xref: /freebsd/sys/dev/e1000/if_em.c (revision 963f5dc7a30624e95d72fb7f87b8892651164e46)
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_82571
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) - lem-class legacy devices */
55 	PVID(0x8086, E1000_DEV_ID_82540EM, "Intel(R) Legacy PRO/1000 MT 82540EM"),
56 	PVID(0x8086, E1000_DEV_ID_82540EM_LOM, "Intel(R) Legacy PRO/1000 MT 82540EM (LOM)"),
57 	PVID(0x8086, E1000_DEV_ID_82540EP, "Intel(R) Legacy PRO/1000 MT 82540EP"),
58 	PVID(0x8086, E1000_DEV_ID_82540EP_LOM, "Intel(R) Legacy PRO/1000 MT 82540EP (LOM)"),
59 	PVID(0x8086, E1000_DEV_ID_82540EP_LP, "Intel(R) Legacy PRO/1000 MT 82540EP (Mobile)"),
60 
61 	PVID(0x8086, E1000_DEV_ID_82541EI, "Intel(R) Legacy PRO/1000 MT 82541EI (Copper)"),
62 	PVID(0x8086, E1000_DEV_ID_82541ER, "Intel(R) Legacy PRO/1000 82541ER"),
63 	PVID(0x8086, E1000_DEV_ID_82541ER_LOM, "Intel(R) Legacy PRO/1000 MT 82541ER"),
64 	PVID(0x8086, E1000_DEV_ID_82541EI_MOBILE, "Intel(R) Legacy PRO/1000 MT 82541EI (Mobile)"),
65 	PVID(0x8086, E1000_DEV_ID_82541GI, "Intel(R) Legacy PRO/1000 MT 82541GI"),
66 	PVID(0x8086, E1000_DEV_ID_82541GI_LF, "Intel(R) Legacy PRO/1000 GT 82541PI"),
67 	PVID(0x8086, E1000_DEV_ID_82541GI_MOBILE, "Intel(R) Legacy PRO/1000 MT 82541GI (Mobile)"),
68 
69 	PVID(0x8086, E1000_DEV_ID_82542, "Intel(R) Legacy PRO/1000 82542 (Fiber)"),
70 
71 	PVID(0x8086, E1000_DEV_ID_82543GC_FIBER, "Intel(R) Legacy PRO/1000 F 82543GC (Fiber)"),
72 	PVID(0x8086, E1000_DEV_ID_82543GC_COPPER, "Intel(R) Legacy PRO/1000 T 82543GC (Copper)"),
73 
74 	PVID(0x8086, E1000_DEV_ID_82544EI_COPPER, "Intel(R) Legacy PRO/1000 XT 82544EI (Copper)"),
75 	PVID(0x8086, E1000_DEV_ID_82544EI_FIBER, "Intel(R) Legacy PRO/1000 XF 82544EI (Fiber)"),
76 	PVID(0x8086, E1000_DEV_ID_82544GC_COPPER, "Intel(R) Legacy PRO/1000 T 82544GC (Copper)"),
77 	PVID(0x8086, E1000_DEV_ID_82544GC_LOM, "Intel(R) Legacy PRO/1000 XT 82544GC (LOM)"),
78 
79 	PVID(0x8086, E1000_DEV_ID_82545EM_COPPER, "Intel(R) Legacy PRO/1000 MT 82545EM (Copper)"),
80 	PVID(0x8086, E1000_DEV_ID_82545EM_FIBER, "Intel(R) Legacy PRO/1000 MF 82545EM (Fiber)"),
81 	PVID(0x8086, E1000_DEV_ID_82545GM_COPPER, "Intel(R) Legacy PRO/1000 MT 82545GM (Copper)"),
82 	PVID(0x8086, E1000_DEV_ID_82545GM_FIBER, "Intel(R) Legacy PRO/1000 MF 82545GM (Fiber)"),
83 	PVID(0x8086, E1000_DEV_ID_82545GM_SERDES, "Intel(R) Legacy PRO/1000 MB 82545GM (SERDES)"),
84 
85 	PVID(0x8086, E1000_DEV_ID_82546EB_COPPER, "Intel(R) Legacy PRO/1000 MT 82546EB (Copper)"),
86 	PVID(0x8086, E1000_DEV_ID_82546EB_FIBER, "Intel(R) Legacy PRO/1000 MF 82546EB (Fiber)"),
87 	PVID(0x8086, E1000_DEV_ID_82546EB_QUAD_COPPER, "Intel(R) Legacy PRO/1000 MT 82546EB (Quad Copper"),
88 	PVID(0x8086, E1000_DEV_ID_82546GB_COPPER, "Intel(R) Legacy PRO/1000 MT 82546GB (Copper)"),
89 	PVID(0x8086, E1000_DEV_ID_82546GB_FIBER, "Intel(R) Legacy PRO/1000 MF 82546GB (Fiber)"),
90 	PVID(0x8086, E1000_DEV_ID_82546GB_SERDES, "Intel(R) Legacy PRO/1000 MB 82546GB (SERDES)"),
91 	PVID(0x8086, E1000_DEV_ID_82546GB_PCIE, "Intel(R) Legacy PRO/1000 P 82546GB (PCIe)"),
92 	PVID(0x8086, E1000_DEV_ID_82546GB_QUAD_COPPER, "Intel(R) Legacy PRO/1000 GT 82546GB (Quad Copper)"),
93 	PVID(0x8086, E1000_DEV_ID_82546GB_QUAD_COPPER_KSP3, "Intel(R) Legacy PRO/1000 GT 82546GB (Quad Copper)"),
94 
95 	PVID(0x8086, E1000_DEV_ID_82547EI, "Intel(R) Legacy PRO/1000 CT 82547EI"),
96 	PVID(0x8086, E1000_DEV_ID_82547EI_MOBILE, "Intel(R) Legacy PRO/1000 CT 82547EI (Mobile)"),
97 	PVID(0x8086, E1000_DEV_ID_82547GI, "Intel(R) Legacy PRO/1000 CT 82547GI"),
98 
99 	/* Intel(R) - em-class devices */
100 	PVID(0x8086, E1000_DEV_ID_82571EB_COPPER, "Intel(R) PRO/1000 PT 82571EB/82571GB (Copper)"),
101 	PVID(0x8086, E1000_DEV_ID_82571EB_FIBER, "Intel(R) PRO/1000 PF 82571EB/82571GB (Fiber)"),
102 	PVID(0x8086, E1000_DEV_ID_82571EB_SERDES, "Intel(R) PRO/1000 PB 82571EB (SERDES)"),
103 	PVID(0x8086, E1000_DEV_ID_82571EB_SERDES_DUAL, "Intel(R) PRO/1000 82571EB (Dual Mezzanine)"),
104 	PVID(0x8086, E1000_DEV_ID_82571EB_SERDES_QUAD, "Intel(R) PRO/1000 82571EB (Quad Mezzanine)"),
105 	PVID(0x8086, E1000_DEV_ID_82571EB_QUAD_COPPER, "Intel(R) PRO/1000 PT 82571EB/82571GB (Quad Copper)"),
106 	PVID(0x8086, E1000_DEV_ID_82571EB_QUAD_COPPER_LP, "Intel(R) PRO/1000 PT 82571EB/82571GB (Quad Copper)"),
107 	PVID(0x8086, E1000_DEV_ID_82571EB_QUAD_FIBER, "Intel(R) PRO/1000 PF 82571EB (Quad Fiber)"),
108 	PVID(0x8086, E1000_DEV_ID_82571PT_QUAD_COPPER, "Intel(R) PRO/1000 PT 82571PT (Quad Copper)"),
109 	PVID(0x8086, E1000_DEV_ID_82572EI, "Intel(R) PRO/1000 PT 82572EI (Copper)"),
110 	PVID(0x8086, E1000_DEV_ID_82572EI_COPPER, "Intel(R) PRO/1000 PT 82572EI (Copper)"),
111 	PVID(0x8086, E1000_DEV_ID_82572EI_FIBER, "Intel(R) PRO/1000 PF 82572EI (Fiber)"),
112 	PVID(0x8086, E1000_DEV_ID_82572EI_SERDES, "Intel(R) PRO/1000 82572EI (SERDES)"),
113 	PVID(0x8086, E1000_DEV_ID_82573E, "Intel(R) PRO/1000 82573E (Copper)"),
114 	PVID(0x8086, E1000_DEV_ID_82573E_IAMT, "Intel(R) PRO/1000 82573E AMT (Copper)"),
115 	PVID(0x8086, E1000_DEV_ID_82573L, "Intel(R) PRO/1000 82573L"),
116 	PVID(0x8086, E1000_DEV_ID_82583V, "Intel(R) 82583V"),
117 	PVID(0x8086, E1000_DEV_ID_80003ES2LAN_COPPER_SPT, "Intel(R) 80003ES2LAN (Copper)"),
118 	PVID(0x8086, E1000_DEV_ID_80003ES2LAN_SERDES_SPT, "Intel(R) 80003ES2LAN (SERDES)"),
119 	PVID(0x8086, E1000_DEV_ID_80003ES2LAN_COPPER_DPT, "Intel(R) 80003ES2LAN (Dual Copper)"),
120 	PVID(0x8086, E1000_DEV_ID_80003ES2LAN_SERDES_DPT, "Intel(R) 80003ES2LAN (Dual SERDES)"),
121 	PVID(0x8086, E1000_DEV_ID_ICH8_IGP_M_AMT, "Intel(R) 82566MM ICH8 AMT (Mobile)"),
122 	PVID(0x8086, E1000_DEV_ID_ICH8_IGP_AMT, "Intel(R) 82566DM ICH8 AMT"),
123 	PVID(0x8086, E1000_DEV_ID_ICH8_IGP_C, "Intel(R) 82566DC ICH8"),
124 	PVID(0x8086, E1000_DEV_ID_ICH8_IFE, "Intel(R) 82562V ICH8"),
125 	PVID(0x8086, E1000_DEV_ID_ICH8_IFE_GT, "Intel(R) 82562GT ICH8"),
126 	PVID(0x8086, E1000_DEV_ID_ICH8_IFE_G, "Intel(R) 82562G ICH8"),
127 	PVID(0x8086, E1000_DEV_ID_ICH8_IGP_M, "Intel(R) 82566MC ICH8"),
128 	PVID(0x8086, E1000_DEV_ID_ICH8_82567V_3, "Intel(R) 82567V-3 ICH8"),
129 	PVID(0x8086, E1000_DEV_ID_ICH9_IGP_M_AMT, "Intel(R) 82567LM ICH9 AMT"),
130 	PVID(0x8086, E1000_DEV_ID_ICH9_IGP_AMT, "Intel(R) 82566DM-2 ICH9 AMT"),
131 	PVID(0x8086, E1000_DEV_ID_ICH9_IGP_C, "Intel(R) 82566DC-2 ICH9"),
132 	PVID(0x8086, E1000_DEV_ID_ICH9_IGP_M, "Intel(R) 82567LF ICH9"),
133 	PVID(0x8086, E1000_DEV_ID_ICH9_IGP_M_V, "Intel(R) 82567V ICH9"),
134 	PVID(0x8086, E1000_DEV_ID_ICH9_IFE, "Intel(R) 82562V-2 ICH9"),
135 	PVID(0x8086, E1000_DEV_ID_ICH9_IFE_GT, "Intel(R) 82562GT-2 ICH9"),
136 	PVID(0x8086, E1000_DEV_ID_ICH9_IFE_G, "Intel(R) 82562G-2 ICH9"),
137 	PVID(0x8086, E1000_DEV_ID_ICH9_BM, "Intel(R) 82567LM-4 ICH9"),
138 	PVID(0x8086, E1000_DEV_ID_82574L, "Intel(R) Gigabit CT 82574L"),
139 	PVID(0x8086, E1000_DEV_ID_82574LA, "Intel(R) 82574L-Apple"),
140 	PVID(0x8086, E1000_DEV_ID_ICH10_R_BM_LM, "Intel(R) 82567LM-2 ICH10"),
141 	PVID(0x8086, E1000_DEV_ID_ICH10_R_BM_LF, "Intel(R) 82567LF-2 ICH10"),
142 	PVID(0x8086, E1000_DEV_ID_ICH10_R_BM_V, "Intel(R) 82567V-2 ICH10"),
143 	PVID(0x8086, E1000_DEV_ID_ICH10_D_BM_LM, "Intel(R) 82567LM-3 ICH10"),
144 	PVID(0x8086, E1000_DEV_ID_ICH10_D_BM_LF, "Intel(R) 82567LF-3 ICH10"),
145 	PVID(0x8086, E1000_DEV_ID_ICH10_D_BM_V, "Intel(R) 82567V-4 ICH10"),
146 	PVID(0x8086, E1000_DEV_ID_PCH_M_HV_LM, "Intel(R) 82577LM"),
147 	PVID(0x8086, E1000_DEV_ID_PCH_M_HV_LC, "Intel(R) 82577LC"),
148 	PVID(0x8086, E1000_DEV_ID_PCH_D_HV_DM, "Intel(R) 82578DM"),
149 	PVID(0x8086, E1000_DEV_ID_PCH_D_HV_DC, "Intel(R) 82578DC"),
150 	PVID(0x8086, E1000_DEV_ID_PCH2_LV_LM, "Intel(R) 82579LM"),
151 	PVID(0x8086, E1000_DEV_ID_PCH2_LV_V, "Intel(R) 82579V"),
152 	PVID(0x8086, E1000_DEV_ID_PCH_LPT_I217_LM, "Intel(R) I217-LM LPT"),
153 	PVID(0x8086, E1000_DEV_ID_PCH_LPT_I217_V, "Intel(R) I217-V LPT"),
154 	PVID(0x8086, E1000_DEV_ID_PCH_LPTLP_I218_LM, "Intel(R) I218-LM LPTLP"),
155 	PVID(0x8086, E1000_DEV_ID_PCH_LPTLP_I218_V, "Intel(R) I218-V LPTLP"),
156 	PVID(0x8086, E1000_DEV_ID_PCH_I218_LM2, "Intel(R) I218-LM (2)"),
157 	PVID(0x8086, E1000_DEV_ID_PCH_I218_V2, "Intel(R) I218-V (2)"),
158 	PVID(0x8086, E1000_DEV_ID_PCH_I218_LM3, "Intel(R) I218-LM (3)"),
159 	PVID(0x8086, E1000_DEV_ID_PCH_I218_V3, "Intel(R) I218-V (3)"),
160 	PVID(0x8086, E1000_DEV_ID_PCH_SPT_I219_LM, "Intel(R) I219-LM SPT"),
161 	PVID(0x8086, E1000_DEV_ID_PCH_SPT_I219_V, "Intel(R) I219-V SPT"),
162 	PVID(0x8086, E1000_DEV_ID_PCH_SPT_I219_LM2, "Intel(R) I219-LM SPT-H(2)"),
163 	PVID(0x8086, E1000_DEV_ID_PCH_SPT_I219_V2, "Intel(R) I219-V SPT-H(2)"),
164 	PVID(0x8086, E1000_DEV_ID_PCH_LBG_I219_LM3, "Intel(R) I219-LM LBG(3)"),
165 	PVID(0x8086, E1000_DEV_ID_PCH_SPT_I219_LM4, "Intel(R) I219-LM SPT(4)"),
166 	PVID(0x8086, E1000_DEV_ID_PCH_SPT_I219_V4, "Intel(R) I219-V SPT(4)"),
167 	PVID(0x8086, E1000_DEV_ID_PCH_SPT_I219_LM5, "Intel(R) I219-LM SPT(5)"),
168 	PVID(0x8086, E1000_DEV_ID_PCH_SPT_I219_V5, "Intel(R) I219-V SPT(5)"),
169 	PVID(0x8086, E1000_DEV_ID_PCH_CNP_I219_LM6, "Intel(R) I219-LM CNP(6)"),
170 	PVID(0x8086, E1000_DEV_ID_PCH_CNP_I219_V6, "Intel(R) I219-V CNP(6)"),
171 	PVID(0x8086, E1000_DEV_ID_PCH_CNP_I219_LM7, "Intel(R) I219-LM CNP(7)"),
172 	PVID(0x8086, E1000_DEV_ID_PCH_CNP_I219_V7, "Intel(R) I219-V CNP(7)"),
173 	PVID(0x8086, E1000_DEV_ID_PCH_ICP_I219_LM8, "Intel(R) I219-LM ICP(8)"),
174 	PVID(0x8086, E1000_DEV_ID_PCH_ICP_I219_V8, "Intel(R) I219-V ICP(8)"),
175 	PVID(0x8086, E1000_DEV_ID_PCH_ICP_I219_LM9, "Intel(R) I219-LM ICP(9)"),
176 	PVID(0x8086, E1000_DEV_ID_PCH_ICP_I219_V9, "Intel(R) I219-V ICP(9)"),
177 	PVID(0x8086, E1000_DEV_ID_PCH_CMP_I219_LM10, "Intel(R) I219-LM CMP(10)"),
178 	PVID(0x8086, E1000_DEV_ID_PCH_CMP_I219_V10, "Intel(R) I219-V CMP(10)"),
179 	PVID(0x8086, E1000_DEV_ID_PCH_CMP_I219_LM11, "Intel(R) I219-LM CMP(11)"),
180 	PVID(0x8086, E1000_DEV_ID_PCH_CMP_I219_V11, "Intel(R) I219-V CMP(11)"),
181 	PVID(0x8086, E1000_DEV_ID_PCH_CMP_I219_LM12, "Intel(R) I219-LM CMP(12)"),
182 	PVID(0x8086, E1000_DEV_ID_PCH_CMP_I219_V12, "Intel(R) I219-V CMP(12)"),
183 	PVID(0x8086, E1000_DEV_ID_PCH_TGP_I219_LM13, "Intel(R) I219-LM TGP(13)"),
184 	PVID(0x8086, E1000_DEV_ID_PCH_TGP_I219_V13, "Intel(R) I219-V TGP(13)"),
185 	PVID(0x8086, E1000_DEV_ID_PCH_TGP_I219_LM14, "Intel(R) I219-LM TGP(14)"),
186 	PVID(0x8086, E1000_DEV_ID_PCH_TGP_I219_V14, "Intel(R) I219-V GTP(14)"),
187 	PVID(0x8086, E1000_DEV_ID_PCH_TGP_I219_LM15, "Intel(R) I219-LM TGP(15)"),
188 	PVID(0x8086, E1000_DEV_ID_PCH_TGP_I219_V15, "Intel(R) I219-V TGP(15)"),
189 	PVID(0x8086, E1000_DEV_ID_PCH_ADL_I219_LM16, "Intel(R) I219-LM ADL(16)"),
190 	PVID(0x8086, E1000_DEV_ID_PCH_ADL_I219_V16, "Intel(R) I219-V ADL(16)"),
191 	PVID(0x8086, E1000_DEV_ID_PCH_ADL_I219_LM17, "Intel(R) I219-LM ADL(17)"),
192 	PVID(0x8086, E1000_DEV_ID_PCH_ADL_I219_V17, "Intel(R) I219-V ADL(17)"),
193 	PVID(0x8086, E1000_DEV_ID_PCH_MTP_I219_LM18, "Intel(R) I219-LM MTP(18)"),
194 	PVID(0x8086, E1000_DEV_ID_PCH_MTP_I219_V18, "Intel(R) I219-V MTP(18)"),
195 	PVID(0x8086, E1000_DEV_ID_PCH_MTP_I219_LM19, "Intel(R) I219-LM MTP(19)"),
196 	PVID(0x8086, E1000_DEV_ID_PCH_MTP_I219_V19, "Intel(R) I219-V MTP(19)"),
197 	/* required last entry */
198 	PVID_END
199 };
200 
201 static pci_vendor_info_t igb_vendor_info_array[] =
202 {
203 	/* Intel(R) - igb-class devices */
204 	PVID(0x8086, E1000_DEV_ID_82575EB_COPPER, "Intel(R) PRO/1000 82575EB (Copper)"),
205 	PVID(0x8086, E1000_DEV_ID_82575EB_FIBER_SERDES, "Intel(R) PRO/1000 82575EB (SERDES)"),
206 	PVID(0x8086, E1000_DEV_ID_82575GB_QUAD_COPPER, "Intel(R) PRO/1000 VT 82575GB (Quad Copper)"),
207 	PVID(0x8086, E1000_DEV_ID_82576, "Intel(R) PRO/1000 82576"),
208 	PVID(0x8086, E1000_DEV_ID_82576_NS, "Intel(R) PRO/1000 82576NS"),
209 	PVID(0x8086, E1000_DEV_ID_82576_NS_SERDES, "Intel(R) PRO/1000 82576NS (SERDES)"),
210 	PVID(0x8086, E1000_DEV_ID_82576_FIBER, "Intel(R) PRO/1000 EF 82576 (Dual Fiber)"),
211 	PVID(0x8086, E1000_DEV_ID_82576_SERDES, "Intel(R) PRO/1000 82576 (Dual SERDES)"),
212 	PVID(0x8086, E1000_DEV_ID_82576_SERDES_QUAD, "Intel(R) PRO/1000 ET 82576 (Quad SERDES)"),
213 	PVID(0x8086, E1000_DEV_ID_82576_QUAD_COPPER, "Intel(R) PRO/1000 ET 82576 (Quad Copper)"),
214 	PVID(0x8086, E1000_DEV_ID_82576_QUAD_COPPER_ET2, "Intel(R) PRO/1000 ET(2) 82576 (Quad Copper)"),
215 	PVID(0x8086, E1000_DEV_ID_82576_VF, "Intel(R) PRO/1000 82576 Virtual Function"),
216 	PVID(0x8086, E1000_DEV_ID_82580_COPPER, "Intel(R) I340 82580 (Copper)"),
217 	PVID(0x8086, E1000_DEV_ID_82580_FIBER, "Intel(R) I340 82580 (Fiber)"),
218 	PVID(0x8086, E1000_DEV_ID_82580_SERDES, "Intel(R) I340 82580 (SERDES)"),
219 	PVID(0x8086, E1000_DEV_ID_82580_SGMII, "Intel(R) I340 82580 (SGMII)"),
220 	PVID(0x8086, E1000_DEV_ID_82580_COPPER_DUAL, "Intel(R) I340-T2 82580 (Dual Copper)"),
221 	PVID(0x8086, E1000_DEV_ID_82580_QUAD_FIBER, "Intel(R) I340-F4 82580 (Quad Fiber)"),
222 	PVID(0x8086, E1000_DEV_ID_DH89XXCC_SERDES, "Intel(R) DH89XXCC (SERDES)"),
223 	PVID(0x8086, E1000_DEV_ID_DH89XXCC_SGMII, "Intel(R) I347-AT4 DH89XXCC"),
224 	PVID(0x8086, E1000_DEV_ID_DH89XXCC_SFP, "Intel(R) DH89XXCC (SFP)"),
225 	PVID(0x8086, E1000_DEV_ID_DH89XXCC_BACKPLANE, "Intel(R) DH89XXCC (Backplane)"),
226 	PVID(0x8086, E1000_DEV_ID_I350_COPPER, "Intel(R) I350 (Copper)"),
227 	PVID(0x8086, E1000_DEV_ID_I350_FIBER, "Intel(R) I350 (Fiber)"),
228 	PVID(0x8086, E1000_DEV_ID_I350_SERDES, "Intel(R) I350 (SERDES)"),
229 	PVID(0x8086, E1000_DEV_ID_I350_SGMII, "Intel(R) I350 (SGMII)"),
230 	PVID(0x8086, E1000_DEV_ID_I350_VF, "Intel(R) I350 Virtual Function"),
231 	PVID(0x8086, E1000_DEV_ID_I210_COPPER, "Intel(R) I210 (Copper)"),
232 	PVID(0x8086, E1000_DEV_ID_I210_COPPER_IT, "Intel(R) I210 IT (Copper)"),
233 	PVID(0x8086, E1000_DEV_ID_I210_COPPER_OEM1, "Intel(R) I210 (OEM)"),
234 	PVID(0x8086, E1000_DEV_ID_I210_COPPER_FLASHLESS, "Intel(R) I210 Flashless (Copper)"),
235 	PVID(0x8086, E1000_DEV_ID_I210_SERDES_FLASHLESS, "Intel(R) I210 Flashless (SERDES)"),
236 	PVID(0x8086, E1000_DEV_ID_I210_SGMII_FLASHLESS, "Intel(R) I210 Flashless (SGMII)"),
237 	PVID(0x8086, E1000_DEV_ID_I210_FIBER, "Intel(R) I210 (Fiber)"),
238 	PVID(0x8086, E1000_DEV_ID_I210_SERDES, "Intel(R) I210 (SERDES)"),
239 	PVID(0x8086, E1000_DEV_ID_I210_SGMII, "Intel(R) I210 (SGMII)"),
240 	PVID(0x8086, E1000_DEV_ID_I211_COPPER, "Intel(R) I211 (Copper)"),
241 	PVID(0x8086, E1000_DEV_ID_I354_BACKPLANE_1GBPS, "Intel(R) I354 (1.0 GbE Backplane)"),
242 	PVID(0x8086, E1000_DEV_ID_I354_BACKPLANE_2_5GBPS, "Intel(R) I354 (2.5 GbE Backplane)"),
243 	PVID(0x8086, E1000_DEV_ID_I354_SGMII, "Intel(R) I354 (SGMII)"),
244 	/* required last entry */
245 	PVID_END
246 };
247 
248 /*********************************************************************
249  *  Function prototypes
250  *********************************************************************/
251 static void	*em_register(device_t);
252 static void	*igb_register(device_t);
253 static int	em_if_attach_pre(if_ctx_t);
254 static int	em_if_attach_post(if_ctx_t);
255 static int	em_if_detach(if_ctx_t);
256 static int	em_if_shutdown(if_ctx_t);
257 static int	em_if_suspend(if_ctx_t);
258 static int	em_if_resume(if_ctx_t);
259 
260 static int	em_if_tx_queues_alloc(if_ctx_t, caddr_t *, uint64_t *, int, int);
261 static int	em_if_rx_queues_alloc(if_ctx_t, caddr_t *, uint64_t *, int, int);
262 static void	em_if_queues_free(if_ctx_t);
263 
264 static uint64_t	em_if_get_counter(if_ctx_t, ift_counter);
265 static void	em_if_init(if_ctx_t);
266 static void	em_if_stop(if_ctx_t);
267 static void	em_if_media_status(if_ctx_t, struct ifmediareq *);
268 static int	em_if_media_change(if_ctx_t);
269 static int	em_if_mtu_set(if_ctx_t, uint32_t);
270 static void	em_if_timer(if_ctx_t, uint16_t);
271 static void	em_if_vlan_register(if_ctx_t, u16);
272 static void	em_if_vlan_unregister(if_ctx_t, u16);
273 static void	em_if_watchdog_reset(if_ctx_t);
274 static bool	em_if_needs_restart(if_ctx_t, enum iflib_restart_event);
275 
276 static void	em_identify_hardware(if_ctx_t);
277 static int	em_allocate_pci_resources(if_ctx_t);
278 static void	em_free_pci_resources(if_ctx_t);
279 static void	em_reset(if_ctx_t);
280 static int	em_setup_interface(if_ctx_t);
281 static int	em_setup_msix(if_ctx_t);
282 
283 static void	em_initialize_transmit_unit(if_ctx_t);
284 static void	em_initialize_receive_unit(if_ctx_t);
285 
286 static void	em_if_intr_enable(if_ctx_t);
287 static void	em_if_intr_disable(if_ctx_t);
288 static void	igb_if_intr_enable(if_ctx_t);
289 static void	igb_if_intr_disable(if_ctx_t);
290 static int	em_if_rx_queue_intr_enable(if_ctx_t, uint16_t);
291 static int	em_if_tx_queue_intr_enable(if_ctx_t, uint16_t);
292 static int	igb_if_rx_queue_intr_enable(if_ctx_t, uint16_t);
293 static int	igb_if_tx_queue_intr_enable(if_ctx_t, uint16_t);
294 static void	em_if_multi_set(if_ctx_t);
295 static void	em_if_update_admin_status(if_ctx_t);
296 static void	em_if_debug(if_ctx_t);
297 static void	em_update_stats_counters(struct e1000_softc *);
298 static void	em_add_hw_stats(struct e1000_softc *);
299 static int	em_if_set_promisc(if_ctx_t, int);
300 static bool	em_if_vlan_filter_capable(if_ctx_t);
301 static bool	em_if_vlan_filter_used(if_ctx_t);
302 static void	em_if_vlan_filter_enable(struct e1000_softc *);
303 static void	em_if_vlan_filter_disable(struct e1000_softc *);
304 static void	em_if_vlan_filter_write(struct e1000_softc *);
305 static void	em_setup_vlan_hw_support(if_ctx_t ctx);
306 static int	em_sysctl_nvm_info(SYSCTL_HANDLER_ARGS);
307 static void	em_print_nvm_info(struct e1000_softc *);
308 static void	em_fw_version_locked(if_ctx_t);
309 static void	em_sbuf_fw_version(struct e1000_fw_version *, struct sbuf *);
310 static void	em_print_fw_version(struct e1000_softc *);
311 static int	em_sysctl_print_fw_version(SYSCTL_HANDLER_ARGS);
312 static int	em_sysctl_debug_info(SYSCTL_HANDLER_ARGS);
313 static int	em_get_rs(SYSCTL_HANDLER_ARGS);
314 static void	em_print_debug_info(struct e1000_softc *);
315 static int 	em_is_valid_ether_addr(u8 *);
316 static int	em_sysctl_int_delay(SYSCTL_HANDLER_ARGS);
317 static void	em_add_int_delay_sysctl(struct e1000_softc *, const char *,
318 		    const char *, struct em_int_delay_info *, int, int);
319 /* Management and WOL Support */
320 static void	em_init_manageability(struct e1000_softc *);
321 static void	em_release_manageability(struct e1000_softc *);
322 static void	em_get_hw_control(struct e1000_softc *);
323 static void	em_release_hw_control(struct e1000_softc *);
324 static void	em_get_wakeup(if_ctx_t);
325 static void	em_enable_wakeup(if_ctx_t);
326 static int	em_enable_phy_wakeup(struct e1000_softc *);
327 static void	em_disable_aspm(struct e1000_softc *);
328 
329 int		em_intr(void *);
330 
331 /* MSI-X handlers */
332 static int	em_if_msix_intr_assign(if_ctx_t, int);
333 static int	em_msix_link(void *);
334 static void	em_handle_link(void *);
335 
336 static void	em_enable_vectors_82574(if_ctx_t);
337 
338 static int	em_set_flowcntl(SYSCTL_HANDLER_ARGS);
339 static int	em_sysctl_eee(SYSCTL_HANDLER_ARGS);
340 static void	em_if_led_func(if_ctx_t, int);
341 
342 static int	em_get_regs(SYSCTL_HANDLER_ARGS);
343 
344 static void	lem_smartspeed(struct e1000_softc *);
345 static void	igb_configure_queues(struct e1000_softc *);
346 
347 
348 /*********************************************************************
349  *  FreeBSD Device Interface Entry Points
350  *********************************************************************/
351 static device_method_t em_methods[] = {
352 	/* Device interface */
353 	DEVMETHOD(device_register, em_register),
354 	DEVMETHOD(device_probe, iflib_device_probe),
355 	DEVMETHOD(device_attach, iflib_device_attach),
356 	DEVMETHOD(device_detach, iflib_device_detach),
357 	DEVMETHOD(device_shutdown, iflib_device_shutdown),
358 	DEVMETHOD(device_suspend, iflib_device_suspend),
359 	DEVMETHOD(device_resume, iflib_device_resume),
360 	DEVMETHOD_END
361 };
362 
363 static device_method_t igb_methods[] = {
364 	/* Device interface */
365 	DEVMETHOD(device_register, igb_register),
366 	DEVMETHOD(device_probe, iflib_device_probe),
367 	DEVMETHOD(device_attach, iflib_device_attach),
368 	DEVMETHOD(device_detach, iflib_device_detach),
369 	DEVMETHOD(device_shutdown, iflib_device_shutdown),
370 	DEVMETHOD(device_suspend, iflib_device_suspend),
371 	DEVMETHOD(device_resume, iflib_device_resume),
372 	DEVMETHOD_END
373 };
374 
375 
376 static driver_t em_driver = {
377 	"em", em_methods, sizeof(struct e1000_softc),
378 };
379 
380 static devclass_t em_devclass;
381 DRIVER_MODULE(em, pci, em_driver, em_devclass, 0, 0);
382 
383 MODULE_DEPEND(em, pci, 1, 1, 1);
384 MODULE_DEPEND(em, ether, 1, 1, 1);
385 MODULE_DEPEND(em, iflib, 1, 1, 1);
386 
387 IFLIB_PNP_INFO(pci, em, em_vendor_info_array);
388 
389 static driver_t igb_driver = {
390 	"igb", igb_methods, sizeof(struct e1000_softc),
391 };
392 
393 static devclass_t igb_devclass;
394 DRIVER_MODULE(igb, pci, igb_driver, igb_devclass, 0, 0);
395 
396 MODULE_DEPEND(igb, pci, 1, 1, 1);
397 MODULE_DEPEND(igb, ether, 1, 1, 1);
398 MODULE_DEPEND(igb, iflib, 1, 1, 1);
399 
400 IFLIB_PNP_INFO(pci, igb, igb_vendor_info_array);
401 
402 static device_method_t em_if_methods[] = {
403 	DEVMETHOD(ifdi_attach_pre, em_if_attach_pre),
404 	DEVMETHOD(ifdi_attach_post, em_if_attach_post),
405 	DEVMETHOD(ifdi_detach, em_if_detach),
406 	DEVMETHOD(ifdi_shutdown, em_if_shutdown),
407 	DEVMETHOD(ifdi_suspend, em_if_suspend),
408 	DEVMETHOD(ifdi_resume, em_if_resume),
409 	DEVMETHOD(ifdi_init, em_if_init),
410 	DEVMETHOD(ifdi_stop, em_if_stop),
411 	DEVMETHOD(ifdi_msix_intr_assign, em_if_msix_intr_assign),
412 	DEVMETHOD(ifdi_intr_enable, em_if_intr_enable),
413 	DEVMETHOD(ifdi_intr_disable, em_if_intr_disable),
414 	DEVMETHOD(ifdi_tx_queues_alloc, em_if_tx_queues_alloc),
415 	DEVMETHOD(ifdi_rx_queues_alloc, em_if_rx_queues_alloc),
416 	DEVMETHOD(ifdi_queues_free, em_if_queues_free),
417 	DEVMETHOD(ifdi_update_admin_status, em_if_update_admin_status),
418 	DEVMETHOD(ifdi_multi_set, em_if_multi_set),
419 	DEVMETHOD(ifdi_media_status, em_if_media_status),
420 	DEVMETHOD(ifdi_media_change, em_if_media_change),
421 	DEVMETHOD(ifdi_mtu_set, em_if_mtu_set),
422 	DEVMETHOD(ifdi_promisc_set, em_if_set_promisc),
423 	DEVMETHOD(ifdi_timer, em_if_timer),
424 	DEVMETHOD(ifdi_watchdog_reset, em_if_watchdog_reset),
425 	DEVMETHOD(ifdi_vlan_register, em_if_vlan_register),
426 	DEVMETHOD(ifdi_vlan_unregister, em_if_vlan_unregister),
427 	DEVMETHOD(ifdi_get_counter, em_if_get_counter),
428 	DEVMETHOD(ifdi_led_func, em_if_led_func),
429 	DEVMETHOD(ifdi_rx_queue_intr_enable, em_if_rx_queue_intr_enable),
430 	DEVMETHOD(ifdi_tx_queue_intr_enable, em_if_tx_queue_intr_enable),
431 	DEVMETHOD(ifdi_debug, em_if_debug),
432 	DEVMETHOD(ifdi_needs_restart, em_if_needs_restart),
433 	DEVMETHOD_END
434 };
435 
436 static driver_t em_if_driver = {
437 	"em_if", em_if_methods, sizeof(struct e1000_softc)
438 };
439 
440 static device_method_t igb_if_methods[] = {
441 	DEVMETHOD(ifdi_attach_pre, em_if_attach_pre),
442 	DEVMETHOD(ifdi_attach_post, em_if_attach_post),
443 	DEVMETHOD(ifdi_detach, em_if_detach),
444 	DEVMETHOD(ifdi_shutdown, em_if_shutdown),
445 	DEVMETHOD(ifdi_suspend, em_if_suspend),
446 	DEVMETHOD(ifdi_resume, em_if_resume),
447 	DEVMETHOD(ifdi_init, em_if_init),
448 	DEVMETHOD(ifdi_stop, em_if_stop),
449 	DEVMETHOD(ifdi_msix_intr_assign, em_if_msix_intr_assign),
450 	DEVMETHOD(ifdi_intr_enable, igb_if_intr_enable),
451 	DEVMETHOD(ifdi_intr_disable, igb_if_intr_disable),
452 	DEVMETHOD(ifdi_tx_queues_alloc, em_if_tx_queues_alloc),
453 	DEVMETHOD(ifdi_rx_queues_alloc, em_if_rx_queues_alloc),
454 	DEVMETHOD(ifdi_queues_free, em_if_queues_free),
455 	DEVMETHOD(ifdi_update_admin_status, em_if_update_admin_status),
456 	DEVMETHOD(ifdi_multi_set, em_if_multi_set),
457 	DEVMETHOD(ifdi_media_status, em_if_media_status),
458 	DEVMETHOD(ifdi_media_change, em_if_media_change),
459 	DEVMETHOD(ifdi_mtu_set, em_if_mtu_set),
460 	DEVMETHOD(ifdi_promisc_set, em_if_set_promisc),
461 	DEVMETHOD(ifdi_timer, em_if_timer),
462 	DEVMETHOD(ifdi_watchdog_reset, em_if_watchdog_reset),
463 	DEVMETHOD(ifdi_vlan_register, em_if_vlan_register),
464 	DEVMETHOD(ifdi_vlan_unregister, em_if_vlan_unregister),
465 	DEVMETHOD(ifdi_get_counter, em_if_get_counter),
466 	DEVMETHOD(ifdi_led_func, em_if_led_func),
467 	DEVMETHOD(ifdi_rx_queue_intr_enable, igb_if_rx_queue_intr_enable),
468 	DEVMETHOD(ifdi_tx_queue_intr_enable, igb_if_tx_queue_intr_enable),
469 	DEVMETHOD(ifdi_debug, em_if_debug),
470 	DEVMETHOD(ifdi_needs_restart, em_if_needs_restart),
471 	DEVMETHOD_END
472 };
473 
474 static driver_t igb_if_driver = {
475 	"igb_if", igb_if_methods, sizeof(struct e1000_softc)
476 };
477 
478 /*********************************************************************
479  *  Tunable default values.
480  *********************************************************************/
481 
482 #define EM_TICKS_TO_USECS(ticks)	((1024 * (ticks) + 500) / 1000)
483 #define EM_USECS_TO_TICKS(usecs)	((1000 * (usecs) + 512) / 1024)
484 
485 #define MAX_INTS_PER_SEC	8000
486 #define DEFAULT_ITR		(1000000000/(MAX_INTS_PER_SEC * 256))
487 
488 /* Allow common code without TSO */
489 #ifndef CSUM_TSO
490 #define CSUM_TSO	0
491 #endif
492 
493 static SYSCTL_NODE(_hw, OID_AUTO, em, CTLFLAG_RD | CTLFLAG_MPSAFE, 0,
494     "EM driver parameters");
495 
496 static int em_disable_crc_stripping = 0;
497 SYSCTL_INT(_hw_em, OID_AUTO, disable_crc_stripping, CTLFLAG_RDTUN,
498     &em_disable_crc_stripping, 0, "Disable CRC Stripping");
499 
500 static int em_tx_int_delay_dflt = EM_TICKS_TO_USECS(EM_TIDV);
501 static int em_rx_int_delay_dflt = EM_TICKS_TO_USECS(EM_RDTR);
502 SYSCTL_INT(_hw_em, OID_AUTO, tx_int_delay, CTLFLAG_RDTUN, &em_tx_int_delay_dflt,
503     0, "Default transmit interrupt delay in usecs");
504 SYSCTL_INT(_hw_em, OID_AUTO, rx_int_delay, CTLFLAG_RDTUN, &em_rx_int_delay_dflt,
505     0, "Default receive interrupt delay in usecs");
506 
507 static int em_tx_abs_int_delay_dflt = EM_TICKS_TO_USECS(EM_TADV);
508 static int em_rx_abs_int_delay_dflt = EM_TICKS_TO_USECS(EM_RADV);
509 SYSCTL_INT(_hw_em, OID_AUTO, tx_abs_int_delay, CTLFLAG_RDTUN,
510     &em_tx_abs_int_delay_dflt, 0,
511     "Default transmit interrupt delay limit in usecs");
512 SYSCTL_INT(_hw_em, OID_AUTO, rx_abs_int_delay, CTLFLAG_RDTUN,
513     &em_rx_abs_int_delay_dflt, 0,
514     "Default receive interrupt delay limit in usecs");
515 
516 static int em_smart_pwr_down = false;
517 SYSCTL_INT(_hw_em, OID_AUTO, smart_pwr_down, CTLFLAG_RDTUN, &em_smart_pwr_down,
518     0, "Set to true to leave smart power down enabled on newer adapters");
519 
520 /* Controls whether promiscuous also shows bad packets */
521 static int em_debug_sbp = false;
522 SYSCTL_INT(_hw_em, OID_AUTO, sbp, CTLFLAG_RDTUN, &em_debug_sbp, 0,
523     "Show bad packets in promiscuous mode");
524 
525 /* How many packets rxeof tries to clean at a time */
526 static int em_rx_process_limit = 100;
527 SYSCTL_INT(_hw_em, OID_AUTO, rx_process_limit, CTLFLAG_RDTUN,
528     &em_rx_process_limit, 0,
529     "Maximum number of received packets to process "
530     "at a time, -1 means unlimited");
531 
532 /* Energy efficient ethernet - default to OFF */
533 static int eee_setting = 1;
534 SYSCTL_INT(_hw_em, OID_AUTO, eee_setting, CTLFLAG_RDTUN, &eee_setting, 0,
535     "Enable Energy Efficient Ethernet");
536 
537 /*
538 ** Tuneable Interrupt rate
539 */
540 static int em_max_interrupt_rate = 8000;
541 SYSCTL_INT(_hw_em, OID_AUTO, max_interrupt_rate, CTLFLAG_RDTUN,
542     &em_max_interrupt_rate, 0, "Maximum interrupts per second");
543 
544 
545 
546 /* Global used in WOL setup with multiport cards */
547 static int global_quad_port_a = 0;
548 
549 extern struct if_txrx igb_txrx;
550 extern struct if_txrx em_txrx;
551 extern struct if_txrx lem_txrx;
552 
553 static struct if_shared_ctx em_sctx_init = {
554 	.isc_magic = IFLIB_MAGIC,
555 	.isc_q_align = PAGE_SIZE,
556 	.isc_tx_maxsize = EM_TSO_SIZE + sizeof(struct ether_vlan_header),
557 	.isc_tx_maxsegsize = PAGE_SIZE,
558 	.isc_tso_maxsize = EM_TSO_SIZE + sizeof(struct ether_vlan_header),
559 	.isc_tso_maxsegsize = EM_TSO_SEG_SIZE,
560 	.isc_rx_maxsize = MJUM9BYTES,
561 	.isc_rx_nsegments = 1,
562 	.isc_rx_maxsegsize = MJUM9BYTES,
563 	.isc_nfl = 1,
564 	.isc_nrxqs = 1,
565 	.isc_ntxqs = 1,
566 	.isc_admin_intrcnt = 1,
567 	.isc_vendor_info = em_vendor_info_array,
568 	.isc_driver_version = em_driver_version,
569 	.isc_driver = &em_if_driver,
570 	.isc_flags = IFLIB_NEED_SCRATCH | IFLIB_TSO_INIT_IP | IFLIB_NEED_ZERO_CSUM,
571 
572 	.isc_nrxd_min = {EM_MIN_RXD},
573 	.isc_ntxd_min = {EM_MIN_TXD},
574 	.isc_nrxd_max = {EM_MAX_RXD},
575 	.isc_ntxd_max = {EM_MAX_TXD},
576 	.isc_nrxd_default = {EM_DEFAULT_RXD},
577 	.isc_ntxd_default = {EM_DEFAULT_TXD},
578 };
579 
580 static struct if_shared_ctx igb_sctx_init = {
581 	.isc_magic = IFLIB_MAGIC,
582 	.isc_q_align = PAGE_SIZE,
583 	.isc_tx_maxsize = EM_TSO_SIZE + sizeof(struct ether_vlan_header),
584 	.isc_tx_maxsegsize = PAGE_SIZE,
585 	.isc_tso_maxsize = EM_TSO_SIZE + sizeof(struct ether_vlan_header),
586 	.isc_tso_maxsegsize = EM_TSO_SEG_SIZE,
587 	.isc_rx_maxsize = MJUM9BYTES,
588 	.isc_rx_nsegments = 1,
589 	.isc_rx_maxsegsize = MJUM9BYTES,
590 	.isc_nfl = 1,
591 	.isc_nrxqs = 1,
592 	.isc_ntxqs = 1,
593 	.isc_admin_intrcnt = 1,
594 	.isc_vendor_info = igb_vendor_info_array,
595 	.isc_driver_version = em_driver_version,
596 	.isc_driver = &igb_if_driver,
597 	.isc_flags = IFLIB_NEED_SCRATCH | IFLIB_TSO_INIT_IP | IFLIB_NEED_ZERO_CSUM,
598 
599 	.isc_nrxd_min = {EM_MIN_RXD},
600 	.isc_ntxd_min = {EM_MIN_TXD},
601 	.isc_nrxd_max = {IGB_MAX_RXD},
602 	.isc_ntxd_max = {IGB_MAX_TXD},
603 	.isc_nrxd_default = {EM_DEFAULT_RXD},
604 	.isc_ntxd_default = {EM_DEFAULT_TXD},
605 };
606 
607 /*****************************************************************
608  *
609  * Dump Registers
610  *
611  ****************************************************************/
612 #define IGB_REGS_LEN 739
613 
614 static int em_get_regs(SYSCTL_HANDLER_ARGS)
615 {
616 	struct e1000_softc *sc = (struct e1000_softc *)arg1;
617 	struct e1000_hw *hw = &sc->hw;
618 	struct sbuf *sb;
619 	u32 *regs_buff;
620 	int rc;
621 
622 	regs_buff = malloc(sizeof(u32) * IGB_REGS_LEN, M_DEVBUF, M_WAITOK);
623 	memset(regs_buff, 0, IGB_REGS_LEN * sizeof(u32));
624 
625 	rc = sysctl_wire_old_buffer(req, 0);
626 	MPASS(rc == 0);
627 	if (rc != 0) {
628 		free(regs_buff, M_DEVBUF);
629 		return (rc);
630 	}
631 
632 	sb = sbuf_new_for_sysctl(NULL, NULL, 32*400, req);
633 	MPASS(sb != NULL);
634 	if (sb == NULL) {
635 		free(regs_buff, M_DEVBUF);
636 		return (ENOMEM);
637 	}
638 
639 	/* General Registers */
640 	regs_buff[0] = E1000_READ_REG(hw, E1000_CTRL);
641 	regs_buff[1] = E1000_READ_REG(hw, E1000_STATUS);
642 	regs_buff[2] = E1000_READ_REG(hw, E1000_CTRL_EXT);
643 	regs_buff[3] = E1000_READ_REG(hw, E1000_ICR);
644 	regs_buff[4] = E1000_READ_REG(hw, E1000_RCTL);
645 	regs_buff[5] = E1000_READ_REG(hw, E1000_RDLEN(0));
646 	regs_buff[6] = E1000_READ_REG(hw, E1000_RDH(0));
647 	regs_buff[7] = E1000_READ_REG(hw, E1000_RDT(0));
648 	regs_buff[8] = E1000_READ_REG(hw, E1000_RXDCTL(0));
649 	regs_buff[9] = E1000_READ_REG(hw, E1000_RDBAL(0));
650 	regs_buff[10] = E1000_READ_REG(hw, E1000_RDBAH(0));
651 	regs_buff[11] = E1000_READ_REG(hw, E1000_TCTL);
652 	regs_buff[12] = E1000_READ_REG(hw, E1000_TDBAL(0));
653 	regs_buff[13] = E1000_READ_REG(hw, E1000_TDBAH(0));
654 	regs_buff[14] = E1000_READ_REG(hw, E1000_TDLEN(0));
655 	regs_buff[15] = E1000_READ_REG(hw, E1000_TDH(0));
656 	regs_buff[16] = E1000_READ_REG(hw, E1000_TDT(0));
657 	regs_buff[17] = E1000_READ_REG(hw, E1000_TXDCTL(0));
658 	regs_buff[18] = E1000_READ_REG(hw, E1000_TDFH);
659 	regs_buff[19] = E1000_READ_REG(hw, E1000_TDFT);
660 	regs_buff[20] = E1000_READ_REG(hw, E1000_TDFHS);
661 	regs_buff[21] = E1000_READ_REG(hw, E1000_TDFPC);
662 
663 	sbuf_printf(sb, "General Registers\n");
664 	sbuf_printf(sb, "\tCTRL\t %08x\n", regs_buff[0]);
665 	sbuf_printf(sb, "\tSTATUS\t %08x\n", regs_buff[1]);
666 	sbuf_printf(sb, "\tCTRL_EXT\t %08x\n\n", regs_buff[2]);
667 
668 	sbuf_printf(sb, "Interrupt Registers\n");
669 	sbuf_printf(sb, "\tICR\t %08x\n\n", regs_buff[3]);
670 
671 	sbuf_printf(sb, "RX Registers\n");
672 	sbuf_printf(sb, "\tRCTL\t %08x\n", regs_buff[4]);
673 	sbuf_printf(sb, "\tRDLEN\t %08x\n", regs_buff[5]);
674 	sbuf_printf(sb, "\tRDH\t %08x\n", regs_buff[6]);
675 	sbuf_printf(sb, "\tRDT\t %08x\n", regs_buff[7]);
676 	sbuf_printf(sb, "\tRXDCTL\t %08x\n", regs_buff[8]);
677 	sbuf_printf(sb, "\tRDBAL\t %08x\n", regs_buff[9]);
678 	sbuf_printf(sb, "\tRDBAH\t %08x\n\n", regs_buff[10]);
679 
680 	sbuf_printf(sb, "TX Registers\n");
681 	sbuf_printf(sb, "\tTCTL\t %08x\n", regs_buff[11]);
682 	sbuf_printf(sb, "\tTDBAL\t %08x\n", regs_buff[12]);
683 	sbuf_printf(sb, "\tTDBAH\t %08x\n", regs_buff[13]);
684 	sbuf_printf(sb, "\tTDLEN\t %08x\n", regs_buff[14]);
685 	sbuf_printf(sb, "\tTDH\t %08x\n", regs_buff[15]);
686 	sbuf_printf(sb, "\tTDT\t %08x\n", regs_buff[16]);
687 	sbuf_printf(sb, "\tTXDCTL\t %08x\n", regs_buff[17]);
688 	sbuf_printf(sb, "\tTDFH\t %08x\n", regs_buff[18]);
689 	sbuf_printf(sb, "\tTDFT\t %08x\n", regs_buff[19]);
690 	sbuf_printf(sb, "\tTDFHS\t %08x\n", regs_buff[20]);
691 	sbuf_printf(sb, "\tTDFPC\t %08x\n\n", regs_buff[21]);
692 
693 	free(regs_buff, M_DEVBUF);
694 
695 #ifdef DUMP_DESCS
696 	{
697 		if_softc_ctx_t scctx = sc->shared;
698 		struct rx_ring *rxr = &rx_que->rxr;
699 		struct tx_ring *txr = &tx_que->txr;
700 		int ntxd = scctx->isc_ntxd[0];
701 		int nrxd = scctx->isc_nrxd[0];
702 		int j;
703 
704 	for (j = 0; j < nrxd; j++) {
705 		u32 staterr = le32toh(rxr->rx_base[j].wb.upper.status_error);
706 		u32 length =  le32toh(rxr->rx_base[j].wb.upper.length);
707 		sbuf_printf(sb, "\tReceive Descriptor Address %d: %08" PRIx64 "  Error:%d  Length:%d\n", j, rxr->rx_base[j].read.buffer_addr, staterr, length);
708 	}
709 
710 	for (j = 0; j < min(ntxd, 256); j++) {
711 		unsigned int *ptr = (unsigned int *)&txr->tx_base[j];
712 
713 		sbuf_printf(sb, "\tTXD[%03d] [0]: %08x [1]: %08x [2]: %08x [3]: %08x  eop: %d DD=%d\n",
714 			    j, ptr[0], ptr[1], ptr[2], ptr[3], buf->eop,
715 			    buf->eop != -1 ? txr->tx_base[buf->eop].upper.fields.status & E1000_TXD_STAT_DD : 0);
716 
717 	}
718 	}
719 #endif
720 
721 	rc = sbuf_finish(sb);
722 	sbuf_delete(sb);
723 	return(rc);
724 }
725 
726 static void *
727 em_register(device_t dev)
728 {
729 	return (&em_sctx_init);
730 }
731 
732 static void *
733 igb_register(device_t dev)
734 {
735 	return (&igb_sctx_init);
736 }
737 
738 static int
739 em_set_num_queues(if_ctx_t ctx)
740 {
741 	struct e1000_softc *sc = iflib_get_softc(ctx);
742 	int maxqueues;
743 
744 	/* Sanity check based on HW */
745 	switch (sc->hw.mac.type) {
746 	case e1000_82576:
747 	case e1000_82580:
748 	case e1000_i350:
749 	case e1000_i354:
750 		maxqueues = 8;
751 		break;
752 	case e1000_i210:
753 	case e1000_82575:
754 		maxqueues = 4;
755 		break;
756 	case e1000_i211:
757 	case e1000_82574:
758 		maxqueues = 2;
759 		break;
760 	default:
761 		maxqueues = 1;
762 		break;
763 	}
764 
765 	return (maxqueues);
766 }
767 
768 #define	LEM_CAPS							\
769     IFCAP_HWCSUM | IFCAP_VLAN_MTU | IFCAP_VLAN_HWTAGGING |		\
770     IFCAP_VLAN_HWCSUM | IFCAP_WOL | IFCAP_VLAN_HWFILTER
771 
772 #define	EM_CAPS								\
773     IFCAP_HWCSUM | IFCAP_VLAN_MTU | IFCAP_VLAN_HWTAGGING |		\
774     IFCAP_VLAN_HWCSUM | IFCAP_WOL | IFCAP_VLAN_HWFILTER | IFCAP_TSO4 |	\
775     IFCAP_LRO | IFCAP_VLAN_HWTSO
776 
777 #define	IGB_CAPS							\
778     IFCAP_HWCSUM | IFCAP_VLAN_MTU | IFCAP_VLAN_HWTAGGING |		\
779     IFCAP_VLAN_HWCSUM | IFCAP_WOL | IFCAP_VLAN_HWFILTER | IFCAP_TSO4 |	\
780     IFCAP_LRO | IFCAP_VLAN_HWTSO | IFCAP_JUMBO_MTU | IFCAP_HWCSUM_IPV6 |\
781     IFCAP_TSO6
782 
783 /*********************************************************************
784  *  Device initialization routine
785  *
786  *  The attach entry point is called when the driver is being loaded.
787  *  This routine identifies the type of hardware, allocates all resources
788  *  and initializes the hardware.
789  *
790  *  return 0 on success, positive on failure
791  *********************************************************************/
792 static int
793 em_if_attach_pre(if_ctx_t ctx)
794 {
795 	struct e1000_softc *sc;
796 	if_softc_ctx_t scctx;
797 	device_t dev;
798 	struct e1000_hw *hw;
799 	struct sysctl_oid_list *child;
800 	struct sysctl_ctx_list *ctx_list;
801 	int error = 0;
802 
803 	INIT_DEBUGOUT("em_if_attach_pre: begin");
804 	dev = iflib_get_dev(ctx);
805 	sc = iflib_get_softc(ctx);
806 
807 	sc->ctx = sc->osdep.ctx = ctx;
808 	sc->dev = sc->osdep.dev = dev;
809 	scctx = sc->shared = iflib_get_softc_ctx(ctx);
810 	sc->media = iflib_get_media(ctx);
811 	hw = &sc->hw;
812 
813 	sc->tx_process_limit = scctx->isc_ntxd[0];
814 
815 	/* Determine hardware and mac info */
816 	em_identify_hardware(ctx);
817 
818 	/* SYSCTL stuff */
819 	ctx_list = device_get_sysctl_ctx(dev);
820 	child = SYSCTL_CHILDREN(device_get_sysctl_tree(dev));
821 
822 	SYSCTL_ADD_PROC(ctx_list, child, OID_AUTO, "nvm",
823 	    CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_NEEDGIANT, sc, 0,
824 	    em_sysctl_nvm_info, "I", "NVM Information");
825 
826 	SYSCTL_ADD_PROC(ctx_list, child, OID_AUTO, "fw_version",
827 	    CTLTYPE_STRING | CTLFLAG_RD | CTLFLAG_NEEDGIANT, sc, 0,
828 	    em_sysctl_print_fw_version, "A",
829 	    "Prints FW/NVM Versions");
830 
831 	SYSCTL_ADD_PROC(ctx_list, child, OID_AUTO, "debug",
832 	    CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_NEEDGIANT, sc, 0,
833 	    em_sysctl_debug_info, "I", "Debug Information");
834 
835 	SYSCTL_ADD_PROC(ctx_list, child, OID_AUTO, "fc",
836 	    CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_NEEDGIANT, sc, 0,
837 	    em_set_flowcntl, "I", "Flow Control");
838 
839 	SYSCTL_ADD_PROC(ctx_list, child, OID_AUTO, "reg_dump",
840 	    CTLTYPE_STRING | CTLFLAG_RD | CTLFLAG_NEEDGIANT, sc, 0,
841 	    em_get_regs, "A", "Dump Registers");
842 
843 	SYSCTL_ADD_PROC(ctx_list, child, OID_AUTO, "rs_dump",
844 	    CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_NEEDGIANT, sc, 0,
845 	    em_get_rs, "I", "Dump RS indexes");
846 
847 	scctx->isc_tx_nsegments = EM_MAX_SCATTER;
848 	scctx->isc_nrxqsets_max = scctx->isc_ntxqsets_max = em_set_num_queues(ctx);
849 	if (bootverbose)
850 		device_printf(dev, "attach_pre capping queues at %d\n",
851 		    scctx->isc_ntxqsets_max);
852 
853 	if (hw->mac.type >= igb_mac_min) {
854 		scctx->isc_txqsizes[0] = roundup2(scctx->isc_ntxd[0] * sizeof(union e1000_adv_tx_desc), EM_DBA_ALIGN);
855 		scctx->isc_rxqsizes[0] = roundup2(scctx->isc_nrxd[0] * sizeof(union e1000_adv_rx_desc), EM_DBA_ALIGN);
856 		scctx->isc_txd_size[0] = sizeof(union e1000_adv_tx_desc);
857 		scctx->isc_rxd_size[0] = sizeof(union e1000_adv_rx_desc);
858 		scctx->isc_txrx = &igb_txrx;
859 		scctx->isc_tx_tso_segments_max = EM_MAX_SCATTER;
860 		scctx->isc_tx_tso_size_max = EM_TSO_SIZE;
861 		scctx->isc_tx_tso_segsize_max = EM_TSO_SEG_SIZE;
862 		scctx->isc_capabilities = scctx->isc_capenable = IGB_CAPS;
863 		scctx->isc_tx_csum_flags = CSUM_TCP | CSUM_UDP | CSUM_TSO |
864 		     CSUM_IP6_TCP | CSUM_IP6_UDP;
865 		if (hw->mac.type != e1000_82575)
866 			scctx->isc_tx_csum_flags |= CSUM_SCTP | CSUM_IP6_SCTP;
867 		/*
868 		** Some new devices, as with ixgbe, now may
869 		** use a different BAR, so we need to keep
870 		** track of which is used.
871 		*/
872 		scctx->isc_msix_bar = pci_msix_table_bar(dev);
873 	} else if (hw->mac.type >= em_mac_min) {
874 		scctx->isc_txqsizes[0] = roundup2(scctx->isc_ntxd[0]* sizeof(struct e1000_tx_desc), EM_DBA_ALIGN);
875 		scctx->isc_rxqsizes[0] = roundup2(scctx->isc_nrxd[0] * sizeof(union e1000_rx_desc_extended), EM_DBA_ALIGN);
876 		scctx->isc_txd_size[0] = sizeof(struct e1000_tx_desc);
877 		scctx->isc_rxd_size[0] = sizeof(union e1000_rx_desc_extended);
878 		scctx->isc_txrx = &em_txrx;
879 		scctx->isc_tx_tso_segments_max = EM_MAX_SCATTER;
880 		scctx->isc_tx_tso_size_max = EM_TSO_SIZE;
881 		scctx->isc_tx_tso_segsize_max = EM_TSO_SEG_SIZE;
882 		scctx->isc_capabilities = scctx->isc_capenable = EM_CAPS;
883 		/*
884 		 * For EM-class devices, don't enable IFCAP_{TSO4,VLAN_HWTSO}
885 		 * by default as we don't have workarounds for all associated
886 		 * silicon errata.  E. g., with several MACs such as 82573E,
887 		 * TSO only works at Gigabit speed and otherwise can cause the
888 		 * hardware to hang (which also would be next to impossible to
889 		 * work around given that already queued TSO-using descriptors
890 		 * would need to be flushed and vlan(4) reconfigured at runtime
891 		 * in case of a link speed change).  Moreover, MACs like 82579
892 		 * still can hang at Gigabit even with all publicly documented
893 		 * TSO workarounds implemented.  Generally, the penality of
894 		 * these workarounds is rather high and may involve copying
895 		 * mbuf data around so advantages of TSO lapse.  Still, TSO may
896 		 * work for a few MACs of this class - at least when sticking
897 		 * with Gigabit - in which case users may enable TSO manually.
898 		 */
899 		scctx->isc_capenable &= ~(IFCAP_TSO4 | IFCAP_VLAN_HWTSO);
900 		scctx->isc_tx_csum_flags = CSUM_TCP | CSUM_UDP | CSUM_IP_TSO;
901 		/*
902 		 * We support MSI-X with 82574 only, but indicate to iflib(4)
903 		 * that it shall give MSI at least a try with other devices.
904 		 */
905 		if (hw->mac.type == e1000_82574) {
906 			scctx->isc_msix_bar = pci_msix_table_bar(dev);;
907 		} else {
908 			scctx->isc_msix_bar = -1;
909 			scctx->isc_disable_msix = 1;
910 		}
911 	} else {
912 		scctx->isc_txqsizes[0] = roundup2((scctx->isc_ntxd[0] + 1) * sizeof(struct e1000_tx_desc), EM_DBA_ALIGN);
913 		scctx->isc_rxqsizes[0] = roundup2((scctx->isc_nrxd[0] + 1) * sizeof(struct e1000_rx_desc), EM_DBA_ALIGN);
914 		scctx->isc_txd_size[0] = sizeof(struct e1000_tx_desc);
915 		scctx->isc_rxd_size[0] = sizeof(struct e1000_rx_desc);
916 		scctx->isc_tx_csum_flags = CSUM_TCP | CSUM_UDP;
917 		scctx->isc_txrx = &lem_txrx;
918 		scctx->isc_capabilities = LEM_CAPS;
919 		if (hw->mac.type < e1000_82543)
920 			scctx->isc_capabilities &= ~(IFCAP_HWCSUM|IFCAP_VLAN_HWCSUM);
921 		/* 82541ER doesn't do HW tagging */
922 		if (hw->device_id == E1000_DEV_ID_82541ER || hw->device_id == E1000_DEV_ID_82541ER_LOM)
923 			scctx->isc_capabilities &= ~IFCAP_VLAN_HWTAGGING;
924 		/* INTx only */
925 		scctx->isc_msix_bar = 0;
926 		scctx->isc_capenable = scctx->isc_capabilities;
927 	}
928 
929 	/* Setup PCI resources */
930 	if (em_allocate_pci_resources(ctx)) {
931 		device_printf(dev, "Allocation of PCI resources failed\n");
932 		error = ENXIO;
933 		goto err_pci;
934 	}
935 
936 	/*
937 	** For ICH8 and family we need to
938 	** map the flash memory, and this
939 	** must happen after the MAC is
940 	** identified
941 	*/
942 	if ((hw->mac.type == e1000_ich8lan) ||
943 	    (hw->mac.type == e1000_ich9lan) ||
944 	    (hw->mac.type == e1000_ich10lan) ||
945 	    (hw->mac.type == e1000_pchlan) ||
946 	    (hw->mac.type == e1000_pch2lan) ||
947 	    (hw->mac.type == e1000_pch_lpt)) {
948 		int rid = EM_BAR_TYPE_FLASH;
949 		sc->flash = bus_alloc_resource_any(dev,
950 		    SYS_RES_MEMORY, &rid, RF_ACTIVE);
951 		if (sc->flash == NULL) {
952 			device_printf(dev, "Mapping of Flash failed\n");
953 			error = ENXIO;
954 			goto err_pci;
955 		}
956 		/* This is used in the shared code */
957 		hw->flash_address = (u8 *)sc->flash;
958 		sc->osdep.flash_bus_space_tag =
959 		    rman_get_bustag(sc->flash);
960 		sc->osdep.flash_bus_space_handle =
961 		    rman_get_bushandle(sc->flash);
962 	}
963 	/*
964 	** In the new SPT device flash is not  a
965 	** separate BAR, rather it is also in BAR0,
966 	** so use the same tag and an offset handle for the
967 	** FLASH read/write macros in the shared code.
968 	*/
969 	else if (hw->mac.type >= e1000_pch_spt) {
970 		sc->osdep.flash_bus_space_tag =
971 		    sc->osdep.mem_bus_space_tag;
972 		sc->osdep.flash_bus_space_handle =
973 		    sc->osdep.mem_bus_space_handle
974 		    + E1000_FLASH_BASE_ADDR;
975 	}
976 
977 	/* Do Shared Code initialization */
978 	error = e1000_setup_init_funcs(hw, true);
979 	if (error) {
980 		device_printf(dev, "Setup of Shared code failed, error %d\n",
981 		    error);
982 		error = ENXIO;
983 		goto err_pci;
984 	}
985 
986 	em_setup_msix(ctx);
987 	e1000_get_bus_info(hw);
988 
989 	/* Set up some sysctls for the tunable interrupt delays */
990 	em_add_int_delay_sysctl(sc, "rx_int_delay",
991 	    "receive interrupt delay in usecs", &sc->rx_int_delay,
992 	    E1000_REGISTER(hw, E1000_RDTR), em_rx_int_delay_dflt);
993 	em_add_int_delay_sysctl(sc, "tx_int_delay",
994 	    "transmit interrupt delay in usecs", &sc->tx_int_delay,
995 	    E1000_REGISTER(hw, E1000_TIDV), em_tx_int_delay_dflt);
996 	em_add_int_delay_sysctl(sc, "rx_abs_int_delay",
997 	    "receive interrupt delay limit in usecs",
998 	    &sc->rx_abs_int_delay,
999 	    E1000_REGISTER(hw, E1000_RADV),
1000 	    em_rx_abs_int_delay_dflt);
1001 	em_add_int_delay_sysctl(sc, "tx_abs_int_delay",
1002 	    "transmit interrupt delay limit in usecs",
1003 	    &sc->tx_abs_int_delay,
1004 	    E1000_REGISTER(hw, E1000_TADV),
1005 	    em_tx_abs_int_delay_dflt);
1006 	em_add_int_delay_sysctl(sc, "itr",
1007 	    "interrupt delay limit in usecs/4",
1008 	    &sc->tx_itr,
1009 	    E1000_REGISTER(hw, E1000_ITR),
1010 	    DEFAULT_ITR);
1011 
1012 	hw->mac.autoneg = DO_AUTO_NEG;
1013 	hw->phy.autoneg_wait_to_complete = false;
1014 	hw->phy.autoneg_advertised = AUTONEG_ADV_DEFAULT;
1015 
1016 	if (hw->mac.type < em_mac_min) {
1017 		e1000_init_script_state_82541(hw, true);
1018 		e1000_set_tbi_compatibility_82543(hw, true);
1019 	}
1020 	/* Copper options */
1021 	if (hw->phy.media_type == e1000_media_type_copper) {
1022 		hw->phy.mdix = AUTO_ALL_MODES;
1023 		hw->phy.disable_polarity_correction = false;
1024 		hw->phy.ms_type = EM_MASTER_SLAVE;
1025 	}
1026 
1027 	/*
1028 	 * Set the frame limits assuming
1029 	 * standard ethernet sized frames.
1030 	 */
1031 	scctx->isc_max_frame_size = hw->mac.max_frame_size =
1032 	    ETHERMTU + ETHER_HDR_LEN + ETHERNET_FCS_SIZE;
1033 
1034 	/*
1035 	 * This controls when hardware reports transmit completion
1036 	 * status.
1037 	 */
1038 	hw->mac.report_tx_early = 1;
1039 
1040 	/* Allocate multicast array memory. */
1041 	sc->mta = malloc(sizeof(u8) * ETHER_ADDR_LEN *
1042 	    MAX_NUM_MULTICAST_ADDRESSES, M_DEVBUF, M_NOWAIT);
1043 	if (sc->mta == NULL) {
1044 		device_printf(dev, "Can not allocate multicast setup array\n");
1045 		error = ENOMEM;
1046 		goto err_late;
1047 	}
1048 
1049 	/* Check SOL/IDER usage */
1050 	if (e1000_check_reset_block(hw))
1051 		device_printf(dev, "PHY reset is blocked"
1052 			      " due to SOL/IDER session.\n");
1053 
1054 	/* Sysctl for setting Energy Efficient Ethernet */
1055 	hw->dev_spec.ich8lan.eee_disable = eee_setting;
1056 	SYSCTL_ADD_PROC(ctx_list, child, OID_AUTO, "eee_control",
1057 	    CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_NEEDGIANT, sc, 0,
1058 	    em_sysctl_eee, "I", "Disable Energy Efficient Ethernet");
1059 
1060 	/*
1061 	** Start from a known state, this is
1062 	** important in reading the nvm and
1063 	** mac from that.
1064 	*/
1065 	e1000_reset_hw(hw);
1066 
1067 	/* Make sure we have a good EEPROM before we read from it */
1068 	if (e1000_validate_nvm_checksum(hw) < 0) {
1069 		/*
1070 		** Some PCI-E parts fail the first check due to
1071 		** the link being in sleep state, call it again,
1072 		** if it fails a second time its a real issue.
1073 		*/
1074 		if (e1000_validate_nvm_checksum(hw) < 0) {
1075 			device_printf(dev,
1076 			    "The EEPROM Checksum Is Not Valid\n");
1077 			error = EIO;
1078 			goto err_late;
1079 		}
1080 	}
1081 
1082 	/* Copy the permanent MAC address out of the EEPROM */
1083 	if (e1000_read_mac_addr(hw) < 0) {
1084 		device_printf(dev, "EEPROM read error while reading MAC"
1085 			      " address\n");
1086 		error = EIO;
1087 		goto err_late;
1088 	}
1089 
1090 	if (!em_is_valid_ether_addr(hw->mac.addr)) {
1091 		if (sc->vf_ifp) {
1092 			ether_gen_addr(iflib_get_ifp(ctx),
1093 			    (struct ether_addr *)hw->mac.addr);
1094 		} else {
1095 			device_printf(dev, "Invalid MAC address\n");
1096 			error = EIO;
1097 			goto err_late;
1098 		}
1099 	}
1100 
1101 	/* Save the EEPROM/NVM versions, must be done under IFLIB_CTX_LOCK */
1102 	em_fw_version_locked(ctx);
1103 
1104 	em_print_fw_version(sc);
1105 
1106 	/* Disable ULP support */
1107 	e1000_disable_ulp_lpt_lp(hw, true);
1108 
1109 	/*
1110 	 * Get Wake-on-Lan and Management info for later use
1111 	 */
1112 	em_get_wakeup(ctx);
1113 
1114 	/* Enable only WOL MAGIC by default */
1115 	scctx->isc_capenable &= ~IFCAP_WOL;
1116 	if (sc->wol != 0)
1117 		scctx->isc_capenable |= IFCAP_WOL_MAGIC;
1118 
1119 	iflib_set_mac(ctx, hw->mac.addr);
1120 
1121 	return (0);
1122 
1123 err_late:
1124 	em_release_hw_control(sc);
1125 err_pci:
1126 	em_free_pci_resources(ctx);
1127 	free(sc->mta, M_DEVBUF);
1128 
1129 	return (error);
1130 }
1131 
1132 static int
1133 em_if_attach_post(if_ctx_t ctx)
1134 {
1135 	struct e1000_softc *sc = iflib_get_softc(ctx);
1136 	struct e1000_hw *hw = &sc->hw;
1137 	int error = 0;
1138 
1139 	/* Setup OS specific network interface */
1140 	error = em_setup_interface(ctx);
1141 	if (error != 0) {
1142 		device_printf(sc->dev, "Interface setup failed: %d\n", error);
1143 		goto err_late;
1144 	}
1145 
1146 	em_reset(ctx);
1147 
1148 	/* Initialize statistics */
1149 	em_update_stats_counters(sc);
1150 	hw->mac.get_link_status = 1;
1151 	em_if_update_admin_status(ctx);
1152 	em_add_hw_stats(sc);
1153 
1154 	/* Non-AMT based hardware can now take control from firmware */
1155 	if (sc->has_manage && !sc->has_amt)
1156 		em_get_hw_control(sc);
1157 
1158 	INIT_DEBUGOUT("em_if_attach_post: end");
1159 
1160 	return (0);
1161 
1162 err_late:
1163 	/* upon attach_post() error, iflib calls _if_detach() to free resources. */
1164 	return (error);
1165 }
1166 
1167 /*********************************************************************
1168  *  Device removal routine
1169  *
1170  *  The detach entry point is called when the driver is being removed.
1171  *  This routine stops the adapter and deallocates all the resources
1172  *  that were allocated for driver operation.
1173  *
1174  *  return 0 on success, positive on failure
1175  *********************************************************************/
1176 static int
1177 em_if_detach(if_ctx_t ctx)
1178 {
1179 	struct e1000_softc	*sc = iflib_get_softc(ctx);
1180 
1181 	INIT_DEBUGOUT("em_if_detach: begin");
1182 
1183 	e1000_phy_hw_reset(&sc->hw);
1184 
1185 	em_release_manageability(sc);
1186 	em_release_hw_control(sc);
1187 	em_free_pci_resources(ctx);
1188 	free(sc->mta, M_DEVBUF);
1189 	sc->mta = NULL;
1190 
1191 	return (0);
1192 }
1193 
1194 /*********************************************************************
1195  *
1196  *  Shutdown entry point
1197  *
1198  **********************************************************************/
1199 
1200 static int
1201 em_if_shutdown(if_ctx_t ctx)
1202 {
1203 	return em_if_suspend(ctx);
1204 }
1205 
1206 /*
1207  * Suspend/resume device methods.
1208  */
1209 static int
1210 em_if_suspend(if_ctx_t ctx)
1211 {
1212 	struct e1000_softc *sc = iflib_get_softc(ctx);
1213 
1214 	em_release_manageability(sc);
1215 	em_release_hw_control(sc);
1216 	em_enable_wakeup(ctx);
1217 	return (0);
1218 }
1219 
1220 static int
1221 em_if_resume(if_ctx_t ctx)
1222 {
1223 	struct e1000_softc *sc = iflib_get_softc(ctx);
1224 
1225 	if (sc->hw.mac.type == e1000_pch2lan)
1226 		e1000_resume_workarounds_pchlan(&sc->hw);
1227 	em_if_init(ctx);
1228 	em_init_manageability(sc);
1229 
1230 	return(0);
1231 }
1232 
1233 static int
1234 em_if_mtu_set(if_ctx_t ctx, uint32_t mtu)
1235 {
1236 	int max_frame_size;
1237 	struct e1000_softc *sc = iflib_get_softc(ctx);
1238 	if_softc_ctx_t scctx = iflib_get_softc_ctx(ctx);
1239 
1240 	IOCTL_DEBUGOUT("ioctl rcv'd: SIOCSIFMTU (Set Interface MTU)");
1241 
1242 	switch (sc->hw.mac.type) {
1243 	case e1000_82571:
1244 	case e1000_82572:
1245 	case e1000_ich9lan:
1246 	case e1000_ich10lan:
1247 	case e1000_pch2lan:
1248 	case e1000_pch_lpt:
1249 	case e1000_pch_spt:
1250 	case e1000_pch_cnp:
1251 	case e1000_pch_tgp:
1252 	case e1000_pch_adp:
1253 	case e1000_pch_mtp:
1254 	case e1000_82574:
1255 	case e1000_82583:
1256 	case e1000_80003es2lan:
1257 		/* 9K Jumbo Frame size */
1258 		max_frame_size = 9234;
1259 		break;
1260 	case e1000_pchlan:
1261 		max_frame_size = 4096;
1262 		break;
1263 	case e1000_82542:
1264 	case e1000_ich8lan:
1265 		/* Adapters that do not support jumbo frames */
1266 		max_frame_size = ETHER_MAX_LEN;
1267 		break;
1268 	default:
1269 		if (sc->hw.mac.type >= igb_mac_min)
1270 			max_frame_size = 9234;
1271 		else /* lem */
1272 			max_frame_size = MAX_JUMBO_FRAME_SIZE;
1273 	}
1274 	if (mtu > max_frame_size - ETHER_HDR_LEN - ETHER_CRC_LEN) {
1275 		return (EINVAL);
1276 	}
1277 
1278 	scctx->isc_max_frame_size = sc->hw.mac.max_frame_size =
1279 	    mtu + ETHER_HDR_LEN + ETHER_CRC_LEN;
1280 	return (0);
1281 }
1282 
1283 /*********************************************************************
1284  *  Init entry point
1285  *
1286  *  This routine is used in two ways. It is used by the stack as
1287  *  init entry point in network interface structure. It is also used
1288  *  by the driver as a hw/sw initialization routine to get to a
1289  *  consistent state.
1290  *
1291  **********************************************************************/
1292 static void
1293 em_if_init(if_ctx_t ctx)
1294 {
1295 	struct e1000_softc *sc = iflib_get_softc(ctx);
1296 	if_softc_ctx_t scctx = sc->shared;
1297 	struct ifnet *ifp = iflib_get_ifp(ctx);
1298 	struct em_tx_queue *tx_que;
1299 	int i;
1300 
1301 	INIT_DEBUGOUT("em_if_init: begin");
1302 
1303 	/* Get the latest mac address, User can use a LAA */
1304 	bcopy(if_getlladdr(ifp), sc->hw.mac.addr,
1305 	    ETHER_ADDR_LEN);
1306 
1307 	/* Put the address into the Receive Address Array */
1308 	e1000_rar_set(&sc->hw, sc->hw.mac.addr, 0);
1309 
1310 	/*
1311 	 * With the 82571 adapter, RAR[0] may be overwritten
1312 	 * when the other port is reset, we make a duplicate
1313 	 * in RAR[14] for that eventuality, this assures
1314 	 * the interface continues to function.
1315 	 */
1316 	if (sc->hw.mac.type == e1000_82571) {
1317 		e1000_set_laa_state_82571(&sc->hw, true);
1318 		e1000_rar_set(&sc->hw, sc->hw.mac.addr,
1319 		    E1000_RAR_ENTRIES - 1);
1320 	}
1321 
1322 
1323 	/* Initialize the hardware */
1324 	em_reset(ctx);
1325 	em_if_update_admin_status(ctx);
1326 
1327 	for (i = 0, tx_que = sc->tx_queues; i < sc->tx_num_queues; i++, tx_que++) {
1328 		struct tx_ring *txr = &tx_que->txr;
1329 
1330 		txr->tx_rs_cidx = txr->tx_rs_pidx;
1331 
1332 		/* Initialize the last processed descriptor to be the end of
1333 		 * the ring, rather than the start, so that we avoid an
1334 		 * off-by-one error when calculating how many descriptors are
1335 		 * done in the credits_update function.
1336 		 */
1337 		txr->tx_cidx_processed = scctx->isc_ntxd[0] - 1;
1338 	}
1339 
1340 	/* Setup VLAN support, basic and offload if available */
1341 	E1000_WRITE_REG(&sc->hw, E1000_VET, ETHERTYPE_VLAN);
1342 
1343 	/* Clear bad data from Rx FIFOs */
1344 	if (sc->hw.mac.type >= igb_mac_min)
1345 		e1000_rx_fifo_flush_base(&sc->hw);
1346 
1347 	/* Configure for OS presence */
1348 	em_init_manageability(sc);
1349 
1350 	/* Prepare transmit descriptors and buffers */
1351 	em_initialize_transmit_unit(ctx);
1352 
1353 	/* Setup Multicast table */
1354 	em_if_multi_set(ctx);
1355 
1356 	sc->rx_mbuf_sz = iflib_get_rx_mbuf_sz(ctx);
1357 	em_initialize_receive_unit(ctx);
1358 
1359 	/* Set up VLAN support and filter */
1360 	em_setup_vlan_hw_support(ctx);
1361 
1362 	/* Don't lose promiscuous settings */
1363 	em_if_set_promisc(ctx, if_getflags(ifp));
1364 	e1000_clear_hw_cntrs_base_generic(&sc->hw);
1365 
1366 	/* MSI-X configuration for 82574 */
1367 	if (sc->hw.mac.type == e1000_82574) {
1368 		int tmp = E1000_READ_REG(&sc->hw, E1000_CTRL_EXT);
1369 
1370 		tmp |= E1000_CTRL_EXT_PBA_CLR;
1371 		E1000_WRITE_REG(&sc->hw, E1000_CTRL_EXT, tmp);
1372 		/* Set the IVAR - interrupt vector routing. */
1373 		E1000_WRITE_REG(&sc->hw, E1000_IVAR, sc->ivars);
1374 	} else if (sc->intr_type == IFLIB_INTR_MSIX) /* Set up queue routing */
1375 		igb_configure_queues(sc);
1376 
1377 	/* this clears any pending interrupts */
1378 	E1000_READ_REG(&sc->hw, E1000_ICR);
1379 	E1000_WRITE_REG(&sc->hw, E1000_ICS, E1000_ICS_LSC);
1380 
1381 	/* AMT based hardware can now take control from firmware */
1382 	if (sc->has_manage && sc->has_amt)
1383 		em_get_hw_control(sc);
1384 
1385 	/* Set Energy Efficient Ethernet */
1386 	if (sc->hw.mac.type >= igb_mac_min &&
1387 	    sc->hw.phy.media_type == e1000_media_type_copper) {
1388 		if (sc->hw.mac.type == e1000_i354)
1389 			e1000_set_eee_i354(&sc->hw, true, true);
1390 		else
1391 			e1000_set_eee_i350(&sc->hw, true, true);
1392 	}
1393 }
1394 
1395 /*********************************************************************
1396  *
1397  *  Fast Legacy/MSI Combined Interrupt Service routine
1398  *
1399  *********************************************************************/
1400 int
1401 em_intr(void *arg)
1402 {
1403 	struct e1000_softc *sc = arg;
1404 	if_ctx_t ctx = sc->ctx;
1405 	u32 reg_icr;
1406 
1407 	reg_icr = E1000_READ_REG(&sc->hw, E1000_ICR);
1408 
1409 	/* Hot eject? */
1410 	if (reg_icr == 0xffffffff)
1411 		return FILTER_STRAY;
1412 
1413 	/* Definitely not our interrupt. */
1414 	if (reg_icr == 0x0)
1415 		return FILTER_STRAY;
1416 
1417 	/*
1418 	 * Starting with the 82571 chip, bit 31 should be used to
1419 	 * determine whether the interrupt belongs to us.
1420 	 */
1421 	if (sc->hw.mac.type >= e1000_82571 &&
1422 	    (reg_icr & E1000_ICR_INT_ASSERTED) == 0)
1423 		return FILTER_STRAY;
1424 
1425 	/*
1426 	 * Only MSI-X interrupts have one-shot behavior by taking advantage
1427 	 * of the EIAC register.  Thus, explicitly disable interrupts.  This
1428 	 * also works around the MSI message reordering errata on certain
1429 	 * systems.
1430 	 */
1431 	IFDI_INTR_DISABLE(ctx);
1432 
1433 	/* Link status change */
1434 	if (reg_icr & (E1000_ICR_RXSEQ | E1000_ICR_LSC))
1435 		em_handle_link(ctx);
1436 
1437 	if (reg_icr & E1000_ICR_RXO)
1438 		sc->rx_overruns++;
1439 
1440 	return (FILTER_SCHEDULE_THREAD);
1441 }
1442 
1443 static int
1444 em_if_rx_queue_intr_enable(if_ctx_t ctx, uint16_t rxqid)
1445 {
1446 	struct e1000_softc *sc = iflib_get_softc(ctx);
1447 	struct em_rx_queue *rxq = &sc->rx_queues[rxqid];
1448 
1449 	E1000_WRITE_REG(&sc->hw, E1000_IMS, rxq->eims);
1450 	return (0);
1451 }
1452 
1453 static int
1454 em_if_tx_queue_intr_enable(if_ctx_t ctx, uint16_t txqid)
1455 {
1456 	struct e1000_softc *sc = iflib_get_softc(ctx);
1457 	struct em_tx_queue *txq = &sc->tx_queues[txqid];
1458 
1459 	E1000_WRITE_REG(&sc->hw, E1000_IMS, txq->eims);
1460 	return (0);
1461 }
1462 
1463 static int
1464 igb_if_rx_queue_intr_enable(if_ctx_t ctx, uint16_t rxqid)
1465 {
1466 	struct e1000_softc *sc = iflib_get_softc(ctx);
1467 	struct em_rx_queue *rxq = &sc->rx_queues[rxqid];
1468 
1469 	E1000_WRITE_REG(&sc->hw, E1000_EIMS, rxq->eims);
1470 	return (0);
1471 }
1472 
1473 static int
1474 igb_if_tx_queue_intr_enable(if_ctx_t ctx, uint16_t txqid)
1475 {
1476 	struct e1000_softc *sc = iflib_get_softc(ctx);
1477 	struct em_tx_queue *txq = &sc->tx_queues[txqid];
1478 
1479 	E1000_WRITE_REG(&sc->hw, E1000_EIMS, txq->eims);
1480 	return (0);
1481 }
1482 
1483 /*********************************************************************
1484  *
1485  *  MSI-X RX Interrupt Service routine
1486  *
1487  **********************************************************************/
1488 static int
1489 em_msix_que(void *arg)
1490 {
1491 	struct em_rx_queue *que = arg;
1492 
1493 	++que->irqs;
1494 
1495 	return (FILTER_SCHEDULE_THREAD);
1496 }
1497 
1498 /*********************************************************************
1499  *
1500  *  MSI-X Link Fast Interrupt Service routine
1501  *
1502  **********************************************************************/
1503 static int
1504 em_msix_link(void *arg)
1505 {
1506 	struct e1000_softc *sc = arg;
1507 	u32 reg_icr;
1508 
1509 	++sc->link_irq;
1510 	MPASS(sc->hw.back != NULL);
1511 	reg_icr = E1000_READ_REG(&sc->hw, E1000_ICR);
1512 
1513 	if (reg_icr & E1000_ICR_RXO)
1514 		sc->rx_overruns++;
1515 
1516 	if (reg_icr & (E1000_ICR_RXSEQ | E1000_ICR_LSC))
1517 		em_handle_link(sc->ctx);
1518 
1519 	/* Re-arm unconditionally */
1520 	if (sc->hw.mac.type >= igb_mac_min) {
1521 		E1000_WRITE_REG(&sc->hw, E1000_IMS, E1000_IMS_LSC);
1522 		E1000_WRITE_REG(&sc->hw, E1000_EIMS, sc->link_mask);
1523 	} else if (sc->hw.mac.type == e1000_82574) {
1524 		E1000_WRITE_REG(&sc->hw, E1000_IMS, E1000_IMS_LSC |
1525 		    E1000_IMS_OTHER);
1526 		/*
1527 		 * Because we must read the ICR for this interrupt it may
1528 		 * clear other causes using autoclear, for this reason we
1529 		 * simply create a soft interrupt for all these vectors.
1530 		 */
1531 		if (reg_icr)
1532 			E1000_WRITE_REG(&sc->hw, E1000_ICS, sc->ims);
1533 	} else
1534 		E1000_WRITE_REG(&sc->hw, E1000_IMS, E1000_IMS_LSC);
1535 
1536 	return (FILTER_HANDLED);
1537 }
1538 
1539 static void
1540 em_handle_link(void *context)
1541 {
1542 	if_ctx_t ctx = context;
1543 	struct e1000_softc *sc = iflib_get_softc(ctx);
1544 
1545 	sc->hw.mac.get_link_status = 1;
1546 	iflib_admin_intr_deferred(ctx);
1547 }
1548 
1549 /*********************************************************************
1550  *
1551  *  Media Ioctl callback
1552  *
1553  *  This routine is called whenever the user queries the status of
1554  *  the interface using ifconfig.
1555  *
1556  **********************************************************************/
1557 static void
1558 em_if_media_status(if_ctx_t ctx, struct ifmediareq *ifmr)
1559 {
1560 	struct e1000_softc *sc = iflib_get_softc(ctx);
1561 	u_char fiber_type = IFM_1000_SX;
1562 
1563 	INIT_DEBUGOUT("em_if_media_status: begin");
1564 
1565 	iflib_admin_intr_deferred(ctx);
1566 
1567 	ifmr->ifm_status = IFM_AVALID;
1568 	ifmr->ifm_active = IFM_ETHER;
1569 
1570 	if (!sc->link_active) {
1571 		return;
1572 	}
1573 
1574 	ifmr->ifm_status |= IFM_ACTIVE;
1575 
1576 	if ((sc->hw.phy.media_type == e1000_media_type_fiber) ||
1577 	    (sc->hw.phy.media_type == e1000_media_type_internal_serdes)) {
1578 		if (sc->hw.mac.type == e1000_82545)
1579 			fiber_type = IFM_1000_LX;
1580 		ifmr->ifm_active |= fiber_type | IFM_FDX;
1581 	} else {
1582 		switch (sc->link_speed) {
1583 		case 10:
1584 			ifmr->ifm_active |= IFM_10_T;
1585 			break;
1586 		case 100:
1587 			ifmr->ifm_active |= IFM_100_TX;
1588 			break;
1589 		case 1000:
1590 			ifmr->ifm_active |= IFM_1000_T;
1591 			break;
1592 		}
1593 		if (sc->link_duplex == FULL_DUPLEX)
1594 			ifmr->ifm_active |= IFM_FDX;
1595 		else
1596 			ifmr->ifm_active |= IFM_HDX;
1597 	}
1598 }
1599 
1600 /*********************************************************************
1601  *
1602  *  Media Ioctl callback
1603  *
1604  *  This routine is called when the user changes speed/duplex using
1605  *  media/mediopt option with ifconfig.
1606  *
1607  **********************************************************************/
1608 static int
1609 em_if_media_change(if_ctx_t ctx)
1610 {
1611 	struct e1000_softc *sc = iflib_get_softc(ctx);
1612 	struct ifmedia *ifm = iflib_get_media(ctx);
1613 
1614 	INIT_DEBUGOUT("em_if_media_change: begin");
1615 
1616 	if (IFM_TYPE(ifm->ifm_media) != IFM_ETHER)
1617 		return (EINVAL);
1618 
1619 	switch (IFM_SUBTYPE(ifm->ifm_media)) {
1620 	case IFM_AUTO:
1621 		sc->hw.mac.autoneg = DO_AUTO_NEG;
1622 		sc->hw.phy.autoneg_advertised = AUTONEG_ADV_DEFAULT;
1623 		break;
1624 	case IFM_1000_LX:
1625 	case IFM_1000_SX:
1626 	case IFM_1000_T:
1627 		sc->hw.mac.autoneg = DO_AUTO_NEG;
1628 		sc->hw.phy.autoneg_advertised = ADVERTISE_1000_FULL;
1629 		break;
1630 	case IFM_100_TX:
1631 		sc->hw.mac.autoneg = false;
1632 		sc->hw.phy.autoneg_advertised = 0;
1633 		if ((ifm->ifm_media & IFM_GMASK) == IFM_FDX)
1634 			sc->hw.mac.forced_speed_duplex = ADVERTISE_100_FULL;
1635 		else
1636 			sc->hw.mac.forced_speed_duplex = ADVERTISE_100_HALF;
1637 		break;
1638 	case IFM_10_T:
1639 		sc->hw.mac.autoneg = false;
1640 		sc->hw.phy.autoneg_advertised = 0;
1641 		if ((ifm->ifm_media & IFM_GMASK) == IFM_FDX)
1642 			sc->hw.mac.forced_speed_duplex = ADVERTISE_10_FULL;
1643 		else
1644 			sc->hw.mac.forced_speed_duplex = ADVERTISE_10_HALF;
1645 		break;
1646 	default:
1647 		device_printf(sc->dev, "Unsupported media type\n");
1648 	}
1649 
1650 	em_if_init(ctx);
1651 
1652 	return (0);
1653 }
1654 
1655 static int
1656 em_if_set_promisc(if_ctx_t ctx, int flags)
1657 {
1658 	struct e1000_softc *sc = iflib_get_softc(ctx);
1659 	struct ifnet *ifp = iflib_get_ifp(ctx);
1660 	u32 reg_rctl;
1661 	int mcnt = 0;
1662 
1663 	reg_rctl = E1000_READ_REG(&sc->hw, E1000_RCTL);
1664 	reg_rctl &= ~(E1000_RCTL_SBP | E1000_RCTL_UPE);
1665 	if (flags & IFF_ALLMULTI)
1666 		mcnt = MAX_NUM_MULTICAST_ADDRESSES;
1667 	else
1668 		mcnt = min(if_llmaddr_count(ifp), MAX_NUM_MULTICAST_ADDRESSES);
1669 
1670 	if (mcnt < MAX_NUM_MULTICAST_ADDRESSES)
1671 		reg_rctl &= (~E1000_RCTL_MPE);
1672 	E1000_WRITE_REG(&sc->hw, E1000_RCTL, reg_rctl);
1673 
1674 	if (flags & IFF_PROMISC) {
1675 		reg_rctl |= (E1000_RCTL_UPE | E1000_RCTL_MPE);
1676 		em_if_vlan_filter_disable(sc);
1677 		/* Turn this on if you want to see bad packets */
1678 		if (em_debug_sbp)
1679 			reg_rctl |= E1000_RCTL_SBP;
1680 		E1000_WRITE_REG(&sc->hw, E1000_RCTL, reg_rctl);
1681 	} else {
1682 		if (flags & IFF_ALLMULTI) {
1683 			reg_rctl |= E1000_RCTL_MPE;
1684 			reg_rctl &= ~E1000_RCTL_UPE;
1685 			E1000_WRITE_REG(&sc->hw, E1000_RCTL, reg_rctl);
1686 		}
1687 		if (em_if_vlan_filter_used(ctx))
1688 			em_if_vlan_filter_enable(sc);
1689 	}
1690 	return (0);
1691 }
1692 
1693 static u_int
1694 em_copy_maddr(void *arg, struct sockaddr_dl *sdl, u_int idx)
1695 {
1696 	u8 *mta = arg;
1697 
1698 	if (idx == MAX_NUM_MULTICAST_ADDRESSES)
1699 		return (0);
1700 
1701 	bcopy(LLADDR(sdl), &mta[idx * ETHER_ADDR_LEN], ETHER_ADDR_LEN);
1702 
1703 	return (1);
1704 }
1705 
1706 /*********************************************************************
1707  *  Multicast Update
1708  *
1709  *  This routine is called whenever multicast address list is updated.
1710  *
1711  **********************************************************************/
1712 static void
1713 em_if_multi_set(if_ctx_t ctx)
1714 {
1715 	struct e1000_softc *sc = iflib_get_softc(ctx);
1716 	struct ifnet *ifp = iflib_get_ifp(ctx);
1717 	u8  *mta; /* Multicast array memory */
1718 	u32 reg_rctl = 0;
1719 	int mcnt = 0;
1720 
1721 	IOCTL_DEBUGOUT("em_set_multi: begin");
1722 
1723 	mta = sc->mta;
1724 	bzero(mta, sizeof(u8) * ETHER_ADDR_LEN * MAX_NUM_MULTICAST_ADDRESSES);
1725 
1726 	if (sc->hw.mac.type == e1000_82542 &&
1727 	    sc->hw.revision_id == E1000_REVISION_2) {
1728 		reg_rctl = E1000_READ_REG(&sc->hw, E1000_RCTL);
1729 		if (sc->hw.bus.pci_cmd_word & CMD_MEM_WRT_INVALIDATE)
1730 			e1000_pci_clear_mwi(&sc->hw);
1731 		reg_rctl |= E1000_RCTL_RST;
1732 		E1000_WRITE_REG(&sc->hw, E1000_RCTL, reg_rctl);
1733 		msec_delay(5);
1734 	}
1735 
1736 	mcnt = if_foreach_llmaddr(ifp, em_copy_maddr, mta);
1737 
1738 	reg_rctl = E1000_READ_REG(&sc->hw, E1000_RCTL);
1739 
1740 	if (if_getflags(ifp) & IFF_PROMISC)
1741 		reg_rctl |= (E1000_RCTL_UPE | E1000_RCTL_MPE);
1742 	else if (mcnt >= MAX_NUM_MULTICAST_ADDRESSES ||
1743 	    if_getflags(ifp) & IFF_ALLMULTI) {
1744 		reg_rctl |= E1000_RCTL_MPE;
1745 		reg_rctl &= ~E1000_RCTL_UPE;
1746 	} else
1747 		reg_rctl &= ~(E1000_RCTL_UPE | E1000_RCTL_MPE);
1748 
1749 	E1000_WRITE_REG(&sc->hw, E1000_RCTL, reg_rctl);
1750 
1751 	if (mcnt < MAX_NUM_MULTICAST_ADDRESSES)
1752 		e1000_update_mc_addr_list(&sc->hw, mta, mcnt);
1753 
1754 	if (sc->hw.mac.type == e1000_82542 &&
1755 	    sc->hw.revision_id == E1000_REVISION_2) {
1756 		reg_rctl = E1000_READ_REG(&sc->hw, E1000_RCTL);
1757 		reg_rctl &= ~E1000_RCTL_RST;
1758 		E1000_WRITE_REG(&sc->hw, E1000_RCTL, reg_rctl);
1759 		msec_delay(5);
1760 		if (sc->hw.bus.pci_cmd_word & CMD_MEM_WRT_INVALIDATE)
1761 			e1000_pci_set_mwi(&sc->hw);
1762 	}
1763 }
1764 
1765 /*********************************************************************
1766  *  Timer routine
1767  *
1768  *  This routine schedules em_if_update_admin_status() to check for
1769  *  link status and to gather statistics as well as to perform some
1770  *  controller-specific hardware patting.
1771  *
1772  **********************************************************************/
1773 static void
1774 em_if_timer(if_ctx_t ctx, uint16_t qid)
1775 {
1776 
1777 	if (qid != 0)
1778 		return;
1779 
1780 	iflib_admin_intr_deferred(ctx);
1781 }
1782 
1783 static void
1784 em_if_update_admin_status(if_ctx_t ctx)
1785 {
1786 	struct e1000_softc *sc = iflib_get_softc(ctx);
1787 	struct e1000_hw *hw = &sc->hw;
1788 	device_t dev = iflib_get_dev(ctx);
1789 	u32 link_check, thstat, ctrl;
1790 
1791 	link_check = thstat = ctrl = 0;
1792 	/* Get the cached link value or read phy for real */
1793 	switch (hw->phy.media_type) {
1794 	case e1000_media_type_copper:
1795 		if (hw->mac.get_link_status) {
1796 			if (hw->mac.type == e1000_pch_spt)
1797 				msec_delay(50);
1798 			/* Do the work to read phy */
1799 			e1000_check_for_link(hw);
1800 			link_check = !hw->mac.get_link_status;
1801 			if (link_check) /* ESB2 fix */
1802 				e1000_cfg_on_link_up(hw);
1803 		} else {
1804 			link_check = true;
1805 		}
1806 		break;
1807 	case e1000_media_type_fiber:
1808 		e1000_check_for_link(hw);
1809 		link_check = (E1000_READ_REG(hw, E1000_STATUS) &
1810 			    E1000_STATUS_LU);
1811 		break;
1812 	case e1000_media_type_internal_serdes:
1813 		e1000_check_for_link(hw);
1814 		link_check = hw->mac.serdes_has_link;
1815 		break;
1816 	/* VF device is type_unknown */
1817 	case e1000_media_type_unknown:
1818 		e1000_check_for_link(hw);
1819 		link_check = !hw->mac.get_link_status;
1820 		/* FALLTHROUGH */
1821 	default:
1822 		break;
1823 	}
1824 
1825 	/* Check for thermal downshift or shutdown */
1826 	if (hw->mac.type == e1000_i350) {
1827 		thstat = E1000_READ_REG(hw, E1000_THSTAT);
1828 		ctrl = E1000_READ_REG(hw, E1000_CTRL_EXT);
1829 	}
1830 
1831 	/* Now check for a transition */
1832 	if (link_check && (sc->link_active == 0)) {
1833 		e1000_get_speed_and_duplex(hw, &sc->link_speed,
1834 		    &sc->link_duplex);
1835 		/* Check if we must disable SPEED_MODE bit on PCI-E */
1836 		if ((sc->link_speed != SPEED_1000) &&
1837 		    ((hw->mac.type == e1000_82571) ||
1838 		    (hw->mac.type == e1000_82572))) {
1839 			int tarc0;
1840 			tarc0 = E1000_READ_REG(hw, E1000_TARC(0));
1841 			tarc0 &= ~TARC_SPEED_MODE_BIT;
1842 			E1000_WRITE_REG(hw, E1000_TARC(0), tarc0);
1843 		}
1844 		if (bootverbose)
1845 			device_printf(dev, "Link is up %d Mbps %s\n",
1846 			    sc->link_speed,
1847 			    ((sc->link_duplex == FULL_DUPLEX) ?
1848 			    "Full Duplex" : "Half Duplex"));
1849 		sc->link_active = 1;
1850 		sc->smartspeed = 0;
1851 		if ((ctrl & E1000_CTRL_EXT_LINK_MODE_MASK) ==
1852 		    E1000_CTRL_EXT_LINK_MODE_GMII &&
1853 		    (thstat & E1000_THSTAT_LINK_THROTTLE))
1854 			device_printf(dev, "Link: thermal downshift\n");
1855 		/* Delay Link Up for Phy update */
1856 		if (((hw->mac.type == e1000_i210) ||
1857 		    (hw->mac.type == e1000_i211)) &&
1858 		    (hw->phy.id == I210_I_PHY_ID))
1859 			msec_delay(I210_LINK_DELAY);
1860 		/* Reset if the media type changed. */
1861 		if (hw->dev_spec._82575.media_changed &&
1862 		    hw->mac.type >= igb_mac_min) {
1863 			hw->dev_spec._82575.media_changed = false;
1864 			sc->flags |= IGB_MEDIA_RESET;
1865 			em_reset(ctx);
1866 		}
1867 		iflib_link_state_change(ctx, LINK_STATE_UP,
1868 		    IF_Mbps(sc->link_speed));
1869 	} else if (!link_check && (sc->link_active == 1)) {
1870 		sc->link_speed = 0;
1871 		sc->link_duplex = 0;
1872 		sc->link_active = 0;
1873 		iflib_link_state_change(ctx, LINK_STATE_DOWN, 0);
1874 	}
1875 	em_update_stats_counters(sc);
1876 
1877 	/* Reset LAA into RAR[0] on 82571 */
1878 	if (hw->mac.type == e1000_82571 && e1000_get_laa_state_82571(hw))
1879 		e1000_rar_set(hw, hw->mac.addr, 0);
1880 
1881 	if (hw->mac.type < em_mac_min)
1882 		lem_smartspeed(sc);
1883 }
1884 
1885 static void
1886 em_if_watchdog_reset(if_ctx_t ctx)
1887 {
1888 	struct e1000_softc *sc = iflib_get_softc(ctx);
1889 
1890 	/*
1891 	 * Just count the event; iflib(4) will already trigger a
1892 	 * sufficient reset of the controller.
1893 	 */
1894 	sc->watchdog_events++;
1895 }
1896 
1897 /*********************************************************************
1898  *
1899  *  This routine disables all traffic on the adapter by issuing a
1900  *  global reset on the MAC.
1901  *
1902  **********************************************************************/
1903 static void
1904 em_if_stop(if_ctx_t ctx)
1905 {
1906 	struct e1000_softc *sc = iflib_get_softc(ctx);
1907 
1908 	INIT_DEBUGOUT("em_if_stop: begin");
1909 
1910 	e1000_reset_hw(&sc->hw);
1911 	if (sc->hw.mac.type >= e1000_82544)
1912 		E1000_WRITE_REG(&sc->hw, E1000_WUFC, 0);
1913 
1914 	e1000_led_off(&sc->hw);
1915 	e1000_cleanup_led(&sc->hw);
1916 }
1917 
1918 /*********************************************************************
1919  *
1920  *  Determine hardware revision.
1921  *
1922  **********************************************************************/
1923 static void
1924 em_identify_hardware(if_ctx_t ctx)
1925 {
1926 	device_t dev = iflib_get_dev(ctx);
1927 	struct e1000_softc *sc = iflib_get_softc(ctx);
1928 
1929 	/* Make sure our PCI config space has the necessary stuff set */
1930 	sc->hw.bus.pci_cmd_word = pci_read_config(dev, PCIR_COMMAND, 2);
1931 
1932 	/* Save off the information about this board */
1933 	sc->hw.vendor_id = pci_get_vendor(dev);
1934 	sc->hw.device_id = pci_get_device(dev);
1935 	sc->hw.revision_id = pci_read_config(dev, PCIR_REVID, 1);
1936 	sc->hw.subsystem_vendor_id =
1937 	    pci_read_config(dev, PCIR_SUBVEND_0, 2);
1938 	sc->hw.subsystem_device_id =
1939 	    pci_read_config(dev, PCIR_SUBDEV_0, 2);
1940 
1941 	/* Do Shared Code Init and Setup */
1942 	if (e1000_set_mac_type(&sc->hw)) {
1943 		device_printf(dev, "Setup init failure\n");
1944 		return;
1945 	}
1946 
1947 	/* Are we a VF device? */
1948 	if ((sc->hw.mac.type == e1000_vfadapt) ||
1949 	    (sc->hw.mac.type == e1000_vfadapt_i350))
1950 		sc->vf_ifp = 1;
1951 	else
1952 		sc->vf_ifp = 0;
1953 }
1954 
1955 static int
1956 em_allocate_pci_resources(if_ctx_t ctx)
1957 {
1958 	struct e1000_softc *sc = iflib_get_softc(ctx);
1959 	device_t dev = iflib_get_dev(ctx);
1960 	int rid, val;
1961 
1962 	rid = PCIR_BAR(0);
1963 	sc->memory = bus_alloc_resource_any(dev, SYS_RES_MEMORY,
1964 	    &rid, RF_ACTIVE);
1965 	if (sc->memory == NULL) {
1966 		device_printf(dev, "Unable to allocate bus resource: memory\n");
1967 		return (ENXIO);
1968 	}
1969 	sc->osdep.mem_bus_space_tag = rman_get_bustag(sc->memory);
1970 	sc->osdep.mem_bus_space_handle =
1971 	    rman_get_bushandle(sc->memory);
1972 	sc->hw.hw_addr = (u8 *)&sc->osdep.mem_bus_space_handle;
1973 
1974 	/* Only older adapters use IO mapping */
1975 	if (sc->hw.mac.type < em_mac_min &&
1976 	    sc->hw.mac.type > e1000_82543) {
1977 		/* Figure our where our IO BAR is ? */
1978 		for (rid = PCIR_BAR(0); rid < PCIR_CIS;) {
1979 			val = pci_read_config(dev, rid, 4);
1980 			if (EM_BAR_TYPE(val) == EM_BAR_TYPE_IO) {
1981 				break;
1982 			}
1983 			rid += 4;
1984 			/* check for 64bit BAR */
1985 			if (EM_BAR_MEM_TYPE(val) == EM_BAR_MEM_TYPE_64BIT)
1986 				rid += 4;
1987 		}
1988 		if (rid >= PCIR_CIS) {
1989 			device_printf(dev, "Unable to locate IO BAR\n");
1990 			return (ENXIO);
1991 		}
1992 		sc->ioport = bus_alloc_resource_any(dev, SYS_RES_IOPORT,
1993 		    &rid, RF_ACTIVE);
1994 		if (sc->ioport == NULL) {
1995 			device_printf(dev, "Unable to allocate bus resource: "
1996 			    "ioport\n");
1997 			return (ENXIO);
1998 		}
1999 		sc->hw.io_base = 0;
2000 		sc->osdep.io_bus_space_tag =
2001 		    rman_get_bustag(sc->ioport);
2002 		sc->osdep.io_bus_space_handle =
2003 		    rman_get_bushandle(sc->ioport);
2004 	}
2005 
2006 	sc->hw.back = &sc->osdep;
2007 
2008 	return (0);
2009 }
2010 
2011 /*********************************************************************
2012  *
2013  *  Set up the MSI-X Interrupt handlers
2014  *
2015  **********************************************************************/
2016 static int
2017 em_if_msix_intr_assign(if_ctx_t ctx, int msix)
2018 {
2019 	struct e1000_softc *sc = iflib_get_softc(ctx);
2020 	struct em_rx_queue *rx_que = sc->rx_queues;
2021 	struct em_tx_queue *tx_que = sc->tx_queues;
2022 	int error, rid, i, vector = 0, rx_vectors;
2023 	char buf[16];
2024 
2025 	/* First set up ring resources */
2026 	for (i = 0; i < sc->rx_num_queues; i++, rx_que++, vector++) {
2027 		rid = vector + 1;
2028 		snprintf(buf, sizeof(buf), "rxq%d", i);
2029 		error = iflib_irq_alloc_generic(ctx, &rx_que->que_irq, rid, IFLIB_INTR_RXTX, em_msix_que, rx_que, rx_que->me, buf);
2030 		if (error) {
2031 			device_printf(iflib_get_dev(ctx), "Failed to allocate que int %d err: %d", i, error);
2032 			sc->rx_num_queues = i + 1;
2033 			goto fail;
2034 		}
2035 
2036 		rx_que->msix =  vector;
2037 
2038 		/*
2039 		 * Set the bit to enable interrupt
2040 		 * in E1000_IMS -- bits 20 and 21
2041 		 * are for RX0 and RX1, note this has
2042 		 * NOTHING to do with the MSI-X vector
2043 		 */
2044 		if (sc->hw.mac.type == e1000_82574) {
2045 			rx_que->eims = 1 << (20 + i);
2046 			sc->ims |= rx_que->eims;
2047 			sc->ivars |= (8 | rx_que->msix) << (i * 4);
2048 		} else if (sc->hw.mac.type == e1000_82575)
2049 			rx_que->eims = E1000_EICR_TX_QUEUE0 << vector;
2050 		else
2051 			rx_que->eims = 1 << vector;
2052 	}
2053 	rx_vectors = vector;
2054 
2055 	vector = 0;
2056 	for (i = 0; i < sc->tx_num_queues; i++, tx_que++, vector++) {
2057 		snprintf(buf, sizeof(buf), "txq%d", i);
2058 		tx_que = &sc->tx_queues[i];
2059 		iflib_softirq_alloc_generic(ctx,
2060 		    &sc->rx_queues[i % sc->rx_num_queues].que_irq,
2061 		    IFLIB_INTR_TX, tx_que, tx_que->me, buf);
2062 
2063 		tx_que->msix = (vector % sc->rx_num_queues);
2064 
2065 		/*
2066 		 * Set the bit to enable interrupt
2067 		 * in E1000_IMS -- bits 22 and 23
2068 		 * are for TX0 and TX1, note this has
2069 		 * NOTHING to do with the MSI-X vector
2070 		 */
2071 		if (sc->hw.mac.type == e1000_82574) {
2072 			tx_que->eims = 1 << (22 + i);
2073 			sc->ims |= tx_que->eims;
2074 			sc->ivars |= (8 | tx_que->msix) << (8 + (i * 4));
2075 		} else if (sc->hw.mac.type == e1000_82575) {
2076 			tx_que->eims = E1000_EICR_TX_QUEUE0 << i;
2077 		} else {
2078 			tx_que->eims = 1 << i;
2079 		}
2080 	}
2081 
2082 	/* Link interrupt */
2083 	rid = rx_vectors + 1;
2084 	error = iflib_irq_alloc_generic(ctx, &sc->irq, rid, IFLIB_INTR_ADMIN, em_msix_link, sc, 0, "aq");
2085 
2086 	if (error) {
2087 		device_printf(iflib_get_dev(ctx), "Failed to register admin handler");
2088 		goto fail;
2089 	}
2090 	sc->linkvec = rx_vectors;
2091 	if (sc->hw.mac.type < igb_mac_min) {
2092 		sc->ivars |=  (8 | rx_vectors) << 16;
2093 		sc->ivars |= 0x80000000;
2094 		/* Enable the "Other" interrupt type for link status change */
2095 		sc->ims |= E1000_IMS_OTHER;
2096 	}
2097 
2098 	return (0);
2099 fail:
2100 	iflib_irq_free(ctx, &sc->irq);
2101 	rx_que = sc->rx_queues;
2102 	for (int i = 0; i < sc->rx_num_queues; i++, rx_que++)
2103 		iflib_irq_free(ctx, &rx_que->que_irq);
2104 	return (error);
2105 }
2106 
2107 static void
2108 igb_configure_queues(struct e1000_softc *sc)
2109 {
2110 	struct e1000_hw *hw = &sc->hw;
2111 	struct em_rx_queue *rx_que;
2112 	struct em_tx_queue *tx_que;
2113 	u32 tmp, ivar = 0, newitr = 0;
2114 
2115 	/* First turn on RSS capability */
2116 	if (hw->mac.type != e1000_82575)
2117 		E1000_WRITE_REG(hw, E1000_GPIE,
2118 		    E1000_GPIE_MSIX_MODE | E1000_GPIE_EIAME |
2119 		    E1000_GPIE_PBA | E1000_GPIE_NSICR);
2120 
2121 	/* Turn on MSI-X */
2122 	switch (hw->mac.type) {
2123 	case e1000_82580:
2124 	case e1000_i350:
2125 	case e1000_i354:
2126 	case e1000_i210:
2127 	case e1000_i211:
2128 	case e1000_vfadapt:
2129 	case e1000_vfadapt_i350:
2130 		/* RX entries */
2131 		for (int i = 0; i < sc->rx_num_queues; i++) {
2132 			u32 index = i >> 1;
2133 			ivar = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index);
2134 			rx_que = &sc->rx_queues[i];
2135 			if (i & 1) {
2136 				ivar &= 0xFF00FFFF;
2137 				ivar |= (rx_que->msix | E1000_IVAR_VALID) << 16;
2138 			} else {
2139 				ivar &= 0xFFFFFF00;
2140 				ivar |= rx_que->msix | E1000_IVAR_VALID;
2141 			}
2142 			E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, ivar);
2143 		}
2144 		/* TX entries */
2145 		for (int i = 0; i < sc->tx_num_queues; i++) {
2146 			u32 index = i >> 1;
2147 			ivar = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index);
2148 			tx_que = &sc->tx_queues[i];
2149 			if (i & 1) {
2150 				ivar &= 0x00FFFFFF;
2151 				ivar |= (tx_que->msix | E1000_IVAR_VALID) << 24;
2152 			} else {
2153 				ivar &= 0xFFFF00FF;
2154 				ivar |= (tx_que->msix | E1000_IVAR_VALID) << 8;
2155 			}
2156 			E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, ivar);
2157 			sc->que_mask |= tx_que->eims;
2158 		}
2159 
2160 		/* And for the link interrupt */
2161 		ivar = (sc->linkvec | E1000_IVAR_VALID) << 8;
2162 		sc->link_mask = 1 << sc->linkvec;
2163 		E1000_WRITE_REG(hw, E1000_IVAR_MISC, ivar);
2164 		break;
2165 	case e1000_82576:
2166 		/* RX entries */
2167 		for (int i = 0; i < sc->rx_num_queues; i++) {
2168 			u32 index = i & 0x7; /* Each IVAR has two entries */
2169 			ivar = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index);
2170 			rx_que = &sc->rx_queues[i];
2171 			if (i < 8) {
2172 				ivar &= 0xFFFFFF00;
2173 				ivar |= rx_que->msix | E1000_IVAR_VALID;
2174 			} else {
2175 				ivar &= 0xFF00FFFF;
2176 				ivar |= (rx_que->msix | E1000_IVAR_VALID) << 16;
2177 			}
2178 			E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, ivar);
2179 			sc->que_mask |= rx_que->eims;
2180 		}
2181 		/* TX entries */
2182 		for (int i = 0; i < sc->tx_num_queues; i++) {
2183 			u32 index = i & 0x7; /* Each IVAR has two entries */
2184 			ivar = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index);
2185 			tx_que = &sc->tx_queues[i];
2186 			if (i < 8) {
2187 				ivar &= 0xFFFF00FF;
2188 				ivar |= (tx_que->msix | E1000_IVAR_VALID) << 8;
2189 			} else {
2190 				ivar &= 0x00FFFFFF;
2191 				ivar |= (tx_que->msix | E1000_IVAR_VALID) << 24;
2192 			}
2193 			E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, ivar);
2194 			sc->que_mask |= tx_que->eims;
2195 		}
2196 
2197 		/* And for the link interrupt */
2198 		ivar = (sc->linkvec | E1000_IVAR_VALID) << 8;
2199 		sc->link_mask = 1 << sc->linkvec;
2200 		E1000_WRITE_REG(hw, E1000_IVAR_MISC, ivar);
2201 		break;
2202 
2203 	case e1000_82575:
2204 		/* enable MSI-X support*/
2205 		tmp = E1000_READ_REG(hw, E1000_CTRL_EXT);
2206 		tmp |= E1000_CTRL_EXT_PBA_CLR;
2207 		/* Auto-Mask interrupts upon ICR read. */
2208 		tmp |= E1000_CTRL_EXT_EIAME;
2209 		tmp |= E1000_CTRL_EXT_IRCA;
2210 		E1000_WRITE_REG(hw, E1000_CTRL_EXT, tmp);
2211 
2212 		/* Queues */
2213 		for (int i = 0; i < sc->rx_num_queues; i++) {
2214 			rx_que = &sc->rx_queues[i];
2215 			tmp = E1000_EICR_RX_QUEUE0 << i;
2216 			tmp |= E1000_EICR_TX_QUEUE0 << i;
2217 			rx_que->eims = tmp;
2218 			E1000_WRITE_REG_ARRAY(hw, E1000_MSIXBM(0),
2219 			    i, rx_que->eims);
2220 			sc->que_mask |= rx_que->eims;
2221 		}
2222 
2223 		/* Link */
2224 		E1000_WRITE_REG(hw, E1000_MSIXBM(sc->linkvec),
2225 		    E1000_EIMS_OTHER);
2226 		sc->link_mask |= E1000_EIMS_OTHER;
2227 	default:
2228 		break;
2229 	}
2230 
2231 	/* Set the starting interrupt rate */
2232 	if (em_max_interrupt_rate > 0)
2233 		newitr = (4000000 / em_max_interrupt_rate) & 0x7FFC;
2234 
2235 	if (hw->mac.type == e1000_82575)
2236 		newitr |= newitr << 16;
2237 	else
2238 		newitr |= E1000_EITR_CNT_IGNR;
2239 
2240 	for (int i = 0; i < sc->rx_num_queues; i++) {
2241 		rx_que = &sc->rx_queues[i];
2242 		E1000_WRITE_REG(hw, E1000_EITR(rx_que->msix), newitr);
2243 	}
2244 
2245 	return;
2246 }
2247 
2248 static void
2249 em_free_pci_resources(if_ctx_t ctx)
2250 {
2251 	struct e1000_softc *sc = iflib_get_softc(ctx);
2252 	struct em_rx_queue *que = sc->rx_queues;
2253 	device_t dev = iflib_get_dev(ctx);
2254 
2255 	/* Release all MSI-X queue resources */
2256 	if (sc->intr_type == IFLIB_INTR_MSIX)
2257 		iflib_irq_free(ctx, &sc->irq);
2258 
2259 	if (que != NULL) {
2260 		for (int i = 0; i < sc->rx_num_queues; i++, que++) {
2261 			iflib_irq_free(ctx, &que->que_irq);
2262 		}
2263 	}
2264 
2265 	if (sc->memory != NULL) {
2266 		bus_release_resource(dev, SYS_RES_MEMORY,
2267 		    rman_get_rid(sc->memory), sc->memory);
2268 		sc->memory = NULL;
2269 	}
2270 
2271 	if (sc->flash != NULL) {
2272 		bus_release_resource(dev, SYS_RES_MEMORY,
2273 		    rman_get_rid(sc->flash), sc->flash);
2274 		sc->flash = NULL;
2275 	}
2276 
2277 	if (sc->ioport != NULL) {
2278 		bus_release_resource(dev, SYS_RES_IOPORT,
2279 		    rman_get_rid(sc->ioport), sc->ioport);
2280 		sc->ioport = NULL;
2281 	}
2282 }
2283 
2284 /* Set up MSI or MSI-X */
2285 static int
2286 em_setup_msix(if_ctx_t ctx)
2287 {
2288 	struct e1000_softc *sc = iflib_get_softc(ctx);
2289 
2290 	if (sc->hw.mac.type == e1000_82574) {
2291 		em_enable_vectors_82574(ctx);
2292 	}
2293 	return (0);
2294 }
2295 
2296 /*********************************************************************
2297  *
2298  *  Workaround for SmartSpeed on 82541 and 82547 controllers
2299  *
2300  **********************************************************************/
2301 static void
2302 lem_smartspeed(struct e1000_softc *sc)
2303 {
2304 	u16 phy_tmp;
2305 
2306 	if (sc->link_active || (sc->hw.phy.type != e1000_phy_igp) ||
2307 	    sc->hw.mac.autoneg == 0 ||
2308 	    (sc->hw.phy.autoneg_advertised & ADVERTISE_1000_FULL) == 0)
2309 		return;
2310 
2311 	if (sc->smartspeed == 0) {
2312 		/* If Master/Slave config fault is asserted twice,
2313 		 * we assume back-to-back */
2314 		e1000_read_phy_reg(&sc->hw, PHY_1000T_STATUS, &phy_tmp);
2315 		if (!(phy_tmp & SR_1000T_MS_CONFIG_FAULT))
2316 			return;
2317 		e1000_read_phy_reg(&sc->hw, PHY_1000T_STATUS, &phy_tmp);
2318 		if (phy_tmp & SR_1000T_MS_CONFIG_FAULT) {
2319 			e1000_read_phy_reg(&sc->hw,
2320 			    PHY_1000T_CTRL, &phy_tmp);
2321 			if(phy_tmp & CR_1000T_MS_ENABLE) {
2322 				phy_tmp &= ~CR_1000T_MS_ENABLE;
2323 				e1000_write_phy_reg(&sc->hw,
2324 				    PHY_1000T_CTRL, phy_tmp);
2325 				sc->smartspeed++;
2326 				if(sc->hw.mac.autoneg &&
2327 				   !e1000_copper_link_autoneg(&sc->hw) &&
2328 				   !e1000_read_phy_reg(&sc->hw,
2329 				    PHY_CONTROL, &phy_tmp)) {
2330 					phy_tmp |= (MII_CR_AUTO_NEG_EN |
2331 						    MII_CR_RESTART_AUTO_NEG);
2332 					e1000_write_phy_reg(&sc->hw,
2333 					    PHY_CONTROL, phy_tmp);
2334 				}
2335 			}
2336 		}
2337 		return;
2338 	} else if(sc->smartspeed == EM_SMARTSPEED_DOWNSHIFT) {
2339 		/* If still no link, perhaps using 2/3 pair cable */
2340 		e1000_read_phy_reg(&sc->hw, PHY_1000T_CTRL, &phy_tmp);
2341 		phy_tmp |= CR_1000T_MS_ENABLE;
2342 		e1000_write_phy_reg(&sc->hw, PHY_1000T_CTRL, phy_tmp);
2343 		if(sc->hw.mac.autoneg &&
2344 		   !e1000_copper_link_autoneg(&sc->hw) &&
2345 		   !e1000_read_phy_reg(&sc->hw, PHY_CONTROL, &phy_tmp)) {
2346 			phy_tmp |= (MII_CR_AUTO_NEG_EN |
2347 				    MII_CR_RESTART_AUTO_NEG);
2348 			e1000_write_phy_reg(&sc->hw, PHY_CONTROL, phy_tmp);
2349 		}
2350 	}
2351 	/* Restart process after EM_SMARTSPEED_MAX iterations */
2352 	if(sc->smartspeed++ == EM_SMARTSPEED_MAX)
2353 		sc->smartspeed = 0;
2354 }
2355 
2356 /*********************************************************************
2357  *
2358  *  Initialize the DMA Coalescing feature
2359  *
2360  **********************************************************************/
2361 static void
2362 igb_init_dmac(struct e1000_softc *sc, u32 pba)
2363 {
2364 	device_t	dev = sc->dev;
2365 	struct e1000_hw *hw = &sc->hw;
2366 	u32 		dmac, reg = ~E1000_DMACR_DMAC_EN;
2367 	u16		hwm;
2368 	u16		max_frame_size;
2369 
2370 	if (hw->mac.type == e1000_i211)
2371 		return;
2372 
2373 	max_frame_size = sc->shared->isc_max_frame_size;
2374 	if (hw->mac.type > e1000_82580) {
2375 
2376 		if (sc->dmac == 0) { /* Disabling it */
2377 			E1000_WRITE_REG(hw, E1000_DMACR, reg);
2378 			return;
2379 		} else
2380 			device_printf(dev, "DMA Coalescing enabled\n");
2381 
2382 		/* Set starting threshold */
2383 		E1000_WRITE_REG(hw, E1000_DMCTXTH, 0);
2384 
2385 		hwm = 64 * pba - max_frame_size / 16;
2386 		if (hwm < 64 * (pba - 6))
2387 			hwm = 64 * (pba - 6);
2388 		reg = E1000_READ_REG(hw, E1000_FCRTC);
2389 		reg &= ~E1000_FCRTC_RTH_COAL_MASK;
2390 		reg |= ((hwm << E1000_FCRTC_RTH_COAL_SHIFT)
2391 		    & E1000_FCRTC_RTH_COAL_MASK);
2392 		E1000_WRITE_REG(hw, E1000_FCRTC, reg);
2393 
2394 
2395 		dmac = pba - max_frame_size / 512;
2396 		if (dmac < pba - 10)
2397 			dmac = pba - 10;
2398 		reg = E1000_READ_REG(hw, E1000_DMACR);
2399 		reg &= ~E1000_DMACR_DMACTHR_MASK;
2400 		reg |= ((dmac << E1000_DMACR_DMACTHR_SHIFT)
2401 		    & E1000_DMACR_DMACTHR_MASK);
2402 
2403 		/* transition to L0x or L1 if available..*/
2404 		reg |= (E1000_DMACR_DMAC_EN | E1000_DMACR_DMAC_LX_MASK);
2405 
2406 		/* Check if status is 2.5Gb backplane connection
2407 		* before configuration of watchdog timer, which is
2408 		* in msec values in 12.8usec intervals
2409 		* watchdog timer= msec values in 32usec intervals
2410 		* for non 2.5Gb connection
2411 		*/
2412 		if (hw->mac.type == e1000_i354) {
2413 			int status = E1000_READ_REG(hw, E1000_STATUS);
2414 			if ((status & E1000_STATUS_2P5_SKU) &&
2415 			    (!(status & E1000_STATUS_2P5_SKU_OVER)))
2416 				reg |= ((sc->dmac * 5) >> 6);
2417 			else
2418 				reg |= (sc->dmac >> 5);
2419 		} else {
2420 			reg |= (sc->dmac >> 5);
2421 		}
2422 
2423 		E1000_WRITE_REG(hw, E1000_DMACR, reg);
2424 
2425 		E1000_WRITE_REG(hw, E1000_DMCRTRH, 0);
2426 
2427 		/* Set the interval before transition */
2428 		reg = E1000_READ_REG(hw, E1000_DMCTLX);
2429 		if (hw->mac.type == e1000_i350)
2430 			reg |= IGB_DMCTLX_DCFLUSH_DIS;
2431 		/*
2432 		** in 2.5Gb connection, TTLX unit is 0.4 usec
2433 		** which is 0x4*2 = 0xA. But delay is still 4 usec
2434 		*/
2435 		if (hw->mac.type == e1000_i354) {
2436 			int status = E1000_READ_REG(hw, E1000_STATUS);
2437 			if ((status & E1000_STATUS_2P5_SKU) &&
2438 			    (!(status & E1000_STATUS_2P5_SKU_OVER)))
2439 				reg |= 0xA;
2440 			else
2441 				reg |= 0x4;
2442 		} else {
2443 			reg |= 0x4;
2444 		}
2445 
2446 		E1000_WRITE_REG(hw, E1000_DMCTLX, reg);
2447 
2448 		/* free space in tx packet buffer to wake from DMA coal */
2449 		E1000_WRITE_REG(hw, E1000_DMCTXTH, (IGB_TXPBSIZE -
2450 		    (2 * max_frame_size)) >> 6);
2451 
2452 		/* make low power state decision controlled by DMA coal */
2453 		reg = E1000_READ_REG(hw, E1000_PCIEMISC);
2454 		reg &= ~E1000_PCIEMISC_LX_DECISION;
2455 		E1000_WRITE_REG(hw, E1000_PCIEMISC, reg);
2456 
2457 	} else if (hw->mac.type == e1000_82580) {
2458 		u32 reg = E1000_READ_REG(hw, E1000_PCIEMISC);
2459 		E1000_WRITE_REG(hw, E1000_PCIEMISC,
2460 		    reg & ~E1000_PCIEMISC_LX_DECISION);
2461 		E1000_WRITE_REG(hw, E1000_DMACR, 0);
2462 	}
2463 }
2464 
2465 /*********************************************************************
2466  *
2467  *  Initialize the hardware to a configuration as specified by the
2468  *  sc structure.
2469  *
2470  **********************************************************************/
2471 static void
2472 em_reset(if_ctx_t ctx)
2473 {
2474 	device_t dev = iflib_get_dev(ctx);
2475 	struct e1000_softc *sc = iflib_get_softc(ctx);
2476 	struct ifnet *ifp = iflib_get_ifp(ctx);
2477 	struct e1000_hw *hw = &sc->hw;
2478 	u16 rx_buffer_size;
2479 	u32 pba;
2480 
2481 	INIT_DEBUGOUT("em_reset: begin");
2482 	/* Let the firmware know the OS is in control */
2483 	em_get_hw_control(sc);
2484 
2485 	/* Set up smart power down as default off on newer adapters. */
2486 	if (!em_smart_pwr_down && (hw->mac.type == e1000_82571 ||
2487 	    hw->mac.type == e1000_82572)) {
2488 		u16 phy_tmp = 0;
2489 
2490 		/* Speed up time to link by disabling smart power down. */
2491 		e1000_read_phy_reg(hw, IGP02E1000_PHY_POWER_MGMT, &phy_tmp);
2492 		phy_tmp &= ~IGP02E1000_PM_SPD;
2493 		e1000_write_phy_reg(hw, IGP02E1000_PHY_POWER_MGMT, phy_tmp);
2494 	}
2495 
2496 	/*
2497 	 * Packet Buffer Allocation (PBA)
2498 	 * Writing PBA sets the receive portion of the buffer
2499 	 * the remainder is used for the transmit buffer.
2500 	 */
2501 	switch (hw->mac.type) {
2502 	/* 82547: Total Packet Buffer is 40K */
2503 	case e1000_82547:
2504 	case e1000_82547_rev_2:
2505 		if (hw->mac.max_frame_size > 8192)
2506 			pba = E1000_PBA_22K; /* 22K for Rx, 18K for Tx */
2507 		else
2508 			pba = E1000_PBA_30K; /* 30K for Rx, 10K for Tx */
2509 		break;
2510 	/* 82571/82572/80003es2lan: Total Packet Buffer is 48K */
2511 	case e1000_82571:
2512 	case e1000_82572:
2513 	case e1000_80003es2lan:
2514 			pba = E1000_PBA_32K; /* 32K for Rx, 16K for Tx */
2515 		break;
2516 	/* 82573: Total Packet Buffer is 32K */
2517 	case e1000_82573:
2518 			pba = E1000_PBA_12K; /* 12K for Rx, 20K for Tx */
2519 		break;
2520 	case e1000_82574:
2521 	case e1000_82583:
2522 			pba = E1000_PBA_20K; /* 20K for Rx, 20K for Tx */
2523 		break;
2524 	case e1000_ich8lan:
2525 		pba = E1000_PBA_8K;
2526 		break;
2527 	case e1000_ich9lan:
2528 	case e1000_ich10lan:
2529 		/* Boost Receive side for jumbo frames */
2530 		if (hw->mac.max_frame_size > 4096)
2531 			pba = E1000_PBA_14K;
2532 		else
2533 			pba = E1000_PBA_10K;
2534 		break;
2535 	case e1000_pchlan:
2536 	case e1000_pch2lan:
2537 	case e1000_pch_lpt:
2538 	case e1000_pch_spt:
2539 	case e1000_pch_cnp:
2540 	case e1000_pch_tgp:
2541 	case e1000_pch_adp:
2542 	case e1000_pch_mtp:
2543 		pba = E1000_PBA_26K;
2544 		break;
2545 	case e1000_82575:
2546 		pba = E1000_PBA_32K;
2547 		break;
2548 	case e1000_82576:
2549 	case e1000_vfadapt:
2550 		pba = E1000_READ_REG(hw, E1000_RXPBS);
2551 		pba &= E1000_RXPBS_SIZE_MASK_82576;
2552 		break;
2553 	case e1000_82580:
2554 	case e1000_i350:
2555 	case e1000_i354:
2556 	case e1000_vfadapt_i350:
2557 		pba = E1000_READ_REG(hw, E1000_RXPBS);
2558 		pba = e1000_rxpbs_adjust_82580(pba);
2559 		break;
2560 	case e1000_i210:
2561 	case e1000_i211:
2562 		pba = E1000_PBA_34K;
2563 		break;
2564 	default:
2565 		/* Remaining devices assumed to have a Packet Buffer of 64K. */
2566 		if (hw->mac.max_frame_size > 8192)
2567 			pba = E1000_PBA_40K; /* 40K for Rx, 24K for Tx */
2568 		else
2569 			pba = E1000_PBA_48K; /* 48K for Rx, 16K for Tx */
2570 	}
2571 
2572 	/* Special needs in case of Jumbo frames */
2573 	if ((hw->mac.type == e1000_82575) && (ifp->if_mtu > ETHERMTU)) {
2574 		u32 tx_space, min_tx, min_rx;
2575 		pba = E1000_READ_REG(hw, E1000_PBA);
2576 		tx_space = pba >> 16;
2577 		pba &= 0xffff;
2578 		min_tx = (hw->mac.max_frame_size +
2579 		    sizeof(struct e1000_tx_desc) - ETHERNET_FCS_SIZE) * 2;
2580 		min_tx = roundup2(min_tx, 1024);
2581 		min_tx >>= 10;
2582 		min_rx = hw->mac.max_frame_size;
2583 		min_rx = roundup2(min_rx, 1024);
2584 		min_rx >>= 10;
2585 		if (tx_space < min_tx &&
2586 		    ((min_tx - tx_space) < pba)) {
2587 			pba = pba - (min_tx - tx_space);
2588 			/*
2589 			 * if short on rx space, rx wins
2590 			 * and must trump tx adjustment
2591 			 */
2592 			if (pba < min_rx)
2593 				pba = min_rx;
2594 		}
2595 		E1000_WRITE_REG(hw, E1000_PBA, pba);
2596 	}
2597 
2598 	if (hw->mac.type < igb_mac_min)
2599 		E1000_WRITE_REG(hw, E1000_PBA, pba);
2600 
2601 	INIT_DEBUGOUT1("em_reset: pba=%dK",pba);
2602 
2603 	/*
2604 	 * These parameters control the automatic generation (Tx) and
2605 	 * response (Rx) to Ethernet PAUSE frames.
2606 	 * - High water mark should allow for at least two frames to be
2607 	 *   received after sending an XOFF.
2608 	 * - Low water mark works best when it is very near the high water mark.
2609 	 *   This allows the receiver to restart by sending XON when it has
2610 	 *   drained a bit. Here we use an arbitrary value of 1500 which will
2611 	 *   restart after one full frame is pulled from the buffer. There
2612 	 *   could be several smaller frames in the buffer and if so they will
2613 	 *   not trigger the XON until their total number reduces the buffer
2614 	 *   by 1500.
2615 	 * - The pause time is fairly large at 1000 x 512ns = 512 usec.
2616 	 */
2617 	rx_buffer_size = (pba & 0xffff) << 10;
2618 	hw->fc.high_water = rx_buffer_size -
2619 	    roundup2(hw->mac.max_frame_size, 1024);
2620 	hw->fc.low_water = hw->fc.high_water - 1500;
2621 
2622 	if (sc->fc) /* locally set flow control value? */
2623 		hw->fc.requested_mode = sc->fc;
2624 	else
2625 		hw->fc.requested_mode = e1000_fc_full;
2626 
2627 	if (hw->mac.type == e1000_80003es2lan)
2628 		hw->fc.pause_time = 0xFFFF;
2629 	else
2630 		hw->fc.pause_time = EM_FC_PAUSE_TIME;
2631 
2632 	hw->fc.send_xon = true;
2633 
2634 	/* Device specific overrides/settings */
2635 	switch (hw->mac.type) {
2636 	case e1000_pchlan:
2637 		/* Workaround: no TX flow ctrl for PCH */
2638 		hw->fc.requested_mode = e1000_fc_rx_pause;
2639 		hw->fc.pause_time = 0xFFFF; /* override */
2640 		if (if_getmtu(ifp) > ETHERMTU) {
2641 			hw->fc.high_water = 0x3500;
2642 			hw->fc.low_water = 0x1500;
2643 		} else {
2644 			hw->fc.high_water = 0x5000;
2645 			hw->fc.low_water = 0x3000;
2646 		}
2647 		hw->fc.refresh_time = 0x1000;
2648 		break;
2649 	case e1000_pch2lan:
2650 	case e1000_pch_lpt:
2651 	case e1000_pch_spt:
2652 	case e1000_pch_cnp:
2653 	case e1000_pch_tgp:
2654 	case e1000_pch_adp:
2655 	case e1000_pch_mtp:
2656 		hw->fc.high_water = 0x5C20;
2657 		hw->fc.low_water = 0x5048;
2658 		hw->fc.pause_time = 0x0650;
2659 		hw->fc.refresh_time = 0x0400;
2660 		/* Jumbos need adjusted PBA */
2661 		if (if_getmtu(ifp) > ETHERMTU)
2662 			E1000_WRITE_REG(hw, E1000_PBA, 12);
2663 		else
2664 			E1000_WRITE_REG(hw, E1000_PBA, 26);
2665 		break;
2666 	case e1000_82575:
2667 	case e1000_82576:
2668 		/* 8-byte granularity */
2669 		hw->fc.low_water = hw->fc.high_water - 8;
2670 		break;
2671 	case e1000_82580:
2672 	case e1000_i350:
2673 	case e1000_i354:
2674 	case e1000_i210:
2675 	case e1000_i211:
2676 	case e1000_vfadapt:
2677 	case e1000_vfadapt_i350:
2678 		/* 16-byte granularity */
2679 		hw->fc.low_water = hw->fc.high_water - 16;
2680 		break;
2681 	case e1000_ich9lan:
2682 	case e1000_ich10lan:
2683 		if (if_getmtu(ifp) > ETHERMTU) {
2684 			hw->fc.high_water = 0x2800;
2685 			hw->fc.low_water = hw->fc.high_water - 8;
2686 			break;
2687 		}
2688 		/* FALLTHROUGH */
2689 	default:
2690 		if (hw->mac.type == e1000_80003es2lan)
2691 			hw->fc.pause_time = 0xFFFF;
2692 		break;
2693 	}
2694 
2695 	/* Issue a global reset */
2696 	e1000_reset_hw(hw);
2697 	if (hw->mac.type >= igb_mac_min) {
2698 		E1000_WRITE_REG(hw, E1000_WUC, 0);
2699 	} else {
2700 		E1000_WRITE_REG(hw, E1000_WUFC, 0);
2701 		em_disable_aspm(sc);
2702 	}
2703 	if (sc->flags & IGB_MEDIA_RESET) {
2704 		e1000_setup_init_funcs(hw, true);
2705 		e1000_get_bus_info(hw);
2706 		sc->flags &= ~IGB_MEDIA_RESET;
2707 	}
2708 	/* and a re-init */
2709 	if (e1000_init_hw(hw) < 0) {
2710 		device_printf(dev, "Hardware Initialization Failed\n");
2711 		return;
2712 	}
2713 	if (hw->mac.type >= igb_mac_min)
2714 		igb_init_dmac(sc, pba);
2715 
2716 	E1000_WRITE_REG(hw, E1000_VET, ETHERTYPE_VLAN);
2717 	e1000_get_phy_info(hw);
2718 	e1000_check_for_link(hw);
2719 }
2720 
2721 /*
2722  * Initialise the RSS mapping for NICs that support multiple transmit/
2723  * receive rings.
2724  */
2725 
2726 #define RSSKEYLEN 10
2727 static void
2728 em_initialize_rss_mapping(struct e1000_softc *sc)
2729 {
2730 	uint8_t  rss_key[4 * RSSKEYLEN];
2731 	uint32_t reta = 0;
2732 	struct e1000_hw	*hw = &sc->hw;
2733 	int i;
2734 
2735 	/*
2736 	 * Configure RSS key
2737 	 */
2738 	arc4rand(rss_key, sizeof(rss_key), 0);
2739 	for (i = 0; i < RSSKEYLEN; ++i) {
2740 		uint32_t rssrk = 0;
2741 
2742 		rssrk = EM_RSSRK_VAL(rss_key, i);
2743 		E1000_WRITE_REG(hw,E1000_RSSRK(i), rssrk);
2744 	}
2745 
2746 	/*
2747 	 * Configure RSS redirect table in following fashion:
2748 	 * (hash & ring_cnt_mask) == rdr_table[(hash & rdr_table_mask)]
2749 	 */
2750 	for (i = 0; i < sizeof(reta); ++i) {
2751 		uint32_t q;
2752 
2753 		q = (i % sc->rx_num_queues) << 7;
2754 		reta |= q << (8 * i);
2755 	}
2756 
2757 	for (i = 0; i < 32; ++i)
2758 		E1000_WRITE_REG(hw, E1000_RETA(i), reta);
2759 
2760 	E1000_WRITE_REG(hw, E1000_MRQC, E1000_MRQC_RSS_ENABLE_2Q |
2761 			E1000_MRQC_RSS_FIELD_IPV4_TCP |
2762 			E1000_MRQC_RSS_FIELD_IPV4 |
2763 			E1000_MRQC_RSS_FIELD_IPV6_TCP_EX |
2764 			E1000_MRQC_RSS_FIELD_IPV6_EX |
2765 			E1000_MRQC_RSS_FIELD_IPV6);
2766 }
2767 
2768 static void
2769 igb_initialize_rss_mapping(struct e1000_softc *sc)
2770 {
2771 	struct e1000_hw *hw = &sc->hw;
2772 	int i;
2773 	int queue_id;
2774 	u32 reta;
2775 	u32 rss_key[10], mrqc, shift = 0;
2776 
2777 	/* XXX? */
2778 	if (hw->mac.type == e1000_82575)
2779 		shift = 6;
2780 
2781 	/*
2782 	 * The redirection table controls which destination
2783 	 * queue each bucket redirects traffic to.
2784 	 * Each DWORD represents four queues, with the LSB
2785 	 * being the first queue in the DWORD.
2786 	 *
2787 	 * This just allocates buckets to queues using round-robin
2788 	 * allocation.
2789 	 *
2790 	 * NOTE: It Just Happens to line up with the default
2791 	 * RSS allocation method.
2792 	 */
2793 
2794 	/* Warning FM follows */
2795 	reta = 0;
2796 	for (i = 0; i < 128; i++) {
2797 #ifdef RSS
2798 		queue_id = rss_get_indirection_to_bucket(i);
2799 		/*
2800 		 * If we have more queues than buckets, we'll
2801 		 * end up mapping buckets to a subset of the
2802 		 * queues.
2803 		 *
2804 		 * If we have more buckets than queues, we'll
2805 		 * end up instead assigning multiple buckets
2806 		 * to queues.
2807 		 *
2808 		 * Both are suboptimal, but we need to handle
2809 		 * the case so we don't go out of bounds
2810 		 * indexing arrays and such.
2811 		 */
2812 		queue_id = queue_id % sc->rx_num_queues;
2813 #else
2814 		queue_id = (i % sc->rx_num_queues);
2815 #endif
2816 		/* Adjust if required */
2817 		queue_id = queue_id << shift;
2818 
2819 		/*
2820 		 * The low 8 bits are for hash value (n+0);
2821 		 * The next 8 bits are for hash value (n+1), etc.
2822 		 */
2823 		reta = reta >> 8;
2824 		reta = reta | ( ((uint32_t) queue_id) << 24);
2825 		if ((i & 3) == 3) {
2826 			E1000_WRITE_REG(hw, E1000_RETA(i >> 2), reta);
2827 			reta = 0;
2828 		}
2829 	}
2830 
2831 	/* Now fill in hash table */
2832 
2833 	/*
2834 	 * MRQC: Multiple Receive Queues Command
2835 	 * Set queuing to RSS control, number depends on the device.
2836 	 */
2837 	mrqc = E1000_MRQC_ENABLE_RSS_MQ;
2838 
2839 #ifdef RSS
2840 	/* XXX ew typecasting */
2841 	rss_getkey((uint8_t *) &rss_key);
2842 #else
2843 	arc4rand(&rss_key, sizeof(rss_key), 0);
2844 #endif
2845 	for (i = 0; i < 10; i++)
2846 		E1000_WRITE_REG_ARRAY(hw, E1000_RSSRK(0), i, rss_key[i]);
2847 
2848 	/*
2849 	 * Configure the RSS fields to hash upon.
2850 	 */
2851 	mrqc |= (E1000_MRQC_RSS_FIELD_IPV4 |
2852 	    E1000_MRQC_RSS_FIELD_IPV4_TCP);
2853 	mrqc |= (E1000_MRQC_RSS_FIELD_IPV6 |
2854 	    E1000_MRQC_RSS_FIELD_IPV6_TCP);
2855 	mrqc |=( E1000_MRQC_RSS_FIELD_IPV4_UDP |
2856 	    E1000_MRQC_RSS_FIELD_IPV6_UDP);
2857 	mrqc |=( E1000_MRQC_RSS_FIELD_IPV6_UDP_EX |
2858 	    E1000_MRQC_RSS_FIELD_IPV6_TCP_EX);
2859 
2860 	E1000_WRITE_REG(hw, E1000_MRQC, mrqc);
2861 }
2862 
2863 /*********************************************************************
2864  *
2865  *  Setup networking device structure and register interface media.
2866  *
2867  **********************************************************************/
2868 static int
2869 em_setup_interface(if_ctx_t ctx)
2870 {
2871 	struct ifnet *ifp = iflib_get_ifp(ctx);
2872 	struct e1000_softc *sc = iflib_get_softc(ctx);
2873 	if_softc_ctx_t scctx = sc->shared;
2874 
2875 	INIT_DEBUGOUT("em_setup_interface: begin");
2876 
2877 	/* Single Queue */
2878 	if (sc->tx_num_queues == 1) {
2879 		if_setsendqlen(ifp, scctx->isc_ntxd[0] - 1);
2880 		if_setsendqready(ifp);
2881 	}
2882 
2883 	/*
2884 	 * Specify the media types supported by this adapter and register
2885 	 * callbacks to update media and link information
2886 	 */
2887 	if (sc->hw.phy.media_type == e1000_media_type_fiber ||
2888 	    sc->hw.phy.media_type == e1000_media_type_internal_serdes) {
2889 		u_char fiber_type = IFM_1000_SX;	/* default type */
2890 
2891 		if (sc->hw.mac.type == e1000_82545)
2892 			fiber_type = IFM_1000_LX;
2893 		ifmedia_add(sc->media, IFM_ETHER | fiber_type | IFM_FDX, 0, NULL);
2894 		ifmedia_add(sc->media, IFM_ETHER | fiber_type, 0, NULL);
2895 	} else {
2896 		ifmedia_add(sc->media, IFM_ETHER | IFM_10_T, 0, NULL);
2897 		ifmedia_add(sc->media, IFM_ETHER | IFM_10_T | IFM_FDX, 0, NULL);
2898 		ifmedia_add(sc->media, IFM_ETHER | IFM_100_TX, 0, NULL);
2899 		ifmedia_add(sc->media, IFM_ETHER | IFM_100_TX | IFM_FDX, 0, NULL);
2900 		if (sc->hw.phy.type != e1000_phy_ife) {
2901 			ifmedia_add(sc->media, IFM_ETHER | IFM_1000_T | IFM_FDX, 0, NULL);
2902 			ifmedia_add(sc->media, IFM_ETHER | IFM_1000_T, 0, NULL);
2903 		}
2904 	}
2905 	ifmedia_add(sc->media, IFM_ETHER | IFM_AUTO, 0, NULL);
2906 	ifmedia_set(sc->media, IFM_ETHER | IFM_AUTO);
2907 	return (0);
2908 }
2909 
2910 static int
2911 em_if_tx_queues_alloc(if_ctx_t ctx, caddr_t *vaddrs, uint64_t *paddrs, int ntxqs, int ntxqsets)
2912 {
2913 	struct e1000_softc *sc = iflib_get_softc(ctx);
2914 	if_softc_ctx_t scctx = sc->shared;
2915 	int error = E1000_SUCCESS;
2916 	struct em_tx_queue *que;
2917 	int i, j;
2918 
2919 	MPASS(sc->tx_num_queues > 0);
2920 	MPASS(sc->tx_num_queues == ntxqsets);
2921 
2922 	/* First allocate the top level queue structs */
2923 	if (!(sc->tx_queues =
2924 	    (struct em_tx_queue *) malloc(sizeof(struct em_tx_queue) *
2925 	    sc->tx_num_queues, M_DEVBUF, M_NOWAIT | M_ZERO))) {
2926 		device_printf(iflib_get_dev(ctx), "Unable to allocate queue memory\n");
2927 		return(ENOMEM);
2928 	}
2929 
2930 	for (i = 0, que = sc->tx_queues; i < sc->tx_num_queues; i++, que++) {
2931 		/* Set up some basics */
2932 
2933 		struct tx_ring *txr = &que->txr;
2934 		txr->sc = que->sc = sc;
2935 		que->me = txr->me =  i;
2936 
2937 		/* Allocate report status array */
2938 		if (!(txr->tx_rsq = (qidx_t *) malloc(sizeof(qidx_t) * scctx->isc_ntxd[0], M_DEVBUF, M_NOWAIT | M_ZERO))) {
2939 			device_printf(iflib_get_dev(ctx), "failed to allocate rs_idxs memory\n");
2940 			error = ENOMEM;
2941 			goto fail;
2942 		}
2943 		for (j = 0; j < scctx->isc_ntxd[0]; j++)
2944 			txr->tx_rsq[j] = QIDX_INVALID;
2945 		/* get the virtual and physical address of the hardware queues */
2946 		txr->tx_base = (struct e1000_tx_desc *)vaddrs[i*ntxqs];
2947 		txr->tx_paddr = paddrs[i*ntxqs];
2948 	}
2949 
2950 	if (bootverbose)
2951 		device_printf(iflib_get_dev(ctx),
2952 		    "allocated for %d tx_queues\n", sc->tx_num_queues);
2953 	return (0);
2954 fail:
2955 	em_if_queues_free(ctx);
2956 	return (error);
2957 }
2958 
2959 static int
2960 em_if_rx_queues_alloc(if_ctx_t ctx, caddr_t *vaddrs, uint64_t *paddrs, int nrxqs, int nrxqsets)
2961 {
2962 	struct e1000_softc *sc = iflib_get_softc(ctx);
2963 	int error = E1000_SUCCESS;
2964 	struct em_rx_queue *que;
2965 	int i;
2966 
2967 	MPASS(sc->rx_num_queues > 0);
2968 	MPASS(sc->rx_num_queues == nrxqsets);
2969 
2970 	/* First allocate the top level queue structs */
2971 	if (!(sc->rx_queues =
2972 	    (struct em_rx_queue *) malloc(sizeof(struct em_rx_queue) *
2973 	    sc->rx_num_queues, M_DEVBUF, M_NOWAIT | M_ZERO))) {
2974 		device_printf(iflib_get_dev(ctx), "Unable to allocate queue memory\n");
2975 		error = ENOMEM;
2976 		goto fail;
2977 	}
2978 
2979 	for (i = 0, que = sc->rx_queues; i < nrxqsets; i++, que++) {
2980 		/* Set up some basics */
2981 		struct rx_ring *rxr = &que->rxr;
2982 		rxr->sc = que->sc = sc;
2983 		rxr->que = que;
2984 		que->me = rxr->me =  i;
2985 
2986 		/* get the virtual and physical address of the hardware queues */
2987 		rxr->rx_base = (union e1000_rx_desc_extended *)vaddrs[i*nrxqs];
2988 		rxr->rx_paddr = paddrs[i*nrxqs];
2989 	}
2990 
2991 	if (bootverbose)
2992 		device_printf(iflib_get_dev(ctx),
2993 		    "allocated for %d rx_queues\n", sc->rx_num_queues);
2994 
2995 	return (0);
2996 fail:
2997 	em_if_queues_free(ctx);
2998 	return (error);
2999 }
3000 
3001 static void
3002 em_if_queues_free(if_ctx_t ctx)
3003 {
3004 	struct e1000_softc *sc = iflib_get_softc(ctx);
3005 	struct em_tx_queue *tx_que = sc->tx_queues;
3006 	struct em_rx_queue *rx_que = sc->rx_queues;
3007 
3008 	if (tx_que != NULL) {
3009 		for (int i = 0; i < sc->tx_num_queues; i++, tx_que++) {
3010 			struct tx_ring *txr = &tx_que->txr;
3011 			if (txr->tx_rsq == NULL)
3012 				break;
3013 
3014 			free(txr->tx_rsq, M_DEVBUF);
3015 			txr->tx_rsq = NULL;
3016 		}
3017 		free(sc->tx_queues, M_DEVBUF);
3018 		sc->tx_queues = NULL;
3019 	}
3020 
3021 	if (rx_que != NULL) {
3022 		free(sc->rx_queues, M_DEVBUF);
3023 		sc->rx_queues = NULL;
3024 	}
3025 }
3026 
3027 /*********************************************************************
3028  *
3029  *  Enable transmit unit.
3030  *
3031  **********************************************************************/
3032 static void
3033 em_initialize_transmit_unit(if_ctx_t ctx)
3034 {
3035 	struct e1000_softc *sc = iflib_get_softc(ctx);
3036 	if_softc_ctx_t scctx = sc->shared;
3037 	struct em_tx_queue *que;
3038 	struct tx_ring	*txr;
3039 	struct e1000_hw	*hw = &sc->hw;
3040 	u32 tctl, txdctl = 0, tarc, tipg = 0;
3041 
3042 	INIT_DEBUGOUT("em_initialize_transmit_unit: begin");
3043 
3044 	for (int i = 0; i < sc->tx_num_queues; i++, txr++) {
3045 		u64 bus_addr;
3046 		caddr_t offp, endp;
3047 
3048 		que = &sc->tx_queues[i];
3049 		txr = &que->txr;
3050 		bus_addr = txr->tx_paddr;
3051 
3052 		/* Clear checksum offload context. */
3053 		offp = (caddr_t)&txr->csum_flags;
3054 		endp = (caddr_t)(txr + 1);
3055 		bzero(offp, endp - offp);
3056 
3057 		/* Base and Len of TX Ring */
3058 		E1000_WRITE_REG(hw, E1000_TDLEN(i),
3059 		    scctx->isc_ntxd[0] * sizeof(struct e1000_tx_desc));
3060 		E1000_WRITE_REG(hw, E1000_TDBAH(i),
3061 		    (u32)(bus_addr >> 32));
3062 		E1000_WRITE_REG(hw, E1000_TDBAL(i),
3063 		    (u32)bus_addr);
3064 		/* Init the HEAD/TAIL indices */
3065 		E1000_WRITE_REG(hw, E1000_TDT(i), 0);
3066 		E1000_WRITE_REG(hw, E1000_TDH(i), 0);
3067 
3068 		HW_DEBUGOUT2("Base = %x, Length = %x\n",
3069 		    E1000_READ_REG(hw, E1000_TDBAL(i)),
3070 		    E1000_READ_REG(hw, E1000_TDLEN(i)));
3071 
3072 		txdctl = 0; /* clear txdctl */
3073 		txdctl |= 0x1f; /* PTHRESH */
3074 		txdctl |= 1 << 8; /* HTHRESH */
3075 		txdctl |= 1 << 16;/* WTHRESH */
3076 		txdctl |= 1 << 22; /* Reserved bit 22 must always be 1 */
3077 		txdctl |= E1000_TXDCTL_GRAN;
3078 		txdctl |= 1 << 25; /* LWTHRESH */
3079 
3080 		E1000_WRITE_REG(hw, E1000_TXDCTL(i), txdctl);
3081 	}
3082 
3083 	/* Set the default values for the Tx Inter Packet Gap timer */
3084 	switch (hw->mac.type) {
3085 	case e1000_80003es2lan:
3086 		tipg = DEFAULT_82543_TIPG_IPGR1;
3087 		tipg |= DEFAULT_80003ES2LAN_TIPG_IPGR2 <<
3088 		    E1000_TIPG_IPGR2_SHIFT;
3089 		break;
3090 	case e1000_82542:
3091 		tipg = DEFAULT_82542_TIPG_IPGT;
3092 		tipg |= DEFAULT_82542_TIPG_IPGR1 << E1000_TIPG_IPGR1_SHIFT;
3093 		tipg |= DEFAULT_82542_TIPG_IPGR2 << E1000_TIPG_IPGR2_SHIFT;
3094 		break;
3095 	default:
3096 		if (hw->phy.media_type == e1000_media_type_fiber ||
3097 		    hw->phy.media_type == e1000_media_type_internal_serdes)
3098 			tipg = DEFAULT_82543_TIPG_IPGT_FIBER;
3099 		else
3100 			tipg = DEFAULT_82543_TIPG_IPGT_COPPER;
3101 		tipg |= DEFAULT_82543_TIPG_IPGR1 << E1000_TIPG_IPGR1_SHIFT;
3102 		tipg |= DEFAULT_82543_TIPG_IPGR2 << E1000_TIPG_IPGR2_SHIFT;
3103 	}
3104 
3105 	E1000_WRITE_REG(hw, E1000_TIPG, tipg);
3106 	E1000_WRITE_REG(hw, E1000_TIDV, sc->tx_int_delay.value);
3107 
3108 	if(hw->mac.type >= e1000_82540)
3109 		E1000_WRITE_REG(hw, E1000_TADV,
3110 		    sc->tx_abs_int_delay.value);
3111 
3112 	if (hw->mac.type == e1000_82571 || hw->mac.type == e1000_82572) {
3113 		tarc = E1000_READ_REG(hw, E1000_TARC(0));
3114 		tarc |= TARC_SPEED_MODE_BIT;
3115 		E1000_WRITE_REG(hw, E1000_TARC(0), tarc);
3116 	} else if (hw->mac.type == e1000_80003es2lan) {
3117 		/* errata: program both queues to unweighted RR */
3118 		tarc = E1000_READ_REG(hw, E1000_TARC(0));
3119 		tarc |= 1;
3120 		E1000_WRITE_REG(hw, E1000_TARC(0), tarc);
3121 		tarc = E1000_READ_REG(hw, E1000_TARC(1));
3122 		tarc |= 1;
3123 		E1000_WRITE_REG(hw, E1000_TARC(1), tarc);
3124 	} else if (hw->mac.type == e1000_82574) {
3125 		tarc = E1000_READ_REG(hw, E1000_TARC(0));
3126 		tarc |= TARC_ERRATA_BIT;
3127 		if ( sc->tx_num_queues > 1) {
3128 			tarc |= (TARC_COMPENSATION_MODE | TARC_MQ_FIX);
3129 			E1000_WRITE_REG(hw, E1000_TARC(0), tarc);
3130 			E1000_WRITE_REG(hw, E1000_TARC(1), tarc);
3131 		} else
3132 			E1000_WRITE_REG(hw, E1000_TARC(0), tarc);
3133 	}
3134 
3135 	if (sc->tx_int_delay.value > 0)
3136 		sc->txd_cmd |= E1000_TXD_CMD_IDE;
3137 
3138 	/* Program the Transmit Control Register */
3139 	tctl = E1000_READ_REG(hw, E1000_TCTL);
3140 	tctl &= ~E1000_TCTL_CT;
3141 	tctl |= (E1000_TCTL_PSP | E1000_TCTL_RTLC | E1000_TCTL_EN |
3142 		   (E1000_COLLISION_THRESHOLD << E1000_CT_SHIFT));
3143 
3144 	if (hw->mac.type >= e1000_82571)
3145 		tctl |= E1000_TCTL_MULR;
3146 
3147 	/* This write will effectively turn on the transmit unit. */
3148 	E1000_WRITE_REG(hw, E1000_TCTL, tctl);
3149 
3150 	/* SPT and KBL errata workarounds */
3151 	if (hw->mac.type == e1000_pch_spt) {
3152 		u32 reg;
3153 		reg = E1000_READ_REG(hw, E1000_IOSFPC);
3154 		reg |= E1000_RCTL_RDMTS_HEX;
3155 		E1000_WRITE_REG(hw, E1000_IOSFPC, reg);
3156 		/* i218-i219 Specification Update 1.5.4.5 */
3157 		reg = E1000_READ_REG(hw, E1000_TARC(0));
3158 		reg &= ~E1000_TARC0_CB_MULTIQ_3_REQ;
3159 		reg |= E1000_TARC0_CB_MULTIQ_2_REQ;
3160 		E1000_WRITE_REG(hw, E1000_TARC(0), reg);
3161 	}
3162 }
3163 
3164 /*********************************************************************
3165  *
3166  *  Enable receive unit.
3167  *
3168  **********************************************************************/
3169 #define BSIZEPKT_ROUNDUP ((1<<E1000_SRRCTL_BSIZEPKT_SHIFT)-1)
3170 
3171 static void
3172 em_initialize_receive_unit(if_ctx_t ctx)
3173 {
3174 	struct e1000_softc *sc = iflib_get_softc(ctx);
3175 	if_softc_ctx_t scctx = sc->shared;
3176 	struct ifnet *ifp = iflib_get_ifp(ctx);
3177 	struct e1000_hw	*hw = &sc->hw;
3178 	struct em_rx_queue *que;
3179 	int i;
3180 	uint32_t rctl, rxcsum;
3181 
3182 	INIT_DEBUGOUT("em_initialize_receive_units: begin");
3183 
3184 	/*
3185 	 * Make sure receives are disabled while setting
3186 	 * up the descriptor ring
3187 	 */
3188 	rctl = E1000_READ_REG(hw, E1000_RCTL);
3189 	/* Do not disable if ever enabled on this hardware */
3190 	if ((hw->mac.type != e1000_82574) && (hw->mac.type != e1000_82583))
3191 		E1000_WRITE_REG(hw, E1000_RCTL, rctl & ~E1000_RCTL_EN);
3192 
3193 	/* Setup the Receive Control Register */
3194 	rctl &= ~(3 << E1000_RCTL_MO_SHIFT);
3195 	rctl |= E1000_RCTL_EN | E1000_RCTL_BAM |
3196 	    E1000_RCTL_LBM_NO | E1000_RCTL_RDMTS_HALF |
3197 	    (hw->mac.mc_filter_type << E1000_RCTL_MO_SHIFT);
3198 
3199 	/* Do not store bad packets */
3200 	rctl &= ~E1000_RCTL_SBP;
3201 
3202 	/* Enable Long Packet receive */
3203 	if (if_getmtu(ifp) > ETHERMTU)
3204 		rctl |= E1000_RCTL_LPE;
3205 	else
3206 		rctl &= ~E1000_RCTL_LPE;
3207 
3208 	/* Strip the CRC */
3209 	if (!em_disable_crc_stripping)
3210 		rctl |= E1000_RCTL_SECRC;
3211 
3212 	if (hw->mac.type >= e1000_82540) {
3213 		E1000_WRITE_REG(hw, E1000_RADV,
3214 		    sc->rx_abs_int_delay.value);
3215 
3216 		/*
3217 		 * Set the interrupt throttling rate. Value is calculated
3218 		 * as DEFAULT_ITR = 1/(MAX_INTS_PER_SEC * 256ns)
3219 		 */
3220 		E1000_WRITE_REG(hw, E1000_ITR, DEFAULT_ITR);
3221 	}
3222 	E1000_WRITE_REG(hw, E1000_RDTR, sc->rx_int_delay.value);
3223 
3224 	if (hw->mac.type >= em_mac_min) {
3225 		uint32_t rfctl;
3226 		/* Use extended rx descriptor formats */
3227 		rfctl = E1000_READ_REG(hw, E1000_RFCTL);
3228 		rfctl |= E1000_RFCTL_EXTEN;
3229 
3230 		/*
3231 		 * When using MSI-X interrupts we need to throttle
3232 		 * using the EITR register (82574 only)
3233 		 */
3234 		if (hw->mac.type == e1000_82574) {
3235 			for (int i = 0; i < 4; i++)
3236 				E1000_WRITE_REG(hw, E1000_EITR_82574(i),
3237 				    DEFAULT_ITR);
3238 			/* Disable accelerated acknowledge */
3239 			rfctl |= E1000_RFCTL_ACK_DIS;
3240 		}
3241 		E1000_WRITE_REG(hw, E1000_RFCTL, rfctl);
3242 	}
3243 
3244 	/* Set up L3 and L4 csum Rx descriptor offloads */
3245 	rxcsum = E1000_READ_REG(hw, E1000_RXCSUM);
3246 	if (if_getcapenable(ifp) & IFCAP_RXCSUM) {
3247 		rxcsum |= E1000_RXCSUM_TUOFL | E1000_RXCSUM_IPOFL;
3248 		if (hw->mac.type > e1000_82575)
3249 			rxcsum |= E1000_RXCSUM_CRCOFL;
3250 		else if (hw->mac.type < em_mac_min &&
3251 		    if_getcapenable(ifp) & IFCAP_HWCSUM_IPV6)
3252 			rxcsum |= E1000_RXCSUM_IPV6OFL;
3253 	} else {
3254 		rxcsum &= ~(E1000_RXCSUM_IPOFL | E1000_RXCSUM_TUOFL);
3255 		if (hw->mac.type > e1000_82575)
3256 			rxcsum &= ~E1000_RXCSUM_CRCOFL;
3257 		else if (hw->mac.type < em_mac_min)
3258 			rxcsum &= ~E1000_RXCSUM_IPV6OFL;
3259 	}
3260 
3261 	if (sc->rx_num_queues > 1) {
3262 		/* RSS hash needed in the Rx descriptor */
3263 		rxcsum |= E1000_RXCSUM_PCSD;
3264 
3265 		if (hw->mac.type >= igb_mac_min)
3266 			igb_initialize_rss_mapping(sc);
3267 		else
3268 			em_initialize_rss_mapping(sc);
3269 	}
3270 	E1000_WRITE_REG(hw, E1000_RXCSUM, rxcsum);
3271 
3272 	/*
3273 	 * XXX TEMPORARY WORKAROUND: on some systems with 82573
3274 	 * long latencies are observed, like Lenovo X60. This
3275 	 * change eliminates the problem, but since having positive
3276 	 * values in RDTR is a known source of problems on other
3277 	 * platforms another solution is being sought.
3278 	 */
3279 	if (hw->mac.type == e1000_82573)
3280 		E1000_WRITE_REG(hw, E1000_RDTR, 0x20);
3281 
3282 	for (i = 0, que = sc->rx_queues; i < sc->rx_num_queues; i++, que++) {
3283 		struct rx_ring *rxr = &que->rxr;
3284 		/* Setup the Base and Length of the Rx Descriptor Ring */
3285 		u64 bus_addr = rxr->rx_paddr;
3286 #if 0
3287 		u32 rdt = sc->rx_num_queues -1;  /* default */
3288 #endif
3289 
3290 		E1000_WRITE_REG(hw, E1000_RDLEN(i),
3291 		    scctx->isc_nrxd[0] * sizeof(union e1000_rx_desc_extended));
3292 		E1000_WRITE_REG(hw, E1000_RDBAH(i), (u32)(bus_addr >> 32));
3293 		E1000_WRITE_REG(hw, E1000_RDBAL(i), (u32)bus_addr);
3294 		/* Setup the Head and Tail Descriptor Pointers */
3295 		E1000_WRITE_REG(hw, E1000_RDH(i), 0);
3296 		E1000_WRITE_REG(hw, E1000_RDT(i), 0);
3297 	}
3298 
3299 	/*
3300 	 * Set PTHRESH for improved jumbo performance
3301 	 * According to 10.2.5.11 of Intel 82574 Datasheet,
3302 	 * RXDCTL(1) is written whenever RXDCTL(0) is written.
3303 	 * Only write to RXDCTL(1) if there is a need for different
3304 	 * settings.
3305 	 */
3306 	if ((hw->mac.type == e1000_ich9lan || hw->mac.type == e1000_pch2lan ||
3307 	    hw->mac.type == e1000_ich10lan) && if_getmtu(ifp) > ETHERMTU) {
3308 		u32 rxdctl = E1000_READ_REG(hw, E1000_RXDCTL(0));
3309 		E1000_WRITE_REG(hw, E1000_RXDCTL(0), rxdctl | 3);
3310 	} else if (hw->mac.type == e1000_82574) {
3311 		for (int i = 0; i < sc->rx_num_queues; i++) {
3312 			u32 rxdctl = E1000_READ_REG(hw, E1000_RXDCTL(i));
3313 			rxdctl |= 0x20; /* PTHRESH */
3314 			rxdctl |= 4 << 8; /* HTHRESH */
3315 			rxdctl |= 4 << 16;/* WTHRESH */
3316 			rxdctl |= 1 << 24; /* Switch to granularity */
3317 			E1000_WRITE_REG(hw, E1000_RXDCTL(i), rxdctl);
3318 		}
3319 	} else if (hw->mac.type >= igb_mac_min) {
3320 		u32 psize, srrctl = 0;
3321 
3322 		if (if_getmtu(ifp) > ETHERMTU) {
3323 			psize = scctx->isc_max_frame_size;
3324 			/* are we on a vlan? */
3325 			if (ifp->if_vlantrunk != NULL)
3326 				psize += VLAN_TAG_SIZE;
3327 
3328 			if (sc->vf_ifp)
3329 				e1000_rlpml_set_vf(hw, psize);
3330 			else
3331 				E1000_WRITE_REG(hw, E1000_RLPML, psize);
3332 		}
3333 
3334 		/* Set maximum packet buffer len */
3335 		srrctl |= (sc->rx_mbuf_sz + BSIZEPKT_ROUNDUP) >>
3336 		    E1000_SRRCTL_BSIZEPKT_SHIFT;
3337 
3338 		/*
3339 		 * If TX flow control is disabled and there's >1 queue defined,
3340 		 * enable DROP.
3341 		 *
3342 		 * This drops frames rather than hanging the RX MAC for all queues.
3343 		 */
3344 		if ((sc->rx_num_queues > 1) &&
3345 		    (sc->fc == e1000_fc_none ||
3346 		     sc->fc == e1000_fc_rx_pause)) {
3347 			srrctl |= E1000_SRRCTL_DROP_EN;
3348 		}
3349 			/* Setup the Base and Length of the Rx Descriptor Rings */
3350 		for (i = 0, que = sc->rx_queues; i < sc->rx_num_queues; i++, que++) {
3351 			struct rx_ring *rxr = &que->rxr;
3352 			u64 bus_addr = rxr->rx_paddr;
3353 			u32 rxdctl;
3354 
3355 #ifdef notyet
3356 			/* Configure for header split? -- ignore for now */
3357 			rxr->hdr_split = igb_header_split;
3358 #else
3359 			srrctl |= E1000_SRRCTL_DESCTYPE_ADV_ONEBUF;
3360 #endif
3361 
3362 			E1000_WRITE_REG(hw, E1000_RDLEN(i),
3363 					scctx->isc_nrxd[0] * sizeof(struct e1000_rx_desc));
3364 			E1000_WRITE_REG(hw, E1000_RDBAH(i),
3365 					(uint32_t)(bus_addr >> 32));
3366 			E1000_WRITE_REG(hw, E1000_RDBAL(i),
3367 					(uint32_t)bus_addr);
3368 			E1000_WRITE_REG(hw, E1000_SRRCTL(i), srrctl);
3369 			/* Enable this Queue */
3370 			rxdctl = E1000_READ_REG(hw, E1000_RXDCTL(i));
3371 			rxdctl |= E1000_RXDCTL_QUEUE_ENABLE;
3372 			rxdctl &= 0xFFF00000;
3373 			rxdctl |= IGB_RX_PTHRESH;
3374 			rxdctl |= IGB_RX_HTHRESH << 8;
3375 			rxdctl |= IGB_RX_WTHRESH << 16;
3376 			E1000_WRITE_REG(hw, E1000_RXDCTL(i), rxdctl);
3377 		}
3378 	} else if (hw->mac.type >= e1000_pch2lan) {
3379 		if (if_getmtu(ifp) > ETHERMTU)
3380 			e1000_lv_jumbo_workaround_ich8lan(hw, true);
3381 		else
3382 			e1000_lv_jumbo_workaround_ich8lan(hw, false);
3383 	}
3384 
3385 	/* Make sure VLAN Filters are off */
3386 	rctl &= ~E1000_RCTL_VFE;
3387 
3388 	/* Set up packet buffer size, overridden by per queue srrctl on igb */
3389 	if (hw->mac.type < igb_mac_min) {
3390 		if (sc->rx_mbuf_sz > 2048 && sc->rx_mbuf_sz <= 4096)
3391 			rctl |= E1000_RCTL_SZ_4096 | E1000_RCTL_BSEX;
3392 		else if (sc->rx_mbuf_sz > 4096 && sc->rx_mbuf_sz <= 8192)
3393 			rctl |= E1000_RCTL_SZ_8192 | E1000_RCTL_BSEX;
3394 		else if (sc->rx_mbuf_sz > 8192)
3395 			rctl |= E1000_RCTL_SZ_16384 | E1000_RCTL_BSEX;
3396 		else {
3397 			rctl |= E1000_RCTL_SZ_2048;
3398 			rctl &= ~E1000_RCTL_BSEX;
3399 		}
3400 	} else
3401 		rctl |= E1000_RCTL_SZ_2048;
3402 
3403 	/*
3404 	 * rctl bits 11:10 are as follows
3405 	 * lem: reserved
3406 	 * em: DTYPE
3407 	 * igb: reserved
3408 	 * and should be 00 on all of the above
3409 	 */
3410 	rctl &= ~0x00000C00;
3411 
3412 	/* Write out the settings */
3413 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
3414 
3415 	return;
3416 }
3417 
3418 static void
3419 em_if_vlan_register(if_ctx_t ctx, u16 vtag)
3420 {
3421 	struct e1000_softc *sc = iflib_get_softc(ctx);
3422 	u32 index, bit;
3423 
3424 	index = (vtag >> 5) & 0x7F;
3425 	bit = vtag & 0x1F;
3426 	sc->shadow_vfta[index] |= (1 << bit);
3427 	++sc->num_vlans;
3428 	em_if_vlan_filter_write(sc);
3429 }
3430 
3431 static void
3432 em_if_vlan_unregister(if_ctx_t ctx, u16 vtag)
3433 {
3434 	struct e1000_softc *sc = iflib_get_softc(ctx);
3435 	u32 index, bit;
3436 
3437 	index = (vtag >> 5) & 0x7F;
3438 	bit = vtag & 0x1F;
3439 	sc->shadow_vfta[index] &= ~(1 << bit);
3440 	--sc->num_vlans;
3441 	em_if_vlan_filter_write(sc);
3442 }
3443 
3444 static bool
3445 em_if_vlan_filter_capable(if_ctx_t ctx)
3446 {
3447 	if_t ifp = iflib_get_ifp(ctx);
3448 
3449 	if ((if_getcapenable(ifp) & IFCAP_VLAN_HWFILTER) &&
3450 	    !em_disable_crc_stripping)
3451 		return (true);
3452 
3453 	return (false);
3454 }
3455 
3456 static bool
3457 em_if_vlan_filter_used(if_ctx_t ctx)
3458 {
3459 	struct e1000_softc *sc = iflib_get_softc(ctx);
3460 
3461 	if (!em_if_vlan_filter_capable(ctx))
3462 		return (false);
3463 
3464 	for (int i = 0; i < EM_VFTA_SIZE; i++)
3465 		if (sc->shadow_vfta[i] != 0)
3466 			return (true);
3467 
3468 	return (false);
3469 }
3470 
3471 static void
3472 em_if_vlan_filter_enable(struct e1000_softc *sc)
3473 {
3474 	struct e1000_hw *hw = &sc->hw;
3475 	u32 reg;
3476 
3477 	reg = E1000_READ_REG(hw, E1000_RCTL);
3478 	reg &= ~E1000_RCTL_CFIEN;
3479 	reg |= E1000_RCTL_VFE;
3480 	E1000_WRITE_REG(hw, E1000_RCTL, reg);
3481 }
3482 
3483 static void
3484 em_if_vlan_filter_disable(struct e1000_softc *sc)
3485 {
3486 	struct e1000_hw *hw = &sc->hw;
3487 	u32 reg;
3488 
3489 	reg = E1000_READ_REG(hw, E1000_RCTL);
3490 	reg &= ~(E1000_RCTL_VFE | E1000_RCTL_CFIEN);
3491 	E1000_WRITE_REG(hw, E1000_RCTL, reg);
3492 }
3493 
3494 static void
3495 em_if_vlan_filter_write(struct e1000_softc *sc)
3496 {
3497 	struct e1000_hw *hw = &sc->hw;
3498 
3499 	if (sc->vf_ifp)
3500 		return;
3501 
3502 	/* Disable interrupts for lem-class devices during the filter change */
3503 	if (hw->mac.type < em_mac_min)
3504 		em_if_intr_disable(sc->ctx);
3505 
3506 	for (int i = 0; i < EM_VFTA_SIZE; i++)
3507 		if (sc->shadow_vfta[i] != 0) {
3508 			/* XXXKB: incomplete VF support, we return early above */
3509 			if (sc->vf_ifp)
3510 				e1000_vfta_set_vf(hw, sc->shadow_vfta[i], true);
3511 			else
3512 				e1000_write_vfta(hw, i, sc->shadow_vfta[i]);
3513 		}
3514 
3515 	/* Re-enable interrupts for lem-class devices */
3516 	if (hw->mac.type < em_mac_min)
3517 		em_if_intr_enable(sc->ctx);
3518 }
3519 
3520 static void
3521 em_setup_vlan_hw_support(if_ctx_t ctx)
3522 {
3523 	struct e1000_softc *sc = iflib_get_softc(ctx);
3524 	struct e1000_hw *hw = &sc->hw;
3525 	struct ifnet *ifp = iflib_get_ifp(ctx);
3526 	u32 reg;
3527 
3528 	/* XXXKB: Return early if we are a VF until VF decap and filter management
3529 	 * is ready and tested.
3530 	 */
3531 	if (sc->vf_ifp)
3532 		return;
3533 
3534 	if (if_getcapenable(ifp) & IFCAP_VLAN_HWTAGGING &&
3535 	    !em_disable_crc_stripping) {
3536 		reg = E1000_READ_REG(hw, E1000_CTRL);
3537 		reg |= E1000_CTRL_VME;
3538 		E1000_WRITE_REG(hw, E1000_CTRL, reg);
3539 	} else {
3540 		reg = E1000_READ_REG(hw, E1000_CTRL);
3541 		reg &= ~E1000_CTRL_VME;
3542 		E1000_WRITE_REG(hw, E1000_CTRL, reg);
3543 	}
3544 
3545 	/* If we aren't doing HW filtering, we're done */
3546 	if (!em_if_vlan_filter_capable(ctx))  {
3547 		em_if_vlan_filter_disable(sc);
3548 		return;
3549 	}
3550 
3551 	/*
3552 	 * A soft reset zero's out the VFTA, so
3553 	 * we need to repopulate it now.
3554 	 */
3555 	em_if_vlan_filter_write(sc);
3556 
3557 	/* Enable the Filter Table */
3558 	em_if_vlan_filter_enable(sc);
3559 }
3560 
3561 static void
3562 em_if_intr_enable(if_ctx_t ctx)
3563 {
3564 	struct e1000_softc *sc = iflib_get_softc(ctx);
3565 	struct e1000_hw *hw = &sc->hw;
3566 	u32 ims_mask = IMS_ENABLE_MASK;
3567 
3568 	if (sc->intr_type == IFLIB_INTR_MSIX) {
3569 		E1000_WRITE_REG(hw, EM_EIAC, sc->ims);
3570 		ims_mask |= sc->ims;
3571 	}
3572 	E1000_WRITE_REG(hw, E1000_IMS, ims_mask);
3573 	E1000_WRITE_FLUSH(hw);
3574 }
3575 
3576 static void
3577 em_if_intr_disable(if_ctx_t ctx)
3578 {
3579 	struct e1000_softc *sc = iflib_get_softc(ctx);
3580 	struct e1000_hw *hw = &sc->hw;
3581 
3582 	if (sc->intr_type == IFLIB_INTR_MSIX)
3583 		E1000_WRITE_REG(hw, EM_EIAC, 0);
3584 	E1000_WRITE_REG(hw, E1000_IMC, 0xffffffff);
3585 	E1000_WRITE_FLUSH(hw);
3586 }
3587 
3588 static void
3589 igb_if_intr_enable(if_ctx_t ctx)
3590 {
3591 	struct e1000_softc *sc = iflib_get_softc(ctx);
3592 	struct e1000_hw *hw = &sc->hw;
3593 	u32 mask;
3594 
3595 	if (__predict_true(sc->intr_type == IFLIB_INTR_MSIX)) {
3596 		mask = (sc->que_mask | sc->link_mask);
3597 		E1000_WRITE_REG(hw, E1000_EIAC, mask);
3598 		E1000_WRITE_REG(hw, E1000_EIAM, mask);
3599 		E1000_WRITE_REG(hw, E1000_EIMS, mask);
3600 		E1000_WRITE_REG(hw, E1000_IMS, E1000_IMS_LSC);
3601 	} else
3602 		E1000_WRITE_REG(hw, E1000_IMS, IMS_ENABLE_MASK);
3603 	E1000_WRITE_FLUSH(hw);
3604 }
3605 
3606 static void
3607 igb_if_intr_disable(if_ctx_t ctx)
3608 {
3609 	struct e1000_softc *sc = iflib_get_softc(ctx);
3610 	struct e1000_hw *hw = &sc->hw;
3611 
3612 	if (__predict_true(sc->intr_type == IFLIB_INTR_MSIX)) {
3613 		E1000_WRITE_REG(hw, E1000_EIMC, 0xffffffff);
3614 		E1000_WRITE_REG(hw, E1000_EIAC, 0);
3615 	}
3616 	E1000_WRITE_REG(hw, E1000_IMC, 0xffffffff);
3617 	E1000_WRITE_FLUSH(hw);
3618 }
3619 
3620 /*
3621  * Bit of a misnomer, what this really means is
3622  * to enable OS management of the system... aka
3623  * to disable special hardware management features
3624  */
3625 static void
3626 em_init_manageability(struct e1000_softc *sc)
3627 {
3628 	/* A shared code workaround */
3629 #define E1000_82542_MANC2H E1000_MANC2H
3630 	if (sc->has_manage) {
3631 		int manc2h = E1000_READ_REG(&sc->hw, E1000_MANC2H);
3632 		int manc = E1000_READ_REG(&sc->hw, E1000_MANC);
3633 
3634 		/* disable hardware interception of ARP */
3635 		manc &= ~(E1000_MANC_ARP_EN);
3636 
3637 		/* enable receiving management packets to the host */
3638 		manc |= E1000_MANC_EN_MNG2HOST;
3639 #define E1000_MNG2HOST_PORT_623 (1 << 5)
3640 #define E1000_MNG2HOST_PORT_664 (1 << 6)
3641 		manc2h |= E1000_MNG2HOST_PORT_623;
3642 		manc2h |= E1000_MNG2HOST_PORT_664;
3643 		E1000_WRITE_REG(&sc->hw, E1000_MANC2H, manc2h);
3644 		E1000_WRITE_REG(&sc->hw, E1000_MANC, manc);
3645 	}
3646 }
3647 
3648 /*
3649  * Give control back to hardware management
3650  * controller if there is one.
3651  */
3652 static void
3653 em_release_manageability(struct e1000_softc *sc)
3654 {
3655 	if (sc->has_manage) {
3656 		int manc = E1000_READ_REG(&sc->hw, E1000_MANC);
3657 
3658 		/* re-enable hardware interception of ARP */
3659 		manc |= E1000_MANC_ARP_EN;
3660 		manc &= ~E1000_MANC_EN_MNG2HOST;
3661 
3662 		E1000_WRITE_REG(&sc->hw, E1000_MANC, manc);
3663 	}
3664 }
3665 
3666 /*
3667  * em_get_hw_control sets the {CTRL_EXT|FWSM}:DRV_LOAD bit.
3668  * For ASF and Pass Through versions of f/w this means
3669  * that the driver is loaded. For AMT version type f/w
3670  * this means that the network i/f is open.
3671  */
3672 static void
3673 em_get_hw_control(struct e1000_softc *sc)
3674 {
3675 	u32 ctrl_ext, swsm;
3676 
3677 	if (sc->vf_ifp)
3678 		return;
3679 
3680 	if (sc->hw.mac.type == e1000_82573) {
3681 		swsm = E1000_READ_REG(&sc->hw, E1000_SWSM);
3682 		E1000_WRITE_REG(&sc->hw, E1000_SWSM,
3683 		    swsm | E1000_SWSM_DRV_LOAD);
3684 		return;
3685 	}
3686 	/* else */
3687 	ctrl_ext = E1000_READ_REG(&sc->hw, E1000_CTRL_EXT);
3688 	E1000_WRITE_REG(&sc->hw, E1000_CTRL_EXT,
3689 	    ctrl_ext | E1000_CTRL_EXT_DRV_LOAD);
3690 }
3691 
3692 /*
3693  * em_release_hw_control resets {CTRL_EXT|FWSM}:DRV_LOAD bit.
3694  * For ASF and Pass Through versions of f/w this means that
3695  * the driver is no longer loaded. For AMT versions of the
3696  * f/w this means that the network i/f is closed.
3697  */
3698 static void
3699 em_release_hw_control(struct e1000_softc *sc)
3700 {
3701 	u32 ctrl_ext, swsm;
3702 
3703 	if (!sc->has_manage)
3704 		return;
3705 
3706 	if (sc->hw.mac.type == e1000_82573) {
3707 		swsm = E1000_READ_REG(&sc->hw, E1000_SWSM);
3708 		E1000_WRITE_REG(&sc->hw, E1000_SWSM,
3709 		    swsm & ~E1000_SWSM_DRV_LOAD);
3710 		return;
3711 	}
3712 	/* else */
3713 	ctrl_ext = E1000_READ_REG(&sc->hw, E1000_CTRL_EXT);
3714 	E1000_WRITE_REG(&sc->hw, E1000_CTRL_EXT,
3715 	    ctrl_ext & ~E1000_CTRL_EXT_DRV_LOAD);
3716 	return;
3717 }
3718 
3719 static int
3720 em_is_valid_ether_addr(u8 *addr)
3721 {
3722 	char zero_addr[6] = { 0, 0, 0, 0, 0, 0 };
3723 
3724 	if ((addr[0] & 1) || (!bcmp(addr, zero_addr, ETHER_ADDR_LEN))) {
3725 		return (false);
3726 	}
3727 
3728 	return (true);
3729 }
3730 
3731 /*
3732 ** Parse the interface capabilities with regard
3733 ** to both system management and wake-on-lan for
3734 ** later use.
3735 */
3736 static void
3737 em_get_wakeup(if_ctx_t ctx)
3738 {
3739 	struct e1000_softc *sc = iflib_get_softc(ctx);
3740 	device_t dev = iflib_get_dev(ctx);
3741 	u16 eeprom_data = 0, device_id, apme_mask;
3742 
3743 	sc->has_manage = e1000_enable_mng_pass_thru(&sc->hw);
3744 	apme_mask = EM_EEPROM_APME;
3745 
3746 	switch (sc->hw.mac.type) {
3747 	case e1000_82542:
3748 	case e1000_82543:
3749 		break;
3750 	case e1000_82544:
3751 		e1000_read_nvm(&sc->hw,
3752 		    NVM_INIT_CONTROL2_REG, 1, &eeprom_data);
3753 		apme_mask = EM_82544_APME;
3754 		break;
3755 	case e1000_82546:
3756 	case e1000_82546_rev_3:
3757 		if (sc->hw.bus.func == 1) {
3758 			e1000_read_nvm(&sc->hw,
3759 			    NVM_INIT_CONTROL3_PORT_B, 1, &eeprom_data);
3760 			break;
3761 		} else
3762 			e1000_read_nvm(&sc->hw,
3763 			    NVM_INIT_CONTROL3_PORT_A, 1, &eeprom_data);
3764 		break;
3765 	case e1000_82573:
3766 	case e1000_82583:
3767 		sc->has_amt = true;
3768 		/* FALLTHROUGH */
3769 	case e1000_82571:
3770 	case e1000_82572:
3771 	case e1000_80003es2lan:
3772 		if (sc->hw.bus.func == 1) {
3773 			e1000_read_nvm(&sc->hw,
3774 			    NVM_INIT_CONTROL3_PORT_B, 1, &eeprom_data);
3775 			break;
3776 		} else
3777 			e1000_read_nvm(&sc->hw,
3778 			    NVM_INIT_CONTROL3_PORT_A, 1, &eeprom_data);
3779 		break;
3780 	case e1000_ich8lan:
3781 	case e1000_ich9lan:
3782 	case e1000_ich10lan:
3783 	case e1000_pchlan:
3784 	case e1000_pch2lan:
3785 	case e1000_pch_lpt:
3786 	case e1000_pch_spt:
3787 	case e1000_82575:	/* listing all igb devices */
3788 	case e1000_82576:
3789 	case e1000_82580:
3790 	case e1000_i350:
3791 	case e1000_i354:
3792 	case e1000_i210:
3793 	case e1000_i211:
3794 	case e1000_vfadapt:
3795 	case e1000_vfadapt_i350:
3796 		apme_mask = E1000_WUC_APME;
3797 		sc->has_amt = true;
3798 		eeprom_data = E1000_READ_REG(&sc->hw, E1000_WUC);
3799 		break;
3800 	default:
3801 		e1000_read_nvm(&sc->hw,
3802 		    NVM_INIT_CONTROL3_PORT_A, 1, &eeprom_data);
3803 		break;
3804 	}
3805 	if (eeprom_data & apme_mask)
3806 		sc->wol = (E1000_WUFC_MAG | E1000_WUFC_MC);
3807 	/*
3808 	 * We have the eeprom settings, now apply the special cases
3809 	 * where the eeprom may be wrong or the board won't support
3810 	 * wake on lan on a particular port
3811 	 */
3812 	device_id = pci_get_device(dev);
3813 	switch (device_id) {
3814 	case E1000_DEV_ID_82546GB_PCIE:
3815 		sc->wol = 0;
3816 		break;
3817 	case E1000_DEV_ID_82546EB_FIBER:
3818 	case E1000_DEV_ID_82546GB_FIBER:
3819 		/* Wake events only supported on port A for dual fiber
3820 		 * regardless of eeprom setting */
3821 		if (E1000_READ_REG(&sc->hw, E1000_STATUS) &
3822 		    E1000_STATUS_FUNC_1)
3823 			sc->wol = 0;
3824 		break;
3825 	case E1000_DEV_ID_82546GB_QUAD_COPPER_KSP3:
3826 		/* if quad port adapter, disable WoL on all but port A */
3827 		if (global_quad_port_a != 0)
3828 			sc->wol = 0;
3829 		/* Reset for multiple quad port adapters */
3830 		if (++global_quad_port_a == 4)
3831 			global_quad_port_a = 0;
3832 		break;
3833 	case E1000_DEV_ID_82571EB_FIBER:
3834 		/* Wake events only supported on port A for dual fiber
3835 		 * regardless of eeprom setting */
3836 		if (E1000_READ_REG(&sc->hw, E1000_STATUS) &
3837 		    E1000_STATUS_FUNC_1)
3838 			sc->wol = 0;
3839 		break;
3840 	case E1000_DEV_ID_82571EB_QUAD_COPPER:
3841 	case E1000_DEV_ID_82571EB_QUAD_FIBER:
3842 	case E1000_DEV_ID_82571EB_QUAD_COPPER_LP:
3843 		/* if quad port adapter, disable WoL on all but port A */
3844 		if (global_quad_port_a != 0)
3845 			sc->wol = 0;
3846 		/* Reset for multiple quad port adapters */
3847 		if (++global_quad_port_a == 4)
3848 			global_quad_port_a = 0;
3849 		break;
3850 	}
3851 	return;
3852 }
3853 
3854 
3855 /*
3856  * Enable PCI Wake On Lan capability
3857  */
3858 static void
3859 em_enable_wakeup(if_ctx_t ctx)
3860 {
3861 	struct e1000_softc *sc = iflib_get_softc(ctx);
3862 	device_t dev = iflib_get_dev(ctx);
3863 	if_t ifp = iflib_get_ifp(ctx);
3864 	int error = 0;
3865 	u32 pmc, ctrl, ctrl_ext, rctl;
3866 	u16 status;
3867 
3868 	if (pci_find_cap(dev, PCIY_PMG, &pmc) != 0)
3869 		return;
3870 
3871 	/*
3872 	 * Determine type of Wakeup: note that wol
3873 	 * is set with all bits on by default.
3874 	 */
3875 	if ((if_getcapenable(ifp) & IFCAP_WOL_MAGIC) == 0)
3876 		sc->wol &= ~E1000_WUFC_MAG;
3877 
3878 	if ((if_getcapenable(ifp) & IFCAP_WOL_UCAST) == 0)
3879 		sc->wol &= ~E1000_WUFC_EX;
3880 
3881 	if ((if_getcapenable(ifp) & IFCAP_WOL_MCAST) == 0)
3882 		sc->wol &= ~E1000_WUFC_MC;
3883 	else {
3884 		rctl = E1000_READ_REG(&sc->hw, E1000_RCTL);
3885 		rctl |= E1000_RCTL_MPE;
3886 		E1000_WRITE_REG(&sc->hw, E1000_RCTL, rctl);
3887 	}
3888 
3889 	if (!(sc->wol & (E1000_WUFC_EX | E1000_WUFC_MAG | E1000_WUFC_MC)))
3890 		goto pme;
3891 
3892 	/* Advertise the wakeup capability */
3893 	ctrl = E1000_READ_REG(&sc->hw, E1000_CTRL);
3894 	ctrl |= (E1000_CTRL_SWDPIN2 | E1000_CTRL_SWDPIN3);
3895 	E1000_WRITE_REG(&sc->hw, E1000_CTRL, ctrl);
3896 
3897 	/* Keep the laser running on Fiber adapters */
3898 	if (sc->hw.phy.media_type == e1000_media_type_fiber ||
3899 	    sc->hw.phy.media_type == e1000_media_type_internal_serdes) {
3900 		ctrl_ext = E1000_READ_REG(&sc->hw, E1000_CTRL_EXT);
3901 		ctrl_ext |= E1000_CTRL_EXT_SDP3_DATA;
3902 		E1000_WRITE_REG(&sc->hw, E1000_CTRL_EXT, ctrl_ext);
3903 	}
3904 
3905 	if ((sc->hw.mac.type == e1000_ich8lan) ||
3906 	    (sc->hw.mac.type == e1000_pchlan) ||
3907 	    (sc->hw.mac.type == e1000_ich9lan) ||
3908 	    (sc->hw.mac.type == e1000_ich10lan))
3909 		e1000_suspend_workarounds_ich8lan(&sc->hw);
3910 
3911 	if ( sc->hw.mac.type >= e1000_pchlan) {
3912 		error = em_enable_phy_wakeup(sc);
3913 		if (error)
3914 			goto pme;
3915 	} else {
3916 		/* Enable wakeup by the MAC */
3917 		E1000_WRITE_REG(&sc->hw, E1000_WUC, E1000_WUC_PME_EN);
3918 		E1000_WRITE_REG(&sc->hw, E1000_WUFC, sc->wol);
3919 	}
3920 
3921 	if (sc->hw.phy.type == e1000_phy_igp_3)
3922 		e1000_igp3_phy_powerdown_workaround_ich8lan(&sc->hw);
3923 
3924 pme:
3925 	status = pci_read_config(dev, pmc + PCIR_POWER_STATUS, 2);
3926 	status &= ~(PCIM_PSTAT_PME | PCIM_PSTAT_PMEENABLE);
3927 	if (!error && (if_getcapenable(ifp) & IFCAP_WOL))
3928 		status |= PCIM_PSTAT_PME | PCIM_PSTAT_PMEENABLE;
3929 	pci_write_config(dev, pmc + PCIR_POWER_STATUS, status, 2);
3930 
3931 	return;
3932 }
3933 
3934 /*
3935  * WOL in the newer chipset interfaces (pchlan)
3936  * require thing to be copied into the phy
3937  */
3938 static int
3939 em_enable_phy_wakeup(struct e1000_softc *sc)
3940 {
3941 	struct e1000_hw *hw = &sc->hw;
3942 	u32 mreg, ret = 0;
3943 	u16 preg;
3944 
3945 	/* copy MAC RARs to PHY RARs */
3946 	e1000_copy_rx_addrs_to_phy_ich8lan(hw);
3947 
3948 	/* copy MAC MTA to PHY MTA */
3949 	for (int i = 0; i < hw->mac.mta_reg_count; i++) {
3950 		mreg = E1000_READ_REG_ARRAY(hw, E1000_MTA, i);
3951 		e1000_write_phy_reg(hw, BM_MTA(i), (u16)(mreg & 0xFFFF));
3952 		e1000_write_phy_reg(hw, BM_MTA(i) + 1,
3953 		    (u16)((mreg >> 16) & 0xFFFF));
3954 	}
3955 
3956 	/* configure PHY Rx Control register */
3957 	e1000_read_phy_reg(hw, BM_RCTL, &preg);
3958 	mreg = E1000_READ_REG(hw, E1000_RCTL);
3959 	if (mreg & E1000_RCTL_UPE)
3960 		preg |= BM_RCTL_UPE;
3961 	if (mreg & E1000_RCTL_MPE)
3962 		preg |= BM_RCTL_MPE;
3963 	preg &= ~(BM_RCTL_MO_MASK);
3964 	if (mreg & E1000_RCTL_MO_3)
3965 		preg |= (((mreg & E1000_RCTL_MO_3) >> E1000_RCTL_MO_SHIFT)
3966 				<< BM_RCTL_MO_SHIFT);
3967 	if (mreg & E1000_RCTL_BAM)
3968 		preg |= BM_RCTL_BAM;
3969 	if (mreg & E1000_RCTL_PMCF)
3970 		preg |= BM_RCTL_PMCF;
3971 	mreg = E1000_READ_REG(hw, E1000_CTRL);
3972 	if (mreg & E1000_CTRL_RFCE)
3973 		preg |= BM_RCTL_RFCE;
3974 	e1000_write_phy_reg(hw, BM_RCTL, preg);
3975 
3976 	/* enable PHY wakeup in MAC register */
3977 	E1000_WRITE_REG(hw, E1000_WUC,
3978 	    E1000_WUC_PHY_WAKE | E1000_WUC_PME_EN | E1000_WUC_APME);
3979 	E1000_WRITE_REG(hw, E1000_WUFC, sc->wol);
3980 
3981 	/* configure and enable PHY wakeup in PHY registers */
3982 	e1000_write_phy_reg(hw, BM_WUFC, sc->wol);
3983 	e1000_write_phy_reg(hw, BM_WUC, E1000_WUC_PME_EN);
3984 
3985 	/* activate PHY wakeup */
3986 	ret = hw->phy.ops.acquire(hw);
3987 	if (ret) {
3988 		printf("Could not acquire PHY\n");
3989 		return ret;
3990 	}
3991 	e1000_write_phy_reg_mdic(hw, IGP01E1000_PHY_PAGE_SELECT,
3992 	                         (BM_WUC_ENABLE_PAGE << IGP_PAGE_SHIFT));
3993 	ret = e1000_read_phy_reg_mdic(hw, BM_WUC_ENABLE_REG, &preg);
3994 	if (ret) {
3995 		printf("Could not read PHY page 769\n");
3996 		goto out;
3997 	}
3998 	preg |= BM_WUC_ENABLE_BIT | BM_WUC_HOST_WU_BIT;
3999 	ret = e1000_write_phy_reg_mdic(hw, BM_WUC_ENABLE_REG, preg);
4000 	if (ret)
4001 		printf("Could not set PHY Host Wakeup bit\n");
4002 out:
4003 	hw->phy.ops.release(hw);
4004 
4005 	return ret;
4006 }
4007 
4008 static void
4009 em_if_led_func(if_ctx_t ctx, int onoff)
4010 {
4011 	struct e1000_softc *sc = iflib_get_softc(ctx);
4012 
4013 	if (onoff) {
4014 		e1000_setup_led(&sc->hw);
4015 		e1000_led_on(&sc->hw);
4016 	} else {
4017 		e1000_led_off(&sc->hw);
4018 		e1000_cleanup_led(&sc->hw);
4019 	}
4020 }
4021 
4022 /*
4023  * Disable the L0S and L1 LINK states
4024  */
4025 static void
4026 em_disable_aspm(struct e1000_softc *sc)
4027 {
4028 	int base, reg;
4029 	u16 link_cap,link_ctrl;
4030 	device_t dev = sc->dev;
4031 
4032 	switch (sc->hw.mac.type) {
4033 	case e1000_82573:
4034 	case e1000_82574:
4035 	case e1000_82583:
4036 		break;
4037 	default:
4038 		return;
4039 	}
4040 	if (pci_find_cap(dev, PCIY_EXPRESS, &base) != 0)
4041 		return;
4042 	reg = base + PCIER_LINK_CAP;
4043 	link_cap = pci_read_config(dev, reg, 2);
4044 	if ((link_cap & PCIEM_LINK_CAP_ASPM) == 0)
4045 		return;
4046 	reg = base + PCIER_LINK_CTL;
4047 	link_ctrl = pci_read_config(dev, reg, 2);
4048 	link_ctrl &= ~PCIEM_LINK_CTL_ASPMC;
4049 	pci_write_config(dev, reg, link_ctrl, 2);
4050 	return;
4051 }
4052 
4053 /**********************************************************************
4054  *
4055  *  Update the board statistics counters.
4056  *
4057  **********************************************************************/
4058 static void
4059 em_update_stats_counters(struct e1000_softc *sc)
4060 {
4061 	u64 prev_xoffrxc = sc->stats.xoffrxc;
4062 
4063 	if(sc->hw.phy.media_type == e1000_media_type_copper ||
4064 	   (E1000_READ_REG(&sc->hw, E1000_STATUS) & E1000_STATUS_LU)) {
4065 		sc->stats.symerrs += E1000_READ_REG(&sc->hw, E1000_SYMERRS);
4066 		sc->stats.sec += E1000_READ_REG(&sc->hw, E1000_SEC);
4067 	}
4068 	sc->stats.crcerrs += E1000_READ_REG(&sc->hw, E1000_CRCERRS);
4069 	sc->stats.mpc += E1000_READ_REG(&sc->hw, E1000_MPC);
4070 	sc->stats.scc += E1000_READ_REG(&sc->hw, E1000_SCC);
4071 	sc->stats.ecol += E1000_READ_REG(&sc->hw, E1000_ECOL);
4072 
4073 	sc->stats.mcc += E1000_READ_REG(&sc->hw, E1000_MCC);
4074 	sc->stats.latecol += E1000_READ_REG(&sc->hw, E1000_LATECOL);
4075 	sc->stats.colc += E1000_READ_REG(&sc->hw, E1000_COLC);
4076 	sc->stats.dc += E1000_READ_REG(&sc->hw, E1000_DC);
4077 	sc->stats.rlec += E1000_READ_REG(&sc->hw, E1000_RLEC);
4078 	sc->stats.xonrxc += E1000_READ_REG(&sc->hw, E1000_XONRXC);
4079 	sc->stats.xontxc += E1000_READ_REG(&sc->hw, E1000_XONTXC);
4080 	sc->stats.xoffrxc += E1000_READ_REG(&sc->hw, E1000_XOFFRXC);
4081 	/*
4082 	 ** For watchdog management we need to know if we have been
4083 	 ** paused during the last interval, so capture that here.
4084 	*/
4085 	if (sc->stats.xoffrxc != prev_xoffrxc)
4086 		sc->shared->isc_pause_frames = 1;
4087 	sc->stats.xofftxc += E1000_READ_REG(&sc->hw, E1000_XOFFTXC);
4088 	sc->stats.fcruc += E1000_READ_REG(&sc->hw, E1000_FCRUC);
4089 	sc->stats.prc64 += E1000_READ_REG(&sc->hw, E1000_PRC64);
4090 	sc->stats.prc127 += E1000_READ_REG(&sc->hw, E1000_PRC127);
4091 	sc->stats.prc255 += E1000_READ_REG(&sc->hw, E1000_PRC255);
4092 	sc->stats.prc511 += E1000_READ_REG(&sc->hw, E1000_PRC511);
4093 	sc->stats.prc1023 += E1000_READ_REG(&sc->hw, E1000_PRC1023);
4094 	sc->stats.prc1522 += E1000_READ_REG(&sc->hw, E1000_PRC1522);
4095 	sc->stats.gprc += E1000_READ_REG(&sc->hw, E1000_GPRC);
4096 	sc->stats.bprc += E1000_READ_REG(&sc->hw, E1000_BPRC);
4097 	sc->stats.mprc += E1000_READ_REG(&sc->hw, E1000_MPRC);
4098 	sc->stats.gptc += E1000_READ_REG(&sc->hw, E1000_GPTC);
4099 
4100 	/* For the 64-bit byte counters the low dword must be read first. */
4101 	/* Both registers clear on the read of the high dword */
4102 
4103 	sc->stats.gorc += E1000_READ_REG(&sc->hw, E1000_GORCL) +
4104 	    ((u64)E1000_READ_REG(&sc->hw, E1000_GORCH) << 32);
4105 	sc->stats.gotc += E1000_READ_REG(&sc->hw, E1000_GOTCL) +
4106 	    ((u64)E1000_READ_REG(&sc->hw, E1000_GOTCH) << 32);
4107 
4108 	sc->stats.rnbc += E1000_READ_REG(&sc->hw, E1000_RNBC);
4109 	sc->stats.ruc += E1000_READ_REG(&sc->hw, E1000_RUC);
4110 	sc->stats.rfc += E1000_READ_REG(&sc->hw, E1000_RFC);
4111 	sc->stats.roc += E1000_READ_REG(&sc->hw, E1000_ROC);
4112 	sc->stats.rjc += E1000_READ_REG(&sc->hw, E1000_RJC);
4113 
4114 	sc->stats.tor += E1000_READ_REG(&sc->hw, E1000_TORH);
4115 	sc->stats.tot += E1000_READ_REG(&sc->hw, E1000_TOTH);
4116 
4117 	sc->stats.tpr += E1000_READ_REG(&sc->hw, E1000_TPR);
4118 	sc->stats.tpt += E1000_READ_REG(&sc->hw, E1000_TPT);
4119 	sc->stats.ptc64 += E1000_READ_REG(&sc->hw, E1000_PTC64);
4120 	sc->stats.ptc127 += E1000_READ_REG(&sc->hw, E1000_PTC127);
4121 	sc->stats.ptc255 += E1000_READ_REG(&sc->hw, E1000_PTC255);
4122 	sc->stats.ptc511 += E1000_READ_REG(&sc->hw, E1000_PTC511);
4123 	sc->stats.ptc1023 += E1000_READ_REG(&sc->hw, E1000_PTC1023);
4124 	sc->stats.ptc1522 += E1000_READ_REG(&sc->hw, E1000_PTC1522);
4125 	sc->stats.mptc += E1000_READ_REG(&sc->hw, E1000_MPTC);
4126 	sc->stats.bptc += E1000_READ_REG(&sc->hw, E1000_BPTC);
4127 
4128 	/* Interrupt Counts */
4129 
4130 	sc->stats.iac += E1000_READ_REG(&sc->hw, E1000_IAC);
4131 	sc->stats.icrxptc += E1000_READ_REG(&sc->hw, E1000_ICRXPTC);
4132 	sc->stats.icrxatc += E1000_READ_REG(&sc->hw, E1000_ICRXATC);
4133 	sc->stats.ictxptc += E1000_READ_REG(&sc->hw, E1000_ICTXPTC);
4134 	sc->stats.ictxatc += E1000_READ_REG(&sc->hw, E1000_ICTXATC);
4135 	sc->stats.ictxqec += E1000_READ_REG(&sc->hw, E1000_ICTXQEC);
4136 	sc->stats.ictxqmtc += E1000_READ_REG(&sc->hw, E1000_ICTXQMTC);
4137 	sc->stats.icrxdmtc += E1000_READ_REG(&sc->hw, E1000_ICRXDMTC);
4138 	sc->stats.icrxoc += E1000_READ_REG(&sc->hw, E1000_ICRXOC);
4139 
4140 	if (sc->hw.mac.type >= e1000_82543) {
4141 		sc->stats.algnerrc +=
4142 		E1000_READ_REG(&sc->hw, E1000_ALGNERRC);
4143 		sc->stats.rxerrc +=
4144 		E1000_READ_REG(&sc->hw, E1000_RXERRC);
4145 		sc->stats.tncrs +=
4146 		E1000_READ_REG(&sc->hw, E1000_TNCRS);
4147 		sc->stats.cexterr +=
4148 		E1000_READ_REG(&sc->hw, E1000_CEXTERR);
4149 		sc->stats.tsctc +=
4150 		E1000_READ_REG(&sc->hw, E1000_TSCTC);
4151 		sc->stats.tsctfc +=
4152 		E1000_READ_REG(&sc->hw, E1000_TSCTFC);
4153 	}
4154 }
4155 
4156 static uint64_t
4157 em_if_get_counter(if_ctx_t ctx, ift_counter cnt)
4158 {
4159 	struct e1000_softc *sc = iflib_get_softc(ctx);
4160 	struct ifnet *ifp = iflib_get_ifp(ctx);
4161 
4162 	switch (cnt) {
4163 	case IFCOUNTER_COLLISIONS:
4164 		return (sc->stats.colc);
4165 	case IFCOUNTER_IERRORS:
4166 		return (sc->dropped_pkts + sc->stats.rxerrc +
4167 		    sc->stats.crcerrs + sc->stats.algnerrc +
4168 		    sc->stats.ruc + sc->stats.roc +
4169 		    sc->stats.mpc + sc->stats.cexterr);
4170 	case IFCOUNTER_OERRORS:
4171 		return (sc->stats.ecol + sc->stats.latecol +
4172 		    sc->watchdog_events);
4173 	default:
4174 		return (if_get_counter_default(ifp, cnt));
4175 	}
4176 }
4177 
4178 /* em_if_needs_restart - Tell iflib when the driver needs to be reinitialized
4179  * @ctx: iflib context
4180  * @event: event code to check
4181  *
4182  * Defaults to returning true for unknown events.
4183  *
4184  * @returns true if iflib needs to reinit the interface
4185  */
4186 static bool
4187 em_if_needs_restart(if_ctx_t ctx __unused, enum iflib_restart_event event)
4188 {
4189 	switch (event) {
4190 	case IFLIB_RESTART_VLAN_CONFIG:
4191 		return (false);
4192 	default:
4193 		return (true);
4194 	}
4195 }
4196 
4197 /* Export a single 32-bit register via a read-only sysctl. */
4198 static int
4199 em_sysctl_reg_handler(SYSCTL_HANDLER_ARGS)
4200 {
4201 	struct e1000_softc *sc;
4202 	u_int val;
4203 
4204 	sc = oidp->oid_arg1;
4205 	val = E1000_READ_REG(&sc->hw, oidp->oid_arg2);
4206 	return (sysctl_handle_int(oidp, &val, 0, req));
4207 }
4208 
4209 /*
4210  * Add sysctl variables, one per statistic, to the system.
4211  */
4212 static void
4213 em_add_hw_stats(struct e1000_softc *sc)
4214 {
4215 	device_t dev = iflib_get_dev(sc->ctx);
4216 	struct em_tx_queue *tx_que = sc->tx_queues;
4217 	struct em_rx_queue *rx_que = sc->rx_queues;
4218 
4219 	struct sysctl_ctx_list *ctx = device_get_sysctl_ctx(dev);
4220 	struct sysctl_oid *tree = device_get_sysctl_tree(dev);
4221 	struct sysctl_oid_list *child = SYSCTL_CHILDREN(tree);
4222 	struct e1000_hw_stats *stats = &sc->stats;
4223 
4224 	struct sysctl_oid *stat_node, *queue_node, *int_node;
4225 	struct sysctl_oid_list *stat_list, *queue_list, *int_list;
4226 
4227 #define QUEUE_NAME_LEN 32
4228 	char namebuf[QUEUE_NAME_LEN];
4229 
4230 	/* Driver Statistics */
4231 	SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "dropped",
4232 			CTLFLAG_RD, &sc->dropped_pkts,
4233 			"Driver dropped packets");
4234 	SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "link_irq",
4235 			CTLFLAG_RD, &sc->link_irq,
4236 			"Link MSI-X IRQ Handled");
4237 	SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "rx_overruns",
4238 			CTLFLAG_RD, &sc->rx_overruns,
4239 			"RX overruns");
4240 	SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "watchdog_timeouts",
4241 			CTLFLAG_RD, &sc->watchdog_events,
4242 			"Watchdog timeouts");
4243 	SYSCTL_ADD_PROC(ctx, child, OID_AUTO, "device_control",
4244 	    CTLTYPE_UINT | CTLFLAG_RD | CTLFLAG_NEEDGIANT,
4245 	    sc, E1000_CTRL, em_sysctl_reg_handler, "IU",
4246 	    "Device Control Register");
4247 	SYSCTL_ADD_PROC(ctx, child, OID_AUTO, "rx_control",
4248 	    CTLTYPE_UINT | CTLFLAG_RD | CTLFLAG_NEEDGIANT,
4249 	    sc, E1000_RCTL, em_sysctl_reg_handler, "IU",
4250 	    "Receiver Control Register");
4251 	SYSCTL_ADD_UINT(ctx, child, OID_AUTO, "fc_high_water",
4252 			CTLFLAG_RD, &sc->hw.fc.high_water, 0,
4253 			"Flow Control High Watermark");
4254 	SYSCTL_ADD_UINT(ctx, child, OID_AUTO, "fc_low_water",
4255 			CTLFLAG_RD, &sc->hw.fc.low_water, 0,
4256 			"Flow Control Low Watermark");
4257 
4258 	for (int i = 0; i < sc->tx_num_queues; i++, tx_que++) {
4259 		struct tx_ring *txr = &tx_que->txr;
4260 		snprintf(namebuf, QUEUE_NAME_LEN, "queue_tx_%d", i);
4261 		queue_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, namebuf,
4262 		    CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, "TX Queue Name");
4263 		queue_list = SYSCTL_CHILDREN(queue_node);
4264 
4265 		SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "txd_head",
4266 		    CTLTYPE_UINT | CTLFLAG_RD | CTLFLAG_NEEDGIANT, sc,
4267 		    E1000_TDH(txr->me), em_sysctl_reg_handler, "IU",
4268 		    "Transmit Descriptor Head");
4269 		SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "txd_tail",
4270 		    CTLTYPE_UINT | CTLFLAG_RD | CTLFLAG_NEEDGIANT, sc,
4271 		    E1000_TDT(txr->me), em_sysctl_reg_handler, "IU",
4272 		    "Transmit Descriptor Tail");
4273 		SYSCTL_ADD_ULONG(ctx, queue_list, OID_AUTO, "tx_irq",
4274 				CTLFLAG_RD, &txr->tx_irq,
4275 				"Queue MSI-X Transmit Interrupts");
4276 	}
4277 
4278 	for (int j = 0; j < sc->rx_num_queues; j++, rx_que++) {
4279 		struct rx_ring *rxr = &rx_que->rxr;
4280 		snprintf(namebuf, QUEUE_NAME_LEN, "queue_rx_%d", j);
4281 		queue_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, namebuf,
4282 		    CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, "RX Queue Name");
4283 		queue_list = SYSCTL_CHILDREN(queue_node);
4284 
4285 		SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "rxd_head",
4286 		    CTLTYPE_UINT | CTLFLAG_RD | CTLFLAG_NEEDGIANT, sc,
4287 		    E1000_RDH(rxr->me), em_sysctl_reg_handler, "IU",
4288 		    "Receive Descriptor Head");
4289 		SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "rxd_tail",
4290 		    CTLTYPE_UINT | CTLFLAG_RD | CTLFLAG_NEEDGIANT, sc,
4291 		    E1000_RDT(rxr->me), em_sysctl_reg_handler, "IU",
4292 		    "Receive Descriptor Tail");
4293 		SYSCTL_ADD_ULONG(ctx, queue_list, OID_AUTO, "rx_irq",
4294 				CTLFLAG_RD, &rxr->rx_irq,
4295 				"Queue MSI-X Receive Interrupts");
4296 	}
4297 
4298 	/* MAC stats get their own sub node */
4299 
4300 	stat_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, "mac_stats",
4301 	    CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, "Statistics");
4302 	stat_list = SYSCTL_CHILDREN(stat_node);
4303 
4304 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "excess_coll",
4305 			CTLFLAG_RD, &stats->ecol,
4306 			"Excessive collisions");
4307 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "single_coll",
4308 			CTLFLAG_RD, &stats->scc,
4309 			"Single collisions");
4310 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "multiple_coll",
4311 			CTLFLAG_RD, &stats->mcc,
4312 			"Multiple collisions");
4313 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "late_coll",
4314 			CTLFLAG_RD, &stats->latecol,
4315 			"Late collisions");
4316 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "collision_count",
4317 			CTLFLAG_RD, &stats->colc,
4318 			"Collision Count");
4319 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "symbol_errors",
4320 			CTLFLAG_RD, &sc->stats.symerrs,
4321 			"Symbol Errors");
4322 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "sequence_errors",
4323 			CTLFLAG_RD, &sc->stats.sec,
4324 			"Sequence Errors");
4325 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "defer_count",
4326 			CTLFLAG_RD, &sc->stats.dc,
4327 			"Defer Count");
4328 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "missed_packets",
4329 			CTLFLAG_RD, &sc->stats.mpc,
4330 			"Missed Packets");
4331 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_no_buff",
4332 			CTLFLAG_RD, &sc->stats.rnbc,
4333 			"Receive No Buffers");
4334 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_undersize",
4335 			CTLFLAG_RD, &sc->stats.ruc,
4336 			"Receive Undersize");
4337 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_fragmented",
4338 			CTLFLAG_RD, &sc->stats.rfc,
4339 			"Fragmented Packets Received ");
4340 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_oversize",
4341 			CTLFLAG_RD, &sc->stats.roc,
4342 			"Oversized Packets Received");
4343 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_jabber",
4344 			CTLFLAG_RD, &sc->stats.rjc,
4345 			"Recevied Jabber");
4346 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_errs",
4347 			CTLFLAG_RD, &sc->stats.rxerrc,
4348 			"Receive Errors");
4349 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "crc_errs",
4350 			CTLFLAG_RD, &sc->stats.crcerrs,
4351 			"CRC errors");
4352 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "alignment_errs",
4353 			CTLFLAG_RD, &sc->stats.algnerrc,
4354 			"Alignment Errors");
4355 	/* On 82575 these are collision counts */
4356 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "coll_ext_errs",
4357 			CTLFLAG_RD, &sc->stats.cexterr,
4358 			"Collision/Carrier extension errors");
4359 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xon_recvd",
4360 			CTLFLAG_RD, &sc->stats.xonrxc,
4361 			"XON Received");
4362 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xon_txd",
4363 			CTLFLAG_RD, &sc->stats.xontxc,
4364 			"XON Transmitted");
4365 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xoff_recvd",
4366 			CTLFLAG_RD, &sc->stats.xoffrxc,
4367 			"XOFF Received");
4368 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xoff_txd",
4369 			CTLFLAG_RD, &sc->stats.xofftxc,
4370 			"XOFF Transmitted");
4371 
4372 	/* Packet Reception Stats */
4373 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "total_pkts_recvd",
4374 			CTLFLAG_RD, &sc->stats.tpr,
4375 			"Total Packets Received ");
4376 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_pkts_recvd",
4377 			CTLFLAG_RD, &sc->stats.gprc,
4378 			"Good Packets Received");
4379 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "bcast_pkts_recvd",
4380 			CTLFLAG_RD, &sc->stats.bprc,
4381 			"Broadcast Packets Received");
4382 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "mcast_pkts_recvd",
4383 			CTLFLAG_RD, &sc->stats.mprc,
4384 			"Multicast Packets Received");
4385 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_64",
4386 			CTLFLAG_RD, &sc->stats.prc64,
4387 			"64 byte frames received ");
4388 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_65_127",
4389 			CTLFLAG_RD, &sc->stats.prc127,
4390 			"65-127 byte frames received");
4391 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_128_255",
4392 			CTLFLAG_RD, &sc->stats.prc255,
4393 			"128-255 byte frames received");
4394 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_256_511",
4395 			CTLFLAG_RD, &sc->stats.prc511,
4396 			"256-511 byte frames received");
4397 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_512_1023",
4398 			CTLFLAG_RD, &sc->stats.prc1023,
4399 			"512-1023 byte frames received");
4400 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_1024_1522",
4401 			CTLFLAG_RD, &sc->stats.prc1522,
4402 			"1023-1522 byte frames received");
4403 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_octets_recvd",
4404 			CTLFLAG_RD, &sc->stats.gorc,
4405 			"Good Octets Received");
4406 
4407 	/* Packet Transmission Stats */
4408 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_octets_txd",
4409 			CTLFLAG_RD, &sc->stats.gotc,
4410 			"Good Octets Transmitted");
4411 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "total_pkts_txd",
4412 			CTLFLAG_RD, &sc->stats.tpt,
4413 			"Total Packets Transmitted");
4414 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_pkts_txd",
4415 			CTLFLAG_RD, &sc->stats.gptc,
4416 			"Good Packets Transmitted");
4417 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "bcast_pkts_txd",
4418 			CTLFLAG_RD, &sc->stats.bptc,
4419 			"Broadcast Packets Transmitted");
4420 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "mcast_pkts_txd",
4421 			CTLFLAG_RD, &sc->stats.mptc,
4422 			"Multicast Packets Transmitted");
4423 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_64",
4424 			CTLFLAG_RD, &sc->stats.ptc64,
4425 			"64 byte frames transmitted ");
4426 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_65_127",
4427 			CTLFLAG_RD, &sc->stats.ptc127,
4428 			"65-127 byte frames transmitted");
4429 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_128_255",
4430 			CTLFLAG_RD, &sc->stats.ptc255,
4431 			"128-255 byte frames transmitted");
4432 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_256_511",
4433 			CTLFLAG_RD, &sc->stats.ptc511,
4434 			"256-511 byte frames transmitted");
4435 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_512_1023",
4436 			CTLFLAG_RD, &sc->stats.ptc1023,
4437 			"512-1023 byte frames transmitted");
4438 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_1024_1522",
4439 			CTLFLAG_RD, &sc->stats.ptc1522,
4440 			"1024-1522 byte frames transmitted");
4441 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tso_txd",
4442 			CTLFLAG_RD, &sc->stats.tsctc,
4443 			"TSO Contexts Transmitted");
4444 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tso_ctx_fail",
4445 			CTLFLAG_RD, &sc->stats.tsctfc,
4446 			"TSO Contexts Failed");
4447 
4448 
4449 	/* Interrupt Stats */
4450 
4451 	int_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, "interrupts",
4452 	    CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, "Interrupt Statistics");
4453 	int_list = SYSCTL_CHILDREN(int_node);
4454 
4455 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "asserts",
4456 			CTLFLAG_RD, &sc->stats.iac,
4457 			"Interrupt Assertion Count");
4458 
4459 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "rx_pkt_timer",
4460 			CTLFLAG_RD, &sc->stats.icrxptc,
4461 			"Interrupt Cause Rx Pkt Timer Expire Count");
4462 
4463 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "rx_abs_timer",
4464 			CTLFLAG_RD, &sc->stats.icrxatc,
4465 			"Interrupt Cause Rx Abs Timer Expire Count");
4466 
4467 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "tx_pkt_timer",
4468 			CTLFLAG_RD, &sc->stats.ictxptc,
4469 			"Interrupt Cause Tx Pkt Timer Expire Count");
4470 
4471 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "tx_abs_timer",
4472 			CTLFLAG_RD, &sc->stats.ictxatc,
4473 			"Interrupt Cause Tx Abs Timer Expire Count");
4474 
4475 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "tx_queue_empty",
4476 			CTLFLAG_RD, &sc->stats.ictxqec,
4477 			"Interrupt Cause Tx Queue Empty Count");
4478 
4479 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "tx_queue_min_thresh",
4480 			CTLFLAG_RD, &sc->stats.ictxqmtc,
4481 			"Interrupt Cause Tx Queue Min Thresh Count");
4482 
4483 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "rx_desc_min_thresh",
4484 			CTLFLAG_RD, &sc->stats.icrxdmtc,
4485 			"Interrupt Cause Rx Desc Min Thresh Count");
4486 
4487 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "rx_overrun",
4488 			CTLFLAG_RD, &sc->stats.icrxoc,
4489 			"Interrupt Cause Receiver Overrun Count");
4490 }
4491 
4492 static void
4493 em_fw_version_locked(if_ctx_t ctx)
4494 {
4495 	struct e1000_softc *sc = iflib_get_softc(ctx);
4496 	struct e1000_hw *hw = &sc->hw;
4497 	struct e1000_fw_version *fw_ver = &sc->fw_ver;
4498 	uint16_t eep = 0;
4499 
4500 	/*
4501 	 * em_fw_version_locked() must run under the IFLIB_CTX_LOCK to meet the
4502 	 * NVM locking model, so we do it in em_if_attach_pre() and store the
4503 	 * info in the softc
4504 	 */
4505 	ASSERT_CTX_LOCK_HELD(hw);
4506 
4507 	*fw_ver = (struct e1000_fw_version){0};
4508 
4509 	if (hw->mac.type >= igb_mac_min) {
4510 		/*
4511 		 * Use the Shared Code for igb(4)
4512 		 */
4513 		e1000_get_fw_version(hw, fw_ver);
4514 	} else {
4515 		/*
4516 		 * Otherwise, EEPROM version should be present on (almost?) all
4517 		 * devices here
4518 		 */
4519 		if(e1000_read_nvm(hw, NVM_VERSION, 1, &eep)) {
4520 			INIT_DEBUGOUT("can't get EEPROM version");
4521 			return;
4522 		}
4523 
4524 		fw_ver->eep_major = (eep & NVM_MAJOR_MASK) >> NVM_MAJOR_SHIFT;
4525 		fw_ver->eep_minor = (eep & NVM_MINOR_MASK) >> NVM_MINOR_SHIFT;
4526 		fw_ver->eep_build = (eep & NVM_IMAGE_ID_MASK);
4527 	}
4528 }
4529 
4530 static void
4531 em_sbuf_fw_version(struct e1000_fw_version *fw_ver, struct sbuf *buf)
4532 {
4533 	const char *space = "";
4534 
4535 	if (fw_ver->eep_major || fw_ver->eep_minor || fw_ver->eep_build) {
4536 		sbuf_printf(buf, "EEPROM V%d.%d-%d", fw_ver->eep_major,
4537 			    fw_ver->eep_minor, fw_ver->eep_build);
4538 		space = " ";
4539 	}
4540 
4541 	if (fw_ver->invm_major || fw_ver->invm_minor || fw_ver->invm_img_type) {
4542 		sbuf_printf(buf, "%sNVM V%d.%d imgtype%d",
4543 			    space, fw_ver->invm_major, fw_ver->invm_minor,
4544 			    fw_ver->invm_img_type);
4545 		space = " ";
4546 	}
4547 
4548 	if (fw_ver->or_valid) {
4549 		sbuf_printf(buf, "%sOption ROM V%d-b%d-p%d",
4550 			    space, fw_ver->or_major, fw_ver->or_build,
4551 			    fw_ver->or_patch);
4552 		space = " ";
4553 	}
4554 
4555 	if (fw_ver->etrack_id)
4556 		sbuf_printf(buf, "%seTrack 0x%08x", space, fw_ver->etrack_id);
4557 }
4558 
4559 static void
4560 em_print_fw_version(struct e1000_softc *sc )
4561 {
4562 	device_t dev = sc->dev;
4563 	struct sbuf *buf;
4564 	int error = 0;
4565 
4566 	buf = sbuf_new_auto();
4567 	if (!buf) {
4568 		device_printf(dev, "Could not allocate sbuf for output.\n");
4569 		return;
4570 	}
4571 
4572 	em_sbuf_fw_version(&sc->fw_ver, buf);
4573 
4574 	error = sbuf_finish(buf);
4575 	if (error)
4576 		device_printf(dev, "Error finishing sbuf: %d\n", error);
4577 	else if (sbuf_len(buf))
4578 		device_printf(dev, "%s\n", sbuf_data(buf));
4579 
4580 	sbuf_delete(buf);
4581 }
4582 
4583 static int
4584 em_sysctl_print_fw_version(SYSCTL_HANDLER_ARGS)
4585 {
4586 	struct e1000_softc *sc = (struct e1000_softc *)arg1;
4587 	device_t dev = sc->dev;
4588 	struct sbuf *buf;
4589 	int error = 0;
4590 
4591 	buf = sbuf_new_for_sysctl(NULL, NULL, 128, req);
4592 	if (!buf) {
4593 		device_printf(dev, "Could not allocate sbuf for output.\n");
4594 		return (ENOMEM);
4595 	}
4596 
4597 	em_sbuf_fw_version(&sc->fw_ver, buf);
4598 
4599 	error = sbuf_finish(buf);
4600 	if (error)
4601 		device_printf(dev, "Error finishing sbuf: %d\n", error);
4602 
4603 	sbuf_delete(buf);
4604 
4605 	return (0);
4606 }
4607 
4608 /**********************************************************************
4609  *
4610  *  This routine provides a way to dump out the adapter eeprom,
4611  *  often a useful debug/service tool. This only dumps the first
4612  *  32 words, stuff that matters is in that extent.
4613  *
4614  **********************************************************************/
4615 static int
4616 em_sysctl_nvm_info(SYSCTL_HANDLER_ARGS)
4617 {
4618 	struct e1000_softc *sc = (struct e1000_softc *)arg1;
4619 	int error;
4620 	int result;
4621 
4622 	result = -1;
4623 	error = sysctl_handle_int(oidp, &result, 0, req);
4624 
4625 	if (error || !req->newptr)
4626 		return (error);
4627 
4628 	/*
4629 	 * This value will cause a hex dump of the
4630 	 * first 32 16-bit words of the EEPROM to
4631 	 * the screen.
4632 	 */
4633 	if (result == 1)
4634 		em_print_nvm_info(sc);
4635 
4636 	return (error);
4637 }
4638 
4639 static void
4640 em_print_nvm_info(struct e1000_softc *sc)
4641 {
4642 	struct e1000_hw *hw = &sc->hw;
4643 	struct sx *iflib_ctx_lock = iflib_ctx_lock_get(sc->ctx);
4644 	u16 eeprom_data;
4645 	int i, j, row = 0;
4646 
4647 	/* Its a bit crude, but it gets the job done */
4648 	printf("\nInterface EEPROM Dump:\n");
4649 	printf("Offset\n0x0000  ");
4650 
4651 	/* We rely on the IFLIB_CTX_LOCK as part of NVM locking model */
4652 	sx_xlock(iflib_ctx_lock);
4653 	ASSERT_CTX_LOCK_HELD(hw);
4654 	for (i = 0, j = 0; i < 32; i++, j++) {
4655 		if (j == 8) { /* Make the offset block */
4656 			j = 0; ++row;
4657 			printf("\n0x00%x0  ",row);
4658 		}
4659 		e1000_read_nvm(hw, i, 1, &eeprom_data);
4660 		printf("%04x ", eeprom_data);
4661 	}
4662 	sx_xunlock(iflib_ctx_lock);
4663 	printf("\n");
4664 }
4665 
4666 static int
4667 em_sysctl_int_delay(SYSCTL_HANDLER_ARGS)
4668 {
4669 	struct em_int_delay_info *info;
4670 	struct e1000_softc *sc;
4671 	u32 regval;
4672 	int error, usecs, ticks;
4673 
4674 	info = (struct em_int_delay_info *) arg1;
4675 	usecs = info->value;
4676 	error = sysctl_handle_int(oidp, &usecs, 0, req);
4677 	if (error != 0 || req->newptr == NULL)
4678 		return (error);
4679 	if (usecs < 0 || usecs > EM_TICKS_TO_USECS(65535))
4680 		return (EINVAL);
4681 	info->value = usecs;
4682 	ticks = EM_USECS_TO_TICKS(usecs);
4683 	if (info->offset == E1000_ITR)	/* units are 256ns here */
4684 		ticks *= 4;
4685 
4686 	sc = info->sc;
4687 
4688 	regval = E1000_READ_OFFSET(&sc->hw, info->offset);
4689 	regval = (regval & ~0xffff) | (ticks & 0xffff);
4690 	/* Handle a few special cases. */
4691 	switch (info->offset) {
4692 	case E1000_RDTR:
4693 		break;
4694 	case E1000_TIDV:
4695 		if (ticks == 0) {
4696 			sc->txd_cmd &= ~E1000_TXD_CMD_IDE;
4697 			/* Don't write 0 into the TIDV register. */
4698 			regval++;
4699 		} else
4700 			sc->txd_cmd |= E1000_TXD_CMD_IDE;
4701 		break;
4702 	}
4703 	E1000_WRITE_OFFSET(&sc->hw, info->offset, regval);
4704 	return (0);
4705 }
4706 
4707 static void
4708 em_add_int_delay_sysctl(struct e1000_softc *sc, const char *name,
4709 	const char *description, struct em_int_delay_info *info,
4710 	int offset, int value)
4711 {
4712 	info->sc = sc;
4713 	info->offset = offset;
4714 	info->value = value;
4715 	SYSCTL_ADD_PROC(device_get_sysctl_ctx(sc->dev),
4716 	    SYSCTL_CHILDREN(device_get_sysctl_tree(sc->dev)),
4717 	    OID_AUTO, name, CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_NEEDGIANT,
4718 	    info, 0, em_sysctl_int_delay, "I", description);
4719 }
4720 
4721 /*
4722  * Set flow control using sysctl:
4723  * Flow control values:
4724  *      0 - off
4725  *      1 - rx pause
4726  *      2 - tx pause
4727  *      3 - full
4728  */
4729 static int
4730 em_set_flowcntl(SYSCTL_HANDLER_ARGS)
4731 {
4732 	int error;
4733 	static int input = 3; /* default is full */
4734 	struct e1000_softc	*sc = (struct e1000_softc *) arg1;
4735 
4736 	error = sysctl_handle_int(oidp, &input, 0, req);
4737 
4738 	if ((error) || (req->newptr == NULL))
4739 		return (error);
4740 
4741 	if (input == sc->fc) /* no change? */
4742 		return (error);
4743 
4744 	switch (input) {
4745 	case e1000_fc_rx_pause:
4746 	case e1000_fc_tx_pause:
4747 	case e1000_fc_full:
4748 	case e1000_fc_none:
4749 		sc->hw.fc.requested_mode = input;
4750 		sc->fc = input;
4751 		break;
4752 	default:
4753 		/* Do nothing */
4754 		return (error);
4755 	}
4756 
4757 	sc->hw.fc.current_mode = sc->hw.fc.requested_mode;
4758 	e1000_force_mac_fc(&sc->hw);
4759 	return (error);
4760 }
4761 
4762 /*
4763  * Manage Energy Efficient Ethernet:
4764  * Control values:
4765  *     0/1 - enabled/disabled
4766  */
4767 static int
4768 em_sysctl_eee(SYSCTL_HANDLER_ARGS)
4769 {
4770 	struct e1000_softc *sc = (struct e1000_softc *) arg1;
4771 	int error, value;
4772 
4773 	value = sc->hw.dev_spec.ich8lan.eee_disable;
4774 	error = sysctl_handle_int(oidp, &value, 0, req);
4775 	if (error || req->newptr == NULL)
4776 		return (error);
4777 	sc->hw.dev_spec.ich8lan.eee_disable = (value != 0);
4778 	em_if_init(sc->ctx);
4779 
4780 	return (0);
4781 }
4782 
4783 static int
4784 em_sysctl_debug_info(SYSCTL_HANDLER_ARGS)
4785 {
4786 	struct e1000_softc *sc;
4787 	int error;
4788 	int result;
4789 
4790 	result = -1;
4791 	error = sysctl_handle_int(oidp, &result, 0, req);
4792 
4793 	if (error || !req->newptr)
4794 		return (error);
4795 
4796 	if (result == 1) {
4797 		sc = (struct e1000_softc *) arg1;
4798 		em_print_debug_info(sc);
4799 	}
4800 
4801 	return (error);
4802 }
4803 
4804 static int
4805 em_get_rs(SYSCTL_HANDLER_ARGS)
4806 {
4807 	struct e1000_softc *sc = (struct e1000_softc *) arg1;
4808 	int error;
4809 	int result;
4810 
4811 	result = 0;
4812 	error = sysctl_handle_int(oidp, &result, 0, req);
4813 
4814 	if (error || !req->newptr || result != 1)
4815 		return (error);
4816 	em_dump_rs(sc);
4817 
4818 	return (error);
4819 }
4820 
4821 static void
4822 em_if_debug(if_ctx_t ctx)
4823 {
4824 	em_dump_rs(iflib_get_softc(ctx));
4825 }
4826 
4827 /*
4828  * This routine is meant to be fluid, add whatever is
4829  * needed for debugging a problem.  -jfv
4830  */
4831 static void
4832 em_print_debug_info(struct e1000_softc *sc)
4833 {
4834 	device_t dev = iflib_get_dev(sc->ctx);
4835 	struct ifnet *ifp = iflib_get_ifp(sc->ctx);
4836 	struct tx_ring *txr = &sc->tx_queues->txr;
4837 	struct rx_ring *rxr = &sc->rx_queues->rxr;
4838 
4839 	if (if_getdrvflags(ifp) & IFF_DRV_RUNNING)
4840 		printf("Interface is RUNNING ");
4841 	else
4842 		printf("Interface is NOT RUNNING\n");
4843 
4844 	if (if_getdrvflags(ifp) & IFF_DRV_OACTIVE)
4845 		printf("and INACTIVE\n");
4846 	else
4847 		printf("and ACTIVE\n");
4848 
4849 	for (int i = 0; i < sc->tx_num_queues; i++, txr++) {
4850 		device_printf(dev, "TX Queue %d ------\n", i);
4851 		device_printf(dev, "hw tdh = %d, hw tdt = %d\n",
4852 			E1000_READ_REG(&sc->hw, E1000_TDH(i)),
4853 			E1000_READ_REG(&sc->hw, E1000_TDT(i)));
4854 
4855 	}
4856 	for (int j=0; j < sc->rx_num_queues; j++, rxr++) {
4857 		device_printf(dev, "RX Queue %d ------\n", j);
4858 		device_printf(dev, "hw rdh = %d, hw rdt = %d\n",
4859 			E1000_READ_REG(&sc->hw, E1000_RDH(j)),
4860 			E1000_READ_REG(&sc->hw, E1000_RDT(j)));
4861 	}
4862 }
4863 
4864 /*
4865  * 82574 only:
4866  * Write a new value to the EEPROM increasing the number of MSI-X
4867  * vectors from 3 to 5, for proper multiqueue support.
4868  */
4869 static void
4870 em_enable_vectors_82574(if_ctx_t ctx)
4871 {
4872 	struct e1000_softc *sc = iflib_get_softc(ctx);
4873 	struct e1000_hw *hw = &sc->hw;
4874 	device_t dev = iflib_get_dev(ctx);
4875 	u16 edata;
4876 
4877 	e1000_read_nvm(hw, EM_NVM_PCIE_CTRL, 1, &edata);
4878 	if (bootverbose)
4879 		device_printf(dev, "EM_NVM_PCIE_CTRL = %#06x\n", edata);
4880 	if (((edata & EM_NVM_MSIX_N_MASK) >> EM_NVM_MSIX_N_SHIFT) != 4) {
4881 		device_printf(dev, "Writing to eeprom: increasing "
4882 		    "reported MSI-X vectors from 3 to 5...\n");
4883 		edata &= ~(EM_NVM_MSIX_N_MASK);
4884 		edata |= 4 << EM_NVM_MSIX_N_SHIFT;
4885 		e1000_write_nvm(hw, EM_NVM_PCIE_CTRL, 1, &edata);
4886 		e1000_update_nvm_checksum(hw);
4887 		device_printf(dev, "Writing to eeprom: done\n");
4888 	}
4889 }
4890