xref: /freebsd/sys/dev/e1000/if_em.c (revision 69c5fa5cd1ec9b09ed88a086607a8a0993818db9)
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 (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 (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 (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 (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 (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 (hw->mac.type < em_mac_min) {
991 		e1000_init_script_state_82541(hw, TRUE);
992 		e1000_set_tbi_compatibility_82543(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 = 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 = 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 		    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(hw, E1000_IMS, EM_MSIX_LINK | E1000_IMS_LSC);
1868 }
1869 
1870 static void
1871 em_if_watchdog_reset(if_ctx_t ctx)
1872 {
1873 	struct adapter *adapter = iflib_get_softc(ctx);
1874 
1875 	/*
1876 	 * Just count the event; iflib(4) will already trigger a
1877 	 * sufficient reset of the controller.
1878 	 */
1879 	adapter->watchdog_events++;
1880 }
1881 
1882 /*********************************************************************
1883  *
1884  *  This routine disables all traffic on the adapter by issuing a
1885  *  global reset on the MAC.
1886  *
1887  **********************************************************************/
1888 static void
1889 em_if_stop(if_ctx_t ctx)
1890 {
1891 	struct adapter *adapter = iflib_get_softc(ctx);
1892 
1893 	INIT_DEBUGOUT("em_if_stop: begin");
1894 
1895 	e1000_reset_hw(&adapter->hw);
1896 	if (adapter->hw.mac.type >= e1000_82544)
1897 		E1000_WRITE_REG(&adapter->hw, E1000_WUFC, 0);
1898 
1899 	e1000_led_off(&adapter->hw);
1900 	e1000_cleanup_led(&adapter->hw);
1901 }
1902 
1903 /*********************************************************************
1904  *
1905  *  Determine hardware revision.
1906  *
1907  **********************************************************************/
1908 static void
1909 em_identify_hardware(if_ctx_t ctx)
1910 {
1911 	device_t dev = iflib_get_dev(ctx);
1912 	struct adapter *adapter = iflib_get_softc(ctx);
1913 
1914 	/* Make sure our PCI config space has the necessary stuff set */
1915 	adapter->hw.bus.pci_cmd_word = pci_read_config(dev, PCIR_COMMAND, 2);
1916 
1917 	/* Save off the information about this board */
1918 	adapter->hw.vendor_id = pci_get_vendor(dev);
1919 	adapter->hw.device_id = pci_get_device(dev);
1920 	adapter->hw.revision_id = pci_read_config(dev, PCIR_REVID, 1);
1921 	adapter->hw.subsystem_vendor_id =
1922 	    pci_read_config(dev, PCIR_SUBVEND_0, 2);
1923 	adapter->hw.subsystem_device_id =
1924 	    pci_read_config(dev, PCIR_SUBDEV_0, 2);
1925 
1926 	/* Do Shared Code Init and Setup */
1927 	if (e1000_set_mac_type(&adapter->hw)) {
1928 		device_printf(dev, "Setup init failure\n");
1929 		return;
1930 	}
1931 }
1932 
1933 static int
1934 em_allocate_pci_resources(if_ctx_t ctx)
1935 {
1936 	struct adapter *adapter = iflib_get_softc(ctx);
1937 	device_t dev = iflib_get_dev(ctx);
1938 	int rid, val;
1939 
1940 	rid = PCIR_BAR(0);
1941 	adapter->memory = bus_alloc_resource_any(dev, SYS_RES_MEMORY,
1942 	    &rid, RF_ACTIVE);
1943 	if (adapter->memory == NULL) {
1944 		device_printf(dev, "Unable to allocate bus resource: memory\n");
1945 		return (ENXIO);
1946 	}
1947 	adapter->osdep.mem_bus_space_tag = rman_get_bustag(adapter->memory);
1948 	adapter->osdep.mem_bus_space_handle =
1949 	    rman_get_bushandle(adapter->memory);
1950 	adapter->hw.hw_addr = (u8 *)&adapter->osdep.mem_bus_space_handle;
1951 
1952 	/* Only older adapters use IO mapping */
1953 	if (adapter->hw.mac.type < em_mac_min &&
1954 	    adapter->hw.mac.type > e1000_82543) {
1955 		/* Figure our where our IO BAR is ? */
1956 		for (rid = PCIR_BAR(0); rid < PCIR_CIS;) {
1957 			val = pci_read_config(dev, rid, 4);
1958 			if (EM_BAR_TYPE(val) == EM_BAR_TYPE_IO) {
1959 				break;
1960 			}
1961 			rid += 4;
1962 			/* check for 64bit BAR */
1963 			if (EM_BAR_MEM_TYPE(val) == EM_BAR_MEM_TYPE_64BIT)
1964 				rid += 4;
1965 		}
1966 		if (rid >= PCIR_CIS) {
1967 			device_printf(dev, "Unable to locate IO BAR\n");
1968 			return (ENXIO);
1969 		}
1970 		adapter->ioport = bus_alloc_resource_any(dev, SYS_RES_IOPORT,
1971 		    &rid, RF_ACTIVE);
1972 		if (adapter->ioport == NULL) {
1973 			device_printf(dev, "Unable to allocate bus resource: "
1974 			    "ioport\n");
1975 			return (ENXIO);
1976 		}
1977 		adapter->hw.io_base = 0;
1978 		adapter->osdep.io_bus_space_tag =
1979 		    rman_get_bustag(adapter->ioport);
1980 		adapter->osdep.io_bus_space_handle =
1981 		    rman_get_bushandle(adapter->ioport);
1982 	}
1983 
1984 	adapter->hw.back = &adapter->osdep;
1985 
1986 	return (0);
1987 }
1988 
1989 /*********************************************************************
1990  *
1991  *  Set up the MSI-X Interrupt handlers
1992  *
1993  **********************************************************************/
1994 static int
1995 em_if_msix_intr_assign(if_ctx_t ctx, int msix)
1996 {
1997 	struct adapter *adapter = iflib_get_softc(ctx);
1998 	struct em_rx_queue *rx_que = adapter->rx_queues;
1999 	struct em_tx_queue *tx_que = adapter->tx_queues;
2000 	int error, rid, i, vector = 0, rx_vectors;
2001 	char buf[16];
2002 
2003 	/* First set up ring resources */
2004 	for (i = 0; i < adapter->rx_num_queues; i++, rx_que++, vector++) {
2005 		rid = vector + 1;
2006 		snprintf(buf, sizeof(buf), "rxq%d", i);
2007 		error = iflib_irq_alloc_generic(ctx, &rx_que->que_irq, rid, IFLIB_INTR_RXTX, em_msix_que, rx_que, rx_que->me, buf);
2008 		if (error) {
2009 			device_printf(iflib_get_dev(ctx), "Failed to allocate que int %d err: %d", i, error);
2010 			adapter->rx_num_queues = i + 1;
2011 			goto fail;
2012 		}
2013 
2014 		rx_que->msix =  vector;
2015 
2016 		/*
2017 		 * Set the bit to enable interrupt
2018 		 * in E1000_IMS -- bits 20 and 21
2019 		 * are for RX0 and RX1, note this has
2020 		 * NOTHING to do with the MSI-X vector
2021 		 */
2022 		if (adapter->hw.mac.type == e1000_82574) {
2023 			rx_que->eims = 1 << (20 + i);
2024 			adapter->ims |= rx_que->eims;
2025 			adapter->ivars |= (8 | rx_que->msix) << (i * 4);
2026 		} else if (adapter->hw.mac.type == e1000_82575)
2027 			rx_que->eims = E1000_EICR_TX_QUEUE0 << vector;
2028 		else
2029 			rx_que->eims = 1 << vector;
2030 	}
2031 	rx_vectors = vector;
2032 
2033 	vector = 0;
2034 	for (i = 0; i < adapter->tx_num_queues; i++, tx_que++, vector++) {
2035 		snprintf(buf, sizeof(buf), "txq%d", i);
2036 		tx_que = &adapter->tx_queues[i];
2037 		iflib_softirq_alloc_generic(ctx,
2038 		    &adapter->rx_queues[i % adapter->rx_num_queues].que_irq,
2039 		    IFLIB_INTR_TX, tx_que, tx_que->me, buf);
2040 
2041 		tx_que->msix = (vector % adapter->rx_num_queues);
2042 
2043 		/*
2044 		 * Set the bit to enable interrupt
2045 		 * in E1000_IMS -- bits 22 and 23
2046 		 * are for TX0 and TX1, note this has
2047 		 * NOTHING to do with the MSI-X vector
2048 		 */
2049 		if (adapter->hw.mac.type == e1000_82574) {
2050 			tx_que->eims = 1 << (22 + i);
2051 			adapter->ims |= tx_que->eims;
2052 			adapter->ivars |= (8 | tx_que->msix) << (8 + (i * 4));
2053 		} else if (adapter->hw.mac.type == e1000_82575) {
2054 			tx_que->eims = E1000_EICR_TX_QUEUE0 << i;
2055 		} else {
2056 			tx_que->eims = 1 << i;
2057 		}
2058 	}
2059 
2060 	/* Link interrupt */
2061 	rid = rx_vectors + 1;
2062 	error = iflib_irq_alloc_generic(ctx, &adapter->irq, rid, IFLIB_INTR_ADMIN, em_msix_link, adapter, 0, "aq");
2063 
2064 	if (error) {
2065 		device_printf(iflib_get_dev(ctx), "Failed to register admin handler");
2066 		goto fail;
2067 	}
2068 	adapter->linkvec = rx_vectors;
2069 	if (adapter->hw.mac.type < igb_mac_min) {
2070 		adapter->ivars |=  (8 | rx_vectors) << 16;
2071 		adapter->ivars |= 0x80000000;
2072 	}
2073 	return (0);
2074 fail:
2075 	iflib_irq_free(ctx, &adapter->irq);
2076 	rx_que = adapter->rx_queues;
2077 	for (int i = 0; i < adapter->rx_num_queues; i++, rx_que++)
2078 		iflib_irq_free(ctx, &rx_que->que_irq);
2079 	return (error);
2080 }
2081 
2082 static void
2083 igb_configure_queues(struct adapter *adapter)
2084 {
2085 	struct e1000_hw *hw = &adapter->hw;
2086 	struct em_rx_queue *rx_que;
2087 	struct em_tx_queue *tx_que;
2088 	u32 tmp, ivar = 0, newitr = 0;
2089 
2090 	/* First turn on RSS capability */
2091 	if (hw->mac.type != e1000_82575)
2092 		E1000_WRITE_REG(hw, E1000_GPIE,
2093 		    E1000_GPIE_MSIX_MODE | E1000_GPIE_EIAME |
2094 		    E1000_GPIE_PBA | E1000_GPIE_NSICR);
2095 
2096 	/* Turn on MSI-X */
2097 	switch (hw->mac.type) {
2098 	case e1000_82580:
2099 	case e1000_i350:
2100 	case e1000_i354:
2101 	case e1000_i210:
2102 	case e1000_i211:
2103 	case e1000_vfadapt:
2104 	case e1000_vfadapt_i350:
2105 		/* RX entries */
2106 		for (int i = 0; i < adapter->rx_num_queues; i++) {
2107 			u32 index = i >> 1;
2108 			ivar = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index);
2109 			rx_que = &adapter->rx_queues[i];
2110 			if (i & 1) {
2111 				ivar &= 0xFF00FFFF;
2112 				ivar |= (rx_que->msix | E1000_IVAR_VALID) << 16;
2113 			} else {
2114 				ivar &= 0xFFFFFF00;
2115 				ivar |= rx_que->msix | E1000_IVAR_VALID;
2116 			}
2117 			E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, ivar);
2118 		}
2119 		/* TX entries */
2120 		for (int i = 0; i < adapter->tx_num_queues; i++) {
2121 			u32 index = i >> 1;
2122 			ivar = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index);
2123 			tx_que = &adapter->tx_queues[i];
2124 			if (i & 1) {
2125 				ivar &= 0x00FFFFFF;
2126 				ivar |= (tx_que->msix | E1000_IVAR_VALID) << 24;
2127 			} else {
2128 				ivar &= 0xFFFF00FF;
2129 				ivar |= (tx_que->msix | E1000_IVAR_VALID) << 8;
2130 			}
2131 			E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, ivar);
2132 			adapter->que_mask |= tx_que->eims;
2133 		}
2134 
2135 		/* And for the link interrupt */
2136 		ivar = (adapter->linkvec | E1000_IVAR_VALID) << 8;
2137 		adapter->link_mask = 1 << adapter->linkvec;
2138 		E1000_WRITE_REG(hw, E1000_IVAR_MISC, ivar);
2139 		break;
2140 	case e1000_82576:
2141 		/* RX entries */
2142 		for (int i = 0; i < adapter->rx_num_queues; i++) {
2143 			u32 index = i & 0x7; /* Each IVAR has two entries */
2144 			ivar = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index);
2145 			rx_que = &adapter->rx_queues[i];
2146 			if (i < 8) {
2147 				ivar &= 0xFFFFFF00;
2148 				ivar |= rx_que->msix | E1000_IVAR_VALID;
2149 			} else {
2150 				ivar &= 0xFF00FFFF;
2151 				ivar |= (rx_que->msix | E1000_IVAR_VALID) << 16;
2152 			}
2153 			E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, ivar);
2154 			adapter->que_mask |= rx_que->eims;
2155 		}
2156 		/* TX entries */
2157 		for (int i = 0; i < adapter->tx_num_queues; i++) {
2158 			u32 index = i & 0x7; /* Each IVAR has two entries */
2159 			ivar = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index);
2160 			tx_que = &adapter->tx_queues[i];
2161 			if (i < 8) {
2162 				ivar &= 0xFFFF00FF;
2163 				ivar |= (tx_que->msix | E1000_IVAR_VALID) << 8;
2164 			} else {
2165 				ivar &= 0x00FFFFFF;
2166 				ivar |= (tx_que->msix | E1000_IVAR_VALID) << 24;
2167 			}
2168 			E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, ivar);
2169 			adapter->que_mask |= tx_que->eims;
2170 		}
2171 
2172 		/* And for the link interrupt */
2173 		ivar = (adapter->linkvec | E1000_IVAR_VALID) << 8;
2174 		adapter->link_mask = 1 << adapter->linkvec;
2175 		E1000_WRITE_REG(hw, E1000_IVAR_MISC, ivar);
2176 		break;
2177 
2178 	case e1000_82575:
2179 		/* enable MSI-X support*/
2180 		tmp = E1000_READ_REG(hw, E1000_CTRL_EXT);
2181 		tmp |= E1000_CTRL_EXT_PBA_CLR;
2182 		/* Auto-Mask interrupts upon ICR read. */
2183 		tmp |= E1000_CTRL_EXT_EIAME;
2184 		tmp |= E1000_CTRL_EXT_IRCA;
2185 		E1000_WRITE_REG(hw, E1000_CTRL_EXT, tmp);
2186 
2187 		/* Queues */
2188 		for (int i = 0; i < adapter->rx_num_queues; i++) {
2189 			rx_que = &adapter->rx_queues[i];
2190 			tmp = E1000_EICR_RX_QUEUE0 << i;
2191 			tmp |= E1000_EICR_TX_QUEUE0 << i;
2192 			rx_que->eims = tmp;
2193 			E1000_WRITE_REG_ARRAY(hw, E1000_MSIXBM(0),
2194 			    i, rx_que->eims);
2195 			adapter->que_mask |= rx_que->eims;
2196 		}
2197 
2198 		/* Link */
2199 		E1000_WRITE_REG(hw, E1000_MSIXBM(adapter->linkvec),
2200 		    E1000_EIMS_OTHER);
2201 		adapter->link_mask |= E1000_EIMS_OTHER;
2202 	default:
2203 		break;
2204 	}
2205 
2206 	/* Set the starting interrupt rate */
2207 	if (em_max_interrupt_rate > 0)
2208 		newitr = (4000000 / em_max_interrupt_rate) & 0x7FFC;
2209 
2210 	if (hw->mac.type == e1000_82575)
2211 		newitr |= newitr << 16;
2212 	else
2213 		newitr |= E1000_EITR_CNT_IGNR;
2214 
2215 	for (int i = 0; i < adapter->rx_num_queues; i++) {
2216 		rx_que = &adapter->rx_queues[i];
2217 		E1000_WRITE_REG(hw, E1000_EITR(rx_que->msix), newitr);
2218 	}
2219 
2220 	return;
2221 }
2222 
2223 static void
2224 em_free_pci_resources(if_ctx_t ctx)
2225 {
2226 	struct adapter *adapter = iflib_get_softc(ctx);
2227 	struct em_rx_queue *que = adapter->rx_queues;
2228 	device_t dev = iflib_get_dev(ctx);
2229 
2230 	/* Release all MSI-X queue resources */
2231 	if (adapter->intr_type == IFLIB_INTR_MSIX)
2232 		iflib_irq_free(ctx, &adapter->irq);
2233 
2234 	if (que != NULL) {
2235 		for (int i = 0; i < adapter->rx_num_queues; i++, que++) {
2236 			iflib_irq_free(ctx, &que->que_irq);
2237 		}
2238 	}
2239 
2240 	if (adapter->memory != NULL) {
2241 		bus_release_resource(dev, SYS_RES_MEMORY,
2242 		    rman_get_rid(adapter->memory), adapter->memory);
2243 		adapter->memory = NULL;
2244 	}
2245 
2246 	if (adapter->flash != NULL) {
2247 		bus_release_resource(dev, SYS_RES_MEMORY,
2248 		    rman_get_rid(adapter->flash), adapter->flash);
2249 		adapter->flash = NULL;
2250 	}
2251 
2252 	if (adapter->ioport != NULL) {
2253 		bus_release_resource(dev, SYS_RES_IOPORT,
2254 		    rman_get_rid(adapter->ioport), adapter->ioport);
2255 		adapter->ioport = NULL;
2256 	}
2257 }
2258 
2259 /* Set up MSI or MSI-X */
2260 static int
2261 em_setup_msix(if_ctx_t ctx)
2262 {
2263 	struct adapter *adapter = iflib_get_softc(ctx);
2264 
2265 	if (adapter->hw.mac.type == e1000_82574) {
2266 		em_enable_vectors_82574(ctx);
2267 	}
2268 	return (0);
2269 }
2270 
2271 /*********************************************************************
2272  *
2273  *  Workaround for SmartSpeed on 82541 and 82547 controllers
2274  *
2275  **********************************************************************/
2276 static void
2277 lem_smartspeed(struct adapter *adapter)
2278 {
2279 	u16 phy_tmp;
2280 
2281 	if (adapter->link_active || (adapter->hw.phy.type != e1000_phy_igp) ||
2282 	    adapter->hw.mac.autoneg == 0 ||
2283 	    (adapter->hw.phy.autoneg_advertised & ADVERTISE_1000_FULL) == 0)
2284 		return;
2285 
2286 	if (adapter->smartspeed == 0) {
2287 		/* If Master/Slave config fault is asserted twice,
2288 		 * we assume back-to-back */
2289 		e1000_read_phy_reg(&adapter->hw, PHY_1000T_STATUS, &phy_tmp);
2290 		if (!(phy_tmp & SR_1000T_MS_CONFIG_FAULT))
2291 			return;
2292 		e1000_read_phy_reg(&adapter->hw, PHY_1000T_STATUS, &phy_tmp);
2293 		if (phy_tmp & SR_1000T_MS_CONFIG_FAULT) {
2294 			e1000_read_phy_reg(&adapter->hw,
2295 			    PHY_1000T_CTRL, &phy_tmp);
2296 			if(phy_tmp & CR_1000T_MS_ENABLE) {
2297 				phy_tmp &= ~CR_1000T_MS_ENABLE;
2298 				e1000_write_phy_reg(&adapter->hw,
2299 				    PHY_1000T_CTRL, phy_tmp);
2300 				adapter->smartspeed++;
2301 				if(adapter->hw.mac.autoneg &&
2302 				   !e1000_copper_link_autoneg(&adapter->hw) &&
2303 				   !e1000_read_phy_reg(&adapter->hw,
2304 				    PHY_CONTROL, &phy_tmp)) {
2305 					phy_tmp |= (MII_CR_AUTO_NEG_EN |
2306 						    MII_CR_RESTART_AUTO_NEG);
2307 					e1000_write_phy_reg(&adapter->hw,
2308 					    PHY_CONTROL, phy_tmp);
2309 				}
2310 			}
2311 		}
2312 		return;
2313 	} else if(adapter->smartspeed == EM_SMARTSPEED_DOWNSHIFT) {
2314 		/* If still no link, perhaps using 2/3 pair cable */
2315 		e1000_read_phy_reg(&adapter->hw, PHY_1000T_CTRL, &phy_tmp);
2316 		phy_tmp |= CR_1000T_MS_ENABLE;
2317 		e1000_write_phy_reg(&adapter->hw, PHY_1000T_CTRL, phy_tmp);
2318 		if(adapter->hw.mac.autoneg &&
2319 		   !e1000_copper_link_autoneg(&adapter->hw) &&
2320 		   !e1000_read_phy_reg(&adapter->hw, PHY_CONTROL, &phy_tmp)) {
2321 			phy_tmp |= (MII_CR_AUTO_NEG_EN |
2322 				    MII_CR_RESTART_AUTO_NEG);
2323 			e1000_write_phy_reg(&adapter->hw, PHY_CONTROL, phy_tmp);
2324 		}
2325 	}
2326 	/* Restart process after EM_SMARTSPEED_MAX iterations */
2327 	if(adapter->smartspeed++ == EM_SMARTSPEED_MAX)
2328 		adapter->smartspeed = 0;
2329 }
2330 
2331 /*********************************************************************
2332  *
2333  *  Initialize the DMA Coalescing feature
2334  *
2335  **********************************************************************/
2336 static void
2337 igb_init_dmac(struct adapter *adapter, u32 pba)
2338 {
2339 	device_t	dev = adapter->dev;
2340 	struct e1000_hw *hw = &adapter->hw;
2341 	u32 		dmac, reg = ~E1000_DMACR_DMAC_EN;
2342 	u16		hwm;
2343 	u16		max_frame_size;
2344 
2345 	if (hw->mac.type == e1000_i211)
2346 		return;
2347 
2348 	max_frame_size = adapter->shared->isc_max_frame_size;
2349 	if (hw->mac.type > e1000_82580) {
2350 
2351 		if (adapter->dmac == 0) { /* Disabling it */
2352 			E1000_WRITE_REG(hw, E1000_DMACR, reg);
2353 			return;
2354 		} else
2355 			device_printf(dev, "DMA Coalescing enabled\n");
2356 
2357 		/* Set starting threshold */
2358 		E1000_WRITE_REG(hw, E1000_DMCTXTH, 0);
2359 
2360 		hwm = 64 * pba - max_frame_size / 16;
2361 		if (hwm < 64 * (pba - 6))
2362 			hwm = 64 * (pba - 6);
2363 		reg = E1000_READ_REG(hw, E1000_FCRTC);
2364 		reg &= ~E1000_FCRTC_RTH_COAL_MASK;
2365 		reg |= ((hwm << E1000_FCRTC_RTH_COAL_SHIFT)
2366 		    & E1000_FCRTC_RTH_COAL_MASK);
2367 		E1000_WRITE_REG(hw, E1000_FCRTC, reg);
2368 
2369 
2370 		dmac = pba - max_frame_size / 512;
2371 		if (dmac < pba - 10)
2372 			dmac = pba - 10;
2373 		reg = E1000_READ_REG(hw, E1000_DMACR);
2374 		reg &= ~E1000_DMACR_DMACTHR_MASK;
2375 		reg |= ((dmac << E1000_DMACR_DMACTHR_SHIFT)
2376 		    & E1000_DMACR_DMACTHR_MASK);
2377 
2378 		/* transition to L0x or L1 if available..*/
2379 		reg |= (E1000_DMACR_DMAC_EN | E1000_DMACR_DMAC_LX_MASK);
2380 
2381 		/* Check if status is 2.5Gb backplane connection
2382 		* before configuration of watchdog timer, which is
2383 		* in msec values in 12.8usec intervals
2384 		* watchdog timer= msec values in 32usec intervals
2385 		* for non 2.5Gb connection
2386 		*/
2387 		if (hw->mac.type == e1000_i354) {
2388 			int status = E1000_READ_REG(hw, E1000_STATUS);
2389 			if ((status & E1000_STATUS_2P5_SKU) &&
2390 			    (!(status & E1000_STATUS_2P5_SKU_OVER)))
2391 				reg |= ((adapter->dmac * 5) >> 6);
2392 			else
2393 				reg |= (adapter->dmac >> 5);
2394 		} else {
2395 			reg |= (adapter->dmac >> 5);
2396 		}
2397 
2398 		E1000_WRITE_REG(hw, E1000_DMACR, reg);
2399 
2400 		E1000_WRITE_REG(hw, E1000_DMCRTRH, 0);
2401 
2402 		/* Set the interval before transition */
2403 		reg = E1000_READ_REG(hw, E1000_DMCTLX);
2404 		if (hw->mac.type == e1000_i350)
2405 			reg |= IGB_DMCTLX_DCFLUSH_DIS;
2406 		/*
2407 		** in 2.5Gb connection, TTLX unit is 0.4 usec
2408 		** which is 0x4*2 = 0xA. But delay is still 4 usec
2409 		*/
2410 		if (hw->mac.type == e1000_i354) {
2411 			int status = E1000_READ_REG(hw, E1000_STATUS);
2412 			if ((status & E1000_STATUS_2P5_SKU) &&
2413 			    (!(status & E1000_STATUS_2P5_SKU_OVER)))
2414 				reg |= 0xA;
2415 			else
2416 				reg |= 0x4;
2417 		} else {
2418 			reg |= 0x4;
2419 		}
2420 
2421 		E1000_WRITE_REG(hw, E1000_DMCTLX, reg);
2422 
2423 		/* free space in tx packet buffer to wake from DMA coal */
2424 		E1000_WRITE_REG(hw, E1000_DMCTXTH, (IGB_TXPBSIZE -
2425 		    (2 * max_frame_size)) >> 6);
2426 
2427 		/* make low power state decision controlled by DMA coal */
2428 		reg = E1000_READ_REG(hw, E1000_PCIEMISC);
2429 		reg &= ~E1000_PCIEMISC_LX_DECISION;
2430 		E1000_WRITE_REG(hw, E1000_PCIEMISC, reg);
2431 
2432 	} else if (hw->mac.type == e1000_82580) {
2433 		u32 reg = E1000_READ_REG(hw, E1000_PCIEMISC);
2434 		E1000_WRITE_REG(hw, E1000_PCIEMISC,
2435 		    reg & ~E1000_PCIEMISC_LX_DECISION);
2436 		E1000_WRITE_REG(hw, E1000_DMACR, 0);
2437 	}
2438 }
2439 
2440 /*********************************************************************
2441  *
2442  *  Initialize the hardware to a configuration as specified by the
2443  *  adapter structure.
2444  *
2445  **********************************************************************/
2446 static void
2447 em_reset(if_ctx_t ctx)
2448 {
2449 	device_t dev = iflib_get_dev(ctx);
2450 	struct adapter *adapter = iflib_get_softc(ctx);
2451 	struct ifnet *ifp = iflib_get_ifp(ctx);
2452 	struct e1000_hw *hw = &adapter->hw;
2453 	u16 rx_buffer_size;
2454 	u32 pba;
2455 
2456 	INIT_DEBUGOUT("em_reset: begin");
2457 	/* Let the firmware know the OS is in control */
2458 	em_get_hw_control(adapter);
2459 
2460 	/* Set up smart power down as default off on newer adapters. */
2461 	if (!em_smart_pwr_down && (hw->mac.type == e1000_82571 ||
2462 	    hw->mac.type == e1000_82572)) {
2463 		u16 phy_tmp = 0;
2464 
2465 		/* Speed up time to link by disabling smart power down. */
2466 		e1000_read_phy_reg(hw, IGP02E1000_PHY_POWER_MGMT, &phy_tmp);
2467 		phy_tmp &= ~IGP02E1000_PM_SPD;
2468 		e1000_write_phy_reg(hw, IGP02E1000_PHY_POWER_MGMT, phy_tmp);
2469 	}
2470 
2471 	/*
2472 	 * Packet Buffer Allocation (PBA)
2473 	 * Writing PBA sets the receive portion of the buffer
2474 	 * the remainder is used for the transmit buffer.
2475 	 */
2476 	switch (hw->mac.type) {
2477 	/* Total Packet Buffer on these is 48K */
2478 	case e1000_82571:
2479 	case e1000_82572:
2480 	case e1000_80003es2lan:
2481 			pba = E1000_PBA_32K; /* 32K for Rx, 16K for Tx */
2482 		break;
2483 	case e1000_82573: /* 82573: Total Packet Buffer is 32K */
2484 			pba = E1000_PBA_12K; /* 12K for Rx, 20K for Tx */
2485 		break;
2486 	case e1000_82574:
2487 	case e1000_82583:
2488 			pba = E1000_PBA_20K; /* 20K for Rx, 20K for Tx */
2489 		break;
2490 	case e1000_ich8lan:
2491 		pba = E1000_PBA_8K;
2492 		break;
2493 	case e1000_ich9lan:
2494 	case e1000_ich10lan:
2495 		/* Boost Receive side for jumbo frames */
2496 		if (hw->mac.max_frame_size > 4096)
2497 			pba = E1000_PBA_14K;
2498 		else
2499 			pba = E1000_PBA_10K;
2500 		break;
2501 	case e1000_pchlan:
2502 	case e1000_pch2lan:
2503 	case e1000_pch_lpt:
2504 	case e1000_pch_spt:
2505 	case e1000_pch_cnp:
2506 		pba = E1000_PBA_26K;
2507 		break;
2508 	case e1000_82575:
2509 		pba = E1000_PBA_32K;
2510 		break;
2511 	case e1000_82576:
2512 	case e1000_vfadapt:
2513 		pba = E1000_READ_REG(hw, E1000_RXPBS);
2514 		pba &= E1000_RXPBS_SIZE_MASK_82576;
2515 		break;
2516 	case e1000_82580:
2517 	case e1000_i350:
2518 	case e1000_i354:
2519 	case e1000_vfadapt_i350:
2520 		pba = E1000_READ_REG(hw, E1000_RXPBS);
2521 		pba = e1000_rxpbs_adjust_82580(pba);
2522 		break;
2523 	case e1000_i210:
2524 	case e1000_i211:
2525 		pba = E1000_PBA_34K;
2526 		break;
2527 	default:
2528 		if (hw->mac.max_frame_size > 8192)
2529 			pba = E1000_PBA_40K; /* 40K for Rx, 24K for Tx */
2530 		else
2531 			pba = E1000_PBA_48K; /* 48K for Rx, 16K for Tx */
2532 	}
2533 
2534 	/* Special needs in case of Jumbo frames */
2535 	if ((hw->mac.type == e1000_82575) && (ifp->if_mtu > ETHERMTU)) {
2536 		u32 tx_space, min_tx, min_rx;
2537 		pba = E1000_READ_REG(hw, E1000_PBA);
2538 		tx_space = pba >> 16;
2539 		pba &= 0xffff;
2540 		min_tx = (hw->mac.max_frame_size +
2541 		    sizeof(struct e1000_tx_desc) - ETHERNET_FCS_SIZE) * 2;
2542 		min_tx = roundup2(min_tx, 1024);
2543 		min_tx >>= 10;
2544 		min_rx = hw->mac.max_frame_size;
2545 		min_rx = roundup2(min_rx, 1024);
2546 		min_rx >>= 10;
2547 		if (tx_space < min_tx &&
2548 		    ((min_tx - tx_space) < pba)) {
2549 			pba = pba - (min_tx - tx_space);
2550 			/*
2551 			 * if short on rx space, rx wins
2552 			 * and must trump tx adjustment
2553 			 */
2554 			if (pba < min_rx)
2555 				pba = min_rx;
2556 		}
2557 		E1000_WRITE_REG(hw, E1000_PBA, pba);
2558 	}
2559 
2560 	if (hw->mac.type < igb_mac_min)
2561 		E1000_WRITE_REG(hw, E1000_PBA, pba);
2562 
2563 	INIT_DEBUGOUT1("em_reset: pba=%dK",pba);
2564 
2565 	/*
2566 	 * These parameters control the automatic generation (Tx) and
2567 	 * response (Rx) to Ethernet PAUSE frames.
2568 	 * - High water mark should allow for at least two frames to be
2569 	 *   received after sending an XOFF.
2570 	 * - Low water mark works best when it is very near the high water mark.
2571 	 *   This allows the receiver to restart by sending XON when it has
2572 	 *   drained a bit. Here we use an arbitrary value of 1500 which will
2573 	 *   restart after one full frame is pulled from the buffer. There
2574 	 *   could be several smaller frames in the buffer and if so they will
2575 	 *   not trigger the XON until their total number reduces the buffer
2576 	 *   by 1500.
2577 	 * - The pause time is fairly large at 1000 x 512ns = 512 usec.
2578 	 */
2579 	rx_buffer_size = (pba & 0xffff) << 10;
2580 	hw->fc.high_water = rx_buffer_size -
2581 	    roundup2(hw->mac.max_frame_size, 1024);
2582 	hw->fc.low_water = hw->fc.high_water - 1500;
2583 
2584 	if (adapter->fc) /* locally set flow control value? */
2585 		hw->fc.requested_mode = adapter->fc;
2586 	else
2587 		hw->fc.requested_mode = e1000_fc_full;
2588 
2589 	if (hw->mac.type == e1000_80003es2lan)
2590 		hw->fc.pause_time = 0xFFFF;
2591 	else
2592 		hw->fc.pause_time = EM_FC_PAUSE_TIME;
2593 
2594 	hw->fc.send_xon = TRUE;
2595 
2596 	/* Device specific overrides/settings */
2597 	switch (hw->mac.type) {
2598 	case e1000_pchlan:
2599 		/* Workaround: no TX flow ctrl for PCH */
2600 		hw->fc.requested_mode = e1000_fc_rx_pause;
2601 		hw->fc.pause_time = 0xFFFF; /* override */
2602 		if (if_getmtu(ifp) > ETHERMTU) {
2603 			hw->fc.high_water = 0x3500;
2604 			hw->fc.low_water = 0x1500;
2605 		} else {
2606 			hw->fc.high_water = 0x5000;
2607 			hw->fc.low_water = 0x3000;
2608 		}
2609 		hw->fc.refresh_time = 0x1000;
2610 		break;
2611 	case e1000_pch2lan:
2612 	case e1000_pch_lpt:
2613 	case e1000_pch_spt:
2614 	case e1000_pch_cnp:
2615 		hw->fc.high_water = 0x5C20;
2616 		hw->fc.low_water = 0x5048;
2617 		hw->fc.pause_time = 0x0650;
2618 		hw->fc.refresh_time = 0x0400;
2619 		/* Jumbos need adjusted PBA */
2620 		if (if_getmtu(ifp) > ETHERMTU)
2621 			E1000_WRITE_REG(hw, E1000_PBA, 12);
2622 		else
2623 			E1000_WRITE_REG(hw, E1000_PBA, 26);
2624 		break;
2625 	case e1000_82575:
2626 	case e1000_82576:
2627 		/* 8-byte granularity */
2628 		hw->fc.low_water = hw->fc.high_water - 8;
2629 		break;
2630 	case e1000_82580:
2631 	case e1000_i350:
2632 	case e1000_i354:
2633 	case e1000_i210:
2634 	case e1000_i211:
2635 	case e1000_vfadapt:
2636 	case e1000_vfadapt_i350:
2637 		/* 16-byte granularity */
2638 		hw->fc.low_water = hw->fc.high_water - 16;
2639 		break;
2640 	case e1000_ich9lan:
2641 	case e1000_ich10lan:
2642 		if (if_getmtu(ifp) > ETHERMTU) {
2643 			hw->fc.high_water = 0x2800;
2644 			hw->fc.low_water = hw->fc.high_water - 8;
2645 			break;
2646 		}
2647 		/* FALLTHROUGH */
2648 	default:
2649 		if (hw->mac.type == e1000_80003es2lan)
2650 			hw->fc.pause_time = 0xFFFF;
2651 		break;
2652 	}
2653 
2654 	/* Issue a global reset */
2655 	e1000_reset_hw(hw);
2656 	if (hw->mac.type >= igb_mac_min) {
2657 		E1000_WRITE_REG(hw, E1000_WUC, 0);
2658 	} else {
2659 		E1000_WRITE_REG(hw, E1000_WUFC, 0);
2660 		em_disable_aspm(adapter);
2661 	}
2662 	if (adapter->flags & IGB_MEDIA_RESET) {
2663 		e1000_setup_init_funcs(hw, TRUE);
2664 		e1000_get_bus_info(hw);
2665 		adapter->flags &= ~IGB_MEDIA_RESET;
2666 	}
2667 	/* and a re-init */
2668 	if (e1000_init_hw(hw) < 0) {
2669 		device_printf(dev, "Hardware Initialization Failed\n");
2670 		return;
2671 	}
2672 	if (hw->mac.type >= igb_mac_min)
2673 		igb_init_dmac(adapter, pba);
2674 
2675 	E1000_WRITE_REG(hw, E1000_VET, ETHERTYPE_VLAN);
2676 	e1000_get_phy_info(hw);
2677 	e1000_check_for_link(hw);
2678 }
2679 
2680 /*
2681  * Initialise the RSS mapping for NICs that support multiple transmit/
2682  * receive rings.
2683  */
2684 
2685 #define RSSKEYLEN 10
2686 static void
2687 em_initialize_rss_mapping(struct adapter *adapter)
2688 {
2689 	uint8_t  rss_key[4 * RSSKEYLEN];
2690 	uint32_t reta = 0;
2691 	struct e1000_hw	*hw = &adapter->hw;
2692 	int i;
2693 
2694 	/*
2695 	 * Configure RSS key
2696 	 */
2697 	arc4rand(rss_key, sizeof(rss_key), 0);
2698 	for (i = 0; i < RSSKEYLEN; ++i) {
2699 		uint32_t rssrk = 0;
2700 
2701 		rssrk = EM_RSSRK_VAL(rss_key, i);
2702 		E1000_WRITE_REG(hw,E1000_RSSRK(i), rssrk);
2703 	}
2704 
2705 	/*
2706 	 * Configure RSS redirect table in following fashion:
2707 	 * (hash & ring_cnt_mask) == rdr_table[(hash & rdr_table_mask)]
2708 	 */
2709 	for (i = 0; i < sizeof(reta); ++i) {
2710 		uint32_t q;
2711 
2712 		q = (i % adapter->rx_num_queues) << 7;
2713 		reta |= q << (8 * i);
2714 	}
2715 
2716 	for (i = 0; i < 32; ++i)
2717 		E1000_WRITE_REG(hw, E1000_RETA(i), reta);
2718 
2719 	E1000_WRITE_REG(hw, E1000_MRQC, E1000_MRQC_RSS_ENABLE_2Q |
2720 			E1000_MRQC_RSS_FIELD_IPV4_TCP |
2721 			E1000_MRQC_RSS_FIELD_IPV4 |
2722 			E1000_MRQC_RSS_FIELD_IPV6_TCP_EX |
2723 			E1000_MRQC_RSS_FIELD_IPV6_EX |
2724 			E1000_MRQC_RSS_FIELD_IPV6);
2725 }
2726 
2727 static void
2728 igb_initialize_rss_mapping(struct adapter *adapter)
2729 {
2730 	struct e1000_hw *hw = &adapter->hw;
2731 	int i;
2732 	int queue_id;
2733 	u32 reta;
2734 	u32 rss_key[10], mrqc, shift = 0;
2735 
2736 	/* XXX? */
2737 	if (hw->mac.type == e1000_82575)
2738 		shift = 6;
2739 
2740 	/*
2741 	 * The redirection table controls which destination
2742 	 * queue each bucket redirects traffic to.
2743 	 * Each DWORD represents four queues, with the LSB
2744 	 * being the first queue in the DWORD.
2745 	 *
2746 	 * This just allocates buckets to queues using round-robin
2747 	 * allocation.
2748 	 *
2749 	 * NOTE: It Just Happens to line up with the default
2750 	 * RSS allocation method.
2751 	 */
2752 
2753 	/* Warning FM follows */
2754 	reta = 0;
2755 	for (i = 0; i < 128; i++) {
2756 #ifdef RSS
2757 		queue_id = rss_get_indirection_to_bucket(i);
2758 		/*
2759 		 * If we have more queues than buckets, we'll
2760 		 * end up mapping buckets to a subset of the
2761 		 * queues.
2762 		 *
2763 		 * If we have more buckets than queues, we'll
2764 		 * end up instead assigning multiple buckets
2765 		 * to queues.
2766 		 *
2767 		 * Both are suboptimal, but we need to handle
2768 		 * the case so we don't go out of bounds
2769 		 * indexing arrays and such.
2770 		 */
2771 		queue_id = queue_id % adapter->rx_num_queues;
2772 #else
2773 		queue_id = (i % adapter->rx_num_queues);
2774 #endif
2775 		/* Adjust if required */
2776 		queue_id = queue_id << shift;
2777 
2778 		/*
2779 		 * The low 8 bits are for hash value (n+0);
2780 		 * The next 8 bits are for hash value (n+1), etc.
2781 		 */
2782 		reta = reta >> 8;
2783 		reta = reta | ( ((uint32_t) queue_id) << 24);
2784 		if ((i & 3) == 3) {
2785 			E1000_WRITE_REG(hw, E1000_RETA(i >> 2), reta);
2786 			reta = 0;
2787 		}
2788 	}
2789 
2790 	/* Now fill in hash table */
2791 
2792 	/*
2793 	 * MRQC: Multiple Receive Queues Command
2794 	 * Set queuing to RSS control, number depends on the device.
2795 	 */
2796 	mrqc = E1000_MRQC_ENABLE_RSS_8Q;
2797 
2798 #ifdef RSS
2799 	/* XXX ew typecasting */
2800 	rss_getkey((uint8_t *) &rss_key);
2801 #else
2802 	arc4rand(&rss_key, sizeof(rss_key), 0);
2803 #endif
2804 	for (i = 0; i < 10; i++)
2805 		E1000_WRITE_REG_ARRAY(hw, E1000_RSSRK(0), i, rss_key[i]);
2806 
2807 	/*
2808 	 * Configure the RSS fields to hash upon.
2809 	 */
2810 	mrqc |= (E1000_MRQC_RSS_FIELD_IPV4 |
2811 	    E1000_MRQC_RSS_FIELD_IPV4_TCP);
2812 	mrqc |= (E1000_MRQC_RSS_FIELD_IPV6 |
2813 	    E1000_MRQC_RSS_FIELD_IPV6_TCP);
2814 	mrqc |=( E1000_MRQC_RSS_FIELD_IPV4_UDP |
2815 	    E1000_MRQC_RSS_FIELD_IPV6_UDP);
2816 	mrqc |=( E1000_MRQC_RSS_FIELD_IPV6_UDP_EX |
2817 	    E1000_MRQC_RSS_FIELD_IPV6_TCP_EX);
2818 
2819 	E1000_WRITE_REG(hw, E1000_MRQC, mrqc);
2820 }
2821 
2822 /*********************************************************************
2823  *
2824  *  Setup networking device structure and register interface media.
2825  *
2826  **********************************************************************/
2827 static int
2828 em_setup_interface(if_ctx_t ctx)
2829 {
2830 	struct ifnet *ifp = iflib_get_ifp(ctx);
2831 	struct adapter *adapter = iflib_get_softc(ctx);
2832 	if_softc_ctx_t scctx = adapter->shared;
2833 
2834 	INIT_DEBUGOUT("em_setup_interface: begin");
2835 
2836 	/* Single Queue */
2837 	if (adapter->tx_num_queues == 1) {
2838 		if_setsendqlen(ifp, scctx->isc_ntxd[0] - 1);
2839 		if_setsendqready(ifp);
2840 	}
2841 
2842 	/*
2843 	 * Specify the media types supported by this adapter and register
2844 	 * callbacks to update media and link information
2845 	 */
2846 	if (adapter->hw.phy.media_type == e1000_media_type_fiber ||
2847 	    adapter->hw.phy.media_type == e1000_media_type_internal_serdes) {
2848 		u_char fiber_type = IFM_1000_SX;	/* default type */
2849 
2850 		if (adapter->hw.mac.type == e1000_82545)
2851 			fiber_type = IFM_1000_LX;
2852 		ifmedia_add(adapter->media, IFM_ETHER | fiber_type | IFM_FDX, 0, NULL);
2853 		ifmedia_add(adapter->media, IFM_ETHER | fiber_type, 0, NULL);
2854 	} else {
2855 		ifmedia_add(adapter->media, IFM_ETHER | IFM_10_T, 0, NULL);
2856 		ifmedia_add(adapter->media, IFM_ETHER | IFM_10_T | IFM_FDX, 0, NULL);
2857 		ifmedia_add(adapter->media, IFM_ETHER | IFM_100_TX, 0, NULL);
2858 		ifmedia_add(adapter->media, IFM_ETHER | IFM_100_TX | IFM_FDX, 0, NULL);
2859 		if (adapter->hw.phy.type != e1000_phy_ife) {
2860 			ifmedia_add(adapter->media, IFM_ETHER | IFM_1000_T | IFM_FDX, 0, NULL);
2861 			ifmedia_add(adapter->media, IFM_ETHER | IFM_1000_T, 0, NULL);
2862 		}
2863 	}
2864 	ifmedia_add(adapter->media, IFM_ETHER | IFM_AUTO, 0, NULL);
2865 	ifmedia_set(adapter->media, IFM_ETHER | IFM_AUTO);
2866 	return (0);
2867 }
2868 
2869 static int
2870 em_if_tx_queues_alloc(if_ctx_t ctx, caddr_t *vaddrs, uint64_t *paddrs, int ntxqs, int ntxqsets)
2871 {
2872 	struct adapter *adapter = iflib_get_softc(ctx);
2873 	if_softc_ctx_t scctx = adapter->shared;
2874 	int error = E1000_SUCCESS;
2875 	struct em_tx_queue *que;
2876 	int i, j;
2877 
2878 	MPASS(adapter->tx_num_queues > 0);
2879 	MPASS(adapter->tx_num_queues == ntxqsets);
2880 
2881 	/* First allocate the top level queue structs */
2882 	if (!(adapter->tx_queues =
2883 	    (struct em_tx_queue *) malloc(sizeof(struct em_tx_queue) *
2884 	    adapter->tx_num_queues, M_DEVBUF, M_NOWAIT | M_ZERO))) {
2885 		device_printf(iflib_get_dev(ctx), "Unable to allocate queue memory\n");
2886 		return(ENOMEM);
2887 	}
2888 
2889 	for (i = 0, que = adapter->tx_queues; i < adapter->tx_num_queues; i++, que++) {
2890 		/* Set up some basics */
2891 
2892 		struct tx_ring *txr = &que->txr;
2893 		txr->adapter = que->adapter = adapter;
2894 		que->me = txr->me =  i;
2895 
2896 		/* Allocate report status array */
2897 		if (!(txr->tx_rsq = (qidx_t *) malloc(sizeof(qidx_t) * scctx->isc_ntxd[0], M_DEVBUF, M_NOWAIT | M_ZERO))) {
2898 			device_printf(iflib_get_dev(ctx), "failed to allocate rs_idxs memory\n");
2899 			error = ENOMEM;
2900 			goto fail;
2901 		}
2902 		for (j = 0; j < scctx->isc_ntxd[0]; j++)
2903 			txr->tx_rsq[j] = QIDX_INVALID;
2904 		/* get the virtual and physical address of the hardware queues */
2905 		txr->tx_base = (struct e1000_tx_desc *)vaddrs[i*ntxqs];
2906 		txr->tx_paddr = paddrs[i*ntxqs];
2907 	}
2908 
2909 	if (bootverbose)
2910 		device_printf(iflib_get_dev(ctx),
2911 		    "allocated for %d tx_queues\n", adapter->tx_num_queues);
2912 	return (0);
2913 fail:
2914 	em_if_queues_free(ctx);
2915 	return (error);
2916 }
2917 
2918 static int
2919 em_if_rx_queues_alloc(if_ctx_t ctx, caddr_t *vaddrs, uint64_t *paddrs, int nrxqs, int nrxqsets)
2920 {
2921 	struct adapter *adapter = iflib_get_softc(ctx);
2922 	int error = E1000_SUCCESS;
2923 	struct em_rx_queue *que;
2924 	int i;
2925 
2926 	MPASS(adapter->rx_num_queues > 0);
2927 	MPASS(adapter->rx_num_queues == nrxqsets);
2928 
2929 	/* First allocate the top level queue structs */
2930 	if (!(adapter->rx_queues =
2931 	    (struct em_rx_queue *) malloc(sizeof(struct em_rx_queue) *
2932 	    adapter->rx_num_queues, M_DEVBUF, M_NOWAIT | M_ZERO))) {
2933 		device_printf(iflib_get_dev(ctx), "Unable to allocate queue memory\n");
2934 		error = ENOMEM;
2935 		goto fail;
2936 	}
2937 
2938 	for (i = 0, que = adapter->rx_queues; i < nrxqsets; i++, que++) {
2939 		/* Set up some basics */
2940 		struct rx_ring *rxr = &que->rxr;
2941 		rxr->adapter = que->adapter = adapter;
2942 		rxr->que = que;
2943 		que->me = rxr->me =  i;
2944 
2945 		/* get the virtual and physical address of the hardware queues */
2946 		rxr->rx_base = (union e1000_rx_desc_extended *)vaddrs[i*nrxqs];
2947 		rxr->rx_paddr = paddrs[i*nrxqs];
2948 	}
2949 
2950 	if (bootverbose)
2951 		device_printf(iflib_get_dev(ctx),
2952 		    "allocated for %d rx_queues\n", adapter->rx_num_queues);
2953 
2954 	return (0);
2955 fail:
2956 	em_if_queues_free(ctx);
2957 	return (error);
2958 }
2959 
2960 static void
2961 em_if_queues_free(if_ctx_t ctx)
2962 {
2963 	struct adapter *adapter = iflib_get_softc(ctx);
2964 	struct em_tx_queue *tx_que = adapter->tx_queues;
2965 	struct em_rx_queue *rx_que = adapter->rx_queues;
2966 
2967 	if (tx_que != NULL) {
2968 		for (int i = 0; i < adapter->tx_num_queues; i++, tx_que++) {
2969 			struct tx_ring *txr = &tx_que->txr;
2970 			if (txr->tx_rsq == NULL)
2971 				break;
2972 
2973 			free(txr->tx_rsq, M_DEVBUF);
2974 			txr->tx_rsq = NULL;
2975 		}
2976 		free(adapter->tx_queues, M_DEVBUF);
2977 		adapter->tx_queues = NULL;
2978 	}
2979 
2980 	if (rx_que != NULL) {
2981 		free(adapter->rx_queues, M_DEVBUF);
2982 		adapter->rx_queues = NULL;
2983 	}
2984 
2985 	em_release_hw_control(adapter);
2986 
2987 	if (adapter->mta != NULL) {
2988 		free(adapter->mta, M_DEVBUF);
2989 	}
2990 }
2991 
2992 /*********************************************************************
2993  *
2994  *  Enable transmit unit.
2995  *
2996  **********************************************************************/
2997 static void
2998 em_initialize_transmit_unit(if_ctx_t ctx)
2999 {
3000 	struct adapter *adapter = iflib_get_softc(ctx);
3001 	if_softc_ctx_t scctx = adapter->shared;
3002 	struct em_tx_queue *que;
3003 	struct tx_ring	*txr;
3004 	struct e1000_hw	*hw = &adapter->hw;
3005 	u32 tctl, txdctl = 0, tarc, tipg = 0;
3006 
3007 	INIT_DEBUGOUT("em_initialize_transmit_unit: begin");
3008 
3009 	for (int i = 0; i < adapter->tx_num_queues; i++, txr++) {
3010 		u64 bus_addr;
3011 		caddr_t offp, endp;
3012 
3013 		que = &adapter->tx_queues[i];
3014 		txr = &que->txr;
3015 		bus_addr = txr->tx_paddr;
3016 
3017 		/* Clear checksum offload context. */
3018 		offp = (caddr_t)&txr->csum_flags;
3019 		endp = (caddr_t)(txr + 1);
3020 		bzero(offp, endp - offp);
3021 
3022 		/* Base and Len of TX Ring */
3023 		E1000_WRITE_REG(hw, E1000_TDLEN(i),
3024 		    scctx->isc_ntxd[0] * sizeof(struct e1000_tx_desc));
3025 		E1000_WRITE_REG(hw, E1000_TDBAH(i),
3026 		    (u32)(bus_addr >> 32));
3027 		E1000_WRITE_REG(hw, E1000_TDBAL(i),
3028 		    (u32)bus_addr);
3029 		/* Init the HEAD/TAIL indices */
3030 		E1000_WRITE_REG(hw, E1000_TDT(i), 0);
3031 		E1000_WRITE_REG(hw, E1000_TDH(i), 0);
3032 
3033 		HW_DEBUGOUT2("Base = %x, Length = %x\n",
3034 		    E1000_READ_REG(hw, E1000_TDBAL(i)),
3035 		    E1000_READ_REG(hw, E1000_TDLEN(i)));
3036 
3037 		txdctl = 0; /* clear txdctl */
3038 		txdctl |= 0x1f; /* PTHRESH */
3039 		txdctl |= 1 << 8; /* HTHRESH */
3040 		txdctl |= 1 << 16;/* WTHRESH */
3041 		txdctl |= 1 << 22; /* Reserved bit 22 must always be 1 */
3042 		txdctl |= E1000_TXDCTL_GRAN;
3043 		txdctl |= 1 << 25; /* LWTHRESH */
3044 
3045 		E1000_WRITE_REG(hw, E1000_TXDCTL(i), txdctl);
3046 	}
3047 
3048 	/* Set the default values for the Tx Inter Packet Gap timer */
3049 	switch (hw->mac.type) {
3050 	case e1000_80003es2lan:
3051 		tipg = DEFAULT_82543_TIPG_IPGR1;
3052 		tipg |= DEFAULT_80003ES2LAN_TIPG_IPGR2 <<
3053 		    E1000_TIPG_IPGR2_SHIFT;
3054 		break;
3055 	case e1000_82542:
3056 		tipg = DEFAULT_82542_TIPG_IPGT;
3057 		tipg |= DEFAULT_82542_TIPG_IPGR1 << E1000_TIPG_IPGR1_SHIFT;
3058 		tipg |= DEFAULT_82542_TIPG_IPGR2 << E1000_TIPG_IPGR2_SHIFT;
3059 		break;
3060 	default:
3061 		if (hw->phy.media_type == e1000_media_type_fiber ||
3062 		    hw->phy.media_type == e1000_media_type_internal_serdes)
3063 			tipg = DEFAULT_82543_TIPG_IPGT_FIBER;
3064 		else
3065 			tipg = DEFAULT_82543_TIPG_IPGT_COPPER;
3066 		tipg |= DEFAULT_82543_TIPG_IPGR1 << E1000_TIPG_IPGR1_SHIFT;
3067 		tipg |= DEFAULT_82543_TIPG_IPGR2 << E1000_TIPG_IPGR2_SHIFT;
3068 	}
3069 
3070 	E1000_WRITE_REG(hw, E1000_TIPG, tipg);
3071 	E1000_WRITE_REG(hw, E1000_TIDV, adapter->tx_int_delay.value);
3072 
3073 	if(hw->mac.type >= e1000_82540)
3074 		E1000_WRITE_REG(hw, E1000_TADV,
3075 		    adapter->tx_abs_int_delay.value);
3076 
3077 	if (hw->mac.type == e1000_82571 || hw->mac.type == e1000_82572) {
3078 		tarc = E1000_READ_REG(hw, E1000_TARC(0));
3079 		tarc |= TARC_SPEED_MODE_BIT;
3080 		E1000_WRITE_REG(hw, E1000_TARC(0), tarc);
3081 	} else if (hw->mac.type == e1000_80003es2lan) {
3082 		/* errata: program both queues to unweighted RR */
3083 		tarc = E1000_READ_REG(hw, E1000_TARC(0));
3084 		tarc |= 1;
3085 		E1000_WRITE_REG(hw, E1000_TARC(0), tarc);
3086 		tarc = E1000_READ_REG(hw, E1000_TARC(1));
3087 		tarc |= 1;
3088 		E1000_WRITE_REG(hw, E1000_TARC(1), tarc);
3089 	} else if (hw->mac.type == e1000_82574) {
3090 		tarc = E1000_READ_REG(hw, E1000_TARC(0));
3091 		tarc |= TARC_ERRATA_BIT;
3092 		if ( adapter->tx_num_queues > 1) {
3093 			tarc |= (TARC_COMPENSATION_MODE | TARC_MQ_FIX);
3094 			E1000_WRITE_REG(hw, E1000_TARC(0), tarc);
3095 			E1000_WRITE_REG(hw, E1000_TARC(1), tarc);
3096 		} else
3097 			E1000_WRITE_REG(hw, E1000_TARC(0), tarc);
3098 	}
3099 
3100 	if (adapter->tx_int_delay.value > 0)
3101 		adapter->txd_cmd |= E1000_TXD_CMD_IDE;
3102 
3103 	/* Program the Transmit Control Register */
3104 	tctl = E1000_READ_REG(hw, E1000_TCTL);
3105 	tctl &= ~E1000_TCTL_CT;
3106 	tctl |= (E1000_TCTL_PSP | E1000_TCTL_RTLC | E1000_TCTL_EN |
3107 		   (E1000_COLLISION_THRESHOLD << E1000_CT_SHIFT));
3108 
3109 	if (hw->mac.type >= e1000_82571)
3110 		tctl |= E1000_TCTL_MULR;
3111 
3112 	/* This write will effectively turn on the transmit unit. */
3113 	E1000_WRITE_REG(hw, E1000_TCTL, tctl);
3114 
3115 	/* SPT and KBL errata workarounds */
3116 	if (hw->mac.type == e1000_pch_spt) {
3117 		u32 reg;
3118 		reg = E1000_READ_REG(hw, E1000_IOSFPC);
3119 		reg |= E1000_RCTL_RDMTS_HEX;
3120 		E1000_WRITE_REG(hw, E1000_IOSFPC, reg);
3121 		/* i218-i219 Specification Update 1.5.4.5 */
3122 		reg = E1000_READ_REG(hw, E1000_TARC(0));
3123 		reg &= ~E1000_TARC0_CB_MULTIQ_3_REQ;
3124 		reg |= E1000_TARC0_CB_MULTIQ_2_REQ;
3125 		E1000_WRITE_REG(hw, E1000_TARC(0), reg);
3126 	}
3127 }
3128 
3129 /*********************************************************************
3130  *
3131  *  Enable receive unit.
3132  *
3133  **********************************************************************/
3134 
3135 static void
3136 em_initialize_receive_unit(if_ctx_t ctx)
3137 {
3138 	struct adapter *adapter = iflib_get_softc(ctx);
3139 	if_softc_ctx_t scctx = adapter->shared;
3140 	struct ifnet *ifp = iflib_get_ifp(ctx);
3141 	struct e1000_hw	*hw = &adapter->hw;
3142 	struct em_rx_queue *que;
3143 	int i;
3144 	u32 rctl, rxcsum, rfctl;
3145 
3146 	INIT_DEBUGOUT("em_initialize_receive_units: begin");
3147 
3148 	/*
3149 	 * Make sure receives are disabled while setting
3150 	 * up the descriptor ring
3151 	 */
3152 	rctl = E1000_READ_REG(hw, E1000_RCTL);
3153 	/* Do not disable if ever enabled on this hardware */
3154 	if ((hw->mac.type != e1000_82574) && (hw->mac.type != e1000_82583))
3155 		E1000_WRITE_REG(hw, E1000_RCTL, rctl & ~E1000_RCTL_EN);
3156 
3157 	/* Setup the Receive Control Register */
3158 	rctl &= ~(3 << E1000_RCTL_MO_SHIFT);
3159 	rctl |= E1000_RCTL_EN | E1000_RCTL_BAM |
3160 	    E1000_RCTL_LBM_NO | E1000_RCTL_RDMTS_HALF |
3161 	    (hw->mac.mc_filter_type << E1000_RCTL_MO_SHIFT);
3162 
3163 	/* Do not store bad packets */
3164 	rctl &= ~E1000_RCTL_SBP;
3165 
3166 	/* Enable Long Packet receive */
3167 	if (if_getmtu(ifp) > ETHERMTU)
3168 		rctl |= E1000_RCTL_LPE;
3169 	else
3170 		rctl &= ~E1000_RCTL_LPE;
3171 
3172 	/* Strip the CRC */
3173 	if (!em_disable_crc_stripping)
3174 		rctl |= E1000_RCTL_SECRC;
3175 
3176 	if (hw->mac.type >= e1000_82540) {
3177 		E1000_WRITE_REG(hw, E1000_RADV,
3178 		    adapter->rx_abs_int_delay.value);
3179 
3180 		/*
3181 		 * Set the interrupt throttling rate. Value is calculated
3182 		 * as DEFAULT_ITR = 1/(MAX_INTS_PER_SEC * 256ns)
3183 		 */
3184 		E1000_WRITE_REG(hw, E1000_ITR, DEFAULT_ITR);
3185 	}
3186 	E1000_WRITE_REG(hw, E1000_RDTR, adapter->rx_int_delay.value);
3187 
3188 	/* Use extended rx descriptor formats */
3189 	rfctl = E1000_READ_REG(hw, E1000_RFCTL);
3190 	rfctl |= E1000_RFCTL_EXTEN;
3191 	/*
3192 	 * When using MSI-X interrupts we need to throttle
3193 	 * using the EITR register (82574 only)
3194 	 */
3195 	if (hw->mac.type == e1000_82574) {
3196 		for (int i = 0; i < 4; i++)
3197 			E1000_WRITE_REG(hw, E1000_EITR_82574(i),
3198 			    DEFAULT_ITR);
3199 		/* Disable accelerated acknowledge */
3200 		rfctl |= E1000_RFCTL_ACK_DIS;
3201 	}
3202 	E1000_WRITE_REG(hw, E1000_RFCTL, rfctl);
3203 
3204 	rxcsum = E1000_READ_REG(hw, E1000_RXCSUM);
3205 	if (if_getcapenable(ifp) & IFCAP_RXCSUM &&
3206 	    hw->mac.type >= e1000_82543) {
3207 		if (adapter->tx_num_queues > 1) {
3208 			if (hw->mac.type >= igb_mac_min) {
3209 				rxcsum |= E1000_RXCSUM_PCSD;
3210 				if (hw->mac.type != e1000_82575)
3211 					rxcsum |= E1000_RXCSUM_CRCOFL;
3212 			} else
3213 				rxcsum |= E1000_RXCSUM_TUOFL |
3214 					E1000_RXCSUM_IPOFL |
3215 					E1000_RXCSUM_PCSD;
3216 		} else {
3217 			if (hw->mac.type >= igb_mac_min)
3218 				rxcsum |= E1000_RXCSUM_IPPCSE;
3219 			else
3220 				rxcsum |= E1000_RXCSUM_TUOFL | E1000_RXCSUM_IPOFL;
3221 			if (hw->mac.type > e1000_82575)
3222 				rxcsum |= E1000_RXCSUM_CRCOFL;
3223 		}
3224 	} else
3225 		rxcsum &= ~E1000_RXCSUM_TUOFL;
3226 
3227 	E1000_WRITE_REG(hw, E1000_RXCSUM, rxcsum);
3228 
3229 	if (adapter->rx_num_queues > 1) {
3230 		if (hw->mac.type >= igb_mac_min)
3231 			igb_initialize_rss_mapping(adapter);
3232 		else
3233 			em_initialize_rss_mapping(adapter);
3234 	}
3235 
3236 	/*
3237 	 * XXX TEMPORARY WORKAROUND: on some systems with 82573
3238 	 * long latencies are observed, like Lenovo X60. This
3239 	 * change eliminates the problem, but since having positive
3240 	 * values in RDTR is a known source of problems on other
3241 	 * platforms another solution is being sought.
3242 	 */
3243 	if (hw->mac.type == e1000_82573)
3244 		E1000_WRITE_REG(hw, E1000_RDTR, 0x20);
3245 
3246 	for (i = 0, que = adapter->rx_queues; i < adapter->rx_num_queues; i++, que++) {
3247 		struct rx_ring *rxr = &que->rxr;
3248 		/* Setup the Base and Length of the Rx Descriptor Ring */
3249 		u64 bus_addr = rxr->rx_paddr;
3250 #if 0
3251 		u32 rdt = adapter->rx_num_queues -1;  /* default */
3252 #endif
3253 
3254 		E1000_WRITE_REG(hw, E1000_RDLEN(i),
3255 		    scctx->isc_nrxd[0] * sizeof(union e1000_rx_desc_extended));
3256 		E1000_WRITE_REG(hw, E1000_RDBAH(i), (u32)(bus_addr >> 32));
3257 		E1000_WRITE_REG(hw, E1000_RDBAL(i), (u32)bus_addr);
3258 		/* Setup the Head and Tail Descriptor Pointers */
3259 		E1000_WRITE_REG(hw, E1000_RDH(i), 0);
3260 		E1000_WRITE_REG(hw, E1000_RDT(i), 0);
3261 	}
3262 
3263 	/*
3264 	 * Set PTHRESH for improved jumbo performance
3265 	 * According to 10.2.5.11 of Intel 82574 Datasheet,
3266 	 * RXDCTL(1) is written whenever RXDCTL(0) is written.
3267 	 * Only write to RXDCTL(1) if there is a need for different
3268 	 * settings.
3269 	 */
3270 	if ((hw->mac.type == e1000_ich9lan || hw->mac.type == e1000_pch2lan ||
3271 	    hw->mac.type == e1000_ich10lan) && if_getmtu(ifp) > ETHERMTU) {
3272 		u32 rxdctl = E1000_READ_REG(hw, E1000_RXDCTL(0));
3273 		E1000_WRITE_REG(hw, E1000_RXDCTL(0), rxdctl | 3);
3274 	} else if (hw->mac.type == e1000_82574) {
3275 		for (int i = 0; i < adapter->rx_num_queues; i++) {
3276 			u32 rxdctl = E1000_READ_REG(hw, E1000_RXDCTL(i));
3277 			rxdctl |= 0x20; /* PTHRESH */
3278 			rxdctl |= 4 << 8; /* HTHRESH */
3279 			rxdctl |= 4 << 16;/* WTHRESH */
3280 			rxdctl |= 1 << 24; /* Switch to granularity */
3281 			E1000_WRITE_REG(hw, E1000_RXDCTL(i), rxdctl);
3282 		}
3283 	} else if (hw->mac.type >= igb_mac_min) {
3284 		u32 psize, srrctl = 0;
3285 
3286 		if (if_getmtu(ifp) > ETHERMTU) {
3287 			/* Set maximum packet len */
3288 			if (adapter->rx_mbuf_sz <= 4096) {
3289 				srrctl |= 4096 >> E1000_SRRCTL_BSIZEPKT_SHIFT;
3290 				rctl |= E1000_RCTL_SZ_4096 | E1000_RCTL_BSEX;
3291 			} else if (adapter->rx_mbuf_sz > 4096) {
3292 				srrctl |= 8192 >> E1000_SRRCTL_BSIZEPKT_SHIFT;
3293 				rctl |= E1000_RCTL_SZ_8192 | E1000_RCTL_BSEX;
3294 			}
3295 			psize = scctx->isc_max_frame_size;
3296 			/* are we on a vlan? */
3297 			if (ifp->if_vlantrunk != NULL)
3298 				psize += VLAN_TAG_SIZE;
3299 			E1000_WRITE_REG(hw, E1000_RLPML, psize);
3300 		} else {
3301 			srrctl |= 2048 >> E1000_SRRCTL_BSIZEPKT_SHIFT;
3302 			rctl |= E1000_RCTL_SZ_2048;
3303 		}
3304 
3305 		/*
3306 		 * If TX flow control is disabled and there's >1 queue defined,
3307 		 * enable DROP.
3308 		 *
3309 		 * This drops frames rather than hanging the RX MAC for all queues.
3310 		 */
3311 		if ((adapter->rx_num_queues > 1) &&
3312 		    (adapter->fc == e1000_fc_none ||
3313 		     adapter->fc == e1000_fc_rx_pause)) {
3314 			srrctl |= E1000_SRRCTL_DROP_EN;
3315 		}
3316 			/* Setup the Base and Length of the Rx Descriptor Rings */
3317 		for (i = 0, que = adapter->rx_queues; i < adapter->rx_num_queues; i++, que++) {
3318 			struct rx_ring *rxr = &que->rxr;
3319 			u64 bus_addr = rxr->rx_paddr;
3320 			u32 rxdctl;
3321 
3322 #ifdef notyet
3323 			/* Configure for header split? -- ignore for now */
3324 			rxr->hdr_split = igb_header_split;
3325 #else
3326 			srrctl |= E1000_SRRCTL_DESCTYPE_ADV_ONEBUF;
3327 #endif
3328 
3329 			E1000_WRITE_REG(hw, E1000_RDLEN(i),
3330 					scctx->isc_nrxd[0] * sizeof(struct e1000_rx_desc));
3331 			E1000_WRITE_REG(hw, E1000_RDBAH(i),
3332 					(uint32_t)(bus_addr >> 32));
3333 			E1000_WRITE_REG(hw, E1000_RDBAL(i),
3334 					(uint32_t)bus_addr);
3335 			E1000_WRITE_REG(hw, E1000_SRRCTL(i), srrctl);
3336 			/* Enable this Queue */
3337 			rxdctl = E1000_READ_REG(hw, E1000_RXDCTL(i));
3338 			rxdctl |= E1000_RXDCTL_QUEUE_ENABLE;
3339 			rxdctl &= 0xFFF00000;
3340 			rxdctl |= IGB_RX_PTHRESH;
3341 			rxdctl |= IGB_RX_HTHRESH << 8;
3342 			rxdctl |= IGB_RX_WTHRESH << 16;
3343 			E1000_WRITE_REG(hw, E1000_RXDCTL(i), rxdctl);
3344 		}
3345 	} else if (hw->mac.type >= e1000_pch2lan) {
3346 		if (if_getmtu(ifp) > ETHERMTU)
3347 			e1000_lv_jumbo_workaround_ich8lan(hw, TRUE);
3348 		else
3349 			e1000_lv_jumbo_workaround_ich8lan(hw, FALSE);
3350 	}
3351 
3352 	/* Make sure VLAN Filters are off */
3353 	rctl &= ~E1000_RCTL_VFE;
3354 
3355 	if (hw->mac.type < igb_mac_min) {
3356 		if (adapter->rx_mbuf_sz == MCLBYTES)
3357 			rctl |= E1000_RCTL_SZ_2048;
3358 		else if (adapter->rx_mbuf_sz == MJUMPAGESIZE)
3359 			rctl |= E1000_RCTL_SZ_4096 | E1000_RCTL_BSEX;
3360 		else if (adapter->rx_mbuf_sz > MJUMPAGESIZE)
3361 			rctl |= E1000_RCTL_SZ_8192 | E1000_RCTL_BSEX;
3362 
3363 		/* ensure we clear use DTYPE of 00 here */
3364 		rctl &= ~0x00000C00;
3365 	}
3366 
3367 	/* Write out the settings */
3368 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
3369 
3370 	return;
3371 }
3372 
3373 static void
3374 em_if_vlan_register(if_ctx_t ctx, u16 vtag)
3375 {
3376 	struct adapter *adapter = iflib_get_softc(ctx);
3377 	u32 index, bit;
3378 
3379 	index = (vtag >> 5) & 0x7F;
3380 	bit = vtag & 0x1F;
3381 	adapter->shadow_vfta[index] |= (1 << bit);
3382 	++adapter->num_vlans;
3383 }
3384 
3385 static void
3386 em_if_vlan_unregister(if_ctx_t ctx, u16 vtag)
3387 {
3388 	struct adapter *adapter = iflib_get_softc(ctx);
3389 	u32 index, bit;
3390 
3391 	index = (vtag >> 5) & 0x7F;
3392 	bit = vtag & 0x1F;
3393 	adapter->shadow_vfta[index] &= ~(1 << bit);
3394 	--adapter->num_vlans;
3395 }
3396 
3397 static void
3398 em_setup_vlan_hw_support(struct adapter *adapter)
3399 {
3400 	struct e1000_hw *hw = &adapter->hw;
3401 	u32 reg;
3402 
3403 	/*
3404 	 * We get here thru init_locked, meaning
3405 	 * a soft reset, this has already cleared
3406 	 * the VFTA and other state, so if there
3407 	 * have been no vlan's registered do nothing.
3408 	 */
3409 	if (adapter->num_vlans == 0)
3410 		return;
3411 
3412 	/*
3413 	 * A soft reset zero's out the VFTA, so
3414 	 * we need to repopulate it now.
3415 	 */
3416 	for (int i = 0; i < EM_VFTA_SIZE; i++)
3417 		if (adapter->shadow_vfta[i] != 0)
3418 			E1000_WRITE_REG_ARRAY(hw, E1000_VFTA,
3419 			    i, adapter->shadow_vfta[i]);
3420 
3421 	reg = E1000_READ_REG(hw, E1000_CTRL);
3422 	reg |= E1000_CTRL_VME;
3423 	E1000_WRITE_REG(hw, E1000_CTRL, reg);
3424 
3425 	/* Enable the Filter Table */
3426 	reg = E1000_READ_REG(hw, E1000_RCTL);
3427 	reg &= ~E1000_RCTL_CFIEN;
3428 	reg |= E1000_RCTL_VFE;
3429 	E1000_WRITE_REG(hw, E1000_RCTL, reg);
3430 }
3431 
3432 static void
3433 em_if_intr_enable(if_ctx_t ctx)
3434 {
3435 	struct adapter *adapter = iflib_get_softc(ctx);
3436 	struct e1000_hw *hw = &adapter->hw;
3437 	u32 ims_mask = IMS_ENABLE_MASK;
3438 
3439 	if (hw->mac.type == e1000_82574) {
3440 		E1000_WRITE_REG(hw, EM_EIAC, EM_MSIX_MASK);
3441 		ims_mask |= adapter->ims;
3442 	}
3443 	E1000_WRITE_REG(hw, E1000_IMS, ims_mask);
3444 }
3445 
3446 static void
3447 em_if_intr_disable(if_ctx_t ctx)
3448 {
3449 	struct adapter *adapter = iflib_get_softc(ctx);
3450 	struct e1000_hw *hw = &adapter->hw;
3451 
3452 	if (hw->mac.type == e1000_82574)
3453 		E1000_WRITE_REG(hw, EM_EIAC, 0);
3454 	E1000_WRITE_REG(hw, E1000_IMC, 0xffffffff);
3455 }
3456 
3457 static void
3458 igb_if_intr_enable(if_ctx_t ctx)
3459 {
3460 	struct adapter *adapter = iflib_get_softc(ctx);
3461 	struct e1000_hw *hw = &adapter->hw;
3462 	u32 mask;
3463 
3464 	if (__predict_true(adapter->intr_type == IFLIB_INTR_MSIX)) {
3465 		mask = (adapter->que_mask | adapter->link_mask);
3466 		E1000_WRITE_REG(hw, E1000_EIAC, mask);
3467 		E1000_WRITE_REG(hw, E1000_EIAM, mask);
3468 		E1000_WRITE_REG(hw, E1000_EIMS, mask);
3469 		E1000_WRITE_REG(hw, E1000_IMS, E1000_IMS_LSC);
3470 	} else
3471 		E1000_WRITE_REG(hw, E1000_IMS, IMS_ENABLE_MASK);
3472 	E1000_WRITE_FLUSH(hw);
3473 }
3474 
3475 static void
3476 igb_if_intr_disable(if_ctx_t ctx)
3477 {
3478 	struct adapter *adapter = iflib_get_softc(ctx);
3479 	struct e1000_hw *hw = &adapter->hw;
3480 
3481 	if (__predict_true(adapter->intr_type == IFLIB_INTR_MSIX)) {
3482 		E1000_WRITE_REG(hw, E1000_EIMC, 0xffffffff);
3483 		E1000_WRITE_REG(hw, E1000_EIAC, 0);
3484 	}
3485 	E1000_WRITE_REG(hw, E1000_IMC, 0xffffffff);
3486 	E1000_WRITE_FLUSH(hw);
3487 }
3488 
3489 /*
3490  * Bit of a misnomer, what this really means is
3491  * to enable OS management of the system... aka
3492  * to disable special hardware management features
3493  */
3494 static void
3495 em_init_manageability(struct adapter *adapter)
3496 {
3497 	/* A shared code workaround */
3498 #define E1000_82542_MANC2H E1000_MANC2H
3499 	if (adapter->has_manage) {
3500 		int manc2h = E1000_READ_REG(&adapter->hw, E1000_MANC2H);
3501 		int manc = E1000_READ_REG(&adapter->hw, E1000_MANC);
3502 
3503 		/* disable hardware interception of ARP */
3504 		manc &= ~(E1000_MANC_ARP_EN);
3505 
3506 		/* enable receiving management packets to the host */
3507 		manc |= E1000_MANC_EN_MNG2HOST;
3508 #define E1000_MNG2HOST_PORT_623 (1 << 5)
3509 #define E1000_MNG2HOST_PORT_664 (1 << 6)
3510 		manc2h |= E1000_MNG2HOST_PORT_623;
3511 		manc2h |= E1000_MNG2HOST_PORT_664;
3512 		E1000_WRITE_REG(&adapter->hw, E1000_MANC2H, manc2h);
3513 		E1000_WRITE_REG(&adapter->hw, E1000_MANC, manc);
3514 	}
3515 }
3516 
3517 /*
3518  * Give control back to hardware management
3519  * controller if there is one.
3520  */
3521 static void
3522 em_release_manageability(struct adapter *adapter)
3523 {
3524 	if (adapter->has_manage) {
3525 		int manc = E1000_READ_REG(&adapter->hw, E1000_MANC);
3526 
3527 		/* re-enable hardware interception of ARP */
3528 		manc |= E1000_MANC_ARP_EN;
3529 		manc &= ~E1000_MANC_EN_MNG2HOST;
3530 
3531 		E1000_WRITE_REG(&adapter->hw, E1000_MANC, manc);
3532 	}
3533 }
3534 
3535 /*
3536  * em_get_hw_control sets the {CTRL_EXT|FWSM}:DRV_LOAD bit.
3537  * For ASF and Pass Through versions of f/w this means
3538  * that the driver is loaded. For AMT version type f/w
3539  * this means that the network i/f is open.
3540  */
3541 static void
3542 em_get_hw_control(struct adapter *adapter)
3543 {
3544 	u32 ctrl_ext, swsm;
3545 
3546 	if (adapter->vf_ifp)
3547 		return;
3548 
3549 	if (adapter->hw.mac.type == e1000_82573) {
3550 		swsm = E1000_READ_REG(&adapter->hw, E1000_SWSM);
3551 		E1000_WRITE_REG(&adapter->hw, E1000_SWSM,
3552 		    swsm | E1000_SWSM_DRV_LOAD);
3553 		return;
3554 	}
3555 	/* else */
3556 	ctrl_ext = E1000_READ_REG(&adapter->hw, E1000_CTRL_EXT);
3557 	E1000_WRITE_REG(&adapter->hw, E1000_CTRL_EXT,
3558 	    ctrl_ext | E1000_CTRL_EXT_DRV_LOAD);
3559 }
3560 
3561 /*
3562  * em_release_hw_control resets {CTRL_EXT|FWSM}:DRV_LOAD bit.
3563  * For ASF and Pass Through versions of f/w this means that
3564  * the driver is no longer loaded. For AMT versions of the
3565  * f/w this means that the network i/f is closed.
3566  */
3567 static void
3568 em_release_hw_control(struct adapter *adapter)
3569 {
3570 	u32 ctrl_ext, swsm;
3571 
3572 	if (!adapter->has_manage)
3573 		return;
3574 
3575 	if (adapter->hw.mac.type == e1000_82573) {
3576 		swsm = E1000_READ_REG(&adapter->hw, E1000_SWSM);
3577 		E1000_WRITE_REG(&adapter->hw, E1000_SWSM,
3578 		    swsm & ~E1000_SWSM_DRV_LOAD);
3579 		return;
3580 	}
3581 	/* else */
3582 	ctrl_ext = E1000_READ_REG(&adapter->hw, E1000_CTRL_EXT);
3583 	E1000_WRITE_REG(&adapter->hw, E1000_CTRL_EXT,
3584 	    ctrl_ext & ~E1000_CTRL_EXT_DRV_LOAD);
3585 	return;
3586 }
3587 
3588 static int
3589 em_is_valid_ether_addr(u8 *addr)
3590 {
3591 	char zero_addr[6] = { 0, 0, 0, 0, 0, 0 };
3592 
3593 	if ((addr[0] & 1) || (!bcmp(addr, zero_addr, ETHER_ADDR_LEN))) {
3594 		return (FALSE);
3595 	}
3596 
3597 	return (TRUE);
3598 }
3599 
3600 /*
3601 ** Parse the interface capabilities with regard
3602 ** to both system management and wake-on-lan for
3603 ** later use.
3604 */
3605 static void
3606 em_get_wakeup(if_ctx_t ctx)
3607 {
3608 	struct adapter *adapter = iflib_get_softc(ctx);
3609 	device_t dev = iflib_get_dev(ctx);
3610 	u16 eeprom_data = 0, device_id, apme_mask;
3611 
3612 	adapter->has_manage = e1000_enable_mng_pass_thru(&adapter->hw);
3613 	apme_mask = EM_EEPROM_APME;
3614 
3615 	switch (adapter->hw.mac.type) {
3616 	case e1000_82542:
3617 	case e1000_82543:
3618 		break;
3619 	case e1000_82544:
3620 		e1000_read_nvm(&adapter->hw,
3621 		    NVM_INIT_CONTROL2_REG, 1, &eeprom_data);
3622 		apme_mask = EM_82544_APME;
3623 		break;
3624 	case e1000_82546:
3625 	case e1000_82546_rev_3:
3626 		if (adapter->hw.bus.func == 1) {
3627 			e1000_read_nvm(&adapter->hw,
3628 			    NVM_INIT_CONTROL3_PORT_B, 1, &eeprom_data);
3629 			break;
3630 		} else
3631 			e1000_read_nvm(&adapter->hw,
3632 			    NVM_INIT_CONTROL3_PORT_A, 1, &eeprom_data);
3633 		break;
3634 	case e1000_82573:
3635 	case e1000_82583:
3636 		adapter->has_amt = TRUE;
3637 		/* FALLTHROUGH */
3638 	case e1000_82571:
3639 	case e1000_82572:
3640 	case e1000_80003es2lan:
3641 		if (adapter->hw.bus.func == 1) {
3642 			e1000_read_nvm(&adapter->hw,
3643 			    NVM_INIT_CONTROL3_PORT_B, 1, &eeprom_data);
3644 			break;
3645 		} else
3646 			e1000_read_nvm(&adapter->hw,
3647 			    NVM_INIT_CONTROL3_PORT_A, 1, &eeprom_data);
3648 		break;
3649 	case e1000_ich8lan:
3650 	case e1000_ich9lan:
3651 	case e1000_ich10lan:
3652 	case e1000_pchlan:
3653 	case e1000_pch2lan:
3654 	case e1000_pch_lpt:
3655 	case e1000_pch_spt:
3656 	case e1000_82575:	/* listing all igb devices */
3657 	case e1000_82576:
3658 	case e1000_82580:
3659 	case e1000_i350:
3660 	case e1000_i354:
3661 	case e1000_i210:
3662 	case e1000_i211:
3663 	case e1000_vfadapt:
3664 	case e1000_vfadapt_i350:
3665 		apme_mask = E1000_WUC_APME;
3666 		adapter->has_amt = TRUE;
3667 		eeprom_data = E1000_READ_REG(&adapter->hw, E1000_WUC);
3668 		break;
3669 	default:
3670 		e1000_read_nvm(&adapter->hw,
3671 		    NVM_INIT_CONTROL3_PORT_A, 1, &eeprom_data);
3672 		break;
3673 	}
3674 	if (eeprom_data & apme_mask)
3675 		adapter->wol = (E1000_WUFC_MAG | E1000_WUFC_MC);
3676 	/*
3677 	 * We have the eeprom settings, now apply the special cases
3678 	 * where the eeprom may be wrong or the board won't support
3679 	 * wake on lan on a particular port
3680 	 */
3681 	device_id = pci_get_device(dev);
3682 	switch (device_id) {
3683 	case E1000_DEV_ID_82546GB_PCIE:
3684 		adapter->wol = 0;
3685 		break;
3686 	case E1000_DEV_ID_82546EB_FIBER:
3687 	case E1000_DEV_ID_82546GB_FIBER:
3688 		/* Wake events only supported on port A for dual fiber
3689 		 * regardless of eeprom setting */
3690 		if (E1000_READ_REG(&adapter->hw, E1000_STATUS) &
3691 		    E1000_STATUS_FUNC_1)
3692 			adapter->wol = 0;
3693 		break;
3694 	case E1000_DEV_ID_82546GB_QUAD_COPPER_KSP3:
3695 		/* if quad port adapter, disable WoL on all but port A */
3696 		if (global_quad_port_a != 0)
3697 			adapter->wol = 0;
3698 		/* Reset for multiple quad port adapters */
3699 		if (++global_quad_port_a == 4)
3700 			global_quad_port_a = 0;
3701 		break;
3702 	case E1000_DEV_ID_82571EB_FIBER:
3703 		/* Wake events only supported on port A for dual fiber
3704 		 * regardless of eeprom setting */
3705 		if (E1000_READ_REG(&adapter->hw, E1000_STATUS) &
3706 		    E1000_STATUS_FUNC_1)
3707 			adapter->wol = 0;
3708 		break;
3709 	case E1000_DEV_ID_82571EB_QUAD_COPPER:
3710 	case E1000_DEV_ID_82571EB_QUAD_FIBER:
3711 	case E1000_DEV_ID_82571EB_QUAD_COPPER_LP:
3712 		/* if quad port adapter, disable WoL on all but port A */
3713 		if (global_quad_port_a != 0)
3714 			adapter->wol = 0;
3715 		/* Reset for multiple quad port adapters */
3716 		if (++global_quad_port_a == 4)
3717 			global_quad_port_a = 0;
3718 		break;
3719 	}
3720 	return;
3721 }
3722 
3723 
3724 /*
3725  * Enable PCI Wake On Lan capability
3726  */
3727 static void
3728 em_enable_wakeup(if_ctx_t ctx)
3729 {
3730 	struct adapter *adapter = iflib_get_softc(ctx);
3731 	device_t dev = iflib_get_dev(ctx);
3732 	if_t ifp = iflib_get_ifp(ctx);
3733 	int error = 0;
3734 	u32 pmc, ctrl, ctrl_ext, rctl;
3735 	u16 status;
3736 
3737 	if (pci_find_cap(dev, PCIY_PMG, &pmc) != 0)
3738 		return;
3739 
3740 	/*
3741 	 * Determine type of Wakeup: note that wol
3742 	 * is set with all bits on by default.
3743 	 */
3744 	if ((if_getcapenable(ifp) & IFCAP_WOL_MAGIC) == 0)
3745 		adapter->wol &= ~E1000_WUFC_MAG;
3746 
3747 	if ((if_getcapenable(ifp) & IFCAP_WOL_UCAST) == 0)
3748 		adapter->wol &= ~E1000_WUFC_EX;
3749 
3750 	if ((if_getcapenable(ifp) & IFCAP_WOL_MCAST) == 0)
3751 		adapter->wol &= ~E1000_WUFC_MC;
3752 	else {
3753 		rctl = E1000_READ_REG(&adapter->hw, E1000_RCTL);
3754 		rctl |= E1000_RCTL_MPE;
3755 		E1000_WRITE_REG(&adapter->hw, E1000_RCTL, rctl);
3756 	}
3757 
3758 	if (!(adapter->wol & (E1000_WUFC_EX | E1000_WUFC_MAG | E1000_WUFC_MC)))
3759 		goto pme;
3760 
3761 	/* Advertise the wakeup capability */
3762 	ctrl = E1000_READ_REG(&adapter->hw, E1000_CTRL);
3763 	ctrl |= (E1000_CTRL_SWDPIN2 | E1000_CTRL_SWDPIN3);
3764 	E1000_WRITE_REG(&adapter->hw, E1000_CTRL, ctrl);
3765 
3766 	/* Keep the laser running on Fiber adapters */
3767 	if (adapter->hw.phy.media_type == e1000_media_type_fiber ||
3768 	    adapter->hw.phy.media_type == e1000_media_type_internal_serdes) {
3769 		ctrl_ext = E1000_READ_REG(&adapter->hw, E1000_CTRL_EXT);
3770 		ctrl_ext |= E1000_CTRL_EXT_SDP3_DATA;
3771 		E1000_WRITE_REG(&adapter->hw, E1000_CTRL_EXT, ctrl_ext);
3772 	}
3773 
3774 	if ((adapter->hw.mac.type == e1000_ich8lan) ||
3775 	    (adapter->hw.mac.type == e1000_pchlan) ||
3776 	    (adapter->hw.mac.type == e1000_ich9lan) ||
3777 	    (adapter->hw.mac.type == e1000_ich10lan))
3778 		e1000_suspend_workarounds_ich8lan(&adapter->hw);
3779 
3780 	if ( adapter->hw.mac.type >= e1000_pchlan) {
3781 		error = em_enable_phy_wakeup(adapter);
3782 		if (error)
3783 			goto pme;
3784 	} else {
3785 		/* Enable wakeup by the MAC */
3786 		E1000_WRITE_REG(&adapter->hw, E1000_WUC, E1000_WUC_PME_EN);
3787 		E1000_WRITE_REG(&adapter->hw, E1000_WUFC, adapter->wol);
3788 	}
3789 
3790 	if (adapter->hw.phy.type == e1000_phy_igp_3)
3791 		e1000_igp3_phy_powerdown_workaround_ich8lan(&adapter->hw);
3792 
3793 pme:
3794 	status = pci_read_config(dev, pmc + PCIR_POWER_STATUS, 2);
3795 	status &= ~(PCIM_PSTAT_PME | PCIM_PSTAT_PMEENABLE);
3796 	if (!error && (if_getcapenable(ifp) & IFCAP_WOL))
3797 		status |= PCIM_PSTAT_PME | PCIM_PSTAT_PMEENABLE;
3798 	pci_write_config(dev, pmc + PCIR_POWER_STATUS, status, 2);
3799 
3800 	return;
3801 }
3802 
3803 /*
3804  * WOL in the newer chipset interfaces (pchlan)
3805  * require thing to be copied into the phy
3806  */
3807 static int
3808 em_enable_phy_wakeup(struct adapter *adapter)
3809 {
3810 	struct e1000_hw *hw = &adapter->hw;
3811 	u32 mreg, ret = 0;
3812 	u16 preg;
3813 
3814 	/* copy MAC RARs to PHY RARs */
3815 	e1000_copy_rx_addrs_to_phy_ich8lan(hw);
3816 
3817 	/* copy MAC MTA to PHY MTA */
3818 	for (int i = 0; i < hw->mac.mta_reg_count; i++) {
3819 		mreg = E1000_READ_REG_ARRAY(hw, E1000_MTA, i);
3820 		e1000_write_phy_reg(hw, BM_MTA(i), (u16)(mreg & 0xFFFF));
3821 		e1000_write_phy_reg(hw, BM_MTA(i) + 1,
3822 		    (u16)((mreg >> 16) & 0xFFFF));
3823 	}
3824 
3825 	/* configure PHY Rx Control register */
3826 	e1000_read_phy_reg(hw, BM_RCTL, &preg);
3827 	mreg = E1000_READ_REG(hw, E1000_RCTL);
3828 	if (mreg & E1000_RCTL_UPE)
3829 		preg |= BM_RCTL_UPE;
3830 	if (mreg & E1000_RCTL_MPE)
3831 		preg |= BM_RCTL_MPE;
3832 	preg &= ~(BM_RCTL_MO_MASK);
3833 	if (mreg & E1000_RCTL_MO_3)
3834 		preg |= (((mreg & E1000_RCTL_MO_3) >> E1000_RCTL_MO_SHIFT)
3835 				<< BM_RCTL_MO_SHIFT);
3836 	if (mreg & E1000_RCTL_BAM)
3837 		preg |= BM_RCTL_BAM;
3838 	if (mreg & E1000_RCTL_PMCF)
3839 		preg |= BM_RCTL_PMCF;
3840 	mreg = E1000_READ_REG(hw, E1000_CTRL);
3841 	if (mreg & E1000_CTRL_RFCE)
3842 		preg |= BM_RCTL_RFCE;
3843 	e1000_write_phy_reg(hw, BM_RCTL, preg);
3844 
3845 	/* enable PHY wakeup in MAC register */
3846 	E1000_WRITE_REG(hw, E1000_WUC,
3847 	    E1000_WUC_PHY_WAKE | E1000_WUC_PME_EN | E1000_WUC_APME);
3848 	E1000_WRITE_REG(hw, E1000_WUFC, adapter->wol);
3849 
3850 	/* configure and enable PHY wakeup in PHY registers */
3851 	e1000_write_phy_reg(hw, BM_WUFC, adapter->wol);
3852 	e1000_write_phy_reg(hw, BM_WUC, E1000_WUC_PME_EN);
3853 
3854 	/* activate PHY wakeup */
3855 	ret = hw->phy.ops.acquire(hw);
3856 	if (ret) {
3857 		printf("Could not acquire PHY\n");
3858 		return ret;
3859 	}
3860 	e1000_write_phy_reg_mdic(hw, IGP01E1000_PHY_PAGE_SELECT,
3861 	                         (BM_WUC_ENABLE_PAGE << IGP_PAGE_SHIFT));
3862 	ret = e1000_read_phy_reg_mdic(hw, BM_WUC_ENABLE_REG, &preg);
3863 	if (ret) {
3864 		printf("Could not read PHY page 769\n");
3865 		goto out;
3866 	}
3867 	preg |= BM_WUC_ENABLE_BIT | BM_WUC_HOST_WU_BIT;
3868 	ret = e1000_write_phy_reg_mdic(hw, BM_WUC_ENABLE_REG, preg);
3869 	if (ret)
3870 		printf("Could not set PHY Host Wakeup bit\n");
3871 out:
3872 	hw->phy.ops.release(hw);
3873 
3874 	return ret;
3875 }
3876 
3877 static void
3878 em_if_led_func(if_ctx_t ctx, int onoff)
3879 {
3880 	struct adapter *adapter = iflib_get_softc(ctx);
3881 
3882 	if (onoff) {
3883 		e1000_setup_led(&adapter->hw);
3884 		e1000_led_on(&adapter->hw);
3885 	} else {
3886 		e1000_led_off(&adapter->hw);
3887 		e1000_cleanup_led(&adapter->hw);
3888 	}
3889 }
3890 
3891 /*
3892  * Disable the L0S and L1 LINK states
3893  */
3894 static void
3895 em_disable_aspm(struct adapter *adapter)
3896 {
3897 	int base, reg;
3898 	u16 link_cap,link_ctrl;
3899 	device_t dev = adapter->dev;
3900 
3901 	switch (adapter->hw.mac.type) {
3902 	case e1000_82573:
3903 	case e1000_82574:
3904 	case e1000_82583:
3905 		break;
3906 	default:
3907 		return;
3908 	}
3909 	if (pci_find_cap(dev, PCIY_EXPRESS, &base) != 0)
3910 		return;
3911 	reg = base + PCIER_LINK_CAP;
3912 	link_cap = pci_read_config(dev, reg, 2);
3913 	if ((link_cap & PCIEM_LINK_CAP_ASPM) == 0)
3914 		return;
3915 	reg = base + PCIER_LINK_CTL;
3916 	link_ctrl = pci_read_config(dev, reg, 2);
3917 	link_ctrl &= ~PCIEM_LINK_CTL_ASPMC;
3918 	pci_write_config(dev, reg, link_ctrl, 2);
3919 	return;
3920 }
3921 
3922 /**********************************************************************
3923  *
3924  *  Update the board statistics counters.
3925  *
3926  **********************************************************************/
3927 static void
3928 em_update_stats_counters(struct adapter *adapter)
3929 {
3930 	u64 prev_xoffrxc = adapter->stats.xoffrxc;
3931 
3932 	if(adapter->hw.phy.media_type == e1000_media_type_copper ||
3933 	   (E1000_READ_REG(&adapter->hw, E1000_STATUS) & E1000_STATUS_LU)) {
3934 		adapter->stats.symerrs += E1000_READ_REG(&adapter->hw, E1000_SYMERRS);
3935 		adapter->stats.sec += E1000_READ_REG(&adapter->hw, E1000_SEC);
3936 	}
3937 	adapter->stats.crcerrs += E1000_READ_REG(&adapter->hw, E1000_CRCERRS);
3938 	adapter->stats.mpc += E1000_READ_REG(&adapter->hw, E1000_MPC);
3939 	adapter->stats.scc += E1000_READ_REG(&adapter->hw, E1000_SCC);
3940 	adapter->stats.ecol += E1000_READ_REG(&adapter->hw, E1000_ECOL);
3941 
3942 	adapter->stats.mcc += E1000_READ_REG(&adapter->hw, E1000_MCC);
3943 	adapter->stats.latecol += E1000_READ_REG(&adapter->hw, E1000_LATECOL);
3944 	adapter->stats.colc += E1000_READ_REG(&adapter->hw, E1000_COLC);
3945 	adapter->stats.dc += E1000_READ_REG(&adapter->hw, E1000_DC);
3946 	adapter->stats.rlec += E1000_READ_REG(&adapter->hw, E1000_RLEC);
3947 	adapter->stats.xonrxc += E1000_READ_REG(&adapter->hw, E1000_XONRXC);
3948 	adapter->stats.xontxc += E1000_READ_REG(&adapter->hw, E1000_XONTXC);
3949 	adapter->stats.xoffrxc += E1000_READ_REG(&adapter->hw, E1000_XOFFRXC);
3950 	/*
3951 	 ** For watchdog management we need to know if we have been
3952 	 ** paused during the last interval, so capture that here.
3953 	*/
3954 	if (adapter->stats.xoffrxc != prev_xoffrxc)
3955 		adapter->shared->isc_pause_frames = 1;
3956 	adapter->stats.xofftxc += E1000_READ_REG(&adapter->hw, E1000_XOFFTXC);
3957 	adapter->stats.fcruc += E1000_READ_REG(&adapter->hw, E1000_FCRUC);
3958 	adapter->stats.prc64 += E1000_READ_REG(&adapter->hw, E1000_PRC64);
3959 	adapter->stats.prc127 += E1000_READ_REG(&adapter->hw, E1000_PRC127);
3960 	adapter->stats.prc255 += E1000_READ_REG(&adapter->hw, E1000_PRC255);
3961 	adapter->stats.prc511 += E1000_READ_REG(&adapter->hw, E1000_PRC511);
3962 	adapter->stats.prc1023 += E1000_READ_REG(&adapter->hw, E1000_PRC1023);
3963 	adapter->stats.prc1522 += E1000_READ_REG(&adapter->hw, E1000_PRC1522);
3964 	adapter->stats.gprc += E1000_READ_REG(&adapter->hw, E1000_GPRC);
3965 	adapter->stats.bprc += E1000_READ_REG(&adapter->hw, E1000_BPRC);
3966 	adapter->stats.mprc += E1000_READ_REG(&adapter->hw, E1000_MPRC);
3967 	adapter->stats.gptc += E1000_READ_REG(&adapter->hw, E1000_GPTC);
3968 
3969 	/* For the 64-bit byte counters the low dword must be read first. */
3970 	/* Both registers clear on the read of the high dword */
3971 
3972 	adapter->stats.gorc += E1000_READ_REG(&adapter->hw, E1000_GORCL) +
3973 	    ((u64)E1000_READ_REG(&adapter->hw, E1000_GORCH) << 32);
3974 	adapter->stats.gotc += E1000_READ_REG(&adapter->hw, E1000_GOTCL) +
3975 	    ((u64)E1000_READ_REG(&adapter->hw, E1000_GOTCH) << 32);
3976 
3977 	adapter->stats.rnbc += E1000_READ_REG(&adapter->hw, E1000_RNBC);
3978 	adapter->stats.ruc += E1000_READ_REG(&adapter->hw, E1000_RUC);
3979 	adapter->stats.rfc += E1000_READ_REG(&adapter->hw, E1000_RFC);
3980 	adapter->stats.roc += E1000_READ_REG(&adapter->hw, E1000_ROC);
3981 	adapter->stats.rjc += E1000_READ_REG(&adapter->hw, E1000_RJC);
3982 
3983 	adapter->stats.tor += E1000_READ_REG(&adapter->hw, E1000_TORH);
3984 	adapter->stats.tot += E1000_READ_REG(&adapter->hw, E1000_TOTH);
3985 
3986 	adapter->stats.tpr += E1000_READ_REG(&adapter->hw, E1000_TPR);
3987 	adapter->stats.tpt += E1000_READ_REG(&adapter->hw, E1000_TPT);
3988 	adapter->stats.ptc64 += E1000_READ_REG(&adapter->hw, E1000_PTC64);
3989 	adapter->stats.ptc127 += E1000_READ_REG(&adapter->hw, E1000_PTC127);
3990 	adapter->stats.ptc255 += E1000_READ_REG(&adapter->hw, E1000_PTC255);
3991 	adapter->stats.ptc511 += E1000_READ_REG(&adapter->hw, E1000_PTC511);
3992 	adapter->stats.ptc1023 += E1000_READ_REG(&adapter->hw, E1000_PTC1023);
3993 	adapter->stats.ptc1522 += E1000_READ_REG(&adapter->hw, E1000_PTC1522);
3994 	adapter->stats.mptc += E1000_READ_REG(&adapter->hw, E1000_MPTC);
3995 	adapter->stats.bptc += E1000_READ_REG(&adapter->hw, E1000_BPTC);
3996 
3997 	/* Interrupt Counts */
3998 
3999 	adapter->stats.iac += E1000_READ_REG(&adapter->hw, E1000_IAC);
4000 	adapter->stats.icrxptc += E1000_READ_REG(&adapter->hw, E1000_ICRXPTC);
4001 	adapter->stats.icrxatc += E1000_READ_REG(&adapter->hw, E1000_ICRXATC);
4002 	adapter->stats.ictxptc += E1000_READ_REG(&adapter->hw, E1000_ICTXPTC);
4003 	adapter->stats.ictxatc += E1000_READ_REG(&adapter->hw, E1000_ICTXATC);
4004 	adapter->stats.ictxqec += E1000_READ_REG(&adapter->hw, E1000_ICTXQEC);
4005 	adapter->stats.ictxqmtc += E1000_READ_REG(&adapter->hw, E1000_ICTXQMTC);
4006 	adapter->stats.icrxdmtc += E1000_READ_REG(&adapter->hw, E1000_ICRXDMTC);
4007 	adapter->stats.icrxoc += E1000_READ_REG(&adapter->hw, E1000_ICRXOC);
4008 
4009 	if (adapter->hw.mac.type >= e1000_82543) {
4010 		adapter->stats.algnerrc +=
4011 		E1000_READ_REG(&adapter->hw, E1000_ALGNERRC);
4012 		adapter->stats.rxerrc +=
4013 		E1000_READ_REG(&adapter->hw, E1000_RXERRC);
4014 		adapter->stats.tncrs +=
4015 		E1000_READ_REG(&adapter->hw, E1000_TNCRS);
4016 		adapter->stats.cexterr +=
4017 		E1000_READ_REG(&adapter->hw, E1000_CEXTERR);
4018 		adapter->stats.tsctc +=
4019 		E1000_READ_REG(&adapter->hw, E1000_TSCTC);
4020 		adapter->stats.tsctfc +=
4021 		E1000_READ_REG(&adapter->hw, E1000_TSCTFC);
4022 	}
4023 }
4024 
4025 static uint64_t
4026 em_if_get_counter(if_ctx_t ctx, ift_counter cnt)
4027 {
4028 	struct adapter *adapter = iflib_get_softc(ctx);
4029 	struct ifnet *ifp = iflib_get_ifp(ctx);
4030 
4031 	switch (cnt) {
4032 	case IFCOUNTER_COLLISIONS:
4033 		return (adapter->stats.colc);
4034 	case IFCOUNTER_IERRORS:
4035 		return (adapter->dropped_pkts + adapter->stats.rxerrc +
4036 		    adapter->stats.crcerrs + adapter->stats.algnerrc +
4037 		    adapter->stats.ruc + adapter->stats.roc +
4038 		    adapter->stats.mpc + adapter->stats.cexterr);
4039 	case IFCOUNTER_OERRORS:
4040 		return (adapter->stats.ecol + adapter->stats.latecol +
4041 		    adapter->watchdog_events);
4042 	default:
4043 		return (if_get_counter_default(ifp, cnt));
4044 	}
4045 }
4046 
4047 /* em_if_needs_restart - Tell iflib when the driver needs to be reinitialized
4048  * @ctx: iflib context
4049  * @event: event code to check
4050  *
4051  * Defaults to returning true for unknown events.
4052  *
4053  * @returns true if iflib needs to reinit the interface
4054  */
4055 static bool
4056 em_if_needs_restart(if_ctx_t ctx __unused, enum iflib_restart_event event)
4057 {
4058 	switch (event) {
4059 	case IFLIB_RESTART_VLAN_CONFIG:
4060 	default:
4061 		return (true);
4062 	}
4063 }
4064 
4065 /* Export a single 32-bit register via a read-only sysctl. */
4066 static int
4067 em_sysctl_reg_handler(SYSCTL_HANDLER_ARGS)
4068 {
4069 	struct adapter *adapter;
4070 	u_int val;
4071 
4072 	adapter = oidp->oid_arg1;
4073 	val = E1000_READ_REG(&adapter->hw, oidp->oid_arg2);
4074 	return (sysctl_handle_int(oidp, &val, 0, req));
4075 }
4076 
4077 /*
4078  * Add sysctl variables, one per statistic, to the system.
4079  */
4080 static void
4081 em_add_hw_stats(struct adapter *adapter)
4082 {
4083 	device_t dev = iflib_get_dev(adapter->ctx);
4084 	struct em_tx_queue *tx_que = adapter->tx_queues;
4085 	struct em_rx_queue *rx_que = adapter->rx_queues;
4086 
4087 	struct sysctl_ctx_list *ctx = device_get_sysctl_ctx(dev);
4088 	struct sysctl_oid *tree = device_get_sysctl_tree(dev);
4089 	struct sysctl_oid_list *child = SYSCTL_CHILDREN(tree);
4090 	struct e1000_hw_stats *stats = &adapter->stats;
4091 
4092 	struct sysctl_oid *stat_node, *queue_node, *int_node;
4093 	struct sysctl_oid_list *stat_list, *queue_list, *int_list;
4094 
4095 #define QUEUE_NAME_LEN 32
4096 	char namebuf[QUEUE_NAME_LEN];
4097 
4098 	/* Driver Statistics */
4099 	SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "dropped",
4100 			CTLFLAG_RD, &adapter->dropped_pkts,
4101 			"Driver dropped packets");
4102 	SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "link_irq",
4103 			CTLFLAG_RD, &adapter->link_irq,
4104 			"Link MSI-X IRQ Handled");
4105 	SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "rx_overruns",
4106 			CTLFLAG_RD, &adapter->rx_overruns,
4107 			"RX overruns");
4108 	SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "watchdog_timeouts",
4109 			CTLFLAG_RD, &adapter->watchdog_events,
4110 			"Watchdog timeouts");
4111 	SYSCTL_ADD_PROC(ctx, child, OID_AUTO, "device_control",
4112 	    CTLTYPE_UINT | CTLFLAG_RD | CTLFLAG_NEEDGIANT,
4113 	    adapter, E1000_CTRL, em_sysctl_reg_handler, "IU",
4114 	    "Device Control Register");
4115 	SYSCTL_ADD_PROC(ctx, child, OID_AUTO, "rx_control",
4116 	    CTLTYPE_UINT | CTLFLAG_RD | CTLFLAG_NEEDGIANT,
4117 	    adapter, E1000_RCTL, em_sysctl_reg_handler, "IU",
4118 	    "Receiver Control Register");
4119 	SYSCTL_ADD_UINT(ctx, child, OID_AUTO, "fc_high_water",
4120 			CTLFLAG_RD, &adapter->hw.fc.high_water, 0,
4121 			"Flow Control High Watermark");
4122 	SYSCTL_ADD_UINT(ctx, child, OID_AUTO, "fc_low_water",
4123 			CTLFLAG_RD, &adapter->hw.fc.low_water, 0,
4124 			"Flow Control Low Watermark");
4125 
4126 	for (int i = 0; i < adapter->tx_num_queues; i++, tx_que++) {
4127 		struct tx_ring *txr = &tx_que->txr;
4128 		snprintf(namebuf, QUEUE_NAME_LEN, "queue_tx_%d", i);
4129 		queue_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, namebuf,
4130 		    CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, "TX Queue Name");
4131 		queue_list = SYSCTL_CHILDREN(queue_node);
4132 
4133 		SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "txd_head",
4134 		    CTLTYPE_UINT | CTLFLAG_RD | CTLFLAG_NEEDGIANT, adapter,
4135 		    E1000_TDH(txr->me), em_sysctl_reg_handler, "IU",
4136 		    "Transmit Descriptor Head");
4137 		SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "txd_tail",
4138 		    CTLTYPE_UINT | CTLFLAG_RD | CTLFLAG_NEEDGIANT, adapter,
4139 		    E1000_TDT(txr->me), em_sysctl_reg_handler, "IU",
4140 		    "Transmit Descriptor Tail");
4141 		SYSCTL_ADD_ULONG(ctx, queue_list, OID_AUTO, "tx_irq",
4142 				CTLFLAG_RD, &txr->tx_irq,
4143 				"Queue MSI-X Transmit Interrupts");
4144 	}
4145 
4146 	for (int j = 0; j < adapter->rx_num_queues; j++, rx_que++) {
4147 		struct rx_ring *rxr = &rx_que->rxr;
4148 		snprintf(namebuf, QUEUE_NAME_LEN, "queue_rx_%d", j);
4149 		queue_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, namebuf,
4150 		    CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, "RX Queue Name");
4151 		queue_list = SYSCTL_CHILDREN(queue_node);
4152 
4153 		SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "rxd_head",
4154 		    CTLTYPE_UINT | CTLFLAG_RD | CTLFLAG_NEEDGIANT, adapter,
4155 		    E1000_RDH(rxr->me), em_sysctl_reg_handler, "IU",
4156 		    "Receive Descriptor Head");
4157 		SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "rxd_tail",
4158 		    CTLTYPE_UINT | CTLFLAG_RD | CTLFLAG_NEEDGIANT, adapter,
4159 		    E1000_RDT(rxr->me), em_sysctl_reg_handler, "IU",
4160 		    "Receive Descriptor Tail");
4161 		SYSCTL_ADD_ULONG(ctx, queue_list, OID_AUTO, "rx_irq",
4162 				CTLFLAG_RD, &rxr->rx_irq,
4163 				"Queue MSI-X Receive Interrupts");
4164 	}
4165 
4166 	/* MAC stats get their own sub node */
4167 
4168 	stat_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, "mac_stats",
4169 	    CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, "Statistics");
4170 	stat_list = SYSCTL_CHILDREN(stat_node);
4171 
4172 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "excess_coll",
4173 			CTLFLAG_RD, &stats->ecol,
4174 			"Excessive collisions");
4175 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "single_coll",
4176 			CTLFLAG_RD, &stats->scc,
4177 			"Single collisions");
4178 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "multiple_coll",
4179 			CTLFLAG_RD, &stats->mcc,
4180 			"Multiple collisions");
4181 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "late_coll",
4182 			CTLFLAG_RD, &stats->latecol,
4183 			"Late collisions");
4184 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "collision_count",
4185 			CTLFLAG_RD, &stats->colc,
4186 			"Collision Count");
4187 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "symbol_errors",
4188 			CTLFLAG_RD, &adapter->stats.symerrs,
4189 			"Symbol Errors");
4190 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "sequence_errors",
4191 			CTLFLAG_RD, &adapter->stats.sec,
4192 			"Sequence Errors");
4193 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "defer_count",
4194 			CTLFLAG_RD, &adapter->stats.dc,
4195 			"Defer Count");
4196 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "missed_packets",
4197 			CTLFLAG_RD, &adapter->stats.mpc,
4198 			"Missed Packets");
4199 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_no_buff",
4200 			CTLFLAG_RD, &adapter->stats.rnbc,
4201 			"Receive No Buffers");
4202 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_undersize",
4203 			CTLFLAG_RD, &adapter->stats.ruc,
4204 			"Receive Undersize");
4205 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_fragmented",
4206 			CTLFLAG_RD, &adapter->stats.rfc,
4207 			"Fragmented Packets Received ");
4208 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_oversize",
4209 			CTLFLAG_RD, &adapter->stats.roc,
4210 			"Oversized Packets Received");
4211 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_jabber",
4212 			CTLFLAG_RD, &adapter->stats.rjc,
4213 			"Recevied Jabber");
4214 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_errs",
4215 			CTLFLAG_RD, &adapter->stats.rxerrc,
4216 			"Receive Errors");
4217 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "crc_errs",
4218 			CTLFLAG_RD, &adapter->stats.crcerrs,
4219 			"CRC errors");
4220 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "alignment_errs",
4221 			CTLFLAG_RD, &adapter->stats.algnerrc,
4222 			"Alignment Errors");
4223 	/* On 82575 these are collision counts */
4224 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "coll_ext_errs",
4225 			CTLFLAG_RD, &adapter->stats.cexterr,
4226 			"Collision/Carrier extension errors");
4227 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xon_recvd",
4228 			CTLFLAG_RD, &adapter->stats.xonrxc,
4229 			"XON Received");
4230 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xon_txd",
4231 			CTLFLAG_RD, &adapter->stats.xontxc,
4232 			"XON Transmitted");
4233 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xoff_recvd",
4234 			CTLFLAG_RD, &adapter->stats.xoffrxc,
4235 			"XOFF Received");
4236 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xoff_txd",
4237 			CTLFLAG_RD, &adapter->stats.xofftxc,
4238 			"XOFF Transmitted");
4239 
4240 	/* Packet Reception Stats */
4241 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "total_pkts_recvd",
4242 			CTLFLAG_RD, &adapter->stats.tpr,
4243 			"Total Packets Received ");
4244 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_pkts_recvd",
4245 			CTLFLAG_RD, &adapter->stats.gprc,
4246 			"Good Packets Received");
4247 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "bcast_pkts_recvd",
4248 			CTLFLAG_RD, &adapter->stats.bprc,
4249 			"Broadcast Packets Received");
4250 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "mcast_pkts_recvd",
4251 			CTLFLAG_RD, &adapter->stats.mprc,
4252 			"Multicast Packets Received");
4253 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_64",
4254 			CTLFLAG_RD, &adapter->stats.prc64,
4255 			"64 byte frames received ");
4256 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_65_127",
4257 			CTLFLAG_RD, &adapter->stats.prc127,
4258 			"65-127 byte frames received");
4259 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_128_255",
4260 			CTLFLAG_RD, &adapter->stats.prc255,
4261 			"128-255 byte frames received");
4262 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_256_511",
4263 			CTLFLAG_RD, &adapter->stats.prc511,
4264 			"256-511 byte frames received");
4265 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_512_1023",
4266 			CTLFLAG_RD, &adapter->stats.prc1023,
4267 			"512-1023 byte frames received");
4268 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_1024_1522",
4269 			CTLFLAG_RD, &adapter->stats.prc1522,
4270 			"1023-1522 byte frames received");
4271 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_octets_recvd",
4272 			CTLFLAG_RD, &adapter->stats.gorc,
4273 			"Good Octets Received");
4274 
4275 	/* Packet Transmission Stats */
4276 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_octets_txd",
4277 			CTLFLAG_RD, &adapter->stats.gotc,
4278 			"Good Octets Transmitted");
4279 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "total_pkts_txd",
4280 			CTLFLAG_RD, &adapter->stats.tpt,
4281 			"Total Packets Transmitted");
4282 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_pkts_txd",
4283 			CTLFLAG_RD, &adapter->stats.gptc,
4284 			"Good Packets Transmitted");
4285 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "bcast_pkts_txd",
4286 			CTLFLAG_RD, &adapter->stats.bptc,
4287 			"Broadcast Packets Transmitted");
4288 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "mcast_pkts_txd",
4289 			CTLFLAG_RD, &adapter->stats.mptc,
4290 			"Multicast Packets Transmitted");
4291 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_64",
4292 			CTLFLAG_RD, &adapter->stats.ptc64,
4293 			"64 byte frames transmitted ");
4294 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_65_127",
4295 			CTLFLAG_RD, &adapter->stats.ptc127,
4296 			"65-127 byte frames transmitted");
4297 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_128_255",
4298 			CTLFLAG_RD, &adapter->stats.ptc255,
4299 			"128-255 byte frames transmitted");
4300 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_256_511",
4301 			CTLFLAG_RD, &adapter->stats.ptc511,
4302 			"256-511 byte frames transmitted");
4303 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_512_1023",
4304 			CTLFLAG_RD, &adapter->stats.ptc1023,
4305 			"512-1023 byte frames transmitted");
4306 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_1024_1522",
4307 			CTLFLAG_RD, &adapter->stats.ptc1522,
4308 			"1024-1522 byte frames transmitted");
4309 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tso_txd",
4310 			CTLFLAG_RD, &adapter->stats.tsctc,
4311 			"TSO Contexts Transmitted");
4312 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tso_ctx_fail",
4313 			CTLFLAG_RD, &adapter->stats.tsctfc,
4314 			"TSO Contexts Failed");
4315 
4316 
4317 	/* Interrupt Stats */
4318 
4319 	int_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, "interrupts",
4320 	    CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, "Interrupt Statistics");
4321 	int_list = SYSCTL_CHILDREN(int_node);
4322 
4323 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "asserts",
4324 			CTLFLAG_RD, &adapter->stats.iac,
4325 			"Interrupt Assertion Count");
4326 
4327 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "rx_pkt_timer",
4328 			CTLFLAG_RD, &adapter->stats.icrxptc,
4329 			"Interrupt Cause Rx Pkt Timer Expire Count");
4330 
4331 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "rx_abs_timer",
4332 			CTLFLAG_RD, &adapter->stats.icrxatc,
4333 			"Interrupt Cause Rx Abs Timer Expire Count");
4334 
4335 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "tx_pkt_timer",
4336 			CTLFLAG_RD, &adapter->stats.ictxptc,
4337 			"Interrupt Cause Tx Pkt Timer Expire Count");
4338 
4339 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "tx_abs_timer",
4340 			CTLFLAG_RD, &adapter->stats.ictxatc,
4341 			"Interrupt Cause Tx Abs Timer Expire Count");
4342 
4343 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "tx_queue_empty",
4344 			CTLFLAG_RD, &adapter->stats.ictxqec,
4345 			"Interrupt Cause Tx Queue Empty Count");
4346 
4347 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "tx_queue_min_thresh",
4348 			CTLFLAG_RD, &adapter->stats.ictxqmtc,
4349 			"Interrupt Cause Tx Queue Min Thresh Count");
4350 
4351 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "rx_desc_min_thresh",
4352 			CTLFLAG_RD, &adapter->stats.icrxdmtc,
4353 			"Interrupt Cause Rx Desc Min Thresh Count");
4354 
4355 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "rx_overrun",
4356 			CTLFLAG_RD, &adapter->stats.icrxoc,
4357 			"Interrupt Cause Receiver Overrun Count");
4358 }
4359 
4360 /**********************************************************************
4361  *
4362  *  This routine provides a way to dump out the adapter eeprom,
4363  *  often a useful debug/service tool. This only dumps the first
4364  *  32 words, stuff that matters is in that extent.
4365  *
4366  **********************************************************************/
4367 static int
4368 em_sysctl_nvm_info(SYSCTL_HANDLER_ARGS)
4369 {
4370 	struct adapter *adapter = (struct adapter *)arg1;
4371 	int error;
4372 	int result;
4373 
4374 	result = -1;
4375 	error = sysctl_handle_int(oidp, &result, 0, req);
4376 
4377 	if (error || !req->newptr)
4378 		return (error);
4379 
4380 	/*
4381 	 * This value will cause a hex dump of the
4382 	 * first 32 16-bit words of the EEPROM to
4383 	 * the screen.
4384 	 */
4385 	if (result == 1)
4386 		em_print_nvm_info(adapter);
4387 
4388 	return (error);
4389 }
4390 
4391 static void
4392 em_print_nvm_info(struct adapter *adapter)
4393 {
4394 	u16 eeprom_data;
4395 	int i, j, row = 0;
4396 
4397 	/* Its a bit crude, but it gets the job done */
4398 	printf("\nInterface EEPROM Dump:\n");
4399 	printf("Offset\n0x0000  ");
4400 	for (i = 0, j = 0; i < 32; i++, j++) {
4401 		if (j == 8) { /* Make the offset block */
4402 			j = 0; ++row;
4403 			printf("\n0x00%x0  ",row);
4404 		}
4405 		e1000_read_nvm(&adapter->hw, i, 1, &eeprom_data);
4406 		printf("%04x ", eeprom_data);
4407 	}
4408 	printf("\n");
4409 }
4410 
4411 static int
4412 em_sysctl_int_delay(SYSCTL_HANDLER_ARGS)
4413 {
4414 	struct em_int_delay_info *info;
4415 	struct adapter *adapter;
4416 	u32 regval;
4417 	int error, usecs, ticks;
4418 
4419 	info = (struct em_int_delay_info *) arg1;
4420 	usecs = info->value;
4421 	error = sysctl_handle_int(oidp, &usecs, 0, req);
4422 	if (error != 0 || req->newptr == NULL)
4423 		return (error);
4424 	if (usecs < 0 || usecs > EM_TICKS_TO_USECS(65535))
4425 		return (EINVAL);
4426 	info->value = usecs;
4427 	ticks = EM_USECS_TO_TICKS(usecs);
4428 	if (info->offset == E1000_ITR)	/* units are 256ns here */
4429 		ticks *= 4;
4430 
4431 	adapter = info->adapter;
4432 
4433 	regval = E1000_READ_OFFSET(&adapter->hw, info->offset);
4434 	regval = (regval & ~0xffff) | (ticks & 0xffff);
4435 	/* Handle a few special cases. */
4436 	switch (info->offset) {
4437 	case E1000_RDTR:
4438 		break;
4439 	case E1000_TIDV:
4440 		if (ticks == 0) {
4441 			adapter->txd_cmd &= ~E1000_TXD_CMD_IDE;
4442 			/* Don't write 0 into the TIDV register. */
4443 			regval++;
4444 		} else
4445 			adapter->txd_cmd |= E1000_TXD_CMD_IDE;
4446 		break;
4447 	}
4448 	E1000_WRITE_OFFSET(&adapter->hw, info->offset, regval);
4449 	return (0);
4450 }
4451 
4452 static void
4453 em_add_int_delay_sysctl(struct adapter *adapter, const char *name,
4454 	const char *description, struct em_int_delay_info *info,
4455 	int offset, int value)
4456 {
4457 	info->adapter = adapter;
4458 	info->offset = offset;
4459 	info->value = value;
4460 	SYSCTL_ADD_PROC(device_get_sysctl_ctx(adapter->dev),
4461 	    SYSCTL_CHILDREN(device_get_sysctl_tree(adapter->dev)),
4462 	    OID_AUTO, name, CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_NEEDGIANT,
4463 	    info, 0, em_sysctl_int_delay, "I", description);
4464 }
4465 
4466 /*
4467  * Set flow control using sysctl:
4468  * Flow control values:
4469  *      0 - off
4470  *      1 - rx pause
4471  *      2 - tx pause
4472  *      3 - full
4473  */
4474 static int
4475 em_set_flowcntl(SYSCTL_HANDLER_ARGS)
4476 {
4477 	int error;
4478 	static int input = 3; /* default is full */
4479 	struct adapter	*adapter = (struct adapter *) arg1;
4480 
4481 	error = sysctl_handle_int(oidp, &input, 0, req);
4482 
4483 	if ((error) || (req->newptr == NULL))
4484 		return (error);
4485 
4486 	if (input == adapter->fc) /* no change? */
4487 		return (error);
4488 
4489 	switch (input) {
4490 	case e1000_fc_rx_pause:
4491 	case e1000_fc_tx_pause:
4492 	case e1000_fc_full:
4493 	case e1000_fc_none:
4494 		adapter->hw.fc.requested_mode = input;
4495 		adapter->fc = input;
4496 		break;
4497 	default:
4498 		/* Do nothing */
4499 		return (error);
4500 	}
4501 
4502 	adapter->hw.fc.current_mode = adapter->hw.fc.requested_mode;
4503 	e1000_force_mac_fc(&adapter->hw);
4504 	return (error);
4505 }
4506 
4507 /*
4508  * Manage Energy Efficient Ethernet:
4509  * Control values:
4510  *     0/1 - enabled/disabled
4511  */
4512 static int
4513 em_sysctl_eee(SYSCTL_HANDLER_ARGS)
4514 {
4515 	struct adapter *adapter = (struct adapter *) arg1;
4516 	int error, value;
4517 
4518 	value = adapter->hw.dev_spec.ich8lan.eee_disable;
4519 	error = sysctl_handle_int(oidp, &value, 0, req);
4520 	if (error || req->newptr == NULL)
4521 		return (error);
4522 	adapter->hw.dev_spec.ich8lan.eee_disable = (value != 0);
4523 	em_if_init(adapter->ctx);
4524 
4525 	return (0);
4526 }
4527 
4528 static int
4529 em_sysctl_debug_info(SYSCTL_HANDLER_ARGS)
4530 {
4531 	struct adapter *adapter;
4532 	int error;
4533 	int result;
4534 
4535 	result = -1;
4536 	error = sysctl_handle_int(oidp, &result, 0, req);
4537 
4538 	if (error || !req->newptr)
4539 		return (error);
4540 
4541 	if (result == 1) {
4542 		adapter = (struct adapter *) arg1;
4543 		em_print_debug_info(adapter);
4544 	}
4545 
4546 	return (error);
4547 }
4548 
4549 static int
4550 em_get_rs(SYSCTL_HANDLER_ARGS)
4551 {
4552 	struct adapter *adapter = (struct adapter *) arg1;
4553 	int error;
4554 	int result;
4555 
4556 	result = 0;
4557 	error = sysctl_handle_int(oidp, &result, 0, req);
4558 
4559 	if (error || !req->newptr || result != 1)
4560 		return (error);
4561 	em_dump_rs(adapter);
4562 
4563 	return (error);
4564 }
4565 
4566 static void
4567 em_if_debug(if_ctx_t ctx)
4568 {
4569 	em_dump_rs(iflib_get_softc(ctx));
4570 }
4571 
4572 /*
4573  * This routine is meant to be fluid, add whatever is
4574  * needed for debugging a problem.  -jfv
4575  */
4576 static void
4577 em_print_debug_info(struct adapter *adapter)
4578 {
4579 	device_t dev = iflib_get_dev(adapter->ctx);
4580 	struct ifnet *ifp = iflib_get_ifp(adapter->ctx);
4581 	struct tx_ring *txr = &adapter->tx_queues->txr;
4582 	struct rx_ring *rxr = &adapter->rx_queues->rxr;
4583 
4584 	if (if_getdrvflags(ifp) & IFF_DRV_RUNNING)
4585 		printf("Interface is RUNNING ");
4586 	else
4587 		printf("Interface is NOT RUNNING\n");
4588 
4589 	if (if_getdrvflags(ifp) & IFF_DRV_OACTIVE)
4590 		printf("and INACTIVE\n");
4591 	else
4592 		printf("and ACTIVE\n");
4593 
4594 	for (int i = 0; i < adapter->tx_num_queues; i++, txr++) {
4595 		device_printf(dev, "TX Queue %d ------\n", i);
4596 		device_printf(dev, "hw tdh = %d, hw tdt = %d\n",
4597 			E1000_READ_REG(&adapter->hw, E1000_TDH(i)),
4598 			E1000_READ_REG(&adapter->hw, E1000_TDT(i)));
4599 
4600 	}
4601 	for (int j=0; j < adapter->rx_num_queues; j++, rxr++) {
4602 		device_printf(dev, "RX Queue %d ------\n", j);
4603 		device_printf(dev, "hw rdh = %d, hw rdt = %d\n",
4604 			E1000_READ_REG(&adapter->hw, E1000_RDH(j)),
4605 			E1000_READ_REG(&adapter->hw, E1000_RDT(j)));
4606 	}
4607 }
4608 
4609 /*
4610  * 82574 only:
4611  * Write a new value to the EEPROM increasing the number of MSI-X
4612  * vectors from 3 to 5, for proper multiqueue support.
4613  */
4614 static void
4615 em_enable_vectors_82574(if_ctx_t ctx)
4616 {
4617 	struct adapter *adapter = iflib_get_softc(ctx);
4618 	struct e1000_hw *hw = &adapter->hw;
4619 	device_t dev = iflib_get_dev(ctx);
4620 	u16 edata;
4621 
4622 	e1000_read_nvm(hw, EM_NVM_PCIE_CTRL, 1, &edata);
4623 	if (bootverbose)
4624 		device_printf(dev, "EM_NVM_PCIE_CTRL = %#06x\n", edata);
4625 	if (((edata & EM_NVM_MSIX_N_MASK) >> EM_NVM_MSIX_N_SHIFT) != 4) {
4626 		device_printf(dev, "Writing to eeprom: increasing "
4627 		    "reported MSI-X vectors from 3 to 5...\n");
4628 		edata &= ~(EM_NVM_MSIX_N_MASK);
4629 		edata |= 4 << EM_NVM_MSIX_N_SHIFT;
4630 		e1000_write_nvm(hw, EM_NVM_PCIE_CTRL, 1, &edata);
4631 		e1000_update_nvm_checksum(hw);
4632 		device_printf(dev, "Writing to eeprom: done\n");
4633 	}
4634 }
4635