xref: /freebsd/sys/dev/e1000/if_em.c (revision 25cdacf79b06356c929e59d5074d26c9dac41bdf)
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 if_shared_ctx_t em_sctx = &em_sctx_init;
558 
559 static struct if_shared_ctx igb_sctx_init = {
560 	.isc_magic = IFLIB_MAGIC,
561 	.isc_q_align = PAGE_SIZE,
562 	.isc_tx_maxsize = EM_TSO_SIZE + sizeof(struct ether_vlan_header),
563 	.isc_tx_maxsegsize = PAGE_SIZE,
564 	.isc_tso_maxsize = EM_TSO_SIZE + sizeof(struct ether_vlan_header),
565 	.isc_tso_maxsegsize = EM_TSO_SEG_SIZE,
566 	.isc_rx_maxsize = MJUM9BYTES,
567 	.isc_rx_nsegments = 1,
568 	.isc_rx_maxsegsize = MJUM9BYTES,
569 	.isc_nfl = 1,
570 	.isc_nrxqs = 1,
571 	.isc_ntxqs = 1,
572 	.isc_admin_intrcnt = 1,
573 	.isc_vendor_info = igb_vendor_info_array,
574 	.isc_driver_version = em_driver_version,
575 	.isc_driver = &igb_if_driver,
576 	.isc_flags = IFLIB_NEED_SCRATCH | IFLIB_TSO_INIT_IP | IFLIB_NEED_ZERO_CSUM,
577 
578 	.isc_nrxd_min = {EM_MIN_RXD},
579 	.isc_ntxd_min = {EM_MIN_TXD},
580 	.isc_nrxd_max = {IGB_MAX_RXD},
581 	.isc_ntxd_max = {IGB_MAX_TXD},
582 	.isc_nrxd_default = {EM_DEFAULT_RXD},
583 	.isc_ntxd_default = {EM_DEFAULT_TXD},
584 };
585 
586 if_shared_ctx_t igb_sctx = &igb_sctx_init;
587 
588 /*****************************************************************
589  *
590  * Dump Registers
591  *
592  ****************************************************************/
593 #define IGB_REGS_LEN 739
594 
595 static int em_get_regs(SYSCTL_HANDLER_ARGS)
596 {
597 	struct adapter *adapter = (struct adapter *)arg1;
598 	struct e1000_hw *hw = &adapter->hw;
599 	struct sbuf *sb;
600 	u32 *regs_buff;
601 	int rc;
602 
603 	regs_buff = malloc(sizeof(u32) * IGB_REGS_LEN, M_DEVBUF, M_WAITOK);
604 	memset(regs_buff, 0, IGB_REGS_LEN * sizeof(u32));
605 
606 	rc = sysctl_wire_old_buffer(req, 0);
607 	MPASS(rc == 0);
608 	if (rc != 0) {
609 		free(regs_buff, M_DEVBUF);
610 		return (rc);
611 	}
612 
613 	sb = sbuf_new_for_sysctl(NULL, NULL, 32*400, req);
614 	MPASS(sb != NULL);
615 	if (sb == NULL) {
616 		free(regs_buff, M_DEVBUF);
617 		return (ENOMEM);
618 	}
619 
620 	/* General Registers */
621 	regs_buff[0] = E1000_READ_REG(hw, E1000_CTRL);
622 	regs_buff[1] = E1000_READ_REG(hw, E1000_STATUS);
623 	regs_buff[2] = E1000_READ_REG(hw, E1000_CTRL_EXT);
624 	regs_buff[3] = E1000_READ_REG(hw, E1000_ICR);
625 	regs_buff[4] = E1000_READ_REG(hw, E1000_RCTL);
626 	regs_buff[5] = E1000_READ_REG(hw, E1000_RDLEN(0));
627 	regs_buff[6] = E1000_READ_REG(hw, E1000_RDH(0));
628 	regs_buff[7] = E1000_READ_REG(hw, E1000_RDT(0));
629 	regs_buff[8] = E1000_READ_REG(hw, E1000_RXDCTL(0));
630 	regs_buff[9] = E1000_READ_REG(hw, E1000_RDBAL(0));
631 	regs_buff[10] = E1000_READ_REG(hw, E1000_RDBAH(0));
632 	regs_buff[11] = E1000_READ_REG(hw, E1000_TCTL);
633 	regs_buff[12] = E1000_READ_REG(hw, E1000_TDBAL(0));
634 	regs_buff[13] = E1000_READ_REG(hw, E1000_TDBAH(0));
635 	regs_buff[14] = E1000_READ_REG(hw, E1000_TDLEN(0));
636 	regs_buff[15] = E1000_READ_REG(hw, E1000_TDH(0));
637 	regs_buff[16] = E1000_READ_REG(hw, E1000_TDT(0));
638 	regs_buff[17] = E1000_READ_REG(hw, E1000_TXDCTL(0));
639 	regs_buff[18] = E1000_READ_REG(hw, E1000_TDFH);
640 	regs_buff[19] = E1000_READ_REG(hw, E1000_TDFT);
641 	regs_buff[20] = E1000_READ_REG(hw, E1000_TDFHS);
642 	regs_buff[21] = E1000_READ_REG(hw, E1000_TDFPC);
643 
644 	sbuf_printf(sb, "General Registers\n");
645 	sbuf_printf(sb, "\tCTRL\t %08x\n", regs_buff[0]);
646 	sbuf_printf(sb, "\tSTATUS\t %08x\n", regs_buff[1]);
647 	sbuf_printf(sb, "\tCTRL_EXIT\t %08x\n\n", regs_buff[2]);
648 
649 	sbuf_printf(sb, "Interrupt Registers\n");
650 	sbuf_printf(sb, "\tICR\t %08x\n\n", regs_buff[3]);
651 
652 	sbuf_printf(sb, "RX Registers\n");
653 	sbuf_printf(sb, "\tRCTL\t %08x\n", regs_buff[4]);
654 	sbuf_printf(sb, "\tRDLEN\t %08x\n", regs_buff[5]);
655 	sbuf_printf(sb, "\tRDH\t %08x\n", regs_buff[6]);
656 	sbuf_printf(sb, "\tRDT\t %08x\n", regs_buff[7]);
657 	sbuf_printf(sb, "\tRXDCTL\t %08x\n", regs_buff[8]);
658 	sbuf_printf(sb, "\tRDBAL\t %08x\n", regs_buff[9]);
659 	sbuf_printf(sb, "\tRDBAH\t %08x\n\n", regs_buff[10]);
660 
661 	sbuf_printf(sb, "TX Registers\n");
662 	sbuf_printf(sb, "\tTCTL\t %08x\n", regs_buff[11]);
663 	sbuf_printf(sb, "\tTDBAL\t %08x\n", regs_buff[12]);
664 	sbuf_printf(sb, "\tTDBAH\t %08x\n", regs_buff[13]);
665 	sbuf_printf(sb, "\tTDLEN\t %08x\n", regs_buff[14]);
666 	sbuf_printf(sb, "\tTDH\t %08x\n", regs_buff[15]);
667 	sbuf_printf(sb, "\tTDT\t %08x\n", regs_buff[16]);
668 	sbuf_printf(sb, "\tTXDCTL\t %08x\n", regs_buff[17]);
669 	sbuf_printf(sb, "\tTDFH\t %08x\n", regs_buff[18]);
670 	sbuf_printf(sb, "\tTDFT\t %08x\n", regs_buff[19]);
671 	sbuf_printf(sb, "\tTDFHS\t %08x\n", regs_buff[20]);
672 	sbuf_printf(sb, "\tTDFPC\t %08x\n\n", regs_buff[21]);
673 
674 	free(regs_buff, M_DEVBUF);
675 
676 #ifdef DUMP_DESCS
677 	{
678 		if_softc_ctx_t scctx = adapter->shared;
679 		struct rx_ring *rxr = &rx_que->rxr;
680 		struct tx_ring *txr = &tx_que->txr;
681 		int ntxd = scctx->isc_ntxd[0];
682 		int nrxd = scctx->isc_nrxd[0];
683 		int j;
684 
685 	for (j = 0; j < nrxd; j++) {
686 		u32 staterr = le32toh(rxr->rx_base[j].wb.upper.status_error);
687 		u32 length =  le32toh(rxr->rx_base[j].wb.upper.length);
688 		sbuf_printf(sb, "\tReceive Descriptor Address %d: %08" PRIx64 "  Error:%d  Length:%d\n", j, rxr->rx_base[j].read.buffer_addr, staterr, length);
689 	}
690 
691 	for (j = 0; j < min(ntxd, 256); j++) {
692 		unsigned int *ptr = (unsigned int *)&txr->tx_base[j];
693 
694 		sbuf_printf(sb, "\tTXD[%03d] [0]: %08x [1]: %08x [2]: %08x [3]: %08x  eop: %d DD=%d\n",
695 			    j, ptr[0], ptr[1], ptr[2], ptr[3], buf->eop,
696 			    buf->eop != -1 ? txr->tx_base[buf->eop].upper.fields.status & E1000_TXD_STAT_DD : 0);
697 
698 	}
699 	}
700 #endif
701 
702 	rc = sbuf_finish(sb);
703 	sbuf_delete(sb);
704 	return(rc);
705 }
706 
707 static void *
708 em_register(device_t dev)
709 {
710 	return (em_sctx);
711 }
712 
713 static void *
714 igb_register(device_t dev)
715 {
716 	return (igb_sctx);
717 }
718 
719 static int
720 em_set_num_queues(if_ctx_t ctx)
721 {
722 	struct adapter *adapter = iflib_get_softc(ctx);
723 	int maxqueues;
724 
725 	/* Sanity check based on HW */
726 	switch (adapter->hw.mac.type) {
727 	case e1000_82576:
728 	case e1000_82580:
729 	case e1000_i350:
730 	case e1000_i354:
731 		maxqueues = 8;
732 		break;
733 	case e1000_i210:
734 	case e1000_82575:
735 		maxqueues = 4;
736 		break;
737 	case e1000_i211:
738 	case e1000_82574:
739 		maxqueues = 2;
740 		break;
741 	default:
742 		maxqueues = 1;
743 		break;
744 	}
745 
746 	return (maxqueues);
747 }
748 
749 #define	LEM_CAPS							\
750     IFCAP_HWCSUM | IFCAP_VLAN_MTU | IFCAP_VLAN_HWTAGGING |		\
751     IFCAP_VLAN_HWCSUM | IFCAP_WOL | IFCAP_VLAN_HWFILTER
752 
753 #define	EM_CAPS								\
754     IFCAP_HWCSUM | IFCAP_VLAN_MTU | IFCAP_VLAN_HWTAGGING |		\
755     IFCAP_VLAN_HWCSUM | IFCAP_WOL | IFCAP_VLAN_HWFILTER | IFCAP_TSO4 |	\
756     IFCAP_LRO | IFCAP_VLAN_HWTSO
757 
758 #define	IGB_CAPS							\
759     IFCAP_HWCSUM | IFCAP_VLAN_MTU | IFCAP_VLAN_HWTAGGING |		\
760     IFCAP_VLAN_HWCSUM | IFCAP_WOL | IFCAP_VLAN_HWFILTER | IFCAP_TSO4 |	\
761     IFCAP_LRO | IFCAP_VLAN_HWTSO | IFCAP_JUMBO_MTU | IFCAP_HWCSUM_IPV6 |\
762     IFCAP_TSO6
763 
764 /*********************************************************************
765  *  Device initialization routine
766  *
767  *  The attach entry point is called when the driver is being loaded.
768  *  This routine identifies the type of hardware, allocates all resources
769  *  and initializes the hardware.
770  *
771  *  return 0 on success, positive on failure
772  *********************************************************************/
773 static int
774 em_if_attach_pre(if_ctx_t ctx)
775 {
776 	struct adapter *adapter;
777 	if_softc_ctx_t scctx;
778 	device_t dev;
779 	struct e1000_hw *hw;
780 	int error = 0;
781 
782 	INIT_DEBUGOUT("em_if_attach_pre: begin");
783 	dev = iflib_get_dev(ctx);
784 	adapter = iflib_get_softc(ctx);
785 
786 	adapter->ctx = adapter->osdep.ctx = ctx;
787 	adapter->dev = adapter->osdep.dev = dev;
788 	scctx = adapter->shared = iflib_get_softc_ctx(ctx);
789 	adapter->media = iflib_get_media(ctx);
790 	hw = &adapter->hw;
791 
792 	adapter->tx_process_limit = scctx->isc_ntxd[0];
793 
794 	/* SYSCTL stuff */
795 	SYSCTL_ADD_PROC(device_get_sysctl_ctx(dev),
796 	    SYSCTL_CHILDREN(device_get_sysctl_tree(dev)),
797 	    OID_AUTO, "nvm", CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_NEEDGIANT,
798 	    adapter, 0, em_sysctl_nvm_info, "I", "NVM Information");
799 
800 	SYSCTL_ADD_PROC(device_get_sysctl_ctx(dev),
801 	    SYSCTL_CHILDREN(device_get_sysctl_tree(dev)),
802 	    OID_AUTO, "debug", CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_NEEDGIANT,
803 	    adapter, 0, em_sysctl_debug_info, "I", "Debug Information");
804 
805 	SYSCTL_ADD_PROC(device_get_sysctl_ctx(dev),
806 	    SYSCTL_CHILDREN(device_get_sysctl_tree(dev)),
807 	    OID_AUTO, "fc", CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_NEEDGIANT,
808 	    adapter, 0, em_set_flowcntl, "I", "Flow Control");
809 
810 	SYSCTL_ADD_PROC(device_get_sysctl_ctx(dev),
811 	    SYSCTL_CHILDREN(device_get_sysctl_tree(dev)),
812 	    OID_AUTO, "reg_dump",
813 	    CTLTYPE_STRING | CTLFLAG_RD | CTLFLAG_NEEDGIANT, adapter, 0,
814 	    em_get_regs, "A", "Dump Registers");
815 
816 	SYSCTL_ADD_PROC(device_get_sysctl_ctx(dev),
817 	    SYSCTL_CHILDREN(device_get_sysctl_tree(dev)),
818 	    OID_AUTO, "rs_dump",
819 	    CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_NEEDGIANT, adapter, 0,
820 	    em_get_rs, "I", "Dump RS indexes");
821 
822 	/* Determine hardware and mac info */
823 	em_identify_hardware(ctx);
824 
825 	scctx->isc_tx_nsegments = EM_MAX_SCATTER;
826 	scctx->isc_nrxqsets_max = scctx->isc_ntxqsets_max = em_set_num_queues(ctx);
827 	if (bootverbose)
828 		device_printf(dev, "attach_pre capping queues at %d\n",
829 		    scctx->isc_ntxqsets_max);
830 
831 	if (adapter->hw.mac.type >= igb_mac_min) {
832 		scctx->isc_txqsizes[0] = roundup2(scctx->isc_ntxd[0] * sizeof(union e1000_adv_tx_desc), EM_DBA_ALIGN);
833 		scctx->isc_rxqsizes[0] = roundup2(scctx->isc_nrxd[0] * sizeof(union e1000_adv_rx_desc), EM_DBA_ALIGN);
834 		scctx->isc_txd_size[0] = sizeof(union e1000_adv_tx_desc);
835 		scctx->isc_rxd_size[0] = sizeof(union e1000_adv_rx_desc);
836 		scctx->isc_txrx = &igb_txrx;
837 		scctx->isc_tx_tso_segments_max = EM_MAX_SCATTER;
838 		scctx->isc_tx_tso_size_max = EM_TSO_SIZE;
839 		scctx->isc_tx_tso_segsize_max = EM_TSO_SEG_SIZE;
840 		scctx->isc_capabilities = scctx->isc_capenable = IGB_CAPS;
841 		scctx->isc_tx_csum_flags = CSUM_TCP | CSUM_UDP | CSUM_TSO |
842 		     CSUM_IP6_TCP | CSUM_IP6_UDP;
843 		if (adapter->hw.mac.type != e1000_82575)
844 			scctx->isc_tx_csum_flags |= CSUM_SCTP | CSUM_IP6_SCTP;
845 		/*
846 		** Some new devices, as with ixgbe, now may
847 		** use a different BAR, so we need to keep
848 		** track of which is used.
849 		*/
850 		scctx->isc_msix_bar = pci_msix_table_bar(dev);
851 	} else if (adapter->hw.mac.type >= em_mac_min) {
852 		scctx->isc_txqsizes[0] = roundup2(scctx->isc_ntxd[0]* sizeof(struct e1000_tx_desc), EM_DBA_ALIGN);
853 		scctx->isc_rxqsizes[0] = roundup2(scctx->isc_nrxd[0] * sizeof(union e1000_rx_desc_extended), EM_DBA_ALIGN);
854 		scctx->isc_txd_size[0] = sizeof(struct e1000_tx_desc);
855 		scctx->isc_rxd_size[0] = sizeof(union e1000_rx_desc_extended);
856 		scctx->isc_txrx = &em_txrx;
857 		scctx->isc_tx_tso_segments_max = EM_MAX_SCATTER;
858 		scctx->isc_tx_tso_size_max = EM_TSO_SIZE;
859 		scctx->isc_tx_tso_segsize_max = EM_TSO_SEG_SIZE;
860 		scctx->isc_capabilities = scctx->isc_capenable = EM_CAPS;
861 		/*
862 		 * For EM-class devices, don't enable IFCAP_{TSO4,VLAN_HWTSO}
863 		 * by default as we don't have workarounds for all associated
864 		 * silicon errata.  E. g., with several MACs such as 82573E,
865 		 * TSO only works at Gigabit speed and otherwise can cause the
866 		 * hardware to hang (which also would be next to impossible to
867 		 * work around given that already queued TSO-using descriptors
868 		 * would need to be flushed and vlan(4) reconfigured at runtime
869 		 * in case of a link speed change).  Moreover, MACs like 82579
870 		 * still can hang at Gigabit even with all publicly documented
871 		 * TSO workarounds implemented.  Generally, the penality of
872 		 * these workarounds is rather high and may involve copying
873 		 * mbuf data around so advantages of TSO lapse.  Still, TSO may
874 		 * work for a few MACs of this class - at least when sticking
875 		 * with Gigabit - in which case users may enable TSO manually.
876 		 */
877 		scctx->isc_capenable &= ~(IFCAP_TSO4 | IFCAP_VLAN_HWTSO);
878 		scctx->isc_tx_csum_flags = CSUM_TCP | CSUM_UDP | CSUM_IP_TSO;
879 		/*
880 		 * We support MSI-X with 82574 only, but indicate to iflib(4)
881 		 * that it shall give MSI at least a try with other devices.
882 		 */
883 		if (adapter->hw.mac.type == e1000_82574) {
884 			scctx->isc_msix_bar = pci_msix_table_bar(dev);;
885 		} else {
886 			scctx->isc_msix_bar = -1;
887 			scctx->isc_disable_msix = 1;
888 		}
889 	} else {
890 		scctx->isc_txqsizes[0] = roundup2((scctx->isc_ntxd[0] + 1) * sizeof(struct e1000_tx_desc), EM_DBA_ALIGN);
891 		scctx->isc_rxqsizes[0] = roundup2((scctx->isc_nrxd[0] + 1) * sizeof(struct e1000_rx_desc), EM_DBA_ALIGN);
892 		scctx->isc_txd_size[0] = sizeof(struct e1000_tx_desc);
893 		scctx->isc_rxd_size[0] = sizeof(struct e1000_rx_desc);
894 		scctx->isc_tx_csum_flags = CSUM_TCP | CSUM_UDP;
895 		scctx->isc_txrx = &lem_txrx;
896 		scctx->isc_capabilities = scctx->isc_capenable = LEM_CAPS;
897 		if (adapter->hw.mac.type < e1000_82543)
898 			scctx->isc_capenable &= ~(IFCAP_HWCSUM|IFCAP_VLAN_HWCSUM);
899 		/* INTx only */
900 		scctx->isc_msix_bar = 0;
901 	}
902 
903 	/* Setup PCI resources */
904 	if (em_allocate_pci_resources(ctx)) {
905 		device_printf(dev, "Allocation of PCI resources failed\n");
906 		error = ENXIO;
907 		goto err_pci;
908 	}
909 
910 	/*
911 	** For ICH8 and family we need to
912 	** map the flash memory, and this
913 	** must happen after the MAC is
914 	** identified
915 	*/
916 	if ((hw->mac.type == e1000_ich8lan) ||
917 	    (hw->mac.type == e1000_ich9lan) ||
918 	    (hw->mac.type == e1000_ich10lan) ||
919 	    (hw->mac.type == e1000_pchlan) ||
920 	    (hw->mac.type == e1000_pch2lan) ||
921 	    (hw->mac.type == e1000_pch_lpt)) {
922 		int rid = EM_BAR_TYPE_FLASH;
923 		adapter->flash = bus_alloc_resource_any(dev,
924 		    SYS_RES_MEMORY, &rid, RF_ACTIVE);
925 		if (adapter->flash == NULL) {
926 			device_printf(dev, "Mapping of Flash failed\n");
927 			error = ENXIO;
928 			goto err_pci;
929 		}
930 		/* This is used in the shared code */
931 		hw->flash_address = (u8 *)adapter->flash;
932 		adapter->osdep.flash_bus_space_tag =
933 		    rman_get_bustag(adapter->flash);
934 		adapter->osdep.flash_bus_space_handle =
935 		    rman_get_bushandle(adapter->flash);
936 	}
937 	/*
938 	** In the new SPT device flash is not  a
939 	** separate BAR, rather it is also in BAR0,
940 	** so use the same tag and an offset handle for the
941 	** FLASH read/write macros in the shared code.
942 	*/
943 	else if (hw->mac.type >= e1000_pch_spt) {
944 		adapter->osdep.flash_bus_space_tag =
945 		    adapter->osdep.mem_bus_space_tag;
946 		adapter->osdep.flash_bus_space_handle =
947 		    adapter->osdep.mem_bus_space_handle
948 		    + E1000_FLASH_BASE_ADDR;
949 	}
950 
951 	/* Do Shared Code initialization */
952 	error = e1000_setup_init_funcs(hw, TRUE);
953 	if (error) {
954 		device_printf(dev, "Setup of Shared code failed, error %d\n",
955 		    error);
956 		error = ENXIO;
957 		goto err_pci;
958 	}
959 
960 	em_setup_msix(ctx);
961 	e1000_get_bus_info(hw);
962 
963 	/* Set up some sysctls for the tunable interrupt delays */
964 	em_add_int_delay_sysctl(adapter, "rx_int_delay",
965 	    "receive interrupt delay in usecs", &adapter->rx_int_delay,
966 	    E1000_REGISTER(hw, E1000_RDTR), em_rx_int_delay_dflt);
967 	em_add_int_delay_sysctl(adapter, "tx_int_delay",
968 	    "transmit interrupt delay in usecs", &adapter->tx_int_delay,
969 	    E1000_REGISTER(hw, E1000_TIDV), em_tx_int_delay_dflt);
970 	em_add_int_delay_sysctl(adapter, "rx_abs_int_delay",
971 	    "receive interrupt delay limit in usecs",
972 	    &adapter->rx_abs_int_delay,
973 	    E1000_REGISTER(hw, E1000_RADV),
974 	    em_rx_abs_int_delay_dflt);
975 	em_add_int_delay_sysctl(adapter, "tx_abs_int_delay",
976 	    "transmit interrupt delay limit in usecs",
977 	    &adapter->tx_abs_int_delay,
978 	    E1000_REGISTER(hw, E1000_TADV),
979 	    em_tx_abs_int_delay_dflt);
980 	em_add_int_delay_sysctl(adapter, "itr",
981 	    "interrupt delay limit in usecs/4",
982 	    &adapter->tx_itr,
983 	    E1000_REGISTER(hw, E1000_ITR),
984 	    DEFAULT_ITR);
985 
986 	hw->mac.autoneg = DO_AUTO_NEG;
987 	hw->phy.autoneg_wait_to_complete = FALSE;
988 	hw->phy.autoneg_advertised = AUTONEG_ADV_DEFAULT;
989 
990 	if (adapter->hw.mac.type < em_mac_min) {
991 		e1000_init_script_state_82541(&adapter->hw, TRUE);
992 		e1000_set_tbi_compatibility_82543(&adapter->hw, TRUE);
993 	}
994 	/* Copper options */
995 	if (hw->phy.media_type == e1000_media_type_copper) {
996 		hw->phy.mdix = AUTO_ALL_MODES;
997 		hw->phy.disable_polarity_correction = FALSE;
998 		hw->phy.ms_type = EM_MASTER_SLAVE;
999 	}
1000 
1001 	/*
1002 	 * Set the frame limits assuming
1003 	 * standard ethernet sized frames.
1004 	 */
1005 	scctx->isc_max_frame_size = adapter->hw.mac.max_frame_size =
1006 	    ETHERMTU + ETHER_HDR_LEN + ETHERNET_FCS_SIZE;
1007 
1008 	/*
1009 	 * This controls when hardware reports transmit completion
1010 	 * status.
1011 	 */
1012 	hw->mac.report_tx_early = 1;
1013 
1014 	/* Allocate multicast array memory. */
1015 	adapter->mta = malloc(sizeof(u8) * ETHER_ADDR_LEN *
1016 	    MAX_NUM_MULTICAST_ADDRESSES, M_DEVBUF, M_NOWAIT);
1017 	if (adapter->mta == NULL) {
1018 		device_printf(dev, "Can not allocate multicast setup array\n");
1019 		error = ENOMEM;
1020 		goto err_late;
1021 	}
1022 
1023 	/* Check SOL/IDER usage */
1024 	if (e1000_check_reset_block(hw))
1025 		device_printf(dev, "PHY reset is blocked"
1026 			      " due to SOL/IDER session.\n");
1027 
1028 	/* Sysctl for setting Energy Efficient Ethernet */
1029 	hw->dev_spec.ich8lan.eee_disable = eee_setting;
1030 	SYSCTL_ADD_PROC(device_get_sysctl_ctx(dev),
1031 	    SYSCTL_CHILDREN(device_get_sysctl_tree(dev)),
1032 	    OID_AUTO, "eee_control",
1033 	    CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_NEEDGIANT,
1034 	    adapter, 0, em_sysctl_eee, "I",
1035 	    "Disable Energy Efficient Ethernet");
1036 
1037 	/*
1038 	** Start from a known state, this is
1039 	** important in reading the nvm and
1040 	** mac from that.
1041 	*/
1042 	e1000_reset_hw(hw);
1043 
1044 	/* Make sure we have a good EEPROM before we read from it */
1045 	if (e1000_validate_nvm_checksum(hw) < 0) {
1046 		/*
1047 		** Some PCI-E parts fail the first check due to
1048 		** the link being in sleep state, call it again,
1049 		** if it fails a second time its a real issue.
1050 		*/
1051 		if (e1000_validate_nvm_checksum(hw) < 0) {
1052 			device_printf(dev,
1053 			    "The EEPROM Checksum Is Not Valid\n");
1054 			error = EIO;
1055 			goto err_late;
1056 		}
1057 	}
1058 
1059 	/* Copy the permanent MAC address out of the EEPROM */
1060 	if (e1000_read_mac_addr(hw) < 0) {
1061 		device_printf(dev, "EEPROM read error while reading MAC"
1062 			      " address\n");
1063 		error = EIO;
1064 		goto err_late;
1065 	}
1066 
1067 	if (!em_is_valid_ether_addr(hw->mac.addr)) {
1068 		device_printf(dev, "Invalid MAC address\n");
1069 		error = EIO;
1070 		goto err_late;
1071 	}
1072 
1073 	/* Disable ULP support */
1074 	e1000_disable_ulp_lpt_lp(hw, TRUE);
1075 
1076 	/*
1077 	 * Get Wake-on-Lan and Management info for later use
1078 	 */
1079 	em_get_wakeup(ctx);
1080 
1081 	/* Enable only WOL MAGIC by default */
1082 	scctx->isc_capenable &= ~IFCAP_WOL;
1083 	if (adapter->wol != 0)
1084 		scctx->isc_capenable |= IFCAP_WOL_MAGIC;
1085 
1086 	iflib_set_mac(ctx, hw->mac.addr);
1087 
1088 	return (0);
1089 
1090 err_late:
1091 	em_release_hw_control(adapter);
1092 err_pci:
1093 	em_free_pci_resources(ctx);
1094 	free(adapter->mta, M_DEVBUF);
1095 
1096 	return (error);
1097 }
1098 
1099 static int
1100 em_if_attach_post(if_ctx_t ctx)
1101 {
1102 	struct adapter *adapter = iflib_get_softc(ctx);
1103 	struct e1000_hw *hw = &adapter->hw;
1104 	int error = 0;
1105 
1106 	/* Setup OS specific network interface */
1107 	error = em_setup_interface(ctx);
1108 	if (error != 0) {
1109 		goto err_late;
1110 	}
1111 
1112 	em_reset(ctx);
1113 
1114 	/* Initialize statistics */
1115 	em_update_stats_counters(adapter);
1116 	hw->mac.get_link_status = 1;
1117 	em_if_update_admin_status(ctx);
1118 	em_add_hw_stats(adapter);
1119 
1120 	/* Non-AMT based hardware can now take control from firmware */
1121 	if (adapter->has_manage && !adapter->has_amt)
1122 		em_get_hw_control(adapter);
1123 
1124 	INIT_DEBUGOUT("em_if_attach_post: end");
1125 
1126 	return (error);
1127 
1128 err_late:
1129 	em_release_hw_control(adapter);
1130 	em_free_pci_resources(ctx);
1131 	em_if_queues_free(ctx);
1132 	free(adapter->mta, M_DEVBUF);
1133 
1134 	return (error);
1135 }
1136 
1137 /*********************************************************************
1138  *  Device removal routine
1139  *
1140  *  The detach entry point is called when the driver is being removed.
1141  *  This routine stops the adapter and deallocates all the resources
1142  *  that were allocated for driver operation.
1143  *
1144  *  return 0 on success, positive on failure
1145  *********************************************************************/
1146 static int
1147 em_if_detach(if_ctx_t ctx)
1148 {
1149 	struct adapter	*adapter = iflib_get_softc(ctx);
1150 
1151 	INIT_DEBUGOUT("em_if_detach: begin");
1152 
1153 	e1000_phy_hw_reset(&adapter->hw);
1154 
1155 	em_release_manageability(adapter);
1156 	em_release_hw_control(adapter);
1157 	em_free_pci_resources(ctx);
1158 
1159 	return (0);
1160 }
1161 
1162 /*********************************************************************
1163  *
1164  *  Shutdown entry point
1165  *
1166  **********************************************************************/
1167 
1168 static int
1169 em_if_shutdown(if_ctx_t ctx)
1170 {
1171 	return em_if_suspend(ctx);
1172 }
1173 
1174 /*
1175  * Suspend/resume device methods.
1176  */
1177 static int
1178 em_if_suspend(if_ctx_t ctx)
1179 {
1180 	struct adapter *adapter = iflib_get_softc(ctx);
1181 
1182 	em_release_manageability(adapter);
1183 	em_release_hw_control(adapter);
1184 	em_enable_wakeup(ctx);
1185 	return (0);
1186 }
1187 
1188 static int
1189 em_if_resume(if_ctx_t ctx)
1190 {
1191 	struct adapter *adapter = iflib_get_softc(ctx);
1192 
1193 	if (adapter->hw.mac.type == e1000_pch2lan)
1194 		e1000_resume_workarounds_pchlan(&adapter->hw);
1195 	em_if_init(ctx);
1196 	em_init_manageability(adapter);
1197 
1198 	return(0);
1199 }
1200 
1201 static int
1202 em_if_mtu_set(if_ctx_t ctx, uint32_t mtu)
1203 {
1204 	int max_frame_size;
1205 	struct adapter *adapter = iflib_get_softc(ctx);
1206 	if_softc_ctx_t scctx = iflib_get_softc_ctx(ctx);
1207 
1208 	 IOCTL_DEBUGOUT("ioctl rcv'd: SIOCSIFMTU (Set Interface MTU)");
1209 
1210 	switch (adapter->hw.mac.type) {
1211 	case e1000_82571:
1212 	case e1000_82572:
1213 	case e1000_ich9lan:
1214 	case e1000_ich10lan:
1215 	case e1000_pch2lan:
1216 	case e1000_pch_lpt:
1217 	case e1000_pch_spt:
1218 	case e1000_pch_cnp:
1219 	case e1000_82574:
1220 	case e1000_82583:
1221 	case e1000_80003es2lan:
1222 		/* 9K Jumbo Frame size */
1223 		max_frame_size = 9234;
1224 		break;
1225 	case e1000_pchlan:
1226 		max_frame_size = 4096;
1227 		break;
1228 	case e1000_82542:
1229 	case e1000_ich8lan:
1230 		/* Adapters that do not support jumbo frames */
1231 		max_frame_size = ETHER_MAX_LEN;
1232 		break;
1233 	default:
1234 		if (adapter->hw.mac.type >= igb_mac_min)
1235 			max_frame_size = 9234;
1236 		else /* lem */
1237 			max_frame_size = MAX_JUMBO_FRAME_SIZE;
1238 	}
1239 	if (mtu > max_frame_size - ETHER_HDR_LEN - ETHER_CRC_LEN) {
1240 		return (EINVAL);
1241 	}
1242 
1243 	scctx->isc_max_frame_size = adapter->hw.mac.max_frame_size =
1244 	    mtu + ETHER_HDR_LEN + ETHER_CRC_LEN;
1245 	return (0);
1246 }
1247 
1248 /*********************************************************************
1249  *  Init entry point
1250  *
1251  *  This routine is used in two ways. It is used by the stack as
1252  *  init entry point in network interface structure. It is also used
1253  *  by the driver as a hw/sw initialization routine to get to a
1254  *  consistent state.
1255  *
1256  **********************************************************************/
1257 static void
1258 em_if_init(if_ctx_t ctx)
1259 {
1260 	struct adapter *adapter = iflib_get_softc(ctx);
1261 	if_softc_ctx_t scctx = adapter->shared;
1262 	struct ifnet *ifp = iflib_get_ifp(ctx);
1263 	struct em_tx_queue *tx_que;
1264 	int i;
1265 
1266 	INIT_DEBUGOUT("em_if_init: begin");
1267 
1268 	/* Get the latest mac address, User can use a LAA */
1269 	bcopy(if_getlladdr(ifp), adapter->hw.mac.addr,
1270 	    ETHER_ADDR_LEN);
1271 
1272 	/* Put the address into the Receive Address Array */
1273 	e1000_rar_set(&adapter->hw, adapter->hw.mac.addr, 0);
1274 
1275 	/*
1276 	 * With the 82571 adapter, RAR[0] may be overwritten
1277 	 * when the other port is reset, we make a duplicate
1278 	 * in RAR[14] for that eventuality, this assures
1279 	 * the interface continues to function.
1280 	 */
1281 	if (adapter->hw.mac.type == e1000_82571) {
1282 		e1000_set_laa_state_82571(&adapter->hw, TRUE);
1283 		e1000_rar_set(&adapter->hw, adapter->hw.mac.addr,
1284 		    E1000_RAR_ENTRIES - 1);
1285 	}
1286 
1287 
1288 	/* Initialize the hardware */
1289 	em_reset(ctx);
1290 	em_if_update_admin_status(ctx);
1291 
1292 	for (i = 0, tx_que = adapter->tx_queues; i < adapter->tx_num_queues; i++, tx_que++) {
1293 		struct tx_ring *txr = &tx_que->txr;
1294 
1295 		txr->tx_rs_cidx = txr->tx_rs_pidx;
1296 
1297 		/* Initialize the last processed descriptor to be the end of
1298 		 * the ring, rather than the start, so that we avoid an
1299 		 * off-by-one error when calculating how many descriptors are
1300 		 * done in the credits_update function.
1301 		 */
1302 		txr->tx_cidx_processed = scctx->isc_ntxd[0] - 1;
1303 	}
1304 
1305 	/* Setup VLAN support, basic and offload if available */
1306 	E1000_WRITE_REG(&adapter->hw, E1000_VET, ETHERTYPE_VLAN);
1307 
1308 	/* Clear bad data from Rx FIFOs */
1309 	if (adapter->hw.mac.type >= igb_mac_min)
1310 		e1000_rx_fifo_flush_82575(&adapter->hw);
1311 
1312 	/* Configure for OS presence */
1313 	em_init_manageability(adapter);
1314 
1315 	/* Prepare transmit descriptors and buffers */
1316 	em_initialize_transmit_unit(ctx);
1317 
1318 	/* Setup Multicast table */
1319 	em_if_multi_set(ctx);
1320 
1321 	adapter->rx_mbuf_sz = iflib_get_rx_mbuf_sz(ctx);
1322 	em_initialize_receive_unit(ctx);
1323 
1324 	/* Use real VLAN Filter support? */
1325 	if (if_getcapenable(ifp) & IFCAP_VLAN_HWTAGGING) {
1326 		if (if_getcapenable(ifp) & IFCAP_VLAN_HWFILTER)
1327 			/* Use real VLAN Filter support */
1328 			em_setup_vlan_hw_support(adapter);
1329 		else {
1330 			u32 ctrl;
1331 			ctrl = E1000_READ_REG(&adapter->hw, E1000_CTRL);
1332 			ctrl |= E1000_CTRL_VME;
1333 			E1000_WRITE_REG(&adapter->hw, E1000_CTRL, ctrl);
1334 		}
1335 	} else {
1336 		u32 ctrl;
1337 		ctrl = E1000_READ_REG(&adapter->hw, E1000_CTRL);
1338 		ctrl &= ~E1000_CTRL_VME;
1339 		E1000_WRITE_REG(&adapter->hw, E1000_CTRL, ctrl);
1340 	}
1341 
1342 	/* Don't lose promiscuous settings */
1343 	em_if_set_promisc(ctx, if_getflags(ifp));
1344 	e1000_clear_hw_cntrs_base_generic(&adapter->hw);
1345 
1346 	/* MSI-X configuration for 82574 */
1347 	if (adapter->hw.mac.type == e1000_82574) {
1348 		int tmp = E1000_READ_REG(&adapter->hw, E1000_CTRL_EXT);
1349 
1350 		tmp |= E1000_CTRL_EXT_PBA_CLR;
1351 		E1000_WRITE_REG(&adapter->hw, E1000_CTRL_EXT, tmp);
1352 		/* Set the IVAR - interrupt vector routing. */
1353 		E1000_WRITE_REG(&adapter->hw, E1000_IVAR, adapter->ivars);
1354 	} else if (adapter->intr_type == IFLIB_INTR_MSIX) /* Set up queue routing */
1355 		igb_configure_queues(adapter);
1356 
1357 	/* this clears any pending interrupts */
1358 	E1000_READ_REG(&adapter->hw, E1000_ICR);
1359 	E1000_WRITE_REG(&adapter->hw, E1000_ICS, E1000_ICS_LSC);
1360 
1361 	/* AMT based hardware can now take control from firmware */
1362 	if (adapter->has_manage && adapter->has_amt)
1363 		em_get_hw_control(adapter);
1364 
1365 	/* Set Energy Efficient Ethernet */
1366 	if (adapter->hw.mac.type >= igb_mac_min &&
1367 	    adapter->hw.phy.media_type == e1000_media_type_copper) {
1368 		if (adapter->hw.mac.type == e1000_i354)
1369 			e1000_set_eee_i354(&adapter->hw, TRUE, TRUE);
1370 		else
1371 			e1000_set_eee_i350(&adapter->hw, TRUE, TRUE);
1372 	}
1373 }
1374 
1375 /*********************************************************************
1376  *
1377  *  Fast Legacy/MSI Combined Interrupt Service routine
1378  *
1379  *********************************************************************/
1380 int
1381 em_intr(void *arg)
1382 {
1383 	struct adapter *adapter = arg;
1384 	if_ctx_t ctx = adapter->ctx;
1385 	u32 reg_icr;
1386 
1387 	reg_icr = E1000_READ_REG(&adapter->hw, E1000_ICR);
1388 
1389 	/* Hot eject? */
1390 	if (reg_icr == 0xffffffff)
1391 		return FILTER_STRAY;
1392 
1393 	/* Definitely not our interrupt. */
1394 	if (reg_icr == 0x0)
1395 		return FILTER_STRAY;
1396 
1397 	/*
1398 	 * Starting with the 82571 chip, bit 31 should be used to
1399 	 * determine whether the interrupt belongs to us.
1400 	 */
1401 	if (adapter->hw.mac.type >= e1000_82571 &&
1402 	    (reg_icr & E1000_ICR_INT_ASSERTED) == 0)
1403 		return FILTER_STRAY;
1404 
1405 	/*
1406 	 * Only MSI-X interrupts have one-shot behavior by taking advantage
1407 	 * of the EIAC register.  Thus, explicitly disable interrupts.  This
1408 	 * also works around the MSI message reordering errata on certain
1409 	 * systems.
1410 	 */
1411 	IFDI_INTR_DISABLE(ctx);
1412 
1413 	/* Link status change */
1414 	if (reg_icr & (E1000_ICR_RXSEQ | E1000_ICR_LSC))
1415 		em_handle_link(ctx);
1416 
1417 	if (reg_icr & E1000_ICR_RXO)
1418 		adapter->rx_overruns++;
1419 
1420 	return (FILTER_SCHEDULE_THREAD);
1421 }
1422 
1423 static int
1424 em_if_rx_queue_intr_enable(if_ctx_t ctx, uint16_t rxqid)
1425 {
1426 	struct adapter *adapter = iflib_get_softc(ctx);
1427 	struct em_rx_queue *rxq = &adapter->rx_queues[rxqid];
1428 
1429 	E1000_WRITE_REG(&adapter->hw, E1000_IMS, rxq->eims);
1430 	return (0);
1431 }
1432 
1433 static int
1434 em_if_tx_queue_intr_enable(if_ctx_t ctx, uint16_t txqid)
1435 {
1436 	struct adapter *adapter = iflib_get_softc(ctx);
1437 	struct em_tx_queue *txq = &adapter->tx_queues[txqid];
1438 
1439 	E1000_WRITE_REG(&adapter->hw, E1000_IMS, txq->eims);
1440 	return (0);
1441 }
1442 
1443 static int
1444 igb_if_rx_queue_intr_enable(if_ctx_t ctx, uint16_t rxqid)
1445 {
1446 	struct adapter *adapter = iflib_get_softc(ctx);
1447 	struct em_rx_queue *rxq = &adapter->rx_queues[rxqid];
1448 
1449 	E1000_WRITE_REG(&adapter->hw, E1000_EIMS, rxq->eims);
1450 	return (0);
1451 }
1452 
1453 static int
1454 igb_if_tx_queue_intr_enable(if_ctx_t ctx, uint16_t txqid)
1455 {
1456 	struct adapter *adapter = iflib_get_softc(ctx);
1457 	struct em_tx_queue *txq = &adapter->tx_queues[txqid];
1458 
1459 	E1000_WRITE_REG(&adapter->hw, E1000_EIMS, txq->eims);
1460 	return (0);
1461 }
1462 
1463 /*********************************************************************
1464  *
1465  *  MSI-X RX Interrupt Service routine
1466  *
1467  **********************************************************************/
1468 static int
1469 em_msix_que(void *arg)
1470 {
1471 	struct em_rx_queue *que = arg;
1472 
1473 	++que->irqs;
1474 
1475 	return (FILTER_SCHEDULE_THREAD);
1476 }
1477 
1478 /*********************************************************************
1479  *
1480  *  MSI-X Link Fast Interrupt Service routine
1481  *
1482  **********************************************************************/
1483 static int
1484 em_msix_link(void *arg)
1485 {
1486 	struct adapter *adapter = arg;
1487 	u32 reg_icr;
1488 
1489 	++adapter->link_irq;
1490 	MPASS(adapter->hw.back != NULL);
1491 	reg_icr = E1000_READ_REG(&adapter->hw, E1000_ICR);
1492 
1493 	if (reg_icr & E1000_ICR_RXO)
1494 		adapter->rx_overruns++;
1495 
1496 	if (reg_icr & (E1000_ICR_RXSEQ | E1000_ICR_LSC)) {
1497 		em_handle_link(adapter->ctx);
1498 	} else if (adapter->hw.mac.type == e1000_82574) {
1499 		/* Only re-arm 82574 if em_if_update_admin_status() won't. */
1500 		E1000_WRITE_REG(&adapter->hw, E1000_IMS, EM_MSIX_LINK |
1501 		    E1000_IMS_LSC);
1502 	}
1503 
1504 	if (adapter->hw.mac.type == e1000_82574) {
1505 		/*
1506 		 * Because we must read the ICR for this interrupt it may
1507 		 * clear other causes using autoclear, for this reason we
1508 		 * simply create a soft interrupt for all these vectors.
1509 		 */
1510 		if (reg_icr)
1511 			E1000_WRITE_REG(&adapter->hw, E1000_ICS, adapter->ims);
1512 	} else {
1513 		/* Re-arm unconditionally */
1514 		E1000_WRITE_REG(&adapter->hw, E1000_IMS, E1000_IMS_LSC);
1515 		E1000_WRITE_REG(&adapter->hw, E1000_EIMS, adapter->link_mask);
1516 	}
1517 
1518 	return (FILTER_HANDLED);
1519 }
1520 
1521 static void
1522 em_handle_link(void *context)
1523 {
1524 	if_ctx_t ctx = context;
1525 	struct adapter *adapter = iflib_get_softc(ctx);
1526 
1527 	adapter->hw.mac.get_link_status = 1;
1528 	iflib_admin_intr_deferred(ctx);
1529 }
1530 
1531 /*********************************************************************
1532  *
1533  *  Media Ioctl callback
1534  *
1535  *  This routine is called whenever the user queries the status of
1536  *  the interface using ifconfig.
1537  *
1538  **********************************************************************/
1539 static void
1540 em_if_media_status(if_ctx_t ctx, struct ifmediareq *ifmr)
1541 {
1542 	struct adapter *adapter = iflib_get_softc(ctx);
1543 	u_char fiber_type = IFM_1000_SX;
1544 
1545 	INIT_DEBUGOUT("em_if_media_status: begin");
1546 
1547 	iflib_admin_intr_deferred(ctx);
1548 
1549 	ifmr->ifm_status = IFM_AVALID;
1550 	ifmr->ifm_active = IFM_ETHER;
1551 
1552 	if (!adapter->link_active) {
1553 		return;
1554 	}
1555 
1556 	ifmr->ifm_status |= IFM_ACTIVE;
1557 
1558 	if ((adapter->hw.phy.media_type == e1000_media_type_fiber) ||
1559 	    (adapter->hw.phy.media_type == e1000_media_type_internal_serdes)) {
1560 		if (adapter->hw.mac.type == e1000_82545)
1561 			fiber_type = IFM_1000_LX;
1562 		ifmr->ifm_active |= fiber_type | IFM_FDX;
1563 	} else {
1564 		switch (adapter->link_speed) {
1565 		case 10:
1566 			ifmr->ifm_active |= IFM_10_T;
1567 			break;
1568 		case 100:
1569 			ifmr->ifm_active |= IFM_100_TX;
1570 			break;
1571 		case 1000:
1572 			ifmr->ifm_active |= IFM_1000_T;
1573 			break;
1574 		}
1575 		if (adapter->link_duplex == FULL_DUPLEX)
1576 			ifmr->ifm_active |= IFM_FDX;
1577 		else
1578 			ifmr->ifm_active |= IFM_HDX;
1579 	}
1580 }
1581 
1582 /*********************************************************************
1583  *
1584  *  Media Ioctl callback
1585  *
1586  *  This routine is called when the user changes speed/duplex using
1587  *  media/mediopt option with ifconfig.
1588  *
1589  **********************************************************************/
1590 static int
1591 em_if_media_change(if_ctx_t ctx)
1592 {
1593 	struct adapter *adapter = iflib_get_softc(ctx);
1594 	struct ifmedia *ifm = iflib_get_media(ctx);
1595 
1596 	INIT_DEBUGOUT("em_if_media_change: begin");
1597 
1598 	if (IFM_TYPE(ifm->ifm_media) != IFM_ETHER)
1599 		return (EINVAL);
1600 
1601 	switch (IFM_SUBTYPE(ifm->ifm_media)) {
1602 	case IFM_AUTO:
1603 		adapter->hw.mac.autoneg = DO_AUTO_NEG;
1604 		adapter->hw.phy.autoneg_advertised = AUTONEG_ADV_DEFAULT;
1605 		break;
1606 	case IFM_1000_LX:
1607 	case IFM_1000_SX:
1608 	case IFM_1000_T:
1609 		adapter->hw.mac.autoneg = DO_AUTO_NEG;
1610 		adapter->hw.phy.autoneg_advertised = ADVERTISE_1000_FULL;
1611 		break;
1612 	case IFM_100_TX:
1613 		adapter->hw.mac.autoneg = FALSE;
1614 		adapter->hw.phy.autoneg_advertised = 0;
1615 		if ((ifm->ifm_media & IFM_GMASK) == IFM_FDX)
1616 			adapter->hw.mac.forced_speed_duplex = ADVERTISE_100_FULL;
1617 		else
1618 			adapter->hw.mac.forced_speed_duplex = ADVERTISE_100_HALF;
1619 		break;
1620 	case IFM_10_T:
1621 		adapter->hw.mac.autoneg = FALSE;
1622 		adapter->hw.phy.autoneg_advertised = 0;
1623 		if ((ifm->ifm_media & IFM_GMASK) == IFM_FDX)
1624 			adapter->hw.mac.forced_speed_duplex = ADVERTISE_10_FULL;
1625 		else
1626 			adapter->hw.mac.forced_speed_duplex = ADVERTISE_10_HALF;
1627 		break;
1628 	default:
1629 		device_printf(adapter->dev, "Unsupported media type\n");
1630 	}
1631 
1632 	em_if_init(ctx);
1633 
1634 	return (0);
1635 }
1636 
1637 static int
1638 em_if_set_promisc(if_ctx_t ctx, int flags)
1639 {
1640 	struct adapter *adapter = iflib_get_softc(ctx);
1641 	u32 reg_rctl;
1642 
1643 	em_disable_promisc(ctx);
1644 
1645 	reg_rctl = E1000_READ_REG(&adapter->hw, E1000_RCTL);
1646 
1647 	if (flags & IFF_PROMISC) {
1648 		reg_rctl |= (E1000_RCTL_UPE | E1000_RCTL_MPE);
1649 		/* Turn this on if you want to see bad packets */
1650 		if (em_debug_sbp)
1651 			reg_rctl |= E1000_RCTL_SBP;
1652 		E1000_WRITE_REG(&adapter->hw, E1000_RCTL, reg_rctl);
1653 	} else if (flags & IFF_ALLMULTI) {
1654 		reg_rctl |= E1000_RCTL_MPE;
1655 		reg_rctl &= ~E1000_RCTL_UPE;
1656 		E1000_WRITE_REG(&adapter->hw, E1000_RCTL, reg_rctl);
1657 	}
1658 	return (0);
1659 }
1660 
1661 static void
1662 em_disable_promisc(if_ctx_t ctx)
1663 {
1664 	struct adapter *adapter = iflib_get_softc(ctx);
1665 	struct ifnet *ifp = iflib_get_ifp(ctx);
1666 	u32 reg_rctl;
1667 	int mcnt = 0;
1668 
1669 	reg_rctl = E1000_READ_REG(&adapter->hw, E1000_RCTL);
1670 	reg_rctl &= (~E1000_RCTL_UPE);
1671 	if (if_getflags(ifp) & IFF_ALLMULTI)
1672 		mcnt = MAX_NUM_MULTICAST_ADDRESSES;
1673 	else
1674 		mcnt = if_llmaddr_count(ifp);
1675 	/* Don't disable if in MAX groups */
1676 	if (mcnt < MAX_NUM_MULTICAST_ADDRESSES)
1677 		reg_rctl &=  (~E1000_RCTL_MPE);
1678 	reg_rctl &=  (~E1000_RCTL_SBP);
1679 	E1000_WRITE_REG(&adapter->hw, E1000_RCTL, reg_rctl);
1680 }
1681 
1682 
1683 static u_int
1684 em_copy_maddr(void *arg, struct sockaddr_dl *sdl, u_int cnt)
1685 {
1686 	u8 *mta = arg;
1687 
1688 	if (cnt == MAX_NUM_MULTICAST_ADDRESSES)
1689 		return (1);
1690 
1691 	bcopy(LLADDR(sdl), &mta[cnt * ETHER_ADDR_LEN], ETHER_ADDR_LEN);
1692 
1693 	return (1);
1694 }
1695 
1696 /*********************************************************************
1697  *  Multicast Update
1698  *
1699  *  This routine is called whenever multicast address list is updated.
1700  *
1701  **********************************************************************/
1702 
1703 static void
1704 em_if_multi_set(if_ctx_t ctx)
1705 {
1706 	struct adapter *adapter = iflib_get_softc(ctx);
1707 	struct ifnet *ifp = iflib_get_ifp(ctx);
1708 	u32 reg_rctl = 0;
1709 	u8  *mta; /* Multicast array memory */
1710 	int mcnt = 0;
1711 
1712 	IOCTL_DEBUGOUT("em_set_multi: begin");
1713 
1714 	mta = adapter->mta;
1715 	bzero(mta, sizeof(u8) * ETHER_ADDR_LEN * MAX_NUM_MULTICAST_ADDRESSES);
1716 
1717 	if (adapter->hw.mac.type == e1000_82542 &&
1718 	    adapter->hw.revision_id == E1000_REVISION_2) {
1719 		reg_rctl = E1000_READ_REG(&adapter->hw, E1000_RCTL);
1720 		if (adapter->hw.bus.pci_cmd_word & CMD_MEM_WRT_INVALIDATE)
1721 			e1000_pci_clear_mwi(&adapter->hw);
1722 		reg_rctl |= E1000_RCTL_RST;
1723 		E1000_WRITE_REG(&adapter->hw, E1000_RCTL, reg_rctl);
1724 		msec_delay(5);
1725 	}
1726 
1727 	mcnt = if_foreach_llmaddr(ifp, em_copy_maddr, mta);
1728 
1729 	if (mcnt >= MAX_NUM_MULTICAST_ADDRESSES) {
1730 		reg_rctl = E1000_READ_REG(&adapter->hw, E1000_RCTL);
1731 		reg_rctl |= E1000_RCTL_MPE;
1732 		E1000_WRITE_REG(&adapter->hw, E1000_RCTL, reg_rctl);
1733 	} else
1734 		e1000_update_mc_addr_list(&adapter->hw, mta, mcnt);
1735 
1736 	if (adapter->hw.mac.type == e1000_82542 &&
1737 	    adapter->hw.revision_id == E1000_REVISION_2) {
1738 		reg_rctl = E1000_READ_REG(&adapter->hw, E1000_RCTL);
1739 		reg_rctl &= ~E1000_RCTL_RST;
1740 		E1000_WRITE_REG(&adapter->hw, E1000_RCTL, reg_rctl);
1741 		msec_delay(5);
1742 		if (adapter->hw.bus.pci_cmd_word & CMD_MEM_WRT_INVALIDATE)
1743 			e1000_pci_set_mwi(&adapter->hw);
1744 	}
1745 }
1746 
1747 /*********************************************************************
1748  *  Timer routine
1749  *
1750  *  This routine schedules em_if_update_admin_status() to check for
1751  *  link status and to gather statistics as well as to perform some
1752  *  controller-specific hardware patting.
1753  *
1754  **********************************************************************/
1755 static void
1756 em_if_timer(if_ctx_t ctx, uint16_t qid)
1757 {
1758 
1759 	if (qid != 0)
1760 		return;
1761 
1762 	iflib_admin_intr_deferred(ctx);
1763 }
1764 
1765 static void
1766 em_if_update_admin_status(if_ctx_t ctx)
1767 {
1768 	struct adapter *adapter = iflib_get_softc(ctx);
1769 	struct e1000_hw *hw = &adapter->hw;
1770 	device_t dev = iflib_get_dev(ctx);
1771 	u32 link_check, thstat, ctrl;
1772 
1773 	link_check = thstat = ctrl = 0;
1774 	/* Get the cached link value or read phy for real */
1775 	switch (hw->phy.media_type) {
1776 	case e1000_media_type_copper:
1777 		if (hw->mac.get_link_status) {
1778 			if (hw->mac.type == e1000_pch_spt)
1779 				msec_delay(50);
1780 			/* Do the work to read phy */
1781 			e1000_check_for_link(hw);
1782 			link_check = !hw->mac.get_link_status;
1783 			if (link_check) /* ESB2 fix */
1784 				e1000_cfg_on_link_up(hw);
1785 		} else {
1786 			link_check = TRUE;
1787 		}
1788 		break;
1789 	case e1000_media_type_fiber:
1790 		e1000_check_for_link(hw);
1791 		link_check = (E1000_READ_REG(hw, E1000_STATUS) &
1792 			    E1000_STATUS_LU);
1793 		break;
1794 	case e1000_media_type_internal_serdes:
1795 		e1000_check_for_link(hw);
1796 		link_check = adapter->hw.mac.serdes_has_link;
1797 		break;
1798 	/* VF device is type_unknown */
1799 	case e1000_media_type_unknown:
1800 		e1000_check_for_link(hw);
1801 		link_check = !hw->mac.get_link_status;
1802 		/* FALLTHROUGH */
1803 	default:
1804 		break;
1805 	}
1806 
1807 	/* Check for thermal downshift or shutdown */
1808 	if (hw->mac.type == e1000_i350) {
1809 		thstat = E1000_READ_REG(hw, E1000_THSTAT);
1810 		ctrl = E1000_READ_REG(hw, E1000_CTRL_EXT);
1811 	}
1812 
1813 	/* Now check for a transition */
1814 	if (link_check && (adapter->link_active == 0)) {
1815 		e1000_get_speed_and_duplex(hw, &adapter->link_speed,
1816 		    &adapter->link_duplex);
1817 		/* Check if we must disable SPEED_MODE bit on PCI-E */
1818 		if ((adapter->link_speed != SPEED_1000) &&
1819 		    ((hw->mac.type == e1000_82571) ||
1820 		    (hw->mac.type == e1000_82572))) {
1821 			int tarc0;
1822 			tarc0 = E1000_READ_REG(hw, E1000_TARC(0));
1823 			tarc0 &= ~TARC_SPEED_MODE_BIT;
1824 			E1000_WRITE_REG(hw, E1000_TARC(0), tarc0);
1825 		}
1826 		if (bootverbose)
1827 			device_printf(dev, "Link is up %d Mbps %s\n",
1828 			    adapter->link_speed,
1829 			    ((adapter->link_duplex == FULL_DUPLEX) ?
1830 			    "Full Duplex" : "Half Duplex"));
1831 		adapter->link_active = 1;
1832 		adapter->smartspeed = 0;
1833 		if ((ctrl & E1000_CTRL_EXT_LINK_MODE_MASK) ==
1834 		    E1000_CTRL_EXT_LINK_MODE_GMII &&
1835 		    (thstat & E1000_THSTAT_LINK_THROTTLE))
1836 			device_printf(dev, "Link: thermal downshift\n");
1837 		/* Delay Link Up for Phy update */
1838 		if (((hw->mac.type == e1000_i210) ||
1839 		    (hw->mac.type == e1000_i211)) &&
1840 		    (hw->phy.id == I210_I_PHY_ID))
1841 			msec_delay(I210_LINK_DELAY);
1842 		/* Reset if the media type changed. */
1843 		if ((hw->dev_spec._82575.media_changed) &&
1844 			(adapter->hw.mac.type >= igb_mac_min)) {
1845 			hw->dev_spec._82575.media_changed = false;
1846 			adapter->flags |= IGB_MEDIA_RESET;
1847 			em_reset(ctx);
1848 		}
1849 		iflib_link_state_change(ctx, LINK_STATE_UP,
1850 		    IF_Mbps(adapter->link_speed));
1851 	} else if (!link_check && (adapter->link_active == 1)) {
1852 		adapter->link_speed = 0;
1853 		adapter->link_duplex = 0;
1854 		adapter->link_active = 0;
1855 		iflib_link_state_change(ctx, LINK_STATE_DOWN, 0);
1856 	}
1857 	em_update_stats_counters(adapter);
1858 
1859 	/* Reset LAA into RAR[0] on 82571 */
1860 	if (hw->mac.type == e1000_82571 && e1000_get_laa_state_82571(hw))
1861 		e1000_rar_set(hw, hw->mac.addr, 0);
1862 
1863 	if (hw->mac.type < em_mac_min)
1864 		lem_smartspeed(adapter);
1865 	else if (hw->mac.type == e1000_82574 &&
1866 	    adapter->intr_type == IFLIB_INTR_MSIX)
1867 		E1000_WRITE_REG(&adapter->hw, E1000_IMS, EM_MSIX_LINK |
1868 		    E1000_IMS_LSC);
1869 }
1870 
1871 static void
1872 em_if_watchdog_reset(if_ctx_t ctx)
1873 {
1874 	struct adapter *adapter = iflib_get_softc(ctx);
1875 
1876 	/*
1877 	 * Just count the event; iflib(4) will already trigger a
1878 	 * sufficient reset of the controller.
1879 	 */
1880 	adapter->watchdog_events++;
1881 }
1882 
1883 /*********************************************************************
1884  *
1885  *  This routine disables all traffic on the adapter by issuing a
1886  *  global reset on the MAC.
1887  *
1888  **********************************************************************/
1889 static void
1890 em_if_stop(if_ctx_t ctx)
1891 {
1892 	struct adapter *adapter = iflib_get_softc(ctx);
1893 
1894 	INIT_DEBUGOUT("em_if_stop: begin");
1895 
1896 	e1000_reset_hw(&adapter->hw);
1897 	if (adapter->hw.mac.type >= e1000_82544)
1898 		E1000_WRITE_REG(&adapter->hw, E1000_WUFC, 0);
1899 
1900 	e1000_led_off(&adapter->hw);
1901 	e1000_cleanup_led(&adapter->hw);
1902 }
1903 
1904 /*********************************************************************
1905  *
1906  *  Determine hardware revision.
1907  *
1908  **********************************************************************/
1909 static void
1910 em_identify_hardware(if_ctx_t ctx)
1911 {
1912 	device_t dev = iflib_get_dev(ctx);
1913 	struct adapter *adapter = iflib_get_softc(ctx);
1914 
1915 	/* Make sure our PCI config space has the necessary stuff set */
1916 	adapter->hw.bus.pci_cmd_word = pci_read_config(dev, PCIR_COMMAND, 2);
1917 
1918 	/* Save off the information about this board */
1919 	adapter->hw.vendor_id = pci_get_vendor(dev);
1920 	adapter->hw.device_id = pci_get_device(dev);
1921 	adapter->hw.revision_id = pci_read_config(dev, PCIR_REVID, 1);
1922 	adapter->hw.subsystem_vendor_id =
1923 	    pci_read_config(dev, PCIR_SUBVEND_0, 2);
1924 	adapter->hw.subsystem_device_id =
1925 	    pci_read_config(dev, PCIR_SUBDEV_0, 2);
1926 
1927 	/* Do Shared Code Init and Setup */
1928 	if (e1000_set_mac_type(&adapter->hw)) {
1929 		device_printf(dev, "Setup init failure\n");
1930 		return;
1931 	}
1932 }
1933 
1934 static int
1935 em_allocate_pci_resources(if_ctx_t ctx)
1936 {
1937 	struct adapter *adapter = iflib_get_softc(ctx);
1938 	device_t dev = iflib_get_dev(ctx);
1939 	int rid, val;
1940 
1941 	rid = PCIR_BAR(0);
1942 	adapter->memory = bus_alloc_resource_any(dev, SYS_RES_MEMORY,
1943 	    &rid, RF_ACTIVE);
1944 	if (adapter->memory == NULL) {
1945 		device_printf(dev, "Unable to allocate bus resource: memory\n");
1946 		return (ENXIO);
1947 	}
1948 	adapter->osdep.mem_bus_space_tag = rman_get_bustag(adapter->memory);
1949 	adapter->osdep.mem_bus_space_handle =
1950 	    rman_get_bushandle(adapter->memory);
1951 	adapter->hw.hw_addr = (u8 *)&adapter->osdep.mem_bus_space_handle;
1952 
1953 	/* Only older adapters use IO mapping */
1954 	if (adapter->hw.mac.type < em_mac_min &&
1955 	    adapter->hw.mac.type > e1000_82543) {
1956 		/* Figure our where our IO BAR is ? */
1957 		for (rid = PCIR_BAR(0); rid < PCIR_CIS;) {
1958 			val = pci_read_config(dev, rid, 4);
1959 			if (EM_BAR_TYPE(val) == EM_BAR_TYPE_IO) {
1960 				break;
1961 			}
1962 			rid += 4;
1963 			/* check for 64bit BAR */
1964 			if (EM_BAR_MEM_TYPE(val) == EM_BAR_MEM_TYPE_64BIT)
1965 				rid += 4;
1966 		}
1967 		if (rid >= PCIR_CIS) {
1968 			device_printf(dev, "Unable to locate IO BAR\n");
1969 			return (ENXIO);
1970 		}
1971 		adapter->ioport = bus_alloc_resource_any(dev, SYS_RES_IOPORT,
1972 		    &rid, RF_ACTIVE);
1973 		if (adapter->ioport == NULL) {
1974 			device_printf(dev, "Unable to allocate bus resource: "
1975 			    "ioport\n");
1976 			return (ENXIO);
1977 		}
1978 		adapter->hw.io_base = 0;
1979 		adapter->osdep.io_bus_space_tag =
1980 		    rman_get_bustag(adapter->ioport);
1981 		adapter->osdep.io_bus_space_handle =
1982 		    rman_get_bushandle(adapter->ioport);
1983 	}
1984 
1985 	adapter->hw.back = &adapter->osdep;
1986 
1987 	return (0);
1988 }
1989 
1990 /*********************************************************************
1991  *
1992  *  Set up the MSI-X Interrupt handlers
1993  *
1994  **********************************************************************/
1995 static int
1996 em_if_msix_intr_assign(if_ctx_t ctx, int msix)
1997 {
1998 	struct adapter *adapter = iflib_get_softc(ctx);
1999 	struct em_rx_queue *rx_que = adapter->rx_queues;
2000 	struct em_tx_queue *tx_que = adapter->tx_queues;
2001 	int error, rid, i, vector = 0, rx_vectors;
2002 	char buf[16];
2003 
2004 	/* First set up ring resources */
2005 	for (i = 0; i < adapter->rx_num_queues; i++, rx_que++, vector++) {
2006 		rid = vector + 1;
2007 		snprintf(buf, sizeof(buf), "rxq%d", i);
2008 		error = iflib_irq_alloc_generic(ctx, &rx_que->que_irq, rid, IFLIB_INTR_RXTX, em_msix_que, rx_que, rx_que->me, buf);
2009 		if (error) {
2010 			device_printf(iflib_get_dev(ctx), "Failed to allocate que int %d err: %d", i, error);
2011 			adapter->rx_num_queues = i + 1;
2012 			goto fail;
2013 		}
2014 
2015 		rx_que->msix =  vector;
2016 
2017 		/*
2018 		 * Set the bit to enable interrupt
2019 		 * in E1000_IMS -- bits 20 and 21
2020 		 * are for RX0 and RX1, note this has
2021 		 * NOTHING to do with the MSI-X vector
2022 		 */
2023 		if (adapter->hw.mac.type == e1000_82574) {
2024 			rx_que->eims = 1 << (20 + i);
2025 			adapter->ims |= rx_que->eims;
2026 			adapter->ivars |= (8 | rx_que->msix) << (i * 4);
2027 		} else if (adapter->hw.mac.type == e1000_82575)
2028 			rx_que->eims = E1000_EICR_TX_QUEUE0 << vector;
2029 		else
2030 			rx_que->eims = 1 << vector;
2031 	}
2032 	rx_vectors = vector;
2033 
2034 	vector = 0;
2035 	for (i = 0; i < adapter->tx_num_queues; i++, tx_que++, vector++) {
2036 		snprintf(buf, sizeof(buf), "txq%d", i);
2037 		tx_que = &adapter->tx_queues[i];
2038 		iflib_softirq_alloc_generic(ctx,
2039 		    &adapter->rx_queues[i % adapter->rx_num_queues].que_irq,
2040 		    IFLIB_INTR_TX, tx_que, tx_que->me, buf);
2041 
2042 		tx_que->msix = (vector % adapter->rx_num_queues);
2043 
2044 		/*
2045 		 * Set the bit to enable interrupt
2046 		 * in E1000_IMS -- bits 22 and 23
2047 		 * are for TX0 and TX1, note this has
2048 		 * NOTHING to do with the MSI-X vector
2049 		 */
2050 		if (adapter->hw.mac.type == e1000_82574) {
2051 			tx_que->eims = 1 << (22 + i);
2052 			adapter->ims |= tx_que->eims;
2053 			adapter->ivars |= (8 | tx_que->msix) << (8 + (i * 4));
2054 		} else if (adapter->hw.mac.type == e1000_82575) {
2055 			tx_que->eims = E1000_EICR_TX_QUEUE0 << i;
2056 		} else {
2057 			tx_que->eims = 1 << i;
2058 		}
2059 	}
2060 
2061 	/* Link interrupt */
2062 	rid = rx_vectors + 1;
2063 	error = iflib_irq_alloc_generic(ctx, &adapter->irq, rid, IFLIB_INTR_ADMIN, em_msix_link, adapter, 0, "aq");
2064 
2065 	if (error) {
2066 		device_printf(iflib_get_dev(ctx), "Failed to register admin handler");
2067 		goto fail;
2068 	}
2069 	adapter->linkvec = rx_vectors;
2070 	if (adapter->hw.mac.type < igb_mac_min) {
2071 		adapter->ivars |=  (8 | rx_vectors) << 16;
2072 		adapter->ivars |= 0x80000000;
2073 	}
2074 	return (0);
2075 fail:
2076 	iflib_irq_free(ctx, &adapter->irq);
2077 	rx_que = adapter->rx_queues;
2078 	for (int i = 0; i < adapter->rx_num_queues; i++, rx_que++)
2079 		iflib_irq_free(ctx, &rx_que->que_irq);
2080 	return (error);
2081 }
2082 
2083 static void
2084 igb_configure_queues(struct adapter *adapter)
2085 {
2086 	struct e1000_hw *hw = &adapter->hw;
2087 	struct em_rx_queue *rx_que;
2088 	struct em_tx_queue *tx_que;
2089 	u32 tmp, ivar = 0, newitr = 0;
2090 
2091 	/* First turn on RSS capability */
2092 	if (adapter->hw.mac.type != e1000_82575)
2093 		E1000_WRITE_REG(hw, E1000_GPIE,
2094 		    E1000_GPIE_MSIX_MODE | E1000_GPIE_EIAME |
2095 		    E1000_GPIE_PBA | E1000_GPIE_NSICR);
2096 
2097 	/* Turn on MSI-X */
2098 	switch (adapter->hw.mac.type) {
2099 	case e1000_82580:
2100 	case e1000_i350:
2101 	case e1000_i354:
2102 	case e1000_i210:
2103 	case e1000_i211:
2104 	case e1000_vfadapt:
2105 	case e1000_vfadapt_i350:
2106 		/* RX entries */
2107 		for (int i = 0; i < adapter->rx_num_queues; i++) {
2108 			u32 index = i >> 1;
2109 			ivar = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index);
2110 			rx_que = &adapter->rx_queues[i];
2111 			if (i & 1) {
2112 				ivar &= 0xFF00FFFF;
2113 				ivar |= (rx_que->msix | E1000_IVAR_VALID) << 16;
2114 			} else {
2115 				ivar &= 0xFFFFFF00;
2116 				ivar |= rx_que->msix | E1000_IVAR_VALID;
2117 			}
2118 			E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, ivar);
2119 		}
2120 		/* TX entries */
2121 		for (int i = 0; i < adapter->tx_num_queues; i++) {
2122 			u32 index = i >> 1;
2123 			ivar = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index);
2124 			tx_que = &adapter->tx_queues[i];
2125 			if (i & 1) {
2126 				ivar &= 0x00FFFFFF;
2127 				ivar |= (tx_que->msix | E1000_IVAR_VALID) << 24;
2128 			} else {
2129 				ivar &= 0xFFFF00FF;
2130 				ivar |= (tx_que->msix | E1000_IVAR_VALID) << 8;
2131 			}
2132 			E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, ivar);
2133 			adapter->que_mask |= tx_que->eims;
2134 		}
2135 
2136 		/* And for the link interrupt */
2137 		ivar = (adapter->linkvec | E1000_IVAR_VALID) << 8;
2138 		adapter->link_mask = 1 << adapter->linkvec;
2139 		E1000_WRITE_REG(hw, E1000_IVAR_MISC, ivar);
2140 		break;
2141 	case e1000_82576:
2142 		/* RX entries */
2143 		for (int i = 0; i < adapter->rx_num_queues; i++) {
2144 			u32 index = i & 0x7; /* Each IVAR has two entries */
2145 			ivar = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index);
2146 			rx_que = &adapter->rx_queues[i];
2147 			if (i < 8) {
2148 				ivar &= 0xFFFFFF00;
2149 				ivar |= rx_que->msix | E1000_IVAR_VALID;
2150 			} else {
2151 				ivar &= 0xFF00FFFF;
2152 				ivar |= (rx_que->msix | E1000_IVAR_VALID) << 16;
2153 			}
2154 			E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, ivar);
2155 			adapter->que_mask |= rx_que->eims;
2156 		}
2157 		/* TX entries */
2158 		for (int i = 0; i < adapter->tx_num_queues; i++) {
2159 			u32 index = i & 0x7; /* Each IVAR has two entries */
2160 			ivar = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index);
2161 			tx_que = &adapter->tx_queues[i];
2162 			if (i < 8) {
2163 				ivar &= 0xFFFF00FF;
2164 				ivar |= (tx_que->msix | E1000_IVAR_VALID) << 8;
2165 			} else {
2166 				ivar &= 0x00FFFFFF;
2167 				ivar |= (tx_que->msix | E1000_IVAR_VALID) << 24;
2168 			}
2169 			E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, ivar);
2170 			adapter->que_mask |= tx_que->eims;
2171 		}
2172 
2173 		/* And for the link interrupt */
2174 		ivar = (adapter->linkvec | E1000_IVAR_VALID) << 8;
2175 		adapter->link_mask = 1 << adapter->linkvec;
2176 		E1000_WRITE_REG(hw, E1000_IVAR_MISC, ivar);
2177 		break;
2178 
2179 	case e1000_82575:
2180 		/* enable MSI-X support*/
2181 		tmp = E1000_READ_REG(hw, E1000_CTRL_EXT);
2182 		tmp |= E1000_CTRL_EXT_PBA_CLR;
2183 		/* Auto-Mask interrupts upon ICR read. */
2184 		tmp |= E1000_CTRL_EXT_EIAME;
2185 		tmp |= E1000_CTRL_EXT_IRCA;
2186 		E1000_WRITE_REG(hw, E1000_CTRL_EXT, tmp);
2187 
2188 		/* Queues */
2189 		for (int i = 0; i < adapter->rx_num_queues; i++) {
2190 			rx_que = &adapter->rx_queues[i];
2191 			tmp = E1000_EICR_RX_QUEUE0 << i;
2192 			tmp |= E1000_EICR_TX_QUEUE0 << i;
2193 			rx_que->eims = tmp;
2194 			E1000_WRITE_REG_ARRAY(hw, E1000_MSIXBM(0),
2195 			    i, rx_que->eims);
2196 			adapter->que_mask |= rx_que->eims;
2197 		}
2198 
2199 		/* Link */
2200 		E1000_WRITE_REG(hw, E1000_MSIXBM(adapter->linkvec),
2201 		    E1000_EIMS_OTHER);
2202 		adapter->link_mask |= E1000_EIMS_OTHER;
2203 	default:
2204 		break;
2205 	}
2206 
2207 	/* Set the starting interrupt rate */
2208 	if (em_max_interrupt_rate > 0)
2209 		newitr = (4000000 / em_max_interrupt_rate) & 0x7FFC;
2210 
2211 	if (hw->mac.type == e1000_82575)
2212 		newitr |= newitr << 16;
2213 	else
2214 		newitr |= E1000_EITR_CNT_IGNR;
2215 
2216 	for (int i = 0; i < adapter->rx_num_queues; i++) {
2217 		rx_que = &adapter->rx_queues[i];
2218 		E1000_WRITE_REG(hw, E1000_EITR(rx_que->msix), newitr);
2219 	}
2220 
2221 	return;
2222 }
2223 
2224 static void
2225 em_free_pci_resources(if_ctx_t ctx)
2226 {
2227 	struct adapter *adapter = iflib_get_softc(ctx);
2228 	struct em_rx_queue *que = adapter->rx_queues;
2229 	device_t dev = iflib_get_dev(ctx);
2230 
2231 	/* Release all MSI-X queue resources */
2232 	if (adapter->intr_type == IFLIB_INTR_MSIX)
2233 		iflib_irq_free(ctx, &adapter->irq);
2234 
2235 	if (que != NULL) {
2236 		for (int i = 0; i < adapter->rx_num_queues; i++, que++) {
2237 			iflib_irq_free(ctx, &que->que_irq);
2238 		}
2239 	}
2240 
2241 	if (adapter->memory != NULL) {
2242 		bus_release_resource(dev, SYS_RES_MEMORY,
2243 		    rman_get_rid(adapter->memory), adapter->memory);
2244 		adapter->memory = NULL;
2245 	}
2246 
2247 	if (adapter->flash != NULL) {
2248 		bus_release_resource(dev, SYS_RES_MEMORY,
2249 		    rman_get_rid(adapter->flash), adapter->flash);
2250 		adapter->flash = NULL;
2251 	}
2252 
2253 	if (adapter->ioport != NULL) {
2254 		bus_release_resource(dev, SYS_RES_IOPORT,
2255 		    rman_get_rid(adapter->ioport), adapter->ioport);
2256 		adapter->ioport = NULL;
2257 	}
2258 }
2259 
2260 /* Set up MSI or MSI-X */
2261 static int
2262 em_setup_msix(if_ctx_t ctx)
2263 {
2264 	struct adapter *adapter = iflib_get_softc(ctx);
2265 
2266 	if (adapter->hw.mac.type == e1000_82574) {
2267 		em_enable_vectors_82574(ctx);
2268 	}
2269 	return (0);
2270 }
2271 
2272 /*********************************************************************
2273  *
2274  *  Workaround for SmartSpeed on 82541 and 82547 controllers
2275  *
2276  **********************************************************************/
2277 static void
2278 lem_smartspeed(struct adapter *adapter)
2279 {
2280 	u16 phy_tmp;
2281 
2282 	if (adapter->link_active || (adapter->hw.phy.type != e1000_phy_igp) ||
2283 	    adapter->hw.mac.autoneg == 0 ||
2284 	    (adapter->hw.phy.autoneg_advertised & ADVERTISE_1000_FULL) == 0)
2285 		return;
2286 
2287 	if (adapter->smartspeed == 0) {
2288 		/* If Master/Slave config fault is asserted twice,
2289 		 * we assume back-to-back */
2290 		e1000_read_phy_reg(&adapter->hw, PHY_1000T_STATUS, &phy_tmp);
2291 		if (!(phy_tmp & SR_1000T_MS_CONFIG_FAULT))
2292 			return;
2293 		e1000_read_phy_reg(&adapter->hw, PHY_1000T_STATUS, &phy_tmp);
2294 		if (phy_tmp & SR_1000T_MS_CONFIG_FAULT) {
2295 			e1000_read_phy_reg(&adapter->hw,
2296 			    PHY_1000T_CTRL, &phy_tmp);
2297 			if(phy_tmp & CR_1000T_MS_ENABLE) {
2298 				phy_tmp &= ~CR_1000T_MS_ENABLE;
2299 				e1000_write_phy_reg(&adapter->hw,
2300 				    PHY_1000T_CTRL, phy_tmp);
2301 				adapter->smartspeed++;
2302 				if(adapter->hw.mac.autoneg &&
2303 				   !e1000_copper_link_autoneg(&adapter->hw) &&
2304 				   !e1000_read_phy_reg(&adapter->hw,
2305 				    PHY_CONTROL, &phy_tmp)) {
2306 					phy_tmp |= (MII_CR_AUTO_NEG_EN |
2307 						    MII_CR_RESTART_AUTO_NEG);
2308 					e1000_write_phy_reg(&adapter->hw,
2309 					    PHY_CONTROL, phy_tmp);
2310 				}
2311 			}
2312 		}
2313 		return;
2314 	} else if(adapter->smartspeed == EM_SMARTSPEED_DOWNSHIFT) {
2315 		/* If still no link, perhaps using 2/3 pair cable */
2316 		e1000_read_phy_reg(&adapter->hw, PHY_1000T_CTRL, &phy_tmp);
2317 		phy_tmp |= CR_1000T_MS_ENABLE;
2318 		e1000_write_phy_reg(&adapter->hw, PHY_1000T_CTRL, phy_tmp);
2319 		if(adapter->hw.mac.autoneg &&
2320 		   !e1000_copper_link_autoneg(&adapter->hw) &&
2321 		   !e1000_read_phy_reg(&adapter->hw, PHY_CONTROL, &phy_tmp)) {
2322 			phy_tmp |= (MII_CR_AUTO_NEG_EN |
2323 				    MII_CR_RESTART_AUTO_NEG);
2324 			e1000_write_phy_reg(&adapter->hw, PHY_CONTROL, phy_tmp);
2325 		}
2326 	}
2327 	/* Restart process after EM_SMARTSPEED_MAX iterations */
2328 	if(adapter->smartspeed++ == EM_SMARTSPEED_MAX)
2329 		adapter->smartspeed = 0;
2330 }
2331 
2332 /*********************************************************************
2333  *
2334  *  Initialize the DMA Coalescing feature
2335  *
2336  **********************************************************************/
2337 static void
2338 igb_init_dmac(struct adapter *adapter, u32 pba)
2339 {
2340 	device_t	dev = adapter->dev;
2341 	struct e1000_hw *hw = &adapter->hw;
2342 	u32 		dmac, reg = ~E1000_DMACR_DMAC_EN;
2343 	u16		hwm;
2344 	u16		max_frame_size;
2345 
2346 	if (hw->mac.type == e1000_i211)
2347 		return;
2348 
2349 	max_frame_size = adapter->shared->isc_max_frame_size;
2350 	if (hw->mac.type > e1000_82580) {
2351 
2352 		if (adapter->dmac == 0) { /* Disabling it */
2353 			E1000_WRITE_REG(hw, E1000_DMACR, reg);
2354 			return;
2355 		} else
2356 			device_printf(dev, "DMA Coalescing enabled\n");
2357 
2358 		/* Set starting threshold */
2359 		E1000_WRITE_REG(hw, E1000_DMCTXTH, 0);
2360 
2361 		hwm = 64 * pba - max_frame_size / 16;
2362 		if (hwm < 64 * (pba - 6))
2363 			hwm = 64 * (pba - 6);
2364 		reg = E1000_READ_REG(hw, E1000_FCRTC);
2365 		reg &= ~E1000_FCRTC_RTH_COAL_MASK;
2366 		reg |= ((hwm << E1000_FCRTC_RTH_COAL_SHIFT)
2367 		    & E1000_FCRTC_RTH_COAL_MASK);
2368 		E1000_WRITE_REG(hw, E1000_FCRTC, reg);
2369 
2370 
2371 		dmac = pba - max_frame_size / 512;
2372 		if (dmac < pba - 10)
2373 			dmac = pba - 10;
2374 		reg = E1000_READ_REG(hw, E1000_DMACR);
2375 		reg &= ~E1000_DMACR_DMACTHR_MASK;
2376 		reg |= ((dmac << E1000_DMACR_DMACTHR_SHIFT)
2377 		    & E1000_DMACR_DMACTHR_MASK);
2378 
2379 		/* transition to L0x or L1 if available..*/
2380 		reg |= (E1000_DMACR_DMAC_EN | E1000_DMACR_DMAC_LX_MASK);
2381 
2382 		/* Check if status is 2.5Gb backplane connection
2383 		* before configuration of watchdog timer, which is
2384 		* in msec values in 12.8usec intervals
2385 		* watchdog timer= msec values in 32usec intervals
2386 		* for non 2.5Gb connection
2387 		*/
2388 		if (hw->mac.type == e1000_i354) {
2389 			int status = E1000_READ_REG(hw, E1000_STATUS);
2390 			if ((status & E1000_STATUS_2P5_SKU) &&
2391 			    (!(status & E1000_STATUS_2P5_SKU_OVER)))
2392 				reg |= ((adapter->dmac * 5) >> 6);
2393 			else
2394 				reg |= (adapter->dmac >> 5);
2395 		} else {
2396 			reg |= (adapter->dmac >> 5);
2397 		}
2398 
2399 		E1000_WRITE_REG(hw, E1000_DMACR, reg);
2400 
2401 		E1000_WRITE_REG(hw, E1000_DMCRTRH, 0);
2402 
2403 		/* Set the interval before transition */
2404 		reg = E1000_READ_REG(hw, E1000_DMCTLX);
2405 		if (hw->mac.type == e1000_i350)
2406 			reg |= IGB_DMCTLX_DCFLUSH_DIS;
2407 		/*
2408 		** in 2.5Gb connection, TTLX unit is 0.4 usec
2409 		** which is 0x4*2 = 0xA. But delay is still 4 usec
2410 		*/
2411 		if (hw->mac.type == e1000_i354) {
2412 			int status = E1000_READ_REG(hw, E1000_STATUS);
2413 			if ((status & E1000_STATUS_2P5_SKU) &&
2414 			    (!(status & E1000_STATUS_2P5_SKU_OVER)))
2415 				reg |= 0xA;
2416 			else
2417 				reg |= 0x4;
2418 		} else {
2419 			reg |= 0x4;
2420 		}
2421 
2422 		E1000_WRITE_REG(hw, E1000_DMCTLX, reg);
2423 
2424 		/* free space in tx packet buffer to wake from DMA coal */
2425 		E1000_WRITE_REG(hw, E1000_DMCTXTH, (IGB_TXPBSIZE -
2426 		    (2 * max_frame_size)) >> 6);
2427 
2428 		/* make low power state decision controlled by DMA coal */
2429 		reg = E1000_READ_REG(hw, E1000_PCIEMISC);
2430 		reg &= ~E1000_PCIEMISC_LX_DECISION;
2431 		E1000_WRITE_REG(hw, E1000_PCIEMISC, reg);
2432 
2433 	} else if (hw->mac.type == e1000_82580) {
2434 		u32 reg = E1000_READ_REG(hw, E1000_PCIEMISC);
2435 		E1000_WRITE_REG(hw, E1000_PCIEMISC,
2436 		    reg & ~E1000_PCIEMISC_LX_DECISION);
2437 		E1000_WRITE_REG(hw, E1000_DMACR, 0);
2438 	}
2439 }
2440 
2441 /*********************************************************************
2442  *
2443  *  Initialize the hardware to a configuration as specified by the
2444  *  adapter structure.
2445  *
2446  **********************************************************************/
2447 static void
2448 em_reset(if_ctx_t ctx)
2449 {
2450 	device_t dev = iflib_get_dev(ctx);
2451 	struct adapter *adapter = iflib_get_softc(ctx);
2452 	struct ifnet *ifp = iflib_get_ifp(ctx);
2453 	struct e1000_hw *hw = &adapter->hw;
2454 	u16 rx_buffer_size;
2455 	u32 pba;
2456 
2457 	INIT_DEBUGOUT("em_reset: begin");
2458 	/* Let the firmware know the OS is in control */
2459 	em_get_hw_control(adapter);
2460 
2461 	/* Set up smart power down as default off on newer adapters. */
2462 	if (!em_smart_pwr_down && (hw->mac.type == e1000_82571 ||
2463 	    hw->mac.type == e1000_82572)) {
2464 		u16 phy_tmp = 0;
2465 
2466 		/* Speed up time to link by disabling smart power down. */
2467 		e1000_read_phy_reg(hw, IGP02E1000_PHY_POWER_MGMT, &phy_tmp);
2468 		phy_tmp &= ~IGP02E1000_PM_SPD;
2469 		e1000_write_phy_reg(hw, IGP02E1000_PHY_POWER_MGMT, phy_tmp);
2470 	}
2471 
2472 	/*
2473 	 * Packet Buffer Allocation (PBA)
2474 	 * Writing PBA sets the receive portion of the buffer
2475 	 * the remainder is used for the transmit buffer.
2476 	 */
2477 	switch (hw->mac.type) {
2478 	/* Total Packet Buffer on these is 48K */
2479 	case e1000_82571:
2480 	case e1000_82572:
2481 	case e1000_80003es2lan:
2482 			pba = E1000_PBA_32K; /* 32K for Rx, 16K for Tx */
2483 		break;
2484 	case e1000_82573: /* 82573: Total Packet Buffer is 32K */
2485 			pba = E1000_PBA_12K; /* 12K for Rx, 20K for Tx */
2486 		break;
2487 	case e1000_82574:
2488 	case e1000_82583:
2489 			pba = E1000_PBA_20K; /* 20K for Rx, 20K for Tx */
2490 		break;
2491 	case e1000_ich8lan:
2492 		pba = E1000_PBA_8K;
2493 		break;
2494 	case e1000_ich9lan:
2495 	case e1000_ich10lan:
2496 		/* Boost Receive side for jumbo frames */
2497 		if (adapter->hw.mac.max_frame_size > 4096)
2498 			pba = E1000_PBA_14K;
2499 		else
2500 			pba = E1000_PBA_10K;
2501 		break;
2502 	case e1000_pchlan:
2503 	case e1000_pch2lan:
2504 	case e1000_pch_lpt:
2505 	case e1000_pch_spt:
2506 	case e1000_pch_cnp:
2507 		pba = E1000_PBA_26K;
2508 		break;
2509 	case e1000_82575:
2510 		pba = E1000_PBA_32K;
2511 		break;
2512 	case e1000_82576:
2513 	case e1000_vfadapt:
2514 		pba = E1000_READ_REG(hw, E1000_RXPBS);
2515 		pba &= E1000_RXPBS_SIZE_MASK_82576;
2516 		break;
2517 	case e1000_82580:
2518 	case e1000_i350:
2519 	case e1000_i354:
2520 	case e1000_vfadapt_i350:
2521 		pba = E1000_READ_REG(hw, E1000_RXPBS);
2522 		pba = e1000_rxpbs_adjust_82580(pba);
2523 		break;
2524 	case e1000_i210:
2525 	case e1000_i211:
2526 		pba = E1000_PBA_34K;
2527 		break;
2528 	default:
2529 		if (adapter->hw.mac.max_frame_size > 8192)
2530 			pba = E1000_PBA_40K; /* 40K for Rx, 24K for Tx */
2531 		else
2532 			pba = E1000_PBA_48K; /* 48K for Rx, 16K for Tx */
2533 	}
2534 
2535 	/* Special needs in case of Jumbo frames */
2536 	if ((hw->mac.type == e1000_82575) && (ifp->if_mtu > ETHERMTU)) {
2537 		u32 tx_space, min_tx, min_rx;
2538 		pba = E1000_READ_REG(hw, E1000_PBA);
2539 		tx_space = pba >> 16;
2540 		pba &= 0xffff;
2541 		min_tx = (adapter->hw.mac.max_frame_size +
2542 		    sizeof(struct e1000_tx_desc) - ETHERNET_FCS_SIZE) * 2;
2543 		min_tx = roundup2(min_tx, 1024);
2544 		min_tx >>= 10;
2545 		min_rx = adapter->hw.mac.max_frame_size;
2546 		min_rx = roundup2(min_rx, 1024);
2547 		min_rx >>= 10;
2548 		if (tx_space < min_tx &&
2549 		    ((min_tx - tx_space) < pba)) {
2550 			pba = pba - (min_tx - tx_space);
2551 			/*
2552 			 * if short on rx space, rx wins
2553 			 * and must trump tx adjustment
2554 			 */
2555 			if (pba < min_rx)
2556 				pba = min_rx;
2557 		}
2558 		E1000_WRITE_REG(hw, E1000_PBA, pba);
2559 	}
2560 
2561 	if (hw->mac.type < igb_mac_min)
2562 		E1000_WRITE_REG(&adapter->hw, E1000_PBA, pba);
2563 
2564 	INIT_DEBUGOUT1("em_reset: pba=%dK",pba);
2565 
2566 	/*
2567 	 * These parameters control the automatic generation (Tx) and
2568 	 * response (Rx) to Ethernet PAUSE frames.
2569 	 * - High water mark should allow for at least two frames to be
2570 	 *   received after sending an XOFF.
2571 	 * - Low water mark works best when it is very near the high water mark.
2572 	 *   This allows the receiver to restart by sending XON when it has
2573 	 *   drained a bit. Here we use an arbitrary value of 1500 which will
2574 	 *   restart after one full frame is pulled from the buffer. There
2575 	 *   could be several smaller frames in the buffer and if so they will
2576 	 *   not trigger the XON until their total number reduces the buffer
2577 	 *   by 1500.
2578 	 * - The pause time is fairly large at 1000 x 512ns = 512 usec.
2579 	 */
2580 	rx_buffer_size = (pba & 0xffff) << 10;
2581 	hw->fc.high_water = rx_buffer_size -
2582 	    roundup2(adapter->hw.mac.max_frame_size, 1024);
2583 	hw->fc.low_water = hw->fc.high_water - 1500;
2584 
2585 	if (adapter->fc) /* locally set flow control value? */
2586 		hw->fc.requested_mode = adapter->fc;
2587 	else
2588 		hw->fc.requested_mode = e1000_fc_full;
2589 
2590 	if (hw->mac.type == e1000_80003es2lan)
2591 		hw->fc.pause_time = 0xFFFF;
2592 	else
2593 		hw->fc.pause_time = EM_FC_PAUSE_TIME;
2594 
2595 	hw->fc.send_xon = TRUE;
2596 
2597 	/* Device specific overrides/settings */
2598 	switch (hw->mac.type) {
2599 	case e1000_pchlan:
2600 		/* Workaround: no TX flow ctrl for PCH */
2601 		hw->fc.requested_mode = e1000_fc_rx_pause;
2602 		hw->fc.pause_time = 0xFFFF; /* override */
2603 		if (if_getmtu(ifp) > ETHERMTU) {
2604 			hw->fc.high_water = 0x3500;
2605 			hw->fc.low_water = 0x1500;
2606 		} else {
2607 			hw->fc.high_water = 0x5000;
2608 			hw->fc.low_water = 0x3000;
2609 		}
2610 		hw->fc.refresh_time = 0x1000;
2611 		break;
2612 	case e1000_pch2lan:
2613 	case e1000_pch_lpt:
2614 	case e1000_pch_spt:
2615 	case e1000_pch_cnp:
2616 		hw->fc.high_water = 0x5C20;
2617 		hw->fc.low_water = 0x5048;
2618 		hw->fc.pause_time = 0x0650;
2619 		hw->fc.refresh_time = 0x0400;
2620 		/* Jumbos need adjusted PBA */
2621 		if (if_getmtu(ifp) > ETHERMTU)
2622 			E1000_WRITE_REG(hw, E1000_PBA, 12);
2623 		else
2624 			E1000_WRITE_REG(hw, E1000_PBA, 26);
2625 		break;
2626 	case e1000_82575:
2627 	case e1000_82576:
2628 		/* 8-byte granularity */
2629 		hw->fc.low_water = hw->fc.high_water - 8;
2630 		break;
2631 	case e1000_82580:
2632 	case e1000_i350:
2633 	case e1000_i354:
2634 	case e1000_i210:
2635 	case e1000_i211:
2636 	case e1000_vfadapt:
2637 	case e1000_vfadapt_i350:
2638 		/* 16-byte granularity */
2639 		hw->fc.low_water = hw->fc.high_water - 16;
2640 		break;
2641 	case e1000_ich9lan:
2642 	case e1000_ich10lan:
2643 		if (if_getmtu(ifp) > ETHERMTU) {
2644 			hw->fc.high_water = 0x2800;
2645 			hw->fc.low_water = hw->fc.high_water - 8;
2646 			break;
2647 		}
2648 		/* FALLTHROUGH */
2649 	default:
2650 		if (hw->mac.type == e1000_80003es2lan)
2651 			hw->fc.pause_time = 0xFFFF;
2652 		break;
2653 	}
2654 
2655 	/* Issue a global reset */
2656 	e1000_reset_hw(hw);
2657 	if (adapter->hw.mac.type >= igb_mac_min) {
2658 		E1000_WRITE_REG(hw, E1000_WUC, 0);
2659 	} else {
2660 		E1000_WRITE_REG(hw, E1000_WUFC, 0);
2661 		em_disable_aspm(adapter);
2662 	}
2663 	if (adapter->flags & IGB_MEDIA_RESET) {
2664 		e1000_setup_init_funcs(hw, TRUE);
2665 		e1000_get_bus_info(hw);
2666 		adapter->flags &= ~IGB_MEDIA_RESET;
2667 	}
2668 	/* and a re-init */
2669 	if (e1000_init_hw(hw) < 0) {
2670 		device_printf(dev, "Hardware Initialization Failed\n");
2671 		return;
2672 	}
2673 	if (adapter->hw.mac.type >= igb_mac_min)
2674 		igb_init_dmac(adapter, pba);
2675 
2676 	E1000_WRITE_REG(hw, E1000_VET, ETHERTYPE_VLAN);
2677 	e1000_get_phy_info(hw);
2678 	e1000_check_for_link(hw);
2679 }
2680 
2681 /*
2682  * Initialise the RSS mapping for NICs that support multiple transmit/
2683  * receive rings.
2684  */
2685 
2686 #define RSSKEYLEN 10
2687 static void
2688 em_initialize_rss_mapping(struct adapter *adapter)
2689 {
2690 	uint8_t  rss_key[4 * RSSKEYLEN];
2691 	uint32_t reta = 0;
2692 	struct e1000_hw	*hw = &adapter->hw;
2693 	int i;
2694 
2695 	/*
2696 	 * Configure RSS key
2697 	 */
2698 	arc4rand(rss_key, sizeof(rss_key), 0);
2699 	for (i = 0; i < RSSKEYLEN; ++i) {
2700 		uint32_t rssrk = 0;
2701 
2702 		rssrk = EM_RSSRK_VAL(rss_key, i);
2703 		E1000_WRITE_REG(hw,E1000_RSSRK(i), rssrk);
2704 	}
2705 
2706 	/*
2707 	 * Configure RSS redirect table in following fashion:
2708 	 * (hash & ring_cnt_mask) == rdr_table[(hash & rdr_table_mask)]
2709 	 */
2710 	for (i = 0; i < sizeof(reta); ++i) {
2711 		uint32_t q;
2712 
2713 		q = (i % adapter->rx_num_queues) << 7;
2714 		reta |= q << (8 * i);
2715 	}
2716 
2717 	for (i = 0; i < 32; ++i)
2718 		E1000_WRITE_REG(hw, E1000_RETA(i), reta);
2719 
2720 	E1000_WRITE_REG(hw, E1000_MRQC, E1000_MRQC_RSS_ENABLE_2Q |
2721 			E1000_MRQC_RSS_FIELD_IPV4_TCP |
2722 			E1000_MRQC_RSS_FIELD_IPV4 |
2723 			E1000_MRQC_RSS_FIELD_IPV6_TCP_EX |
2724 			E1000_MRQC_RSS_FIELD_IPV6_EX |
2725 			E1000_MRQC_RSS_FIELD_IPV6);
2726 }
2727 
2728 static void
2729 igb_initialize_rss_mapping(struct adapter *adapter)
2730 {
2731 	struct e1000_hw *hw = &adapter->hw;
2732 	int i;
2733 	int queue_id;
2734 	u32 reta;
2735 	u32 rss_key[10], mrqc, shift = 0;
2736 
2737 	/* XXX? */
2738 	if (adapter->hw.mac.type == e1000_82575)
2739 		shift = 6;
2740 
2741 	/*
2742 	 * The redirection table controls which destination
2743 	 * queue each bucket redirects traffic to.
2744 	 * Each DWORD represents four queues, with the LSB
2745 	 * being the first queue in the DWORD.
2746 	 *
2747 	 * This just allocates buckets to queues using round-robin
2748 	 * allocation.
2749 	 *
2750 	 * NOTE: It Just Happens to line up with the default
2751 	 * RSS allocation method.
2752 	 */
2753 
2754 	/* Warning FM follows */
2755 	reta = 0;
2756 	for (i = 0; i < 128; i++) {
2757 #ifdef RSS
2758 		queue_id = rss_get_indirection_to_bucket(i);
2759 		/*
2760 		 * If we have more queues than buckets, we'll
2761 		 * end up mapping buckets to a subset of the
2762 		 * queues.
2763 		 *
2764 		 * If we have more buckets than queues, we'll
2765 		 * end up instead assigning multiple buckets
2766 		 * to queues.
2767 		 *
2768 		 * Both are suboptimal, but we need to handle
2769 		 * the case so we don't go out of bounds
2770 		 * indexing arrays and such.
2771 		 */
2772 		queue_id = queue_id % adapter->rx_num_queues;
2773 #else
2774 		queue_id = (i % adapter->rx_num_queues);
2775 #endif
2776 		/* Adjust if required */
2777 		queue_id = queue_id << shift;
2778 
2779 		/*
2780 		 * The low 8 bits are for hash value (n+0);
2781 		 * The next 8 bits are for hash value (n+1), etc.
2782 		 */
2783 		reta = reta >> 8;
2784 		reta = reta | ( ((uint32_t) queue_id) << 24);
2785 		if ((i & 3) == 3) {
2786 			E1000_WRITE_REG(hw, E1000_RETA(i >> 2), reta);
2787 			reta = 0;
2788 		}
2789 	}
2790 
2791 	/* Now fill in hash table */
2792 
2793 	/*
2794 	 * MRQC: Multiple Receive Queues Command
2795 	 * Set queuing to RSS control, number depends on the device.
2796 	 */
2797 	mrqc = E1000_MRQC_ENABLE_RSS_8Q;
2798 
2799 #ifdef RSS
2800 	/* XXX ew typecasting */
2801 	rss_getkey((uint8_t *) &rss_key);
2802 #else
2803 	arc4rand(&rss_key, sizeof(rss_key), 0);
2804 #endif
2805 	for (i = 0; i < 10; i++)
2806 		E1000_WRITE_REG_ARRAY(hw, E1000_RSSRK(0), i, rss_key[i]);
2807 
2808 	/*
2809 	 * Configure the RSS fields to hash upon.
2810 	 */
2811 	mrqc |= (E1000_MRQC_RSS_FIELD_IPV4 |
2812 	    E1000_MRQC_RSS_FIELD_IPV4_TCP);
2813 	mrqc |= (E1000_MRQC_RSS_FIELD_IPV6 |
2814 	    E1000_MRQC_RSS_FIELD_IPV6_TCP);
2815 	mrqc |=( E1000_MRQC_RSS_FIELD_IPV4_UDP |
2816 	    E1000_MRQC_RSS_FIELD_IPV6_UDP);
2817 	mrqc |=( E1000_MRQC_RSS_FIELD_IPV6_UDP_EX |
2818 	    E1000_MRQC_RSS_FIELD_IPV6_TCP_EX);
2819 
2820 	E1000_WRITE_REG(hw, E1000_MRQC, mrqc);
2821 }
2822 
2823 /*********************************************************************
2824  *
2825  *  Setup networking device structure and register interface media.
2826  *
2827  **********************************************************************/
2828 static int
2829 em_setup_interface(if_ctx_t ctx)
2830 {
2831 	struct ifnet *ifp = iflib_get_ifp(ctx);
2832 	struct adapter *adapter = iflib_get_softc(ctx);
2833 	if_softc_ctx_t scctx = adapter->shared;
2834 
2835 	INIT_DEBUGOUT("em_setup_interface: begin");
2836 
2837 	/* Single Queue */
2838 	if (adapter->tx_num_queues == 1) {
2839 		if_setsendqlen(ifp, scctx->isc_ntxd[0] - 1);
2840 		if_setsendqready(ifp);
2841 	}
2842 
2843 	/*
2844 	 * Specify the media types supported by this adapter and register
2845 	 * callbacks to update media and link information
2846 	 */
2847 	if ((adapter->hw.phy.media_type == e1000_media_type_fiber) ||
2848 	    (adapter->hw.phy.media_type == e1000_media_type_internal_serdes)) {
2849 		u_char fiber_type = IFM_1000_SX;	/* default type */
2850 
2851 		if (adapter->hw.mac.type == e1000_82545)
2852 			fiber_type = IFM_1000_LX;
2853 		ifmedia_add(adapter->media, IFM_ETHER | fiber_type | IFM_FDX, 0, NULL);
2854 		ifmedia_add(adapter->media, IFM_ETHER | fiber_type, 0, NULL);
2855 	} else {
2856 		ifmedia_add(adapter->media, IFM_ETHER | IFM_10_T, 0, NULL);
2857 		ifmedia_add(adapter->media, IFM_ETHER | IFM_10_T | IFM_FDX, 0, NULL);
2858 		ifmedia_add(adapter->media, IFM_ETHER | IFM_100_TX, 0, NULL);
2859 		ifmedia_add(adapter->media, IFM_ETHER | IFM_100_TX | IFM_FDX, 0, NULL);
2860 		if (adapter->hw.phy.type != e1000_phy_ife) {
2861 			ifmedia_add(adapter->media, IFM_ETHER | IFM_1000_T | IFM_FDX, 0, NULL);
2862 			ifmedia_add(adapter->media, IFM_ETHER | IFM_1000_T, 0, NULL);
2863 		}
2864 	}
2865 	ifmedia_add(adapter->media, IFM_ETHER | IFM_AUTO, 0, NULL);
2866 	ifmedia_set(adapter->media, IFM_ETHER | IFM_AUTO);
2867 	return (0);
2868 }
2869 
2870 static int
2871 em_if_tx_queues_alloc(if_ctx_t ctx, caddr_t *vaddrs, uint64_t *paddrs, int ntxqs, int ntxqsets)
2872 {
2873 	struct adapter *adapter = iflib_get_softc(ctx);
2874 	if_softc_ctx_t scctx = adapter->shared;
2875 	int error = E1000_SUCCESS;
2876 	struct em_tx_queue *que;
2877 	int i, j;
2878 
2879 	MPASS(adapter->tx_num_queues > 0);
2880 	MPASS(adapter->tx_num_queues == ntxqsets);
2881 
2882 	/* First allocate the top level queue structs */
2883 	if (!(adapter->tx_queues =
2884 	    (struct em_tx_queue *) malloc(sizeof(struct em_tx_queue) *
2885 	    adapter->tx_num_queues, M_DEVBUF, M_NOWAIT | M_ZERO))) {
2886 		device_printf(iflib_get_dev(ctx), "Unable to allocate queue memory\n");
2887 		return(ENOMEM);
2888 	}
2889 
2890 	for (i = 0, que = adapter->tx_queues; i < adapter->tx_num_queues; i++, que++) {
2891 		/* Set up some basics */
2892 
2893 		struct tx_ring *txr = &que->txr;
2894 		txr->adapter = que->adapter = adapter;
2895 		que->me = txr->me =  i;
2896 
2897 		/* Allocate report status array */
2898 		if (!(txr->tx_rsq = (qidx_t *) malloc(sizeof(qidx_t) * scctx->isc_ntxd[0], M_DEVBUF, M_NOWAIT | M_ZERO))) {
2899 			device_printf(iflib_get_dev(ctx), "failed to allocate rs_idxs memory\n");
2900 			error = ENOMEM;
2901 			goto fail;
2902 		}
2903 		for (j = 0; j < scctx->isc_ntxd[0]; j++)
2904 			txr->tx_rsq[j] = QIDX_INVALID;
2905 		/* get the virtual and physical address of the hardware queues */
2906 		txr->tx_base = (struct e1000_tx_desc *)vaddrs[i*ntxqs];
2907 		txr->tx_paddr = paddrs[i*ntxqs];
2908 	}
2909 
2910 	if (bootverbose)
2911 		device_printf(iflib_get_dev(ctx),
2912 		    "allocated for %d tx_queues\n", adapter->tx_num_queues);
2913 	return (0);
2914 fail:
2915 	em_if_queues_free(ctx);
2916 	return (error);
2917 }
2918 
2919 static int
2920 em_if_rx_queues_alloc(if_ctx_t ctx, caddr_t *vaddrs, uint64_t *paddrs, int nrxqs, int nrxqsets)
2921 {
2922 	struct adapter *adapter = iflib_get_softc(ctx);
2923 	int error = E1000_SUCCESS;
2924 	struct em_rx_queue *que;
2925 	int i;
2926 
2927 	MPASS(adapter->rx_num_queues > 0);
2928 	MPASS(adapter->rx_num_queues == nrxqsets);
2929 
2930 	/* First allocate the top level queue structs */
2931 	if (!(adapter->rx_queues =
2932 	    (struct em_rx_queue *) malloc(sizeof(struct em_rx_queue) *
2933 	    adapter->rx_num_queues, M_DEVBUF, M_NOWAIT | M_ZERO))) {
2934 		device_printf(iflib_get_dev(ctx), "Unable to allocate queue memory\n");
2935 		error = ENOMEM;
2936 		goto fail;
2937 	}
2938 
2939 	for (i = 0, que = adapter->rx_queues; i < nrxqsets; i++, que++) {
2940 		/* Set up some basics */
2941 		struct rx_ring *rxr = &que->rxr;
2942 		rxr->adapter = que->adapter = adapter;
2943 		rxr->que = que;
2944 		que->me = rxr->me =  i;
2945 
2946 		/* get the virtual and physical address of the hardware queues */
2947 		rxr->rx_base = (union e1000_rx_desc_extended *)vaddrs[i*nrxqs];
2948 		rxr->rx_paddr = paddrs[i*nrxqs];
2949 	}
2950 
2951 	if (bootverbose)
2952 		device_printf(iflib_get_dev(ctx),
2953 		    "allocated for %d rx_queues\n", adapter->rx_num_queues);
2954 
2955 	return (0);
2956 fail:
2957 	em_if_queues_free(ctx);
2958 	return (error);
2959 }
2960 
2961 static void
2962 em_if_queues_free(if_ctx_t ctx)
2963 {
2964 	struct adapter *adapter = iflib_get_softc(ctx);
2965 	struct em_tx_queue *tx_que = adapter->tx_queues;
2966 	struct em_rx_queue *rx_que = adapter->rx_queues;
2967 
2968 	if (tx_que != NULL) {
2969 		for (int i = 0; i < adapter->tx_num_queues; i++, tx_que++) {
2970 			struct tx_ring *txr = &tx_que->txr;
2971 			if (txr->tx_rsq == NULL)
2972 				break;
2973 
2974 			free(txr->tx_rsq, M_DEVBUF);
2975 			txr->tx_rsq = NULL;
2976 		}
2977 		free(adapter->tx_queues, M_DEVBUF);
2978 		adapter->tx_queues = NULL;
2979 	}
2980 
2981 	if (rx_que != NULL) {
2982 		free(adapter->rx_queues, M_DEVBUF);
2983 		adapter->rx_queues = NULL;
2984 	}
2985 
2986 	em_release_hw_control(adapter);
2987 
2988 	if (adapter->mta != NULL) {
2989 		free(adapter->mta, M_DEVBUF);
2990 	}
2991 }
2992 
2993 /*********************************************************************
2994  *
2995  *  Enable transmit unit.
2996  *
2997  **********************************************************************/
2998 static void
2999 em_initialize_transmit_unit(if_ctx_t ctx)
3000 {
3001 	struct adapter *adapter = iflib_get_softc(ctx);
3002 	if_softc_ctx_t scctx = adapter->shared;
3003 	struct em_tx_queue *que;
3004 	struct tx_ring	*txr;
3005 	struct e1000_hw	*hw = &adapter->hw;
3006 	u32 tctl, txdctl = 0, tarc, tipg = 0;
3007 
3008 	INIT_DEBUGOUT("em_initialize_transmit_unit: begin");
3009 
3010 	for (int i = 0; i < adapter->tx_num_queues; i++, txr++) {
3011 		u64 bus_addr;
3012 		caddr_t offp, endp;
3013 
3014 		que = &adapter->tx_queues[i];
3015 		txr = &que->txr;
3016 		bus_addr = txr->tx_paddr;
3017 
3018 		/* Clear checksum offload context. */
3019 		offp = (caddr_t)&txr->csum_flags;
3020 		endp = (caddr_t)(txr + 1);
3021 		bzero(offp, endp - offp);
3022 
3023 		/* Base and Len of TX Ring */
3024 		E1000_WRITE_REG(hw, E1000_TDLEN(i),
3025 		    scctx->isc_ntxd[0] * sizeof(struct e1000_tx_desc));
3026 		E1000_WRITE_REG(hw, E1000_TDBAH(i),
3027 		    (u32)(bus_addr >> 32));
3028 		E1000_WRITE_REG(hw, E1000_TDBAL(i),
3029 		    (u32)bus_addr);
3030 		/* Init the HEAD/TAIL indices */
3031 		E1000_WRITE_REG(hw, E1000_TDT(i), 0);
3032 		E1000_WRITE_REG(hw, E1000_TDH(i), 0);
3033 
3034 		HW_DEBUGOUT2("Base = %x, Length = %x\n",
3035 		    E1000_READ_REG(&adapter->hw, E1000_TDBAL(i)),
3036 		    E1000_READ_REG(&adapter->hw, E1000_TDLEN(i)));
3037 
3038 		txdctl = 0; /* clear txdctl */
3039 		txdctl |= 0x1f; /* PTHRESH */
3040 		txdctl |= 1 << 8; /* HTHRESH */
3041 		txdctl |= 1 << 16;/* WTHRESH */
3042 		txdctl |= 1 << 22; /* Reserved bit 22 must always be 1 */
3043 		txdctl |= E1000_TXDCTL_GRAN;
3044 		txdctl |= 1 << 25; /* LWTHRESH */
3045 
3046 		E1000_WRITE_REG(hw, E1000_TXDCTL(i), txdctl);
3047 	}
3048 
3049 	/* Set the default values for the Tx Inter Packet Gap timer */
3050 	switch (adapter->hw.mac.type) {
3051 	case e1000_80003es2lan:
3052 		tipg = DEFAULT_82543_TIPG_IPGR1;
3053 		tipg |= DEFAULT_80003ES2LAN_TIPG_IPGR2 <<
3054 		    E1000_TIPG_IPGR2_SHIFT;
3055 		break;
3056 	case e1000_82542:
3057 		tipg = DEFAULT_82542_TIPG_IPGT;
3058 		tipg |= DEFAULT_82542_TIPG_IPGR1 << E1000_TIPG_IPGR1_SHIFT;
3059 		tipg |= DEFAULT_82542_TIPG_IPGR2 << E1000_TIPG_IPGR2_SHIFT;
3060 		break;
3061 	default:
3062 		if ((adapter->hw.phy.media_type == e1000_media_type_fiber) ||
3063 		    (adapter->hw.phy.media_type ==
3064 		    e1000_media_type_internal_serdes))
3065 			tipg = DEFAULT_82543_TIPG_IPGT_FIBER;
3066 		else
3067 			tipg = DEFAULT_82543_TIPG_IPGT_COPPER;
3068 		tipg |= DEFAULT_82543_TIPG_IPGR1 << E1000_TIPG_IPGR1_SHIFT;
3069 		tipg |= DEFAULT_82543_TIPG_IPGR2 << E1000_TIPG_IPGR2_SHIFT;
3070 	}
3071 
3072 	E1000_WRITE_REG(&adapter->hw, E1000_TIPG, tipg);
3073 	E1000_WRITE_REG(&adapter->hw, E1000_TIDV, adapter->tx_int_delay.value);
3074 
3075 	if(adapter->hw.mac.type >= e1000_82540)
3076 		E1000_WRITE_REG(&adapter->hw, E1000_TADV,
3077 		    adapter->tx_abs_int_delay.value);
3078 
3079 	if ((adapter->hw.mac.type == e1000_82571) ||
3080 	    (adapter->hw.mac.type == e1000_82572)) {
3081 		tarc = E1000_READ_REG(&adapter->hw, E1000_TARC(0));
3082 		tarc |= TARC_SPEED_MODE_BIT;
3083 		E1000_WRITE_REG(&adapter->hw, E1000_TARC(0), tarc);
3084 	} else if (adapter->hw.mac.type == e1000_80003es2lan) {
3085 		/* errata: program both queues to unweighted RR */
3086 		tarc = E1000_READ_REG(&adapter->hw, E1000_TARC(0));
3087 		tarc |= 1;
3088 		E1000_WRITE_REG(&adapter->hw, E1000_TARC(0), tarc);
3089 		tarc = E1000_READ_REG(&adapter->hw, E1000_TARC(1));
3090 		tarc |= 1;
3091 		E1000_WRITE_REG(&adapter->hw, E1000_TARC(1), tarc);
3092 	} else if (adapter->hw.mac.type == e1000_82574) {
3093 		tarc = E1000_READ_REG(&adapter->hw, E1000_TARC(0));
3094 		tarc |= TARC_ERRATA_BIT;
3095 		if ( adapter->tx_num_queues > 1) {
3096 			tarc |= (TARC_COMPENSATION_MODE | TARC_MQ_FIX);
3097 			E1000_WRITE_REG(&adapter->hw, E1000_TARC(0), tarc);
3098 			E1000_WRITE_REG(&adapter->hw, E1000_TARC(1), tarc);
3099 		} else
3100 			E1000_WRITE_REG(&adapter->hw, E1000_TARC(0), tarc);
3101 	}
3102 
3103 	if (adapter->tx_int_delay.value > 0)
3104 		adapter->txd_cmd |= E1000_TXD_CMD_IDE;
3105 
3106 	/* Program the Transmit Control Register */
3107 	tctl = E1000_READ_REG(&adapter->hw, E1000_TCTL);
3108 	tctl &= ~E1000_TCTL_CT;
3109 	tctl |= (E1000_TCTL_PSP | E1000_TCTL_RTLC | E1000_TCTL_EN |
3110 		   (E1000_COLLISION_THRESHOLD << E1000_CT_SHIFT));
3111 
3112 	if (adapter->hw.mac.type >= e1000_82571)
3113 		tctl |= E1000_TCTL_MULR;
3114 
3115 	/* This write will effectively turn on the transmit unit. */
3116 	E1000_WRITE_REG(&adapter->hw, E1000_TCTL, tctl);
3117 
3118 	/* SPT and KBL errata workarounds */
3119 	if (hw->mac.type == e1000_pch_spt) {
3120 		u32 reg;
3121 		reg = E1000_READ_REG(hw, E1000_IOSFPC);
3122 		reg |= E1000_RCTL_RDMTS_HEX;
3123 		E1000_WRITE_REG(hw, E1000_IOSFPC, reg);
3124 		/* i218-i219 Specification Update 1.5.4.5 */
3125 		reg = E1000_READ_REG(hw, E1000_TARC(0));
3126 		reg &= ~E1000_TARC0_CB_MULTIQ_3_REQ;
3127 		reg |= E1000_TARC0_CB_MULTIQ_2_REQ;
3128 		E1000_WRITE_REG(hw, E1000_TARC(0), reg);
3129 	}
3130 }
3131 
3132 /*********************************************************************
3133  *
3134  *  Enable receive unit.
3135  *
3136  **********************************************************************/
3137 
3138 static void
3139 em_initialize_receive_unit(if_ctx_t ctx)
3140 {
3141 	struct adapter *adapter = iflib_get_softc(ctx);
3142 	if_softc_ctx_t scctx = adapter->shared;
3143 	struct ifnet *ifp = iflib_get_ifp(ctx);
3144 	struct e1000_hw	*hw = &adapter->hw;
3145 	struct em_rx_queue *que;
3146 	int i;
3147 	u32 rctl, rxcsum, rfctl;
3148 
3149 	INIT_DEBUGOUT("em_initialize_receive_units: begin");
3150 
3151 	/*
3152 	 * Make sure receives are disabled while setting
3153 	 * up the descriptor ring
3154 	 */
3155 	rctl = E1000_READ_REG(hw, E1000_RCTL);
3156 	/* Do not disable if ever enabled on this hardware */
3157 	if ((hw->mac.type != e1000_82574) && (hw->mac.type != e1000_82583))
3158 		E1000_WRITE_REG(hw, E1000_RCTL, rctl & ~E1000_RCTL_EN);
3159 
3160 	/* Setup the Receive Control Register */
3161 	rctl &= ~(3 << E1000_RCTL_MO_SHIFT);
3162 	rctl |= E1000_RCTL_EN | E1000_RCTL_BAM |
3163 	    E1000_RCTL_LBM_NO | E1000_RCTL_RDMTS_HALF |
3164 	    (hw->mac.mc_filter_type << E1000_RCTL_MO_SHIFT);
3165 
3166 	/* Do not store bad packets */
3167 	rctl &= ~E1000_RCTL_SBP;
3168 
3169 	/* Enable Long Packet receive */
3170 	if (if_getmtu(ifp) > ETHERMTU)
3171 		rctl |= E1000_RCTL_LPE;
3172 	else
3173 		rctl &= ~E1000_RCTL_LPE;
3174 
3175 	/* Strip the CRC */
3176 	if (!em_disable_crc_stripping)
3177 		rctl |= E1000_RCTL_SECRC;
3178 
3179 	if (adapter->hw.mac.type >= e1000_82540) {
3180 		E1000_WRITE_REG(&adapter->hw, E1000_RADV,
3181 			    adapter->rx_abs_int_delay.value);
3182 
3183 		/*
3184 		 * Set the interrupt throttling rate. Value is calculated
3185 		 * as DEFAULT_ITR = 1/(MAX_INTS_PER_SEC * 256ns)
3186 		 */
3187 		E1000_WRITE_REG(hw, E1000_ITR, DEFAULT_ITR);
3188 	}
3189 	E1000_WRITE_REG(&adapter->hw, E1000_RDTR,
3190 	    adapter->rx_int_delay.value);
3191 
3192 	/* Use extended rx descriptor formats */
3193 	rfctl = E1000_READ_REG(hw, E1000_RFCTL);
3194 	rfctl |= E1000_RFCTL_EXTEN;
3195 	/*
3196 	 * When using MSI-X interrupts we need to throttle
3197 	 * using the EITR register (82574 only)
3198 	 */
3199 	if (hw->mac.type == e1000_82574) {
3200 		for (int i = 0; i < 4; i++)
3201 			E1000_WRITE_REG(hw, E1000_EITR_82574(i),
3202 			    DEFAULT_ITR);
3203 		/* Disable accelerated acknowledge */
3204 		rfctl |= E1000_RFCTL_ACK_DIS;
3205 	}
3206 	E1000_WRITE_REG(hw, E1000_RFCTL, rfctl);
3207 
3208 	rxcsum = E1000_READ_REG(hw, E1000_RXCSUM);
3209 	if (if_getcapenable(ifp) & IFCAP_RXCSUM &&
3210 	    adapter->hw.mac.type >= e1000_82543) {
3211 		if (adapter->tx_num_queues > 1) {
3212 			if (adapter->hw.mac.type >= igb_mac_min) {
3213 				rxcsum |= E1000_RXCSUM_PCSD;
3214 				if (hw->mac.type != e1000_82575)
3215 					rxcsum |= E1000_RXCSUM_CRCOFL;
3216 			} else
3217 				rxcsum |= E1000_RXCSUM_TUOFL |
3218 					E1000_RXCSUM_IPOFL |
3219 					E1000_RXCSUM_PCSD;
3220 		} else {
3221 			if (adapter->hw.mac.type >= igb_mac_min)
3222 				rxcsum |= E1000_RXCSUM_IPPCSE;
3223 			else
3224 				rxcsum |= E1000_RXCSUM_TUOFL | E1000_RXCSUM_IPOFL;
3225 			if (adapter->hw.mac.type > e1000_82575)
3226 				rxcsum |= E1000_RXCSUM_CRCOFL;
3227 		}
3228 	} else
3229 		rxcsum &= ~E1000_RXCSUM_TUOFL;
3230 
3231 	E1000_WRITE_REG(hw, E1000_RXCSUM, rxcsum);
3232 
3233 	if (adapter->rx_num_queues > 1) {
3234 		if (adapter->hw.mac.type >= igb_mac_min)
3235 			igb_initialize_rss_mapping(adapter);
3236 		else
3237 			em_initialize_rss_mapping(adapter);
3238 	}
3239 
3240 	/*
3241 	 * XXX TEMPORARY WORKAROUND: on some systems with 82573
3242 	 * long latencies are observed, like Lenovo X60. This
3243 	 * change eliminates the problem, but since having positive
3244 	 * values in RDTR is a known source of problems on other
3245 	 * platforms another solution is being sought.
3246 	 */
3247 	if (hw->mac.type == e1000_82573)
3248 		E1000_WRITE_REG(hw, E1000_RDTR, 0x20);
3249 
3250 	for (i = 0, que = adapter->rx_queues; i < adapter->rx_num_queues; i++, que++) {
3251 		struct rx_ring *rxr = &que->rxr;
3252 		/* Setup the Base and Length of the Rx Descriptor Ring */
3253 		u64 bus_addr = rxr->rx_paddr;
3254 #if 0
3255 		u32 rdt = adapter->rx_num_queues -1;  /* default */
3256 #endif
3257 
3258 		E1000_WRITE_REG(hw, E1000_RDLEN(i),
3259 		    scctx->isc_nrxd[0] * sizeof(union e1000_rx_desc_extended));
3260 		E1000_WRITE_REG(hw, E1000_RDBAH(i), (u32)(bus_addr >> 32));
3261 		E1000_WRITE_REG(hw, E1000_RDBAL(i), (u32)bus_addr);
3262 		/* Setup the Head and Tail Descriptor Pointers */
3263 		E1000_WRITE_REG(hw, E1000_RDH(i), 0);
3264 		E1000_WRITE_REG(hw, E1000_RDT(i), 0);
3265 	}
3266 
3267 	/*
3268 	 * Set PTHRESH for improved jumbo performance
3269 	 * According to 10.2.5.11 of Intel 82574 Datasheet,
3270 	 * RXDCTL(1) is written whenever RXDCTL(0) is written.
3271 	 * Only write to RXDCTL(1) if there is a need for different
3272 	 * settings.
3273 	 */
3274 
3275 	if (((adapter->hw.mac.type == e1000_ich9lan) ||
3276 	    (adapter->hw.mac.type == e1000_pch2lan) ||
3277 	    (adapter->hw.mac.type == e1000_ich10lan)) &&
3278 	    (if_getmtu(ifp) > ETHERMTU)) {
3279 		u32 rxdctl = E1000_READ_REG(hw, E1000_RXDCTL(0));
3280 		E1000_WRITE_REG(hw, E1000_RXDCTL(0), rxdctl | 3);
3281 	} else if (adapter->hw.mac.type == e1000_82574) {
3282 		for (int i = 0; i < adapter->rx_num_queues; i++) {
3283 			u32 rxdctl = E1000_READ_REG(hw, E1000_RXDCTL(i));
3284 			rxdctl |= 0x20; /* PTHRESH */
3285 			rxdctl |= 4 << 8; /* HTHRESH */
3286 			rxdctl |= 4 << 16;/* WTHRESH */
3287 			rxdctl |= 1 << 24; /* Switch to granularity */
3288 			E1000_WRITE_REG(hw, E1000_RXDCTL(i), rxdctl);
3289 		}
3290 	} else if (adapter->hw.mac.type >= igb_mac_min) {
3291 		u32 psize, srrctl = 0;
3292 
3293 		if (if_getmtu(ifp) > ETHERMTU) {
3294 			/* Set maximum packet len */
3295 			if (adapter->rx_mbuf_sz <= 4096) {
3296 				srrctl |= 4096 >> E1000_SRRCTL_BSIZEPKT_SHIFT;
3297 				rctl |= E1000_RCTL_SZ_4096 | E1000_RCTL_BSEX;
3298 			} else if (adapter->rx_mbuf_sz > 4096) {
3299 				srrctl |= 8192 >> E1000_SRRCTL_BSIZEPKT_SHIFT;
3300 				rctl |= E1000_RCTL_SZ_8192 | E1000_RCTL_BSEX;
3301 			}
3302 			psize = scctx->isc_max_frame_size;
3303 			/* are we on a vlan? */
3304 			if (ifp->if_vlantrunk != NULL)
3305 				psize += VLAN_TAG_SIZE;
3306 			E1000_WRITE_REG(&adapter->hw, E1000_RLPML, psize);
3307 		} else {
3308 			srrctl |= 2048 >> E1000_SRRCTL_BSIZEPKT_SHIFT;
3309 			rctl |= E1000_RCTL_SZ_2048;
3310 		}
3311 
3312 		/*
3313 		 * If TX flow control is disabled and there's >1 queue defined,
3314 		 * enable DROP.
3315 		 *
3316 		 * This drops frames rather than hanging the RX MAC for all queues.
3317 		 */
3318 		if ((adapter->rx_num_queues > 1) &&
3319 		    (adapter->fc == e1000_fc_none ||
3320 		     adapter->fc == e1000_fc_rx_pause)) {
3321 			srrctl |= E1000_SRRCTL_DROP_EN;
3322 		}
3323 			/* Setup the Base and Length of the Rx Descriptor Rings */
3324 		for (i = 0, que = adapter->rx_queues; i < adapter->rx_num_queues; i++, que++) {
3325 			struct rx_ring *rxr = &que->rxr;
3326 			u64 bus_addr = rxr->rx_paddr;
3327 			u32 rxdctl;
3328 
3329 #ifdef notyet
3330 			/* Configure for header split? -- ignore for now */
3331 			rxr->hdr_split = igb_header_split;
3332 #else
3333 			srrctl |= E1000_SRRCTL_DESCTYPE_ADV_ONEBUF;
3334 #endif
3335 
3336 			E1000_WRITE_REG(hw, E1000_RDLEN(i),
3337 					scctx->isc_nrxd[0] * sizeof(struct e1000_rx_desc));
3338 			E1000_WRITE_REG(hw, E1000_RDBAH(i),
3339 					(uint32_t)(bus_addr >> 32));
3340 			E1000_WRITE_REG(hw, E1000_RDBAL(i),
3341 					(uint32_t)bus_addr);
3342 			E1000_WRITE_REG(hw, E1000_SRRCTL(i), srrctl);
3343 			/* Enable this Queue */
3344 			rxdctl = E1000_READ_REG(hw, E1000_RXDCTL(i));
3345 			rxdctl |= E1000_RXDCTL_QUEUE_ENABLE;
3346 			rxdctl &= 0xFFF00000;
3347 			rxdctl |= IGB_RX_PTHRESH;
3348 			rxdctl |= IGB_RX_HTHRESH << 8;
3349 			rxdctl |= IGB_RX_WTHRESH << 16;
3350 			E1000_WRITE_REG(hw, E1000_RXDCTL(i), rxdctl);
3351 		}
3352 	} else if (adapter->hw.mac.type >= e1000_pch2lan) {
3353 		if (if_getmtu(ifp) > ETHERMTU)
3354 			e1000_lv_jumbo_workaround_ich8lan(hw, TRUE);
3355 		else
3356 			e1000_lv_jumbo_workaround_ich8lan(hw, FALSE);
3357 	}
3358 
3359 	/* Make sure VLAN Filters are off */
3360 	rctl &= ~E1000_RCTL_VFE;
3361 
3362 	if (adapter->hw.mac.type < igb_mac_min) {
3363 		if (adapter->rx_mbuf_sz == MCLBYTES)
3364 			rctl |= E1000_RCTL_SZ_2048;
3365 		else if (adapter->rx_mbuf_sz == MJUMPAGESIZE)
3366 			rctl |= E1000_RCTL_SZ_4096 | E1000_RCTL_BSEX;
3367 		else if (adapter->rx_mbuf_sz > MJUMPAGESIZE)
3368 			rctl |= E1000_RCTL_SZ_8192 | E1000_RCTL_BSEX;
3369 
3370 		/* ensure we clear use DTYPE of 00 here */
3371 		rctl &= ~0x00000C00;
3372 	}
3373 
3374 	/* Write out the settings */
3375 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
3376 
3377 	return;
3378 }
3379 
3380 static void
3381 em_if_vlan_register(if_ctx_t ctx, u16 vtag)
3382 {
3383 	struct adapter *adapter = iflib_get_softc(ctx);
3384 	u32 index, bit;
3385 
3386 	index = (vtag >> 5) & 0x7F;
3387 	bit = vtag & 0x1F;
3388 	adapter->shadow_vfta[index] |= (1 << bit);
3389 	++adapter->num_vlans;
3390 }
3391 
3392 static void
3393 em_if_vlan_unregister(if_ctx_t ctx, u16 vtag)
3394 {
3395 	struct adapter *adapter = iflib_get_softc(ctx);
3396 	u32 index, bit;
3397 
3398 	index = (vtag >> 5) & 0x7F;
3399 	bit = vtag & 0x1F;
3400 	adapter->shadow_vfta[index] &= ~(1 << bit);
3401 	--adapter->num_vlans;
3402 }
3403 
3404 static void
3405 em_setup_vlan_hw_support(struct adapter *adapter)
3406 {
3407 	struct e1000_hw *hw = &adapter->hw;
3408 	u32 reg;
3409 
3410 	/*
3411 	 * We get here thru init_locked, meaning
3412 	 * a soft reset, this has already cleared
3413 	 * the VFTA and other state, so if there
3414 	 * have been no vlan's registered do nothing.
3415 	 */
3416 	if (adapter->num_vlans == 0)
3417 		return;
3418 
3419 	/*
3420 	 * A soft reset zero's out the VFTA, so
3421 	 * we need to repopulate it now.
3422 	 */
3423 	for (int i = 0; i < EM_VFTA_SIZE; i++)
3424 		if (adapter->shadow_vfta[i] != 0)
3425 			E1000_WRITE_REG_ARRAY(hw, E1000_VFTA,
3426 			    i, adapter->shadow_vfta[i]);
3427 
3428 	reg = E1000_READ_REG(hw, E1000_CTRL);
3429 	reg |= E1000_CTRL_VME;
3430 	E1000_WRITE_REG(hw, E1000_CTRL, reg);
3431 
3432 	/* Enable the Filter Table */
3433 	reg = E1000_READ_REG(hw, E1000_RCTL);
3434 	reg &= ~E1000_RCTL_CFIEN;
3435 	reg |= E1000_RCTL_VFE;
3436 	E1000_WRITE_REG(hw, E1000_RCTL, reg);
3437 }
3438 
3439 static void
3440 em_if_intr_enable(if_ctx_t ctx)
3441 {
3442 	struct adapter *adapter = iflib_get_softc(ctx);
3443 	struct e1000_hw *hw = &adapter->hw;
3444 	u32 ims_mask = IMS_ENABLE_MASK;
3445 
3446 	if (hw->mac.type == e1000_82574) {
3447 		E1000_WRITE_REG(hw, EM_EIAC, EM_MSIX_MASK);
3448 		ims_mask |= adapter->ims;
3449 	}
3450 	E1000_WRITE_REG(hw, E1000_IMS, ims_mask);
3451 }
3452 
3453 static void
3454 em_if_intr_disable(if_ctx_t ctx)
3455 {
3456 	struct adapter *adapter = iflib_get_softc(ctx);
3457 	struct e1000_hw *hw = &adapter->hw;
3458 
3459 	if (hw->mac.type == e1000_82574)
3460 		E1000_WRITE_REG(hw, EM_EIAC, 0);
3461 	E1000_WRITE_REG(hw, E1000_IMC, 0xffffffff);
3462 }
3463 
3464 static void
3465 igb_if_intr_enable(if_ctx_t ctx)
3466 {
3467 	struct adapter *adapter = iflib_get_softc(ctx);
3468 	struct e1000_hw *hw = &adapter->hw;
3469 	u32 mask;
3470 
3471 	if (__predict_true(adapter->intr_type == IFLIB_INTR_MSIX)) {
3472 		mask = (adapter->que_mask | adapter->link_mask);
3473 		E1000_WRITE_REG(hw, E1000_EIAC, mask);
3474 		E1000_WRITE_REG(hw, E1000_EIAM, mask);
3475 		E1000_WRITE_REG(hw, E1000_EIMS, mask);
3476 		E1000_WRITE_REG(hw, E1000_IMS, E1000_IMS_LSC);
3477 	} else
3478 		E1000_WRITE_REG(hw, E1000_IMS, IMS_ENABLE_MASK);
3479 	E1000_WRITE_FLUSH(hw);
3480 }
3481 
3482 static void
3483 igb_if_intr_disable(if_ctx_t ctx)
3484 {
3485 	struct adapter *adapter = iflib_get_softc(ctx);
3486 	struct e1000_hw *hw = &adapter->hw;
3487 
3488 	if (__predict_true(adapter->intr_type == IFLIB_INTR_MSIX)) {
3489 		E1000_WRITE_REG(hw, E1000_EIMC, 0xffffffff);
3490 		E1000_WRITE_REG(hw, E1000_EIAC, 0);
3491 	}
3492 	E1000_WRITE_REG(hw, E1000_IMC, 0xffffffff);
3493 	E1000_WRITE_FLUSH(hw);
3494 }
3495 
3496 /*
3497  * Bit of a misnomer, what this really means is
3498  * to enable OS management of the system... aka
3499  * to disable special hardware management features
3500  */
3501 static void
3502 em_init_manageability(struct adapter *adapter)
3503 {
3504 	/* A shared code workaround */
3505 #define E1000_82542_MANC2H E1000_MANC2H
3506 	if (adapter->has_manage) {
3507 		int manc2h = E1000_READ_REG(&adapter->hw, E1000_MANC2H);
3508 		int manc = E1000_READ_REG(&adapter->hw, E1000_MANC);
3509 
3510 		/* disable hardware interception of ARP */
3511 		manc &= ~(E1000_MANC_ARP_EN);
3512 
3513 		/* enable receiving management packets to the host */
3514 		manc |= E1000_MANC_EN_MNG2HOST;
3515 #define E1000_MNG2HOST_PORT_623 (1 << 5)
3516 #define E1000_MNG2HOST_PORT_664 (1 << 6)
3517 		manc2h |= E1000_MNG2HOST_PORT_623;
3518 		manc2h |= E1000_MNG2HOST_PORT_664;
3519 		E1000_WRITE_REG(&adapter->hw, E1000_MANC2H, manc2h);
3520 		E1000_WRITE_REG(&adapter->hw, E1000_MANC, manc);
3521 	}
3522 }
3523 
3524 /*
3525  * Give control back to hardware management
3526  * controller if there is one.
3527  */
3528 static void
3529 em_release_manageability(struct adapter *adapter)
3530 {
3531 	if (adapter->has_manage) {
3532 		int manc = E1000_READ_REG(&adapter->hw, E1000_MANC);
3533 
3534 		/* re-enable hardware interception of ARP */
3535 		manc |= E1000_MANC_ARP_EN;
3536 		manc &= ~E1000_MANC_EN_MNG2HOST;
3537 
3538 		E1000_WRITE_REG(&adapter->hw, E1000_MANC, manc);
3539 	}
3540 }
3541 
3542 /*
3543  * em_get_hw_control sets the {CTRL_EXT|FWSM}:DRV_LOAD bit.
3544  * For ASF and Pass Through versions of f/w this means
3545  * that the driver is loaded. For AMT version type f/w
3546  * this means that the network i/f is open.
3547  */
3548 static void
3549 em_get_hw_control(struct adapter *adapter)
3550 {
3551 	u32 ctrl_ext, swsm;
3552 
3553 	if (adapter->vf_ifp)
3554 		return;
3555 
3556 	if (adapter->hw.mac.type == e1000_82573) {
3557 		swsm = E1000_READ_REG(&adapter->hw, E1000_SWSM);
3558 		E1000_WRITE_REG(&adapter->hw, E1000_SWSM,
3559 		    swsm | E1000_SWSM_DRV_LOAD);
3560 		return;
3561 	}
3562 	/* else */
3563 	ctrl_ext = E1000_READ_REG(&adapter->hw, E1000_CTRL_EXT);
3564 	E1000_WRITE_REG(&adapter->hw, E1000_CTRL_EXT,
3565 	    ctrl_ext | E1000_CTRL_EXT_DRV_LOAD);
3566 }
3567 
3568 /*
3569  * em_release_hw_control resets {CTRL_EXT|FWSM}:DRV_LOAD bit.
3570  * For ASF and Pass Through versions of f/w this means that
3571  * the driver is no longer loaded. For AMT versions of the
3572  * f/w this means that the network i/f is closed.
3573  */
3574 static void
3575 em_release_hw_control(struct adapter *adapter)
3576 {
3577 	u32 ctrl_ext, swsm;
3578 
3579 	if (!adapter->has_manage)
3580 		return;
3581 
3582 	if (adapter->hw.mac.type == e1000_82573) {
3583 		swsm = E1000_READ_REG(&adapter->hw, E1000_SWSM);
3584 		E1000_WRITE_REG(&adapter->hw, E1000_SWSM,
3585 		    swsm & ~E1000_SWSM_DRV_LOAD);
3586 		return;
3587 	}
3588 	/* else */
3589 	ctrl_ext = E1000_READ_REG(&adapter->hw, E1000_CTRL_EXT);
3590 	E1000_WRITE_REG(&adapter->hw, E1000_CTRL_EXT,
3591 	    ctrl_ext & ~E1000_CTRL_EXT_DRV_LOAD);
3592 	return;
3593 }
3594 
3595 static int
3596 em_is_valid_ether_addr(u8 *addr)
3597 {
3598 	char zero_addr[6] = { 0, 0, 0, 0, 0, 0 };
3599 
3600 	if ((addr[0] & 1) || (!bcmp(addr, zero_addr, ETHER_ADDR_LEN))) {
3601 		return (FALSE);
3602 	}
3603 
3604 	return (TRUE);
3605 }
3606 
3607 /*
3608 ** Parse the interface capabilities with regard
3609 ** to both system management and wake-on-lan for
3610 ** later use.
3611 */
3612 static void
3613 em_get_wakeup(if_ctx_t ctx)
3614 {
3615 	struct adapter *adapter = iflib_get_softc(ctx);
3616 	device_t dev = iflib_get_dev(ctx);
3617 	u16 eeprom_data = 0, device_id, apme_mask;
3618 
3619 	adapter->has_manage = e1000_enable_mng_pass_thru(&adapter->hw);
3620 	apme_mask = EM_EEPROM_APME;
3621 
3622 	switch (adapter->hw.mac.type) {
3623 	case e1000_82542:
3624 	case e1000_82543:
3625 		break;
3626 	case e1000_82544:
3627 		e1000_read_nvm(&adapter->hw,
3628 		    NVM_INIT_CONTROL2_REG, 1, &eeprom_data);
3629 		apme_mask = EM_82544_APME;
3630 		break;
3631 	case e1000_82546:
3632 	case e1000_82546_rev_3:
3633 		if (adapter->hw.bus.func == 1) {
3634 			e1000_read_nvm(&adapter->hw,
3635 			    NVM_INIT_CONTROL3_PORT_B, 1, &eeprom_data);
3636 			break;
3637 		} else
3638 			e1000_read_nvm(&adapter->hw,
3639 			    NVM_INIT_CONTROL3_PORT_A, 1, &eeprom_data);
3640 		break;
3641 	case e1000_82573:
3642 	case e1000_82583:
3643 		adapter->has_amt = TRUE;
3644 		/* FALLTHROUGH */
3645 	case e1000_82571:
3646 	case e1000_82572:
3647 	case e1000_80003es2lan:
3648 		if (adapter->hw.bus.func == 1) {
3649 			e1000_read_nvm(&adapter->hw,
3650 			    NVM_INIT_CONTROL3_PORT_B, 1, &eeprom_data);
3651 			break;
3652 		} else
3653 			e1000_read_nvm(&adapter->hw,
3654 			    NVM_INIT_CONTROL3_PORT_A, 1, &eeprom_data);
3655 		break;
3656 	case e1000_ich8lan:
3657 	case e1000_ich9lan:
3658 	case e1000_ich10lan:
3659 	case e1000_pchlan:
3660 	case e1000_pch2lan:
3661 	case e1000_pch_lpt:
3662 	case e1000_pch_spt:
3663 	case e1000_82575:	/* listing all igb devices */
3664 	case e1000_82576:
3665 	case e1000_82580:
3666 	case e1000_i350:
3667 	case e1000_i354:
3668 	case e1000_i210:
3669 	case e1000_i211:
3670 	case e1000_vfadapt:
3671 	case e1000_vfadapt_i350:
3672 		apme_mask = E1000_WUC_APME;
3673 		adapter->has_amt = TRUE;
3674 		eeprom_data = E1000_READ_REG(&adapter->hw, E1000_WUC);
3675 		break;
3676 	default:
3677 		e1000_read_nvm(&adapter->hw,
3678 		    NVM_INIT_CONTROL3_PORT_A, 1, &eeprom_data);
3679 		break;
3680 	}
3681 	if (eeprom_data & apme_mask)
3682 		adapter->wol = (E1000_WUFC_MAG | E1000_WUFC_MC);
3683 	/*
3684 	 * We have the eeprom settings, now apply the special cases
3685 	 * where the eeprom may be wrong or the board won't support
3686 	 * wake on lan on a particular port
3687 	 */
3688 	device_id = pci_get_device(dev);
3689 	switch (device_id) {
3690 	case E1000_DEV_ID_82546GB_PCIE:
3691 		adapter->wol = 0;
3692 		break;
3693 	case E1000_DEV_ID_82546EB_FIBER:
3694 	case E1000_DEV_ID_82546GB_FIBER:
3695 		/* Wake events only supported on port A for dual fiber
3696 		 * regardless of eeprom setting */
3697 		if (E1000_READ_REG(&adapter->hw, E1000_STATUS) &
3698 		    E1000_STATUS_FUNC_1)
3699 			adapter->wol = 0;
3700 		break;
3701 	case E1000_DEV_ID_82546GB_QUAD_COPPER_KSP3:
3702 		/* if quad port adapter, disable WoL on all but port A */
3703 		if (global_quad_port_a != 0)
3704 			adapter->wol = 0;
3705 		/* Reset for multiple quad port adapters */
3706 		if (++global_quad_port_a == 4)
3707 			global_quad_port_a = 0;
3708 		break;
3709 	case E1000_DEV_ID_82571EB_FIBER:
3710 		/* Wake events only supported on port A for dual fiber
3711 		 * regardless of eeprom setting */
3712 		if (E1000_READ_REG(&adapter->hw, E1000_STATUS) &
3713 		    E1000_STATUS_FUNC_1)
3714 			adapter->wol = 0;
3715 		break;
3716 	case E1000_DEV_ID_82571EB_QUAD_COPPER:
3717 	case E1000_DEV_ID_82571EB_QUAD_FIBER:
3718 	case E1000_DEV_ID_82571EB_QUAD_COPPER_LP:
3719 		/* if quad port adapter, disable WoL on all but port A */
3720 		if (global_quad_port_a != 0)
3721 			adapter->wol = 0;
3722 		/* Reset for multiple quad port adapters */
3723 		if (++global_quad_port_a == 4)
3724 			global_quad_port_a = 0;
3725 		break;
3726 	}
3727 	return;
3728 }
3729 
3730 
3731 /*
3732  * Enable PCI Wake On Lan capability
3733  */
3734 static void
3735 em_enable_wakeup(if_ctx_t ctx)
3736 {
3737 	struct adapter *adapter = iflib_get_softc(ctx);
3738 	device_t dev = iflib_get_dev(ctx);
3739 	if_t ifp = iflib_get_ifp(ctx);
3740 	int error = 0;
3741 	u32 pmc, ctrl, ctrl_ext, rctl;
3742 	u16 status;
3743 
3744 	if (pci_find_cap(dev, PCIY_PMG, &pmc) != 0)
3745 		return;
3746 
3747 	/*
3748 	 * Determine type of Wakeup: note that wol
3749 	 * is set with all bits on by default.
3750 	 */
3751 	if ((if_getcapenable(ifp) & IFCAP_WOL_MAGIC) == 0)
3752 		adapter->wol &= ~E1000_WUFC_MAG;
3753 
3754 	if ((if_getcapenable(ifp) & IFCAP_WOL_UCAST) == 0)
3755 		adapter->wol &= ~E1000_WUFC_EX;
3756 
3757 	if ((if_getcapenable(ifp) & IFCAP_WOL_MCAST) == 0)
3758 		adapter->wol &= ~E1000_WUFC_MC;
3759 	else {
3760 		rctl = E1000_READ_REG(&adapter->hw, E1000_RCTL);
3761 		rctl |= E1000_RCTL_MPE;
3762 		E1000_WRITE_REG(&adapter->hw, E1000_RCTL, rctl);
3763 	}
3764 
3765 	if (!(adapter->wol & (E1000_WUFC_EX | E1000_WUFC_MAG | E1000_WUFC_MC)))
3766 		goto pme;
3767 
3768 	/* Advertise the wakeup capability */
3769 	ctrl = E1000_READ_REG(&adapter->hw, E1000_CTRL);
3770 	ctrl |= (E1000_CTRL_SWDPIN2 | E1000_CTRL_SWDPIN3);
3771 	E1000_WRITE_REG(&adapter->hw, E1000_CTRL, ctrl);
3772 
3773 	/* Keep the laser running on Fiber adapters */
3774 	if (adapter->hw.phy.media_type == e1000_media_type_fiber ||
3775 	    adapter->hw.phy.media_type == e1000_media_type_internal_serdes) {
3776 		ctrl_ext = E1000_READ_REG(&adapter->hw, E1000_CTRL_EXT);
3777 		ctrl_ext |= E1000_CTRL_EXT_SDP3_DATA;
3778 		E1000_WRITE_REG(&adapter->hw, E1000_CTRL_EXT, ctrl_ext);
3779 	}
3780 
3781 	if ((adapter->hw.mac.type == e1000_ich8lan) ||
3782 	    (adapter->hw.mac.type == e1000_pchlan) ||
3783 	    (adapter->hw.mac.type == e1000_ich9lan) ||
3784 	    (adapter->hw.mac.type == e1000_ich10lan))
3785 		e1000_suspend_workarounds_ich8lan(&adapter->hw);
3786 
3787 	if ( adapter->hw.mac.type >= e1000_pchlan) {
3788 		error = em_enable_phy_wakeup(adapter);
3789 		if (error)
3790 			goto pme;
3791 	} else {
3792 		/* Enable wakeup by the MAC */
3793 		E1000_WRITE_REG(&adapter->hw, E1000_WUC, E1000_WUC_PME_EN);
3794 		E1000_WRITE_REG(&adapter->hw, E1000_WUFC, adapter->wol);
3795 	}
3796 
3797 	if (adapter->hw.phy.type == e1000_phy_igp_3)
3798 		e1000_igp3_phy_powerdown_workaround_ich8lan(&adapter->hw);
3799 
3800 pme:
3801 	status = pci_read_config(dev, pmc + PCIR_POWER_STATUS, 2);
3802 	status &= ~(PCIM_PSTAT_PME | PCIM_PSTAT_PMEENABLE);
3803 	if (!error && (if_getcapenable(ifp) & IFCAP_WOL))
3804 		status |= PCIM_PSTAT_PME | PCIM_PSTAT_PMEENABLE;
3805 	pci_write_config(dev, pmc + PCIR_POWER_STATUS, status, 2);
3806 
3807 	return;
3808 }
3809 
3810 /*
3811  * WOL in the newer chipset interfaces (pchlan)
3812  * require thing to be copied into the phy
3813  */
3814 static int
3815 em_enable_phy_wakeup(struct adapter *adapter)
3816 {
3817 	struct e1000_hw *hw = &adapter->hw;
3818 	u32 mreg, ret = 0;
3819 	u16 preg;
3820 
3821 	/* copy MAC RARs to PHY RARs */
3822 	e1000_copy_rx_addrs_to_phy_ich8lan(hw);
3823 
3824 	/* copy MAC MTA to PHY MTA */
3825 	for (int i = 0; i < adapter->hw.mac.mta_reg_count; i++) {
3826 		mreg = E1000_READ_REG_ARRAY(hw, E1000_MTA, i);
3827 		e1000_write_phy_reg(hw, BM_MTA(i), (u16)(mreg & 0xFFFF));
3828 		e1000_write_phy_reg(hw, BM_MTA(i) + 1,
3829 		    (u16)((mreg >> 16) & 0xFFFF));
3830 	}
3831 
3832 	/* configure PHY Rx Control register */
3833 	e1000_read_phy_reg(&adapter->hw, BM_RCTL, &preg);
3834 	mreg = E1000_READ_REG(hw, E1000_RCTL);
3835 	if (mreg & E1000_RCTL_UPE)
3836 		preg |= BM_RCTL_UPE;
3837 	if (mreg & E1000_RCTL_MPE)
3838 		preg |= BM_RCTL_MPE;
3839 	preg &= ~(BM_RCTL_MO_MASK);
3840 	if (mreg & E1000_RCTL_MO_3)
3841 		preg |= (((mreg & E1000_RCTL_MO_3) >> E1000_RCTL_MO_SHIFT)
3842 				<< BM_RCTL_MO_SHIFT);
3843 	if (mreg & E1000_RCTL_BAM)
3844 		preg |= BM_RCTL_BAM;
3845 	if (mreg & E1000_RCTL_PMCF)
3846 		preg |= BM_RCTL_PMCF;
3847 	mreg = E1000_READ_REG(hw, E1000_CTRL);
3848 	if (mreg & E1000_CTRL_RFCE)
3849 		preg |= BM_RCTL_RFCE;
3850 	e1000_write_phy_reg(&adapter->hw, BM_RCTL, preg);
3851 
3852 	/* enable PHY wakeup in MAC register */
3853 	E1000_WRITE_REG(hw, E1000_WUC,
3854 	    E1000_WUC_PHY_WAKE | E1000_WUC_PME_EN | E1000_WUC_APME);
3855 	E1000_WRITE_REG(hw, E1000_WUFC, adapter->wol);
3856 
3857 	/* configure and enable PHY wakeup in PHY registers */
3858 	e1000_write_phy_reg(&adapter->hw, BM_WUFC, adapter->wol);
3859 	e1000_write_phy_reg(&adapter->hw, BM_WUC, E1000_WUC_PME_EN);
3860 
3861 	/* activate PHY wakeup */
3862 	ret = hw->phy.ops.acquire(hw);
3863 	if (ret) {
3864 		printf("Could not acquire PHY\n");
3865 		return ret;
3866 	}
3867 	e1000_write_phy_reg_mdic(hw, IGP01E1000_PHY_PAGE_SELECT,
3868 	                         (BM_WUC_ENABLE_PAGE << IGP_PAGE_SHIFT));
3869 	ret = e1000_read_phy_reg_mdic(hw, BM_WUC_ENABLE_REG, &preg);
3870 	if (ret) {
3871 		printf("Could not read PHY page 769\n");
3872 		goto out;
3873 	}
3874 	preg |= BM_WUC_ENABLE_BIT | BM_WUC_HOST_WU_BIT;
3875 	ret = e1000_write_phy_reg_mdic(hw, BM_WUC_ENABLE_REG, preg);
3876 	if (ret)
3877 		printf("Could not set PHY Host Wakeup bit\n");
3878 out:
3879 	hw->phy.ops.release(hw);
3880 
3881 	return ret;
3882 }
3883 
3884 static void
3885 em_if_led_func(if_ctx_t ctx, int onoff)
3886 {
3887 	struct adapter *adapter = iflib_get_softc(ctx);
3888 
3889 	if (onoff) {
3890 		e1000_setup_led(&adapter->hw);
3891 		e1000_led_on(&adapter->hw);
3892 	} else {
3893 		e1000_led_off(&adapter->hw);
3894 		e1000_cleanup_led(&adapter->hw);
3895 	}
3896 }
3897 
3898 /*
3899  * Disable the L0S and L1 LINK states
3900  */
3901 static void
3902 em_disable_aspm(struct adapter *adapter)
3903 {
3904 	int base, reg;
3905 	u16 link_cap,link_ctrl;
3906 	device_t dev = adapter->dev;
3907 
3908 	switch (adapter->hw.mac.type) {
3909 	case e1000_82573:
3910 	case e1000_82574:
3911 	case e1000_82583:
3912 		break;
3913 	default:
3914 		return;
3915 	}
3916 	if (pci_find_cap(dev, PCIY_EXPRESS, &base) != 0)
3917 		return;
3918 	reg = base + PCIER_LINK_CAP;
3919 	link_cap = pci_read_config(dev, reg, 2);
3920 	if ((link_cap & PCIEM_LINK_CAP_ASPM) == 0)
3921 		return;
3922 	reg = base + PCIER_LINK_CTL;
3923 	link_ctrl = pci_read_config(dev, reg, 2);
3924 	link_ctrl &= ~PCIEM_LINK_CTL_ASPMC;
3925 	pci_write_config(dev, reg, link_ctrl, 2);
3926 	return;
3927 }
3928 
3929 /**********************************************************************
3930  *
3931  *  Update the board statistics counters.
3932  *
3933  **********************************************************************/
3934 static void
3935 em_update_stats_counters(struct adapter *adapter)
3936 {
3937 	u64 prev_xoffrxc = adapter->stats.xoffrxc;
3938 
3939 	if(adapter->hw.phy.media_type == e1000_media_type_copper ||
3940 	   (E1000_READ_REG(&adapter->hw, E1000_STATUS) & E1000_STATUS_LU)) {
3941 		adapter->stats.symerrs += E1000_READ_REG(&adapter->hw, E1000_SYMERRS);
3942 		adapter->stats.sec += E1000_READ_REG(&adapter->hw, E1000_SEC);
3943 	}
3944 	adapter->stats.crcerrs += E1000_READ_REG(&adapter->hw, E1000_CRCERRS);
3945 	adapter->stats.mpc += E1000_READ_REG(&adapter->hw, E1000_MPC);
3946 	adapter->stats.scc += E1000_READ_REG(&adapter->hw, E1000_SCC);
3947 	adapter->stats.ecol += E1000_READ_REG(&adapter->hw, E1000_ECOL);
3948 
3949 	adapter->stats.mcc += E1000_READ_REG(&adapter->hw, E1000_MCC);
3950 	adapter->stats.latecol += E1000_READ_REG(&adapter->hw, E1000_LATECOL);
3951 	adapter->stats.colc += E1000_READ_REG(&adapter->hw, E1000_COLC);
3952 	adapter->stats.dc += E1000_READ_REG(&adapter->hw, E1000_DC);
3953 	adapter->stats.rlec += E1000_READ_REG(&adapter->hw, E1000_RLEC);
3954 	adapter->stats.xonrxc += E1000_READ_REG(&adapter->hw, E1000_XONRXC);
3955 	adapter->stats.xontxc += E1000_READ_REG(&adapter->hw, E1000_XONTXC);
3956 	adapter->stats.xoffrxc += E1000_READ_REG(&adapter->hw, E1000_XOFFRXC);
3957 	/*
3958 	 ** For watchdog management we need to know if we have been
3959 	 ** paused during the last interval, so capture that here.
3960 	*/
3961 	if (adapter->stats.xoffrxc != prev_xoffrxc)
3962 		adapter->shared->isc_pause_frames = 1;
3963 	adapter->stats.xofftxc += E1000_READ_REG(&adapter->hw, E1000_XOFFTXC);
3964 	adapter->stats.fcruc += E1000_READ_REG(&adapter->hw, E1000_FCRUC);
3965 	adapter->stats.prc64 += E1000_READ_REG(&adapter->hw, E1000_PRC64);
3966 	adapter->stats.prc127 += E1000_READ_REG(&adapter->hw, E1000_PRC127);
3967 	adapter->stats.prc255 += E1000_READ_REG(&adapter->hw, E1000_PRC255);
3968 	adapter->stats.prc511 += E1000_READ_REG(&adapter->hw, E1000_PRC511);
3969 	adapter->stats.prc1023 += E1000_READ_REG(&adapter->hw, E1000_PRC1023);
3970 	adapter->stats.prc1522 += E1000_READ_REG(&adapter->hw, E1000_PRC1522);
3971 	adapter->stats.gprc += E1000_READ_REG(&adapter->hw, E1000_GPRC);
3972 	adapter->stats.bprc += E1000_READ_REG(&adapter->hw, E1000_BPRC);
3973 	adapter->stats.mprc += E1000_READ_REG(&adapter->hw, E1000_MPRC);
3974 	adapter->stats.gptc += E1000_READ_REG(&adapter->hw, E1000_GPTC);
3975 
3976 	/* For the 64-bit byte counters the low dword must be read first. */
3977 	/* Both registers clear on the read of the high dword */
3978 
3979 	adapter->stats.gorc += E1000_READ_REG(&adapter->hw, E1000_GORCL) +
3980 	    ((u64)E1000_READ_REG(&adapter->hw, E1000_GORCH) << 32);
3981 	adapter->stats.gotc += E1000_READ_REG(&adapter->hw, E1000_GOTCL) +
3982 	    ((u64)E1000_READ_REG(&adapter->hw, E1000_GOTCH) << 32);
3983 
3984 	adapter->stats.rnbc += E1000_READ_REG(&adapter->hw, E1000_RNBC);
3985 	adapter->stats.ruc += E1000_READ_REG(&adapter->hw, E1000_RUC);
3986 	adapter->stats.rfc += E1000_READ_REG(&adapter->hw, E1000_RFC);
3987 	adapter->stats.roc += E1000_READ_REG(&adapter->hw, E1000_ROC);
3988 	adapter->stats.rjc += E1000_READ_REG(&adapter->hw, E1000_RJC);
3989 
3990 	adapter->stats.tor += E1000_READ_REG(&adapter->hw, E1000_TORH);
3991 	adapter->stats.tot += E1000_READ_REG(&adapter->hw, E1000_TOTH);
3992 
3993 	adapter->stats.tpr += E1000_READ_REG(&adapter->hw, E1000_TPR);
3994 	adapter->stats.tpt += E1000_READ_REG(&adapter->hw, E1000_TPT);
3995 	adapter->stats.ptc64 += E1000_READ_REG(&adapter->hw, E1000_PTC64);
3996 	adapter->stats.ptc127 += E1000_READ_REG(&adapter->hw, E1000_PTC127);
3997 	adapter->stats.ptc255 += E1000_READ_REG(&adapter->hw, E1000_PTC255);
3998 	adapter->stats.ptc511 += E1000_READ_REG(&adapter->hw, E1000_PTC511);
3999 	adapter->stats.ptc1023 += E1000_READ_REG(&adapter->hw, E1000_PTC1023);
4000 	adapter->stats.ptc1522 += E1000_READ_REG(&adapter->hw, E1000_PTC1522);
4001 	adapter->stats.mptc += E1000_READ_REG(&adapter->hw, E1000_MPTC);
4002 	adapter->stats.bptc += E1000_READ_REG(&adapter->hw, E1000_BPTC);
4003 
4004 	/* Interrupt Counts */
4005 
4006 	adapter->stats.iac += E1000_READ_REG(&adapter->hw, E1000_IAC);
4007 	adapter->stats.icrxptc += E1000_READ_REG(&adapter->hw, E1000_ICRXPTC);
4008 	adapter->stats.icrxatc += E1000_READ_REG(&adapter->hw, E1000_ICRXATC);
4009 	adapter->stats.ictxptc += E1000_READ_REG(&adapter->hw, E1000_ICTXPTC);
4010 	adapter->stats.ictxatc += E1000_READ_REG(&adapter->hw, E1000_ICTXATC);
4011 	adapter->stats.ictxqec += E1000_READ_REG(&adapter->hw, E1000_ICTXQEC);
4012 	adapter->stats.ictxqmtc += E1000_READ_REG(&adapter->hw, E1000_ICTXQMTC);
4013 	adapter->stats.icrxdmtc += E1000_READ_REG(&adapter->hw, E1000_ICRXDMTC);
4014 	adapter->stats.icrxoc += E1000_READ_REG(&adapter->hw, E1000_ICRXOC);
4015 
4016 	if (adapter->hw.mac.type >= e1000_82543) {
4017 		adapter->stats.algnerrc +=
4018 		E1000_READ_REG(&adapter->hw, E1000_ALGNERRC);
4019 		adapter->stats.rxerrc +=
4020 		E1000_READ_REG(&adapter->hw, E1000_RXERRC);
4021 		adapter->stats.tncrs +=
4022 		E1000_READ_REG(&adapter->hw, E1000_TNCRS);
4023 		adapter->stats.cexterr +=
4024 		E1000_READ_REG(&adapter->hw, E1000_CEXTERR);
4025 		adapter->stats.tsctc +=
4026 		E1000_READ_REG(&adapter->hw, E1000_TSCTC);
4027 		adapter->stats.tsctfc +=
4028 		E1000_READ_REG(&adapter->hw, E1000_TSCTFC);
4029 	}
4030 }
4031 
4032 static uint64_t
4033 em_if_get_counter(if_ctx_t ctx, ift_counter cnt)
4034 {
4035 	struct adapter *adapter = iflib_get_softc(ctx);
4036 	struct ifnet *ifp = iflib_get_ifp(ctx);
4037 
4038 	switch (cnt) {
4039 	case IFCOUNTER_COLLISIONS:
4040 		return (adapter->stats.colc);
4041 	case IFCOUNTER_IERRORS:
4042 		return (adapter->dropped_pkts + adapter->stats.rxerrc +
4043 		    adapter->stats.crcerrs + adapter->stats.algnerrc +
4044 		    adapter->stats.ruc + adapter->stats.roc +
4045 		    adapter->stats.mpc + adapter->stats.cexterr);
4046 	case IFCOUNTER_OERRORS:
4047 		return (adapter->stats.ecol + adapter->stats.latecol +
4048 		    adapter->watchdog_events);
4049 	default:
4050 		return (if_get_counter_default(ifp, cnt));
4051 	}
4052 }
4053 
4054 /* em_if_needs_restart - Tell iflib when the driver needs to be reinitialized
4055  * @ctx: iflib context
4056  * @event: event code to check
4057  *
4058  * Defaults to returning true for unknown events.
4059  *
4060  * @returns true if iflib needs to reinit the interface
4061  */
4062 static bool
4063 em_if_needs_restart(if_ctx_t ctx __unused, enum iflib_restart_event event)
4064 {
4065 	switch (event) {
4066 	case IFLIB_RESTART_VLAN_CONFIG:
4067 	default:
4068 		return (true);
4069 	}
4070 }
4071 
4072 /* Export a single 32-bit register via a read-only sysctl. */
4073 static int
4074 em_sysctl_reg_handler(SYSCTL_HANDLER_ARGS)
4075 {
4076 	struct adapter *adapter;
4077 	u_int val;
4078 
4079 	adapter = oidp->oid_arg1;
4080 	val = E1000_READ_REG(&adapter->hw, oidp->oid_arg2);
4081 	return (sysctl_handle_int(oidp, &val, 0, req));
4082 }
4083 
4084 /*
4085  * Add sysctl variables, one per statistic, to the system.
4086  */
4087 static void
4088 em_add_hw_stats(struct adapter *adapter)
4089 {
4090 	device_t dev = iflib_get_dev(adapter->ctx);
4091 	struct em_tx_queue *tx_que = adapter->tx_queues;
4092 	struct em_rx_queue *rx_que = adapter->rx_queues;
4093 
4094 	struct sysctl_ctx_list *ctx = device_get_sysctl_ctx(dev);
4095 	struct sysctl_oid *tree = device_get_sysctl_tree(dev);
4096 	struct sysctl_oid_list *child = SYSCTL_CHILDREN(tree);
4097 	struct e1000_hw_stats *stats = &adapter->stats;
4098 
4099 	struct sysctl_oid *stat_node, *queue_node, *int_node;
4100 	struct sysctl_oid_list *stat_list, *queue_list, *int_list;
4101 
4102 #define QUEUE_NAME_LEN 32
4103 	char namebuf[QUEUE_NAME_LEN];
4104 
4105 	/* Driver Statistics */
4106 	SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "dropped",
4107 			CTLFLAG_RD, &adapter->dropped_pkts,
4108 			"Driver dropped packets");
4109 	SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "link_irq",
4110 			CTLFLAG_RD, &adapter->link_irq,
4111 			"Link MSI-X IRQ Handled");
4112 	SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "rx_overruns",
4113 			CTLFLAG_RD, &adapter->rx_overruns,
4114 			"RX overruns");
4115 	SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "watchdog_timeouts",
4116 			CTLFLAG_RD, &adapter->watchdog_events,
4117 			"Watchdog timeouts");
4118 	SYSCTL_ADD_PROC(ctx, child, OID_AUTO, "device_control",
4119 	    CTLTYPE_UINT | CTLFLAG_RD | CTLFLAG_NEEDGIANT,
4120 	    adapter, E1000_CTRL, em_sysctl_reg_handler, "IU",
4121 	    "Device Control Register");
4122 	SYSCTL_ADD_PROC(ctx, child, OID_AUTO, "rx_control",
4123 	    CTLTYPE_UINT | CTLFLAG_RD | CTLFLAG_NEEDGIANT,
4124 	    adapter, E1000_RCTL, em_sysctl_reg_handler, "IU",
4125 	    "Receiver Control Register");
4126 	SYSCTL_ADD_UINT(ctx, child, OID_AUTO, "fc_high_water",
4127 			CTLFLAG_RD, &adapter->hw.fc.high_water, 0,
4128 			"Flow Control High Watermark");
4129 	SYSCTL_ADD_UINT(ctx, child, OID_AUTO, "fc_low_water",
4130 			CTLFLAG_RD, &adapter->hw.fc.low_water, 0,
4131 			"Flow Control Low Watermark");
4132 
4133 	for (int i = 0; i < adapter->tx_num_queues; i++, tx_que++) {
4134 		struct tx_ring *txr = &tx_que->txr;
4135 		snprintf(namebuf, QUEUE_NAME_LEN, "queue_tx_%d", i);
4136 		queue_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, namebuf,
4137 		    CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, "TX Queue Name");
4138 		queue_list = SYSCTL_CHILDREN(queue_node);
4139 
4140 		SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "txd_head",
4141 		    CTLTYPE_UINT | CTLFLAG_RD | CTLFLAG_NEEDGIANT, adapter,
4142 		    E1000_TDH(txr->me), em_sysctl_reg_handler, "IU",
4143 		    "Transmit Descriptor Head");
4144 		SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "txd_tail",
4145 		    CTLTYPE_UINT | CTLFLAG_RD | CTLFLAG_NEEDGIANT, adapter,
4146 		    E1000_TDT(txr->me), em_sysctl_reg_handler, "IU",
4147 		    "Transmit Descriptor Tail");
4148 		SYSCTL_ADD_ULONG(ctx, queue_list, OID_AUTO, "tx_irq",
4149 				CTLFLAG_RD, &txr->tx_irq,
4150 				"Queue MSI-X Transmit Interrupts");
4151 	}
4152 
4153 	for (int j = 0; j < adapter->rx_num_queues; j++, rx_que++) {
4154 		struct rx_ring *rxr = &rx_que->rxr;
4155 		snprintf(namebuf, QUEUE_NAME_LEN, "queue_rx_%d", j);
4156 		queue_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, namebuf,
4157 		    CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, "RX Queue Name");
4158 		queue_list = SYSCTL_CHILDREN(queue_node);
4159 
4160 		SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "rxd_head",
4161 		    CTLTYPE_UINT | CTLFLAG_RD | CTLFLAG_NEEDGIANT, adapter,
4162 		    E1000_RDH(rxr->me), em_sysctl_reg_handler, "IU",
4163 		    "Receive Descriptor Head");
4164 		SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "rxd_tail",
4165 		    CTLTYPE_UINT | CTLFLAG_RD | CTLFLAG_NEEDGIANT, adapter,
4166 		    E1000_RDT(rxr->me), em_sysctl_reg_handler, "IU",
4167 		    "Receive Descriptor Tail");
4168 		SYSCTL_ADD_ULONG(ctx, queue_list, OID_AUTO, "rx_irq",
4169 				CTLFLAG_RD, &rxr->rx_irq,
4170 				"Queue MSI-X Receive Interrupts");
4171 	}
4172 
4173 	/* MAC stats get their own sub node */
4174 
4175 	stat_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, "mac_stats",
4176 	    CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, "Statistics");
4177 	stat_list = SYSCTL_CHILDREN(stat_node);
4178 
4179 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "excess_coll",
4180 			CTLFLAG_RD, &stats->ecol,
4181 			"Excessive collisions");
4182 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "single_coll",
4183 			CTLFLAG_RD, &stats->scc,
4184 			"Single collisions");
4185 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "multiple_coll",
4186 			CTLFLAG_RD, &stats->mcc,
4187 			"Multiple collisions");
4188 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "late_coll",
4189 			CTLFLAG_RD, &stats->latecol,
4190 			"Late collisions");
4191 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "collision_count",
4192 			CTLFLAG_RD, &stats->colc,
4193 			"Collision Count");
4194 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "symbol_errors",
4195 			CTLFLAG_RD, &adapter->stats.symerrs,
4196 			"Symbol Errors");
4197 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "sequence_errors",
4198 			CTLFLAG_RD, &adapter->stats.sec,
4199 			"Sequence Errors");
4200 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "defer_count",
4201 			CTLFLAG_RD, &adapter->stats.dc,
4202 			"Defer Count");
4203 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "missed_packets",
4204 			CTLFLAG_RD, &adapter->stats.mpc,
4205 			"Missed Packets");
4206 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_no_buff",
4207 			CTLFLAG_RD, &adapter->stats.rnbc,
4208 			"Receive No Buffers");
4209 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_undersize",
4210 			CTLFLAG_RD, &adapter->stats.ruc,
4211 			"Receive Undersize");
4212 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_fragmented",
4213 			CTLFLAG_RD, &adapter->stats.rfc,
4214 			"Fragmented Packets Received ");
4215 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_oversize",
4216 			CTLFLAG_RD, &adapter->stats.roc,
4217 			"Oversized Packets Received");
4218 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_jabber",
4219 			CTLFLAG_RD, &adapter->stats.rjc,
4220 			"Recevied Jabber");
4221 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_errs",
4222 			CTLFLAG_RD, &adapter->stats.rxerrc,
4223 			"Receive Errors");
4224 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "crc_errs",
4225 			CTLFLAG_RD, &adapter->stats.crcerrs,
4226 			"CRC errors");
4227 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "alignment_errs",
4228 			CTLFLAG_RD, &adapter->stats.algnerrc,
4229 			"Alignment Errors");
4230 	/* On 82575 these are collision counts */
4231 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "coll_ext_errs",
4232 			CTLFLAG_RD, &adapter->stats.cexterr,
4233 			"Collision/Carrier extension errors");
4234 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xon_recvd",
4235 			CTLFLAG_RD, &adapter->stats.xonrxc,
4236 			"XON Received");
4237 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xon_txd",
4238 			CTLFLAG_RD, &adapter->stats.xontxc,
4239 			"XON Transmitted");
4240 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xoff_recvd",
4241 			CTLFLAG_RD, &adapter->stats.xoffrxc,
4242 			"XOFF Received");
4243 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xoff_txd",
4244 			CTLFLAG_RD, &adapter->stats.xofftxc,
4245 			"XOFF Transmitted");
4246 
4247 	/* Packet Reception Stats */
4248 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "total_pkts_recvd",
4249 			CTLFLAG_RD, &adapter->stats.tpr,
4250 			"Total Packets Received ");
4251 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_pkts_recvd",
4252 			CTLFLAG_RD, &adapter->stats.gprc,
4253 			"Good Packets Received");
4254 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "bcast_pkts_recvd",
4255 			CTLFLAG_RD, &adapter->stats.bprc,
4256 			"Broadcast Packets Received");
4257 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "mcast_pkts_recvd",
4258 			CTLFLAG_RD, &adapter->stats.mprc,
4259 			"Multicast Packets Received");
4260 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_64",
4261 			CTLFLAG_RD, &adapter->stats.prc64,
4262 			"64 byte frames received ");
4263 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_65_127",
4264 			CTLFLAG_RD, &adapter->stats.prc127,
4265 			"65-127 byte frames received");
4266 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_128_255",
4267 			CTLFLAG_RD, &adapter->stats.prc255,
4268 			"128-255 byte frames received");
4269 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_256_511",
4270 			CTLFLAG_RD, &adapter->stats.prc511,
4271 			"256-511 byte frames received");
4272 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_512_1023",
4273 			CTLFLAG_RD, &adapter->stats.prc1023,
4274 			"512-1023 byte frames received");
4275 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_1024_1522",
4276 			CTLFLAG_RD, &adapter->stats.prc1522,
4277 			"1023-1522 byte frames received");
4278 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_octets_recvd",
4279 			CTLFLAG_RD, &adapter->stats.gorc,
4280 			"Good Octets Received");
4281 
4282 	/* Packet Transmission Stats */
4283 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_octets_txd",
4284 			CTLFLAG_RD, &adapter->stats.gotc,
4285 			"Good Octets Transmitted");
4286 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "total_pkts_txd",
4287 			CTLFLAG_RD, &adapter->stats.tpt,
4288 			"Total Packets Transmitted");
4289 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_pkts_txd",
4290 			CTLFLAG_RD, &adapter->stats.gptc,
4291 			"Good Packets Transmitted");
4292 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "bcast_pkts_txd",
4293 			CTLFLAG_RD, &adapter->stats.bptc,
4294 			"Broadcast Packets Transmitted");
4295 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "mcast_pkts_txd",
4296 			CTLFLAG_RD, &adapter->stats.mptc,
4297 			"Multicast Packets Transmitted");
4298 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_64",
4299 			CTLFLAG_RD, &adapter->stats.ptc64,
4300 			"64 byte frames transmitted ");
4301 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_65_127",
4302 			CTLFLAG_RD, &adapter->stats.ptc127,
4303 			"65-127 byte frames transmitted");
4304 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_128_255",
4305 			CTLFLAG_RD, &adapter->stats.ptc255,
4306 			"128-255 byte frames transmitted");
4307 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_256_511",
4308 			CTLFLAG_RD, &adapter->stats.ptc511,
4309 			"256-511 byte frames transmitted");
4310 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_512_1023",
4311 			CTLFLAG_RD, &adapter->stats.ptc1023,
4312 			"512-1023 byte frames transmitted");
4313 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_1024_1522",
4314 			CTLFLAG_RD, &adapter->stats.ptc1522,
4315 			"1024-1522 byte frames transmitted");
4316 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tso_txd",
4317 			CTLFLAG_RD, &adapter->stats.tsctc,
4318 			"TSO Contexts Transmitted");
4319 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tso_ctx_fail",
4320 			CTLFLAG_RD, &adapter->stats.tsctfc,
4321 			"TSO Contexts Failed");
4322 
4323 
4324 	/* Interrupt Stats */
4325 
4326 	int_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, "interrupts",
4327 	    CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, "Interrupt Statistics");
4328 	int_list = SYSCTL_CHILDREN(int_node);
4329 
4330 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "asserts",
4331 			CTLFLAG_RD, &adapter->stats.iac,
4332 			"Interrupt Assertion Count");
4333 
4334 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "rx_pkt_timer",
4335 			CTLFLAG_RD, &adapter->stats.icrxptc,
4336 			"Interrupt Cause Rx Pkt Timer Expire Count");
4337 
4338 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "rx_abs_timer",
4339 			CTLFLAG_RD, &adapter->stats.icrxatc,
4340 			"Interrupt Cause Rx Abs Timer Expire Count");
4341 
4342 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "tx_pkt_timer",
4343 			CTLFLAG_RD, &adapter->stats.ictxptc,
4344 			"Interrupt Cause Tx Pkt Timer Expire Count");
4345 
4346 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "tx_abs_timer",
4347 			CTLFLAG_RD, &adapter->stats.ictxatc,
4348 			"Interrupt Cause Tx Abs Timer Expire Count");
4349 
4350 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "tx_queue_empty",
4351 			CTLFLAG_RD, &adapter->stats.ictxqec,
4352 			"Interrupt Cause Tx Queue Empty Count");
4353 
4354 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "tx_queue_min_thresh",
4355 			CTLFLAG_RD, &adapter->stats.ictxqmtc,
4356 			"Interrupt Cause Tx Queue Min Thresh Count");
4357 
4358 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "rx_desc_min_thresh",
4359 			CTLFLAG_RD, &adapter->stats.icrxdmtc,
4360 			"Interrupt Cause Rx Desc Min Thresh Count");
4361 
4362 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "rx_overrun",
4363 			CTLFLAG_RD, &adapter->stats.icrxoc,
4364 			"Interrupt Cause Receiver Overrun Count");
4365 }
4366 
4367 /**********************************************************************
4368  *
4369  *  This routine provides a way to dump out the adapter eeprom,
4370  *  often a useful debug/service tool. This only dumps the first
4371  *  32 words, stuff that matters is in that extent.
4372  *
4373  **********************************************************************/
4374 static int
4375 em_sysctl_nvm_info(SYSCTL_HANDLER_ARGS)
4376 {
4377 	struct adapter *adapter = (struct adapter *)arg1;
4378 	int error;
4379 	int result;
4380 
4381 	result = -1;
4382 	error = sysctl_handle_int(oidp, &result, 0, req);
4383 
4384 	if (error || !req->newptr)
4385 		return (error);
4386 
4387 	/*
4388 	 * This value will cause a hex dump of the
4389 	 * first 32 16-bit words of the EEPROM to
4390 	 * the screen.
4391 	 */
4392 	if (result == 1)
4393 		em_print_nvm_info(adapter);
4394 
4395 	return (error);
4396 }
4397 
4398 static void
4399 em_print_nvm_info(struct adapter *adapter)
4400 {
4401 	u16 eeprom_data;
4402 	int i, j, row = 0;
4403 
4404 	/* Its a bit crude, but it gets the job done */
4405 	printf("\nInterface EEPROM Dump:\n");
4406 	printf("Offset\n0x0000  ");
4407 	for (i = 0, j = 0; i < 32; i++, j++) {
4408 		if (j == 8) { /* Make the offset block */
4409 			j = 0; ++row;
4410 			printf("\n0x00%x0  ",row);
4411 		}
4412 		e1000_read_nvm(&adapter->hw, i, 1, &eeprom_data);
4413 		printf("%04x ", eeprom_data);
4414 	}
4415 	printf("\n");
4416 }
4417 
4418 static int
4419 em_sysctl_int_delay(SYSCTL_HANDLER_ARGS)
4420 {
4421 	struct em_int_delay_info *info;
4422 	struct adapter *adapter;
4423 	u32 regval;
4424 	int error, usecs, ticks;
4425 
4426 	info = (struct em_int_delay_info *) arg1;
4427 	usecs = info->value;
4428 	error = sysctl_handle_int(oidp, &usecs, 0, req);
4429 	if (error != 0 || req->newptr == NULL)
4430 		return (error);
4431 	if (usecs < 0 || usecs > EM_TICKS_TO_USECS(65535))
4432 		return (EINVAL);
4433 	info->value = usecs;
4434 	ticks = EM_USECS_TO_TICKS(usecs);
4435 	if (info->offset == E1000_ITR)	/* units are 256ns here */
4436 		ticks *= 4;
4437 
4438 	adapter = info->adapter;
4439 
4440 	regval = E1000_READ_OFFSET(&adapter->hw, info->offset);
4441 	regval = (regval & ~0xffff) | (ticks & 0xffff);
4442 	/* Handle a few special cases. */
4443 	switch (info->offset) {
4444 	case E1000_RDTR:
4445 		break;
4446 	case E1000_TIDV:
4447 		if (ticks == 0) {
4448 			adapter->txd_cmd &= ~E1000_TXD_CMD_IDE;
4449 			/* Don't write 0 into the TIDV register. */
4450 			regval++;
4451 		} else
4452 			adapter->txd_cmd |= E1000_TXD_CMD_IDE;
4453 		break;
4454 	}
4455 	E1000_WRITE_OFFSET(&adapter->hw, info->offset, regval);
4456 	return (0);
4457 }
4458 
4459 static void
4460 em_add_int_delay_sysctl(struct adapter *adapter, const char *name,
4461 	const char *description, struct em_int_delay_info *info,
4462 	int offset, int value)
4463 {
4464 	info->adapter = adapter;
4465 	info->offset = offset;
4466 	info->value = value;
4467 	SYSCTL_ADD_PROC(device_get_sysctl_ctx(adapter->dev),
4468 	    SYSCTL_CHILDREN(device_get_sysctl_tree(adapter->dev)),
4469 	    OID_AUTO, name, CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_NEEDGIANT,
4470 	    info, 0, em_sysctl_int_delay, "I", description);
4471 }
4472 
4473 /*
4474  * Set flow control using sysctl:
4475  * Flow control values:
4476  *      0 - off
4477  *      1 - rx pause
4478  *      2 - tx pause
4479  *      3 - full
4480  */
4481 static int
4482 em_set_flowcntl(SYSCTL_HANDLER_ARGS)
4483 {
4484 	int error;
4485 	static int input = 3; /* default is full */
4486 	struct adapter	*adapter = (struct adapter *) arg1;
4487 
4488 	error = sysctl_handle_int(oidp, &input, 0, req);
4489 
4490 	if ((error) || (req->newptr == NULL))
4491 		return (error);
4492 
4493 	if (input == adapter->fc) /* no change? */
4494 		return (error);
4495 
4496 	switch (input) {
4497 	case e1000_fc_rx_pause:
4498 	case e1000_fc_tx_pause:
4499 	case e1000_fc_full:
4500 	case e1000_fc_none:
4501 		adapter->hw.fc.requested_mode = input;
4502 		adapter->fc = input;
4503 		break;
4504 	default:
4505 		/* Do nothing */
4506 		return (error);
4507 	}
4508 
4509 	adapter->hw.fc.current_mode = adapter->hw.fc.requested_mode;
4510 	e1000_force_mac_fc(&adapter->hw);
4511 	return (error);
4512 }
4513 
4514 /*
4515  * Manage Energy Efficient Ethernet:
4516  * Control values:
4517  *     0/1 - enabled/disabled
4518  */
4519 static int
4520 em_sysctl_eee(SYSCTL_HANDLER_ARGS)
4521 {
4522 	struct adapter *adapter = (struct adapter *) arg1;
4523 	int error, value;
4524 
4525 	value = adapter->hw.dev_spec.ich8lan.eee_disable;
4526 	error = sysctl_handle_int(oidp, &value, 0, req);
4527 	if (error || req->newptr == NULL)
4528 		return (error);
4529 	adapter->hw.dev_spec.ich8lan.eee_disable = (value != 0);
4530 	em_if_init(adapter->ctx);
4531 
4532 	return (0);
4533 }
4534 
4535 static int
4536 em_sysctl_debug_info(SYSCTL_HANDLER_ARGS)
4537 {
4538 	struct adapter *adapter;
4539 	int error;
4540 	int result;
4541 
4542 	result = -1;
4543 	error = sysctl_handle_int(oidp, &result, 0, req);
4544 
4545 	if (error || !req->newptr)
4546 		return (error);
4547 
4548 	if (result == 1) {
4549 		adapter = (struct adapter *) arg1;
4550 		em_print_debug_info(adapter);
4551 	}
4552 
4553 	return (error);
4554 }
4555 
4556 static int
4557 em_get_rs(SYSCTL_HANDLER_ARGS)
4558 {
4559 	struct adapter *adapter = (struct adapter *) arg1;
4560 	int error;
4561 	int result;
4562 
4563 	result = 0;
4564 	error = sysctl_handle_int(oidp, &result, 0, req);
4565 
4566 	if (error || !req->newptr || result != 1)
4567 		return (error);
4568 	em_dump_rs(adapter);
4569 
4570 	return (error);
4571 }
4572 
4573 static void
4574 em_if_debug(if_ctx_t ctx)
4575 {
4576 	em_dump_rs(iflib_get_softc(ctx));
4577 }
4578 
4579 /*
4580  * This routine is meant to be fluid, add whatever is
4581  * needed for debugging a problem.  -jfv
4582  */
4583 static void
4584 em_print_debug_info(struct adapter *adapter)
4585 {
4586 	device_t dev = iflib_get_dev(adapter->ctx);
4587 	struct ifnet *ifp = iflib_get_ifp(adapter->ctx);
4588 	struct tx_ring *txr = &adapter->tx_queues->txr;
4589 	struct rx_ring *rxr = &adapter->rx_queues->rxr;
4590 
4591 	if (if_getdrvflags(ifp) & IFF_DRV_RUNNING)
4592 		printf("Interface is RUNNING ");
4593 	else
4594 		printf("Interface is NOT RUNNING\n");
4595 
4596 	if (if_getdrvflags(ifp) & IFF_DRV_OACTIVE)
4597 		printf("and INACTIVE\n");
4598 	else
4599 		printf("and ACTIVE\n");
4600 
4601 	for (int i = 0; i < adapter->tx_num_queues; i++, txr++) {
4602 		device_printf(dev, "TX Queue %d ------\n", i);
4603 		device_printf(dev, "hw tdh = %d, hw tdt = %d\n",
4604 			E1000_READ_REG(&adapter->hw, E1000_TDH(i)),
4605 			E1000_READ_REG(&adapter->hw, E1000_TDT(i)));
4606 
4607 	}
4608 	for (int j=0; j < adapter->rx_num_queues; j++, rxr++) {
4609 		device_printf(dev, "RX Queue %d ------\n", j);
4610 		device_printf(dev, "hw rdh = %d, hw rdt = %d\n",
4611 			E1000_READ_REG(&adapter->hw, E1000_RDH(j)),
4612 			E1000_READ_REG(&adapter->hw, E1000_RDT(j)));
4613 	}
4614 }
4615 
4616 /*
4617  * 82574 only:
4618  * Write a new value to the EEPROM increasing the number of MSI-X
4619  * vectors from 3 to 5, for proper multiqueue support.
4620  */
4621 static void
4622 em_enable_vectors_82574(if_ctx_t ctx)
4623 {
4624 	struct adapter *adapter = iflib_get_softc(ctx);
4625 	struct e1000_hw *hw = &adapter->hw;
4626 	device_t dev = iflib_get_dev(ctx);
4627 	u16 edata;
4628 
4629 	e1000_read_nvm(hw, EM_NVM_PCIE_CTRL, 1, &edata);
4630 	if (bootverbose)
4631 		device_printf(dev, "EM_NVM_PCIE_CTRL = %#06x\n", edata);
4632 	if (((edata & EM_NVM_MSIX_N_MASK) >> EM_NVM_MSIX_N_SHIFT) != 4) {
4633 		device_printf(dev, "Writing to eeprom: increasing "
4634 		    "reported MSI-X vectors from 3 to 5...\n");
4635 		edata &= ~(EM_NVM_MSIX_N_MASK);
4636 		edata |= 4 << EM_NVM_MSIX_N_SHIFT;
4637 		e1000_write_nvm(hw, EM_NVM_PCIE_CTRL, 1, &edata);
4638 		e1000_update_nvm_checksum(hw);
4639 		device_printf(dev, "Writing to eeprom: done\n");
4640 	}
4641 }
4642