xref: /freebsd/sys/dev/e1000/if_em.c (revision ecf4106237505fa9459ae871793b754334989c17)
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 #include "if_em.h"
30 #include <sys/sbuf.h>
31 #include <machine/_inttypes.h>
32 
33 #define em_mac_min e1000_82571
34 #define igb_mac_min e1000_82575
35 
36 /*********************************************************************
37  *  Driver version:
38  *********************************************************************/
39 static const char em_driver_version[] = "7.7.8-fbsd";
40 static const char igb_driver_version[] = "2.5.19-fbsd";
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 const 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 	PVID(0x8086, E1000_DEV_ID_PCH_LNL_I219_LM20, "Intel(R) I219-LM LNL(20)"),
198 	PVID(0x8086, E1000_DEV_ID_PCH_LNL_I219_V20, "Intel(R) I219-V LNL(20)"),
199 	PVID(0x8086, E1000_DEV_ID_PCH_LNL_I219_LM21, "Intel(R) I219-LM LNL(21)"),
200 	PVID(0x8086, E1000_DEV_ID_PCH_LNL_I219_V21, "Intel(R) I219-V LNL(21)"),
201 	PVID(0x8086, E1000_DEV_ID_PCH_RPL_I219_LM22, "Intel(R) I219-LM RPL(22)"),
202 	PVID(0x8086, E1000_DEV_ID_PCH_RPL_I219_V22, "Intel(R) I219-V RPL(22)"),
203 	PVID(0x8086, E1000_DEV_ID_PCH_RPL_I219_LM23, "Intel(R) I219-LM RPL(23)"),
204 	PVID(0x8086, E1000_DEV_ID_PCH_RPL_I219_V23, "Intel(R) I219-V RPL(23)"),
205 	PVID(0x8086, E1000_DEV_ID_PCH_ARL_I219_LM24, "Intel(R) I219-LM ARL(24)"),
206 	PVID(0x8086, E1000_DEV_ID_PCH_ARL_I219_V24, "Intel(R) I219-V ARL(24)"),
207 	PVID(0x8086, E1000_DEV_ID_PCH_PTP_I219_LM25, "Intel(R) I219-LM PTP(25)"),
208 	PVID(0x8086, E1000_DEV_ID_PCH_PTP_I219_V25, "Intel(R) I219-V PTP(25)"),
209 	PVID(0x8086, E1000_DEV_ID_PCH_PTP_I219_LM26, "Intel(R) I219-LM PTP(26)"),
210 	PVID(0x8086, E1000_DEV_ID_PCH_PTP_I219_V26, "Intel(R) I219-V PTP(26)"),
211 	PVID(0x8086, E1000_DEV_ID_PCH_PTP_I219_LM27, "Intel(R) I219-LM PTP(27)"),
212 	PVID(0x8086, E1000_DEV_ID_PCH_PTP_I219_V27, "Intel(R) I219-V PTP(27)"),
213 	/* required last entry */
214 	PVID_END
215 };
216 
217 static const pci_vendor_info_t igb_vendor_info_array[] =
218 {
219 	/* Intel(R) - igb-class devices */
220 	PVID(0x8086, E1000_DEV_ID_82575EB_COPPER, "Intel(R) PRO/1000 82575EB (Copper)"),
221 	PVID(0x8086, E1000_DEV_ID_82575EB_FIBER_SERDES, "Intel(R) PRO/1000 82575EB (SERDES)"),
222 	PVID(0x8086, E1000_DEV_ID_82575GB_QUAD_COPPER, "Intel(R) PRO/1000 VT 82575GB (Quad Copper)"),
223 	PVID(0x8086, E1000_DEV_ID_82576, "Intel(R) PRO/1000 82576"),
224 	PVID(0x8086, E1000_DEV_ID_82576_NS, "Intel(R) PRO/1000 82576NS"),
225 	PVID(0x8086, E1000_DEV_ID_82576_NS_SERDES, "Intel(R) PRO/1000 82576NS (SERDES)"),
226 	PVID(0x8086, E1000_DEV_ID_82576_FIBER, "Intel(R) PRO/1000 EF 82576 (Dual Fiber)"),
227 	PVID(0x8086, E1000_DEV_ID_82576_SERDES, "Intel(R) PRO/1000 82576 (Dual SERDES)"),
228 	PVID(0x8086, E1000_DEV_ID_82576_SERDES_QUAD, "Intel(R) PRO/1000 ET 82576 (Quad SERDES)"),
229 	PVID(0x8086, E1000_DEV_ID_82576_QUAD_COPPER, "Intel(R) PRO/1000 ET 82576 (Quad Copper)"),
230 	PVID(0x8086, E1000_DEV_ID_82576_QUAD_COPPER_ET2, "Intel(R) PRO/1000 ET(2) 82576 (Quad Copper)"),
231 	PVID(0x8086, E1000_DEV_ID_82576_VF, "Intel(R) PRO/1000 82576 Virtual Function"),
232 	PVID(0x8086, E1000_DEV_ID_82580_COPPER, "Intel(R) I340 82580 (Copper)"),
233 	PVID(0x8086, E1000_DEV_ID_82580_FIBER, "Intel(R) I340 82580 (Fiber)"),
234 	PVID(0x8086, E1000_DEV_ID_82580_SERDES, "Intel(R) I340 82580 (SERDES)"),
235 	PVID(0x8086, E1000_DEV_ID_82580_SGMII, "Intel(R) I340 82580 (SGMII)"),
236 	PVID(0x8086, E1000_DEV_ID_82580_COPPER_DUAL, "Intel(R) I340-T2 82580 (Dual Copper)"),
237 	PVID(0x8086, E1000_DEV_ID_82580_QUAD_FIBER, "Intel(R) I340-F4 82580 (Quad Fiber)"),
238 	PVID(0x8086, E1000_DEV_ID_DH89XXCC_SERDES, "Intel(R) DH89XXCC (SERDES)"),
239 	PVID(0x8086, E1000_DEV_ID_DH89XXCC_SGMII, "Intel(R) I347-AT4 DH89XXCC"),
240 	PVID(0x8086, E1000_DEV_ID_DH89XXCC_SFP, "Intel(R) DH89XXCC (SFP)"),
241 	PVID(0x8086, E1000_DEV_ID_DH89XXCC_BACKPLANE, "Intel(R) DH89XXCC (Backplane)"),
242 	PVID(0x8086, E1000_DEV_ID_I350_COPPER, "Intel(R) I350 (Copper)"),
243 	PVID(0x8086, E1000_DEV_ID_I350_FIBER, "Intel(R) I350 (Fiber)"),
244 	PVID(0x8086, E1000_DEV_ID_I350_SERDES, "Intel(R) I350 (SERDES)"),
245 	PVID(0x8086, E1000_DEV_ID_I350_SGMII, "Intel(R) I350 (SGMII)"),
246 	PVID(0x8086, E1000_DEV_ID_I350_VF, "Intel(R) I350 Virtual Function"),
247 	PVID(0x8086, E1000_DEV_ID_I210_COPPER, "Intel(R) I210 (Copper)"),
248 	PVID(0x8086, E1000_DEV_ID_I210_COPPER_IT, "Intel(R) I210 IT (Copper)"),
249 	PVID(0x8086, E1000_DEV_ID_I210_COPPER_OEM1, "Intel(R) I210 (OEM)"),
250 	PVID(0x8086, E1000_DEV_ID_I210_COPPER_FLASHLESS, "Intel(R) I210 Flashless (Copper)"),
251 	PVID(0x8086, E1000_DEV_ID_I210_SERDES_FLASHLESS, "Intel(R) I210 Flashless (SERDES)"),
252 	PVID(0x8086, E1000_DEV_ID_I210_SGMII_FLASHLESS, "Intel(R) I210 Flashless (SGMII)"),
253 	PVID(0x8086, E1000_DEV_ID_I210_FIBER, "Intel(R) I210 (Fiber)"),
254 	PVID(0x8086, E1000_DEV_ID_I210_SERDES, "Intel(R) I210 (SERDES)"),
255 	PVID(0x8086, E1000_DEV_ID_I210_SGMII, "Intel(R) I210 (SGMII)"),
256 	PVID(0x8086, E1000_DEV_ID_I211_COPPER, "Intel(R) I211 (Copper)"),
257 	PVID(0x8086, E1000_DEV_ID_I354_BACKPLANE_1GBPS, "Intel(R) I354 (1.0 GbE Backplane)"),
258 	PVID(0x8086, E1000_DEV_ID_I354_BACKPLANE_2_5GBPS, "Intel(R) I354 (2.5 GbE Backplane)"),
259 	PVID(0x8086, E1000_DEV_ID_I354_SGMII, "Intel(R) I354 (SGMII)"),
260 	/* required last entry */
261 	PVID_END
262 };
263 
264 /*********************************************************************
265  *  Function prototypes
266  *********************************************************************/
267 static void	*em_register(device_t);
268 static void	*igb_register(device_t);
269 static int	em_if_attach_pre(if_ctx_t);
270 static int	em_if_attach_post(if_ctx_t);
271 static int	em_if_detach(if_ctx_t);
272 static int	em_if_shutdown(if_ctx_t);
273 static int	em_if_suspend(if_ctx_t);
274 static int	em_if_resume(if_ctx_t);
275 
276 static int	em_if_tx_queues_alloc(if_ctx_t, caddr_t *, uint64_t *, int, int);
277 static int	em_if_rx_queues_alloc(if_ctx_t, caddr_t *, uint64_t *, int, int);
278 static void	em_if_queues_free(if_ctx_t);
279 
280 static uint64_t	em_if_get_counter(if_ctx_t, ift_counter);
281 static void	em_if_init(if_ctx_t);
282 static void	em_if_stop(if_ctx_t);
283 static void	em_if_media_status(if_ctx_t, struct ifmediareq *);
284 static int	em_if_media_change(if_ctx_t);
285 static int	em_if_mtu_set(if_ctx_t, uint32_t);
286 static void	em_if_timer(if_ctx_t, uint16_t);
287 static void	em_if_vlan_register(if_ctx_t, u16);
288 static void	em_if_vlan_unregister(if_ctx_t, u16);
289 static void	em_if_watchdog_reset(if_ctx_t);
290 static bool	em_if_needs_restart(if_ctx_t, enum iflib_restart_event);
291 
292 static void	em_identify_hardware(if_ctx_t);
293 static int	em_allocate_pci_resources(if_ctx_t);
294 static void	em_free_pci_resources(if_ctx_t);
295 static void	em_reset(if_ctx_t);
296 static int	em_setup_interface(if_ctx_t);
297 static int	em_setup_msix(if_ctx_t);
298 
299 static void	em_initialize_transmit_unit(if_ctx_t);
300 static void	em_initialize_receive_unit(if_ctx_t);
301 
302 static void	em_if_intr_enable(if_ctx_t);
303 static void	em_if_intr_disable(if_ctx_t);
304 static void	igb_if_intr_enable(if_ctx_t);
305 static void	igb_if_intr_disable(if_ctx_t);
306 static int	em_if_rx_queue_intr_enable(if_ctx_t, uint16_t);
307 static int	em_if_tx_queue_intr_enable(if_ctx_t, uint16_t);
308 static int	igb_if_rx_queue_intr_enable(if_ctx_t, uint16_t);
309 static int	igb_if_tx_queue_intr_enable(if_ctx_t, uint16_t);
310 static void	em_if_multi_set(if_ctx_t);
311 static void	em_if_update_admin_status(if_ctx_t);
312 static void	em_if_debug(if_ctx_t);
313 static void	em_update_stats_counters(struct e1000_softc *);
314 static void	em_add_hw_stats(struct e1000_softc *);
315 static int	em_if_set_promisc(if_ctx_t, int);
316 static bool	em_if_vlan_filter_capable(if_ctx_t);
317 static bool	em_if_vlan_filter_used(if_ctx_t);
318 static void	em_if_vlan_filter_enable(struct e1000_softc *);
319 static void	em_if_vlan_filter_disable(struct e1000_softc *);
320 static void	em_if_vlan_filter_write(struct e1000_softc *);
321 static void	em_setup_vlan_hw_support(if_ctx_t ctx);
322 static int	em_sysctl_nvm_info(SYSCTL_HANDLER_ARGS);
323 static void	em_print_nvm_info(struct e1000_softc *);
324 static void	em_fw_version_locked(if_ctx_t);
325 static void	em_sbuf_fw_version(struct e1000_fw_version *, struct sbuf *);
326 static void	em_print_fw_version(struct e1000_softc *);
327 static int	em_sysctl_print_fw_version(SYSCTL_HANDLER_ARGS);
328 static int	em_sysctl_debug_info(SYSCTL_HANDLER_ARGS);
329 static int	em_get_rs(SYSCTL_HANDLER_ARGS);
330 static void	em_print_debug_info(struct e1000_softc *);
331 static int 	em_is_valid_ether_addr(u8 *);
332 static bool	em_automask_tso(if_ctx_t);
333 static int	em_sysctl_int_delay(SYSCTL_HANDLER_ARGS);
334 static void	em_add_int_delay_sysctl(struct e1000_softc *, const char *,
335 		    const char *, struct em_int_delay_info *, int, int);
336 /* Management and WOL Support */
337 static void	em_init_manageability(struct e1000_softc *);
338 static void	em_release_manageability(struct e1000_softc *);
339 static void	em_get_hw_control(struct e1000_softc *);
340 static void	em_release_hw_control(struct e1000_softc *);
341 static void	em_get_wakeup(if_ctx_t);
342 static void	em_enable_wakeup(if_ctx_t);
343 static int	em_enable_phy_wakeup(struct e1000_softc *);
344 static void	em_disable_aspm(struct e1000_softc *);
345 
346 int		em_intr(void *);
347 
348 /* MSI-X handlers */
349 static int	em_if_msix_intr_assign(if_ctx_t, int);
350 static int	em_msix_link(void *);
351 static void	em_handle_link(void *);
352 
353 static void	em_enable_vectors_82574(if_ctx_t);
354 
355 static int	em_set_flowcntl(SYSCTL_HANDLER_ARGS);
356 static int	em_sysctl_eee(SYSCTL_HANDLER_ARGS);
357 static void	em_if_led_func(if_ctx_t, int);
358 
359 static int	em_get_regs(SYSCTL_HANDLER_ARGS);
360 
361 static void	lem_smartspeed(struct e1000_softc *);
362 static void	igb_configure_queues(struct e1000_softc *);
363 static void	em_flush_desc_rings(struct e1000_softc *);
364 
365 
366 /*********************************************************************
367  *  FreeBSD Device Interface Entry Points
368  *********************************************************************/
369 static device_method_t em_methods[] = {
370 	/* Device interface */
371 	DEVMETHOD(device_register, em_register),
372 	DEVMETHOD(device_probe, iflib_device_probe),
373 	DEVMETHOD(device_attach, iflib_device_attach),
374 	DEVMETHOD(device_detach, iflib_device_detach),
375 	DEVMETHOD(device_shutdown, iflib_device_shutdown),
376 	DEVMETHOD(device_suspend, iflib_device_suspend),
377 	DEVMETHOD(device_resume, iflib_device_resume),
378 	DEVMETHOD_END
379 };
380 
381 static device_method_t igb_methods[] = {
382 	/* Device interface */
383 	DEVMETHOD(device_register, igb_register),
384 	DEVMETHOD(device_probe, iflib_device_probe),
385 	DEVMETHOD(device_attach, iflib_device_attach),
386 	DEVMETHOD(device_detach, iflib_device_detach),
387 	DEVMETHOD(device_shutdown, iflib_device_shutdown),
388 	DEVMETHOD(device_suspend, iflib_device_suspend),
389 	DEVMETHOD(device_resume, iflib_device_resume),
390 	DEVMETHOD_END
391 };
392 
393 
394 static driver_t em_driver = {
395 	"em", em_methods, sizeof(struct e1000_softc),
396 };
397 
398 DRIVER_MODULE(em, pci, em_driver, 0, 0);
399 
400 MODULE_DEPEND(em, pci, 1, 1, 1);
401 MODULE_DEPEND(em, ether, 1, 1, 1);
402 MODULE_DEPEND(em, iflib, 1, 1, 1);
403 
404 IFLIB_PNP_INFO(pci, em, em_vendor_info_array);
405 
406 static driver_t igb_driver = {
407 	"igb", igb_methods, sizeof(struct e1000_softc),
408 };
409 
410 DRIVER_MODULE(igb, pci, igb_driver, 0, 0);
411 
412 MODULE_DEPEND(igb, pci, 1, 1, 1);
413 MODULE_DEPEND(igb, ether, 1, 1, 1);
414 MODULE_DEPEND(igb, iflib, 1, 1, 1);
415 
416 IFLIB_PNP_INFO(pci, igb, igb_vendor_info_array);
417 
418 static device_method_t em_if_methods[] = {
419 	DEVMETHOD(ifdi_attach_pre, em_if_attach_pre),
420 	DEVMETHOD(ifdi_attach_post, em_if_attach_post),
421 	DEVMETHOD(ifdi_detach, em_if_detach),
422 	DEVMETHOD(ifdi_shutdown, em_if_shutdown),
423 	DEVMETHOD(ifdi_suspend, em_if_suspend),
424 	DEVMETHOD(ifdi_resume, em_if_resume),
425 	DEVMETHOD(ifdi_init, em_if_init),
426 	DEVMETHOD(ifdi_stop, em_if_stop),
427 	DEVMETHOD(ifdi_msix_intr_assign, em_if_msix_intr_assign),
428 	DEVMETHOD(ifdi_intr_enable, em_if_intr_enable),
429 	DEVMETHOD(ifdi_intr_disable, em_if_intr_disable),
430 	DEVMETHOD(ifdi_tx_queues_alloc, em_if_tx_queues_alloc),
431 	DEVMETHOD(ifdi_rx_queues_alloc, em_if_rx_queues_alloc),
432 	DEVMETHOD(ifdi_queues_free, em_if_queues_free),
433 	DEVMETHOD(ifdi_update_admin_status, em_if_update_admin_status),
434 	DEVMETHOD(ifdi_multi_set, em_if_multi_set),
435 	DEVMETHOD(ifdi_media_status, em_if_media_status),
436 	DEVMETHOD(ifdi_media_change, em_if_media_change),
437 	DEVMETHOD(ifdi_mtu_set, em_if_mtu_set),
438 	DEVMETHOD(ifdi_promisc_set, em_if_set_promisc),
439 	DEVMETHOD(ifdi_timer, em_if_timer),
440 	DEVMETHOD(ifdi_watchdog_reset, em_if_watchdog_reset),
441 	DEVMETHOD(ifdi_vlan_register, em_if_vlan_register),
442 	DEVMETHOD(ifdi_vlan_unregister, em_if_vlan_unregister),
443 	DEVMETHOD(ifdi_get_counter, em_if_get_counter),
444 	DEVMETHOD(ifdi_led_func, em_if_led_func),
445 	DEVMETHOD(ifdi_rx_queue_intr_enable, em_if_rx_queue_intr_enable),
446 	DEVMETHOD(ifdi_tx_queue_intr_enable, em_if_tx_queue_intr_enable),
447 	DEVMETHOD(ifdi_debug, em_if_debug),
448 	DEVMETHOD(ifdi_needs_restart, em_if_needs_restart),
449 	DEVMETHOD_END
450 };
451 
452 static driver_t em_if_driver = {
453 	"em_if", em_if_methods, sizeof(struct e1000_softc)
454 };
455 
456 static device_method_t igb_if_methods[] = {
457 	DEVMETHOD(ifdi_attach_pre, em_if_attach_pre),
458 	DEVMETHOD(ifdi_attach_post, em_if_attach_post),
459 	DEVMETHOD(ifdi_detach, em_if_detach),
460 	DEVMETHOD(ifdi_shutdown, em_if_shutdown),
461 	DEVMETHOD(ifdi_suspend, em_if_suspend),
462 	DEVMETHOD(ifdi_resume, em_if_resume),
463 	DEVMETHOD(ifdi_init, em_if_init),
464 	DEVMETHOD(ifdi_stop, em_if_stop),
465 	DEVMETHOD(ifdi_msix_intr_assign, em_if_msix_intr_assign),
466 	DEVMETHOD(ifdi_intr_enable, igb_if_intr_enable),
467 	DEVMETHOD(ifdi_intr_disable, igb_if_intr_disable),
468 	DEVMETHOD(ifdi_tx_queues_alloc, em_if_tx_queues_alloc),
469 	DEVMETHOD(ifdi_rx_queues_alloc, em_if_rx_queues_alloc),
470 	DEVMETHOD(ifdi_queues_free, em_if_queues_free),
471 	DEVMETHOD(ifdi_update_admin_status, em_if_update_admin_status),
472 	DEVMETHOD(ifdi_multi_set, em_if_multi_set),
473 	DEVMETHOD(ifdi_media_status, em_if_media_status),
474 	DEVMETHOD(ifdi_media_change, em_if_media_change),
475 	DEVMETHOD(ifdi_mtu_set, em_if_mtu_set),
476 	DEVMETHOD(ifdi_promisc_set, em_if_set_promisc),
477 	DEVMETHOD(ifdi_timer, em_if_timer),
478 	DEVMETHOD(ifdi_watchdog_reset, em_if_watchdog_reset),
479 	DEVMETHOD(ifdi_vlan_register, em_if_vlan_register),
480 	DEVMETHOD(ifdi_vlan_unregister, em_if_vlan_unregister),
481 	DEVMETHOD(ifdi_get_counter, em_if_get_counter),
482 	DEVMETHOD(ifdi_led_func, em_if_led_func),
483 	DEVMETHOD(ifdi_rx_queue_intr_enable, igb_if_rx_queue_intr_enable),
484 	DEVMETHOD(ifdi_tx_queue_intr_enable, igb_if_tx_queue_intr_enable),
485 	DEVMETHOD(ifdi_debug, em_if_debug),
486 	DEVMETHOD(ifdi_needs_restart, em_if_needs_restart),
487 	DEVMETHOD_END
488 };
489 
490 static driver_t igb_if_driver = {
491 	"igb_if", igb_if_methods, sizeof(struct e1000_softc)
492 };
493 
494 /*********************************************************************
495  *  Tunable default values.
496  *********************************************************************/
497 
498 #define EM_TICKS_TO_USECS(ticks)	((1024 * (ticks) + 500) / 1000)
499 #define EM_USECS_TO_TICKS(usecs)	((1000 * (usecs) + 512) / 1024)
500 
501 #define MAX_INTS_PER_SEC	8000
502 #define DEFAULT_ITR		(1000000000/(MAX_INTS_PER_SEC * 256))
503 
504 /* Allow common code without TSO */
505 #ifndef CSUM_TSO
506 #define CSUM_TSO	0
507 #endif
508 
509 static SYSCTL_NODE(_hw, OID_AUTO, em, CTLFLAG_RD | CTLFLAG_MPSAFE, 0,
510     "EM driver parameters");
511 
512 static int em_disable_crc_stripping = 0;
513 SYSCTL_INT(_hw_em, OID_AUTO, disable_crc_stripping, CTLFLAG_RDTUN,
514     &em_disable_crc_stripping, 0, "Disable CRC Stripping");
515 
516 static int em_tx_int_delay_dflt = EM_TICKS_TO_USECS(EM_TIDV);
517 static int em_rx_int_delay_dflt = EM_TICKS_TO_USECS(EM_RDTR);
518 SYSCTL_INT(_hw_em, OID_AUTO, tx_int_delay, CTLFLAG_RDTUN, &em_tx_int_delay_dflt,
519     0, "Default transmit interrupt delay in usecs");
520 SYSCTL_INT(_hw_em, OID_AUTO, rx_int_delay, CTLFLAG_RDTUN, &em_rx_int_delay_dflt,
521     0, "Default receive interrupt delay in usecs");
522 
523 static int em_tx_abs_int_delay_dflt = EM_TICKS_TO_USECS(EM_TADV);
524 static int em_rx_abs_int_delay_dflt = EM_TICKS_TO_USECS(EM_RADV);
525 SYSCTL_INT(_hw_em, OID_AUTO, tx_abs_int_delay, CTLFLAG_RDTUN,
526     &em_tx_abs_int_delay_dflt, 0,
527     "Default transmit interrupt delay limit in usecs");
528 SYSCTL_INT(_hw_em, OID_AUTO, rx_abs_int_delay, CTLFLAG_RDTUN,
529     &em_rx_abs_int_delay_dflt, 0,
530     "Default receive interrupt delay limit in usecs");
531 
532 static int em_smart_pwr_down = false;
533 SYSCTL_INT(_hw_em, OID_AUTO, smart_pwr_down, CTLFLAG_RDTUN, &em_smart_pwr_down,
534     0, "Set to true to leave smart power down enabled on newer adapters");
535 
536 static bool em_unsupported_tso = false;
537 SYSCTL_BOOL(_hw_em, OID_AUTO, unsupported_tso, CTLFLAG_RDTUN,
538     &em_unsupported_tso, 0, "Allow unsupported em(4) TSO configurations");
539 
540 /* Controls whether promiscuous also shows bad packets */
541 static int em_debug_sbp = false;
542 SYSCTL_INT(_hw_em, OID_AUTO, sbp, CTLFLAG_RDTUN, &em_debug_sbp, 0,
543     "Show bad packets in promiscuous mode");
544 
545 /* Energy efficient ethernet - default to OFF */
546 static int eee_setting = 1;
547 SYSCTL_INT(_hw_em, OID_AUTO, eee_setting, CTLFLAG_RDTUN, &eee_setting, 0,
548     "Enable Energy Efficient Ethernet");
549 
550 /*
551 ** Tuneable Interrupt rate
552 */
553 static int em_max_interrupt_rate = 8000;
554 SYSCTL_INT(_hw_em, OID_AUTO, max_interrupt_rate, CTLFLAG_RDTUN,
555     &em_max_interrupt_rate, 0, "Maximum interrupts per second");
556 
557 /* Global used in WOL setup with multiport cards */
558 static int global_quad_port_a = 0;
559 
560 extern struct if_txrx igb_txrx;
561 extern struct if_txrx em_txrx;
562 extern struct if_txrx lem_txrx;
563 
564 static struct if_shared_ctx em_sctx_init = {
565 	.isc_magic = IFLIB_MAGIC,
566 	.isc_q_align = PAGE_SIZE,
567 	.isc_tx_maxsize = EM_TSO_SIZE + sizeof(struct ether_vlan_header),
568 	.isc_tx_maxsegsize = PAGE_SIZE,
569 	.isc_tso_maxsize = EM_TSO_SIZE + sizeof(struct ether_vlan_header),
570 	.isc_tso_maxsegsize = EM_TSO_SEG_SIZE,
571 	.isc_rx_maxsize = MJUM9BYTES,
572 	.isc_rx_nsegments = 1,
573 	.isc_rx_maxsegsize = MJUM9BYTES,
574 	.isc_nfl = 1,
575 	.isc_nrxqs = 1,
576 	.isc_ntxqs = 1,
577 	.isc_admin_intrcnt = 1,
578 	.isc_vendor_info = em_vendor_info_array,
579 	.isc_driver_version = em_driver_version,
580 	.isc_driver = &em_if_driver,
581 	.isc_flags = IFLIB_NEED_SCRATCH | IFLIB_TSO_INIT_IP | IFLIB_NEED_ZERO_CSUM,
582 
583 	.isc_nrxd_min = {EM_MIN_RXD},
584 	.isc_ntxd_min = {EM_MIN_TXD},
585 	.isc_nrxd_max = {EM_MAX_RXD},
586 	.isc_ntxd_max = {EM_MAX_TXD},
587 	.isc_nrxd_default = {EM_DEFAULT_RXD},
588 	.isc_ntxd_default = {EM_DEFAULT_TXD},
589 };
590 
591 static struct if_shared_ctx igb_sctx_init = {
592 	.isc_magic = IFLIB_MAGIC,
593 	.isc_q_align = PAGE_SIZE,
594 	.isc_tx_maxsize = EM_TSO_SIZE + sizeof(struct ether_vlan_header),
595 	.isc_tx_maxsegsize = PAGE_SIZE,
596 	.isc_tso_maxsize = EM_TSO_SIZE + sizeof(struct ether_vlan_header),
597 	.isc_tso_maxsegsize = EM_TSO_SEG_SIZE,
598 	.isc_rx_maxsize = MJUM9BYTES,
599 	.isc_rx_nsegments = 1,
600 	.isc_rx_maxsegsize = MJUM9BYTES,
601 	.isc_nfl = 1,
602 	.isc_nrxqs = 1,
603 	.isc_ntxqs = 1,
604 	.isc_admin_intrcnt = 1,
605 	.isc_vendor_info = igb_vendor_info_array,
606 	.isc_driver_version = igb_driver_version,
607 	.isc_driver = &igb_if_driver,
608 	.isc_flags = IFLIB_NEED_SCRATCH | IFLIB_TSO_INIT_IP | IFLIB_NEED_ZERO_CSUM,
609 
610 	.isc_nrxd_min = {EM_MIN_RXD},
611 	.isc_ntxd_min = {EM_MIN_TXD},
612 	.isc_nrxd_max = {IGB_MAX_RXD},
613 	.isc_ntxd_max = {IGB_MAX_TXD},
614 	.isc_nrxd_default = {EM_DEFAULT_RXD},
615 	.isc_ntxd_default = {EM_DEFAULT_TXD},
616 };
617 
618 /*****************************************************************
619  *
620  * Dump Registers
621  *
622  ****************************************************************/
623 #define IGB_REGS_LEN 739
624 
625 static int em_get_regs(SYSCTL_HANDLER_ARGS)
626 {
627 	struct e1000_softc *sc = (struct e1000_softc *)arg1;
628 	struct e1000_hw *hw = &sc->hw;
629 	struct sbuf *sb;
630 	u32 *regs_buff;
631 	int rc;
632 
633 	regs_buff = malloc(sizeof(u32) * IGB_REGS_LEN, M_DEVBUF, M_WAITOK);
634 	memset(regs_buff, 0, IGB_REGS_LEN * sizeof(u32));
635 
636 	rc = sysctl_wire_old_buffer(req, 0);
637 	MPASS(rc == 0);
638 	if (rc != 0) {
639 		free(regs_buff, M_DEVBUF);
640 		return (rc);
641 	}
642 
643 	sb = sbuf_new_for_sysctl(NULL, NULL, 32*400, req);
644 	MPASS(sb != NULL);
645 	if (sb == NULL) {
646 		free(regs_buff, M_DEVBUF);
647 		return (ENOMEM);
648 	}
649 
650 	/* General Registers */
651 	regs_buff[0] = E1000_READ_REG(hw, E1000_CTRL);
652 	regs_buff[1] = E1000_READ_REG(hw, E1000_STATUS);
653 	regs_buff[2] = E1000_READ_REG(hw, E1000_CTRL_EXT);
654 	regs_buff[3] = E1000_READ_REG(hw, E1000_ICR);
655 	regs_buff[4] = E1000_READ_REG(hw, E1000_RCTL);
656 	regs_buff[5] = E1000_READ_REG(hw, E1000_RDLEN(0));
657 	regs_buff[6] = E1000_READ_REG(hw, E1000_RDH(0));
658 	regs_buff[7] = E1000_READ_REG(hw, E1000_RDT(0));
659 	regs_buff[8] = E1000_READ_REG(hw, E1000_RXDCTL(0));
660 	regs_buff[9] = E1000_READ_REG(hw, E1000_RDBAL(0));
661 	regs_buff[10] = E1000_READ_REG(hw, E1000_RDBAH(0));
662 	regs_buff[11] = E1000_READ_REG(hw, E1000_TCTL);
663 	regs_buff[12] = E1000_READ_REG(hw, E1000_TDBAL(0));
664 	regs_buff[13] = E1000_READ_REG(hw, E1000_TDBAH(0));
665 	regs_buff[14] = E1000_READ_REG(hw, E1000_TDLEN(0));
666 	regs_buff[15] = E1000_READ_REG(hw, E1000_TDH(0));
667 	regs_buff[16] = E1000_READ_REG(hw, E1000_TDT(0));
668 	regs_buff[17] = E1000_READ_REG(hw, E1000_TXDCTL(0));
669 	regs_buff[18] = E1000_READ_REG(hw, E1000_TDFH);
670 	regs_buff[19] = E1000_READ_REG(hw, E1000_TDFT);
671 	regs_buff[20] = E1000_READ_REG(hw, E1000_TDFHS);
672 	regs_buff[21] = E1000_READ_REG(hw, E1000_TDFPC);
673 
674 	sbuf_printf(sb, "General Registers\n");
675 	sbuf_printf(sb, "\tCTRL\t %08x\n", regs_buff[0]);
676 	sbuf_printf(sb, "\tSTATUS\t %08x\n", regs_buff[1]);
677 	sbuf_printf(sb, "\tCTRL_EXT\t %08x\n\n", regs_buff[2]);
678 
679 	sbuf_printf(sb, "Interrupt Registers\n");
680 	sbuf_printf(sb, "\tICR\t %08x\n\n", regs_buff[3]);
681 
682 	sbuf_printf(sb, "RX Registers\n");
683 	sbuf_printf(sb, "\tRCTL\t %08x\n", regs_buff[4]);
684 	sbuf_printf(sb, "\tRDLEN\t %08x\n", regs_buff[5]);
685 	sbuf_printf(sb, "\tRDH\t %08x\n", regs_buff[6]);
686 	sbuf_printf(sb, "\tRDT\t %08x\n", regs_buff[7]);
687 	sbuf_printf(sb, "\tRXDCTL\t %08x\n", regs_buff[8]);
688 	sbuf_printf(sb, "\tRDBAL\t %08x\n", regs_buff[9]);
689 	sbuf_printf(sb, "\tRDBAH\t %08x\n\n", regs_buff[10]);
690 
691 	sbuf_printf(sb, "TX Registers\n");
692 	sbuf_printf(sb, "\tTCTL\t %08x\n", regs_buff[11]);
693 	sbuf_printf(sb, "\tTDBAL\t %08x\n", regs_buff[12]);
694 	sbuf_printf(sb, "\tTDBAH\t %08x\n", regs_buff[13]);
695 	sbuf_printf(sb, "\tTDLEN\t %08x\n", regs_buff[14]);
696 	sbuf_printf(sb, "\tTDH\t %08x\n", regs_buff[15]);
697 	sbuf_printf(sb, "\tTDT\t %08x\n", regs_buff[16]);
698 	sbuf_printf(sb, "\tTXDCTL\t %08x\n", regs_buff[17]);
699 	sbuf_printf(sb, "\tTDFH\t %08x\n", regs_buff[18]);
700 	sbuf_printf(sb, "\tTDFT\t %08x\n", regs_buff[19]);
701 	sbuf_printf(sb, "\tTDFHS\t %08x\n", regs_buff[20]);
702 	sbuf_printf(sb, "\tTDFPC\t %08x\n\n", regs_buff[21]);
703 
704 	free(regs_buff, M_DEVBUF);
705 
706 #ifdef DUMP_DESCS
707 	{
708 		if_softc_ctx_t scctx = sc->shared;
709 		struct rx_ring *rxr = &rx_que->rxr;
710 		struct tx_ring *txr = &tx_que->txr;
711 		int ntxd = scctx->isc_ntxd[0];
712 		int nrxd = scctx->isc_nrxd[0];
713 		int j;
714 
715 	for (j = 0; j < nrxd; j++) {
716 		u32 staterr = le32toh(rxr->rx_base[j].wb.upper.status_error);
717 		u32 length =  le32toh(rxr->rx_base[j].wb.upper.length);
718 		sbuf_printf(sb, "\tReceive Descriptor Address %d: %08" PRIx64 "  Error:%d  Length:%d\n", j, rxr->rx_base[j].read.buffer_addr, staterr, length);
719 	}
720 
721 	for (j = 0; j < min(ntxd, 256); j++) {
722 		unsigned int *ptr = (unsigned int *)&txr->tx_base[j];
723 
724 		sbuf_printf(sb, "\tTXD[%03d] [0]: %08x [1]: %08x [2]: %08x [3]: %08x  eop: %d DD=%d\n",
725 			    j, ptr[0], ptr[1], ptr[2], ptr[3], buf->eop,
726 			    buf->eop != -1 ? txr->tx_base[buf->eop].upper.fields.status & E1000_TXD_STAT_DD : 0);
727 
728 	}
729 	}
730 #endif
731 
732 	rc = sbuf_finish(sb);
733 	sbuf_delete(sb);
734 	return(rc);
735 }
736 
737 static void *
738 em_register(device_t dev)
739 {
740 	return (&em_sctx_init);
741 }
742 
743 static void *
744 igb_register(device_t dev)
745 {
746 	return (&igb_sctx_init);
747 }
748 
749 static int
750 em_set_num_queues(if_ctx_t ctx)
751 {
752 	struct e1000_softc *sc = iflib_get_softc(ctx);
753 	int maxqueues;
754 
755 	/* Sanity check based on HW */
756 	switch (sc->hw.mac.type) {
757 	case e1000_82576:
758 	case e1000_82580:
759 	case e1000_i350:
760 	case e1000_i354:
761 		maxqueues = 8;
762 		break;
763 	case e1000_i210:
764 	case e1000_82575:
765 		maxqueues = 4;
766 		break;
767 	case e1000_i211:
768 	case e1000_82574:
769 		maxqueues = 2;
770 		break;
771 	default:
772 		maxqueues = 1;
773 		break;
774 	}
775 
776 	return (maxqueues);
777 }
778 
779 #define LEM_CAPS \
780     IFCAP_HWCSUM | IFCAP_VLAN_MTU | IFCAP_VLAN_HWTAGGING | \
781     IFCAP_VLAN_HWCSUM | IFCAP_WOL | IFCAP_VLAN_HWFILTER | IFCAP_TSO4 | \
782     IFCAP_LRO | IFCAP_VLAN_HWTSO | IFCAP_JUMBO_MTU | IFCAP_HWCSUM_IPV6
783 
784 #define EM_CAPS \
785     IFCAP_HWCSUM | IFCAP_VLAN_MTU | IFCAP_VLAN_HWTAGGING | \
786     IFCAP_VLAN_HWCSUM | IFCAP_WOL | IFCAP_VLAN_HWFILTER | IFCAP_TSO4 | \
787     IFCAP_LRO | IFCAP_VLAN_HWTSO | IFCAP_JUMBO_MTU | IFCAP_HWCSUM_IPV6 | \
788     IFCAP_TSO6
789 
790 #define IGB_CAPS \
791     IFCAP_HWCSUM | IFCAP_VLAN_MTU | IFCAP_VLAN_HWTAGGING | \
792     IFCAP_VLAN_HWCSUM | IFCAP_WOL | IFCAP_VLAN_HWFILTER | IFCAP_TSO4 | \
793     IFCAP_LRO | IFCAP_VLAN_HWTSO | IFCAP_JUMBO_MTU | IFCAP_HWCSUM_IPV6 | \
794     IFCAP_TSO6
795 
796 /*********************************************************************
797  *  Device initialization routine
798  *
799  *  The attach entry point is called when the driver is being loaded.
800  *  This routine identifies the type of hardware, allocates all resources
801  *  and initializes the hardware.
802  *
803  *  return 0 on success, positive on failure
804  *********************************************************************/
805 static int
806 em_if_attach_pre(if_ctx_t ctx)
807 {
808 	struct e1000_softc *sc;
809 	if_softc_ctx_t scctx;
810 	device_t dev;
811 	struct e1000_hw *hw;
812 	struct sysctl_oid_list *child;
813 	struct sysctl_ctx_list *ctx_list;
814 	int error = 0;
815 
816 	INIT_DEBUGOUT("em_if_attach_pre: begin");
817 	dev = iflib_get_dev(ctx);
818 	sc = iflib_get_softc(ctx);
819 
820 	sc->ctx = sc->osdep.ctx = ctx;
821 	sc->dev = sc->osdep.dev = dev;
822 	scctx = sc->shared = iflib_get_softc_ctx(ctx);
823 	sc->media = iflib_get_media(ctx);
824 	hw = &sc->hw;
825 
826 	/* Determine hardware and mac info */
827 	em_identify_hardware(ctx);
828 
829 	/* SYSCTL stuff */
830 	ctx_list = device_get_sysctl_ctx(dev);
831 	child = SYSCTL_CHILDREN(device_get_sysctl_tree(dev));
832 
833 	SYSCTL_ADD_PROC(ctx_list, child, OID_AUTO, "nvm",
834 	    CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_NEEDGIANT, sc, 0,
835 	    em_sysctl_nvm_info, "I", "NVM Information");
836 
837 	SYSCTL_ADD_PROC(ctx_list, child, OID_AUTO, "fw_version",
838 	    CTLTYPE_STRING | CTLFLAG_RD | CTLFLAG_NEEDGIANT, sc, 0,
839 	    em_sysctl_print_fw_version, "A",
840 	    "Prints FW/NVM Versions");
841 
842 	SYSCTL_ADD_PROC(ctx_list, child, OID_AUTO, "debug",
843 	    CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_NEEDGIANT, sc, 0,
844 	    em_sysctl_debug_info, "I", "Debug Information");
845 
846 	SYSCTL_ADD_PROC(ctx_list, child, OID_AUTO, "fc",
847 	    CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_NEEDGIANT, sc, 0,
848 	    em_set_flowcntl, "I", "Flow Control");
849 
850 	SYSCTL_ADD_PROC(ctx_list, child, OID_AUTO, "reg_dump",
851 	    CTLTYPE_STRING | CTLFLAG_RD | CTLFLAG_NEEDGIANT, sc, 0,
852 	    em_get_regs, "A", "Dump Registers");
853 
854 	SYSCTL_ADD_PROC(ctx_list, child, OID_AUTO, "rs_dump",
855 	    CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_NEEDGIANT, sc, 0,
856 	    em_get_rs, "I", "Dump RS indexes");
857 
858 	scctx->isc_tx_nsegments = EM_MAX_SCATTER;
859 	scctx->isc_nrxqsets_max = scctx->isc_ntxqsets_max = em_set_num_queues(ctx);
860 	if (bootverbose)
861 		device_printf(dev, "attach_pre capping queues at %d\n",
862 		    scctx->isc_ntxqsets_max);
863 
864 	if (hw->mac.type >= igb_mac_min) {
865 		scctx->isc_txqsizes[0] = roundup2(scctx->isc_ntxd[0] * sizeof(union e1000_adv_tx_desc), EM_DBA_ALIGN);
866 		scctx->isc_rxqsizes[0] = roundup2(scctx->isc_nrxd[0] * sizeof(union e1000_adv_rx_desc), EM_DBA_ALIGN);
867 		scctx->isc_txd_size[0] = sizeof(union e1000_adv_tx_desc);
868 		scctx->isc_rxd_size[0] = sizeof(union e1000_adv_rx_desc);
869 		scctx->isc_txrx = &igb_txrx;
870 		scctx->isc_tx_tso_segments_max = EM_MAX_SCATTER;
871 		scctx->isc_tx_tso_size_max = EM_TSO_SIZE;
872 		scctx->isc_tx_tso_segsize_max = EM_TSO_SEG_SIZE;
873 		scctx->isc_capabilities = scctx->isc_capenable = IGB_CAPS;
874 		scctx->isc_tx_csum_flags = CSUM_TCP | CSUM_UDP | CSUM_TSO |
875 		     CSUM_IP6_TCP | CSUM_IP6_UDP;
876 		if (hw->mac.type != e1000_82575)
877 			scctx->isc_tx_csum_flags |= CSUM_SCTP | CSUM_IP6_SCTP;
878 		/*
879 		** Some new devices, as with ixgbe, now may
880 		** use a different BAR, so we need to keep
881 		** track of which is used.
882 		*/
883 		scctx->isc_msix_bar = pci_msix_table_bar(dev);
884 	} else if (hw->mac.type >= em_mac_min) {
885 		scctx->isc_txqsizes[0] = roundup2(scctx->isc_ntxd[0]* sizeof(struct e1000_tx_desc), EM_DBA_ALIGN);
886 		scctx->isc_rxqsizes[0] = roundup2(scctx->isc_nrxd[0] * sizeof(union e1000_rx_desc_extended), EM_DBA_ALIGN);
887 		scctx->isc_txd_size[0] = sizeof(struct e1000_tx_desc);
888 		scctx->isc_rxd_size[0] = sizeof(union e1000_rx_desc_extended);
889 		scctx->isc_txrx = &em_txrx;
890 		scctx->isc_tx_tso_segments_max = EM_MAX_SCATTER;
891 		scctx->isc_tx_tso_size_max = EM_TSO_SIZE;
892 		scctx->isc_tx_tso_segsize_max = EM_TSO_SEG_SIZE;
893 		scctx->isc_capabilities = scctx->isc_capenable = EM_CAPS;
894 		scctx->isc_tx_csum_flags = CSUM_TCP | CSUM_UDP | CSUM_IP_TSO |
895 		    CSUM_IP6_TCP | CSUM_IP6_UDP;
896 
897 		/* Disable TSO on all em(4) until ring stalls can be debugged */
898 		scctx->isc_capenable &= ~IFCAP_TSO;
899 
900 		/*
901 		 * Disable TSO on SPT due to errata that downclocks DMA performance
902 		 * i218-i219 Specification Update 1.5.4.5
903 		 */
904 		if (hw->mac.type == e1000_pch_spt)
905 			scctx->isc_capenable &= ~IFCAP_TSO;
906 
907 		/*
908 		 * We support MSI-X with 82574 only, but indicate to iflib(4)
909 		 * that it shall give MSI at least a try with other devices.
910 		 */
911 		if (hw->mac.type == e1000_82574) {
912 			scctx->isc_msix_bar = pci_msix_table_bar(dev);
913 		} else {
914 			scctx->isc_msix_bar = -1;
915 			scctx->isc_disable_msix = 1;
916 		}
917 	} else {
918 		scctx->isc_txqsizes[0] = roundup2((scctx->isc_ntxd[0] + 1) * sizeof(struct e1000_tx_desc), EM_DBA_ALIGN);
919 		scctx->isc_rxqsizes[0] = roundup2((scctx->isc_nrxd[0] + 1) * sizeof(struct e1000_rx_desc), EM_DBA_ALIGN);
920 		scctx->isc_txd_size[0] = sizeof(struct e1000_tx_desc);
921 		scctx->isc_rxd_size[0] = sizeof(struct e1000_rx_desc);
922 		scctx->isc_txrx = &lem_txrx;
923 		scctx->isc_tx_tso_segments_max = EM_MAX_SCATTER;
924 		scctx->isc_tx_tso_size_max = EM_TSO_SIZE;
925 		scctx->isc_tx_tso_segsize_max = EM_TSO_SEG_SIZE;
926 		scctx->isc_capabilities = scctx->isc_capenable = LEM_CAPS;
927 		if (em_unsupported_tso)
928 			scctx->isc_capabilities |= IFCAP_TSO6;
929 		scctx->isc_tx_csum_flags = CSUM_TCP | CSUM_UDP | CSUM_IP_TSO |
930 		    CSUM_IP6_TCP | CSUM_IP6_UDP;
931 
932 		/* Disable TSO on all lem(4) until ring stalls can be debugged */
933 		scctx->isc_capenable &= ~IFCAP_TSO;
934 
935 		/* 82541ER doesn't do HW tagging */
936 		if (hw->device_id == E1000_DEV_ID_82541ER ||
937 		    hw->device_id == E1000_DEV_ID_82541ER_LOM) {
938 			scctx->isc_capabilities &= ~IFCAP_VLAN_HWTAGGING;
939 			scctx->isc_capenable = scctx->isc_capabilities;
940 		}
941 		/* This is the first e1000 chip and it does not do offloads */
942 		if (hw->mac.type == e1000_82542) {
943 			scctx->isc_capabilities &= ~(IFCAP_HWCSUM | IFCAP_VLAN_HWCSUM |
944 			    IFCAP_HWCSUM_IPV6 | IFCAP_VLAN_HWTAGGING |
945 			    IFCAP_VLAN_HWFILTER | IFCAP_TSO | IFCAP_VLAN_HWTSO);
946 			scctx->isc_capenable = scctx->isc_capabilities;
947 		}
948 		/* These can't do TSO for various reasons */
949 		if (hw->mac.type < e1000_82544 || hw->mac.type == e1000_82547 ||
950 		    hw->mac.type == e1000_82547_rev_2) {
951 			scctx->isc_capabilities &= ~(IFCAP_TSO | IFCAP_VLAN_HWTSO);
952 			scctx->isc_capenable = scctx->isc_capabilities;
953 		}
954 		/* XXXKB: No IPv6 before this? */
955 		if (hw->mac.type < e1000_82545){
956 			scctx->isc_capabilities &= ~IFCAP_HWCSUM_IPV6;
957 			scctx->isc_capenable = scctx->isc_capabilities;
958 		}
959 		/* "PCI/PCI-X SDM 4.0" page 33 (b) - FDX requirement on these chips */
960 		if (hw->mac.type == e1000_82547 || hw->mac.type == e1000_82547_rev_2)
961 			scctx->isc_capenable &= ~(IFCAP_HWCSUM | IFCAP_VLAN_HWCSUM |
962 			    IFCAP_HWCSUM_IPV6);
963 
964 		/* INTx only */
965 		scctx->isc_msix_bar = 0;
966 	}
967 
968 	/* Setup PCI resources */
969 	if (em_allocate_pci_resources(ctx)) {
970 		device_printf(dev, "Allocation of PCI resources failed\n");
971 		error = ENXIO;
972 		goto err_pci;
973 	}
974 
975 	/*
976 	** For ICH8 and family we need to
977 	** map the flash memory, and this
978 	** must happen after the MAC is
979 	** identified
980 	*/
981 	if ((hw->mac.type == e1000_ich8lan) ||
982 	    (hw->mac.type == e1000_ich9lan) ||
983 	    (hw->mac.type == e1000_ich10lan) ||
984 	    (hw->mac.type == e1000_pchlan) ||
985 	    (hw->mac.type == e1000_pch2lan) ||
986 	    (hw->mac.type == e1000_pch_lpt)) {
987 		int rid = EM_BAR_TYPE_FLASH;
988 		sc->flash = bus_alloc_resource_any(dev,
989 		    SYS_RES_MEMORY, &rid, RF_ACTIVE);
990 		if (sc->flash == NULL) {
991 			device_printf(dev, "Mapping of Flash failed\n");
992 			error = ENXIO;
993 			goto err_pci;
994 		}
995 		/* This is used in the shared code */
996 		hw->flash_address = (u8 *)sc->flash;
997 		sc->osdep.flash_bus_space_tag =
998 		    rman_get_bustag(sc->flash);
999 		sc->osdep.flash_bus_space_handle =
1000 		    rman_get_bushandle(sc->flash);
1001 	}
1002 	/*
1003 	** In the new SPT device flash is not  a
1004 	** separate BAR, rather it is also in BAR0,
1005 	** so use the same tag and an offset handle for the
1006 	** FLASH read/write macros in the shared code.
1007 	*/
1008 	else if (hw->mac.type >= e1000_pch_spt) {
1009 		sc->osdep.flash_bus_space_tag =
1010 		    sc->osdep.mem_bus_space_tag;
1011 		sc->osdep.flash_bus_space_handle =
1012 		    sc->osdep.mem_bus_space_handle
1013 		    + E1000_FLASH_BASE_ADDR;
1014 	}
1015 
1016 	/* Do Shared Code initialization */
1017 	error = e1000_setup_init_funcs(hw, true);
1018 	if (error) {
1019 		device_printf(dev, "Setup of Shared code failed, error %d\n",
1020 		    error);
1021 		error = ENXIO;
1022 		goto err_pci;
1023 	}
1024 
1025 	em_setup_msix(ctx);
1026 	e1000_get_bus_info(hw);
1027 
1028 	/* Set up some sysctls for the tunable interrupt delays */
1029 	em_add_int_delay_sysctl(sc, "rx_int_delay",
1030 	    "receive interrupt delay in usecs", &sc->rx_int_delay,
1031 	    E1000_REGISTER(hw, E1000_RDTR), em_rx_int_delay_dflt);
1032 	em_add_int_delay_sysctl(sc, "tx_int_delay",
1033 	    "transmit interrupt delay in usecs", &sc->tx_int_delay,
1034 	    E1000_REGISTER(hw, E1000_TIDV), em_tx_int_delay_dflt);
1035 	em_add_int_delay_sysctl(sc, "rx_abs_int_delay",
1036 	    "receive interrupt delay limit in usecs",
1037 	    &sc->rx_abs_int_delay,
1038 	    E1000_REGISTER(hw, E1000_RADV),
1039 	    em_rx_abs_int_delay_dflt);
1040 	em_add_int_delay_sysctl(sc, "tx_abs_int_delay",
1041 	    "transmit interrupt delay limit in usecs",
1042 	    &sc->tx_abs_int_delay,
1043 	    E1000_REGISTER(hw, E1000_TADV),
1044 	    em_tx_abs_int_delay_dflt);
1045 	em_add_int_delay_sysctl(sc, "itr",
1046 	    "interrupt delay limit in usecs/4",
1047 	    &sc->tx_itr,
1048 	    E1000_REGISTER(hw, E1000_ITR),
1049 	    DEFAULT_ITR);
1050 
1051 	hw->mac.autoneg = DO_AUTO_NEG;
1052 	hw->phy.autoneg_wait_to_complete = false;
1053 	hw->phy.autoneg_advertised = AUTONEG_ADV_DEFAULT;
1054 
1055 	if (hw->mac.type < em_mac_min) {
1056 		e1000_init_script_state_82541(hw, true);
1057 		e1000_set_tbi_compatibility_82543(hw, true);
1058 	}
1059 	/* Copper options */
1060 	if (hw->phy.media_type == e1000_media_type_copper) {
1061 		hw->phy.mdix = AUTO_ALL_MODES;
1062 		hw->phy.disable_polarity_correction = false;
1063 		hw->phy.ms_type = EM_MASTER_SLAVE;
1064 	}
1065 
1066 	/*
1067 	 * Set the frame limits assuming
1068 	 * standard ethernet sized frames.
1069 	 */
1070 	scctx->isc_max_frame_size = hw->mac.max_frame_size =
1071 	    ETHERMTU + ETHER_HDR_LEN + ETHERNET_FCS_SIZE;
1072 
1073 	/*
1074 	 * This controls when hardware reports transmit completion
1075 	 * status.
1076 	 */
1077 	hw->mac.report_tx_early = 1;
1078 
1079 	/* Allocate multicast array memory. */
1080 	sc->mta = malloc(sizeof(u8) * ETHER_ADDR_LEN *
1081 	    MAX_NUM_MULTICAST_ADDRESSES, M_DEVBUF, M_NOWAIT);
1082 	if (sc->mta == NULL) {
1083 		device_printf(dev, "Can not allocate multicast setup array\n");
1084 		error = ENOMEM;
1085 		goto err_late;
1086 	}
1087 
1088 	/* Clear the IFCAP_TSO auto mask */
1089 	sc->tso_automasked = 0;
1090 
1091 	/* Check SOL/IDER usage */
1092 	if (e1000_check_reset_block(hw))
1093 		device_printf(dev, "PHY reset is blocked"
1094 			      " due to SOL/IDER session.\n");
1095 
1096 	/* Sysctl for setting Energy Efficient Ethernet */
1097 	hw->dev_spec.ich8lan.eee_disable = eee_setting;
1098 	SYSCTL_ADD_PROC(ctx_list, child, OID_AUTO, "eee_control",
1099 	    CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_NEEDGIANT, sc, 0,
1100 	    em_sysctl_eee, "I", "Disable Energy Efficient Ethernet");
1101 
1102 	/*
1103 	** Start from a known state, this is
1104 	** important in reading the nvm and
1105 	** mac from that.
1106 	*/
1107 	e1000_reset_hw(hw);
1108 
1109 	/* Make sure we have a good EEPROM before we read from it */
1110 	if (e1000_validate_nvm_checksum(hw) < 0) {
1111 		/*
1112 		** Some PCI-E parts fail the first check due to
1113 		** the link being in sleep state, call it again,
1114 		** if it fails a second time its a real issue.
1115 		*/
1116 		if (e1000_validate_nvm_checksum(hw) < 0) {
1117 			device_printf(dev,
1118 			    "The EEPROM Checksum Is Not Valid\n");
1119 			error = EIO;
1120 			goto err_late;
1121 		}
1122 	}
1123 
1124 	/* Copy the permanent MAC address out of the EEPROM */
1125 	if (e1000_read_mac_addr(hw) < 0) {
1126 		device_printf(dev, "EEPROM read error while reading MAC"
1127 			      " address\n");
1128 		error = EIO;
1129 		goto err_late;
1130 	}
1131 
1132 	if (!em_is_valid_ether_addr(hw->mac.addr)) {
1133 		if (sc->vf_ifp) {
1134 			ether_gen_addr(iflib_get_ifp(ctx),
1135 			    (struct ether_addr *)hw->mac.addr);
1136 		} else {
1137 			device_printf(dev, "Invalid MAC address\n");
1138 			error = EIO;
1139 			goto err_late;
1140 		}
1141 	}
1142 
1143 	/* Save the EEPROM/NVM versions, must be done under IFLIB_CTX_LOCK */
1144 	em_fw_version_locked(ctx);
1145 
1146 	em_print_fw_version(sc);
1147 
1148 	/*
1149 	 * Get Wake-on-Lan and Management info for later use
1150 	 */
1151 	em_get_wakeup(ctx);
1152 
1153 	/* Enable only WOL MAGIC by default */
1154 	scctx->isc_capenable &= ~IFCAP_WOL;
1155 	if (sc->wol != 0)
1156 		scctx->isc_capenable |= IFCAP_WOL_MAGIC;
1157 
1158 	iflib_set_mac(ctx, hw->mac.addr);
1159 
1160 	return (0);
1161 
1162 err_late:
1163 	em_release_hw_control(sc);
1164 err_pci:
1165 	em_free_pci_resources(ctx);
1166 	free(sc->mta, M_DEVBUF);
1167 
1168 	return (error);
1169 }
1170 
1171 static int
1172 em_if_attach_post(if_ctx_t ctx)
1173 {
1174 	struct e1000_softc *sc = iflib_get_softc(ctx);
1175 	struct e1000_hw *hw = &sc->hw;
1176 	int error = 0;
1177 
1178 	/* Setup OS specific network interface */
1179 	error = em_setup_interface(ctx);
1180 	if (error != 0) {
1181 		device_printf(sc->dev, "Interface setup failed: %d\n", error);
1182 		goto err_late;
1183 	}
1184 
1185 	em_reset(ctx);
1186 
1187 	/* Initialize statistics */
1188 	em_update_stats_counters(sc);
1189 	hw->mac.get_link_status = 1;
1190 	em_if_update_admin_status(ctx);
1191 	em_add_hw_stats(sc);
1192 
1193 	/* Non-AMT based hardware can now take control from firmware */
1194 	if (sc->has_manage && !sc->has_amt)
1195 		em_get_hw_control(sc);
1196 
1197 	INIT_DEBUGOUT("em_if_attach_post: end");
1198 
1199 	return (0);
1200 
1201 err_late:
1202 	/* upon attach_post() error, iflib calls _if_detach() to free resources. */
1203 	return (error);
1204 }
1205 
1206 /*********************************************************************
1207  *  Device removal routine
1208  *
1209  *  The detach entry point is called when the driver is being removed.
1210  *  This routine stops the adapter and deallocates all the resources
1211  *  that were allocated for driver operation.
1212  *
1213  *  return 0 on success, positive on failure
1214  *********************************************************************/
1215 static int
1216 em_if_detach(if_ctx_t ctx)
1217 {
1218 	struct e1000_softc	*sc = iflib_get_softc(ctx);
1219 
1220 	INIT_DEBUGOUT("em_if_detach: begin");
1221 
1222 	e1000_phy_hw_reset(&sc->hw);
1223 
1224 	em_release_manageability(sc);
1225 	em_release_hw_control(sc);
1226 	em_free_pci_resources(ctx);
1227 	free(sc->mta, M_DEVBUF);
1228 	sc->mta = NULL;
1229 
1230 	return (0);
1231 }
1232 
1233 /*********************************************************************
1234  *
1235  *  Shutdown entry point
1236  *
1237  **********************************************************************/
1238 
1239 static int
1240 em_if_shutdown(if_ctx_t ctx)
1241 {
1242 	return em_if_suspend(ctx);
1243 }
1244 
1245 /*
1246  * Suspend/resume device methods.
1247  */
1248 static int
1249 em_if_suspend(if_ctx_t ctx)
1250 {
1251 	struct e1000_softc *sc = iflib_get_softc(ctx);
1252 
1253 	em_release_manageability(sc);
1254 	em_release_hw_control(sc);
1255 	em_enable_wakeup(ctx);
1256 	return (0);
1257 }
1258 
1259 static int
1260 em_if_resume(if_ctx_t ctx)
1261 {
1262 	struct e1000_softc *sc = iflib_get_softc(ctx);
1263 
1264 	if (sc->hw.mac.type == e1000_pch2lan)
1265 		e1000_resume_workarounds_pchlan(&sc->hw);
1266 	em_if_init(ctx);
1267 	em_init_manageability(sc);
1268 
1269 	return(0);
1270 }
1271 
1272 static int
1273 em_if_mtu_set(if_ctx_t ctx, uint32_t mtu)
1274 {
1275 	int max_frame_size;
1276 	struct e1000_softc *sc = iflib_get_softc(ctx);
1277 	if_softc_ctx_t scctx = iflib_get_softc_ctx(ctx);
1278 
1279 	IOCTL_DEBUGOUT("ioctl rcv'd: SIOCSIFMTU (Set Interface MTU)");
1280 
1281 	switch (sc->hw.mac.type) {
1282 	case e1000_82571:
1283 	case e1000_82572:
1284 	case e1000_ich9lan:
1285 	case e1000_ich10lan:
1286 	case e1000_pch2lan:
1287 	case e1000_pch_lpt:
1288 	case e1000_pch_spt:
1289 	case e1000_pch_cnp:
1290 	case e1000_pch_tgp:
1291 	case e1000_pch_adp:
1292 	case e1000_pch_mtp:
1293 	case e1000_pch_ptp:
1294 	case e1000_82574:
1295 	case e1000_82583:
1296 	case e1000_80003es2lan:
1297 		/* 9K Jumbo Frame size */
1298 		max_frame_size = 9234;
1299 		break;
1300 	case e1000_pchlan:
1301 		max_frame_size = 4096;
1302 		break;
1303 	case e1000_82542:
1304 	case e1000_ich8lan:
1305 		/* Adapters that do not support jumbo frames */
1306 		max_frame_size = ETHER_MAX_LEN;
1307 		break;
1308 	default:
1309 		if (sc->hw.mac.type >= igb_mac_min)
1310 			max_frame_size = 9234;
1311 		else /* lem */
1312 			max_frame_size = MAX_JUMBO_FRAME_SIZE;
1313 	}
1314 	if (mtu > max_frame_size - ETHER_HDR_LEN - ETHER_CRC_LEN) {
1315 		return (EINVAL);
1316 	}
1317 
1318 	scctx->isc_max_frame_size = sc->hw.mac.max_frame_size =
1319 	    mtu + ETHER_HDR_LEN + ETHER_CRC_LEN;
1320 	return (0);
1321 }
1322 
1323 /*********************************************************************
1324  *  Init entry point
1325  *
1326  *  This routine is used in two ways. It is used by the stack as
1327  *  init entry point in network interface structure. It is also used
1328  *  by the driver as a hw/sw initialization routine to get to a
1329  *  consistent state.
1330  *
1331  **********************************************************************/
1332 static void
1333 em_if_init(if_ctx_t ctx)
1334 {
1335 	struct e1000_softc *sc = iflib_get_softc(ctx);
1336 	if_softc_ctx_t scctx = sc->shared;
1337 	if_t ifp = iflib_get_ifp(ctx);
1338 	struct em_tx_queue *tx_que;
1339 	int i;
1340 
1341 	INIT_DEBUGOUT("em_if_init: begin");
1342 
1343 	/* Get the latest mac address, User can use a LAA */
1344 	bcopy(if_getlladdr(ifp), sc->hw.mac.addr,
1345 	    ETHER_ADDR_LEN);
1346 
1347 	/* Put the address into the Receive Address Array */
1348 	e1000_rar_set(&sc->hw, sc->hw.mac.addr, 0);
1349 
1350 	/*
1351 	 * With the 82571 adapter, RAR[0] may be overwritten
1352 	 * when the other port is reset, we make a duplicate
1353 	 * in RAR[14] for that eventuality, this assures
1354 	 * the interface continues to function.
1355 	 */
1356 	if (sc->hw.mac.type == e1000_82571) {
1357 		e1000_set_laa_state_82571(&sc->hw, true);
1358 		e1000_rar_set(&sc->hw, sc->hw.mac.addr,
1359 		    E1000_RAR_ENTRIES - 1);
1360 	}
1361 
1362 	/* Initialize the hardware */
1363 	em_reset(ctx);
1364 	em_if_update_admin_status(ctx);
1365 
1366 	for (i = 0, tx_que = sc->tx_queues; i < sc->tx_num_queues; i++, tx_que++) {
1367 		struct tx_ring *txr = &tx_que->txr;
1368 
1369 		txr->tx_rs_cidx = txr->tx_rs_pidx;
1370 
1371 		/* Initialize the last processed descriptor to be the end of
1372 		 * the ring, rather than the start, so that we avoid an
1373 		 * off-by-one error when calculating how many descriptors are
1374 		 * done in the credits_update function.
1375 		 */
1376 		txr->tx_cidx_processed = scctx->isc_ntxd[0] - 1;
1377 	}
1378 
1379 	/* Setup VLAN support, basic and offload if available */
1380 	E1000_WRITE_REG(&sc->hw, E1000_VET, ETHERTYPE_VLAN);
1381 
1382 	/* Clear bad data from Rx FIFOs */
1383 	if (sc->hw.mac.type >= igb_mac_min)
1384 		e1000_rx_fifo_flush_base(&sc->hw);
1385 
1386 	/* Configure for OS presence */
1387 	em_init_manageability(sc);
1388 
1389 	/* Prepare transmit descriptors and buffers */
1390 	em_initialize_transmit_unit(ctx);
1391 
1392 	/* Setup Multicast table */
1393 	em_if_multi_set(ctx);
1394 
1395 	sc->rx_mbuf_sz = iflib_get_rx_mbuf_sz(ctx);
1396 	em_initialize_receive_unit(ctx);
1397 
1398 	/* Set up VLAN support and filter */
1399 	em_setup_vlan_hw_support(ctx);
1400 
1401 	/* Don't lose promiscuous settings */
1402 	em_if_set_promisc(ctx, if_getflags(ifp));
1403 	e1000_clear_hw_cntrs_base_generic(&sc->hw);
1404 
1405 	/* MSI-X configuration for 82574 */
1406 	if (sc->hw.mac.type == e1000_82574) {
1407 		int tmp = E1000_READ_REG(&sc->hw, E1000_CTRL_EXT);
1408 
1409 		tmp |= E1000_CTRL_EXT_PBA_CLR;
1410 		E1000_WRITE_REG(&sc->hw, E1000_CTRL_EXT, tmp);
1411 		/* Set the IVAR - interrupt vector routing. */
1412 		E1000_WRITE_REG(&sc->hw, E1000_IVAR, sc->ivars);
1413 	} else if (sc->intr_type == IFLIB_INTR_MSIX) /* Set up queue routing */
1414 		igb_configure_queues(sc);
1415 
1416 	/* this clears any pending interrupts */
1417 	E1000_READ_REG(&sc->hw, E1000_ICR);
1418 	E1000_WRITE_REG(&sc->hw, E1000_ICS, E1000_ICS_LSC);
1419 
1420 	/* AMT based hardware can now take control from firmware */
1421 	if (sc->has_manage && sc->has_amt)
1422 		em_get_hw_control(sc);
1423 
1424 	/* Set Energy Efficient Ethernet */
1425 	if (sc->hw.mac.type >= igb_mac_min &&
1426 	    sc->hw.phy.media_type == e1000_media_type_copper) {
1427 		if (sc->hw.mac.type == e1000_i354)
1428 			e1000_set_eee_i354(&sc->hw, true, true);
1429 		else
1430 			e1000_set_eee_i350(&sc->hw, true, true);
1431 	}
1432 }
1433 
1434 /*********************************************************************
1435  *
1436  *  Fast Legacy/MSI Combined Interrupt Service routine
1437  *
1438  *********************************************************************/
1439 int
1440 em_intr(void *arg)
1441 {
1442 	struct e1000_softc *sc = arg;
1443 	if_ctx_t ctx = sc->ctx;
1444 	u32 reg_icr;
1445 
1446 	reg_icr = E1000_READ_REG(&sc->hw, E1000_ICR);
1447 
1448 	/* Hot eject? */
1449 	if (reg_icr == 0xffffffff)
1450 		return FILTER_STRAY;
1451 
1452 	/* Definitely not our interrupt. */
1453 	if (reg_icr == 0x0)
1454 		return FILTER_STRAY;
1455 
1456 	/*
1457 	 * Starting with the 82571 chip, bit 31 should be used to
1458 	 * determine whether the interrupt belongs to us.
1459 	 */
1460 	if (sc->hw.mac.type >= e1000_82571 &&
1461 	    (reg_icr & E1000_ICR_INT_ASSERTED) == 0)
1462 		return FILTER_STRAY;
1463 
1464 	/*
1465 	 * Only MSI-X interrupts have one-shot behavior by taking advantage
1466 	 * of the EIAC register.  Thus, explicitly disable interrupts.  This
1467 	 * also works around the MSI message reordering errata on certain
1468 	 * systems.
1469 	 */
1470 	IFDI_INTR_DISABLE(ctx);
1471 
1472 	/* Link status change */
1473 	if (reg_icr & (E1000_ICR_RXSEQ | E1000_ICR_LSC))
1474 		em_handle_link(ctx);
1475 
1476 	if (reg_icr & E1000_ICR_RXO)
1477 		sc->rx_overruns++;
1478 
1479 	return (FILTER_SCHEDULE_THREAD);
1480 }
1481 
1482 static int
1483 em_if_rx_queue_intr_enable(if_ctx_t ctx, uint16_t rxqid)
1484 {
1485 	struct e1000_softc *sc = iflib_get_softc(ctx);
1486 	struct em_rx_queue *rxq = &sc->rx_queues[rxqid];
1487 
1488 	E1000_WRITE_REG(&sc->hw, E1000_IMS, rxq->eims);
1489 	return (0);
1490 }
1491 
1492 static int
1493 em_if_tx_queue_intr_enable(if_ctx_t ctx, uint16_t txqid)
1494 {
1495 	struct e1000_softc *sc = iflib_get_softc(ctx);
1496 	struct em_tx_queue *txq = &sc->tx_queues[txqid];
1497 
1498 	E1000_WRITE_REG(&sc->hw, E1000_IMS, txq->eims);
1499 	return (0);
1500 }
1501 
1502 static int
1503 igb_if_rx_queue_intr_enable(if_ctx_t ctx, uint16_t rxqid)
1504 {
1505 	struct e1000_softc *sc = iflib_get_softc(ctx);
1506 	struct em_rx_queue *rxq = &sc->rx_queues[rxqid];
1507 
1508 	E1000_WRITE_REG(&sc->hw, E1000_EIMS, rxq->eims);
1509 	return (0);
1510 }
1511 
1512 static int
1513 igb_if_tx_queue_intr_enable(if_ctx_t ctx, uint16_t txqid)
1514 {
1515 	struct e1000_softc *sc = iflib_get_softc(ctx);
1516 	struct em_tx_queue *txq = &sc->tx_queues[txqid];
1517 
1518 	E1000_WRITE_REG(&sc->hw, E1000_EIMS, txq->eims);
1519 	return (0);
1520 }
1521 
1522 /*********************************************************************
1523  *
1524  *  MSI-X RX Interrupt Service routine
1525  *
1526  **********************************************************************/
1527 static int
1528 em_msix_que(void *arg)
1529 {
1530 	struct em_rx_queue *que = arg;
1531 
1532 	++que->irqs;
1533 
1534 	return (FILTER_SCHEDULE_THREAD);
1535 }
1536 
1537 /*********************************************************************
1538  *
1539  *  MSI-X Link Fast Interrupt Service routine
1540  *
1541  **********************************************************************/
1542 static int
1543 em_msix_link(void *arg)
1544 {
1545 	struct e1000_softc *sc = arg;
1546 	u32 reg_icr;
1547 
1548 	++sc->link_irq;
1549 	MPASS(sc->hw.back != NULL);
1550 	reg_icr = E1000_READ_REG(&sc->hw, E1000_ICR);
1551 
1552 	if (reg_icr & E1000_ICR_RXO)
1553 		sc->rx_overruns++;
1554 
1555 	if (reg_icr & (E1000_ICR_RXSEQ | E1000_ICR_LSC))
1556 		em_handle_link(sc->ctx);
1557 
1558 	/* Re-arm unconditionally */
1559 	if (sc->hw.mac.type >= igb_mac_min) {
1560 		E1000_WRITE_REG(&sc->hw, E1000_IMS, E1000_IMS_LSC);
1561 		E1000_WRITE_REG(&sc->hw, E1000_EIMS, sc->link_mask);
1562 	} else if (sc->hw.mac.type == e1000_82574) {
1563 		E1000_WRITE_REG(&sc->hw, E1000_IMS, E1000_IMS_LSC |
1564 		    E1000_IMS_OTHER);
1565 		/*
1566 		 * Because we must read the ICR for this interrupt it may
1567 		 * clear other causes using autoclear, for this reason we
1568 		 * simply create a soft interrupt for all these vectors.
1569 		 */
1570 		if (reg_icr)
1571 			E1000_WRITE_REG(&sc->hw, E1000_ICS, sc->ims);
1572 	} else
1573 		E1000_WRITE_REG(&sc->hw, E1000_IMS, E1000_IMS_LSC);
1574 
1575 	return (FILTER_HANDLED);
1576 }
1577 
1578 static void
1579 em_handle_link(void *context)
1580 {
1581 	if_ctx_t ctx = context;
1582 	struct e1000_softc *sc = iflib_get_softc(ctx);
1583 
1584 	sc->hw.mac.get_link_status = 1;
1585 	iflib_admin_intr_deferred(ctx);
1586 }
1587 
1588 /*********************************************************************
1589  *
1590  *  Media Ioctl callback
1591  *
1592  *  This routine is called whenever the user queries the status of
1593  *  the interface using ifconfig.
1594  *
1595  **********************************************************************/
1596 static void
1597 em_if_media_status(if_ctx_t ctx, struct ifmediareq *ifmr)
1598 {
1599 	struct e1000_softc *sc = iflib_get_softc(ctx);
1600 	u_char fiber_type = IFM_1000_SX;
1601 
1602 	INIT_DEBUGOUT("em_if_media_status: begin");
1603 
1604 	iflib_admin_intr_deferred(ctx);
1605 
1606 	ifmr->ifm_status = IFM_AVALID;
1607 	ifmr->ifm_active = IFM_ETHER;
1608 
1609 	if (!sc->link_active) {
1610 		return;
1611 	}
1612 
1613 	ifmr->ifm_status |= IFM_ACTIVE;
1614 
1615 	if ((sc->hw.phy.media_type == e1000_media_type_fiber) ||
1616 	    (sc->hw.phy.media_type == e1000_media_type_internal_serdes)) {
1617 		if (sc->hw.mac.type == e1000_82545)
1618 			fiber_type = IFM_1000_LX;
1619 		ifmr->ifm_active |= fiber_type | IFM_FDX;
1620 	} else {
1621 		switch (sc->link_speed) {
1622 		case 10:
1623 			ifmr->ifm_active |= IFM_10_T;
1624 			break;
1625 		case 100:
1626 			ifmr->ifm_active |= IFM_100_TX;
1627 			break;
1628 		case 1000:
1629 			ifmr->ifm_active |= IFM_1000_T;
1630 			break;
1631 		}
1632 		if (sc->link_duplex == FULL_DUPLEX)
1633 			ifmr->ifm_active |= IFM_FDX;
1634 		else
1635 			ifmr->ifm_active |= IFM_HDX;
1636 	}
1637 }
1638 
1639 /*********************************************************************
1640  *
1641  *  Media Ioctl callback
1642  *
1643  *  This routine is called when the user changes speed/duplex using
1644  *  media/mediopt option with ifconfig.
1645  *
1646  **********************************************************************/
1647 static int
1648 em_if_media_change(if_ctx_t ctx)
1649 {
1650 	struct e1000_softc *sc = iflib_get_softc(ctx);
1651 	struct ifmedia *ifm = iflib_get_media(ctx);
1652 
1653 	INIT_DEBUGOUT("em_if_media_change: begin");
1654 
1655 	if (IFM_TYPE(ifm->ifm_media) != IFM_ETHER)
1656 		return (EINVAL);
1657 
1658 	switch (IFM_SUBTYPE(ifm->ifm_media)) {
1659 	case IFM_AUTO:
1660 		sc->hw.mac.autoneg = DO_AUTO_NEG;
1661 		sc->hw.phy.autoneg_advertised = AUTONEG_ADV_DEFAULT;
1662 		break;
1663 	case IFM_1000_LX:
1664 	case IFM_1000_SX:
1665 	case IFM_1000_T:
1666 		sc->hw.mac.autoneg = DO_AUTO_NEG;
1667 		sc->hw.phy.autoneg_advertised = ADVERTISE_1000_FULL;
1668 		break;
1669 	case IFM_100_TX:
1670 		sc->hw.mac.autoneg = false;
1671 		sc->hw.phy.autoneg_advertised = 0;
1672 		if ((ifm->ifm_media & IFM_GMASK) == IFM_FDX)
1673 			sc->hw.mac.forced_speed_duplex = ADVERTISE_100_FULL;
1674 		else
1675 			sc->hw.mac.forced_speed_duplex = ADVERTISE_100_HALF;
1676 		break;
1677 	case IFM_10_T:
1678 		sc->hw.mac.autoneg = false;
1679 		sc->hw.phy.autoneg_advertised = 0;
1680 		if ((ifm->ifm_media & IFM_GMASK) == IFM_FDX)
1681 			sc->hw.mac.forced_speed_duplex = ADVERTISE_10_FULL;
1682 		else
1683 			sc->hw.mac.forced_speed_duplex = ADVERTISE_10_HALF;
1684 		break;
1685 	default:
1686 		device_printf(sc->dev, "Unsupported media type\n");
1687 	}
1688 
1689 	em_if_init(ctx);
1690 
1691 	return (0);
1692 }
1693 
1694 static int
1695 em_if_set_promisc(if_ctx_t ctx, int flags)
1696 {
1697 	struct e1000_softc *sc = iflib_get_softc(ctx);
1698 	if_t ifp = iflib_get_ifp(ctx);
1699 	u32 reg_rctl;
1700 	int mcnt = 0;
1701 
1702 	reg_rctl = E1000_READ_REG(&sc->hw, E1000_RCTL);
1703 	reg_rctl &= ~(E1000_RCTL_SBP | E1000_RCTL_UPE);
1704 	if (flags & IFF_ALLMULTI)
1705 		mcnt = MAX_NUM_MULTICAST_ADDRESSES;
1706 	else
1707 		mcnt = min(if_llmaddr_count(ifp), MAX_NUM_MULTICAST_ADDRESSES);
1708 
1709 	if (mcnt < MAX_NUM_MULTICAST_ADDRESSES)
1710 		reg_rctl &= (~E1000_RCTL_MPE);
1711 	E1000_WRITE_REG(&sc->hw, E1000_RCTL, reg_rctl);
1712 
1713 	if (flags & IFF_PROMISC) {
1714 		reg_rctl |= (E1000_RCTL_UPE | E1000_RCTL_MPE);
1715 		em_if_vlan_filter_disable(sc);
1716 		/* Turn this on if you want to see bad packets */
1717 		if (em_debug_sbp)
1718 			reg_rctl |= E1000_RCTL_SBP;
1719 		E1000_WRITE_REG(&sc->hw, E1000_RCTL, reg_rctl);
1720 	} else {
1721 		if (flags & IFF_ALLMULTI) {
1722 			reg_rctl |= E1000_RCTL_MPE;
1723 			reg_rctl &= ~E1000_RCTL_UPE;
1724 			E1000_WRITE_REG(&sc->hw, E1000_RCTL, reg_rctl);
1725 		}
1726 		if (em_if_vlan_filter_used(ctx))
1727 			em_if_vlan_filter_enable(sc);
1728 	}
1729 	return (0);
1730 }
1731 
1732 static u_int
1733 em_copy_maddr(void *arg, struct sockaddr_dl *sdl, u_int idx)
1734 {
1735 	u8 *mta = arg;
1736 
1737 	if (idx == MAX_NUM_MULTICAST_ADDRESSES)
1738 		return (0);
1739 
1740 	bcopy(LLADDR(sdl), &mta[idx * ETHER_ADDR_LEN], ETHER_ADDR_LEN);
1741 
1742 	return (1);
1743 }
1744 
1745 /*********************************************************************
1746  *  Multicast Update
1747  *
1748  *  This routine is called whenever multicast address list is updated.
1749  *
1750  **********************************************************************/
1751 static void
1752 em_if_multi_set(if_ctx_t ctx)
1753 {
1754 	struct e1000_softc *sc = iflib_get_softc(ctx);
1755 	if_t ifp = iflib_get_ifp(ctx);
1756 	u8  *mta; /* Multicast array memory */
1757 	u32 reg_rctl = 0;
1758 	int mcnt = 0;
1759 
1760 	IOCTL_DEBUGOUT("em_set_multi: begin");
1761 
1762 	mta = sc->mta;
1763 	bzero(mta, sizeof(u8) * ETHER_ADDR_LEN * MAX_NUM_MULTICAST_ADDRESSES);
1764 
1765 	if (sc->hw.mac.type == e1000_82542 &&
1766 	    sc->hw.revision_id == E1000_REVISION_2) {
1767 		reg_rctl = E1000_READ_REG(&sc->hw, E1000_RCTL);
1768 		if (sc->hw.bus.pci_cmd_word & CMD_MEM_WRT_INVALIDATE)
1769 			e1000_pci_clear_mwi(&sc->hw);
1770 		reg_rctl |= E1000_RCTL_RST;
1771 		E1000_WRITE_REG(&sc->hw, E1000_RCTL, reg_rctl);
1772 		msec_delay(5);
1773 	}
1774 
1775 	mcnt = if_foreach_llmaddr(ifp, em_copy_maddr, mta);
1776 
1777 	if (mcnt < MAX_NUM_MULTICAST_ADDRESSES)
1778 		e1000_update_mc_addr_list(&sc->hw, mta, mcnt);
1779 
1780 	reg_rctl = E1000_READ_REG(&sc->hw, E1000_RCTL);
1781 
1782 	if (if_getflags(ifp) & IFF_PROMISC)
1783 		reg_rctl |= (E1000_RCTL_UPE | E1000_RCTL_MPE);
1784 	else if (mcnt >= MAX_NUM_MULTICAST_ADDRESSES ||
1785 	    if_getflags(ifp) & IFF_ALLMULTI) {
1786 		reg_rctl |= E1000_RCTL_MPE;
1787 		reg_rctl &= ~E1000_RCTL_UPE;
1788 	} else
1789 		reg_rctl &= ~(E1000_RCTL_UPE | E1000_RCTL_MPE);
1790 
1791 	E1000_WRITE_REG(&sc->hw, E1000_RCTL, reg_rctl);
1792 
1793 	if (sc->hw.mac.type == e1000_82542 &&
1794 	    sc->hw.revision_id == E1000_REVISION_2) {
1795 		reg_rctl = E1000_READ_REG(&sc->hw, E1000_RCTL);
1796 		reg_rctl &= ~E1000_RCTL_RST;
1797 		E1000_WRITE_REG(&sc->hw, E1000_RCTL, reg_rctl);
1798 		msec_delay(5);
1799 		if (sc->hw.bus.pci_cmd_word & CMD_MEM_WRT_INVALIDATE)
1800 			e1000_pci_set_mwi(&sc->hw);
1801 	}
1802 }
1803 
1804 /*********************************************************************
1805  *  Timer routine
1806  *
1807  *  This routine schedules em_if_update_admin_status() to check for
1808  *  link status and to gather statistics as well as to perform some
1809  *  controller-specific hardware patting.
1810  *
1811  **********************************************************************/
1812 static void
1813 em_if_timer(if_ctx_t ctx, uint16_t qid)
1814 {
1815 
1816 	if (qid != 0)
1817 		return;
1818 
1819 	iflib_admin_intr_deferred(ctx);
1820 }
1821 
1822 static void
1823 em_if_update_admin_status(if_ctx_t ctx)
1824 {
1825 	struct e1000_softc *sc = iflib_get_softc(ctx);
1826 	struct e1000_hw *hw = &sc->hw;
1827 	device_t dev = iflib_get_dev(ctx);
1828 	u32 link_check, thstat, ctrl;
1829 	bool automasked = false;
1830 
1831 	link_check = thstat = ctrl = 0;
1832 	/* Get the cached link value or read phy for real */
1833 	switch (hw->phy.media_type) {
1834 	case e1000_media_type_copper:
1835 		if (hw->mac.get_link_status) {
1836 			if (hw->mac.type == e1000_pch_spt)
1837 				msec_delay(50);
1838 			/* Do the work to read phy */
1839 			e1000_check_for_link(hw);
1840 			link_check = !hw->mac.get_link_status;
1841 			if (link_check) /* ESB2 fix */
1842 				e1000_cfg_on_link_up(hw);
1843 		} else {
1844 			link_check = true;
1845 		}
1846 		break;
1847 	case e1000_media_type_fiber:
1848 		e1000_check_for_link(hw);
1849 		link_check = (E1000_READ_REG(hw, E1000_STATUS) &
1850 			    E1000_STATUS_LU);
1851 		break;
1852 	case e1000_media_type_internal_serdes:
1853 		e1000_check_for_link(hw);
1854 		link_check = hw->mac.serdes_has_link;
1855 		break;
1856 	/* VF device is type_unknown */
1857 	case e1000_media_type_unknown:
1858 		e1000_check_for_link(hw);
1859 		link_check = !hw->mac.get_link_status;
1860 		/* FALLTHROUGH */
1861 	default:
1862 		break;
1863 	}
1864 
1865 	/* Check for thermal downshift or shutdown */
1866 	if (hw->mac.type == e1000_i350) {
1867 		thstat = E1000_READ_REG(hw, E1000_THSTAT);
1868 		ctrl = E1000_READ_REG(hw, E1000_CTRL_EXT);
1869 	}
1870 
1871 	/* Now check for a transition */
1872 	if (link_check && (sc->link_active == 0)) {
1873 		e1000_get_speed_and_duplex(hw, &sc->link_speed,
1874 		    &sc->link_duplex);
1875 		/* Check if we must disable SPEED_MODE bit on PCI-E */
1876 		if ((sc->link_speed != SPEED_1000) &&
1877 		    ((hw->mac.type == e1000_82571) ||
1878 		    (hw->mac.type == e1000_82572))) {
1879 			int tarc0;
1880 			tarc0 = E1000_READ_REG(hw, E1000_TARC(0));
1881 			tarc0 &= ~TARC_SPEED_MODE_BIT;
1882 			E1000_WRITE_REG(hw, E1000_TARC(0), tarc0);
1883 		}
1884 		if (bootverbose)
1885 			device_printf(dev, "Link is up %d Mbps %s\n",
1886 			    sc->link_speed,
1887 			    ((sc->link_duplex == FULL_DUPLEX) ?
1888 			    "Full Duplex" : "Half Duplex"));
1889 		sc->link_active = 1;
1890 		sc->smartspeed = 0;
1891 		if ((ctrl & E1000_CTRL_EXT_LINK_MODE_MASK) ==
1892 		    E1000_CTRL_EXT_LINK_MODE_GMII &&
1893 		    (thstat & E1000_THSTAT_LINK_THROTTLE))
1894 			device_printf(dev, "Link: thermal downshift\n");
1895 		/* Delay Link Up for Phy update */
1896 		if (((hw->mac.type == e1000_i210) ||
1897 		    (hw->mac.type == e1000_i211)) &&
1898 		    (hw->phy.id == I210_I_PHY_ID))
1899 			msec_delay(I210_LINK_DELAY);
1900 		/* Reset if the media type changed. */
1901 		if (hw->dev_spec._82575.media_changed &&
1902 		    hw->mac.type >= igb_mac_min) {
1903 			hw->dev_spec._82575.media_changed = false;
1904 			sc->flags |= IGB_MEDIA_RESET;
1905 			em_reset(ctx);
1906 		}
1907 		/* Only do TSO on gigabit Ethernet for older chips due to errata */
1908 		if (hw->mac.type < igb_mac_min)
1909 			automasked = em_automask_tso(ctx);
1910 
1911 		/* Automasking resets the interface, so don't mark it up yet */
1912 		if (!automasked)
1913 			iflib_link_state_change(ctx, LINK_STATE_UP,
1914 			    IF_Mbps(sc->link_speed));
1915 	} else if (!link_check && (sc->link_active == 1)) {
1916 		sc->link_speed = 0;
1917 		sc->link_duplex = 0;
1918 		sc->link_active = 0;
1919 		iflib_link_state_change(ctx, LINK_STATE_DOWN, 0);
1920 	}
1921 	em_update_stats_counters(sc);
1922 
1923 	/* Reset LAA into RAR[0] on 82571 */
1924 	if (hw->mac.type == e1000_82571 && e1000_get_laa_state_82571(hw))
1925 		e1000_rar_set(hw, hw->mac.addr, 0);
1926 
1927 	if (hw->mac.type < em_mac_min)
1928 		lem_smartspeed(sc);
1929 }
1930 
1931 static void
1932 em_if_watchdog_reset(if_ctx_t ctx)
1933 {
1934 	struct e1000_softc *sc = iflib_get_softc(ctx);
1935 
1936 	/*
1937 	 * Just count the event; iflib(4) will already trigger a
1938 	 * sufficient reset of the controller.
1939 	 */
1940 	sc->watchdog_events++;
1941 }
1942 
1943 /*********************************************************************
1944  *
1945  *  This routine disables all traffic on the adapter by issuing a
1946  *  global reset on the MAC.
1947  *
1948  **********************************************************************/
1949 static void
1950 em_if_stop(if_ctx_t ctx)
1951 {
1952 	struct e1000_softc *sc = iflib_get_softc(ctx);
1953 
1954 	INIT_DEBUGOUT("em_if_stop: begin");
1955 
1956 	/* I219 needs special flushing to avoid hangs */
1957 	if (sc->hw.mac.type >= e1000_pch_spt && sc->hw.mac.type < igb_mac_min)
1958 		em_flush_desc_rings(sc);
1959 
1960 	e1000_reset_hw(&sc->hw);
1961 	if (sc->hw.mac.type >= e1000_82544)
1962 		E1000_WRITE_REG(&sc->hw, E1000_WUFC, 0);
1963 
1964 	e1000_led_off(&sc->hw);
1965 	e1000_cleanup_led(&sc->hw);
1966 }
1967 
1968 /*********************************************************************
1969  *
1970  *  Determine hardware revision.
1971  *
1972  **********************************************************************/
1973 static void
1974 em_identify_hardware(if_ctx_t ctx)
1975 {
1976 	device_t dev = iflib_get_dev(ctx);
1977 	struct e1000_softc *sc = iflib_get_softc(ctx);
1978 
1979 	/* Make sure our PCI config space has the necessary stuff set */
1980 	sc->hw.bus.pci_cmd_word = pci_read_config(dev, PCIR_COMMAND, 2);
1981 
1982 	/* Save off the information about this board */
1983 	sc->hw.vendor_id = pci_get_vendor(dev);
1984 	sc->hw.device_id = pci_get_device(dev);
1985 	sc->hw.revision_id = pci_read_config(dev, PCIR_REVID, 1);
1986 	sc->hw.subsystem_vendor_id =
1987 	    pci_read_config(dev, PCIR_SUBVEND_0, 2);
1988 	sc->hw.subsystem_device_id =
1989 	    pci_read_config(dev, PCIR_SUBDEV_0, 2);
1990 
1991 	/* Do Shared Code Init and Setup */
1992 	if (e1000_set_mac_type(&sc->hw)) {
1993 		device_printf(dev, "Setup init failure\n");
1994 		return;
1995 	}
1996 
1997 	/* Are we a VF device? */
1998 	if ((sc->hw.mac.type == e1000_vfadapt) ||
1999 	    (sc->hw.mac.type == e1000_vfadapt_i350))
2000 		sc->vf_ifp = 1;
2001 	else
2002 		sc->vf_ifp = 0;
2003 }
2004 
2005 static int
2006 em_allocate_pci_resources(if_ctx_t ctx)
2007 {
2008 	struct e1000_softc *sc = iflib_get_softc(ctx);
2009 	device_t dev = iflib_get_dev(ctx);
2010 	int rid, val;
2011 
2012 	rid = PCIR_BAR(0);
2013 	sc->memory = bus_alloc_resource_any(dev, SYS_RES_MEMORY,
2014 	    &rid, RF_ACTIVE);
2015 	if (sc->memory == NULL) {
2016 		device_printf(dev, "Unable to allocate bus resource: memory\n");
2017 		return (ENXIO);
2018 	}
2019 	sc->osdep.mem_bus_space_tag = rman_get_bustag(sc->memory);
2020 	sc->osdep.mem_bus_space_handle =
2021 	    rman_get_bushandle(sc->memory);
2022 	sc->hw.hw_addr = (u8 *)&sc->osdep.mem_bus_space_handle;
2023 
2024 	/* Only older adapters use IO mapping */
2025 	if (sc->hw.mac.type < em_mac_min && sc->hw.mac.type > e1000_82543) {
2026 		/* Figure our where our IO BAR is ? */
2027 		for (rid = PCIR_BAR(0); rid < PCIR_CIS;) {
2028 			val = pci_read_config(dev, rid, 4);
2029 			if (EM_BAR_TYPE(val) == EM_BAR_TYPE_IO) {
2030 				break;
2031 			}
2032 			rid += 4;
2033 			/* check for 64bit BAR */
2034 			if (EM_BAR_MEM_TYPE(val) == EM_BAR_MEM_TYPE_64BIT)
2035 				rid += 4;
2036 		}
2037 		if (rid >= PCIR_CIS) {
2038 			device_printf(dev, "Unable to locate IO BAR\n");
2039 			return (ENXIO);
2040 		}
2041 		sc->ioport = bus_alloc_resource_any(dev, SYS_RES_IOPORT,
2042 		    &rid, RF_ACTIVE);
2043 		if (sc->ioport == NULL) {
2044 			device_printf(dev, "Unable to allocate bus resource: "
2045 			    "ioport\n");
2046 			return (ENXIO);
2047 		}
2048 		sc->hw.io_base = 0;
2049 		sc->osdep.io_bus_space_tag =
2050 		    rman_get_bustag(sc->ioport);
2051 		sc->osdep.io_bus_space_handle =
2052 		    rman_get_bushandle(sc->ioport);
2053 	}
2054 
2055 	sc->hw.back = &sc->osdep;
2056 
2057 	return (0);
2058 }
2059 
2060 /*********************************************************************
2061  *
2062  *  Set up the MSI-X Interrupt handlers
2063  *
2064  **********************************************************************/
2065 static int
2066 em_if_msix_intr_assign(if_ctx_t ctx, int msix)
2067 {
2068 	struct e1000_softc *sc = iflib_get_softc(ctx);
2069 	struct em_rx_queue *rx_que = sc->rx_queues;
2070 	struct em_tx_queue *tx_que = sc->tx_queues;
2071 	int error, rid, i, vector = 0, rx_vectors;
2072 	char buf[16];
2073 
2074 	/* First set up ring resources */
2075 	for (i = 0; i < sc->rx_num_queues; i++, rx_que++, vector++) {
2076 		rid = vector + 1;
2077 		snprintf(buf, sizeof(buf), "rxq%d", i);
2078 		error = iflib_irq_alloc_generic(ctx, &rx_que->que_irq, rid, IFLIB_INTR_RXTX, em_msix_que, rx_que, rx_que->me, buf);
2079 		if (error) {
2080 			device_printf(iflib_get_dev(ctx), "Failed to allocate que int %d err: %d", i, error);
2081 			sc->rx_num_queues = i + 1;
2082 			goto fail;
2083 		}
2084 
2085 		rx_que->msix =  vector;
2086 
2087 		/*
2088 		 * Set the bit to enable interrupt
2089 		 * in E1000_IMS -- bits 20 and 21
2090 		 * are for RX0 and RX1, note this has
2091 		 * NOTHING to do with the MSI-X vector
2092 		 */
2093 		if (sc->hw.mac.type == e1000_82574) {
2094 			rx_que->eims = 1 << (20 + i);
2095 			sc->ims |= rx_que->eims;
2096 			sc->ivars |= (8 | rx_que->msix) << (i * 4);
2097 		} else if (sc->hw.mac.type == e1000_82575)
2098 			rx_que->eims = E1000_EICR_TX_QUEUE0 << vector;
2099 		else
2100 			rx_que->eims = 1 << vector;
2101 	}
2102 	rx_vectors = vector;
2103 
2104 	vector = 0;
2105 	for (i = 0; i < sc->tx_num_queues; i++, tx_que++, vector++) {
2106 		snprintf(buf, sizeof(buf), "txq%d", i);
2107 		tx_que = &sc->tx_queues[i];
2108 		iflib_softirq_alloc_generic(ctx,
2109 		    &sc->rx_queues[i % sc->rx_num_queues].que_irq,
2110 		    IFLIB_INTR_TX, tx_que, tx_que->me, buf);
2111 
2112 		tx_que->msix = (vector % sc->rx_num_queues);
2113 
2114 		/*
2115 		 * Set the bit to enable interrupt
2116 		 * in E1000_IMS -- bits 22 and 23
2117 		 * are for TX0 and TX1, note this has
2118 		 * NOTHING to do with the MSI-X vector
2119 		 */
2120 		if (sc->hw.mac.type == e1000_82574) {
2121 			tx_que->eims = 1 << (22 + i);
2122 			sc->ims |= tx_que->eims;
2123 			sc->ivars |= (8 | tx_que->msix) << (8 + (i * 4));
2124 		} else if (sc->hw.mac.type == e1000_82575) {
2125 			tx_que->eims = E1000_EICR_TX_QUEUE0 << i;
2126 		} else {
2127 			tx_que->eims = 1 << i;
2128 		}
2129 	}
2130 
2131 	/* Link interrupt */
2132 	rid = rx_vectors + 1;
2133 	error = iflib_irq_alloc_generic(ctx, &sc->irq, rid, IFLIB_INTR_ADMIN, em_msix_link, sc, 0, "aq");
2134 
2135 	if (error) {
2136 		device_printf(iflib_get_dev(ctx), "Failed to register admin handler");
2137 		goto fail;
2138 	}
2139 	sc->linkvec = rx_vectors;
2140 	if (sc->hw.mac.type < igb_mac_min) {
2141 		sc->ivars |=  (8 | rx_vectors) << 16;
2142 		sc->ivars |= 0x80000000;
2143 		/* Enable the "Other" interrupt type for link status change */
2144 		sc->ims |= E1000_IMS_OTHER;
2145 	}
2146 
2147 	return (0);
2148 fail:
2149 	iflib_irq_free(ctx, &sc->irq);
2150 	rx_que = sc->rx_queues;
2151 	for (int i = 0; i < sc->rx_num_queues; i++, rx_que++)
2152 		iflib_irq_free(ctx, &rx_que->que_irq);
2153 	return (error);
2154 }
2155 
2156 static void
2157 igb_configure_queues(struct e1000_softc *sc)
2158 {
2159 	struct e1000_hw *hw = &sc->hw;
2160 	struct em_rx_queue *rx_que;
2161 	struct em_tx_queue *tx_que;
2162 	u32 tmp, ivar = 0, newitr = 0;
2163 
2164 	/* First turn on RSS capability */
2165 	if (hw->mac.type != e1000_82575)
2166 		E1000_WRITE_REG(hw, E1000_GPIE,
2167 		    E1000_GPIE_MSIX_MODE | E1000_GPIE_EIAME |
2168 		    E1000_GPIE_PBA | E1000_GPIE_NSICR);
2169 
2170 	/* Turn on MSI-X */
2171 	switch (hw->mac.type) {
2172 	case e1000_82580:
2173 	case e1000_i350:
2174 	case e1000_i354:
2175 	case e1000_i210:
2176 	case e1000_i211:
2177 	case e1000_vfadapt:
2178 	case e1000_vfadapt_i350:
2179 		/* RX entries */
2180 		for (int i = 0; i < sc->rx_num_queues; i++) {
2181 			u32 index = i >> 1;
2182 			ivar = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index);
2183 			rx_que = &sc->rx_queues[i];
2184 			if (i & 1) {
2185 				ivar &= 0xFF00FFFF;
2186 				ivar |= (rx_que->msix | E1000_IVAR_VALID) << 16;
2187 			} else {
2188 				ivar &= 0xFFFFFF00;
2189 				ivar |= rx_que->msix | E1000_IVAR_VALID;
2190 			}
2191 			E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, ivar);
2192 		}
2193 		/* TX entries */
2194 		for (int i = 0; i < sc->tx_num_queues; i++) {
2195 			u32 index = i >> 1;
2196 			ivar = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index);
2197 			tx_que = &sc->tx_queues[i];
2198 			if (i & 1) {
2199 				ivar &= 0x00FFFFFF;
2200 				ivar |= (tx_que->msix | E1000_IVAR_VALID) << 24;
2201 			} else {
2202 				ivar &= 0xFFFF00FF;
2203 				ivar |= (tx_que->msix | E1000_IVAR_VALID) << 8;
2204 			}
2205 			E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, ivar);
2206 			sc->que_mask |= tx_que->eims;
2207 		}
2208 
2209 		/* And for the link interrupt */
2210 		ivar = (sc->linkvec | E1000_IVAR_VALID) << 8;
2211 		sc->link_mask = 1 << sc->linkvec;
2212 		E1000_WRITE_REG(hw, E1000_IVAR_MISC, ivar);
2213 		break;
2214 	case e1000_82576:
2215 		/* RX entries */
2216 		for (int i = 0; i < sc->rx_num_queues; i++) {
2217 			u32 index = i & 0x7; /* Each IVAR has two entries */
2218 			ivar = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index);
2219 			rx_que = &sc->rx_queues[i];
2220 			if (i < 8) {
2221 				ivar &= 0xFFFFFF00;
2222 				ivar |= rx_que->msix | E1000_IVAR_VALID;
2223 			} else {
2224 				ivar &= 0xFF00FFFF;
2225 				ivar |= (rx_que->msix | E1000_IVAR_VALID) << 16;
2226 			}
2227 			E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, ivar);
2228 			sc->que_mask |= rx_que->eims;
2229 		}
2230 		/* TX entries */
2231 		for (int i = 0; i < sc->tx_num_queues; i++) {
2232 			u32 index = i & 0x7; /* Each IVAR has two entries */
2233 			ivar = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index);
2234 			tx_que = &sc->tx_queues[i];
2235 			if (i < 8) {
2236 				ivar &= 0xFFFF00FF;
2237 				ivar |= (tx_que->msix | E1000_IVAR_VALID) << 8;
2238 			} else {
2239 				ivar &= 0x00FFFFFF;
2240 				ivar |= (tx_que->msix | E1000_IVAR_VALID) << 24;
2241 			}
2242 			E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, ivar);
2243 			sc->que_mask |= tx_que->eims;
2244 		}
2245 
2246 		/* And for the link interrupt */
2247 		ivar = (sc->linkvec | E1000_IVAR_VALID) << 8;
2248 		sc->link_mask = 1 << sc->linkvec;
2249 		E1000_WRITE_REG(hw, E1000_IVAR_MISC, ivar);
2250 		break;
2251 
2252 	case e1000_82575:
2253 		/* enable MSI-X support*/
2254 		tmp = E1000_READ_REG(hw, E1000_CTRL_EXT);
2255 		tmp |= E1000_CTRL_EXT_PBA_CLR;
2256 		/* Auto-Mask interrupts upon ICR read. */
2257 		tmp |= E1000_CTRL_EXT_EIAME;
2258 		tmp |= E1000_CTRL_EXT_IRCA;
2259 		E1000_WRITE_REG(hw, E1000_CTRL_EXT, tmp);
2260 
2261 		/* Queues */
2262 		for (int i = 0; i < sc->rx_num_queues; i++) {
2263 			rx_que = &sc->rx_queues[i];
2264 			tmp = E1000_EICR_RX_QUEUE0 << i;
2265 			tmp |= E1000_EICR_TX_QUEUE0 << i;
2266 			rx_que->eims = tmp;
2267 			E1000_WRITE_REG_ARRAY(hw, E1000_MSIXBM(0),
2268 			    i, rx_que->eims);
2269 			sc->que_mask |= rx_que->eims;
2270 		}
2271 
2272 		/* Link */
2273 		E1000_WRITE_REG(hw, E1000_MSIXBM(sc->linkvec),
2274 		    E1000_EIMS_OTHER);
2275 		sc->link_mask |= E1000_EIMS_OTHER;
2276 	default:
2277 		break;
2278 	}
2279 
2280 	/* Set the starting interrupt rate */
2281 	if (em_max_interrupt_rate > 0)
2282 		newitr = (4000000 / em_max_interrupt_rate) & 0x7FFC;
2283 
2284 	if (hw->mac.type == e1000_82575)
2285 		newitr |= newitr << 16;
2286 	else
2287 		newitr |= E1000_EITR_CNT_IGNR;
2288 
2289 	for (int i = 0; i < sc->rx_num_queues; i++) {
2290 		rx_que = &sc->rx_queues[i];
2291 		E1000_WRITE_REG(hw, E1000_EITR(rx_que->msix), newitr);
2292 	}
2293 
2294 	return;
2295 }
2296 
2297 static void
2298 em_free_pci_resources(if_ctx_t ctx)
2299 {
2300 	struct e1000_softc *sc = iflib_get_softc(ctx);
2301 	struct em_rx_queue *que = sc->rx_queues;
2302 	device_t dev = iflib_get_dev(ctx);
2303 
2304 	/* Release all MSI-X queue resources */
2305 	if (sc->intr_type == IFLIB_INTR_MSIX)
2306 		iflib_irq_free(ctx, &sc->irq);
2307 
2308 	if (que != NULL) {
2309 		for (int i = 0; i < sc->rx_num_queues; i++, que++) {
2310 			iflib_irq_free(ctx, &que->que_irq);
2311 		}
2312 	}
2313 
2314 	if (sc->memory != NULL) {
2315 		bus_release_resource(dev, SYS_RES_MEMORY,
2316 		    rman_get_rid(sc->memory), sc->memory);
2317 		sc->memory = NULL;
2318 	}
2319 
2320 	if (sc->flash != NULL) {
2321 		bus_release_resource(dev, SYS_RES_MEMORY,
2322 		    rman_get_rid(sc->flash), sc->flash);
2323 		sc->flash = NULL;
2324 	}
2325 
2326 	if (sc->ioport != NULL) {
2327 		bus_release_resource(dev, SYS_RES_IOPORT,
2328 		    rman_get_rid(sc->ioport), sc->ioport);
2329 		sc->ioport = NULL;
2330 	}
2331 }
2332 
2333 /* Set up MSI or MSI-X */
2334 static int
2335 em_setup_msix(if_ctx_t ctx)
2336 {
2337 	struct e1000_softc *sc = iflib_get_softc(ctx);
2338 
2339 	if (sc->hw.mac.type == e1000_82574) {
2340 		em_enable_vectors_82574(ctx);
2341 	}
2342 	return (0);
2343 }
2344 
2345 /*********************************************************************
2346  *
2347  *  Workaround for SmartSpeed on 82541 and 82547 controllers
2348  *
2349  **********************************************************************/
2350 static void
2351 lem_smartspeed(struct e1000_softc *sc)
2352 {
2353 	u16 phy_tmp;
2354 
2355 	if (sc->link_active || (sc->hw.phy.type != e1000_phy_igp) ||
2356 	    sc->hw.mac.autoneg == 0 ||
2357 	    (sc->hw.phy.autoneg_advertised & ADVERTISE_1000_FULL) == 0)
2358 		return;
2359 
2360 	if (sc->smartspeed == 0) {
2361 		/* If Master/Slave config fault is asserted twice,
2362 		 * we assume back-to-back */
2363 		e1000_read_phy_reg(&sc->hw, PHY_1000T_STATUS, &phy_tmp);
2364 		if (!(phy_tmp & SR_1000T_MS_CONFIG_FAULT))
2365 			return;
2366 		e1000_read_phy_reg(&sc->hw, PHY_1000T_STATUS, &phy_tmp);
2367 		if (phy_tmp & SR_1000T_MS_CONFIG_FAULT) {
2368 			e1000_read_phy_reg(&sc->hw,
2369 			    PHY_1000T_CTRL, &phy_tmp);
2370 			if(phy_tmp & CR_1000T_MS_ENABLE) {
2371 				phy_tmp &= ~CR_1000T_MS_ENABLE;
2372 				e1000_write_phy_reg(&sc->hw,
2373 				    PHY_1000T_CTRL, phy_tmp);
2374 				sc->smartspeed++;
2375 				if(sc->hw.mac.autoneg &&
2376 				   !e1000_copper_link_autoneg(&sc->hw) &&
2377 				   !e1000_read_phy_reg(&sc->hw,
2378 				    PHY_CONTROL, &phy_tmp)) {
2379 					phy_tmp |= (MII_CR_AUTO_NEG_EN |
2380 						    MII_CR_RESTART_AUTO_NEG);
2381 					e1000_write_phy_reg(&sc->hw,
2382 					    PHY_CONTROL, phy_tmp);
2383 				}
2384 			}
2385 		}
2386 		return;
2387 	} else if(sc->smartspeed == EM_SMARTSPEED_DOWNSHIFT) {
2388 		/* If still no link, perhaps using 2/3 pair cable */
2389 		e1000_read_phy_reg(&sc->hw, PHY_1000T_CTRL, &phy_tmp);
2390 		phy_tmp |= CR_1000T_MS_ENABLE;
2391 		e1000_write_phy_reg(&sc->hw, PHY_1000T_CTRL, phy_tmp);
2392 		if(sc->hw.mac.autoneg &&
2393 		   !e1000_copper_link_autoneg(&sc->hw) &&
2394 		   !e1000_read_phy_reg(&sc->hw, PHY_CONTROL, &phy_tmp)) {
2395 			phy_tmp |= (MII_CR_AUTO_NEG_EN |
2396 				    MII_CR_RESTART_AUTO_NEG);
2397 			e1000_write_phy_reg(&sc->hw, PHY_CONTROL, phy_tmp);
2398 		}
2399 	}
2400 	/* Restart process after EM_SMARTSPEED_MAX iterations */
2401 	if(sc->smartspeed++ == EM_SMARTSPEED_MAX)
2402 		sc->smartspeed = 0;
2403 }
2404 
2405 /*********************************************************************
2406  *
2407  *  Initialize the DMA Coalescing feature
2408  *
2409  **********************************************************************/
2410 static void
2411 igb_init_dmac(struct e1000_softc *sc, u32 pba)
2412 {
2413 	device_t	dev = sc->dev;
2414 	struct e1000_hw *hw = &sc->hw;
2415 	u32 		dmac, reg = ~E1000_DMACR_DMAC_EN;
2416 	u16		hwm;
2417 	u16		max_frame_size;
2418 
2419 	if (hw->mac.type == e1000_i211)
2420 		return;
2421 
2422 	max_frame_size = sc->shared->isc_max_frame_size;
2423 	if (hw->mac.type > e1000_82580) {
2424 
2425 		if (sc->dmac == 0) { /* Disabling it */
2426 			E1000_WRITE_REG(hw, E1000_DMACR, reg);
2427 			return;
2428 		} else
2429 			device_printf(dev, "DMA Coalescing enabled\n");
2430 
2431 		/* Set starting threshold */
2432 		E1000_WRITE_REG(hw, E1000_DMCTXTH, 0);
2433 
2434 		hwm = 64 * pba - max_frame_size / 16;
2435 		if (hwm < 64 * (pba - 6))
2436 			hwm = 64 * (pba - 6);
2437 		reg = E1000_READ_REG(hw, E1000_FCRTC);
2438 		reg &= ~E1000_FCRTC_RTH_COAL_MASK;
2439 		reg |= ((hwm << E1000_FCRTC_RTH_COAL_SHIFT)
2440 		    & E1000_FCRTC_RTH_COAL_MASK);
2441 		E1000_WRITE_REG(hw, E1000_FCRTC, reg);
2442 
2443 
2444 		dmac = pba - max_frame_size / 512;
2445 		if (dmac < pba - 10)
2446 			dmac = pba - 10;
2447 		reg = E1000_READ_REG(hw, E1000_DMACR);
2448 		reg &= ~E1000_DMACR_DMACTHR_MASK;
2449 		reg |= ((dmac << E1000_DMACR_DMACTHR_SHIFT)
2450 		    & E1000_DMACR_DMACTHR_MASK);
2451 
2452 		/* transition to L0x or L1 if available..*/
2453 		reg |= (E1000_DMACR_DMAC_EN | E1000_DMACR_DMAC_LX_MASK);
2454 
2455 		/* Check if status is 2.5Gb backplane connection
2456 		* before configuration of watchdog timer, which is
2457 		* in msec values in 12.8usec intervals
2458 		* watchdog timer= msec values in 32usec intervals
2459 		* for non 2.5Gb connection
2460 		*/
2461 		if (hw->mac.type == e1000_i354) {
2462 			int status = E1000_READ_REG(hw, E1000_STATUS);
2463 			if ((status & E1000_STATUS_2P5_SKU) &&
2464 			    (!(status & E1000_STATUS_2P5_SKU_OVER)))
2465 				reg |= ((sc->dmac * 5) >> 6);
2466 			else
2467 				reg |= (sc->dmac >> 5);
2468 		} else {
2469 			reg |= (sc->dmac >> 5);
2470 		}
2471 
2472 		E1000_WRITE_REG(hw, E1000_DMACR, reg);
2473 
2474 		E1000_WRITE_REG(hw, E1000_DMCRTRH, 0);
2475 
2476 		/* Set the interval before transition */
2477 		reg = E1000_READ_REG(hw, E1000_DMCTLX);
2478 		if (hw->mac.type == e1000_i350)
2479 			reg |= IGB_DMCTLX_DCFLUSH_DIS;
2480 		/*
2481 		** in 2.5Gb connection, TTLX unit is 0.4 usec
2482 		** which is 0x4*2 = 0xA. But delay is still 4 usec
2483 		*/
2484 		if (hw->mac.type == e1000_i354) {
2485 			int status = E1000_READ_REG(hw, E1000_STATUS);
2486 			if ((status & E1000_STATUS_2P5_SKU) &&
2487 			    (!(status & E1000_STATUS_2P5_SKU_OVER)))
2488 				reg |= 0xA;
2489 			else
2490 				reg |= 0x4;
2491 		} else {
2492 			reg |= 0x4;
2493 		}
2494 
2495 		E1000_WRITE_REG(hw, E1000_DMCTLX, reg);
2496 
2497 		/* free space in tx packet buffer to wake from DMA coal */
2498 		E1000_WRITE_REG(hw, E1000_DMCTXTH, (IGB_TXPBSIZE -
2499 		    (2 * max_frame_size)) >> 6);
2500 
2501 		/* make low power state decision controlled by DMA coal */
2502 		reg = E1000_READ_REG(hw, E1000_PCIEMISC);
2503 		reg &= ~E1000_PCIEMISC_LX_DECISION;
2504 		E1000_WRITE_REG(hw, E1000_PCIEMISC, reg);
2505 
2506 	} else if (hw->mac.type == e1000_82580) {
2507 		u32 reg = E1000_READ_REG(hw, E1000_PCIEMISC);
2508 		E1000_WRITE_REG(hw, E1000_PCIEMISC,
2509 		    reg & ~E1000_PCIEMISC_LX_DECISION);
2510 		E1000_WRITE_REG(hw, E1000_DMACR, 0);
2511 	}
2512 }
2513 /*********************************************************************
2514  * The 3 following flush routines are used as a workaround in the
2515  * I219 client parts and only for them.
2516  *
2517  * em_flush_tx_ring - remove all descriptors from the tx_ring
2518  *
2519  * We want to clear all pending descriptors from the TX ring.
2520  * zeroing happens when the HW reads the regs. We assign the ring itself as
2521  * the data of the next descriptor. We don't care about the data we are about
2522  * to reset the HW.
2523  **********************************************************************/
2524 static void
2525 em_flush_tx_ring(struct e1000_softc *sc)
2526 {
2527 	struct e1000_hw		*hw = &sc->hw;
2528 	struct tx_ring		*txr = &sc->tx_queues->txr;
2529 	struct e1000_tx_desc	*txd;
2530 	u32			tctl, txd_lower = E1000_TXD_CMD_IFCS;
2531 	u16			size = 512;
2532 
2533 	tctl = E1000_READ_REG(hw, E1000_TCTL);
2534 	E1000_WRITE_REG(hw, E1000_TCTL, tctl | E1000_TCTL_EN);
2535 
2536 	txd = &txr->tx_base[txr->tx_cidx_processed];
2537 
2538 	/* Just use the ring as a dummy buffer addr */
2539 	txd->buffer_addr = txr->tx_paddr;
2540 	txd->lower.data = htole32(txd_lower | size);
2541 	txd->upper.data = 0;
2542 
2543 	/* flush descriptors to memory before notifying the HW */
2544 	wmb();
2545 
2546 	E1000_WRITE_REG(hw, E1000_TDT(0), txr->tx_cidx_processed);
2547 	mb();
2548 	usec_delay(250);
2549 }
2550 
2551 /*********************************************************************
2552  * em_flush_rx_ring - remove all descriptors from the rx_ring
2553  *
2554  * Mark all descriptors in the RX ring as consumed and disable the rx ring
2555  **********************************************************************/
2556 static void
2557 em_flush_rx_ring(struct e1000_softc *sc)
2558 {
2559 	struct e1000_hw	*hw = &sc->hw;
2560 	u32		rctl, rxdctl;
2561 
2562 	rctl = E1000_READ_REG(hw, E1000_RCTL);
2563 	E1000_WRITE_REG(hw, E1000_RCTL, rctl & ~E1000_RCTL_EN);
2564 	E1000_WRITE_FLUSH(hw);
2565 	usec_delay(150);
2566 
2567 	rxdctl = E1000_READ_REG(hw, E1000_RXDCTL(0));
2568 	/* zero the lower 14 bits (prefetch and host thresholds) */
2569 	rxdctl &= 0xffffc000;
2570 	/*
2571 	 * update thresholds: prefetch threshold to 31, host threshold to 1
2572 	 * and make sure the granularity is "descriptors" and not "cache lines"
2573 	 */
2574 	rxdctl |= (0x1F | (1 << 8) | E1000_RXDCTL_THRESH_UNIT_DESC);
2575 	E1000_WRITE_REG(hw, E1000_RXDCTL(0), rxdctl);
2576 
2577 	/* momentarily enable the RX ring for the changes to take effect */
2578 	E1000_WRITE_REG(hw, E1000_RCTL, rctl | E1000_RCTL_EN);
2579 	E1000_WRITE_FLUSH(hw);
2580 	usec_delay(150);
2581 	E1000_WRITE_REG(hw, E1000_RCTL, rctl & ~E1000_RCTL_EN);
2582 }
2583 
2584 /*********************************************************************
2585  * em_flush_desc_rings - remove all descriptors from the descriptor rings
2586  *
2587  * In I219, the descriptor rings must be emptied before resetting the HW
2588  * or before changing the device state to D3 during runtime (runtime PM).
2589  *
2590  * Failure to do this will cause the HW to enter a unit hang state which can
2591  * only be released by PCI reset on the device
2592  *
2593  **********************************************************************/
2594 static void
2595 em_flush_desc_rings(struct e1000_softc *sc)
2596 {
2597 	struct e1000_hw	*hw = &sc->hw;
2598 	device_t dev = sc->dev;
2599 	u16		hang_state;
2600 	u32		fext_nvm11, tdlen;
2601 
2602 	/* First, disable MULR fix in FEXTNVM11 */
2603 	fext_nvm11 = E1000_READ_REG(hw, E1000_FEXTNVM11);
2604 	fext_nvm11 |= E1000_FEXTNVM11_DISABLE_MULR_FIX;
2605 	E1000_WRITE_REG(hw, E1000_FEXTNVM11, fext_nvm11);
2606 
2607 	/* do nothing if we're not in faulty state, or if the queue is empty */
2608 	tdlen = E1000_READ_REG(hw, E1000_TDLEN(0));
2609 	hang_state = pci_read_config(dev, PCICFG_DESC_RING_STATUS, 2);
2610 	if (!(hang_state & FLUSH_DESC_REQUIRED) || !tdlen)
2611 		return;
2612 	em_flush_tx_ring(sc);
2613 
2614 	/* recheck, maybe the fault is caused by the rx ring */
2615 	hang_state = pci_read_config(dev, PCICFG_DESC_RING_STATUS, 2);
2616 	if (hang_state & FLUSH_DESC_REQUIRED)
2617 		em_flush_rx_ring(sc);
2618 }
2619 
2620 
2621 /*********************************************************************
2622  *
2623  *  Initialize the hardware to a configuration as specified by the
2624  *  sc structure.
2625  *
2626  **********************************************************************/
2627 static void
2628 em_reset(if_ctx_t ctx)
2629 {
2630 	device_t dev = iflib_get_dev(ctx);
2631 	struct e1000_softc *sc = iflib_get_softc(ctx);
2632 	if_t ifp = iflib_get_ifp(ctx);
2633 	struct e1000_hw *hw = &sc->hw;
2634 	u32 rx_buffer_size;
2635 	u32 pba;
2636 
2637 	INIT_DEBUGOUT("em_reset: begin");
2638 	/* Let the firmware know the OS is in control */
2639 	em_get_hw_control(sc);
2640 
2641 	/* Set up smart power down as default off on newer adapters. */
2642 	if (!em_smart_pwr_down && (hw->mac.type == e1000_82571 ||
2643 	    hw->mac.type == e1000_82572)) {
2644 		u16 phy_tmp = 0;
2645 
2646 		/* Speed up time to link by disabling smart power down. */
2647 		e1000_read_phy_reg(hw, IGP02E1000_PHY_POWER_MGMT, &phy_tmp);
2648 		phy_tmp &= ~IGP02E1000_PM_SPD;
2649 		e1000_write_phy_reg(hw, IGP02E1000_PHY_POWER_MGMT, phy_tmp);
2650 	}
2651 
2652 	/*
2653 	 * Packet Buffer Allocation (PBA)
2654 	 * Writing PBA sets the receive portion of the buffer
2655 	 * the remainder is used for the transmit buffer.
2656 	 */
2657 	switch (hw->mac.type) {
2658 	/* 82547: Total Packet Buffer is 40K */
2659 	case e1000_82547:
2660 	case e1000_82547_rev_2:
2661 		if (hw->mac.max_frame_size > 8192)
2662 			pba = E1000_PBA_22K; /* 22K for Rx, 18K for Tx */
2663 		else
2664 			pba = E1000_PBA_30K; /* 30K for Rx, 10K for Tx */
2665 		break;
2666 	/* 82571/82572/80003es2lan: Total Packet Buffer is 48K */
2667 	case e1000_82571:
2668 	case e1000_82572:
2669 	case e1000_80003es2lan:
2670 			pba = E1000_PBA_32K; /* 32K for Rx, 16K for Tx */
2671 		break;
2672 	/* 82573: Total Packet Buffer is 32K */
2673 	case e1000_82573:
2674 			pba = E1000_PBA_12K; /* 12K for Rx, 20K for Tx */
2675 		break;
2676 	case e1000_82574:
2677 	case e1000_82583:
2678 			pba = E1000_PBA_20K; /* 20K for Rx, 20K for Tx */
2679 		break;
2680 	case e1000_ich8lan:
2681 		pba = E1000_PBA_8K;
2682 		break;
2683 	case e1000_ich9lan:
2684 	case e1000_ich10lan:
2685 		/* Boost Receive side for jumbo frames */
2686 		if (hw->mac.max_frame_size > 4096)
2687 			pba = E1000_PBA_14K;
2688 		else
2689 			pba = E1000_PBA_10K;
2690 		break;
2691 	case e1000_pchlan:
2692 	case e1000_pch2lan:
2693 	case e1000_pch_lpt:
2694 	case e1000_pch_spt:
2695 	case e1000_pch_cnp:
2696 	case e1000_pch_tgp:
2697 	case e1000_pch_adp:
2698 	case e1000_pch_mtp:
2699 	case e1000_pch_ptp:
2700 		pba = E1000_PBA_26K;
2701 		break;
2702 	case e1000_82575:
2703 		pba = E1000_PBA_32K;
2704 		break;
2705 	case e1000_82576:
2706 	case e1000_vfadapt:
2707 		pba = E1000_READ_REG(hw, E1000_RXPBS);
2708 		pba &= E1000_RXPBS_SIZE_MASK_82576;
2709 		break;
2710 	case e1000_82580:
2711 	case e1000_i350:
2712 	case e1000_i354:
2713 	case e1000_vfadapt_i350:
2714 		pba = E1000_READ_REG(hw, E1000_RXPBS);
2715 		pba = e1000_rxpbs_adjust_82580(pba);
2716 		break;
2717 	case e1000_i210:
2718 	case e1000_i211:
2719 		pba = E1000_PBA_34K;
2720 		break;
2721 	default:
2722 		/* Remaining devices assumed to have a Packet Buffer of 64K. */
2723 		if (hw->mac.max_frame_size > 8192)
2724 			pba = E1000_PBA_40K; /* 40K for Rx, 24K for Tx */
2725 		else
2726 			pba = E1000_PBA_48K; /* 48K for Rx, 16K for Tx */
2727 	}
2728 
2729 	/* Special needs in case of Jumbo frames */
2730 	if ((hw->mac.type == e1000_82575) && (if_getmtu(ifp) > ETHERMTU)) {
2731 		u32 tx_space, min_tx, min_rx;
2732 		pba = E1000_READ_REG(hw, E1000_PBA);
2733 		tx_space = pba >> 16;
2734 		pba &= 0xffff;
2735 		min_tx = (hw->mac.max_frame_size +
2736 		    sizeof(struct e1000_tx_desc) - ETHERNET_FCS_SIZE) * 2;
2737 		min_tx = roundup2(min_tx, 1024);
2738 		min_tx >>= 10;
2739 		min_rx = hw->mac.max_frame_size;
2740 		min_rx = roundup2(min_rx, 1024);
2741 		min_rx >>= 10;
2742 		if (tx_space < min_tx &&
2743 		    ((min_tx - tx_space) < pba)) {
2744 			pba = pba - (min_tx - tx_space);
2745 			/*
2746 			 * if short on rx space, rx wins
2747 			 * and must trump tx adjustment
2748 			 */
2749 			if (pba < min_rx)
2750 				pba = min_rx;
2751 		}
2752 		E1000_WRITE_REG(hw, E1000_PBA, pba);
2753 	}
2754 
2755 	if (hw->mac.type < igb_mac_min)
2756 		E1000_WRITE_REG(hw, E1000_PBA, pba);
2757 
2758 	INIT_DEBUGOUT1("em_reset: pba=%dK",pba);
2759 
2760 	/*
2761 	 * These parameters control the automatic generation (Tx) and
2762 	 * response (Rx) to Ethernet PAUSE frames.
2763 	 * - High water mark should allow for at least two frames to be
2764 	 *   received after sending an XOFF.
2765 	 * - Low water mark works best when it is very near the high water mark.
2766 	 *   This allows the receiver to restart by sending XON when it has
2767 	 *   drained a bit. Here we use an arbitrary value of 1500 which will
2768 	 *   restart after one full frame is pulled from the buffer. There
2769 	 *   could be several smaller frames in the buffer and if so they will
2770 	 *   not trigger the XON until their total number reduces the buffer
2771 	 *   by 1500.
2772 	 * - The pause time is fairly large at 1000 x 512ns = 512 usec.
2773 	 */
2774 	rx_buffer_size = (pba & 0xffff) << 10;
2775 	hw->fc.high_water = rx_buffer_size -
2776 	    roundup2(hw->mac.max_frame_size, 1024);
2777 	hw->fc.low_water = hw->fc.high_water - 1500;
2778 
2779 	if (sc->fc) /* locally set flow control value? */
2780 		hw->fc.requested_mode = sc->fc;
2781 	else
2782 		hw->fc.requested_mode = e1000_fc_full;
2783 
2784 	if (hw->mac.type == e1000_80003es2lan)
2785 		hw->fc.pause_time = 0xFFFF;
2786 	else
2787 		hw->fc.pause_time = EM_FC_PAUSE_TIME;
2788 
2789 	hw->fc.send_xon = true;
2790 
2791 	/* Device specific overrides/settings */
2792 	switch (hw->mac.type) {
2793 	case e1000_pchlan:
2794 		/* Workaround: no TX flow ctrl for PCH */
2795 		hw->fc.requested_mode = e1000_fc_rx_pause;
2796 		hw->fc.pause_time = 0xFFFF; /* override */
2797 		if (if_getmtu(ifp) > ETHERMTU) {
2798 			hw->fc.high_water = 0x3500;
2799 			hw->fc.low_water = 0x1500;
2800 		} else {
2801 			hw->fc.high_water = 0x5000;
2802 			hw->fc.low_water = 0x3000;
2803 		}
2804 		hw->fc.refresh_time = 0x1000;
2805 		break;
2806 	case e1000_pch2lan:
2807 	case e1000_pch_lpt:
2808 	case e1000_pch_spt:
2809 	case e1000_pch_cnp:
2810 	case e1000_pch_tgp:
2811 	case e1000_pch_adp:
2812 	case e1000_pch_mtp:
2813 	case e1000_pch_ptp:
2814 		hw->fc.high_water = 0x5C20;
2815 		hw->fc.low_water = 0x5048;
2816 		hw->fc.pause_time = 0x0650;
2817 		hw->fc.refresh_time = 0x0400;
2818 		/* Jumbos need adjusted PBA */
2819 		if (if_getmtu(ifp) > ETHERMTU)
2820 			E1000_WRITE_REG(hw, E1000_PBA, 12);
2821 		else
2822 			E1000_WRITE_REG(hw, E1000_PBA, 26);
2823 		break;
2824 	case e1000_82575:
2825 	case e1000_82576:
2826 		/* 8-byte granularity */
2827 		hw->fc.low_water = hw->fc.high_water - 8;
2828 		break;
2829 	case e1000_82580:
2830 	case e1000_i350:
2831 	case e1000_i354:
2832 	case e1000_i210:
2833 	case e1000_i211:
2834 	case e1000_vfadapt:
2835 	case e1000_vfadapt_i350:
2836 		/* 16-byte granularity */
2837 		hw->fc.low_water = hw->fc.high_water - 16;
2838 		break;
2839 	case e1000_ich9lan:
2840 	case e1000_ich10lan:
2841 		if (if_getmtu(ifp) > ETHERMTU) {
2842 			hw->fc.high_water = 0x2800;
2843 			hw->fc.low_water = hw->fc.high_water - 8;
2844 			break;
2845 		}
2846 		/* FALLTHROUGH */
2847 	default:
2848 		if (hw->mac.type == e1000_80003es2lan)
2849 			hw->fc.pause_time = 0xFFFF;
2850 		break;
2851 	}
2852 
2853 	/* I219 needs some special flushing to avoid hangs */
2854 	if (sc->hw.mac.type >= e1000_pch_spt && sc->hw.mac.type < igb_mac_min)
2855 		em_flush_desc_rings(sc);
2856 
2857 	/* Issue a global reset */
2858 	e1000_reset_hw(hw);
2859 	if (hw->mac.type >= igb_mac_min) {
2860 		E1000_WRITE_REG(hw, E1000_WUC, 0);
2861 	} else {
2862 		E1000_WRITE_REG(hw, E1000_WUFC, 0);
2863 		em_disable_aspm(sc);
2864 	}
2865 	if (sc->flags & IGB_MEDIA_RESET) {
2866 		e1000_setup_init_funcs(hw, true);
2867 		e1000_get_bus_info(hw);
2868 		sc->flags &= ~IGB_MEDIA_RESET;
2869 	}
2870 	/* and a re-init */
2871 	if (e1000_init_hw(hw) < 0) {
2872 		device_printf(dev, "Hardware Initialization Failed\n");
2873 		return;
2874 	}
2875 	if (hw->mac.type >= igb_mac_min)
2876 		igb_init_dmac(sc, pba);
2877 
2878 	E1000_WRITE_REG(hw, E1000_VET, ETHERTYPE_VLAN);
2879 	e1000_get_phy_info(hw);
2880 	e1000_check_for_link(hw);
2881 }
2882 
2883 /*
2884  * Initialise the RSS mapping for NICs that support multiple transmit/
2885  * receive rings.
2886  */
2887 
2888 #define RSSKEYLEN 10
2889 static void
2890 em_initialize_rss_mapping(struct e1000_softc *sc)
2891 {
2892 	uint8_t  rss_key[4 * RSSKEYLEN];
2893 	uint32_t reta = 0;
2894 	struct e1000_hw	*hw = &sc->hw;
2895 	int i;
2896 
2897 	/*
2898 	 * Configure RSS key
2899 	 */
2900 	arc4rand(rss_key, sizeof(rss_key), 0);
2901 	for (i = 0; i < RSSKEYLEN; ++i) {
2902 		uint32_t rssrk = 0;
2903 
2904 		rssrk = EM_RSSRK_VAL(rss_key, i);
2905 		E1000_WRITE_REG(hw,E1000_RSSRK(i), rssrk);
2906 	}
2907 
2908 	/*
2909 	 * Configure RSS redirect table in following fashion:
2910 	 * (hash & ring_cnt_mask) == rdr_table[(hash & rdr_table_mask)]
2911 	 */
2912 	for (i = 0; i < sizeof(reta); ++i) {
2913 		uint32_t q;
2914 
2915 		q = (i % sc->rx_num_queues) << 7;
2916 		reta |= q << (8 * i);
2917 	}
2918 
2919 	for (i = 0; i < 32; ++i)
2920 		E1000_WRITE_REG(hw, E1000_RETA(i), reta);
2921 
2922 	E1000_WRITE_REG(hw, E1000_MRQC, E1000_MRQC_RSS_ENABLE_2Q |
2923 			E1000_MRQC_RSS_FIELD_IPV4_TCP |
2924 			E1000_MRQC_RSS_FIELD_IPV4 |
2925 			E1000_MRQC_RSS_FIELD_IPV6_TCP_EX |
2926 			E1000_MRQC_RSS_FIELD_IPV6_EX |
2927 			E1000_MRQC_RSS_FIELD_IPV6);
2928 }
2929 
2930 static void
2931 igb_initialize_rss_mapping(struct e1000_softc *sc)
2932 {
2933 	struct e1000_hw *hw = &sc->hw;
2934 	int i;
2935 	int queue_id;
2936 	u32 reta;
2937 	u32 rss_key[10], mrqc, shift = 0;
2938 
2939 	/* XXX? */
2940 	if (hw->mac.type == e1000_82575)
2941 		shift = 6;
2942 
2943 	/*
2944 	 * The redirection table controls which destination
2945 	 * queue each bucket redirects traffic to.
2946 	 * Each DWORD represents four queues, with the LSB
2947 	 * being the first queue in the DWORD.
2948 	 *
2949 	 * This just allocates buckets to queues using round-robin
2950 	 * allocation.
2951 	 *
2952 	 * NOTE: It Just Happens to line up with the default
2953 	 * RSS allocation method.
2954 	 */
2955 
2956 	/* Warning FM follows */
2957 	reta = 0;
2958 	for (i = 0; i < 128; i++) {
2959 #ifdef RSS
2960 		queue_id = rss_get_indirection_to_bucket(i);
2961 		/*
2962 		 * If we have more queues than buckets, we'll
2963 		 * end up mapping buckets to a subset of the
2964 		 * queues.
2965 		 *
2966 		 * If we have more buckets than queues, we'll
2967 		 * end up instead assigning multiple buckets
2968 		 * to queues.
2969 		 *
2970 		 * Both are suboptimal, but we need to handle
2971 		 * the case so we don't go out of bounds
2972 		 * indexing arrays and such.
2973 		 */
2974 		queue_id = queue_id % sc->rx_num_queues;
2975 #else
2976 		queue_id = (i % sc->rx_num_queues);
2977 #endif
2978 		/* Adjust if required */
2979 		queue_id = queue_id << shift;
2980 
2981 		/*
2982 		 * The low 8 bits are for hash value (n+0);
2983 		 * The next 8 bits are for hash value (n+1), etc.
2984 		 */
2985 		reta = reta >> 8;
2986 		reta = reta | ( ((uint32_t) queue_id) << 24);
2987 		if ((i & 3) == 3) {
2988 			E1000_WRITE_REG(hw, E1000_RETA(i >> 2), reta);
2989 			reta = 0;
2990 		}
2991 	}
2992 
2993 	/* Now fill in hash table */
2994 
2995 	/*
2996 	 * MRQC: Multiple Receive Queues Command
2997 	 * Set queuing to RSS control, number depends on the device.
2998 	 */
2999 	mrqc = E1000_MRQC_ENABLE_RSS_MQ;
3000 
3001 #ifdef RSS
3002 	/* XXX ew typecasting */
3003 	rss_getkey((uint8_t *) &rss_key);
3004 #else
3005 	arc4rand(&rss_key, sizeof(rss_key), 0);
3006 #endif
3007 	for (i = 0; i < 10; i++)
3008 		E1000_WRITE_REG_ARRAY(hw, E1000_RSSRK(0), i, rss_key[i]);
3009 
3010 	/*
3011 	 * Configure the RSS fields to hash upon.
3012 	 */
3013 	mrqc |= (E1000_MRQC_RSS_FIELD_IPV4 |
3014 	    E1000_MRQC_RSS_FIELD_IPV4_TCP);
3015 	mrqc |= (E1000_MRQC_RSS_FIELD_IPV6 |
3016 	    E1000_MRQC_RSS_FIELD_IPV6_TCP);
3017 	mrqc |=( E1000_MRQC_RSS_FIELD_IPV4_UDP |
3018 	    E1000_MRQC_RSS_FIELD_IPV6_UDP);
3019 	mrqc |=( E1000_MRQC_RSS_FIELD_IPV6_UDP_EX |
3020 	    E1000_MRQC_RSS_FIELD_IPV6_TCP_EX);
3021 
3022 	E1000_WRITE_REG(hw, E1000_MRQC, mrqc);
3023 }
3024 
3025 /*********************************************************************
3026  *
3027  *  Setup networking device structure and register interface media.
3028  *
3029  **********************************************************************/
3030 static int
3031 em_setup_interface(if_ctx_t ctx)
3032 {
3033 	if_t ifp = iflib_get_ifp(ctx);
3034 	struct e1000_softc *sc = iflib_get_softc(ctx);
3035 	if_softc_ctx_t scctx = sc->shared;
3036 
3037 	INIT_DEBUGOUT("em_setup_interface: begin");
3038 
3039 	/* Single Queue */
3040 	if (sc->tx_num_queues == 1) {
3041 		if_setsendqlen(ifp, scctx->isc_ntxd[0] - 1);
3042 		if_setsendqready(ifp);
3043 	}
3044 
3045 	/*
3046 	 * Specify the media types supported by this adapter and register
3047 	 * callbacks to update media and link information
3048 	 */
3049 	if (sc->hw.phy.media_type == e1000_media_type_fiber ||
3050 	    sc->hw.phy.media_type == e1000_media_type_internal_serdes) {
3051 		u_char fiber_type = IFM_1000_SX;	/* default type */
3052 
3053 		if (sc->hw.mac.type == e1000_82545)
3054 			fiber_type = IFM_1000_LX;
3055 		ifmedia_add(sc->media, IFM_ETHER | fiber_type | IFM_FDX, 0, NULL);
3056 		ifmedia_add(sc->media, IFM_ETHER | fiber_type, 0, NULL);
3057 	} else {
3058 		ifmedia_add(sc->media, IFM_ETHER | IFM_10_T, 0, NULL);
3059 		ifmedia_add(sc->media, IFM_ETHER | IFM_10_T | IFM_FDX, 0, NULL);
3060 		ifmedia_add(sc->media, IFM_ETHER | IFM_100_TX, 0, NULL);
3061 		ifmedia_add(sc->media, IFM_ETHER | IFM_100_TX | IFM_FDX, 0, NULL);
3062 		if (sc->hw.phy.type != e1000_phy_ife) {
3063 			ifmedia_add(sc->media, IFM_ETHER | IFM_1000_T | IFM_FDX, 0, NULL);
3064 			ifmedia_add(sc->media, IFM_ETHER | IFM_1000_T, 0, NULL);
3065 		}
3066 	}
3067 	ifmedia_add(sc->media, IFM_ETHER | IFM_AUTO, 0, NULL);
3068 	ifmedia_set(sc->media, IFM_ETHER | IFM_AUTO);
3069 	return (0);
3070 }
3071 
3072 static int
3073 em_if_tx_queues_alloc(if_ctx_t ctx, caddr_t *vaddrs, uint64_t *paddrs, int ntxqs, int ntxqsets)
3074 {
3075 	struct e1000_softc *sc = iflib_get_softc(ctx);
3076 	if_softc_ctx_t scctx = sc->shared;
3077 	int error = E1000_SUCCESS;
3078 	struct em_tx_queue *que;
3079 	int i, j;
3080 
3081 	MPASS(sc->tx_num_queues > 0);
3082 	MPASS(sc->tx_num_queues == ntxqsets);
3083 
3084 	/* First allocate the top level queue structs */
3085 	if (!(sc->tx_queues =
3086 	    (struct em_tx_queue *) malloc(sizeof(struct em_tx_queue) *
3087 	    sc->tx_num_queues, M_DEVBUF, M_NOWAIT | M_ZERO))) {
3088 		device_printf(iflib_get_dev(ctx), "Unable to allocate queue memory\n");
3089 		return(ENOMEM);
3090 	}
3091 
3092 	for (i = 0, que = sc->tx_queues; i < sc->tx_num_queues; i++, que++) {
3093 		/* Set up some basics */
3094 
3095 		struct tx_ring *txr = &que->txr;
3096 		txr->sc = que->sc = sc;
3097 		que->me = txr->me =  i;
3098 
3099 		/* Allocate report status array */
3100 		if (!(txr->tx_rsq = (qidx_t *) malloc(sizeof(qidx_t) * scctx->isc_ntxd[0], M_DEVBUF, M_NOWAIT | M_ZERO))) {
3101 			device_printf(iflib_get_dev(ctx), "failed to allocate rs_idxs memory\n");
3102 			error = ENOMEM;
3103 			goto fail;
3104 		}
3105 		for (j = 0; j < scctx->isc_ntxd[0]; j++)
3106 			txr->tx_rsq[j] = QIDX_INVALID;
3107 		/* get the virtual and physical address of the hardware queues */
3108 		txr->tx_base = (struct e1000_tx_desc *)vaddrs[i*ntxqs];
3109 		txr->tx_paddr = paddrs[i*ntxqs];
3110 	}
3111 
3112 	if (bootverbose)
3113 		device_printf(iflib_get_dev(ctx),
3114 		    "allocated for %d tx_queues\n", sc->tx_num_queues);
3115 	return (0);
3116 fail:
3117 	em_if_queues_free(ctx);
3118 	return (error);
3119 }
3120 
3121 static int
3122 em_if_rx_queues_alloc(if_ctx_t ctx, caddr_t *vaddrs, uint64_t *paddrs, int nrxqs, int nrxqsets)
3123 {
3124 	struct e1000_softc *sc = iflib_get_softc(ctx);
3125 	int error = E1000_SUCCESS;
3126 	struct em_rx_queue *que;
3127 	int i;
3128 
3129 	MPASS(sc->rx_num_queues > 0);
3130 	MPASS(sc->rx_num_queues == nrxqsets);
3131 
3132 	/* First allocate the top level queue structs */
3133 	if (!(sc->rx_queues =
3134 	    (struct em_rx_queue *) malloc(sizeof(struct em_rx_queue) *
3135 	    sc->rx_num_queues, M_DEVBUF, M_NOWAIT | M_ZERO))) {
3136 		device_printf(iflib_get_dev(ctx), "Unable to allocate queue memory\n");
3137 		error = ENOMEM;
3138 		goto fail;
3139 	}
3140 
3141 	for (i = 0, que = sc->rx_queues; i < nrxqsets; i++, que++) {
3142 		/* Set up some basics */
3143 		struct rx_ring *rxr = &que->rxr;
3144 		rxr->sc = que->sc = sc;
3145 		rxr->que = que;
3146 		que->me = rxr->me =  i;
3147 
3148 		/* get the virtual and physical address of the hardware queues */
3149 		rxr->rx_base = (union e1000_rx_desc_extended *)vaddrs[i*nrxqs];
3150 		rxr->rx_paddr = paddrs[i*nrxqs];
3151 	}
3152 
3153 	if (bootverbose)
3154 		device_printf(iflib_get_dev(ctx),
3155 		    "allocated for %d rx_queues\n", sc->rx_num_queues);
3156 
3157 	return (0);
3158 fail:
3159 	em_if_queues_free(ctx);
3160 	return (error);
3161 }
3162 
3163 static void
3164 em_if_queues_free(if_ctx_t ctx)
3165 {
3166 	struct e1000_softc *sc = iflib_get_softc(ctx);
3167 	struct em_tx_queue *tx_que = sc->tx_queues;
3168 	struct em_rx_queue *rx_que = sc->rx_queues;
3169 
3170 	if (tx_que != NULL) {
3171 		for (int i = 0; i < sc->tx_num_queues; i++, tx_que++) {
3172 			struct tx_ring *txr = &tx_que->txr;
3173 			if (txr->tx_rsq == NULL)
3174 				break;
3175 
3176 			free(txr->tx_rsq, M_DEVBUF);
3177 			txr->tx_rsq = NULL;
3178 		}
3179 		free(sc->tx_queues, M_DEVBUF);
3180 		sc->tx_queues = NULL;
3181 	}
3182 
3183 	if (rx_que != NULL) {
3184 		free(sc->rx_queues, M_DEVBUF);
3185 		sc->rx_queues = NULL;
3186 	}
3187 }
3188 
3189 /*********************************************************************
3190  *
3191  *  Enable transmit unit.
3192  *
3193  **********************************************************************/
3194 static void
3195 em_initialize_transmit_unit(if_ctx_t ctx)
3196 {
3197 	struct e1000_softc *sc = iflib_get_softc(ctx);
3198 	if_softc_ctx_t scctx = sc->shared;
3199 	struct em_tx_queue *que;
3200 	struct tx_ring	*txr;
3201 	struct e1000_hw	*hw = &sc->hw;
3202 	u32 tctl, txdctl = 0, tarc, tipg = 0;
3203 
3204 	INIT_DEBUGOUT("em_initialize_transmit_unit: begin");
3205 
3206 	for (int i = 0; i < sc->tx_num_queues; i++, txr++) {
3207 		u64 bus_addr;
3208 		caddr_t offp, endp;
3209 
3210 		que = &sc->tx_queues[i];
3211 		txr = &que->txr;
3212 		bus_addr = txr->tx_paddr;
3213 
3214 		/* Clear checksum offload context. */
3215 		offp = (caddr_t)&txr->csum_flags;
3216 		endp = (caddr_t)(txr + 1);
3217 		bzero(offp, endp - offp);
3218 
3219 		/* Base and Len of TX Ring */
3220 		E1000_WRITE_REG(hw, E1000_TDLEN(i),
3221 		    scctx->isc_ntxd[0] * sizeof(struct e1000_tx_desc));
3222 		E1000_WRITE_REG(hw, E1000_TDBAH(i),
3223 		    (u32)(bus_addr >> 32));
3224 		E1000_WRITE_REG(hw, E1000_TDBAL(i),
3225 		    (u32)bus_addr);
3226 		/* Init the HEAD/TAIL indices */
3227 		E1000_WRITE_REG(hw, E1000_TDT(i), 0);
3228 		E1000_WRITE_REG(hw, E1000_TDH(i), 0);
3229 
3230 		HW_DEBUGOUT2("Base = %x, Length = %x\n",
3231 		    E1000_READ_REG(hw, E1000_TDBAL(i)),
3232 		    E1000_READ_REG(hw, E1000_TDLEN(i)));
3233 
3234 		txdctl = 0; /* clear txdctl */
3235 		txdctl |= 0x1f; /* PTHRESH */
3236 		txdctl |= 1 << 8; /* HTHRESH */
3237 		txdctl |= 1 << 16;/* WTHRESH */
3238 		txdctl |= 1 << 22; /* Reserved bit 22 must always be 1 */
3239 		txdctl |= E1000_TXDCTL_GRAN;
3240 		txdctl |= 1 << 25; /* LWTHRESH */
3241 
3242 		E1000_WRITE_REG(hw, E1000_TXDCTL(i), txdctl);
3243 	}
3244 
3245 	/* Set the default values for the Tx Inter Packet Gap timer */
3246 	switch (hw->mac.type) {
3247 	case e1000_80003es2lan:
3248 		tipg = DEFAULT_82543_TIPG_IPGR1;
3249 		tipg |= DEFAULT_80003ES2LAN_TIPG_IPGR2 <<
3250 		    E1000_TIPG_IPGR2_SHIFT;
3251 		break;
3252 	case e1000_82542:
3253 		tipg = DEFAULT_82542_TIPG_IPGT;
3254 		tipg |= DEFAULT_82542_TIPG_IPGR1 << E1000_TIPG_IPGR1_SHIFT;
3255 		tipg |= DEFAULT_82542_TIPG_IPGR2 << E1000_TIPG_IPGR2_SHIFT;
3256 		break;
3257 	default:
3258 		if (hw->phy.media_type == e1000_media_type_fiber ||
3259 		    hw->phy.media_type == e1000_media_type_internal_serdes)
3260 			tipg = DEFAULT_82543_TIPG_IPGT_FIBER;
3261 		else
3262 			tipg = DEFAULT_82543_TIPG_IPGT_COPPER;
3263 		tipg |= DEFAULT_82543_TIPG_IPGR1 << E1000_TIPG_IPGR1_SHIFT;
3264 		tipg |= DEFAULT_82543_TIPG_IPGR2 << E1000_TIPG_IPGR2_SHIFT;
3265 	}
3266 
3267 	E1000_WRITE_REG(hw, E1000_TIPG, tipg);
3268 	E1000_WRITE_REG(hw, E1000_TIDV, sc->tx_int_delay.value);
3269 
3270 	if(hw->mac.type >= e1000_82540)
3271 		E1000_WRITE_REG(hw, E1000_TADV,
3272 		    sc->tx_abs_int_delay.value);
3273 
3274 	if (hw->mac.type == e1000_82571 || hw->mac.type == e1000_82572) {
3275 		tarc = E1000_READ_REG(hw, E1000_TARC(0));
3276 		tarc |= TARC_SPEED_MODE_BIT;
3277 		E1000_WRITE_REG(hw, E1000_TARC(0), tarc);
3278 	} else if (hw->mac.type == e1000_80003es2lan) {
3279 		/* errata: program both queues to unweighted RR */
3280 		tarc = E1000_READ_REG(hw, E1000_TARC(0));
3281 		tarc |= 1;
3282 		E1000_WRITE_REG(hw, E1000_TARC(0), tarc);
3283 		tarc = E1000_READ_REG(hw, E1000_TARC(1));
3284 		tarc |= 1;
3285 		E1000_WRITE_REG(hw, E1000_TARC(1), tarc);
3286 	} else if (hw->mac.type == e1000_82574) {
3287 		tarc = E1000_READ_REG(hw, E1000_TARC(0));
3288 		tarc |= TARC_ERRATA_BIT;
3289 		if ( sc->tx_num_queues > 1) {
3290 			tarc |= (TARC_COMPENSATION_MODE | TARC_MQ_FIX);
3291 			E1000_WRITE_REG(hw, E1000_TARC(0), tarc);
3292 			E1000_WRITE_REG(hw, E1000_TARC(1), tarc);
3293 		} else
3294 			E1000_WRITE_REG(hw, E1000_TARC(0), tarc);
3295 	}
3296 
3297 	if (sc->tx_int_delay.value > 0)
3298 		sc->txd_cmd |= E1000_TXD_CMD_IDE;
3299 
3300 	/* Program the Transmit Control Register */
3301 	tctl = E1000_READ_REG(hw, E1000_TCTL);
3302 	tctl &= ~E1000_TCTL_CT;
3303 	tctl |= (E1000_TCTL_PSP | E1000_TCTL_RTLC | E1000_TCTL_EN |
3304 		   (E1000_COLLISION_THRESHOLD << E1000_CT_SHIFT));
3305 
3306 	if (hw->mac.type >= e1000_82571)
3307 		tctl |= E1000_TCTL_MULR;
3308 
3309 	/* This write will effectively turn on the transmit unit. */
3310 	E1000_WRITE_REG(hw, E1000_TCTL, tctl);
3311 
3312 	/* SPT and KBL errata workarounds */
3313 	if (hw->mac.type == e1000_pch_spt) {
3314 		u32 reg;
3315 		reg = E1000_READ_REG(hw, E1000_IOSFPC);
3316 		reg |= E1000_RCTL_RDMTS_HEX;
3317 		E1000_WRITE_REG(hw, E1000_IOSFPC, reg);
3318 		/* i218-i219 Specification Update 1.5.4.5 */
3319 		reg = E1000_READ_REG(hw, E1000_TARC(0));
3320 		reg &= ~E1000_TARC0_CB_MULTIQ_3_REQ;
3321 		reg |= E1000_TARC0_CB_MULTIQ_2_REQ;
3322 		E1000_WRITE_REG(hw, E1000_TARC(0), reg);
3323 	}
3324 }
3325 
3326 /*********************************************************************
3327  *
3328  *  Enable receive unit.
3329  *
3330  **********************************************************************/
3331 #define BSIZEPKT_ROUNDUP ((1<<E1000_SRRCTL_BSIZEPKT_SHIFT)-1)
3332 
3333 static void
3334 em_initialize_receive_unit(if_ctx_t ctx)
3335 {
3336 	struct e1000_softc *sc = iflib_get_softc(ctx);
3337 	if_softc_ctx_t scctx = sc->shared;
3338 	if_t ifp = iflib_get_ifp(ctx);
3339 	struct e1000_hw	*hw = &sc->hw;
3340 	struct em_rx_queue *que;
3341 	int i;
3342 	uint32_t rctl, rxcsum;
3343 
3344 	INIT_DEBUGOUT("em_initialize_receive_units: begin");
3345 
3346 	/*
3347 	 * Make sure receives are disabled while setting
3348 	 * up the descriptor ring
3349 	 */
3350 	rctl = E1000_READ_REG(hw, E1000_RCTL);
3351 	/* Do not disable if ever enabled on this hardware */
3352 	if ((hw->mac.type != e1000_82574) && (hw->mac.type != e1000_82583))
3353 		E1000_WRITE_REG(hw, E1000_RCTL, rctl & ~E1000_RCTL_EN);
3354 
3355 	/* Setup the Receive Control Register */
3356 	rctl &= ~(3 << E1000_RCTL_MO_SHIFT);
3357 	rctl |= E1000_RCTL_EN | E1000_RCTL_BAM |
3358 	    E1000_RCTL_LBM_NO | E1000_RCTL_RDMTS_HALF |
3359 	    (hw->mac.mc_filter_type << E1000_RCTL_MO_SHIFT);
3360 
3361 	/* Do not store bad packets */
3362 	rctl &= ~E1000_RCTL_SBP;
3363 
3364 	/* Enable Long Packet receive */
3365 	if (if_getmtu(ifp) > ETHERMTU)
3366 		rctl |= E1000_RCTL_LPE;
3367 	else
3368 		rctl &= ~E1000_RCTL_LPE;
3369 
3370 	/* Strip the CRC */
3371 	if (!em_disable_crc_stripping)
3372 		rctl |= E1000_RCTL_SECRC;
3373 
3374 	if (hw->mac.type >= e1000_82540) {
3375 		E1000_WRITE_REG(hw, E1000_RADV,
3376 		    sc->rx_abs_int_delay.value);
3377 
3378 		/*
3379 		 * Set the interrupt throttling rate. Value is calculated
3380 		 * as DEFAULT_ITR = 1/(MAX_INTS_PER_SEC * 256ns)
3381 		 */
3382 		E1000_WRITE_REG(hw, E1000_ITR, DEFAULT_ITR);
3383 	}
3384 	E1000_WRITE_REG(hw, E1000_RDTR, sc->rx_int_delay.value);
3385 
3386 	if (hw->mac.type >= em_mac_min) {
3387 		uint32_t rfctl;
3388 		/* Use extended rx descriptor formats */
3389 		rfctl = E1000_READ_REG(hw, E1000_RFCTL);
3390 		rfctl |= E1000_RFCTL_EXTEN;
3391 
3392 		/*
3393 		 * When using MSI-X interrupts we need to throttle
3394 		 * using the EITR register (82574 only)
3395 		 */
3396 		if (hw->mac.type == e1000_82574) {
3397 			for (int i = 0; i < 4; i++)
3398 				E1000_WRITE_REG(hw, E1000_EITR_82574(i),
3399 				    DEFAULT_ITR);
3400 			/* Disable accelerated acknowledge */
3401 			rfctl |= E1000_RFCTL_ACK_DIS;
3402 		}
3403 		E1000_WRITE_REG(hw, E1000_RFCTL, rfctl);
3404 	}
3405 
3406 	/* Set up L3 and L4 csum Rx descriptor offloads */
3407 	rxcsum = E1000_READ_REG(hw, E1000_RXCSUM);
3408 	if (if_getcapenable(ifp) & IFCAP_RXCSUM) {
3409 		rxcsum |= E1000_RXCSUM_TUOFL | E1000_RXCSUM_IPOFL;
3410 		if (hw->mac.type > e1000_82575)
3411 			rxcsum |= E1000_RXCSUM_CRCOFL;
3412 		else if (hw->mac.type < em_mac_min &&
3413 		    if_getcapenable(ifp) & IFCAP_HWCSUM_IPV6)
3414 			rxcsum |= E1000_RXCSUM_IPV6OFL;
3415 	} else {
3416 		rxcsum &= ~(E1000_RXCSUM_IPOFL | E1000_RXCSUM_TUOFL);
3417 		if (hw->mac.type > e1000_82575)
3418 			rxcsum &= ~E1000_RXCSUM_CRCOFL;
3419 		else if (hw->mac.type < em_mac_min)
3420 			rxcsum &= ~E1000_RXCSUM_IPV6OFL;
3421 	}
3422 
3423 	if (sc->rx_num_queues > 1) {
3424 		/* RSS hash needed in the Rx descriptor */
3425 		rxcsum |= E1000_RXCSUM_PCSD;
3426 
3427 		if (hw->mac.type >= igb_mac_min)
3428 			igb_initialize_rss_mapping(sc);
3429 		else
3430 			em_initialize_rss_mapping(sc);
3431 	}
3432 	E1000_WRITE_REG(hw, E1000_RXCSUM, rxcsum);
3433 
3434 	/*
3435 	 * XXX TEMPORARY WORKAROUND: on some systems with 82573
3436 	 * long latencies are observed, like Lenovo X60. This
3437 	 * change eliminates the problem, but since having positive
3438 	 * values in RDTR is a known source of problems on other
3439 	 * platforms another solution is being sought.
3440 	 */
3441 	if (hw->mac.type == e1000_82573)
3442 		E1000_WRITE_REG(hw, E1000_RDTR, 0x20);
3443 
3444 	for (i = 0, que = sc->rx_queues; i < sc->rx_num_queues; i++, que++) {
3445 		struct rx_ring *rxr = &que->rxr;
3446 		/* Setup the Base and Length of the Rx Descriptor Ring */
3447 		u64 bus_addr = rxr->rx_paddr;
3448 #if 0
3449 		u32 rdt = sc->rx_num_queues -1;  /* default */
3450 #endif
3451 
3452 		E1000_WRITE_REG(hw, E1000_RDLEN(i),
3453 		    scctx->isc_nrxd[0] * sizeof(union e1000_rx_desc_extended));
3454 		E1000_WRITE_REG(hw, E1000_RDBAH(i), (u32)(bus_addr >> 32));
3455 		E1000_WRITE_REG(hw, E1000_RDBAL(i), (u32)bus_addr);
3456 		/* Setup the Head and Tail Descriptor Pointers */
3457 		E1000_WRITE_REG(hw, E1000_RDH(i), 0);
3458 		E1000_WRITE_REG(hw, E1000_RDT(i), 0);
3459 	}
3460 
3461 	/*
3462 	 * Set PTHRESH for improved jumbo performance
3463 	 * According to 10.2.5.11 of Intel 82574 Datasheet,
3464 	 * RXDCTL(1) is written whenever RXDCTL(0) is written.
3465 	 * Only write to RXDCTL(1) if there is a need for different
3466 	 * settings.
3467 	 */
3468 	if ((hw->mac.type == e1000_ich9lan || hw->mac.type == e1000_pch2lan ||
3469 	    hw->mac.type == e1000_ich10lan) && if_getmtu(ifp) > ETHERMTU) {
3470 		u32 rxdctl = E1000_READ_REG(hw, E1000_RXDCTL(0));
3471 		E1000_WRITE_REG(hw, E1000_RXDCTL(0), rxdctl | 3);
3472 	} else if (hw->mac.type == e1000_82574) {
3473 		for (int i = 0; i < sc->rx_num_queues; i++) {
3474 			u32 rxdctl = E1000_READ_REG(hw, E1000_RXDCTL(i));
3475 			rxdctl |= 0x20; /* PTHRESH */
3476 			rxdctl |= 4 << 8; /* HTHRESH */
3477 			rxdctl |= 4 << 16;/* WTHRESH */
3478 			rxdctl |= 1 << 24; /* Switch to granularity */
3479 			E1000_WRITE_REG(hw, E1000_RXDCTL(i), rxdctl);
3480 		}
3481 	} else if (hw->mac.type >= igb_mac_min) {
3482 		u32 psize, srrctl = 0;
3483 
3484 		if (if_getmtu(ifp) > ETHERMTU) {
3485 			psize = scctx->isc_max_frame_size;
3486 			/* are we on a vlan? */
3487 			if (if_vlantrunkinuse(ifp))
3488 				psize += VLAN_TAG_SIZE;
3489 
3490 			if (sc->vf_ifp)
3491 				e1000_rlpml_set_vf(hw, psize);
3492 			else
3493 				E1000_WRITE_REG(hw, E1000_RLPML, psize);
3494 		}
3495 
3496 		/* Set maximum packet buffer len */
3497 		srrctl |= (sc->rx_mbuf_sz + BSIZEPKT_ROUNDUP) >>
3498 		    E1000_SRRCTL_BSIZEPKT_SHIFT;
3499 
3500 		/*
3501 		 * If TX flow control is disabled and there's >1 queue defined,
3502 		 * enable DROP.
3503 		 *
3504 		 * This drops frames rather than hanging the RX MAC for all queues.
3505 		 */
3506 		if ((sc->rx_num_queues > 1) &&
3507 		    (sc->fc == e1000_fc_none ||
3508 		     sc->fc == e1000_fc_rx_pause)) {
3509 			srrctl |= E1000_SRRCTL_DROP_EN;
3510 		}
3511 			/* Setup the Base and Length of the Rx Descriptor Rings */
3512 		for (i = 0, que = sc->rx_queues; i < sc->rx_num_queues; i++, que++) {
3513 			struct rx_ring *rxr = &que->rxr;
3514 			u64 bus_addr = rxr->rx_paddr;
3515 			u32 rxdctl;
3516 
3517 #ifdef notyet
3518 			/* Configure for header split? -- ignore for now */
3519 			rxr->hdr_split = igb_header_split;
3520 #else
3521 			srrctl |= E1000_SRRCTL_DESCTYPE_ADV_ONEBUF;
3522 #endif
3523 
3524 			E1000_WRITE_REG(hw, E1000_RDLEN(i),
3525 					scctx->isc_nrxd[0] * sizeof(struct e1000_rx_desc));
3526 			E1000_WRITE_REG(hw, E1000_RDBAH(i),
3527 					(uint32_t)(bus_addr >> 32));
3528 			E1000_WRITE_REG(hw, E1000_RDBAL(i),
3529 					(uint32_t)bus_addr);
3530 			E1000_WRITE_REG(hw, E1000_SRRCTL(i), srrctl);
3531 			/* Enable this Queue */
3532 			rxdctl = E1000_READ_REG(hw, E1000_RXDCTL(i));
3533 			rxdctl |= E1000_RXDCTL_QUEUE_ENABLE;
3534 			rxdctl &= 0xFFF00000;
3535 			rxdctl |= IGB_RX_PTHRESH;
3536 			rxdctl |= IGB_RX_HTHRESH << 8;
3537 			rxdctl |= IGB_RX_WTHRESH << 16;
3538 			E1000_WRITE_REG(hw, E1000_RXDCTL(i), rxdctl);
3539 		}
3540 	} else if (hw->mac.type >= e1000_pch2lan) {
3541 		if (if_getmtu(ifp) > ETHERMTU)
3542 			e1000_lv_jumbo_workaround_ich8lan(hw, true);
3543 		else
3544 			e1000_lv_jumbo_workaround_ich8lan(hw, false);
3545 	}
3546 
3547 	/* Make sure VLAN Filters are off */
3548 	rctl &= ~E1000_RCTL_VFE;
3549 
3550 	/* Set up packet buffer size, overridden by per queue srrctl on igb */
3551 	if (hw->mac.type < igb_mac_min) {
3552 		if (sc->rx_mbuf_sz > 2048 && sc->rx_mbuf_sz <= 4096)
3553 			rctl |= E1000_RCTL_SZ_4096 | E1000_RCTL_BSEX;
3554 		else if (sc->rx_mbuf_sz > 4096 && sc->rx_mbuf_sz <= 8192)
3555 			rctl |= E1000_RCTL_SZ_8192 | E1000_RCTL_BSEX;
3556 		else if (sc->rx_mbuf_sz > 8192)
3557 			rctl |= E1000_RCTL_SZ_16384 | E1000_RCTL_BSEX;
3558 		else {
3559 			rctl |= E1000_RCTL_SZ_2048;
3560 			rctl &= ~E1000_RCTL_BSEX;
3561 		}
3562 	} else
3563 		rctl |= E1000_RCTL_SZ_2048;
3564 
3565 	/*
3566 	 * rctl bits 11:10 are as follows
3567 	 * lem: reserved
3568 	 * em: DTYPE
3569 	 * igb: reserved
3570 	 * and should be 00 on all of the above
3571 	 */
3572 	rctl &= ~0x00000C00;
3573 
3574 	/* Write out the settings */
3575 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
3576 
3577 	return;
3578 }
3579 
3580 static void
3581 em_if_vlan_register(if_ctx_t ctx, u16 vtag)
3582 {
3583 	struct e1000_softc *sc = iflib_get_softc(ctx);
3584 	u32 index, bit;
3585 
3586 	index = (vtag >> 5) & 0x7F;
3587 	bit = vtag & 0x1F;
3588 	sc->shadow_vfta[index] |= (1 << bit);
3589 	++sc->num_vlans;
3590 	em_if_vlan_filter_write(sc);
3591 }
3592 
3593 static void
3594 em_if_vlan_unregister(if_ctx_t ctx, u16 vtag)
3595 {
3596 	struct e1000_softc *sc = iflib_get_softc(ctx);
3597 	u32 index, bit;
3598 
3599 	index = (vtag >> 5) & 0x7F;
3600 	bit = vtag & 0x1F;
3601 	sc->shadow_vfta[index] &= ~(1 << bit);
3602 	--sc->num_vlans;
3603 	em_if_vlan_filter_write(sc);
3604 }
3605 
3606 static bool
3607 em_if_vlan_filter_capable(if_ctx_t ctx)
3608 {
3609 	if_t ifp = iflib_get_ifp(ctx);
3610 
3611 	if ((if_getcapenable(ifp) & IFCAP_VLAN_HWFILTER) &&
3612 	    !em_disable_crc_stripping)
3613 		return (true);
3614 
3615 	return (false);
3616 }
3617 
3618 static bool
3619 em_if_vlan_filter_used(if_ctx_t ctx)
3620 {
3621 	struct e1000_softc *sc = iflib_get_softc(ctx);
3622 
3623 	if (!em_if_vlan_filter_capable(ctx))
3624 		return (false);
3625 
3626 	for (int i = 0; i < EM_VFTA_SIZE; i++)
3627 		if (sc->shadow_vfta[i] != 0)
3628 			return (true);
3629 
3630 	return (false);
3631 }
3632 
3633 static void
3634 em_if_vlan_filter_enable(struct e1000_softc *sc)
3635 {
3636 	struct e1000_hw *hw = &sc->hw;
3637 	u32 reg;
3638 
3639 	reg = E1000_READ_REG(hw, E1000_RCTL);
3640 	reg &= ~E1000_RCTL_CFIEN;
3641 	reg |= E1000_RCTL_VFE;
3642 	E1000_WRITE_REG(hw, E1000_RCTL, reg);
3643 }
3644 
3645 static void
3646 em_if_vlan_filter_disable(struct e1000_softc *sc)
3647 {
3648 	struct e1000_hw *hw = &sc->hw;
3649 	u32 reg;
3650 
3651 	reg = E1000_READ_REG(hw, E1000_RCTL);
3652 	reg &= ~(E1000_RCTL_VFE | E1000_RCTL_CFIEN);
3653 	E1000_WRITE_REG(hw, E1000_RCTL, reg);
3654 }
3655 
3656 static void
3657 em_if_vlan_filter_write(struct e1000_softc *sc)
3658 {
3659 	struct e1000_hw *hw = &sc->hw;
3660 
3661 	if (sc->vf_ifp)
3662 		return;
3663 
3664 	/* Disable interrupts for lem-class devices during the filter change */
3665 	if (hw->mac.type < em_mac_min)
3666 		em_if_intr_disable(sc->ctx);
3667 
3668 	for (int i = 0; i < EM_VFTA_SIZE; i++)
3669 		if (sc->shadow_vfta[i] != 0) {
3670 			/* XXXKB: incomplete VF support, we return early above */
3671 			if (sc->vf_ifp)
3672 				e1000_vfta_set_vf(hw, sc->shadow_vfta[i], true);
3673 			else
3674 				e1000_write_vfta(hw, i, sc->shadow_vfta[i]);
3675 		}
3676 
3677 	/* Re-enable interrupts for lem-class devices */
3678 	if (hw->mac.type < em_mac_min)
3679 		em_if_intr_enable(sc->ctx);
3680 }
3681 
3682 static void
3683 em_setup_vlan_hw_support(if_ctx_t ctx)
3684 {
3685 	struct e1000_softc *sc = iflib_get_softc(ctx);
3686 	struct e1000_hw *hw = &sc->hw;
3687 	if_t ifp = iflib_get_ifp(ctx);
3688 	u32 reg;
3689 
3690 	/* XXXKB: Return early if we are a VF until VF decap and filter management
3691 	 * is ready and tested.
3692 	 */
3693 	if (sc->vf_ifp)
3694 		return;
3695 
3696 	if (if_getcapenable(ifp) & IFCAP_VLAN_HWTAGGING &&
3697 	    !em_disable_crc_stripping) {
3698 		reg = E1000_READ_REG(hw, E1000_CTRL);
3699 		reg |= E1000_CTRL_VME;
3700 		E1000_WRITE_REG(hw, E1000_CTRL, reg);
3701 	} else {
3702 		reg = E1000_READ_REG(hw, E1000_CTRL);
3703 		reg &= ~E1000_CTRL_VME;
3704 		E1000_WRITE_REG(hw, E1000_CTRL, reg);
3705 	}
3706 
3707 	/* If we aren't doing HW filtering, we're done */
3708 	if (!em_if_vlan_filter_capable(ctx))  {
3709 		em_if_vlan_filter_disable(sc);
3710 		return;
3711 	}
3712 
3713 	/*
3714 	 * A soft reset zero's out the VFTA, so
3715 	 * we need to repopulate it now.
3716 	 * We also insert VLAN 0 in the filter list, so we pass VLAN 0 tagged
3717 	 * traffic through. This will write the entire table.
3718 	 */
3719 	em_if_vlan_register(ctx, 0);
3720 
3721 	/* Enable the Filter Table */
3722 	em_if_vlan_filter_enable(sc);
3723 }
3724 
3725 static void
3726 em_if_intr_enable(if_ctx_t ctx)
3727 {
3728 	struct e1000_softc *sc = iflib_get_softc(ctx);
3729 	struct e1000_hw *hw = &sc->hw;
3730 	u32 ims_mask = IMS_ENABLE_MASK;
3731 
3732 	if (sc->intr_type == IFLIB_INTR_MSIX) {
3733 		E1000_WRITE_REG(hw, EM_EIAC, sc->ims);
3734 		ims_mask |= sc->ims;
3735 	}
3736 	E1000_WRITE_REG(hw, E1000_IMS, ims_mask);
3737 	E1000_WRITE_FLUSH(hw);
3738 }
3739 
3740 static void
3741 em_if_intr_disable(if_ctx_t ctx)
3742 {
3743 	struct e1000_softc *sc = iflib_get_softc(ctx);
3744 	struct e1000_hw *hw = &sc->hw;
3745 
3746 	if (sc->intr_type == IFLIB_INTR_MSIX)
3747 		E1000_WRITE_REG(hw, EM_EIAC, 0);
3748 	E1000_WRITE_REG(hw, E1000_IMC, 0xffffffff);
3749 	E1000_WRITE_FLUSH(hw);
3750 }
3751 
3752 static void
3753 igb_if_intr_enable(if_ctx_t ctx)
3754 {
3755 	struct e1000_softc *sc = iflib_get_softc(ctx);
3756 	struct e1000_hw *hw = &sc->hw;
3757 	u32 mask;
3758 
3759 	if (__predict_true(sc->intr_type == IFLIB_INTR_MSIX)) {
3760 		mask = (sc->que_mask | sc->link_mask);
3761 		E1000_WRITE_REG(hw, E1000_EIAC, mask);
3762 		E1000_WRITE_REG(hw, E1000_EIAM, mask);
3763 		E1000_WRITE_REG(hw, E1000_EIMS, mask);
3764 		E1000_WRITE_REG(hw, E1000_IMS, E1000_IMS_LSC);
3765 	} else
3766 		E1000_WRITE_REG(hw, E1000_IMS, IMS_ENABLE_MASK);
3767 	E1000_WRITE_FLUSH(hw);
3768 }
3769 
3770 static void
3771 igb_if_intr_disable(if_ctx_t ctx)
3772 {
3773 	struct e1000_softc *sc = iflib_get_softc(ctx);
3774 	struct e1000_hw *hw = &sc->hw;
3775 
3776 	if (__predict_true(sc->intr_type == IFLIB_INTR_MSIX)) {
3777 		E1000_WRITE_REG(hw, E1000_EIMC, 0xffffffff);
3778 		E1000_WRITE_REG(hw, E1000_EIAC, 0);
3779 	}
3780 	E1000_WRITE_REG(hw, E1000_IMC, 0xffffffff);
3781 	E1000_WRITE_FLUSH(hw);
3782 }
3783 
3784 /*
3785  * Bit of a misnomer, what this really means is
3786  * to enable OS management of the system... aka
3787  * to disable special hardware management features
3788  */
3789 static void
3790 em_init_manageability(struct e1000_softc *sc)
3791 {
3792 	/* A shared code workaround */
3793 #define E1000_82542_MANC2H E1000_MANC2H
3794 	if (sc->has_manage) {
3795 		int manc2h = E1000_READ_REG(&sc->hw, E1000_MANC2H);
3796 		int manc = E1000_READ_REG(&sc->hw, E1000_MANC);
3797 
3798 		/* disable hardware interception of ARP */
3799 		manc &= ~(E1000_MANC_ARP_EN);
3800 
3801 		/* enable receiving management packets to the host */
3802 		manc |= E1000_MANC_EN_MNG2HOST;
3803 #define E1000_MNG2HOST_PORT_623 (1 << 5)
3804 #define E1000_MNG2HOST_PORT_664 (1 << 6)
3805 		manc2h |= E1000_MNG2HOST_PORT_623;
3806 		manc2h |= E1000_MNG2HOST_PORT_664;
3807 		E1000_WRITE_REG(&sc->hw, E1000_MANC2H, manc2h);
3808 		E1000_WRITE_REG(&sc->hw, E1000_MANC, manc);
3809 	}
3810 }
3811 
3812 /*
3813  * Give control back to hardware management
3814  * controller if there is one.
3815  */
3816 static void
3817 em_release_manageability(struct e1000_softc *sc)
3818 {
3819 	if (sc->has_manage) {
3820 		int manc = E1000_READ_REG(&sc->hw, E1000_MANC);
3821 
3822 		/* re-enable hardware interception of ARP */
3823 		manc |= E1000_MANC_ARP_EN;
3824 		manc &= ~E1000_MANC_EN_MNG2HOST;
3825 
3826 		E1000_WRITE_REG(&sc->hw, E1000_MANC, manc);
3827 	}
3828 }
3829 
3830 /*
3831  * em_get_hw_control sets the {CTRL_EXT|FWSM}:DRV_LOAD bit.
3832  * For ASF and Pass Through versions of f/w this means
3833  * that the driver is loaded. For AMT version type f/w
3834  * this means that the network i/f is open.
3835  */
3836 static void
3837 em_get_hw_control(struct e1000_softc *sc)
3838 {
3839 	u32 ctrl_ext, swsm;
3840 
3841 	if (sc->vf_ifp)
3842 		return;
3843 
3844 	if (sc->hw.mac.type == e1000_82573) {
3845 		swsm = E1000_READ_REG(&sc->hw, E1000_SWSM);
3846 		E1000_WRITE_REG(&sc->hw, E1000_SWSM,
3847 		    swsm | E1000_SWSM_DRV_LOAD);
3848 		return;
3849 	}
3850 	/* else */
3851 	ctrl_ext = E1000_READ_REG(&sc->hw, E1000_CTRL_EXT);
3852 	E1000_WRITE_REG(&sc->hw, E1000_CTRL_EXT,
3853 	    ctrl_ext | E1000_CTRL_EXT_DRV_LOAD);
3854 }
3855 
3856 /*
3857  * em_release_hw_control resets {CTRL_EXT|FWSM}:DRV_LOAD bit.
3858  * For ASF and Pass Through versions of f/w this means that
3859  * the driver is no longer loaded. For AMT versions of the
3860  * f/w this means that the network i/f is closed.
3861  */
3862 static void
3863 em_release_hw_control(struct e1000_softc *sc)
3864 {
3865 	u32 ctrl_ext, swsm;
3866 
3867 	if (!sc->has_manage)
3868 		return;
3869 
3870 	if (sc->hw.mac.type == e1000_82573) {
3871 		swsm = E1000_READ_REG(&sc->hw, E1000_SWSM);
3872 		E1000_WRITE_REG(&sc->hw, E1000_SWSM,
3873 		    swsm & ~E1000_SWSM_DRV_LOAD);
3874 		return;
3875 	}
3876 	/* else */
3877 	ctrl_ext = E1000_READ_REG(&sc->hw, E1000_CTRL_EXT);
3878 	E1000_WRITE_REG(&sc->hw, E1000_CTRL_EXT,
3879 	    ctrl_ext & ~E1000_CTRL_EXT_DRV_LOAD);
3880 	return;
3881 }
3882 
3883 static int
3884 em_is_valid_ether_addr(u8 *addr)
3885 {
3886 	char zero_addr[6] = { 0, 0, 0, 0, 0, 0 };
3887 
3888 	if ((addr[0] & 1) || (!bcmp(addr, zero_addr, ETHER_ADDR_LEN))) {
3889 		return (false);
3890 	}
3891 
3892 	return (true);
3893 }
3894 
3895 static bool
3896 em_automask_tso(if_ctx_t ctx)
3897 {
3898 	struct e1000_softc *sc = iflib_get_softc(ctx);
3899 	if_softc_ctx_t scctx = iflib_get_softc_ctx(ctx);
3900 	if_t ifp = iflib_get_ifp(ctx);
3901 
3902 	if (!em_unsupported_tso && sc->link_speed &&
3903 	    sc->link_speed != SPEED_1000 && scctx->isc_capenable & IFCAP_TSO) {
3904 		device_printf(sc->dev, "Disabling TSO for 10/100 Ethernet.\n");
3905 		sc->tso_automasked = scctx->isc_capenable & IFCAP_TSO;
3906 		scctx->isc_capenable &= ~IFCAP_TSO;
3907 		if_setcapenablebit(ifp, 0, IFCAP_TSO);
3908 		/* iflib_init_locked handles ifnet hwassistbits */
3909 		iflib_request_reset(ctx);
3910 		return true;
3911 	} else if (sc->link_speed == SPEED_1000 && sc->tso_automasked) {
3912 		device_printf(sc->dev, "Re-enabling TSO for GbE.\n");
3913 		scctx->isc_capenable |= sc->tso_automasked;
3914 		if_setcapenablebit(ifp, sc->tso_automasked, 0);
3915 		sc->tso_automasked = 0;
3916 		/* iflib_init_locked handles ifnet hwassistbits */
3917 		iflib_request_reset(ctx);
3918 		return true;
3919 	}
3920 
3921 	return false;
3922 }
3923 
3924 /*
3925 ** Parse the interface capabilities with regard
3926 ** to both system management and wake-on-lan for
3927 ** later use.
3928 */
3929 static void
3930 em_get_wakeup(if_ctx_t ctx)
3931 {
3932 	struct e1000_softc *sc = iflib_get_softc(ctx);
3933 	device_t dev = iflib_get_dev(ctx);
3934 	u16 eeprom_data = 0, device_id, apme_mask;
3935 
3936 	sc->has_manage = e1000_enable_mng_pass_thru(&sc->hw);
3937 	apme_mask = EM_EEPROM_APME;
3938 
3939 	switch (sc->hw.mac.type) {
3940 	case e1000_82542:
3941 	case e1000_82543:
3942 		break;
3943 	case e1000_82544:
3944 		e1000_read_nvm(&sc->hw,
3945 		    NVM_INIT_CONTROL2_REG, 1, &eeprom_data);
3946 		apme_mask = EM_82544_APME;
3947 		break;
3948 	case e1000_82546:
3949 	case e1000_82546_rev_3:
3950 		if (sc->hw.bus.func == 1) {
3951 			e1000_read_nvm(&sc->hw,
3952 			    NVM_INIT_CONTROL3_PORT_B, 1, &eeprom_data);
3953 			break;
3954 		} else
3955 			e1000_read_nvm(&sc->hw,
3956 			    NVM_INIT_CONTROL3_PORT_A, 1, &eeprom_data);
3957 		break;
3958 	case e1000_82573:
3959 	case e1000_82583:
3960 		sc->has_amt = true;
3961 		/* FALLTHROUGH */
3962 	case e1000_82571:
3963 	case e1000_82572:
3964 	case e1000_80003es2lan:
3965 		if (sc->hw.bus.func == 1) {
3966 			e1000_read_nvm(&sc->hw,
3967 			    NVM_INIT_CONTROL3_PORT_B, 1, &eeprom_data);
3968 			break;
3969 		} else
3970 			e1000_read_nvm(&sc->hw,
3971 			    NVM_INIT_CONTROL3_PORT_A, 1, &eeprom_data);
3972 		break;
3973 	case e1000_ich8lan:
3974 	case e1000_ich9lan:
3975 	case e1000_ich10lan:
3976 	case e1000_pchlan:
3977 	case e1000_pch2lan:
3978 	case e1000_pch_lpt:
3979 	case e1000_pch_spt:
3980 	case e1000_82575:	/* listing all igb devices */
3981 	case e1000_82576:
3982 	case e1000_82580:
3983 	case e1000_i350:
3984 	case e1000_i354:
3985 	case e1000_i210:
3986 	case e1000_i211:
3987 	case e1000_vfadapt:
3988 	case e1000_vfadapt_i350:
3989 		apme_mask = E1000_WUC_APME;
3990 		sc->has_amt = true;
3991 		eeprom_data = E1000_READ_REG(&sc->hw, E1000_WUC);
3992 		break;
3993 	default:
3994 		e1000_read_nvm(&sc->hw,
3995 		    NVM_INIT_CONTROL3_PORT_A, 1, &eeprom_data);
3996 		break;
3997 	}
3998 	if (eeprom_data & apme_mask)
3999 		sc->wol = (E1000_WUFC_MAG | E1000_WUFC_MC);
4000 	/*
4001 	 * We have the eeprom settings, now apply the special cases
4002 	 * where the eeprom may be wrong or the board won't support
4003 	 * wake on lan on a particular port
4004 	 */
4005 	device_id = pci_get_device(dev);
4006 	switch (device_id) {
4007 	case E1000_DEV_ID_82546GB_PCIE:
4008 		sc->wol = 0;
4009 		break;
4010 	case E1000_DEV_ID_82546EB_FIBER:
4011 	case E1000_DEV_ID_82546GB_FIBER:
4012 		/* Wake events only supported on port A for dual fiber
4013 		 * regardless of eeprom setting */
4014 		if (E1000_READ_REG(&sc->hw, E1000_STATUS) &
4015 		    E1000_STATUS_FUNC_1)
4016 			sc->wol = 0;
4017 		break;
4018 	case E1000_DEV_ID_82546GB_QUAD_COPPER_KSP3:
4019 		/* if quad port adapter, disable WoL on all but port A */
4020 		if (global_quad_port_a != 0)
4021 			sc->wol = 0;
4022 		/* Reset for multiple quad port adapters */
4023 		if (++global_quad_port_a == 4)
4024 			global_quad_port_a = 0;
4025 		break;
4026 	case E1000_DEV_ID_82571EB_FIBER:
4027 		/* Wake events only supported on port A for dual fiber
4028 		 * regardless of eeprom setting */
4029 		if (E1000_READ_REG(&sc->hw, E1000_STATUS) &
4030 		    E1000_STATUS_FUNC_1)
4031 			sc->wol = 0;
4032 		break;
4033 	case E1000_DEV_ID_82571EB_QUAD_COPPER:
4034 	case E1000_DEV_ID_82571EB_QUAD_FIBER:
4035 	case E1000_DEV_ID_82571EB_QUAD_COPPER_LP:
4036 		/* if quad port adapter, disable WoL on all but port A */
4037 		if (global_quad_port_a != 0)
4038 			sc->wol = 0;
4039 		/* Reset for multiple quad port adapters */
4040 		if (++global_quad_port_a == 4)
4041 			global_quad_port_a = 0;
4042 		break;
4043 	}
4044 	return;
4045 }
4046 
4047 
4048 /*
4049  * Enable PCI Wake On Lan capability
4050  */
4051 static void
4052 em_enable_wakeup(if_ctx_t ctx)
4053 {
4054 	struct e1000_softc *sc = iflib_get_softc(ctx);
4055 	device_t dev = iflib_get_dev(ctx);
4056 	if_t ifp = iflib_get_ifp(ctx);
4057 	int error = 0;
4058 	u32 pmc, ctrl, ctrl_ext, rctl;
4059 	u16 status;
4060 
4061 	if (pci_find_cap(dev, PCIY_PMG, &pmc) != 0)
4062 		return;
4063 
4064 	/*
4065 	 * Determine type of Wakeup: note that wol
4066 	 * is set with all bits on by default.
4067 	 */
4068 	if ((if_getcapenable(ifp) & IFCAP_WOL_MAGIC) == 0)
4069 		sc->wol &= ~E1000_WUFC_MAG;
4070 
4071 	if ((if_getcapenable(ifp) & IFCAP_WOL_UCAST) == 0)
4072 		sc->wol &= ~E1000_WUFC_EX;
4073 
4074 	if ((if_getcapenable(ifp) & IFCAP_WOL_MCAST) == 0)
4075 		sc->wol &= ~E1000_WUFC_MC;
4076 	else {
4077 		rctl = E1000_READ_REG(&sc->hw, E1000_RCTL);
4078 		rctl |= E1000_RCTL_MPE;
4079 		E1000_WRITE_REG(&sc->hw, E1000_RCTL, rctl);
4080 	}
4081 
4082 	if (!(sc->wol & (E1000_WUFC_EX | E1000_WUFC_MAG | E1000_WUFC_MC)))
4083 		goto pme;
4084 
4085 	/* Advertise the wakeup capability */
4086 	ctrl = E1000_READ_REG(&sc->hw, E1000_CTRL);
4087 	ctrl |= (E1000_CTRL_SWDPIN2 | E1000_CTRL_SWDPIN3);
4088 	E1000_WRITE_REG(&sc->hw, E1000_CTRL, ctrl);
4089 
4090 	/* Keep the laser running on Fiber adapters */
4091 	if (sc->hw.phy.media_type == e1000_media_type_fiber ||
4092 	    sc->hw.phy.media_type == e1000_media_type_internal_serdes) {
4093 		ctrl_ext = E1000_READ_REG(&sc->hw, E1000_CTRL_EXT);
4094 		ctrl_ext |= E1000_CTRL_EXT_SDP3_DATA;
4095 		E1000_WRITE_REG(&sc->hw, E1000_CTRL_EXT, ctrl_ext);
4096 	}
4097 
4098 	if ((sc->hw.mac.type == e1000_ich8lan) ||
4099 	    (sc->hw.mac.type == e1000_pchlan) ||
4100 	    (sc->hw.mac.type == e1000_ich9lan) ||
4101 	    (sc->hw.mac.type == e1000_ich10lan))
4102 		e1000_suspend_workarounds_ich8lan(&sc->hw);
4103 
4104 	if ( sc->hw.mac.type >= e1000_pchlan) {
4105 		error = em_enable_phy_wakeup(sc);
4106 		if (error)
4107 			goto pme;
4108 	} else {
4109 		/* Enable wakeup by the MAC */
4110 		E1000_WRITE_REG(&sc->hw, E1000_WUC, E1000_WUC_PME_EN);
4111 		E1000_WRITE_REG(&sc->hw, E1000_WUFC, sc->wol);
4112 	}
4113 
4114 	if (sc->hw.phy.type == e1000_phy_igp_3)
4115 		e1000_igp3_phy_powerdown_workaround_ich8lan(&sc->hw);
4116 
4117 pme:
4118 	status = pci_read_config(dev, pmc + PCIR_POWER_STATUS, 2);
4119 	status &= ~(PCIM_PSTAT_PME | PCIM_PSTAT_PMEENABLE);
4120 	if (!error && (if_getcapenable(ifp) & IFCAP_WOL))
4121 		status |= PCIM_PSTAT_PME | PCIM_PSTAT_PMEENABLE;
4122 	pci_write_config(dev, pmc + PCIR_POWER_STATUS, status, 2);
4123 
4124 	return;
4125 }
4126 
4127 /*
4128  * WOL in the newer chipset interfaces (pchlan)
4129  * require thing to be copied into the phy
4130  */
4131 static int
4132 em_enable_phy_wakeup(struct e1000_softc *sc)
4133 {
4134 	struct e1000_hw *hw = &sc->hw;
4135 	u32 mreg, ret = 0;
4136 	u16 preg;
4137 
4138 	/* copy MAC RARs to PHY RARs */
4139 	e1000_copy_rx_addrs_to_phy_ich8lan(hw);
4140 
4141 	/* copy MAC MTA to PHY MTA */
4142 	for (int i = 0; i < hw->mac.mta_reg_count; i++) {
4143 		mreg = E1000_READ_REG_ARRAY(hw, E1000_MTA, i);
4144 		e1000_write_phy_reg(hw, BM_MTA(i), (u16)(mreg & 0xFFFF));
4145 		e1000_write_phy_reg(hw, BM_MTA(i) + 1,
4146 		    (u16)((mreg >> 16) & 0xFFFF));
4147 	}
4148 
4149 	/* configure PHY Rx Control register */
4150 	e1000_read_phy_reg(hw, BM_RCTL, &preg);
4151 	mreg = E1000_READ_REG(hw, E1000_RCTL);
4152 	if (mreg & E1000_RCTL_UPE)
4153 		preg |= BM_RCTL_UPE;
4154 	if (mreg & E1000_RCTL_MPE)
4155 		preg |= BM_RCTL_MPE;
4156 	preg &= ~(BM_RCTL_MO_MASK);
4157 	if (mreg & E1000_RCTL_MO_3)
4158 		preg |= (((mreg & E1000_RCTL_MO_3) >> E1000_RCTL_MO_SHIFT)
4159 				<< BM_RCTL_MO_SHIFT);
4160 	if (mreg & E1000_RCTL_BAM)
4161 		preg |= BM_RCTL_BAM;
4162 	if (mreg & E1000_RCTL_PMCF)
4163 		preg |= BM_RCTL_PMCF;
4164 	mreg = E1000_READ_REG(hw, E1000_CTRL);
4165 	if (mreg & E1000_CTRL_RFCE)
4166 		preg |= BM_RCTL_RFCE;
4167 	e1000_write_phy_reg(hw, BM_RCTL, preg);
4168 
4169 	/* enable PHY wakeup in MAC register */
4170 	E1000_WRITE_REG(hw, E1000_WUC,
4171 	    E1000_WUC_PHY_WAKE | E1000_WUC_PME_EN | E1000_WUC_APME);
4172 	E1000_WRITE_REG(hw, E1000_WUFC, sc->wol);
4173 
4174 	/* configure and enable PHY wakeup in PHY registers */
4175 	e1000_write_phy_reg(hw, BM_WUFC, sc->wol);
4176 	e1000_write_phy_reg(hw, BM_WUC, E1000_WUC_PME_EN);
4177 
4178 	/* activate PHY wakeup */
4179 	ret = hw->phy.ops.acquire(hw);
4180 	if (ret) {
4181 		printf("Could not acquire PHY\n");
4182 		return ret;
4183 	}
4184 	e1000_write_phy_reg_mdic(hw, IGP01E1000_PHY_PAGE_SELECT,
4185 	                         (BM_WUC_ENABLE_PAGE << IGP_PAGE_SHIFT));
4186 	ret = e1000_read_phy_reg_mdic(hw, BM_WUC_ENABLE_REG, &preg);
4187 	if (ret) {
4188 		printf("Could not read PHY page 769\n");
4189 		goto out;
4190 	}
4191 	preg |= BM_WUC_ENABLE_BIT | BM_WUC_HOST_WU_BIT;
4192 	ret = e1000_write_phy_reg_mdic(hw, BM_WUC_ENABLE_REG, preg);
4193 	if (ret)
4194 		printf("Could not set PHY Host Wakeup bit\n");
4195 out:
4196 	hw->phy.ops.release(hw);
4197 
4198 	return ret;
4199 }
4200 
4201 static void
4202 em_if_led_func(if_ctx_t ctx, int onoff)
4203 {
4204 	struct e1000_softc *sc = iflib_get_softc(ctx);
4205 
4206 	if (onoff) {
4207 		e1000_setup_led(&sc->hw);
4208 		e1000_led_on(&sc->hw);
4209 	} else {
4210 		e1000_led_off(&sc->hw);
4211 		e1000_cleanup_led(&sc->hw);
4212 	}
4213 }
4214 
4215 /*
4216  * Disable the L0S and L1 LINK states
4217  */
4218 static void
4219 em_disable_aspm(struct e1000_softc *sc)
4220 {
4221 	int base, reg;
4222 	u16 link_cap,link_ctrl;
4223 	device_t dev = sc->dev;
4224 
4225 	switch (sc->hw.mac.type) {
4226 	case e1000_82573:
4227 	case e1000_82574:
4228 	case e1000_82583:
4229 		break;
4230 	default:
4231 		return;
4232 	}
4233 	if (pci_find_cap(dev, PCIY_EXPRESS, &base) != 0)
4234 		return;
4235 	reg = base + PCIER_LINK_CAP;
4236 	link_cap = pci_read_config(dev, reg, 2);
4237 	if ((link_cap & PCIEM_LINK_CAP_ASPM) == 0)
4238 		return;
4239 	reg = base + PCIER_LINK_CTL;
4240 	link_ctrl = pci_read_config(dev, reg, 2);
4241 	link_ctrl &= ~PCIEM_LINK_CTL_ASPMC;
4242 	pci_write_config(dev, reg, link_ctrl, 2);
4243 	return;
4244 }
4245 
4246 /**********************************************************************
4247  *
4248  *  Update the board statistics counters.
4249  *
4250  **********************************************************************/
4251 static void
4252 em_update_stats_counters(struct e1000_softc *sc)
4253 {
4254 	u64 prev_xoffrxc = sc->stats.xoffrxc;
4255 
4256 	if(sc->hw.phy.media_type == e1000_media_type_copper ||
4257 	   (E1000_READ_REG(&sc->hw, E1000_STATUS) & E1000_STATUS_LU)) {
4258 		sc->stats.symerrs += E1000_READ_REG(&sc->hw, E1000_SYMERRS);
4259 		sc->stats.sec += E1000_READ_REG(&sc->hw, E1000_SEC);
4260 	}
4261 	sc->stats.crcerrs += E1000_READ_REG(&sc->hw, E1000_CRCERRS);
4262 	sc->stats.mpc += E1000_READ_REG(&sc->hw, E1000_MPC);
4263 	sc->stats.scc += E1000_READ_REG(&sc->hw, E1000_SCC);
4264 	sc->stats.ecol += E1000_READ_REG(&sc->hw, E1000_ECOL);
4265 
4266 	sc->stats.mcc += E1000_READ_REG(&sc->hw, E1000_MCC);
4267 	sc->stats.latecol += E1000_READ_REG(&sc->hw, E1000_LATECOL);
4268 	sc->stats.colc += E1000_READ_REG(&sc->hw, E1000_COLC);
4269 	sc->stats.dc += E1000_READ_REG(&sc->hw, E1000_DC);
4270 	sc->stats.rlec += E1000_READ_REG(&sc->hw, E1000_RLEC);
4271 	sc->stats.xonrxc += E1000_READ_REG(&sc->hw, E1000_XONRXC);
4272 	sc->stats.xontxc += E1000_READ_REG(&sc->hw, E1000_XONTXC);
4273 	sc->stats.xoffrxc += E1000_READ_REG(&sc->hw, E1000_XOFFRXC);
4274 	/*
4275 	 ** For watchdog management we need to know if we have been
4276 	 ** paused during the last interval, so capture that here.
4277 	*/
4278 	if (sc->stats.xoffrxc != prev_xoffrxc)
4279 		sc->shared->isc_pause_frames = 1;
4280 	sc->stats.xofftxc += E1000_READ_REG(&sc->hw, E1000_XOFFTXC);
4281 	sc->stats.fcruc += E1000_READ_REG(&sc->hw, E1000_FCRUC);
4282 	sc->stats.prc64 += E1000_READ_REG(&sc->hw, E1000_PRC64);
4283 	sc->stats.prc127 += E1000_READ_REG(&sc->hw, E1000_PRC127);
4284 	sc->stats.prc255 += E1000_READ_REG(&sc->hw, E1000_PRC255);
4285 	sc->stats.prc511 += E1000_READ_REG(&sc->hw, E1000_PRC511);
4286 	sc->stats.prc1023 += E1000_READ_REG(&sc->hw, E1000_PRC1023);
4287 	sc->stats.prc1522 += E1000_READ_REG(&sc->hw, E1000_PRC1522);
4288 	sc->stats.gprc += E1000_READ_REG(&sc->hw, E1000_GPRC);
4289 	sc->stats.bprc += E1000_READ_REG(&sc->hw, E1000_BPRC);
4290 	sc->stats.mprc += E1000_READ_REG(&sc->hw, E1000_MPRC);
4291 	sc->stats.gptc += E1000_READ_REG(&sc->hw, E1000_GPTC);
4292 
4293 	/* For the 64-bit byte counters the low dword must be read first. */
4294 	/* Both registers clear on the read of the high dword */
4295 
4296 	sc->stats.gorc += E1000_READ_REG(&sc->hw, E1000_GORCL) +
4297 	    ((u64)E1000_READ_REG(&sc->hw, E1000_GORCH) << 32);
4298 	sc->stats.gotc += E1000_READ_REG(&sc->hw, E1000_GOTCL) +
4299 	    ((u64)E1000_READ_REG(&sc->hw, E1000_GOTCH) << 32);
4300 
4301 	sc->stats.rnbc += E1000_READ_REG(&sc->hw, E1000_RNBC);
4302 	sc->stats.ruc += E1000_READ_REG(&sc->hw, E1000_RUC);
4303 	sc->stats.rfc += E1000_READ_REG(&sc->hw, E1000_RFC);
4304 	sc->stats.roc += E1000_READ_REG(&sc->hw, E1000_ROC);
4305 	sc->stats.rjc += E1000_READ_REG(&sc->hw, E1000_RJC);
4306 
4307 	sc->stats.tor += E1000_READ_REG(&sc->hw, E1000_TORH);
4308 	sc->stats.tot += E1000_READ_REG(&sc->hw, E1000_TOTH);
4309 
4310 	sc->stats.tpr += E1000_READ_REG(&sc->hw, E1000_TPR);
4311 	sc->stats.tpt += E1000_READ_REG(&sc->hw, E1000_TPT);
4312 	sc->stats.ptc64 += E1000_READ_REG(&sc->hw, E1000_PTC64);
4313 	sc->stats.ptc127 += E1000_READ_REG(&sc->hw, E1000_PTC127);
4314 	sc->stats.ptc255 += E1000_READ_REG(&sc->hw, E1000_PTC255);
4315 	sc->stats.ptc511 += E1000_READ_REG(&sc->hw, E1000_PTC511);
4316 	sc->stats.ptc1023 += E1000_READ_REG(&sc->hw, E1000_PTC1023);
4317 	sc->stats.ptc1522 += E1000_READ_REG(&sc->hw, E1000_PTC1522);
4318 	sc->stats.mptc += E1000_READ_REG(&sc->hw, E1000_MPTC);
4319 	sc->stats.bptc += E1000_READ_REG(&sc->hw, E1000_BPTC);
4320 
4321 	/* Interrupt Counts */
4322 
4323 	sc->stats.iac += E1000_READ_REG(&sc->hw, E1000_IAC);
4324 	sc->stats.icrxptc += E1000_READ_REG(&sc->hw, E1000_ICRXPTC);
4325 	sc->stats.icrxatc += E1000_READ_REG(&sc->hw, E1000_ICRXATC);
4326 	sc->stats.ictxptc += E1000_READ_REG(&sc->hw, E1000_ICTXPTC);
4327 	sc->stats.ictxatc += E1000_READ_REG(&sc->hw, E1000_ICTXATC);
4328 	sc->stats.ictxqec += E1000_READ_REG(&sc->hw, E1000_ICTXQEC);
4329 	sc->stats.ictxqmtc += E1000_READ_REG(&sc->hw, E1000_ICTXQMTC);
4330 	sc->stats.icrxdmtc += E1000_READ_REG(&sc->hw, E1000_ICRXDMTC);
4331 	sc->stats.icrxoc += E1000_READ_REG(&sc->hw, E1000_ICRXOC);
4332 
4333 	if (sc->hw.mac.type >= e1000_82543) {
4334 		sc->stats.algnerrc +=
4335 		E1000_READ_REG(&sc->hw, E1000_ALGNERRC);
4336 		sc->stats.rxerrc +=
4337 		E1000_READ_REG(&sc->hw, E1000_RXERRC);
4338 		sc->stats.tncrs +=
4339 		E1000_READ_REG(&sc->hw, E1000_TNCRS);
4340 		sc->stats.cexterr +=
4341 		E1000_READ_REG(&sc->hw, E1000_CEXTERR);
4342 		sc->stats.tsctc +=
4343 		E1000_READ_REG(&sc->hw, E1000_TSCTC);
4344 		sc->stats.tsctfc +=
4345 		E1000_READ_REG(&sc->hw, E1000_TSCTFC);
4346 	}
4347 }
4348 
4349 static uint64_t
4350 em_if_get_counter(if_ctx_t ctx, ift_counter cnt)
4351 {
4352 	struct e1000_softc *sc = iflib_get_softc(ctx);
4353 	if_t ifp = iflib_get_ifp(ctx);
4354 
4355 	switch (cnt) {
4356 	case IFCOUNTER_COLLISIONS:
4357 		return (sc->stats.colc);
4358 	case IFCOUNTER_IERRORS:
4359 		return (sc->dropped_pkts + sc->stats.rxerrc +
4360 		    sc->stats.crcerrs + sc->stats.algnerrc +
4361 		    sc->stats.ruc + sc->stats.roc +
4362 		    sc->stats.mpc + sc->stats.cexterr);
4363 	case IFCOUNTER_OERRORS:
4364 		return (sc->stats.ecol + sc->stats.latecol +
4365 		    sc->watchdog_events);
4366 	default:
4367 		return (if_get_counter_default(ifp, cnt));
4368 	}
4369 }
4370 
4371 /* em_if_needs_restart - Tell iflib when the driver needs to be reinitialized
4372  * @ctx: iflib context
4373  * @event: event code to check
4374  *
4375  * Defaults to returning false for unknown events.
4376  *
4377  * @returns true if iflib needs to reinit the interface
4378  */
4379 static bool
4380 em_if_needs_restart(if_ctx_t ctx __unused, enum iflib_restart_event event)
4381 {
4382 	switch (event) {
4383 	case IFLIB_RESTART_VLAN_CONFIG:
4384 	default:
4385 		return (false);
4386 	}
4387 }
4388 
4389 /* Export a single 32-bit register via a read-only sysctl. */
4390 static int
4391 em_sysctl_reg_handler(SYSCTL_HANDLER_ARGS)
4392 {
4393 	struct e1000_softc *sc;
4394 	u_int val;
4395 
4396 	sc = oidp->oid_arg1;
4397 	val = E1000_READ_REG(&sc->hw, oidp->oid_arg2);
4398 	return (sysctl_handle_int(oidp, &val, 0, req));
4399 }
4400 
4401 /*
4402  * Add sysctl variables, one per statistic, to the system.
4403  */
4404 static void
4405 em_add_hw_stats(struct e1000_softc *sc)
4406 {
4407 	device_t dev = iflib_get_dev(sc->ctx);
4408 	struct em_tx_queue *tx_que = sc->tx_queues;
4409 	struct em_rx_queue *rx_que = sc->rx_queues;
4410 
4411 	struct sysctl_ctx_list *ctx = device_get_sysctl_ctx(dev);
4412 	struct sysctl_oid *tree = device_get_sysctl_tree(dev);
4413 	struct sysctl_oid_list *child = SYSCTL_CHILDREN(tree);
4414 	struct e1000_hw_stats *stats = &sc->stats;
4415 
4416 	struct sysctl_oid *stat_node, *queue_node, *int_node;
4417 	struct sysctl_oid_list *stat_list, *queue_list, *int_list;
4418 
4419 #define QUEUE_NAME_LEN 32
4420 	char namebuf[QUEUE_NAME_LEN];
4421 
4422 	/* Driver Statistics */
4423 	SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "dropped",
4424 			CTLFLAG_RD, &sc->dropped_pkts,
4425 			"Driver dropped packets");
4426 	SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "link_irq",
4427 			CTLFLAG_RD, &sc->link_irq,
4428 			"Link MSI-X IRQ Handled");
4429 	SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "rx_overruns",
4430 			CTLFLAG_RD, &sc->rx_overruns,
4431 			"RX overruns");
4432 	SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "watchdog_timeouts",
4433 			CTLFLAG_RD, &sc->watchdog_events,
4434 			"Watchdog timeouts");
4435 	SYSCTL_ADD_PROC(ctx, child, OID_AUTO, "device_control",
4436 	    CTLTYPE_UINT | CTLFLAG_RD | CTLFLAG_NEEDGIANT,
4437 	    sc, E1000_CTRL, em_sysctl_reg_handler, "IU",
4438 	    "Device Control Register");
4439 	SYSCTL_ADD_PROC(ctx, child, OID_AUTO, "rx_control",
4440 	    CTLTYPE_UINT | CTLFLAG_RD | CTLFLAG_NEEDGIANT,
4441 	    sc, E1000_RCTL, em_sysctl_reg_handler, "IU",
4442 	    "Receiver Control Register");
4443 	SYSCTL_ADD_UINT(ctx, child, OID_AUTO, "fc_high_water",
4444 			CTLFLAG_RD, &sc->hw.fc.high_water, 0,
4445 			"Flow Control High Watermark");
4446 	SYSCTL_ADD_UINT(ctx, child, OID_AUTO, "fc_low_water",
4447 			CTLFLAG_RD, &sc->hw.fc.low_water, 0,
4448 			"Flow Control Low Watermark");
4449 
4450 	for (int i = 0; i < sc->tx_num_queues; i++, tx_que++) {
4451 		struct tx_ring *txr = &tx_que->txr;
4452 		snprintf(namebuf, QUEUE_NAME_LEN, "queue_tx_%d", i);
4453 		queue_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, namebuf,
4454 		    CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, "TX Queue Name");
4455 		queue_list = SYSCTL_CHILDREN(queue_node);
4456 
4457 		SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "txd_head",
4458 		    CTLTYPE_UINT | CTLFLAG_RD | CTLFLAG_NEEDGIANT, sc,
4459 		    E1000_TDH(txr->me), em_sysctl_reg_handler, "IU",
4460 		    "Transmit Descriptor Head");
4461 		SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "txd_tail",
4462 		    CTLTYPE_UINT | CTLFLAG_RD | CTLFLAG_NEEDGIANT, sc,
4463 		    E1000_TDT(txr->me), em_sysctl_reg_handler, "IU",
4464 		    "Transmit Descriptor Tail");
4465 		SYSCTL_ADD_ULONG(ctx, queue_list, OID_AUTO, "tx_irq",
4466 				CTLFLAG_RD, &txr->tx_irq,
4467 				"Queue MSI-X Transmit Interrupts");
4468 	}
4469 
4470 	for (int j = 0; j < sc->rx_num_queues; j++, rx_que++) {
4471 		struct rx_ring *rxr = &rx_que->rxr;
4472 		snprintf(namebuf, QUEUE_NAME_LEN, "queue_rx_%d", j);
4473 		queue_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, namebuf,
4474 		    CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, "RX Queue Name");
4475 		queue_list = SYSCTL_CHILDREN(queue_node);
4476 
4477 		SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "rxd_head",
4478 		    CTLTYPE_UINT | CTLFLAG_RD | CTLFLAG_NEEDGIANT, sc,
4479 		    E1000_RDH(rxr->me), em_sysctl_reg_handler, "IU",
4480 		    "Receive Descriptor Head");
4481 		SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "rxd_tail",
4482 		    CTLTYPE_UINT | CTLFLAG_RD | CTLFLAG_NEEDGIANT, sc,
4483 		    E1000_RDT(rxr->me), em_sysctl_reg_handler, "IU",
4484 		    "Receive Descriptor Tail");
4485 		SYSCTL_ADD_ULONG(ctx, queue_list, OID_AUTO, "rx_irq",
4486 				CTLFLAG_RD, &rxr->rx_irq,
4487 				"Queue MSI-X Receive Interrupts");
4488 	}
4489 
4490 	/* MAC stats get their own sub node */
4491 
4492 	stat_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, "mac_stats",
4493 	    CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, "Statistics");
4494 	stat_list = SYSCTL_CHILDREN(stat_node);
4495 
4496 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "excess_coll",
4497 			CTLFLAG_RD, &stats->ecol,
4498 			"Excessive collisions");
4499 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "single_coll",
4500 			CTLFLAG_RD, &stats->scc,
4501 			"Single collisions");
4502 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "multiple_coll",
4503 			CTLFLAG_RD, &stats->mcc,
4504 			"Multiple collisions");
4505 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "late_coll",
4506 			CTLFLAG_RD, &stats->latecol,
4507 			"Late collisions");
4508 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "collision_count",
4509 			CTLFLAG_RD, &stats->colc,
4510 			"Collision Count");
4511 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "symbol_errors",
4512 			CTLFLAG_RD, &sc->stats.symerrs,
4513 			"Symbol Errors");
4514 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "sequence_errors",
4515 			CTLFLAG_RD, &sc->stats.sec,
4516 			"Sequence Errors");
4517 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "defer_count",
4518 			CTLFLAG_RD, &sc->stats.dc,
4519 			"Defer Count");
4520 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "missed_packets",
4521 			CTLFLAG_RD, &sc->stats.mpc,
4522 			"Missed Packets");
4523 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_no_buff",
4524 			CTLFLAG_RD, &sc->stats.rnbc,
4525 			"Receive No Buffers");
4526 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_undersize",
4527 			CTLFLAG_RD, &sc->stats.ruc,
4528 			"Receive Undersize");
4529 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_fragmented",
4530 			CTLFLAG_RD, &sc->stats.rfc,
4531 			"Fragmented Packets Received ");
4532 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_oversize",
4533 			CTLFLAG_RD, &sc->stats.roc,
4534 			"Oversized Packets Received");
4535 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_jabber",
4536 			CTLFLAG_RD, &sc->stats.rjc,
4537 			"Recevied Jabber");
4538 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_errs",
4539 			CTLFLAG_RD, &sc->stats.rxerrc,
4540 			"Receive Errors");
4541 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "crc_errs",
4542 			CTLFLAG_RD, &sc->stats.crcerrs,
4543 			"CRC errors");
4544 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "alignment_errs",
4545 			CTLFLAG_RD, &sc->stats.algnerrc,
4546 			"Alignment Errors");
4547 	/* On 82575 these are collision counts */
4548 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "coll_ext_errs",
4549 			CTLFLAG_RD, &sc->stats.cexterr,
4550 			"Collision/Carrier extension errors");
4551 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xon_recvd",
4552 			CTLFLAG_RD, &sc->stats.xonrxc,
4553 			"XON Received");
4554 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xon_txd",
4555 			CTLFLAG_RD, &sc->stats.xontxc,
4556 			"XON Transmitted");
4557 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xoff_recvd",
4558 			CTLFLAG_RD, &sc->stats.xoffrxc,
4559 			"XOFF Received");
4560 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xoff_txd",
4561 			CTLFLAG_RD, &sc->stats.xofftxc,
4562 			"XOFF Transmitted");
4563 
4564 	/* Packet Reception Stats */
4565 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "total_pkts_recvd",
4566 			CTLFLAG_RD, &sc->stats.tpr,
4567 			"Total Packets Received ");
4568 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_pkts_recvd",
4569 			CTLFLAG_RD, &sc->stats.gprc,
4570 			"Good Packets Received");
4571 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "bcast_pkts_recvd",
4572 			CTLFLAG_RD, &sc->stats.bprc,
4573 			"Broadcast Packets Received");
4574 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "mcast_pkts_recvd",
4575 			CTLFLAG_RD, &sc->stats.mprc,
4576 			"Multicast Packets Received");
4577 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_64",
4578 			CTLFLAG_RD, &sc->stats.prc64,
4579 			"64 byte frames received ");
4580 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_65_127",
4581 			CTLFLAG_RD, &sc->stats.prc127,
4582 			"65-127 byte frames received");
4583 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_128_255",
4584 			CTLFLAG_RD, &sc->stats.prc255,
4585 			"128-255 byte frames received");
4586 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_256_511",
4587 			CTLFLAG_RD, &sc->stats.prc511,
4588 			"256-511 byte frames received");
4589 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_512_1023",
4590 			CTLFLAG_RD, &sc->stats.prc1023,
4591 			"512-1023 byte frames received");
4592 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_1024_1522",
4593 			CTLFLAG_RD, &sc->stats.prc1522,
4594 			"1023-1522 byte frames received");
4595 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_octets_recvd",
4596 			CTLFLAG_RD, &sc->stats.gorc,
4597 			"Good Octets Received");
4598 
4599 	/* Packet Transmission Stats */
4600 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_octets_txd",
4601 			CTLFLAG_RD, &sc->stats.gotc,
4602 			"Good Octets Transmitted");
4603 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "total_pkts_txd",
4604 			CTLFLAG_RD, &sc->stats.tpt,
4605 			"Total Packets Transmitted");
4606 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_pkts_txd",
4607 			CTLFLAG_RD, &sc->stats.gptc,
4608 			"Good Packets Transmitted");
4609 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "bcast_pkts_txd",
4610 			CTLFLAG_RD, &sc->stats.bptc,
4611 			"Broadcast Packets Transmitted");
4612 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "mcast_pkts_txd",
4613 			CTLFLAG_RD, &sc->stats.mptc,
4614 			"Multicast Packets Transmitted");
4615 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_64",
4616 			CTLFLAG_RD, &sc->stats.ptc64,
4617 			"64 byte frames transmitted ");
4618 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_65_127",
4619 			CTLFLAG_RD, &sc->stats.ptc127,
4620 			"65-127 byte frames transmitted");
4621 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_128_255",
4622 			CTLFLAG_RD, &sc->stats.ptc255,
4623 			"128-255 byte frames transmitted");
4624 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_256_511",
4625 			CTLFLAG_RD, &sc->stats.ptc511,
4626 			"256-511 byte frames transmitted");
4627 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_512_1023",
4628 			CTLFLAG_RD, &sc->stats.ptc1023,
4629 			"512-1023 byte frames transmitted");
4630 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_1024_1522",
4631 			CTLFLAG_RD, &sc->stats.ptc1522,
4632 			"1024-1522 byte frames transmitted");
4633 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tso_txd",
4634 			CTLFLAG_RD, &sc->stats.tsctc,
4635 			"TSO Contexts Transmitted");
4636 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tso_ctx_fail",
4637 			CTLFLAG_RD, &sc->stats.tsctfc,
4638 			"TSO Contexts Failed");
4639 
4640 
4641 	/* Interrupt Stats */
4642 
4643 	int_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, "interrupts",
4644 	    CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, "Interrupt Statistics");
4645 	int_list = SYSCTL_CHILDREN(int_node);
4646 
4647 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "asserts",
4648 			CTLFLAG_RD, &sc->stats.iac,
4649 			"Interrupt Assertion Count");
4650 
4651 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "rx_pkt_timer",
4652 			CTLFLAG_RD, &sc->stats.icrxptc,
4653 			"Interrupt Cause Rx Pkt Timer Expire Count");
4654 
4655 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "rx_abs_timer",
4656 			CTLFLAG_RD, &sc->stats.icrxatc,
4657 			"Interrupt Cause Rx Abs Timer Expire Count");
4658 
4659 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "tx_pkt_timer",
4660 			CTLFLAG_RD, &sc->stats.ictxptc,
4661 			"Interrupt Cause Tx Pkt Timer Expire Count");
4662 
4663 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "tx_abs_timer",
4664 			CTLFLAG_RD, &sc->stats.ictxatc,
4665 			"Interrupt Cause Tx Abs Timer Expire Count");
4666 
4667 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "tx_queue_empty",
4668 			CTLFLAG_RD, &sc->stats.ictxqec,
4669 			"Interrupt Cause Tx Queue Empty Count");
4670 
4671 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "tx_queue_min_thresh",
4672 			CTLFLAG_RD, &sc->stats.ictxqmtc,
4673 			"Interrupt Cause Tx Queue Min Thresh Count");
4674 
4675 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "rx_desc_min_thresh",
4676 			CTLFLAG_RD, &sc->stats.icrxdmtc,
4677 			"Interrupt Cause Rx Desc Min Thresh Count");
4678 
4679 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "rx_overrun",
4680 			CTLFLAG_RD, &sc->stats.icrxoc,
4681 			"Interrupt Cause Receiver Overrun Count");
4682 }
4683 
4684 static void
4685 em_fw_version_locked(if_ctx_t ctx)
4686 {
4687 	struct e1000_softc *sc = iflib_get_softc(ctx);
4688 	struct e1000_hw *hw = &sc->hw;
4689 	struct e1000_fw_version *fw_ver = &sc->fw_ver;
4690 	uint16_t eep = 0;
4691 
4692 	/*
4693 	 * em_fw_version_locked() must run under the IFLIB_CTX_LOCK to meet the
4694 	 * NVM locking model, so we do it in em_if_attach_pre() and store the
4695 	 * info in the softc
4696 	 */
4697 	ASSERT_CTX_LOCK_HELD(hw);
4698 
4699 	*fw_ver = (struct e1000_fw_version){0};
4700 
4701 	if (hw->mac.type >= igb_mac_min) {
4702 		/*
4703 		 * Use the Shared Code for igb(4)
4704 		 */
4705 		e1000_get_fw_version(hw, fw_ver);
4706 	} else {
4707 		/*
4708 		 * Otherwise, EEPROM version should be present on (almost?) all
4709 		 * devices here
4710 		 */
4711 		if(e1000_read_nvm(hw, NVM_VERSION, 1, &eep)) {
4712 			INIT_DEBUGOUT("can't get EEPROM version");
4713 			return;
4714 		}
4715 
4716 		fw_ver->eep_major = (eep & NVM_MAJOR_MASK) >> NVM_MAJOR_SHIFT;
4717 		fw_ver->eep_minor = (eep & NVM_MINOR_MASK) >> NVM_MINOR_SHIFT;
4718 		fw_ver->eep_build = (eep & NVM_IMAGE_ID_MASK);
4719 	}
4720 }
4721 
4722 static void
4723 em_sbuf_fw_version(struct e1000_fw_version *fw_ver, struct sbuf *buf)
4724 {
4725 	const char *space = "";
4726 
4727 	if (fw_ver->eep_major || fw_ver->eep_minor || fw_ver->eep_build) {
4728 		sbuf_printf(buf, "EEPROM V%d.%d-%d", fw_ver->eep_major,
4729 			    fw_ver->eep_minor, fw_ver->eep_build);
4730 		space = " ";
4731 	}
4732 
4733 	if (fw_ver->invm_major || fw_ver->invm_minor || fw_ver->invm_img_type) {
4734 		sbuf_printf(buf, "%sNVM V%d.%d imgtype%d",
4735 			    space, fw_ver->invm_major, fw_ver->invm_minor,
4736 			    fw_ver->invm_img_type);
4737 		space = " ";
4738 	}
4739 
4740 	if (fw_ver->or_valid) {
4741 		sbuf_printf(buf, "%sOption ROM V%d-b%d-p%d",
4742 			    space, fw_ver->or_major, fw_ver->or_build,
4743 			    fw_ver->or_patch);
4744 		space = " ";
4745 	}
4746 
4747 	if (fw_ver->etrack_id)
4748 		sbuf_printf(buf, "%seTrack 0x%08x", space, fw_ver->etrack_id);
4749 }
4750 
4751 static void
4752 em_print_fw_version(struct e1000_softc *sc )
4753 {
4754 	device_t dev = sc->dev;
4755 	struct sbuf *buf;
4756 	int error = 0;
4757 
4758 	buf = sbuf_new_auto();
4759 	if (!buf) {
4760 		device_printf(dev, "Could not allocate sbuf for output.\n");
4761 		return;
4762 	}
4763 
4764 	em_sbuf_fw_version(&sc->fw_ver, buf);
4765 
4766 	error = sbuf_finish(buf);
4767 	if (error)
4768 		device_printf(dev, "Error finishing sbuf: %d\n", error);
4769 	else if (sbuf_len(buf))
4770 		device_printf(dev, "%s\n", sbuf_data(buf));
4771 
4772 	sbuf_delete(buf);
4773 }
4774 
4775 static int
4776 em_sysctl_print_fw_version(SYSCTL_HANDLER_ARGS)
4777 {
4778 	struct e1000_softc *sc = (struct e1000_softc *)arg1;
4779 	device_t dev = sc->dev;
4780 	struct sbuf *buf;
4781 	int error = 0;
4782 
4783 	buf = sbuf_new_for_sysctl(NULL, NULL, 128, req);
4784 	if (!buf) {
4785 		device_printf(dev, "Could not allocate sbuf for output.\n");
4786 		return (ENOMEM);
4787 	}
4788 
4789 	em_sbuf_fw_version(&sc->fw_ver, buf);
4790 
4791 	error = sbuf_finish(buf);
4792 	if (error)
4793 		device_printf(dev, "Error finishing sbuf: %d\n", error);
4794 
4795 	sbuf_delete(buf);
4796 
4797 	return (0);
4798 }
4799 
4800 /**********************************************************************
4801  *
4802  *  This routine provides a way to dump out the adapter eeprom,
4803  *  often a useful debug/service tool. This only dumps the first
4804  *  32 words, stuff that matters is in that extent.
4805  *
4806  **********************************************************************/
4807 static int
4808 em_sysctl_nvm_info(SYSCTL_HANDLER_ARGS)
4809 {
4810 	struct e1000_softc *sc = (struct e1000_softc *)arg1;
4811 	int error;
4812 	int result;
4813 
4814 	result = -1;
4815 	error = sysctl_handle_int(oidp, &result, 0, req);
4816 
4817 	if (error || !req->newptr)
4818 		return (error);
4819 
4820 	/*
4821 	 * This value will cause a hex dump of the
4822 	 * first 32 16-bit words of the EEPROM to
4823 	 * the screen.
4824 	 */
4825 	if (result == 1)
4826 		em_print_nvm_info(sc);
4827 
4828 	return (error);
4829 }
4830 
4831 static void
4832 em_print_nvm_info(struct e1000_softc *sc)
4833 {
4834 	struct e1000_hw *hw = &sc->hw;
4835 	struct sx *iflib_ctx_lock = iflib_ctx_lock_get(sc->ctx);
4836 	u16 eeprom_data;
4837 	int i, j, row = 0;
4838 
4839 	/* Its a bit crude, but it gets the job done */
4840 	printf("\nInterface EEPROM Dump:\n");
4841 	printf("Offset\n0x0000  ");
4842 
4843 	/* We rely on the IFLIB_CTX_LOCK as part of NVM locking model */
4844 	sx_xlock(iflib_ctx_lock);
4845 	ASSERT_CTX_LOCK_HELD(hw);
4846 	for (i = 0, j = 0; i < 32; i++, j++) {
4847 		if (j == 8) { /* Make the offset block */
4848 			j = 0; ++row;
4849 			printf("\n0x00%x0  ",row);
4850 		}
4851 		e1000_read_nvm(hw, i, 1, &eeprom_data);
4852 		printf("%04x ", eeprom_data);
4853 	}
4854 	sx_xunlock(iflib_ctx_lock);
4855 	printf("\n");
4856 }
4857 
4858 static int
4859 em_sysctl_int_delay(SYSCTL_HANDLER_ARGS)
4860 {
4861 	struct em_int_delay_info *info;
4862 	struct e1000_softc *sc;
4863 	u32 regval;
4864 	int error, usecs, ticks;
4865 
4866 	info = (struct em_int_delay_info *) arg1;
4867 	usecs = info->value;
4868 	error = sysctl_handle_int(oidp, &usecs, 0, req);
4869 	if (error != 0 || req->newptr == NULL)
4870 		return (error);
4871 	if (usecs < 0 || usecs > EM_TICKS_TO_USECS(65535))
4872 		return (EINVAL);
4873 	info->value = usecs;
4874 	ticks = EM_USECS_TO_TICKS(usecs);
4875 	if (info->offset == E1000_ITR)	/* units are 256ns here */
4876 		ticks *= 4;
4877 
4878 	sc = info->sc;
4879 
4880 	regval = E1000_READ_OFFSET(&sc->hw, info->offset);
4881 	regval = (regval & ~0xffff) | (ticks & 0xffff);
4882 	/* Handle a few special cases. */
4883 	switch (info->offset) {
4884 	case E1000_RDTR:
4885 		break;
4886 	case E1000_TIDV:
4887 		if (ticks == 0) {
4888 			sc->txd_cmd &= ~E1000_TXD_CMD_IDE;
4889 			/* Don't write 0 into the TIDV register. */
4890 			regval++;
4891 		} else
4892 			sc->txd_cmd |= E1000_TXD_CMD_IDE;
4893 		break;
4894 	}
4895 	E1000_WRITE_OFFSET(&sc->hw, info->offset, regval);
4896 	return (0);
4897 }
4898 
4899 static void
4900 em_add_int_delay_sysctl(struct e1000_softc *sc, const char *name,
4901 	const char *description, struct em_int_delay_info *info,
4902 	int offset, int value)
4903 {
4904 	info->sc = sc;
4905 	info->offset = offset;
4906 	info->value = value;
4907 	SYSCTL_ADD_PROC(device_get_sysctl_ctx(sc->dev),
4908 	    SYSCTL_CHILDREN(device_get_sysctl_tree(sc->dev)),
4909 	    OID_AUTO, name, CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_NEEDGIANT,
4910 	    info, 0, em_sysctl_int_delay, "I", description);
4911 }
4912 
4913 /*
4914  * Set flow control using sysctl:
4915  * Flow control values:
4916  *      0 - off
4917  *      1 - rx pause
4918  *      2 - tx pause
4919  *      3 - full
4920  */
4921 static int
4922 em_set_flowcntl(SYSCTL_HANDLER_ARGS)
4923 {
4924 	int error;
4925 	static int input = 3; /* default is full */
4926 	struct e1000_softc	*sc = (struct e1000_softc *) arg1;
4927 
4928 	error = sysctl_handle_int(oidp, &input, 0, req);
4929 
4930 	if ((error) || (req->newptr == NULL))
4931 		return (error);
4932 
4933 	if (input == sc->fc) /* no change? */
4934 		return (error);
4935 
4936 	switch (input) {
4937 	case e1000_fc_rx_pause:
4938 	case e1000_fc_tx_pause:
4939 	case e1000_fc_full:
4940 	case e1000_fc_none:
4941 		sc->hw.fc.requested_mode = input;
4942 		sc->fc = input;
4943 		break;
4944 	default:
4945 		/* Do nothing */
4946 		return (error);
4947 	}
4948 
4949 	sc->hw.fc.current_mode = sc->hw.fc.requested_mode;
4950 	e1000_force_mac_fc(&sc->hw);
4951 	return (error);
4952 }
4953 
4954 /*
4955  * Manage Energy Efficient Ethernet:
4956  * Control values:
4957  *     0/1 - enabled/disabled
4958  */
4959 static int
4960 em_sysctl_eee(SYSCTL_HANDLER_ARGS)
4961 {
4962 	struct e1000_softc *sc = (struct e1000_softc *) arg1;
4963 	int error, value;
4964 
4965 	value = sc->hw.dev_spec.ich8lan.eee_disable;
4966 	error = sysctl_handle_int(oidp, &value, 0, req);
4967 	if (error || req->newptr == NULL)
4968 		return (error);
4969 	sc->hw.dev_spec.ich8lan.eee_disable = (value != 0);
4970 	em_if_init(sc->ctx);
4971 
4972 	return (0);
4973 }
4974 
4975 static int
4976 em_sysctl_debug_info(SYSCTL_HANDLER_ARGS)
4977 {
4978 	struct e1000_softc *sc;
4979 	int error;
4980 	int result;
4981 
4982 	result = -1;
4983 	error = sysctl_handle_int(oidp, &result, 0, req);
4984 
4985 	if (error || !req->newptr)
4986 		return (error);
4987 
4988 	if (result == 1) {
4989 		sc = (struct e1000_softc *) arg1;
4990 		em_print_debug_info(sc);
4991 	}
4992 
4993 	return (error);
4994 }
4995 
4996 static int
4997 em_get_rs(SYSCTL_HANDLER_ARGS)
4998 {
4999 	struct e1000_softc *sc = (struct e1000_softc *) arg1;
5000 	int error;
5001 	int result;
5002 
5003 	result = 0;
5004 	error = sysctl_handle_int(oidp, &result, 0, req);
5005 
5006 	if (error || !req->newptr || result != 1)
5007 		return (error);
5008 	em_dump_rs(sc);
5009 
5010 	return (error);
5011 }
5012 
5013 static void
5014 em_if_debug(if_ctx_t ctx)
5015 {
5016 	em_dump_rs(iflib_get_softc(ctx));
5017 }
5018 
5019 /*
5020  * This routine is meant to be fluid, add whatever is
5021  * needed for debugging a problem.  -jfv
5022  */
5023 static void
5024 em_print_debug_info(struct e1000_softc *sc)
5025 {
5026 	device_t dev = iflib_get_dev(sc->ctx);
5027 	if_t ifp = iflib_get_ifp(sc->ctx);
5028 	struct tx_ring *txr = &sc->tx_queues->txr;
5029 	struct rx_ring *rxr = &sc->rx_queues->rxr;
5030 
5031 	if (if_getdrvflags(ifp) & IFF_DRV_RUNNING)
5032 		printf("Interface is RUNNING ");
5033 	else
5034 		printf("Interface is NOT RUNNING\n");
5035 
5036 	if (if_getdrvflags(ifp) & IFF_DRV_OACTIVE)
5037 		printf("and INACTIVE\n");
5038 	else
5039 		printf("and ACTIVE\n");
5040 
5041 	for (int i = 0; i < sc->tx_num_queues; i++, txr++) {
5042 		device_printf(dev, "TX Queue %d ------\n", i);
5043 		device_printf(dev, "hw tdh = %d, hw tdt = %d\n",
5044 			E1000_READ_REG(&sc->hw, E1000_TDH(i)),
5045 			E1000_READ_REG(&sc->hw, E1000_TDT(i)));
5046 
5047 	}
5048 	for (int j=0; j < sc->rx_num_queues; j++, rxr++) {
5049 		device_printf(dev, "RX Queue %d ------\n", j);
5050 		device_printf(dev, "hw rdh = %d, hw rdt = %d\n",
5051 			E1000_READ_REG(&sc->hw, E1000_RDH(j)),
5052 			E1000_READ_REG(&sc->hw, E1000_RDT(j)));
5053 	}
5054 }
5055 
5056 /*
5057  * 82574 only:
5058  * Write a new value to the EEPROM increasing the number of MSI-X
5059  * vectors from 3 to 5, for proper multiqueue support.
5060  */
5061 static void
5062 em_enable_vectors_82574(if_ctx_t ctx)
5063 {
5064 	struct e1000_softc *sc = iflib_get_softc(ctx);
5065 	struct e1000_hw *hw = &sc->hw;
5066 	device_t dev = iflib_get_dev(ctx);
5067 	u16 edata;
5068 
5069 	e1000_read_nvm(hw, EM_NVM_PCIE_CTRL, 1, &edata);
5070 	if (bootverbose)
5071 		device_printf(dev, "EM_NVM_PCIE_CTRL = %#06x\n", edata);
5072 	if (((edata & EM_NVM_MSIX_N_MASK) >> EM_NVM_MSIX_N_SHIFT) != 4) {
5073 		device_printf(dev, "Writing to eeprom: increasing "
5074 		    "reported MSI-X vectors from 3 to 5...\n");
5075 		edata &= ~(EM_NVM_MSIX_N_MASK);
5076 		edata |= 4 << EM_NVM_MSIX_N_SHIFT;
5077 		e1000_write_nvm(hw, EM_NVM_PCIE_CTRL, 1, &edata);
5078 		e1000_update_nvm_checksum(hw);
5079 		device_printf(dev, "Writing to eeprom: done\n");
5080 	}
5081 }
5082