xref: /freebsd/sys/dev/e1000/if_em.c (revision 90853dfac851afa9e8840f5a38383256d75458b6)
1 /*-
2  * SPDX-License-Identifier: BSD-2-Clause
3  *
4  * Copyright (c) 2001-2024, Intel Corporation
5  * Copyright (c) 2016 Nicole Graziano <nicole@nextbsd.org>
6  * Copyright (c) 2024 Kevin Bowling <kbowling@FreeBSD.org>
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27  * SUCH DAMAGE.
28  */
29 
30 #include "if_em.h"
31 #include <sys/sbuf.h>
32 #include <machine/_inttypes.h>
33 
34 #define em_mac_min e1000_82571
35 #define igb_mac_min e1000_82575
36 
37 /*********************************************************************
38  *  Driver version:
39  *********************************************************************/
40 static const char em_driver_version[] = "7.7.8-fbsd";
41 static const char igb_driver_version[] = "2.5.28-fbsd";
42 
43 /*********************************************************************
44  *  PCI Device ID Table
45  *
46  *  Used by probe to select devices to load on
47  *  Last field stores an index into e1000_strings
48  *  Last entry must be all 0s
49  *
50  *  { Vendor ID, Device ID, SubVendor ID, SubDevice ID, String Index }
51  *********************************************************************/
52 
53 static const pci_vendor_info_t em_vendor_info_array[] =
54 {
55 	/* Intel(R) - lem-class legacy devices */
56 	PVID(0x8086, E1000_DEV_ID_82540EM, "Intel(R) Legacy PRO/1000 MT 82540EM"),
57 	PVID(0x8086, E1000_DEV_ID_82540EM_LOM, "Intel(R) Legacy PRO/1000 MT 82540EM (LOM)"),
58 	PVID(0x8086, E1000_DEV_ID_82540EP, "Intel(R) Legacy PRO/1000 MT 82540EP"),
59 	PVID(0x8086, E1000_DEV_ID_82540EP_LOM, "Intel(R) Legacy PRO/1000 MT 82540EP (LOM)"),
60 	PVID(0x8086, E1000_DEV_ID_82540EP_LP, "Intel(R) Legacy PRO/1000 MT 82540EP (Mobile)"),
61 
62 	PVID(0x8086, E1000_DEV_ID_82541EI, "Intel(R) Legacy PRO/1000 MT 82541EI (Copper)"),
63 	PVID(0x8086, E1000_DEV_ID_82541ER, "Intel(R) Legacy PRO/1000 82541ER"),
64 	PVID(0x8086, E1000_DEV_ID_82541ER_LOM, "Intel(R) Legacy PRO/1000 MT 82541ER"),
65 	PVID(0x8086, E1000_DEV_ID_82541EI_MOBILE, "Intel(R) Legacy PRO/1000 MT 82541EI (Mobile)"),
66 	PVID(0x8086, E1000_DEV_ID_82541GI, "Intel(R) Legacy PRO/1000 MT 82541GI"),
67 	PVID(0x8086, E1000_DEV_ID_82541GI_LF, "Intel(R) Legacy PRO/1000 GT 82541PI"),
68 	PVID(0x8086, E1000_DEV_ID_82541GI_MOBILE, "Intel(R) Legacy PRO/1000 MT 82541GI (Mobile)"),
69 
70 	PVID(0x8086, E1000_DEV_ID_82542, "Intel(R) Legacy PRO/1000 82542 (Fiber)"),
71 
72 	PVID(0x8086, E1000_DEV_ID_82543GC_FIBER, "Intel(R) Legacy PRO/1000 F 82543GC (Fiber)"),
73 	PVID(0x8086, E1000_DEV_ID_82543GC_COPPER, "Intel(R) Legacy PRO/1000 T 82543GC (Copper)"),
74 
75 	PVID(0x8086, E1000_DEV_ID_82544EI_COPPER, "Intel(R) Legacy PRO/1000 XT 82544EI (Copper)"),
76 	PVID(0x8086, E1000_DEV_ID_82544EI_FIBER, "Intel(R) Legacy PRO/1000 XF 82544EI (Fiber)"),
77 	PVID(0x8086, E1000_DEV_ID_82544GC_COPPER, "Intel(R) Legacy PRO/1000 T 82544GC (Copper)"),
78 	PVID(0x8086, E1000_DEV_ID_82544GC_LOM, "Intel(R) Legacy PRO/1000 XT 82544GC (LOM)"),
79 
80 	PVID(0x8086, E1000_DEV_ID_82545EM_COPPER, "Intel(R) Legacy PRO/1000 MT 82545EM (Copper)"),
81 	PVID(0x8086, E1000_DEV_ID_82545EM_FIBER, "Intel(R) Legacy PRO/1000 MF 82545EM (Fiber)"),
82 	PVID(0x8086, E1000_DEV_ID_82545GM_COPPER, "Intel(R) Legacy PRO/1000 MT 82545GM (Copper)"),
83 	PVID(0x8086, E1000_DEV_ID_82545GM_FIBER, "Intel(R) Legacy PRO/1000 MF 82545GM (Fiber)"),
84 	PVID(0x8086, E1000_DEV_ID_82545GM_SERDES, "Intel(R) Legacy PRO/1000 MB 82545GM (SERDES)"),
85 
86 	PVID(0x8086, E1000_DEV_ID_82546EB_COPPER, "Intel(R) Legacy PRO/1000 MT 82546EB (Copper)"),
87 	PVID(0x8086, E1000_DEV_ID_82546EB_FIBER, "Intel(R) Legacy PRO/1000 MF 82546EB (Fiber)"),
88 	PVID(0x8086, E1000_DEV_ID_82546EB_QUAD_COPPER, "Intel(R) Legacy PRO/1000 MT 82546EB (Quad Copper"),
89 	PVID(0x8086, E1000_DEV_ID_82546GB_COPPER, "Intel(R) Legacy PRO/1000 MT 82546GB (Copper)"),
90 	PVID(0x8086, E1000_DEV_ID_82546GB_FIBER, "Intel(R) Legacy PRO/1000 MF 82546GB (Fiber)"),
91 	PVID(0x8086, E1000_DEV_ID_82546GB_SERDES, "Intel(R) Legacy PRO/1000 MB 82546GB (SERDES)"),
92 	PVID(0x8086, E1000_DEV_ID_82546GB_PCIE, "Intel(R) Legacy PRO/1000 P 82546GB (PCIe)"),
93 	PVID(0x8086, E1000_DEV_ID_82546GB_QUAD_COPPER, "Intel(R) Legacy PRO/1000 GT 82546GB (Quad Copper)"),
94 	PVID(0x8086, E1000_DEV_ID_82546GB_QUAD_COPPER_KSP3, "Intel(R) Legacy PRO/1000 GT 82546GB (Quad Copper)"),
95 
96 	PVID(0x8086, E1000_DEV_ID_82547EI, "Intel(R) Legacy PRO/1000 CT 82547EI"),
97 	PVID(0x8086, E1000_DEV_ID_82547EI_MOBILE, "Intel(R) Legacy PRO/1000 CT 82547EI (Mobile)"),
98 	PVID(0x8086, E1000_DEV_ID_82547GI, "Intel(R) Legacy PRO/1000 CT 82547GI"),
99 
100 	/* Intel(R) - em-class devices */
101 	PVID(0x8086, E1000_DEV_ID_82571EB_COPPER, "Intel(R) PRO/1000 PT 82571EB/82571GB (Copper)"),
102 	PVID(0x8086, E1000_DEV_ID_82571EB_FIBER, "Intel(R) PRO/1000 PF 82571EB/82571GB (Fiber)"),
103 	PVID(0x8086, E1000_DEV_ID_82571EB_SERDES, "Intel(R) PRO/1000 PB 82571EB (SERDES)"),
104 	PVID(0x8086, E1000_DEV_ID_82571EB_SERDES_DUAL, "Intel(R) PRO/1000 82571EB (Dual Mezzanine)"),
105 	PVID(0x8086, E1000_DEV_ID_82571EB_SERDES_QUAD, "Intel(R) PRO/1000 82571EB (Quad Mezzanine)"),
106 	PVID(0x8086, E1000_DEV_ID_82571EB_QUAD_COPPER, "Intel(R) PRO/1000 PT 82571EB/82571GB (Quad Copper)"),
107 	PVID(0x8086, E1000_DEV_ID_82571EB_QUAD_COPPER_LP, "Intel(R) PRO/1000 PT 82571EB/82571GB (Quad Copper)"),
108 	PVID(0x8086, E1000_DEV_ID_82571EB_QUAD_FIBER, "Intel(R) PRO/1000 PF 82571EB (Quad Fiber)"),
109 	PVID(0x8086, E1000_DEV_ID_82571PT_QUAD_COPPER, "Intel(R) PRO/1000 PT 82571PT (Quad Copper)"),
110 	PVID(0x8086, E1000_DEV_ID_82572EI, "Intel(R) PRO/1000 PT 82572EI (Copper)"),
111 	PVID(0x8086, E1000_DEV_ID_82572EI_COPPER, "Intel(R) PRO/1000 PT 82572EI (Copper)"),
112 	PVID(0x8086, E1000_DEV_ID_82572EI_FIBER, "Intel(R) PRO/1000 PF 82572EI (Fiber)"),
113 	PVID(0x8086, E1000_DEV_ID_82572EI_SERDES, "Intel(R) PRO/1000 82572EI (SERDES)"),
114 	PVID(0x8086, E1000_DEV_ID_82573E, "Intel(R) PRO/1000 82573E (Copper)"),
115 	PVID(0x8086, E1000_DEV_ID_82573E_IAMT, "Intel(R) PRO/1000 82573E AMT (Copper)"),
116 	PVID(0x8086, E1000_DEV_ID_82573L, "Intel(R) PRO/1000 82573L"),
117 	PVID(0x8086, E1000_DEV_ID_82583V, "Intel(R) 82583V"),
118 	PVID(0x8086, E1000_DEV_ID_80003ES2LAN_COPPER_SPT, "Intel(R) 80003ES2LAN (Copper)"),
119 	PVID(0x8086, E1000_DEV_ID_80003ES2LAN_SERDES_SPT, "Intel(R) 80003ES2LAN (SERDES)"),
120 	PVID(0x8086, E1000_DEV_ID_80003ES2LAN_COPPER_DPT, "Intel(R) 80003ES2LAN (Dual Copper)"),
121 	PVID(0x8086, E1000_DEV_ID_80003ES2LAN_SERDES_DPT, "Intel(R) 80003ES2LAN (Dual SERDES)"),
122 	PVID(0x8086, E1000_DEV_ID_ICH8_IGP_M_AMT, "Intel(R) 82566MM ICH8 AMT (Mobile)"),
123 	PVID(0x8086, E1000_DEV_ID_ICH8_IGP_AMT, "Intel(R) 82566DM ICH8 AMT"),
124 	PVID(0x8086, E1000_DEV_ID_ICH8_IGP_C, "Intel(R) 82566DC ICH8"),
125 	PVID(0x8086, E1000_DEV_ID_ICH8_IFE, "Intel(R) 82562V ICH8"),
126 	PVID(0x8086, E1000_DEV_ID_ICH8_IFE_GT, "Intel(R) 82562GT ICH8"),
127 	PVID(0x8086, E1000_DEV_ID_ICH8_IFE_G, "Intel(R) 82562G ICH8"),
128 	PVID(0x8086, E1000_DEV_ID_ICH8_IGP_M, "Intel(R) 82566MC ICH8"),
129 	PVID(0x8086, E1000_DEV_ID_ICH8_82567V_3, "Intel(R) 82567V-3 ICH8"),
130 	PVID(0x8086, E1000_DEV_ID_ICH9_IGP_M_AMT, "Intel(R) 82567LM ICH9 AMT"),
131 	PVID(0x8086, E1000_DEV_ID_ICH9_IGP_AMT, "Intel(R) 82566DM-2 ICH9 AMT"),
132 	PVID(0x8086, E1000_DEV_ID_ICH9_IGP_C, "Intel(R) 82566DC-2 ICH9"),
133 	PVID(0x8086, E1000_DEV_ID_ICH9_IGP_M, "Intel(R) 82567LF ICH9"),
134 	PVID(0x8086, E1000_DEV_ID_ICH9_IGP_M_V, "Intel(R) 82567V ICH9"),
135 	PVID(0x8086, E1000_DEV_ID_ICH9_IFE, "Intel(R) 82562V-2 ICH9"),
136 	PVID(0x8086, E1000_DEV_ID_ICH9_IFE_GT, "Intel(R) 82562GT-2 ICH9"),
137 	PVID(0x8086, E1000_DEV_ID_ICH9_IFE_G, "Intel(R) 82562G-2 ICH9"),
138 	PVID(0x8086, E1000_DEV_ID_ICH9_BM, "Intel(R) 82567LM-4 ICH9"),
139 	PVID(0x8086, E1000_DEV_ID_82574L, "Intel(R) Gigabit CT 82574L"),
140 	PVID(0x8086, E1000_DEV_ID_82574LA, "Intel(R) 82574L-Apple"),
141 	PVID(0x8086, E1000_DEV_ID_ICH10_R_BM_LM, "Intel(R) 82567LM-2 ICH10"),
142 	PVID(0x8086, E1000_DEV_ID_ICH10_R_BM_LF, "Intel(R) 82567LF-2 ICH10"),
143 	PVID(0x8086, E1000_DEV_ID_ICH10_R_BM_V, "Intel(R) 82567V-2 ICH10"),
144 	PVID(0x8086, E1000_DEV_ID_ICH10_D_BM_LM, "Intel(R) 82567LM-3 ICH10"),
145 	PVID(0x8086, E1000_DEV_ID_ICH10_D_BM_LF, "Intel(R) 82567LF-3 ICH10"),
146 	PVID(0x8086, E1000_DEV_ID_ICH10_D_BM_V, "Intel(R) 82567V-4 ICH10"),
147 	PVID(0x8086, E1000_DEV_ID_PCH_M_HV_LM, "Intel(R) 82577LM"),
148 	PVID(0x8086, E1000_DEV_ID_PCH_M_HV_LC, "Intel(R) 82577LC"),
149 	PVID(0x8086, E1000_DEV_ID_PCH_D_HV_DM, "Intel(R) 82578DM"),
150 	PVID(0x8086, E1000_DEV_ID_PCH_D_HV_DC, "Intel(R) 82578DC"),
151 	PVID(0x8086, E1000_DEV_ID_PCH2_LV_LM, "Intel(R) 82579LM"),
152 	PVID(0x8086, E1000_DEV_ID_PCH2_LV_V, "Intel(R) 82579V"),
153 	PVID(0x8086, E1000_DEV_ID_PCH_LPT_I217_LM, "Intel(R) I217-LM LPT"),
154 	PVID(0x8086, E1000_DEV_ID_PCH_LPT_I217_V, "Intel(R) I217-V LPT"),
155 	PVID(0x8086, E1000_DEV_ID_PCH_LPTLP_I218_LM, "Intel(R) I218-LM LPTLP"),
156 	PVID(0x8086, E1000_DEV_ID_PCH_LPTLP_I218_V, "Intel(R) I218-V LPTLP"),
157 	PVID(0x8086, E1000_DEV_ID_PCH_I218_LM2, "Intel(R) I218-LM (2)"),
158 	PVID(0x8086, E1000_DEV_ID_PCH_I218_V2, "Intel(R) I218-V (2)"),
159 	PVID(0x8086, E1000_DEV_ID_PCH_I218_LM3, "Intel(R) I218-LM (3)"),
160 	PVID(0x8086, E1000_DEV_ID_PCH_I218_V3, "Intel(R) I218-V (3)"),
161 	PVID(0x8086, E1000_DEV_ID_PCH_SPT_I219_LM, "Intel(R) I219-LM SPT"),
162 	PVID(0x8086, E1000_DEV_ID_PCH_SPT_I219_V, "Intel(R) I219-V SPT"),
163 	PVID(0x8086, E1000_DEV_ID_PCH_SPT_I219_LM2, "Intel(R) I219-LM SPT-H(2)"),
164 	PVID(0x8086, E1000_DEV_ID_PCH_SPT_I219_V2, "Intel(R) I219-V SPT-H(2)"),
165 	PVID(0x8086, E1000_DEV_ID_PCH_LBG_I219_LM3, "Intel(R) I219-LM LBG(3)"),
166 	PVID(0x8086, E1000_DEV_ID_PCH_SPT_I219_LM4, "Intel(R) I219-LM SPT(4)"),
167 	PVID(0x8086, E1000_DEV_ID_PCH_SPT_I219_V4, "Intel(R) I219-V SPT(4)"),
168 	PVID(0x8086, E1000_DEV_ID_PCH_SPT_I219_LM5, "Intel(R) I219-LM SPT(5)"),
169 	PVID(0x8086, E1000_DEV_ID_PCH_SPT_I219_V5, "Intel(R) I219-V SPT(5)"),
170 	PVID(0x8086, E1000_DEV_ID_PCH_CNP_I219_LM6, "Intel(R) I219-LM CNP(6)"),
171 	PVID(0x8086, E1000_DEV_ID_PCH_CNP_I219_V6, "Intel(R) I219-V CNP(6)"),
172 	PVID(0x8086, E1000_DEV_ID_PCH_CNP_I219_LM7, "Intel(R) I219-LM CNP(7)"),
173 	PVID(0x8086, E1000_DEV_ID_PCH_CNP_I219_V7, "Intel(R) I219-V CNP(7)"),
174 	PVID(0x8086, E1000_DEV_ID_PCH_ICP_I219_LM8, "Intel(R) I219-LM ICP(8)"),
175 	PVID(0x8086, E1000_DEV_ID_PCH_ICP_I219_V8, "Intel(R) I219-V ICP(8)"),
176 	PVID(0x8086, E1000_DEV_ID_PCH_ICP_I219_LM9, "Intel(R) I219-LM ICP(9)"),
177 	PVID(0x8086, E1000_DEV_ID_PCH_ICP_I219_V9, "Intel(R) I219-V ICP(9)"),
178 	PVID(0x8086, E1000_DEV_ID_PCH_CMP_I219_LM10, "Intel(R) I219-LM CMP(10)"),
179 	PVID(0x8086, E1000_DEV_ID_PCH_CMP_I219_V10, "Intel(R) I219-V CMP(10)"),
180 	PVID(0x8086, E1000_DEV_ID_PCH_CMP_I219_LM11, "Intel(R) I219-LM CMP(11)"),
181 	PVID(0x8086, E1000_DEV_ID_PCH_CMP_I219_V11, "Intel(R) I219-V CMP(11)"),
182 	PVID(0x8086, E1000_DEV_ID_PCH_CMP_I219_LM12, "Intel(R) I219-LM CMP(12)"),
183 	PVID(0x8086, E1000_DEV_ID_PCH_CMP_I219_V12, "Intel(R) I219-V CMP(12)"),
184 	PVID(0x8086, E1000_DEV_ID_PCH_TGP_I219_LM13, "Intel(R) I219-LM TGP(13)"),
185 	PVID(0x8086, E1000_DEV_ID_PCH_TGP_I219_V13, "Intel(R) I219-V TGP(13)"),
186 	PVID(0x8086, E1000_DEV_ID_PCH_TGP_I219_LM14, "Intel(R) I219-LM TGP(14)"),
187 	PVID(0x8086, E1000_DEV_ID_PCH_TGP_I219_V14, "Intel(R) I219-V GTP(14)"),
188 	PVID(0x8086, E1000_DEV_ID_PCH_TGP_I219_LM15, "Intel(R) I219-LM TGP(15)"),
189 	PVID(0x8086, E1000_DEV_ID_PCH_TGP_I219_V15, "Intel(R) I219-V TGP(15)"),
190 	PVID(0x8086, E1000_DEV_ID_PCH_ADL_I219_LM16, "Intel(R) I219-LM ADL(16)"),
191 	PVID(0x8086, E1000_DEV_ID_PCH_ADL_I219_V16, "Intel(R) I219-V ADL(16)"),
192 	PVID(0x8086, E1000_DEV_ID_PCH_ADL_I219_LM17, "Intel(R) I219-LM ADL(17)"),
193 	PVID(0x8086, E1000_DEV_ID_PCH_ADL_I219_V17, "Intel(R) I219-V ADL(17)"),
194 	PVID(0x8086, E1000_DEV_ID_PCH_MTP_I219_LM18, "Intel(R) I219-LM MTP(18)"),
195 	PVID(0x8086, E1000_DEV_ID_PCH_MTP_I219_V18, "Intel(R) I219-V MTP(18)"),
196 	PVID(0x8086, E1000_DEV_ID_PCH_ADL_I219_LM19, "Intel(R) I219-LM ADL(19)"),
197 	PVID(0x8086, E1000_DEV_ID_PCH_ADL_I219_V19, "Intel(R) I219-V ADL(19)"),
198 	PVID(0x8086, E1000_DEV_ID_PCH_LNL_I219_LM20, "Intel(R) I219-LM LNL(20)"),
199 	PVID(0x8086, E1000_DEV_ID_PCH_LNL_I219_V20, "Intel(R) I219-V LNL(20)"),
200 	PVID(0x8086, E1000_DEV_ID_PCH_LNL_I219_LM21, "Intel(R) I219-LM LNL(21)"),
201 	PVID(0x8086, E1000_DEV_ID_PCH_LNL_I219_V21, "Intel(R) I219-V LNL(21)"),
202 	PVID(0x8086, E1000_DEV_ID_PCH_RPL_I219_LM22, "Intel(R) I219-LM RPL(22)"),
203 	PVID(0x8086, E1000_DEV_ID_PCH_RPL_I219_V22, "Intel(R) I219-V RPL(22)"),
204 	PVID(0x8086, E1000_DEV_ID_PCH_RPL_I219_LM23, "Intel(R) I219-LM RPL(23)"),
205 	PVID(0x8086, E1000_DEV_ID_PCH_RPL_I219_V23, "Intel(R) I219-V RPL(23)"),
206 	PVID(0x8086, E1000_DEV_ID_PCH_ARL_I219_LM24, "Intel(R) I219-LM ARL(24)"),
207 	PVID(0x8086, E1000_DEV_ID_PCH_ARL_I219_V24, "Intel(R) I219-V ARL(24)"),
208 	PVID(0x8086, E1000_DEV_ID_PCH_PTP_I219_LM25, "Intel(R) I219-LM PTP(25)"),
209 	PVID(0x8086, E1000_DEV_ID_PCH_PTP_I219_V25, "Intel(R) I219-V PTP(25)"),
210 	PVID(0x8086, E1000_DEV_ID_PCH_PTP_I219_LM26, "Intel(R) I219-LM PTP(26)"),
211 	PVID(0x8086, E1000_DEV_ID_PCH_PTP_I219_V26, "Intel(R) I219-V PTP(26)"),
212 	PVID(0x8086, E1000_DEV_ID_PCH_PTP_I219_LM27, "Intel(R) I219-LM PTP(27)"),
213 	PVID(0x8086, E1000_DEV_ID_PCH_PTP_I219_V27, "Intel(R) I219-V PTP(27)"),
214 	/* required last entry */
215 	PVID_END
216 };
217 
218 static const pci_vendor_info_t igb_vendor_info_array[] =
219 {
220 	/* Intel(R) - igb-class devices */
221 	PVID(0x8086, E1000_DEV_ID_82575EB_COPPER, "Intel(R) PRO/1000 82575EB (Copper)"),
222 	PVID(0x8086, E1000_DEV_ID_82575EB_FIBER_SERDES, "Intel(R) PRO/1000 82575EB (SERDES)"),
223 	PVID(0x8086, E1000_DEV_ID_82575GB_QUAD_COPPER, "Intel(R) PRO/1000 VT 82575GB (Quad Copper)"),
224 	PVID(0x8086, E1000_DEV_ID_82576, "Intel(R) PRO/1000 82576"),
225 	PVID(0x8086, E1000_DEV_ID_82576_NS, "Intel(R) PRO/1000 82576NS"),
226 	PVID(0x8086, E1000_DEV_ID_82576_NS_SERDES, "Intel(R) PRO/1000 82576NS (SERDES)"),
227 	PVID(0x8086, E1000_DEV_ID_82576_FIBER, "Intel(R) PRO/1000 EF 82576 (Dual Fiber)"),
228 	PVID(0x8086, E1000_DEV_ID_82576_SERDES, "Intel(R) PRO/1000 82576 (Dual SERDES)"),
229 	PVID(0x8086, E1000_DEV_ID_82576_SERDES_QUAD, "Intel(R) PRO/1000 ET 82576 (Quad SERDES)"),
230 	PVID(0x8086, E1000_DEV_ID_82576_QUAD_COPPER, "Intel(R) PRO/1000 ET 82576 (Quad Copper)"),
231 	PVID(0x8086, E1000_DEV_ID_82576_QUAD_COPPER_ET2, "Intel(R) PRO/1000 ET(2) 82576 (Quad Copper)"),
232 	PVID(0x8086, E1000_DEV_ID_82576_VF, "Intel(R) PRO/1000 82576 Virtual Function"),
233 	PVID(0x8086, E1000_DEV_ID_82580_COPPER, "Intel(R) I340 82580 (Copper)"),
234 	PVID(0x8086, E1000_DEV_ID_82580_FIBER, "Intel(R) I340 82580 (Fiber)"),
235 	PVID(0x8086, E1000_DEV_ID_82580_SERDES, "Intel(R) I340 82580 (SERDES)"),
236 	PVID(0x8086, E1000_DEV_ID_82580_SGMII, "Intel(R) I340 82580 (SGMII)"),
237 	PVID(0x8086, E1000_DEV_ID_82580_COPPER_DUAL, "Intel(R) I340-T2 82580 (Dual Copper)"),
238 	PVID(0x8086, E1000_DEV_ID_82580_QUAD_FIBER, "Intel(R) I340-F4 82580 (Quad Fiber)"),
239 	PVID(0x8086, E1000_DEV_ID_DH89XXCC_SERDES, "Intel(R) DH89XXCC (SERDES)"),
240 	PVID(0x8086, E1000_DEV_ID_DH89XXCC_SGMII, "Intel(R) I347-AT4 DH89XXCC"),
241 	PVID(0x8086, E1000_DEV_ID_DH89XXCC_SFP, "Intel(R) DH89XXCC (SFP)"),
242 	PVID(0x8086, E1000_DEV_ID_DH89XXCC_BACKPLANE, "Intel(R) DH89XXCC (Backplane)"),
243 	PVID(0x8086, E1000_DEV_ID_I350_COPPER, "Intel(R) I350 (Copper)"),
244 	PVID(0x8086, E1000_DEV_ID_I350_FIBER, "Intel(R) I350 (Fiber)"),
245 	PVID(0x8086, E1000_DEV_ID_I350_SERDES, "Intel(R) I350 (SERDES)"),
246 	PVID(0x8086, E1000_DEV_ID_I350_SGMII, "Intel(R) I350 (SGMII)"),
247 	PVID(0x8086, E1000_DEV_ID_I350_VF, "Intel(R) I350 Virtual Function"),
248 	PVID(0x8086, E1000_DEV_ID_I210_COPPER, "Intel(R) I210 (Copper)"),
249 	PVID(0x8086, E1000_DEV_ID_I210_COPPER_IT, "Intel(R) I210 IT (Copper)"),
250 	PVID(0x8086, E1000_DEV_ID_I210_COPPER_OEM1, "Intel(R) I210 (OEM)"),
251 	PVID(0x8086, E1000_DEV_ID_I210_COPPER_FLASHLESS, "Intel(R) I210 Flashless (Copper)"),
252 	PVID(0x8086, E1000_DEV_ID_I210_SERDES_FLASHLESS, "Intel(R) I210 Flashless (SERDES)"),
253 	PVID(0x8086, E1000_DEV_ID_I210_SGMII_FLASHLESS, "Intel(R) I210 Flashless (SGMII)"),
254 	PVID(0x8086, E1000_DEV_ID_I210_FIBER, "Intel(R) I210 (Fiber)"),
255 	PVID(0x8086, E1000_DEV_ID_I210_SERDES, "Intel(R) I210 (SERDES)"),
256 	PVID(0x8086, E1000_DEV_ID_I210_SGMII, "Intel(R) I210 (SGMII)"),
257 	PVID(0x8086, E1000_DEV_ID_I211_COPPER, "Intel(R) I211 (Copper)"),
258 	PVID(0x8086, E1000_DEV_ID_I354_BACKPLANE_1GBPS, "Intel(R) I354 (1.0 GbE Backplane)"),
259 	PVID(0x8086, E1000_DEV_ID_I354_BACKPLANE_2_5GBPS, "Intel(R) I354 (2.5 GbE Backplane)"),
260 	PVID(0x8086, E1000_DEV_ID_I354_SGMII, "Intel(R) I354 (SGMII)"),
261 	/* required last entry */
262 	PVID_END
263 };
264 
265 /*********************************************************************
266  *  Function prototypes
267  *********************************************************************/
268 static void	*em_register(device_t);
269 static void	*igb_register(device_t);
270 static int	em_if_attach_pre(if_ctx_t);
271 static int	em_if_attach_post(if_ctx_t);
272 static int	em_if_detach(if_ctx_t);
273 static int	em_if_shutdown(if_ctx_t);
274 static int	em_if_suspend(if_ctx_t);
275 static int	em_if_resume(if_ctx_t);
276 
277 static int	em_if_tx_queues_alloc(if_ctx_t, caddr_t *, uint64_t *, int, int);
278 static int	em_if_rx_queues_alloc(if_ctx_t, caddr_t *, uint64_t *, int, int);
279 static void	em_if_queues_free(if_ctx_t);
280 
281 static uint64_t	em_if_get_counter(if_ctx_t, ift_counter);
282 static void	em_if_init(if_ctx_t);
283 static void	em_if_stop(if_ctx_t);
284 static void	em_if_media_status(if_ctx_t, struct ifmediareq *);
285 static int	em_if_media_change(if_ctx_t);
286 static int	em_if_mtu_set(if_ctx_t, uint32_t);
287 static void	em_if_timer(if_ctx_t, uint16_t);
288 static void	em_if_vlan_register(if_ctx_t, u16);
289 static void	em_if_vlan_unregister(if_ctx_t, u16);
290 static void	em_if_watchdog_reset(if_ctx_t);
291 static bool	em_if_needs_restart(if_ctx_t, enum iflib_restart_event);
292 
293 static void	em_identify_hardware(if_ctx_t);
294 static int	em_allocate_pci_resources(if_ctx_t);
295 static void	em_free_pci_resources(if_ctx_t);
296 static void	em_reset(if_ctx_t);
297 static int	em_setup_interface(if_ctx_t);
298 static int	em_setup_msix(if_ctx_t);
299 
300 static void	em_initialize_transmit_unit(if_ctx_t);
301 static void	em_initialize_receive_unit(if_ctx_t);
302 
303 static void	em_if_intr_enable(if_ctx_t);
304 static void	em_if_intr_disable(if_ctx_t);
305 static void	igb_if_intr_enable(if_ctx_t);
306 static void	igb_if_intr_disable(if_ctx_t);
307 static int	em_if_rx_queue_intr_enable(if_ctx_t, uint16_t);
308 static int	em_if_tx_queue_intr_enable(if_ctx_t, uint16_t);
309 static int	igb_if_rx_queue_intr_enable(if_ctx_t, uint16_t);
310 static int	igb_if_tx_queue_intr_enable(if_ctx_t, uint16_t);
311 static void	em_if_multi_set(if_ctx_t);
312 static void	em_if_update_admin_status(if_ctx_t);
313 static void	em_if_debug(if_ctx_t);
314 static void	em_update_stats_counters(struct e1000_softc *);
315 static void	em_add_hw_stats(struct e1000_softc *);
316 static int	em_if_set_promisc(if_ctx_t, int);
317 static bool	em_if_vlan_filter_capable(if_ctx_t);
318 static bool	em_if_vlan_filter_used(if_ctx_t);
319 static void	em_if_vlan_filter_enable(struct e1000_softc *);
320 static void	em_if_vlan_filter_disable(struct e1000_softc *);
321 static void	em_if_vlan_filter_write(struct e1000_softc *);
322 static void	em_setup_vlan_hw_support(if_ctx_t ctx);
323 static int	em_sysctl_nvm_info(SYSCTL_HANDLER_ARGS);
324 static void	em_print_nvm_info(struct e1000_softc *);
325 static void	em_fw_version_locked(if_ctx_t);
326 static void	em_sbuf_fw_version(struct e1000_fw_version *, struct sbuf *);
327 static void	em_print_fw_version(struct e1000_softc *);
328 static int	em_sysctl_print_fw_version(SYSCTL_HANDLER_ARGS);
329 static int	em_sysctl_debug_info(SYSCTL_HANDLER_ARGS);
330 static int	em_get_rs(SYSCTL_HANDLER_ARGS);
331 static void	em_print_debug_info(struct e1000_softc *);
332 static int 	em_is_valid_ether_addr(u8 *);
333 static void	em_newitr(struct e1000_softc *, struct em_rx_queue *,
334     struct tx_ring *, struct rx_ring *);
335 static bool	em_automask_tso(if_ctx_t);
336 static int	em_sysctl_tso_tcp_flags_mask(SYSCTL_HANDLER_ARGS);
337 static int	em_sysctl_int_delay(SYSCTL_HANDLER_ARGS);
338 static void	em_add_int_delay_sysctl(struct e1000_softc *, const char *,
339     const char *, struct em_int_delay_info *, int, int);
340 /* Management and WOL Support */
341 static void	em_init_manageability(struct e1000_softc *);
342 static void	em_release_manageability(struct e1000_softc *);
343 static void	em_get_hw_control(struct e1000_softc *);
344 static void	em_release_hw_control(struct e1000_softc *);
345 static void	em_get_wakeup(if_ctx_t);
346 static void	em_enable_wakeup(if_ctx_t);
347 static int	em_enable_phy_wakeup(struct e1000_softc *);
348 static void	em_disable_aspm(struct e1000_softc *);
349 
350 int		em_intr(void *);
351 
352 /* MSI-X handlers */
353 static int	em_if_msix_intr_assign(if_ctx_t, int);
354 static int	em_msix_link(void *);
355 static void	em_handle_link(void *);
356 
357 static void	em_enable_vectors_82574(if_ctx_t);
358 
359 static int	em_set_flowcntl(SYSCTL_HANDLER_ARGS);
360 static int	em_sysctl_eee(SYSCTL_HANDLER_ARGS);
361 static int	igb_sysctl_dmac(SYSCTL_HANDLER_ARGS);
362 static void	em_if_led_func(if_ctx_t, int);
363 
364 static int	em_get_regs(SYSCTL_HANDLER_ARGS);
365 
366 static void	lem_smartspeed(struct e1000_softc *);
367 static void	igb_configure_queues(struct e1000_softc *);
368 static void	em_flush_desc_rings(struct e1000_softc *);
369 
370 
371 /*********************************************************************
372  *  FreeBSD Device Interface Entry Points
373  *********************************************************************/
374 static device_method_t em_methods[] = {
375 	/* Device interface */
376 	DEVMETHOD(device_register, em_register),
377 	DEVMETHOD(device_probe, iflib_device_probe),
378 	DEVMETHOD(device_attach, iflib_device_attach),
379 	DEVMETHOD(device_detach, iflib_device_detach),
380 	DEVMETHOD(device_shutdown, iflib_device_shutdown),
381 	DEVMETHOD(device_suspend, iflib_device_suspend),
382 	DEVMETHOD(device_resume, iflib_device_resume),
383 	DEVMETHOD_END
384 };
385 
386 static device_method_t igb_methods[] = {
387 	/* Device interface */
388 	DEVMETHOD(device_register, igb_register),
389 	DEVMETHOD(device_probe, iflib_device_probe),
390 	DEVMETHOD(device_attach, iflib_device_attach),
391 	DEVMETHOD(device_detach, iflib_device_detach),
392 	DEVMETHOD(device_shutdown, iflib_device_shutdown),
393 	DEVMETHOD(device_suspend, iflib_device_suspend),
394 	DEVMETHOD(device_resume, iflib_device_resume),
395 	DEVMETHOD_END
396 };
397 
398 
399 static driver_t em_driver = {
400 	"em", em_methods, sizeof(struct e1000_softc),
401 };
402 
403 DRIVER_MODULE(em, pci, em_driver, 0, 0);
404 
405 MODULE_DEPEND(em, pci, 1, 1, 1);
406 MODULE_DEPEND(em, ether, 1, 1, 1);
407 MODULE_DEPEND(em, iflib, 1, 1, 1);
408 
409 IFLIB_PNP_INFO(pci, em, em_vendor_info_array);
410 
411 static driver_t igb_driver = {
412 	"igb", igb_methods, sizeof(struct e1000_softc),
413 };
414 
415 DRIVER_MODULE(igb, pci, igb_driver, 0, 0);
416 
417 MODULE_DEPEND(igb, pci, 1, 1, 1);
418 MODULE_DEPEND(igb, ether, 1, 1, 1);
419 MODULE_DEPEND(igb, iflib, 1, 1, 1);
420 
421 IFLIB_PNP_INFO(pci, igb, igb_vendor_info_array);
422 
423 static device_method_t em_if_methods[] = {
424 	DEVMETHOD(ifdi_attach_pre, em_if_attach_pre),
425 	DEVMETHOD(ifdi_attach_post, em_if_attach_post),
426 	DEVMETHOD(ifdi_detach, em_if_detach),
427 	DEVMETHOD(ifdi_shutdown, em_if_shutdown),
428 	DEVMETHOD(ifdi_suspend, em_if_suspend),
429 	DEVMETHOD(ifdi_resume, em_if_resume),
430 	DEVMETHOD(ifdi_init, em_if_init),
431 	DEVMETHOD(ifdi_stop, em_if_stop),
432 	DEVMETHOD(ifdi_msix_intr_assign, em_if_msix_intr_assign),
433 	DEVMETHOD(ifdi_intr_enable, em_if_intr_enable),
434 	DEVMETHOD(ifdi_intr_disable, em_if_intr_disable),
435 	DEVMETHOD(ifdi_tx_queues_alloc, em_if_tx_queues_alloc),
436 	DEVMETHOD(ifdi_rx_queues_alloc, em_if_rx_queues_alloc),
437 	DEVMETHOD(ifdi_queues_free, em_if_queues_free),
438 	DEVMETHOD(ifdi_update_admin_status, em_if_update_admin_status),
439 	DEVMETHOD(ifdi_multi_set, em_if_multi_set),
440 	DEVMETHOD(ifdi_media_status, em_if_media_status),
441 	DEVMETHOD(ifdi_media_change, em_if_media_change),
442 	DEVMETHOD(ifdi_mtu_set, em_if_mtu_set),
443 	DEVMETHOD(ifdi_promisc_set, em_if_set_promisc),
444 	DEVMETHOD(ifdi_timer, em_if_timer),
445 	DEVMETHOD(ifdi_watchdog_reset, em_if_watchdog_reset),
446 	DEVMETHOD(ifdi_vlan_register, em_if_vlan_register),
447 	DEVMETHOD(ifdi_vlan_unregister, em_if_vlan_unregister),
448 	DEVMETHOD(ifdi_get_counter, em_if_get_counter),
449 	DEVMETHOD(ifdi_led_func, em_if_led_func),
450 	DEVMETHOD(ifdi_rx_queue_intr_enable, em_if_rx_queue_intr_enable),
451 	DEVMETHOD(ifdi_tx_queue_intr_enable, em_if_tx_queue_intr_enable),
452 	DEVMETHOD(ifdi_debug, em_if_debug),
453 	DEVMETHOD(ifdi_needs_restart, em_if_needs_restart),
454 	DEVMETHOD_END
455 };
456 
457 static driver_t em_if_driver = {
458 	"em_if", em_if_methods, sizeof(struct e1000_softc)
459 };
460 
461 static device_method_t igb_if_methods[] = {
462 	DEVMETHOD(ifdi_attach_pre, em_if_attach_pre),
463 	DEVMETHOD(ifdi_attach_post, em_if_attach_post),
464 	DEVMETHOD(ifdi_detach, em_if_detach),
465 	DEVMETHOD(ifdi_shutdown, em_if_shutdown),
466 	DEVMETHOD(ifdi_suspend, em_if_suspend),
467 	DEVMETHOD(ifdi_resume, em_if_resume),
468 	DEVMETHOD(ifdi_init, em_if_init),
469 	DEVMETHOD(ifdi_stop, em_if_stop),
470 	DEVMETHOD(ifdi_msix_intr_assign, em_if_msix_intr_assign),
471 	DEVMETHOD(ifdi_intr_enable, igb_if_intr_enable),
472 	DEVMETHOD(ifdi_intr_disable, igb_if_intr_disable),
473 	DEVMETHOD(ifdi_tx_queues_alloc, em_if_tx_queues_alloc),
474 	DEVMETHOD(ifdi_rx_queues_alloc, em_if_rx_queues_alloc),
475 	DEVMETHOD(ifdi_queues_free, em_if_queues_free),
476 	DEVMETHOD(ifdi_update_admin_status, em_if_update_admin_status),
477 	DEVMETHOD(ifdi_multi_set, em_if_multi_set),
478 	DEVMETHOD(ifdi_media_status, em_if_media_status),
479 	DEVMETHOD(ifdi_media_change, em_if_media_change),
480 	DEVMETHOD(ifdi_mtu_set, em_if_mtu_set),
481 	DEVMETHOD(ifdi_promisc_set, em_if_set_promisc),
482 	DEVMETHOD(ifdi_timer, em_if_timer),
483 	DEVMETHOD(ifdi_watchdog_reset, em_if_watchdog_reset),
484 	DEVMETHOD(ifdi_vlan_register, em_if_vlan_register),
485 	DEVMETHOD(ifdi_vlan_unregister, em_if_vlan_unregister),
486 	DEVMETHOD(ifdi_get_counter, em_if_get_counter),
487 	DEVMETHOD(ifdi_led_func, em_if_led_func),
488 	DEVMETHOD(ifdi_rx_queue_intr_enable, igb_if_rx_queue_intr_enable),
489 	DEVMETHOD(ifdi_tx_queue_intr_enable, igb_if_tx_queue_intr_enable),
490 	DEVMETHOD(ifdi_debug, em_if_debug),
491 	DEVMETHOD(ifdi_needs_restart, em_if_needs_restart),
492 	DEVMETHOD_END
493 };
494 
495 static driver_t igb_if_driver = {
496 	"igb_if", igb_if_methods, sizeof(struct e1000_softc)
497 };
498 
499 /*********************************************************************
500  *  Tunable default values.
501  *********************************************************************/
502 
503 #define EM_TICKS_TO_USECS(ticks)	((1024 * (ticks) + 500) / 1000)
504 #define EM_USECS_TO_TICKS(usecs)	((1000 * (usecs) + 512) / 1024)
505 
506 /* Allow common code without TSO */
507 #ifndef CSUM_TSO
508 #define CSUM_TSO	0
509 #endif
510 
511 static SYSCTL_NODE(_hw, OID_AUTO, em, CTLFLAG_RD | CTLFLAG_MPSAFE, 0,
512     "EM driver parameters");
513 
514 static int em_disable_crc_stripping = 0;
515 SYSCTL_INT(_hw_em, OID_AUTO, disable_crc_stripping, CTLFLAG_RDTUN,
516     &em_disable_crc_stripping, 0, "Disable CRC Stripping");
517 
518 static int em_tx_int_delay_dflt = EM_TICKS_TO_USECS(EM_TIDV);
519 static int em_rx_int_delay_dflt = EM_TICKS_TO_USECS(EM_RDTR);
520 SYSCTL_INT(_hw_em, OID_AUTO, tx_int_delay, CTLFLAG_RDTUN, &em_tx_int_delay_dflt,
521     0, "Default transmit interrupt delay in usecs");
522 SYSCTL_INT(_hw_em, OID_AUTO, rx_int_delay, CTLFLAG_RDTUN, &em_rx_int_delay_dflt,
523     0, "Default receive interrupt delay in usecs");
524 
525 static int em_tx_abs_int_delay_dflt = EM_TICKS_TO_USECS(EM_TADV);
526 static int em_rx_abs_int_delay_dflt = EM_TICKS_TO_USECS(EM_RADV);
527 SYSCTL_INT(_hw_em, OID_AUTO, tx_abs_int_delay, CTLFLAG_RDTUN,
528     &em_tx_abs_int_delay_dflt, 0,
529     "Default transmit interrupt delay limit in usecs");
530 SYSCTL_INT(_hw_em, OID_AUTO, rx_abs_int_delay, CTLFLAG_RDTUN,
531     &em_rx_abs_int_delay_dflt, 0,
532     "Default receive interrupt delay limit in usecs");
533 
534 static int em_smart_pwr_down = false;
535 SYSCTL_INT(_hw_em, OID_AUTO, smart_pwr_down, CTLFLAG_RDTUN, &em_smart_pwr_down,
536     0, "Set to true to leave smart power down enabled on newer adapters");
537 
538 static bool em_unsupported_tso = false;
539 SYSCTL_BOOL(_hw_em, OID_AUTO, unsupported_tso, CTLFLAG_RDTUN,
540     &em_unsupported_tso, 0, "Allow unsupported em(4) TSO configurations");
541 
542 /* Controls whether promiscuous also shows bad packets */
543 static int em_debug_sbp = false;
544 SYSCTL_INT(_hw_em, OID_AUTO, sbp, CTLFLAG_RDTUN, &em_debug_sbp, 0,
545     "Show bad packets in promiscuous mode");
546 
547 /* Energy efficient ethernet - default to OFF */
548 static int eee_setting = 1;
549 SYSCTL_INT(_hw_em, OID_AUTO, eee_setting, CTLFLAG_RDTUN, &eee_setting, 0,
550     "Enable Energy Efficient Ethernet");
551 
552 /*
553  * AIM: Adaptive Interrupt Moderation
554  * which means that the interrupt rate is varied over time based on the
555  * traffic for that interrupt vector
556  */
557 static int em_enable_aim = 1;
558 SYSCTL_INT(_hw_em, OID_AUTO, enable_aim, CTLFLAG_RWTUN, &em_enable_aim,
559     0, "Enable adaptive interrupt moderation (1=normal, 2=lowlatency)");
560 
561 /*
562 ** Tuneable Interrupt rate
563 */
564 static int em_max_interrupt_rate = EM_INTS_DEFAULT;
565 SYSCTL_INT(_hw_em, OID_AUTO, max_interrupt_rate, CTLFLAG_RDTUN,
566     &em_max_interrupt_rate, 0, "Maximum interrupts per second");
567 
568 /* Global used in WOL setup with multiport cards */
569 static int global_quad_port_a = 0;
570 
571 extern struct if_txrx igb_txrx;
572 extern struct if_txrx em_txrx;
573 extern struct if_txrx lem_txrx;
574 
575 static struct if_shared_ctx em_sctx_init = {
576 	.isc_magic = IFLIB_MAGIC,
577 	.isc_q_align = PAGE_SIZE,
578 	.isc_tx_maxsize = EM_TSO_SIZE + sizeof(struct ether_vlan_header),
579 	.isc_tx_maxsegsize = PAGE_SIZE,
580 	.isc_tso_maxsize = EM_TSO_SIZE + sizeof(struct ether_vlan_header),
581 	.isc_tso_maxsegsize = EM_TSO_SEG_SIZE,
582 	.isc_rx_maxsize = MJUM9BYTES,
583 	.isc_rx_nsegments = 1,
584 	.isc_rx_maxsegsize = MJUM9BYTES,
585 	.isc_nfl = 1,
586 	.isc_nrxqs = 1,
587 	.isc_ntxqs = 1,
588 	.isc_admin_intrcnt = 1,
589 	.isc_vendor_info = em_vendor_info_array,
590 	.isc_driver_version = em_driver_version,
591 	.isc_driver = &em_if_driver,
592 	.isc_flags = IFLIB_NEED_SCRATCH | IFLIB_TSO_INIT_IP | IFLIB_NEED_ZERO_CSUM,
593 
594 	.isc_nrxd_min = {EM_MIN_RXD},
595 	.isc_ntxd_min = {EM_MIN_TXD},
596 	.isc_nrxd_max = {EM_MAX_RXD},
597 	.isc_ntxd_max = {EM_MAX_TXD},
598 	.isc_nrxd_default = {EM_DEFAULT_RXD},
599 	.isc_ntxd_default = {EM_DEFAULT_TXD},
600 };
601 
602 static struct if_shared_ctx igb_sctx_init = {
603 	.isc_magic = IFLIB_MAGIC,
604 	.isc_q_align = PAGE_SIZE,
605 	.isc_tx_maxsize = EM_TSO_SIZE + sizeof(struct ether_vlan_header),
606 	.isc_tx_maxsegsize = PAGE_SIZE,
607 	.isc_tso_maxsize = EM_TSO_SIZE + sizeof(struct ether_vlan_header),
608 	.isc_tso_maxsegsize = EM_TSO_SEG_SIZE,
609 	.isc_rx_maxsize = MJUM9BYTES,
610 	.isc_rx_nsegments = 1,
611 	.isc_rx_maxsegsize = MJUM9BYTES,
612 	.isc_nfl = 1,
613 	.isc_nrxqs = 1,
614 	.isc_ntxqs = 1,
615 	.isc_admin_intrcnt = 1,
616 	.isc_vendor_info = igb_vendor_info_array,
617 	.isc_driver_version = igb_driver_version,
618 	.isc_driver = &igb_if_driver,
619 	.isc_flags = IFLIB_NEED_SCRATCH | IFLIB_TSO_INIT_IP | IFLIB_NEED_ZERO_CSUM,
620 
621 	.isc_nrxd_min = {EM_MIN_RXD},
622 	.isc_ntxd_min = {EM_MIN_TXD},
623 	.isc_nrxd_max = {IGB_MAX_RXD},
624 	.isc_ntxd_max = {IGB_MAX_TXD},
625 	.isc_nrxd_default = {EM_DEFAULT_RXD},
626 	.isc_ntxd_default = {EM_DEFAULT_TXD},
627 };
628 
629 /*****************************************************************
630  *
631  * Dump Registers
632  *
633  ****************************************************************/
634 #define IGB_REGS_LEN 739
635 
636 static int em_get_regs(SYSCTL_HANDLER_ARGS)
637 {
638 	struct e1000_softc *sc = (struct e1000_softc *)arg1;
639 	struct e1000_hw *hw = &sc->hw;
640 	struct sbuf *sb;
641 	u32 *regs_buff;
642 	int rc;
643 
644 	regs_buff = malloc(sizeof(u32) * IGB_REGS_LEN, M_DEVBUF, M_WAITOK);
645 	memset(regs_buff, 0, IGB_REGS_LEN * sizeof(u32));
646 
647 	rc = sysctl_wire_old_buffer(req, 0);
648 	MPASS(rc == 0);
649 	if (rc != 0) {
650 		free(regs_buff, M_DEVBUF);
651 		return (rc);
652 	}
653 
654 	sb = sbuf_new_for_sysctl(NULL, NULL, 32*400, req);
655 	MPASS(sb != NULL);
656 	if (sb == NULL) {
657 		free(regs_buff, M_DEVBUF);
658 		return (ENOMEM);
659 	}
660 
661 	/* General Registers */
662 	regs_buff[0] = E1000_READ_REG(hw, E1000_CTRL);
663 	regs_buff[1] = E1000_READ_REG(hw, E1000_STATUS);
664 	regs_buff[2] = E1000_READ_REG(hw, E1000_CTRL_EXT);
665 	regs_buff[3] = E1000_READ_REG(hw, E1000_ICR);
666 	regs_buff[4] = E1000_READ_REG(hw, E1000_RCTL);
667 	regs_buff[5] = E1000_READ_REG(hw, E1000_RDLEN(0));
668 	regs_buff[6] = E1000_READ_REG(hw, E1000_RDH(0));
669 	regs_buff[7] = E1000_READ_REG(hw, E1000_RDT(0));
670 	regs_buff[8] = E1000_READ_REG(hw, E1000_RXDCTL(0));
671 	regs_buff[9] = E1000_READ_REG(hw, E1000_RDBAL(0));
672 	regs_buff[10] = E1000_READ_REG(hw, E1000_RDBAH(0));
673 	regs_buff[11] = E1000_READ_REG(hw, E1000_TCTL);
674 	regs_buff[12] = E1000_READ_REG(hw, E1000_TDBAL(0));
675 	regs_buff[13] = E1000_READ_REG(hw, E1000_TDBAH(0));
676 	regs_buff[14] = E1000_READ_REG(hw, E1000_TDLEN(0));
677 	regs_buff[15] = E1000_READ_REG(hw, E1000_TDH(0));
678 	regs_buff[16] = E1000_READ_REG(hw, E1000_TDT(0));
679 	regs_buff[17] = E1000_READ_REG(hw, E1000_TXDCTL(0));
680 	regs_buff[18] = E1000_READ_REG(hw, E1000_TDFH);
681 	regs_buff[19] = E1000_READ_REG(hw, E1000_TDFT);
682 	regs_buff[20] = E1000_READ_REG(hw, E1000_TDFHS);
683 	regs_buff[21] = E1000_READ_REG(hw, E1000_TDFPC);
684 
685 	sbuf_printf(sb, "General Registers\n");
686 	sbuf_printf(sb, "\tCTRL\t %08x\n", regs_buff[0]);
687 	sbuf_printf(sb, "\tSTATUS\t %08x\n", regs_buff[1]);
688 	sbuf_printf(sb, "\tCTRL_EXT\t %08x\n\n", regs_buff[2]);
689 
690 	sbuf_printf(sb, "Interrupt Registers\n");
691 	sbuf_printf(sb, "\tICR\t %08x\n\n", regs_buff[3]);
692 
693 	sbuf_printf(sb, "RX Registers\n");
694 	sbuf_printf(sb, "\tRCTL\t %08x\n", regs_buff[4]);
695 	sbuf_printf(sb, "\tRDLEN\t %08x\n", regs_buff[5]);
696 	sbuf_printf(sb, "\tRDH\t %08x\n", regs_buff[6]);
697 	sbuf_printf(sb, "\tRDT\t %08x\n", regs_buff[7]);
698 	sbuf_printf(sb, "\tRXDCTL\t %08x\n", regs_buff[8]);
699 	sbuf_printf(sb, "\tRDBAL\t %08x\n", regs_buff[9]);
700 	sbuf_printf(sb, "\tRDBAH\t %08x\n\n", regs_buff[10]);
701 
702 	sbuf_printf(sb, "TX Registers\n");
703 	sbuf_printf(sb, "\tTCTL\t %08x\n", regs_buff[11]);
704 	sbuf_printf(sb, "\tTDBAL\t %08x\n", regs_buff[12]);
705 	sbuf_printf(sb, "\tTDBAH\t %08x\n", regs_buff[13]);
706 	sbuf_printf(sb, "\tTDLEN\t %08x\n", regs_buff[14]);
707 	sbuf_printf(sb, "\tTDH\t %08x\n", regs_buff[15]);
708 	sbuf_printf(sb, "\tTDT\t %08x\n", regs_buff[16]);
709 	sbuf_printf(sb, "\tTXDCTL\t %08x\n", regs_buff[17]);
710 	sbuf_printf(sb, "\tTDFH\t %08x\n", regs_buff[18]);
711 	sbuf_printf(sb, "\tTDFT\t %08x\n", regs_buff[19]);
712 	sbuf_printf(sb, "\tTDFHS\t %08x\n", regs_buff[20]);
713 	sbuf_printf(sb, "\tTDFPC\t %08x\n\n", regs_buff[21]);
714 
715 	free(regs_buff, M_DEVBUF);
716 
717 #ifdef DUMP_DESCS
718 	{
719 		if_softc_ctx_t scctx = sc->shared;
720 		struct rx_ring *rxr = &rx_que->rxr;
721 		struct tx_ring *txr = &tx_que->txr;
722 		int ntxd = scctx->isc_ntxd[0];
723 		int nrxd = scctx->isc_nrxd[0];
724 		int j;
725 
726 	for (j = 0; j < nrxd; j++) {
727 		u32 staterr = le32toh(rxr->rx_base[j].wb.upper.status_error);
728 		u32 length =  le32toh(rxr->rx_base[j].wb.upper.length);
729 		sbuf_printf(sb, "\tReceive Descriptor Address %d: %08" PRIx64 "  Error:%d  Length:%d\n", j, rxr->rx_base[j].read.buffer_addr, staterr, length);
730 	}
731 
732 	for (j = 0; j < min(ntxd, 256); j++) {
733 		unsigned int *ptr = (unsigned int *)&txr->tx_base[j];
734 
735 		sbuf_printf(sb, "\tTXD[%03d] [0]: %08x [1]: %08x [2]: %08x [3]: %08x  eop: %d DD=%d\n",
736 			    j, ptr[0], ptr[1], ptr[2], ptr[3], buf->eop,
737 			    buf->eop != -1 ? txr->tx_base[buf->eop].upper.fields.status & E1000_TXD_STAT_DD : 0);
738 
739 	}
740 	}
741 #endif
742 
743 	rc = sbuf_finish(sb);
744 	sbuf_delete(sb);
745 	return(rc);
746 }
747 
748 static void *
749 em_register(device_t dev)
750 {
751 	return (&em_sctx_init);
752 }
753 
754 static void *
755 igb_register(device_t dev)
756 {
757 	return (&igb_sctx_init);
758 }
759 
760 static int
761 em_set_num_queues(if_ctx_t ctx)
762 {
763 	struct e1000_softc *sc = iflib_get_softc(ctx);
764 	int maxqueues;
765 
766 	/* Sanity check based on HW */
767 	switch (sc->hw.mac.type) {
768 	case e1000_82576:
769 	case e1000_82580:
770 	case e1000_i350:
771 	case e1000_i354:
772 		maxqueues = 8;
773 		break;
774 	case e1000_i210:
775 	case e1000_82575:
776 		maxqueues = 4;
777 		break;
778 	case e1000_i211:
779 	case e1000_82574:
780 		maxqueues = 2;
781 		break;
782 	default:
783 		maxqueues = 1;
784 		break;
785 	}
786 
787 	return (maxqueues);
788 }
789 
790 #define LEM_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 
795 #define EM_CAPS \
796     IFCAP_HWCSUM | IFCAP_VLAN_MTU | IFCAP_VLAN_HWTAGGING | \
797     IFCAP_VLAN_HWCSUM | IFCAP_WOL | IFCAP_VLAN_HWFILTER | IFCAP_TSO4 | \
798     IFCAP_LRO | IFCAP_VLAN_HWTSO | IFCAP_JUMBO_MTU | IFCAP_HWCSUM_IPV6 | \
799     IFCAP_TSO6
800 
801 #define IGB_CAPS \
802     IFCAP_HWCSUM | IFCAP_VLAN_MTU | IFCAP_VLAN_HWTAGGING | \
803     IFCAP_VLAN_HWCSUM | IFCAP_WOL | IFCAP_VLAN_HWFILTER | IFCAP_TSO4 | \
804     IFCAP_LRO | IFCAP_VLAN_HWTSO | IFCAP_JUMBO_MTU | IFCAP_HWCSUM_IPV6 | \
805     IFCAP_TSO6
806 
807 /*********************************************************************
808  *  Device initialization routine
809  *
810  *  The attach entry point is called when the driver is being loaded.
811  *  This routine identifies the type of hardware, allocates all resources
812  *  and initializes the hardware.
813  *
814  *  return 0 on success, positive on failure
815  *********************************************************************/
816 static int
817 em_if_attach_pre(if_ctx_t ctx)
818 {
819 	struct e1000_softc *sc;
820 	if_softc_ctx_t scctx;
821 	device_t dev;
822 	struct e1000_hw *hw;
823 	struct sysctl_oid_list *child;
824 	struct sysctl_ctx_list *ctx_list;
825 	int error = 0;
826 
827 	INIT_DEBUGOUT("em_if_attach_pre: begin");
828 	dev = iflib_get_dev(ctx);
829 	sc = iflib_get_softc(ctx);
830 
831 	sc->ctx = sc->osdep.ctx = ctx;
832 	sc->dev = sc->osdep.dev = dev;
833 	scctx = sc->shared = iflib_get_softc_ctx(ctx);
834 	sc->media = iflib_get_media(ctx);
835 	hw = &sc->hw;
836 
837 	/* Determine hardware and mac info */
838 	em_identify_hardware(ctx);
839 
840 	/* SYSCTL stuff */
841 	ctx_list = device_get_sysctl_ctx(dev);
842 	child = SYSCTL_CHILDREN(device_get_sysctl_tree(dev));
843 
844 	SYSCTL_ADD_PROC(ctx_list, child, OID_AUTO, "nvm",
845 	    CTLTYPE_INT | CTLFLAG_RW, sc, 0,
846 	    em_sysctl_nvm_info, "I", "NVM Information");
847 
848 	sc->enable_aim = em_enable_aim;
849 	SYSCTL_ADD_INT(ctx_list, child, OID_AUTO, "enable_aim",
850 	    CTLFLAG_RW, &sc->enable_aim, 0,
851 		"Interrupt Moderation (1=normal, 2=lowlatency)");
852 
853 	SYSCTL_ADD_PROC(ctx_list, child, OID_AUTO, "fw_version",
854 	    CTLTYPE_STRING | CTLFLAG_RD, sc, 0,
855 	    em_sysctl_print_fw_version, "A",
856 	    "Prints FW/NVM Versions");
857 
858 	SYSCTL_ADD_PROC(ctx_list, child, OID_AUTO, "debug",
859 	    CTLTYPE_INT | CTLFLAG_RW, sc, 0,
860 	    em_sysctl_debug_info, "I", "Debug Information");
861 
862 	SYSCTL_ADD_PROC(ctx_list, child, OID_AUTO, "fc",
863 	    CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_NEEDGIANT, sc, 0,
864 	    em_set_flowcntl, "I", "Flow Control");
865 
866 	SYSCTL_ADD_PROC(ctx_list, child, OID_AUTO, "reg_dump",
867 	    CTLTYPE_STRING | CTLFLAG_RD | CTLFLAG_NEEDGIANT, sc, 0,
868 	    em_get_regs, "A", "Dump Registers");
869 
870 	SYSCTL_ADD_PROC(ctx_list, child, OID_AUTO, "rs_dump",
871 	    CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_NEEDGIANT, sc, 0,
872 	    em_get_rs, "I", "Dump RS indexes");
873 
874 	if (hw->mac.type >= e1000_i350) {
875 		SYSCTL_ADD_PROC(ctx_list, child, OID_AUTO, "dmac",
876 		    CTLTYPE_INT | CTLFLAG_RW, sc, 0,
877 		    igb_sysctl_dmac, "I", "DMA Coalesce");
878 	}
879 
880 	SYSCTL_ADD_PROC(ctx_list, child, OID_AUTO,
881 	    "tso_tcp_flags_mask_first_segment",
882 	    CTLTYPE_UINT | CTLFLAG_RW | CTLFLAG_NEEDGIANT,
883 	    sc, 0, em_sysctl_tso_tcp_flags_mask, "IU",
884 	    "TSO TCP flags mask for first segment");
885 
886 	SYSCTL_ADD_PROC(ctx_list, child, OID_AUTO,
887 	    "tso_tcp_flags_mask_middle_segment",
888 	    CTLTYPE_UINT | CTLFLAG_RW | CTLFLAG_NEEDGIANT,
889 	    sc, 1, em_sysctl_tso_tcp_flags_mask, "IU",
890 	    "TSO TCP flags mask for middle segment");
891 
892 	SYSCTL_ADD_PROC(ctx_list, child, OID_AUTO,
893 	    "tso_tcp_flags_mask_last_segment",
894 	    CTLTYPE_UINT | CTLFLAG_RW | CTLFLAG_NEEDGIANT,
895 	    sc, 2, em_sysctl_tso_tcp_flags_mask, "IU",
896 	    "TSO TCP flags mask for last segment");
897 
898 	scctx->isc_tx_nsegments = EM_MAX_SCATTER;
899 	scctx->isc_nrxqsets_max = scctx->isc_ntxqsets_max = em_set_num_queues(ctx);
900 	if (bootverbose)
901 		device_printf(dev, "attach_pre capping queues at %d\n",
902 		    scctx->isc_ntxqsets_max);
903 
904 	if (hw->mac.type >= igb_mac_min) {
905 		scctx->isc_txqsizes[0] = roundup2(scctx->isc_ntxd[0] * sizeof(union e1000_adv_tx_desc), EM_DBA_ALIGN);
906 		scctx->isc_rxqsizes[0] = roundup2(scctx->isc_nrxd[0] * sizeof(union e1000_adv_rx_desc), EM_DBA_ALIGN);
907 		scctx->isc_txd_size[0] = sizeof(union e1000_adv_tx_desc);
908 		scctx->isc_rxd_size[0] = sizeof(union e1000_adv_rx_desc);
909 		scctx->isc_txrx = &igb_txrx;
910 		scctx->isc_tx_tso_segments_max = EM_MAX_SCATTER;
911 		scctx->isc_tx_tso_size_max = EM_TSO_SIZE;
912 		scctx->isc_tx_tso_segsize_max = EM_TSO_SEG_SIZE;
913 		scctx->isc_capabilities = scctx->isc_capenable = IGB_CAPS;
914 		scctx->isc_tx_csum_flags = CSUM_TCP | CSUM_UDP | CSUM_TSO |
915 		     CSUM_IP6_TCP | CSUM_IP6_UDP;
916 		if (hw->mac.type != e1000_82575)
917 			scctx->isc_tx_csum_flags |= CSUM_SCTP | CSUM_IP6_SCTP;
918 		/*
919 		** Some new devices, as with ixgbe, now may
920 		** use a different BAR, so we need to keep
921 		** track of which is used.
922 		*/
923 		scctx->isc_msix_bar = pci_msix_table_bar(dev);
924 	} else if (hw->mac.type >= em_mac_min) {
925 		scctx->isc_txqsizes[0] = roundup2(scctx->isc_ntxd[0]* sizeof(struct e1000_tx_desc), EM_DBA_ALIGN);
926 		scctx->isc_rxqsizes[0] = roundup2(scctx->isc_nrxd[0] * sizeof(union e1000_rx_desc_extended), EM_DBA_ALIGN);
927 		scctx->isc_txd_size[0] = sizeof(struct e1000_tx_desc);
928 		scctx->isc_rxd_size[0] = sizeof(union e1000_rx_desc_extended);
929 		scctx->isc_txrx = &em_txrx;
930 		scctx->isc_tx_tso_segments_max = EM_MAX_SCATTER;
931 		scctx->isc_tx_tso_size_max = EM_TSO_SIZE;
932 		scctx->isc_tx_tso_segsize_max = EM_TSO_SEG_SIZE;
933 		scctx->isc_capabilities = scctx->isc_capenable = EM_CAPS;
934 		scctx->isc_tx_csum_flags = CSUM_TCP | CSUM_UDP | CSUM_IP_TSO |
935 		    CSUM_IP6_TCP | CSUM_IP6_UDP;
936 
937 		/* Disable TSO on all em(4) until ring stalls can be debugged */
938 		scctx->isc_capenable &= ~IFCAP_TSO;
939 
940 		/*
941 		 * Disable TSO on SPT due to errata that downclocks DMA performance
942 		 * i218-i219 Specification Update 1.5.4.5
943 		 */
944 		if (hw->mac.type == e1000_pch_spt)
945 			scctx->isc_capenable &= ~IFCAP_TSO;
946 
947 		/*
948 		 * We support MSI-X with 82574 only, but indicate to iflib(4)
949 		 * that it shall give MSI at least a try with other devices.
950 		 */
951 		if (hw->mac.type == e1000_82574) {
952 			scctx->isc_msix_bar = pci_msix_table_bar(dev);
953 		} else {
954 			scctx->isc_msix_bar = -1;
955 			scctx->isc_disable_msix = 1;
956 		}
957 	} else {
958 		scctx->isc_txqsizes[0] = roundup2((scctx->isc_ntxd[0] + 1) * sizeof(struct e1000_tx_desc), EM_DBA_ALIGN);
959 		scctx->isc_rxqsizes[0] = roundup2((scctx->isc_nrxd[0] + 1) * sizeof(struct e1000_rx_desc), EM_DBA_ALIGN);
960 		scctx->isc_txd_size[0] = sizeof(struct e1000_tx_desc);
961 		scctx->isc_rxd_size[0] = sizeof(struct e1000_rx_desc);
962 		scctx->isc_txrx = &lem_txrx;
963 		scctx->isc_tx_tso_segments_max = EM_MAX_SCATTER;
964 		scctx->isc_tx_tso_size_max = EM_TSO_SIZE;
965 		scctx->isc_tx_tso_segsize_max = EM_TSO_SEG_SIZE;
966 		scctx->isc_capabilities = scctx->isc_capenable = LEM_CAPS;
967 		if (em_unsupported_tso)
968 			scctx->isc_capabilities |= IFCAP_TSO6;
969 		scctx->isc_tx_csum_flags = CSUM_TCP | CSUM_UDP | CSUM_IP_TSO |
970 		    CSUM_IP6_TCP | CSUM_IP6_UDP;
971 
972 		/* Disable TSO on all lem(4) until ring stalls can be debugged */
973 		scctx->isc_capenable &= ~IFCAP_TSO;
974 
975 		/* 82541ER doesn't do HW tagging */
976 		if (hw->device_id == E1000_DEV_ID_82541ER ||
977 		    hw->device_id == E1000_DEV_ID_82541ER_LOM) {
978 			scctx->isc_capabilities &= ~IFCAP_VLAN_HWTAGGING;
979 			scctx->isc_capenable = scctx->isc_capabilities;
980 		}
981 		/* This is the first e1000 chip and it does not do offloads */
982 		if (hw->mac.type == e1000_82542) {
983 			scctx->isc_capabilities &= ~(IFCAP_HWCSUM | IFCAP_VLAN_HWCSUM |
984 			    IFCAP_HWCSUM_IPV6 | IFCAP_VLAN_HWTAGGING |
985 			    IFCAP_VLAN_HWFILTER | IFCAP_TSO | IFCAP_VLAN_HWTSO);
986 			scctx->isc_capenable = scctx->isc_capabilities;
987 		}
988 		/* These can't do TSO for various reasons */
989 		if (hw->mac.type < e1000_82544 || hw->mac.type == e1000_82547 ||
990 		    hw->mac.type == e1000_82547_rev_2) {
991 			scctx->isc_capabilities &= ~(IFCAP_TSO | IFCAP_VLAN_HWTSO);
992 			scctx->isc_capenable = scctx->isc_capabilities;
993 		}
994 		/* XXXKB: No IPv6 before this? */
995 		if (hw->mac.type < e1000_82545){
996 			scctx->isc_capabilities &= ~IFCAP_HWCSUM_IPV6;
997 			scctx->isc_capenable = scctx->isc_capabilities;
998 		}
999 		/* "PCI/PCI-X SDM 4.0" page 33 (b) - FDX requirement on these chips */
1000 		if (hw->mac.type == e1000_82547 || hw->mac.type == e1000_82547_rev_2)
1001 			scctx->isc_capenable &= ~(IFCAP_HWCSUM | IFCAP_VLAN_HWCSUM |
1002 			    IFCAP_HWCSUM_IPV6);
1003 
1004 		/* INTx only */
1005 		scctx->isc_msix_bar = 0;
1006 	}
1007 
1008 	/* Setup PCI resources */
1009 	if (em_allocate_pci_resources(ctx)) {
1010 		device_printf(dev, "Allocation of PCI resources failed\n");
1011 		error = ENXIO;
1012 		goto err_pci;
1013 	}
1014 
1015 	/*
1016 	** For ICH8 and family we need to
1017 	** map the flash memory, and this
1018 	** must happen after the MAC is
1019 	** identified
1020 	*/
1021 	if ((hw->mac.type == e1000_ich8lan) ||
1022 	    (hw->mac.type == e1000_ich9lan) ||
1023 	    (hw->mac.type == e1000_ich10lan) ||
1024 	    (hw->mac.type == e1000_pchlan) ||
1025 	    (hw->mac.type == e1000_pch2lan) ||
1026 	    (hw->mac.type == e1000_pch_lpt)) {
1027 		int rid = EM_BAR_TYPE_FLASH;
1028 		sc->flash = bus_alloc_resource_any(dev,
1029 		    SYS_RES_MEMORY, &rid, RF_ACTIVE);
1030 		if (sc->flash == NULL) {
1031 			device_printf(dev, "Mapping of Flash failed\n");
1032 			error = ENXIO;
1033 			goto err_pci;
1034 		}
1035 		/* This is used in the shared code */
1036 		hw->flash_address = (u8 *)sc->flash;
1037 		sc->osdep.flash_bus_space_tag =
1038 		    rman_get_bustag(sc->flash);
1039 		sc->osdep.flash_bus_space_handle =
1040 		    rman_get_bushandle(sc->flash);
1041 	}
1042 	/*
1043 	** In the new SPT device flash is not  a
1044 	** separate BAR, rather it is also in BAR0,
1045 	** so use the same tag and an offset handle for the
1046 	** FLASH read/write macros in the shared code.
1047 	*/
1048 	else if (hw->mac.type >= e1000_pch_spt) {
1049 		sc->osdep.flash_bus_space_tag =
1050 		    sc->osdep.mem_bus_space_tag;
1051 		sc->osdep.flash_bus_space_handle =
1052 		    sc->osdep.mem_bus_space_handle
1053 		    + E1000_FLASH_BASE_ADDR;
1054 	}
1055 
1056 	/* Do Shared Code initialization */
1057 	error = e1000_setup_init_funcs(hw, true);
1058 	if (error) {
1059 		device_printf(dev, "Setup of Shared code failed, error %d\n",
1060 		    error);
1061 		error = ENXIO;
1062 		goto err_pci;
1063 	}
1064 
1065 	em_setup_msix(ctx);
1066 	e1000_get_bus_info(hw);
1067 
1068 	/* Set up some sysctls for the tunable interrupt delays */
1069 	if (hw->mac.type < igb_mac_min) {
1070 		em_add_int_delay_sysctl(sc, "rx_int_delay",
1071 		    "receive interrupt delay in usecs", &sc->rx_int_delay,
1072 		    E1000_REGISTER(hw, E1000_RDTR), em_rx_int_delay_dflt);
1073 		em_add_int_delay_sysctl(sc, "tx_int_delay",
1074 		    "transmit interrupt delay in usecs", &sc->tx_int_delay,
1075 		    E1000_REGISTER(hw, E1000_TIDV), em_tx_int_delay_dflt);
1076 	}
1077 	if (hw->mac.type >= e1000_82540 && hw->mac.type < igb_mac_min) {
1078 		em_add_int_delay_sysctl(sc, "rx_abs_int_delay",
1079 		    "receive interrupt delay limit in usecs", &sc->rx_abs_int_delay,
1080 		    E1000_REGISTER(hw, E1000_RADV), em_rx_abs_int_delay_dflt);
1081 		em_add_int_delay_sysctl(sc, "tx_abs_int_delay",
1082 		    "transmit interrupt delay limit in usecs", &sc->tx_abs_int_delay,
1083 		    E1000_REGISTER(hw, E1000_TADV), em_tx_abs_int_delay_dflt);
1084 		em_add_int_delay_sysctl(sc, "itr",
1085 		    "interrupt delay limit in usecs/4", &sc->tx_itr,
1086 		    E1000_REGISTER(hw, E1000_ITR),
1087 			EM_INTS_TO_ITR(em_max_interrupt_rate));
1088 	}
1089 
1090 	hw->mac.autoneg = DO_AUTO_NEG;
1091 	hw->phy.autoneg_wait_to_complete = false;
1092 	hw->phy.autoneg_advertised = AUTONEG_ADV_DEFAULT;
1093 
1094 	if (hw->mac.type < em_mac_min) {
1095 		e1000_init_script_state_82541(hw, true);
1096 		e1000_set_tbi_compatibility_82543(hw, true);
1097 	}
1098 	/* Copper options */
1099 	if (hw->phy.media_type == e1000_media_type_copper) {
1100 		hw->phy.mdix = AUTO_ALL_MODES;
1101 		hw->phy.disable_polarity_correction = false;
1102 		hw->phy.ms_type = EM_MASTER_SLAVE;
1103 	}
1104 
1105 	/*
1106 	 * Set the frame limits assuming
1107 	 * standard ethernet sized frames.
1108 	 */
1109 	scctx->isc_max_frame_size = hw->mac.max_frame_size =
1110 	    ETHERMTU + ETHER_HDR_LEN + ETHERNET_FCS_SIZE;
1111 
1112 	/*
1113 	 * This controls when hardware reports transmit completion
1114 	 * status.
1115 	 */
1116 	hw->mac.report_tx_early = 1;
1117 
1118 	/* Allocate multicast array memory. */
1119 	sc->mta = malloc(sizeof(u8) * ETHER_ADDR_LEN *
1120 	    MAX_NUM_MULTICAST_ADDRESSES, M_DEVBUF, M_NOWAIT);
1121 	if (sc->mta == NULL) {
1122 		device_printf(dev, "Can not allocate multicast setup array\n");
1123 		error = ENOMEM;
1124 		goto err_late;
1125 	}
1126 
1127 	/* Clear the IFCAP_TSO auto mask */
1128 	sc->tso_automasked = 0;
1129 
1130 	/* Check SOL/IDER usage */
1131 	if (e1000_check_reset_block(hw))
1132 		device_printf(dev, "PHY reset is blocked"
1133 			      " due to SOL/IDER session.\n");
1134 
1135 	/* Sysctl for setting Energy Efficient Ethernet */
1136 	if (hw->mac.type < igb_mac_min)
1137 		hw->dev_spec.ich8lan.eee_disable = eee_setting;
1138 	else
1139 		hw->dev_spec._82575.eee_disable = eee_setting;
1140 	SYSCTL_ADD_PROC(ctx_list, child, OID_AUTO, "eee_control",
1141 	    CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_NEEDGIANT, sc, 0,
1142 	    em_sysctl_eee, "I", "Disable Energy Efficient Ethernet");
1143 
1144 	/*
1145 	** Start from a known state, this is
1146 	** important in reading the nvm and
1147 	** mac from that.
1148 	*/
1149 	e1000_reset_hw(hw);
1150 
1151 	/* Make sure we have a good EEPROM before we read from it */
1152 	if (e1000_validate_nvm_checksum(hw) < 0) {
1153 		/*
1154 		** Some PCI-E parts fail the first check due to
1155 		** the link being in sleep state, call it again,
1156 		** if it fails a second time its a real issue.
1157 		*/
1158 		if (e1000_validate_nvm_checksum(hw) < 0) {
1159 			device_printf(dev,
1160 			    "The EEPROM Checksum Is Not Valid\n");
1161 			error = EIO;
1162 			goto err_late;
1163 		}
1164 	}
1165 
1166 	/* Copy the permanent MAC address out of the EEPROM */
1167 	if (e1000_read_mac_addr(hw) < 0) {
1168 		device_printf(dev, "EEPROM read error while reading MAC"
1169 			      " address\n");
1170 		error = EIO;
1171 		goto err_late;
1172 	}
1173 
1174 	if (!em_is_valid_ether_addr(hw->mac.addr)) {
1175 		if (sc->vf_ifp) {
1176 			ether_gen_addr(iflib_get_ifp(ctx),
1177 			    (struct ether_addr *)hw->mac.addr);
1178 		} else {
1179 			device_printf(dev, "Invalid MAC address\n");
1180 			error = EIO;
1181 			goto err_late;
1182 		}
1183 	}
1184 
1185 	/* Save the EEPROM/NVM versions, must be done under IFLIB_CTX_LOCK */
1186 	em_fw_version_locked(ctx);
1187 
1188 	em_print_fw_version(sc);
1189 
1190 	/*
1191 	 * Get Wake-on-Lan and Management info for later use
1192 	 */
1193 	em_get_wakeup(ctx);
1194 
1195 	/* Enable only WOL MAGIC by default */
1196 	scctx->isc_capenable &= ~IFCAP_WOL;
1197 	if (sc->wol != 0)
1198 		scctx->isc_capenable |= IFCAP_WOL_MAGIC;
1199 
1200 	iflib_set_mac(ctx, hw->mac.addr);
1201 
1202 	return (0);
1203 
1204 err_late:
1205 	em_release_hw_control(sc);
1206 err_pci:
1207 	em_free_pci_resources(ctx);
1208 	free(sc->mta, M_DEVBUF);
1209 
1210 	return (error);
1211 }
1212 
1213 static int
1214 em_if_attach_post(if_ctx_t ctx)
1215 {
1216 	struct e1000_softc *sc = iflib_get_softc(ctx);
1217 	struct e1000_hw *hw = &sc->hw;
1218 	int error = 0;
1219 
1220 	/* Setup OS specific network interface */
1221 	error = em_setup_interface(ctx);
1222 	if (error != 0) {
1223 		device_printf(sc->dev, "Interface setup failed: %d\n", error);
1224 		goto err_late;
1225 	}
1226 
1227 	em_reset(ctx);
1228 
1229 	/* Initialize statistics */
1230 	em_update_stats_counters(sc);
1231 	hw->mac.get_link_status = 1;
1232 	em_if_update_admin_status(ctx);
1233 	em_add_hw_stats(sc);
1234 
1235 	/* Non-AMT based hardware can now take control from firmware */
1236 	if (sc->has_manage && !sc->has_amt)
1237 		em_get_hw_control(sc);
1238 
1239 	INIT_DEBUGOUT("em_if_attach_post: end");
1240 
1241 	return (0);
1242 
1243 err_late:
1244 	/* upon attach_post() error, iflib calls _if_detach() to free resources. */
1245 	return (error);
1246 }
1247 
1248 /*********************************************************************
1249  *  Device removal routine
1250  *
1251  *  The detach entry point is called when the driver is being removed.
1252  *  This routine stops the adapter and deallocates all the resources
1253  *  that were allocated for driver operation.
1254  *
1255  *  return 0 on success, positive on failure
1256  *********************************************************************/
1257 static int
1258 em_if_detach(if_ctx_t ctx)
1259 {
1260 	struct e1000_softc	*sc = iflib_get_softc(ctx);
1261 
1262 	INIT_DEBUGOUT("em_if_detach: begin");
1263 
1264 	e1000_phy_hw_reset(&sc->hw);
1265 
1266 	em_release_manageability(sc);
1267 	em_release_hw_control(sc);
1268 	em_free_pci_resources(ctx);
1269 	free(sc->mta, M_DEVBUF);
1270 	sc->mta = NULL;
1271 
1272 	return (0);
1273 }
1274 
1275 /*********************************************************************
1276  *
1277  *  Shutdown entry point
1278  *
1279  **********************************************************************/
1280 
1281 static int
1282 em_if_shutdown(if_ctx_t ctx)
1283 {
1284 	return em_if_suspend(ctx);
1285 }
1286 
1287 /*
1288  * Suspend/resume device methods.
1289  */
1290 static int
1291 em_if_suspend(if_ctx_t ctx)
1292 {
1293 	struct e1000_softc *sc = iflib_get_softc(ctx);
1294 
1295 	em_release_manageability(sc);
1296 	em_release_hw_control(sc);
1297 	em_enable_wakeup(ctx);
1298 	return (0);
1299 }
1300 
1301 static int
1302 em_if_resume(if_ctx_t ctx)
1303 {
1304 	struct e1000_softc *sc = iflib_get_softc(ctx);
1305 
1306 	if (sc->hw.mac.type == e1000_pch2lan)
1307 		e1000_resume_workarounds_pchlan(&sc->hw);
1308 	em_if_init(ctx);
1309 	em_init_manageability(sc);
1310 
1311 	return(0);
1312 }
1313 
1314 static int
1315 em_if_mtu_set(if_ctx_t ctx, uint32_t mtu)
1316 {
1317 	int max_frame_size;
1318 	struct e1000_softc *sc = iflib_get_softc(ctx);
1319 	if_softc_ctx_t scctx = iflib_get_softc_ctx(ctx);
1320 
1321 	IOCTL_DEBUGOUT("ioctl rcv'd: SIOCSIFMTU (Set Interface MTU)");
1322 
1323 	switch (sc->hw.mac.type) {
1324 	case e1000_82571:
1325 	case e1000_82572:
1326 	case e1000_ich9lan:
1327 	case e1000_ich10lan:
1328 	case e1000_pch2lan:
1329 	case e1000_pch_lpt:
1330 	case e1000_pch_spt:
1331 	case e1000_pch_cnp:
1332 	case e1000_pch_tgp:
1333 	case e1000_pch_adp:
1334 	case e1000_pch_mtp:
1335 	case e1000_pch_ptp:
1336 	case e1000_82574:
1337 	case e1000_82583:
1338 	case e1000_80003es2lan:
1339 		/* 9K Jumbo Frame size */
1340 		max_frame_size = 9234;
1341 		break;
1342 	case e1000_pchlan:
1343 		max_frame_size = 4096;
1344 		break;
1345 	case e1000_82542:
1346 	case e1000_ich8lan:
1347 		/* Adapters that do not support jumbo frames */
1348 		max_frame_size = ETHER_MAX_LEN;
1349 		break;
1350 	default:
1351 		if (sc->hw.mac.type >= igb_mac_min)
1352 			max_frame_size = 9234;
1353 		else /* lem */
1354 			max_frame_size = MAX_JUMBO_FRAME_SIZE;
1355 	}
1356 	if (mtu > max_frame_size - ETHER_HDR_LEN - ETHER_CRC_LEN) {
1357 		return (EINVAL);
1358 	}
1359 
1360 	scctx->isc_max_frame_size = sc->hw.mac.max_frame_size =
1361 	    mtu + ETHER_HDR_LEN + ETHER_CRC_LEN;
1362 	return (0);
1363 }
1364 
1365 /*********************************************************************
1366  *  Init entry point
1367  *
1368  *  This routine is used in two ways. It is used by the stack as
1369  *  init entry point in network interface structure. It is also used
1370  *  by the driver as a hw/sw initialization routine to get to a
1371  *  consistent state.
1372  *
1373  **********************************************************************/
1374 static void
1375 em_if_init(if_ctx_t ctx)
1376 {
1377 	struct e1000_softc *sc = iflib_get_softc(ctx);
1378 	if_softc_ctx_t scctx = sc->shared;
1379 	if_t ifp = iflib_get_ifp(ctx);
1380 	struct em_tx_queue *tx_que;
1381 	int i;
1382 
1383 	INIT_DEBUGOUT("em_if_init: begin");
1384 
1385 	/* Get the latest mac address, User can use a LAA */
1386 	bcopy(if_getlladdr(ifp), sc->hw.mac.addr,
1387 	    ETHER_ADDR_LEN);
1388 
1389 	/* Put the address into the Receive Address Array */
1390 	e1000_rar_set(&sc->hw, sc->hw.mac.addr, 0);
1391 
1392 	/*
1393 	 * With the 82571 adapter, RAR[0] may be overwritten
1394 	 * when the other port is reset, we make a duplicate
1395 	 * in RAR[14] for that eventuality, this assures
1396 	 * the interface continues to function.
1397 	 */
1398 	if (sc->hw.mac.type == e1000_82571) {
1399 		e1000_set_laa_state_82571(&sc->hw, true);
1400 		e1000_rar_set(&sc->hw, sc->hw.mac.addr,
1401 		    E1000_RAR_ENTRIES - 1);
1402 	}
1403 
1404 	/* Initialize the hardware */
1405 	em_reset(ctx);
1406 	em_if_update_admin_status(ctx);
1407 
1408 	for (i = 0, tx_que = sc->tx_queues; i < sc->tx_num_queues; i++, tx_que++) {
1409 		struct tx_ring *txr = &tx_que->txr;
1410 
1411 		txr->tx_rs_cidx = txr->tx_rs_pidx;
1412 
1413 		/* Initialize the last processed descriptor to be the end of
1414 		 * the ring, rather than the start, so that we avoid an
1415 		 * off-by-one error when calculating how many descriptors are
1416 		 * done in the credits_update function.
1417 		 */
1418 		txr->tx_cidx_processed = scctx->isc_ntxd[0] - 1;
1419 	}
1420 
1421 	/* Setup VLAN support, basic and offload if available */
1422 	E1000_WRITE_REG(&sc->hw, E1000_VET, ETHERTYPE_VLAN);
1423 
1424 	/* Clear bad data from Rx FIFOs */
1425 	if (sc->hw.mac.type >= igb_mac_min)
1426 		e1000_rx_fifo_flush_base(&sc->hw);
1427 
1428 	/* Configure for OS presence */
1429 	em_init_manageability(sc);
1430 
1431 	/* Prepare transmit descriptors and buffers */
1432 	em_initialize_transmit_unit(ctx);
1433 
1434 	/* Setup Multicast table */
1435 	em_if_multi_set(ctx);
1436 
1437 	sc->rx_mbuf_sz = iflib_get_rx_mbuf_sz(ctx);
1438 	em_initialize_receive_unit(ctx);
1439 
1440 	/* Set up VLAN support and filter */
1441 	em_setup_vlan_hw_support(ctx);
1442 
1443 	/* Don't lose promiscuous settings */
1444 	em_if_set_promisc(ctx, if_getflags(ifp));
1445 	e1000_clear_hw_cntrs_base_generic(&sc->hw);
1446 
1447 	/* MSI-X configuration for 82574 */
1448 	if (sc->hw.mac.type == e1000_82574) {
1449 		int tmp = E1000_READ_REG(&sc->hw, E1000_CTRL_EXT);
1450 
1451 		tmp |= E1000_CTRL_EXT_PBA_CLR;
1452 		E1000_WRITE_REG(&sc->hw, E1000_CTRL_EXT, tmp);
1453 		/* Set the IVAR - interrupt vector routing. */
1454 		E1000_WRITE_REG(&sc->hw, E1000_IVAR, sc->ivars);
1455 	} else if (sc->intr_type == IFLIB_INTR_MSIX) /* Set up queue routing */
1456 		igb_configure_queues(sc);
1457 
1458 	/* this clears any pending interrupts */
1459 	E1000_READ_REG(&sc->hw, E1000_ICR);
1460 	E1000_WRITE_REG(&sc->hw, E1000_ICS, E1000_ICS_LSC);
1461 
1462 	/* AMT based hardware can now take control from firmware */
1463 	if (sc->has_manage && sc->has_amt)
1464 		em_get_hw_control(sc);
1465 
1466 	/* Set Energy Efficient Ethernet */
1467 	if (sc->hw.mac.type >= igb_mac_min &&
1468 	    sc->hw.phy.media_type == e1000_media_type_copper) {
1469 		if (sc->hw.mac.type == e1000_i354)
1470 			e1000_set_eee_i354(&sc->hw, true, true);
1471 		else
1472 			e1000_set_eee_i350(&sc->hw, true, true);
1473 	}
1474 }
1475 
1476 enum itr_latency_target {
1477 	itr_latency_disabled = 0,
1478 	itr_latency_lowest = 1,
1479 	itr_latency_low = 2,
1480 	itr_latency_bulk = 3
1481 };
1482 /*********************************************************************
1483  *
1484  *  Helper to calculate next (E)ITR value for AIM
1485  *
1486  *********************************************************************/
1487 static void
1488 em_newitr(struct e1000_softc *sc, struct em_rx_queue *que,
1489     struct tx_ring *txr, struct rx_ring *rxr)
1490 {
1491 	struct e1000_hw *hw = &sc->hw;
1492 	u32 newitr;
1493 	u32 bytes;
1494 	u32 bytes_packets;
1495 	u32 packets;
1496 	u8 nextlatency;
1497 
1498 	/* Idle, do nothing */
1499 	if ((txr->tx_bytes == 0) && (rxr->rx_bytes == 0))
1500 		return;
1501 
1502 	newitr = 0;
1503 
1504 	if (sc->enable_aim) {
1505 		nextlatency = rxr->rx_nextlatency;
1506 
1507 		/* Use half default (4K) ITR if sub-gig */
1508 		if (sc->link_speed != 1000) {
1509 			newitr = EM_INTS_4K;
1510 			goto em_set_next_itr;
1511 		}
1512 		/* Want at least enough packet buffer for two frames to AIM */
1513 		if (sc->shared->isc_max_frame_size * 2 > (sc->pba << 10)) {
1514 			newitr = em_max_interrupt_rate;
1515 			sc->enable_aim = 0;
1516 			goto em_set_next_itr;
1517 		}
1518 
1519 		/* Get the largest values from the associated tx and rx ring */
1520 		if (txr->tx_bytes && txr->tx_packets) {
1521 			bytes = txr->tx_bytes;
1522 			bytes_packets = txr->tx_bytes/txr->tx_packets;
1523 			packets = txr->tx_packets;
1524 		}
1525 		if (rxr->rx_bytes && rxr->rx_packets) {
1526 			bytes = max(bytes, rxr->rx_bytes);
1527 			bytes_packets = max(bytes_packets, rxr->rx_bytes/rxr->rx_packets);
1528 			packets = max(packets, rxr->rx_packets);
1529 		}
1530 
1531 		/* Latency state machine */
1532 		switch (nextlatency) {
1533 		case itr_latency_disabled: /* Bootstrapping */
1534 			nextlatency = itr_latency_low;
1535 			break;
1536 		case itr_latency_lowest: /* 70k ints/s */
1537 			/* TSO and jumbo frames */
1538 			if (bytes_packets > 8000)
1539 				nextlatency = itr_latency_bulk;
1540 			else if ((packets < 5) && (bytes > 512))
1541 				nextlatency = itr_latency_low;
1542 			break;
1543 		case itr_latency_low: /* 20k ints/s */
1544 			if (bytes > 10000) {
1545 				/* Handle TSO */
1546 				if (bytes_packets > 8000)
1547 					nextlatency = itr_latency_bulk;
1548 				else if ((packets < 10) || (bytes_packets > 1200))
1549 					nextlatency = itr_latency_bulk;
1550 				else if (packets > 35)
1551 					nextlatency = itr_latency_lowest;
1552 			} else if (bytes_packets > 2000) {
1553 				nextlatency = itr_latency_bulk;
1554 			} else if (packets < 3 && bytes < 512) {
1555 				nextlatency = itr_latency_lowest;
1556 			}
1557 			break;
1558 		case itr_latency_bulk: /* 4k ints/s */
1559 			if (bytes > 25000) {
1560 				if (packets > 35)
1561 					nextlatency = itr_latency_low;
1562 			} else if (bytes < 1500)
1563 				nextlatency = itr_latency_low;
1564 			break;
1565 		default:
1566 			nextlatency = itr_latency_low;
1567 			device_printf(sc->dev, "Unexpected newitr transition %d\n",
1568 			    nextlatency);
1569 			break;
1570 		}
1571 
1572 		/* Trim itr_latency_lowest for default AIM setting */
1573 		if (sc->enable_aim == 1 && nextlatency == itr_latency_lowest)
1574 			nextlatency = itr_latency_low;
1575 
1576 		/* Request new latency */
1577 		rxr->rx_nextlatency = nextlatency;
1578 	} else {
1579 		/* We may have toggled to AIM disabled */
1580 		nextlatency = itr_latency_disabled;
1581 		rxr->rx_nextlatency = nextlatency;
1582 	}
1583 
1584 	/* ITR state machine */
1585 	switch(nextlatency) {
1586 	case itr_latency_lowest:
1587 		newitr = EM_INTS_70K;
1588 		break;
1589 	case itr_latency_low:
1590 		newitr = EM_INTS_20K;
1591 		break;
1592 	case itr_latency_bulk:
1593 		newitr = EM_INTS_4K;
1594 		break;
1595 	case itr_latency_disabled:
1596 	default:
1597 		newitr = em_max_interrupt_rate;
1598 		break;
1599 	}
1600 
1601 em_set_next_itr:
1602 	if (hw->mac.type >= igb_mac_min) {
1603 		newitr = IGB_INTS_TO_EITR(newitr);
1604 
1605 		if (hw->mac.type == e1000_82575)
1606 			newitr |= newitr << 16;
1607 		else
1608 			newitr |= E1000_EITR_CNT_IGNR;
1609 
1610 		if (newitr != que->itr_setting) {
1611 			que->itr_setting = newitr;
1612 			E1000_WRITE_REG(hw, E1000_EITR(que->msix), que->itr_setting);
1613 		}
1614 	} else {
1615 		newitr = EM_INTS_TO_ITR(newitr);
1616 
1617 		if (newitr != que->itr_setting) {
1618 			que->itr_setting = newitr;
1619 			if (hw->mac.type == e1000_82574 && que->msix) {
1620 				E1000_WRITE_REG(hw,
1621 				    E1000_EITR_82574(que->msix), que->itr_setting);
1622 			} else {
1623 				E1000_WRITE_REG(hw, E1000_ITR, que->itr_setting);
1624 			}
1625 		}
1626 	}
1627 }
1628 
1629 /*********************************************************************
1630  *
1631  *  Fast Legacy/MSI Combined Interrupt Service routine
1632  *
1633  *********************************************************************/
1634 int
1635 em_intr(void *arg)
1636 {
1637 	struct e1000_softc *sc = arg;
1638 	struct e1000_hw *hw = &sc->hw;
1639 	struct em_rx_queue *que = &sc->rx_queues[0];
1640 	struct tx_ring *txr = &sc->tx_queues[0].txr;
1641 	struct rx_ring *rxr = &que->rxr;
1642 	if_ctx_t ctx = sc->ctx;
1643 	u32 reg_icr;
1644 
1645 	reg_icr = E1000_READ_REG(hw, E1000_ICR);
1646 
1647 	/* Hot eject? */
1648 	if (reg_icr == 0xffffffff)
1649 		return FILTER_STRAY;
1650 
1651 	/* Definitely not our interrupt. */
1652 	if (reg_icr == 0x0)
1653 		return FILTER_STRAY;
1654 
1655 	/*
1656 	 * Starting with the 82571 chip, bit 31 should be used to
1657 	 * determine whether the interrupt belongs to us.
1658 	 */
1659 	if (hw->mac.type >= e1000_82571 &&
1660 	    (reg_icr & E1000_ICR_INT_ASSERTED) == 0)
1661 		return FILTER_STRAY;
1662 
1663 	/*
1664 	 * Only MSI-X interrupts have one-shot behavior by taking advantage
1665 	 * of the EIAC register.  Thus, explicitly disable interrupts.  This
1666 	 * also works around the MSI message reordering errata on certain
1667 	 * systems.
1668 	 */
1669 	IFDI_INTR_DISABLE(ctx);
1670 
1671 	/* Link status change */
1672 	if (reg_icr & (E1000_ICR_RXSEQ | E1000_ICR_LSC))
1673 		em_handle_link(ctx);
1674 
1675 	if (reg_icr & E1000_ICR_RXO)
1676 		sc->rx_overruns++;
1677 
1678 	if (hw->mac.type >= e1000_82540)
1679 		em_newitr(sc, que, txr, rxr);
1680 
1681 	/* Reset state */
1682 	txr->tx_bytes = 0;
1683 	txr->tx_packets = 0;
1684 	rxr->rx_bytes = 0;
1685 	rxr->rx_packets = 0;
1686 
1687 	return (FILTER_SCHEDULE_THREAD);
1688 }
1689 
1690 static int
1691 em_if_rx_queue_intr_enable(if_ctx_t ctx, uint16_t rxqid)
1692 {
1693 	struct e1000_softc *sc = iflib_get_softc(ctx);
1694 	struct em_rx_queue *rxq = &sc->rx_queues[rxqid];
1695 
1696 	E1000_WRITE_REG(&sc->hw, E1000_IMS, rxq->eims);
1697 	return (0);
1698 }
1699 
1700 static int
1701 em_if_tx_queue_intr_enable(if_ctx_t ctx, uint16_t txqid)
1702 {
1703 	struct e1000_softc *sc = iflib_get_softc(ctx);
1704 	struct em_tx_queue *txq = &sc->tx_queues[txqid];
1705 
1706 	E1000_WRITE_REG(&sc->hw, E1000_IMS, txq->eims);
1707 	return (0);
1708 }
1709 
1710 static int
1711 igb_if_rx_queue_intr_enable(if_ctx_t ctx, uint16_t rxqid)
1712 {
1713 	struct e1000_softc *sc = iflib_get_softc(ctx);
1714 	struct em_rx_queue *rxq = &sc->rx_queues[rxqid];
1715 
1716 	E1000_WRITE_REG(&sc->hw, E1000_EIMS, rxq->eims);
1717 	return (0);
1718 }
1719 
1720 static int
1721 igb_if_tx_queue_intr_enable(if_ctx_t ctx, uint16_t txqid)
1722 {
1723 	struct e1000_softc *sc = iflib_get_softc(ctx);
1724 	struct em_tx_queue *txq = &sc->tx_queues[txqid];
1725 
1726 	E1000_WRITE_REG(&sc->hw, E1000_EIMS, txq->eims);
1727 	return (0);
1728 }
1729 
1730 /*********************************************************************
1731  *
1732  *  MSI-X RX Interrupt Service routine
1733  *
1734  **********************************************************************/
1735 static int
1736 em_msix_que(void *arg)
1737 {
1738 	struct em_rx_queue *que = arg;
1739 	struct e1000_softc *sc = que->sc;
1740 	struct tx_ring *txr = &sc->tx_queues[que->msix].txr;
1741 	struct rx_ring *rxr = &que->rxr;
1742 
1743 	++que->irqs;
1744 
1745 	em_newitr(sc, que, txr, rxr);
1746 
1747 	/* Reset state */
1748 	txr->tx_bytes = 0;
1749 	txr->tx_packets = 0;
1750 	rxr->rx_bytes = 0;
1751 	rxr->rx_packets = 0;
1752 
1753 	return (FILTER_SCHEDULE_THREAD);
1754 }
1755 
1756 /*********************************************************************
1757  *
1758  *  MSI-X Link Fast Interrupt Service routine
1759  *
1760  **********************************************************************/
1761 static int
1762 em_msix_link(void *arg)
1763 {
1764 	struct e1000_softc *sc = arg;
1765 	u32 reg_icr;
1766 
1767 	++sc->link_irq;
1768 	MPASS(sc->hw.back != NULL);
1769 	reg_icr = E1000_READ_REG(&sc->hw, E1000_ICR);
1770 
1771 	if (reg_icr & E1000_ICR_RXO)
1772 		sc->rx_overruns++;
1773 
1774 	if (reg_icr & (E1000_ICR_RXSEQ | E1000_ICR_LSC))
1775 		em_handle_link(sc->ctx);
1776 
1777 	/* Re-arm unconditionally */
1778 	if (sc->hw.mac.type >= igb_mac_min) {
1779 		E1000_WRITE_REG(&sc->hw, E1000_IMS, E1000_IMS_LSC);
1780 		E1000_WRITE_REG(&sc->hw, E1000_EIMS, sc->link_mask);
1781 	} else if (sc->hw.mac.type == e1000_82574) {
1782 		E1000_WRITE_REG(&sc->hw, E1000_IMS, E1000_IMS_LSC |
1783 		    E1000_IMS_OTHER);
1784 		/*
1785 		 * Because we must read the ICR for this interrupt it may
1786 		 * clear other causes using autoclear, for this reason we
1787 		 * simply create a soft interrupt for all these vectors.
1788 		 */
1789 		if (reg_icr)
1790 			E1000_WRITE_REG(&sc->hw, E1000_ICS, sc->ims);
1791 	} else
1792 		E1000_WRITE_REG(&sc->hw, E1000_IMS, E1000_IMS_LSC);
1793 
1794 	return (FILTER_HANDLED);
1795 }
1796 
1797 static void
1798 em_handle_link(void *context)
1799 {
1800 	if_ctx_t ctx = context;
1801 	struct e1000_softc *sc = iflib_get_softc(ctx);
1802 
1803 	sc->hw.mac.get_link_status = 1;
1804 	iflib_admin_intr_deferred(ctx);
1805 }
1806 
1807 /*********************************************************************
1808  *
1809  *  Media Ioctl callback
1810  *
1811  *  This routine is called whenever the user queries the status of
1812  *  the interface using ifconfig.
1813  *
1814  **********************************************************************/
1815 static void
1816 em_if_media_status(if_ctx_t ctx, struct ifmediareq *ifmr)
1817 {
1818 	struct e1000_softc *sc = iflib_get_softc(ctx);
1819 	u_char fiber_type = IFM_1000_SX;
1820 
1821 	INIT_DEBUGOUT("em_if_media_status: begin");
1822 
1823 	iflib_admin_intr_deferred(ctx);
1824 
1825 	ifmr->ifm_status = IFM_AVALID;
1826 	ifmr->ifm_active = IFM_ETHER;
1827 
1828 	if (!sc->link_active) {
1829 		return;
1830 	}
1831 
1832 	ifmr->ifm_status |= IFM_ACTIVE;
1833 
1834 	if ((sc->hw.phy.media_type == e1000_media_type_fiber) ||
1835 	    (sc->hw.phy.media_type == e1000_media_type_internal_serdes)) {
1836 		if (sc->hw.mac.type == e1000_82545)
1837 			fiber_type = IFM_1000_LX;
1838 		ifmr->ifm_active |= fiber_type | IFM_FDX;
1839 	} else {
1840 		switch (sc->link_speed) {
1841 		case 10:
1842 			ifmr->ifm_active |= IFM_10_T;
1843 			break;
1844 		case 100:
1845 			ifmr->ifm_active |= IFM_100_TX;
1846 			break;
1847 		case 1000:
1848 			ifmr->ifm_active |= IFM_1000_T;
1849 			break;
1850 		}
1851 		if (sc->link_duplex == FULL_DUPLEX)
1852 			ifmr->ifm_active |= IFM_FDX;
1853 		else
1854 			ifmr->ifm_active |= IFM_HDX;
1855 	}
1856 }
1857 
1858 /*********************************************************************
1859  *
1860  *  Media Ioctl callback
1861  *
1862  *  This routine is called when the user changes speed/duplex using
1863  *  media/mediopt option with ifconfig.
1864  *
1865  **********************************************************************/
1866 static int
1867 em_if_media_change(if_ctx_t ctx)
1868 {
1869 	struct e1000_softc *sc = iflib_get_softc(ctx);
1870 	struct ifmedia *ifm = iflib_get_media(ctx);
1871 
1872 	INIT_DEBUGOUT("em_if_media_change: begin");
1873 
1874 	if (IFM_TYPE(ifm->ifm_media) != IFM_ETHER)
1875 		return (EINVAL);
1876 
1877 	switch (IFM_SUBTYPE(ifm->ifm_media)) {
1878 	case IFM_AUTO:
1879 		sc->hw.mac.autoneg = DO_AUTO_NEG;
1880 		sc->hw.phy.autoneg_advertised = AUTONEG_ADV_DEFAULT;
1881 		break;
1882 	case IFM_1000_LX:
1883 	case IFM_1000_SX:
1884 	case IFM_1000_T:
1885 		sc->hw.mac.autoneg = DO_AUTO_NEG;
1886 		sc->hw.phy.autoneg_advertised = ADVERTISE_1000_FULL;
1887 		break;
1888 	case IFM_100_TX:
1889 		sc->hw.mac.autoneg = false;
1890 		sc->hw.phy.autoneg_advertised = 0;
1891 		if ((ifm->ifm_media & IFM_GMASK) == IFM_FDX)
1892 			sc->hw.mac.forced_speed_duplex = ADVERTISE_100_FULL;
1893 		else
1894 			sc->hw.mac.forced_speed_duplex = ADVERTISE_100_HALF;
1895 		break;
1896 	case IFM_10_T:
1897 		sc->hw.mac.autoneg = false;
1898 		sc->hw.phy.autoneg_advertised = 0;
1899 		if ((ifm->ifm_media & IFM_GMASK) == IFM_FDX)
1900 			sc->hw.mac.forced_speed_duplex = ADVERTISE_10_FULL;
1901 		else
1902 			sc->hw.mac.forced_speed_duplex = ADVERTISE_10_HALF;
1903 		break;
1904 	default:
1905 		device_printf(sc->dev, "Unsupported media type\n");
1906 	}
1907 
1908 	em_if_init(ctx);
1909 
1910 	return (0);
1911 }
1912 
1913 static int
1914 em_if_set_promisc(if_ctx_t ctx, int flags)
1915 {
1916 	struct e1000_softc *sc = iflib_get_softc(ctx);
1917 	if_t ifp = iflib_get_ifp(ctx);
1918 	u32 reg_rctl;
1919 	int mcnt = 0;
1920 
1921 	reg_rctl = E1000_READ_REG(&sc->hw, E1000_RCTL);
1922 	reg_rctl &= ~(E1000_RCTL_SBP | E1000_RCTL_UPE);
1923 	if (flags & IFF_ALLMULTI)
1924 		mcnt = MAX_NUM_MULTICAST_ADDRESSES;
1925 	else
1926 		mcnt = min(if_llmaddr_count(ifp), MAX_NUM_MULTICAST_ADDRESSES);
1927 
1928 	if (mcnt < MAX_NUM_MULTICAST_ADDRESSES)
1929 		reg_rctl &= (~E1000_RCTL_MPE);
1930 	E1000_WRITE_REG(&sc->hw, E1000_RCTL, reg_rctl);
1931 
1932 	if (flags & IFF_PROMISC) {
1933 		reg_rctl |= (E1000_RCTL_UPE | E1000_RCTL_MPE);
1934 		em_if_vlan_filter_disable(sc);
1935 		/* Turn this on if you want to see bad packets */
1936 		if (em_debug_sbp)
1937 			reg_rctl |= E1000_RCTL_SBP;
1938 		E1000_WRITE_REG(&sc->hw, E1000_RCTL, reg_rctl);
1939 	} else {
1940 		if (flags & IFF_ALLMULTI) {
1941 			reg_rctl |= E1000_RCTL_MPE;
1942 			reg_rctl &= ~E1000_RCTL_UPE;
1943 			E1000_WRITE_REG(&sc->hw, E1000_RCTL, reg_rctl);
1944 		}
1945 		if (em_if_vlan_filter_used(ctx))
1946 			em_if_vlan_filter_enable(sc);
1947 	}
1948 	return (0);
1949 }
1950 
1951 static u_int
1952 em_copy_maddr(void *arg, struct sockaddr_dl *sdl, u_int idx)
1953 {
1954 	u8 *mta = arg;
1955 
1956 	if (idx == MAX_NUM_MULTICAST_ADDRESSES)
1957 		return (0);
1958 
1959 	bcopy(LLADDR(sdl), &mta[idx * ETHER_ADDR_LEN], ETHER_ADDR_LEN);
1960 
1961 	return (1);
1962 }
1963 
1964 /*********************************************************************
1965  *  Multicast Update
1966  *
1967  *  This routine is called whenever multicast address list is updated.
1968  *
1969  **********************************************************************/
1970 static void
1971 em_if_multi_set(if_ctx_t ctx)
1972 {
1973 	struct e1000_softc *sc = iflib_get_softc(ctx);
1974 	if_t ifp = iflib_get_ifp(ctx);
1975 	u8  *mta; /* Multicast array memory */
1976 	u32 reg_rctl = 0;
1977 	int mcnt = 0;
1978 
1979 	IOCTL_DEBUGOUT("em_set_multi: begin");
1980 
1981 	mta = sc->mta;
1982 	bzero(mta, sizeof(u8) * ETHER_ADDR_LEN * MAX_NUM_MULTICAST_ADDRESSES);
1983 
1984 	if (sc->hw.mac.type == e1000_82542 &&
1985 	    sc->hw.revision_id == E1000_REVISION_2) {
1986 		reg_rctl = E1000_READ_REG(&sc->hw, E1000_RCTL);
1987 		if (sc->hw.bus.pci_cmd_word & CMD_MEM_WRT_INVALIDATE)
1988 			e1000_pci_clear_mwi(&sc->hw);
1989 		reg_rctl |= E1000_RCTL_RST;
1990 		E1000_WRITE_REG(&sc->hw, E1000_RCTL, reg_rctl);
1991 		msec_delay(5);
1992 	}
1993 
1994 	mcnt = if_foreach_llmaddr(ifp, em_copy_maddr, mta);
1995 
1996 	if (mcnt < MAX_NUM_MULTICAST_ADDRESSES)
1997 		e1000_update_mc_addr_list(&sc->hw, mta, mcnt);
1998 
1999 	reg_rctl = E1000_READ_REG(&sc->hw, E1000_RCTL);
2000 
2001 	if (if_getflags(ifp) & IFF_PROMISC)
2002 		reg_rctl |= (E1000_RCTL_UPE | E1000_RCTL_MPE);
2003 	else if (mcnt >= MAX_NUM_MULTICAST_ADDRESSES ||
2004 	    if_getflags(ifp) & IFF_ALLMULTI) {
2005 		reg_rctl |= E1000_RCTL_MPE;
2006 		reg_rctl &= ~E1000_RCTL_UPE;
2007 	} else
2008 		reg_rctl &= ~(E1000_RCTL_UPE | E1000_RCTL_MPE);
2009 
2010 	E1000_WRITE_REG(&sc->hw, E1000_RCTL, reg_rctl);
2011 
2012 	if (sc->hw.mac.type == e1000_82542 &&
2013 	    sc->hw.revision_id == E1000_REVISION_2) {
2014 		reg_rctl = E1000_READ_REG(&sc->hw, E1000_RCTL);
2015 		reg_rctl &= ~E1000_RCTL_RST;
2016 		E1000_WRITE_REG(&sc->hw, E1000_RCTL, reg_rctl);
2017 		msec_delay(5);
2018 		if (sc->hw.bus.pci_cmd_word & CMD_MEM_WRT_INVALIDATE)
2019 			e1000_pci_set_mwi(&sc->hw);
2020 	}
2021 }
2022 
2023 /*********************************************************************
2024  *  Timer routine
2025  *
2026  *  This routine schedules em_if_update_admin_status() to check for
2027  *  link status and to gather statistics as well as to perform some
2028  *  controller-specific hardware patting.
2029  *
2030  **********************************************************************/
2031 static void
2032 em_if_timer(if_ctx_t ctx, uint16_t qid)
2033 {
2034 
2035 	if (qid != 0)
2036 		return;
2037 
2038 	iflib_admin_intr_deferred(ctx);
2039 }
2040 
2041 static void
2042 em_if_update_admin_status(if_ctx_t ctx)
2043 {
2044 	struct e1000_softc *sc = iflib_get_softc(ctx);
2045 	struct e1000_hw *hw = &sc->hw;
2046 	device_t dev = iflib_get_dev(ctx);
2047 	u32 link_check, thstat, ctrl;
2048 	bool automasked = false;
2049 
2050 	link_check = thstat = ctrl = 0;
2051 	/* Get the cached link value or read phy for real */
2052 	switch (hw->phy.media_type) {
2053 	case e1000_media_type_copper:
2054 		if (hw->mac.get_link_status) {
2055 			if (hw->mac.type == e1000_pch_spt)
2056 				msec_delay(50);
2057 			/* Do the work to read phy */
2058 			e1000_check_for_link(hw);
2059 			link_check = !hw->mac.get_link_status;
2060 			if (link_check) /* ESB2 fix */
2061 				e1000_cfg_on_link_up(hw);
2062 		} else {
2063 			link_check = true;
2064 		}
2065 		break;
2066 	case e1000_media_type_fiber:
2067 		e1000_check_for_link(hw);
2068 		link_check = (E1000_READ_REG(hw, E1000_STATUS) &
2069 			    E1000_STATUS_LU);
2070 		break;
2071 	case e1000_media_type_internal_serdes:
2072 		e1000_check_for_link(hw);
2073 		link_check = hw->mac.serdes_has_link;
2074 		break;
2075 	/* VF device is type_unknown */
2076 	case e1000_media_type_unknown:
2077 		e1000_check_for_link(hw);
2078 		link_check = !hw->mac.get_link_status;
2079 		/* FALLTHROUGH */
2080 	default:
2081 		break;
2082 	}
2083 
2084 	/* Check for thermal downshift or shutdown */
2085 	if (hw->mac.type == e1000_i350) {
2086 		thstat = E1000_READ_REG(hw, E1000_THSTAT);
2087 		ctrl = E1000_READ_REG(hw, E1000_CTRL_EXT);
2088 	}
2089 
2090 	/* Now check for a transition */
2091 	if (link_check && (sc->link_active == 0)) {
2092 		e1000_get_speed_and_duplex(hw, &sc->link_speed,
2093 		    &sc->link_duplex);
2094 		/* Check if we must disable SPEED_MODE bit on PCI-E */
2095 		if ((sc->link_speed != SPEED_1000) &&
2096 		    ((hw->mac.type == e1000_82571) ||
2097 		    (hw->mac.type == e1000_82572))) {
2098 			int tarc0;
2099 			tarc0 = E1000_READ_REG(hw, E1000_TARC(0));
2100 			tarc0 &= ~TARC_SPEED_MODE_BIT;
2101 			E1000_WRITE_REG(hw, E1000_TARC(0), tarc0);
2102 		}
2103 		if (bootverbose)
2104 			device_printf(dev, "Link is up %d Mbps %s\n",
2105 			    sc->link_speed,
2106 			    ((sc->link_duplex == FULL_DUPLEX) ?
2107 			    "Full Duplex" : "Half Duplex"));
2108 		sc->link_active = 1;
2109 		sc->smartspeed = 0;
2110 		if ((ctrl & E1000_CTRL_EXT_LINK_MODE_MASK) ==
2111 		    E1000_CTRL_EXT_LINK_MODE_GMII &&
2112 		    (thstat & E1000_THSTAT_LINK_THROTTLE))
2113 			device_printf(dev, "Link: thermal downshift\n");
2114 		/* Delay Link Up for Phy update */
2115 		if (((hw->mac.type == e1000_i210) ||
2116 		    (hw->mac.type == e1000_i211)) &&
2117 		    (hw->phy.id == I210_I_PHY_ID))
2118 			msec_delay(I210_LINK_DELAY);
2119 		/* Reset if the media type changed. */
2120 		if (hw->dev_spec._82575.media_changed &&
2121 		    hw->mac.type >= igb_mac_min) {
2122 			hw->dev_spec._82575.media_changed = false;
2123 			sc->flags |= IGB_MEDIA_RESET;
2124 			em_reset(ctx);
2125 		}
2126 		/* Only do TSO on gigabit Ethernet for older chips due to errata */
2127 		if (hw->mac.type < igb_mac_min)
2128 			automasked = em_automask_tso(ctx);
2129 
2130 		/* Automasking resets the interface, so don't mark it up yet */
2131 		if (!automasked)
2132 			iflib_link_state_change(ctx, LINK_STATE_UP,
2133 			    IF_Mbps(sc->link_speed));
2134 	} else if (!link_check && (sc->link_active == 1)) {
2135 		sc->link_speed = 0;
2136 		sc->link_duplex = 0;
2137 		sc->link_active = 0;
2138 		iflib_link_state_change(ctx, LINK_STATE_DOWN, 0);
2139 	}
2140 	em_update_stats_counters(sc);
2141 
2142 	/* Reset LAA into RAR[0] on 82571 */
2143 	if (hw->mac.type == e1000_82571 && e1000_get_laa_state_82571(hw))
2144 		e1000_rar_set(hw, hw->mac.addr, 0);
2145 
2146 	if (hw->mac.type < em_mac_min)
2147 		lem_smartspeed(sc);
2148 }
2149 
2150 static void
2151 em_if_watchdog_reset(if_ctx_t ctx)
2152 {
2153 	struct e1000_softc *sc = iflib_get_softc(ctx);
2154 
2155 	/*
2156 	 * Just count the event; iflib(4) will already trigger a
2157 	 * sufficient reset of the controller.
2158 	 */
2159 	sc->watchdog_events++;
2160 }
2161 
2162 /*********************************************************************
2163  *
2164  *  This routine disables all traffic on the adapter by issuing a
2165  *  global reset on the MAC.
2166  *
2167  **********************************************************************/
2168 static void
2169 em_if_stop(if_ctx_t ctx)
2170 {
2171 	struct e1000_softc *sc = iflib_get_softc(ctx);
2172 
2173 	INIT_DEBUGOUT("em_if_stop: begin");
2174 
2175 	/* I219 needs special flushing to avoid hangs */
2176 	if (sc->hw.mac.type >= e1000_pch_spt && sc->hw.mac.type < igb_mac_min)
2177 		em_flush_desc_rings(sc);
2178 
2179 	e1000_reset_hw(&sc->hw);
2180 	if (sc->hw.mac.type >= e1000_82544)
2181 		E1000_WRITE_REG(&sc->hw, E1000_WUFC, 0);
2182 
2183 	e1000_led_off(&sc->hw);
2184 	e1000_cleanup_led(&sc->hw);
2185 }
2186 
2187 /*********************************************************************
2188  *
2189  *  Determine hardware revision.
2190  *
2191  **********************************************************************/
2192 static void
2193 em_identify_hardware(if_ctx_t ctx)
2194 {
2195 	device_t dev = iflib_get_dev(ctx);
2196 	struct e1000_softc *sc = iflib_get_softc(ctx);
2197 
2198 	/* Make sure our PCI config space has the necessary stuff set */
2199 	sc->hw.bus.pci_cmd_word = pci_read_config(dev, PCIR_COMMAND, 2);
2200 
2201 	/* Save off the information about this board */
2202 	sc->hw.vendor_id = pci_get_vendor(dev);
2203 	sc->hw.device_id = pci_get_device(dev);
2204 	sc->hw.revision_id = pci_read_config(dev, PCIR_REVID, 1);
2205 	sc->hw.subsystem_vendor_id =
2206 	    pci_read_config(dev, PCIR_SUBVEND_0, 2);
2207 	sc->hw.subsystem_device_id =
2208 	    pci_read_config(dev, PCIR_SUBDEV_0, 2);
2209 
2210 	/* Do Shared Code Init and Setup */
2211 	if (e1000_set_mac_type(&sc->hw)) {
2212 		device_printf(dev, "Setup init failure\n");
2213 		return;
2214 	}
2215 
2216 	/* Are we a VF device? */
2217 	if ((sc->hw.mac.type == e1000_vfadapt) ||
2218 	    (sc->hw.mac.type == e1000_vfadapt_i350))
2219 		sc->vf_ifp = 1;
2220 	else
2221 		sc->vf_ifp = 0;
2222 }
2223 
2224 static int
2225 em_allocate_pci_resources(if_ctx_t ctx)
2226 {
2227 	struct e1000_softc *sc = iflib_get_softc(ctx);
2228 	device_t dev = iflib_get_dev(ctx);
2229 	int rid, val;
2230 
2231 	rid = PCIR_BAR(0);
2232 	sc->memory = bus_alloc_resource_any(dev, SYS_RES_MEMORY,
2233 	    &rid, RF_ACTIVE);
2234 	if (sc->memory == NULL) {
2235 		device_printf(dev, "Unable to allocate bus resource: memory\n");
2236 		return (ENXIO);
2237 	}
2238 	sc->osdep.mem_bus_space_tag = rman_get_bustag(sc->memory);
2239 	sc->osdep.mem_bus_space_handle =
2240 	    rman_get_bushandle(sc->memory);
2241 	sc->hw.hw_addr = (u8 *)&sc->osdep.mem_bus_space_handle;
2242 
2243 	/* Only older adapters use IO mapping */
2244 	if (sc->hw.mac.type < em_mac_min && sc->hw.mac.type > e1000_82543) {
2245 		/* Figure our where our IO BAR is ? */
2246 		for (rid = PCIR_BAR(0); rid < PCIR_CIS;) {
2247 			val = pci_read_config(dev, rid, 4);
2248 			if (EM_BAR_TYPE(val) == EM_BAR_TYPE_IO) {
2249 				break;
2250 			}
2251 			rid += 4;
2252 			/* check for 64bit BAR */
2253 			if (EM_BAR_MEM_TYPE(val) == EM_BAR_MEM_TYPE_64BIT)
2254 				rid += 4;
2255 		}
2256 		if (rid >= PCIR_CIS) {
2257 			device_printf(dev, "Unable to locate IO BAR\n");
2258 			return (ENXIO);
2259 		}
2260 		sc->ioport = bus_alloc_resource_any(dev, SYS_RES_IOPORT,
2261 		    &rid, RF_ACTIVE);
2262 		if (sc->ioport == NULL) {
2263 			device_printf(dev, "Unable to allocate bus resource: "
2264 			    "ioport\n");
2265 			return (ENXIO);
2266 		}
2267 		sc->hw.io_base = 0;
2268 		sc->osdep.io_bus_space_tag =
2269 		    rman_get_bustag(sc->ioport);
2270 		sc->osdep.io_bus_space_handle =
2271 		    rman_get_bushandle(sc->ioport);
2272 	}
2273 
2274 	sc->hw.back = &sc->osdep;
2275 
2276 	return (0);
2277 }
2278 
2279 /*********************************************************************
2280  *
2281  *  Set up the MSI-X Interrupt handlers
2282  *
2283  **********************************************************************/
2284 static int
2285 em_if_msix_intr_assign(if_ctx_t ctx, int msix)
2286 {
2287 	struct e1000_softc *sc = iflib_get_softc(ctx);
2288 	struct em_rx_queue *rx_que = sc->rx_queues;
2289 	struct em_tx_queue *tx_que = sc->tx_queues;
2290 	int error, rid, i, vector = 0, rx_vectors;
2291 	char buf[16];
2292 
2293 	/* First set up ring resources */
2294 	for (i = 0; i < sc->rx_num_queues; i++, rx_que++, vector++) {
2295 		rid = vector + 1;
2296 		snprintf(buf, sizeof(buf), "rxq%d", i);
2297 		error = iflib_irq_alloc_generic(ctx, &rx_que->que_irq, rid, IFLIB_INTR_RXTX, em_msix_que, rx_que, rx_que->me, buf);
2298 		if (error) {
2299 			device_printf(iflib_get_dev(ctx), "Failed to allocate que int %d err: %d", i, error);
2300 			sc->rx_num_queues = i + 1;
2301 			goto fail;
2302 		}
2303 
2304 		rx_que->msix =  vector;
2305 
2306 		/*
2307 		 * Set the bit to enable interrupt
2308 		 * in E1000_IMS -- bits 20 and 21
2309 		 * are for RX0 and RX1, note this has
2310 		 * NOTHING to do with the MSI-X vector
2311 		 */
2312 		if (sc->hw.mac.type == e1000_82574) {
2313 			rx_que->eims = 1 << (20 + i);
2314 			sc->ims |= rx_que->eims;
2315 			sc->ivars |= (8 | rx_que->msix) << (i * 4);
2316 		} else if (sc->hw.mac.type == e1000_82575)
2317 			rx_que->eims = E1000_EICR_TX_QUEUE0 << vector;
2318 		else
2319 			rx_que->eims = 1 << vector;
2320 	}
2321 	rx_vectors = vector;
2322 
2323 	vector = 0;
2324 	for (i = 0; i < sc->tx_num_queues; i++, tx_que++, vector++) {
2325 		snprintf(buf, sizeof(buf), "txq%d", i);
2326 		tx_que = &sc->tx_queues[i];
2327 		iflib_softirq_alloc_generic(ctx,
2328 		    &sc->rx_queues[i % sc->rx_num_queues].que_irq,
2329 		    IFLIB_INTR_TX, tx_que, tx_que->me, buf);
2330 
2331 		tx_que->msix = (vector % sc->rx_num_queues);
2332 
2333 		/*
2334 		 * Set the bit to enable interrupt
2335 		 * in E1000_IMS -- bits 22 and 23
2336 		 * are for TX0 and TX1, note this has
2337 		 * NOTHING to do with the MSI-X vector
2338 		 */
2339 		if (sc->hw.mac.type == e1000_82574) {
2340 			tx_que->eims = 1 << (22 + i);
2341 			sc->ims |= tx_que->eims;
2342 			sc->ivars |= (8 | tx_que->msix) << (8 + (i * 4));
2343 		} else if (sc->hw.mac.type == e1000_82575) {
2344 			tx_que->eims = E1000_EICR_TX_QUEUE0 << i;
2345 		} else {
2346 			tx_que->eims = 1 << i;
2347 		}
2348 	}
2349 
2350 	/* Link interrupt */
2351 	rid = rx_vectors + 1;
2352 	error = iflib_irq_alloc_generic(ctx, &sc->irq, rid, IFLIB_INTR_ADMIN, em_msix_link, sc, 0, "aq");
2353 
2354 	if (error) {
2355 		device_printf(iflib_get_dev(ctx), "Failed to register admin handler");
2356 		goto fail;
2357 	}
2358 	sc->linkvec = rx_vectors;
2359 	if (sc->hw.mac.type < igb_mac_min) {
2360 		sc->ivars |=  (8 | rx_vectors) << 16;
2361 		sc->ivars |= 0x80000000;
2362 		/* Enable the "Other" interrupt type for link status change */
2363 		sc->ims |= E1000_IMS_OTHER;
2364 	}
2365 
2366 	return (0);
2367 fail:
2368 	iflib_irq_free(ctx, &sc->irq);
2369 	rx_que = sc->rx_queues;
2370 	for (int i = 0; i < sc->rx_num_queues; i++, rx_que++)
2371 		iflib_irq_free(ctx, &rx_que->que_irq);
2372 	return (error);
2373 }
2374 
2375 static void
2376 igb_configure_queues(struct e1000_softc *sc)
2377 {
2378 	struct e1000_hw *hw = &sc->hw;
2379 	struct em_rx_queue *rx_que;
2380 	struct em_tx_queue *tx_que;
2381 	u32 tmp, ivar = 0, newitr = 0;
2382 
2383 	/* First turn on RSS capability */
2384 	if (hw->mac.type != e1000_82575)
2385 		E1000_WRITE_REG(hw, E1000_GPIE,
2386 		    E1000_GPIE_MSIX_MODE | E1000_GPIE_EIAME |
2387 		    E1000_GPIE_PBA | E1000_GPIE_NSICR);
2388 
2389 	/* Turn on MSI-X */
2390 	switch (hw->mac.type) {
2391 	case e1000_82580:
2392 	case e1000_i350:
2393 	case e1000_i354:
2394 	case e1000_i210:
2395 	case e1000_i211:
2396 	case e1000_vfadapt:
2397 	case e1000_vfadapt_i350:
2398 		/* RX entries */
2399 		for (int i = 0; i < sc->rx_num_queues; i++) {
2400 			u32 index = i >> 1;
2401 			ivar = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index);
2402 			rx_que = &sc->rx_queues[i];
2403 			if (i & 1) {
2404 				ivar &= 0xFF00FFFF;
2405 				ivar |= (rx_que->msix | E1000_IVAR_VALID) << 16;
2406 			} else {
2407 				ivar &= 0xFFFFFF00;
2408 				ivar |= rx_que->msix | E1000_IVAR_VALID;
2409 			}
2410 			E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, ivar);
2411 		}
2412 		/* TX entries */
2413 		for (int i = 0; i < sc->tx_num_queues; i++) {
2414 			u32 index = i >> 1;
2415 			ivar = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index);
2416 			tx_que = &sc->tx_queues[i];
2417 			if (i & 1) {
2418 				ivar &= 0x00FFFFFF;
2419 				ivar |= (tx_que->msix | E1000_IVAR_VALID) << 24;
2420 			} else {
2421 				ivar &= 0xFFFF00FF;
2422 				ivar |= (tx_que->msix | E1000_IVAR_VALID) << 8;
2423 			}
2424 			E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, ivar);
2425 			sc->que_mask |= tx_que->eims;
2426 		}
2427 
2428 		/* And for the link interrupt */
2429 		ivar = (sc->linkvec | E1000_IVAR_VALID) << 8;
2430 		sc->link_mask = 1 << sc->linkvec;
2431 		E1000_WRITE_REG(hw, E1000_IVAR_MISC, ivar);
2432 		break;
2433 	case e1000_82576:
2434 		/* RX entries */
2435 		for (int i = 0; i < sc->rx_num_queues; i++) {
2436 			u32 index = i & 0x7; /* Each IVAR has two entries */
2437 			ivar = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index);
2438 			rx_que = &sc->rx_queues[i];
2439 			if (i < 8) {
2440 				ivar &= 0xFFFFFF00;
2441 				ivar |= rx_que->msix | E1000_IVAR_VALID;
2442 			} else {
2443 				ivar &= 0xFF00FFFF;
2444 				ivar |= (rx_que->msix | E1000_IVAR_VALID) << 16;
2445 			}
2446 			E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, ivar);
2447 			sc->que_mask |= rx_que->eims;
2448 		}
2449 		/* TX entries */
2450 		for (int i = 0; i < sc->tx_num_queues; i++) {
2451 			u32 index = i & 0x7; /* Each IVAR has two entries */
2452 			ivar = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index);
2453 			tx_que = &sc->tx_queues[i];
2454 			if (i < 8) {
2455 				ivar &= 0xFFFF00FF;
2456 				ivar |= (tx_que->msix | E1000_IVAR_VALID) << 8;
2457 			} else {
2458 				ivar &= 0x00FFFFFF;
2459 				ivar |= (tx_que->msix | E1000_IVAR_VALID) << 24;
2460 			}
2461 			E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, ivar);
2462 			sc->que_mask |= tx_que->eims;
2463 		}
2464 
2465 		/* And for the link interrupt */
2466 		ivar = (sc->linkvec | E1000_IVAR_VALID) << 8;
2467 		sc->link_mask = 1 << sc->linkvec;
2468 		E1000_WRITE_REG(hw, E1000_IVAR_MISC, ivar);
2469 		break;
2470 
2471 	case e1000_82575:
2472 		/* enable MSI-X support*/
2473 		tmp = E1000_READ_REG(hw, E1000_CTRL_EXT);
2474 		tmp |= E1000_CTRL_EXT_PBA_CLR;
2475 		/* Auto-Mask interrupts upon ICR read. */
2476 		tmp |= E1000_CTRL_EXT_EIAME;
2477 		tmp |= E1000_CTRL_EXT_IRCA;
2478 		E1000_WRITE_REG(hw, E1000_CTRL_EXT, tmp);
2479 
2480 		/* Queues */
2481 		for (int i = 0; i < sc->rx_num_queues; i++) {
2482 			rx_que = &sc->rx_queues[i];
2483 			tmp = E1000_EICR_RX_QUEUE0 << i;
2484 			tmp |= E1000_EICR_TX_QUEUE0 << i;
2485 			rx_que->eims = tmp;
2486 			E1000_WRITE_REG_ARRAY(hw, E1000_MSIXBM(0),
2487 			    i, rx_que->eims);
2488 			sc->que_mask |= rx_que->eims;
2489 		}
2490 
2491 		/* Link */
2492 		E1000_WRITE_REG(hw, E1000_MSIXBM(sc->linkvec),
2493 		    E1000_EIMS_OTHER);
2494 		sc->link_mask |= E1000_EIMS_OTHER;
2495 	default:
2496 		break;
2497 	}
2498 
2499 	/* Set the igb starting interrupt rate */
2500 	if (em_max_interrupt_rate > 0) {
2501 		newitr = IGB_INTS_TO_EITR(em_max_interrupt_rate);
2502 
2503 		if (hw->mac.type == e1000_82575)
2504 			newitr |= newitr << 16;
2505 		else
2506 			newitr |= E1000_EITR_CNT_IGNR;
2507 
2508 		for (int i = 0; i < sc->rx_num_queues; i++) {
2509 			rx_que = &sc->rx_queues[i];
2510 			E1000_WRITE_REG(hw, E1000_EITR(rx_que->msix), newitr);
2511 		}
2512 	}
2513 
2514 	return;
2515 }
2516 
2517 static void
2518 em_free_pci_resources(if_ctx_t ctx)
2519 {
2520 	struct e1000_softc *sc = iflib_get_softc(ctx);
2521 	struct em_rx_queue *que = sc->rx_queues;
2522 	device_t dev = iflib_get_dev(ctx);
2523 
2524 	/* Release all MSI-X queue resources */
2525 	if (sc->intr_type == IFLIB_INTR_MSIX)
2526 		iflib_irq_free(ctx, &sc->irq);
2527 
2528 	if (que != NULL) {
2529 		for (int i = 0; i < sc->rx_num_queues; i++, que++) {
2530 			iflib_irq_free(ctx, &que->que_irq);
2531 		}
2532 	}
2533 
2534 	if (sc->memory != NULL) {
2535 		bus_release_resource(dev, SYS_RES_MEMORY,
2536 		    rman_get_rid(sc->memory), sc->memory);
2537 		sc->memory = NULL;
2538 	}
2539 
2540 	if (sc->flash != NULL) {
2541 		bus_release_resource(dev, SYS_RES_MEMORY,
2542 		    rman_get_rid(sc->flash), sc->flash);
2543 		sc->flash = NULL;
2544 	}
2545 
2546 	if (sc->ioport != NULL) {
2547 		bus_release_resource(dev, SYS_RES_IOPORT,
2548 		    rman_get_rid(sc->ioport), sc->ioport);
2549 		sc->ioport = NULL;
2550 	}
2551 }
2552 
2553 /* Set up MSI or MSI-X */
2554 static int
2555 em_setup_msix(if_ctx_t ctx)
2556 {
2557 	struct e1000_softc *sc = iflib_get_softc(ctx);
2558 
2559 	if (sc->hw.mac.type == e1000_82574) {
2560 		em_enable_vectors_82574(ctx);
2561 	}
2562 	return (0);
2563 }
2564 
2565 /*********************************************************************
2566  *
2567  *  Workaround for SmartSpeed on 82541 and 82547 controllers
2568  *
2569  **********************************************************************/
2570 static void
2571 lem_smartspeed(struct e1000_softc *sc)
2572 {
2573 	u16 phy_tmp;
2574 
2575 	if (sc->link_active || (sc->hw.phy.type != e1000_phy_igp) ||
2576 	    sc->hw.mac.autoneg == 0 ||
2577 	    (sc->hw.phy.autoneg_advertised & ADVERTISE_1000_FULL) == 0)
2578 		return;
2579 
2580 	if (sc->smartspeed == 0) {
2581 		/* If Master/Slave config fault is asserted twice,
2582 		 * we assume back-to-back */
2583 		e1000_read_phy_reg(&sc->hw, PHY_1000T_STATUS, &phy_tmp);
2584 		if (!(phy_tmp & SR_1000T_MS_CONFIG_FAULT))
2585 			return;
2586 		e1000_read_phy_reg(&sc->hw, PHY_1000T_STATUS, &phy_tmp);
2587 		if (phy_tmp & SR_1000T_MS_CONFIG_FAULT) {
2588 			e1000_read_phy_reg(&sc->hw,
2589 			    PHY_1000T_CTRL, &phy_tmp);
2590 			if(phy_tmp & CR_1000T_MS_ENABLE) {
2591 				phy_tmp &= ~CR_1000T_MS_ENABLE;
2592 				e1000_write_phy_reg(&sc->hw,
2593 				    PHY_1000T_CTRL, phy_tmp);
2594 				sc->smartspeed++;
2595 				if(sc->hw.mac.autoneg &&
2596 				   !e1000_copper_link_autoneg(&sc->hw) &&
2597 				   !e1000_read_phy_reg(&sc->hw,
2598 				    PHY_CONTROL, &phy_tmp)) {
2599 					phy_tmp |= (MII_CR_AUTO_NEG_EN |
2600 						    MII_CR_RESTART_AUTO_NEG);
2601 					e1000_write_phy_reg(&sc->hw,
2602 					    PHY_CONTROL, phy_tmp);
2603 				}
2604 			}
2605 		}
2606 		return;
2607 	} else if(sc->smartspeed == EM_SMARTSPEED_DOWNSHIFT) {
2608 		/* If still no link, perhaps using 2/3 pair cable */
2609 		e1000_read_phy_reg(&sc->hw, PHY_1000T_CTRL, &phy_tmp);
2610 		phy_tmp |= CR_1000T_MS_ENABLE;
2611 		e1000_write_phy_reg(&sc->hw, PHY_1000T_CTRL, phy_tmp);
2612 		if(sc->hw.mac.autoneg &&
2613 		   !e1000_copper_link_autoneg(&sc->hw) &&
2614 		   !e1000_read_phy_reg(&sc->hw, PHY_CONTROL, &phy_tmp)) {
2615 			phy_tmp |= (MII_CR_AUTO_NEG_EN |
2616 				    MII_CR_RESTART_AUTO_NEG);
2617 			e1000_write_phy_reg(&sc->hw, PHY_CONTROL, phy_tmp);
2618 		}
2619 	}
2620 	/* Restart process after EM_SMARTSPEED_MAX iterations */
2621 	if(sc->smartspeed++ == EM_SMARTSPEED_MAX)
2622 		sc->smartspeed = 0;
2623 }
2624 
2625 /*********************************************************************
2626  *
2627  *  Initialize the DMA Coalescing feature
2628  *
2629  **********************************************************************/
2630 static void
2631 igb_init_dmac(struct e1000_softc *sc, u32 pba)
2632 {
2633 	device_t	dev = sc->dev;
2634 	struct e1000_hw *hw = &sc->hw;
2635 	u32 		dmac, reg = ~E1000_DMACR_DMAC_EN;
2636 	u16		hwm;
2637 	u16		max_frame_size;
2638 
2639 	if (hw->mac.type == e1000_i211)
2640 		return;
2641 
2642 	max_frame_size = sc->shared->isc_max_frame_size;
2643 	if (hw->mac.type > e1000_82580) {
2644 
2645 		if (sc->dmac == 0) { /* Disabling it */
2646 			E1000_WRITE_REG(hw, E1000_DMACR, reg);
2647 			return;
2648 		} else
2649 			device_printf(dev, "DMA Coalescing enabled\n");
2650 
2651 		/* Set starting threshold */
2652 		E1000_WRITE_REG(hw, E1000_DMCTXTH, 0);
2653 
2654 		hwm = 64 * pba - max_frame_size / 16;
2655 		if (hwm < 64 * (pba - 6))
2656 			hwm = 64 * (pba - 6);
2657 		reg = E1000_READ_REG(hw, E1000_FCRTC);
2658 		reg &= ~E1000_FCRTC_RTH_COAL_MASK;
2659 		reg |= ((hwm << E1000_FCRTC_RTH_COAL_SHIFT)
2660 		    & E1000_FCRTC_RTH_COAL_MASK);
2661 		E1000_WRITE_REG(hw, E1000_FCRTC, reg);
2662 
2663 
2664 		dmac = pba - max_frame_size / 512;
2665 		if (dmac < pba - 10)
2666 			dmac = pba - 10;
2667 		reg = E1000_READ_REG(hw, E1000_DMACR);
2668 		reg &= ~E1000_DMACR_DMACTHR_MASK;
2669 		reg |= ((dmac << E1000_DMACR_DMACTHR_SHIFT)
2670 		    & E1000_DMACR_DMACTHR_MASK);
2671 
2672 		/* transition to L0x or L1 if available..*/
2673 		reg |= (E1000_DMACR_DMAC_EN | E1000_DMACR_DMAC_LX_MASK);
2674 
2675 		/* Check if status is 2.5Gb backplane connection
2676 		* before configuration of watchdog timer, which is
2677 		* in msec values in 12.8usec intervals
2678 		* watchdog timer= msec values in 32usec intervals
2679 		* for non 2.5Gb connection
2680 		*/
2681 		if (hw->mac.type == e1000_i354) {
2682 			int status = E1000_READ_REG(hw, E1000_STATUS);
2683 			if ((status & E1000_STATUS_2P5_SKU) &&
2684 			    (!(status & E1000_STATUS_2P5_SKU_OVER)))
2685 				reg |= ((sc->dmac * 5) >> 6);
2686 			else
2687 				reg |= (sc->dmac >> 5);
2688 		} else {
2689 			reg |= (sc->dmac >> 5);
2690 		}
2691 
2692 		E1000_WRITE_REG(hw, E1000_DMACR, reg);
2693 
2694 		E1000_WRITE_REG(hw, E1000_DMCRTRH, 0);
2695 
2696 		/* Set the interval before transition */
2697 		reg = E1000_READ_REG(hw, E1000_DMCTLX);
2698 		if (hw->mac.type == e1000_i350)
2699 			reg |= IGB_DMCTLX_DCFLUSH_DIS;
2700 		/*
2701 		** in 2.5Gb connection, TTLX unit is 0.4 usec
2702 		** which is 0x4*2 = 0xA. But delay is still 4 usec
2703 		*/
2704 		if (hw->mac.type == e1000_i354) {
2705 			int status = E1000_READ_REG(hw, E1000_STATUS);
2706 			if ((status & E1000_STATUS_2P5_SKU) &&
2707 			    (!(status & E1000_STATUS_2P5_SKU_OVER)))
2708 				reg |= 0xA;
2709 			else
2710 				reg |= 0x4;
2711 		} else {
2712 			reg |= 0x4;
2713 		}
2714 
2715 		E1000_WRITE_REG(hw, E1000_DMCTLX, reg);
2716 
2717 		/* free space in tx packet buffer to wake from DMA coal */
2718 		E1000_WRITE_REG(hw, E1000_DMCTXTH, (IGB_TXPBSIZE -
2719 		    (2 * max_frame_size)) >> 6);
2720 
2721 		/* make low power state decision controlled by DMA coal */
2722 		reg = E1000_READ_REG(hw, E1000_PCIEMISC);
2723 		reg &= ~E1000_PCIEMISC_LX_DECISION;
2724 		E1000_WRITE_REG(hw, E1000_PCIEMISC, reg);
2725 
2726 	} else if (hw->mac.type == e1000_82580) {
2727 		u32 reg = E1000_READ_REG(hw, E1000_PCIEMISC);
2728 		E1000_WRITE_REG(hw, E1000_PCIEMISC,
2729 		    reg & ~E1000_PCIEMISC_LX_DECISION);
2730 		E1000_WRITE_REG(hw, E1000_DMACR, 0);
2731 	}
2732 }
2733 /*********************************************************************
2734  * The 3 following flush routines are used as a workaround in the
2735  * I219 client parts and only for them.
2736  *
2737  * em_flush_tx_ring - remove all descriptors from the tx_ring
2738  *
2739  * We want to clear all pending descriptors from the TX ring.
2740  * zeroing happens when the HW reads the regs. We assign the ring itself as
2741  * the data of the next descriptor. We don't care about the data we are about
2742  * to reset the HW.
2743  **********************************************************************/
2744 static void
2745 em_flush_tx_ring(struct e1000_softc *sc)
2746 {
2747 	struct e1000_hw		*hw = &sc->hw;
2748 	struct tx_ring		*txr = &sc->tx_queues->txr;
2749 	struct e1000_tx_desc	*txd;
2750 	u32			tctl, txd_lower = E1000_TXD_CMD_IFCS;
2751 	u16			size = 512;
2752 
2753 	tctl = E1000_READ_REG(hw, E1000_TCTL);
2754 	E1000_WRITE_REG(hw, E1000_TCTL, tctl | E1000_TCTL_EN);
2755 
2756 	txd = &txr->tx_base[txr->tx_cidx_processed];
2757 
2758 	/* Just use the ring as a dummy buffer addr */
2759 	txd->buffer_addr = txr->tx_paddr;
2760 	txd->lower.data = htole32(txd_lower | size);
2761 	txd->upper.data = 0;
2762 
2763 	/* flush descriptors to memory before notifying the HW */
2764 	wmb();
2765 
2766 	E1000_WRITE_REG(hw, E1000_TDT(0), txr->tx_cidx_processed);
2767 	mb();
2768 	usec_delay(250);
2769 }
2770 
2771 /*********************************************************************
2772  * em_flush_rx_ring - remove all descriptors from the rx_ring
2773  *
2774  * Mark all descriptors in the RX ring as consumed and disable the rx ring
2775  **********************************************************************/
2776 static void
2777 em_flush_rx_ring(struct e1000_softc *sc)
2778 {
2779 	struct e1000_hw	*hw = &sc->hw;
2780 	u32		rctl, rxdctl;
2781 
2782 	rctl = E1000_READ_REG(hw, E1000_RCTL);
2783 	E1000_WRITE_REG(hw, E1000_RCTL, rctl & ~E1000_RCTL_EN);
2784 	E1000_WRITE_FLUSH(hw);
2785 	usec_delay(150);
2786 
2787 	rxdctl = E1000_READ_REG(hw, E1000_RXDCTL(0));
2788 	/* zero the lower 14 bits (prefetch and host thresholds) */
2789 	rxdctl &= 0xffffc000;
2790 	/*
2791 	 * update thresholds: prefetch threshold to 31, host threshold to 1
2792 	 * and make sure the granularity is "descriptors" and not "cache lines"
2793 	 */
2794 	rxdctl |= (0x1F | (1 << 8) | E1000_RXDCTL_THRESH_UNIT_DESC);
2795 	E1000_WRITE_REG(hw, E1000_RXDCTL(0), rxdctl);
2796 
2797 	/* momentarily enable the RX ring for the changes to take effect */
2798 	E1000_WRITE_REG(hw, E1000_RCTL, rctl | E1000_RCTL_EN);
2799 	E1000_WRITE_FLUSH(hw);
2800 	usec_delay(150);
2801 	E1000_WRITE_REG(hw, E1000_RCTL, rctl & ~E1000_RCTL_EN);
2802 }
2803 
2804 /*********************************************************************
2805  * em_flush_desc_rings - remove all descriptors from the descriptor rings
2806  *
2807  * In I219, the descriptor rings must be emptied before resetting the HW
2808  * or before changing the device state to D3 during runtime (runtime PM).
2809  *
2810  * Failure to do this will cause the HW to enter a unit hang state which can
2811  * only be released by PCI reset on the device
2812  *
2813  **********************************************************************/
2814 static void
2815 em_flush_desc_rings(struct e1000_softc *sc)
2816 {
2817 	struct e1000_hw	*hw = &sc->hw;
2818 	device_t dev = sc->dev;
2819 	u16		hang_state;
2820 	u32		fext_nvm11, tdlen;
2821 
2822 	/* First, disable MULR fix in FEXTNVM11 */
2823 	fext_nvm11 = E1000_READ_REG(hw, E1000_FEXTNVM11);
2824 	fext_nvm11 |= E1000_FEXTNVM11_DISABLE_MULR_FIX;
2825 	E1000_WRITE_REG(hw, E1000_FEXTNVM11, fext_nvm11);
2826 
2827 	/* do nothing if we're not in faulty state, or if the queue is empty */
2828 	tdlen = E1000_READ_REG(hw, E1000_TDLEN(0));
2829 	hang_state = pci_read_config(dev, PCICFG_DESC_RING_STATUS, 2);
2830 	if (!(hang_state & FLUSH_DESC_REQUIRED) || !tdlen)
2831 		return;
2832 	em_flush_tx_ring(sc);
2833 
2834 	/* recheck, maybe the fault is caused by the rx ring */
2835 	hang_state = pci_read_config(dev, PCICFG_DESC_RING_STATUS, 2);
2836 	if (hang_state & FLUSH_DESC_REQUIRED)
2837 		em_flush_rx_ring(sc);
2838 }
2839 
2840 
2841 /*********************************************************************
2842  *
2843  *  Initialize the hardware to a configuration as specified by the
2844  *  sc structure.
2845  *
2846  **********************************************************************/
2847 static void
2848 em_reset(if_ctx_t ctx)
2849 {
2850 	device_t dev = iflib_get_dev(ctx);
2851 	struct e1000_softc *sc = iflib_get_softc(ctx);
2852 	if_t ifp = iflib_get_ifp(ctx);
2853 	struct e1000_hw *hw = &sc->hw;
2854 	u32 rx_buffer_size;
2855 	u32 pba;
2856 
2857 	INIT_DEBUGOUT("em_reset: begin");
2858 	/* Let the firmware know the OS is in control */
2859 	em_get_hw_control(sc);
2860 
2861 	/* Set up smart power down as default off on newer adapters. */
2862 	if (!em_smart_pwr_down && (hw->mac.type == e1000_82571 ||
2863 	    hw->mac.type == e1000_82572)) {
2864 		u16 phy_tmp = 0;
2865 
2866 		/* Speed up time to link by disabling smart power down. */
2867 		e1000_read_phy_reg(hw, IGP02E1000_PHY_POWER_MGMT, &phy_tmp);
2868 		phy_tmp &= ~IGP02E1000_PM_SPD;
2869 		e1000_write_phy_reg(hw, IGP02E1000_PHY_POWER_MGMT, phy_tmp);
2870 	}
2871 
2872 	/*
2873 	 * Packet Buffer Allocation (PBA)
2874 	 * Writing PBA sets the receive portion of the buffer
2875 	 * the remainder is used for the transmit buffer.
2876 	 */
2877 	switch (hw->mac.type) {
2878 	/* 82547: Total Packet Buffer is 40K */
2879 	case e1000_82547:
2880 	case e1000_82547_rev_2:
2881 		if (hw->mac.max_frame_size > 8192)
2882 			pba = E1000_PBA_22K; /* 22K for Rx, 18K for Tx */
2883 		else
2884 			pba = E1000_PBA_30K; /* 30K for Rx, 10K for Tx */
2885 		break;
2886 	/* 82571/82572/80003es2lan: Total Packet Buffer is 48K */
2887 	case e1000_82571:
2888 	case e1000_82572:
2889 	case e1000_80003es2lan:
2890 			pba = E1000_PBA_32K; /* 32K for Rx, 16K for Tx */
2891 		break;
2892 	/* 82573: Total Packet Buffer is 32K */
2893 	case e1000_82573:
2894 			pba = E1000_PBA_12K; /* 12K for Rx, 20K for Tx */
2895 		break;
2896 	case e1000_82574:
2897 	case e1000_82583:
2898 			pba = E1000_PBA_20K; /* 20K for Rx, 20K for Tx */
2899 		break;
2900 	case e1000_ich8lan:
2901 		pba = E1000_PBA_8K;
2902 		break;
2903 	case e1000_ich9lan:
2904 	case e1000_ich10lan:
2905 		/* Boost Receive side for jumbo frames */
2906 		if (hw->mac.max_frame_size > 4096)
2907 			pba = E1000_PBA_14K;
2908 		else
2909 			pba = E1000_PBA_10K;
2910 		break;
2911 	case e1000_pchlan:
2912 	case e1000_pch2lan:
2913 	case e1000_pch_lpt:
2914 	case e1000_pch_spt:
2915 	case e1000_pch_cnp:
2916 	case e1000_pch_tgp:
2917 	case e1000_pch_adp:
2918 	case e1000_pch_mtp:
2919 	case e1000_pch_ptp:
2920 		pba = E1000_PBA_26K;
2921 		break;
2922 	case e1000_82575:
2923 		pba = E1000_PBA_32K;
2924 		break;
2925 	case e1000_82576:
2926 	case e1000_vfadapt:
2927 		pba = E1000_READ_REG(hw, E1000_RXPBS);
2928 		pba &= E1000_RXPBS_SIZE_MASK_82576;
2929 		break;
2930 	case e1000_82580:
2931 	case e1000_i350:
2932 	case e1000_i354:
2933 	case e1000_vfadapt_i350:
2934 		pba = E1000_READ_REG(hw, E1000_RXPBS);
2935 		pba = e1000_rxpbs_adjust_82580(pba);
2936 		break;
2937 	case e1000_i210:
2938 	case e1000_i211:
2939 		pba = E1000_PBA_34K;
2940 		break;
2941 	default:
2942 		/* Remaining devices assumed to have a Packet Buffer of 64K. */
2943 		if (hw->mac.max_frame_size > 8192)
2944 			pba = E1000_PBA_40K; /* 40K for Rx, 24K for Tx */
2945 		else
2946 			pba = E1000_PBA_48K; /* 48K for Rx, 16K for Tx */
2947 	}
2948 
2949 	/* Special needs in case of Jumbo frames */
2950 	if ((hw->mac.type == e1000_82575) && (if_getmtu(ifp) > ETHERMTU)) {
2951 		u32 tx_space, min_tx, min_rx;
2952 		pba = E1000_READ_REG(hw, E1000_PBA);
2953 		tx_space = pba >> 16;
2954 		pba &= 0xffff;
2955 		min_tx = (hw->mac.max_frame_size +
2956 		    sizeof(struct e1000_tx_desc) - ETHERNET_FCS_SIZE) * 2;
2957 		min_tx = roundup2(min_tx, 1024);
2958 		min_tx >>= 10;
2959 		min_rx = hw->mac.max_frame_size;
2960 		min_rx = roundup2(min_rx, 1024);
2961 		min_rx >>= 10;
2962 		if (tx_space < min_tx &&
2963 		    ((min_tx - tx_space) < pba)) {
2964 			pba = pba - (min_tx - tx_space);
2965 			/*
2966 			 * if short on rx space, rx wins
2967 			 * and must trump tx adjustment
2968 			 */
2969 			if (pba < min_rx)
2970 				pba = min_rx;
2971 		}
2972 		E1000_WRITE_REG(hw, E1000_PBA, pba);
2973 	}
2974 
2975 	if (hw->mac.type < igb_mac_min)
2976 		E1000_WRITE_REG(hw, E1000_PBA, pba);
2977 
2978 	INIT_DEBUGOUT1("em_reset: pba=%dK",pba);
2979 
2980 	/*
2981 	 * These parameters control the automatic generation (Tx) and
2982 	 * response (Rx) to Ethernet PAUSE frames.
2983 	 * - High water mark should allow for at least two frames to be
2984 	 *   received after sending an XOFF.
2985 	 * - Low water mark works best when it is very near the high water mark.
2986 	 *   This allows the receiver to restart by sending XON when it has
2987 	 *   drained a bit. Here we use an arbitrary value of 1500 which will
2988 	 *   restart after one full frame is pulled from the buffer. There
2989 	 *   could be several smaller frames in the buffer and if so they will
2990 	 *   not trigger the XON until their total number reduces the buffer
2991 	 *   by 1500.
2992 	 * - The pause time is fairly large at 1000 x 512ns = 512 usec.
2993 	 */
2994 	rx_buffer_size = (pba & 0xffff) << 10;
2995 	hw->fc.high_water = rx_buffer_size -
2996 	    roundup2(hw->mac.max_frame_size, 1024);
2997 	hw->fc.low_water = hw->fc.high_water - 1500;
2998 
2999 	if (sc->fc) /* locally set flow control value? */
3000 		hw->fc.requested_mode = sc->fc;
3001 	else
3002 		hw->fc.requested_mode = e1000_fc_full;
3003 
3004 	if (hw->mac.type == e1000_80003es2lan)
3005 		hw->fc.pause_time = 0xFFFF;
3006 	else
3007 		hw->fc.pause_time = EM_FC_PAUSE_TIME;
3008 
3009 	hw->fc.send_xon = true;
3010 
3011 	/* Device specific overrides/settings */
3012 	switch (hw->mac.type) {
3013 	case e1000_pchlan:
3014 		/* Workaround: no TX flow ctrl for PCH */
3015 		hw->fc.requested_mode = e1000_fc_rx_pause;
3016 		hw->fc.pause_time = 0xFFFF; /* override */
3017 		if (if_getmtu(ifp) > ETHERMTU) {
3018 			hw->fc.high_water = 0x3500;
3019 			hw->fc.low_water = 0x1500;
3020 		} else {
3021 			hw->fc.high_water = 0x5000;
3022 			hw->fc.low_water = 0x3000;
3023 		}
3024 		hw->fc.refresh_time = 0x1000;
3025 		break;
3026 	case e1000_pch2lan:
3027 	case e1000_pch_lpt:
3028 	case e1000_pch_spt:
3029 	case e1000_pch_cnp:
3030 	case e1000_pch_tgp:
3031 	case e1000_pch_adp:
3032 	case e1000_pch_mtp:
3033 	case e1000_pch_ptp:
3034 		hw->fc.high_water = 0x5C20;
3035 		hw->fc.low_water = 0x5048;
3036 		hw->fc.pause_time = 0x0650;
3037 		hw->fc.refresh_time = 0x0400;
3038 		/* Jumbos need adjusted PBA */
3039 		if (if_getmtu(ifp) > ETHERMTU)
3040 			E1000_WRITE_REG(hw, E1000_PBA, 12);
3041 		else
3042 			E1000_WRITE_REG(hw, E1000_PBA, 26);
3043 		break;
3044 	case e1000_82575:
3045 	case e1000_82576:
3046 		/* 8-byte granularity */
3047 		hw->fc.low_water = hw->fc.high_water - 8;
3048 		break;
3049 	case e1000_82580:
3050 	case e1000_i350:
3051 	case e1000_i354:
3052 	case e1000_i210:
3053 	case e1000_i211:
3054 	case e1000_vfadapt:
3055 	case e1000_vfadapt_i350:
3056 		/* 16-byte granularity */
3057 		hw->fc.low_water = hw->fc.high_water - 16;
3058 		break;
3059 	case e1000_ich9lan:
3060 	case e1000_ich10lan:
3061 		if (if_getmtu(ifp) > ETHERMTU) {
3062 			hw->fc.high_water = 0x2800;
3063 			hw->fc.low_water = hw->fc.high_water - 8;
3064 			break;
3065 		}
3066 		/* FALLTHROUGH */
3067 	default:
3068 		if (hw->mac.type == e1000_80003es2lan)
3069 			hw->fc.pause_time = 0xFFFF;
3070 		break;
3071 	}
3072 
3073 	/* I219 needs some special flushing to avoid hangs */
3074 	if (sc->hw.mac.type >= e1000_pch_spt && sc->hw.mac.type < igb_mac_min)
3075 		em_flush_desc_rings(sc);
3076 
3077 	/* Issue a global reset */
3078 	e1000_reset_hw(hw);
3079 	if (hw->mac.type >= igb_mac_min) {
3080 		E1000_WRITE_REG(hw, E1000_WUC, 0);
3081 	} else {
3082 		E1000_WRITE_REG(hw, E1000_WUFC, 0);
3083 		em_disable_aspm(sc);
3084 	}
3085 	if (sc->flags & IGB_MEDIA_RESET) {
3086 		e1000_setup_init_funcs(hw, true);
3087 		e1000_get_bus_info(hw);
3088 		sc->flags &= ~IGB_MEDIA_RESET;
3089 	}
3090 	/* and a re-init */
3091 	if (e1000_init_hw(hw) < 0) {
3092 		device_printf(dev, "Hardware Initialization Failed\n");
3093 		return;
3094 	}
3095 	if (hw->mac.type >= igb_mac_min)
3096 		igb_init_dmac(sc, pba);
3097 
3098 	/* Save the final PBA off if it needs to be used elsewhere i.e. AIM */
3099 	sc->pba = pba;
3100 
3101 	E1000_WRITE_REG(hw, E1000_VET, ETHERTYPE_VLAN);
3102 	e1000_get_phy_info(hw);
3103 	e1000_check_for_link(hw);
3104 }
3105 
3106 /*
3107  * Initialise the RSS mapping for NICs that support multiple transmit/
3108  * receive rings.
3109  */
3110 
3111 #define RSSKEYLEN 10
3112 static void
3113 em_initialize_rss_mapping(struct e1000_softc *sc)
3114 {
3115 	uint8_t  rss_key[4 * RSSKEYLEN];
3116 	uint32_t reta = 0;
3117 	struct e1000_hw	*hw = &sc->hw;
3118 	int i;
3119 
3120 	/*
3121 	 * Configure RSS key
3122 	 */
3123 	arc4rand(rss_key, sizeof(rss_key), 0);
3124 	for (i = 0; i < RSSKEYLEN; ++i) {
3125 		uint32_t rssrk = 0;
3126 
3127 		rssrk = EM_RSSRK_VAL(rss_key, i);
3128 		E1000_WRITE_REG(hw,E1000_RSSRK(i), rssrk);
3129 	}
3130 
3131 	/*
3132 	 * Configure RSS redirect table in following fashion:
3133 	 * (hash & ring_cnt_mask) == rdr_table[(hash & rdr_table_mask)]
3134 	 */
3135 	for (i = 0; i < sizeof(reta); ++i) {
3136 		uint32_t q;
3137 
3138 		q = (i % sc->rx_num_queues) << 7;
3139 		reta |= q << (8 * i);
3140 	}
3141 
3142 	for (i = 0; i < 32; ++i)
3143 		E1000_WRITE_REG(hw, E1000_RETA(i), reta);
3144 
3145 	E1000_WRITE_REG(hw, E1000_MRQC, E1000_MRQC_RSS_ENABLE_2Q |
3146 			E1000_MRQC_RSS_FIELD_IPV4_TCP |
3147 			E1000_MRQC_RSS_FIELD_IPV4 |
3148 			E1000_MRQC_RSS_FIELD_IPV6_TCP_EX |
3149 			E1000_MRQC_RSS_FIELD_IPV6_EX |
3150 			E1000_MRQC_RSS_FIELD_IPV6);
3151 }
3152 
3153 static void
3154 igb_initialize_rss_mapping(struct e1000_softc *sc)
3155 {
3156 	struct e1000_hw *hw = &sc->hw;
3157 	int i;
3158 	int queue_id;
3159 	u32 reta;
3160 	u32 rss_key[10], mrqc, shift = 0;
3161 
3162 	/* XXX? */
3163 	if (hw->mac.type == e1000_82575)
3164 		shift = 6;
3165 
3166 	/*
3167 	 * The redirection table controls which destination
3168 	 * queue each bucket redirects traffic to.
3169 	 * Each DWORD represents four queues, with the LSB
3170 	 * being the first queue in the DWORD.
3171 	 *
3172 	 * This just allocates buckets to queues using round-robin
3173 	 * allocation.
3174 	 *
3175 	 * NOTE: It Just Happens to line up with the default
3176 	 * RSS allocation method.
3177 	 */
3178 
3179 	/* Warning FM follows */
3180 	reta = 0;
3181 	for (i = 0; i < 128; i++) {
3182 #ifdef RSS
3183 		queue_id = rss_get_indirection_to_bucket(i);
3184 		/*
3185 		 * If we have more queues than buckets, we'll
3186 		 * end up mapping buckets to a subset of the
3187 		 * queues.
3188 		 *
3189 		 * If we have more buckets than queues, we'll
3190 		 * end up instead assigning multiple buckets
3191 		 * to queues.
3192 		 *
3193 		 * Both are suboptimal, but we need to handle
3194 		 * the case so we don't go out of bounds
3195 		 * indexing arrays and such.
3196 		 */
3197 		queue_id = queue_id % sc->rx_num_queues;
3198 #else
3199 		queue_id = (i % sc->rx_num_queues);
3200 #endif
3201 		/* Adjust if required */
3202 		queue_id = queue_id << shift;
3203 
3204 		/*
3205 		 * The low 8 bits are for hash value (n+0);
3206 		 * The next 8 bits are for hash value (n+1), etc.
3207 		 */
3208 		reta = reta >> 8;
3209 		reta = reta | ( ((uint32_t) queue_id) << 24);
3210 		if ((i & 3) == 3) {
3211 			E1000_WRITE_REG(hw, E1000_RETA(i >> 2), reta);
3212 			reta = 0;
3213 		}
3214 	}
3215 
3216 	/* Now fill in hash table */
3217 
3218 	/*
3219 	 * MRQC: Multiple Receive Queues Command
3220 	 * Set queuing to RSS control, number depends on the device.
3221 	 */
3222 	mrqc = E1000_MRQC_ENABLE_RSS_MQ;
3223 
3224 #ifdef RSS
3225 	/* XXX ew typecasting */
3226 	rss_getkey((uint8_t *) &rss_key);
3227 #else
3228 	arc4rand(&rss_key, sizeof(rss_key), 0);
3229 #endif
3230 	for (i = 0; i < 10; i++)
3231 		E1000_WRITE_REG_ARRAY(hw, E1000_RSSRK(0), i, rss_key[i]);
3232 
3233 	/*
3234 	 * Configure the RSS fields to hash upon.
3235 	 */
3236 	mrqc |= (E1000_MRQC_RSS_FIELD_IPV4 |
3237 	    E1000_MRQC_RSS_FIELD_IPV4_TCP);
3238 	mrqc |= (E1000_MRQC_RSS_FIELD_IPV6 |
3239 	    E1000_MRQC_RSS_FIELD_IPV6_TCP);
3240 	mrqc |=( E1000_MRQC_RSS_FIELD_IPV4_UDP |
3241 	    E1000_MRQC_RSS_FIELD_IPV6_UDP);
3242 	mrqc |=( E1000_MRQC_RSS_FIELD_IPV6_UDP_EX |
3243 	    E1000_MRQC_RSS_FIELD_IPV6_TCP_EX);
3244 
3245 	E1000_WRITE_REG(hw, E1000_MRQC, mrqc);
3246 }
3247 
3248 /*********************************************************************
3249  *
3250  *  Setup networking device structure and register interface media.
3251  *
3252  **********************************************************************/
3253 static int
3254 em_setup_interface(if_ctx_t ctx)
3255 {
3256 	if_t ifp = iflib_get_ifp(ctx);
3257 	struct e1000_softc *sc = iflib_get_softc(ctx);
3258 	if_softc_ctx_t scctx = sc->shared;
3259 
3260 	INIT_DEBUGOUT("em_setup_interface: begin");
3261 
3262 	/* Single Queue */
3263 	if (sc->tx_num_queues == 1) {
3264 		if_setsendqlen(ifp, scctx->isc_ntxd[0] - 1);
3265 		if_setsendqready(ifp);
3266 	}
3267 
3268 	/*
3269 	 * Specify the media types supported by this adapter and register
3270 	 * callbacks to update media and link information
3271 	 */
3272 	if (sc->hw.phy.media_type == e1000_media_type_fiber ||
3273 	    sc->hw.phy.media_type == e1000_media_type_internal_serdes) {
3274 		u_char fiber_type = IFM_1000_SX;	/* default type */
3275 
3276 		if (sc->hw.mac.type == e1000_82545)
3277 			fiber_type = IFM_1000_LX;
3278 		ifmedia_add(sc->media, IFM_ETHER | fiber_type | IFM_FDX, 0, NULL);
3279 		ifmedia_add(sc->media, IFM_ETHER | fiber_type, 0, NULL);
3280 	} else {
3281 		ifmedia_add(sc->media, IFM_ETHER | IFM_10_T, 0, NULL);
3282 		ifmedia_add(sc->media, IFM_ETHER | IFM_10_T | IFM_FDX, 0, NULL);
3283 		ifmedia_add(sc->media, IFM_ETHER | IFM_100_TX, 0, NULL);
3284 		ifmedia_add(sc->media, IFM_ETHER | IFM_100_TX | IFM_FDX, 0, NULL);
3285 		if (sc->hw.phy.type != e1000_phy_ife) {
3286 			ifmedia_add(sc->media, IFM_ETHER | IFM_1000_T | IFM_FDX, 0, NULL);
3287 			ifmedia_add(sc->media, IFM_ETHER | IFM_1000_T, 0, NULL);
3288 		}
3289 	}
3290 	ifmedia_add(sc->media, IFM_ETHER | IFM_AUTO, 0, NULL);
3291 	ifmedia_set(sc->media, IFM_ETHER | IFM_AUTO);
3292 	return (0);
3293 }
3294 
3295 static int
3296 em_if_tx_queues_alloc(if_ctx_t ctx, caddr_t *vaddrs, uint64_t *paddrs, int ntxqs, int ntxqsets)
3297 {
3298 	struct e1000_softc *sc = iflib_get_softc(ctx);
3299 	if_softc_ctx_t scctx = sc->shared;
3300 	int error = E1000_SUCCESS;
3301 	struct em_tx_queue *que;
3302 	int i, j;
3303 
3304 	MPASS(sc->tx_num_queues > 0);
3305 	MPASS(sc->tx_num_queues == ntxqsets);
3306 
3307 	/* First allocate the top level queue structs */
3308 	if (!(sc->tx_queues =
3309 	    (struct em_tx_queue *) malloc(sizeof(struct em_tx_queue) *
3310 	    sc->tx_num_queues, M_DEVBUF, M_NOWAIT | M_ZERO))) {
3311 		device_printf(iflib_get_dev(ctx), "Unable to allocate queue memory\n");
3312 		return(ENOMEM);
3313 	}
3314 
3315 	for (i = 0, que = sc->tx_queues; i < sc->tx_num_queues; i++, que++) {
3316 		/* Set up some basics */
3317 
3318 		struct tx_ring *txr = &que->txr;
3319 		txr->sc = que->sc = sc;
3320 		que->me = txr->me =  i;
3321 
3322 		/* Allocate report status array */
3323 		if (!(txr->tx_rsq = (qidx_t *) malloc(sizeof(qidx_t) * scctx->isc_ntxd[0], M_DEVBUF, M_NOWAIT | M_ZERO))) {
3324 			device_printf(iflib_get_dev(ctx), "failed to allocate rs_idxs memory\n");
3325 			error = ENOMEM;
3326 			goto fail;
3327 		}
3328 		for (j = 0; j < scctx->isc_ntxd[0]; j++)
3329 			txr->tx_rsq[j] = QIDX_INVALID;
3330 		/* get the virtual and physical address of the hardware queues */
3331 		txr->tx_base = (struct e1000_tx_desc *)vaddrs[i*ntxqs];
3332 		txr->tx_paddr = paddrs[i*ntxqs];
3333 	}
3334 
3335 	if (bootverbose)
3336 		device_printf(iflib_get_dev(ctx),
3337 		    "allocated for %d tx_queues\n", sc->tx_num_queues);
3338 	return (0);
3339 fail:
3340 	em_if_queues_free(ctx);
3341 	return (error);
3342 }
3343 
3344 static int
3345 em_if_rx_queues_alloc(if_ctx_t ctx, caddr_t *vaddrs, uint64_t *paddrs, int nrxqs, int nrxqsets)
3346 {
3347 	struct e1000_softc *sc = iflib_get_softc(ctx);
3348 	int error = E1000_SUCCESS;
3349 	struct em_rx_queue *que;
3350 	int i;
3351 
3352 	MPASS(sc->rx_num_queues > 0);
3353 	MPASS(sc->rx_num_queues == nrxqsets);
3354 
3355 	/* First allocate the top level queue structs */
3356 	if (!(sc->rx_queues =
3357 	    (struct em_rx_queue *) malloc(sizeof(struct em_rx_queue) *
3358 	    sc->rx_num_queues, M_DEVBUF, M_NOWAIT | M_ZERO))) {
3359 		device_printf(iflib_get_dev(ctx), "Unable to allocate queue memory\n");
3360 		error = ENOMEM;
3361 		goto fail;
3362 	}
3363 
3364 	for (i = 0, que = sc->rx_queues; i < nrxqsets; i++, que++) {
3365 		/* Set up some basics */
3366 		struct rx_ring *rxr = &que->rxr;
3367 		rxr->sc = que->sc = sc;
3368 		rxr->que = que;
3369 		que->me = rxr->me =  i;
3370 
3371 		/* get the virtual and physical address of the hardware queues */
3372 		rxr->rx_base = (union e1000_rx_desc_extended *)vaddrs[i*nrxqs];
3373 		rxr->rx_paddr = paddrs[i*nrxqs];
3374 	}
3375 
3376 	if (bootverbose)
3377 		device_printf(iflib_get_dev(ctx),
3378 		    "allocated for %d rx_queues\n", sc->rx_num_queues);
3379 
3380 	return (0);
3381 fail:
3382 	em_if_queues_free(ctx);
3383 	return (error);
3384 }
3385 
3386 static void
3387 em_if_queues_free(if_ctx_t ctx)
3388 {
3389 	struct e1000_softc *sc = iflib_get_softc(ctx);
3390 	struct em_tx_queue *tx_que = sc->tx_queues;
3391 	struct em_rx_queue *rx_que = sc->rx_queues;
3392 
3393 	if (tx_que != NULL) {
3394 		for (int i = 0; i < sc->tx_num_queues; i++, tx_que++) {
3395 			struct tx_ring *txr = &tx_que->txr;
3396 			if (txr->tx_rsq == NULL)
3397 				break;
3398 
3399 			free(txr->tx_rsq, M_DEVBUF);
3400 			txr->tx_rsq = NULL;
3401 		}
3402 		free(sc->tx_queues, M_DEVBUF);
3403 		sc->tx_queues = NULL;
3404 	}
3405 
3406 	if (rx_que != NULL) {
3407 		free(sc->rx_queues, M_DEVBUF);
3408 		sc->rx_queues = NULL;
3409 	}
3410 }
3411 
3412 /*********************************************************************
3413  *
3414  *  Enable transmit unit.
3415  *
3416  **********************************************************************/
3417 static void
3418 em_initialize_transmit_unit(if_ctx_t ctx)
3419 {
3420 	struct e1000_softc *sc = iflib_get_softc(ctx);
3421 	if_softc_ctx_t scctx = sc->shared;
3422 	struct em_tx_queue *que;
3423 	struct tx_ring	*txr;
3424 	struct e1000_hw	*hw = &sc->hw;
3425 	u32 tctl, txdctl = 0, tarc, tipg = 0;
3426 
3427 	INIT_DEBUGOUT("em_initialize_transmit_unit: begin");
3428 
3429 	for (int i = 0; i < sc->tx_num_queues; i++, txr++) {
3430 		u64 bus_addr;
3431 		caddr_t offp, endp;
3432 
3433 		que = &sc->tx_queues[i];
3434 		txr = &que->txr;
3435 		bus_addr = txr->tx_paddr;
3436 
3437 		/* Clear checksum offload context. */
3438 		offp = (caddr_t)&txr->csum_flags;
3439 		endp = (caddr_t)(txr + 1);
3440 		bzero(offp, endp - offp);
3441 
3442 		/* Base and Len of TX Ring */
3443 		E1000_WRITE_REG(hw, E1000_TDLEN(i),
3444 		    scctx->isc_ntxd[0] * sizeof(struct e1000_tx_desc));
3445 		E1000_WRITE_REG(hw, E1000_TDBAH(i),
3446 		    (u32)(bus_addr >> 32));
3447 		E1000_WRITE_REG(hw, E1000_TDBAL(i),
3448 		    (u32)bus_addr);
3449 		/* Init the HEAD/TAIL indices */
3450 		E1000_WRITE_REG(hw, E1000_TDT(i), 0);
3451 		E1000_WRITE_REG(hw, E1000_TDH(i), 0);
3452 
3453 		HW_DEBUGOUT2("Base = %x, Length = %x\n",
3454 		    E1000_READ_REG(hw, E1000_TDBAL(i)),
3455 		    E1000_READ_REG(hw, E1000_TDLEN(i)));
3456 
3457 		txdctl = 0; /* clear txdctl */
3458 		txdctl |= 0x1f; /* PTHRESH */
3459 		txdctl |= 1 << 8; /* HTHRESH */
3460 		txdctl |= 1 << 16;/* WTHRESH */
3461 		txdctl |= 1 << 22; /* Reserved bit 22 must always be 1 */
3462 		txdctl |= E1000_TXDCTL_GRAN;
3463 		txdctl |= 1 << 25; /* LWTHRESH */
3464 
3465 		E1000_WRITE_REG(hw, E1000_TXDCTL(i), txdctl);
3466 	}
3467 
3468 	/* Set the default values for the Tx Inter Packet Gap timer */
3469 	switch (hw->mac.type) {
3470 	case e1000_80003es2lan:
3471 		tipg = DEFAULT_82543_TIPG_IPGR1;
3472 		tipg |= DEFAULT_80003ES2LAN_TIPG_IPGR2 <<
3473 		    E1000_TIPG_IPGR2_SHIFT;
3474 		break;
3475 	case e1000_82542:
3476 		tipg = DEFAULT_82542_TIPG_IPGT;
3477 		tipg |= DEFAULT_82542_TIPG_IPGR1 << E1000_TIPG_IPGR1_SHIFT;
3478 		tipg |= DEFAULT_82542_TIPG_IPGR2 << E1000_TIPG_IPGR2_SHIFT;
3479 		break;
3480 	default:
3481 		if (hw->phy.media_type == e1000_media_type_fiber ||
3482 		    hw->phy.media_type == e1000_media_type_internal_serdes)
3483 			tipg = DEFAULT_82543_TIPG_IPGT_FIBER;
3484 		else
3485 			tipg = DEFAULT_82543_TIPG_IPGT_COPPER;
3486 		tipg |= DEFAULT_82543_TIPG_IPGR1 << E1000_TIPG_IPGR1_SHIFT;
3487 		tipg |= DEFAULT_82543_TIPG_IPGR2 << E1000_TIPG_IPGR2_SHIFT;
3488 	}
3489 
3490 	if (hw->mac.type < igb_mac_min) {
3491 		E1000_WRITE_REG(hw, E1000_TIPG, tipg);
3492 		E1000_WRITE_REG(hw, E1000_TIDV, sc->tx_int_delay.value);
3493 
3494 		if (sc->tx_int_delay.value > 0)
3495 			sc->txd_cmd |= E1000_TXD_CMD_IDE;
3496 	}
3497 
3498 	if (hw->mac.type >= e1000_82540)
3499 		E1000_WRITE_REG(hw, E1000_TADV, sc->tx_abs_int_delay.value);
3500 
3501 	if (hw->mac.type == e1000_82571 || hw->mac.type == e1000_82572) {
3502 		tarc = E1000_READ_REG(hw, E1000_TARC(0));
3503 		tarc |= TARC_SPEED_MODE_BIT;
3504 		E1000_WRITE_REG(hw, E1000_TARC(0), tarc);
3505 	} else if (hw->mac.type == e1000_80003es2lan) {
3506 		/* errata: program both queues to unweighted RR */
3507 		tarc = E1000_READ_REG(hw, E1000_TARC(0));
3508 		tarc |= 1;
3509 		E1000_WRITE_REG(hw, E1000_TARC(0), tarc);
3510 		tarc = E1000_READ_REG(hw, E1000_TARC(1));
3511 		tarc |= 1;
3512 		E1000_WRITE_REG(hw, E1000_TARC(1), tarc);
3513 	} else if (hw->mac.type == e1000_82574) {
3514 		tarc = E1000_READ_REG(hw, E1000_TARC(0));
3515 		tarc |= TARC_ERRATA_BIT;
3516 		if ( sc->tx_num_queues > 1) {
3517 			tarc |= (TARC_COMPENSATION_MODE | TARC_MQ_FIX);
3518 			E1000_WRITE_REG(hw, E1000_TARC(0), tarc);
3519 			E1000_WRITE_REG(hw, E1000_TARC(1), tarc);
3520 		} else
3521 			E1000_WRITE_REG(hw, E1000_TARC(0), tarc);
3522 	}
3523 
3524 	/* Program the Transmit Control Register */
3525 	tctl = E1000_READ_REG(hw, E1000_TCTL);
3526 	tctl &= ~E1000_TCTL_CT;
3527 	tctl |= (E1000_TCTL_PSP | E1000_TCTL_RTLC | E1000_TCTL_EN |
3528 		   (E1000_COLLISION_THRESHOLD << E1000_CT_SHIFT));
3529 
3530 	if (hw->mac.type >= e1000_82571 && hw->mac.type < igb_mac_min)
3531 		tctl |= E1000_TCTL_MULR;
3532 
3533 	/* This write will effectively turn on the transmit unit. */
3534 	E1000_WRITE_REG(hw, E1000_TCTL, tctl);
3535 
3536 	/* SPT and KBL errata workarounds */
3537 	if (hw->mac.type == e1000_pch_spt) {
3538 		u32 reg;
3539 		reg = E1000_READ_REG(hw, E1000_IOSFPC);
3540 		reg |= E1000_RCTL_RDMTS_HEX;
3541 		E1000_WRITE_REG(hw, E1000_IOSFPC, reg);
3542 		/* i218-i219 Specification Update 1.5.4.5 */
3543 		reg = E1000_READ_REG(hw, E1000_TARC(0));
3544 		reg &= ~E1000_TARC0_CB_MULTIQ_3_REQ;
3545 		reg |= E1000_TARC0_CB_MULTIQ_2_REQ;
3546 		E1000_WRITE_REG(hw, E1000_TARC(0), reg);
3547 	}
3548 }
3549 
3550 /*********************************************************************
3551  *
3552  *  Enable receive unit.
3553  *
3554  **********************************************************************/
3555 #define BSIZEPKT_ROUNDUP ((1<<E1000_SRRCTL_BSIZEPKT_SHIFT)-1)
3556 
3557 static void
3558 em_initialize_receive_unit(if_ctx_t ctx)
3559 {
3560 	struct e1000_softc *sc = iflib_get_softc(ctx);
3561 	if_softc_ctx_t scctx = sc->shared;
3562 	if_t ifp = iflib_get_ifp(ctx);
3563 	struct e1000_hw	*hw = &sc->hw;
3564 	struct em_rx_queue *que;
3565 	int i;
3566 	uint32_t rctl, rxcsum;
3567 
3568 	INIT_DEBUGOUT("em_initialize_receive_units: begin");
3569 
3570 	/*
3571 	 * Make sure receives are disabled while setting
3572 	 * up the descriptor ring
3573 	 */
3574 	rctl = E1000_READ_REG(hw, E1000_RCTL);
3575 	/* Do not disable if ever enabled on this hardware */
3576 	if ((hw->mac.type != e1000_82574) && (hw->mac.type != e1000_82583))
3577 		E1000_WRITE_REG(hw, E1000_RCTL, rctl & ~E1000_RCTL_EN);
3578 
3579 	/* Setup the Receive Control Register */
3580 	rctl &= ~(3 << E1000_RCTL_MO_SHIFT);
3581 	rctl |= E1000_RCTL_EN | E1000_RCTL_BAM |
3582 	    E1000_RCTL_LBM_NO | E1000_RCTL_RDMTS_HALF |
3583 	    (hw->mac.mc_filter_type << E1000_RCTL_MO_SHIFT);
3584 
3585 	/* Do not store bad packets */
3586 	rctl &= ~E1000_RCTL_SBP;
3587 
3588 	/* Enable Long Packet receive */
3589 	if (if_getmtu(ifp) > ETHERMTU)
3590 		rctl |= E1000_RCTL_LPE;
3591 	else
3592 		rctl &= ~E1000_RCTL_LPE;
3593 
3594 	/* Strip the CRC */
3595 	if (!em_disable_crc_stripping)
3596 		rctl |= E1000_RCTL_SECRC;
3597 
3598 	/* lem/em default interrupt moderation */
3599 	if (hw->mac.type < igb_mac_min) {
3600 		if (hw->mac.type >= e1000_82540) {
3601 			E1000_WRITE_REG(hw, E1000_RADV, sc->rx_abs_int_delay.value);
3602 
3603 			/* Set the default interrupt throttling rate */
3604 			E1000_WRITE_REG(hw, E1000_ITR,
3605 			    EM_INTS_TO_ITR(em_max_interrupt_rate));
3606 		}
3607 
3608 		/* XXX TEMPORARY WORKAROUND: on some systems with 82573
3609 		 * long latencies are observed, like Lenovo X60. This
3610 		 * change eliminates the problem, but since having positive
3611 		 * values in RDTR is a known source of problems on other
3612 		 * platforms another solution is being sought.
3613 		 */
3614 		if (hw->mac.type == e1000_82573)
3615 			E1000_WRITE_REG(hw, E1000_RDTR, 0x20);
3616 		else
3617 			E1000_WRITE_REG(hw, E1000_RDTR, sc->rx_int_delay.value);
3618 	}
3619 
3620 	if (hw->mac.type >= em_mac_min) {
3621 		uint32_t rfctl;
3622 		/* Use extended rx descriptor formats */
3623 		rfctl = E1000_READ_REG(hw, E1000_RFCTL);
3624 		rfctl |= E1000_RFCTL_EXTEN;
3625 
3626 		/*
3627 		 * When using MSI-X interrupts we need to throttle
3628 		 * using the EITR register (82574 only)
3629 		 */
3630 		if (hw->mac.type == e1000_82574) {
3631 			for (int i = 0; i < 4; i++)
3632 				E1000_WRITE_REG(hw, E1000_EITR_82574(i),
3633 				    EM_INTS_TO_ITR(em_max_interrupt_rate));
3634 			/* Disable accelerated acknowledge */
3635 			rfctl |= E1000_RFCTL_ACK_DIS;
3636 		}
3637 		E1000_WRITE_REG(hw, E1000_RFCTL, rfctl);
3638 	}
3639 
3640 	/* Set up L3 and L4 csum Rx descriptor offloads */
3641 	rxcsum = E1000_READ_REG(hw, E1000_RXCSUM);
3642 	if (if_getcapenable(ifp) & IFCAP_RXCSUM) {
3643 		rxcsum |= E1000_RXCSUM_TUOFL | E1000_RXCSUM_IPOFL;
3644 		if (hw->mac.type > e1000_82575)
3645 			rxcsum |= E1000_RXCSUM_CRCOFL;
3646 		else if (hw->mac.type < em_mac_min &&
3647 		    if_getcapenable(ifp) & IFCAP_HWCSUM_IPV6)
3648 			rxcsum |= E1000_RXCSUM_IPV6OFL;
3649 	} else {
3650 		rxcsum &= ~(E1000_RXCSUM_IPOFL | E1000_RXCSUM_TUOFL);
3651 		if (hw->mac.type > e1000_82575)
3652 			rxcsum &= ~E1000_RXCSUM_CRCOFL;
3653 		else if (hw->mac.type < em_mac_min)
3654 			rxcsum &= ~E1000_RXCSUM_IPV6OFL;
3655 	}
3656 
3657 	if (sc->rx_num_queues > 1) {
3658 		/* RSS hash needed in the Rx descriptor */
3659 		rxcsum |= E1000_RXCSUM_PCSD;
3660 
3661 		if (hw->mac.type >= igb_mac_min)
3662 			igb_initialize_rss_mapping(sc);
3663 		else
3664 			em_initialize_rss_mapping(sc);
3665 	}
3666 	E1000_WRITE_REG(hw, E1000_RXCSUM, rxcsum);
3667 
3668 	for (i = 0, que = sc->rx_queues; i < sc->rx_num_queues; i++, que++) {
3669 		struct rx_ring *rxr = &que->rxr;
3670 		/* Setup the Base and Length of the Rx Descriptor Ring */
3671 		u64 bus_addr = rxr->rx_paddr;
3672 #if 0
3673 		u32 rdt = sc->rx_num_queues -1;  /* default */
3674 #endif
3675 
3676 		E1000_WRITE_REG(hw, E1000_RDLEN(i),
3677 		    scctx->isc_nrxd[0] * sizeof(union e1000_rx_desc_extended));
3678 		E1000_WRITE_REG(hw, E1000_RDBAH(i), (u32)(bus_addr >> 32));
3679 		E1000_WRITE_REG(hw, E1000_RDBAL(i), (u32)bus_addr);
3680 		/* Setup the Head and Tail Descriptor Pointers */
3681 		E1000_WRITE_REG(hw, E1000_RDH(i), 0);
3682 		E1000_WRITE_REG(hw, E1000_RDT(i), 0);
3683 	}
3684 
3685 	/*
3686 	 * Set PTHRESH for improved jumbo performance
3687 	 * According to 10.2.5.11 of Intel 82574 Datasheet,
3688 	 * RXDCTL(1) is written whenever RXDCTL(0) is written.
3689 	 * Only write to RXDCTL(1) if there is a need for different
3690 	 * settings.
3691 	 */
3692 	if ((hw->mac.type == e1000_ich9lan || hw->mac.type == e1000_pch2lan ||
3693 	    hw->mac.type == e1000_ich10lan) && if_getmtu(ifp) > ETHERMTU) {
3694 		u32 rxdctl = E1000_READ_REG(hw, E1000_RXDCTL(0));
3695 		E1000_WRITE_REG(hw, E1000_RXDCTL(0), rxdctl | 3);
3696 	} else if (hw->mac.type == e1000_82574) {
3697 		for (int i = 0; i < sc->rx_num_queues; i++) {
3698 			u32 rxdctl = E1000_READ_REG(hw, E1000_RXDCTL(i));
3699 			rxdctl |= 0x20; /* PTHRESH */
3700 			rxdctl |= 4 << 8; /* HTHRESH */
3701 			rxdctl |= 4 << 16;/* WTHRESH */
3702 			rxdctl |= 1 << 24; /* Switch to granularity */
3703 			E1000_WRITE_REG(hw, E1000_RXDCTL(i), rxdctl);
3704 		}
3705 	} else if (hw->mac.type >= igb_mac_min) {
3706 		u32 psize, srrctl = 0;
3707 
3708 		if (if_getmtu(ifp) > ETHERMTU) {
3709 			psize = scctx->isc_max_frame_size;
3710 			/* are we on a vlan? */
3711 			if (if_vlantrunkinuse(ifp))
3712 				psize += VLAN_TAG_SIZE;
3713 
3714 			if (sc->vf_ifp)
3715 				e1000_rlpml_set_vf(hw, psize);
3716 			else
3717 				E1000_WRITE_REG(hw, E1000_RLPML, psize);
3718 		}
3719 
3720 		/* Set maximum packet buffer len */
3721 		srrctl |= (sc->rx_mbuf_sz + BSIZEPKT_ROUNDUP) >>
3722 		    E1000_SRRCTL_BSIZEPKT_SHIFT;
3723 
3724 		/*
3725 		 * If TX flow control is disabled and there's >1 queue defined,
3726 		 * enable DROP.
3727 		 *
3728 		 * This drops frames rather than hanging the RX MAC for all queues.
3729 		 */
3730 		if ((sc->rx_num_queues > 1) &&
3731 		    (sc->fc == e1000_fc_none ||
3732 		     sc->fc == e1000_fc_rx_pause)) {
3733 			srrctl |= E1000_SRRCTL_DROP_EN;
3734 		}
3735 			/* Setup the Base and Length of the Rx Descriptor Rings */
3736 		for (i = 0, que = sc->rx_queues; i < sc->rx_num_queues; i++, que++) {
3737 			struct rx_ring *rxr = &que->rxr;
3738 			u64 bus_addr = rxr->rx_paddr;
3739 			u32 rxdctl;
3740 
3741 #ifdef notyet
3742 			/* Configure for header split? -- ignore for now */
3743 			rxr->hdr_split = igb_header_split;
3744 #else
3745 			srrctl |= E1000_SRRCTL_DESCTYPE_ADV_ONEBUF;
3746 #endif
3747 
3748 			E1000_WRITE_REG(hw, E1000_RDLEN(i),
3749 					scctx->isc_nrxd[0] * sizeof(struct e1000_rx_desc));
3750 			E1000_WRITE_REG(hw, E1000_RDBAH(i),
3751 					(uint32_t)(bus_addr >> 32));
3752 			E1000_WRITE_REG(hw, E1000_RDBAL(i),
3753 					(uint32_t)bus_addr);
3754 			E1000_WRITE_REG(hw, E1000_SRRCTL(i), srrctl);
3755 			/* Enable this Queue */
3756 			rxdctl = E1000_READ_REG(hw, E1000_RXDCTL(i));
3757 			rxdctl |= E1000_RXDCTL_QUEUE_ENABLE;
3758 			rxdctl &= 0xFFF00000;
3759 			rxdctl |= IGB_RX_PTHRESH;
3760 			rxdctl |= IGB_RX_HTHRESH << 8;
3761 			rxdctl |= IGB_RX_WTHRESH << 16;
3762 			E1000_WRITE_REG(hw, E1000_RXDCTL(i), rxdctl);
3763 		}
3764 	} else if (hw->mac.type >= e1000_pch2lan) {
3765 		if (if_getmtu(ifp) > ETHERMTU)
3766 			e1000_lv_jumbo_workaround_ich8lan(hw, true);
3767 		else
3768 			e1000_lv_jumbo_workaround_ich8lan(hw, false);
3769 	}
3770 
3771 	/* Make sure VLAN Filters are off */
3772 	rctl &= ~E1000_RCTL_VFE;
3773 
3774 	/* Set up packet buffer size, overridden by per queue srrctl on igb */
3775 	if (hw->mac.type < igb_mac_min) {
3776 		if (sc->rx_mbuf_sz > 2048 && sc->rx_mbuf_sz <= 4096)
3777 			rctl |= E1000_RCTL_SZ_4096 | E1000_RCTL_BSEX;
3778 		else if (sc->rx_mbuf_sz > 4096 && sc->rx_mbuf_sz <= 8192)
3779 			rctl |= E1000_RCTL_SZ_8192 | E1000_RCTL_BSEX;
3780 		else if (sc->rx_mbuf_sz > 8192)
3781 			rctl |= E1000_RCTL_SZ_16384 | E1000_RCTL_BSEX;
3782 		else {
3783 			rctl |= E1000_RCTL_SZ_2048;
3784 			rctl &= ~E1000_RCTL_BSEX;
3785 		}
3786 	} else
3787 		rctl |= E1000_RCTL_SZ_2048;
3788 
3789 	/*
3790 	 * rctl bits 11:10 are as follows
3791 	 * lem: reserved
3792 	 * em: DTYPE
3793 	 * igb: reserved
3794 	 * and should be 00 on all of the above
3795 	 */
3796 	rctl &= ~0x00000C00;
3797 
3798 	/* Write out the settings */
3799 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
3800 
3801 	return;
3802 }
3803 
3804 static void
3805 em_if_vlan_register(if_ctx_t ctx, u16 vtag)
3806 {
3807 	struct e1000_softc *sc = iflib_get_softc(ctx);
3808 	u32 index, bit;
3809 
3810 	index = (vtag >> 5) & 0x7F;
3811 	bit = vtag & 0x1F;
3812 	sc->shadow_vfta[index] |= (1 << bit);
3813 	++sc->num_vlans;
3814 	em_if_vlan_filter_write(sc);
3815 }
3816 
3817 static void
3818 em_if_vlan_unregister(if_ctx_t ctx, u16 vtag)
3819 {
3820 	struct e1000_softc *sc = iflib_get_softc(ctx);
3821 	u32 index, bit;
3822 
3823 	index = (vtag >> 5) & 0x7F;
3824 	bit = vtag & 0x1F;
3825 	sc->shadow_vfta[index] &= ~(1 << bit);
3826 	--sc->num_vlans;
3827 	em_if_vlan_filter_write(sc);
3828 }
3829 
3830 static bool
3831 em_if_vlan_filter_capable(if_ctx_t ctx)
3832 {
3833 	if_t ifp = iflib_get_ifp(ctx);
3834 
3835 	if ((if_getcapenable(ifp) & IFCAP_VLAN_HWFILTER) &&
3836 	    !em_disable_crc_stripping)
3837 		return (true);
3838 
3839 	return (false);
3840 }
3841 
3842 static bool
3843 em_if_vlan_filter_used(if_ctx_t ctx)
3844 {
3845 	struct e1000_softc *sc = iflib_get_softc(ctx);
3846 
3847 	if (!em_if_vlan_filter_capable(ctx))
3848 		return (false);
3849 
3850 	for (int i = 0; i < EM_VFTA_SIZE; i++)
3851 		if (sc->shadow_vfta[i] != 0)
3852 			return (true);
3853 
3854 	return (false);
3855 }
3856 
3857 static void
3858 em_if_vlan_filter_enable(struct e1000_softc *sc)
3859 {
3860 	struct e1000_hw *hw = &sc->hw;
3861 	u32 reg;
3862 
3863 	reg = E1000_READ_REG(hw, E1000_RCTL);
3864 	reg &= ~E1000_RCTL_CFIEN;
3865 	reg |= E1000_RCTL_VFE;
3866 	E1000_WRITE_REG(hw, E1000_RCTL, reg);
3867 }
3868 
3869 static void
3870 em_if_vlan_filter_disable(struct e1000_softc *sc)
3871 {
3872 	struct e1000_hw *hw = &sc->hw;
3873 	u32 reg;
3874 
3875 	reg = E1000_READ_REG(hw, E1000_RCTL);
3876 	reg &= ~(E1000_RCTL_VFE | E1000_RCTL_CFIEN);
3877 	E1000_WRITE_REG(hw, E1000_RCTL, reg);
3878 }
3879 
3880 static void
3881 em_if_vlan_filter_write(struct e1000_softc *sc)
3882 {
3883 	struct e1000_hw *hw = &sc->hw;
3884 
3885 	if (sc->vf_ifp)
3886 		return;
3887 
3888 	/* Disable interrupts for lem-class devices during the filter change */
3889 	if (hw->mac.type < em_mac_min)
3890 		em_if_intr_disable(sc->ctx);
3891 
3892 	for (int i = 0; i < EM_VFTA_SIZE; i++)
3893 		if (sc->shadow_vfta[i] != 0) {
3894 			/* XXXKB: incomplete VF support, we return early above */
3895 			if (sc->vf_ifp)
3896 				e1000_vfta_set_vf(hw, sc->shadow_vfta[i], true);
3897 			else
3898 				e1000_write_vfta(hw, i, sc->shadow_vfta[i]);
3899 		}
3900 
3901 	/* Re-enable interrupts for lem-class devices */
3902 	if (hw->mac.type < em_mac_min)
3903 		em_if_intr_enable(sc->ctx);
3904 }
3905 
3906 static void
3907 em_setup_vlan_hw_support(if_ctx_t ctx)
3908 {
3909 	struct e1000_softc *sc = iflib_get_softc(ctx);
3910 	struct e1000_hw *hw = &sc->hw;
3911 	if_t ifp = iflib_get_ifp(ctx);
3912 	u32 reg;
3913 
3914 	/* XXXKB: Return early if we are a VF until VF decap and filter management
3915 	 * is ready and tested.
3916 	 */
3917 	if (sc->vf_ifp)
3918 		return;
3919 
3920 	if (if_getcapenable(ifp) & IFCAP_VLAN_HWTAGGING &&
3921 	    !em_disable_crc_stripping) {
3922 		reg = E1000_READ_REG(hw, E1000_CTRL);
3923 		reg |= E1000_CTRL_VME;
3924 		E1000_WRITE_REG(hw, E1000_CTRL, reg);
3925 	} else {
3926 		reg = E1000_READ_REG(hw, E1000_CTRL);
3927 		reg &= ~E1000_CTRL_VME;
3928 		E1000_WRITE_REG(hw, E1000_CTRL, reg);
3929 	}
3930 
3931 	/* If we aren't doing HW filtering, we're done */
3932 	if (!em_if_vlan_filter_capable(ctx))  {
3933 		em_if_vlan_filter_disable(sc);
3934 		return;
3935 	}
3936 
3937 	/*
3938 	 * A soft reset zero's out the VFTA, so
3939 	 * we need to repopulate it now.
3940 	 * We also insert VLAN 0 in the filter list, so we pass VLAN 0 tagged
3941 	 * traffic through. This will write the entire table.
3942 	 */
3943 	em_if_vlan_register(ctx, 0);
3944 
3945 	/* Enable the Filter Table */
3946 	em_if_vlan_filter_enable(sc);
3947 }
3948 
3949 static void
3950 em_if_intr_enable(if_ctx_t ctx)
3951 {
3952 	struct e1000_softc *sc = iflib_get_softc(ctx);
3953 	struct e1000_hw *hw = &sc->hw;
3954 	u32 ims_mask = IMS_ENABLE_MASK;
3955 
3956 	if (sc->intr_type == IFLIB_INTR_MSIX) {
3957 		E1000_WRITE_REG(hw, EM_EIAC, sc->ims);
3958 		ims_mask |= sc->ims;
3959 	}
3960 
3961 	E1000_WRITE_REG(hw, E1000_IMS, ims_mask);
3962 	E1000_WRITE_FLUSH(hw);
3963 }
3964 
3965 static void
3966 em_if_intr_disable(if_ctx_t ctx)
3967 {
3968 	struct e1000_softc *sc = iflib_get_softc(ctx);
3969 	struct e1000_hw *hw = &sc->hw;
3970 
3971 	if (sc->intr_type == IFLIB_INTR_MSIX)
3972 		E1000_WRITE_REG(hw, EM_EIAC, 0);
3973 	E1000_WRITE_REG(hw, E1000_IMC, 0xffffffff);
3974 	E1000_WRITE_FLUSH(hw);
3975 }
3976 
3977 static void
3978 igb_if_intr_enable(if_ctx_t ctx)
3979 {
3980 	struct e1000_softc *sc = iflib_get_softc(ctx);
3981 	struct e1000_hw *hw = &sc->hw;
3982 	u32 mask;
3983 
3984 	if (__predict_true(sc->intr_type == IFLIB_INTR_MSIX)) {
3985 		mask = (sc->que_mask | sc->link_mask);
3986 		E1000_WRITE_REG(hw, E1000_EIAC, mask);
3987 		E1000_WRITE_REG(hw, E1000_EIAM, mask);
3988 		E1000_WRITE_REG(hw, E1000_EIMS, mask);
3989 		E1000_WRITE_REG(hw, E1000_IMS, E1000_IMS_LSC);
3990 	} else
3991 		E1000_WRITE_REG(hw, E1000_IMS, IMS_ENABLE_MASK);
3992 	E1000_WRITE_FLUSH(hw);
3993 }
3994 
3995 static void
3996 igb_if_intr_disable(if_ctx_t ctx)
3997 {
3998 	struct e1000_softc *sc = iflib_get_softc(ctx);
3999 	struct e1000_hw *hw = &sc->hw;
4000 
4001 	if (__predict_true(sc->intr_type == IFLIB_INTR_MSIX)) {
4002 		E1000_WRITE_REG(hw, E1000_EIMC, 0xffffffff);
4003 		E1000_WRITE_REG(hw, E1000_EIAC, 0);
4004 	}
4005 	E1000_WRITE_REG(hw, E1000_IMC, 0xffffffff);
4006 	E1000_WRITE_FLUSH(hw);
4007 }
4008 
4009 /*
4010  * Bit of a misnomer, what this really means is
4011  * to enable OS management of the system... aka
4012  * to disable special hardware management features
4013  */
4014 static void
4015 em_init_manageability(struct e1000_softc *sc)
4016 {
4017 	/* A shared code workaround */
4018 #define E1000_82542_MANC2H E1000_MANC2H
4019 	if (sc->has_manage) {
4020 		int manc2h = E1000_READ_REG(&sc->hw, E1000_MANC2H);
4021 		int manc = E1000_READ_REG(&sc->hw, E1000_MANC);
4022 
4023 		/* disable hardware interception of ARP */
4024 		manc &= ~(E1000_MANC_ARP_EN);
4025 
4026 		/* enable receiving management packets to the host */
4027 		manc |= E1000_MANC_EN_MNG2HOST;
4028 #define E1000_MNG2HOST_PORT_623 (1 << 5)
4029 #define E1000_MNG2HOST_PORT_664 (1 << 6)
4030 		manc2h |= E1000_MNG2HOST_PORT_623;
4031 		manc2h |= E1000_MNG2HOST_PORT_664;
4032 		E1000_WRITE_REG(&sc->hw, E1000_MANC2H, manc2h);
4033 		E1000_WRITE_REG(&sc->hw, E1000_MANC, manc);
4034 	}
4035 }
4036 
4037 /*
4038  * Give control back to hardware management
4039  * controller if there is one.
4040  */
4041 static void
4042 em_release_manageability(struct e1000_softc *sc)
4043 {
4044 	if (sc->has_manage) {
4045 		int manc = E1000_READ_REG(&sc->hw, E1000_MANC);
4046 
4047 		/* re-enable hardware interception of ARP */
4048 		manc |= E1000_MANC_ARP_EN;
4049 		manc &= ~E1000_MANC_EN_MNG2HOST;
4050 
4051 		E1000_WRITE_REG(&sc->hw, E1000_MANC, manc);
4052 	}
4053 }
4054 
4055 /*
4056  * em_get_hw_control sets the {CTRL_EXT|FWSM}:DRV_LOAD bit.
4057  * For ASF and Pass Through versions of f/w this means
4058  * that the driver is loaded. For AMT version type f/w
4059  * this means that the network i/f is open.
4060  */
4061 static void
4062 em_get_hw_control(struct e1000_softc *sc)
4063 {
4064 	u32 ctrl_ext, swsm;
4065 
4066 	if (sc->vf_ifp)
4067 		return;
4068 
4069 	if (sc->hw.mac.type == e1000_82573) {
4070 		swsm = E1000_READ_REG(&sc->hw, E1000_SWSM);
4071 		E1000_WRITE_REG(&sc->hw, E1000_SWSM,
4072 		    swsm | E1000_SWSM_DRV_LOAD);
4073 		return;
4074 	}
4075 	/* else */
4076 	ctrl_ext = E1000_READ_REG(&sc->hw, E1000_CTRL_EXT);
4077 	E1000_WRITE_REG(&sc->hw, E1000_CTRL_EXT,
4078 	    ctrl_ext | E1000_CTRL_EXT_DRV_LOAD);
4079 }
4080 
4081 /*
4082  * em_release_hw_control resets {CTRL_EXT|FWSM}:DRV_LOAD bit.
4083  * For ASF and Pass Through versions of f/w this means that
4084  * the driver is no longer loaded. For AMT versions of the
4085  * f/w this means that the network i/f is closed.
4086  */
4087 static void
4088 em_release_hw_control(struct e1000_softc *sc)
4089 {
4090 	u32 ctrl_ext, swsm;
4091 
4092 	if (!sc->has_manage)
4093 		return;
4094 
4095 	if (sc->hw.mac.type == e1000_82573) {
4096 		swsm = E1000_READ_REG(&sc->hw, E1000_SWSM);
4097 		E1000_WRITE_REG(&sc->hw, E1000_SWSM,
4098 		    swsm & ~E1000_SWSM_DRV_LOAD);
4099 		return;
4100 	}
4101 	/* else */
4102 	ctrl_ext = E1000_READ_REG(&sc->hw, E1000_CTRL_EXT);
4103 	E1000_WRITE_REG(&sc->hw, E1000_CTRL_EXT,
4104 	    ctrl_ext & ~E1000_CTRL_EXT_DRV_LOAD);
4105 	return;
4106 }
4107 
4108 static int
4109 em_is_valid_ether_addr(u8 *addr)
4110 {
4111 	char zero_addr[6] = { 0, 0, 0, 0, 0, 0 };
4112 
4113 	if ((addr[0] & 1) || (!bcmp(addr, zero_addr, ETHER_ADDR_LEN))) {
4114 		return (false);
4115 	}
4116 
4117 	return (true);
4118 }
4119 
4120 static bool
4121 em_automask_tso(if_ctx_t ctx)
4122 {
4123 	struct e1000_softc *sc = iflib_get_softc(ctx);
4124 	if_softc_ctx_t scctx = iflib_get_softc_ctx(ctx);
4125 	if_t ifp = iflib_get_ifp(ctx);
4126 
4127 	if (!em_unsupported_tso && sc->link_speed &&
4128 	    sc->link_speed != SPEED_1000 && scctx->isc_capenable & IFCAP_TSO) {
4129 		device_printf(sc->dev, "Disabling TSO for 10/100 Ethernet.\n");
4130 		sc->tso_automasked = scctx->isc_capenable & IFCAP_TSO;
4131 		scctx->isc_capenable &= ~IFCAP_TSO;
4132 		if_setcapenablebit(ifp, 0, IFCAP_TSO);
4133 		/* iflib_init_locked handles ifnet hwassistbits */
4134 		iflib_request_reset(ctx);
4135 		return true;
4136 	} else if (sc->link_speed == SPEED_1000 && sc->tso_automasked) {
4137 		device_printf(sc->dev, "Re-enabling TSO for GbE.\n");
4138 		scctx->isc_capenable |= sc->tso_automasked;
4139 		if_setcapenablebit(ifp, sc->tso_automasked, 0);
4140 		sc->tso_automasked = 0;
4141 		/* iflib_init_locked handles ifnet hwassistbits */
4142 		iflib_request_reset(ctx);
4143 		return true;
4144 	}
4145 
4146 	return false;
4147 }
4148 
4149 /*
4150 ** Parse the interface capabilities with regard
4151 ** to both system management and wake-on-lan for
4152 ** later use.
4153 */
4154 static void
4155 em_get_wakeup(if_ctx_t ctx)
4156 {
4157 	struct e1000_softc *sc = iflib_get_softc(ctx);
4158 	device_t dev = iflib_get_dev(ctx);
4159 	u16 eeprom_data = 0, device_id, apme_mask;
4160 
4161 	sc->has_manage = e1000_enable_mng_pass_thru(&sc->hw);
4162 	apme_mask = EM_EEPROM_APME;
4163 
4164 	switch (sc->hw.mac.type) {
4165 	case e1000_82542:
4166 	case e1000_82543:
4167 		break;
4168 	case e1000_82544:
4169 		e1000_read_nvm(&sc->hw,
4170 		    NVM_INIT_CONTROL2_REG, 1, &eeprom_data);
4171 		apme_mask = EM_82544_APME;
4172 		break;
4173 	case e1000_82546:
4174 	case e1000_82546_rev_3:
4175 		if (sc->hw.bus.func == 1) {
4176 			e1000_read_nvm(&sc->hw,
4177 			    NVM_INIT_CONTROL3_PORT_B, 1, &eeprom_data);
4178 			break;
4179 		} else
4180 			e1000_read_nvm(&sc->hw,
4181 			    NVM_INIT_CONTROL3_PORT_A, 1, &eeprom_data);
4182 		break;
4183 	case e1000_82573:
4184 	case e1000_82583:
4185 		sc->has_amt = true;
4186 		/* FALLTHROUGH */
4187 	case e1000_82571:
4188 	case e1000_82572:
4189 	case e1000_80003es2lan:
4190 		if (sc->hw.bus.func == 1) {
4191 			e1000_read_nvm(&sc->hw,
4192 			    NVM_INIT_CONTROL3_PORT_B, 1, &eeprom_data);
4193 			break;
4194 		} else
4195 			e1000_read_nvm(&sc->hw,
4196 			    NVM_INIT_CONTROL3_PORT_A, 1, &eeprom_data);
4197 		break;
4198 	case e1000_ich8lan:
4199 	case e1000_ich9lan:
4200 	case e1000_ich10lan:
4201 	case e1000_pchlan:
4202 	case e1000_pch2lan:
4203 	case e1000_pch_lpt:
4204 	case e1000_pch_spt:
4205 	case e1000_82575:	/* listing all igb devices */
4206 	case e1000_82576:
4207 	case e1000_82580:
4208 	case e1000_i350:
4209 	case e1000_i354:
4210 	case e1000_i210:
4211 	case e1000_i211:
4212 	case e1000_vfadapt:
4213 	case e1000_vfadapt_i350:
4214 		apme_mask = E1000_WUC_APME;
4215 		sc->has_amt = true;
4216 		eeprom_data = E1000_READ_REG(&sc->hw, E1000_WUC);
4217 		break;
4218 	default:
4219 		e1000_read_nvm(&sc->hw,
4220 		    NVM_INIT_CONTROL3_PORT_A, 1, &eeprom_data);
4221 		break;
4222 	}
4223 	if (eeprom_data & apme_mask)
4224 		sc->wol = (E1000_WUFC_MAG | E1000_WUFC_MC);
4225 	/*
4226 	 * We have the eeprom settings, now apply the special cases
4227 	 * where the eeprom may be wrong or the board won't support
4228 	 * wake on lan on a particular port
4229 	 */
4230 	device_id = pci_get_device(dev);
4231 	switch (device_id) {
4232 	case E1000_DEV_ID_82546GB_PCIE:
4233 		sc->wol = 0;
4234 		break;
4235 	case E1000_DEV_ID_82546EB_FIBER:
4236 	case E1000_DEV_ID_82546GB_FIBER:
4237 		/* Wake events only supported on port A for dual fiber
4238 		 * regardless of eeprom setting */
4239 		if (E1000_READ_REG(&sc->hw, E1000_STATUS) &
4240 		    E1000_STATUS_FUNC_1)
4241 			sc->wol = 0;
4242 		break;
4243 	case E1000_DEV_ID_82546GB_QUAD_COPPER_KSP3:
4244 		/* if quad port adapter, disable WoL on all but port A */
4245 		if (global_quad_port_a != 0)
4246 			sc->wol = 0;
4247 		/* Reset for multiple quad port adapters */
4248 		if (++global_quad_port_a == 4)
4249 			global_quad_port_a = 0;
4250 		break;
4251 	case E1000_DEV_ID_82571EB_FIBER:
4252 		/* Wake events only supported on port A for dual fiber
4253 		 * regardless of eeprom setting */
4254 		if (E1000_READ_REG(&sc->hw, E1000_STATUS) &
4255 		    E1000_STATUS_FUNC_1)
4256 			sc->wol = 0;
4257 		break;
4258 	case E1000_DEV_ID_82571EB_QUAD_COPPER:
4259 	case E1000_DEV_ID_82571EB_QUAD_FIBER:
4260 	case E1000_DEV_ID_82571EB_QUAD_COPPER_LP:
4261 		/* if quad port adapter, disable WoL on all but port A */
4262 		if (global_quad_port_a != 0)
4263 			sc->wol = 0;
4264 		/* Reset for multiple quad port adapters */
4265 		if (++global_quad_port_a == 4)
4266 			global_quad_port_a = 0;
4267 		break;
4268 	}
4269 	return;
4270 }
4271 
4272 
4273 /*
4274  * Enable PCI Wake On Lan capability
4275  */
4276 static void
4277 em_enable_wakeup(if_ctx_t ctx)
4278 {
4279 	struct e1000_softc *sc = iflib_get_softc(ctx);
4280 	device_t dev = iflib_get_dev(ctx);
4281 	if_t ifp = iflib_get_ifp(ctx);
4282 	int error = 0;
4283 	u32 pmc, ctrl, ctrl_ext, rctl;
4284 	u16 status;
4285 
4286 	if (pci_find_cap(dev, PCIY_PMG, &pmc) != 0)
4287 		return;
4288 
4289 	/*
4290 	 * Determine type of Wakeup: note that wol
4291 	 * is set with all bits on by default.
4292 	 */
4293 	if ((if_getcapenable(ifp) & IFCAP_WOL_MAGIC) == 0)
4294 		sc->wol &= ~E1000_WUFC_MAG;
4295 
4296 	if ((if_getcapenable(ifp) & IFCAP_WOL_UCAST) == 0)
4297 		sc->wol &= ~E1000_WUFC_EX;
4298 
4299 	if ((if_getcapenable(ifp) & IFCAP_WOL_MCAST) == 0)
4300 		sc->wol &= ~E1000_WUFC_MC;
4301 	else {
4302 		rctl = E1000_READ_REG(&sc->hw, E1000_RCTL);
4303 		rctl |= E1000_RCTL_MPE;
4304 		E1000_WRITE_REG(&sc->hw, E1000_RCTL, rctl);
4305 	}
4306 
4307 	if (!(sc->wol & (E1000_WUFC_EX | E1000_WUFC_MAG | E1000_WUFC_MC)))
4308 		goto pme;
4309 
4310 	/* Advertise the wakeup capability */
4311 	ctrl = E1000_READ_REG(&sc->hw, E1000_CTRL);
4312 	ctrl |= (E1000_CTRL_SWDPIN2 | E1000_CTRL_SWDPIN3);
4313 	E1000_WRITE_REG(&sc->hw, E1000_CTRL, ctrl);
4314 
4315 	/* Keep the laser running on Fiber adapters */
4316 	if (sc->hw.phy.media_type == e1000_media_type_fiber ||
4317 	    sc->hw.phy.media_type == e1000_media_type_internal_serdes) {
4318 		ctrl_ext = E1000_READ_REG(&sc->hw, E1000_CTRL_EXT);
4319 		ctrl_ext |= E1000_CTRL_EXT_SDP3_DATA;
4320 		E1000_WRITE_REG(&sc->hw, E1000_CTRL_EXT, ctrl_ext);
4321 	}
4322 
4323 	if ((sc->hw.mac.type == e1000_ich8lan) ||
4324 	    (sc->hw.mac.type == e1000_pchlan) ||
4325 	    (sc->hw.mac.type == e1000_ich9lan) ||
4326 	    (sc->hw.mac.type == e1000_ich10lan))
4327 		e1000_suspend_workarounds_ich8lan(&sc->hw);
4328 
4329 	if ( sc->hw.mac.type >= e1000_pchlan) {
4330 		error = em_enable_phy_wakeup(sc);
4331 		if (error)
4332 			goto pme;
4333 	} else {
4334 		/* Enable wakeup by the MAC */
4335 		E1000_WRITE_REG(&sc->hw, E1000_WUC, E1000_WUC_PME_EN);
4336 		E1000_WRITE_REG(&sc->hw, E1000_WUFC, sc->wol);
4337 	}
4338 
4339 	if (sc->hw.phy.type == e1000_phy_igp_3)
4340 		e1000_igp3_phy_powerdown_workaround_ich8lan(&sc->hw);
4341 
4342 pme:
4343 	status = pci_read_config(dev, pmc + PCIR_POWER_STATUS, 2);
4344 	status &= ~(PCIM_PSTAT_PME | PCIM_PSTAT_PMEENABLE);
4345 	if (!error && (if_getcapenable(ifp) & IFCAP_WOL))
4346 		status |= PCIM_PSTAT_PME | PCIM_PSTAT_PMEENABLE;
4347 	pci_write_config(dev, pmc + PCIR_POWER_STATUS, status, 2);
4348 
4349 	return;
4350 }
4351 
4352 /*
4353  * WOL in the newer chipset interfaces (pchlan)
4354  * require thing to be copied into the phy
4355  */
4356 static int
4357 em_enable_phy_wakeup(struct e1000_softc *sc)
4358 {
4359 	struct e1000_hw *hw = &sc->hw;
4360 	u32 mreg, ret = 0;
4361 	u16 preg;
4362 
4363 	/* copy MAC RARs to PHY RARs */
4364 	e1000_copy_rx_addrs_to_phy_ich8lan(hw);
4365 
4366 	/* copy MAC MTA to PHY MTA */
4367 	for (int i = 0; i < hw->mac.mta_reg_count; i++) {
4368 		mreg = E1000_READ_REG_ARRAY(hw, E1000_MTA, i);
4369 		e1000_write_phy_reg(hw, BM_MTA(i), (u16)(mreg & 0xFFFF));
4370 		e1000_write_phy_reg(hw, BM_MTA(i) + 1,
4371 		    (u16)((mreg >> 16) & 0xFFFF));
4372 	}
4373 
4374 	/* configure PHY Rx Control register */
4375 	e1000_read_phy_reg(hw, BM_RCTL, &preg);
4376 	mreg = E1000_READ_REG(hw, E1000_RCTL);
4377 	if (mreg & E1000_RCTL_UPE)
4378 		preg |= BM_RCTL_UPE;
4379 	if (mreg & E1000_RCTL_MPE)
4380 		preg |= BM_RCTL_MPE;
4381 	preg &= ~(BM_RCTL_MO_MASK);
4382 	if (mreg & E1000_RCTL_MO_3)
4383 		preg |= (((mreg & E1000_RCTL_MO_3) >> E1000_RCTL_MO_SHIFT)
4384 				<< BM_RCTL_MO_SHIFT);
4385 	if (mreg & E1000_RCTL_BAM)
4386 		preg |= BM_RCTL_BAM;
4387 	if (mreg & E1000_RCTL_PMCF)
4388 		preg |= BM_RCTL_PMCF;
4389 	mreg = E1000_READ_REG(hw, E1000_CTRL);
4390 	if (mreg & E1000_CTRL_RFCE)
4391 		preg |= BM_RCTL_RFCE;
4392 	e1000_write_phy_reg(hw, BM_RCTL, preg);
4393 
4394 	/* enable PHY wakeup in MAC register */
4395 	E1000_WRITE_REG(hw, E1000_WUC,
4396 	    E1000_WUC_PHY_WAKE | E1000_WUC_PME_EN | E1000_WUC_APME);
4397 	E1000_WRITE_REG(hw, E1000_WUFC, sc->wol);
4398 
4399 	/* configure and enable PHY wakeup in PHY registers */
4400 	e1000_write_phy_reg(hw, BM_WUFC, sc->wol);
4401 	e1000_write_phy_reg(hw, BM_WUC, E1000_WUC_PME_EN);
4402 
4403 	/* activate PHY wakeup */
4404 	ret = hw->phy.ops.acquire(hw);
4405 	if (ret) {
4406 		printf("Could not acquire PHY\n");
4407 		return ret;
4408 	}
4409 	e1000_write_phy_reg_mdic(hw, IGP01E1000_PHY_PAGE_SELECT,
4410 	                         (BM_WUC_ENABLE_PAGE << IGP_PAGE_SHIFT));
4411 	ret = e1000_read_phy_reg_mdic(hw, BM_WUC_ENABLE_REG, &preg);
4412 	if (ret) {
4413 		printf("Could not read PHY page 769\n");
4414 		goto out;
4415 	}
4416 	preg |= BM_WUC_ENABLE_BIT | BM_WUC_HOST_WU_BIT;
4417 	ret = e1000_write_phy_reg_mdic(hw, BM_WUC_ENABLE_REG, preg);
4418 	if (ret)
4419 		printf("Could not set PHY Host Wakeup bit\n");
4420 out:
4421 	hw->phy.ops.release(hw);
4422 
4423 	return ret;
4424 }
4425 
4426 static void
4427 em_if_led_func(if_ctx_t ctx, int onoff)
4428 {
4429 	struct e1000_softc *sc = iflib_get_softc(ctx);
4430 
4431 	if (onoff) {
4432 		e1000_setup_led(&sc->hw);
4433 		e1000_led_on(&sc->hw);
4434 	} else {
4435 		e1000_led_off(&sc->hw);
4436 		e1000_cleanup_led(&sc->hw);
4437 	}
4438 }
4439 
4440 /*
4441  * Disable the L0S and L1 LINK states
4442  */
4443 static void
4444 em_disable_aspm(struct e1000_softc *sc)
4445 {
4446 	int base, reg;
4447 	u16 link_cap,link_ctrl;
4448 	device_t dev = sc->dev;
4449 
4450 	switch (sc->hw.mac.type) {
4451 	case e1000_82573:
4452 	case e1000_82574:
4453 	case e1000_82583:
4454 		break;
4455 	default:
4456 		return;
4457 	}
4458 	if (pci_find_cap(dev, PCIY_EXPRESS, &base) != 0)
4459 		return;
4460 	reg = base + PCIER_LINK_CAP;
4461 	link_cap = pci_read_config(dev, reg, 2);
4462 	if ((link_cap & PCIEM_LINK_CAP_ASPM) == 0)
4463 		return;
4464 	reg = base + PCIER_LINK_CTL;
4465 	link_ctrl = pci_read_config(dev, reg, 2);
4466 	link_ctrl &= ~PCIEM_LINK_CTL_ASPMC;
4467 	pci_write_config(dev, reg, link_ctrl, 2);
4468 	return;
4469 }
4470 
4471 /**********************************************************************
4472  *
4473  *  Update the board statistics counters.
4474  *
4475  **********************************************************************/
4476 static void
4477 em_update_stats_counters(struct e1000_softc *sc)
4478 {
4479 	u64 prev_xoffrxc = sc->stats.xoffrxc;
4480 
4481 	if(sc->hw.phy.media_type == e1000_media_type_copper ||
4482 	   (E1000_READ_REG(&sc->hw, E1000_STATUS) & E1000_STATUS_LU)) {
4483 		sc->stats.symerrs += E1000_READ_REG(&sc->hw, E1000_SYMERRS);
4484 		sc->stats.sec += E1000_READ_REG(&sc->hw, E1000_SEC);
4485 	}
4486 	sc->stats.crcerrs += E1000_READ_REG(&sc->hw, E1000_CRCERRS);
4487 	sc->stats.mpc += E1000_READ_REG(&sc->hw, E1000_MPC);
4488 	sc->stats.scc += E1000_READ_REG(&sc->hw, E1000_SCC);
4489 	sc->stats.ecol += E1000_READ_REG(&sc->hw, E1000_ECOL);
4490 
4491 	sc->stats.mcc += E1000_READ_REG(&sc->hw, E1000_MCC);
4492 	sc->stats.latecol += E1000_READ_REG(&sc->hw, E1000_LATECOL);
4493 	sc->stats.colc += E1000_READ_REG(&sc->hw, E1000_COLC);
4494 	sc->stats.dc += E1000_READ_REG(&sc->hw, E1000_DC);
4495 	sc->stats.rlec += E1000_READ_REG(&sc->hw, E1000_RLEC);
4496 	sc->stats.xonrxc += E1000_READ_REG(&sc->hw, E1000_XONRXC);
4497 	sc->stats.xontxc += E1000_READ_REG(&sc->hw, E1000_XONTXC);
4498 	sc->stats.xoffrxc += E1000_READ_REG(&sc->hw, E1000_XOFFRXC);
4499 	/*
4500 	 ** For watchdog management we need to know if we have been
4501 	 ** paused during the last interval, so capture that here.
4502 	*/
4503 	if (sc->stats.xoffrxc != prev_xoffrxc)
4504 		sc->shared->isc_pause_frames = 1;
4505 	sc->stats.xofftxc += E1000_READ_REG(&sc->hw, E1000_XOFFTXC);
4506 	sc->stats.fcruc += E1000_READ_REG(&sc->hw, E1000_FCRUC);
4507 	sc->stats.prc64 += E1000_READ_REG(&sc->hw, E1000_PRC64);
4508 	sc->stats.prc127 += E1000_READ_REG(&sc->hw, E1000_PRC127);
4509 	sc->stats.prc255 += E1000_READ_REG(&sc->hw, E1000_PRC255);
4510 	sc->stats.prc511 += E1000_READ_REG(&sc->hw, E1000_PRC511);
4511 	sc->stats.prc1023 += E1000_READ_REG(&sc->hw, E1000_PRC1023);
4512 	sc->stats.prc1522 += E1000_READ_REG(&sc->hw, E1000_PRC1522);
4513 	sc->stats.gprc += E1000_READ_REG(&sc->hw, E1000_GPRC);
4514 	sc->stats.bprc += E1000_READ_REG(&sc->hw, E1000_BPRC);
4515 	sc->stats.mprc += E1000_READ_REG(&sc->hw, E1000_MPRC);
4516 	sc->stats.gptc += E1000_READ_REG(&sc->hw, E1000_GPTC);
4517 
4518 	/* For the 64-bit byte counters the low dword must be read first. */
4519 	/* Both registers clear on the read of the high dword */
4520 
4521 	sc->stats.gorc += E1000_READ_REG(&sc->hw, E1000_GORCL) +
4522 	    ((u64)E1000_READ_REG(&sc->hw, E1000_GORCH) << 32);
4523 	sc->stats.gotc += E1000_READ_REG(&sc->hw, E1000_GOTCL) +
4524 	    ((u64)E1000_READ_REG(&sc->hw, E1000_GOTCH) << 32);
4525 
4526 	sc->stats.rnbc += E1000_READ_REG(&sc->hw, E1000_RNBC);
4527 	sc->stats.ruc += E1000_READ_REG(&sc->hw, E1000_RUC);
4528 	sc->stats.rfc += E1000_READ_REG(&sc->hw, E1000_RFC);
4529 	sc->stats.roc += E1000_READ_REG(&sc->hw, E1000_ROC);
4530 	sc->stats.rjc += E1000_READ_REG(&sc->hw, E1000_RJC);
4531 
4532 	sc->stats.mgprc += E1000_READ_REG(&sc->hw, E1000_MGTPRC);
4533 	sc->stats.mgpdc += E1000_READ_REG(&sc->hw, E1000_MGTPDC);
4534 	sc->stats.mgptc += E1000_READ_REG(&sc->hw, E1000_MGTPTC);
4535 
4536 	sc->stats.tor += E1000_READ_REG(&sc->hw, E1000_TORH);
4537 	sc->stats.tot += E1000_READ_REG(&sc->hw, E1000_TOTH);
4538 
4539 	sc->stats.tpr += E1000_READ_REG(&sc->hw, E1000_TPR);
4540 	sc->stats.tpt += E1000_READ_REG(&sc->hw, E1000_TPT);
4541 	sc->stats.ptc64 += E1000_READ_REG(&sc->hw, E1000_PTC64);
4542 	sc->stats.ptc127 += E1000_READ_REG(&sc->hw, E1000_PTC127);
4543 	sc->stats.ptc255 += E1000_READ_REG(&sc->hw, E1000_PTC255);
4544 	sc->stats.ptc511 += E1000_READ_REG(&sc->hw, E1000_PTC511);
4545 	sc->stats.ptc1023 += E1000_READ_REG(&sc->hw, E1000_PTC1023);
4546 	sc->stats.ptc1522 += E1000_READ_REG(&sc->hw, E1000_PTC1522);
4547 	sc->stats.mptc += E1000_READ_REG(&sc->hw, E1000_MPTC);
4548 	sc->stats.bptc += E1000_READ_REG(&sc->hw, E1000_BPTC);
4549 
4550 	/* Interrupt Counts */
4551 
4552 	sc->stats.iac += E1000_READ_REG(&sc->hw, E1000_IAC);
4553 	sc->stats.icrxptc += E1000_READ_REG(&sc->hw, E1000_ICRXPTC);
4554 	sc->stats.icrxatc += E1000_READ_REG(&sc->hw, E1000_ICRXATC);
4555 	sc->stats.ictxptc += E1000_READ_REG(&sc->hw, E1000_ICTXPTC);
4556 	sc->stats.ictxatc += E1000_READ_REG(&sc->hw, E1000_ICTXATC);
4557 	sc->stats.ictxqec += E1000_READ_REG(&sc->hw, E1000_ICTXQEC);
4558 	sc->stats.ictxqmtc += E1000_READ_REG(&sc->hw, E1000_ICTXQMTC);
4559 	sc->stats.icrxdmtc += E1000_READ_REG(&sc->hw, E1000_ICRXDMTC);
4560 	sc->stats.icrxoc += E1000_READ_REG(&sc->hw, E1000_ICRXOC);
4561 
4562 	if (sc->hw.mac.type >= e1000_82543) {
4563 		sc->stats.algnerrc +=
4564 		E1000_READ_REG(&sc->hw, E1000_ALGNERRC);
4565 		sc->stats.rxerrc +=
4566 		E1000_READ_REG(&sc->hw, E1000_RXERRC);
4567 		sc->stats.tncrs +=
4568 		E1000_READ_REG(&sc->hw, E1000_TNCRS);
4569 		sc->stats.cexterr +=
4570 		E1000_READ_REG(&sc->hw, E1000_CEXTERR);
4571 		sc->stats.tsctc +=
4572 		E1000_READ_REG(&sc->hw, E1000_TSCTC);
4573 		sc->stats.tsctfc +=
4574 		E1000_READ_REG(&sc->hw, E1000_TSCTFC);
4575 	}
4576 }
4577 
4578 static uint64_t
4579 em_if_get_counter(if_ctx_t ctx, ift_counter cnt)
4580 {
4581 	struct e1000_softc *sc = iflib_get_softc(ctx);
4582 	if_t ifp = iflib_get_ifp(ctx);
4583 
4584 	switch (cnt) {
4585 	case IFCOUNTER_COLLISIONS:
4586 		return (sc->stats.colc);
4587 	case IFCOUNTER_IERRORS:
4588 		return (sc->dropped_pkts + sc->stats.rxerrc +
4589 		    sc->stats.crcerrs + sc->stats.algnerrc +
4590 		    sc->stats.ruc + sc->stats.roc +
4591 		    sc->stats.mpc + sc->stats.cexterr);
4592 	case IFCOUNTER_OERRORS:
4593 		return (sc->stats.ecol + sc->stats.latecol +
4594 		    sc->watchdog_events);
4595 	default:
4596 		return (if_get_counter_default(ifp, cnt));
4597 	}
4598 }
4599 
4600 /* em_if_needs_restart - Tell iflib when the driver needs to be reinitialized
4601  * @ctx: iflib context
4602  * @event: event code to check
4603  *
4604  * Defaults to returning false for unknown events.
4605  *
4606  * @returns true if iflib needs to reinit the interface
4607  */
4608 static bool
4609 em_if_needs_restart(if_ctx_t ctx __unused, enum iflib_restart_event event)
4610 {
4611 	switch (event) {
4612 	case IFLIB_RESTART_VLAN_CONFIG:
4613 	default:
4614 		return (false);
4615 	}
4616 }
4617 
4618 /* Export a single 32-bit register via a read-only sysctl. */
4619 static int
4620 em_sysctl_reg_handler(SYSCTL_HANDLER_ARGS)
4621 {
4622 	struct e1000_softc *sc;
4623 	u_int val;
4624 
4625 	sc = oidp->oid_arg1;
4626 	val = E1000_READ_REG(&sc->hw, oidp->oid_arg2);
4627 	return (sysctl_handle_int(oidp, &val, 0, req));
4628 }
4629 
4630 /* Per queue holdoff interrupt rate handler */
4631 static int
4632 em_sysctl_interrupt_rate_handler(SYSCTL_HANDLER_ARGS)
4633 {
4634 	struct em_rx_queue *rque;
4635 	struct em_tx_queue *tque;
4636 	struct e1000_hw *hw;
4637 	int error;
4638 	u32 reg, usec, rate;
4639 
4640 	bool tx = oidp->oid_arg2;
4641 
4642 	if (tx) {
4643 		tque = oidp->oid_arg1;
4644 		hw = &tque->sc->hw;
4645 		if (hw->mac.type >= igb_mac_min)
4646 			reg = E1000_READ_REG(hw, E1000_EITR(tque->me));
4647 		else if (hw->mac.type == e1000_82574 && tque->msix)
4648 			reg = E1000_READ_REG(hw, E1000_EITR_82574(tque->me));
4649 		else
4650 			reg = E1000_READ_REG(hw, E1000_ITR);
4651 	} else {
4652 		rque = oidp->oid_arg1;
4653 		hw = &rque->sc->hw;
4654 		if (hw->mac.type >= igb_mac_min)
4655 			reg = E1000_READ_REG(hw, E1000_EITR(rque->msix));
4656 		else if (hw->mac.type == e1000_82574 && rque->msix)
4657 			reg = E1000_READ_REG(hw, E1000_EITR_82574(rque->msix));
4658 		else
4659 			reg = E1000_READ_REG(hw, E1000_ITR);
4660 	}
4661 
4662 	if (hw->mac.type < igb_mac_min) {
4663 		if (reg > 0)
4664 			rate = EM_INTS_TO_ITR(reg);
4665 		else
4666 			rate = 0;
4667 	} else {
4668 		usec = (reg & IGB_QVECTOR_MASK);
4669 		if (usec > 0)
4670 			rate = IGB_INTS_TO_EITR(usec);
4671 		else
4672 			rate = 0;
4673 	}
4674 
4675 	error = sysctl_handle_int(oidp, &rate, 0, req);
4676 	if (error || !req->newptr)
4677 		return error;
4678 	return 0;
4679 }
4680 
4681 /*
4682  * Add sysctl variables, one per statistic, to the system.
4683  */
4684 static void
4685 em_add_hw_stats(struct e1000_softc *sc)
4686 {
4687 	device_t dev = iflib_get_dev(sc->ctx);
4688 	struct em_tx_queue *tx_que = sc->tx_queues;
4689 	struct em_rx_queue *rx_que = sc->rx_queues;
4690 
4691 	struct sysctl_ctx_list *ctx = device_get_sysctl_ctx(dev);
4692 	struct sysctl_oid *tree = device_get_sysctl_tree(dev);
4693 	struct sysctl_oid_list *child = SYSCTL_CHILDREN(tree);
4694 	struct e1000_hw_stats *stats = &sc->stats;
4695 
4696 	struct sysctl_oid *stat_node, *queue_node, *int_node;
4697 	struct sysctl_oid_list *stat_list, *queue_list, *int_list;
4698 
4699 #define QUEUE_NAME_LEN 32
4700 	char namebuf[QUEUE_NAME_LEN];
4701 
4702 	/* Driver Statistics */
4703 	SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "dropped",
4704 			CTLFLAG_RD, &sc->dropped_pkts,
4705 			"Driver dropped packets");
4706 	SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "link_irq",
4707 			CTLFLAG_RD, &sc->link_irq,
4708 			"Link MSI-X IRQ Handled");
4709 	SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "rx_overruns",
4710 			CTLFLAG_RD, &sc->rx_overruns,
4711 			"RX overruns");
4712 	SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "watchdog_timeouts",
4713 			CTLFLAG_RD, &sc->watchdog_events,
4714 			"Watchdog timeouts");
4715 	SYSCTL_ADD_PROC(ctx, child, OID_AUTO, "device_control",
4716 	    CTLTYPE_UINT | CTLFLAG_RD,
4717 	    sc, E1000_CTRL, em_sysctl_reg_handler, "IU",
4718 	    "Device Control Register");
4719 	SYSCTL_ADD_PROC(ctx, child, OID_AUTO, "rx_control",
4720 	    CTLTYPE_UINT | CTLFLAG_RD,
4721 	    sc, E1000_RCTL, em_sysctl_reg_handler, "IU",
4722 	    "Receiver Control Register");
4723 	SYSCTL_ADD_UINT(ctx, child, OID_AUTO, "fc_high_water",
4724 			CTLFLAG_RD, &sc->hw.fc.high_water, 0,
4725 			"Flow Control High Watermark");
4726 	SYSCTL_ADD_UINT(ctx, child, OID_AUTO, "fc_low_water",
4727 			CTLFLAG_RD, &sc->hw.fc.low_water, 0,
4728 			"Flow Control Low Watermark");
4729 
4730 	for (int i = 0; i < sc->tx_num_queues; i++, tx_que++) {
4731 		struct tx_ring *txr = &tx_que->txr;
4732 		snprintf(namebuf, QUEUE_NAME_LEN, "queue_tx_%d", i);
4733 		queue_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, namebuf,
4734 		    CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, "TX Queue Name");
4735 		queue_list = SYSCTL_CHILDREN(queue_node);
4736 
4737 		SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "interrupt_rate",
4738 		    CTLTYPE_UINT | CTLFLAG_RD, tx_que,
4739 		    true, em_sysctl_interrupt_rate_handler,
4740 		    "IU", "Interrupt Rate");
4741 
4742 		SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "txd_head",
4743 		    CTLTYPE_UINT | CTLFLAG_RD, sc,
4744 		    E1000_TDH(txr->me), em_sysctl_reg_handler, "IU",
4745 		    "Transmit Descriptor Head");
4746 		SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "txd_tail",
4747 		    CTLTYPE_UINT | CTLFLAG_RD, sc,
4748 		    E1000_TDT(txr->me), em_sysctl_reg_handler, "IU",
4749 		    "Transmit Descriptor Tail");
4750 		SYSCTL_ADD_ULONG(ctx, queue_list, OID_AUTO, "tx_irq",
4751 				CTLFLAG_RD, &txr->tx_irq,
4752 				"Queue MSI-X Transmit Interrupts");
4753 	}
4754 
4755 	for (int j = 0; j < sc->rx_num_queues; j++, rx_que++) {
4756 		struct rx_ring *rxr = &rx_que->rxr;
4757 		snprintf(namebuf, QUEUE_NAME_LEN, "queue_rx_%d", j);
4758 		queue_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, namebuf,
4759 		    CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, "RX Queue Name");
4760 		queue_list = SYSCTL_CHILDREN(queue_node);
4761 
4762 		SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "interrupt_rate",
4763 		    CTLTYPE_UINT | CTLFLAG_RD, rx_que,
4764 		    false, em_sysctl_interrupt_rate_handler,
4765 		    "IU", "Interrupt Rate");
4766 
4767 		SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "rxd_head",
4768 		    CTLTYPE_UINT | CTLFLAG_RD, sc,
4769 		    E1000_RDH(rxr->me), em_sysctl_reg_handler, "IU",
4770 		    "Receive Descriptor Head");
4771 		SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "rxd_tail",
4772 		    CTLTYPE_UINT | CTLFLAG_RD, sc,
4773 		    E1000_RDT(rxr->me), em_sysctl_reg_handler, "IU",
4774 		    "Receive Descriptor Tail");
4775 		SYSCTL_ADD_ULONG(ctx, queue_list, OID_AUTO, "rx_irq",
4776 				CTLFLAG_RD, &rxr->rx_irq,
4777 				"Queue MSI-X Receive Interrupts");
4778 	}
4779 
4780 	/* MAC stats get their own sub node */
4781 
4782 	stat_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, "mac_stats",
4783 	    CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, "Statistics");
4784 	stat_list = SYSCTL_CHILDREN(stat_node);
4785 
4786 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "excess_coll",
4787 			CTLFLAG_RD, &stats->ecol,
4788 			"Excessive collisions");
4789 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "single_coll",
4790 			CTLFLAG_RD, &stats->scc,
4791 			"Single collisions");
4792 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "multiple_coll",
4793 			CTLFLAG_RD, &stats->mcc,
4794 			"Multiple collisions");
4795 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "late_coll",
4796 			CTLFLAG_RD, &stats->latecol,
4797 			"Late collisions");
4798 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "collision_count",
4799 			CTLFLAG_RD, &stats->colc,
4800 			"Collision Count");
4801 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "symbol_errors",
4802 			CTLFLAG_RD, &sc->stats.symerrs,
4803 			"Symbol Errors");
4804 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "sequence_errors",
4805 			CTLFLAG_RD, &sc->stats.sec,
4806 			"Sequence Errors");
4807 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "defer_count",
4808 			CTLFLAG_RD, &sc->stats.dc,
4809 			"Defer Count");
4810 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "missed_packets",
4811 			CTLFLAG_RD, &sc->stats.mpc,
4812 			"Missed Packets");
4813 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_length_errors",
4814 			CTLFLAG_RD, &sc->stats.rlec,
4815 			"Receive Length Errors");
4816 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_no_buff",
4817 			CTLFLAG_RD, &sc->stats.rnbc,
4818 			"Receive No Buffers");
4819 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_undersize",
4820 			CTLFLAG_RD, &sc->stats.ruc,
4821 			"Receive Undersize");
4822 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_fragmented",
4823 			CTLFLAG_RD, &sc->stats.rfc,
4824 			"Fragmented Packets Received ");
4825 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_oversize",
4826 			CTLFLAG_RD, &sc->stats.roc,
4827 			"Oversized Packets Received");
4828 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_jabber",
4829 			CTLFLAG_RD, &sc->stats.rjc,
4830 			"Recevied Jabber");
4831 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_errs",
4832 			CTLFLAG_RD, &sc->stats.rxerrc,
4833 			"Receive Errors");
4834 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "crc_errs",
4835 			CTLFLAG_RD, &sc->stats.crcerrs,
4836 			"CRC errors");
4837 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "alignment_errs",
4838 			CTLFLAG_RD, &sc->stats.algnerrc,
4839 			"Alignment Errors");
4840 	/* On 82575 these are collision counts */
4841 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "coll_ext_errs",
4842 			CTLFLAG_RD, &sc->stats.cexterr,
4843 			"Collision/Carrier extension errors");
4844 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xon_recvd",
4845 			CTLFLAG_RD, &sc->stats.xonrxc,
4846 			"XON Received");
4847 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xon_txd",
4848 			CTLFLAG_RD, &sc->stats.xontxc,
4849 			"XON Transmitted");
4850 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xoff_recvd",
4851 			CTLFLAG_RD, &sc->stats.xoffrxc,
4852 			"XOFF Received");
4853 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xoff_txd",
4854 			CTLFLAG_RD, &sc->stats.xofftxc,
4855 			"XOFF Transmitted");
4856 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "unsupported_fc_recvd",
4857 			CTLFLAG_RD, &sc->stats.fcruc,
4858 			"Unsupported Flow Control Received");
4859 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "mgmt_pkts_recvd",
4860 			CTLFLAG_RD, &sc->stats.mgprc,
4861 			"Management Packets Received");
4862 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "mgmt_pkts_drop",
4863 			CTLFLAG_RD, &sc->stats.mgpdc,
4864 			"Management Packets Dropped");
4865 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "mgmt_pkts_txd",
4866 			CTLFLAG_RD, &sc->stats.mgptc,
4867 			"Management Packets Transmitted");
4868 
4869 	/* Packet Reception Stats */
4870 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "total_pkts_recvd",
4871 			CTLFLAG_RD, &sc->stats.tpr,
4872 			"Total Packets Received ");
4873 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_pkts_recvd",
4874 			CTLFLAG_RD, &sc->stats.gprc,
4875 			"Good Packets Received");
4876 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "bcast_pkts_recvd",
4877 			CTLFLAG_RD, &sc->stats.bprc,
4878 			"Broadcast Packets Received");
4879 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "mcast_pkts_recvd",
4880 			CTLFLAG_RD, &sc->stats.mprc,
4881 			"Multicast Packets Received");
4882 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_64",
4883 			CTLFLAG_RD, &sc->stats.prc64,
4884 			"64 byte frames received ");
4885 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_65_127",
4886 			CTLFLAG_RD, &sc->stats.prc127,
4887 			"65-127 byte frames received");
4888 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_128_255",
4889 			CTLFLAG_RD, &sc->stats.prc255,
4890 			"128-255 byte frames received");
4891 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_256_511",
4892 			CTLFLAG_RD, &sc->stats.prc511,
4893 			"256-511 byte frames received");
4894 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_512_1023",
4895 			CTLFLAG_RD, &sc->stats.prc1023,
4896 			"512-1023 byte frames received");
4897 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_1024_1522",
4898 			CTLFLAG_RD, &sc->stats.prc1522,
4899 			"1023-1522 byte frames received");
4900 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_octets_recvd",
4901 			CTLFLAG_RD, &sc->stats.gorc,
4902 			"Good Octets Received");
4903 
4904 	/* Packet Transmission Stats */
4905 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_octets_txd",
4906 			CTLFLAG_RD, &sc->stats.gotc,
4907 			"Good Octets Transmitted");
4908 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "total_pkts_txd",
4909 			CTLFLAG_RD, &sc->stats.tpt,
4910 			"Total Packets Transmitted");
4911 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_pkts_txd",
4912 			CTLFLAG_RD, &sc->stats.gptc,
4913 			"Good Packets Transmitted");
4914 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "bcast_pkts_txd",
4915 			CTLFLAG_RD, &sc->stats.bptc,
4916 			"Broadcast Packets Transmitted");
4917 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "mcast_pkts_txd",
4918 			CTLFLAG_RD, &sc->stats.mptc,
4919 			"Multicast Packets Transmitted");
4920 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_64",
4921 			CTLFLAG_RD, &sc->stats.ptc64,
4922 			"64 byte frames transmitted ");
4923 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_65_127",
4924 			CTLFLAG_RD, &sc->stats.ptc127,
4925 			"65-127 byte frames transmitted");
4926 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_128_255",
4927 			CTLFLAG_RD, &sc->stats.ptc255,
4928 			"128-255 byte frames transmitted");
4929 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_256_511",
4930 			CTLFLAG_RD, &sc->stats.ptc511,
4931 			"256-511 byte frames transmitted");
4932 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_512_1023",
4933 			CTLFLAG_RD, &sc->stats.ptc1023,
4934 			"512-1023 byte frames transmitted");
4935 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_1024_1522",
4936 			CTLFLAG_RD, &sc->stats.ptc1522,
4937 			"1024-1522 byte frames transmitted");
4938 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tso_txd",
4939 			CTLFLAG_RD, &sc->stats.tsctc,
4940 			"TSO Contexts Transmitted");
4941 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tso_ctx_fail",
4942 			CTLFLAG_RD, &sc->stats.tsctfc,
4943 			"TSO Contexts Failed");
4944 
4945 
4946 	/* Interrupt Stats */
4947 
4948 	int_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, "interrupts",
4949 	    CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, "Interrupt Statistics");
4950 	int_list = SYSCTL_CHILDREN(int_node);
4951 
4952 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "asserts",
4953 			CTLFLAG_RD, &sc->stats.iac,
4954 			"Interrupt Assertion Count");
4955 
4956 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "rx_pkt_timer",
4957 			CTLFLAG_RD, &sc->stats.icrxptc,
4958 			"Interrupt Cause Rx Pkt Timer Expire Count");
4959 
4960 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "rx_abs_timer",
4961 			CTLFLAG_RD, &sc->stats.icrxatc,
4962 			"Interrupt Cause Rx Abs Timer Expire Count");
4963 
4964 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "tx_pkt_timer",
4965 			CTLFLAG_RD, &sc->stats.ictxptc,
4966 			"Interrupt Cause Tx Pkt Timer Expire Count");
4967 
4968 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "tx_abs_timer",
4969 			CTLFLAG_RD, &sc->stats.ictxatc,
4970 			"Interrupt Cause Tx Abs Timer Expire Count");
4971 
4972 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "tx_queue_empty",
4973 			CTLFLAG_RD, &sc->stats.ictxqec,
4974 			"Interrupt Cause Tx Queue Empty Count");
4975 
4976 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "tx_queue_min_thresh",
4977 			CTLFLAG_RD, &sc->stats.ictxqmtc,
4978 			"Interrupt Cause Tx Queue Min Thresh Count");
4979 
4980 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "rx_desc_min_thresh",
4981 			CTLFLAG_RD, &sc->stats.icrxdmtc,
4982 			"Interrupt Cause Rx Desc Min Thresh Count");
4983 
4984 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "rx_overrun",
4985 			CTLFLAG_RD, &sc->stats.icrxoc,
4986 			"Interrupt Cause Receiver Overrun Count");
4987 }
4988 
4989 static void
4990 em_fw_version_locked(if_ctx_t ctx)
4991 {
4992 	struct e1000_softc *sc = iflib_get_softc(ctx);
4993 	struct e1000_hw *hw = &sc->hw;
4994 	struct e1000_fw_version *fw_ver = &sc->fw_ver;
4995 	uint16_t eep = 0;
4996 
4997 	/*
4998 	 * em_fw_version_locked() must run under the IFLIB_CTX_LOCK to meet the
4999 	 * NVM locking model, so we do it in em_if_attach_pre() and store the
5000 	 * info in the softc
5001 	 */
5002 	ASSERT_CTX_LOCK_HELD(hw);
5003 
5004 	*fw_ver = (struct e1000_fw_version){0};
5005 
5006 	if (hw->mac.type >= igb_mac_min) {
5007 		/*
5008 		 * Use the Shared Code for igb(4)
5009 		 */
5010 		e1000_get_fw_version(hw, fw_ver);
5011 	} else {
5012 		/*
5013 		 * Otherwise, EEPROM version should be present on (almost?) all
5014 		 * devices here
5015 		 */
5016 		if(e1000_read_nvm(hw, NVM_VERSION, 1, &eep)) {
5017 			INIT_DEBUGOUT("can't get EEPROM version");
5018 			return;
5019 		}
5020 
5021 		fw_ver->eep_major = (eep & NVM_MAJOR_MASK) >> NVM_MAJOR_SHIFT;
5022 		fw_ver->eep_minor = (eep & NVM_MINOR_MASK) >> NVM_MINOR_SHIFT;
5023 		fw_ver->eep_build = (eep & NVM_IMAGE_ID_MASK);
5024 	}
5025 }
5026 
5027 static void
5028 em_sbuf_fw_version(struct e1000_fw_version *fw_ver, struct sbuf *buf)
5029 {
5030 	const char *space = "";
5031 
5032 	if (fw_ver->eep_major || fw_ver->eep_minor || fw_ver->eep_build) {
5033 		sbuf_printf(buf, "EEPROM V%d.%d-%d", fw_ver->eep_major,
5034 			    fw_ver->eep_minor, fw_ver->eep_build);
5035 		space = " ";
5036 	}
5037 
5038 	if (fw_ver->invm_major || fw_ver->invm_minor || fw_ver->invm_img_type) {
5039 		sbuf_printf(buf, "%sNVM V%d.%d imgtype%d",
5040 			    space, fw_ver->invm_major, fw_ver->invm_minor,
5041 			    fw_ver->invm_img_type);
5042 		space = " ";
5043 	}
5044 
5045 	if (fw_ver->or_valid) {
5046 		sbuf_printf(buf, "%sOption ROM V%d-b%d-p%d",
5047 			    space, fw_ver->or_major, fw_ver->or_build,
5048 			    fw_ver->or_patch);
5049 		space = " ";
5050 	}
5051 
5052 	if (fw_ver->etrack_id)
5053 		sbuf_printf(buf, "%seTrack 0x%08x", space, fw_ver->etrack_id);
5054 }
5055 
5056 static void
5057 em_print_fw_version(struct e1000_softc *sc )
5058 {
5059 	device_t dev = sc->dev;
5060 	struct sbuf *buf;
5061 	int error = 0;
5062 
5063 	buf = sbuf_new_auto();
5064 	if (!buf) {
5065 		device_printf(dev, "Could not allocate sbuf for output.\n");
5066 		return;
5067 	}
5068 
5069 	em_sbuf_fw_version(&sc->fw_ver, buf);
5070 
5071 	error = sbuf_finish(buf);
5072 	if (error)
5073 		device_printf(dev, "Error finishing sbuf: %d\n", error);
5074 	else if (sbuf_len(buf))
5075 		device_printf(dev, "%s\n", sbuf_data(buf));
5076 
5077 	sbuf_delete(buf);
5078 }
5079 
5080 static int
5081 em_sysctl_print_fw_version(SYSCTL_HANDLER_ARGS)
5082 {
5083 	struct e1000_softc *sc = (struct e1000_softc *)arg1;
5084 	device_t dev = sc->dev;
5085 	struct sbuf *buf;
5086 	int error = 0;
5087 
5088 	buf = sbuf_new_for_sysctl(NULL, NULL, 128, req);
5089 	if (!buf) {
5090 		device_printf(dev, "Could not allocate sbuf for output.\n");
5091 		return (ENOMEM);
5092 	}
5093 
5094 	em_sbuf_fw_version(&sc->fw_ver, buf);
5095 
5096 	error = sbuf_finish(buf);
5097 	if (error)
5098 		device_printf(dev, "Error finishing sbuf: %d\n", error);
5099 
5100 	sbuf_delete(buf);
5101 
5102 	return (0);
5103 }
5104 
5105 /**********************************************************************
5106  *
5107  *  This routine provides a way to dump out the adapter eeprom,
5108  *  often a useful debug/service tool. This only dumps the first
5109  *  32 words, stuff that matters is in that extent.
5110  *
5111  **********************************************************************/
5112 static int
5113 em_sysctl_nvm_info(SYSCTL_HANDLER_ARGS)
5114 {
5115 	struct e1000_softc *sc = (struct e1000_softc *)arg1;
5116 	int error;
5117 	int result;
5118 
5119 	result = -1;
5120 	error = sysctl_handle_int(oidp, &result, 0, req);
5121 
5122 	if (error || !req->newptr)
5123 		return (error);
5124 
5125 	/*
5126 	 * This value will cause a hex dump of the
5127 	 * first 32 16-bit words of the EEPROM to
5128 	 * the screen.
5129 	 */
5130 	if (result == 1)
5131 		em_print_nvm_info(sc);
5132 
5133 	return (error);
5134 }
5135 
5136 static void
5137 em_print_nvm_info(struct e1000_softc *sc)
5138 {
5139 	struct e1000_hw *hw = &sc->hw;
5140 	struct sx *iflib_ctx_lock = iflib_ctx_lock_get(sc->ctx);
5141 	u16 eeprom_data;
5142 	int i, j, row = 0;
5143 
5144 	/* Its a bit crude, but it gets the job done */
5145 	printf("\nInterface EEPROM Dump:\n");
5146 	printf("Offset\n0x0000  ");
5147 
5148 	/* We rely on the IFLIB_CTX_LOCK as part of NVM locking model */
5149 	sx_xlock(iflib_ctx_lock);
5150 	ASSERT_CTX_LOCK_HELD(hw);
5151 	for (i = 0, j = 0; i < 32; i++, j++) {
5152 		if (j == 8) { /* Make the offset block */
5153 			j = 0; ++row;
5154 			printf("\n0x00%x0  ",row);
5155 		}
5156 		e1000_read_nvm(hw, i, 1, &eeprom_data);
5157 		printf("%04x ", eeprom_data);
5158 	}
5159 	sx_xunlock(iflib_ctx_lock);
5160 	printf("\n");
5161 }
5162 
5163 static int
5164 em_sysctl_int_delay(SYSCTL_HANDLER_ARGS)
5165 {
5166 	struct em_int_delay_info *info;
5167 	struct e1000_softc *sc;
5168 	u32 regval;
5169 	int error, usecs, ticks;
5170 
5171 	info = (struct em_int_delay_info *) arg1;
5172 	usecs = info->value;
5173 	error = sysctl_handle_int(oidp, &usecs, 0, req);
5174 	if (error != 0 || req->newptr == NULL)
5175 		return (error);
5176 	if (usecs < 0 || usecs > EM_TICKS_TO_USECS(65535))
5177 		return (EINVAL);
5178 	info->value = usecs;
5179 	ticks = EM_USECS_TO_TICKS(usecs);
5180 	if (info->offset == E1000_ITR)	/* units are 256ns here */
5181 		ticks *= 4;
5182 
5183 	sc = info->sc;
5184 
5185 	regval = E1000_READ_OFFSET(&sc->hw, info->offset);
5186 	regval = (regval & ~0xffff) | (ticks & 0xffff);
5187 	/* Handle a few special cases. */
5188 	switch (info->offset) {
5189 	case E1000_RDTR:
5190 		break;
5191 	case E1000_TIDV:
5192 		if (ticks == 0) {
5193 			sc->txd_cmd &= ~E1000_TXD_CMD_IDE;
5194 			/* Don't write 0 into the TIDV register. */
5195 			regval++;
5196 		} else
5197 			sc->txd_cmd |= E1000_TXD_CMD_IDE;
5198 		break;
5199 	}
5200 	E1000_WRITE_OFFSET(&sc->hw, info->offset, regval);
5201 	return (0);
5202 }
5203 
5204 static int
5205 em_sysctl_tso_tcp_flags_mask(SYSCTL_HANDLER_ARGS)
5206 {
5207 	struct e1000_softc *sc;
5208 	u32 reg, val, shift;
5209 	int error, mask;
5210 
5211 	sc = oidp->oid_arg1;
5212 	switch (oidp->oid_arg2) {
5213 	case 0:
5214 		reg = E1000_DTXTCPFLGL;
5215 		shift = 0;
5216 		break;
5217 	case 1:
5218 		reg = E1000_DTXTCPFLGL;
5219 		shift = 16;
5220 		break;
5221 	case 2:
5222 		reg = E1000_DTXTCPFLGH;
5223 		shift = 0;
5224 		break;
5225 	default:
5226 		return (EINVAL);
5227 		break;
5228 	}
5229 	val = E1000_READ_REG(&sc->hw, reg);
5230 	mask = (val >> shift) & 0xfff;
5231 	error = sysctl_handle_int(oidp, &mask, 0, req);
5232 	if (error != 0 || req->newptr == NULL)
5233 		return (error);
5234 	if (mask < 0 || mask > 0xfff)
5235 		return (EINVAL);
5236 	val = (val & ~(0xfff << shift)) | (mask << shift);
5237 	E1000_WRITE_REG(&sc->hw, reg, val);
5238 	return (0);
5239 }
5240 
5241 static void
5242 em_add_int_delay_sysctl(struct e1000_softc *sc, const char *name,
5243 	const char *description, struct em_int_delay_info *info,
5244 	int offset, int value)
5245 {
5246 	info->sc = sc;
5247 	info->offset = offset;
5248 	info->value = value;
5249 	SYSCTL_ADD_PROC(device_get_sysctl_ctx(sc->dev),
5250 	    SYSCTL_CHILDREN(device_get_sysctl_tree(sc->dev)),
5251 	    OID_AUTO, name, CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_NEEDGIANT,
5252 	    info, 0, em_sysctl_int_delay, "I", description);
5253 }
5254 
5255 /*
5256  * Set flow control using sysctl:
5257  * Flow control values:
5258  *      0 - off
5259  *      1 - rx pause
5260  *      2 - tx pause
5261  *      3 - full
5262  */
5263 static int
5264 em_set_flowcntl(SYSCTL_HANDLER_ARGS)
5265 {
5266 	int error;
5267 	static int input = 3; /* default is full */
5268 	struct e1000_softc	*sc = (struct e1000_softc *) arg1;
5269 
5270 	error = sysctl_handle_int(oidp, &input, 0, req);
5271 
5272 	if ((error) || (req->newptr == NULL))
5273 		return (error);
5274 
5275 	if (input == sc->fc) /* no change? */
5276 		return (error);
5277 
5278 	switch (input) {
5279 	case e1000_fc_rx_pause:
5280 	case e1000_fc_tx_pause:
5281 	case e1000_fc_full:
5282 	case e1000_fc_none:
5283 		sc->hw.fc.requested_mode = input;
5284 		sc->fc = input;
5285 		break;
5286 	default:
5287 		/* Do nothing */
5288 		return (error);
5289 	}
5290 
5291 	sc->hw.fc.current_mode = sc->hw.fc.requested_mode;
5292 	e1000_force_mac_fc(&sc->hw);
5293 	return (error);
5294 }
5295 
5296 /*
5297  * Manage DMA Coalesce:
5298  * Control values:
5299  * 	0/1 - off/on
5300  *	Legal timer values are:
5301  *	250,500,1000-10000 in thousands
5302  */
5303 static int
5304 igb_sysctl_dmac(SYSCTL_HANDLER_ARGS)
5305 {
5306 	struct e1000_softc *sc = (struct e1000_softc *) arg1;
5307 	int error;
5308 
5309 	error = sysctl_handle_int(oidp, &sc->dmac, 0, req);
5310 
5311 	if ((error) || (req->newptr == NULL))
5312 		return (error);
5313 
5314 	switch (sc->dmac) {
5315 		case 0:
5316 			/* Disabling */
5317 			break;
5318 		case 1: /* Just enable and use default */
5319 			sc->dmac = 1000;
5320 			break;
5321 		case 250:
5322 		case 500:
5323 		case 1000:
5324 		case 2000:
5325 		case 3000:
5326 		case 4000:
5327 		case 5000:
5328 		case 6000:
5329 		case 7000:
5330 		case 8000:
5331 		case 9000:
5332 		case 10000:
5333 			/* Legal values - allow */
5334 			break;
5335 		default:
5336 			/* Do nothing, illegal value */
5337 			sc->dmac = 0;
5338 			return (EINVAL);
5339 	}
5340 	/* Reinit the interface */
5341 	em_if_init(sc->ctx);
5342 	return (error);
5343 }
5344 
5345 /*
5346  * Manage Energy Efficient Ethernet:
5347  * Control values:
5348  *     0/1 - enabled/disabled
5349  */
5350 static int
5351 em_sysctl_eee(SYSCTL_HANDLER_ARGS)
5352 {
5353 	struct e1000_softc *sc = (struct e1000_softc *) arg1;
5354 	int error, value;
5355 
5356 	if (sc->hw.mac.type < igb_mac_min)
5357 		value = sc->hw.dev_spec.ich8lan.eee_disable;
5358 	else
5359 		value = sc->hw.dev_spec._82575.eee_disable;
5360 	error = sysctl_handle_int(oidp, &value, 0, req);
5361 	if (error || req->newptr == NULL)
5362 		return (error);
5363 	if (sc->hw.mac.type < igb_mac_min)
5364 		sc->hw.dev_spec.ich8lan.eee_disable = (value != 0);
5365 	else
5366 		sc->hw.dev_spec._82575.eee_disable = (value != 0);
5367 	em_if_init(sc->ctx);
5368 
5369 	return (0);
5370 }
5371 
5372 static int
5373 em_sysctl_debug_info(SYSCTL_HANDLER_ARGS)
5374 {
5375 	struct e1000_softc *sc;
5376 	int error;
5377 	int result;
5378 
5379 	result = -1;
5380 	error = sysctl_handle_int(oidp, &result, 0, req);
5381 
5382 	if (error || !req->newptr)
5383 		return (error);
5384 
5385 	if (result == 1) {
5386 		sc = (struct e1000_softc *) arg1;
5387 		em_print_debug_info(sc);
5388 	}
5389 
5390 	return (error);
5391 }
5392 
5393 static int
5394 em_get_rs(SYSCTL_HANDLER_ARGS)
5395 {
5396 	struct e1000_softc *sc = (struct e1000_softc *) arg1;
5397 	int error;
5398 	int result;
5399 
5400 	result = 0;
5401 	error = sysctl_handle_int(oidp, &result, 0, req);
5402 
5403 	if (error || !req->newptr || result != 1)
5404 		return (error);
5405 	em_dump_rs(sc);
5406 
5407 	return (error);
5408 }
5409 
5410 static void
5411 em_if_debug(if_ctx_t ctx)
5412 {
5413 	em_dump_rs(iflib_get_softc(ctx));
5414 }
5415 
5416 /*
5417  * This routine is meant to be fluid, add whatever is
5418  * needed for debugging a problem.  -jfv
5419  */
5420 static void
5421 em_print_debug_info(struct e1000_softc *sc)
5422 {
5423 	device_t dev = iflib_get_dev(sc->ctx);
5424 	if_t ifp = iflib_get_ifp(sc->ctx);
5425 	struct tx_ring *txr = &sc->tx_queues->txr;
5426 	struct rx_ring *rxr = &sc->rx_queues->rxr;
5427 
5428 	if (if_getdrvflags(ifp) & IFF_DRV_RUNNING)
5429 		printf("Interface is RUNNING ");
5430 	else
5431 		printf("Interface is NOT RUNNING\n");
5432 
5433 	if (if_getdrvflags(ifp) & IFF_DRV_OACTIVE)
5434 		printf("and INACTIVE\n");
5435 	else
5436 		printf("and ACTIVE\n");
5437 
5438 	for (int i = 0; i < sc->tx_num_queues; i++, txr++) {
5439 		device_printf(dev, "TX Queue %d ------\n", i);
5440 		device_printf(dev, "hw tdh = %d, hw tdt = %d\n",
5441 			E1000_READ_REG(&sc->hw, E1000_TDH(i)),
5442 			E1000_READ_REG(&sc->hw, E1000_TDT(i)));
5443 
5444 	}
5445 	for (int j=0; j < sc->rx_num_queues; j++, rxr++) {
5446 		device_printf(dev, "RX Queue %d ------\n", j);
5447 		device_printf(dev, "hw rdh = %d, hw rdt = %d\n",
5448 			E1000_READ_REG(&sc->hw, E1000_RDH(j)),
5449 			E1000_READ_REG(&sc->hw, E1000_RDT(j)));
5450 	}
5451 }
5452 
5453 /*
5454  * 82574 only:
5455  * Write a new value to the EEPROM increasing the number of MSI-X
5456  * vectors from 3 to 5, for proper multiqueue support.
5457  */
5458 static void
5459 em_enable_vectors_82574(if_ctx_t ctx)
5460 {
5461 	struct e1000_softc *sc = iflib_get_softc(ctx);
5462 	struct e1000_hw *hw = &sc->hw;
5463 	device_t dev = iflib_get_dev(ctx);
5464 	u16 edata;
5465 
5466 	e1000_read_nvm(hw, EM_NVM_PCIE_CTRL, 1, &edata);
5467 	if (bootverbose)
5468 		device_printf(dev, "EM_NVM_PCIE_CTRL = %#06x\n", edata);
5469 	if (((edata & EM_NVM_MSIX_N_MASK) >> EM_NVM_MSIX_N_SHIFT) != 4) {
5470 		device_printf(dev, "Writing to eeprom: increasing "
5471 		    "reported MSI-X vectors from 3 to 5...\n");
5472 		edata &= ~(EM_NVM_MSIX_N_MASK);
5473 		edata |= 4 << EM_NVM_MSIX_N_SHIFT;
5474 		e1000_write_nvm(hw, EM_NVM_PCIE_CTRL, 1, &edata);
5475 		e1000_update_nvm_checksum(hw);
5476 		device_printf(dev, "Writing to eeprom: done\n");
5477 	}
5478 }
5479