_min) {
3345                 if (adapter->rx_mbuf_sz == MCLBYTES)
3346                         rctl |= E1000_RCTL_SZ_2048;
3347                 else if (adapter->rx_mbuf_sz == MJUMPAGESIZE)
3348                         rctl |= E1000_RCTL_SZ_4096 | E1000_RCTL_BSEX;
3349                 else if (adapter->rx_mbuf_sz > MJUMPAGESIZE)
3350                         rctl |= E1000_RCTL_SZ_8192 | E1000_RCTL_BSEX;
3351
3352                 /* ensure we clear use DTYPE of 00 here */
3353                 rctl &= ~0x00000C00;
3354         }
3355
3356         /* Write out the settings */
3357         E1000_WRITE_REG(hw, E1000_RCTL, rctl);
3358
3359         return;
3360 }
3361
3362 static void
3363 em_if_vlan_register(if_ctx_t ctx, u16 vtag)
3364 {
3365         struct adapter *adapter = iflib_get_softc(ctx);
3366         u32 index, bit;
3367
3368         index = (vtag >> 5) & 0x7F;
3369         bit = vtag & 0x1F;
3370         adapter->shadow_vfta[index] |= (1 << bit);
3371         ++adapter->num_vlans;
3372 }
3373
3374 static void
3375 em_if_vlan_unregister(if_ctx_t ctx, u16 vtag)
3376 {
3377         struct adapter *adapter = iflib_get_softc(ctx);
3378         u32 index, bit;
3379
3380         index = (vtag >> 5) & 0x7F;
3381         bit = vtag & 0x1F;
3382         adapter->shadow_vfta[index] &= ~(1 << bit);
3383         --adapter->num_vlans;
3384 }
3385
3386 static void
3387 em_setup_vlan_hw_support(struct adapter *adapter)
3388 {
3389         struct e1000_hw *hw = &adapter->hw;
3390         u32 reg;
3391
3392         /*
3393          * We get here thru init_locked, meaning
3394          * a soft reset, this has already cleared
3395          * the VFTA and other state, so if there
3396          * have been no vlan's registered do nothing.
3397          */
3398         if (adapter->num_vlans == 0)
3399                 return;
3400
3401         /*
3402          * A soft reset zero's out the VFTA, so
3403          * we need to repopulate it now.
3404          */
3405         for (int i = 0; i < EM_VFTA_SIZE; i++)
3406                 if (adapter->shadow_vfta[i] != 0)
3407                         E1000_WRITE_REG_ARRAY(hw, E1000_VFTA,
3408                             i, adapter->shadow_vfta[i]);
3409
3410         reg = E1000_READ_REG(hw, E1000_CTRL);
3411         reg |= E1000_CTRL_VME;
3412         E1000_WRITE_REG(hw, E1000_CTRL, reg);
3413
3414         /* Enable the Filter Table */
3415         reg = E1000_READ_REG(hw, E1000_RCTL);
3416         reg &= ~E1000_RCTL_CFIEN;
3417         reg |= E1000_RCTL_VFE;
3418         E1000_WRITE_REG(hw, E1000_RCTL, reg);
3419 }
3420
3421 static void
3422 em_if_intr_enable(if_ctx_t ctx)
3423 {
3424         struct adapter *adapter = iflib_get_softc(ctx);
3425         struct e1000_hw *hw = &adapter->hw;
3426         u32 ims_mask = IMS_ENABLE_MASK;
3427
3428         if (hw->mac.type == e1000_82574) {
3429                 E1000_WRITE_REG(hw, EM_EIAC, EM_MSIX_MASK);
3430                 ims_mask |= adapter->ims;
3431         }
3432         E1000_WRITE_REG(hw, E1000_IMS, ims_mask);
3433 }
3434
3435 static void
3436 em_if_intr_disable(if_ctx_t ctx)
3437 {
3438         struct adapter *adapter = iflib_get_softc(ctx);
3439         struct e1000_hw *hw = &adapter->hw;
3440
3441         if (hw->mac.type == e1000_82574)
3442                 E1000_WRITE_REG(hw, EM_EIAC, 0);
3443         E1000_WRITE_REG(hw, E1000_IMC, 0xffffffff);
3444 }
3445
3446 static void
3447 igb_if_intr_enable(if_ctx_t ctx)
3448 {
3449         struct adapter *adapter = iflib_get_softc(ctx);
3450         struct e1000_hw *hw = &adapter->hw;
3451         u32 mask;
3452
3453         if (__predict_true(adapter->intr_type == IFLIB_INTR_MSIX)) {
3454                 mask = (adapter->que_mask | adapter->link_mask);
3455                 E1000_WRITE_REG(hw, E1000_EIAC, mask);
3456                 E1000_WRITE_REG(hw, E1000_EIAM, mask);
3457                 E1000_WRITE_REG(hw, E1000_EIMS, mask);
3458                 E1000_WRITE_REG(hw, E1000_IMS, E1000_IMS_LSC);
3459         } else
3460                 E1000_WRITE_REG(hw, E1000_IMS, IMS_ENABLE_MASK);
3461         E1000_WRITE_FLUSH(hw);
3462 }
3463
3464 static void
3465 igb_if_intr_disable(if_ctx_t ctx)
3466 {
3467         struct adapter *adapter = iflib_get_softc(ctx);
3468         struct e1000_hw *hw = &adapter->hw;
3469
3470         if (__predict_true(adapter->intr_type == IFLIB_INTR_MSIX)) {
3471                 E1000_WRITE_REG(hw, E1000_EIMC, 0xffffffff);
3472                 E1000_WRITE_REG(hw, E1000_EIAC, 0);
3473         }
3474         E1000_WRITE_REG(hw, E1000_IMC, 0xffffffff);
3475         E1000_WRITE_FLUSH(hw);
3476 }
3477
3478 /*
3479  * Bit of a misnomer, what this really means is
3480  * to enable OS management of the system... aka
3481  * to disable special hardware management features
3482  */
3483 static void
3484 em_init_manageability(struct adapter *adapter)
3485 {
3486         /* A shared code workaround */
3487 #define E1000_82542_MANC2H E1000_MANC2H
3488         if (adapter->has_manage) {
3489                 int manc2h = E1000_READ_REG(&adapter->hw, E1000_MANC2H);
3490                 int manc = E1000_READ_REG(&adapter->hw, E1000_MANC);
3491
3492                 /* disable hardware interception of ARP */
3493                 manc &= ~(E1000_MANC_ARP_EN);
3494
3495                 /* enable receiving management packets to the host */
3496                 manc |= E1000_MANC_EN_MNG2HOST;
3497 #define E1000_MNG2HOST_PORT_623 (1 << 5)
3498 #define E1000_MNG2HOST_PORT_664 (1 << 6)
3499                 manc2h |= E1000_MNG2HOST_PORT_623;
3500                 manc2h |= E1000_MNG2HOST_PORT_664;
3501                 E1000_WRITE_REG(&adapter->hw, E1000_MANC2H, manc2h);
3502                 E1000_WRITE_REG(&adapter->hw, E1000_MANC, manc);
3503         }
3504 }
3505
3506 /*
3507  * Give control back to hardware management
3508  * controller if there is one.
3509  */
3510 static void
3511 em_release_manageability(struct adapter *adapter)
3512 {
3513         if (adapter->has_manage) {
3514                 int manc = E1000_READ_REG(&adapter->hw, E1000_MANC);
3515
3516                 /* re-enable hardware interception of ARP */
3517                 manc |= E1000_MANC_ARP_EN;
3518                 manc &= ~E1000_MANC_EN_MNG2HOST;
3519
3520                 E1000_WRITE_REG(&adapter->hw, E1000_MANC, manc);
3521         }
3522 }
3523
3524 /*
3525  * em_get_hw_control sets the {CTRL_EXT|FWSM}:DRV_LOAD bit.
3526  * For ASF and Pass Through versions of f/w this means
3527  * that the driver is loaded. For AMT version type f/w
3528  * this means that the network i/f is open.
3529  */
3530 static void
3531 em_get_hw_control(struct adapter *adapter)
3532 {
3533         u32 ctrl_ext, swsm;
3534
3535         if (adapter->vf_ifp)
3536                 return;
3537
3538         if (adapter->hw.mac.type == e1000_82573) {
3539                 swsm = E1000_READ_REG(&adapter->hw, E1000_SWSM);
3540                 E1000_WRITE_REG(&adapter->hw, E1000_SWSM,
3541                     swsm | E1000_SWSM_DRV_LOAD);
3542                 return;
3543         }
3544         /* else */
3545         ctrl_ext = E1000_READ_REG(&adapter->hw, E1000_CTRL_EXT);
3546         E1000_WRITE_REG(&adapter->hw, E1000_CTRL_EXT,
3547             ctrl_ext | E1000_CTRL_EXT_DRV_LOAD);
3548 }
3549
3550 /*
3551  * em_release_hw_control resets {CTRL_EXT|FWSM}:DRV_LOAD bit.
3552  * For ASF and Pass Through versions of f/w this means that
3553  * the driver is no longer loaded. For AMT versions of the
3554  * f/w this means that the network i/f is closed.
3555  */
3556 static void
3557 em_release_hw_control(struct adapter *adapter)
3558 {
3559         u32 ctrl_ext, swsm;
3560
3561         if (!adapter->has_manage)
3562                 return;
3563
3564         if (adapter->hw.mac.type == e1000_82573) {
3565                 swsm = E1000_READ_REG(&adapter->hw, E1000_SWSM);
3566                 E1000_WRITE_REG(&adapter->hw, E1000_SWSM,
3567                     swsm & ~E1000_SWSM_DRV_LOAD);
3568                 return;
3569         }
3570         /* else */
3571         ctrl_ext = E1000_READ_REG(&adapter->hw, E1000_CTRL_EXT);
3572         E1000_WRITE_REG(&adapter->hw, E1000_CTRL_EXT,
3573             ctrl_ext & ~E1000_CTRL_EXT_DRV_LOAD);
3574         return;
3575 }
3576
3577 static int
3578 em_is_valid_ether_addr(u8 *addr)
3579 {
3580         char zero_addr[6] = { 0, 0, 0, 0, 0, 0 };
3581
3582         if ((addr[0] & 1) || (!bcmp(addr, zero_addr, ETHER_ADDR_LEN))) {
3583                 return (FALSE);
3584         }
3585
3586         return (TRUE);
3587 }
3588
3589 /*
3590 ** Parse the interface capabilities with regard
3591 ** to both system management and wake-on-lan for
3592 ** later use.
3593 */
3594 static void
3595 em_get_wakeup(if_ctx_t ctx)
3596 {
3597         struct adapter *adapter = iflib_get_softc(ctx);
3598         device_t dev = iflib_get_dev(ctx);
3599         u16 eeprom_data = 0, device_id, apme_mask;
3600
3601         adapter->has_manage = e1000_enable_mng_pass_thru(&adapter->hw);
3602         apme_mask = EM_EEPROM_APME;
3603
3604         switch (adapter->hw.mac.type) {
3605         case e1000_82542:
3606         case e1000_82543:
3607                 break;
3608         case e1000_82544:
3609                 e1000_read_nvm(&adapter->hw,
3610                     NVM_INIT_CONTROL2_REG, 1, &eeprom_data);
3611                 apme_mask = EM_82544_APME;
3612                 break;
3613         case e1000_82546:
3614         case e1000_82546_rev_3:
3615                 if (adapter->hw.bus.func == 1) {
3616                         e1000_read_nvm(&adapter->hw,
3617                             NVM_INIT_CONTROL3_PORT_B, 1, &eeprom_data);
3618                         break;
3619                 } else
3620                         e1000_read_nvm(&adapter->hw,
3621                             NVM_INIT_CONTROL3_PORT_A, 1, &eeprom_data);
3622                 break;
3623         case e1000_82573:
3624         case e1000_82583:
3625                 adapter->has_amt = TRUE;
3626                 /* FALLTHROUGH */
3627         case e1000_82571:
3628         case e1000_82572:
3629         case e1000_80003es2lan:
3630                 if (adapter->hw.bus.func == 1) {
3631                         e1000_read_nvm(&adapter->hw,
3632                             NVM_INIT_CONTROL3_PORT_B, 1, &eeprom_data);
3633                         break;
3634                 } else
3635                         e1000_read_nvm(&adapter->hw,
3636                             NVM_INIT_CONTROL3_PORT_A, 1, &eeprom_data);
3637                 break;
3638         case e1000_ich8lan:
3639         case e1000_ich9lan:
3640         case e1000_ich10lan:
3641         case e1000_pchlan:
3642         case e1000_pch2lan:
3643         case e1000_pch_lpt:
3644         case e1000_pch_spt:
3645         case e1000_82575:       /* listing all igb devices */
3646         case e1000_82576:
3647         case e1000_82580:
3648         case e1000_i350:
3649         case e1000_i354:
3650         case e1000_i210:
3651         case e1000_i211:
3652         case e1000_vfadapt:
3653         case e1000_vfadapt_i350:
3654                 apme_mask = E1000_WUC_APME;
3655                 adapter->has_amt = TRUE;
3656                 eeprom_data = E1000_READ_REG(&adapter->hw, E1000_WUC);
3657                 break;
3658         default:
3659                 e1000_read_nvm(&adapter->hw,
3660                     NVM_INIT_CONTROL3_PORT_A, 1, &eeprom_data);
3661                 break;
3662         }
3663         if (eeprom_data & apme_mask)
3664                 adapter->wol = (E1000_WUFC_MAG | E1000_WUFC_MC);
3665         /*
3666          * We have the eeprom settings, now apply the special cases
3667          * where the eeprom may be wrong or the board won't support
3668          * wake on lan on a particular port
3669          */
3670         device_id = pci_get_device(dev);
3671         switch (device_id) {
3672         case E1000_DEV_ID_82546GB_PCIE:
3673                 adapter->wol = 0;
3674                 break;
3675         case E1000_DEV_ID_82546EB_FIBER:
3676         case E1000_DEV_ID_82546GB_FIBER:
3677                 /* Wake events only supported on port A for dual fiber
3678                  * regardless of eeprom setting */
3679                 if (E1000_READ_REG(&adapter->hw, E1000_STATUS) &
3680                     E1000_STATUS_FUNC_1)
3681                         adapter->wol = 0;
3682                 break;
3683         case E1000_DEV_ID_82546GB_QUAD_COPPER_KSP3:
3684                 /* if quad port adapter, disable WoL on all but port A */
3685                 if (global_quad_port_a != 0)
3686                         adapter->wol = 0;
3687                 /* Reset for multiple quad port adapters */
3688                 if (++global_quad_port_a == 4)
3689                         global_quad_port_a = 0;
3690                 break;
3691         case E1000_DEV_ID_82571EB_FIBER:
3692                 /* Wake events only supported on port A for dual fiber
3693                  * regardless of eeprom setting */
3694                 if (E1000_READ_REG(&adapter->hw, E1000_STATUS) &
3695                     E1000_STATUS_FUNC_1)
3696                         adapter->wol = 0;
3697                 break;
3698         case E1000_DEV_ID_82571EB_QUAD_COPPER:
3699         case E1000_DEV_ID_82571EB_QUAD_FIBER:
3700         case E1000_DEV_ID_82571EB_QUAD_COPPER_LP:
3701                 /* if quad port adapter, disable WoL on all but port A */
3702                 if (global_quad_port_a != 0)
3703                         adapter->wol = 0;
3704                 /* Reset for multiple quad port adapters */
3705                 if (++global_quad_port_a == 4)
3706                         global_quad_port_a = 0;
3707                 break;
3708         }
3709         return;
3710 }
3711
3712
3713 /*
3714  * Enable PCI Wake On Lan capability
3715  */
3716 static void
3717 em_enable_wakeup(if_ctx_t ctx)
3718 {
3719         struct adapter *adapter = iflib_get_softc(ctx);
3720         device_t dev = iflib_get_dev(ctx);
3721         if_t ifp = iflib_get_ifp(ctx);
3722         int error = 0;
3723         u32 pmc, ctrl, ctrl_ext, rctl;
3724         u16 status;
3725
3726         if (pci_find_cap(dev, PCIY_PMG, &pmc) != 0)
3727                 return;
3728
3729         /*
3730          * Determine type of Wakeup: note that wol
3731          * is set with all bits on by default.
3732          */
3733         if ((if_getcapenable(ifp) & IFCAP_WOL_MAGIC) == 0)
3734                 adapter->wol &= ~E1000_WUFC_MAG;
3735
3736         if ((if_getcapenable(ifp) & IFCAP_WOL_UCAST) == 0)
3737                 adapter->wol &= ~E1000_WUFC_EX;
3738
3739         if ((if_getcapenable(ifp) & IFCAP_WOL_MCAST) == 0)
3740                 adapter->wol &= ~E1000_WUFC_MC;
3741         else {
3742                 rctl = E1000_READ_REG(&adapter->hw, E1000_RCTL);
3743                 rctl |= E1000_RCTL_MPE;
3744                 E1000_WRITE_REG(&adapter->hw, E1000_RCTL, rctl);
3745         }
3746
3747         if (!(adapter->wol & (E1000_WUFC_EX | E1000_WUFC_MAG | E1000_WUFC_MC)))
3748                 goto pme;
3749
3750         /* Advertise the wakeup capability */
3751         ctrl = E1000_READ_REG(&adapter->hw, E1000_CTRL);
3752         ctrl |= (E1000_CTRL_SWDPIN2 | E1000_CTRL_SWDPIN3);
3753         E1000_WRITE_REG(&adapter->hw, E1000_CTRL, ctrl);
3754
3755         /* Keep the laser running on Fiber adapters */
3756         if (adapter->hw.phy.media_type == e1000_media_type_fiber ||
3757             adapter->hw.phy.media_type == e1000_media_type_internal_serdes) {
3758                 ctrl_ext = E1000_READ_REG(&adapter->hw, E1000_CTRL_EXT);
3759                 ctrl_ext |= E1000_CTRL_EXT_SDP3_DATA;
3760                 E1000_WRITE_REG(&adapter->hw, E1000_CTRL_EXT, ctrl_ext);
3761         }
3762
3763         if ((adapter->hw.mac.type == e1000_ich8lan) ||
3764             (adapter->hw.mac.type == e1000_pchlan) ||
3765             (adapter->hw.mac.type == e1000_ich9lan) ||
3766             (adapter->hw.mac.type == e1000_ich10lan))
3767                 e1000_suspend_workarounds_ich8lan(&adapter->hw);
3768
3769         if ( adapter->hw.mac.type >= e1000_pchlan) {
3770                 error = em_enable_phy_wakeup(adapter);
3771                 if (error)
3772                         goto pme;
3773         } else {
3774                 /* Enable wakeup by the MAC */
3775                 E1000_WRITE_REG(&adapter->hw, E1000_WUC, E1000_WUC_PME_EN);
3776                 E1000_WRITE_REG(&adapter->hw, E1000_WUFC, adapter->wol);
3777         }
3778
3779         if (adapter->hw.phy.type == e1000_phy_igp_3)
3780                 e1000_igp3_phy_powerdown_workaround_ich8lan(&adapter->hw);
3781
3782 pme:
3783         status = pci_read_config(dev, pmc + PCIR_POWER_STATUS, 2);
3784         status &= ~(PCIM_PSTAT_PME | PCIM_PSTAT_PMEENABLE);
3785         if (!error && (if_getcapenable(ifp) & IFCAP_WOL))
3786                 status |= PCIM_PSTAT_PME | PCIM_PSTAT_PMEENABLE;
3787         pci_write_config(dev, pmc + PCIR_POWER_STATUS, status, 2);
3788
3789         return;
3790 }
3791
3792 /*
3793  * WOL in the newer chipset interfaces (pchlan)
3794  * require thing to be copied into the phy
3795  */
3796 static int
3797 em_enable_phy_wakeup(struct adapter *adapter)
3798 {
3799         struct e1000_hw *hw = &adapter->hw;
3800         u32 mreg, ret = 0;
3801         u16 preg;
3802
3803         /* copy MAC RARs to PHY RARs */
3804         e1000_copy_rx_addrs_to_phy_ich8lan(hw);
3805
3806         /* copy MAC MTA to PHY MTA */
3807         for (int i = 0; i < hw->mac.mta_reg_count; i++) {
3808                 mreg = E1000_READ_REG_ARRAY(hw, E1000_MTA, i);
3809                 e1000_write_phy_reg(hw, BM_MTA(i), (u16)(mreg & 0xFFFF));
3810                 e1000_write_phy_reg(hw, BM_MTA(i) + 1,
3811                     (u16)((mreg >> 16) & 0xFFFF));
3812         }
3813
3814         /* configure PHY Rx Control register */
3815         e1000_read_phy_reg(hw, BM_RCTL, &preg);
3816         mreg = E1000_READ_REG(hw, E1000_RCTL);
3817         if (mreg & E1000_RCTL_UPE)
3818                 preg |= BM_RCTL_UPE;
3819         if (mreg & E1000_RCTL_MPE)
3820                 preg |= BM_RCTL_MPE;
3821         preg &= ~(BM_RCTL_MO_MASK);
3822         if (mreg & E1000_RCTL_MO_3)
3823                 preg |= (((mreg & E1000_RCTL_MO_3) >> E1000_RCTL_MO_SHIFT)
3824                                 << BM_RCTL_MO_SHIFT);
3825         if (mreg & E1000_RCTL_BAM)
3826                 preg |= BM_RCTL_BAM;
3827         if (mreg & E1000_RCTL_PMCF)
3828                 preg |= BM_RCTL_PMCF;
3829         mreg = E1000_READ_REG(hw, E1000_CTRL);
3830         if (mreg & E1000_CTRL_RFCE)
3831                 preg |= BM_RCTL_RFCE;
3832         e1000_write_phy_reg(hw, BM_RCTL, preg);
3833
3834         /* enable PHY wakeup in MAC register */
3835         E1000_WRITE_REG(hw, E1000_WUC,
3836             E1000_WUC_PHY_WAKE | E1000_WUC_PME_EN | E1000_WUC_APME);
3837         E1000_WRITE_REG(hw, E1000_WUFC, adapter->wol);
3838
3839         /* configure and enable PHY wakeup in PHY registers */
3840         e1000_write_phy_reg(hw, BM_WUFC, adapter->wol);
3841         e1000_write_phy_reg(hw, BM_WUC, E1000_WUC_PME_EN);
3842
3843         /* activate PHY wakeup */
3844         ret = hw->phy.ops.acquire(hw);
3845         if (ret) {
3846                 printf("Could not acquire PHY\n");
3847                 return ret;
3848         }
3849         e1000_write_phy_reg_mdic(hw, IGP01E1000_PHY_PAGE_SELECT,
3850                                  (BM_WUC_ENABLE_PAGE << IGP_PAGE_SHIFT));
3851         ret = e1000_read_phy_reg_mdic(hw, BM_WUC_ENABLE_REG, &preg);
3852         if (ret) {
3853                 printf("Could not read PHY page 769\n");
3854                 goto out;
3855         }
3856         preg |= BM_WUC_ENABLE_BIT | BM_WUC_HOST_WU_BIT;
3857         ret = e1000_write_phy_reg_mdic(hw, BM_WUC_ENABLE_REG, preg);
3858         if (ret)
3859                 printf("Could not set PHY Host Wakeup bit\n");
3860 out:
3861         hw->phy.ops.release(hw);
3862
3863         return ret;
3864 }
3865
3866 static void
3867 em_if_led_func(if_ctx_t ctx, int onoff)
3868 {
3869         struct adapter *adapter = iflib_get_softc(ctx);
3870
3871         if (onoff) {
3872                 e1000_setup_led(&adapter->hw);
3873                 e1000_led_on(&adapter->hw);
3874         } else {
3875                 e1000_led_off(&adapter->hw);
3876                 e1000_cleanup_led(&adapter->hw);
3877         }
3878 }
3879
3880 /*
3881  * Disable the L0S and L1 LINK states
3882  */
3883 static void
3884 em_disable_aspm(struct adapter *adapter)
3885 {
3886         int base, reg;
3887         u16 link_cap,link_ctrl;
3888         device_t dev = adapter->dev;
3889
3890         switch (adapter->hw.mac.type) {
3891         case e1000_82573:
3892         case e1000_82574:
3893         case e1000_82583:
3894                 break;
3895         default:
3896                 return;
3897         }
3898         if (pci_find_cap(dev, PCIY_EXPRESS, &base) != 0)
3899                 return;
3900         reg = base + PCIER_LINK_CAP;
3901         link_cap = pci_read_config(dev, reg, 2);
3902         if ((link_cap & PCIEM_LINK_CAP_ASPM) == 0)
3903                 return;
3904         reg = base + PCIER_LINK_CTL;
3905         link_ctrl = pci_read_config(dev, reg, 2);
3906         link_ctrl &= ~PCIEM_LINK_CTL_ASPMC;
3907         pci_write_config(dev, reg, link_ctrl, 2);
3908         return;
3909 }
3910
3911 /**********************************************************************
3912  *
3913  *  Update the board statistics counters.
3914  *
3915  **********************************************************************/
3916 static void
3917 em_update_stats_counters(struct adapter *adapter)
3918 {
3919         u64 prev_xoffrxc = adapter->stats.xoffrxc;
3920
3921         if(adapter->hw.phy.media_type == e1000_media_type_copper ||
3922            (E1000_READ_REG(&adapter->hw, E1000_STATUS) & E1000_STATUS_LU)) {
3923                 adapter->stats.symerrs += E1000_READ_REG(&adapter->hw, E1000_SYMERRS);
3924                 adapter->stats.sec += E1000_READ_REG(&adapter->hw, E1000_SEC);
3925         }
3926         adapter->stats.crcerrs += E1000_READ_REG(&adapter->hw, E1000_CRCERRS);
3927         adapter->stats.mpc += E1000_READ_REG(&adapter->hw, E1000_MPC);
3928         adapter->stats.scc += E1000_READ_REG(&adapter->hw, E1000_SCC);
3929         adapter->stats.ecol += E1000_READ_REG(&adapter->hw, E1000_ECOL);
3930
3931         adapter->stats.mcc += E1000_READ_REG(&adapter->hw, E1000_MCC);
3932         adapter->stats.latecol += E1000_READ_REG(&adapter->hw, E1000_LATECOL);
3933         adapter->stats.colc += E1000_READ_REG(&adapter->hw, E1000_COLC);
3934         adapter->stats.dc += E1000_READ_REG(&adapter->hw, E1000_DC);
3935         adapter->stats.rlec += E1000_READ_REG(&adapter->hw, E1000_RLEC);
3936         adapter->stats.xonrxc += E1000_READ_REG(&adapter->hw, E1000_XONRXC);
3937         adapter->stats.xontxc += E1000_READ_REG(&adapter->hw, E1000_XONTXC);
3938         adapter->stats.xoffrxc += E1000_READ_REG(&adapter->hw, E1000_XOFFRXC);
3939         /*
3940          ** For watchdog management we need to know if we have been
3941          ** paused during the last interval, so capture that here.
3942         */
3943         if (adapter->stats.xoffrxc != prev_xoffrxc)
3944                 adapter->shared->isc_pause_frames = 1;
3945         adapter->stats.xofftxc += E1000_READ_REG(&adapter->hw, E1000_XOFFTXC);
3946         adapter->stats.fcruc += E1000_READ_REG(&adapter->hw, E1000_FCRUC);
3947         adapter->stats.prc64 += E1000_READ_REG(&adapter->hw, E1000_PRC64);
3948         adapter->stats.prc127 += E1000_READ_REG(&adapter->hw, E1000_PRC127);
3949         adapter->stats.prc255 += E1000_READ_REG(&adapter->hw, E1000_PRC255);
3950         adapter->stats.prc511 += E1000_READ_REG(&adapter->hw, E1000_PRC511);
3951         adapter->stats.prc1023 += E1000_READ_REG(&adapter->hw, E1000_PRC1023);
3952         adapter->stats.prc1522 += E1000_READ_REG(&adapter->hw, E1000_PRC1522);
3953         adapter->stats.gprc += E1000_READ_REG(&adapter->hw, E1000_GPRC);
3954         adapter->stats.bprc += E1000_READ_REG(&adapter->hw, E1000_BPRC);
3955         adapter->stats.mprc += E1000_READ_REG(&adapter->hw, E1000_MPRC);
3956         adapter->stats.gptc += E1000_READ_REG(&adapter->hw, E1000_GPTC);
3957
3958         /* For the 64-bit byte counters the low dword must be read first. */
3959         /* Both registers clear on the read of the high dword */
3960
3961         adapter->stats.gorc += E1000_READ_REG(&adapter->hw, E1000_GORCL) +
3962             ((u64)E1000_READ_REG(&adapter->hw, E1000_GORCH) << 32);
3963         adapter->stats.gotc += E1000_READ_REG(&adapter->hw, E1000_GOTCL) +
3964             ((u64)E1000_READ_REG(&adapter->hw, E1000_GOTCH) << 32);
3965
3966         adapter->stats.rnbc += E1000_READ_REG(&adapter->hw, E1000_RNBC);
3967         adapter->stats.ruc += E1000_READ_REG(&adapter->hw, E1000_RUC);
3968         adapter->stats.rfc += E1000_READ_REG(&adapter->hw, E1000_RFC);
3969         adapter->stats.roc += E1000_READ_REG(&adapter->hw, E1000_ROC);
3970         adapter->stats.rjc += E1000_READ_REG(&adapter->hw, E1000_RJC);
3971
3972         adapter->stats.tor += E1000_READ_REG(&adapter->hw, E1000_TORH);
3973         adapter->stats.tot += E1000_READ_REG(&adapter->hw, E1000_TOTH);
3974
3975         adapter->stats.tpr += E1000_READ_REG(&adapter->hw, E1000_TPR);
3976         adapter->stats.tpt += E1000_READ_REG(&adapter->hw, E1000_TPT);
3977         adapter->stats.ptc64 += E1000_READ_REG(&adapter->hw, E1000_PTC64);
3978         adapter->stats.ptc127 += E1000_READ_REG(&adapter->hw, E1000_PTC127);
3979         adapter->stats.ptc255 += E1000_READ_REG(&adapter->hw, E1000_PTC255);
3980         adapter->stats.ptc511 += E1000_READ_REG(&adapter->hw, E1000_PTC511);
3981         adapter->stats.ptc1023 += E1000_READ_REG(&adapter->hw, E1000_PTC1023);
3982         adapter->stats.ptc1522 += E1000_READ_REG(&adapter->hw, E1000_PTC1522);
3983         adapter->stats.mptc += E1000_READ_REG(&adapter->hw, E1000_MPTC);
3984         adapter->stats.bptc += E1000_READ_REG(&adapter->hw, E1000_BPTC);
3985
3986         /* Interrupt Counts */
3987
3988         adapter->stats.iac += E1000_READ_REG(&adapter->hw, E1000_IAC);
3989         adapter->stats.icrxptc += E1000_READ_REG(&adapter->hw, E1000_ICRXPTC);
3990         adapter->stats.icrxatc += E1000_READ_REG(&adapter->hw, E1000_ICRXATC);
3991         adapter->stats.ictxptc += E1000_READ_REG(&adapter->hw, E1000_ICTXPTC);
3992         adapter->stats.ictxatc += E1000_READ_REG(&adapter->hw, E1000_ICTXATC);
3993         adapter->stats.ictxqec += E1000_READ_REG(&adapter->hw, E1000_ICTXQEC);
3994         adapter->stats.ictxqmtc += E1000_READ_REG(&adapter->hw, E1000_ICTXQMTC);
3995         adapter->stats.icrxdmtc += E1000_READ_REG(&adapter->hw, E1000_ICRXDMTC);
3996         adapter->stats.icrxoc += E1000_READ_REG(&adapter->hw, E1000_ICRXOC);
3997
3998         if (adapter->hw.mac.type >= e1000_82543) {
3999                 adapter->stats.algnerrc +=
4000                 E1000_READ_REG(&adapter->hw, E1000_ALGNERRC);
4001                 adapter->stats.rxerrc +=
4002                 E1000_READ_REG(&adapter->hw, E1000_RXERRC);
4003                 adapter->stats.tncrs +=
4004                 E1000_READ_REG(&adapter->hw, E1000_TNCRS);
4005                 adapter->stats.cexterr +=
4006                 E1000_READ_REG(&adapter->hw, E1000_CEXTERR);
4007                 adapter->stats.tsctc +=
4008                 E1000_READ_REG(&adapter->hw, E1000_TSCTC);
4009                 adapter->stats.tsctfc +=
4010                 E1000_READ_REG(&adapter->hw, E1000_TSCTFC);
4011         }
4012 }
4013
4014 static uint64_t
4015 em_if_get_counter(if_ctx_t ctx, ift_counter cnt)
4016 {
4017         struct adapter *adapter = iflib_get_softc(ctx);
4018         struct ifnet *ifp = iflib_get_ifp(ctx);
4019
4020         switch (cnt) {
4021         case IFCOUNTER_COLLISIONS:
4022                 return (adapter->stats.colc);
4023         case IFCOUNTER_IERRORS:
4024                 return (adapter->dropped_pkts + adapter->stats.rxerrc +
4025                     adapter->stats.crcerrs + adapter->stats.algnerrc +
4026                     adapter->stats.ruc + adapter->stats.roc +
4027                     adapter->stats.mpc + adapter->stats.cexterr);
4028         case IFCOUNTER_OERRORS:
4029                 return (adapter->stats.ecol + adapter->stats.latecol +
4030                     adapter->watchdog_events);
4031         default:
4032                 return (if_get_counter_default(ifp, cnt));
4033         }
4034 }
4035
4036 /* em_if_needs_restart - Tell iflib when the driver needs to be reinitialized
4037  * @ctx: iflib context
4038  * @event: event code to check
4039  *
4040  * Defaults to returning true for unknown events.
4041  *
4042  * @returns true if iflib needs to reinit the interface
4043  */
4044 static bool
4045 em_if_needs_restart(if_ctx_t ctx __unused, enum iflib_restart_event event)
4046 {
4047         switch (event) {
4048         case IFLIB_RESTART_VLAN_CONFIG:
4049         default:
4050                 return (true);
4051         }
4052 }
4053
4054 /* Export a single 32-bit register via a read-only sysctl. */
4055 static int
4056 em_sysctl_reg_handler(SYSCTL_HANDLER_ARGS)
4057 {
4058         struct adapter *adapter;
4059         u_int val;
4060
4061         adapter = oidp->oid_arg1;
4062         val = E1000_READ_REG(&adapter->hw, oidp->oid_arg2);
4063         return (sysctl_handle_int(oidp, &val, 0, req));
4064 }
4065
4066 /*
4067  * Add sysctl variables, one per statistic, to the system.
4068  */
4069 static void
4070 em_add_hw_stats(struct adapter *adapter)
4071 {
4072         device_t dev = iflib_get_dev(adapter->ctx);
4073         struct em_tx_queue *tx_que = adapter->tx_queues;
4074         struct em_rx_queue *rx_que = adapter->rx_queues;
4075
4076         struct sysctl_ctx_list *ctx = device_get_sysctl_ctx(dev);
4077         struct sysctl_oid *tree = device_get_sysctl_tree(dev);
4078         struct sysctl_oid_list *child = SYSCTL_CHILDREN(tree);
4079         struct e1000_hw_stats *stats = &adapter->stats;
4080
4081         struct sysctl_oid *stat_node, *queue_node, *int_node;
4082         struct sysctl_oid_list *stat_list, *queue_list, *int_list;
4083
4084 #define QUEUE_NAME_LEN 32
4085         char namebuf[QUEUE_NAME_LEN];
4086
4087         /* Driver Statistics */
4088         SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "dropped",
4089                         CTLFLAG_RD, &adapter->dropped_pkts,
4090                         "Driver dropped packets");
4091         SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "link_irq",
4092                         CTLFLAG_RD, &adapter->link_irq,
4093                         "Link MSI-X IRQ Handled");
4094         SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "rx_overruns",
4095                         CTLFLAG_RD, &adapter->rx_overruns,
4096                         "RX overruns");
4097         SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "watchdog_timeouts",
4098                         CTLFLAG_RD, &adapter->watchdog_events,
4099                         "Watchdog timeouts");
4100         SYSCTL_ADD_PROC(ctx, child, OID_AUTO, "device_control",
4101             CTLTYPE_UINT | CTLFLAG_RD | CTLFLAG_NEEDGIANT,
4102             adapter, E1000_CTRL, em_sysctl_reg_handler, "IU",
4103             "Device Control Register");
4104         SYSCTL_ADD_PROC(ctx, child, OID_AUTO, "rx_control",
4105             CTLTYPE_UINT | CTLFLAG_RD | CTLFLAG_NEEDGIANT,
4106             adapter, E1000_RCTL, em_sysctl_reg_handler, "IU",
4107             "Receiver Control Register");
4108         SYSCTL_ADD_UINT(ctx, child, OID_AUTO, "fc_high_water",
4109                         CTLFLAG_RD, &adapter->hw.fc.high_water, 0,
4110                         "Flow Control High Watermark");
4111         SYSCTL_ADD_UINT(ctx, child, OID_AUTO, "fc_low_water",
4112                         CTLFLAG_RD, &adapter->hw.fc.low_water, 0,
4113                         "Flow Control Low Watermark");
4114
4115         for (int i = 0; i < adapter->tx_num_queues; i++, tx_que++) {
4116                 struct tx_ring *txr = &tx_que->txr;
4117                 snprintf(namebuf, QUEUE_NAME_LEN, "queue_tx_%d", i);
4118                 queue_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, namebuf,
4119                     CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, "TX Queue Name");
4120                 queue_list = SYSCTL_CHILDREN(queue_node);
4121
4122                 SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "txd_head",
4123                     CTLTYPE_UINT | CTLFLAG_RD | CTLFLAG_NEEDGIANT, adapter,
4124                     E1000_TDH(txr->me), em_sysctl_reg_handler, "IU",
4125                     "Transmit Descriptor Head");
4126                 SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "txd_tail",
4127                     CTLTYPE_UINT | CTLFLAG_RD | CTLFLAG_NEEDGIANT, adapter,
4128                     E1000_TDT(txr->me), em_sysctl_reg_handler, "IU",
4129                     "Transmit Descriptor Tail");
4130                 SYSCTL_ADD_ULONG(ctx, queue_list, OID_AUTO, "tx_irq",
4131                                 CTLFLAG_RD, &txr->tx_irq,
4132                                 "Queue MSI-X Transmit Interrupts");
4133         }
4134
4135         for (int j = 0; j < adapter->rx_num_queues; j++, rx_que++) {
4136                 struct rx_ring *rxr = &rx_que->rxr;
4137                 snprintf(namebuf, QUEUE_NAME_LEN, "queue_rx_%d", j);
4138                 queue_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, namebuf,
4139                     CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, "RX Queue Name");
4140                 queue_list = SYSCTL_CHILDREN(queue_node);
4141
4142                 SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "rxd_head",
4143                     CTLTYPE_UINT | CTLFLAG_RD | CTLFLAG_NEEDGIANT, adapter,
4144                     E1000_RDH(rxr->me), em_sysctl_reg_handler, "IU",
4145                     "Receive Descriptor Head");
4146                 SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "rxd_tail",
4147                     CTLTYPE_UINT | CTLFLAG_RD | CTLFLAG_NEEDGIANT, adapter,
4148                     E1000_RDT(rxr->me), em_sysctl_reg_handler, "IU",
4149                     "Receive Descriptor Tail");
4150                 SYSCTL_ADD_ULONG(ctx, queue_list, OID_AUTO, "rx_irq",
4151                                 CTLFLAG_RD, &rxr->rx_irq,
4152                                 "Queue MSI-X Receive Interrupts");
4153         }
4154
4155         /* MAC stats get their own sub node */
4156
4157         stat_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, "mac_stats",
4158             CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, "Statistics");
4159         stat_list = SYSCTL_CHILDREN(stat_node);
4160
4161         SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "excess_coll",
4162                         CTLFLAG_RD, &stats->ecol,
4163                         "Excessive collisions");
4164         SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "single_coll",
4165                         CTLFLAG_RD, &stats->scc,
4166                         "Single collisions");
4167         SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "multiple_coll",
4168                         CTLFLAG_RD, &stats->mcc,
4169                         "Multiple collisions");
4170         SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "late_coll",
4171                         CTLFLAG_RD, &stats->latecol,
4172                         "Late collisions");
4173         SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "collision_count",
4174                         CTLFLAG_RD, &stats->colc,
4175                         "Collision Count");
4176         SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "symbol_errors",
4177                         CTLFLAG_RD, &adapter->stats.symerrs,
4178                         "Symbol Errors");
4179         SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "sequence_errors",
4180                         CTLFLAG_RD, &adapter->stats.sec,
4181                         "Sequence Errors");
4182         SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "defer_count",
4183                         CTLFLAG_RD, &adapter->stats.dc,
4184                         "Defer Count");
4185         SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "missed_packets",
4186                         CTLFLAG_RD, &adapter->stats.mpc,
4187                         "Missed Packets");
4188         SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_no_buff",
4189                         CTLFLAG_RD, &adapter->stats.rnbc,
4190                         "Receive No Buffers");
4191         SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_undersize",
4192                         CTLFLAG_RD, &adapter->stats.ruc,
4193                         "Receive Undersize");
4194         SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_fragmented",
4195                         CTLFLAG_RD, &adapter->stats.rfc,
4196                         "Fragmented Packets Received ");
4197         SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_oversize",
4198                         CTLFLAG_RD, &adapter->stats.roc,
4199                         "Oversized Packets Received");
4200         SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_jabber",
4201                         CTLFLAG_RD, &adapter->stats.rjc,
4202                         "Recevied Jabber");
4203         SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_errs",
4204                         CTLFLAG_RD, &adapter->stats.rxerrc,
4205                         "Receive Errors");
4206         SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "crc_errs",
4207                         CTLFLAG_RD, &adapter->stats.crcerrs,
4208                         "CRC errors");
4209         SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "alignment_errs",
4210                         CTLFLAG_RD, &adapter->stats.algnerrc,
4211                         "Alignment Errors");
4212         /* On 82575 these are collision counts */
4213         SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "coll_ext_errs",
4214                         CTLFLAG_RD, &adapter->stats.cexterr,
4215                         "Collision/Carrier extension errors");
4216         SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xon_recvd",
4217                         CTLFLAG_RD, &adapter->stats.xonrxc,
4218                         "XON Received");
4219         SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xon_txd",
4220                         CTLFLAG_RD, &adapter->stats.xontxc,
4221                         "XON Transmitted");
4222         SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xoff_recvd",
4223                         CTLFLAG_RD, &adapter->stats.xoffrxc,
4224                         "XOFF Received");
4225         SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xoff_txd",
4226                         CTLFLAG_RD, &adapter->stats.xofftxc,
4227                         "XOFF Transmitted");
4228
4229         /* Packet Reception Stats */
4230         SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "total_pkts_recvd",
4231                         CTLFLAG_RD, &adapter->stats.tpr,
4232                         "Total Packets Received ");
4233         SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_pkts_recvd",
4234                         CTLFLAG_RD, &adapter->stats.gprc,
4235                         "Good Packets Received");
4236         SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "bcast_pkts_recvd",
4237                         CTLFLAG_RD, &adapter->stats.bprc,
4238                         "Broadcast Packets Received");
4239         SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "mcast_pkts_recvd",
4240                         CTLFLAG_RD, &adapter->stats.mprc,
4241                         "Multicast Packets Received");
4242         SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_64",
4243                         CTLFLAG_RD, &adapter->stats.prc64,
4244                         "64 byte frames received ");
4245         SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_65_127",
4246                         CTLFLAG_RD, &adapter->stats.prc127,
4247                         "65-127 byte frames received");
4248         SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_128_255",
4249                         CTLFLAG_RD, &adapter->stats.prc255,
4250                         "128-255 byte frames received");
4251         SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_256_511",
4252                         CTLFLAG_RD, &adapter->stats.prc511,
4253                         "256-511 byte frames received");
4254         SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_512_1023",
4255                         CTLFLAG_RD, &adapter->stats.prc1023,
4256                         "512-1023 byte frames received");
4257         SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_1024_1522",
4258                         CTLFLAG_RD, &adapter->stats.prc1522,
4259                         "1023-1522 byte frames received");
4260         SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_octets_recvd",
4261                         CTLFLAG_RD, &adapter->stats.gorc,
4262                         "Good Octets Received");
4263
4264         /* Packet Transmission Stats */
4265         SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_octets_txd",
4266                         CTLFLAG_RD, &adapter->stats.gotc,
4267                         "Good Octets Transmitted");
4268         SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "total_pkts_txd",
4269                         CTLFLAG_RD, &adapter->stats.tpt,
4270                         "Total Packets Transmitted");
4271         SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_pkts_txd",
4272                         CTLFLAG_RD, &adapter->stats.gptc,
4273                         "Good Packets Transmitted");
4274         SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "bcast_pkts_txd",
4275                         CTLFLAG_RD, &adapter->stats.bptc,
4276                         "Broadcast Packets Transmitted");
4277         SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "mcast_pkts_txd",
4278                         CTLFLAG_RD, &adapter->stats.mptc,
4279                         "Multicast Packets Transmitted");
4280         SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_64",
4281                         CTLFLAG_RD, &adapter->stats.ptc64,
4282                         "64 byte frames transmitted ");
4283         SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_65_127",
4284                         CTLFLAG_RD, &adapter->stats.ptc127,
4285                         "65-127 byte frames transmitted");
4286         SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_128_255",
4287                         CTLFLAG_RD, &adapter->stats.ptc255,
4288                         "128-255 byte frames transmitted");
4289         SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_256_511",
4290                         CTLFLAG_RD, &adapter->stats.ptc511,
4291                         "256-511 byte frames transmitted");
4292         SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_512_1023",
4293                         CTLFLAG_RD, &adapter->stats.ptc1023,
4294                         "512-1023 byte frames transmitted");
4295         SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_1024_1522",
4296                         CTLFLAG_RD, &adapter->stats.ptc1522,
4297                         "1024-1522 byte frames transmitted");
4298         SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tso_txd",
4299                         CTLFLAG_RD, &adapter->stats.tsctc,
4300                         "TSO Contexts Transmitted");
4301         SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tso_ctx_fail",
4302                         CTLFLAG_RD, &adapter->stats.tsctfc,
4303                         "TSO Contexts Failed");
4304
4305
4306         /* Interrupt Stats */
4307
4308         int_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, "interrupts",
4309             CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, "Interrupt Statistics");
4310         int_list = SYSCTL_CHILDREN(int_node);
4311
4312         SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "asserts",
4313                         CTLFLAG_RD, &adapter->stats.iac,
4314                         "Interrupt Assertion Count");
4315
4316         SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "rx_pkt_timer",
4317                         CTLFLAG_RD, &adapter->stats.icrxptc,
4318                         "Interrupt Cause Rx Pkt Timer Expire Count");
4319
4320         SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "rx_abs_timer",
4321                         CTLFLAG_RD, &adapter->stats.icrxatc,
4322                         "Interrupt Cause Rx Abs Timer Expire Count");
4323
4324         SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "tx_pkt_timer",
4325                         CTLFLAG_RD, &adapter->stats.ictxptc,
4326                         "Interrupt Cause Tx Pkt Timer Expire Count");
4327
4328         SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "tx_abs_timer",
4329                         CTLFLAG_RD, &adapter->stats.ictxatc,
4330                         "Interrupt Cause Tx Abs Timer Expire Count");
4331
4332         SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "tx_queue_empty",
4333                         CTLFLAG_RD, &adapter->stats.ictxqec,
4334                         "Interrupt Cause Tx Queue Empty Count");
4335
4336         SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "tx_queue_min_thresh",
4337                         CTLFLAG_RD, &adapter->stats.ictxqmtc,
4338                         "Interrupt Cause Tx Queue Min Thresh Count");
4339
4340         SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "rx_desc_min_thresh",
4341                         CTLFLAG_RD, &adapter->stats.icrxdmtc,
4342                         "Interrupt Cause Rx Desc Min Thresh Count");
4343
4344         SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "rx_overrun",
4345                         CTLFLAG_RD, &adapter->stats.icrxoc,
4346                         "Interrupt Cause Receiver Overrun Count");
4347 }
4348
4349 /**********************************************************************
4350  *
4351  *  This routine provides a way to dump out the adapter eeprom,
4352  *  often a useful debug/service tool. This only dumps the first
4353  *  32 words, stuff that matters is in that extent.
4354  *
4355  **********************************************************************/
4356 static int
4357 em_sysctl_nvm_info(SYSCTL_HANDLER_ARGS)
4358 {
4359         struct adapter *adapter = (struct adapter *)arg1;
4360         int error;
4361         int result;
4362
4363         result = -1;
4364         error = sysctl_handle_int(oidp, &result, 0, req);
4365
4366         if (error || !req->newptr)
4367                 return (error);
4368
4369         /*
4370          * This value will cause a hex dump of the
4371          * first 32 16-bit words of the EEPROM to
4372          * the screen.
4373          */
4374         if (result == 1)
4375                 em_print_nvm_info(adapter);
4376
4377         return (error);
4378 }
4379
4380 static void
4381 em_print_nvm_info(struct adapter *adapter)
4382 {
4383         u16 eeprom_data;
4384         int i, j, row = 0;
4385
4386         /* Its a bit crude, but it gets the job done */
4387         printf("\nInterface EEPROM Dump:\n");
4388         printf("Offset\n0x0000  ");
4389         for (i = 0, j = 0; i < 32; i++, j++) {
4390                 if (j == 8) { /* Make the offset block */
4391                         j = 0; ++row;
4392                         printf("\n0x00%x0  ",row);
4393                 }
4394                 e1000_read_nvm(&adapter->hw, i, 1, &eeprom_data);
4395                 printf("%04x ", eeprom_data);
4396         }
4397         printf("\n");
4398 }
4399
4400 static int
4401 em_sysctl_int_delay(SYSCTL_HANDLER_ARGS)
4402 {
4403         struct em_int_delay_info *info;
4404         struct adapter *adapter;
4405         u32 regval;
4406         int error, usecs, ticks;
4407
4408         info = (struct em_int_delay_info *) arg1;
4409         usecs = info->value;
4410         error = sysctl_handle_int(oidp, &usecs, 0, req);
4411         if (error != 0 || req->newptr == NULL)
4412                 return (error);
4413         if (usecs < 0 || usecs > EM_TICKS_TO_USECS(65535))
4414                 return (EINVAL);
4415         info->value = usecs;
4416         ticks = EM_USECS_TO_TICKS(usecs);
4417         if (info->offset == E1000_ITR)  /* units are 256ns here */
4418                 ticks *= 4;
4419
4420         adapter = info->adapter;
4421
4422         regval = E1000_READ_OFFSET(&adapter->hw, info->offset);
4423         regval = (regval & ~0xffff) | (ticks & 0xffff);
4424         /* Handle a few special cases. */
4425         switch (info->offset) {
4426         case E1000_RDTR:
4427                 break;
4428         case E1000_TIDV:
4429                 if (ticks == 0) {
4430                         adapter->txd_cmd &= ~E1000_TXD_CMD_IDE;
4431                         /* Don't write 0 into the TIDV register. */
4432                         regval++;
4433                 } else
4434                         adapter->txd_cmd |= E1000_TXD_CMD_IDE;
4435                 break;
4436         }
4437         E1000_WRITE_OFFSET(&adapter->hw, info->offset, regval);
4438         return (0);
4439 }
4440
4441 static void
4442 em_add_int_delay_sysctl(struct adapter *adapter, const char *name,
4443         const char *description, struct em_int_delay_info *info,
4444         int offset, int value)
4445 {
4446         info->adapter = adapter;
4447         info->offset = offset;
4448         info->value = value;
4449         SYSCTL_ADD_PROC(device_get_sysctl_ctx(adapter->dev),
4450             SYSCTL_CHILDREN(device_get_sysctl_tree(adapter->dev)),
4451             OID_AUTO, name, CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_NEEDGIANT,
4452             info, 0, em_sysctl_int_delay, "I", description);
4453 }
4454
4455 /*
4456  * Set flow control using sysctl:
4457  * Flow control values:
4458  *      0 - off
4459  *      1 - rx pause
4460  *      2 - tx pause
4461  *      3 - full
4462  */
4463 static int
4464 em_set_flowcntl(SYSCTL_HANDLER_ARGS)
4465 {
4466         int error;
4467         static int input = 3; /* default is full */
4468         struct adapter  *adapter = (struct adapter *) arg1;
4469
4470         error = sysctl_handle_int(oidp, &input, 0, req);
4471
4472         if ((error) || (req->newptr == NULL))
4473                 return (error);
4474
4475         if (input == adapter->fc) /* no change? */
4476                 return (error);
4477
4478         switch (input) {
4479         case e1000_fc_rx_pause:
4480         case e1000_fc_tx_pause:
4481         case e1000_fc_full:
4482         case e1000_fc_none:
4483                 adapter->hw.fc.requested_mode = input;
4484                 adapter->fc = input;
4485                 break;
4486         default:
4487                 /* Do nothing */
4488                 return (error);
4489         }
4490
4491         adapter->hw.fc.current_mode = adapter->hw.fc.requested_mode;
4492         e1000_force_mac_fc(&adapter->hw);
4493         return (error);
4494 }
4495
4496 /*
4497  * Manage Energy Efficient Ethernet:
4498  * Control values:
4499  *     0/1 - enabled/disabled
4500  */
4501 static int
4502 em_sysctl_eee(SYSCTL_HANDLER_ARGS)
4503 {
4504         struct adapter *adapter = (struct adapter *) arg1;
4505         int error, value;
4506
4507         value = adapter->hw.dev_spec.ich8lan.eee_disable;
4508         error = sysctl_handle_int(oidp, &value, 0, req);
4509         if (error || req->newptr == NULL)
4510                 return (error);
4511         adapter->hw.dev_spec.ich8lan.eee_disable = (value != 0);
4512         em_if_init(adapter->ctx);
4513
4514         return (0);
4515 }
4516
4517 static int
4518 em_sysctl_debug_info(SYSCTL_HANDLER_ARGS)
4519 {
4520         struct adapter *adapter;
4521         int error;
4522         int result;
4523
4524         result = -1;
4525         error = sysctl_handle_int(oidp, &result, 0, req);
4526
4527         if (error || !req->newptr)
4528                 return (error);
4529
4530         if (result == 1) {
4531                 adapter = (struct adapter *) arg1;
4532                 em_print_debug_info(adapter);
4533         }
4534
4535         return (error);
4536 }
4537
4538 static int
4539 em_get_rs(SYSCTL_HANDLER_ARGS)
4540 {
4541         struct adapter *adapter = (struct adapter *) arg1;
4542         int error;
4543         int result;
4544
4545         result = 0;
4546         error = sysctl_handle_int(oidp, &result, 0, req);
4547
4548         if (error || !req->newptr || result != 1)
4549                 return (error);
4550         em_dump_rs(adapter);
4551
4552         return (error);
4553 }
4554
4555 static void
4556 em_if_debug(if_ctx_t ctx)
4557 {
4558         em_dump_rs(iflib_get_softc(ctx));
4559 }
4560
4561 /*
4562  * This routine is meant to be fluid, add whatever is
4563  * needed for debugging a problem.  -jfv
4564  */
4565 static void
4566 em_print_debug_info(struct adapter *adapter)
4567 {
4568         device_t dev = iflib_get_dev(adapter->ctx);
4569         struct ifnet *ifp = iflib_get_ifp(adapter->ctx);
4570         struct tx_ring *txr = &adapter->tx_queues->txr;
4571         struct rx_ring *rxr = &adapter->rx_queues->rxr;
4572
4573         if (if_getdrvflags(ifp) & IFF_DRV_RUNNING)
4574                 printf("Interface is RUNNING ");
4575         else
4576                 printf("Interface is NOT RUNNING\n");
4577
4578         if (if_getdrvflags(ifp) & IFF_DRV_OACTIVE)
4579                 printf("and INACTIVE\n");
4580         else
4581                 printf("and ACTIVE\n");
4582
4583         for (int i = 0; i < adapter->tx_num_queues; i++, txr++) {
4584                 device_printf(dev, "TX Queue %d ------\n", i);
4585                 device_printf(dev, "hw tdh = %d, hw tdt = %d\n",
4586                         E1000_READ_REG(&adapter->hw, E1000_TDH(i)),
4587                         E1000_READ_REG(&adapter->hw, E1000_TDT(i)));
4588
4589         }
4590         for (int j=0; j < adapter->rx_num_queues; j++, rxr++) {
4591                 device_printf(dev, "RX Queue %d ------\n", j);
4592                 device_printf(dev, "hw rdh = %d, hw rdt = %d\n",
4593                         E1000_READ_REG(&adapter->hw, E1000_RDH(j)),
4594                         E1000_READ_REG(&adapter->hw, E1000_RDT(j)));
4595         }
4596 }
4597
4598 /*
4599  * 82574 only:
4600  * Write a new value to the EEPROM increasing the number of MSI-X
4601  * vectors from 3 to 5, for proper multiqueue support.
4602  */
4603 static void
4604 em_enable_vectors_82574(if_ctx_t ctx)
4605 {
4606         struct adapter *adapter = iflib_get_softc(ctx);
4607         struct e1000_hw *hw = &adapter->hw;
4608         device_t dev = iflib_get_dev(ctx);
4609         u16 edata;
4610
4611         e1000_read_nvm(hw, EM_NVM_PCIE_CTRL, 1, &edata);
4612         if (bootverbose)
4613                 device_printf(dev, "EM_NVM_PCIE_CTRL = %#06x\n", edata);
4614         if (((edata & EM_NVM_MSIX_N_MASK) >> EM_NVM_MSIX_N_SHIFT) != 4) {
4615                 device_printf(dev, "Writing to eeprom: increasing "
4616                     "reported MSI-X vectors from 3 to 5...\n");
4617                 edata &= ~(EM_NVM_MSIX_N_MASK);
4618                 edata |= 4 << EM_NVM_MSIX_N_SHIFT;
4619                 e1000_write_nvm(hw, EM_NVM_PCIE_CTRL, 1, &edata);
4620                 e1000_update_nvm_checksum(hw);
4621                 device_printf(dev, "Writing to eeprom: done\n");
4622         }
4623 }