xref: /freebsd/sys/dev/e1000/if_em.c (revision 38a52bd3b5cac3da6f7f6eef3dd050e6aa08ebb3)
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 DRIVER_MODULE(em, pci, em_driver, 0, 0);
381 
382 MODULE_DEPEND(em, pci, 1, 1, 1);
383 MODULE_DEPEND(em, ether, 1, 1, 1);
384 MODULE_DEPEND(em, iflib, 1, 1, 1);
385 
386 IFLIB_PNP_INFO(pci, em, em_vendor_info_array);
387 
388 static driver_t igb_driver = {
389 	"igb", igb_methods, sizeof(struct e1000_softc),
390 };
391 
392 DRIVER_MODULE(igb, pci, igb_driver, 0, 0);
393 
394 MODULE_DEPEND(igb, pci, 1, 1, 1);
395 MODULE_DEPEND(igb, ether, 1, 1, 1);
396 MODULE_DEPEND(igb, iflib, 1, 1, 1);
397 
398 IFLIB_PNP_INFO(pci, igb, igb_vendor_info_array);
399 
400 static device_method_t em_if_methods[] = {
401 	DEVMETHOD(ifdi_attach_pre, em_if_attach_pre),
402 	DEVMETHOD(ifdi_attach_post, em_if_attach_post),
403 	DEVMETHOD(ifdi_detach, em_if_detach),
404 	DEVMETHOD(ifdi_shutdown, em_if_shutdown),
405 	DEVMETHOD(ifdi_suspend, em_if_suspend),
406 	DEVMETHOD(ifdi_resume, em_if_resume),
407 	DEVMETHOD(ifdi_init, em_if_init),
408 	DEVMETHOD(ifdi_stop, em_if_stop),
409 	DEVMETHOD(ifdi_msix_intr_assign, em_if_msix_intr_assign),
410 	DEVMETHOD(ifdi_intr_enable, em_if_intr_enable),
411 	DEVMETHOD(ifdi_intr_disable, em_if_intr_disable),
412 	DEVMETHOD(ifdi_tx_queues_alloc, em_if_tx_queues_alloc),
413 	DEVMETHOD(ifdi_rx_queues_alloc, em_if_rx_queues_alloc),
414 	DEVMETHOD(ifdi_queues_free, em_if_queues_free),
415 	DEVMETHOD(ifdi_update_admin_status, em_if_update_admin_status),
416 	DEVMETHOD(ifdi_multi_set, em_if_multi_set),
417 	DEVMETHOD(ifdi_media_status, em_if_media_status),
418 	DEVMETHOD(ifdi_media_change, em_if_media_change),
419 	DEVMETHOD(ifdi_mtu_set, em_if_mtu_set),
420 	DEVMETHOD(ifdi_promisc_set, em_if_set_promisc),
421 	DEVMETHOD(ifdi_timer, em_if_timer),
422 	DEVMETHOD(ifdi_watchdog_reset, em_if_watchdog_reset),
423 	DEVMETHOD(ifdi_vlan_register, em_if_vlan_register),
424 	DEVMETHOD(ifdi_vlan_unregister, em_if_vlan_unregister),
425 	DEVMETHOD(ifdi_get_counter, em_if_get_counter),
426 	DEVMETHOD(ifdi_led_func, em_if_led_func),
427 	DEVMETHOD(ifdi_rx_queue_intr_enable, em_if_rx_queue_intr_enable),
428 	DEVMETHOD(ifdi_tx_queue_intr_enable, em_if_tx_queue_intr_enable),
429 	DEVMETHOD(ifdi_debug, em_if_debug),
430 	DEVMETHOD(ifdi_needs_restart, em_if_needs_restart),
431 	DEVMETHOD_END
432 };
433 
434 static driver_t em_if_driver = {
435 	"em_if", em_if_methods, sizeof(struct e1000_softc)
436 };
437 
438 static device_method_t igb_if_methods[] = {
439 	DEVMETHOD(ifdi_attach_pre, em_if_attach_pre),
440 	DEVMETHOD(ifdi_attach_post, em_if_attach_post),
441 	DEVMETHOD(ifdi_detach, em_if_detach),
442 	DEVMETHOD(ifdi_shutdown, em_if_shutdown),
443 	DEVMETHOD(ifdi_suspend, em_if_suspend),
444 	DEVMETHOD(ifdi_resume, em_if_resume),
445 	DEVMETHOD(ifdi_init, em_if_init),
446 	DEVMETHOD(ifdi_stop, em_if_stop),
447 	DEVMETHOD(ifdi_msix_intr_assign, em_if_msix_intr_assign),
448 	DEVMETHOD(ifdi_intr_enable, igb_if_intr_enable),
449 	DEVMETHOD(ifdi_intr_disable, igb_if_intr_disable),
450 	DEVMETHOD(ifdi_tx_queues_alloc, em_if_tx_queues_alloc),
451 	DEVMETHOD(ifdi_rx_queues_alloc, em_if_rx_queues_alloc),
452 	DEVMETHOD(ifdi_queues_free, em_if_queues_free),
453 	DEVMETHOD(ifdi_update_admin_status, em_if_update_admin_status),
454 	DEVMETHOD(ifdi_multi_set, em_if_multi_set),
455 	DEVMETHOD(ifdi_media_status, em_if_media_status),
456 	DEVMETHOD(ifdi_media_change, em_if_media_change),
457 	DEVMETHOD(ifdi_mtu_set, em_if_mtu_set),
458 	DEVMETHOD(ifdi_promisc_set, em_if_set_promisc),
459 	DEVMETHOD(ifdi_timer, em_if_timer),
460 	DEVMETHOD(ifdi_watchdog_reset, em_if_watchdog_reset),
461 	DEVMETHOD(ifdi_vlan_register, em_if_vlan_register),
462 	DEVMETHOD(ifdi_vlan_unregister, em_if_vlan_unregister),
463 	DEVMETHOD(ifdi_get_counter, em_if_get_counter),
464 	DEVMETHOD(ifdi_led_func, em_if_led_func),
465 	DEVMETHOD(ifdi_rx_queue_intr_enable, igb_if_rx_queue_intr_enable),
466 	DEVMETHOD(ifdi_tx_queue_intr_enable, igb_if_tx_queue_intr_enable),
467 	DEVMETHOD(ifdi_debug, em_if_debug),
468 	DEVMETHOD(ifdi_needs_restart, em_if_needs_restart),
469 	DEVMETHOD_END
470 };
471 
472 static driver_t igb_if_driver = {
473 	"igb_if", igb_if_methods, sizeof(struct e1000_softc)
474 };
475 
476 /*********************************************************************
477  *  Tunable default values.
478  *********************************************************************/
479 
480 #define EM_TICKS_TO_USECS(ticks)	((1024 * (ticks) + 500) / 1000)
481 #define EM_USECS_TO_TICKS(usecs)	((1000 * (usecs) + 512) / 1024)
482 
483 #define MAX_INTS_PER_SEC	8000
484 #define DEFAULT_ITR		(1000000000/(MAX_INTS_PER_SEC * 256))
485 
486 /* Allow common code without TSO */
487 #ifndef CSUM_TSO
488 #define CSUM_TSO	0
489 #endif
490 
491 static SYSCTL_NODE(_hw, OID_AUTO, em, CTLFLAG_RD | CTLFLAG_MPSAFE, 0,
492     "EM driver parameters");
493 
494 static int em_disable_crc_stripping = 0;
495 SYSCTL_INT(_hw_em, OID_AUTO, disable_crc_stripping, CTLFLAG_RDTUN,
496     &em_disable_crc_stripping, 0, "Disable CRC Stripping");
497 
498 static int em_tx_int_delay_dflt = EM_TICKS_TO_USECS(EM_TIDV);
499 static int em_rx_int_delay_dflt = EM_TICKS_TO_USECS(EM_RDTR);
500 SYSCTL_INT(_hw_em, OID_AUTO, tx_int_delay, CTLFLAG_RDTUN, &em_tx_int_delay_dflt,
501     0, "Default transmit interrupt delay in usecs");
502 SYSCTL_INT(_hw_em, OID_AUTO, rx_int_delay, CTLFLAG_RDTUN, &em_rx_int_delay_dflt,
503     0, "Default receive interrupt delay in usecs");
504 
505 static int em_tx_abs_int_delay_dflt = EM_TICKS_TO_USECS(EM_TADV);
506 static int em_rx_abs_int_delay_dflt = EM_TICKS_TO_USECS(EM_RADV);
507 SYSCTL_INT(_hw_em, OID_AUTO, tx_abs_int_delay, CTLFLAG_RDTUN,
508     &em_tx_abs_int_delay_dflt, 0,
509     "Default transmit interrupt delay limit in usecs");
510 SYSCTL_INT(_hw_em, OID_AUTO, rx_abs_int_delay, CTLFLAG_RDTUN,
511     &em_rx_abs_int_delay_dflt, 0,
512     "Default receive interrupt delay limit in usecs");
513 
514 static int em_smart_pwr_down = false;
515 SYSCTL_INT(_hw_em, OID_AUTO, smart_pwr_down, CTLFLAG_RDTUN, &em_smart_pwr_down,
516     0, "Set to true to leave smart power down enabled on newer adapters");
517 
518 /* Controls whether promiscuous also shows bad packets */
519 static int em_debug_sbp = false;
520 SYSCTL_INT(_hw_em, OID_AUTO, sbp, CTLFLAG_RDTUN, &em_debug_sbp, 0,
521     "Show bad packets in promiscuous mode");
522 
523 /* How many packets rxeof tries to clean at a time */
524 static int em_rx_process_limit = 100;
525 SYSCTL_INT(_hw_em, OID_AUTO, rx_process_limit, CTLFLAG_RDTUN,
526     &em_rx_process_limit, 0,
527     "Maximum number of received packets to process "
528     "at a time, -1 means unlimited");
529 
530 /* Energy efficient ethernet - default to OFF */
531 static int eee_setting = 1;
532 SYSCTL_INT(_hw_em, OID_AUTO, eee_setting, CTLFLAG_RDTUN, &eee_setting, 0,
533     "Enable Energy Efficient Ethernet");
534 
535 /*
536 ** Tuneable Interrupt rate
537 */
538 static int em_max_interrupt_rate = 8000;
539 SYSCTL_INT(_hw_em, OID_AUTO, max_interrupt_rate, CTLFLAG_RDTUN,
540     &em_max_interrupt_rate, 0, "Maximum interrupts per second");
541 
542 
543 
544 /* Global used in WOL setup with multiport cards */
545 static int global_quad_port_a = 0;
546 
547 extern struct if_txrx igb_txrx;
548 extern struct if_txrx em_txrx;
549 extern struct if_txrx lem_txrx;
550 
551 static struct if_shared_ctx em_sctx_init = {
552 	.isc_magic = IFLIB_MAGIC,
553 	.isc_q_align = PAGE_SIZE,
554 	.isc_tx_maxsize = EM_TSO_SIZE + sizeof(struct ether_vlan_header),
555 	.isc_tx_maxsegsize = PAGE_SIZE,
556 	.isc_tso_maxsize = EM_TSO_SIZE + sizeof(struct ether_vlan_header),
557 	.isc_tso_maxsegsize = EM_TSO_SEG_SIZE,
558 	.isc_rx_maxsize = MJUM9BYTES,
559 	.isc_rx_nsegments = 1,
560 	.isc_rx_maxsegsize = MJUM9BYTES,
561 	.isc_nfl = 1,
562 	.isc_nrxqs = 1,
563 	.isc_ntxqs = 1,
564 	.isc_admin_intrcnt = 1,
565 	.isc_vendor_info = em_vendor_info_array,
566 	.isc_driver_version = em_driver_version,
567 	.isc_driver = &em_if_driver,
568 	.isc_flags = IFLIB_NEED_SCRATCH | IFLIB_TSO_INIT_IP | IFLIB_NEED_ZERO_CSUM,
569 
570 	.isc_nrxd_min = {EM_MIN_RXD},
571 	.isc_ntxd_min = {EM_MIN_TXD},
572 	.isc_nrxd_max = {EM_MAX_RXD},
573 	.isc_ntxd_max = {EM_MAX_TXD},
574 	.isc_nrxd_default = {EM_DEFAULT_RXD},
575 	.isc_ntxd_default = {EM_DEFAULT_TXD},
576 };
577 
578 static struct if_shared_ctx igb_sctx_init = {
579 	.isc_magic = IFLIB_MAGIC,
580 	.isc_q_align = PAGE_SIZE,
581 	.isc_tx_maxsize = EM_TSO_SIZE + sizeof(struct ether_vlan_header),
582 	.isc_tx_maxsegsize = PAGE_SIZE,
583 	.isc_tso_maxsize = EM_TSO_SIZE + sizeof(struct ether_vlan_header),
584 	.isc_tso_maxsegsize = EM_TSO_SEG_SIZE,
585 	.isc_rx_maxsize = MJUM9BYTES,
586 	.isc_rx_nsegments = 1,
587 	.isc_rx_maxsegsize = MJUM9BYTES,
588 	.isc_nfl = 1,
589 	.isc_nrxqs = 1,
590 	.isc_ntxqs = 1,
591 	.isc_admin_intrcnt = 1,
592 	.isc_vendor_info = igb_vendor_info_array,
593 	.isc_driver_version = em_driver_version,
594 	.isc_driver = &igb_if_driver,
595 	.isc_flags = IFLIB_NEED_SCRATCH | IFLIB_TSO_INIT_IP | IFLIB_NEED_ZERO_CSUM,
596 
597 	.isc_nrxd_min = {EM_MIN_RXD},
598 	.isc_ntxd_min = {EM_MIN_TXD},
599 	.isc_nrxd_max = {IGB_MAX_RXD},
600 	.isc_ntxd_max = {IGB_MAX_TXD},
601 	.isc_nrxd_default = {EM_DEFAULT_RXD},
602 	.isc_ntxd_default = {EM_DEFAULT_TXD},
603 };
604 
605 /*****************************************************************
606  *
607  * Dump Registers
608  *
609  ****************************************************************/
610 #define IGB_REGS_LEN 739
611 
612 static int em_get_regs(SYSCTL_HANDLER_ARGS)
613 {
614 	struct e1000_softc *sc = (struct e1000_softc *)arg1;
615 	struct e1000_hw *hw = &sc->hw;
616 	struct sbuf *sb;
617 	u32 *regs_buff;
618 	int rc;
619 
620 	regs_buff = malloc(sizeof(u32) * IGB_REGS_LEN, M_DEVBUF, M_WAITOK);
621 	memset(regs_buff, 0, IGB_REGS_LEN * sizeof(u32));
622 
623 	rc = sysctl_wire_old_buffer(req, 0);
624 	MPASS(rc == 0);
625 	if (rc != 0) {
626 		free(regs_buff, M_DEVBUF);
627 		return (rc);
628 	}
629 
630 	sb = sbuf_new_for_sysctl(NULL, NULL, 32*400, req);
631 	MPASS(sb != NULL);
632 	if (sb == NULL) {
633 		free(regs_buff, M_DEVBUF);
634 		return (ENOMEM);
635 	}
636 
637 	/* General Registers */
638 	regs_buff[0] = E1000_READ_REG(hw, E1000_CTRL);
639 	regs_buff[1] = E1000_READ_REG(hw, E1000_STATUS);
640 	regs_buff[2] = E1000_READ_REG(hw, E1000_CTRL_EXT);
641 	regs_buff[3] = E1000_READ_REG(hw, E1000_ICR);
642 	regs_buff[4] = E1000_READ_REG(hw, E1000_RCTL);
643 	regs_buff[5] = E1000_READ_REG(hw, E1000_RDLEN(0));
644 	regs_buff[6] = E1000_READ_REG(hw, E1000_RDH(0));
645 	regs_buff[7] = E1000_READ_REG(hw, E1000_RDT(0));
646 	regs_buff[8] = E1000_READ_REG(hw, E1000_RXDCTL(0));
647 	regs_buff[9] = E1000_READ_REG(hw, E1000_RDBAL(0));
648 	regs_buff[10] = E1000_READ_REG(hw, E1000_RDBAH(0));
649 	regs_buff[11] = E1000_READ_REG(hw, E1000_TCTL);
650 	regs_buff[12] = E1000_READ_REG(hw, E1000_TDBAL(0));
651 	regs_buff[13] = E1000_READ_REG(hw, E1000_TDBAH(0));
652 	regs_buff[14] = E1000_READ_REG(hw, E1000_TDLEN(0));
653 	regs_buff[15] = E1000_READ_REG(hw, E1000_TDH(0));
654 	regs_buff[16] = E1000_READ_REG(hw, E1000_TDT(0));
655 	regs_buff[17] = E1000_READ_REG(hw, E1000_TXDCTL(0));
656 	regs_buff[18] = E1000_READ_REG(hw, E1000_TDFH);
657 	regs_buff[19] = E1000_READ_REG(hw, E1000_TDFT);
658 	regs_buff[20] = E1000_READ_REG(hw, E1000_TDFHS);
659 	regs_buff[21] = E1000_READ_REG(hw, E1000_TDFPC);
660 
661 	sbuf_printf(sb, "General Registers\n");
662 	sbuf_printf(sb, "\tCTRL\t %08x\n", regs_buff[0]);
663 	sbuf_printf(sb, "\tSTATUS\t %08x\n", regs_buff[1]);
664 	sbuf_printf(sb, "\tCTRL_EXT\t %08x\n\n", regs_buff[2]);
665 
666 	sbuf_printf(sb, "Interrupt Registers\n");
667 	sbuf_printf(sb, "\tICR\t %08x\n\n", regs_buff[3]);
668 
669 	sbuf_printf(sb, "RX Registers\n");
670 	sbuf_printf(sb, "\tRCTL\t %08x\n", regs_buff[4]);
671 	sbuf_printf(sb, "\tRDLEN\t %08x\n", regs_buff[5]);
672 	sbuf_printf(sb, "\tRDH\t %08x\n", regs_buff[6]);
673 	sbuf_printf(sb, "\tRDT\t %08x\n", regs_buff[7]);
674 	sbuf_printf(sb, "\tRXDCTL\t %08x\n", regs_buff[8]);
675 	sbuf_printf(sb, "\tRDBAL\t %08x\n", regs_buff[9]);
676 	sbuf_printf(sb, "\tRDBAH\t %08x\n\n", regs_buff[10]);
677 
678 	sbuf_printf(sb, "TX Registers\n");
679 	sbuf_printf(sb, "\tTCTL\t %08x\n", regs_buff[11]);
680 	sbuf_printf(sb, "\tTDBAL\t %08x\n", regs_buff[12]);
681 	sbuf_printf(sb, "\tTDBAH\t %08x\n", regs_buff[13]);
682 	sbuf_printf(sb, "\tTDLEN\t %08x\n", regs_buff[14]);
683 	sbuf_printf(sb, "\tTDH\t %08x\n", regs_buff[15]);
684 	sbuf_printf(sb, "\tTDT\t %08x\n", regs_buff[16]);
685 	sbuf_printf(sb, "\tTXDCTL\t %08x\n", regs_buff[17]);
686 	sbuf_printf(sb, "\tTDFH\t %08x\n", regs_buff[18]);
687 	sbuf_printf(sb, "\tTDFT\t %08x\n", regs_buff[19]);
688 	sbuf_printf(sb, "\tTDFHS\t %08x\n", regs_buff[20]);
689 	sbuf_printf(sb, "\tTDFPC\t %08x\n\n", regs_buff[21]);
690 
691 	free(regs_buff, M_DEVBUF);
692 
693 #ifdef DUMP_DESCS
694 	{
695 		if_softc_ctx_t scctx = sc->shared;
696 		struct rx_ring *rxr = &rx_que->rxr;
697 		struct tx_ring *txr = &tx_que->txr;
698 		int ntxd = scctx->isc_ntxd[0];
699 		int nrxd = scctx->isc_nrxd[0];
700 		int j;
701 
702 	for (j = 0; j < nrxd; j++) {
703 		u32 staterr = le32toh(rxr->rx_base[j].wb.upper.status_error);
704 		u32 length =  le32toh(rxr->rx_base[j].wb.upper.length);
705 		sbuf_printf(sb, "\tReceive Descriptor Address %d: %08" PRIx64 "  Error:%d  Length:%d\n", j, rxr->rx_base[j].read.buffer_addr, staterr, length);
706 	}
707 
708 	for (j = 0; j < min(ntxd, 256); j++) {
709 		unsigned int *ptr = (unsigned int *)&txr->tx_base[j];
710 
711 		sbuf_printf(sb, "\tTXD[%03d] [0]: %08x [1]: %08x [2]: %08x [3]: %08x  eop: %d DD=%d\n",
712 			    j, ptr[0], ptr[1], ptr[2], ptr[3], buf->eop,
713 			    buf->eop != -1 ? txr->tx_base[buf->eop].upper.fields.status & E1000_TXD_STAT_DD : 0);
714 
715 	}
716 	}
717 #endif
718 
719 	rc = sbuf_finish(sb);
720 	sbuf_delete(sb);
721 	return(rc);
722 }
723 
724 static void *
725 em_register(device_t dev)
726 {
727 	return (&em_sctx_init);
728 }
729 
730 static void *
731 igb_register(device_t dev)
732 {
733 	return (&igb_sctx_init);
734 }
735 
736 static int
737 em_set_num_queues(if_ctx_t ctx)
738 {
739 	struct e1000_softc *sc = iflib_get_softc(ctx);
740 	int maxqueues;
741 
742 	/* Sanity check based on HW */
743 	switch (sc->hw.mac.type) {
744 	case e1000_82576:
745 	case e1000_82580:
746 	case e1000_i350:
747 	case e1000_i354:
748 		maxqueues = 8;
749 		break;
750 	case e1000_i210:
751 	case e1000_82575:
752 		maxqueues = 4;
753 		break;
754 	case e1000_i211:
755 	case e1000_82574:
756 		maxqueues = 2;
757 		break;
758 	default:
759 		maxqueues = 1;
760 		break;
761 	}
762 
763 	return (maxqueues);
764 }
765 
766 #define	LEM_CAPS							\
767     IFCAP_HWCSUM | IFCAP_VLAN_MTU | IFCAP_VLAN_HWTAGGING |		\
768     IFCAP_VLAN_HWCSUM | IFCAP_WOL | IFCAP_VLAN_HWFILTER
769 
770 #define	EM_CAPS								\
771     IFCAP_HWCSUM | IFCAP_VLAN_MTU | IFCAP_VLAN_HWTAGGING |		\
772     IFCAP_VLAN_HWCSUM | IFCAP_WOL | IFCAP_VLAN_HWFILTER | IFCAP_TSO4 |	\
773     IFCAP_LRO | IFCAP_VLAN_HWTSO
774 
775 #define	IGB_CAPS							\
776     IFCAP_HWCSUM | IFCAP_VLAN_MTU | IFCAP_VLAN_HWTAGGING |		\
777     IFCAP_VLAN_HWCSUM | IFCAP_WOL | IFCAP_VLAN_HWFILTER | IFCAP_TSO4 |	\
778     IFCAP_LRO | IFCAP_VLAN_HWTSO | IFCAP_JUMBO_MTU | IFCAP_HWCSUM_IPV6 |\
779     IFCAP_TSO6
780 
781 /*********************************************************************
782  *  Device initialization routine
783  *
784  *  The attach entry point is called when the driver is being loaded.
785  *  This routine identifies the type of hardware, allocates all resources
786  *  and initializes the hardware.
787  *
788  *  return 0 on success, positive on failure
789  *********************************************************************/
790 static int
791 em_if_attach_pre(if_ctx_t ctx)
792 {
793 	struct e1000_softc *sc;
794 	if_softc_ctx_t scctx;
795 	device_t dev;
796 	struct e1000_hw *hw;
797 	struct sysctl_oid_list *child;
798 	struct sysctl_ctx_list *ctx_list;
799 	int error = 0;
800 
801 	INIT_DEBUGOUT("em_if_attach_pre: begin");
802 	dev = iflib_get_dev(ctx);
803 	sc = iflib_get_softc(ctx);
804 
805 	sc->ctx = sc->osdep.ctx = ctx;
806 	sc->dev = sc->osdep.dev = dev;
807 	scctx = sc->shared = iflib_get_softc_ctx(ctx);
808 	sc->media = iflib_get_media(ctx);
809 	hw = &sc->hw;
810 
811 	sc->tx_process_limit = scctx->isc_ntxd[0];
812 
813 	/* Determine hardware and mac info */
814 	em_identify_hardware(ctx);
815 
816 	/* SYSCTL stuff */
817 	ctx_list = device_get_sysctl_ctx(dev);
818 	child = SYSCTL_CHILDREN(device_get_sysctl_tree(dev));
819 
820 	SYSCTL_ADD_PROC(ctx_list, child, OID_AUTO, "nvm",
821 	    CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_NEEDGIANT, sc, 0,
822 	    em_sysctl_nvm_info, "I", "NVM Information");
823 
824 	SYSCTL_ADD_PROC(ctx_list, child, OID_AUTO, "fw_version",
825 	    CTLTYPE_STRING | CTLFLAG_RD | CTLFLAG_NEEDGIANT, sc, 0,
826 	    em_sysctl_print_fw_version, "A",
827 	    "Prints FW/NVM Versions");
828 
829 	SYSCTL_ADD_PROC(ctx_list, child, OID_AUTO, "debug",
830 	    CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_NEEDGIANT, sc, 0,
831 	    em_sysctl_debug_info, "I", "Debug Information");
832 
833 	SYSCTL_ADD_PROC(ctx_list, child, OID_AUTO, "fc",
834 	    CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_NEEDGIANT, sc, 0,
835 	    em_set_flowcntl, "I", "Flow Control");
836 
837 	SYSCTL_ADD_PROC(ctx_list, child, OID_AUTO, "reg_dump",
838 	    CTLTYPE_STRING | CTLFLAG_RD | CTLFLAG_NEEDGIANT, sc, 0,
839 	    em_get_regs, "A", "Dump Registers");
840 
841 	SYSCTL_ADD_PROC(ctx_list, child, OID_AUTO, "rs_dump",
842 	    CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_NEEDGIANT, sc, 0,
843 	    em_get_rs, "I", "Dump RS indexes");
844 
845 	scctx->isc_tx_nsegments = EM_MAX_SCATTER;
846 	scctx->isc_nrxqsets_max = scctx->isc_ntxqsets_max = em_set_num_queues(ctx);
847 	if (bootverbose)
848 		device_printf(dev, "attach_pre capping queues at %d\n",
849 		    scctx->isc_ntxqsets_max);
850 
851 	if (hw->mac.type >= igb_mac_min) {
852 		scctx->isc_txqsizes[0] = roundup2(scctx->isc_ntxd[0] * sizeof(union e1000_adv_tx_desc), EM_DBA_ALIGN);
853 		scctx->isc_rxqsizes[0] = roundup2(scctx->isc_nrxd[0] * sizeof(union e1000_adv_rx_desc), EM_DBA_ALIGN);
854 		scctx->isc_txd_size[0] = sizeof(union e1000_adv_tx_desc);
855 		scctx->isc_rxd_size[0] = sizeof(union e1000_adv_rx_desc);
856 		scctx->isc_txrx = &igb_txrx;
857 		scctx->isc_tx_tso_segments_max = EM_MAX_SCATTER;
858 		scctx->isc_tx_tso_size_max = EM_TSO_SIZE;
859 		scctx->isc_tx_tso_segsize_max = EM_TSO_SEG_SIZE;
860 		scctx->isc_capabilities = scctx->isc_capenable = IGB_CAPS;
861 		scctx->isc_tx_csum_flags = CSUM_TCP | CSUM_UDP | CSUM_TSO |
862 		     CSUM_IP6_TCP | CSUM_IP6_UDP;
863 		if (hw->mac.type != e1000_82575)
864 			scctx->isc_tx_csum_flags |= CSUM_SCTP | CSUM_IP6_SCTP;
865 		/*
866 		** Some new devices, as with ixgbe, now may
867 		** use a different BAR, so we need to keep
868 		** track of which is used.
869 		*/
870 		scctx->isc_msix_bar = pci_msix_table_bar(dev);
871 	} else if (hw->mac.type >= em_mac_min) {
872 		scctx->isc_txqsizes[0] = roundup2(scctx->isc_ntxd[0]* sizeof(struct e1000_tx_desc), EM_DBA_ALIGN);
873 		scctx->isc_rxqsizes[0] = roundup2(scctx->isc_nrxd[0] * sizeof(union e1000_rx_desc_extended), EM_DBA_ALIGN);
874 		scctx->isc_txd_size[0] = sizeof(struct e1000_tx_desc);
875 		scctx->isc_rxd_size[0] = sizeof(union e1000_rx_desc_extended);
876 		scctx->isc_txrx = &em_txrx;
877 		scctx->isc_tx_tso_segments_max = EM_MAX_SCATTER;
878 		scctx->isc_tx_tso_size_max = EM_TSO_SIZE;
879 		scctx->isc_tx_tso_segsize_max = EM_TSO_SEG_SIZE;
880 		scctx->isc_capabilities = scctx->isc_capenable = EM_CAPS;
881 		/*
882 		 * For EM-class devices, don't enable IFCAP_{TSO4,VLAN_HWTSO}
883 		 * by default as we don't have workarounds for all associated
884 		 * silicon errata.  E. g., with several MACs such as 82573E,
885 		 * TSO only works at Gigabit speed and otherwise can cause the
886 		 * hardware to hang (which also would be next to impossible to
887 		 * work around given that already queued TSO-using descriptors
888 		 * would need to be flushed and vlan(4) reconfigured at runtime
889 		 * in case of a link speed change).  Moreover, MACs like 82579
890 		 * still can hang at Gigabit even with all publicly documented
891 		 * TSO workarounds implemented.  Generally, the penality of
892 		 * these workarounds is rather high and may involve copying
893 		 * mbuf data around so advantages of TSO lapse.  Still, TSO may
894 		 * work for a few MACs of this class - at least when sticking
895 		 * with Gigabit - in which case users may enable TSO manually.
896 		 */
897 		scctx->isc_capenable &= ~(IFCAP_TSO4 | IFCAP_VLAN_HWTSO);
898 		scctx->isc_tx_csum_flags = CSUM_TCP | CSUM_UDP | CSUM_IP_TSO;
899 		/*
900 		 * We support MSI-X with 82574 only, but indicate to iflib(4)
901 		 * that it shall give MSI at least a try with other devices.
902 		 */
903 		if (hw->mac.type == e1000_82574) {
904 			scctx->isc_msix_bar = pci_msix_table_bar(dev);;
905 		} else {
906 			scctx->isc_msix_bar = -1;
907 			scctx->isc_disable_msix = 1;
908 		}
909 	} else {
910 		scctx->isc_txqsizes[0] = roundup2((scctx->isc_ntxd[0] + 1) * sizeof(struct e1000_tx_desc), EM_DBA_ALIGN);
911 		scctx->isc_rxqsizes[0] = roundup2((scctx->isc_nrxd[0] + 1) * sizeof(struct e1000_rx_desc), EM_DBA_ALIGN);
912 		scctx->isc_txd_size[0] = sizeof(struct e1000_tx_desc);
913 		scctx->isc_rxd_size[0] = sizeof(struct e1000_rx_desc);
914 		scctx->isc_tx_csum_flags = CSUM_TCP | CSUM_UDP;
915 		scctx->isc_txrx = &lem_txrx;
916 		scctx->isc_capabilities = LEM_CAPS;
917 		if (hw->mac.type < e1000_82543)
918 			scctx->isc_capabilities &= ~(IFCAP_HWCSUM|IFCAP_VLAN_HWCSUM);
919 		/* 82541ER doesn't do HW tagging */
920 		if (hw->device_id == E1000_DEV_ID_82541ER || hw->device_id == E1000_DEV_ID_82541ER_LOM)
921 			scctx->isc_capabilities &= ~IFCAP_VLAN_HWTAGGING;
922 		/* INTx only */
923 		scctx->isc_msix_bar = 0;
924 		scctx->isc_capenable = scctx->isc_capabilities;
925 	}
926 
927 	/* Setup PCI resources */
928 	if (em_allocate_pci_resources(ctx)) {
929 		device_printf(dev, "Allocation of PCI resources failed\n");
930 		error = ENXIO;
931 		goto err_pci;
932 	}
933 
934 	/*
935 	** For ICH8 and family we need to
936 	** map the flash memory, and this
937 	** must happen after the MAC is
938 	** identified
939 	*/
940 	if ((hw->mac.type == e1000_ich8lan) ||
941 	    (hw->mac.type == e1000_ich9lan) ||
942 	    (hw->mac.type == e1000_ich10lan) ||
943 	    (hw->mac.type == e1000_pchlan) ||
944 	    (hw->mac.type == e1000_pch2lan) ||
945 	    (hw->mac.type == e1000_pch_lpt)) {
946 		int rid = EM_BAR_TYPE_FLASH;
947 		sc->flash = bus_alloc_resource_any(dev,
948 		    SYS_RES_MEMORY, &rid, RF_ACTIVE);
949 		if (sc->flash == NULL) {
950 			device_printf(dev, "Mapping of Flash failed\n");
951 			error = ENXIO;
952 			goto err_pci;
953 		}
954 		/* This is used in the shared code */
955 		hw->flash_address = (u8 *)sc->flash;
956 		sc->osdep.flash_bus_space_tag =
957 		    rman_get_bustag(sc->flash);
958 		sc->osdep.flash_bus_space_handle =
959 		    rman_get_bushandle(sc->flash);
960 	}
961 	/*
962 	** In the new SPT device flash is not  a
963 	** separate BAR, rather it is also in BAR0,
964 	** so use the same tag and an offset handle for the
965 	** FLASH read/write macros in the shared code.
966 	*/
967 	else if (hw->mac.type >= e1000_pch_spt) {
968 		sc->osdep.flash_bus_space_tag =
969 		    sc->osdep.mem_bus_space_tag;
970 		sc->osdep.flash_bus_space_handle =
971 		    sc->osdep.mem_bus_space_handle
972 		    + E1000_FLASH_BASE_ADDR;
973 	}
974 
975 	/* Do Shared Code initialization */
976 	error = e1000_setup_init_funcs(hw, true);
977 	if (error) {
978 		device_printf(dev, "Setup of Shared code failed, error %d\n",
979 		    error);
980 		error = ENXIO;
981 		goto err_pci;
982 	}
983 
984 	em_setup_msix(ctx);
985 	e1000_get_bus_info(hw);
986 
987 	/* Set up some sysctls for the tunable interrupt delays */
988 	em_add_int_delay_sysctl(sc, "rx_int_delay",
989 	    "receive interrupt delay in usecs", &sc->rx_int_delay,
990 	    E1000_REGISTER(hw, E1000_RDTR), em_rx_int_delay_dflt);
991 	em_add_int_delay_sysctl(sc, "tx_int_delay",
992 	    "transmit interrupt delay in usecs", &sc->tx_int_delay,
993 	    E1000_REGISTER(hw, E1000_TIDV), em_tx_int_delay_dflt);
994 	em_add_int_delay_sysctl(sc, "rx_abs_int_delay",
995 	    "receive interrupt delay limit in usecs",
996 	    &sc->rx_abs_int_delay,
997 	    E1000_REGISTER(hw, E1000_RADV),
998 	    em_rx_abs_int_delay_dflt);
999 	em_add_int_delay_sysctl(sc, "tx_abs_int_delay",
1000 	    "transmit interrupt delay limit in usecs",
1001 	    &sc->tx_abs_int_delay,
1002 	    E1000_REGISTER(hw, E1000_TADV),
1003 	    em_tx_abs_int_delay_dflt);
1004 	em_add_int_delay_sysctl(sc, "itr",
1005 	    "interrupt delay limit in usecs/4",
1006 	    &sc->tx_itr,
1007 	    E1000_REGISTER(hw, E1000_ITR),
1008 	    DEFAULT_ITR);
1009 
1010 	hw->mac.autoneg = DO_AUTO_NEG;
1011 	hw->phy.autoneg_wait_to_complete = false;
1012 	hw->phy.autoneg_advertised = AUTONEG_ADV_DEFAULT;
1013 
1014 	if (hw->mac.type < em_mac_min) {
1015 		e1000_init_script_state_82541(hw, true);
1016 		e1000_set_tbi_compatibility_82543(hw, true);
1017 	}
1018 	/* Copper options */
1019 	if (hw->phy.media_type == e1000_media_type_copper) {
1020 		hw->phy.mdix = AUTO_ALL_MODES;
1021 		hw->phy.disable_polarity_correction = false;
1022 		hw->phy.ms_type = EM_MASTER_SLAVE;
1023 	}
1024 
1025 	/*
1026 	 * Set the frame limits assuming
1027 	 * standard ethernet sized frames.
1028 	 */
1029 	scctx->isc_max_frame_size = hw->mac.max_frame_size =
1030 	    ETHERMTU + ETHER_HDR_LEN + ETHERNET_FCS_SIZE;
1031 
1032 	/*
1033 	 * This controls when hardware reports transmit completion
1034 	 * status.
1035 	 */
1036 	hw->mac.report_tx_early = 1;
1037 
1038 	/* Allocate multicast array memory. */
1039 	sc->mta = malloc(sizeof(u8) * ETHER_ADDR_LEN *
1040 	    MAX_NUM_MULTICAST_ADDRESSES, M_DEVBUF, M_NOWAIT);
1041 	if (sc->mta == NULL) {
1042 		device_printf(dev, "Can not allocate multicast setup array\n");
1043 		error = ENOMEM;
1044 		goto err_late;
1045 	}
1046 
1047 	/* Check SOL/IDER usage */
1048 	if (e1000_check_reset_block(hw))
1049 		device_printf(dev, "PHY reset is blocked"
1050 			      " due to SOL/IDER session.\n");
1051 
1052 	/* Sysctl for setting Energy Efficient Ethernet */
1053 	hw->dev_spec.ich8lan.eee_disable = eee_setting;
1054 	SYSCTL_ADD_PROC(ctx_list, child, OID_AUTO, "eee_control",
1055 	    CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_NEEDGIANT, sc, 0,
1056 	    em_sysctl_eee, "I", "Disable Energy Efficient Ethernet");
1057 
1058 	/*
1059 	** Start from a known state, this is
1060 	** important in reading the nvm and
1061 	** mac from that.
1062 	*/
1063 	e1000_reset_hw(hw);
1064 
1065 	/* Make sure we have a good EEPROM before we read from it */
1066 	if (e1000_validate_nvm_checksum(hw) < 0) {
1067 		/*
1068 		** Some PCI-E parts fail the first check due to
1069 		** the link being in sleep state, call it again,
1070 		** if it fails a second time its a real issue.
1071 		*/
1072 		if (e1000_validate_nvm_checksum(hw) < 0) {
1073 			device_printf(dev,
1074 			    "The EEPROM Checksum Is Not Valid\n");
1075 			error = EIO;
1076 			goto err_late;
1077 		}
1078 	}
1079 
1080 	/* Copy the permanent MAC address out of the EEPROM */
1081 	if (e1000_read_mac_addr(hw) < 0) {
1082 		device_printf(dev, "EEPROM read error while reading MAC"
1083 			      " address\n");
1084 		error = EIO;
1085 		goto err_late;
1086 	}
1087 
1088 	if (!em_is_valid_ether_addr(hw->mac.addr)) {
1089 		if (sc->vf_ifp) {
1090 			ether_gen_addr(iflib_get_ifp(ctx),
1091 			    (struct ether_addr *)hw->mac.addr);
1092 		} else {
1093 			device_printf(dev, "Invalid MAC address\n");
1094 			error = EIO;
1095 			goto err_late;
1096 		}
1097 	}
1098 
1099 	/* Save the EEPROM/NVM versions, must be done under IFLIB_CTX_LOCK */
1100 	em_fw_version_locked(ctx);
1101 
1102 	em_print_fw_version(sc);
1103 
1104 	/* Disable ULP support */
1105 	e1000_disable_ulp_lpt_lp(hw, true);
1106 
1107 	/*
1108 	 * Get Wake-on-Lan and Management info for later use
1109 	 */
1110 	em_get_wakeup(ctx);
1111 
1112 	/* Enable only WOL MAGIC by default */
1113 	scctx->isc_capenable &= ~IFCAP_WOL;
1114 	if (sc->wol != 0)
1115 		scctx->isc_capenable |= IFCAP_WOL_MAGIC;
1116 
1117 	iflib_set_mac(ctx, hw->mac.addr);
1118 
1119 	return (0);
1120 
1121 err_late:
1122 	em_release_hw_control(sc);
1123 err_pci:
1124 	em_free_pci_resources(ctx);
1125 	free(sc->mta, M_DEVBUF);
1126 
1127 	return (error);
1128 }
1129 
1130 static int
1131 em_if_attach_post(if_ctx_t ctx)
1132 {
1133 	struct e1000_softc *sc = iflib_get_softc(ctx);
1134 	struct e1000_hw *hw = &sc->hw;
1135 	int error = 0;
1136 
1137 	/* Setup OS specific network interface */
1138 	error = em_setup_interface(ctx);
1139 	if (error != 0) {
1140 		device_printf(sc->dev, "Interface setup failed: %d\n", error);
1141 		goto err_late;
1142 	}
1143 
1144 	em_reset(ctx);
1145 
1146 	/* Initialize statistics */
1147 	em_update_stats_counters(sc);
1148 	hw->mac.get_link_status = 1;
1149 	em_if_update_admin_status(ctx);
1150 	em_add_hw_stats(sc);
1151 
1152 	/* Non-AMT based hardware can now take control from firmware */
1153 	if (sc->has_manage && !sc->has_amt)
1154 		em_get_hw_control(sc);
1155 
1156 	INIT_DEBUGOUT("em_if_attach_post: end");
1157 
1158 	return (0);
1159 
1160 err_late:
1161 	/* upon attach_post() error, iflib calls _if_detach() to free resources. */
1162 	return (error);
1163 }
1164 
1165 /*********************************************************************
1166  *  Device removal routine
1167  *
1168  *  The detach entry point is called when the driver is being removed.
1169  *  This routine stops the adapter and deallocates all the resources
1170  *  that were allocated for driver operation.
1171  *
1172  *  return 0 on success, positive on failure
1173  *********************************************************************/
1174 static int
1175 em_if_detach(if_ctx_t ctx)
1176 {
1177 	struct e1000_softc	*sc = iflib_get_softc(ctx);
1178 
1179 	INIT_DEBUGOUT("em_if_detach: begin");
1180 
1181 	e1000_phy_hw_reset(&sc->hw);
1182 
1183 	em_release_manageability(sc);
1184 	em_release_hw_control(sc);
1185 	em_free_pci_resources(ctx);
1186 	free(sc->mta, M_DEVBUF);
1187 	sc->mta = NULL;
1188 
1189 	return (0);
1190 }
1191 
1192 /*********************************************************************
1193  *
1194  *  Shutdown entry point
1195  *
1196  **********************************************************************/
1197 
1198 static int
1199 em_if_shutdown(if_ctx_t ctx)
1200 {
1201 	return em_if_suspend(ctx);
1202 }
1203 
1204 /*
1205  * Suspend/resume device methods.
1206  */
1207 static int
1208 em_if_suspend(if_ctx_t ctx)
1209 {
1210 	struct e1000_softc *sc = iflib_get_softc(ctx);
1211 
1212 	em_release_manageability(sc);
1213 	em_release_hw_control(sc);
1214 	em_enable_wakeup(ctx);
1215 	return (0);
1216 }
1217 
1218 static int
1219 em_if_resume(if_ctx_t ctx)
1220 {
1221 	struct e1000_softc *sc = iflib_get_softc(ctx);
1222 
1223 	if (sc->hw.mac.type == e1000_pch2lan)
1224 		e1000_resume_workarounds_pchlan(&sc->hw);
1225 	em_if_init(ctx);
1226 	em_init_manageability(sc);
1227 
1228 	return(0);
1229 }
1230 
1231 static int
1232 em_if_mtu_set(if_ctx_t ctx, uint32_t mtu)
1233 {
1234 	int max_frame_size;
1235 	struct e1000_softc *sc = iflib_get_softc(ctx);
1236 	if_softc_ctx_t scctx = iflib_get_softc_ctx(ctx);
1237 
1238 	IOCTL_DEBUGOUT("ioctl rcv'd: SIOCSIFMTU (Set Interface MTU)");
1239 
1240 	switch (sc->hw.mac.type) {
1241 	case e1000_82571:
1242 	case e1000_82572:
1243 	case e1000_ich9lan:
1244 	case e1000_ich10lan:
1245 	case e1000_pch2lan:
1246 	case e1000_pch_lpt:
1247 	case e1000_pch_spt:
1248 	case e1000_pch_cnp:
1249 	case e1000_pch_tgp:
1250 	case e1000_pch_adp:
1251 	case e1000_pch_mtp:
1252 	case e1000_82574:
1253 	case e1000_82583:
1254 	case e1000_80003es2lan:
1255 		/* 9K Jumbo Frame size */
1256 		max_frame_size = 9234;
1257 		break;
1258 	case e1000_pchlan:
1259 		max_frame_size = 4096;
1260 		break;
1261 	case e1000_82542:
1262 	case e1000_ich8lan:
1263 		/* Adapters that do not support jumbo frames */
1264 		max_frame_size = ETHER_MAX_LEN;
1265 		break;
1266 	default:
1267 		if (sc->hw.mac.type >= igb_mac_min)
1268 			max_frame_size = 9234;
1269 		else /* lem */
1270 			max_frame_size = MAX_JUMBO_FRAME_SIZE;
1271 	}
1272 	if (mtu > max_frame_size - ETHER_HDR_LEN - ETHER_CRC_LEN) {
1273 		return (EINVAL);
1274 	}
1275 
1276 	scctx->isc_max_frame_size = sc->hw.mac.max_frame_size =
1277 	    mtu + ETHER_HDR_LEN + ETHER_CRC_LEN;
1278 	return (0);
1279 }
1280 
1281 /*********************************************************************
1282  *  Init entry point
1283  *
1284  *  This routine is used in two ways. It is used by the stack as
1285  *  init entry point in network interface structure. It is also used
1286  *  by the driver as a hw/sw initialization routine to get to a
1287  *  consistent state.
1288  *
1289  **********************************************************************/
1290 static void
1291 em_if_init(if_ctx_t ctx)
1292 {
1293 	struct e1000_softc *sc = iflib_get_softc(ctx);
1294 	if_softc_ctx_t scctx = sc->shared;
1295 	struct ifnet *ifp = iflib_get_ifp(ctx);
1296 	struct em_tx_queue *tx_que;
1297 	int i;
1298 
1299 	INIT_DEBUGOUT("em_if_init: begin");
1300 
1301 	/* Get the latest mac address, User can use a LAA */
1302 	bcopy(if_getlladdr(ifp), sc->hw.mac.addr,
1303 	    ETHER_ADDR_LEN);
1304 
1305 	/* Put the address into the Receive Address Array */
1306 	e1000_rar_set(&sc->hw, sc->hw.mac.addr, 0);
1307 
1308 	/*
1309 	 * With the 82571 adapter, RAR[0] may be overwritten
1310 	 * when the other port is reset, we make a duplicate
1311 	 * in RAR[14] for that eventuality, this assures
1312 	 * the interface continues to function.
1313 	 */
1314 	if (sc->hw.mac.type == e1000_82571) {
1315 		e1000_set_laa_state_82571(&sc->hw, true);
1316 		e1000_rar_set(&sc->hw, sc->hw.mac.addr,
1317 		    E1000_RAR_ENTRIES - 1);
1318 	}
1319 
1320 
1321 	/* Initialize the hardware */
1322 	em_reset(ctx);
1323 	em_if_update_admin_status(ctx);
1324 
1325 	for (i = 0, tx_que = sc->tx_queues; i < sc->tx_num_queues; i++, tx_que++) {
1326 		struct tx_ring *txr = &tx_que->txr;
1327 
1328 		txr->tx_rs_cidx = txr->tx_rs_pidx;
1329 
1330 		/* Initialize the last processed descriptor to be the end of
1331 		 * the ring, rather than the start, so that we avoid an
1332 		 * off-by-one error when calculating how many descriptors are
1333 		 * done in the credits_update function.
1334 		 */
1335 		txr->tx_cidx_processed = scctx->isc_ntxd[0] - 1;
1336 	}
1337 
1338 	/* Setup VLAN support, basic and offload if available */
1339 	E1000_WRITE_REG(&sc->hw, E1000_VET, ETHERTYPE_VLAN);
1340 
1341 	/* Clear bad data from Rx FIFOs */
1342 	if (sc->hw.mac.type >= igb_mac_min)
1343 		e1000_rx_fifo_flush_base(&sc->hw);
1344 
1345 	/* Configure for OS presence */
1346 	em_init_manageability(sc);
1347 
1348 	/* Prepare transmit descriptors and buffers */
1349 	em_initialize_transmit_unit(ctx);
1350 
1351 	/* Setup Multicast table */
1352 	em_if_multi_set(ctx);
1353 
1354 	sc->rx_mbuf_sz = iflib_get_rx_mbuf_sz(ctx);
1355 	em_initialize_receive_unit(ctx);
1356 
1357 	/* Set up VLAN support and filter */
1358 	em_setup_vlan_hw_support(ctx);
1359 
1360 	/* Don't lose promiscuous settings */
1361 	em_if_set_promisc(ctx, if_getflags(ifp));
1362 	e1000_clear_hw_cntrs_base_generic(&sc->hw);
1363 
1364 	/* MSI-X configuration for 82574 */
1365 	if (sc->hw.mac.type == e1000_82574) {
1366 		int tmp = E1000_READ_REG(&sc->hw, E1000_CTRL_EXT);
1367 
1368 		tmp |= E1000_CTRL_EXT_PBA_CLR;
1369 		E1000_WRITE_REG(&sc->hw, E1000_CTRL_EXT, tmp);
1370 		/* Set the IVAR - interrupt vector routing. */
1371 		E1000_WRITE_REG(&sc->hw, E1000_IVAR, sc->ivars);
1372 	} else if (sc->intr_type == IFLIB_INTR_MSIX) /* Set up queue routing */
1373 		igb_configure_queues(sc);
1374 
1375 	/* this clears any pending interrupts */
1376 	E1000_READ_REG(&sc->hw, E1000_ICR);
1377 	E1000_WRITE_REG(&sc->hw, E1000_ICS, E1000_ICS_LSC);
1378 
1379 	/* AMT based hardware can now take control from firmware */
1380 	if (sc->has_manage && sc->has_amt)
1381 		em_get_hw_control(sc);
1382 
1383 	/* Set Energy Efficient Ethernet */
1384 	if (sc->hw.mac.type >= igb_mac_min &&
1385 	    sc->hw.phy.media_type == e1000_media_type_copper) {
1386 		if (sc->hw.mac.type == e1000_i354)
1387 			e1000_set_eee_i354(&sc->hw, true, true);
1388 		else
1389 			e1000_set_eee_i350(&sc->hw, true, true);
1390 	}
1391 }
1392 
1393 /*********************************************************************
1394  *
1395  *  Fast Legacy/MSI Combined Interrupt Service routine
1396  *
1397  *********************************************************************/
1398 int
1399 em_intr(void *arg)
1400 {
1401 	struct e1000_softc *sc = arg;
1402 	if_ctx_t ctx = sc->ctx;
1403 	u32 reg_icr;
1404 
1405 	reg_icr = E1000_READ_REG(&sc->hw, E1000_ICR);
1406 
1407 	/* Hot eject? */
1408 	if (reg_icr == 0xffffffff)
1409 		return FILTER_STRAY;
1410 
1411 	/* Definitely not our interrupt. */
1412 	if (reg_icr == 0x0)
1413 		return FILTER_STRAY;
1414 
1415 	/*
1416 	 * Starting with the 82571 chip, bit 31 should be used to
1417 	 * determine whether the interrupt belongs to us.
1418 	 */
1419 	if (sc->hw.mac.type >= e1000_82571 &&
1420 	    (reg_icr & E1000_ICR_INT_ASSERTED) == 0)
1421 		return FILTER_STRAY;
1422 
1423 	/*
1424 	 * Only MSI-X interrupts have one-shot behavior by taking advantage
1425 	 * of the EIAC register.  Thus, explicitly disable interrupts.  This
1426 	 * also works around the MSI message reordering errata on certain
1427 	 * systems.
1428 	 */
1429 	IFDI_INTR_DISABLE(ctx);
1430 
1431 	/* Link status change */
1432 	if (reg_icr & (E1000_ICR_RXSEQ | E1000_ICR_LSC))
1433 		em_handle_link(ctx);
1434 
1435 	if (reg_icr & E1000_ICR_RXO)
1436 		sc->rx_overruns++;
1437 
1438 	return (FILTER_SCHEDULE_THREAD);
1439 }
1440 
1441 static int
1442 em_if_rx_queue_intr_enable(if_ctx_t ctx, uint16_t rxqid)
1443 {
1444 	struct e1000_softc *sc = iflib_get_softc(ctx);
1445 	struct em_rx_queue *rxq = &sc->rx_queues[rxqid];
1446 
1447 	E1000_WRITE_REG(&sc->hw, E1000_IMS, rxq->eims);
1448 	return (0);
1449 }
1450 
1451 static int
1452 em_if_tx_queue_intr_enable(if_ctx_t ctx, uint16_t txqid)
1453 {
1454 	struct e1000_softc *sc = iflib_get_softc(ctx);
1455 	struct em_tx_queue *txq = &sc->tx_queues[txqid];
1456 
1457 	E1000_WRITE_REG(&sc->hw, E1000_IMS, txq->eims);
1458 	return (0);
1459 }
1460 
1461 static int
1462 igb_if_rx_queue_intr_enable(if_ctx_t ctx, uint16_t rxqid)
1463 {
1464 	struct e1000_softc *sc = iflib_get_softc(ctx);
1465 	struct em_rx_queue *rxq = &sc->rx_queues[rxqid];
1466 
1467 	E1000_WRITE_REG(&sc->hw, E1000_EIMS, rxq->eims);
1468 	return (0);
1469 }
1470 
1471 static int
1472 igb_if_tx_queue_intr_enable(if_ctx_t ctx, uint16_t txqid)
1473 {
1474 	struct e1000_softc *sc = iflib_get_softc(ctx);
1475 	struct em_tx_queue *txq = &sc->tx_queues[txqid];
1476 
1477 	E1000_WRITE_REG(&sc->hw, E1000_EIMS, txq->eims);
1478 	return (0);
1479 }
1480 
1481 /*********************************************************************
1482  *
1483  *  MSI-X RX Interrupt Service routine
1484  *
1485  **********************************************************************/
1486 static int
1487 em_msix_que(void *arg)
1488 {
1489 	struct em_rx_queue *que = arg;
1490 
1491 	++que->irqs;
1492 
1493 	return (FILTER_SCHEDULE_THREAD);
1494 }
1495 
1496 /*********************************************************************
1497  *
1498  *  MSI-X Link Fast Interrupt Service routine
1499  *
1500  **********************************************************************/
1501 static int
1502 em_msix_link(void *arg)
1503 {
1504 	struct e1000_softc *sc = arg;
1505 	u32 reg_icr;
1506 
1507 	++sc->link_irq;
1508 	MPASS(sc->hw.back != NULL);
1509 	reg_icr = E1000_READ_REG(&sc->hw, E1000_ICR);
1510 
1511 	if (reg_icr & E1000_ICR_RXO)
1512 		sc->rx_overruns++;
1513 
1514 	if (reg_icr & (E1000_ICR_RXSEQ | E1000_ICR_LSC))
1515 		em_handle_link(sc->ctx);
1516 
1517 	/* Re-arm unconditionally */
1518 	if (sc->hw.mac.type >= igb_mac_min) {
1519 		E1000_WRITE_REG(&sc->hw, E1000_IMS, E1000_IMS_LSC);
1520 		E1000_WRITE_REG(&sc->hw, E1000_EIMS, sc->link_mask);
1521 	} else if (sc->hw.mac.type == e1000_82574) {
1522 		E1000_WRITE_REG(&sc->hw, E1000_IMS, E1000_IMS_LSC |
1523 		    E1000_IMS_OTHER);
1524 		/*
1525 		 * Because we must read the ICR for this interrupt it may
1526 		 * clear other causes using autoclear, for this reason we
1527 		 * simply create a soft interrupt for all these vectors.
1528 		 */
1529 		if (reg_icr)
1530 			E1000_WRITE_REG(&sc->hw, E1000_ICS, sc->ims);
1531 	} else
1532 		E1000_WRITE_REG(&sc->hw, E1000_IMS, E1000_IMS_LSC);
1533 
1534 	return (FILTER_HANDLED);
1535 }
1536 
1537 static void
1538 em_handle_link(void *context)
1539 {
1540 	if_ctx_t ctx = context;
1541 	struct e1000_softc *sc = iflib_get_softc(ctx);
1542 
1543 	sc->hw.mac.get_link_status = 1;
1544 	iflib_admin_intr_deferred(ctx);
1545 }
1546 
1547 /*********************************************************************
1548  *
1549  *  Media Ioctl callback
1550  *
1551  *  This routine is called whenever the user queries the status of
1552  *  the interface using ifconfig.
1553  *
1554  **********************************************************************/
1555 static void
1556 em_if_media_status(if_ctx_t ctx, struct ifmediareq *ifmr)
1557 {
1558 	struct e1000_softc *sc = iflib_get_softc(ctx);
1559 	u_char fiber_type = IFM_1000_SX;
1560 
1561 	INIT_DEBUGOUT("em_if_media_status: begin");
1562 
1563 	iflib_admin_intr_deferred(ctx);
1564 
1565 	ifmr->ifm_status = IFM_AVALID;
1566 	ifmr->ifm_active = IFM_ETHER;
1567 
1568 	if (!sc->link_active) {
1569 		return;
1570 	}
1571 
1572 	ifmr->ifm_status |= IFM_ACTIVE;
1573 
1574 	if ((sc->hw.phy.media_type == e1000_media_type_fiber) ||
1575 	    (sc->hw.phy.media_type == e1000_media_type_internal_serdes)) {
1576 		if (sc->hw.mac.type == e1000_82545)
1577 			fiber_type = IFM_1000_LX;
1578 		ifmr->ifm_active |= fiber_type | IFM_FDX;
1579 	} else {
1580 		switch (sc->link_speed) {
1581 		case 10:
1582 			ifmr->ifm_active |= IFM_10_T;
1583 			break;
1584 		case 100:
1585 			ifmr->ifm_active |= IFM_100_TX;
1586 			break;
1587 		case 1000:
1588 			ifmr->ifm_active |= IFM_1000_T;
1589 			break;
1590 		}
1591 		if (sc->link_duplex == FULL_DUPLEX)
1592 			ifmr->ifm_active |= IFM_FDX;
1593 		else
1594 			ifmr->ifm_active |= IFM_HDX;
1595 	}
1596 }
1597 
1598 /*********************************************************************
1599  *
1600  *  Media Ioctl callback
1601  *
1602  *  This routine is called when the user changes speed/duplex using
1603  *  media/mediopt option with ifconfig.
1604  *
1605  **********************************************************************/
1606 static int
1607 em_if_media_change(if_ctx_t ctx)
1608 {
1609 	struct e1000_softc *sc = iflib_get_softc(ctx);
1610 	struct ifmedia *ifm = iflib_get_media(ctx);
1611 
1612 	INIT_DEBUGOUT("em_if_media_change: begin");
1613 
1614 	if (IFM_TYPE(ifm->ifm_media) != IFM_ETHER)
1615 		return (EINVAL);
1616 
1617 	switch (IFM_SUBTYPE(ifm->ifm_media)) {
1618 	case IFM_AUTO:
1619 		sc->hw.mac.autoneg = DO_AUTO_NEG;
1620 		sc->hw.phy.autoneg_advertised = AUTONEG_ADV_DEFAULT;
1621 		break;
1622 	case IFM_1000_LX:
1623 	case IFM_1000_SX:
1624 	case IFM_1000_T:
1625 		sc->hw.mac.autoneg = DO_AUTO_NEG;
1626 		sc->hw.phy.autoneg_advertised = ADVERTISE_1000_FULL;
1627 		break;
1628 	case IFM_100_TX:
1629 		sc->hw.mac.autoneg = false;
1630 		sc->hw.phy.autoneg_advertised = 0;
1631 		if ((ifm->ifm_media & IFM_GMASK) == IFM_FDX)
1632 			sc->hw.mac.forced_speed_duplex = ADVERTISE_100_FULL;
1633 		else
1634 			sc->hw.mac.forced_speed_duplex = ADVERTISE_100_HALF;
1635 		break;
1636 	case IFM_10_T:
1637 		sc->hw.mac.autoneg = false;
1638 		sc->hw.phy.autoneg_advertised = 0;
1639 		if ((ifm->ifm_media & IFM_GMASK) == IFM_FDX)
1640 			sc->hw.mac.forced_speed_duplex = ADVERTISE_10_FULL;
1641 		else
1642 			sc->hw.mac.forced_speed_duplex = ADVERTISE_10_HALF;
1643 		break;
1644 	default:
1645 		device_printf(sc->dev, "Unsupported media type\n");
1646 	}
1647 
1648 	em_if_init(ctx);
1649 
1650 	return (0);
1651 }
1652 
1653 static int
1654 em_if_set_promisc(if_ctx_t ctx, int flags)
1655 {
1656 	struct e1000_softc *sc = iflib_get_softc(ctx);
1657 	struct ifnet *ifp = iflib_get_ifp(ctx);
1658 	u32 reg_rctl;
1659 	int mcnt = 0;
1660 
1661 	reg_rctl = E1000_READ_REG(&sc->hw, E1000_RCTL);
1662 	reg_rctl &= ~(E1000_RCTL_SBP | E1000_RCTL_UPE);
1663 	if (flags & IFF_ALLMULTI)
1664 		mcnt = MAX_NUM_MULTICAST_ADDRESSES;
1665 	else
1666 		mcnt = min(if_llmaddr_count(ifp), MAX_NUM_MULTICAST_ADDRESSES);
1667 
1668 	if (mcnt < MAX_NUM_MULTICAST_ADDRESSES)
1669 		reg_rctl &= (~E1000_RCTL_MPE);
1670 	E1000_WRITE_REG(&sc->hw, E1000_RCTL, reg_rctl);
1671 
1672 	if (flags & IFF_PROMISC) {
1673 		reg_rctl |= (E1000_RCTL_UPE | E1000_RCTL_MPE);
1674 		em_if_vlan_filter_disable(sc);
1675 		/* Turn this on if you want to see bad packets */
1676 		if (em_debug_sbp)
1677 			reg_rctl |= E1000_RCTL_SBP;
1678 		E1000_WRITE_REG(&sc->hw, E1000_RCTL, reg_rctl);
1679 	} else {
1680 		if (flags & IFF_ALLMULTI) {
1681 			reg_rctl |= E1000_RCTL_MPE;
1682 			reg_rctl &= ~E1000_RCTL_UPE;
1683 			E1000_WRITE_REG(&sc->hw, E1000_RCTL, reg_rctl);
1684 		}
1685 		if (em_if_vlan_filter_used(ctx))
1686 			em_if_vlan_filter_enable(sc);
1687 	}
1688 	return (0);
1689 }
1690 
1691 static u_int
1692 em_copy_maddr(void *arg, struct sockaddr_dl *sdl, u_int idx)
1693 {
1694 	u8 *mta = arg;
1695 
1696 	if (idx == MAX_NUM_MULTICAST_ADDRESSES)
1697 		return (0);
1698 
1699 	bcopy(LLADDR(sdl), &mta[idx * ETHER_ADDR_LEN], ETHER_ADDR_LEN);
1700 
1701 	return (1);
1702 }
1703 
1704 /*********************************************************************
1705  *  Multicast Update
1706  *
1707  *  This routine is called whenever multicast address list is updated.
1708  *
1709  **********************************************************************/
1710 static void
1711 em_if_multi_set(if_ctx_t ctx)
1712 {
1713 	struct e1000_softc *sc = iflib_get_softc(ctx);
1714 	struct ifnet *ifp = iflib_get_ifp(ctx);
1715 	u8  *mta; /* Multicast array memory */
1716 	u32 reg_rctl = 0;
1717 	int mcnt = 0;
1718 
1719 	IOCTL_DEBUGOUT("em_set_multi: begin");
1720 
1721 	mta = sc->mta;
1722 	bzero(mta, sizeof(u8) * ETHER_ADDR_LEN * MAX_NUM_MULTICAST_ADDRESSES);
1723 
1724 	if (sc->hw.mac.type == e1000_82542 &&
1725 	    sc->hw.revision_id == E1000_REVISION_2) {
1726 		reg_rctl = E1000_READ_REG(&sc->hw, E1000_RCTL);
1727 		if (sc->hw.bus.pci_cmd_word & CMD_MEM_WRT_INVALIDATE)
1728 			e1000_pci_clear_mwi(&sc->hw);
1729 		reg_rctl |= E1000_RCTL_RST;
1730 		E1000_WRITE_REG(&sc->hw, E1000_RCTL, reg_rctl);
1731 		msec_delay(5);
1732 	}
1733 
1734 	mcnt = if_foreach_llmaddr(ifp, em_copy_maddr, mta);
1735 
1736 	if (mcnt < MAX_NUM_MULTICAST_ADDRESSES)
1737 		e1000_update_mc_addr_list(&sc->hw, mta, mcnt);
1738 
1739 	reg_rctl = E1000_READ_REG(&sc->hw, E1000_RCTL);
1740 
1741 	if (if_getflags(ifp) & IFF_PROMISC)
1742 		reg_rctl |= (E1000_RCTL_UPE | E1000_RCTL_MPE);
1743 	else if (mcnt >= MAX_NUM_MULTICAST_ADDRESSES ||
1744 	    if_getflags(ifp) & IFF_ALLMULTI) {
1745 		reg_rctl |= E1000_RCTL_MPE;
1746 		reg_rctl &= ~E1000_RCTL_UPE;
1747 	} else
1748 		reg_rctl &= ~(E1000_RCTL_UPE | E1000_RCTL_MPE);
1749 
1750 	E1000_WRITE_REG(&sc->hw, E1000_RCTL, reg_rctl);
1751 
1752 	if (sc->hw.mac.type == e1000_82542 &&
1753 	    sc->hw.revision_id == E1000_REVISION_2) {
1754 		reg_rctl = E1000_READ_REG(&sc->hw, E1000_RCTL);
1755 		reg_rctl &= ~E1000_RCTL_RST;
1756 		E1000_WRITE_REG(&sc->hw, E1000_RCTL, reg_rctl);
1757 		msec_delay(5);
1758 		if (sc->hw.bus.pci_cmd_word & CMD_MEM_WRT_INVALIDATE)
1759 			e1000_pci_set_mwi(&sc->hw);
1760 	}
1761 }
1762 
1763 /*********************************************************************
1764  *  Timer routine
1765  *
1766  *  This routine schedules em_if_update_admin_status() to check for
1767  *  link status and to gather statistics as well as to perform some
1768  *  controller-specific hardware patting.
1769  *
1770  **********************************************************************/
1771 static void
1772 em_if_timer(if_ctx_t ctx, uint16_t qid)
1773 {
1774 
1775 	if (qid != 0)
1776 		return;
1777 
1778 	iflib_admin_intr_deferred(ctx);
1779 }
1780 
1781 static void
1782 em_if_update_admin_status(if_ctx_t ctx)
1783 {
1784 	struct e1000_softc *sc = iflib_get_softc(ctx);
1785 	struct e1000_hw *hw = &sc->hw;
1786 	device_t dev = iflib_get_dev(ctx);
1787 	u32 link_check, thstat, ctrl;
1788 
1789 	link_check = thstat = ctrl = 0;
1790 	/* Get the cached link value or read phy for real */
1791 	switch (hw->phy.media_type) {
1792 	case e1000_media_type_copper:
1793 		if (hw->mac.get_link_status) {
1794 			if (hw->mac.type == e1000_pch_spt)
1795 				msec_delay(50);
1796 			/* Do the work to read phy */
1797 			e1000_check_for_link(hw);
1798 			link_check = !hw->mac.get_link_status;
1799 			if (link_check) /* ESB2 fix */
1800 				e1000_cfg_on_link_up(hw);
1801 		} else {
1802 			link_check = true;
1803 		}
1804 		break;
1805 	case e1000_media_type_fiber:
1806 		e1000_check_for_link(hw);
1807 		link_check = (E1000_READ_REG(hw, E1000_STATUS) &
1808 			    E1000_STATUS_LU);
1809 		break;
1810 	case e1000_media_type_internal_serdes:
1811 		e1000_check_for_link(hw);
1812 		link_check = hw->mac.serdes_has_link;
1813 		break;
1814 	/* VF device is type_unknown */
1815 	case e1000_media_type_unknown:
1816 		e1000_check_for_link(hw);
1817 		link_check = !hw->mac.get_link_status;
1818 		/* FALLTHROUGH */
1819 	default:
1820 		break;
1821 	}
1822 
1823 	/* Check for thermal downshift or shutdown */
1824 	if (hw->mac.type == e1000_i350) {
1825 		thstat = E1000_READ_REG(hw, E1000_THSTAT);
1826 		ctrl = E1000_READ_REG(hw, E1000_CTRL_EXT);
1827 	}
1828 
1829 	/* Now check for a transition */
1830 	if (link_check && (sc->link_active == 0)) {
1831 		e1000_get_speed_and_duplex(hw, &sc->link_speed,
1832 		    &sc->link_duplex);
1833 		/* Check if we must disable SPEED_MODE bit on PCI-E */
1834 		if ((sc->link_speed != SPEED_1000) &&
1835 		    ((hw->mac.type == e1000_82571) ||
1836 		    (hw->mac.type == e1000_82572))) {
1837 			int tarc0;
1838 			tarc0 = E1000_READ_REG(hw, E1000_TARC(0));
1839 			tarc0 &= ~TARC_SPEED_MODE_BIT;
1840 			E1000_WRITE_REG(hw, E1000_TARC(0), tarc0);
1841 		}
1842 		if (bootverbose)
1843 			device_printf(dev, "Link is up %d Mbps %s\n",
1844 			    sc->link_speed,
1845 			    ((sc->link_duplex == FULL_DUPLEX) ?
1846 			    "Full Duplex" : "Half Duplex"));
1847 		sc->link_active = 1;
1848 		sc->smartspeed = 0;
1849 		if ((ctrl & E1000_CTRL_EXT_LINK_MODE_MASK) ==
1850 		    E1000_CTRL_EXT_LINK_MODE_GMII &&
1851 		    (thstat & E1000_THSTAT_LINK_THROTTLE))
1852 			device_printf(dev, "Link: thermal downshift\n");
1853 		/* Delay Link Up for Phy update */
1854 		if (((hw->mac.type == e1000_i210) ||
1855 		    (hw->mac.type == e1000_i211)) &&
1856 		    (hw->phy.id == I210_I_PHY_ID))
1857 			msec_delay(I210_LINK_DELAY);
1858 		/* Reset if the media type changed. */
1859 		if (hw->dev_spec._82575.media_changed &&
1860 		    hw->mac.type >= igb_mac_min) {
1861 			hw->dev_spec._82575.media_changed = false;
1862 			sc->flags |= IGB_MEDIA_RESET;
1863 			em_reset(ctx);
1864 		}
1865 		iflib_link_state_change(ctx, LINK_STATE_UP,
1866 		    IF_Mbps(sc->link_speed));
1867 	} else if (!link_check && (sc->link_active == 1)) {
1868 		sc->link_speed = 0;
1869 		sc->link_duplex = 0;
1870 		sc->link_active = 0;
1871 		iflib_link_state_change(ctx, LINK_STATE_DOWN, 0);
1872 	}
1873 	em_update_stats_counters(sc);
1874 
1875 	/* Reset LAA into RAR[0] on 82571 */
1876 	if (hw->mac.type == e1000_82571 && e1000_get_laa_state_82571(hw))
1877 		e1000_rar_set(hw, hw->mac.addr, 0);
1878 
1879 	if (hw->mac.type < em_mac_min)
1880 		lem_smartspeed(sc);
1881 }
1882 
1883 static void
1884 em_if_watchdog_reset(if_ctx_t ctx)
1885 {
1886 	struct e1000_softc *sc = iflib_get_softc(ctx);
1887 
1888 	/*
1889 	 * Just count the event; iflib(4) will already trigger a
1890 	 * sufficient reset of the controller.
1891 	 */
1892 	sc->watchdog_events++;
1893 }
1894 
1895 /*********************************************************************
1896  *
1897  *  This routine disables all traffic on the adapter by issuing a
1898  *  global reset on the MAC.
1899  *
1900  **********************************************************************/
1901 static void
1902 em_if_stop(if_ctx_t ctx)
1903 {
1904 	struct e1000_softc *sc = iflib_get_softc(ctx);
1905 
1906 	INIT_DEBUGOUT("em_if_stop: begin");
1907 
1908 	e1000_reset_hw(&sc->hw);
1909 	if (sc->hw.mac.type >= e1000_82544)
1910 		E1000_WRITE_REG(&sc->hw, E1000_WUFC, 0);
1911 
1912 	e1000_led_off(&sc->hw);
1913 	e1000_cleanup_led(&sc->hw);
1914 }
1915 
1916 /*********************************************************************
1917  *
1918  *  Determine hardware revision.
1919  *
1920  **********************************************************************/
1921 static void
1922 em_identify_hardware(if_ctx_t ctx)
1923 {
1924 	device_t dev = iflib_get_dev(ctx);
1925 	struct e1000_softc *sc = iflib_get_softc(ctx);
1926 
1927 	/* Make sure our PCI config space has the necessary stuff set */
1928 	sc->hw.bus.pci_cmd_word = pci_read_config(dev, PCIR_COMMAND, 2);
1929 
1930 	/* Save off the information about this board */
1931 	sc->hw.vendor_id = pci_get_vendor(dev);
1932 	sc->hw.device_id = pci_get_device(dev);
1933 	sc->hw.revision_id = pci_read_config(dev, PCIR_REVID, 1);
1934 	sc->hw.subsystem_vendor_id =
1935 	    pci_read_config(dev, PCIR_SUBVEND_0, 2);
1936 	sc->hw.subsystem_device_id =
1937 	    pci_read_config(dev, PCIR_SUBDEV_0, 2);
1938 
1939 	/* Do Shared Code Init and Setup */
1940 	if (e1000_set_mac_type(&sc->hw)) {
1941 		device_printf(dev, "Setup init failure\n");
1942 		return;
1943 	}
1944 
1945 	/* Are we a VF device? */
1946 	if ((sc->hw.mac.type == e1000_vfadapt) ||
1947 	    (sc->hw.mac.type == e1000_vfadapt_i350))
1948 		sc->vf_ifp = 1;
1949 	else
1950 		sc->vf_ifp = 0;
1951 }
1952 
1953 static int
1954 em_allocate_pci_resources(if_ctx_t ctx)
1955 {
1956 	struct e1000_softc *sc = iflib_get_softc(ctx);
1957 	device_t dev = iflib_get_dev(ctx);
1958 	int rid, val;
1959 
1960 	rid = PCIR_BAR(0);
1961 	sc->memory = bus_alloc_resource_any(dev, SYS_RES_MEMORY,
1962 	    &rid, RF_ACTIVE);
1963 	if (sc->memory == NULL) {
1964 		device_printf(dev, "Unable to allocate bus resource: memory\n");
1965 		return (ENXIO);
1966 	}
1967 	sc->osdep.mem_bus_space_tag = rman_get_bustag(sc->memory);
1968 	sc->osdep.mem_bus_space_handle =
1969 	    rman_get_bushandle(sc->memory);
1970 	sc->hw.hw_addr = (u8 *)&sc->osdep.mem_bus_space_handle;
1971 
1972 	/* Only older adapters use IO mapping */
1973 	if (sc->hw.mac.type < em_mac_min &&
1974 	    sc->hw.mac.type > e1000_82543) {
1975 		/* Figure our where our IO BAR is ? */
1976 		for (rid = PCIR_BAR(0); rid < PCIR_CIS;) {
1977 			val = pci_read_config(dev, rid, 4);
1978 			if (EM_BAR_TYPE(val) == EM_BAR_TYPE_IO) {
1979 				break;
1980 			}
1981 			rid += 4;
1982 			/* check for 64bit BAR */
1983 			if (EM_BAR_MEM_TYPE(val) == EM_BAR_MEM_TYPE_64BIT)
1984 				rid += 4;
1985 		}
1986 		if (rid >= PCIR_CIS) {
1987 			device_printf(dev, "Unable to locate IO BAR\n");
1988 			return (ENXIO);
1989 		}
1990 		sc->ioport = bus_alloc_resource_any(dev, SYS_RES_IOPORT,
1991 		    &rid, RF_ACTIVE);
1992 		if (sc->ioport == NULL) {
1993 			device_printf(dev, "Unable to allocate bus resource: "
1994 			    "ioport\n");
1995 			return (ENXIO);
1996 		}
1997 		sc->hw.io_base = 0;
1998 		sc->osdep.io_bus_space_tag =
1999 		    rman_get_bustag(sc->ioport);
2000 		sc->osdep.io_bus_space_handle =
2001 		    rman_get_bushandle(sc->ioport);
2002 	}
2003 
2004 	sc->hw.back = &sc->osdep;
2005 
2006 	return (0);
2007 }
2008 
2009 /*********************************************************************
2010  *
2011  *  Set up the MSI-X Interrupt handlers
2012  *
2013  **********************************************************************/
2014 static int
2015 em_if_msix_intr_assign(if_ctx_t ctx, int msix)
2016 {
2017 	struct e1000_softc *sc = iflib_get_softc(ctx);
2018 	struct em_rx_queue *rx_que = sc->rx_queues;
2019 	struct em_tx_queue *tx_que = sc->tx_queues;
2020 	int error, rid, i, vector = 0, rx_vectors;
2021 	char buf[16];
2022 
2023 	/* First set up ring resources */
2024 	for (i = 0; i < sc->rx_num_queues; i++, rx_que++, vector++) {
2025 		rid = vector + 1;
2026 		snprintf(buf, sizeof(buf), "rxq%d", i);
2027 		error = iflib_irq_alloc_generic(ctx, &rx_que->que_irq, rid, IFLIB_INTR_RXTX, em_msix_que, rx_que, rx_que->me, buf);
2028 		if (error) {
2029 			device_printf(iflib_get_dev(ctx), "Failed to allocate que int %d err: %d", i, error);
2030 			sc->rx_num_queues = i + 1;
2031 			goto fail;
2032 		}
2033 
2034 		rx_que->msix =  vector;
2035 
2036 		/*
2037 		 * Set the bit to enable interrupt
2038 		 * in E1000_IMS -- bits 20 and 21
2039 		 * are for RX0 and RX1, note this has
2040 		 * NOTHING to do with the MSI-X vector
2041 		 */
2042 		if (sc->hw.mac.type == e1000_82574) {
2043 			rx_que->eims = 1 << (20 + i);
2044 			sc->ims |= rx_que->eims;
2045 			sc->ivars |= (8 | rx_que->msix) << (i * 4);
2046 		} else if (sc->hw.mac.type == e1000_82575)
2047 			rx_que->eims = E1000_EICR_TX_QUEUE0 << vector;
2048 		else
2049 			rx_que->eims = 1 << vector;
2050 	}
2051 	rx_vectors = vector;
2052 
2053 	vector = 0;
2054 	for (i = 0; i < sc->tx_num_queues; i++, tx_que++, vector++) {
2055 		snprintf(buf, sizeof(buf), "txq%d", i);
2056 		tx_que = &sc->tx_queues[i];
2057 		iflib_softirq_alloc_generic(ctx,
2058 		    &sc->rx_queues[i % sc->rx_num_queues].que_irq,
2059 		    IFLIB_INTR_TX, tx_que, tx_que->me, buf);
2060 
2061 		tx_que->msix = (vector % sc->rx_num_queues);
2062 
2063 		/*
2064 		 * Set the bit to enable interrupt
2065 		 * in E1000_IMS -- bits 22 and 23
2066 		 * are for TX0 and TX1, note this has
2067 		 * NOTHING to do with the MSI-X vector
2068 		 */
2069 		if (sc->hw.mac.type == e1000_82574) {
2070 			tx_que->eims = 1 << (22 + i);
2071 			sc->ims |= tx_que->eims;
2072 			sc->ivars |= (8 | tx_que->msix) << (8 + (i * 4));
2073 		} else if (sc->hw.mac.type == e1000_82575) {
2074 			tx_que->eims = E1000_EICR_TX_QUEUE0 << i;
2075 		} else {
2076 			tx_que->eims = 1 << i;
2077 		}
2078 	}
2079 
2080 	/* Link interrupt */
2081 	rid = rx_vectors + 1;
2082 	error = iflib_irq_alloc_generic(ctx, &sc->irq, rid, IFLIB_INTR_ADMIN, em_msix_link, sc, 0, "aq");
2083 
2084 	if (error) {
2085 		device_printf(iflib_get_dev(ctx), "Failed to register admin handler");
2086 		goto fail;
2087 	}
2088 	sc->linkvec = rx_vectors;
2089 	if (sc->hw.mac.type < igb_mac_min) {
2090 		sc->ivars |=  (8 | rx_vectors) << 16;
2091 		sc->ivars |= 0x80000000;
2092 		/* Enable the "Other" interrupt type for link status change */
2093 		sc->ims |= E1000_IMS_OTHER;
2094 	}
2095 
2096 	return (0);
2097 fail:
2098 	iflib_irq_free(ctx, &sc->irq);
2099 	rx_que = sc->rx_queues;
2100 	for (int i = 0; i < sc->rx_num_queues; i++, rx_que++)
2101 		iflib_irq_free(ctx, &rx_que->que_irq);
2102 	return (error);
2103 }
2104 
2105 static void
2106 igb_configure_queues(struct e1000_softc *sc)
2107 {
2108 	struct e1000_hw *hw = &sc->hw;
2109 	struct em_rx_queue *rx_que;
2110 	struct em_tx_queue *tx_que;
2111 	u32 tmp, ivar = 0, newitr = 0;
2112 
2113 	/* First turn on RSS capability */
2114 	if (hw->mac.type != e1000_82575)
2115 		E1000_WRITE_REG(hw, E1000_GPIE,
2116 		    E1000_GPIE_MSIX_MODE | E1000_GPIE_EIAME |
2117 		    E1000_GPIE_PBA | E1000_GPIE_NSICR);
2118 
2119 	/* Turn on MSI-X */
2120 	switch (hw->mac.type) {
2121 	case e1000_82580:
2122 	case e1000_i350:
2123 	case e1000_i354:
2124 	case e1000_i210:
2125 	case e1000_i211:
2126 	case e1000_vfadapt:
2127 	case e1000_vfadapt_i350:
2128 		/* RX entries */
2129 		for (int i = 0; i < sc->rx_num_queues; i++) {
2130 			u32 index = i >> 1;
2131 			ivar = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index);
2132 			rx_que = &sc->rx_queues[i];
2133 			if (i & 1) {
2134 				ivar &= 0xFF00FFFF;
2135 				ivar |= (rx_que->msix | E1000_IVAR_VALID) << 16;
2136 			} else {
2137 				ivar &= 0xFFFFFF00;
2138 				ivar |= rx_que->msix | E1000_IVAR_VALID;
2139 			}
2140 			E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, ivar);
2141 		}
2142 		/* TX entries */
2143 		for (int i = 0; i < sc->tx_num_queues; i++) {
2144 			u32 index = i >> 1;
2145 			ivar = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index);
2146 			tx_que = &sc->tx_queues[i];
2147 			if (i & 1) {
2148 				ivar &= 0x00FFFFFF;
2149 				ivar |= (tx_que->msix | E1000_IVAR_VALID) << 24;
2150 			} else {
2151 				ivar &= 0xFFFF00FF;
2152 				ivar |= (tx_que->msix | E1000_IVAR_VALID) << 8;
2153 			}
2154 			E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, ivar);
2155 			sc->que_mask |= tx_que->eims;
2156 		}
2157 
2158 		/* And for the link interrupt */
2159 		ivar = (sc->linkvec | E1000_IVAR_VALID) << 8;
2160 		sc->link_mask = 1 << sc->linkvec;
2161 		E1000_WRITE_REG(hw, E1000_IVAR_MISC, ivar);
2162 		break;
2163 	case e1000_82576:
2164 		/* RX entries */
2165 		for (int i = 0; i < sc->rx_num_queues; i++) {
2166 			u32 index = i & 0x7; /* Each IVAR has two entries */
2167 			ivar = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index);
2168 			rx_que = &sc->rx_queues[i];
2169 			if (i < 8) {
2170 				ivar &= 0xFFFFFF00;
2171 				ivar |= rx_que->msix | E1000_IVAR_VALID;
2172 			} else {
2173 				ivar &= 0xFF00FFFF;
2174 				ivar |= (rx_que->msix | E1000_IVAR_VALID) << 16;
2175 			}
2176 			E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, ivar);
2177 			sc->que_mask |= rx_que->eims;
2178 		}
2179 		/* TX entries */
2180 		for (int i = 0; i < sc->tx_num_queues; i++) {
2181 			u32 index = i & 0x7; /* Each IVAR has two entries */
2182 			ivar = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index);
2183 			tx_que = &sc->tx_queues[i];
2184 			if (i < 8) {
2185 				ivar &= 0xFFFF00FF;
2186 				ivar |= (tx_que->msix | E1000_IVAR_VALID) << 8;
2187 			} else {
2188 				ivar &= 0x00FFFFFF;
2189 				ivar |= (tx_que->msix | E1000_IVAR_VALID) << 24;
2190 			}
2191 			E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, ivar);
2192 			sc->que_mask |= tx_que->eims;
2193 		}
2194 
2195 		/* And for the link interrupt */
2196 		ivar = (sc->linkvec | E1000_IVAR_VALID) << 8;
2197 		sc->link_mask = 1 << sc->linkvec;
2198 		E1000_WRITE_REG(hw, E1000_IVAR_MISC, ivar);
2199 		break;
2200 
2201 	case e1000_82575:
2202 		/* enable MSI-X support*/
2203 		tmp = E1000_READ_REG(hw, E1000_CTRL_EXT);
2204 		tmp |= E1000_CTRL_EXT_PBA_CLR;
2205 		/* Auto-Mask interrupts upon ICR read. */
2206 		tmp |= E1000_CTRL_EXT_EIAME;
2207 		tmp |= E1000_CTRL_EXT_IRCA;
2208 		E1000_WRITE_REG(hw, E1000_CTRL_EXT, tmp);
2209 
2210 		/* Queues */
2211 		for (int i = 0; i < sc->rx_num_queues; i++) {
2212 			rx_que = &sc->rx_queues[i];
2213 			tmp = E1000_EICR_RX_QUEUE0 << i;
2214 			tmp |= E1000_EICR_TX_QUEUE0 << i;
2215 			rx_que->eims = tmp;
2216 			E1000_WRITE_REG_ARRAY(hw, E1000_MSIXBM(0),
2217 			    i, rx_que->eims);
2218 			sc->que_mask |= rx_que->eims;
2219 		}
2220 
2221 		/* Link */
2222 		E1000_WRITE_REG(hw, E1000_MSIXBM(sc->linkvec),
2223 		    E1000_EIMS_OTHER);
2224 		sc->link_mask |= E1000_EIMS_OTHER;
2225 	default:
2226 		break;
2227 	}
2228 
2229 	/* Set the starting interrupt rate */
2230 	if (em_max_interrupt_rate > 0)
2231 		newitr = (4000000 / em_max_interrupt_rate) & 0x7FFC;
2232 
2233 	if (hw->mac.type == e1000_82575)
2234 		newitr |= newitr << 16;
2235 	else
2236 		newitr |= E1000_EITR_CNT_IGNR;
2237 
2238 	for (int i = 0; i < sc->rx_num_queues; i++) {
2239 		rx_que = &sc->rx_queues[i];
2240 		E1000_WRITE_REG(hw, E1000_EITR(rx_que->msix), newitr);
2241 	}
2242 
2243 	return;
2244 }
2245 
2246 static void
2247 em_free_pci_resources(if_ctx_t ctx)
2248 {
2249 	struct e1000_softc *sc = iflib_get_softc(ctx);
2250 	struct em_rx_queue *que = sc->rx_queues;
2251 	device_t dev = iflib_get_dev(ctx);
2252 
2253 	/* Release all MSI-X queue resources */
2254 	if (sc->intr_type == IFLIB_INTR_MSIX)
2255 		iflib_irq_free(ctx, &sc->irq);
2256 
2257 	if (que != NULL) {
2258 		for (int i = 0; i < sc->rx_num_queues; i++, que++) {
2259 			iflib_irq_free(ctx, &que->que_irq);
2260 		}
2261 	}
2262 
2263 	if (sc->memory != NULL) {
2264 		bus_release_resource(dev, SYS_RES_MEMORY,
2265 		    rman_get_rid(sc->memory), sc->memory);
2266 		sc->memory = NULL;
2267 	}
2268 
2269 	if (sc->flash != NULL) {
2270 		bus_release_resource(dev, SYS_RES_MEMORY,
2271 		    rman_get_rid(sc->flash), sc->flash);
2272 		sc->flash = NULL;
2273 	}
2274 
2275 	if (sc->ioport != NULL) {
2276 		bus_release_resource(dev, SYS_RES_IOPORT,
2277 		    rman_get_rid(sc->ioport), sc->ioport);
2278 		sc->ioport = NULL;
2279 	}
2280 }
2281 
2282 /* Set up MSI or MSI-X */
2283 static int
2284 em_setup_msix(if_ctx_t ctx)
2285 {
2286 	struct e1000_softc *sc = iflib_get_softc(ctx);
2287 
2288 	if (sc->hw.mac.type == e1000_82574) {
2289 		em_enable_vectors_82574(ctx);
2290 	}
2291 	return (0);
2292 }
2293 
2294 /*********************************************************************
2295  *
2296  *  Workaround for SmartSpeed on 82541 and 82547 controllers
2297  *
2298  **********************************************************************/
2299 static void
2300 lem_smartspeed(struct e1000_softc *sc)
2301 {
2302 	u16 phy_tmp;
2303 
2304 	if (sc->link_active || (sc->hw.phy.type != e1000_phy_igp) ||
2305 	    sc->hw.mac.autoneg == 0 ||
2306 	    (sc->hw.phy.autoneg_advertised & ADVERTISE_1000_FULL) == 0)
2307 		return;
2308 
2309 	if (sc->smartspeed == 0) {
2310 		/* If Master/Slave config fault is asserted twice,
2311 		 * we assume back-to-back */
2312 		e1000_read_phy_reg(&sc->hw, PHY_1000T_STATUS, &phy_tmp);
2313 		if (!(phy_tmp & SR_1000T_MS_CONFIG_FAULT))
2314 			return;
2315 		e1000_read_phy_reg(&sc->hw, PHY_1000T_STATUS, &phy_tmp);
2316 		if (phy_tmp & SR_1000T_MS_CONFIG_FAULT) {
2317 			e1000_read_phy_reg(&sc->hw,
2318 			    PHY_1000T_CTRL, &phy_tmp);
2319 			if(phy_tmp & CR_1000T_MS_ENABLE) {
2320 				phy_tmp &= ~CR_1000T_MS_ENABLE;
2321 				e1000_write_phy_reg(&sc->hw,
2322 				    PHY_1000T_CTRL, phy_tmp);
2323 				sc->smartspeed++;
2324 				if(sc->hw.mac.autoneg &&
2325 				   !e1000_copper_link_autoneg(&sc->hw) &&
2326 				   !e1000_read_phy_reg(&sc->hw,
2327 				    PHY_CONTROL, &phy_tmp)) {
2328 					phy_tmp |= (MII_CR_AUTO_NEG_EN |
2329 						    MII_CR_RESTART_AUTO_NEG);
2330 					e1000_write_phy_reg(&sc->hw,
2331 					    PHY_CONTROL, phy_tmp);
2332 				}
2333 			}
2334 		}
2335 		return;
2336 	} else if(sc->smartspeed == EM_SMARTSPEED_DOWNSHIFT) {
2337 		/* If still no link, perhaps using 2/3 pair cable */
2338 		e1000_read_phy_reg(&sc->hw, PHY_1000T_CTRL, &phy_tmp);
2339 		phy_tmp |= CR_1000T_MS_ENABLE;
2340 		e1000_write_phy_reg(&sc->hw, PHY_1000T_CTRL, phy_tmp);
2341 		if(sc->hw.mac.autoneg &&
2342 		   !e1000_copper_link_autoneg(&sc->hw) &&
2343 		   !e1000_read_phy_reg(&sc->hw, PHY_CONTROL, &phy_tmp)) {
2344 			phy_tmp |= (MII_CR_AUTO_NEG_EN |
2345 				    MII_CR_RESTART_AUTO_NEG);
2346 			e1000_write_phy_reg(&sc->hw, PHY_CONTROL, phy_tmp);
2347 		}
2348 	}
2349 	/* Restart process after EM_SMARTSPEED_MAX iterations */
2350 	if(sc->smartspeed++ == EM_SMARTSPEED_MAX)
2351 		sc->smartspeed = 0;
2352 }
2353 
2354 /*********************************************************************
2355  *
2356  *  Initialize the DMA Coalescing feature
2357  *
2358  **********************************************************************/
2359 static void
2360 igb_init_dmac(struct e1000_softc *sc, u32 pba)
2361 {
2362 	device_t	dev = sc->dev;
2363 	struct e1000_hw *hw = &sc->hw;
2364 	u32 		dmac, reg = ~E1000_DMACR_DMAC_EN;
2365 	u16		hwm;
2366 	u16		max_frame_size;
2367 
2368 	if (hw->mac.type == e1000_i211)
2369 		return;
2370 
2371 	max_frame_size = sc->shared->isc_max_frame_size;
2372 	if (hw->mac.type > e1000_82580) {
2373 
2374 		if (sc->dmac == 0) { /* Disabling it */
2375 			E1000_WRITE_REG(hw, E1000_DMACR, reg);
2376 			return;
2377 		} else
2378 			device_printf(dev, "DMA Coalescing enabled\n");
2379 
2380 		/* Set starting threshold */
2381 		E1000_WRITE_REG(hw, E1000_DMCTXTH, 0);
2382 
2383 		hwm = 64 * pba - max_frame_size / 16;
2384 		if (hwm < 64 * (pba - 6))
2385 			hwm = 64 * (pba - 6);
2386 		reg = E1000_READ_REG(hw, E1000_FCRTC);
2387 		reg &= ~E1000_FCRTC_RTH_COAL_MASK;
2388 		reg |= ((hwm << E1000_FCRTC_RTH_COAL_SHIFT)
2389 		    & E1000_FCRTC_RTH_COAL_MASK);
2390 		E1000_WRITE_REG(hw, E1000_FCRTC, reg);
2391 
2392 
2393 		dmac = pba - max_frame_size / 512;
2394 		if (dmac < pba - 10)
2395 			dmac = pba - 10;
2396 		reg = E1000_READ_REG(hw, E1000_DMACR);
2397 		reg &= ~E1000_DMACR_DMACTHR_MASK;
2398 		reg |= ((dmac << E1000_DMACR_DMACTHR_SHIFT)
2399 		    & E1000_DMACR_DMACTHR_MASK);
2400 
2401 		/* transition to L0x or L1 if available..*/
2402 		reg |= (E1000_DMACR_DMAC_EN | E1000_DMACR_DMAC_LX_MASK);
2403 
2404 		/* Check if status is 2.5Gb backplane connection
2405 		* before configuration of watchdog timer, which is
2406 		* in msec values in 12.8usec intervals
2407 		* watchdog timer= msec values in 32usec intervals
2408 		* for non 2.5Gb connection
2409 		*/
2410 		if (hw->mac.type == e1000_i354) {
2411 			int status = E1000_READ_REG(hw, E1000_STATUS);
2412 			if ((status & E1000_STATUS_2P5_SKU) &&
2413 			    (!(status & E1000_STATUS_2P5_SKU_OVER)))
2414 				reg |= ((sc->dmac * 5) >> 6);
2415 			else
2416 				reg |= (sc->dmac >> 5);
2417 		} else {
2418 			reg |= (sc->dmac >> 5);
2419 		}
2420 
2421 		E1000_WRITE_REG(hw, E1000_DMACR, reg);
2422 
2423 		E1000_WRITE_REG(hw, E1000_DMCRTRH, 0);
2424 
2425 		/* Set the interval before transition */
2426 		reg = E1000_READ_REG(hw, E1000_DMCTLX);
2427 		if (hw->mac.type == e1000_i350)
2428 			reg |= IGB_DMCTLX_DCFLUSH_DIS;
2429 		/*
2430 		** in 2.5Gb connection, TTLX unit is 0.4 usec
2431 		** which is 0x4*2 = 0xA. But delay is still 4 usec
2432 		*/
2433 		if (hw->mac.type == e1000_i354) {
2434 			int status = E1000_READ_REG(hw, E1000_STATUS);
2435 			if ((status & E1000_STATUS_2P5_SKU) &&
2436 			    (!(status & E1000_STATUS_2P5_SKU_OVER)))
2437 				reg |= 0xA;
2438 			else
2439 				reg |= 0x4;
2440 		} else {
2441 			reg |= 0x4;
2442 		}
2443 
2444 		E1000_WRITE_REG(hw, E1000_DMCTLX, reg);
2445 
2446 		/* free space in tx packet buffer to wake from DMA coal */
2447 		E1000_WRITE_REG(hw, E1000_DMCTXTH, (IGB_TXPBSIZE -
2448 		    (2 * max_frame_size)) >> 6);
2449 
2450 		/* make low power state decision controlled by DMA coal */
2451 		reg = E1000_READ_REG(hw, E1000_PCIEMISC);
2452 		reg &= ~E1000_PCIEMISC_LX_DECISION;
2453 		E1000_WRITE_REG(hw, E1000_PCIEMISC, reg);
2454 
2455 	} else if (hw->mac.type == e1000_82580) {
2456 		u32 reg = E1000_READ_REG(hw, E1000_PCIEMISC);
2457 		E1000_WRITE_REG(hw, E1000_PCIEMISC,
2458 		    reg & ~E1000_PCIEMISC_LX_DECISION);
2459 		E1000_WRITE_REG(hw, E1000_DMACR, 0);
2460 	}
2461 }
2462 
2463 /*********************************************************************
2464  *
2465  *  Initialize the hardware to a configuration as specified by the
2466  *  sc structure.
2467  *
2468  **********************************************************************/
2469 static void
2470 em_reset(if_ctx_t ctx)
2471 {
2472 	device_t dev = iflib_get_dev(ctx);
2473 	struct e1000_softc *sc = iflib_get_softc(ctx);
2474 	struct ifnet *ifp = iflib_get_ifp(ctx);
2475 	struct e1000_hw *hw = &sc->hw;
2476 	u32 rx_buffer_size;
2477 	u32 pba;
2478 
2479 	INIT_DEBUGOUT("em_reset: begin");
2480 	/* Let the firmware know the OS is in control */
2481 	em_get_hw_control(sc);
2482 
2483 	/* Set up smart power down as default off on newer adapters. */
2484 	if (!em_smart_pwr_down && (hw->mac.type == e1000_82571 ||
2485 	    hw->mac.type == e1000_82572)) {
2486 		u16 phy_tmp = 0;
2487 
2488 		/* Speed up time to link by disabling smart power down. */
2489 		e1000_read_phy_reg(hw, IGP02E1000_PHY_POWER_MGMT, &phy_tmp);
2490 		phy_tmp &= ~IGP02E1000_PM_SPD;
2491 		e1000_write_phy_reg(hw, IGP02E1000_PHY_POWER_MGMT, phy_tmp);
2492 	}
2493 
2494 	/*
2495 	 * Packet Buffer Allocation (PBA)
2496 	 * Writing PBA sets the receive portion of the buffer
2497 	 * the remainder is used for the transmit buffer.
2498 	 */
2499 	switch (hw->mac.type) {
2500 	/* 82547: Total Packet Buffer is 40K */
2501 	case e1000_82547:
2502 	case e1000_82547_rev_2:
2503 		if (hw->mac.max_frame_size > 8192)
2504 			pba = E1000_PBA_22K; /* 22K for Rx, 18K for Tx */
2505 		else
2506 			pba = E1000_PBA_30K; /* 30K for Rx, 10K for Tx */
2507 		break;
2508 	/* 82571/82572/80003es2lan: Total Packet Buffer is 48K */
2509 	case e1000_82571:
2510 	case e1000_82572:
2511 	case e1000_80003es2lan:
2512 			pba = E1000_PBA_32K; /* 32K for Rx, 16K for Tx */
2513 		break;
2514 	/* 82573: Total Packet Buffer is 32K */
2515 	case e1000_82573:
2516 			pba = E1000_PBA_12K; /* 12K for Rx, 20K for Tx */
2517 		break;
2518 	case e1000_82574:
2519 	case e1000_82583:
2520 			pba = E1000_PBA_20K; /* 20K for Rx, 20K for Tx */
2521 		break;
2522 	case e1000_ich8lan:
2523 		pba = E1000_PBA_8K;
2524 		break;
2525 	case e1000_ich9lan:
2526 	case e1000_ich10lan:
2527 		/* Boost Receive side for jumbo frames */
2528 		if (hw->mac.max_frame_size > 4096)
2529 			pba = E1000_PBA_14K;
2530 		else
2531 			pba = E1000_PBA_10K;
2532 		break;
2533 	case e1000_pchlan:
2534 	case e1000_pch2lan:
2535 	case e1000_pch_lpt:
2536 	case e1000_pch_spt:
2537 	case e1000_pch_cnp:
2538 	case e1000_pch_tgp:
2539 	case e1000_pch_adp:
2540 	case e1000_pch_mtp:
2541 		pba = E1000_PBA_26K;
2542 		break;
2543 	case e1000_82575:
2544 		pba = E1000_PBA_32K;
2545 		break;
2546 	case e1000_82576:
2547 	case e1000_vfadapt:
2548 		pba = E1000_READ_REG(hw, E1000_RXPBS);
2549 		pba &= E1000_RXPBS_SIZE_MASK_82576;
2550 		break;
2551 	case e1000_82580:
2552 	case e1000_i350:
2553 	case e1000_i354:
2554 	case e1000_vfadapt_i350:
2555 		pba = E1000_READ_REG(hw, E1000_RXPBS);
2556 		pba = e1000_rxpbs_adjust_82580(pba);
2557 		break;
2558 	case e1000_i210:
2559 	case e1000_i211:
2560 		pba = E1000_PBA_34K;
2561 		break;
2562 	default:
2563 		/* Remaining devices assumed to have a Packet Buffer of 64K. */
2564 		if (hw->mac.max_frame_size > 8192)
2565 			pba = E1000_PBA_40K; /* 40K for Rx, 24K for Tx */
2566 		else
2567 			pba = E1000_PBA_48K; /* 48K for Rx, 16K for Tx */
2568 	}
2569 
2570 	/* Special needs in case of Jumbo frames */
2571 	if ((hw->mac.type == e1000_82575) && (ifp->if_mtu > ETHERMTU)) {
2572 		u32 tx_space, min_tx, min_rx;
2573 		pba = E1000_READ_REG(hw, E1000_PBA);
2574 		tx_space = pba >> 16;
2575 		pba &= 0xffff;
2576 		min_tx = (hw->mac.max_frame_size +
2577 		    sizeof(struct e1000_tx_desc) - ETHERNET_FCS_SIZE) * 2;
2578 		min_tx = roundup2(min_tx, 1024);
2579 		min_tx >>= 10;
2580 		min_rx = hw->mac.max_frame_size;
2581 		min_rx = roundup2(min_rx, 1024);
2582 		min_rx >>= 10;
2583 		if (tx_space < min_tx &&
2584 		    ((min_tx - tx_space) < pba)) {
2585 			pba = pba - (min_tx - tx_space);
2586 			/*
2587 			 * if short on rx space, rx wins
2588 			 * and must trump tx adjustment
2589 			 */
2590 			if (pba < min_rx)
2591 				pba = min_rx;
2592 		}
2593 		E1000_WRITE_REG(hw, E1000_PBA, pba);
2594 	}
2595 
2596 	if (hw->mac.type < igb_mac_min)
2597 		E1000_WRITE_REG(hw, E1000_PBA, pba);
2598 
2599 	INIT_DEBUGOUT1("em_reset: pba=%dK",pba);
2600 
2601 	/*
2602 	 * These parameters control the automatic generation (Tx) and
2603 	 * response (Rx) to Ethernet PAUSE frames.
2604 	 * - High water mark should allow for at least two frames to be
2605 	 *   received after sending an XOFF.
2606 	 * - Low water mark works best when it is very near the high water mark.
2607 	 *   This allows the receiver to restart by sending XON when it has
2608 	 *   drained a bit. Here we use an arbitrary value of 1500 which will
2609 	 *   restart after one full frame is pulled from the buffer. There
2610 	 *   could be several smaller frames in the buffer and if so they will
2611 	 *   not trigger the XON until their total number reduces the buffer
2612 	 *   by 1500.
2613 	 * - The pause time is fairly large at 1000 x 512ns = 512 usec.
2614 	 */
2615 	rx_buffer_size = (pba & 0xffff) << 10;
2616 	hw->fc.high_water = rx_buffer_size -
2617 	    roundup2(hw->mac.max_frame_size, 1024);
2618 	hw->fc.low_water = hw->fc.high_water - 1500;
2619 
2620 	if (sc->fc) /* locally set flow control value? */
2621 		hw->fc.requested_mode = sc->fc;
2622 	else
2623 		hw->fc.requested_mode = e1000_fc_full;
2624 
2625 	if (hw->mac.type == e1000_80003es2lan)
2626 		hw->fc.pause_time = 0xFFFF;
2627 	else
2628 		hw->fc.pause_time = EM_FC_PAUSE_TIME;
2629 
2630 	hw->fc.send_xon = true;
2631 
2632 	/* Device specific overrides/settings */
2633 	switch (hw->mac.type) {
2634 	case e1000_pchlan:
2635 		/* Workaround: no TX flow ctrl for PCH */
2636 		hw->fc.requested_mode = e1000_fc_rx_pause;
2637 		hw->fc.pause_time = 0xFFFF; /* override */
2638 		if (if_getmtu(ifp) > ETHERMTU) {
2639 			hw->fc.high_water = 0x3500;
2640 			hw->fc.low_water = 0x1500;
2641 		} else {
2642 			hw->fc.high_water = 0x5000;
2643 			hw->fc.low_water = 0x3000;
2644 		}
2645 		hw->fc.refresh_time = 0x1000;
2646 		break;
2647 	case e1000_pch2lan:
2648 	case e1000_pch_lpt:
2649 	case e1000_pch_spt:
2650 	case e1000_pch_cnp:
2651 	case e1000_pch_tgp:
2652 	case e1000_pch_adp:
2653 	case e1000_pch_mtp:
2654 		hw->fc.high_water = 0x5C20;
2655 		hw->fc.low_water = 0x5048;
2656 		hw->fc.pause_time = 0x0650;
2657 		hw->fc.refresh_time = 0x0400;
2658 		/* Jumbos need adjusted PBA */
2659 		if (if_getmtu(ifp) > ETHERMTU)
2660 			E1000_WRITE_REG(hw, E1000_PBA, 12);
2661 		else
2662 			E1000_WRITE_REG(hw, E1000_PBA, 26);
2663 		break;
2664 	case e1000_82575:
2665 	case e1000_82576:
2666 		/* 8-byte granularity */
2667 		hw->fc.low_water = hw->fc.high_water - 8;
2668 		break;
2669 	case e1000_82580:
2670 	case e1000_i350:
2671 	case e1000_i354:
2672 	case e1000_i210:
2673 	case e1000_i211:
2674 	case e1000_vfadapt:
2675 	case e1000_vfadapt_i350:
2676 		/* 16-byte granularity */
2677 		hw->fc.low_water = hw->fc.high_water - 16;
2678 		break;
2679 	case e1000_ich9lan:
2680 	case e1000_ich10lan:
2681 		if (if_getmtu(ifp) > ETHERMTU) {
2682 			hw->fc.high_water = 0x2800;
2683 			hw->fc.low_water = hw->fc.high_water - 8;
2684 			break;
2685 		}
2686 		/* FALLTHROUGH */
2687 	default:
2688 		if (hw->mac.type == e1000_80003es2lan)
2689 			hw->fc.pause_time = 0xFFFF;
2690 		break;
2691 	}
2692 
2693 	/* Issue a global reset */
2694 	e1000_reset_hw(hw);
2695 	if (hw->mac.type >= igb_mac_min) {
2696 		E1000_WRITE_REG(hw, E1000_WUC, 0);
2697 	} else {
2698 		E1000_WRITE_REG(hw, E1000_WUFC, 0);
2699 		em_disable_aspm(sc);
2700 	}
2701 	if (sc->flags & IGB_MEDIA_RESET) {
2702 		e1000_setup_init_funcs(hw, true);
2703 		e1000_get_bus_info(hw);
2704 		sc->flags &= ~IGB_MEDIA_RESET;
2705 	}
2706 	/* and a re-init */
2707 	if (e1000_init_hw(hw) < 0) {
2708 		device_printf(dev, "Hardware Initialization Failed\n");
2709 		return;
2710 	}
2711 	if (hw->mac.type >= igb_mac_min)
2712 		igb_init_dmac(sc, pba);
2713 
2714 	E1000_WRITE_REG(hw, E1000_VET, ETHERTYPE_VLAN);
2715 	e1000_get_phy_info(hw);
2716 	e1000_check_for_link(hw);
2717 }
2718 
2719 /*
2720  * Initialise the RSS mapping for NICs that support multiple transmit/
2721  * receive rings.
2722  */
2723 
2724 #define RSSKEYLEN 10
2725 static void
2726 em_initialize_rss_mapping(struct e1000_softc *sc)
2727 {
2728 	uint8_t  rss_key[4 * RSSKEYLEN];
2729 	uint32_t reta = 0;
2730 	struct e1000_hw	*hw = &sc->hw;
2731 	int i;
2732 
2733 	/*
2734 	 * Configure RSS key
2735 	 */
2736 	arc4rand(rss_key, sizeof(rss_key), 0);
2737 	for (i = 0; i < RSSKEYLEN; ++i) {
2738 		uint32_t rssrk = 0;
2739 
2740 		rssrk = EM_RSSRK_VAL(rss_key, i);
2741 		E1000_WRITE_REG(hw,E1000_RSSRK(i), rssrk);
2742 	}
2743 
2744 	/*
2745 	 * Configure RSS redirect table in following fashion:
2746 	 * (hash & ring_cnt_mask) == rdr_table[(hash & rdr_table_mask)]
2747 	 */
2748 	for (i = 0; i < sizeof(reta); ++i) {
2749 		uint32_t q;
2750 
2751 		q = (i % sc->rx_num_queues) << 7;
2752 		reta |= q << (8 * i);
2753 	}
2754 
2755 	for (i = 0; i < 32; ++i)
2756 		E1000_WRITE_REG(hw, E1000_RETA(i), reta);
2757 
2758 	E1000_WRITE_REG(hw, E1000_MRQC, E1000_MRQC_RSS_ENABLE_2Q |
2759 			E1000_MRQC_RSS_FIELD_IPV4_TCP |
2760 			E1000_MRQC_RSS_FIELD_IPV4 |
2761 			E1000_MRQC_RSS_FIELD_IPV6_TCP_EX |
2762 			E1000_MRQC_RSS_FIELD_IPV6_EX |
2763 			E1000_MRQC_RSS_FIELD_IPV6);
2764 }
2765 
2766 static void
2767 igb_initialize_rss_mapping(struct e1000_softc *sc)
2768 {
2769 	struct e1000_hw *hw = &sc->hw;
2770 	int i;
2771 	int queue_id;
2772 	u32 reta;
2773 	u32 rss_key[10], mrqc, shift = 0;
2774 
2775 	/* XXX? */
2776 	if (hw->mac.type == e1000_82575)
2777 		shift = 6;
2778 
2779 	/*
2780 	 * The redirection table controls which destination
2781 	 * queue each bucket redirects traffic to.
2782 	 * Each DWORD represents four queues, with the LSB
2783 	 * being the first queue in the DWORD.
2784 	 *
2785 	 * This just allocates buckets to queues using round-robin
2786 	 * allocation.
2787 	 *
2788 	 * NOTE: It Just Happens to line up with the default
2789 	 * RSS allocation method.
2790 	 */
2791 
2792 	/* Warning FM follows */
2793 	reta = 0;
2794 	for (i = 0; i < 128; i++) {
2795 #ifdef RSS
2796 		queue_id = rss_get_indirection_to_bucket(i);
2797 		/*
2798 		 * If we have more queues than buckets, we'll
2799 		 * end up mapping buckets to a subset of the
2800 		 * queues.
2801 		 *
2802 		 * If we have more buckets than queues, we'll
2803 		 * end up instead assigning multiple buckets
2804 		 * to queues.
2805 		 *
2806 		 * Both are suboptimal, but we need to handle
2807 		 * the case so we don't go out of bounds
2808 		 * indexing arrays and such.
2809 		 */
2810 		queue_id = queue_id % sc->rx_num_queues;
2811 #else
2812 		queue_id = (i % sc->rx_num_queues);
2813 #endif
2814 		/* Adjust if required */
2815 		queue_id = queue_id << shift;
2816 
2817 		/*
2818 		 * The low 8 bits are for hash value (n+0);
2819 		 * The next 8 bits are for hash value (n+1), etc.
2820 		 */
2821 		reta = reta >> 8;
2822 		reta = reta | ( ((uint32_t) queue_id) << 24);
2823 		if ((i & 3) == 3) {
2824 			E1000_WRITE_REG(hw, E1000_RETA(i >> 2), reta);
2825 			reta = 0;
2826 		}
2827 	}
2828 
2829 	/* Now fill in hash table */
2830 
2831 	/*
2832 	 * MRQC: Multiple Receive Queues Command
2833 	 * Set queuing to RSS control, number depends on the device.
2834 	 */
2835 	mrqc = E1000_MRQC_ENABLE_RSS_MQ;
2836 
2837 #ifdef RSS
2838 	/* XXX ew typecasting */
2839 	rss_getkey((uint8_t *) &rss_key);
2840 #else
2841 	arc4rand(&rss_key, sizeof(rss_key), 0);
2842 #endif
2843 	for (i = 0; i < 10; i++)
2844 		E1000_WRITE_REG_ARRAY(hw, E1000_RSSRK(0), i, rss_key[i]);
2845 
2846 	/*
2847 	 * Configure the RSS fields to hash upon.
2848 	 */
2849 	mrqc |= (E1000_MRQC_RSS_FIELD_IPV4 |
2850 	    E1000_MRQC_RSS_FIELD_IPV4_TCP);
2851 	mrqc |= (E1000_MRQC_RSS_FIELD_IPV6 |
2852 	    E1000_MRQC_RSS_FIELD_IPV6_TCP);
2853 	mrqc |=( E1000_MRQC_RSS_FIELD_IPV4_UDP |
2854 	    E1000_MRQC_RSS_FIELD_IPV6_UDP);
2855 	mrqc |=( E1000_MRQC_RSS_FIELD_IPV6_UDP_EX |
2856 	    E1000_MRQC_RSS_FIELD_IPV6_TCP_EX);
2857 
2858 	E1000_WRITE_REG(hw, E1000_MRQC, mrqc);
2859 }
2860 
2861 /*********************************************************************
2862  *
2863  *  Setup networking device structure and register interface media.
2864  *
2865  **********************************************************************/
2866 static int
2867 em_setup_interface(if_ctx_t ctx)
2868 {
2869 	struct ifnet *ifp = iflib_get_ifp(ctx);
2870 	struct e1000_softc *sc = iflib_get_softc(ctx);
2871 	if_softc_ctx_t scctx = sc->shared;
2872 
2873 	INIT_DEBUGOUT("em_setup_interface: begin");
2874 
2875 	/* Single Queue */
2876 	if (sc->tx_num_queues == 1) {
2877 		if_setsendqlen(ifp, scctx->isc_ntxd[0] - 1);
2878 		if_setsendqready(ifp);
2879 	}
2880 
2881 	/*
2882 	 * Specify the media types supported by this adapter and register
2883 	 * callbacks to update media and link information
2884 	 */
2885 	if (sc->hw.phy.media_type == e1000_media_type_fiber ||
2886 	    sc->hw.phy.media_type == e1000_media_type_internal_serdes) {
2887 		u_char fiber_type = IFM_1000_SX;	/* default type */
2888 
2889 		if (sc->hw.mac.type == e1000_82545)
2890 			fiber_type = IFM_1000_LX;
2891 		ifmedia_add(sc->media, IFM_ETHER | fiber_type | IFM_FDX, 0, NULL);
2892 		ifmedia_add(sc->media, IFM_ETHER | fiber_type, 0, NULL);
2893 	} else {
2894 		ifmedia_add(sc->media, IFM_ETHER | IFM_10_T, 0, NULL);
2895 		ifmedia_add(sc->media, IFM_ETHER | IFM_10_T | IFM_FDX, 0, NULL);
2896 		ifmedia_add(sc->media, IFM_ETHER | IFM_100_TX, 0, NULL);
2897 		ifmedia_add(sc->media, IFM_ETHER | IFM_100_TX | IFM_FDX, 0, NULL);
2898 		if (sc->hw.phy.type != e1000_phy_ife) {
2899 			ifmedia_add(sc->media, IFM_ETHER | IFM_1000_T | IFM_FDX, 0, NULL);
2900 			ifmedia_add(sc->media, IFM_ETHER | IFM_1000_T, 0, NULL);
2901 		}
2902 	}
2903 	ifmedia_add(sc->media, IFM_ETHER | IFM_AUTO, 0, NULL);
2904 	ifmedia_set(sc->media, IFM_ETHER | IFM_AUTO);
2905 	return (0);
2906 }
2907 
2908 static int
2909 em_if_tx_queues_alloc(if_ctx_t ctx, caddr_t *vaddrs, uint64_t *paddrs, int ntxqs, int ntxqsets)
2910 {
2911 	struct e1000_softc *sc = iflib_get_softc(ctx);
2912 	if_softc_ctx_t scctx = sc->shared;
2913 	int error = E1000_SUCCESS;
2914 	struct em_tx_queue *que;
2915 	int i, j;
2916 
2917 	MPASS(sc->tx_num_queues > 0);
2918 	MPASS(sc->tx_num_queues == ntxqsets);
2919 
2920 	/* First allocate the top level queue structs */
2921 	if (!(sc->tx_queues =
2922 	    (struct em_tx_queue *) malloc(sizeof(struct em_tx_queue) *
2923 	    sc->tx_num_queues, M_DEVBUF, M_NOWAIT | M_ZERO))) {
2924 		device_printf(iflib_get_dev(ctx), "Unable to allocate queue memory\n");
2925 		return(ENOMEM);
2926 	}
2927 
2928 	for (i = 0, que = sc->tx_queues; i < sc->tx_num_queues; i++, que++) {
2929 		/* Set up some basics */
2930 
2931 		struct tx_ring *txr = &que->txr;
2932 		txr->sc = que->sc = sc;
2933 		que->me = txr->me =  i;
2934 
2935 		/* Allocate report status array */
2936 		if (!(txr->tx_rsq = (qidx_t *) malloc(sizeof(qidx_t) * scctx->isc_ntxd[0], M_DEVBUF, M_NOWAIT | M_ZERO))) {
2937 			device_printf(iflib_get_dev(ctx), "failed to allocate rs_idxs memory\n");
2938 			error = ENOMEM;
2939 			goto fail;
2940 		}
2941 		for (j = 0; j < scctx->isc_ntxd[0]; j++)
2942 			txr->tx_rsq[j] = QIDX_INVALID;
2943 		/* get the virtual and physical address of the hardware queues */
2944 		txr->tx_base = (struct e1000_tx_desc *)vaddrs[i*ntxqs];
2945 		txr->tx_paddr = paddrs[i*ntxqs];
2946 	}
2947 
2948 	if (bootverbose)
2949 		device_printf(iflib_get_dev(ctx),
2950 		    "allocated for %d tx_queues\n", sc->tx_num_queues);
2951 	return (0);
2952 fail:
2953 	em_if_queues_free(ctx);
2954 	return (error);
2955 }
2956 
2957 static int
2958 em_if_rx_queues_alloc(if_ctx_t ctx, caddr_t *vaddrs, uint64_t *paddrs, int nrxqs, int nrxqsets)
2959 {
2960 	struct e1000_softc *sc = iflib_get_softc(ctx);
2961 	int error = E1000_SUCCESS;
2962 	struct em_rx_queue *que;
2963 	int i;
2964 
2965 	MPASS(sc->rx_num_queues > 0);
2966 	MPASS(sc->rx_num_queues == nrxqsets);
2967 
2968 	/* First allocate the top level queue structs */
2969 	if (!(sc->rx_queues =
2970 	    (struct em_rx_queue *) malloc(sizeof(struct em_rx_queue) *
2971 	    sc->rx_num_queues, M_DEVBUF, M_NOWAIT | M_ZERO))) {
2972 		device_printf(iflib_get_dev(ctx), "Unable to allocate queue memory\n");
2973 		error = ENOMEM;
2974 		goto fail;
2975 	}
2976 
2977 	for (i = 0, que = sc->rx_queues; i < nrxqsets; i++, que++) {
2978 		/* Set up some basics */
2979 		struct rx_ring *rxr = &que->rxr;
2980 		rxr->sc = que->sc = sc;
2981 		rxr->que = que;
2982 		que->me = rxr->me =  i;
2983 
2984 		/* get the virtual and physical address of the hardware queues */
2985 		rxr->rx_base = (union e1000_rx_desc_extended *)vaddrs[i*nrxqs];
2986 		rxr->rx_paddr = paddrs[i*nrxqs];
2987 	}
2988 
2989 	if (bootverbose)
2990 		device_printf(iflib_get_dev(ctx),
2991 		    "allocated for %d rx_queues\n", sc->rx_num_queues);
2992 
2993 	return (0);
2994 fail:
2995 	em_if_queues_free(ctx);
2996 	return (error);
2997 }
2998 
2999 static void
3000 em_if_queues_free(if_ctx_t ctx)
3001 {
3002 	struct e1000_softc *sc = iflib_get_softc(ctx);
3003 	struct em_tx_queue *tx_que = sc->tx_queues;
3004 	struct em_rx_queue *rx_que = sc->rx_queues;
3005 
3006 	if (tx_que != NULL) {
3007 		for (int i = 0; i < sc->tx_num_queues; i++, tx_que++) {
3008 			struct tx_ring *txr = &tx_que->txr;
3009 			if (txr->tx_rsq == NULL)
3010 				break;
3011 
3012 			free(txr->tx_rsq, M_DEVBUF);
3013 			txr->tx_rsq = NULL;
3014 		}
3015 		free(sc->tx_queues, M_DEVBUF);
3016 		sc->tx_queues = NULL;
3017 	}
3018 
3019 	if (rx_que != NULL) {
3020 		free(sc->rx_queues, M_DEVBUF);
3021 		sc->rx_queues = NULL;
3022 	}
3023 }
3024 
3025 /*********************************************************************
3026  *
3027  *  Enable transmit unit.
3028  *
3029  **********************************************************************/
3030 static void
3031 em_initialize_transmit_unit(if_ctx_t ctx)
3032 {
3033 	struct e1000_softc *sc = iflib_get_softc(ctx);
3034 	if_softc_ctx_t scctx = sc->shared;
3035 	struct em_tx_queue *que;
3036 	struct tx_ring	*txr;
3037 	struct e1000_hw	*hw = &sc->hw;
3038 	u32 tctl, txdctl = 0, tarc, tipg = 0;
3039 
3040 	INIT_DEBUGOUT("em_initialize_transmit_unit: begin");
3041 
3042 	for (int i = 0; i < sc->tx_num_queues; i++, txr++) {
3043 		u64 bus_addr;
3044 		caddr_t offp, endp;
3045 
3046 		que = &sc->tx_queues[i];
3047 		txr = &que->txr;
3048 		bus_addr = txr->tx_paddr;
3049 
3050 		/* Clear checksum offload context. */
3051 		offp = (caddr_t)&txr->csum_flags;
3052 		endp = (caddr_t)(txr + 1);
3053 		bzero(offp, endp - offp);
3054 
3055 		/* Base and Len of TX Ring */
3056 		E1000_WRITE_REG(hw, E1000_TDLEN(i),
3057 		    scctx->isc_ntxd[0] * sizeof(struct e1000_tx_desc));
3058 		E1000_WRITE_REG(hw, E1000_TDBAH(i),
3059 		    (u32)(bus_addr >> 32));
3060 		E1000_WRITE_REG(hw, E1000_TDBAL(i),
3061 		    (u32)bus_addr);
3062 		/* Init the HEAD/TAIL indices */
3063 		E1000_WRITE_REG(hw, E1000_TDT(i), 0);
3064 		E1000_WRITE_REG(hw, E1000_TDH(i), 0);
3065 
3066 		HW_DEBUGOUT2("Base = %x, Length = %x\n",
3067 		    E1000_READ_REG(hw, E1000_TDBAL(i)),
3068 		    E1000_READ_REG(hw, E1000_TDLEN(i)));
3069 
3070 		txdctl = 0; /* clear txdctl */
3071 		txdctl |= 0x1f; /* PTHRESH */
3072 		txdctl |= 1 << 8; /* HTHRESH */
3073 		txdctl |= 1 << 16;/* WTHRESH */
3074 		txdctl |= 1 << 22; /* Reserved bit 22 must always be 1 */
3075 		txdctl |= E1000_TXDCTL_GRAN;
3076 		txdctl |= 1 << 25; /* LWTHRESH */
3077 
3078 		E1000_WRITE_REG(hw, E1000_TXDCTL(i), txdctl);
3079 	}
3080 
3081 	/* Set the default values for the Tx Inter Packet Gap timer */
3082 	switch (hw->mac.type) {
3083 	case e1000_80003es2lan:
3084 		tipg = DEFAULT_82543_TIPG_IPGR1;
3085 		tipg |= DEFAULT_80003ES2LAN_TIPG_IPGR2 <<
3086 		    E1000_TIPG_IPGR2_SHIFT;
3087 		break;
3088 	case e1000_82542:
3089 		tipg = DEFAULT_82542_TIPG_IPGT;
3090 		tipg |= DEFAULT_82542_TIPG_IPGR1 << E1000_TIPG_IPGR1_SHIFT;
3091 		tipg |= DEFAULT_82542_TIPG_IPGR2 << E1000_TIPG_IPGR2_SHIFT;
3092 		break;
3093 	default:
3094 		if (hw->phy.media_type == e1000_media_type_fiber ||
3095 		    hw->phy.media_type == e1000_media_type_internal_serdes)
3096 			tipg = DEFAULT_82543_TIPG_IPGT_FIBER;
3097 		else
3098 			tipg = DEFAULT_82543_TIPG_IPGT_COPPER;
3099 		tipg |= DEFAULT_82543_TIPG_IPGR1 << E1000_TIPG_IPGR1_SHIFT;
3100 		tipg |= DEFAULT_82543_TIPG_IPGR2 << E1000_TIPG_IPGR2_SHIFT;
3101 	}
3102 
3103 	E1000_WRITE_REG(hw, E1000_TIPG, tipg);
3104 	E1000_WRITE_REG(hw, E1000_TIDV, sc->tx_int_delay.value);
3105 
3106 	if(hw->mac.type >= e1000_82540)
3107 		E1000_WRITE_REG(hw, E1000_TADV,
3108 		    sc->tx_abs_int_delay.value);
3109 
3110 	if (hw->mac.type == e1000_82571 || hw->mac.type == e1000_82572) {
3111 		tarc = E1000_READ_REG(hw, E1000_TARC(0));
3112 		tarc |= TARC_SPEED_MODE_BIT;
3113 		E1000_WRITE_REG(hw, E1000_TARC(0), tarc);
3114 	} else if (hw->mac.type == e1000_80003es2lan) {
3115 		/* errata: program both queues to unweighted RR */
3116 		tarc = E1000_READ_REG(hw, E1000_TARC(0));
3117 		tarc |= 1;
3118 		E1000_WRITE_REG(hw, E1000_TARC(0), tarc);
3119 		tarc = E1000_READ_REG(hw, E1000_TARC(1));
3120 		tarc |= 1;
3121 		E1000_WRITE_REG(hw, E1000_TARC(1), tarc);
3122 	} else if (hw->mac.type == e1000_82574) {
3123 		tarc = E1000_READ_REG(hw, E1000_TARC(0));
3124 		tarc |= TARC_ERRATA_BIT;
3125 		if ( sc->tx_num_queues > 1) {
3126 			tarc |= (TARC_COMPENSATION_MODE | TARC_MQ_FIX);
3127 			E1000_WRITE_REG(hw, E1000_TARC(0), tarc);
3128 			E1000_WRITE_REG(hw, E1000_TARC(1), tarc);
3129 		} else
3130 			E1000_WRITE_REG(hw, E1000_TARC(0), tarc);
3131 	}
3132 
3133 	if (sc->tx_int_delay.value > 0)
3134 		sc->txd_cmd |= E1000_TXD_CMD_IDE;
3135 
3136 	/* Program the Transmit Control Register */
3137 	tctl = E1000_READ_REG(hw, E1000_TCTL);
3138 	tctl &= ~E1000_TCTL_CT;
3139 	tctl |= (E1000_TCTL_PSP | E1000_TCTL_RTLC | E1000_TCTL_EN |
3140 		   (E1000_COLLISION_THRESHOLD << E1000_CT_SHIFT));
3141 
3142 	if (hw->mac.type >= e1000_82571)
3143 		tctl |= E1000_TCTL_MULR;
3144 
3145 	/* This write will effectively turn on the transmit unit. */
3146 	E1000_WRITE_REG(hw, E1000_TCTL, tctl);
3147 
3148 	/* SPT and KBL errata workarounds */
3149 	if (hw->mac.type == e1000_pch_spt) {
3150 		u32 reg;
3151 		reg = E1000_READ_REG(hw, E1000_IOSFPC);
3152 		reg |= E1000_RCTL_RDMTS_HEX;
3153 		E1000_WRITE_REG(hw, E1000_IOSFPC, reg);
3154 		/* i218-i219 Specification Update 1.5.4.5 */
3155 		reg = E1000_READ_REG(hw, E1000_TARC(0));
3156 		reg &= ~E1000_TARC0_CB_MULTIQ_3_REQ;
3157 		reg |= E1000_TARC0_CB_MULTIQ_2_REQ;
3158 		E1000_WRITE_REG(hw, E1000_TARC(0), reg);
3159 	}
3160 }
3161 
3162 /*********************************************************************
3163  *
3164  *  Enable receive unit.
3165  *
3166  **********************************************************************/
3167 #define BSIZEPKT_ROUNDUP ((1<<E1000_SRRCTL_BSIZEPKT_SHIFT)-1)
3168 
3169 static void
3170 em_initialize_receive_unit(if_ctx_t ctx)
3171 {
3172 	struct e1000_softc *sc = iflib_get_softc(ctx);
3173 	if_softc_ctx_t scctx = sc->shared;
3174 	struct ifnet *ifp = iflib_get_ifp(ctx);
3175 	struct e1000_hw	*hw = &sc->hw;
3176 	struct em_rx_queue *que;
3177 	int i;
3178 	uint32_t rctl, rxcsum;
3179 
3180 	INIT_DEBUGOUT("em_initialize_receive_units: begin");
3181 
3182 	/*
3183 	 * Make sure receives are disabled while setting
3184 	 * up the descriptor ring
3185 	 */
3186 	rctl = E1000_READ_REG(hw, E1000_RCTL);
3187 	/* Do not disable if ever enabled on this hardware */
3188 	if ((hw->mac.type != e1000_82574) && (hw->mac.type != e1000_82583))
3189 		E1000_WRITE_REG(hw, E1000_RCTL, rctl & ~E1000_RCTL_EN);
3190 
3191 	/* Setup the Receive Control Register */
3192 	rctl &= ~(3 << E1000_RCTL_MO_SHIFT);
3193 	rctl |= E1000_RCTL_EN | E1000_RCTL_BAM |
3194 	    E1000_RCTL_LBM_NO | E1000_RCTL_RDMTS_HALF |
3195 	    (hw->mac.mc_filter_type << E1000_RCTL_MO_SHIFT);
3196 
3197 	/* Do not store bad packets */
3198 	rctl &= ~E1000_RCTL_SBP;
3199 
3200 	/* Enable Long Packet receive */
3201 	if (if_getmtu(ifp) > ETHERMTU)
3202 		rctl |= E1000_RCTL_LPE;
3203 	else
3204 		rctl &= ~E1000_RCTL_LPE;
3205 
3206 	/* Strip the CRC */
3207 	if (!em_disable_crc_stripping)
3208 		rctl |= E1000_RCTL_SECRC;
3209 
3210 	if (hw->mac.type >= e1000_82540) {
3211 		E1000_WRITE_REG(hw, E1000_RADV,
3212 		    sc->rx_abs_int_delay.value);
3213 
3214 		/*
3215 		 * Set the interrupt throttling rate. Value is calculated
3216 		 * as DEFAULT_ITR = 1/(MAX_INTS_PER_SEC * 256ns)
3217 		 */
3218 		E1000_WRITE_REG(hw, E1000_ITR, DEFAULT_ITR);
3219 	}
3220 	E1000_WRITE_REG(hw, E1000_RDTR, sc->rx_int_delay.value);
3221 
3222 	if (hw->mac.type >= em_mac_min) {
3223 		uint32_t rfctl;
3224 		/* Use extended rx descriptor formats */
3225 		rfctl = E1000_READ_REG(hw, E1000_RFCTL);
3226 		rfctl |= E1000_RFCTL_EXTEN;
3227 
3228 		/*
3229 		 * When using MSI-X interrupts we need to throttle
3230 		 * using the EITR register (82574 only)
3231 		 */
3232 		if (hw->mac.type == e1000_82574) {
3233 			for (int i = 0; i < 4; i++)
3234 				E1000_WRITE_REG(hw, E1000_EITR_82574(i),
3235 				    DEFAULT_ITR);
3236 			/* Disable accelerated acknowledge */
3237 			rfctl |= E1000_RFCTL_ACK_DIS;
3238 		}
3239 		E1000_WRITE_REG(hw, E1000_RFCTL, rfctl);
3240 	}
3241 
3242 	/* Set up L3 and L4 csum Rx descriptor offloads */
3243 	rxcsum = E1000_READ_REG(hw, E1000_RXCSUM);
3244 	if (if_getcapenable(ifp) & IFCAP_RXCSUM) {
3245 		rxcsum |= E1000_RXCSUM_TUOFL | E1000_RXCSUM_IPOFL;
3246 		if (hw->mac.type > e1000_82575)
3247 			rxcsum |= E1000_RXCSUM_CRCOFL;
3248 		else if (hw->mac.type < em_mac_min &&
3249 		    if_getcapenable(ifp) & IFCAP_HWCSUM_IPV6)
3250 			rxcsum |= E1000_RXCSUM_IPV6OFL;
3251 	} else {
3252 		rxcsum &= ~(E1000_RXCSUM_IPOFL | E1000_RXCSUM_TUOFL);
3253 		if (hw->mac.type > e1000_82575)
3254 			rxcsum &= ~E1000_RXCSUM_CRCOFL;
3255 		else if (hw->mac.type < em_mac_min)
3256 			rxcsum &= ~E1000_RXCSUM_IPV6OFL;
3257 	}
3258 
3259 	if (sc->rx_num_queues > 1) {
3260 		/* RSS hash needed in the Rx descriptor */
3261 		rxcsum |= E1000_RXCSUM_PCSD;
3262 
3263 		if (hw->mac.type >= igb_mac_min)
3264 			igb_initialize_rss_mapping(sc);
3265 		else
3266 			em_initialize_rss_mapping(sc);
3267 	}
3268 	E1000_WRITE_REG(hw, E1000_RXCSUM, rxcsum);
3269 
3270 	/*
3271 	 * XXX TEMPORARY WORKAROUND: on some systems with 82573
3272 	 * long latencies are observed, like Lenovo X60. This
3273 	 * change eliminates the problem, but since having positive
3274 	 * values in RDTR is a known source of problems on other
3275 	 * platforms another solution is being sought.
3276 	 */
3277 	if (hw->mac.type == e1000_82573)
3278 		E1000_WRITE_REG(hw, E1000_RDTR, 0x20);
3279 
3280 	for (i = 0, que = sc->rx_queues; i < sc->rx_num_queues; i++, que++) {
3281 		struct rx_ring *rxr = &que->rxr;
3282 		/* Setup the Base and Length of the Rx Descriptor Ring */
3283 		u64 bus_addr = rxr->rx_paddr;
3284 #if 0
3285 		u32 rdt = sc->rx_num_queues -1;  /* default */
3286 #endif
3287 
3288 		E1000_WRITE_REG(hw, E1000_RDLEN(i),
3289 		    scctx->isc_nrxd[0] * sizeof(union e1000_rx_desc_extended));
3290 		E1000_WRITE_REG(hw, E1000_RDBAH(i), (u32)(bus_addr >> 32));
3291 		E1000_WRITE_REG(hw, E1000_RDBAL(i), (u32)bus_addr);
3292 		/* Setup the Head and Tail Descriptor Pointers */
3293 		E1000_WRITE_REG(hw, E1000_RDH(i), 0);
3294 		E1000_WRITE_REG(hw, E1000_RDT(i), 0);
3295 	}
3296 
3297 	/*
3298 	 * Set PTHRESH for improved jumbo performance
3299 	 * According to 10.2.5.11 of Intel 82574 Datasheet,
3300 	 * RXDCTL(1) is written whenever RXDCTL(0) is written.
3301 	 * Only write to RXDCTL(1) if there is a need for different
3302 	 * settings.
3303 	 */
3304 	if ((hw->mac.type == e1000_ich9lan || hw->mac.type == e1000_pch2lan ||
3305 	    hw->mac.type == e1000_ich10lan) && if_getmtu(ifp) > ETHERMTU) {
3306 		u32 rxdctl = E1000_READ_REG(hw, E1000_RXDCTL(0));
3307 		E1000_WRITE_REG(hw, E1000_RXDCTL(0), rxdctl | 3);
3308 	} else if (hw->mac.type == e1000_82574) {
3309 		for (int i = 0; i < sc->rx_num_queues; i++) {
3310 			u32 rxdctl = E1000_READ_REG(hw, E1000_RXDCTL(i));
3311 			rxdctl |= 0x20; /* PTHRESH */
3312 			rxdctl |= 4 << 8; /* HTHRESH */
3313 			rxdctl |= 4 << 16;/* WTHRESH */
3314 			rxdctl |= 1 << 24; /* Switch to granularity */
3315 			E1000_WRITE_REG(hw, E1000_RXDCTL(i), rxdctl);
3316 		}
3317 	} else if (hw->mac.type >= igb_mac_min) {
3318 		u32 psize, srrctl = 0;
3319 
3320 		if (if_getmtu(ifp) > ETHERMTU) {
3321 			psize = scctx->isc_max_frame_size;
3322 			/* are we on a vlan? */
3323 			if (ifp->if_vlantrunk != NULL)
3324 				psize += VLAN_TAG_SIZE;
3325 
3326 			if (sc->vf_ifp)
3327 				e1000_rlpml_set_vf(hw, psize);
3328 			else
3329 				E1000_WRITE_REG(hw, E1000_RLPML, psize);
3330 		}
3331 
3332 		/* Set maximum packet buffer len */
3333 		srrctl |= (sc->rx_mbuf_sz + BSIZEPKT_ROUNDUP) >>
3334 		    E1000_SRRCTL_BSIZEPKT_SHIFT;
3335 
3336 		/*
3337 		 * If TX flow control is disabled and there's >1 queue defined,
3338 		 * enable DROP.
3339 		 *
3340 		 * This drops frames rather than hanging the RX MAC for all queues.
3341 		 */
3342 		if ((sc->rx_num_queues > 1) &&
3343 		    (sc->fc == e1000_fc_none ||
3344 		     sc->fc == e1000_fc_rx_pause)) {
3345 			srrctl |= E1000_SRRCTL_DROP_EN;
3346 		}
3347 			/* Setup the Base and Length of the Rx Descriptor Rings */
3348 		for (i = 0, que = sc->rx_queues; i < sc->rx_num_queues; i++, que++) {
3349 			struct rx_ring *rxr = &que->rxr;
3350 			u64 bus_addr = rxr->rx_paddr;
3351 			u32 rxdctl;
3352 
3353 #ifdef notyet
3354 			/* Configure for header split? -- ignore for now */
3355 			rxr->hdr_split = igb_header_split;
3356 #else
3357 			srrctl |= E1000_SRRCTL_DESCTYPE_ADV_ONEBUF;
3358 #endif
3359 
3360 			E1000_WRITE_REG(hw, E1000_RDLEN(i),
3361 					scctx->isc_nrxd[0] * sizeof(struct e1000_rx_desc));
3362 			E1000_WRITE_REG(hw, E1000_RDBAH(i),
3363 					(uint32_t)(bus_addr >> 32));
3364 			E1000_WRITE_REG(hw, E1000_RDBAL(i),
3365 					(uint32_t)bus_addr);
3366 			E1000_WRITE_REG(hw, E1000_SRRCTL(i), srrctl);
3367 			/* Enable this Queue */
3368 			rxdctl = E1000_READ_REG(hw, E1000_RXDCTL(i));
3369 			rxdctl |= E1000_RXDCTL_QUEUE_ENABLE;
3370 			rxdctl &= 0xFFF00000;
3371 			rxdctl |= IGB_RX_PTHRESH;
3372 			rxdctl |= IGB_RX_HTHRESH << 8;
3373 			rxdctl |= IGB_RX_WTHRESH << 16;
3374 			E1000_WRITE_REG(hw, E1000_RXDCTL(i), rxdctl);
3375 		}
3376 	} else if (hw->mac.type >= e1000_pch2lan) {
3377 		if (if_getmtu(ifp) > ETHERMTU)
3378 			e1000_lv_jumbo_workaround_ich8lan(hw, true);
3379 		else
3380 			e1000_lv_jumbo_workaround_ich8lan(hw, false);
3381 	}
3382 
3383 	/* Make sure VLAN Filters are off */
3384 	rctl &= ~E1000_RCTL_VFE;
3385 
3386 	/* Set up packet buffer size, overridden by per queue srrctl on igb */
3387 	if (hw->mac.type < igb_mac_min) {
3388 		if (sc->rx_mbuf_sz > 2048 && sc->rx_mbuf_sz <= 4096)
3389 			rctl |= E1000_RCTL_SZ_4096 | E1000_RCTL_BSEX;
3390 		else if (sc->rx_mbuf_sz > 4096 && sc->rx_mbuf_sz <= 8192)
3391 			rctl |= E1000_RCTL_SZ_8192 | E1000_RCTL_BSEX;
3392 		else if (sc->rx_mbuf_sz > 8192)
3393 			rctl |= E1000_RCTL_SZ_16384 | E1000_RCTL_BSEX;
3394 		else {
3395 			rctl |= E1000_RCTL_SZ_2048;
3396 			rctl &= ~E1000_RCTL_BSEX;
3397 		}
3398 	} else
3399 		rctl |= E1000_RCTL_SZ_2048;
3400 
3401 	/*
3402 	 * rctl bits 11:10 are as follows
3403 	 * lem: reserved
3404 	 * em: DTYPE
3405 	 * igb: reserved
3406 	 * and should be 00 on all of the above
3407 	 */
3408 	rctl &= ~0x00000C00;
3409 
3410 	/* Write out the settings */
3411 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
3412 
3413 	return;
3414 }
3415 
3416 static void
3417 em_if_vlan_register(if_ctx_t ctx, u16 vtag)
3418 {
3419 	struct e1000_softc *sc = iflib_get_softc(ctx);
3420 	u32 index, bit;
3421 
3422 	index = (vtag >> 5) & 0x7F;
3423 	bit = vtag & 0x1F;
3424 	sc->shadow_vfta[index] |= (1 << bit);
3425 	++sc->num_vlans;
3426 	em_if_vlan_filter_write(sc);
3427 }
3428 
3429 static void
3430 em_if_vlan_unregister(if_ctx_t ctx, u16 vtag)
3431 {
3432 	struct e1000_softc *sc = iflib_get_softc(ctx);
3433 	u32 index, bit;
3434 
3435 	index = (vtag >> 5) & 0x7F;
3436 	bit = vtag & 0x1F;
3437 	sc->shadow_vfta[index] &= ~(1 << bit);
3438 	--sc->num_vlans;
3439 	em_if_vlan_filter_write(sc);
3440 }
3441 
3442 static bool
3443 em_if_vlan_filter_capable(if_ctx_t ctx)
3444 {
3445 	if_t ifp = iflib_get_ifp(ctx);
3446 
3447 	if ((if_getcapenable(ifp) & IFCAP_VLAN_HWFILTER) &&
3448 	    !em_disable_crc_stripping)
3449 		return (true);
3450 
3451 	return (false);
3452 }
3453 
3454 static bool
3455 em_if_vlan_filter_used(if_ctx_t ctx)
3456 {
3457 	struct e1000_softc *sc = iflib_get_softc(ctx);
3458 
3459 	if (!em_if_vlan_filter_capable(ctx))
3460 		return (false);
3461 
3462 	for (int i = 0; i < EM_VFTA_SIZE; i++)
3463 		if (sc->shadow_vfta[i] != 0)
3464 			return (true);
3465 
3466 	return (false);
3467 }
3468 
3469 static void
3470 em_if_vlan_filter_enable(struct e1000_softc *sc)
3471 {
3472 	struct e1000_hw *hw = &sc->hw;
3473 	u32 reg;
3474 
3475 	reg = E1000_READ_REG(hw, E1000_RCTL);
3476 	reg &= ~E1000_RCTL_CFIEN;
3477 	reg |= E1000_RCTL_VFE;
3478 	E1000_WRITE_REG(hw, E1000_RCTL, reg);
3479 }
3480 
3481 static void
3482 em_if_vlan_filter_disable(struct e1000_softc *sc)
3483 {
3484 	struct e1000_hw *hw = &sc->hw;
3485 	u32 reg;
3486 
3487 	reg = E1000_READ_REG(hw, E1000_RCTL);
3488 	reg &= ~(E1000_RCTL_VFE | E1000_RCTL_CFIEN);
3489 	E1000_WRITE_REG(hw, E1000_RCTL, reg);
3490 }
3491 
3492 static void
3493 em_if_vlan_filter_write(struct e1000_softc *sc)
3494 {
3495 	struct e1000_hw *hw = &sc->hw;
3496 
3497 	if (sc->vf_ifp)
3498 		return;
3499 
3500 	/* Disable interrupts for lem-class devices during the filter change */
3501 	if (hw->mac.type < em_mac_min)
3502 		em_if_intr_disable(sc->ctx);
3503 
3504 	for (int i = 0; i < EM_VFTA_SIZE; i++)
3505 		if (sc->shadow_vfta[i] != 0) {
3506 			/* XXXKB: incomplete VF support, we return early above */
3507 			if (sc->vf_ifp)
3508 				e1000_vfta_set_vf(hw, sc->shadow_vfta[i], true);
3509 			else
3510 				e1000_write_vfta(hw, i, sc->shadow_vfta[i]);
3511 		}
3512 
3513 	/* Re-enable interrupts for lem-class devices */
3514 	if (hw->mac.type < em_mac_min)
3515 		em_if_intr_enable(sc->ctx);
3516 }
3517 
3518 static void
3519 em_setup_vlan_hw_support(if_ctx_t ctx)
3520 {
3521 	struct e1000_softc *sc = iflib_get_softc(ctx);
3522 	struct e1000_hw *hw = &sc->hw;
3523 	struct ifnet *ifp = iflib_get_ifp(ctx);
3524 	u32 reg;
3525 
3526 	/* XXXKB: Return early if we are a VF until VF decap and filter management
3527 	 * is ready and tested.
3528 	 */
3529 	if (sc->vf_ifp)
3530 		return;
3531 
3532 	if (if_getcapenable(ifp) & IFCAP_VLAN_HWTAGGING &&
3533 	    !em_disable_crc_stripping) {
3534 		reg = E1000_READ_REG(hw, E1000_CTRL);
3535 		reg |= E1000_CTRL_VME;
3536 		E1000_WRITE_REG(hw, E1000_CTRL, reg);
3537 	} else {
3538 		reg = E1000_READ_REG(hw, E1000_CTRL);
3539 		reg &= ~E1000_CTRL_VME;
3540 		E1000_WRITE_REG(hw, E1000_CTRL, reg);
3541 	}
3542 
3543 	/* If we aren't doing HW filtering, we're done */
3544 	if (!em_if_vlan_filter_capable(ctx))  {
3545 		em_if_vlan_filter_disable(sc);
3546 		return;
3547 	}
3548 
3549 	/*
3550 	 * A soft reset zero's out the VFTA, so
3551 	 * we need to repopulate it now.
3552 	 */
3553 	em_if_vlan_filter_write(sc);
3554 
3555 	/* Enable the Filter Table */
3556 	em_if_vlan_filter_enable(sc);
3557 }
3558 
3559 static void
3560 em_if_intr_enable(if_ctx_t ctx)
3561 {
3562 	struct e1000_softc *sc = iflib_get_softc(ctx);
3563 	struct e1000_hw *hw = &sc->hw;
3564 	u32 ims_mask = IMS_ENABLE_MASK;
3565 
3566 	if (sc->intr_type == IFLIB_INTR_MSIX) {
3567 		E1000_WRITE_REG(hw, EM_EIAC, sc->ims);
3568 		ims_mask |= sc->ims;
3569 	}
3570 	E1000_WRITE_REG(hw, E1000_IMS, ims_mask);
3571 	E1000_WRITE_FLUSH(hw);
3572 }
3573 
3574 static void
3575 em_if_intr_disable(if_ctx_t ctx)
3576 {
3577 	struct e1000_softc *sc = iflib_get_softc(ctx);
3578 	struct e1000_hw *hw = &sc->hw;
3579 
3580 	if (sc->intr_type == IFLIB_INTR_MSIX)
3581 		E1000_WRITE_REG(hw, EM_EIAC, 0);
3582 	E1000_WRITE_REG(hw, E1000_IMC, 0xffffffff);
3583 	E1000_WRITE_FLUSH(hw);
3584 }
3585 
3586 static void
3587 igb_if_intr_enable(if_ctx_t ctx)
3588 {
3589 	struct e1000_softc *sc = iflib_get_softc(ctx);
3590 	struct e1000_hw *hw = &sc->hw;
3591 	u32 mask;
3592 
3593 	if (__predict_true(sc->intr_type == IFLIB_INTR_MSIX)) {
3594 		mask = (sc->que_mask | sc->link_mask);
3595 		E1000_WRITE_REG(hw, E1000_EIAC, mask);
3596 		E1000_WRITE_REG(hw, E1000_EIAM, mask);
3597 		E1000_WRITE_REG(hw, E1000_EIMS, mask);
3598 		E1000_WRITE_REG(hw, E1000_IMS, E1000_IMS_LSC);
3599 	} else
3600 		E1000_WRITE_REG(hw, E1000_IMS, IMS_ENABLE_MASK);
3601 	E1000_WRITE_FLUSH(hw);
3602 }
3603 
3604 static void
3605 igb_if_intr_disable(if_ctx_t ctx)
3606 {
3607 	struct e1000_softc *sc = iflib_get_softc(ctx);
3608 	struct e1000_hw *hw = &sc->hw;
3609 
3610 	if (__predict_true(sc->intr_type == IFLIB_INTR_MSIX)) {
3611 		E1000_WRITE_REG(hw, E1000_EIMC, 0xffffffff);
3612 		E1000_WRITE_REG(hw, E1000_EIAC, 0);
3613 	}
3614 	E1000_WRITE_REG(hw, E1000_IMC, 0xffffffff);
3615 	E1000_WRITE_FLUSH(hw);
3616 }
3617 
3618 /*
3619  * Bit of a misnomer, what this really means is
3620  * to enable OS management of the system... aka
3621  * to disable special hardware management features
3622  */
3623 static void
3624 em_init_manageability(struct e1000_softc *sc)
3625 {
3626 	/* A shared code workaround */
3627 #define E1000_82542_MANC2H E1000_MANC2H
3628 	if (sc->has_manage) {
3629 		int manc2h = E1000_READ_REG(&sc->hw, E1000_MANC2H);
3630 		int manc = E1000_READ_REG(&sc->hw, E1000_MANC);
3631 
3632 		/* disable hardware interception of ARP */
3633 		manc &= ~(E1000_MANC_ARP_EN);
3634 
3635 		/* enable receiving management packets to the host */
3636 		manc |= E1000_MANC_EN_MNG2HOST;
3637 #define E1000_MNG2HOST_PORT_623 (1 << 5)
3638 #define E1000_MNG2HOST_PORT_664 (1 << 6)
3639 		manc2h |= E1000_MNG2HOST_PORT_623;
3640 		manc2h |= E1000_MNG2HOST_PORT_664;
3641 		E1000_WRITE_REG(&sc->hw, E1000_MANC2H, manc2h);
3642 		E1000_WRITE_REG(&sc->hw, E1000_MANC, manc);
3643 	}
3644 }
3645 
3646 /*
3647  * Give control back to hardware management
3648  * controller if there is one.
3649  */
3650 static void
3651 em_release_manageability(struct e1000_softc *sc)
3652 {
3653 	if (sc->has_manage) {
3654 		int manc = E1000_READ_REG(&sc->hw, E1000_MANC);
3655 
3656 		/* re-enable hardware interception of ARP */
3657 		manc |= E1000_MANC_ARP_EN;
3658 		manc &= ~E1000_MANC_EN_MNG2HOST;
3659 
3660 		E1000_WRITE_REG(&sc->hw, E1000_MANC, manc);
3661 	}
3662 }
3663 
3664 /*
3665  * em_get_hw_control sets the {CTRL_EXT|FWSM}:DRV_LOAD bit.
3666  * For ASF and Pass Through versions of f/w this means
3667  * that the driver is loaded. For AMT version type f/w
3668  * this means that the network i/f is open.
3669  */
3670 static void
3671 em_get_hw_control(struct e1000_softc *sc)
3672 {
3673 	u32 ctrl_ext, swsm;
3674 
3675 	if (sc->vf_ifp)
3676 		return;
3677 
3678 	if (sc->hw.mac.type == e1000_82573) {
3679 		swsm = E1000_READ_REG(&sc->hw, E1000_SWSM);
3680 		E1000_WRITE_REG(&sc->hw, E1000_SWSM,
3681 		    swsm | E1000_SWSM_DRV_LOAD);
3682 		return;
3683 	}
3684 	/* else */
3685 	ctrl_ext = E1000_READ_REG(&sc->hw, E1000_CTRL_EXT);
3686 	E1000_WRITE_REG(&sc->hw, E1000_CTRL_EXT,
3687 	    ctrl_ext | E1000_CTRL_EXT_DRV_LOAD);
3688 }
3689 
3690 /*
3691  * em_release_hw_control resets {CTRL_EXT|FWSM}:DRV_LOAD bit.
3692  * For ASF and Pass Through versions of f/w this means that
3693  * the driver is no longer loaded. For AMT versions of the
3694  * f/w this means that the network i/f is closed.
3695  */
3696 static void
3697 em_release_hw_control(struct e1000_softc *sc)
3698 {
3699 	u32 ctrl_ext, swsm;
3700 
3701 	if (!sc->has_manage)
3702 		return;
3703 
3704 	if (sc->hw.mac.type == e1000_82573) {
3705 		swsm = E1000_READ_REG(&sc->hw, E1000_SWSM);
3706 		E1000_WRITE_REG(&sc->hw, E1000_SWSM,
3707 		    swsm & ~E1000_SWSM_DRV_LOAD);
3708 		return;
3709 	}
3710 	/* else */
3711 	ctrl_ext = E1000_READ_REG(&sc->hw, E1000_CTRL_EXT);
3712 	E1000_WRITE_REG(&sc->hw, E1000_CTRL_EXT,
3713 	    ctrl_ext & ~E1000_CTRL_EXT_DRV_LOAD);
3714 	return;
3715 }
3716 
3717 static int
3718 em_is_valid_ether_addr(u8 *addr)
3719 {
3720 	char zero_addr[6] = { 0, 0, 0, 0, 0, 0 };
3721 
3722 	if ((addr[0] & 1) || (!bcmp(addr, zero_addr, ETHER_ADDR_LEN))) {
3723 		return (false);
3724 	}
3725 
3726 	return (true);
3727 }
3728 
3729 /*
3730 ** Parse the interface capabilities with regard
3731 ** to both system management and wake-on-lan for
3732 ** later use.
3733 */
3734 static void
3735 em_get_wakeup(if_ctx_t ctx)
3736 {
3737 	struct e1000_softc *sc = iflib_get_softc(ctx);
3738 	device_t dev = iflib_get_dev(ctx);
3739 	u16 eeprom_data = 0, device_id, apme_mask;
3740 
3741 	sc->has_manage = e1000_enable_mng_pass_thru(&sc->hw);
3742 	apme_mask = EM_EEPROM_APME;
3743 
3744 	switch (sc->hw.mac.type) {
3745 	case e1000_82542:
3746 	case e1000_82543:
3747 		break;
3748 	case e1000_82544:
3749 		e1000_read_nvm(&sc->hw,
3750 		    NVM_INIT_CONTROL2_REG, 1, &eeprom_data);
3751 		apme_mask = EM_82544_APME;
3752 		break;
3753 	case e1000_82546:
3754 	case e1000_82546_rev_3:
3755 		if (sc->hw.bus.func == 1) {
3756 			e1000_read_nvm(&sc->hw,
3757 			    NVM_INIT_CONTROL3_PORT_B, 1, &eeprom_data);
3758 			break;
3759 		} else
3760 			e1000_read_nvm(&sc->hw,
3761 			    NVM_INIT_CONTROL3_PORT_A, 1, &eeprom_data);
3762 		break;
3763 	case e1000_82573:
3764 	case e1000_82583:
3765 		sc->has_amt = true;
3766 		/* FALLTHROUGH */
3767 	case e1000_82571:
3768 	case e1000_82572:
3769 	case e1000_80003es2lan:
3770 		if (sc->hw.bus.func == 1) {
3771 			e1000_read_nvm(&sc->hw,
3772 			    NVM_INIT_CONTROL3_PORT_B, 1, &eeprom_data);
3773 			break;
3774 		} else
3775 			e1000_read_nvm(&sc->hw,
3776 			    NVM_INIT_CONTROL3_PORT_A, 1, &eeprom_data);
3777 		break;
3778 	case e1000_ich8lan:
3779 	case e1000_ich9lan:
3780 	case e1000_ich10lan:
3781 	case e1000_pchlan:
3782 	case e1000_pch2lan:
3783 	case e1000_pch_lpt:
3784 	case e1000_pch_spt:
3785 	case e1000_82575:	/* listing all igb devices */
3786 	case e1000_82576:
3787 	case e1000_82580:
3788 	case e1000_i350:
3789 	case e1000_i354:
3790 	case e1000_i210:
3791 	case e1000_i211:
3792 	case e1000_vfadapt:
3793 	case e1000_vfadapt_i350:
3794 		apme_mask = E1000_WUC_APME;
3795 		sc->has_amt = true;
3796 		eeprom_data = E1000_READ_REG(&sc->hw, E1000_WUC);
3797 		break;
3798 	default:
3799 		e1000_read_nvm(&sc->hw,
3800 		    NVM_INIT_CONTROL3_PORT_A, 1, &eeprom_data);
3801 		break;
3802 	}
3803 	if (eeprom_data & apme_mask)
3804 		sc->wol = (E1000_WUFC_MAG | E1000_WUFC_MC);
3805 	/*
3806 	 * We have the eeprom settings, now apply the special cases
3807 	 * where the eeprom may be wrong or the board won't support
3808 	 * wake on lan on a particular port
3809 	 */
3810 	device_id = pci_get_device(dev);
3811 	switch (device_id) {
3812 	case E1000_DEV_ID_82546GB_PCIE:
3813 		sc->wol = 0;
3814 		break;
3815 	case E1000_DEV_ID_82546EB_FIBER:
3816 	case E1000_DEV_ID_82546GB_FIBER:
3817 		/* Wake events only supported on port A for dual fiber
3818 		 * regardless of eeprom setting */
3819 		if (E1000_READ_REG(&sc->hw, E1000_STATUS) &
3820 		    E1000_STATUS_FUNC_1)
3821 			sc->wol = 0;
3822 		break;
3823 	case E1000_DEV_ID_82546GB_QUAD_COPPER_KSP3:
3824 		/* if quad port adapter, disable WoL on all but port A */
3825 		if (global_quad_port_a != 0)
3826 			sc->wol = 0;
3827 		/* Reset for multiple quad port adapters */
3828 		if (++global_quad_port_a == 4)
3829 			global_quad_port_a = 0;
3830 		break;
3831 	case E1000_DEV_ID_82571EB_FIBER:
3832 		/* Wake events only supported on port A for dual fiber
3833 		 * regardless of eeprom setting */
3834 		if (E1000_READ_REG(&sc->hw, E1000_STATUS) &
3835 		    E1000_STATUS_FUNC_1)
3836 			sc->wol = 0;
3837 		break;
3838 	case E1000_DEV_ID_82571EB_QUAD_COPPER:
3839 	case E1000_DEV_ID_82571EB_QUAD_FIBER:
3840 	case E1000_DEV_ID_82571EB_QUAD_COPPER_LP:
3841 		/* if quad port adapter, disable WoL on all but port A */
3842 		if (global_quad_port_a != 0)
3843 			sc->wol = 0;
3844 		/* Reset for multiple quad port adapters */
3845 		if (++global_quad_port_a == 4)
3846 			global_quad_port_a = 0;
3847 		break;
3848 	}
3849 	return;
3850 }
3851 
3852 
3853 /*
3854  * Enable PCI Wake On Lan capability
3855  */
3856 static void
3857 em_enable_wakeup(if_ctx_t ctx)
3858 {
3859 	struct e1000_softc *sc = iflib_get_softc(ctx);
3860 	device_t dev = iflib_get_dev(ctx);
3861 	if_t ifp = iflib_get_ifp(ctx);
3862 	int error = 0;
3863 	u32 pmc, ctrl, ctrl_ext, rctl;
3864 	u16 status;
3865 
3866 	if (pci_find_cap(dev, PCIY_PMG, &pmc) != 0)
3867 		return;
3868 
3869 	/*
3870 	 * Determine type of Wakeup: note that wol
3871 	 * is set with all bits on by default.
3872 	 */
3873 	if ((if_getcapenable(ifp) & IFCAP_WOL_MAGIC) == 0)
3874 		sc->wol &= ~E1000_WUFC_MAG;
3875 
3876 	if ((if_getcapenable(ifp) & IFCAP_WOL_UCAST) == 0)
3877 		sc->wol &= ~E1000_WUFC_EX;
3878 
3879 	if ((if_getcapenable(ifp) & IFCAP_WOL_MCAST) == 0)
3880 		sc->wol &= ~E1000_WUFC_MC;
3881 	else {
3882 		rctl = E1000_READ_REG(&sc->hw, E1000_RCTL);
3883 		rctl |= E1000_RCTL_MPE;
3884 		E1000_WRITE_REG(&sc->hw, E1000_RCTL, rctl);
3885 	}
3886 
3887 	if (!(sc->wol & (E1000_WUFC_EX | E1000_WUFC_MAG | E1000_WUFC_MC)))
3888 		goto pme;
3889 
3890 	/* Advertise the wakeup capability */
3891 	ctrl = E1000_READ_REG(&sc->hw, E1000_CTRL);
3892 	ctrl |= (E1000_CTRL_SWDPIN2 | E1000_CTRL_SWDPIN3);
3893 	E1000_WRITE_REG(&sc->hw, E1000_CTRL, ctrl);
3894 
3895 	/* Keep the laser running on Fiber adapters */
3896 	if (sc->hw.phy.media_type == e1000_media_type_fiber ||
3897 	    sc->hw.phy.media_type == e1000_media_type_internal_serdes) {
3898 		ctrl_ext = E1000_READ_REG(&sc->hw, E1000_CTRL_EXT);
3899 		ctrl_ext |= E1000_CTRL_EXT_SDP3_DATA;
3900 		E1000_WRITE_REG(&sc->hw, E1000_CTRL_EXT, ctrl_ext);
3901 	}
3902 
3903 	if ((sc->hw.mac.type == e1000_ich8lan) ||
3904 	    (sc->hw.mac.type == e1000_pchlan) ||
3905 	    (sc->hw.mac.type == e1000_ich9lan) ||
3906 	    (sc->hw.mac.type == e1000_ich10lan))
3907 		e1000_suspend_workarounds_ich8lan(&sc->hw);
3908 
3909 	if ( sc->hw.mac.type >= e1000_pchlan) {
3910 		error = em_enable_phy_wakeup(sc);
3911 		if (error)
3912 			goto pme;
3913 	} else {
3914 		/* Enable wakeup by the MAC */
3915 		E1000_WRITE_REG(&sc->hw, E1000_WUC, E1000_WUC_PME_EN);
3916 		E1000_WRITE_REG(&sc->hw, E1000_WUFC, sc->wol);
3917 	}
3918 
3919 	if (sc->hw.phy.type == e1000_phy_igp_3)
3920 		e1000_igp3_phy_powerdown_workaround_ich8lan(&sc->hw);
3921 
3922 pme:
3923 	status = pci_read_config(dev, pmc + PCIR_POWER_STATUS, 2);
3924 	status &= ~(PCIM_PSTAT_PME | PCIM_PSTAT_PMEENABLE);
3925 	if (!error && (if_getcapenable(ifp) & IFCAP_WOL))
3926 		status |= PCIM_PSTAT_PME | PCIM_PSTAT_PMEENABLE;
3927 	pci_write_config(dev, pmc + PCIR_POWER_STATUS, status, 2);
3928 
3929 	return;
3930 }
3931 
3932 /*
3933  * WOL in the newer chipset interfaces (pchlan)
3934  * require thing to be copied into the phy
3935  */
3936 static int
3937 em_enable_phy_wakeup(struct e1000_softc *sc)
3938 {
3939 	struct e1000_hw *hw = &sc->hw;
3940 	u32 mreg, ret = 0;
3941 	u16 preg;
3942 
3943 	/* copy MAC RARs to PHY RARs */
3944 	e1000_copy_rx_addrs_to_phy_ich8lan(hw);
3945 
3946 	/* copy MAC MTA to PHY MTA */
3947 	for (int i = 0; i < hw->mac.mta_reg_count; i++) {
3948 		mreg = E1000_READ_REG_ARRAY(hw, E1000_MTA, i);
3949 		e1000_write_phy_reg(hw, BM_MTA(i), (u16)(mreg & 0xFFFF));
3950 		e1000_write_phy_reg(hw, BM_MTA(i) + 1,
3951 		    (u16)((mreg >> 16) & 0xFFFF));
3952 	}
3953 
3954 	/* configure PHY Rx Control register */
3955 	e1000_read_phy_reg(hw, BM_RCTL, &preg);
3956 	mreg = E1000_READ_REG(hw, E1000_RCTL);
3957 	if (mreg & E1000_RCTL_UPE)
3958 		preg |= BM_RCTL_UPE;
3959 	if (mreg & E1000_RCTL_MPE)
3960 		preg |= BM_RCTL_MPE;
3961 	preg &= ~(BM_RCTL_MO_MASK);
3962 	if (mreg & E1000_RCTL_MO_3)
3963 		preg |= (((mreg & E1000_RCTL_MO_3) >> E1000_RCTL_MO_SHIFT)
3964 				<< BM_RCTL_MO_SHIFT);
3965 	if (mreg & E1000_RCTL_BAM)
3966 		preg |= BM_RCTL_BAM;
3967 	if (mreg & E1000_RCTL_PMCF)
3968 		preg |= BM_RCTL_PMCF;
3969 	mreg = E1000_READ_REG(hw, E1000_CTRL);
3970 	if (mreg & E1000_CTRL_RFCE)
3971 		preg |= BM_RCTL_RFCE;
3972 	e1000_write_phy_reg(hw, BM_RCTL, preg);
3973 
3974 	/* enable PHY wakeup in MAC register */
3975 	E1000_WRITE_REG(hw, E1000_WUC,
3976 	    E1000_WUC_PHY_WAKE | E1000_WUC_PME_EN | E1000_WUC_APME);
3977 	E1000_WRITE_REG(hw, E1000_WUFC, sc->wol);
3978 
3979 	/* configure and enable PHY wakeup in PHY registers */
3980 	e1000_write_phy_reg(hw, BM_WUFC, sc->wol);
3981 	e1000_write_phy_reg(hw, BM_WUC, E1000_WUC_PME_EN);
3982 
3983 	/* activate PHY wakeup */
3984 	ret = hw->phy.ops.acquire(hw);
3985 	if (ret) {
3986 		printf("Could not acquire PHY\n");
3987 		return ret;
3988 	}
3989 	e1000_write_phy_reg_mdic(hw, IGP01E1000_PHY_PAGE_SELECT,
3990 	                         (BM_WUC_ENABLE_PAGE << IGP_PAGE_SHIFT));
3991 	ret = e1000_read_phy_reg_mdic(hw, BM_WUC_ENABLE_REG, &preg);
3992 	if (ret) {
3993 		printf("Could not read PHY page 769\n");
3994 		goto out;
3995 	}
3996 	preg |= BM_WUC_ENABLE_BIT | BM_WUC_HOST_WU_BIT;
3997 	ret = e1000_write_phy_reg_mdic(hw, BM_WUC_ENABLE_REG, preg);
3998 	if (ret)
3999 		printf("Could not set PHY Host Wakeup bit\n");
4000 out:
4001 	hw->phy.ops.release(hw);
4002 
4003 	return ret;
4004 }
4005 
4006 static void
4007 em_if_led_func(if_ctx_t ctx, int onoff)
4008 {
4009 	struct e1000_softc *sc = iflib_get_softc(ctx);
4010 
4011 	if (onoff) {
4012 		e1000_setup_led(&sc->hw);
4013 		e1000_led_on(&sc->hw);
4014 	} else {
4015 		e1000_led_off(&sc->hw);
4016 		e1000_cleanup_led(&sc->hw);
4017 	}
4018 }
4019 
4020 /*
4021  * Disable the L0S and L1 LINK states
4022  */
4023 static void
4024 em_disable_aspm(struct e1000_softc *sc)
4025 {
4026 	int base, reg;
4027 	u16 link_cap,link_ctrl;
4028 	device_t dev = sc->dev;
4029 
4030 	switch (sc->hw.mac.type) {
4031 	case e1000_82573:
4032 	case e1000_82574:
4033 	case e1000_82583:
4034 		break;
4035 	default:
4036 		return;
4037 	}
4038 	if (pci_find_cap(dev, PCIY_EXPRESS, &base) != 0)
4039 		return;
4040 	reg = base + PCIER_LINK_CAP;
4041 	link_cap = pci_read_config(dev, reg, 2);
4042 	if ((link_cap & PCIEM_LINK_CAP_ASPM) == 0)
4043 		return;
4044 	reg = base + PCIER_LINK_CTL;
4045 	link_ctrl = pci_read_config(dev, reg, 2);
4046 	link_ctrl &= ~PCIEM_LINK_CTL_ASPMC;
4047 	pci_write_config(dev, reg, link_ctrl, 2);
4048 	return;
4049 }
4050 
4051 /**********************************************************************
4052  *
4053  *  Update the board statistics counters.
4054  *
4055  **********************************************************************/
4056 static void
4057 em_update_stats_counters(struct e1000_softc *sc)
4058 {
4059 	u64 prev_xoffrxc = sc->stats.xoffrxc;
4060 
4061 	if(sc->hw.phy.media_type == e1000_media_type_copper ||
4062 	   (E1000_READ_REG(&sc->hw, E1000_STATUS) & E1000_STATUS_LU)) {
4063 		sc->stats.symerrs += E1000_READ_REG(&sc->hw, E1000_SYMERRS);
4064 		sc->stats.sec += E1000_READ_REG(&sc->hw, E1000_SEC);
4065 	}
4066 	sc->stats.crcerrs += E1000_READ_REG(&sc->hw, E1000_CRCERRS);
4067 	sc->stats.mpc += E1000_READ_REG(&sc->hw, E1000_MPC);
4068 	sc->stats.scc += E1000_READ_REG(&sc->hw, E1000_SCC);
4069 	sc->stats.ecol += E1000_READ_REG(&sc->hw, E1000_ECOL);
4070 
4071 	sc->stats.mcc += E1000_READ_REG(&sc->hw, E1000_MCC);
4072 	sc->stats.latecol += E1000_READ_REG(&sc->hw, E1000_LATECOL);
4073 	sc->stats.colc += E1000_READ_REG(&sc->hw, E1000_COLC);
4074 	sc->stats.dc += E1000_READ_REG(&sc->hw, E1000_DC);
4075 	sc->stats.rlec += E1000_READ_REG(&sc->hw, E1000_RLEC);
4076 	sc->stats.xonrxc += E1000_READ_REG(&sc->hw, E1000_XONRXC);
4077 	sc->stats.xontxc += E1000_READ_REG(&sc->hw, E1000_XONTXC);
4078 	sc->stats.xoffrxc += E1000_READ_REG(&sc->hw, E1000_XOFFRXC);
4079 	/*
4080 	 ** For watchdog management we need to know if we have been
4081 	 ** paused during the last interval, so capture that here.
4082 	*/
4083 	if (sc->stats.xoffrxc != prev_xoffrxc)
4084 		sc->shared->isc_pause_frames = 1;
4085 	sc->stats.xofftxc += E1000_READ_REG(&sc->hw, E1000_XOFFTXC);
4086 	sc->stats.fcruc += E1000_READ_REG(&sc->hw, E1000_FCRUC);
4087 	sc->stats.prc64 += E1000_READ_REG(&sc->hw, E1000_PRC64);
4088 	sc->stats.prc127 += E1000_READ_REG(&sc->hw, E1000_PRC127);
4089 	sc->stats.prc255 += E1000_READ_REG(&sc->hw, E1000_PRC255);
4090 	sc->stats.prc511 += E1000_READ_REG(&sc->hw, E1000_PRC511);
4091 	sc->stats.prc1023 += E1000_READ_REG(&sc->hw, E1000_PRC1023);
4092 	sc->stats.prc1522 += E1000_READ_REG(&sc->hw, E1000_PRC1522);
4093 	sc->stats.gprc += E1000_READ_REG(&sc->hw, E1000_GPRC);
4094 	sc->stats.bprc += E1000_READ_REG(&sc->hw, E1000_BPRC);
4095 	sc->stats.mprc += E1000_READ_REG(&sc->hw, E1000_MPRC);
4096 	sc->stats.gptc += E1000_READ_REG(&sc->hw, E1000_GPTC);
4097 
4098 	/* For the 64-bit byte counters the low dword must be read first. */
4099 	/* Both registers clear on the read of the high dword */
4100 
4101 	sc->stats.gorc += E1000_READ_REG(&sc->hw, E1000_GORCL) +
4102 	    ((u64)E1000_READ_REG(&sc->hw, E1000_GORCH) << 32);
4103 	sc->stats.gotc += E1000_READ_REG(&sc->hw, E1000_GOTCL) +
4104 	    ((u64)E1000_READ_REG(&sc->hw, E1000_GOTCH) << 32);
4105 
4106 	sc->stats.rnbc += E1000_READ_REG(&sc->hw, E1000_RNBC);
4107 	sc->stats.ruc += E1000_READ_REG(&sc->hw, E1000_RUC);
4108 	sc->stats.rfc += E1000_READ_REG(&sc->hw, E1000_RFC);
4109 	sc->stats.roc += E1000_READ_REG(&sc->hw, E1000_ROC);
4110 	sc->stats.rjc += E1000_READ_REG(&sc->hw, E1000_RJC);
4111 
4112 	sc->stats.tor += E1000_READ_REG(&sc->hw, E1000_TORH);
4113 	sc->stats.tot += E1000_READ_REG(&sc->hw, E1000_TOTH);
4114 
4115 	sc->stats.tpr += E1000_READ_REG(&sc->hw, E1000_TPR);
4116 	sc->stats.tpt += E1000_READ_REG(&sc->hw, E1000_TPT);
4117 	sc->stats.ptc64 += E1000_READ_REG(&sc->hw, E1000_PTC64);
4118 	sc->stats.ptc127 += E1000_READ_REG(&sc->hw, E1000_PTC127);
4119 	sc->stats.ptc255 += E1000_READ_REG(&sc->hw, E1000_PTC255);
4120 	sc->stats.ptc511 += E1000_READ_REG(&sc->hw, E1000_PTC511);
4121 	sc->stats.ptc1023 += E1000_READ_REG(&sc->hw, E1000_PTC1023);
4122 	sc->stats.ptc1522 += E1000_READ_REG(&sc->hw, E1000_PTC1522);
4123 	sc->stats.mptc += E1000_READ_REG(&sc->hw, E1000_MPTC);
4124 	sc->stats.bptc += E1000_READ_REG(&sc->hw, E1000_BPTC);
4125 
4126 	/* Interrupt Counts */
4127 
4128 	sc->stats.iac += E1000_READ_REG(&sc->hw, E1000_IAC);
4129 	sc->stats.icrxptc += E1000_READ_REG(&sc->hw, E1000_ICRXPTC);
4130 	sc->stats.icrxatc += E1000_READ_REG(&sc->hw, E1000_ICRXATC);
4131 	sc->stats.ictxptc += E1000_READ_REG(&sc->hw, E1000_ICTXPTC);
4132 	sc->stats.ictxatc += E1000_READ_REG(&sc->hw, E1000_ICTXATC);
4133 	sc->stats.ictxqec += E1000_READ_REG(&sc->hw, E1000_ICTXQEC);
4134 	sc->stats.ictxqmtc += E1000_READ_REG(&sc->hw, E1000_ICTXQMTC);
4135 	sc->stats.icrxdmtc += E1000_READ_REG(&sc->hw, E1000_ICRXDMTC);
4136 	sc->stats.icrxoc += E1000_READ_REG(&sc->hw, E1000_ICRXOC);
4137 
4138 	if (sc->hw.mac.type >= e1000_82543) {
4139 		sc->stats.algnerrc +=
4140 		E1000_READ_REG(&sc->hw, E1000_ALGNERRC);
4141 		sc->stats.rxerrc +=
4142 		E1000_READ_REG(&sc->hw, E1000_RXERRC);
4143 		sc->stats.tncrs +=
4144 		E1000_READ_REG(&sc->hw, E1000_TNCRS);
4145 		sc->stats.cexterr +=
4146 		E1000_READ_REG(&sc->hw, E1000_CEXTERR);
4147 		sc->stats.tsctc +=
4148 		E1000_READ_REG(&sc->hw, E1000_TSCTC);
4149 		sc->stats.tsctfc +=
4150 		E1000_READ_REG(&sc->hw, E1000_TSCTFC);
4151 	}
4152 }
4153 
4154 static uint64_t
4155 em_if_get_counter(if_ctx_t ctx, ift_counter cnt)
4156 {
4157 	struct e1000_softc *sc = iflib_get_softc(ctx);
4158 	struct ifnet *ifp = iflib_get_ifp(ctx);
4159 
4160 	switch (cnt) {
4161 	case IFCOUNTER_COLLISIONS:
4162 		return (sc->stats.colc);
4163 	case IFCOUNTER_IERRORS:
4164 		return (sc->dropped_pkts + sc->stats.rxerrc +
4165 		    sc->stats.crcerrs + sc->stats.algnerrc +
4166 		    sc->stats.ruc + sc->stats.roc +
4167 		    sc->stats.mpc + sc->stats.cexterr);
4168 	case IFCOUNTER_OERRORS:
4169 		return (sc->stats.ecol + sc->stats.latecol +
4170 		    sc->watchdog_events);
4171 	default:
4172 		return (if_get_counter_default(ifp, cnt));
4173 	}
4174 }
4175 
4176 /* em_if_needs_restart - Tell iflib when the driver needs to be reinitialized
4177  * @ctx: iflib context
4178  * @event: event code to check
4179  *
4180  * Defaults to returning true for unknown events.
4181  *
4182  * @returns true if iflib needs to reinit the interface
4183  */
4184 static bool
4185 em_if_needs_restart(if_ctx_t ctx __unused, enum iflib_restart_event event)
4186 {
4187 	switch (event) {
4188 	case IFLIB_RESTART_VLAN_CONFIG:
4189 		return (false);
4190 	default:
4191 		return (true);
4192 	}
4193 }
4194 
4195 /* Export a single 32-bit register via a read-only sysctl. */
4196 static int
4197 em_sysctl_reg_handler(SYSCTL_HANDLER_ARGS)
4198 {
4199 	struct e1000_softc *sc;
4200 	u_int val;
4201 
4202 	sc = oidp->oid_arg1;
4203 	val = E1000_READ_REG(&sc->hw, oidp->oid_arg2);
4204 	return (sysctl_handle_int(oidp, &val, 0, req));
4205 }
4206 
4207 /*
4208  * Add sysctl variables, one per statistic, to the system.
4209  */
4210 static void
4211 em_add_hw_stats(struct e1000_softc *sc)
4212 {
4213 	device_t dev = iflib_get_dev(sc->ctx);
4214 	struct em_tx_queue *tx_que = sc->tx_queues;
4215 	struct em_rx_queue *rx_que = sc->rx_queues;
4216 
4217 	struct sysctl_ctx_list *ctx = device_get_sysctl_ctx(dev);
4218 	struct sysctl_oid *tree = device_get_sysctl_tree(dev);
4219 	struct sysctl_oid_list *child = SYSCTL_CHILDREN(tree);
4220 	struct e1000_hw_stats *stats = &sc->stats;
4221 
4222 	struct sysctl_oid *stat_node, *queue_node, *int_node;
4223 	struct sysctl_oid_list *stat_list, *queue_list, *int_list;
4224 
4225 #define QUEUE_NAME_LEN 32
4226 	char namebuf[QUEUE_NAME_LEN];
4227 
4228 	/* Driver Statistics */
4229 	SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "dropped",
4230 			CTLFLAG_RD, &sc->dropped_pkts,
4231 			"Driver dropped packets");
4232 	SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "link_irq",
4233 			CTLFLAG_RD, &sc->link_irq,
4234 			"Link MSI-X IRQ Handled");
4235 	SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "rx_overruns",
4236 			CTLFLAG_RD, &sc->rx_overruns,
4237 			"RX overruns");
4238 	SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "watchdog_timeouts",
4239 			CTLFLAG_RD, &sc->watchdog_events,
4240 			"Watchdog timeouts");
4241 	SYSCTL_ADD_PROC(ctx, child, OID_AUTO, "device_control",
4242 	    CTLTYPE_UINT | CTLFLAG_RD | CTLFLAG_NEEDGIANT,
4243 	    sc, E1000_CTRL, em_sysctl_reg_handler, "IU",
4244 	    "Device Control Register");
4245 	SYSCTL_ADD_PROC(ctx, child, OID_AUTO, "rx_control",
4246 	    CTLTYPE_UINT | CTLFLAG_RD | CTLFLAG_NEEDGIANT,
4247 	    sc, E1000_RCTL, em_sysctl_reg_handler, "IU",
4248 	    "Receiver Control Register");
4249 	SYSCTL_ADD_UINT(ctx, child, OID_AUTO, "fc_high_water",
4250 			CTLFLAG_RD, &sc->hw.fc.high_water, 0,
4251 			"Flow Control High Watermark");
4252 	SYSCTL_ADD_UINT(ctx, child, OID_AUTO, "fc_low_water",
4253 			CTLFLAG_RD, &sc->hw.fc.low_water, 0,
4254 			"Flow Control Low Watermark");
4255 
4256 	for (int i = 0; i < sc->tx_num_queues; i++, tx_que++) {
4257 		struct tx_ring *txr = &tx_que->txr;
4258 		snprintf(namebuf, QUEUE_NAME_LEN, "queue_tx_%d", i);
4259 		queue_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, namebuf,
4260 		    CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, "TX Queue Name");
4261 		queue_list = SYSCTL_CHILDREN(queue_node);
4262 
4263 		SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "txd_head",
4264 		    CTLTYPE_UINT | CTLFLAG_RD | CTLFLAG_NEEDGIANT, sc,
4265 		    E1000_TDH(txr->me), em_sysctl_reg_handler, "IU",
4266 		    "Transmit Descriptor Head");
4267 		SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "txd_tail",
4268 		    CTLTYPE_UINT | CTLFLAG_RD | CTLFLAG_NEEDGIANT, sc,
4269 		    E1000_TDT(txr->me), em_sysctl_reg_handler, "IU",
4270 		    "Transmit Descriptor Tail");
4271 		SYSCTL_ADD_ULONG(ctx, queue_list, OID_AUTO, "tx_irq",
4272 				CTLFLAG_RD, &txr->tx_irq,
4273 				"Queue MSI-X Transmit Interrupts");
4274 	}
4275 
4276 	for (int j = 0; j < sc->rx_num_queues; j++, rx_que++) {
4277 		struct rx_ring *rxr = &rx_que->rxr;
4278 		snprintf(namebuf, QUEUE_NAME_LEN, "queue_rx_%d", j);
4279 		queue_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, namebuf,
4280 		    CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, "RX Queue Name");
4281 		queue_list = SYSCTL_CHILDREN(queue_node);
4282 
4283 		SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "rxd_head",
4284 		    CTLTYPE_UINT | CTLFLAG_RD | CTLFLAG_NEEDGIANT, sc,
4285 		    E1000_RDH(rxr->me), em_sysctl_reg_handler, "IU",
4286 		    "Receive Descriptor Head");
4287 		SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "rxd_tail",
4288 		    CTLTYPE_UINT | CTLFLAG_RD | CTLFLAG_NEEDGIANT, sc,
4289 		    E1000_RDT(rxr->me), em_sysctl_reg_handler, "IU",
4290 		    "Receive Descriptor Tail");
4291 		SYSCTL_ADD_ULONG(ctx, queue_list, OID_AUTO, "rx_irq",
4292 				CTLFLAG_RD, &rxr->rx_irq,
4293 				"Queue MSI-X Receive Interrupts");
4294 	}
4295 
4296 	/* MAC stats get their own sub node */
4297 
4298 	stat_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, "mac_stats",
4299 	    CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, "Statistics");
4300 	stat_list = SYSCTL_CHILDREN(stat_node);
4301 
4302 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "excess_coll",
4303 			CTLFLAG_RD, &stats->ecol,
4304 			"Excessive collisions");
4305 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "single_coll",
4306 			CTLFLAG_RD, &stats->scc,
4307 			"Single collisions");
4308 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "multiple_coll",
4309 			CTLFLAG_RD, &stats->mcc,
4310 			"Multiple collisions");
4311 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "late_coll",
4312 			CTLFLAG_RD, &stats->latecol,
4313 			"Late collisions");
4314 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "collision_count",
4315 			CTLFLAG_RD, &stats->colc,
4316 			"Collision Count");
4317 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "symbol_errors",
4318 			CTLFLAG_RD, &sc->stats.symerrs,
4319 			"Symbol Errors");
4320 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "sequence_errors",
4321 			CTLFLAG_RD, &sc->stats.sec,
4322 			"Sequence Errors");
4323 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "defer_count",
4324 			CTLFLAG_RD, &sc->stats.dc,
4325 			"Defer Count");
4326 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "missed_packets",
4327 			CTLFLAG_RD, &sc->stats.mpc,
4328 			"Missed Packets");
4329 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_no_buff",
4330 			CTLFLAG_RD, &sc->stats.rnbc,
4331 			"Receive No Buffers");
4332 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_undersize",
4333 			CTLFLAG_RD, &sc->stats.ruc,
4334 			"Receive Undersize");
4335 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_fragmented",
4336 			CTLFLAG_RD, &sc->stats.rfc,
4337 			"Fragmented Packets Received ");
4338 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_oversize",
4339 			CTLFLAG_RD, &sc->stats.roc,
4340 			"Oversized Packets Received");
4341 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_jabber",
4342 			CTLFLAG_RD, &sc->stats.rjc,
4343 			"Recevied Jabber");
4344 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_errs",
4345 			CTLFLAG_RD, &sc->stats.rxerrc,
4346 			"Receive Errors");
4347 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "crc_errs",
4348 			CTLFLAG_RD, &sc->stats.crcerrs,
4349 			"CRC errors");
4350 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "alignment_errs",
4351 			CTLFLAG_RD, &sc->stats.algnerrc,
4352 			"Alignment Errors");
4353 	/* On 82575 these are collision counts */
4354 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "coll_ext_errs",
4355 			CTLFLAG_RD, &sc->stats.cexterr,
4356 			"Collision/Carrier extension errors");
4357 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xon_recvd",
4358 			CTLFLAG_RD, &sc->stats.xonrxc,
4359 			"XON Received");
4360 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xon_txd",
4361 			CTLFLAG_RD, &sc->stats.xontxc,
4362 			"XON Transmitted");
4363 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xoff_recvd",
4364 			CTLFLAG_RD, &sc->stats.xoffrxc,
4365 			"XOFF Received");
4366 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xoff_txd",
4367 			CTLFLAG_RD, &sc->stats.xofftxc,
4368 			"XOFF Transmitted");
4369 
4370 	/* Packet Reception Stats */
4371 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "total_pkts_recvd",
4372 			CTLFLAG_RD, &sc->stats.tpr,
4373 			"Total Packets Received ");
4374 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_pkts_recvd",
4375 			CTLFLAG_RD, &sc->stats.gprc,
4376 			"Good Packets Received");
4377 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "bcast_pkts_recvd",
4378 			CTLFLAG_RD, &sc->stats.bprc,
4379 			"Broadcast Packets Received");
4380 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "mcast_pkts_recvd",
4381 			CTLFLAG_RD, &sc->stats.mprc,
4382 			"Multicast Packets Received");
4383 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_64",
4384 			CTLFLAG_RD, &sc->stats.prc64,
4385 			"64 byte frames received ");
4386 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_65_127",
4387 			CTLFLAG_RD, &sc->stats.prc127,
4388 			"65-127 byte frames received");
4389 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_128_255",
4390 			CTLFLAG_RD, &sc->stats.prc255,
4391 			"128-255 byte frames received");
4392 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_256_511",
4393 			CTLFLAG_RD, &sc->stats.prc511,
4394 			"256-511 byte frames received");
4395 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_512_1023",
4396 			CTLFLAG_RD, &sc->stats.prc1023,
4397 			"512-1023 byte frames received");
4398 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_1024_1522",
4399 			CTLFLAG_RD, &sc->stats.prc1522,
4400 			"1023-1522 byte frames received");
4401 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_octets_recvd",
4402 			CTLFLAG_RD, &sc->stats.gorc,
4403 			"Good Octets Received");
4404 
4405 	/* Packet Transmission Stats */
4406 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_octets_txd",
4407 			CTLFLAG_RD, &sc->stats.gotc,
4408 			"Good Octets Transmitted");
4409 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "total_pkts_txd",
4410 			CTLFLAG_RD, &sc->stats.tpt,
4411 			"Total Packets Transmitted");
4412 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_pkts_txd",
4413 			CTLFLAG_RD, &sc->stats.gptc,
4414 			"Good Packets Transmitted");
4415 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "bcast_pkts_txd",
4416 			CTLFLAG_RD, &sc->stats.bptc,
4417 			"Broadcast Packets Transmitted");
4418 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "mcast_pkts_txd",
4419 			CTLFLAG_RD, &sc->stats.mptc,
4420 			"Multicast Packets Transmitted");
4421 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_64",
4422 			CTLFLAG_RD, &sc->stats.ptc64,
4423 			"64 byte frames transmitted ");
4424 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_65_127",
4425 			CTLFLAG_RD, &sc->stats.ptc127,
4426 			"65-127 byte frames transmitted");
4427 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_128_255",
4428 			CTLFLAG_RD, &sc->stats.ptc255,
4429 			"128-255 byte frames transmitted");
4430 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_256_511",
4431 			CTLFLAG_RD, &sc->stats.ptc511,
4432 			"256-511 byte frames transmitted");
4433 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_512_1023",
4434 			CTLFLAG_RD, &sc->stats.ptc1023,
4435 			"512-1023 byte frames transmitted");
4436 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_1024_1522",
4437 			CTLFLAG_RD, &sc->stats.ptc1522,
4438 			"1024-1522 byte frames transmitted");
4439 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tso_txd",
4440 			CTLFLAG_RD, &sc->stats.tsctc,
4441 			"TSO Contexts Transmitted");
4442 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tso_ctx_fail",
4443 			CTLFLAG_RD, &sc->stats.tsctfc,
4444 			"TSO Contexts Failed");
4445 
4446 
4447 	/* Interrupt Stats */
4448 
4449 	int_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, "interrupts",
4450 	    CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, "Interrupt Statistics");
4451 	int_list = SYSCTL_CHILDREN(int_node);
4452 
4453 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "asserts",
4454 			CTLFLAG_RD, &sc->stats.iac,
4455 			"Interrupt Assertion Count");
4456 
4457 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "rx_pkt_timer",
4458 			CTLFLAG_RD, &sc->stats.icrxptc,
4459 			"Interrupt Cause Rx Pkt Timer Expire Count");
4460 
4461 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "rx_abs_timer",
4462 			CTLFLAG_RD, &sc->stats.icrxatc,
4463 			"Interrupt Cause Rx Abs Timer Expire Count");
4464 
4465 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "tx_pkt_timer",
4466 			CTLFLAG_RD, &sc->stats.ictxptc,
4467 			"Interrupt Cause Tx Pkt Timer Expire Count");
4468 
4469 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "tx_abs_timer",
4470 			CTLFLAG_RD, &sc->stats.ictxatc,
4471 			"Interrupt Cause Tx Abs Timer Expire Count");
4472 
4473 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "tx_queue_empty",
4474 			CTLFLAG_RD, &sc->stats.ictxqec,
4475 			"Interrupt Cause Tx Queue Empty Count");
4476 
4477 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "tx_queue_min_thresh",
4478 			CTLFLAG_RD, &sc->stats.ictxqmtc,
4479 			"Interrupt Cause Tx Queue Min Thresh Count");
4480 
4481 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "rx_desc_min_thresh",
4482 			CTLFLAG_RD, &sc->stats.icrxdmtc,
4483 			"Interrupt Cause Rx Desc Min Thresh Count");
4484 
4485 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "rx_overrun",
4486 			CTLFLAG_RD, &sc->stats.icrxoc,
4487 			"Interrupt Cause Receiver Overrun Count");
4488 }
4489 
4490 static void
4491 em_fw_version_locked(if_ctx_t ctx)
4492 {
4493 	struct e1000_softc *sc = iflib_get_softc(ctx);
4494 	struct e1000_hw *hw = &sc->hw;
4495 	struct e1000_fw_version *fw_ver = &sc->fw_ver;
4496 	uint16_t eep = 0;
4497 
4498 	/*
4499 	 * em_fw_version_locked() must run under the IFLIB_CTX_LOCK to meet the
4500 	 * NVM locking model, so we do it in em_if_attach_pre() and store the
4501 	 * info in the softc
4502 	 */
4503 	ASSERT_CTX_LOCK_HELD(hw);
4504 
4505 	*fw_ver = (struct e1000_fw_version){0};
4506 
4507 	if (hw->mac.type >= igb_mac_min) {
4508 		/*
4509 		 * Use the Shared Code for igb(4)
4510 		 */
4511 		e1000_get_fw_version(hw, fw_ver);
4512 	} else {
4513 		/*
4514 		 * Otherwise, EEPROM version should be present on (almost?) all
4515 		 * devices here
4516 		 */
4517 		if(e1000_read_nvm(hw, NVM_VERSION, 1, &eep)) {
4518 			INIT_DEBUGOUT("can't get EEPROM version");
4519 			return;
4520 		}
4521 
4522 		fw_ver->eep_major = (eep & NVM_MAJOR_MASK) >> NVM_MAJOR_SHIFT;
4523 		fw_ver->eep_minor = (eep & NVM_MINOR_MASK) >> NVM_MINOR_SHIFT;
4524 		fw_ver->eep_build = (eep & NVM_IMAGE_ID_MASK);
4525 	}
4526 }
4527 
4528 static void
4529 em_sbuf_fw_version(struct e1000_fw_version *fw_ver, struct sbuf *buf)
4530 {
4531 	const char *space = "";
4532 
4533 	if (fw_ver->eep_major || fw_ver->eep_minor || fw_ver->eep_build) {
4534 		sbuf_printf(buf, "EEPROM V%d.%d-%d", fw_ver->eep_major,
4535 			    fw_ver->eep_minor, fw_ver->eep_build);
4536 		space = " ";
4537 	}
4538 
4539 	if (fw_ver->invm_major || fw_ver->invm_minor || fw_ver->invm_img_type) {
4540 		sbuf_printf(buf, "%sNVM V%d.%d imgtype%d",
4541 			    space, fw_ver->invm_major, fw_ver->invm_minor,
4542 			    fw_ver->invm_img_type);
4543 		space = " ";
4544 	}
4545 
4546 	if (fw_ver->or_valid) {
4547 		sbuf_printf(buf, "%sOption ROM V%d-b%d-p%d",
4548 			    space, fw_ver->or_major, fw_ver->or_build,
4549 			    fw_ver->or_patch);
4550 		space = " ";
4551 	}
4552 
4553 	if (fw_ver->etrack_id)
4554 		sbuf_printf(buf, "%seTrack 0x%08x", space, fw_ver->etrack_id);
4555 }
4556 
4557 static void
4558 em_print_fw_version(struct e1000_softc *sc )
4559 {
4560 	device_t dev = sc->dev;
4561 	struct sbuf *buf;
4562 	int error = 0;
4563 
4564 	buf = sbuf_new_auto();
4565 	if (!buf) {
4566 		device_printf(dev, "Could not allocate sbuf for output.\n");
4567 		return;
4568 	}
4569 
4570 	em_sbuf_fw_version(&sc->fw_ver, buf);
4571 
4572 	error = sbuf_finish(buf);
4573 	if (error)
4574 		device_printf(dev, "Error finishing sbuf: %d\n", error);
4575 	else if (sbuf_len(buf))
4576 		device_printf(dev, "%s\n", sbuf_data(buf));
4577 
4578 	sbuf_delete(buf);
4579 }
4580 
4581 static int
4582 em_sysctl_print_fw_version(SYSCTL_HANDLER_ARGS)
4583 {
4584 	struct e1000_softc *sc = (struct e1000_softc *)arg1;
4585 	device_t dev = sc->dev;
4586 	struct sbuf *buf;
4587 	int error = 0;
4588 
4589 	buf = sbuf_new_for_sysctl(NULL, NULL, 128, req);
4590 	if (!buf) {
4591 		device_printf(dev, "Could not allocate sbuf for output.\n");
4592 		return (ENOMEM);
4593 	}
4594 
4595 	em_sbuf_fw_version(&sc->fw_ver, buf);
4596 
4597 	error = sbuf_finish(buf);
4598 	if (error)
4599 		device_printf(dev, "Error finishing sbuf: %d\n", error);
4600 
4601 	sbuf_delete(buf);
4602 
4603 	return (0);
4604 }
4605 
4606 /**********************************************************************
4607  *
4608  *  This routine provides a way to dump out the adapter eeprom,
4609  *  often a useful debug/service tool. This only dumps the first
4610  *  32 words, stuff that matters is in that extent.
4611  *
4612  **********************************************************************/
4613 static int
4614 em_sysctl_nvm_info(SYSCTL_HANDLER_ARGS)
4615 {
4616 	struct e1000_softc *sc = (struct e1000_softc *)arg1;
4617 	int error;
4618 	int result;
4619 
4620 	result = -1;
4621 	error = sysctl_handle_int(oidp, &result, 0, req);
4622 
4623 	if (error || !req->newptr)
4624 		return (error);
4625 
4626 	/*
4627 	 * This value will cause a hex dump of the
4628 	 * first 32 16-bit words of the EEPROM to
4629 	 * the screen.
4630 	 */
4631 	if (result == 1)
4632 		em_print_nvm_info(sc);
4633 
4634 	return (error);
4635 }
4636 
4637 static void
4638 em_print_nvm_info(struct e1000_softc *sc)
4639 {
4640 	struct e1000_hw *hw = &sc->hw;
4641 	struct sx *iflib_ctx_lock = iflib_ctx_lock_get(sc->ctx);
4642 	u16 eeprom_data;
4643 	int i, j, row = 0;
4644 
4645 	/* Its a bit crude, but it gets the job done */
4646 	printf("\nInterface EEPROM Dump:\n");
4647 	printf("Offset\n0x0000  ");
4648 
4649 	/* We rely on the IFLIB_CTX_LOCK as part of NVM locking model */
4650 	sx_xlock(iflib_ctx_lock);
4651 	ASSERT_CTX_LOCK_HELD(hw);
4652 	for (i = 0, j = 0; i < 32; i++, j++) {
4653 		if (j == 8) { /* Make the offset block */
4654 			j = 0; ++row;
4655 			printf("\n0x00%x0  ",row);
4656 		}
4657 		e1000_read_nvm(hw, i, 1, &eeprom_data);
4658 		printf("%04x ", eeprom_data);
4659 	}
4660 	sx_xunlock(iflib_ctx_lock);
4661 	printf("\n");
4662 }
4663 
4664 static int
4665 em_sysctl_int_delay(SYSCTL_HANDLER_ARGS)
4666 {
4667 	struct em_int_delay_info *info;
4668 	struct e1000_softc *sc;
4669 	u32 regval;
4670 	int error, usecs, ticks;
4671 
4672 	info = (struct em_int_delay_info *) arg1;
4673 	usecs = info->value;
4674 	error = sysctl_handle_int(oidp, &usecs, 0, req);
4675 	if (error != 0 || req->newptr == NULL)
4676 		return (error);
4677 	if (usecs < 0 || usecs > EM_TICKS_TO_USECS(65535))
4678 		return (EINVAL);
4679 	info->value = usecs;
4680 	ticks = EM_USECS_TO_TICKS(usecs);
4681 	if (info->offset == E1000_ITR)	/* units are 256ns here */
4682 		ticks *= 4;
4683 
4684 	sc = info->sc;
4685 
4686 	regval = E1000_READ_OFFSET(&sc->hw, info->offset);
4687 	regval = (regval & ~0xffff) | (ticks & 0xffff);
4688 	/* Handle a few special cases. */
4689 	switch (info->offset) {
4690 	case E1000_RDTR:
4691 		break;
4692 	case E1000_TIDV:
4693 		if (ticks == 0) {
4694 			sc->txd_cmd &= ~E1000_TXD_CMD_IDE;
4695 			/* Don't write 0 into the TIDV register. */
4696 			regval++;
4697 		} else
4698 			sc->txd_cmd |= E1000_TXD_CMD_IDE;
4699 		break;
4700 	}
4701 	E1000_WRITE_OFFSET(&sc->hw, info->offset, regval);
4702 	return (0);
4703 }
4704 
4705 static void
4706 em_add_int_delay_sysctl(struct e1000_softc *sc, const char *name,
4707 	const char *description, struct em_int_delay_info *info,
4708 	int offset, int value)
4709 {
4710 	info->sc = sc;
4711 	info->offset = offset;
4712 	info->value = value;
4713 	SYSCTL_ADD_PROC(device_get_sysctl_ctx(sc->dev),
4714 	    SYSCTL_CHILDREN(device_get_sysctl_tree(sc->dev)),
4715 	    OID_AUTO, name, CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_NEEDGIANT,
4716 	    info, 0, em_sysctl_int_delay, "I", description);
4717 }
4718 
4719 /*
4720  * Set flow control using sysctl:
4721  * Flow control values:
4722  *      0 - off
4723  *      1 - rx pause
4724  *      2 - tx pause
4725  *      3 - full
4726  */
4727 static int
4728 em_set_flowcntl(SYSCTL_HANDLER_ARGS)
4729 {
4730 	int error;
4731 	static int input = 3; /* default is full */
4732 	struct e1000_softc	*sc = (struct e1000_softc *) arg1;
4733 
4734 	error = sysctl_handle_int(oidp, &input, 0, req);
4735 
4736 	if ((error) || (req->newptr == NULL))
4737 		return (error);
4738 
4739 	if (input == sc->fc) /* no change? */
4740 		return (error);
4741 
4742 	switch (input) {
4743 	case e1000_fc_rx_pause:
4744 	case e1000_fc_tx_pause:
4745 	case e1000_fc_full:
4746 	case e1000_fc_none:
4747 		sc->hw.fc.requested_mode = input;
4748 		sc->fc = input;
4749 		break;
4750 	default:
4751 		/* Do nothing */
4752 		return (error);
4753 	}
4754 
4755 	sc->hw.fc.current_mode = sc->hw.fc.requested_mode;
4756 	e1000_force_mac_fc(&sc->hw);
4757 	return (error);
4758 }
4759 
4760 /*
4761  * Manage Energy Efficient Ethernet:
4762  * Control values:
4763  *     0/1 - enabled/disabled
4764  */
4765 static int
4766 em_sysctl_eee(SYSCTL_HANDLER_ARGS)
4767 {
4768 	struct e1000_softc *sc = (struct e1000_softc *) arg1;
4769 	int error, value;
4770 
4771 	value = sc->hw.dev_spec.ich8lan.eee_disable;
4772 	error = sysctl_handle_int(oidp, &value, 0, req);
4773 	if (error || req->newptr == NULL)
4774 		return (error);
4775 	sc->hw.dev_spec.ich8lan.eee_disable = (value != 0);
4776 	em_if_init(sc->ctx);
4777 
4778 	return (0);
4779 }
4780 
4781 static int
4782 em_sysctl_debug_info(SYSCTL_HANDLER_ARGS)
4783 {
4784 	struct e1000_softc *sc;
4785 	int error;
4786 	int result;
4787 
4788 	result = -1;
4789 	error = sysctl_handle_int(oidp, &result, 0, req);
4790 
4791 	if (error || !req->newptr)
4792 		return (error);
4793 
4794 	if (result == 1) {
4795 		sc = (struct e1000_softc *) arg1;
4796 		em_print_debug_info(sc);
4797 	}
4798 
4799 	return (error);
4800 }
4801 
4802 static int
4803 em_get_rs(SYSCTL_HANDLER_ARGS)
4804 {
4805 	struct e1000_softc *sc = (struct e1000_softc *) arg1;
4806 	int error;
4807 	int result;
4808 
4809 	result = 0;
4810 	error = sysctl_handle_int(oidp, &result, 0, req);
4811 
4812 	if (error || !req->newptr || result != 1)
4813 		return (error);
4814 	em_dump_rs(sc);
4815 
4816 	return (error);
4817 }
4818 
4819 static void
4820 em_if_debug(if_ctx_t ctx)
4821 {
4822 	em_dump_rs(iflib_get_softc(ctx));
4823 }
4824 
4825 /*
4826  * This routine is meant to be fluid, add whatever is
4827  * needed for debugging a problem.  -jfv
4828  */
4829 static void
4830 em_print_debug_info(struct e1000_softc *sc)
4831 {
4832 	device_t dev = iflib_get_dev(sc->ctx);
4833 	struct ifnet *ifp = iflib_get_ifp(sc->ctx);
4834 	struct tx_ring *txr = &sc->tx_queues->txr;
4835 	struct rx_ring *rxr = &sc->rx_queues->rxr;
4836 
4837 	if (if_getdrvflags(ifp) & IFF_DRV_RUNNING)
4838 		printf("Interface is RUNNING ");
4839 	else
4840 		printf("Interface is NOT RUNNING\n");
4841 
4842 	if (if_getdrvflags(ifp) & IFF_DRV_OACTIVE)
4843 		printf("and INACTIVE\n");
4844 	else
4845 		printf("and ACTIVE\n");
4846 
4847 	for (int i = 0; i < sc->tx_num_queues; i++, txr++) {
4848 		device_printf(dev, "TX Queue %d ------\n", i);
4849 		device_printf(dev, "hw tdh = %d, hw tdt = %d\n",
4850 			E1000_READ_REG(&sc->hw, E1000_TDH(i)),
4851 			E1000_READ_REG(&sc->hw, E1000_TDT(i)));
4852 
4853 	}
4854 	for (int j=0; j < sc->rx_num_queues; j++, rxr++) {
4855 		device_printf(dev, "RX Queue %d ------\n", j);
4856 		device_printf(dev, "hw rdh = %d, hw rdt = %d\n",
4857 			E1000_READ_REG(&sc->hw, E1000_RDH(j)),
4858 			E1000_READ_REG(&sc->hw, E1000_RDT(j)));
4859 	}
4860 }
4861 
4862 /*
4863  * 82574 only:
4864  * Write a new value to the EEPROM increasing the number of MSI-X
4865  * vectors from 3 to 5, for proper multiqueue support.
4866  */
4867 static void
4868 em_enable_vectors_82574(if_ctx_t ctx)
4869 {
4870 	struct e1000_softc *sc = iflib_get_softc(ctx);
4871 	struct e1000_hw *hw = &sc->hw;
4872 	device_t dev = iflib_get_dev(ctx);
4873 	u16 edata;
4874 
4875 	e1000_read_nvm(hw, EM_NVM_PCIE_CTRL, 1, &edata);
4876 	if (bootverbose)
4877 		device_printf(dev, "EM_NVM_PCIE_CTRL = %#06x\n", edata);
4878 	if (((edata & EM_NVM_MSIX_N_MASK) >> EM_NVM_MSIX_N_SHIFT) != 4) {
4879 		device_printf(dev, "Writing to eeprom: increasing "
4880 		    "reported MSI-X vectors from 3 to 5...\n");
4881 		edata &= ~(EM_NVM_MSIX_N_MASK);
4882 		edata |= 4 << EM_NVM_MSIX_N_SHIFT;
4883 		e1000_write_nvm(hw, EM_NVM_PCIE_CTRL, 1, &edata);
4884 		e1000_update_nvm_checksum(hw);
4885 		device_printf(dev, "Writing to eeprom: done\n");
4886 	}
4887 }
4888