xref: /freebsd/sys/dev/e1000/if_em.c (revision 5e3190f700637fcfc1a52daeaa4a031fdd2557c7)
1 /*-
2  * SPDX-License-Identifier: BSD-2-Clause
3  *
4  * Copyright (c) 2016 Nicole Graziano <nicole@nextbsd.org>
5  * All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  *
16  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
17  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
18  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
19  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
20  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
21  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
22  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
23  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
24  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
25  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
26  * SUCH DAMAGE.
27  */
28 
29 #include "if_em.h"
30 #include <sys/sbuf.h>
31 #include <machine/_inttypes.h>
32 
33 #define em_mac_min e1000_82571
34 #define igb_mac_min e1000_82575
35 
36 /*********************************************************************
37  *  Driver version:
38  *********************************************************************/
39 static const char em_driver_version[] = "7.7.8-fbsd";
40 static const char igb_driver_version[] = "2.5.19-fbsd";
41 
42 /*********************************************************************
43  *  PCI Device ID Table
44  *
45  *  Used by probe to select devices to load on
46  *  Last field stores an index into e1000_strings
47  *  Last entry must be all 0s
48  *
49  *  { Vendor ID, Device ID, SubVendor ID, SubDevice ID, String Index }
50  *********************************************************************/
51 
52 static const pci_vendor_info_t em_vendor_info_array[] =
53 {
54 	/* Intel(R) - lem-class legacy devices */
55 	PVID(0x8086, E1000_DEV_ID_82540EM, "Intel(R) Legacy PRO/1000 MT 82540EM"),
56 	PVID(0x8086, E1000_DEV_ID_82540EM_LOM, "Intel(R) Legacy PRO/1000 MT 82540EM (LOM)"),
57 	PVID(0x8086, E1000_DEV_ID_82540EP, "Intel(R) Legacy PRO/1000 MT 82540EP"),
58 	PVID(0x8086, E1000_DEV_ID_82540EP_LOM, "Intel(R) Legacy PRO/1000 MT 82540EP (LOM)"),
59 	PVID(0x8086, E1000_DEV_ID_82540EP_LP, "Intel(R) Legacy PRO/1000 MT 82540EP (Mobile)"),
60 
61 	PVID(0x8086, E1000_DEV_ID_82541EI, "Intel(R) Legacy PRO/1000 MT 82541EI (Copper)"),
62 	PVID(0x8086, E1000_DEV_ID_82541ER, "Intel(R) Legacy PRO/1000 82541ER"),
63 	PVID(0x8086, E1000_DEV_ID_82541ER_LOM, "Intel(R) Legacy PRO/1000 MT 82541ER"),
64 	PVID(0x8086, E1000_DEV_ID_82541EI_MOBILE, "Intel(R) Legacy PRO/1000 MT 82541EI (Mobile)"),
65 	PVID(0x8086, E1000_DEV_ID_82541GI, "Intel(R) Legacy PRO/1000 MT 82541GI"),
66 	PVID(0x8086, E1000_DEV_ID_82541GI_LF, "Intel(R) Legacy PRO/1000 GT 82541PI"),
67 	PVID(0x8086, E1000_DEV_ID_82541GI_MOBILE, "Intel(R) Legacy PRO/1000 MT 82541GI (Mobile)"),
68 
69 	PVID(0x8086, E1000_DEV_ID_82542, "Intel(R) Legacy PRO/1000 82542 (Fiber)"),
70 
71 	PVID(0x8086, E1000_DEV_ID_82543GC_FIBER, "Intel(R) Legacy PRO/1000 F 82543GC (Fiber)"),
72 	PVID(0x8086, E1000_DEV_ID_82543GC_COPPER, "Intel(R) Legacy PRO/1000 T 82543GC (Copper)"),
73 
74 	PVID(0x8086, E1000_DEV_ID_82544EI_COPPER, "Intel(R) Legacy PRO/1000 XT 82544EI (Copper)"),
75 	PVID(0x8086, E1000_DEV_ID_82544EI_FIBER, "Intel(R) Legacy PRO/1000 XF 82544EI (Fiber)"),
76 	PVID(0x8086, E1000_DEV_ID_82544GC_COPPER, "Intel(R) Legacy PRO/1000 T 82544GC (Copper)"),
77 	PVID(0x8086, E1000_DEV_ID_82544GC_LOM, "Intel(R) Legacy PRO/1000 XT 82544GC (LOM)"),
78 
79 	PVID(0x8086, E1000_DEV_ID_82545EM_COPPER, "Intel(R) Legacy PRO/1000 MT 82545EM (Copper)"),
80 	PVID(0x8086, E1000_DEV_ID_82545EM_FIBER, "Intel(R) Legacy PRO/1000 MF 82545EM (Fiber)"),
81 	PVID(0x8086, E1000_DEV_ID_82545GM_COPPER, "Intel(R) Legacy PRO/1000 MT 82545GM (Copper)"),
82 	PVID(0x8086, E1000_DEV_ID_82545GM_FIBER, "Intel(R) Legacy PRO/1000 MF 82545GM (Fiber)"),
83 	PVID(0x8086, E1000_DEV_ID_82545GM_SERDES, "Intel(R) Legacy PRO/1000 MB 82545GM (SERDES)"),
84 
85 	PVID(0x8086, E1000_DEV_ID_82546EB_COPPER, "Intel(R) Legacy PRO/1000 MT 82546EB (Copper)"),
86 	PVID(0x8086, E1000_DEV_ID_82546EB_FIBER, "Intel(R) Legacy PRO/1000 MF 82546EB (Fiber)"),
87 	PVID(0x8086, E1000_DEV_ID_82546EB_QUAD_COPPER, "Intel(R) Legacy PRO/1000 MT 82546EB (Quad Copper"),
88 	PVID(0x8086, E1000_DEV_ID_82546GB_COPPER, "Intel(R) Legacy PRO/1000 MT 82546GB (Copper)"),
89 	PVID(0x8086, E1000_DEV_ID_82546GB_FIBER, "Intel(R) Legacy PRO/1000 MF 82546GB (Fiber)"),
90 	PVID(0x8086, E1000_DEV_ID_82546GB_SERDES, "Intel(R) Legacy PRO/1000 MB 82546GB (SERDES)"),
91 	PVID(0x8086, E1000_DEV_ID_82546GB_PCIE, "Intel(R) Legacy PRO/1000 P 82546GB (PCIe)"),
92 	PVID(0x8086, E1000_DEV_ID_82546GB_QUAD_COPPER, "Intel(R) Legacy PRO/1000 GT 82546GB (Quad Copper)"),
93 	PVID(0x8086, E1000_DEV_ID_82546GB_QUAD_COPPER_KSP3, "Intel(R) Legacy PRO/1000 GT 82546GB (Quad Copper)"),
94 
95 	PVID(0x8086, E1000_DEV_ID_82547EI, "Intel(R) Legacy PRO/1000 CT 82547EI"),
96 	PVID(0x8086, E1000_DEV_ID_82547EI_MOBILE, "Intel(R) Legacy PRO/1000 CT 82547EI (Mobile)"),
97 	PVID(0x8086, E1000_DEV_ID_82547GI, "Intel(R) Legacy PRO/1000 CT 82547GI"),
98 
99 	/* Intel(R) - em-class devices */
100 	PVID(0x8086, E1000_DEV_ID_82571EB_COPPER, "Intel(R) PRO/1000 PT 82571EB/82571GB (Copper)"),
101 	PVID(0x8086, E1000_DEV_ID_82571EB_FIBER, "Intel(R) PRO/1000 PF 82571EB/82571GB (Fiber)"),
102 	PVID(0x8086, E1000_DEV_ID_82571EB_SERDES, "Intel(R) PRO/1000 PB 82571EB (SERDES)"),
103 	PVID(0x8086, E1000_DEV_ID_82571EB_SERDES_DUAL, "Intel(R) PRO/1000 82571EB (Dual Mezzanine)"),
104 	PVID(0x8086, E1000_DEV_ID_82571EB_SERDES_QUAD, "Intel(R) PRO/1000 82571EB (Quad Mezzanine)"),
105 	PVID(0x8086, E1000_DEV_ID_82571EB_QUAD_COPPER, "Intel(R) PRO/1000 PT 82571EB/82571GB (Quad Copper)"),
106 	PVID(0x8086, E1000_DEV_ID_82571EB_QUAD_COPPER_LP, "Intel(R) PRO/1000 PT 82571EB/82571GB (Quad Copper)"),
107 	PVID(0x8086, E1000_DEV_ID_82571EB_QUAD_FIBER, "Intel(R) PRO/1000 PF 82571EB (Quad Fiber)"),
108 	PVID(0x8086, E1000_DEV_ID_82571PT_QUAD_COPPER, "Intel(R) PRO/1000 PT 82571PT (Quad Copper)"),
109 	PVID(0x8086, E1000_DEV_ID_82572EI, "Intel(R) PRO/1000 PT 82572EI (Copper)"),
110 	PVID(0x8086, E1000_DEV_ID_82572EI_COPPER, "Intel(R) PRO/1000 PT 82572EI (Copper)"),
111 	PVID(0x8086, E1000_DEV_ID_82572EI_FIBER, "Intel(R) PRO/1000 PF 82572EI (Fiber)"),
112 	PVID(0x8086, E1000_DEV_ID_82572EI_SERDES, "Intel(R) PRO/1000 82572EI (SERDES)"),
113 	PVID(0x8086, E1000_DEV_ID_82573E, "Intel(R) PRO/1000 82573E (Copper)"),
114 	PVID(0x8086, E1000_DEV_ID_82573E_IAMT, "Intel(R) PRO/1000 82573E AMT (Copper)"),
115 	PVID(0x8086, E1000_DEV_ID_82573L, "Intel(R) PRO/1000 82573L"),
116 	PVID(0x8086, E1000_DEV_ID_82583V, "Intel(R) 82583V"),
117 	PVID(0x8086, E1000_DEV_ID_80003ES2LAN_COPPER_SPT, "Intel(R) 80003ES2LAN (Copper)"),
118 	PVID(0x8086, E1000_DEV_ID_80003ES2LAN_SERDES_SPT, "Intel(R) 80003ES2LAN (SERDES)"),
119 	PVID(0x8086, E1000_DEV_ID_80003ES2LAN_COPPER_DPT, "Intel(R) 80003ES2LAN (Dual Copper)"),
120 	PVID(0x8086, E1000_DEV_ID_80003ES2LAN_SERDES_DPT, "Intel(R) 80003ES2LAN (Dual SERDES)"),
121 	PVID(0x8086, E1000_DEV_ID_ICH8_IGP_M_AMT, "Intel(R) 82566MM ICH8 AMT (Mobile)"),
122 	PVID(0x8086, E1000_DEV_ID_ICH8_IGP_AMT, "Intel(R) 82566DM ICH8 AMT"),
123 	PVID(0x8086, E1000_DEV_ID_ICH8_IGP_C, "Intel(R) 82566DC ICH8"),
124 	PVID(0x8086, E1000_DEV_ID_ICH8_IFE, "Intel(R) 82562V ICH8"),
125 	PVID(0x8086, E1000_DEV_ID_ICH8_IFE_GT, "Intel(R) 82562GT ICH8"),
126 	PVID(0x8086, E1000_DEV_ID_ICH8_IFE_G, "Intel(R) 82562G ICH8"),
127 	PVID(0x8086, E1000_DEV_ID_ICH8_IGP_M, "Intel(R) 82566MC ICH8"),
128 	PVID(0x8086, E1000_DEV_ID_ICH8_82567V_3, "Intel(R) 82567V-3 ICH8"),
129 	PVID(0x8086, E1000_DEV_ID_ICH9_IGP_M_AMT, "Intel(R) 82567LM ICH9 AMT"),
130 	PVID(0x8086, E1000_DEV_ID_ICH9_IGP_AMT, "Intel(R) 82566DM-2 ICH9 AMT"),
131 	PVID(0x8086, E1000_DEV_ID_ICH9_IGP_C, "Intel(R) 82566DC-2 ICH9"),
132 	PVID(0x8086, E1000_DEV_ID_ICH9_IGP_M, "Intel(R) 82567LF ICH9"),
133 	PVID(0x8086, E1000_DEV_ID_ICH9_IGP_M_V, "Intel(R) 82567V ICH9"),
134 	PVID(0x8086, E1000_DEV_ID_ICH9_IFE, "Intel(R) 82562V-2 ICH9"),
135 	PVID(0x8086, E1000_DEV_ID_ICH9_IFE_GT, "Intel(R) 82562GT-2 ICH9"),
136 	PVID(0x8086, E1000_DEV_ID_ICH9_IFE_G, "Intel(R) 82562G-2 ICH9"),
137 	PVID(0x8086, E1000_DEV_ID_ICH9_BM, "Intel(R) 82567LM-4 ICH9"),
138 	PVID(0x8086, E1000_DEV_ID_82574L, "Intel(R) Gigabit CT 82574L"),
139 	PVID(0x8086, E1000_DEV_ID_82574LA, "Intel(R) 82574L-Apple"),
140 	PVID(0x8086, E1000_DEV_ID_ICH10_R_BM_LM, "Intel(R) 82567LM-2 ICH10"),
141 	PVID(0x8086, E1000_DEV_ID_ICH10_R_BM_LF, "Intel(R) 82567LF-2 ICH10"),
142 	PVID(0x8086, E1000_DEV_ID_ICH10_R_BM_V, "Intel(R) 82567V-2 ICH10"),
143 	PVID(0x8086, E1000_DEV_ID_ICH10_D_BM_LM, "Intel(R) 82567LM-3 ICH10"),
144 	PVID(0x8086, E1000_DEV_ID_ICH10_D_BM_LF, "Intel(R) 82567LF-3 ICH10"),
145 	PVID(0x8086, E1000_DEV_ID_ICH10_D_BM_V, "Intel(R) 82567V-4 ICH10"),
146 	PVID(0x8086, E1000_DEV_ID_PCH_M_HV_LM, "Intel(R) 82577LM"),
147 	PVID(0x8086, E1000_DEV_ID_PCH_M_HV_LC, "Intel(R) 82577LC"),
148 	PVID(0x8086, E1000_DEV_ID_PCH_D_HV_DM, "Intel(R) 82578DM"),
149 	PVID(0x8086, E1000_DEV_ID_PCH_D_HV_DC, "Intel(R) 82578DC"),
150 	PVID(0x8086, E1000_DEV_ID_PCH2_LV_LM, "Intel(R) 82579LM"),
151 	PVID(0x8086, E1000_DEV_ID_PCH2_LV_V, "Intel(R) 82579V"),
152 	PVID(0x8086, E1000_DEV_ID_PCH_LPT_I217_LM, "Intel(R) I217-LM LPT"),
153 	PVID(0x8086, E1000_DEV_ID_PCH_LPT_I217_V, "Intel(R) I217-V LPT"),
154 	PVID(0x8086, E1000_DEV_ID_PCH_LPTLP_I218_LM, "Intel(R) I218-LM LPTLP"),
155 	PVID(0x8086, E1000_DEV_ID_PCH_LPTLP_I218_V, "Intel(R) I218-V LPTLP"),
156 	PVID(0x8086, E1000_DEV_ID_PCH_I218_LM2, "Intel(R) I218-LM (2)"),
157 	PVID(0x8086, E1000_DEV_ID_PCH_I218_V2, "Intel(R) I218-V (2)"),
158 	PVID(0x8086, E1000_DEV_ID_PCH_I218_LM3, "Intel(R) I218-LM (3)"),
159 	PVID(0x8086, E1000_DEV_ID_PCH_I218_V3, "Intel(R) I218-V (3)"),
160 	PVID(0x8086, E1000_DEV_ID_PCH_SPT_I219_LM, "Intel(R) I219-LM SPT"),
161 	PVID(0x8086, E1000_DEV_ID_PCH_SPT_I219_V, "Intel(R) I219-V SPT"),
162 	PVID(0x8086, E1000_DEV_ID_PCH_SPT_I219_LM2, "Intel(R) I219-LM SPT-H(2)"),
163 	PVID(0x8086, E1000_DEV_ID_PCH_SPT_I219_V2, "Intel(R) I219-V SPT-H(2)"),
164 	PVID(0x8086, E1000_DEV_ID_PCH_LBG_I219_LM3, "Intel(R) I219-LM LBG(3)"),
165 	PVID(0x8086, E1000_DEV_ID_PCH_SPT_I219_LM4, "Intel(R) I219-LM SPT(4)"),
166 	PVID(0x8086, E1000_DEV_ID_PCH_SPT_I219_V4, "Intel(R) I219-V SPT(4)"),
167 	PVID(0x8086, E1000_DEV_ID_PCH_SPT_I219_LM5, "Intel(R) I219-LM SPT(5)"),
168 	PVID(0x8086, E1000_DEV_ID_PCH_SPT_I219_V5, "Intel(R) I219-V SPT(5)"),
169 	PVID(0x8086, E1000_DEV_ID_PCH_CNP_I219_LM6, "Intel(R) I219-LM CNP(6)"),
170 	PVID(0x8086, E1000_DEV_ID_PCH_CNP_I219_V6, "Intel(R) I219-V CNP(6)"),
171 	PVID(0x8086, E1000_DEV_ID_PCH_CNP_I219_LM7, "Intel(R) I219-LM CNP(7)"),
172 	PVID(0x8086, E1000_DEV_ID_PCH_CNP_I219_V7, "Intel(R) I219-V CNP(7)"),
173 	PVID(0x8086, E1000_DEV_ID_PCH_ICP_I219_LM8, "Intel(R) I219-LM ICP(8)"),
174 	PVID(0x8086, E1000_DEV_ID_PCH_ICP_I219_V8, "Intel(R) I219-V ICP(8)"),
175 	PVID(0x8086, E1000_DEV_ID_PCH_ICP_I219_LM9, "Intel(R) I219-LM ICP(9)"),
176 	PVID(0x8086, E1000_DEV_ID_PCH_ICP_I219_V9, "Intel(R) I219-V ICP(9)"),
177 	PVID(0x8086, E1000_DEV_ID_PCH_CMP_I219_LM10, "Intel(R) I219-LM CMP(10)"),
178 	PVID(0x8086, E1000_DEV_ID_PCH_CMP_I219_V10, "Intel(R) I219-V CMP(10)"),
179 	PVID(0x8086, E1000_DEV_ID_PCH_CMP_I219_LM11, "Intel(R) I219-LM CMP(11)"),
180 	PVID(0x8086, E1000_DEV_ID_PCH_CMP_I219_V11, "Intel(R) I219-V CMP(11)"),
181 	PVID(0x8086, E1000_DEV_ID_PCH_CMP_I219_LM12, "Intel(R) I219-LM CMP(12)"),
182 	PVID(0x8086, E1000_DEV_ID_PCH_CMP_I219_V12, "Intel(R) I219-V CMP(12)"),
183 	PVID(0x8086, E1000_DEV_ID_PCH_TGP_I219_LM13, "Intel(R) I219-LM TGP(13)"),
184 	PVID(0x8086, E1000_DEV_ID_PCH_TGP_I219_V13, "Intel(R) I219-V TGP(13)"),
185 	PVID(0x8086, E1000_DEV_ID_PCH_TGP_I219_LM14, "Intel(R) I219-LM TGP(14)"),
186 	PVID(0x8086, E1000_DEV_ID_PCH_TGP_I219_V14, "Intel(R) I219-V GTP(14)"),
187 	PVID(0x8086, E1000_DEV_ID_PCH_TGP_I219_LM15, "Intel(R) I219-LM TGP(15)"),
188 	PVID(0x8086, E1000_DEV_ID_PCH_TGP_I219_V15, "Intel(R) I219-V TGP(15)"),
189 	PVID(0x8086, E1000_DEV_ID_PCH_ADL_I219_LM16, "Intel(R) I219-LM ADL(16)"),
190 	PVID(0x8086, E1000_DEV_ID_PCH_ADL_I219_V16, "Intel(R) I219-V ADL(16)"),
191 	PVID(0x8086, E1000_DEV_ID_PCH_ADL_I219_LM17, "Intel(R) I219-LM ADL(17)"),
192 	PVID(0x8086, E1000_DEV_ID_PCH_ADL_I219_V17, "Intel(R) I219-V ADL(17)"),
193 	PVID(0x8086, E1000_DEV_ID_PCH_MTP_I219_LM18, "Intel(R) I219-LM MTP(18)"),
194 	PVID(0x8086, E1000_DEV_ID_PCH_MTP_I219_V18, "Intel(R) I219-V MTP(18)"),
195 	PVID(0x8086, E1000_DEV_ID_PCH_MTP_I219_LM19, "Intel(R) I219-LM MTP(19)"),
196 	PVID(0x8086, E1000_DEV_ID_PCH_MTP_I219_V19, "Intel(R) I219-V MTP(19)"),
197 	PVID(0x8086, E1000_DEV_ID_PCH_LNL_I219_LM20, "Intel(R) I219-LM LNL(20)"),
198 	PVID(0x8086, E1000_DEV_ID_PCH_LNL_I219_V20, "Intel(R) I219-V LNL(20)"),
199 	PVID(0x8086, E1000_DEV_ID_PCH_LNL_I219_LM21, "Intel(R) I219-LM LNL(21)"),
200 	PVID(0x8086, E1000_DEV_ID_PCH_LNL_I219_V21, "Intel(R) I219-V LNL(21)"),
201 	PVID(0x8086, E1000_DEV_ID_PCH_RPL_I219_LM22, "Intel(R) I219-LM RPL(22)"),
202 	PVID(0x8086, E1000_DEV_ID_PCH_RPL_I219_V22, "Intel(R) I219-V RPL(22)"),
203 	PVID(0x8086, E1000_DEV_ID_PCH_RPL_I219_LM23, "Intel(R) I219-LM RPL(23)"),
204 	PVID(0x8086, E1000_DEV_ID_PCH_RPL_I219_V23, "Intel(R) I219-V RPL(23)"),
205 	PVID(0x8086, E1000_DEV_ID_PCH_ARL_I219_LM24, "Intel(R) I219-LM ARL(24)"),
206 	PVID(0x8086, E1000_DEV_ID_PCH_ARL_I219_V24, "Intel(R) I219-V ARL(24)"),
207 	PVID(0x8086, E1000_DEV_ID_PCH_PTP_I219_LM25, "Intel(R) I219-LM PTP(25)"),
208 	PVID(0x8086, E1000_DEV_ID_PCH_PTP_I219_V25, "Intel(R) I219-V PTP(25)"),
209 	PVID(0x8086, E1000_DEV_ID_PCH_PTP_I219_LM26, "Intel(R) I219-LM PTP(26)"),
210 	PVID(0x8086, E1000_DEV_ID_PCH_PTP_I219_V26, "Intel(R) I219-V PTP(26)"),
211 	PVID(0x8086, E1000_DEV_ID_PCH_PTP_I219_LM27, "Intel(R) I219-LM PTP(27)"),
212 	PVID(0x8086, E1000_DEV_ID_PCH_PTP_I219_V27, "Intel(R) I219-V PTP(27)"),
213 	/* required last entry */
214 	PVID_END
215 };
216 
217 static const pci_vendor_info_t igb_vendor_info_array[] =
218 {
219 	/* Intel(R) - igb-class devices */
220 	PVID(0x8086, E1000_DEV_ID_82575EB_COPPER, "Intel(R) PRO/1000 82575EB (Copper)"),
221 	PVID(0x8086, E1000_DEV_ID_82575EB_FIBER_SERDES, "Intel(R) PRO/1000 82575EB (SERDES)"),
222 	PVID(0x8086, E1000_DEV_ID_82575GB_QUAD_COPPER, "Intel(R) PRO/1000 VT 82575GB (Quad Copper)"),
223 	PVID(0x8086, E1000_DEV_ID_82576, "Intel(R) PRO/1000 82576"),
224 	PVID(0x8086, E1000_DEV_ID_82576_NS, "Intel(R) PRO/1000 82576NS"),
225 	PVID(0x8086, E1000_DEV_ID_82576_NS_SERDES, "Intel(R) PRO/1000 82576NS (SERDES)"),
226 	PVID(0x8086, E1000_DEV_ID_82576_FIBER, "Intel(R) PRO/1000 EF 82576 (Dual Fiber)"),
227 	PVID(0x8086, E1000_DEV_ID_82576_SERDES, "Intel(R) PRO/1000 82576 (Dual SERDES)"),
228 	PVID(0x8086, E1000_DEV_ID_82576_SERDES_QUAD, "Intel(R) PRO/1000 ET 82576 (Quad SERDES)"),
229 	PVID(0x8086, E1000_DEV_ID_82576_QUAD_COPPER, "Intel(R) PRO/1000 ET 82576 (Quad Copper)"),
230 	PVID(0x8086, E1000_DEV_ID_82576_QUAD_COPPER_ET2, "Intel(R) PRO/1000 ET(2) 82576 (Quad Copper)"),
231 	PVID(0x8086, E1000_DEV_ID_82576_VF, "Intel(R) PRO/1000 82576 Virtual Function"),
232 	PVID(0x8086, E1000_DEV_ID_82580_COPPER, "Intel(R) I340 82580 (Copper)"),
233 	PVID(0x8086, E1000_DEV_ID_82580_FIBER, "Intel(R) I340 82580 (Fiber)"),
234 	PVID(0x8086, E1000_DEV_ID_82580_SERDES, "Intel(R) I340 82580 (SERDES)"),
235 	PVID(0x8086, E1000_DEV_ID_82580_SGMII, "Intel(R) I340 82580 (SGMII)"),
236 	PVID(0x8086, E1000_DEV_ID_82580_COPPER_DUAL, "Intel(R) I340-T2 82580 (Dual Copper)"),
237 	PVID(0x8086, E1000_DEV_ID_82580_QUAD_FIBER, "Intel(R) I340-F4 82580 (Quad Fiber)"),
238 	PVID(0x8086, E1000_DEV_ID_DH89XXCC_SERDES, "Intel(R) DH89XXCC (SERDES)"),
239 	PVID(0x8086, E1000_DEV_ID_DH89XXCC_SGMII, "Intel(R) I347-AT4 DH89XXCC"),
240 	PVID(0x8086, E1000_DEV_ID_DH89XXCC_SFP, "Intel(R) DH89XXCC (SFP)"),
241 	PVID(0x8086, E1000_DEV_ID_DH89XXCC_BACKPLANE, "Intel(R) DH89XXCC (Backplane)"),
242 	PVID(0x8086, E1000_DEV_ID_I350_COPPER, "Intel(R) I350 (Copper)"),
243 	PVID(0x8086, E1000_DEV_ID_I350_FIBER, "Intel(R) I350 (Fiber)"),
244 	PVID(0x8086, E1000_DEV_ID_I350_SERDES, "Intel(R) I350 (SERDES)"),
245 	PVID(0x8086, E1000_DEV_ID_I350_SGMII, "Intel(R) I350 (SGMII)"),
246 	PVID(0x8086, E1000_DEV_ID_I350_VF, "Intel(R) I350 Virtual Function"),
247 	PVID(0x8086, E1000_DEV_ID_I210_COPPER, "Intel(R) I210 (Copper)"),
248 	PVID(0x8086, E1000_DEV_ID_I210_COPPER_IT, "Intel(R) I210 IT (Copper)"),
249 	PVID(0x8086, E1000_DEV_ID_I210_COPPER_OEM1, "Intel(R) I210 (OEM)"),
250 	PVID(0x8086, E1000_DEV_ID_I210_COPPER_FLASHLESS, "Intel(R) I210 Flashless (Copper)"),
251 	PVID(0x8086, E1000_DEV_ID_I210_SERDES_FLASHLESS, "Intel(R) I210 Flashless (SERDES)"),
252 	PVID(0x8086, E1000_DEV_ID_I210_SGMII_FLASHLESS, "Intel(R) I210 Flashless (SGMII)"),
253 	PVID(0x8086, E1000_DEV_ID_I210_FIBER, "Intel(R) I210 (Fiber)"),
254 	PVID(0x8086, E1000_DEV_ID_I210_SERDES, "Intel(R) I210 (SERDES)"),
255 	PVID(0x8086, E1000_DEV_ID_I210_SGMII, "Intel(R) I210 (SGMII)"),
256 	PVID(0x8086, E1000_DEV_ID_I211_COPPER, "Intel(R) I211 (Copper)"),
257 	PVID(0x8086, E1000_DEV_ID_I354_BACKPLANE_1GBPS, "Intel(R) I354 (1.0 GbE Backplane)"),
258 	PVID(0x8086, E1000_DEV_ID_I354_BACKPLANE_2_5GBPS, "Intel(R) I354 (2.5 GbE Backplane)"),
259 	PVID(0x8086, E1000_DEV_ID_I354_SGMII, "Intel(R) I354 (SGMII)"),
260 	/* required last entry */
261 	PVID_END
262 };
263 
264 /*********************************************************************
265  *  Function prototypes
266  *********************************************************************/
267 static void	*em_register(device_t);
268 static void	*igb_register(device_t);
269 static int	em_if_attach_pre(if_ctx_t);
270 static int	em_if_attach_post(if_ctx_t);
271 static int	em_if_detach(if_ctx_t);
272 static int	em_if_shutdown(if_ctx_t);
273 static int	em_if_suspend(if_ctx_t);
274 static int	em_if_resume(if_ctx_t);
275 
276 static int	em_if_tx_queues_alloc(if_ctx_t, caddr_t *, uint64_t *, int, int);
277 static int	em_if_rx_queues_alloc(if_ctx_t, caddr_t *, uint64_t *, int, int);
278 static void	em_if_queues_free(if_ctx_t);
279 
280 static uint64_t	em_if_get_counter(if_ctx_t, ift_counter);
281 static void	em_if_init(if_ctx_t);
282 static void	em_if_stop(if_ctx_t);
283 static void	em_if_media_status(if_ctx_t, struct ifmediareq *);
284 static int	em_if_media_change(if_ctx_t);
285 static int	em_if_mtu_set(if_ctx_t, uint32_t);
286 static void	em_if_timer(if_ctx_t, uint16_t);
287 static void	em_if_vlan_register(if_ctx_t, u16);
288 static void	em_if_vlan_unregister(if_ctx_t, u16);
289 static void	em_if_watchdog_reset(if_ctx_t);
290 static bool	em_if_needs_restart(if_ctx_t, enum iflib_restart_event);
291 
292 static void	em_identify_hardware(if_ctx_t);
293 static int	em_allocate_pci_resources(if_ctx_t);
294 static void	em_free_pci_resources(if_ctx_t);
295 static void	em_reset(if_ctx_t);
296 static int	em_setup_interface(if_ctx_t);
297 static int	em_setup_msix(if_ctx_t);
298 
299 static void	em_initialize_transmit_unit(if_ctx_t);
300 static void	em_initialize_receive_unit(if_ctx_t);
301 
302 static void	em_if_intr_enable(if_ctx_t);
303 static void	em_if_intr_disable(if_ctx_t);
304 static void	igb_if_intr_enable(if_ctx_t);
305 static void	igb_if_intr_disable(if_ctx_t);
306 static int	em_if_rx_queue_intr_enable(if_ctx_t, uint16_t);
307 static int	em_if_tx_queue_intr_enable(if_ctx_t, uint16_t);
308 static int	igb_if_rx_queue_intr_enable(if_ctx_t, uint16_t);
309 static int	igb_if_tx_queue_intr_enable(if_ctx_t, uint16_t);
310 static void	em_if_multi_set(if_ctx_t);
311 static void	em_if_update_admin_status(if_ctx_t);
312 static void	em_if_debug(if_ctx_t);
313 static void	em_update_stats_counters(struct e1000_softc *);
314 static void	em_add_hw_stats(struct e1000_softc *);
315 static int	em_if_set_promisc(if_ctx_t, int);
316 static bool	em_if_vlan_filter_capable(if_ctx_t);
317 static bool	em_if_vlan_filter_used(if_ctx_t);
318 static void	em_if_vlan_filter_enable(struct e1000_softc *);
319 static void	em_if_vlan_filter_disable(struct e1000_softc *);
320 static void	em_if_vlan_filter_write(struct e1000_softc *);
321 static void	em_setup_vlan_hw_support(if_ctx_t ctx);
322 static int	em_sysctl_nvm_info(SYSCTL_HANDLER_ARGS);
323 static void	em_print_nvm_info(struct e1000_softc *);
324 static void	em_fw_version_locked(if_ctx_t);
325 static void	em_sbuf_fw_version(struct e1000_fw_version *, struct sbuf *);
326 static void	em_print_fw_version(struct e1000_softc *);
327 static int	em_sysctl_print_fw_version(SYSCTL_HANDLER_ARGS);
328 static int	em_sysctl_debug_info(SYSCTL_HANDLER_ARGS);
329 static int	em_get_rs(SYSCTL_HANDLER_ARGS);
330 static void	em_print_debug_info(struct e1000_softc *);
331 static int 	em_is_valid_ether_addr(u8 *);
332 static bool	em_automask_tso(if_ctx_t);
333 static int	em_sysctl_int_delay(SYSCTL_HANDLER_ARGS);
334 static void	em_add_int_delay_sysctl(struct e1000_softc *, const char *,
335 		    const char *, struct em_int_delay_info *, int, int);
336 /* Management and WOL Support */
337 static void	em_init_manageability(struct e1000_softc *);
338 static void	em_release_manageability(struct e1000_softc *);
339 static void	em_get_hw_control(struct e1000_softc *);
340 static void	em_release_hw_control(struct e1000_softc *);
341 static void	em_get_wakeup(if_ctx_t);
342 static void	em_enable_wakeup(if_ctx_t);
343 static int	em_enable_phy_wakeup(struct e1000_softc *);
344 static void	em_disable_aspm(struct e1000_softc *);
345 
346 int		em_intr(void *);
347 
348 /* MSI-X handlers */
349 static int	em_if_msix_intr_assign(if_ctx_t, int);
350 static int	em_msix_link(void *);
351 static void	em_handle_link(void *);
352 
353 static void	em_enable_vectors_82574(if_ctx_t);
354 
355 static int	em_set_flowcntl(SYSCTL_HANDLER_ARGS);
356 static int	em_sysctl_eee(SYSCTL_HANDLER_ARGS);
357 static void	em_if_led_func(if_ctx_t, int);
358 
359 static int	em_get_regs(SYSCTL_HANDLER_ARGS);
360 
361 static void	lem_smartspeed(struct e1000_softc *);
362 static void	igb_configure_queues(struct e1000_softc *);
363 static void	em_flush_desc_rings(struct e1000_softc *);
364 
365 
366 /*********************************************************************
367  *  FreeBSD Device Interface Entry Points
368  *********************************************************************/
369 static device_method_t em_methods[] = {
370 	/* Device interface */
371 	DEVMETHOD(device_register, em_register),
372 	DEVMETHOD(device_probe, iflib_device_probe),
373 	DEVMETHOD(device_attach, iflib_device_attach),
374 	DEVMETHOD(device_detach, iflib_device_detach),
375 	DEVMETHOD(device_shutdown, iflib_device_shutdown),
376 	DEVMETHOD(device_suspend, iflib_device_suspend),
377 	DEVMETHOD(device_resume, iflib_device_resume),
378 	DEVMETHOD_END
379 };
380 
381 static device_method_t igb_methods[] = {
382 	/* Device interface */
383 	DEVMETHOD(device_register, igb_register),
384 	DEVMETHOD(device_probe, iflib_device_probe),
385 	DEVMETHOD(device_attach, iflib_device_attach),
386 	DEVMETHOD(device_detach, iflib_device_detach),
387 	DEVMETHOD(device_shutdown, iflib_device_shutdown),
388 	DEVMETHOD(device_suspend, iflib_device_suspend),
389 	DEVMETHOD(device_resume, iflib_device_resume),
390 	DEVMETHOD_END
391 };
392 
393 
394 static driver_t em_driver = {
395 	"em", em_methods, sizeof(struct e1000_softc),
396 };
397 
398 DRIVER_MODULE(em, pci, em_driver, 0, 0);
399 
400 MODULE_DEPEND(em, pci, 1, 1, 1);
401 MODULE_DEPEND(em, ether, 1, 1, 1);
402 MODULE_DEPEND(em, iflib, 1, 1, 1);
403 
404 IFLIB_PNP_INFO(pci, em, em_vendor_info_array);
405 
406 static driver_t igb_driver = {
407 	"igb", igb_methods, sizeof(struct e1000_softc),
408 };
409 
410 DRIVER_MODULE(igb, pci, igb_driver, 0, 0);
411 
412 MODULE_DEPEND(igb, pci, 1, 1, 1);
413 MODULE_DEPEND(igb, ether, 1, 1, 1);
414 MODULE_DEPEND(igb, iflib, 1, 1, 1);
415 
416 IFLIB_PNP_INFO(pci, igb, igb_vendor_info_array);
417 
418 static device_method_t em_if_methods[] = {
419 	DEVMETHOD(ifdi_attach_pre, em_if_attach_pre),
420 	DEVMETHOD(ifdi_attach_post, em_if_attach_post),
421 	DEVMETHOD(ifdi_detach, em_if_detach),
422 	DEVMETHOD(ifdi_shutdown, em_if_shutdown),
423 	DEVMETHOD(ifdi_suspend, em_if_suspend),
424 	DEVMETHOD(ifdi_resume, em_if_resume),
425 	DEVMETHOD(ifdi_init, em_if_init),
426 	DEVMETHOD(ifdi_stop, em_if_stop),
427 	DEVMETHOD(ifdi_msix_intr_assign, em_if_msix_intr_assign),
428 	DEVMETHOD(ifdi_intr_enable, em_if_intr_enable),
429 	DEVMETHOD(ifdi_intr_disable, em_if_intr_disable),
430 	DEVMETHOD(ifdi_tx_queues_alloc, em_if_tx_queues_alloc),
431 	DEVMETHOD(ifdi_rx_queues_alloc, em_if_rx_queues_alloc),
432 	DEVMETHOD(ifdi_queues_free, em_if_queues_free),
433 	DEVMETHOD(ifdi_update_admin_status, em_if_update_admin_status),
434 	DEVMETHOD(ifdi_multi_set, em_if_multi_set),
435 	DEVMETHOD(ifdi_media_status, em_if_media_status),
436 	DEVMETHOD(ifdi_media_change, em_if_media_change),
437 	DEVMETHOD(ifdi_mtu_set, em_if_mtu_set),
438 	DEVMETHOD(ifdi_promisc_set, em_if_set_promisc),
439 	DEVMETHOD(ifdi_timer, em_if_timer),
440 	DEVMETHOD(ifdi_watchdog_reset, em_if_watchdog_reset),
441 	DEVMETHOD(ifdi_vlan_register, em_if_vlan_register),
442 	DEVMETHOD(ifdi_vlan_unregister, em_if_vlan_unregister),
443 	DEVMETHOD(ifdi_get_counter, em_if_get_counter),
444 	DEVMETHOD(ifdi_led_func, em_if_led_func),
445 	DEVMETHOD(ifdi_rx_queue_intr_enable, em_if_rx_queue_intr_enable),
446 	DEVMETHOD(ifdi_tx_queue_intr_enable, em_if_tx_queue_intr_enable),
447 	DEVMETHOD(ifdi_debug, em_if_debug),
448 	DEVMETHOD(ifdi_needs_restart, em_if_needs_restart),
449 	DEVMETHOD_END
450 };
451 
452 static driver_t em_if_driver = {
453 	"em_if", em_if_methods, sizeof(struct e1000_softc)
454 };
455 
456 static device_method_t igb_if_methods[] = {
457 	DEVMETHOD(ifdi_attach_pre, em_if_attach_pre),
458 	DEVMETHOD(ifdi_attach_post, em_if_attach_post),
459 	DEVMETHOD(ifdi_detach, em_if_detach),
460 	DEVMETHOD(ifdi_shutdown, em_if_shutdown),
461 	DEVMETHOD(ifdi_suspend, em_if_suspend),
462 	DEVMETHOD(ifdi_resume, em_if_resume),
463 	DEVMETHOD(ifdi_init, em_if_init),
464 	DEVMETHOD(ifdi_stop, em_if_stop),
465 	DEVMETHOD(ifdi_msix_intr_assign, em_if_msix_intr_assign),
466 	DEVMETHOD(ifdi_intr_enable, igb_if_intr_enable),
467 	DEVMETHOD(ifdi_intr_disable, igb_if_intr_disable),
468 	DEVMETHOD(ifdi_tx_queues_alloc, em_if_tx_queues_alloc),
469 	DEVMETHOD(ifdi_rx_queues_alloc, em_if_rx_queues_alloc),
470 	DEVMETHOD(ifdi_queues_free, em_if_queues_free),
471 	DEVMETHOD(ifdi_update_admin_status, em_if_update_admin_status),
472 	DEVMETHOD(ifdi_multi_set, em_if_multi_set),
473 	DEVMETHOD(ifdi_media_status, em_if_media_status),
474 	DEVMETHOD(ifdi_media_change, em_if_media_change),
475 	DEVMETHOD(ifdi_mtu_set, em_if_mtu_set),
476 	DEVMETHOD(ifdi_promisc_set, em_if_set_promisc),
477 	DEVMETHOD(ifdi_timer, em_if_timer),
478 	DEVMETHOD(ifdi_watchdog_reset, em_if_watchdog_reset),
479 	DEVMETHOD(ifdi_vlan_register, em_if_vlan_register),
480 	DEVMETHOD(ifdi_vlan_unregister, em_if_vlan_unregister),
481 	DEVMETHOD(ifdi_get_counter, em_if_get_counter),
482 	DEVMETHOD(ifdi_led_func, em_if_led_func),
483 	DEVMETHOD(ifdi_rx_queue_intr_enable, igb_if_rx_queue_intr_enable),
484 	DEVMETHOD(ifdi_tx_queue_intr_enable, igb_if_tx_queue_intr_enable),
485 	DEVMETHOD(ifdi_debug, em_if_debug),
486 	DEVMETHOD(ifdi_needs_restart, em_if_needs_restart),
487 	DEVMETHOD_END
488 };
489 
490 static driver_t igb_if_driver = {
491 	"igb_if", igb_if_methods, sizeof(struct e1000_softc)
492 };
493 
494 /*********************************************************************
495  *  Tunable default values.
496  *********************************************************************/
497 
498 #define EM_TICKS_TO_USECS(ticks)	((1024 * (ticks) + 500) / 1000)
499 #define EM_USECS_TO_TICKS(usecs)	((1000 * (usecs) + 512) / 1024)
500 
501 #define MAX_INTS_PER_SEC	8000
502 #define DEFAULT_ITR		(1000000000/(MAX_INTS_PER_SEC * 256))
503 
504 /* Allow common code without TSO */
505 #ifndef CSUM_TSO
506 #define CSUM_TSO	0
507 #endif
508 
509 static SYSCTL_NODE(_hw, OID_AUTO, em, CTLFLAG_RD | CTLFLAG_MPSAFE, 0,
510     "EM driver parameters");
511 
512 static int em_disable_crc_stripping = 0;
513 SYSCTL_INT(_hw_em, OID_AUTO, disable_crc_stripping, CTLFLAG_RDTUN,
514     &em_disable_crc_stripping, 0, "Disable CRC Stripping");
515 
516 static int em_tx_int_delay_dflt = EM_TICKS_TO_USECS(EM_TIDV);
517 static int em_rx_int_delay_dflt = EM_TICKS_TO_USECS(EM_RDTR);
518 SYSCTL_INT(_hw_em, OID_AUTO, tx_int_delay, CTLFLAG_RDTUN, &em_tx_int_delay_dflt,
519     0, "Default transmit interrupt delay in usecs");
520 SYSCTL_INT(_hw_em, OID_AUTO, rx_int_delay, CTLFLAG_RDTUN, &em_rx_int_delay_dflt,
521     0, "Default receive interrupt delay in usecs");
522 
523 static int em_tx_abs_int_delay_dflt = EM_TICKS_TO_USECS(EM_TADV);
524 static int em_rx_abs_int_delay_dflt = EM_TICKS_TO_USECS(EM_RADV);
525 SYSCTL_INT(_hw_em, OID_AUTO, tx_abs_int_delay, CTLFLAG_RDTUN,
526     &em_tx_abs_int_delay_dflt, 0,
527     "Default transmit interrupt delay limit in usecs");
528 SYSCTL_INT(_hw_em, OID_AUTO, rx_abs_int_delay, CTLFLAG_RDTUN,
529     &em_rx_abs_int_delay_dflt, 0,
530     "Default receive interrupt delay limit in usecs");
531 
532 static int em_smart_pwr_down = false;
533 SYSCTL_INT(_hw_em, OID_AUTO, smart_pwr_down, CTLFLAG_RDTUN, &em_smart_pwr_down,
534     0, "Set to true to leave smart power down enabled on newer adapters");
535 
536 static bool em_unsupported_tso = false;
537 SYSCTL_BOOL(_hw_em, OID_AUTO, unsupported_tso, CTLFLAG_RDTUN,
538     &em_unsupported_tso, 0, "Allow unsupported em(4) TSO configurations");
539 
540 /* Controls whether promiscuous also shows bad packets */
541 static int em_debug_sbp = false;
542 SYSCTL_INT(_hw_em, OID_AUTO, sbp, CTLFLAG_RDTUN, &em_debug_sbp, 0,
543     "Show bad packets in promiscuous mode");
544 
545 /* How many packets rxeof tries to clean at a time */
546 static int em_rx_process_limit = 100;
547 SYSCTL_INT(_hw_em, OID_AUTO, rx_process_limit, CTLFLAG_RDTUN,
548     &em_rx_process_limit, 0,
549     "Maximum number of received packets to process "
550     "at a time, -1 means unlimited");
551 
552 /* Energy efficient ethernet - default to OFF */
553 static int eee_setting = 1;
554 SYSCTL_INT(_hw_em, OID_AUTO, eee_setting, CTLFLAG_RDTUN, &eee_setting, 0,
555     "Enable Energy Efficient Ethernet");
556 
557 /*
558 ** Tuneable Interrupt rate
559 */
560 static int em_max_interrupt_rate = 8000;
561 SYSCTL_INT(_hw_em, OID_AUTO, max_interrupt_rate, CTLFLAG_RDTUN,
562     &em_max_interrupt_rate, 0, "Maximum interrupts per second");
563 
564 
565 
566 /* Global used in WOL setup with multiport cards */
567 static int global_quad_port_a = 0;
568 
569 extern struct if_txrx igb_txrx;
570 extern struct if_txrx em_txrx;
571 extern struct if_txrx lem_txrx;
572 
573 static struct if_shared_ctx em_sctx_init = {
574 	.isc_magic = IFLIB_MAGIC,
575 	.isc_q_align = PAGE_SIZE,
576 	.isc_tx_maxsize = EM_TSO_SIZE + sizeof(struct ether_vlan_header),
577 	.isc_tx_maxsegsize = PAGE_SIZE,
578 	.isc_tso_maxsize = EM_TSO_SIZE + sizeof(struct ether_vlan_header),
579 	.isc_tso_maxsegsize = EM_TSO_SEG_SIZE,
580 	.isc_rx_maxsize = MJUM9BYTES,
581 	.isc_rx_nsegments = 1,
582 	.isc_rx_maxsegsize = MJUM9BYTES,
583 	.isc_nfl = 1,
584 	.isc_nrxqs = 1,
585 	.isc_ntxqs = 1,
586 	.isc_admin_intrcnt = 1,
587 	.isc_vendor_info = em_vendor_info_array,
588 	.isc_driver_version = em_driver_version,
589 	.isc_driver = &em_if_driver,
590 	.isc_flags = IFLIB_NEED_SCRATCH | IFLIB_TSO_INIT_IP | IFLIB_NEED_ZERO_CSUM,
591 
592 	.isc_nrxd_min = {EM_MIN_RXD},
593 	.isc_ntxd_min = {EM_MIN_TXD},
594 	.isc_nrxd_max = {EM_MAX_RXD},
595 	.isc_ntxd_max = {EM_MAX_TXD},
596 	.isc_nrxd_default = {EM_DEFAULT_RXD},
597 	.isc_ntxd_default = {EM_DEFAULT_TXD},
598 };
599 
600 static struct if_shared_ctx igb_sctx_init = {
601 	.isc_magic = IFLIB_MAGIC,
602 	.isc_q_align = PAGE_SIZE,
603 	.isc_tx_maxsize = EM_TSO_SIZE + sizeof(struct ether_vlan_header),
604 	.isc_tx_maxsegsize = PAGE_SIZE,
605 	.isc_tso_maxsize = EM_TSO_SIZE + sizeof(struct ether_vlan_header),
606 	.isc_tso_maxsegsize = EM_TSO_SEG_SIZE,
607 	.isc_rx_maxsize = MJUM9BYTES,
608 	.isc_rx_nsegments = 1,
609 	.isc_rx_maxsegsize = MJUM9BYTES,
610 	.isc_nfl = 1,
611 	.isc_nrxqs = 1,
612 	.isc_ntxqs = 1,
613 	.isc_admin_intrcnt = 1,
614 	.isc_vendor_info = igb_vendor_info_array,
615 	.isc_driver_version = igb_driver_version,
616 	.isc_driver = &igb_if_driver,
617 	.isc_flags = IFLIB_NEED_SCRATCH | IFLIB_TSO_INIT_IP | IFLIB_NEED_ZERO_CSUM,
618 
619 	.isc_nrxd_min = {EM_MIN_RXD},
620 	.isc_ntxd_min = {EM_MIN_TXD},
621 	.isc_nrxd_max = {IGB_MAX_RXD},
622 	.isc_ntxd_max = {IGB_MAX_TXD},
623 	.isc_nrxd_default = {EM_DEFAULT_RXD},
624 	.isc_ntxd_default = {EM_DEFAULT_TXD},
625 };
626 
627 /*****************************************************************
628  *
629  * Dump Registers
630  *
631  ****************************************************************/
632 #define IGB_REGS_LEN 739
633 
634 static int em_get_regs(SYSCTL_HANDLER_ARGS)
635 {
636 	struct e1000_softc *sc = (struct e1000_softc *)arg1;
637 	struct e1000_hw *hw = &sc->hw;
638 	struct sbuf *sb;
639 	u32 *regs_buff;
640 	int rc;
641 
642 	regs_buff = malloc(sizeof(u32) * IGB_REGS_LEN, M_DEVBUF, M_WAITOK);
643 	memset(regs_buff, 0, IGB_REGS_LEN * sizeof(u32));
644 
645 	rc = sysctl_wire_old_buffer(req, 0);
646 	MPASS(rc == 0);
647 	if (rc != 0) {
648 		free(regs_buff, M_DEVBUF);
649 		return (rc);
650 	}
651 
652 	sb = sbuf_new_for_sysctl(NULL, NULL, 32*400, req);
653 	MPASS(sb != NULL);
654 	if (sb == NULL) {
655 		free(regs_buff, M_DEVBUF);
656 		return (ENOMEM);
657 	}
658 
659 	/* General Registers */
660 	regs_buff[0] = E1000_READ_REG(hw, E1000_CTRL);
661 	regs_buff[1] = E1000_READ_REG(hw, E1000_STATUS);
662 	regs_buff[2] = E1000_READ_REG(hw, E1000_CTRL_EXT);
663 	regs_buff[3] = E1000_READ_REG(hw, E1000_ICR);
664 	regs_buff[4] = E1000_READ_REG(hw, E1000_RCTL);
665 	regs_buff[5] = E1000_READ_REG(hw, E1000_RDLEN(0));
666 	regs_buff[6] = E1000_READ_REG(hw, E1000_RDH(0));
667 	regs_buff[7] = E1000_READ_REG(hw, E1000_RDT(0));
668 	regs_buff[8] = E1000_READ_REG(hw, E1000_RXDCTL(0));
669 	regs_buff[9] = E1000_READ_REG(hw, E1000_RDBAL(0));
670 	regs_buff[10] = E1000_READ_REG(hw, E1000_RDBAH(0));
671 	regs_buff[11] = E1000_READ_REG(hw, E1000_TCTL);
672 	regs_buff[12] = E1000_READ_REG(hw, E1000_TDBAL(0));
673 	regs_buff[13] = E1000_READ_REG(hw, E1000_TDBAH(0));
674 	regs_buff[14] = E1000_READ_REG(hw, E1000_TDLEN(0));
675 	regs_buff[15] = E1000_READ_REG(hw, E1000_TDH(0));
676 	regs_buff[16] = E1000_READ_REG(hw, E1000_TDT(0));
677 	regs_buff[17] = E1000_READ_REG(hw, E1000_TXDCTL(0));
678 	regs_buff[18] = E1000_READ_REG(hw, E1000_TDFH);
679 	regs_buff[19] = E1000_READ_REG(hw, E1000_TDFT);
680 	regs_buff[20] = E1000_READ_REG(hw, E1000_TDFHS);
681 	regs_buff[21] = E1000_READ_REG(hw, E1000_TDFPC);
682 
683 	sbuf_printf(sb, "General Registers\n");
684 	sbuf_printf(sb, "\tCTRL\t %08x\n", regs_buff[0]);
685 	sbuf_printf(sb, "\tSTATUS\t %08x\n", regs_buff[1]);
686 	sbuf_printf(sb, "\tCTRL_EXT\t %08x\n\n", regs_buff[2]);
687 
688 	sbuf_printf(sb, "Interrupt Registers\n");
689 	sbuf_printf(sb, "\tICR\t %08x\n\n", regs_buff[3]);
690 
691 	sbuf_printf(sb, "RX Registers\n");
692 	sbuf_printf(sb, "\tRCTL\t %08x\n", regs_buff[4]);
693 	sbuf_printf(sb, "\tRDLEN\t %08x\n", regs_buff[5]);
694 	sbuf_printf(sb, "\tRDH\t %08x\n", regs_buff[6]);
695 	sbuf_printf(sb, "\tRDT\t %08x\n", regs_buff[7]);
696 	sbuf_printf(sb, "\tRXDCTL\t %08x\n", regs_buff[8]);
697 	sbuf_printf(sb, "\tRDBAL\t %08x\n", regs_buff[9]);
698 	sbuf_printf(sb, "\tRDBAH\t %08x\n\n", regs_buff[10]);
699 
700 	sbuf_printf(sb, "TX Registers\n");
701 	sbuf_printf(sb, "\tTCTL\t %08x\n", regs_buff[11]);
702 	sbuf_printf(sb, "\tTDBAL\t %08x\n", regs_buff[12]);
703 	sbuf_printf(sb, "\tTDBAH\t %08x\n", regs_buff[13]);
704 	sbuf_printf(sb, "\tTDLEN\t %08x\n", regs_buff[14]);
705 	sbuf_printf(sb, "\tTDH\t %08x\n", regs_buff[15]);
706 	sbuf_printf(sb, "\tTDT\t %08x\n", regs_buff[16]);
707 	sbuf_printf(sb, "\tTXDCTL\t %08x\n", regs_buff[17]);
708 	sbuf_printf(sb, "\tTDFH\t %08x\n", regs_buff[18]);
709 	sbuf_printf(sb, "\tTDFT\t %08x\n", regs_buff[19]);
710 	sbuf_printf(sb, "\tTDFHS\t %08x\n", regs_buff[20]);
711 	sbuf_printf(sb, "\tTDFPC\t %08x\n\n", regs_buff[21]);
712 
713 	free(regs_buff, M_DEVBUF);
714 
715 #ifdef DUMP_DESCS
716 	{
717 		if_softc_ctx_t scctx = sc->shared;
718 		struct rx_ring *rxr = &rx_que->rxr;
719 		struct tx_ring *txr = &tx_que->txr;
720 		int ntxd = scctx->isc_ntxd[0];
721 		int nrxd = scctx->isc_nrxd[0];
722 		int j;
723 
724 	for (j = 0; j < nrxd; j++) {
725 		u32 staterr = le32toh(rxr->rx_base[j].wb.upper.status_error);
726 		u32 length =  le32toh(rxr->rx_base[j].wb.upper.length);
727 		sbuf_printf(sb, "\tReceive Descriptor Address %d: %08" PRIx64 "  Error:%d  Length:%d\n", j, rxr->rx_base[j].read.buffer_addr, staterr, length);
728 	}
729 
730 	for (j = 0; j < min(ntxd, 256); j++) {
731 		unsigned int *ptr = (unsigned int *)&txr->tx_base[j];
732 
733 		sbuf_printf(sb, "\tTXD[%03d] [0]: %08x [1]: %08x [2]: %08x [3]: %08x  eop: %d DD=%d\n",
734 			    j, ptr[0], ptr[1], ptr[2], ptr[3], buf->eop,
735 			    buf->eop != -1 ? txr->tx_base[buf->eop].upper.fields.status & E1000_TXD_STAT_DD : 0);
736 
737 	}
738 	}
739 #endif
740 
741 	rc = sbuf_finish(sb);
742 	sbuf_delete(sb);
743 	return(rc);
744 }
745 
746 static void *
747 em_register(device_t dev)
748 {
749 	return (&em_sctx_init);
750 }
751 
752 static void *
753 igb_register(device_t dev)
754 {
755 	return (&igb_sctx_init);
756 }
757 
758 static int
759 em_set_num_queues(if_ctx_t ctx)
760 {
761 	struct e1000_softc *sc = iflib_get_softc(ctx);
762 	int maxqueues;
763 
764 	/* Sanity check based on HW */
765 	switch (sc->hw.mac.type) {
766 	case e1000_82576:
767 	case e1000_82580:
768 	case e1000_i350:
769 	case e1000_i354:
770 		maxqueues = 8;
771 		break;
772 	case e1000_i210:
773 	case e1000_82575:
774 		maxqueues = 4;
775 		break;
776 	case e1000_i211:
777 	case e1000_82574:
778 		maxqueues = 2;
779 		break;
780 	default:
781 		maxqueues = 1;
782 		break;
783 	}
784 
785 	return (maxqueues);
786 }
787 
788 #define LEM_CAPS \
789     IFCAP_HWCSUM | IFCAP_VLAN_MTU | IFCAP_VLAN_HWTAGGING | \
790     IFCAP_VLAN_HWCSUM | IFCAP_WOL | IFCAP_VLAN_HWFILTER | IFCAP_TSO4 | \
791     IFCAP_LRO | IFCAP_VLAN_HWTSO | IFCAP_JUMBO_MTU | IFCAP_HWCSUM_IPV6
792 
793 #define EM_CAPS \
794     IFCAP_HWCSUM | IFCAP_VLAN_MTU | IFCAP_VLAN_HWTAGGING | \
795     IFCAP_VLAN_HWCSUM | IFCAP_WOL | IFCAP_VLAN_HWFILTER | IFCAP_TSO4 | \
796     IFCAP_LRO | IFCAP_VLAN_HWTSO | IFCAP_JUMBO_MTU | IFCAP_HWCSUM_IPV6 | \
797     IFCAP_TSO6
798 
799 #define IGB_CAPS \
800     IFCAP_HWCSUM | IFCAP_VLAN_MTU | IFCAP_VLAN_HWTAGGING | \
801     IFCAP_VLAN_HWCSUM | IFCAP_WOL | IFCAP_VLAN_HWFILTER | IFCAP_TSO4 | \
802     IFCAP_LRO | IFCAP_VLAN_HWTSO | IFCAP_JUMBO_MTU | IFCAP_HWCSUM_IPV6 | \
803     IFCAP_TSO6
804 
805 /*********************************************************************
806  *  Device initialization routine
807  *
808  *  The attach entry point is called when the driver is being loaded.
809  *  This routine identifies the type of hardware, allocates all resources
810  *  and initializes the hardware.
811  *
812  *  return 0 on success, positive on failure
813  *********************************************************************/
814 static int
815 em_if_attach_pre(if_ctx_t ctx)
816 {
817 	struct e1000_softc *sc;
818 	if_softc_ctx_t scctx;
819 	device_t dev;
820 	struct e1000_hw *hw;
821 	struct sysctl_oid_list *child;
822 	struct sysctl_ctx_list *ctx_list;
823 	int error = 0;
824 
825 	INIT_DEBUGOUT("em_if_attach_pre: begin");
826 	dev = iflib_get_dev(ctx);
827 	sc = iflib_get_softc(ctx);
828 
829 	sc->ctx = sc->osdep.ctx = ctx;
830 	sc->dev = sc->osdep.dev = dev;
831 	scctx = sc->shared = iflib_get_softc_ctx(ctx);
832 	sc->media = iflib_get_media(ctx);
833 	hw = &sc->hw;
834 
835 	sc->tx_process_limit = scctx->isc_ntxd[0];
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 | CTLFLAG_NEEDGIANT, sc, 0,
846 	    em_sysctl_nvm_info, "I", "NVM Information");
847 
848 	SYSCTL_ADD_PROC(ctx_list, child, OID_AUTO, "fw_version",
849 	    CTLTYPE_STRING | CTLFLAG_RD | CTLFLAG_NEEDGIANT, sc, 0,
850 	    em_sysctl_print_fw_version, "A",
851 	    "Prints FW/NVM Versions");
852 
853 	SYSCTL_ADD_PROC(ctx_list, child, OID_AUTO, "debug",
854 	    CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_NEEDGIANT, sc, 0,
855 	    em_sysctl_debug_info, "I", "Debug Information");
856 
857 	SYSCTL_ADD_PROC(ctx_list, child, OID_AUTO, "fc",
858 	    CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_NEEDGIANT, sc, 0,
859 	    em_set_flowcntl, "I", "Flow Control");
860 
861 	SYSCTL_ADD_PROC(ctx_list, child, OID_AUTO, "reg_dump",
862 	    CTLTYPE_STRING | CTLFLAG_RD | CTLFLAG_NEEDGIANT, sc, 0,
863 	    em_get_regs, "A", "Dump Registers");
864 
865 	SYSCTL_ADD_PROC(ctx_list, child, OID_AUTO, "rs_dump",
866 	    CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_NEEDGIANT, sc, 0,
867 	    em_get_rs, "I", "Dump RS indexes");
868 
869 	scctx->isc_tx_nsegments = EM_MAX_SCATTER;
870 	scctx->isc_nrxqsets_max = scctx->isc_ntxqsets_max = em_set_num_queues(ctx);
871 	if (bootverbose)
872 		device_printf(dev, "attach_pre capping queues at %d\n",
873 		    scctx->isc_ntxqsets_max);
874 
875 	if (hw->mac.type >= igb_mac_min) {
876 		scctx->isc_txqsizes[0] = roundup2(scctx->isc_ntxd[0] * sizeof(union e1000_adv_tx_desc), EM_DBA_ALIGN);
877 		scctx->isc_rxqsizes[0] = roundup2(scctx->isc_nrxd[0] * sizeof(union e1000_adv_rx_desc), EM_DBA_ALIGN);
878 		scctx->isc_txd_size[0] = sizeof(union e1000_adv_tx_desc);
879 		scctx->isc_rxd_size[0] = sizeof(union e1000_adv_rx_desc);
880 		scctx->isc_txrx = &igb_txrx;
881 		scctx->isc_tx_tso_segments_max = EM_MAX_SCATTER;
882 		scctx->isc_tx_tso_size_max = EM_TSO_SIZE;
883 		scctx->isc_tx_tso_segsize_max = EM_TSO_SEG_SIZE;
884 		scctx->isc_capabilities = scctx->isc_capenable = IGB_CAPS;
885 		scctx->isc_tx_csum_flags = CSUM_TCP | CSUM_UDP | CSUM_TSO |
886 		     CSUM_IP6_TCP | CSUM_IP6_UDP;
887 		if (hw->mac.type != e1000_82575)
888 			scctx->isc_tx_csum_flags |= CSUM_SCTP | CSUM_IP6_SCTP;
889 		/*
890 		** Some new devices, as with ixgbe, now may
891 		** use a different BAR, so we need to keep
892 		** track of which is used.
893 		*/
894 		scctx->isc_msix_bar = pci_msix_table_bar(dev);
895 	} else if (hw->mac.type >= em_mac_min) {
896 		scctx->isc_txqsizes[0] = roundup2(scctx->isc_ntxd[0]* sizeof(struct e1000_tx_desc), EM_DBA_ALIGN);
897 		scctx->isc_rxqsizes[0] = roundup2(scctx->isc_nrxd[0] * sizeof(union e1000_rx_desc_extended), EM_DBA_ALIGN);
898 		scctx->isc_txd_size[0] = sizeof(struct e1000_tx_desc);
899 		scctx->isc_rxd_size[0] = sizeof(union e1000_rx_desc_extended);
900 		scctx->isc_txrx = &em_txrx;
901 		scctx->isc_tx_tso_segments_max = EM_MAX_SCATTER;
902 		scctx->isc_tx_tso_size_max = EM_TSO_SIZE;
903 		scctx->isc_tx_tso_segsize_max = EM_TSO_SEG_SIZE;
904 		scctx->isc_capabilities = scctx->isc_capenable = EM_CAPS;
905 		scctx->isc_tx_csum_flags = CSUM_TCP | CSUM_UDP | CSUM_IP_TSO |
906 		    CSUM_IP6_TCP | CSUM_IP6_UDP;
907 
908 		/* Disable TSO on all em(4) until ring stalls can be debugged */
909 		scctx->isc_capenable &= ~IFCAP_TSO;
910 
911 		/*
912 		 * Disable TSO on SPT due to errata that downclocks DMA performance
913 		 * i218-i219 Specification Update 1.5.4.5
914 		 */
915 		if (hw->mac.type == e1000_pch_spt)
916 			scctx->isc_capenable &= ~IFCAP_TSO;
917 
918 		/*
919 		 * We support MSI-X with 82574 only, but indicate to iflib(4)
920 		 * that it shall give MSI at least a try with other devices.
921 		 */
922 		if (hw->mac.type == e1000_82574) {
923 			scctx->isc_msix_bar = pci_msix_table_bar(dev);
924 		} else {
925 			scctx->isc_msix_bar = -1;
926 			scctx->isc_disable_msix = 1;
927 		}
928 	} else {
929 		scctx->isc_txqsizes[0] = roundup2((scctx->isc_ntxd[0] + 1) * sizeof(struct e1000_tx_desc), EM_DBA_ALIGN);
930 		scctx->isc_rxqsizes[0] = roundup2((scctx->isc_nrxd[0] + 1) * sizeof(struct e1000_rx_desc), EM_DBA_ALIGN);
931 		scctx->isc_txd_size[0] = sizeof(struct e1000_tx_desc);
932 		scctx->isc_rxd_size[0] = sizeof(struct e1000_rx_desc);
933 		scctx->isc_txrx = &lem_txrx;
934 		scctx->isc_tx_tso_segments_max = EM_MAX_SCATTER;
935 		scctx->isc_tx_tso_size_max = EM_TSO_SIZE;
936 		scctx->isc_tx_tso_segsize_max = EM_TSO_SEG_SIZE;
937 		scctx->isc_capabilities = scctx->isc_capenable = LEM_CAPS;
938 		if (em_unsupported_tso)
939 			scctx->isc_capabilities |= IFCAP_TSO6;
940 		scctx->isc_tx_csum_flags = CSUM_TCP | CSUM_UDP | CSUM_IP_TSO |
941 		    CSUM_IP6_TCP | CSUM_IP6_UDP;
942 
943 		/* Disable TSO on all lem(4) until ring stalls can be debugged */
944 		scctx->isc_capenable &= ~IFCAP_TSO;
945 
946 		/* 82541ER doesn't do HW tagging */
947 		if (hw->device_id == E1000_DEV_ID_82541ER ||
948 		    hw->device_id == E1000_DEV_ID_82541ER_LOM) {
949 			scctx->isc_capabilities &= ~IFCAP_VLAN_HWTAGGING;
950 			scctx->isc_capenable = scctx->isc_capabilities;
951 		}
952 		/* This is the first e1000 chip and it does not do offloads */
953 		if (hw->mac.type == e1000_82542) {
954 			scctx->isc_capabilities &= ~(IFCAP_HWCSUM | IFCAP_VLAN_HWCSUM |
955 			    IFCAP_HWCSUM_IPV6 | IFCAP_VLAN_HWTAGGING |
956 			    IFCAP_VLAN_HWFILTER | IFCAP_TSO | IFCAP_VLAN_HWTSO);
957 			scctx->isc_capenable = scctx->isc_capabilities;
958 		}
959 		/* These can't do TSO for various reasons */
960 		if (hw->mac.type < e1000_82544 || hw->mac.type == e1000_82547 ||
961 		    hw->mac.type == e1000_82547_rev_2) {
962 			scctx->isc_capabilities &= ~(IFCAP_TSO | IFCAP_VLAN_HWTSO);
963 			scctx->isc_capenable = scctx->isc_capabilities;
964 		}
965 		/* XXXKB: No IPv6 before this? */
966 		if (hw->mac.type < e1000_82545){
967 			scctx->isc_capabilities &= ~IFCAP_HWCSUM_IPV6;
968 			scctx->isc_capenable = scctx->isc_capabilities;
969 		}
970 		/* "PCI/PCI-X SDM 4.0" page 33 (b) - FDX requirement on these chips */
971 		if (hw->mac.type == e1000_82547 || hw->mac.type == e1000_82547_rev_2)
972 			scctx->isc_capenable &= ~(IFCAP_HWCSUM | IFCAP_VLAN_HWCSUM |
973 			    IFCAP_HWCSUM_IPV6);
974 
975 		/* INTx only */
976 		scctx->isc_msix_bar = 0;
977 	}
978 
979 	/* Setup PCI resources */
980 	if (em_allocate_pci_resources(ctx)) {
981 		device_printf(dev, "Allocation of PCI resources failed\n");
982 		error = ENXIO;
983 		goto err_pci;
984 	}
985 
986 	/*
987 	** For ICH8 and family we need to
988 	** map the flash memory, and this
989 	** must happen after the MAC is
990 	** identified
991 	*/
992 	if ((hw->mac.type == e1000_ich8lan) ||
993 	    (hw->mac.type == e1000_ich9lan) ||
994 	    (hw->mac.type == e1000_ich10lan) ||
995 	    (hw->mac.type == e1000_pchlan) ||
996 	    (hw->mac.type == e1000_pch2lan) ||
997 	    (hw->mac.type == e1000_pch_lpt)) {
998 		int rid = EM_BAR_TYPE_FLASH;
999 		sc->flash = bus_alloc_resource_any(dev,
1000 		    SYS_RES_MEMORY, &rid, RF_ACTIVE);
1001 		if (sc->flash == NULL) {
1002 			device_printf(dev, "Mapping of Flash failed\n");
1003 			error = ENXIO;
1004 			goto err_pci;
1005 		}
1006 		/* This is used in the shared code */
1007 		hw->flash_address = (u8 *)sc->flash;
1008 		sc->osdep.flash_bus_space_tag =
1009 		    rman_get_bustag(sc->flash);
1010 		sc->osdep.flash_bus_space_handle =
1011 		    rman_get_bushandle(sc->flash);
1012 	}
1013 	/*
1014 	** In the new SPT device flash is not  a
1015 	** separate BAR, rather it is also in BAR0,
1016 	** so use the same tag and an offset handle for the
1017 	** FLASH read/write macros in the shared code.
1018 	*/
1019 	else if (hw->mac.type >= e1000_pch_spt) {
1020 		sc->osdep.flash_bus_space_tag =
1021 		    sc->osdep.mem_bus_space_tag;
1022 		sc->osdep.flash_bus_space_handle =
1023 		    sc->osdep.mem_bus_space_handle
1024 		    + E1000_FLASH_BASE_ADDR;
1025 	}
1026 
1027 	/* Do Shared Code initialization */
1028 	error = e1000_setup_init_funcs(hw, true);
1029 	if (error) {
1030 		device_printf(dev, "Setup of Shared code failed, error %d\n",
1031 		    error);
1032 		error = ENXIO;
1033 		goto err_pci;
1034 	}
1035 
1036 	em_setup_msix(ctx);
1037 	e1000_get_bus_info(hw);
1038 
1039 	/* Set up some sysctls for the tunable interrupt delays */
1040 	em_add_int_delay_sysctl(sc, "rx_int_delay",
1041 	    "receive interrupt delay in usecs", &sc->rx_int_delay,
1042 	    E1000_REGISTER(hw, E1000_RDTR), em_rx_int_delay_dflt);
1043 	em_add_int_delay_sysctl(sc, "tx_int_delay",
1044 	    "transmit interrupt delay in usecs", &sc->tx_int_delay,
1045 	    E1000_REGISTER(hw, E1000_TIDV), em_tx_int_delay_dflt);
1046 	em_add_int_delay_sysctl(sc, "rx_abs_int_delay",
1047 	    "receive interrupt delay limit in usecs",
1048 	    &sc->rx_abs_int_delay,
1049 	    E1000_REGISTER(hw, E1000_RADV),
1050 	    em_rx_abs_int_delay_dflt);
1051 	em_add_int_delay_sysctl(sc, "tx_abs_int_delay",
1052 	    "transmit interrupt delay limit in usecs",
1053 	    &sc->tx_abs_int_delay,
1054 	    E1000_REGISTER(hw, E1000_TADV),
1055 	    em_tx_abs_int_delay_dflt);
1056 	em_add_int_delay_sysctl(sc, "itr",
1057 	    "interrupt delay limit in usecs/4",
1058 	    &sc->tx_itr,
1059 	    E1000_REGISTER(hw, E1000_ITR),
1060 	    DEFAULT_ITR);
1061 
1062 	hw->mac.autoneg = DO_AUTO_NEG;
1063 	hw->phy.autoneg_wait_to_complete = false;
1064 	hw->phy.autoneg_advertised = AUTONEG_ADV_DEFAULT;
1065 
1066 	if (hw->mac.type < em_mac_min) {
1067 		e1000_init_script_state_82541(hw, true);
1068 		e1000_set_tbi_compatibility_82543(hw, true);
1069 	}
1070 	/* Copper options */
1071 	if (hw->phy.media_type == e1000_media_type_copper) {
1072 		hw->phy.mdix = AUTO_ALL_MODES;
1073 		hw->phy.disable_polarity_correction = false;
1074 		hw->phy.ms_type = EM_MASTER_SLAVE;
1075 	}
1076 
1077 	/*
1078 	 * Set the frame limits assuming
1079 	 * standard ethernet sized frames.
1080 	 */
1081 	scctx->isc_max_frame_size = hw->mac.max_frame_size =
1082 	    ETHERMTU + ETHER_HDR_LEN + ETHERNET_FCS_SIZE;
1083 
1084 	/*
1085 	 * This controls when hardware reports transmit completion
1086 	 * status.
1087 	 */
1088 	hw->mac.report_tx_early = 1;
1089 
1090 	/* Allocate multicast array memory. */
1091 	sc->mta = malloc(sizeof(u8) * ETHER_ADDR_LEN *
1092 	    MAX_NUM_MULTICAST_ADDRESSES, M_DEVBUF, M_NOWAIT);
1093 	if (sc->mta == NULL) {
1094 		device_printf(dev, "Can not allocate multicast setup array\n");
1095 		error = ENOMEM;
1096 		goto err_late;
1097 	}
1098 
1099 	/* Clear the IFCAP_TSO auto mask */
1100 	sc->tso_automasked = 0;
1101 
1102 	/* Check SOL/IDER usage */
1103 	if (e1000_check_reset_block(hw))
1104 		device_printf(dev, "PHY reset is blocked"
1105 			      " due to SOL/IDER session.\n");
1106 
1107 	/* Sysctl for setting Energy Efficient Ethernet */
1108 	hw->dev_spec.ich8lan.eee_disable = eee_setting;
1109 	SYSCTL_ADD_PROC(ctx_list, child, OID_AUTO, "eee_control",
1110 	    CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_NEEDGIANT, sc, 0,
1111 	    em_sysctl_eee, "I", "Disable Energy Efficient Ethernet");
1112 
1113 	/*
1114 	** Start from a known state, this is
1115 	** important in reading the nvm and
1116 	** mac from that.
1117 	*/
1118 	e1000_reset_hw(hw);
1119 
1120 	/* Make sure we have a good EEPROM before we read from it */
1121 	if (e1000_validate_nvm_checksum(hw) < 0) {
1122 		/*
1123 		** Some PCI-E parts fail the first check due to
1124 		** the link being in sleep state, call it again,
1125 		** if it fails a second time its a real issue.
1126 		*/
1127 		if (e1000_validate_nvm_checksum(hw) < 0) {
1128 			device_printf(dev,
1129 			    "The EEPROM Checksum Is Not Valid\n");
1130 			error = EIO;
1131 			goto err_late;
1132 		}
1133 	}
1134 
1135 	/* Copy the permanent MAC address out of the EEPROM */
1136 	if (e1000_read_mac_addr(hw) < 0) {
1137 		device_printf(dev, "EEPROM read error while reading MAC"
1138 			      " address\n");
1139 		error = EIO;
1140 		goto err_late;
1141 	}
1142 
1143 	if (!em_is_valid_ether_addr(hw->mac.addr)) {
1144 		if (sc->vf_ifp) {
1145 			ether_gen_addr(iflib_get_ifp(ctx),
1146 			    (struct ether_addr *)hw->mac.addr);
1147 		} else {
1148 			device_printf(dev, "Invalid MAC address\n");
1149 			error = EIO;
1150 			goto err_late;
1151 		}
1152 	}
1153 
1154 	/* Save the EEPROM/NVM versions, must be done under IFLIB_CTX_LOCK */
1155 	em_fw_version_locked(ctx);
1156 
1157 	em_print_fw_version(sc);
1158 
1159 	/*
1160 	 * Get Wake-on-Lan and Management info for later use
1161 	 */
1162 	em_get_wakeup(ctx);
1163 
1164 	/* Enable only WOL MAGIC by default */
1165 	scctx->isc_capenable &= ~IFCAP_WOL;
1166 	if (sc->wol != 0)
1167 		scctx->isc_capenable |= IFCAP_WOL_MAGIC;
1168 
1169 	iflib_set_mac(ctx, hw->mac.addr);
1170 
1171 	return (0);
1172 
1173 err_late:
1174 	em_release_hw_control(sc);
1175 err_pci:
1176 	em_free_pci_resources(ctx);
1177 	free(sc->mta, M_DEVBUF);
1178 
1179 	return (error);
1180 }
1181 
1182 static int
1183 em_if_attach_post(if_ctx_t ctx)
1184 {
1185 	struct e1000_softc *sc = iflib_get_softc(ctx);
1186 	struct e1000_hw *hw = &sc->hw;
1187 	int error = 0;
1188 
1189 	/* Setup OS specific network interface */
1190 	error = em_setup_interface(ctx);
1191 	if (error != 0) {
1192 		device_printf(sc->dev, "Interface setup failed: %d\n", error);
1193 		goto err_late;
1194 	}
1195 
1196 	em_reset(ctx);
1197 
1198 	/* Initialize statistics */
1199 	em_update_stats_counters(sc);
1200 	hw->mac.get_link_status = 1;
1201 	em_if_update_admin_status(ctx);
1202 	em_add_hw_stats(sc);
1203 
1204 	/* Non-AMT based hardware can now take control from firmware */
1205 	if (sc->has_manage && !sc->has_amt)
1206 		em_get_hw_control(sc);
1207 
1208 	INIT_DEBUGOUT("em_if_attach_post: end");
1209 
1210 	return (0);
1211 
1212 err_late:
1213 	/* upon attach_post() error, iflib calls _if_detach() to free resources. */
1214 	return (error);
1215 }
1216 
1217 /*********************************************************************
1218  *  Device removal routine
1219  *
1220  *  The detach entry point is called when the driver is being removed.
1221  *  This routine stops the adapter and deallocates all the resources
1222  *  that were allocated for driver operation.
1223  *
1224  *  return 0 on success, positive on failure
1225  *********************************************************************/
1226 static int
1227 em_if_detach(if_ctx_t ctx)
1228 {
1229 	struct e1000_softc	*sc = iflib_get_softc(ctx);
1230 
1231 	INIT_DEBUGOUT("em_if_detach: begin");
1232 
1233 	e1000_phy_hw_reset(&sc->hw);
1234 
1235 	em_release_manageability(sc);
1236 	em_release_hw_control(sc);
1237 	em_free_pci_resources(ctx);
1238 	free(sc->mta, M_DEVBUF);
1239 	sc->mta = NULL;
1240 
1241 	return (0);
1242 }
1243 
1244 /*********************************************************************
1245  *
1246  *  Shutdown entry point
1247  *
1248  **********************************************************************/
1249 
1250 static int
1251 em_if_shutdown(if_ctx_t ctx)
1252 {
1253 	return em_if_suspend(ctx);
1254 }
1255 
1256 /*
1257  * Suspend/resume device methods.
1258  */
1259 static int
1260 em_if_suspend(if_ctx_t ctx)
1261 {
1262 	struct e1000_softc *sc = iflib_get_softc(ctx);
1263 
1264 	em_release_manageability(sc);
1265 	em_release_hw_control(sc);
1266 	em_enable_wakeup(ctx);
1267 	return (0);
1268 }
1269 
1270 static int
1271 em_if_resume(if_ctx_t ctx)
1272 {
1273 	struct e1000_softc *sc = iflib_get_softc(ctx);
1274 
1275 	if (sc->hw.mac.type == e1000_pch2lan)
1276 		e1000_resume_workarounds_pchlan(&sc->hw);
1277 	em_if_init(ctx);
1278 	em_init_manageability(sc);
1279 
1280 	return(0);
1281 }
1282 
1283 static int
1284 em_if_mtu_set(if_ctx_t ctx, uint32_t mtu)
1285 {
1286 	int max_frame_size;
1287 	struct e1000_softc *sc = iflib_get_softc(ctx);
1288 	if_softc_ctx_t scctx = iflib_get_softc_ctx(ctx);
1289 
1290 	IOCTL_DEBUGOUT("ioctl rcv'd: SIOCSIFMTU (Set Interface MTU)");
1291 
1292 	switch (sc->hw.mac.type) {
1293 	case e1000_82571:
1294 	case e1000_82572:
1295 	case e1000_ich9lan:
1296 	case e1000_ich10lan:
1297 	case e1000_pch2lan:
1298 	case e1000_pch_lpt:
1299 	case e1000_pch_spt:
1300 	case e1000_pch_cnp:
1301 	case e1000_pch_tgp:
1302 	case e1000_pch_adp:
1303 	case e1000_pch_mtp:
1304 	case e1000_pch_ptp:
1305 	case e1000_82574:
1306 	case e1000_82583:
1307 	case e1000_80003es2lan:
1308 		/* 9K Jumbo Frame size */
1309 		max_frame_size = 9234;
1310 		break;
1311 	case e1000_pchlan:
1312 		max_frame_size = 4096;
1313 		break;
1314 	case e1000_82542:
1315 	case e1000_ich8lan:
1316 		/* Adapters that do not support jumbo frames */
1317 		max_frame_size = ETHER_MAX_LEN;
1318 		break;
1319 	default:
1320 		if (sc->hw.mac.type >= igb_mac_min)
1321 			max_frame_size = 9234;
1322 		else /* lem */
1323 			max_frame_size = MAX_JUMBO_FRAME_SIZE;
1324 	}
1325 	if (mtu > max_frame_size - ETHER_HDR_LEN - ETHER_CRC_LEN) {
1326 		return (EINVAL);
1327 	}
1328 
1329 	scctx->isc_max_frame_size = sc->hw.mac.max_frame_size =
1330 	    mtu + ETHER_HDR_LEN + ETHER_CRC_LEN;
1331 	return (0);
1332 }
1333 
1334 /*********************************************************************
1335  *  Init entry point
1336  *
1337  *  This routine is used in two ways. It is used by the stack as
1338  *  init entry point in network interface structure. It is also used
1339  *  by the driver as a hw/sw initialization routine to get to a
1340  *  consistent state.
1341  *
1342  **********************************************************************/
1343 static void
1344 em_if_init(if_ctx_t ctx)
1345 {
1346 	struct e1000_softc *sc = iflib_get_softc(ctx);
1347 	if_softc_ctx_t scctx = sc->shared;
1348 	if_t ifp = iflib_get_ifp(ctx);
1349 	struct em_tx_queue *tx_que;
1350 	int i;
1351 
1352 	INIT_DEBUGOUT("em_if_init: begin");
1353 
1354 	/* Get the latest mac address, User can use a LAA */
1355 	bcopy(if_getlladdr(ifp), sc->hw.mac.addr,
1356 	    ETHER_ADDR_LEN);
1357 
1358 	/* Put the address into the Receive Address Array */
1359 	e1000_rar_set(&sc->hw, sc->hw.mac.addr, 0);
1360 
1361 	/*
1362 	 * With the 82571 adapter, RAR[0] may be overwritten
1363 	 * when the other port is reset, we make a duplicate
1364 	 * in RAR[14] for that eventuality, this assures
1365 	 * the interface continues to function.
1366 	 */
1367 	if (sc->hw.mac.type == e1000_82571) {
1368 		e1000_set_laa_state_82571(&sc->hw, true);
1369 		e1000_rar_set(&sc->hw, sc->hw.mac.addr,
1370 		    E1000_RAR_ENTRIES - 1);
1371 	}
1372 
1373 	/* Initialize the hardware */
1374 	em_reset(ctx);
1375 	em_if_update_admin_status(ctx);
1376 
1377 	for (i = 0, tx_que = sc->tx_queues; i < sc->tx_num_queues; i++, tx_que++) {
1378 		struct tx_ring *txr = &tx_que->txr;
1379 
1380 		txr->tx_rs_cidx = txr->tx_rs_pidx;
1381 
1382 		/* Initialize the last processed descriptor to be the end of
1383 		 * the ring, rather than the start, so that we avoid an
1384 		 * off-by-one error when calculating how many descriptors are
1385 		 * done in the credits_update function.
1386 		 */
1387 		txr->tx_cidx_processed = scctx->isc_ntxd[0] - 1;
1388 	}
1389 
1390 	/* Setup VLAN support, basic and offload if available */
1391 	E1000_WRITE_REG(&sc->hw, E1000_VET, ETHERTYPE_VLAN);
1392 
1393 	/* Clear bad data from Rx FIFOs */
1394 	if (sc->hw.mac.type >= igb_mac_min)
1395 		e1000_rx_fifo_flush_base(&sc->hw);
1396 
1397 	/* Configure for OS presence */
1398 	em_init_manageability(sc);
1399 
1400 	/* Prepare transmit descriptors and buffers */
1401 	em_initialize_transmit_unit(ctx);
1402 
1403 	/* Setup Multicast table */
1404 	em_if_multi_set(ctx);
1405 
1406 	sc->rx_mbuf_sz = iflib_get_rx_mbuf_sz(ctx);
1407 	em_initialize_receive_unit(ctx);
1408 
1409 	/* Set up VLAN support and filter */
1410 	em_setup_vlan_hw_support(ctx);
1411 
1412 	/* Don't lose promiscuous settings */
1413 	em_if_set_promisc(ctx, if_getflags(ifp));
1414 	e1000_clear_hw_cntrs_base_generic(&sc->hw);
1415 
1416 	/* MSI-X configuration for 82574 */
1417 	if (sc->hw.mac.type == e1000_82574) {
1418 		int tmp = E1000_READ_REG(&sc->hw, E1000_CTRL_EXT);
1419 
1420 		tmp |= E1000_CTRL_EXT_PBA_CLR;
1421 		E1000_WRITE_REG(&sc->hw, E1000_CTRL_EXT, tmp);
1422 		/* Set the IVAR - interrupt vector routing. */
1423 		E1000_WRITE_REG(&sc->hw, E1000_IVAR, sc->ivars);
1424 	} else if (sc->intr_type == IFLIB_INTR_MSIX) /* Set up queue routing */
1425 		igb_configure_queues(sc);
1426 
1427 	/* this clears any pending interrupts */
1428 	E1000_READ_REG(&sc->hw, E1000_ICR);
1429 	E1000_WRITE_REG(&sc->hw, E1000_ICS, E1000_ICS_LSC);
1430 
1431 	/* AMT based hardware can now take control from firmware */
1432 	if (sc->has_manage && sc->has_amt)
1433 		em_get_hw_control(sc);
1434 
1435 	/* Set Energy Efficient Ethernet */
1436 	if (sc->hw.mac.type >= igb_mac_min &&
1437 	    sc->hw.phy.media_type == e1000_media_type_copper) {
1438 		if (sc->hw.mac.type == e1000_i354)
1439 			e1000_set_eee_i354(&sc->hw, true, true);
1440 		else
1441 			e1000_set_eee_i350(&sc->hw, true, true);
1442 	}
1443 }
1444 
1445 /*********************************************************************
1446  *
1447  *  Fast Legacy/MSI Combined Interrupt Service routine
1448  *
1449  *********************************************************************/
1450 int
1451 em_intr(void *arg)
1452 {
1453 	struct e1000_softc *sc = arg;
1454 	if_ctx_t ctx = sc->ctx;
1455 	u32 reg_icr;
1456 
1457 	reg_icr = E1000_READ_REG(&sc->hw, E1000_ICR);
1458 
1459 	/* Hot eject? */
1460 	if (reg_icr == 0xffffffff)
1461 		return FILTER_STRAY;
1462 
1463 	/* Definitely not our interrupt. */
1464 	if (reg_icr == 0x0)
1465 		return FILTER_STRAY;
1466 
1467 	/*
1468 	 * Starting with the 82571 chip, bit 31 should be used to
1469 	 * determine whether the interrupt belongs to us.
1470 	 */
1471 	if (sc->hw.mac.type >= e1000_82571 &&
1472 	    (reg_icr & E1000_ICR_INT_ASSERTED) == 0)
1473 		return FILTER_STRAY;
1474 
1475 	/*
1476 	 * Only MSI-X interrupts have one-shot behavior by taking advantage
1477 	 * of the EIAC register.  Thus, explicitly disable interrupts.  This
1478 	 * also works around the MSI message reordering errata on certain
1479 	 * systems.
1480 	 */
1481 	IFDI_INTR_DISABLE(ctx);
1482 
1483 	/* Link status change */
1484 	if (reg_icr & (E1000_ICR_RXSEQ | E1000_ICR_LSC))
1485 		em_handle_link(ctx);
1486 
1487 	if (reg_icr & E1000_ICR_RXO)
1488 		sc->rx_overruns++;
1489 
1490 	return (FILTER_SCHEDULE_THREAD);
1491 }
1492 
1493 static int
1494 em_if_rx_queue_intr_enable(if_ctx_t ctx, uint16_t rxqid)
1495 {
1496 	struct e1000_softc *sc = iflib_get_softc(ctx);
1497 	struct em_rx_queue *rxq = &sc->rx_queues[rxqid];
1498 
1499 	E1000_WRITE_REG(&sc->hw, E1000_IMS, rxq->eims);
1500 	return (0);
1501 }
1502 
1503 static int
1504 em_if_tx_queue_intr_enable(if_ctx_t ctx, uint16_t txqid)
1505 {
1506 	struct e1000_softc *sc = iflib_get_softc(ctx);
1507 	struct em_tx_queue *txq = &sc->tx_queues[txqid];
1508 
1509 	E1000_WRITE_REG(&sc->hw, E1000_IMS, txq->eims);
1510 	return (0);
1511 }
1512 
1513 static int
1514 igb_if_rx_queue_intr_enable(if_ctx_t ctx, uint16_t rxqid)
1515 {
1516 	struct e1000_softc *sc = iflib_get_softc(ctx);
1517 	struct em_rx_queue *rxq = &sc->rx_queues[rxqid];
1518 
1519 	E1000_WRITE_REG(&sc->hw, E1000_EIMS, rxq->eims);
1520 	return (0);
1521 }
1522 
1523 static int
1524 igb_if_tx_queue_intr_enable(if_ctx_t ctx, uint16_t txqid)
1525 {
1526 	struct e1000_softc *sc = iflib_get_softc(ctx);
1527 	struct em_tx_queue *txq = &sc->tx_queues[txqid];
1528 
1529 	E1000_WRITE_REG(&sc->hw, E1000_EIMS, txq->eims);
1530 	return (0);
1531 }
1532 
1533 /*********************************************************************
1534  *
1535  *  MSI-X RX Interrupt Service routine
1536  *
1537  **********************************************************************/
1538 static int
1539 em_msix_que(void *arg)
1540 {
1541 	struct em_rx_queue *que = arg;
1542 
1543 	++que->irqs;
1544 
1545 	return (FILTER_SCHEDULE_THREAD);
1546 }
1547 
1548 /*********************************************************************
1549  *
1550  *  MSI-X Link Fast Interrupt Service routine
1551  *
1552  **********************************************************************/
1553 static int
1554 em_msix_link(void *arg)
1555 {
1556 	struct e1000_softc *sc = arg;
1557 	u32 reg_icr;
1558 
1559 	++sc->link_irq;
1560 	MPASS(sc->hw.back != NULL);
1561 	reg_icr = E1000_READ_REG(&sc->hw, E1000_ICR);
1562 
1563 	if (reg_icr & E1000_ICR_RXO)
1564 		sc->rx_overruns++;
1565 
1566 	if (reg_icr & (E1000_ICR_RXSEQ | E1000_ICR_LSC))
1567 		em_handle_link(sc->ctx);
1568 
1569 	/* Re-arm unconditionally */
1570 	if (sc->hw.mac.type >= igb_mac_min) {
1571 		E1000_WRITE_REG(&sc->hw, E1000_IMS, E1000_IMS_LSC);
1572 		E1000_WRITE_REG(&sc->hw, E1000_EIMS, sc->link_mask);
1573 	} else if (sc->hw.mac.type == e1000_82574) {
1574 		E1000_WRITE_REG(&sc->hw, E1000_IMS, E1000_IMS_LSC |
1575 		    E1000_IMS_OTHER);
1576 		/*
1577 		 * Because we must read the ICR for this interrupt it may
1578 		 * clear other causes using autoclear, for this reason we
1579 		 * simply create a soft interrupt for all these vectors.
1580 		 */
1581 		if (reg_icr)
1582 			E1000_WRITE_REG(&sc->hw, E1000_ICS, sc->ims);
1583 	} else
1584 		E1000_WRITE_REG(&sc->hw, E1000_IMS, E1000_IMS_LSC);
1585 
1586 	return (FILTER_HANDLED);
1587 }
1588 
1589 static void
1590 em_handle_link(void *context)
1591 {
1592 	if_ctx_t ctx = context;
1593 	struct e1000_softc *sc = iflib_get_softc(ctx);
1594 
1595 	sc->hw.mac.get_link_status = 1;
1596 	iflib_admin_intr_deferred(ctx);
1597 }
1598 
1599 /*********************************************************************
1600  *
1601  *  Media Ioctl callback
1602  *
1603  *  This routine is called whenever the user queries the status of
1604  *  the interface using ifconfig.
1605  *
1606  **********************************************************************/
1607 static void
1608 em_if_media_status(if_ctx_t ctx, struct ifmediareq *ifmr)
1609 {
1610 	struct e1000_softc *sc = iflib_get_softc(ctx);
1611 	u_char fiber_type = IFM_1000_SX;
1612 
1613 	INIT_DEBUGOUT("em_if_media_status: begin");
1614 
1615 	iflib_admin_intr_deferred(ctx);
1616 
1617 	ifmr->ifm_status = IFM_AVALID;
1618 	ifmr->ifm_active = IFM_ETHER;
1619 
1620 	if (!sc->link_active) {
1621 		return;
1622 	}
1623 
1624 	ifmr->ifm_status |= IFM_ACTIVE;
1625 
1626 	if ((sc->hw.phy.media_type == e1000_media_type_fiber) ||
1627 	    (sc->hw.phy.media_type == e1000_media_type_internal_serdes)) {
1628 		if (sc->hw.mac.type == e1000_82545)
1629 			fiber_type = IFM_1000_LX;
1630 		ifmr->ifm_active |= fiber_type | IFM_FDX;
1631 	} else {
1632 		switch (sc->link_speed) {
1633 		case 10:
1634 			ifmr->ifm_active |= IFM_10_T;
1635 			break;
1636 		case 100:
1637 			ifmr->ifm_active |= IFM_100_TX;
1638 			break;
1639 		case 1000:
1640 			ifmr->ifm_active |= IFM_1000_T;
1641 			break;
1642 		}
1643 		if (sc->link_duplex == FULL_DUPLEX)
1644 			ifmr->ifm_active |= IFM_FDX;
1645 		else
1646 			ifmr->ifm_active |= IFM_HDX;
1647 	}
1648 }
1649 
1650 /*********************************************************************
1651  *
1652  *  Media Ioctl callback
1653  *
1654  *  This routine is called when the user changes speed/duplex using
1655  *  media/mediopt option with ifconfig.
1656  *
1657  **********************************************************************/
1658 static int
1659 em_if_media_change(if_ctx_t ctx)
1660 {
1661 	struct e1000_softc *sc = iflib_get_softc(ctx);
1662 	struct ifmedia *ifm = iflib_get_media(ctx);
1663 
1664 	INIT_DEBUGOUT("em_if_media_change: begin");
1665 
1666 	if (IFM_TYPE(ifm->ifm_media) != IFM_ETHER)
1667 		return (EINVAL);
1668 
1669 	switch (IFM_SUBTYPE(ifm->ifm_media)) {
1670 	case IFM_AUTO:
1671 		sc->hw.mac.autoneg = DO_AUTO_NEG;
1672 		sc->hw.phy.autoneg_advertised = AUTONEG_ADV_DEFAULT;
1673 		break;
1674 	case IFM_1000_LX:
1675 	case IFM_1000_SX:
1676 	case IFM_1000_T:
1677 		sc->hw.mac.autoneg = DO_AUTO_NEG;
1678 		sc->hw.phy.autoneg_advertised = ADVERTISE_1000_FULL;
1679 		break;
1680 	case IFM_100_TX:
1681 		sc->hw.mac.autoneg = false;
1682 		sc->hw.phy.autoneg_advertised = 0;
1683 		if ((ifm->ifm_media & IFM_GMASK) == IFM_FDX)
1684 			sc->hw.mac.forced_speed_duplex = ADVERTISE_100_FULL;
1685 		else
1686 			sc->hw.mac.forced_speed_duplex = ADVERTISE_100_HALF;
1687 		break;
1688 	case IFM_10_T:
1689 		sc->hw.mac.autoneg = false;
1690 		sc->hw.phy.autoneg_advertised = 0;
1691 		if ((ifm->ifm_media & IFM_GMASK) == IFM_FDX)
1692 			sc->hw.mac.forced_speed_duplex = ADVERTISE_10_FULL;
1693 		else
1694 			sc->hw.mac.forced_speed_duplex = ADVERTISE_10_HALF;
1695 		break;
1696 	default:
1697 		device_printf(sc->dev, "Unsupported media type\n");
1698 	}
1699 
1700 	em_if_init(ctx);
1701 
1702 	return (0);
1703 }
1704 
1705 static int
1706 em_if_set_promisc(if_ctx_t ctx, int flags)
1707 {
1708 	struct e1000_softc *sc = iflib_get_softc(ctx);
1709 	if_t ifp = iflib_get_ifp(ctx);
1710 	u32 reg_rctl;
1711 	int mcnt = 0;
1712 
1713 	reg_rctl = E1000_READ_REG(&sc->hw, E1000_RCTL);
1714 	reg_rctl &= ~(E1000_RCTL_SBP | E1000_RCTL_UPE);
1715 	if (flags & IFF_ALLMULTI)
1716 		mcnt = MAX_NUM_MULTICAST_ADDRESSES;
1717 	else
1718 		mcnt = min(if_llmaddr_count(ifp), MAX_NUM_MULTICAST_ADDRESSES);
1719 
1720 	if (mcnt < MAX_NUM_MULTICAST_ADDRESSES)
1721 		reg_rctl &= (~E1000_RCTL_MPE);
1722 	E1000_WRITE_REG(&sc->hw, E1000_RCTL, reg_rctl);
1723 
1724 	if (flags & IFF_PROMISC) {
1725 		reg_rctl |= (E1000_RCTL_UPE | E1000_RCTL_MPE);
1726 		em_if_vlan_filter_disable(sc);
1727 		/* Turn this on if you want to see bad packets */
1728 		if (em_debug_sbp)
1729 			reg_rctl |= E1000_RCTL_SBP;
1730 		E1000_WRITE_REG(&sc->hw, E1000_RCTL, reg_rctl);
1731 	} else {
1732 		if (flags & IFF_ALLMULTI) {
1733 			reg_rctl |= E1000_RCTL_MPE;
1734 			reg_rctl &= ~E1000_RCTL_UPE;
1735 			E1000_WRITE_REG(&sc->hw, E1000_RCTL, reg_rctl);
1736 		}
1737 		if (em_if_vlan_filter_used(ctx))
1738 			em_if_vlan_filter_enable(sc);
1739 	}
1740 	return (0);
1741 }
1742 
1743 static u_int
1744 em_copy_maddr(void *arg, struct sockaddr_dl *sdl, u_int idx)
1745 {
1746 	u8 *mta = arg;
1747 
1748 	if (idx == MAX_NUM_MULTICAST_ADDRESSES)
1749 		return (0);
1750 
1751 	bcopy(LLADDR(sdl), &mta[idx * ETHER_ADDR_LEN], ETHER_ADDR_LEN);
1752 
1753 	return (1);
1754 }
1755 
1756 /*********************************************************************
1757  *  Multicast Update
1758  *
1759  *  This routine is called whenever multicast address list is updated.
1760  *
1761  **********************************************************************/
1762 static void
1763 em_if_multi_set(if_ctx_t ctx)
1764 {
1765 	struct e1000_softc *sc = iflib_get_softc(ctx);
1766 	if_t ifp = iflib_get_ifp(ctx);
1767 	u8  *mta; /* Multicast array memory */
1768 	u32 reg_rctl = 0;
1769 	int mcnt = 0;
1770 
1771 	IOCTL_DEBUGOUT("em_set_multi: begin");
1772 
1773 	mta = sc->mta;
1774 	bzero(mta, sizeof(u8) * ETHER_ADDR_LEN * MAX_NUM_MULTICAST_ADDRESSES);
1775 
1776 	if (sc->hw.mac.type == e1000_82542 &&
1777 	    sc->hw.revision_id == E1000_REVISION_2) {
1778 		reg_rctl = E1000_READ_REG(&sc->hw, E1000_RCTL);
1779 		if (sc->hw.bus.pci_cmd_word & CMD_MEM_WRT_INVALIDATE)
1780 			e1000_pci_clear_mwi(&sc->hw);
1781 		reg_rctl |= E1000_RCTL_RST;
1782 		E1000_WRITE_REG(&sc->hw, E1000_RCTL, reg_rctl);
1783 		msec_delay(5);
1784 	}
1785 
1786 	mcnt = if_foreach_llmaddr(ifp, em_copy_maddr, mta);
1787 
1788 	if (mcnt < MAX_NUM_MULTICAST_ADDRESSES)
1789 		e1000_update_mc_addr_list(&sc->hw, mta, mcnt);
1790 
1791 	reg_rctl = E1000_READ_REG(&sc->hw, E1000_RCTL);
1792 
1793 	if (if_getflags(ifp) & IFF_PROMISC)
1794 		reg_rctl |= (E1000_RCTL_UPE | E1000_RCTL_MPE);
1795 	else if (mcnt >= MAX_NUM_MULTICAST_ADDRESSES ||
1796 	    if_getflags(ifp) & IFF_ALLMULTI) {
1797 		reg_rctl |= E1000_RCTL_MPE;
1798 		reg_rctl &= ~E1000_RCTL_UPE;
1799 	} else
1800 		reg_rctl &= ~(E1000_RCTL_UPE | E1000_RCTL_MPE);
1801 
1802 	E1000_WRITE_REG(&sc->hw, E1000_RCTL, reg_rctl);
1803 
1804 	if (sc->hw.mac.type == e1000_82542 &&
1805 	    sc->hw.revision_id == E1000_REVISION_2) {
1806 		reg_rctl = E1000_READ_REG(&sc->hw, E1000_RCTL);
1807 		reg_rctl &= ~E1000_RCTL_RST;
1808 		E1000_WRITE_REG(&sc->hw, E1000_RCTL, reg_rctl);
1809 		msec_delay(5);
1810 		if (sc->hw.bus.pci_cmd_word & CMD_MEM_WRT_INVALIDATE)
1811 			e1000_pci_set_mwi(&sc->hw);
1812 	}
1813 }
1814 
1815 /*********************************************************************
1816  *  Timer routine
1817  *
1818  *  This routine schedules em_if_update_admin_status() to check for
1819  *  link status and to gather statistics as well as to perform some
1820  *  controller-specific hardware patting.
1821  *
1822  **********************************************************************/
1823 static void
1824 em_if_timer(if_ctx_t ctx, uint16_t qid)
1825 {
1826 
1827 	if (qid != 0)
1828 		return;
1829 
1830 	iflib_admin_intr_deferred(ctx);
1831 }
1832 
1833 static void
1834 em_if_update_admin_status(if_ctx_t ctx)
1835 {
1836 	struct e1000_softc *sc = iflib_get_softc(ctx);
1837 	struct e1000_hw *hw = &sc->hw;
1838 	device_t dev = iflib_get_dev(ctx);
1839 	u32 link_check, thstat, ctrl;
1840 	bool automasked = false;
1841 
1842 	link_check = thstat = ctrl = 0;
1843 	/* Get the cached link value or read phy for real */
1844 	switch (hw->phy.media_type) {
1845 	case e1000_media_type_copper:
1846 		if (hw->mac.get_link_status) {
1847 			if (hw->mac.type == e1000_pch_spt)
1848 				msec_delay(50);
1849 			/* Do the work to read phy */
1850 			e1000_check_for_link(hw);
1851 			link_check = !hw->mac.get_link_status;
1852 			if (link_check) /* ESB2 fix */
1853 				e1000_cfg_on_link_up(hw);
1854 		} else {
1855 			link_check = true;
1856 		}
1857 		break;
1858 	case e1000_media_type_fiber:
1859 		e1000_check_for_link(hw);
1860 		link_check = (E1000_READ_REG(hw, E1000_STATUS) &
1861 			    E1000_STATUS_LU);
1862 		break;
1863 	case e1000_media_type_internal_serdes:
1864 		e1000_check_for_link(hw);
1865 		link_check = hw->mac.serdes_has_link;
1866 		break;
1867 	/* VF device is type_unknown */
1868 	case e1000_media_type_unknown:
1869 		e1000_check_for_link(hw);
1870 		link_check = !hw->mac.get_link_status;
1871 		/* FALLTHROUGH */
1872 	default:
1873 		break;
1874 	}
1875 
1876 	/* Check for thermal downshift or shutdown */
1877 	if (hw->mac.type == e1000_i350) {
1878 		thstat = E1000_READ_REG(hw, E1000_THSTAT);
1879 		ctrl = E1000_READ_REG(hw, E1000_CTRL_EXT);
1880 	}
1881 
1882 	/* Now check for a transition */
1883 	if (link_check && (sc->link_active == 0)) {
1884 		e1000_get_speed_and_duplex(hw, &sc->link_speed,
1885 		    &sc->link_duplex);
1886 		/* Check if we must disable SPEED_MODE bit on PCI-E */
1887 		if ((sc->link_speed != SPEED_1000) &&
1888 		    ((hw->mac.type == e1000_82571) ||
1889 		    (hw->mac.type == e1000_82572))) {
1890 			int tarc0;
1891 			tarc0 = E1000_READ_REG(hw, E1000_TARC(0));
1892 			tarc0 &= ~TARC_SPEED_MODE_BIT;
1893 			E1000_WRITE_REG(hw, E1000_TARC(0), tarc0);
1894 		}
1895 		if (bootverbose)
1896 			device_printf(dev, "Link is up %d Mbps %s\n",
1897 			    sc->link_speed,
1898 			    ((sc->link_duplex == FULL_DUPLEX) ?
1899 			    "Full Duplex" : "Half Duplex"));
1900 		sc->link_active = 1;
1901 		sc->smartspeed = 0;
1902 		if ((ctrl & E1000_CTRL_EXT_LINK_MODE_MASK) ==
1903 		    E1000_CTRL_EXT_LINK_MODE_GMII &&
1904 		    (thstat & E1000_THSTAT_LINK_THROTTLE))
1905 			device_printf(dev, "Link: thermal downshift\n");
1906 		/* Delay Link Up for Phy update */
1907 		if (((hw->mac.type == e1000_i210) ||
1908 		    (hw->mac.type == e1000_i211)) &&
1909 		    (hw->phy.id == I210_I_PHY_ID))
1910 			msec_delay(I210_LINK_DELAY);
1911 		/* Reset if the media type changed. */
1912 		if (hw->dev_spec._82575.media_changed &&
1913 		    hw->mac.type >= igb_mac_min) {
1914 			hw->dev_spec._82575.media_changed = false;
1915 			sc->flags |= IGB_MEDIA_RESET;
1916 			em_reset(ctx);
1917 		}
1918 		/* Only do TSO on gigabit Ethernet for older chips due to errata */
1919 		if (hw->mac.type < igb_mac_min)
1920 			automasked = em_automask_tso(ctx);
1921 
1922 		/* Automasking resets the interface, so don't mark it up yet */
1923 		if (!automasked)
1924 			iflib_link_state_change(ctx, LINK_STATE_UP,
1925 			    IF_Mbps(sc->link_speed));
1926 	} else if (!link_check && (sc->link_active == 1)) {
1927 		sc->link_speed = 0;
1928 		sc->link_duplex = 0;
1929 		sc->link_active = 0;
1930 		iflib_link_state_change(ctx, LINK_STATE_DOWN, 0);
1931 	}
1932 	em_update_stats_counters(sc);
1933 
1934 	/* Reset LAA into RAR[0] on 82571 */
1935 	if (hw->mac.type == e1000_82571 && e1000_get_laa_state_82571(hw))
1936 		e1000_rar_set(hw, hw->mac.addr, 0);
1937 
1938 	if (hw->mac.type < em_mac_min)
1939 		lem_smartspeed(sc);
1940 }
1941 
1942 static void
1943 em_if_watchdog_reset(if_ctx_t ctx)
1944 {
1945 	struct e1000_softc *sc = iflib_get_softc(ctx);
1946 
1947 	/*
1948 	 * Just count the event; iflib(4) will already trigger a
1949 	 * sufficient reset of the controller.
1950 	 */
1951 	sc->watchdog_events++;
1952 }
1953 
1954 /*********************************************************************
1955  *
1956  *  This routine disables all traffic on the adapter by issuing a
1957  *  global reset on the MAC.
1958  *
1959  **********************************************************************/
1960 static void
1961 em_if_stop(if_ctx_t ctx)
1962 {
1963 	struct e1000_softc *sc = iflib_get_softc(ctx);
1964 
1965 	INIT_DEBUGOUT("em_if_stop: begin");
1966 
1967 	/* I219 needs special flushing to avoid hangs */
1968 	if (sc->hw.mac.type >= e1000_pch_spt && sc->hw.mac.type < igb_mac_min)
1969 		em_flush_desc_rings(sc);
1970 
1971 	e1000_reset_hw(&sc->hw);
1972 	if (sc->hw.mac.type >= e1000_82544)
1973 		E1000_WRITE_REG(&sc->hw, E1000_WUFC, 0);
1974 
1975 	e1000_led_off(&sc->hw);
1976 	e1000_cleanup_led(&sc->hw);
1977 }
1978 
1979 /*********************************************************************
1980  *
1981  *  Determine hardware revision.
1982  *
1983  **********************************************************************/
1984 static void
1985 em_identify_hardware(if_ctx_t ctx)
1986 {
1987 	device_t dev = iflib_get_dev(ctx);
1988 	struct e1000_softc *sc = iflib_get_softc(ctx);
1989 
1990 	/* Make sure our PCI config space has the necessary stuff set */
1991 	sc->hw.bus.pci_cmd_word = pci_read_config(dev, PCIR_COMMAND, 2);
1992 
1993 	/* Save off the information about this board */
1994 	sc->hw.vendor_id = pci_get_vendor(dev);
1995 	sc->hw.device_id = pci_get_device(dev);
1996 	sc->hw.revision_id = pci_read_config(dev, PCIR_REVID, 1);
1997 	sc->hw.subsystem_vendor_id =
1998 	    pci_read_config(dev, PCIR_SUBVEND_0, 2);
1999 	sc->hw.subsystem_device_id =
2000 	    pci_read_config(dev, PCIR_SUBDEV_0, 2);
2001 
2002 	/* Do Shared Code Init and Setup */
2003 	if (e1000_set_mac_type(&sc->hw)) {
2004 		device_printf(dev, "Setup init failure\n");
2005 		return;
2006 	}
2007 
2008 	/* Are we a VF device? */
2009 	if ((sc->hw.mac.type == e1000_vfadapt) ||
2010 	    (sc->hw.mac.type == e1000_vfadapt_i350))
2011 		sc->vf_ifp = 1;
2012 	else
2013 		sc->vf_ifp = 0;
2014 }
2015 
2016 static int
2017 em_allocate_pci_resources(if_ctx_t ctx)
2018 {
2019 	struct e1000_softc *sc = iflib_get_softc(ctx);
2020 	device_t dev = iflib_get_dev(ctx);
2021 	int rid, val;
2022 
2023 	rid = PCIR_BAR(0);
2024 	sc->memory = bus_alloc_resource_any(dev, SYS_RES_MEMORY,
2025 	    &rid, RF_ACTIVE);
2026 	if (sc->memory == NULL) {
2027 		device_printf(dev, "Unable to allocate bus resource: memory\n");
2028 		return (ENXIO);
2029 	}
2030 	sc->osdep.mem_bus_space_tag = rman_get_bustag(sc->memory);
2031 	sc->osdep.mem_bus_space_handle =
2032 	    rman_get_bushandle(sc->memory);
2033 	sc->hw.hw_addr = (u8 *)&sc->osdep.mem_bus_space_handle;
2034 
2035 	/* Only older adapters use IO mapping */
2036 	if (sc->hw.mac.type < em_mac_min && sc->hw.mac.type > e1000_82543) {
2037 		/* Figure our where our IO BAR is ? */
2038 		for (rid = PCIR_BAR(0); rid < PCIR_CIS;) {
2039 			val = pci_read_config(dev, rid, 4);
2040 			if (EM_BAR_TYPE(val) == EM_BAR_TYPE_IO) {
2041 				break;
2042 			}
2043 			rid += 4;
2044 			/* check for 64bit BAR */
2045 			if (EM_BAR_MEM_TYPE(val) == EM_BAR_MEM_TYPE_64BIT)
2046 				rid += 4;
2047 		}
2048 		if (rid >= PCIR_CIS) {
2049 			device_printf(dev, "Unable to locate IO BAR\n");
2050 			return (ENXIO);
2051 		}
2052 		sc->ioport = bus_alloc_resource_any(dev, SYS_RES_IOPORT,
2053 		    &rid, RF_ACTIVE);
2054 		if (sc->ioport == NULL) {
2055 			device_printf(dev, "Unable to allocate bus resource: "
2056 			    "ioport\n");
2057 			return (ENXIO);
2058 		}
2059 		sc->hw.io_base = 0;
2060 		sc->osdep.io_bus_space_tag =
2061 		    rman_get_bustag(sc->ioport);
2062 		sc->osdep.io_bus_space_handle =
2063 		    rman_get_bushandle(sc->ioport);
2064 	}
2065 
2066 	sc->hw.back = &sc->osdep;
2067 
2068 	return (0);
2069 }
2070 
2071 /*********************************************************************
2072  *
2073  *  Set up the MSI-X Interrupt handlers
2074  *
2075  **********************************************************************/
2076 static int
2077 em_if_msix_intr_assign(if_ctx_t ctx, int msix)
2078 {
2079 	struct e1000_softc *sc = iflib_get_softc(ctx);
2080 	struct em_rx_queue *rx_que = sc->rx_queues;
2081 	struct em_tx_queue *tx_que = sc->tx_queues;
2082 	int error, rid, i, vector = 0, rx_vectors;
2083 	char buf[16];
2084 
2085 	/* First set up ring resources */
2086 	for (i = 0; i < sc->rx_num_queues; i++, rx_que++, vector++) {
2087 		rid = vector + 1;
2088 		snprintf(buf, sizeof(buf), "rxq%d", i);
2089 		error = iflib_irq_alloc_generic(ctx, &rx_que->que_irq, rid, IFLIB_INTR_RXTX, em_msix_que, rx_que, rx_que->me, buf);
2090 		if (error) {
2091 			device_printf(iflib_get_dev(ctx), "Failed to allocate que int %d err: %d", i, error);
2092 			sc->rx_num_queues = i + 1;
2093 			goto fail;
2094 		}
2095 
2096 		rx_que->msix =  vector;
2097 
2098 		/*
2099 		 * Set the bit to enable interrupt
2100 		 * in E1000_IMS -- bits 20 and 21
2101 		 * are for RX0 and RX1, note this has
2102 		 * NOTHING to do with the MSI-X vector
2103 		 */
2104 		if (sc->hw.mac.type == e1000_82574) {
2105 			rx_que->eims = 1 << (20 + i);
2106 			sc->ims |= rx_que->eims;
2107 			sc->ivars |= (8 | rx_que->msix) << (i * 4);
2108 		} else if (sc->hw.mac.type == e1000_82575)
2109 			rx_que->eims = E1000_EICR_TX_QUEUE0 << vector;
2110 		else
2111 			rx_que->eims = 1 << vector;
2112 	}
2113 	rx_vectors = vector;
2114 
2115 	vector = 0;
2116 	for (i = 0; i < sc->tx_num_queues; i++, tx_que++, vector++) {
2117 		snprintf(buf, sizeof(buf), "txq%d", i);
2118 		tx_que = &sc->tx_queues[i];
2119 		iflib_softirq_alloc_generic(ctx,
2120 		    &sc->rx_queues[i % sc->rx_num_queues].que_irq,
2121 		    IFLIB_INTR_TX, tx_que, tx_que->me, buf);
2122 
2123 		tx_que->msix = (vector % sc->rx_num_queues);
2124 
2125 		/*
2126 		 * Set the bit to enable interrupt
2127 		 * in E1000_IMS -- bits 22 and 23
2128 		 * are for TX0 and TX1, note this has
2129 		 * NOTHING to do with the MSI-X vector
2130 		 */
2131 		if (sc->hw.mac.type == e1000_82574) {
2132 			tx_que->eims = 1 << (22 + i);
2133 			sc->ims |= tx_que->eims;
2134 			sc->ivars |= (8 | tx_que->msix) << (8 + (i * 4));
2135 		} else if (sc->hw.mac.type == e1000_82575) {
2136 			tx_que->eims = E1000_EICR_TX_QUEUE0 << i;
2137 		} else {
2138 			tx_que->eims = 1 << i;
2139 		}
2140 	}
2141 
2142 	/* Link interrupt */
2143 	rid = rx_vectors + 1;
2144 	error = iflib_irq_alloc_generic(ctx, &sc->irq, rid, IFLIB_INTR_ADMIN, em_msix_link, sc, 0, "aq");
2145 
2146 	if (error) {
2147 		device_printf(iflib_get_dev(ctx), "Failed to register admin handler");
2148 		goto fail;
2149 	}
2150 	sc->linkvec = rx_vectors;
2151 	if (sc->hw.mac.type < igb_mac_min) {
2152 		sc->ivars |=  (8 | rx_vectors) << 16;
2153 		sc->ivars |= 0x80000000;
2154 		/* Enable the "Other" interrupt type for link status change */
2155 		sc->ims |= E1000_IMS_OTHER;
2156 	}
2157 
2158 	return (0);
2159 fail:
2160 	iflib_irq_free(ctx, &sc->irq);
2161 	rx_que = sc->rx_queues;
2162 	for (int i = 0; i < sc->rx_num_queues; i++, rx_que++)
2163 		iflib_irq_free(ctx, &rx_que->que_irq);
2164 	return (error);
2165 }
2166 
2167 static void
2168 igb_configure_queues(struct e1000_softc *sc)
2169 {
2170 	struct e1000_hw *hw = &sc->hw;
2171 	struct em_rx_queue *rx_que;
2172 	struct em_tx_queue *tx_que;
2173 	u32 tmp, ivar = 0, newitr = 0;
2174 
2175 	/* First turn on RSS capability */
2176 	if (hw->mac.type != e1000_82575)
2177 		E1000_WRITE_REG(hw, E1000_GPIE,
2178 		    E1000_GPIE_MSIX_MODE | E1000_GPIE_EIAME |
2179 		    E1000_GPIE_PBA | E1000_GPIE_NSICR);
2180 
2181 	/* Turn on MSI-X */
2182 	switch (hw->mac.type) {
2183 	case e1000_82580:
2184 	case e1000_i350:
2185 	case e1000_i354:
2186 	case e1000_i210:
2187 	case e1000_i211:
2188 	case e1000_vfadapt:
2189 	case e1000_vfadapt_i350:
2190 		/* RX entries */
2191 		for (int i = 0; i < sc->rx_num_queues; i++) {
2192 			u32 index = i >> 1;
2193 			ivar = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index);
2194 			rx_que = &sc->rx_queues[i];
2195 			if (i & 1) {
2196 				ivar &= 0xFF00FFFF;
2197 				ivar |= (rx_que->msix | E1000_IVAR_VALID) << 16;
2198 			} else {
2199 				ivar &= 0xFFFFFF00;
2200 				ivar |= rx_que->msix | E1000_IVAR_VALID;
2201 			}
2202 			E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, ivar);
2203 		}
2204 		/* TX entries */
2205 		for (int i = 0; i < sc->tx_num_queues; i++) {
2206 			u32 index = i >> 1;
2207 			ivar = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index);
2208 			tx_que = &sc->tx_queues[i];
2209 			if (i & 1) {
2210 				ivar &= 0x00FFFFFF;
2211 				ivar |= (tx_que->msix | E1000_IVAR_VALID) << 24;
2212 			} else {
2213 				ivar &= 0xFFFF00FF;
2214 				ivar |= (tx_que->msix | E1000_IVAR_VALID) << 8;
2215 			}
2216 			E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, ivar);
2217 			sc->que_mask |= tx_que->eims;
2218 		}
2219 
2220 		/* And for the link interrupt */
2221 		ivar = (sc->linkvec | E1000_IVAR_VALID) << 8;
2222 		sc->link_mask = 1 << sc->linkvec;
2223 		E1000_WRITE_REG(hw, E1000_IVAR_MISC, ivar);
2224 		break;
2225 	case e1000_82576:
2226 		/* RX entries */
2227 		for (int i = 0; i < sc->rx_num_queues; i++) {
2228 			u32 index = i & 0x7; /* Each IVAR has two entries */
2229 			ivar = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index);
2230 			rx_que = &sc->rx_queues[i];
2231 			if (i < 8) {
2232 				ivar &= 0xFFFFFF00;
2233 				ivar |= rx_que->msix | E1000_IVAR_VALID;
2234 			} else {
2235 				ivar &= 0xFF00FFFF;
2236 				ivar |= (rx_que->msix | E1000_IVAR_VALID) << 16;
2237 			}
2238 			E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, ivar);
2239 			sc->que_mask |= rx_que->eims;
2240 		}
2241 		/* TX entries */
2242 		for (int i = 0; i < sc->tx_num_queues; i++) {
2243 			u32 index = i & 0x7; /* Each IVAR has two entries */
2244 			ivar = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index);
2245 			tx_que = &sc->tx_queues[i];
2246 			if (i < 8) {
2247 				ivar &= 0xFFFF00FF;
2248 				ivar |= (tx_que->msix | E1000_IVAR_VALID) << 8;
2249 			} else {
2250 				ivar &= 0x00FFFFFF;
2251 				ivar |= (tx_que->msix | E1000_IVAR_VALID) << 24;
2252 			}
2253 			E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, ivar);
2254 			sc->que_mask |= tx_que->eims;
2255 		}
2256 
2257 		/* And for the link interrupt */
2258 		ivar = (sc->linkvec | E1000_IVAR_VALID) << 8;
2259 		sc->link_mask = 1 << sc->linkvec;
2260 		E1000_WRITE_REG(hw, E1000_IVAR_MISC, ivar);
2261 		break;
2262 
2263 	case e1000_82575:
2264 		/* enable MSI-X support*/
2265 		tmp = E1000_READ_REG(hw, E1000_CTRL_EXT);
2266 		tmp |= E1000_CTRL_EXT_PBA_CLR;
2267 		/* Auto-Mask interrupts upon ICR read. */
2268 		tmp |= E1000_CTRL_EXT_EIAME;
2269 		tmp |= E1000_CTRL_EXT_IRCA;
2270 		E1000_WRITE_REG(hw, E1000_CTRL_EXT, tmp);
2271 
2272 		/* Queues */
2273 		for (int i = 0; i < sc->rx_num_queues; i++) {
2274 			rx_que = &sc->rx_queues[i];
2275 			tmp = E1000_EICR_RX_QUEUE0 << i;
2276 			tmp |= E1000_EICR_TX_QUEUE0 << i;
2277 			rx_que->eims = tmp;
2278 			E1000_WRITE_REG_ARRAY(hw, E1000_MSIXBM(0),
2279 			    i, rx_que->eims);
2280 			sc->que_mask |= rx_que->eims;
2281 		}
2282 
2283 		/* Link */
2284 		E1000_WRITE_REG(hw, E1000_MSIXBM(sc->linkvec),
2285 		    E1000_EIMS_OTHER);
2286 		sc->link_mask |= E1000_EIMS_OTHER;
2287 	default:
2288 		break;
2289 	}
2290 
2291 	/* Set the starting interrupt rate */
2292 	if (em_max_interrupt_rate > 0)
2293 		newitr = (4000000 / em_max_interrupt_rate) & 0x7FFC;
2294 
2295 	if (hw->mac.type == e1000_82575)
2296 		newitr |= newitr << 16;
2297 	else
2298 		newitr |= E1000_EITR_CNT_IGNR;
2299 
2300 	for (int i = 0; i < sc->rx_num_queues; i++) {
2301 		rx_que = &sc->rx_queues[i];
2302 		E1000_WRITE_REG(hw, E1000_EITR(rx_que->msix), newitr);
2303 	}
2304 
2305 	return;
2306 }
2307 
2308 static void
2309 em_free_pci_resources(if_ctx_t ctx)
2310 {
2311 	struct e1000_softc *sc = iflib_get_softc(ctx);
2312 	struct em_rx_queue *que = sc->rx_queues;
2313 	device_t dev = iflib_get_dev(ctx);
2314 
2315 	/* Release all MSI-X queue resources */
2316 	if (sc->intr_type == IFLIB_INTR_MSIX)
2317 		iflib_irq_free(ctx, &sc->irq);
2318 
2319 	if (que != NULL) {
2320 		for (int i = 0; i < sc->rx_num_queues; i++, que++) {
2321 			iflib_irq_free(ctx, &que->que_irq);
2322 		}
2323 	}
2324 
2325 	if (sc->memory != NULL) {
2326 		bus_release_resource(dev, SYS_RES_MEMORY,
2327 		    rman_get_rid(sc->memory), sc->memory);
2328 		sc->memory = NULL;
2329 	}
2330 
2331 	if (sc->flash != NULL) {
2332 		bus_release_resource(dev, SYS_RES_MEMORY,
2333 		    rman_get_rid(sc->flash), sc->flash);
2334 		sc->flash = NULL;
2335 	}
2336 
2337 	if (sc->ioport != NULL) {
2338 		bus_release_resource(dev, SYS_RES_IOPORT,
2339 		    rman_get_rid(sc->ioport), sc->ioport);
2340 		sc->ioport = NULL;
2341 	}
2342 }
2343 
2344 /* Set up MSI or MSI-X */
2345 static int
2346 em_setup_msix(if_ctx_t ctx)
2347 {
2348 	struct e1000_softc *sc = iflib_get_softc(ctx);
2349 
2350 	if (sc->hw.mac.type == e1000_82574) {
2351 		em_enable_vectors_82574(ctx);
2352 	}
2353 	return (0);
2354 }
2355 
2356 /*********************************************************************
2357  *
2358  *  Workaround for SmartSpeed on 82541 and 82547 controllers
2359  *
2360  **********************************************************************/
2361 static void
2362 lem_smartspeed(struct e1000_softc *sc)
2363 {
2364 	u16 phy_tmp;
2365 
2366 	if (sc->link_active || (sc->hw.phy.type != e1000_phy_igp) ||
2367 	    sc->hw.mac.autoneg == 0 ||
2368 	    (sc->hw.phy.autoneg_advertised & ADVERTISE_1000_FULL) == 0)
2369 		return;
2370 
2371 	if (sc->smartspeed == 0) {
2372 		/* If Master/Slave config fault is asserted twice,
2373 		 * we assume back-to-back */
2374 		e1000_read_phy_reg(&sc->hw, PHY_1000T_STATUS, &phy_tmp);
2375 		if (!(phy_tmp & SR_1000T_MS_CONFIG_FAULT))
2376 			return;
2377 		e1000_read_phy_reg(&sc->hw, PHY_1000T_STATUS, &phy_tmp);
2378 		if (phy_tmp & SR_1000T_MS_CONFIG_FAULT) {
2379 			e1000_read_phy_reg(&sc->hw,
2380 			    PHY_1000T_CTRL, &phy_tmp);
2381 			if(phy_tmp & CR_1000T_MS_ENABLE) {
2382 				phy_tmp &= ~CR_1000T_MS_ENABLE;
2383 				e1000_write_phy_reg(&sc->hw,
2384 				    PHY_1000T_CTRL, phy_tmp);
2385 				sc->smartspeed++;
2386 				if(sc->hw.mac.autoneg &&
2387 				   !e1000_copper_link_autoneg(&sc->hw) &&
2388 				   !e1000_read_phy_reg(&sc->hw,
2389 				    PHY_CONTROL, &phy_tmp)) {
2390 					phy_tmp |= (MII_CR_AUTO_NEG_EN |
2391 						    MII_CR_RESTART_AUTO_NEG);
2392 					e1000_write_phy_reg(&sc->hw,
2393 					    PHY_CONTROL, phy_tmp);
2394 				}
2395 			}
2396 		}
2397 		return;
2398 	} else if(sc->smartspeed == EM_SMARTSPEED_DOWNSHIFT) {
2399 		/* If still no link, perhaps using 2/3 pair cable */
2400 		e1000_read_phy_reg(&sc->hw, PHY_1000T_CTRL, &phy_tmp);
2401 		phy_tmp |= CR_1000T_MS_ENABLE;
2402 		e1000_write_phy_reg(&sc->hw, PHY_1000T_CTRL, phy_tmp);
2403 		if(sc->hw.mac.autoneg &&
2404 		   !e1000_copper_link_autoneg(&sc->hw) &&
2405 		   !e1000_read_phy_reg(&sc->hw, PHY_CONTROL, &phy_tmp)) {
2406 			phy_tmp |= (MII_CR_AUTO_NEG_EN |
2407 				    MII_CR_RESTART_AUTO_NEG);
2408 			e1000_write_phy_reg(&sc->hw, PHY_CONTROL, phy_tmp);
2409 		}
2410 	}
2411 	/* Restart process after EM_SMARTSPEED_MAX iterations */
2412 	if(sc->smartspeed++ == EM_SMARTSPEED_MAX)
2413 		sc->smartspeed = 0;
2414 }
2415 
2416 /*********************************************************************
2417  *
2418  *  Initialize the DMA Coalescing feature
2419  *
2420  **********************************************************************/
2421 static void
2422 igb_init_dmac(struct e1000_softc *sc, u32 pba)
2423 {
2424 	device_t	dev = sc->dev;
2425 	struct e1000_hw *hw = &sc->hw;
2426 	u32 		dmac, reg = ~E1000_DMACR_DMAC_EN;
2427 	u16		hwm;
2428 	u16		max_frame_size;
2429 
2430 	if (hw->mac.type == e1000_i211)
2431 		return;
2432 
2433 	max_frame_size = sc->shared->isc_max_frame_size;
2434 	if (hw->mac.type > e1000_82580) {
2435 
2436 		if (sc->dmac == 0) { /* Disabling it */
2437 			E1000_WRITE_REG(hw, E1000_DMACR, reg);
2438 			return;
2439 		} else
2440 			device_printf(dev, "DMA Coalescing enabled\n");
2441 
2442 		/* Set starting threshold */
2443 		E1000_WRITE_REG(hw, E1000_DMCTXTH, 0);
2444 
2445 		hwm = 64 * pba - max_frame_size / 16;
2446 		if (hwm < 64 * (pba - 6))
2447 			hwm = 64 * (pba - 6);
2448 		reg = E1000_READ_REG(hw, E1000_FCRTC);
2449 		reg &= ~E1000_FCRTC_RTH_COAL_MASK;
2450 		reg |= ((hwm << E1000_FCRTC_RTH_COAL_SHIFT)
2451 		    & E1000_FCRTC_RTH_COAL_MASK);
2452 		E1000_WRITE_REG(hw, E1000_FCRTC, reg);
2453 
2454 
2455 		dmac = pba - max_frame_size / 512;
2456 		if (dmac < pba - 10)
2457 			dmac = pba - 10;
2458 		reg = E1000_READ_REG(hw, E1000_DMACR);
2459 		reg &= ~E1000_DMACR_DMACTHR_MASK;
2460 		reg |= ((dmac << E1000_DMACR_DMACTHR_SHIFT)
2461 		    & E1000_DMACR_DMACTHR_MASK);
2462 
2463 		/* transition to L0x or L1 if available..*/
2464 		reg |= (E1000_DMACR_DMAC_EN | E1000_DMACR_DMAC_LX_MASK);
2465 
2466 		/* Check if status is 2.5Gb backplane connection
2467 		* before configuration of watchdog timer, which is
2468 		* in msec values in 12.8usec intervals
2469 		* watchdog timer= msec values in 32usec intervals
2470 		* for non 2.5Gb connection
2471 		*/
2472 		if (hw->mac.type == e1000_i354) {
2473 			int status = E1000_READ_REG(hw, E1000_STATUS);
2474 			if ((status & E1000_STATUS_2P5_SKU) &&
2475 			    (!(status & E1000_STATUS_2P5_SKU_OVER)))
2476 				reg |= ((sc->dmac * 5) >> 6);
2477 			else
2478 				reg |= (sc->dmac >> 5);
2479 		} else {
2480 			reg |= (sc->dmac >> 5);
2481 		}
2482 
2483 		E1000_WRITE_REG(hw, E1000_DMACR, reg);
2484 
2485 		E1000_WRITE_REG(hw, E1000_DMCRTRH, 0);
2486 
2487 		/* Set the interval before transition */
2488 		reg = E1000_READ_REG(hw, E1000_DMCTLX);
2489 		if (hw->mac.type == e1000_i350)
2490 			reg |= IGB_DMCTLX_DCFLUSH_DIS;
2491 		/*
2492 		** in 2.5Gb connection, TTLX unit is 0.4 usec
2493 		** which is 0x4*2 = 0xA. But delay is still 4 usec
2494 		*/
2495 		if (hw->mac.type == e1000_i354) {
2496 			int status = E1000_READ_REG(hw, E1000_STATUS);
2497 			if ((status & E1000_STATUS_2P5_SKU) &&
2498 			    (!(status & E1000_STATUS_2P5_SKU_OVER)))
2499 				reg |= 0xA;
2500 			else
2501 				reg |= 0x4;
2502 		} else {
2503 			reg |= 0x4;
2504 		}
2505 
2506 		E1000_WRITE_REG(hw, E1000_DMCTLX, reg);
2507 
2508 		/* free space in tx packet buffer to wake from DMA coal */
2509 		E1000_WRITE_REG(hw, E1000_DMCTXTH, (IGB_TXPBSIZE -
2510 		    (2 * max_frame_size)) >> 6);
2511 
2512 		/* make low power state decision controlled by DMA coal */
2513 		reg = E1000_READ_REG(hw, E1000_PCIEMISC);
2514 		reg &= ~E1000_PCIEMISC_LX_DECISION;
2515 		E1000_WRITE_REG(hw, E1000_PCIEMISC, reg);
2516 
2517 	} else if (hw->mac.type == e1000_82580) {
2518 		u32 reg = E1000_READ_REG(hw, E1000_PCIEMISC);
2519 		E1000_WRITE_REG(hw, E1000_PCIEMISC,
2520 		    reg & ~E1000_PCIEMISC_LX_DECISION);
2521 		E1000_WRITE_REG(hw, E1000_DMACR, 0);
2522 	}
2523 }
2524 /*********************************************************************
2525  * The 3 following flush routines are used as a workaround in the
2526  * I219 client parts and only for them.
2527  *
2528  * em_flush_tx_ring - remove all descriptors from the tx_ring
2529  *
2530  * We want to clear all pending descriptors from the TX ring.
2531  * zeroing happens when the HW reads the regs. We assign the ring itself as
2532  * the data of the next descriptor. We don't care about the data we are about
2533  * to reset the HW.
2534  **********************************************************************/
2535 static void
2536 em_flush_tx_ring(struct e1000_softc *sc)
2537 {
2538 	struct e1000_hw		*hw = &sc->hw;
2539 	struct tx_ring		*txr = &sc->tx_queues->txr;
2540 	struct e1000_tx_desc	*txd;
2541 	u32			tctl, txd_lower = E1000_TXD_CMD_IFCS;
2542 	u16			size = 512;
2543 
2544 	tctl = E1000_READ_REG(hw, E1000_TCTL);
2545 	E1000_WRITE_REG(hw, E1000_TCTL, tctl | E1000_TCTL_EN);
2546 
2547 	txd = &txr->tx_base[txr->tx_cidx_processed];
2548 
2549 	/* Just use the ring as a dummy buffer addr */
2550 	txd->buffer_addr = txr->tx_paddr;
2551 	txd->lower.data = htole32(txd_lower | size);
2552 	txd->upper.data = 0;
2553 
2554 	/* flush descriptors to memory before notifying the HW */
2555 	wmb();
2556 
2557 	E1000_WRITE_REG(hw, E1000_TDT(0), txr->tx_cidx_processed);
2558 	mb();
2559 	usec_delay(250);
2560 }
2561 
2562 /*********************************************************************
2563  * em_flush_rx_ring - remove all descriptors from the rx_ring
2564  *
2565  * Mark all descriptors in the RX ring as consumed and disable the rx ring
2566  **********************************************************************/
2567 static void
2568 em_flush_rx_ring(struct e1000_softc *sc)
2569 {
2570 	struct e1000_hw	*hw = &sc->hw;
2571 	u32		rctl, rxdctl;
2572 
2573 	rctl = E1000_READ_REG(hw, E1000_RCTL);
2574 	E1000_WRITE_REG(hw, E1000_RCTL, rctl & ~E1000_RCTL_EN);
2575 	E1000_WRITE_FLUSH(hw);
2576 	usec_delay(150);
2577 
2578 	rxdctl = E1000_READ_REG(hw, E1000_RXDCTL(0));
2579 	/* zero the lower 14 bits (prefetch and host thresholds) */
2580 	rxdctl &= 0xffffc000;
2581 	/*
2582 	 * update thresholds: prefetch threshold to 31, host threshold to 1
2583 	 * and make sure the granularity is "descriptors" and not "cache lines"
2584 	 */
2585 	rxdctl |= (0x1F | (1 << 8) | E1000_RXDCTL_THRESH_UNIT_DESC);
2586 	E1000_WRITE_REG(hw, E1000_RXDCTL(0), rxdctl);
2587 
2588 	/* momentarily enable the RX ring for the changes to take effect */
2589 	E1000_WRITE_REG(hw, E1000_RCTL, rctl | E1000_RCTL_EN);
2590 	E1000_WRITE_FLUSH(hw);
2591 	usec_delay(150);
2592 	E1000_WRITE_REG(hw, E1000_RCTL, rctl & ~E1000_RCTL_EN);
2593 }
2594 
2595 /*********************************************************************
2596  * em_flush_desc_rings - remove all descriptors from the descriptor rings
2597  *
2598  * In I219, the descriptor rings must be emptied before resetting the HW
2599  * or before changing the device state to D3 during runtime (runtime PM).
2600  *
2601  * Failure to do this will cause the HW to enter a unit hang state which can
2602  * only be released by PCI reset on the device
2603  *
2604  **********************************************************************/
2605 static void
2606 em_flush_desc_rings(struct e1000_softc *sc)
2607 {
2608 	struct e1000_hw	*hw = &sc->hw;
2609 	device_t dev = sc->dev;
2610 	u16		hang_state;
2611 	u32		fext_nvm11, tdlen;
2612 
2613 	/* First, disable MULR fix in FEXTNVM11 */
2614 	fext_nvm11 = E1000_READ_REG(hw, E1000_FEXTNVM11);
2615 	fext_nvm11 |= E1000_FEXTNVM11_DISABLE_MULR_FIX;
2616 	E1000_WRITE_REG(hw, E1000_FEXTNVM11, fext_nvm11);
2617 
2618 	/* do nothing if we're not in faulty state, or if the queue is empty */
2619 	tdlen = E1000_READ_REG(hw, E1000_TDLEN(0));
2620 	hang_state = pci_read_config(dev, PCICFG_DESC_RING_STATUS, 2);
2621 	if (!(hang_state & FLUSH_DESC_REQUIRED) || !tdlen)
2622 		return;
2623 	em_flush_tx_ring(sc);
2624 
2625 	/* recheck, maybe the fault is caused by the rx ring */
2626 	hang_state = pci_read_config(dev, PCICFG_DESC_RING_STATUS, 2);
2627 	if (hang_state & FLUSH_DESC_REQUIRED)
2628 		em_flush_rx_ring(sc);
2629 }
2630 
2631 
2632 /*********************************************************************
2633  *
2634  *  Initialize the hardware to a configuration as specified by the
2635  *  sc structure.
2636  *
2637  **********************************************************************/
2638 static void
2639 em_reset(if_ctx_t ctx)
2640 {
2641 	device_t dev = iflib_get_dev(ctx);
2642 	struct e1000_softc *sc = iflib_get_softc(ctx);
2643 	if_t ifp = iflib_get_ifp(ctx);
2644 	struct e1000_hw *hw = &sc->hw;
2645 	u32 rx_buffer_size;
2646 	u32 pba;
2647 
2648 	INIT_DEBUGOUT("em_reset: begin");
2649 	/* Let the firmware know the OS is in control */
2650 	em_get_hw_control(sc);
2651 
2652 	/* Set up smart power down as default off on newer adapters. */
2653 	if (!em_smart_pwr_down && (hw->mac.type == e1000_82571 ||
2654 	    hw->mac.type == e1000_82572)) {
2655 		u16 phy_tmp = 0;
2656 
2657 		/* Speed up time to link by disabling smart power down. */
2658 		e1000_read_phy_reg(hw, IGP02E1000_PHY_POWER_MGMT, &phy_tmp);
2659 		phy_tmp &= ~IGP02E1000_PM_SPD;
2660 		e1000_write_phy_reg(hw, IGP02E1000_PHY_POWER_MGMT, phy_tmp);
2661 	}
2662 
2663 	/*
2664 	 * Packet Buffer Allocation (PBA)
2665 	 * Writing PBA sets the receive portion of the buffer
2666 	 * the remainder is used for the transmit buffer.
2667 	 */
2668 	switch (hw->mac.type) {
2669 	/* 82547: Total Packet Buffer is 40K */
2670 	case e1000_82547:
2671 	case e1000_82547_rev_2:
2672 		if (hw->mac.max_frame_size > 8192)
2673 			pba = E1000_PBA_22K; /* 22K for Rx, 18K for Tx */
2674 		else
2675 			pba = E1000_PBA_30K; /* 30K for Rx, 10K for Tx */
2676 		break;
2677 	/* 82571/82572/80003es2lan: Total Packet Buffer is 48K */
2678 	case e1000_82571:
2679 	case e1000_82572:
2680 	case e1000_80003es2lan:
2681 			pba = E1000_PBA_32K; /* 32K for Rx, 16K for Tx */
2682 		break;
2683 	/* 82573: Total Packet Buffer is 32K */
2684 	case e1000_82573:
2685 			pba = E1000_PBA_12K; /* 12K for Rx, 20K for Tx */
2686 		break;
2687 	case e1000_82574:
2688 	case e1000_82583:
2689 			pba = E1000_PBA_20K; /* 20K for Rx, 20K for Tx */
2690 		break;
2691 	case e1000_ich8lan:
2692 		pba = E1000_PBA_8K;
2693 		break;
2694 	case e1000_ich9lan:
2695 	case e1000_ich10lan:
2696 		/* Boost Receive side for jumbo frames */
2697 		if (hw->mac.max_frame_size > 4096)
2698 			pba = E1000_PBA_14K;
2699 		else
2700 			pba = E1000_PBA_10K;
2701 		break;
2702 	case e1000_pchlan:
2703 	case e1000_pch2lan:
2704 	case e1000_pch_lpt:
2705 	case e1000_pch_spt:
2706 	case e1000_pch_cnp:
2707 	case e1000_pch_tgp:
2708 	case e1000_pch_adp:
2709 	case e1000_pch_mtp:
2710 	case e1000_pch_ptp:
2711 		pba = E1000_PBA_26K;
2712 		break;
2713 	case e1000_82575:
2714 		pba = E1000_PBA_32K;
2715 		break;
2716 	case e1000_82576:
2717 	case e1000_vfadapt:
2718 		pba = E1000_READ_REG(hw, E1000_RXPBS);
2719 		pba &= E1000_RXPBS_SIZE_MASK_82576;
2720 		break;
2721 	case e1000_82580:
2722 	case e1000_i350:
2723 	case e1000_i354:
2724 	case e1000_vfadapt_i350:
2725 		pba = E1000_READ_REG(hw, E1000_RXPBS);
2726 		pba = e1000_rxpbs_adjust_82580(pba);
2727 		break;
2728 	case e1000_i210:
2729 	case e1000_i211:
2730 		pba = E1000_PBA_34K;
2731 		break;
2732 	default:
2733 		/* Remaining devices assumed to have a Packet Buffer of 64K. */
2734 		if (hw->mac.max_frame_size > 8192)
2735 			pba = E1000_PBA_40K; /* 40K for Rx, 24K for Tx */
2736 		else
2737 			pba = E1000_PBA_48K; /* 48K for Rx, 16K for Tx */
2738 	}
2739 
2740 	/* Special needs in case of Jumbo frames */
2741 	if ((hw->mac.type == e1000_82575) && (if_getmtu(ifp) > ETHERMTU)) {
2742 		u32 tx_space, min_tx, min_rx;
2743 		pba = E1000_READ_REG(hw, E1000_PBA);
2744 		tx_space = pba >> 16;
2745 		pba &= 0xffff;
2746 		min_tx = (hw->mac.max_frame_size +
2747 		    sizeof(struct e1000_tx_desc) - ETHERNET_FCS_SIZE) * 2;
2748 		min_tx = roundup2(min_tx, 1024);
2749 		min_tx >>= 10;
2750 		min_rx = hw->mac.max_frame_size;
2751 		min_rx = roundup2(min_rx, 1024);
2752 		min_rx >>= 10;
2753 		if (tx_space < min_tx &&
2754 		    ((min_tx - tx_space) < pba)) {
2755 			pba = pba - (min_tx - tx_space);
2756 			/*
2757 			 * if short on rx space, rx wins
2758 			 * and must trump tx adjustment
2759 			 */
2760 			if (pba < min_rx)
2761 				pba = min_rx;
2762 		}
2763 		E1000_WRITE_REG(hw, E1000_PBA, pba);
2764 	}
2765 
2766 	if (hw->mac.type < igb_mac_min)
2767 		E1000_WRITE_REG(hw, E1000_PBA, pba);
2768 
2769 	INIT_DEBUGOUT1("em_reset: pba=%dK",pba);
2770 
2771 	/*
2772 	 * These parameters control the automatic generation (Tx) and
2773 	 * response (Rx) to Ethernet PAUSE frames.
2774 	 * - High water mark should allow for at least two frames to be
2775 	 *   received after sending an XOFF.
2776 	 * - Low water mark works best when it is very near the high water mark.
2777 	 *   This allows the receiver to restart by sending XON when it has
2778 	 *   drained a bit. Here we use an arbitrary value of 1500 which will
2779 	 *   restart after one full frame is pulled from the buffer. There
2780 	 *   could be several smaller frames in the buffer and if so they will
2781 	 *   not trigger the XON until their total number reduces the buffer
2782 	 *   by 1500.
2783 	 * - The pause time is fairly large at 1000 x 512ns = 512 usec.
2784 	 */
2785 	rx_buffer_size = (pba & 0xffff) << 10;
2786 	hw->fc.high_water = rx_buffer_size -
2787 	    roundup2(hw->mac.max_frame_size, 1024);
2788 	hw->fc.low_water = hw->fc.high_water - 1500;
2789 
2790 	if (sc->fc) /* locally set flow control value? */
2791 		hw->fc.requested_mode = sc->fc;
2792 	else
2793 		hw->fc.requested_mode = e1000_fc_full;
2794 
2795 	if (hw->mac.type == e1000_80003es2lan)
2796 		hw->fc.pause_time = 0xFFFF;
2797 	else
2798 		hw->fc.pause_time = EM_FC_PAUSE_TIME;
2799 
2800 	hw->fc.send_xon = true;
2801 
2802 	/* Device specific overrides/settings */
2803 	switch (hw->mac.type) {
2804 	case e1000_pchlan:
2805 		/* Workaround: no TX flow ctrl for PCH */
2806 		hw->fc.requested_mode = e1000_fc_rx_pause;
2807 		hw->fc.pause_time = 0xFFFF; /* override */
2808 		if (if_getmtu(ifp) > ETHERMTU) {
2809 			hw->fc.high_water = 0x3500;
2810 			hw->fc.low_water = 0x1500;
2811 		} else {
2812 			hw->fc.high_water = 0x5000;
2813 			hw->fc.low_water = 0x3000;
2814 		}
2815 		hw->fc.refresh_time = 0x1000;
2816 		break;
2817 	case e1000_pch2lan:
2818 	case e1000_pch_lpt:
2819 	case e1000_pch_spt:
2820 	case e1000_pch_cnp:
2821 	case e1000_pch_tgp:
2822 	case e1000_pch_adp:
2823 	case e1000_pch_mtp:
2824 	case e1000_pch_ptp:
2825 		hw->fc.high_water = 0x5C20;
2826 		hw->fc.low_water = 0x5048;
2827 		hw->fc.pause_time = 0x0650;
2828 		hw->fc.refresh_time = 0x0400;
2829 		/* Jumbos need adjusted PBA */
2830 		if (if_getmtu(ifp) > ETHERMTU)
2831 			E1000_WRITE_REG(hw, E1000_PBA, 12);
2832 		else
2833 			E1000_WRITE_REG(hw, E1000_PBA, 26);
2834 		break;
2835 	case e1000_82575:
2836 	case e1000_82576:
2837 		/* 8-byte granularity */
2838 		hw->fc.low_water = hw->fc.high_water - 8;
2839 		break;
2840 	case e1000_82580:
2841 	case e1000_i350:
2842 	case e1000_i354:
2843 	case e1000_i210:
2844 	case e1000_i211:
2845 	case e1000_vfadapt:
2846 	case e1000_vfadapt_i350:
2847 		/* 16-byte granularity */
2848 		hw->fc.low_water = hw->fc.high_water - 16;
2849 		break;
2850 	case e1000_ich9lan:
2851 	case e1000_ich10lan:
2852 		if (if_getmtu(ifp) > ETHERMTU) {
2853 			hw->fc.high_water = 0x2800;
2854 			hw->fc.low_water = hw->fc.high_water - 8;
2855 			break;
2856 		}
2857 		/* FALLTHROUGH */
2858 	default:
2859 		if (hw->mac.type == e1000_80003es2lan)
2860 			hw->fc.pause_time = 0xFFFF;
2861 		break;
2862 	}
2863 
2864 	/* I219 needs some special flushing to avoid hangs */
2865 	if (sc->hw.mac.type >= e1000_pch_spt && sc->hw.mac.type < igb_mac_min)
2866 		em_flush_desc_rings(sc);
2867 
2868 	/* Issue a global reset */
2869 	e1000_reset_hw(hw);
2870 	if (hw->mac.type >= igb_mac_min) {
2871 		E1000_WRITE_REG(hw, E1000_WUC, 0);
2872 	} else {
2873 		E1000_WRITE_REG(hw, E1000_WUFC, 0);
2874 		em_disable_aspm(sc);
2875 	}
2876 	if (sc->flags & IGB_MEDIA_RESET) {
2877 		e1000_setup_init_funcs(hw, true);
2878 		e1000_get_bus_info(hw);
2879 		sc->flags &= ~IGB_MEDIA_RESET;
2880 	}
2881 	/* and a re-init */
2882 	if (e1000_init_hw(hw) < 0) {
2883 		device_printf(dev, "Hardware Initialization Failed\n");
2884 		return;
2885 	}
2886 	if (hw->mac.type >= igb_mac_min)
2887 		igb_init_dmac(sc, pba);
2888 
2889 	E1000_WRITE_REG(hw, E1000_VET, ETHERTYPE_VLAN);
2890 	e1000_get_phy_info(hw);
2891 	e1000_check_for_link(hw);
2892 }
2893 
2894 /*
2895  * Initialise the RSS mapping for NICs that support multiple transmit/
2896  * receive rings.
2897  */
2898 
2899 #define RSSKEYLEN 10
2900 static void
2901 em_initialize_rss_mapping(struct e1000_softc *sc)
2902 {
2903 	uint8_t  rss_key[4 * RSSKEYLEN];
2904 	uint32_t reta = 0;
2905 	struct e1000_hw	*hw = &sc->hw;
2906 	int i;
2907 
2908 	/*
2909 	 * Configure RSS key
2910 	 */
2911 	arc4rand(rss_key, sizeof(rss_key), 0);
2912 	for (i = 0; i < RSSKEYLEN; ++i) {
2913 		uint32_t rssrk = 0;
2914 
2915 		rssrk = EM_RSSRK_VAL(rss_key, i);
2916 		E1000_WRITE_REG(hw,E1000_RSSRK(i), rssrk);
2917 	}
2918 
2919 	/*
2920 	 * Configure RSS redirect table in following fashion:
2921 	 * (hash & ring_cnt_mask) == rdr_table[(hash & rdr_table_mask)]
2922 	 */
2923 	for (i = 0; i < sizeof(reta); ++i) {
2924 		uint32_t q;
2925 
2926 		q = (i % sc->rx_num_queues) << 7;
2927 		reta |= q << (8 * i);
2928 	}
2929 
2930 	for (i = 0; i < 32; ++i)
2931 		E1000_WRITE_REG(hw, E1000_RETA(i), reta);
2932 
2933 	E1000_WRITE_REG(hw, E1000_MRQC, E1000_MRQC_RSS_ENABLE_2Q |
2934 			E1000_MRQC_RSS_FIELD_IPV4_TCP |
2935 			E1000_MRQC_RSS_FIELD_IPV4 |
2936 			E1000_MRQC_RSS_FIELD_IPV6_TCP_EX |
2937 			E1000_MRQC_RSS_FIELD_IPV6_EX |
2938 			E1000_MRQC_RSS_FIELD_IPV6);
2939 }
2940 
2941 static void
2942 igb_initialize_rss_mapping(struct e1000_softc *sc)
2943 {
2944 	struct e1000_hw *hw = &sc->hw;
2945 	int i;
2946 	int queue_id;
2947 	u32 reta;
2948 	u32 rss_key[10], mrqc, shift = 0;
2949 
2950 	/* XXX? */
2951 	if (hw->mac.type == e1000_82575)
2952 		shift = 6;
2953 
2954 	/*
2955 	 * The redirection table controls which destination
2956 	 * queue each bucket redirects traffic to.
2957 	 * Each DWORD represents four queues, with the LSB
2958 	 * being the first queue in the DWORD.
2959 	 *
2960 	 * This just allocates buckets to queues using round-robin
2961 	 * allocation.
2962 	 *
2963 	 * NOTE: It Just Happens to line up with the default
2964 	 * RSS allocation method.
2965 	 */
2966 
2967 	/* Warning FM follows */
2968 	reta = 0;
2969 	for (i = 0; i < 128; i++) {
2970 #ifdef RSS
2971 		queue_id = rss_get_indirection_to_bucket(i);
2972 		/*
2973 		 * If we have more queues than buckets, we'll
2974 		 * end up mapping buckets to a subset of the
2975 		 * queues.
2976 		 *
2977 		 * If we have more buckets than queues, we'll
2978 		 * end up instead assigning multiple buckets
2979 		 * to queues.
2980 		 *
2981 		 * Both are suboptimal, but we need to handle
2982 		 * the case so we don't go out of bounds
2983 		 * indexing arrays and such.
2984 		 */
2985 		queue_id = queue_id % sc->rx_num_queues;
2986 #else
2987 		queue_id = (i % sc->rx_num_queues);
2988 #endif
2989 		/* Adjust if required */
2990 		queue_id = queue_id << shift;
2991 
2992 		/*
2993 		 * The low 8 bits are for hash value (n+0);
2994 		 * The next 8 bits are for hash value (n+1), etc.
2995 		 */
2996 		reta = reta >> 8;
2997 		reta = reta | ( ((uint32_t) queue_id) << 24);
2998 		if ((i & 3) == 3) {
2999 			E1000_WRITE_REG(hw, E1000_RETA(i >> 2), reta);
3000 			reta = 0;
3001 		}
3002 	}
3003 
3004 	/* Now fill in hash table */
3005 
3006 	/*
3007 	 * MRQC: Multiple Receive Queues Command
3008 	 * Set queuing to RSS control, number depends on the device.
3009 	 */
3010 	mrqc = E1000_MRQC_ENABLE_RSS_MQ;
3011 
3012 #ifdef RSS
3013 	/* XXX ew typecasting */
3014 	rss_getkey((uint8_t *) &rss_key);
3015 #else
3016 	arc4rand(&rss_key, sizeof(rss_key), 0);
3017 #endif
3018 	for (i = 0; i < 10; i++)
3019 		E1000_WRITE_REG_ARRAY(hw, E1000_RSSRK(0), i, rss_key[i]);
3020 
3021 	/*
3022 	 * Configure the RSS fields to hash upon.
3023 	 */
3024 	mrqc |= (E1000_MRQC_RSS_FIELD_IPV4 |
3025 	    E1000_MRQC_RSS_FIELD_IPV4_TCP);
3026 	mrqc |= (E1000_MRQC_RSS_FIELD_IPV6 |
3027 	    E1000_MRQC_RSS_FIELD_IPV6_TCP);
3028 	mrqc |=( E1000_MRQC_RSS_FIELD_IPV4_UDP |
3029 	    E1000_MRQC_RSS_FIELD_IPV6_UDP);
3030 	mrqc |=( E1000_MRQC_RSS_FIELD_IPV6_UDP_EX |
3031 	    E1000_MRQC_RSS_FIELD_IPV6_TCP_EX);
3032 
3033 	E1000_WRITE_REG(hw, E1000_MRQC, mrqc);
3034 }
3035 
3036 /*********************************************************************
3037  *
3038  *  Setup networking device structure and register interface media.
3039  *
3040  **********************************************************************/
3041 static int
3042 em_setup_interface(if_ctx_t ctx)
3043 {
3044 	if_t ifp = iflib_get_ifp(ctx);
3045 	struct e1000_softc *sc = iflib_get_softc(ctx);
3046 	if_softc_ctx_t scctx = sc->shared;
3047 
3048 	INIT_DEBUGOUT("em_setup_interface: begin");
3049 
3050 	/* Single Queue */
3051 	if (sc->tx_num_queues == 1) {
3052 		if_setsendqlen(ifp, scctx->isc_ntxd[0] - 1);
3053 		if_setsendqready(ifp);
3054 	}
3055 
3056 	/*
3057 	 * Specify the media types supported by this adapter and register
3058 	 * callbacks to update media and link information
3059 	 */
3060 	if (sc->hw.phy.media_type == e1000_media_type_fiber ||
3061 	    sc->hw.phy.media_type == e1000_media_type_internal_serdes) {
3062 		u_char fiber_type = IFM_1000_SX;	/* default type */
3063 
3064 		if (sc->hw.mac.type == e1000_82545)
3065 			fiber_type = IFM_1000_LX;
3066 		ifmedia_add(sc->media, IFM_ETHER | fiber_type | IFM_FDX, 0, NULL);
3067 		ifmedia_add(sc->media, IFM_ETHER | fiber_type, 0, NULL);
3068 	} else {
3069 		ifmedia_add(sc->media, IFM_ETHER | IFM_10_T, 0, NULL);
3070 		ifmedia_add(sc->media, IFM_ETHER | IFM_10_T | IFM_FDX, 0, NULL);
3071 		ifmedia_add(sc->media, IFM_ETHER | IFM_100_TX, 0, NULL);
3072 		ifmedia_add(sc->media, IFM_ETHER | IFM_100_TX | IFM_FDX, 0, NULL);
3073 		if (sc->hw.phy.type != e1000_phy_ife) {
3074 			ifmedia_add(sc->media, IFM_ETHER | IFM_1000_T | IFM_FDX, 0, NULL);
3075 			ifmedia_add(sc->media, IFM_ETHER | IFM_1000_T, 0, NULL);
3076 		}
3077 	}
3078 	ifmedia_add(sc->media, IFM_ETHER | IFM_AUTO, 0, NULL);
3079 	ifmedia_set(sc->media, IFM_ETHER | IFM_AUTO);
3080 	return (0);
3081 }
3082 
3083 static int
3084 em_if_tx_queues_alloc(if_ctx_t ctx, caddr_t *vaddrs, uint64_t *paddrs, int ntxqs, int ntxqsets)
3085 {
3086 	struct e1000_softc *sc = iflib_get_softc(ctx);
3087 	if_softc_ctx_t scctx = sc->shared;
3088 	int error = E1000_SUCCESS;
3089 	struct em_tx_queue *que;
3090 	int i, j;
3091 
3092 	MPASS(sc->tx_num_queues > 0);
3093 	MPASS(sc->tx_num_queues == ntxqsets);
3094 
3095 	/* First allocate the top level queue structs */
3096 	if (!(sc->tx_queues =
3097 	    (struct em_tx_queue *) malloc(sizeof(struct em_tx_queue) *
3098 	    sc->tx_num_queues, M_DEVBUF, M_NOWAIT | M_ZERO))) {
3099 		device_printf(iflib_get_dev(ctx), "Unable to allocate queue memory\n");
3100 		return(ENOMEM);
3101 	}
3102 
3103 	for (i = 0, que = sc->tx_queues; i < sc->tx_num_queues; i++, que++) {
3104 		/* Set up some basics */
3105 
3106 		struct tx_ring *txr = &que->txr;
3107 		txr->sc = que->sc = sc;
3108 		que->me = txr->me =  i;
3109 
3110 		/* Allocate report status array */
3111 		if (!(txr->tx_rsq = (qidx_t *) malloc(sizeof(qidx_t) * scctx->isc_ntxd[0], M_DEVBUF, M_NOWAIT | M_ZERO))) {
3112 			device_printf(iflib_get_dev(ctx), "failed to allocate rs_idxs memory\n");
3113 			error = ENOMEM;
3114 			goto fail;
3115 		}
3116 		for (j = 0; j < scctx->isc_ntxd[0]; j++)
3117 			txr->tx_rsq[j] = QIDX_INVALID;
3118 		/* get the virtual and physical address of the hardware queues */
3119 		txr->tx_base = (struct e1000_tx_desc *)vaddrs[i*ntxqs];
3120 		txr->tx_paddr = paddrs[i*ntxqs];
3121 	}
3122 
3123 	if (bootverbose)
3124 		device_printf(iflib_get_dev(ctx),
3125 		    "allocated for %d tx_queues\n", sc->tx_num_queues);
3126 	return (0);
3127 fail:
3128 	em_if_queues_free(ctx);
3129 	return (error);
3130 }
3131 
3132 static int
3133 em_if_rx_queues_alloc(if_ctx_t ctx, caddr_t *vaddrs, uint64_t *paddrs, int nrxqs, int nrxqsets)
3134 {
3135 	struct e1000_softc *sc = iflib_get_softc(ctx);
3136 	int error = E1000_SUCCESS;
3137 	struct em_rx_queue *que;
3138 	int i;
3139 
3140 	MPASS(sc->rx_num_queues > 0);
3141 	MPASS(sc->rx_num_queues == nrxqsets);
3142 
3143 	/* First allocate the top level queue structs */
3144 	if (!(sc->rx_queues =
3145 	    (struct em_rx_queue *) malloc(sizeof(struct em_rx_queue) *
3146 	    sc->rx_num_queues, M_DEVBUF, M_NOWAIT | M_ZERO))) {
3147 		device_printf(iflib_get_dev(ctx), "Unable to allocate queue memory\n");
3148 		error = ENOMEM;
3149 		goto fail;
3150 	}
3151 
3152 	for (i = 0, que = sc->rx_queues; i < nrxqsets; i++, que++) {
3153 		/* Set up some basics */
3154 		struct rx_ring *rxr = &que->rxr;
3155 		rxr->sc = que->sc = sc;
3156 		rxr->que = que;
3157 		que->me = rxr->me =  i;
3158 
3159 		/* get the virtual and physical address of the hardware queues */
3160 		rxr->rx_base = (union e1000_rx_desc_extended *)vaddrs[i*nrxqs];
3161 		rxr->rx_paddr = paddrs[i*nrxqs];
3162 	}
3163 
3164 	if (bootverbose)
3165 		device_printf(iflib_get_dev(ctx),
3166 		    "allocated for %d rx_queues\n", sc->rx_num_queues);
3167 
3168 	return (0);
3169 fail:
3170 	em_if_queues_free(ctx);
3171 	return (error);
3172 }
3173 
3174 static void
3175 em_if_queues_free(if_ctx_t ctx)
3176 {
3177 	struct e1000_softc *sc = iflib_get_softc(ctx);
3178 	struct em_tx_queue *tx_que = sc->tx_queues;
3179 	struct em_rx_queue *rx_que = sc->rx_queues;
3180 
3181 	if (tx_que != NULL) {
3182 		for (int i = 0; i < sc->tx_num_queues; i++, tx_que++) {
3183 			struct tx_ring *txr = &tx_que->txr;
3184 			if (txr->tx_rsq == NULL)
3185 				break;
3186 
3187 			free(txr->tx_rsq, M_DEVBUF);
3188 			txr->tx_rsq = NULL;
3189 		}
3190 		free(sc->tx_queues, M_DEVBUF);
3191 		sc->tx_queues = NULL;
3192 	}
3193 
3194 	if (rx_que != NULL) {
3195 		free(sc->rx_queues, M_DEVBUF);
3196 		sc->rx_queues = NULL;
3197 	}
3198 }
3199 
3200 /*********************************************************************
3201  *
3202  *  Enable transmit unit.
3203  *
3204  **********************************************************************/
3205 static void
3206 em_initialize_transmit_unit(if_ctx_t ctx)
3207 {
3208 	struct e1000_softc *sc = iflib_get_softc(ctx);
3209 	if_softc_ctx_t scctx = sc->shared;
3210 	struct em_tx_queue *que;
3211 	struct tx_ring	*txr;
3212 	struct e1000_hw	*hw = &sc->hw;
3213 	u32 tctl, txdctl = 0, tarc, tipg = 0;
3214 
3215 	INIT_DEBUGOUT("em_initialize_transmit_unit: begin");
3216 
3217 	for (int i = 0; i < sc->tx_num_queues; i++, txr++) {
3218 		u64 bus_addr;
3219 		caddr_t offp, endp;
3220 
3221 		que = &sc->tx_queues[i];
3222 		txr = &que->txr;
3223 		bus_addr = txr->tx_paddr;
3224 
3225 		/* Clear checksum offload context. */
3226 		offp = (caddr_t)&txr->csum_flags;
3227 		endp = (caddr_t)(txr + 1);
3228 		bzero(offp, endp - offp);
3229 
3230 		/* Base and Len of TX Ring */
3231 		E1000_WRITE_REG(hw, E1000_TDLEN(i),
3232 		    scctx->isc_ntxd[0] * sizeof(struct e1000_tx_desc));
3233 		E1000_WRITE_REG(hw, E1000_TDBAH(i),
3234 		    (u32)(bus_addr >> 32));
3235 		E1000_WRITE_REG(hw, E1000_TDBAL(i),
3236 		    (u32)bus_addr);
3237 		/* Init the HEAD/TAIL indices */
3238 		E1000_WRITE_REG(hw, E1000_TDT(i), 0);
3239 		E1000_WRITE_REG(hw, E1000_TDH(i), 0);
3240 
3241 		HW_DEBUGOUT2("Base = %x, Length = %x\n",
3242 		    E1000_READ_REG(hw, E1000_TDBAL(i)),
3243 		    E1000_READ_REG(hw, E1000_TDLEN(i)));
3244 
3245 		txdctl = 0; /* clear txdctl */
3246 		txdctl |= 0x1f; /* PTHRESH */
3247 		txdctl |= 1 << 8; /* HTHRESH */
3248 		txdctl |= 1 << 16;/* WTHRESH */
3249 		txdctl |= 1 << 22; /* Reserved bit 22 must always be 1 */
3250 		txdctl |= E1000_TXDCTL_GRAN;
3251 		txdctl |= 1 << 25; /* LWTHRESH */
3252 
3253 		E1000_WRITE_REG(hw, E1000_TXDCTL(i), txdctl);
3254 	}
3255 
3256 	/* Set the default values for the Tx Inter Packet Gap timer */
3257 	switch (hw->mac.type) {
3258 	case e1000_80003es2lan:
3259 		tipg = DEFAULT_82543_TIPG_IPGR1;
3260 		tipg |= DEFAULT_80003ES2LAN_TIPG_IPGR2 <<
3261 		    E1000_TIPG_IPGR2_SHIFT;
3262 		break;
3263 	case e1000_82542:
3264 		tipg = DEFAULT_82542_TIPG_IPGT;
3265 		tipg |= DEFAULT_82542_TIPG_IPGR1 << E1000_TIPG_IPGR1_SHIFT;
3266 		tipg |= DEFAULT_82542_TIPG_IPGR2 << E1000_TIPG_IPGR2_SHIFT;
3267 		break;
3268 	default:
3269 		if (hw->phy.media_type == e1000_media_type_fiber ||
3270 		    hw->phy.media_type == e1000_media_type_internal_serdes)
3271 			tipg = DEFAULT_82543_TIPG_IPGT_FIBER;
3272 		else
3273 			tipg = DEFAULT_82543_TIPG_IPGT_COPPER;
3274 		tipg |= DEFAULT_82543_TIPG_IPGR1 << E1000_TIPG_IPGR1_SHIFT;
3275 		tipg |= DEFAULT_82543_TIPG_IPGR2 << E1000_TIPG_IPGR2_SHIFT;
3276 	}
3277 
3278 	E1000_WRITE_REG(hw, E1000_TIPG, tipg);
3279 	E1000_WRITE_REG(hw, E1000_TIDV, sc->tx_int_delay.value);
3280 
3281 	if(hw->mac.type >= e1000_82540)
3282 		E1000_WRITE_REG(hw, E1000_TADV,
3283 		    sc->tx_abs_int_delay.value);
3284 
3285 	if (hw->mac.type == e1000_82571 || hw->mac.type == e1000_82572) {
3286 		tarc = E1000_READ_REG(hw, E1000_TARC(0));
3287 		tarc |= TARC_SPEED_MODE_BIT;
3288 		E1000_WRITE_REG(hw, E1000_TARC(0), tarc);
3289 	} else if (hw->mac.type == e1000_80003es2lan) {
3290 		/* errata: program both queues to unweighted RR */
3291 		tarc = E1000_READ_REG(hw, E1000_TARC(0));
3292 		tarc |= 1;
3293 		E1000_WRITE_REG(hw, E1000_TARC(0), tarc);
3294 		tarc = E1000_READ_REG(hw, E1000_TARC(1));
3295 		tarc |= 1;
3296 		E1000_WRITE_REG(hw, E1000_TARC(1), tarc);
3297 	} else if (hw->mac.type == e1000_82574) {
3298 		tarc = E1000_READ_REG(hw, E1000_TARC(0));
3299 		tarc |= TARC_ERRATA_BIT;
3300 		if ( sc->tx_num_queues > 1) {
3301 			tarc |= (TARC_COMPENSATION_MODE | TARC_MQ_FIX);
3302 			E1000_WRITE_REG(hw, E1000_TARC(0), tarc);
3303 			E1000_WRITE_REG(hw, E1000_TARC(1), tarc);
3304 		} else
3305 			E1000_WRITE_REG(hw, E1000_TARC(0), tarc);
3306 	}
3307 
3308 	if (sc->tx_int_delay.value > 0)
3309 		sc->txd_cmd |= E1000_TXD_CMD_IDE;
3310 
3311 	/* Program the Transmit Control Register */
3312 	tctl = E1000_READ_REG(hw, E1000_TCTL);
3313 	tctl &= ~E1000_TCTL_CT;
3314 	tctl |= (E1000_TCTL_PSP | E1000_TCTL_RTLC | E1000_TCTL_EN |
3315 		   (E1000_COLLISION_THRESHOLD << E1000_CT_SHIFT));
3316 
3317 	if (hw->mac.type >= e1000_82571)
3318 		tctl |= E1000_TCTL_MULR;
3319 
3320 	/* This write will effectively turn on the transmit unit. */
3321 	E1000_WRITE_REG(hw, E1000_TCTL, tctl);
3322 
3323 	/* SPT and KBL errata workarounds */
3324 	if (hw->mac.type == e1000_pch_spt) {
3325 		u32 reg;
3326 		reg = E1000_READ_REG(hw, E1000_IOSFPC);
3327 		reg |= E1000_RCTL_RDMTS_HEX;
3328 		E1000_WRITE_REG(hw, E1000_IOSFPC, reg);
3329 		/* i218-i219 Specification Update 1.5.4.5 */
3330 		reg = E1000_READ_REG(hw, E1000_TARC(0));
3331 		reg &= ~E1000_TARC0_CB_MULTIQ_3_REQ;
3332 		reg |= E1000_TARC0_CB_MULTIQ_2_REQ;
3333 		E1000_WRITE_REG(hw, E1000_TARC(0), reg);
3334 	}
3335 }
3336 
3337 /*********************************************************************
3338  *
3339  *  Enable receive unit.
3340  *
3341  **********************************************************************/
3342 #define BSIZEPKT_ROUNDUP ((1<<E1000_SRRCTL_BSIZEPKT_SHIFT)-1)
3343 
3344 static void
3345 em_initialize_receive_unit(if_ctx_t ctx)
3346 {
3347 	struct e1000_softc *sc = iflib_get_softc(ctx);
3348 	if_softc_ctx_t scctx = sc->shared;
3349 	if_t ifp = iflib_get_ifp(ctx);
3350 	struct e1000_hw	*hw = &sc->hw;
3351 	struct em_rx_queue *que;
3352 	int i;
3353 	uint32_t rctl, rxcsum;
3354 
3355 	INIT_DEBUGOUT("em_initialize_receive_units: begin");
3356 
3357 	/*
3358 	 * Make sure receives are disabled while setting
3359 	 * up the descriptor ring
3360 	 */
3361 	rctl = E1000_READ_REG(hw, E1000_RCTL);
3362 	/* Do not disable if ever enabled on this hardware */
3363 	if ((hw->mac.type != e1000_82574) && (hw->mac.type != e1000_82583))
3364 		E1000_WRITE_REG(hw, E1000_RCTL, rctl & ~E1000_RCTL_EN);
3365 
3366 	/* Setup the Receive Control Register */
3367 	rctl &= ~(3 << E1000_RCTL_MO_SHIFT);
3368 	rctl |= E1000_RCTL_EN | E1000_RCTL_BAM |
3369 	    E1000_RCTL_LBM_NO | E1000_RCTL_RDMTS_HALF |
3370 	    (hw->mac.mc_filter_type << E1000_RCTL_MO_SHIFT);
3371 
3372 	/* Do not store bad packets */
3373 	rctl &= ~E1000_RCTL_SBP;
3374 
3375 	/* Enable Long Packet receive */
3376 	if (if_getmtu(ifp) > ETHERMTU)
3377 		rctl |= E1000_RCTL_LPE;
3378 	else
3379 		rctl &= ~E1000_RCTL_LPE;
3380 
3381 	/* Strip the CRC */
3382 	if (!em_disable_crc_stripping)
3383 		rctl |= E1000_RCTL_SECRC;
3384 
3385 	if (hw->mac.type >= e1000_82540) {
3386 		E1000_WRITE_REG(hw, E1000_RADV,
3387 		    sc->rx_abs_int_delay.value);
3388 
3389 		/*
3390 		 * Set the interrupt throttling rate. Value is calculated
3391 		 * as DEFAULT_ITR = 1/(MAX_INTS_PER_SEC * 256ns)
3392 		 */
3393 		E1000_WRITE_REG(hw, E1000_ITR, DEFAULT_ITR);
3394 	}
3395 	E1000_WRITE_REG(hw, E1000_RDTR, sc->rx_int_delay.value);
3396 
3397 	if (hw->mac.type >= em_mac_min) {
3398 		uint32_t rfctl;
3399 		/* Use extended rx descriptor formats */
3400 		rfctl = E1000_READ_REG(hw, E1000_RFCTL);
3401 		rfctl |= E1000_RFCTL_EXTEN;
3402 
3403 		/*
3404 		 * When using MSI-X interrupts we need to throttle
3405 		 * using the EITR register (82574 only)
3406 		 */
3407 		if (hw->mac.type == e1000_82574) {
3408 			for (int i = 0; i < 4; i++)
3409 				E1000_WRITE_REG(hw, E1000_EITR_82574(i),
3410 				    DEFAULT_ITR);
3411 			/* Disable accelerated acknowledge */
3412 			rfctl |= E1000_RFCTL_ACK_DIS;
3413 		}
3414 		E1000_WRITE_REG(hw, E1000_RFCTL, rfctl);
3415 	}
3416 
3417 	/* Set up L3 and L4 csum Rx descriptor offloads */
3418 	rxcsum = E1000_READ_REG(hw, E1000_RXCSUM);
3419 	if (if_getcapenable(ifp) & IFCAP_RXCSUM) {
3420 		rxcsum |= E1000_RXCSUM_TUOFL | E1000_RXCSUM_IPOFL;
3421 		if (hw->mac.type > e1000_82575)
3422 			rxcsum |= E1000_RXCSUM_CRCOFL;
3423 		else if (hw->mac.type < em_mac_min &&
3424 		    if_getcapenable(ifp) & IFCAP_HWCSUM_IPV6)
3425 			rxcsum |= E1000_RXCSUM_IPV6OFL;
3426 	} else {
3427 		rxcsum &= ~(E1000_RXCSUM_IPOFL | E1000_RXCSUM_TUOFL);
3428 		if (hw->mac.type > e1000_82575)
3429 			rxcsum &= ~E1000_RXCSUM_CRCOFL;
3430 		else if (hw->mac.type < em_mac_min)
3431 			rxcsum &= ~E1000_RXCSUM_IPV6OFL;
3432 	}
3433 
3434 	if (sc->rx_num_queues > 1) {
3435 		/* RSS hash needed in the Rx descriptor */
3436 		rxcsum |= E1000_RXCSUM_PCSD;
3437 
3438 		if (hw->mac.type >= igb_mac_min)
3439 			igb_initialize_rss_mapping(sc);
3440 		else
3441 			em_initialize_rss_mapping(sc);
3442 	}
3443 	E1000_WRITE_REG(hw, E1000_RXCSUM, rxcsum);
3444 
3445 	/*
3446 	 * XXX TEMPORARY WORKAROUND: on some systems with 82573
3447 	 * long latencies are observed, like Lenovo X60. This
3448 	 * change eliminates the problem, but since having positive
3449 	 * values in RDTR is a known source of problems on other
3450 	 * platforms another solution is being sought.
3451 	 */
3452 	if (hw->mac.type == e1000_82573)
3453 		E1000_WRITE_REG(hw, E1000_RDTR, 0x20);
3454 
3455 	for (i = 0, que = sc->rx_queues; i < sc->rx_num_queues; i++, que++) {
3456 		struct rx_ring *rxr = &que->rxr;
3457 		/* Setup the Base and Length of the Rx Descriptor Ring */
3458 		u64 bus_addr = rxr->rx_paddr;
3459 #if 0
3460 		u32 rdt = sc->rx_num_queues -1;  /* default */
3461 #endif
3462 
3463 		E1000_WRITE_REG(hw, E1000_RDLEN(i),
3464 		    scctx->isc_nrxd[0] * sizeof(union e1000_rx_desc_extended));
3465 		E1000_WRITE_REG(hw, E1000_RDBAH(i), (u32)(bus_addr >> 32));
3466 		E1000_WRITE_REG(hw, E1000_RDBAL(i), (u32)bus_addr);
3467 		/* Setup the Head and Tail Descriptor Pointers */
3468 		E1000_WRITE_REG(hw, E1000_RDH(i), 0);
3469 		E1000_WRITE_REG(hw, E1000_RDT(i), 0);
3470 	}
3471 
3472 	/*
3473 	 * Set PTHRESH for improved jumbo performance
3474 	 * According to 10.2.5.11 of Intel 82574 Datasheet,
3475 	 * RXDCTL(1) is written whenever RXDCTL(0) is written.
3476 	 * Only write to RXDCTL(1) if there is a need for different
3477 	 * settings.
3478 	 */
3479 	if ((hw->mac.type == e1000_ich9lan || hw->mac.type == e1000_pch2lan ||
3480 	    hw->mac.type == e1000_ich10lan) && if_getmtu(ifp) > ETHERMTU) {
3481 		u32 rxdctl = E1000_READ_REG(hw, E1000_RXDCTL(0));
3482 		E1000_WRITE_REG(hw, E1000_RXDCTL(0), rxdctl | 3);
3483 	} else if (hw->mac.type == e1000_82574) {
3484 		for (int i = 0; i < sc->rx_num_queues; i++) {
3485 			u32 rxdctl = E1000_READ_REG(hw, E1000_RXDCTL(i));
3486 			rxdctl |= 0x20; /* PTHRESH */
3487 			rxdctl |= 4 << 8; /* HTHRESH */
3488 			rxdctl |= 4 << 16;/* WTHRESH */
3489 			rxdctl |= 1 << 24; /* Switch to granularity */
3490 			E1000_WRITE_REG(hw, E1000_RXDCTL(i), rxdctl);
3491 		}
3492 	} else if (hw->mac.type >= igb_mac_min) {
3493 		u32 psize, srrctl = 0;
3494 
3495 		if (if_getmtu(ifp) > ETHERMTU) {
3496 			psize = scctx->isc_max_frame_size;
3497 			/* are we on a vlan? */
3498 			if (if_vlantrunkinuse(ifp))
3499 				psize += VLAN_TAG_SIZE;
3500 
3501 			if (sc->vf_ifp)
3502 				e1000_rlpml_set_vf(hw, psize);
3503 			else
3504 				E1000_WRITE_REG(hw, E1000_RLPML, psize);
3505 		}
3506 
3507 		/* Set maximum packet buffer len */
3508 		srrctl |= (sc->rx_mbuf_sz + BSIZEPKT_ROUNDUP) >>
3509 		    E1000_SRRCTL_BSIZEPKT_SHIFT;
3510 
3511 		/*
3512 		 * If TX flow control is disabled and there's >1 queue defined,
3513 		 * enable DROP.
3514 		 *
3515 		 * This drops frames rather than hanging the RX MAC for all queues.
3516 		 */
3517 		if ((sc->rx_num_queues > 1) &&
3518 		    (sc->fc == e1000_fc_none ||
3519 		     sc->fc == e1000_fc_rx_pause)) {
3520 			srrctl |= E1000_SRRCTL_DROP_EN;
3521 		}
3522 			/* Setup the Base and Length of the Rx Descriptor Rings */
3523 		for (i = 0, que = sc->rx_queues; i < sc->rx_num_queues; i++, que++) {
3524 			struct rx_ring *rxr = &que->rxr;
3525 			u64 bus_addr = rxr->rx_paddr;
3526 			u32 rxdctl;
3527 
3528 #ifdef notyet
3529 			/* Configure for header split? -- ignore for now */
3530 			rxr->hdr_split = igb_header_split;
3531 #else
3532 			srrctl |= E1000_SRRCTL_DESCTYPE_ADV_ONEBUF;
3533 #endif
3534 
3535 			E1000_WRITE_REG(hw, E1000_RDLEN(i),
3536 					scctx->isc_nrxd[0] * sizeof(struct e1000_rx_desc));
3537 			E1000_WRITE_REG(hw, E1000_RDBAH(i),
3538 					(uint32_t)(bus_addr >> 32));
3539 			E1000_WRITE_REG(hw, E1000_RDBAL(i),
3540 					(uint32_t)bus_addr);
3541 			E1000_WRITE_REG(hw, E1000_SRRCTL(i), srrctl);
3542 			/* Enable this Queue */
3543 			rxdctl = E1000_READ_REG(hw, E1000_RXDCTL(i));
3544 			rxdctl |= E1000_RXDCTL_QUEUE_ENABLE;
3545 			rxdctl &= 0xFFF00000;
3546 			rxdctl |= IGB_RX_PTHRESH;
3547 			rxdctl |= IGB_RX_HTHRESH << 8;
3548 			rxdctl |= IGB_RX_WTHRESH << 16;
3549 			E1000_WRITE_REG(hw, E1000_RXDCTL(i), rxdctl);
3550 		}
3551 	} else if (hw->mac.type >= e1000_pch2lan) {
3552 		if (if_getmtu(ifp) > ETHERMTU)
3553 			e1000_lv_jumbo_workaround_ich8lan(hw, true);
3554 		else
3555 			e1000_lv_jumbo_workaround_ich8lan(hw, false);
3556 	}
3557 
3558 	/* Make sure VLAN Filters are off */
3559 	rctl &= ~E1000_RCTL_VFE;
3560 
3561 	/* Set up packet buffer size, overridden by per queue srrctl on igb */
3562 	if (hw->mac.type < igb_mac_min) {
3563 		if (sc->rx_mbuf_sz > 2048 && sc->rx_mbuf_sz <= 4096)
3564 			rctl |= E1000_RCTL_SZ_4096 | E1000_RCTL_BSEX;
3565 		else if (sc->rx_mbuf_sz > 4096 && sc->rx_mbuf_sz <= 8192)
3566 			rctl |= E1000_RCTL_SZ_8192 | E1000_RCTL_BSEX;
3567 		else if (sc->rx_mbuf_sz > 8192)
3568 			rctl |= E1000_RCTL_SZ_16384 | E1000_RCTL_BSEX;
3569 		else {
3570 			rctl |= E1000_RCTL_SZ_2048;
3571 			rctl &= ~E1000_RCTL_BSEX;
3572 		}
3573 	} else
3574 		rctl |= E1000_RCTL_SZ_2048;
3575 
3576 	/*
3577 	 * rctl bits 11:10 are as follows
3578 	 * lem: reserved
3579 	 * em: DTYPE
3580 	 * igb: reserved
3581 	 * and should be 00 on all of the above
3582 	 */
3583 	rctl &= ~0x00000C00;
3584 
3585 	/* Write out the settings */
3586 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
3587 
3588 	return;
3589 }
3590 
3591 static void
3592 em_if_vlan_register(if_ctx_t ctx, u16 vtag)
3593 {
3594 	struct e1000_softc *sc = iflib_get_softc(ctx);
3595 	u32 index, bit;
3596 
3597 	index = (vtag >> 5) & 0x7F;
3598 	bit = vtag & 0x1F;
3599 	sc->shadow_vfta[index] |= (1 << bit);
3600 	++sc->num_vlans;
3601 	em_if_vlan_filter_write(sc);
3602 }
3603 
3604 static void
3605 em_if_vlan_unregister(if_ctx_t ctx, u16 vtag)
3606 {
3607 	struct e1000_softc *sc = iflib_get_softc(ctx);
3608 	u32 index, bit;
3609 
3610 	index = (vtag >> 5) & 0x7F;
3611 	bit = vtag & 0x1F;
3612 	sc->shadow_vfta[index] &= ~(1 << bit);
3613 	--sc->num_vlans;
3614 	em_if_vlan_filter_write(sc);
3615 }
3616 
3617 static bool
3618 em_if_vlan_filter_capable(if_ctx_t ctx)
3619 {
3620 	if_t ifp = iflib_get_ifp(ctx);
3621 
3622 	if ((if_getcapenable(ifp) & IFCAP_VLAN_HWFILTER) &&
3623 	    !em_disable_crc_stripping)
3624 		return (true);
3625 
3626 	return (false);
3627 }
3628 
3629 static bool
3630 em_if_vlan_filter_used(if_ctx_t ctx)
3631 {
3632 	struct e1000_softc *sc = iflib_get_softc(ctx);
3633 
3634 	if (!em_if_vlan_filter_capable(ctx))
3635 		return (false);
3636 
3637 	for (int i = 0; i < EM_VFTA_SIZE; i++)
3638 		if (sc->shadow_vfta[i] != 0)
3639 			return (true);
3640 
3641 	return (false);
3642 }
3643 
3644 static void
3645 em_if_vlan_filter_enable(struct e1000_softc *sc)
3646 {
3647 	struct e1000_hw *hw = &sc->hw;
3648 	u32 reg;
3649 
3650 	reg = E1000_READ_REG(hw, E1000_RCTL);
3651 	reg &= ~E1000_RCTL_CFIEN;
3652 	reg |= E1000_RCTL_VFE;
3653 	E1000_WRITE_REG(hw, E1000_RCTL, reg);
3654 }
3655 
3656 static void
3657 em_if_vlan_filter_disable(struct e1000_softc *sc)
3658 {
3659 	struct e1000_hw *hw = &sc->hw;
3660 	u32 reg;
3661 
3662 	reg = E1000_READ_REG(hw, E1000_RCTL);
3663 	reg &= ~(E1000_RCTL_VFE | E1000_RCTL_CFIEN);
3664 	E1000_WRITE_REG(hw, E1000_RCTL, reg);
3665 }
3666 
3667 static void
3668 em_if_vlan_filter_write(struct e1000_softc *sc)
3669 {
3670 	struct e1000_hw *hw = &sc->hw;
3671 
3672 	if (sc->vf_ifp)
3673 		return;
3674 
3675 	/* Disable interrupts for lem-class devices during the filter change */
3676 	if (hw->mac.type < em_mac_min)
3677 		em_if_intr_disable(sc->ctx);
3678 
3679 	for (int i = 0; i < EM_VFTA_SIZE; i++)
3680 		if (sc->shadow_vfta[i] != 0) {
3681 			/* XXXKB: incomplete VF support, we return early above */
3682 			if (sc->vf_ifp)
3683 				e1000_vfta_set_vf(hw, sc->shadow_vfta[i], true);
3684 			else
3685 				e1000_write_vfta(hw, i, sc->shadow_vfta[i]);
3686 		}
3687 
3688 	/* Re-enable interrupts for lem-class devices */
3689 	if (hw->mac.type < em_mac_min)
3690 		em_if_intr_enable(sc->ctx);
3691 }
3692 
3693 static void
3694 em_setup_vlan_hw_support(if_ctx_t ctx)
3695 {
3696 	struct e1000_softc *sc = iflib_get_softc(ctx);
3697 	struct e1000_hw *hw = &sc->hw;
3698 	if_t ifp = iflib_get_ifp(ctx);
3699 	u32 reg;
3700 
3701 	/* XXXKB: Return early if we are a VF until VF decap and filter management
3702 	 * is ready and tested.
3703 	 */
3704 	if (sc->vf_ifp)
3705 		return;
3706 
3707 	if (if_getcapenable(ifp) & IFCAP_VLAN_HWTAGGING &&
3708 	    !em_disable_crc_stripping) {
3709 		reg = E1000_READ_REG(hw, E1000_CTRL);
3710 		reg |= E1000_CTRL_VME;
3711 		E1000_WRITE_REG(hw, E1000_CTRL, reg);
3712 	} else {
3713 		reg = E1000_READ_REG(hw, E1000_CTRL);
3714 		reg &= ~E1000_CTRL_VME;
3715 		E1000_WRITE_REG(hw, E1000_CTRL, reg);
3716 	}
3717 
3718 	/* If we aren't doing HW filtering, we're done */
3719 	if (!em_if_vlan_filter_capable(ctx))  {
3720 		em_if_vlan_filter_disable(sc);
3721 		return;
3722 	}
3723 
3724 	/*
3725 	 * A soft reset zero's out the VFTA, so
3726 	 * we need to repopulate it now.
3727 	 * We also insert VLAN 0 in the filter list, so we pass VLAN 0 tagged
3728 	 * traffic through. This will write the entire table.
3729 	 */
3730 	em_if_vlan_register(ctx, 0);
3731 
3732 	/* Enable the Filter Table */
3733 	em_if_vlan_filter_enable(sc);
3734 }
3735 
3736 static void
3737 em_if_intr_enable(if_ctx_t ctx)
3738 {
3739 	struct e1000_softc *sc = iflib_get_softc(ctx);
3740 	struct e1000_hw *hw = &sc->hw;
3741 	u32 ims_mask = IMS_ENABLE_MASK;
3742 
3743 	if (sc->intr_type == IFLIB_INTR_MSIX) {
3744 		E1000_WRITE_REG(hw, EM_EIAC, sc->ims);
3745 		ims_mask |= sc->ims;
3746 	}
3747 	E1000_WRITE_REG(hw, E1000_IMS, ims_mask);
3748 	E1000_WRITE_FLUSH(hw);
3749 }
3750 
3751 static void
3752 em_if_intr_disable(if_ctx_t ctx)
3753 {
3754 	struct e1000_softc *sc = iflib_get_softc(ctx);
3755 	struct e1000_hw *hw = &sc->hw;
3756 
3757 	if (sc->intr_type == IFLIB_INTR_MSIX)
3758 		E1000_WRITE_REG(hw, EM_EIAC, 0);
3759 	E1000_WRITE_REG(hw, E1000_IMC, 0xffffffff);
3760 	E1000_WRITE_FLUSH(hw);
3761 }
3762 
3763 static void
3764 igb_if_intr_enable(if_ctx_t ctx)
3765 {
3766 	struct e1000_softc *sc = iflib_get_softc(ctx);
3767 	struct e1000_hw *hw = &sc->hw;
3768 	u32 mask;
3769 
3770 	if (__predict_true(sc->intr_type == IFLIB_INTR_MSIX)) {
3771 		mask = (sc->que_mask | sc->link_mask);
3772 		E1000_WRITE_REG(hw, E1000_EIAC, mask);
3773 		E1000_WRITE_REG(hw, E1000_EIAM, mask);
3774 		E1000_WRITE_REG(hw, E1000_EIMS, mask);
3775 		E1000_WRITE_REG(hw, E1000_IMS, E1000_IMS_LSC);
3776 	} else
3777 		E1000_WRITE_REG(hw, E1000_IMS, IMS_ENABLE_MASK);
3778 	E1000_WRITE_FLUSH(hw);
3779 }
3780 
3781 static void
3782 igb_if_intr_disable(if_ctx_t ctx)
3783 {
3784 	struct e1000_softc *sc = iflib_get_softc(ctx);
3785 	struct e1000_hw *hw = &sc->hw;
3786 
3787 	if (__predict_true(sc->intr_type == IFLIB_INTR_MSIX)) {
3788 		E1000_WRITE_REG(hw, E1000_EIMC, 0xffffffff);
3789 		E1000_WRITE_REG(hw, E1000_EIAC, 0);
3790 	}
3791 	E1000_WRITE_REG(hw, E1000_IMC, 0xffffffff);
3792 	E1000_WRITE_FLUSH(hw);
3793 }
3794 
3795 /*
3796  * Bit of a misnomer, what this really means is
3797  * to enable OS management of the system... aka
3798  * to disable special hardware management features
3799  */
3800 static void
3801 em_init_manageability(struct e1000_softc *sc)
3802 {
3803 	/* A shared code workaround */
3804 #define E1000_82542_MANC2H E1000_MANC2H
3805 	if (sc->has_manage) {
3806 		int manc2h = E1000_READ_REG(&sc->hw, E1000_MANC2H);
3807 		int manc = E1000_READ_REG(&sc->hw, E1000_MANC);
3808 
3809 		/* disable hardware interception of ARP */
3810 		manc &= ~(E1000_MANC_ARP_EN);
3811 
3812 		/* enable receiving management packets to the host */
3813 		manc |= E1000_MANC_EN_MNG2HOST;
3814 #define E1000_MNG2HOST_PORT_623 (1 << 5)
3815 #define E1000_MNG2HOST_PORT_664 (1 << 6)
3816 		manc2h |= E1000_MNG2HOST_PORT_623;
3817 		manc2h |= E1000_MNG2HOST_PORT_664;
3818 		E1000_WRITE_REG(&sc->hw, E1000_MANC2H, manc2h);
3819 		E1000_WRITE_REG(&sc->hw, E1000_MANC, manc);
3820 	}
3821 }
3822 
3823 /*
3824  * Give control back to hardware management
3825  * controller if there is one.
3826  */
3827 static void
3828 em_release_manageability(struct e1000_softc *sc)
3829 {
3830 	if (sc->has_manage) {
3831 		int manc = E1000_READ_REG(&sc->hw, E1000_MANC);
3832 
3833 		/* re-enable hardware interception of ARP */
3834 		manc |= E1000_MANC_ARP_EN;
3835 		manc &= ~E1000_MANC_EN_MNG2HOST;
3836 
3837 		E1000_WRITE_REG(&sc->hw, E1000_MANC, manc);
3838 	}
3839 }
3840 
3841 /*
3842  * em_get_hw_control sets the {CTRL_EXT|FWSM}:DRV_LOAD bit.
3843  * For ASF and Pass Through versions of f/w this means
3844  * that the driver is loaded. For AMT version type f/w
3845  * this means that the network i/f is open.
3846  */
3847 static void
3848 em_get_hw_control(struct e1000_softc *sc)
3849 {
3850 	u32 ctrl_ext, swsm;
3851 
3852 	if (sc->vf_ifp)
3853 		return;
3854 
3855 	if (sc->hw.mac.type == e1000_82573) {
3856 		swsm = E1000_READ_REG(&sc->hw, E1000_SWSM);
3857 		E1000_WRITE_REG(&sc->hw, E1000_SWSM,
3858 		    swsm | E1000_SWSM_DRV_LOAD);
3859 		return;
3860 	}
3861 	/* else */
3862 	ctrl_ext = E1000_READ_REG(&sc->hw, E1000_CTRL_EXT);
3863 	E1000_WRITE_REG(&sc->hw, E1000_CTRL_EXT,
3864 	    ctrl_ext | E1000_CTRL_EXT_DRV_LOAD);
3865 }
3866 
3867 /*
3868  * em_release_hw_control resets {CTRL_EXT|FWSM}:DRV_LOAD bit.
3869  * For ASF and Pass Through versions of f/w this means that
3870  * the driver is no longer loaded. For AMT versions of the
3871  * f/w this means that the network i/f is closed.
3872  */
3873 static void
3874 em_release_hw_control(struct e1000_softc *sc)
3875 {
3876 	u32 ctrl_ext, swsm;
3877 
3878 	if (!sc->has_manage)
3879 		return;
3880 
3881 	if (sc->hw.mac.type == e1000_82573) {
3882 		swsm = E1000_READ_REG(&sc->hw, E1000_SWSM);
3883 		E1000_WRITE_REG(&sc->hw, E1000_SWSM,
3884 		    swsm & ~E1000_SWSM_DRV_LOAD);
3885 		return;
3886 	}
3887 	/* else */
3888 	ctrl_ext = E1000_READ_REG(&sc->hw, E1000_CTRL_EXT);
3889 	E1000_WRITE_REG(&sc->hw, E1000_CTRL_EXT,
3890 	    ctrl_ext & ~E1000_CTRL_EXT_DRV_LOAD);
3891 	return;
3892 }
3893 
3894 static int
3895 em_is_valid_ether_addr(u8 *addr)
3896 {
3897 	char zero_addr[6] = { 0, 0, 0, 0, 0, 0 };
3898 
3899 	if ((addr[0] & 1) || (!bcmp(addr, zero_addr, ETHER_ADDR_LEN))) {
3900 		return (false);
3901 	}
3902 
3903 	return (true);
3904 }
3905 
3906 static bool
3907 em_automask_tso(if_ctx_t ctx)
3908 {
3909 	struct e1000_softc *sc = iflib_get_softc(ctx);
3910 	if_softc_ctx_t scctx = iflib_get_softc_ctx(ctx);
3911 	if_t ifp = iflib_get_ifp(ctx);
3912 
3913 	if (!em_unsupported_tso && sc->link_speed &&
3914 	    sc->link_speed != SPEED_1000 && scctx->isc_capenable & IFCAP_TSO) {
3915 		device_printf(sc->dev, "Disabling TSO for 10/100 Ethernet.\n");
3916 		sc->tso_automasked = scctx->isc_capenable & IFCAP_TSO;
3917 		scctx->isc_capenable &= ~IFCAP_TSO;
3918 		if_setcapenablebit(ifp, 0, IFCAP_TSO);
3919 		/* iflib_init_locked handles ifnet hwassistbits */
3920 		iflib_request_reset(ctx);
3921 		return true;
3922 	} else if (sc->link_speed == SPEED_1000 && sc->tso_automasked) {
3923 		device_printf(sc->dev, "Re-enabling TSO for GbE.\n");
3924 		scctx->isc_capenable |= sc->tso_automasked;
3925 		if_setcapenablebit(ifp, sc->tso_automasked, 0);
3926 		sc->tso_automasked = 0;
3927 		/* iflib_init_locked handles ifnet hwassistbits */
3928 		iflib_request_reset(ctx);
3929 		return true;
3930 	}
3931 
3932 	return false;
3933 }
3934 
3935 /*
3936 ** Parse the interface capabilities with regard
3937 ** to both system management and wake-on-lan for
3938 ** later use.
3939 */
3940 static void
3941 em_get_wakeup(if_ctx_t ctx)
3942 {
3943 	struct e1000_softc *sc = iflib_get_softc(ctx);
3944 	device_t dev = iflib_get_dev(ctx);
3945 	u16 eeprom_data = 0, device_id, apme_mask;
3946 
3947 	sc->has_manage = e1000_enable_mng_pass_thru(&sc->hw);
3948 	apme_mask = EM_EEPROM_APME;
3949 
3950 	switch (sc->hw.mac.type) {
3951 	case e1000_82542:
3952 	case e1000_82543:
3953 		break;
3954 	case e1000_82544:
3955 		e1000_read_nvm(&sc->hw,
3956 		    NVM_INIT_CONTROL2_REG, 1, &eeprom_data);
3957 		apme_mask = EM_82544_APME;
3958 		break;
3959 	case e1000_82546:
3960 	case e1000_82546_rev_3:
3961 		if (sc->hw.bus.func == 1) {
3962 			e1000_read_nvm(&sc->hw,
3963 			    NVM_INIT_CONTROL3_PORT_B, 1, &eeprom_data);
3964 			break;
3965 		} else
3966 			e1000_read_nvm(&sc->hw,
3967 			    NVM_INIT_CONTROL3_PORT_A, 1, &eeprom_data);
3968 		break;
3969 	case e1000_82573:
3970 	case e1000_82583:
3971 		sc->has_amt = true;
3972 		/* FALLTHROUGH */
3973 	case e1000_82571:
3974 	case e1000_82572:
3975 	case e1000_80003es2lan:
3976 		if (sc->hw.bus.func == 1) {
3977 			e1000_read_nvm(&sc->hw,
3978 			    NVM_INIT_CONTROL3_PORT_B, 1, &eeprom_data);
3979 			break;
3980 		} else
3981 			e1000_read_nvm(&sc->hw,
3982 			    NVM_INIT_CONTROL3_PORT_A, 1, &eeprom_data);
3983 		break;
3984 	case e1000_ich8lan:
3985 	case e1000_ich9lan:
3986 	case e1000_ich10lan:
3987 	case e1000_pchlan:
3988 	case e1000_pch2lan:
3989 	case e1000_pch_lpt:
3990 	case e1000_pch_spt:
3991 	case e1000_82575:	/* listing all igb devices */
3992 	case e1000_82576:
3993 	case e1000_82580:
3994 	case e1000_i350:
3995 	case e1000_i354:
3996 	case e1000_i210:
3997 	case e1000_i211:
3998 	case e1000_vfadapt:
3999 	case e1000_vfadapt_i350:
4000 		apme_mask = E1000_WUC_APME;
4001 		sc->has_amt = true;
4002 		eeprom_data = E1000_READ_REG(&sc->hw, E1000_WUC);
4003 		break;
4004 	default:
4005 		e1000_read_nvm(&sc->hw,
4006 		    NVM_INIT_CONTROL3_PORT_A, 1, &eeprom_data);
4007 		break;
4008 	}
4009 	if (eeprom_data & apme_mask)
4010 		sc->wol = (E1000_WUFC_MAG | E1000_WUFC_MC);
4011 	/*
4012 	 * We have the eeprom settings, now apply the special cases
4013 	 * where the eeprom may be wrong or the board won't support
4014 	 * wake on lan on a particular port
4015 	 */
4016 	device_id = pci_get_device(dev);
4017 	switch (device_id) {
4018 	case E1000_DEV_ID_82546GB_PCIE:
4019 		sc->wol = 0;
4020 		break;
4021 	case E1000_DEV_ID_82546EB_FIBER:
4022 	case E1000_DEV_ID_82546GB_FIBER:
4023 		/* Wake events only supported on port A for dual fiber
4024 		 * regardless of eeprom setting */
4025 		if (E1000_READ_REG(&sc->hw, E1000_STATUS) &
4026 		    E1000_STATUS_FUNC_1)
4027 			sc->wol = 0;
4028 		break;
4029 	case E1000_DEV_ID_82546GB_QUAD_COPPER_KSP3:
4030 		/* if quad port adapter, disable WoL on all but port A */
4031 		if (global_quad_port_a != 0)
4032 			sc->wol = 0;
4033 		/* Reset for multiple quad port adapters */
4034 		if (++global_quad_port_a == 4)
4035 			global_quad_port_a = 0;
4036 		break;
4037 	case E1000_DEV_ID_82571EB_FIBER:
4038 		/* Wake events only supported on port A for dual fiber
4039 		 * regardless of eeprom setting */
4040 		if (E1000_READ_REG(&sc->hw, E1000_STATUS) &
4041 		    E1000_STATUS_FUNC_1)
4042 			sc->wol = 0;
4043 		break;
4044 	case E1000_DEV_ID_82571EB_QUAD_COPPER:
4045 	case E1000_DEV_ID_82571EB_QUAD_FIBER:
4046 	case E1000_DEV_ID_82571EB_QUAD_COPPER_LP:
4047 		/* if quad port adapter, disable WoL on all but port A */
4048 		if (global_quad_port_a != 0)
4049 			sc->wol = 0;
4050 		/* Reset for multiple quad port adapters */
4051 		if (++global_quad_port_a == 4)
4052 			global_quad_port_a = 0;
4053 		break;
4054 	}
4055 	return;
4056 }
4057 
4058 
4059 /*
4060  * Enable PCI Wake On Lan capability
4061  */
4062 static void
4063 em_enable_wakeup(if_ctx_t ctx)
4064 {
4065 	struct e1000_softc *sc = iflib_get_softc(ctx);
4066 	device_t dev = iflib_get_dev(ctx);
4067 	if_t ifp = iflib_get_ifp(ctx);
4068 	int error = 0;
4069 	u32 pmc, ctrl, ctrl_ext, rctl;
4070 	u16 status;
4071 
4072 	if (pci_find_cap(dev, PCIY_PMG, &pmc) != 0)
4073 		return;
4074 
4075 	/*
4076 	 * Determine type of Wakeup: note that wol
4077 	 * is set with all bits on by default.
4078 	 */
4079 	if ((if_getcapenable(ifp) & IFCAP_WOL_MAGIC) == 0)
4080 		sc->wol &= ~E1000_WUFC_MAG;
4081 
4082 	if ((if_getcapenable(ifp) & IFCAP_WOL_UCAST) == 0)
4083 		sc->wol &= ~E1000_WUFC_EX;
4084 
4085 	if ((if_getcapenable(ifp) & IFCAP_WOL_MCAST) == 0)
4086 		sc->wol &= ~E1000_WUFC_MC;
4087 	else {
4088 		rctl = E1000_READ_REG(&sc->hw, E1000_RCTL);
4089 		rctl |= E1000_RCTL_MPE;
4090 		E1000_WRITE_REG(&sc->hw, E1000_RCTL, rctl);
4091 	}
4092 
4093 	if (!(sc->wol & (E1000_WUFC_EX | E1000_WUFC_MAG | E1000_WUFC_MC)))
4094 		goto pme;
4095 
4096 	/* Advertise the wakeup capability */
4097 	ctrl = E1000_READ_REG(&sc->hw, E1000_CTRL);
4098 	ctrl |= (E1000_CTRL_SWDPIN2 | E1000_CTRL_SWDPIN3);
4099 	E1000_WRITE_REG(&sc->hw, E1000_CTRL, ctrl);
4100 
4101 	/* Keep the laser running on Fiber adapters */
4102 	if (sc->hw.phy.media_type == e1000_media_type_fiber ||
4103 	    sc->hw.phy.media_type == e1000_media_type_internal_serdes) {
4104 		ctrl_ext = E1000_READ_REG(&sc->hw, E1000_CTRL_EXT);
4105 		ctrl_ext |= E1000_CTRL_EXT_SDP3_DATA;
4106 		E1000_WRITE_REG(&sc->hw, E1000_CTRL_EXT, ctrl_ext);
4107 	}
4108 
4109 	if ((sc->hw.mac.type == e1000_ich8lan) ||
4110 	    (sc->hw.mac.type == e1000_pchlan) ||
4111 	    (sc->hw.mac.type == e1000_ich9lan) ||
4112 	    (sc->hw.mac.type == e1000_ich10lan))
4113 		e1000_suspend_workarounds_ich8lan(&sc->hw);
4114 
4115 	if ( sc->hw.mac.type >= e1000_pchlan) {
4116 		error = em_enable_phy_wakeup(sc);
4117 		if (error)
4118 			goto pme;
4119 	} else {
4120 		/* Enable wakeup by the MAC */
4121 		E1000_WRITE_REG(&sc->hw, E1000_WUC, E1000_WUC_PME_EN);
4122 		E1000_WRITE_REG(&sc->hw, E1000_WUFC, sc->wol);
4123 	}
4124 
4125 	if (sc->hw.phy.type == e1000_phy_igp_3)
4126 		e1000_igp3_phy_powerdown_workaround_ich8lan(&sc->hw);
4127 
4128 pme:
4129 	status = pci_read_config(dev, pmc + PCIR_POWER_STATUS, 2);
4130 	status &= ~(PCIM_PSTAT_PME | PCIM_PSTAT_PMEENABLE);
4131 	if (!error && (if_getcapenable(ifp) & IFCAP_WOL))
4132 		status |= PCIM_PSTAT_PME | PCIM_PSTAT_PMEENABLE;
4133 	pci_write_config(dev, pmc + PCIR_POWER_STATUS, status, 2);
4134 
4135 	return;
4136 }
4137 
4138 /*
4139  * WOL in the newer chipset interfaces (pchlan)
4140  * require thing to be copied into the phy
4141  */
4142 static int
4143 em_enable_phy_wakeup(struct e1000_softc *sc)
4144 {
4145 	struct e1000_hw *hw = &sc->hw;
4146 	u32 mreg, ret = 0;
4147 	u16 preg;
4148 
4149 	/* copy MAC RARs to PHY RARs */
4150 	e1000_copy_rx_addrs_to_phy_ich8lan(hw);
4151 
4152 	/* copy MAC MTA to PHY MTA */
4153 	for (int i = 0; i < hw->mac.mta_reg_count; i++) {
4154 		mreg = E1000_READ_REG_ARRAY(hw, E1000_MTA, i);
4155 		e1000_write_phy_reg(hw, BM_MTA(i), (u16)(mreg & 0xFFFF));
4156 		e1000_write_phy_reg(hw, BM_MTA(i) + 1,
4157 		    (u16)((mreg >> 16) & 0xFFFF));
4158 	}
4159 
4160 	/* configure PHY Rx Control register */
4161 	e1000_read_phy_reg(hw, BM_RCTL, &preg);
4162 	mreg = E1000_READ_REG(hw, E1000_RCTL);
4163 	if (mreg & E1000_RCTL_UPE)
4164 		preg |= BM_RCTL_UPE;
4165 	if (mreg & E1000_RCTL_MPE)
4166 		preg |= BM_RCTL_MPE;
4167 	preg &= ~(BM_RCTL_MO_MASK);
4168 	if (mreg & E1000_RCTL_MO_3)
4169 		preg |= (((mreg & E1000_RCTL_MO_3) >> E1000_RCTL_MO_SHIFT)
4170 				<< BM_RCTL_MO_SHIFT);
4171 	if (mreg & E1000_RCTL_BAM)
4172 		preg |= BM_RCTL_BAM;
4173 	if (mreg & E1000_RCTL_PMCF)
4174 		preg |= BM_RCTL_PMCF;
4175 	mreg = E1000_READ_REG(hw, E1000_CTRL);
4176 	if (mreg & E1000_CTRL_RFCE)
4177 		preg |= BM_RCTL_RFCE;
4178 	e1000_write_phy_reg(hw, BM_RCTL, preg);
4179 
4180 	/* enable PHY wakeup in MAC register */
4181 	E1000_WRITE_REG(hw, E1000_WUC,
4182 	    E1000_WUC_PHY_WAKE | E1000_WUC_PME_EN | E1000_WUC_APME);
4183 	E1000_WRITE_REG(hw, E1000_WUFC, sc->wol);
4184 
4185 	/* configure and enable PHY wakeup in PHY registers */
4186 	e1000_write_phy_reg(hw, BM_WUFC, sc->wol);
4187 	e1000_write_phy_reg(hw, BM_WUC, E1000_WUC_PME_EN);
4188 
4189 	/* activate PHY wakeup */
4190 	ret = hw->phy.ops.acquire(hw);
4191 	if (ret) {
4192 		printf("Could not acquire PHY\n");
4193 		return ret;
4194 	}
4195 	e1000_write_phy_reg_mdic(hw, IGP01E1000_PHY_PAGE_SELECT,
4196 	                         (BM_WUC_ENABLE_PAGE << IGP_PAGE_SHIFT));
4197 	ret = e1000_read_phy_reg_mdic(hw, BM_WUC_ENABLE_REG, &preg);
4198 	if (ret) {
4199 		printf("Could not read PHY page 769\n");
4200 		goto out;
4201 	}
4202 	preg |= BM_WUC_ENABLE_BIT | BM_WUC_HOST_WU_BIT;
4203 	ret = e1000_write_phy_reg_mdic(hw, BM_WUC_ENABLE_REG, preg);
4204 	if (ret)
4205 		printf("Could not set PHY Host Wakeup bit\n");
4206 out:
4207 	hw->phy.ops.release(hw);
4208 
4209 	return ret;
4210 }
4211 
4212 static void
4213 em_if_led_func(if_ctx_t ctx, int onoff)
4214 {
4215 	struct e1000_softc *sc = iflib_get_softc(ctx);
4216 
4217 	if (onoff) {
4218 		e1000_setup_led(&sc->hw);
4219 		e1000_led_on(&sc->hw);
4220 	} else {
4221 		e1000_led_off(&sc->hw);
4222 		e1000_cleanup_led(&sc->hw);
4223 	}
4224 }
4225 
4226 /*
4227  * Disable the L0S and L1 LINK states
4228  */
4229 static void
4230 em_disable_aspm(struct e1000_softc *sc)
4231 {
4232 	int base, reg;
4233 	u16 link_cap,link_ctrl;
4234 	device_t dev = sc->dev;
4235 
4236 	switch (sc->hw.mac.type) {
4237 	case e1000_82573:
4238 	case e1000_82574:
4239 	case e1000_82583:
4240 		break;
4241 	default:
4242 		return;
4243 	}
4244 	if (pci_find_cap(dev, PCIY_EXPRESS, &base) != 0)
4245 		return;
4246 	reg = base + PCIER_LINK_CAP;
4247 	link_cap = pci_read_config(dev, reg, 2);
4248 	if ((link_cap & PCIEM_LINK_CAP_ASPM) == 0)
4249 		return;
4250 	reg = base + PCIER_LINK_CTL;
4251 	link_ctrl = pci_read_config(dev, reg, 2);
4252 	link_ctrl &= ~PCIEM_LINK_CTL_ASPMC;
4253 	pci_write_config(dev, reg, link_ctrl, 2);
4254 	return;
4255 }
4256 
4257 /**********************************************************************
4258  *
4259  *  Update the board statistics counters.
4260  *
4261  **********************************************************************/
4262 static void
4263 em_update_stats_counters(struct e1000_softc *sc)
4264 {
4265 	u64 prev_xoffrxc = sc->stats.xoffrxc;
4266 
4267 	if(sc->hw.phy.media_type == e1000_media_type_copper ||
4268 	   (E1000_READ_REG(&sc->hw, E1000_STATUS) & E1000_STATUS_LU)) {
4269 		sc->stats.symerrs += E1000_READ_REG(&sc->hw, E1000_SYMERRS);
4270 		sc->stats.sec += E1000_READ_REG(&sc->hw, E1000_SEC);
4271 	}
4272 	sc->stats.crcerrs += E1000_READ_REG(&sc->hw, E1000_CRCERRS);
4273 	sc->stats.mpc += E1000_READ_REG(&sc->hw, E1000_MPC);
4274 	sc->stats.scc += E1000_READ_REG(&sc->hw, E1000_SCC);
4275 	sc->stats.ecol += E1000_READ_REG(&sc->hw, E1000_ECOL);
4276 
4277 	sc->stats.mcc += E1000_READ_REG(&sc->hw, E1000_MCC);
4278 	sc->stats.latecol += E1000_READ_REG(&sc->hw, E1000_LATECOL);
4279 	sc->stats.colc += E1000_READ_REG(&sc->hw, E1000_COLC);
4280 	sc->stats.dc += E1000_READ_REG(&sc->hw, E1000_DC);
4281 	sc->stats.rlec += E1000_READ_REG(&sc->hw, E1000_RLEC);
4282 	sc->stats.xonrxc += E1000_READ_REG(&sc->hw, E1000_XONRXC);
4283 	sc->stats.xontxc += E1000_READ_REG(&sc->hw, E1000_XONTXC);
4284 	sc->stats.xoffrxc += E1000_READ_REG(&sc->hw, E1000_XOFFRXC);
4285 	/*
4286 	 ** For watchdog management we need to know if we have been
4287 	 ** paused during the last interval, so capture that here.
4288 	*/
4289 	if (sc->stats.xoffrxc != prev_xoffrxc)
4290 		sc->shared->isc_pause_frames = 1;
4291 	sc->stats.xofftxc += E1000_READ_REG(&sc->hw, E1000_XOFFTXC);
4292 	sc->stats.fcruc += E1000_READ_REG(&sc->hw, E1000_FCRUC);
4293 	sc->stats.prc64 += E1000_READ_REG(&sc->hw, E1000_PRC64);
4294 	sc->stats.prc127 += E1000_READ_REG(&sc->hw, E1000_PRC127);
4295 	sc->stats.prc255 += E1000_READ_REG(&sc->hw, E1000_PRC255);
4296 	sc->stats.prc511 += E1000_READ_REG(&sc->hw, E1000_PRC511);
4297 	sc->stats.prc1023 += E1000_READ_REG(&sc->hw, E1000_PRC1023);
4298 	sc->stats.prc1522 += E1000_READ_REG(&sc->hw, E1000_PRC1522);
4299 	sc->stats.gprc += E1000_READ_REG(&sc->hw, E1000_GPRC);
4300 	sc->stats.bprc += E1000_READ_REG(&sc->hw, E1000_BPRC);
4301 	sc->stats.mprc += E1000_READ_REG(&sc->hw, E1000_MPRC);
4302 	sc->stats.gptc += E1000_READ_REG(&sc->hw, E1000_GPTC);
4303 
4304 	/* For the 64-bit byte counters the low dword must be read first. */
4305 	/* Both registers clear on the read of the high dword */
4306 
4307 	sc->stats.gorc += E1000_READ_REG(&sc->hw, E1000_GORCL) +
4308 	    ((u64)E1000_READ_REG(&sc->hw, E1000_GORCH) << 32);
4309 	sc->stats.gotc += E1000_READ_REG(&sc->hw, E1000_GOTCL) +
4310 	    ((u64)E1000_READ_REG(&sc->hw, E1000_GOTCH) << 32);
4311 
4312 	sc->stats.rnbc += E1000_READ_REG(&sc->hw, E1000_RNBC);
4313 	sc->stats.ruc += E1000_READ_REG(&sc->hw, E1000_RUC);
4314 	sc->stats.rfc += E1000_READ_REG(&sc->hw, E1000_RFC);
4315 	sc->stats.roc += E1000_READ_REG(&sc->hw, E1000_ROC);
4316 	sc->stats.rjc += E1000_READ_REG(&sc->hw, E1000_RJC);
4317 
4318 	sc->stats.tor += E1000_READ_REG(&sc->hw, E1000_TORH);
4319 	sc->stats.tot += E1000_READ_REG(&sc->hw, E1000_TOTH);
4320 
4321 	sc->stats.tpr += E1000_READ_REG(&sc->hw, E1000_TPR);
4322 	sc->stats.tpt += E1000_READ_REG(&sc->hw, E1000_TPT);
4323 	sc->stats.ptc64 += E1000_READ_REG(&sc->hw, E1000_PTC64);
4324 	sc->stats.ptc127 += E1000_READ_REG(&sc->hw, E1000_PTC127);
4325 	sc->stats.ptc255 += E1000_READ_REG(&sc->hw, E1000_PTC255);
4326 	sc->stats.ptc511 += E1000_READ_REG(&sc->hw, E1000_PTC511);
4327 	sc->stats.ptc1023 += E1000_READ_REG(&sc->hw, E1000_PTC1023);
4328 	sc->stats.ptc1522 += E1000_READ_REG(&sc->hw, E1000_PTC1522);
4329 	sc->stats.mptc += E1000_READ_REG(&sc->hw, E1000_MPTC);
4330 	sc->stats.bptc += E1000_READ_REG(&sc->hw, E1000_BPTC);
4331 
4332 	/* Interrupt Counts */
4333 
4334 	sc->stats.iac += E1000_READ_REG(&sc->hw, E1000_IAC);
4335 	sc->stats.icrxptc += E1000_READ_REG(&sc->hw, E1000_ICRXPTC);
4336 	sc->stats.icrxatc += E1000_READ_REG(&sc->hw, E1000_ICRXATC);
4337 	sc->stats.ictxptc += E1000_READ_REG(&sc->hw, E1000_ICTXPTC);
4338 	sc->stats.ictxatc += E1000_READ_REG(&sc->hw, E1000_ICTXATC);
4339 	sc->stats.ictxqec += E1000_READ_REG(&sc->hw, E1000_ICTXQEC);
4340 	sc->stats.ictxqmtc += E1000_READ_REG(&sc->hw, E1000_ICTXQMTC);
4341 	sc->stats.icrxdmtc += E1000_READ_REG(&sc->hw, E1000_ICRXDMTC);
4342 	sc->stats.icrxoc += E1000_READ_REG(&sc->hw, E1000_ICRXOC);
4343 
4344 	if (sc->hw.mac.type >= e1000_82543) {
4345 		sc->stats.algnerrc +=
4346 		E1000_READ_REG(&sc->hw, E1000_ALGNERRC);
4347 		sc->stats.rxerrc +=
4348 		E1000_READ_REG(&sc->hw, E1000_RXERRC);
4349 		sc->stats.tncrs +=
4350 		E1000_READ_REG(&sc->hw, E1000_TNCRS);
4351 		sc->stats.cexterr +=
4352 		E1000_READ_REG(&sc->hw, E1000_CEXTERR);
4353 		sc->stats.tsctc +=
4354 		E1000_READ_REG(&sc->hw, E1000_TSCTC);
4355 		sc->stats.tsctfc +=
4356 		E1000_READ_REG(&sc->hw, E1000_TSCTFC);
4357 	}
4358 }
4359 
4360 static uint64_t
4361 em_if_get_counter(if_ctx_t ctx, ift_counter cnt)
4362 {
4363 	struct e1000_softc *sc = iflib_get_softc(ctx);
4364 	if_t ifp = iflib_get_ifp(ctx);
4365 
4366 	switch (cnt) {
4367 	case IFCOUNTER_COLLISIONS:
4368 		return (sc->stats.colc);
4369 	case IFCOUNTER_IERRORS:
4370 		return (sc->dropped_pkts + sc->stats.rxerrc +
4371 		    sc->stats.crcerrs + sc->stats.algnerrc +
4372 		    sc->stats.ruc + sc->stats.roc +
4373 		    sc->stats.mpc + sc->stats.cexterr);
4374 	case IFCOUNTER_OERRORS:
4375 		return (sc->stats.ecol + sc->stats.latecol +
4376 		    sc->watchdog_events);
4377 	default:
4378 		return (if_get_counter_default(ifp, cnt));
4379 	}
4380 }
4381 
4382 /* em_if_needs_restart - Tell iflib when the driver needs to be reinitialized
4383  * @ctx: iflib context
4384  * @event: event code to check
4385  *
4386  * Defaults to returning false for unknown events.
4387  *
4388  * @returns true if iflib needs to reinit the interface
4389  */
4390 static bool
4391 em_if_needs_restart(if_ctx_t ctx __unused, enum iflib_restart_event event)
4392 {
4393 	switch (event) {
4394 	case IFLIB_RESTART_VLAN_CONFIG:
4395 	default:
4396 		return (false);
4397 	}
4398 }
4399 
4400 /* Export a single 32-bit register via a read-only sysctl. */
4401 static int
4402 em_sysctl_reg_handler(SYSCTL_HANDLER_ARGS)
4403 {
4404 	struct e1000_softc *sc;
4405 	u_int val;
4406 
4407 	sc = oidp->oid_arg1;
4408 	val = E1000_READ_REG(&sc->hw, oidp->oid_arg2);
4409 	return (sysctl_handle_int(oidp, &val, 0, req));
4410 }
4411 
4412 /*
4413  * Add sysctl variables, one per statistic, to the system.
4414  */
4415 static void
4416 em_add_hw_stats(struct e1000_softc *sc)
4417 {
4418 	device_t dev = iflib_get_dev(sc->ctx);
4419 	struct em_tx_queue *tx_que = sc->tx_queues;
4420 	struct em_rx_queue *rx_que = sc->rx_queues;
4421 
4422 	struct sysctl_ctx_list *ctx = device_get_sysctl_ctx(dev);
4423 	struct sysctl_oid *tree = device_get_sysctl_tree(dev);
4424 	struct sysctl_oid_list *child = SYSCTL_CHILDREN(tree);
4425 	struct e1000_hw_stats *stats = &sc->stats;
4426 
4427 	struct sysctl_oid *stat_node, *queue_node, *int_node;
4428 	struct sysctl_oid_list *stat_list, *queue_list, *int_list;
4429 
4430 #define QUEUE_NAME_LEN 32
4431 	char namebuf[QUEUE_NAME_LEN];
4432 
4433 	/* Driver Statistics */
4434 	SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "dropped",
4435 			CTLFLAG_RD, &sc->dropped_pkts,
4436 			"Driver dropped packets");
4437 	SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "link_irq",
4438 			CTLFLAG_RD, &sc->link_irq,
4439 			"Link MSI-X IRQ Handled");
4440 	SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "rx_overruns",
4441 			CTLFLAG_RD, &sc->rx_overruns,
4442 			"RX overruns");
4443 	SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "watchdog_timeouts",
4444 			CTLFLAG_RD, &sc->watchdog_events,
4445 			"Watchdog timeouts");
4446 	SYSCTL_ADD_PROC(ctx, child, OID_AUTO, "device_control",
4447 	    CTLTYPE_UINT | CTLFLAG_RD | CTLFLAG_NEEDGIANT,
4448 	    sc, E1000_CTRL, em_sysctl_reg_handler, "IU",
4449 	    "Device Control Register");
4450 	SYSCTL_ADD_PROC(ctx, child, OID_AUTO, "rx_control",
4451 	    CTLTYPE_UINT | CTLFLAG_RD | CTLFLAG_NEEDGIANT,
4452 	    sc, E1000_RCTL, em_sysctl_reg_handler, "IU",
4453 	    "Receiver Control Register");
4454 	SYSCTL_ADD_UINT(ctx, child, OID_AUTO, "fc_high_water",
4455 			CTLFLAG_RD, &sc->hw.fc.high_water, 0,
4456 			"Flow Control High Watermark");
4457 	SYSCTL_ADD_UINT(ctx, child, OID_AUTO, "fc_low_water",
4458 			CTLFLAG_RD, &sc->hw.fc.low_water, 0,
4459 			"Flow Control Low Watermark");
4460 
4461 	for (int i = 0; i < sc->tx_num_queues; i++, tx_que++) {
4462 		struct tx_ring *txr = &tx_que->txr;
4463 		snprintf(namebuf, QUEUE_NAME_LEN, "queue_tx_%d", i);
4464 		queue_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, namebuf,
4465 		    CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, "TX Queue Name");
4466 		queue_list = SYSCTL_CHILDREN(queue_node);
4467 
4468 		SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "txd_head",
4469 		    CTLTYPE_UINT | CTLFLAG_RD | CTLFLAG_NEEDGIANT, sc,
4470 		    E1000_TDH(txr->me), em_sysctl_reg_handler, "IU",
4471 		    "Transmit Descriptor Head");
4472 		SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "txd_tail",
4473 		    CTLTYPE_UINT | CTLFLAG_RD | CTLFLAG_NEEDGIANT, sc,
4474 		    E1000_TDT(txr->me), em_sysctl_reg_handler, "IU",
4475 		    "Transmit Descriptor Tail");
4476 		SYSCTL_ADD_ULONG(ctx, queue_list, OID_AUTO, "tx_irq",
4477 				CTLFLAG_RD, &txr->tx_irq,
4478 				"Queue MSI-X Transmit Interrupts");
4479 	}
4480 
4481 	for (int j = 0; j < sc->rx_num_queues; j++, rx_que++) {
4482 		struct rx_ring *rxr = &rx_que->rxr;
4483 		snprintf(namebuf, QUEUE_NAME_LEN, "queue_rx_%d", j);
4484 		queue_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, namebuf,
4485 		    CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, "RX Queue Name");
4486 		queue_list = SYSCTL_CHILDREN(queue_node);
4487 
4488 		SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "rxd_head",
4489 		    CTLTYPE_UINT | CTLFLAG_RD | CTLFLAG_NEEDGIANT, sc,
4490 		    E1000_RDH(rxr->me), em_sysctl_reg_handler, "IU",
4491 		    "Receive Descriptor Head");
4492 		SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "rxd_tail",
4493 		    CTLTYPE_UINT | CTLFLAG_RD | CTLFLAG_NEEDGIANT, sc,
4494 		    E1000_RDT(rxr->me), em_sysctl_reg_handler, "IU",
4495 		    "Receive Descriptor Tail");
4496 		SYSCTL_ADD_ULONG(ctx, queue_list, OID_AUTO, "rx_irq",
4497 				CTLFLAG_RD, &rxr->rx_irq,
4498 				"Queue MSI-X Receive Interrupts");
4499 	}
4500 
4501 	/* MAC stats get their own sub node */
4502 
4503 	stat_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, "mac_stats",
4504 	    CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, "Statistics");
4505 	stat_list = SYSCTL_CHILDREN(stat_node);
4506 
4507 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "excess_coll",
4508 			CTLFLAG_RD, &stats->ecol,
4509 			"Excessive collisions");
4510 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "single_coll",
4511 			CTLFLAG_RD, &stats->scc,
4512 			"Single collisions");
4513 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "multiple_coll",
4514 			CTLFLAG_RD, &stats->mcc,
4515 			"Multiple collisions");
4516 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "late_coll",
4517 			CTLFLAG_RD, &stats->latecol,
4518 			"Late collisions");
4519 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "collision_count",
4520 			CTLFLAG_RD, &stats->colc,
4521 			"Collision Count");
4522 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "symbol_errors",
4523 			CTLFLAG_RD, &sc->stats.symerrs,
4524 			"Symbol Errors");
4525 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "sequence_errors",
4526 			CTLFLAG_RD, &sc->stats.sec,
4527 			"Sequence Errors");
4528 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "defer_count",
4529 			CTLFLAG_RD, &sc->stats.dc,
4530 			"Defer Count");
4531 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "missed_packets",
4532 			CTLFLAG_RD, &sc->stats.mpc,
4533 			"Missed Packets");
4534 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_no_buff",
4535 			CTLFLAG_RD, &sc->stats.rnbc,
4536 			"Receive No Buffers");
4537 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_undersize",
4538 			CTLFLAG_RD, &sc->stats.ruc,
4539 			"Receive Undersize");
4540 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_fragmented",
4541 			CTLFLAG_RD, &sc->stats.rfc,
4542 			"Fragmented Packets Received ");
4543 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_oversize",
4544 			CTLFLAG_RD, &sc->stats.roc,
4545 			"Oversized Packets Received");
4546 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_jabber",
4547 			CTLFLAG_RD, &sc->stats.rjc,
4548 			"Recevied Jabber");
4549 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_errs",
4550 			CTLFLAG_RD, &sc->stats.rxerrc,
4551 			"Receive Errors");
4552 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "crc_errs",
4553 			CTLFLAG_RD, &sc->stats.crcerrs,
4554 			"CRC errors");
4555 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "alignment_errs",
4556 			CTLFLAG_RD, &sc->stats.algnerrc,
4557 			"Alignment Errors");
4558 	/* On 82575 these are collision counts */
4559 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "coll_ext_errs",
4560 			CTLFLAG_RD, &sc->stats.cexterr,
4561 			"Collision/Carrier extension errors");
4562 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xon_recvd",
4563 			CTLFLAG_RD, &sc->stats.xonrxc,
4564 			"XON Received");
4565 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xon_txd",
4566 			CTLFLAG_RD, &sc->stats.xontxc,
4567 			"XON Transmitted");
4568 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xoff_recvd",
4569 			CTLFLAG_RD, &sc->stats.xoffrxc,
4570 			"XOFF Received");
4571 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xoff_txd",
4572 			CTLFLAG_RD, &sc->stats.xofftxc,
4573 			"XOFF Transmitted");
4574 
4575 	/* Packet Reception Stats */
4576 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "total_pkts_recvd",
4577 			CTLFLAG_RD, &sc->stats.tpr,
4578 			"Total Packets Received ");
4579 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_pkts_recvd",
4580 			CTLFLAG_RD, &sc->stats.gprc,
4581 			"Good Packets Received");
4582 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "bcast_pkts_recvd",
4583 			CTLFLAG_RD, &sc->stats.bprc,
4584 			"Broadcast Packets Received");
4585 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "mcast_pkts_recvd",
4586 			CTLFLAG_RD, &sc->stats.mprc,
4587 			"Multicast Packets Received");
4588 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_64",
4589 			CTLFLAG_RD, &sc->stats.prc64,
4590 			"64 byte frames received ");
4591 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_65_127",
4592 			CTLFLAG_RD, &sc->stats.prc127,
4593 			"65-127 byte frames received");
4594 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_128_255",
4595 			CTLFLAG_RD, &sc->stats.prc255,
4596 			"128-255 byte frames received");
4597 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_256_511",
4598 			CTLFLAG_RD, &sc->stats.prc511,
4599 			"256-511 byte frames received");
4600 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_512_1023",
4601 			CTLFLAG_RD, &sc->stats.prc1023,
4602 			"512-1023 byte frames received");
4603 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_1024_1522",
4604 			CTLFLAG_RD, &sc->stats.prc1522,
4605 			"1023-1522 byte frames received");
4606 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_octets_recvd",
4607 			CTLFLAG_RD, &sc->stats.gorc,
4608 			"Good Octets Received");
4609 
4610 	/* Packet Transmission Stats */
4611 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_octets_txd",
4612 			CTLFLAG_RD, &sc->stats.gotc,
4613 			"Good Octets Transmitted");
4614 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "total_pkts_txd",
4615 			CTLFLAG_RD, &sc->stats.tpt,
4616 			"Total Packets Transmitted");
4617 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_pkts_txd",
4618 			CTLFLAG_RD, &sc->stats.gptc,
4619 			"Good Packets Transmitted");
4620 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "bcast_pkts_txd",
4621 			CTLFLAG_RD, &sc->stats.bptc,
4622 			"Broadcast Packets Transmitted");
4623 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "mcast_pkts_txd",
4624 			CTLFLAG_RD, &sc->stats.mptc,
4625 			"Multicast Packets Transmitted");
4626 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_64",
4627 			CTLFLAG_RD, &sc->stats.ptc64,
4628 			"64 byte frames transmitted ");
4629 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_65_127",
4630 			CTLFLAG_RD, &sc->stats.ptc127,
4631 			"65-127 byte frames transmitted");
4632 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_128_255",
4633 			CTLFLAG_RD, &sc->stats.ptc255,
4634 			"128-255 byte frames transmitted");
4635 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_256_511",
4636 			CTLFLAG_RD, &sc->stats.ptc511,
4637 			"256-511 byte frames transmitted");
4638 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_512_1023",
4639 			CTLFLAG_RD, &sc->stats.ptc1023,
4640 			"512-1023 byte frames transmitted");
4641 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_1024_1522",
4642 			CTLFLAG_RD, &sc->stats.ptc1522,
4643 			"1024-1522 byte frames transmitted");
4644 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tso_txd",
4645 			CTLFLAG_RD, &sc->stats.tsctc,
4646 			"TSO Contexts Transmitted");
4647 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tso_ctx_fail",
4648 			CTLFLAG_RD, &sc->stats.tsctfc,
4649 			"TSO Contexts Failed");
4650 
4651 
4652 	/* Interrupt Stats */
4653 
4654 	int_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, "interrupts",
4655 	    CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, "Interrupt Statistics");
4656 	int_list = SYSCTL_CHILDREN(int_node);
4657 
4658 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "asserts",
4659 			CTLFLAG_RD, &sc->stats.iac,
4660 			"Interrupt Assertion Count");
4661 
4662 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "rx_pkt_timer",
4663 			CTLFLAG_RD, &sc->stats.icrxptc,
4664 			"Interrupt Cause Rx Pkt Timer Expire Count");
4665 
4666 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "rx_abs_timer",
4667 			CTLFLAG_RD, &sc->stats.icrxatc,
4668 			"Interrupt Cause Rx Abs Timer Expire Count");
4669 
4670 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "tx_pkt_timer",
4671 			CTLFLAG_RD, &sc->stats.ictxptc,
4672 			"Interrupt Cause Tx Pkt Timer Expire Count");
4673 
4674 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "tx_abs_timer",
4675 			CTLFLAG_RD, &sc->stats.ictxatc,
4676 			"Interrupt Cause Tx Abs Timer Expire Count");
4677 
4678 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "tx_queue_empty",
4679 			CTLFLAG_RD, &sc->stats.ictxqec,
4680 			"Interrupt Cause Tx Queue Empty Count");
4681 
4682 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "tx_queue_min_thresh",
4683 			CTLFLAG_RD, &sc->stats.ictxqmtc,
4684 			"Interrupt Cause Tx Queue Min Thresh Count");
4685 
4686 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "rx_desc_min_thresh",
4687 			CTLFLAG_RD, &sc->stats.icrxdmtc,
4688 			"Interrupt Cause Rx Desc Min Thresh Count");
4689 
4690 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "rx_overrun",
4691 			CTLFLAG_RD, &sc->stats.icrxoc,
4692 			"Interrupt Cause Receiver Overrun Count");
4693 }
4694 
4695 static void
4696 em_fw_version_locked(if_ctx_t ctx)
4697 {
4698 	struct e1000_softc *sc = iflib_get_softc(ctx);
4699 	struct e1000_hw *hw = &sc->hw;
4700 	struct e1000_fw_version *fw_ver = &sc->fw_ver;
4701 	uint16_t eep = 0;
4702 
4703 	/*
4704 	 * em_fw_version_locked() must run under the IFLIB_CTX_LOCK to meet the
4705 	 * NVM locking model, so we do it in em_if_attach_pre() and store the
4706 	 * info in the softc
4707 	 */
4708 	ASSERT_CTX_LOCK_HELD(hw);
4709 
4710 	*fw_ver = (struct e1000_fw_version){0};
4711 
4712 	if (hw->mac.type >= igb_mac_min) {
4713 		/*
4714 		 * Use the Shared Code for igb(4)
4715 		 */
4716 		e1000_get_fw_version(hw, fw_ver);
4717 	} else {
4718 		/*
4719 		 * Otherwise, EEPROM version should be present on (almost?) all
4720 		 * devices here
4721 		 */
4722 		if(e1000_read_nvm(hw, NVM_VERSION, 1, &eep)) {
4723 			INIT_DEBUGOUT("can't get EEPROM version");
4724 			return;
4725 		}
4726 
4727 		fw_ver->eep_major = (eep & NVM_MAJOR_MASK) >> NVM_MAJOR_SHIFT;
4728 		fw_ver->eep_minor = (eep & NVM_MINOR_MASK) >> NVM_MINOR_SHIFT;
4729 		fw_ver->eep_build = (eep & NVM_IMAGE_ID_MASK);
4730 	}
4731 }
4732 
4733 static void
4734 em_sbuf_fw_version(struct e1000_fw_version *fw_ver, struct sbuf *buf)
4735 {
4736 	const char *space = "";
4737 
4738 	if (fw_ver->eep_major || fw_ver->eep_minor || fw_ver->eep_build) {
4739 		sbuf_printf(buf, "EEPROM V%d.%d-%d", fw_ver->eep_major,
4740 			    fw_ver->eep_minor, fw_ver->eep_build);
4741 		space = " ";
4742 	}
4743 
4744 	if (fw_ver->invm_major || fw_ver->invm_minor || fw_ver->invm_img_type) {
4745 		sbuf_printf(buf, "%sNVM V%d.%d imgtype%d",
4746 			    space, fw_ver->invm_major, fw_ver->invm_minor,
4747 			    fw_ver->invm_img_type);
4748 		space = " ";
4749 	}
4750 
4751 	if (fw_ver->or_valid) {
4752 		sbuf_printf(buf, "%sOption ROM V%d-b%d-p%d",
4753 			    space, fw_ver->or_major, fw_ver->or_build,
4754 			    fw_ver->or_patch);
4755 		space = " ";
4756 	}
4757 
4758 	if (fw_ver->etrack_id)
4759 		sbuf_printf(buf, "%seTrack 0x%08x", space, fw_ver->etrack_id);
4760 }
4761 
4762 static void
4763 em_print_fw_version(struct e1000_softc *sc )
4764 {
4765 	device_t dev = sc->dev;
4766 	struct sbuf *buf;
4767 	int error = 0;
4768 
4769 	buf = sbuf_new_auto();
4770 	if (!buf) {
4771 		device_printf(dev, "Could not allocate sbuf for output.\n");
4772 		return;
4773 	}
4774 
4775 	em_sbuf_fw_version(&sc->fw_ver, buf);
4776 
4777 	error = sbuf_finish(buf);
4778 	if (error)
4779 		device_printf(dev, "Error finishing sbuf: %d\n", error);
4780 	else if (sbuf_len(buf))
4781 		device_printf(dev, "%s\n", sbuf_data(buf));
4782 
4783 	sbuf_delete(buf);
4784 }
4785 
4786 static int
4787 em_sysctl_print_fw_version(SYSCTL_HANDLER_ARGS)
4788 {
4789 	struct e1000_softc *sc = (struct e1000_softc *)arg1;
4790 	device_t dev = sc->dev;
4791 	struct sbuf *buf;
4792 	int error = 0;
4793 
4794 	buf = sbuf_new_for_sysctl(NULL, NULL, 128, req);
4795 	if (!buf) {
4796 		device_printf(dev, "Could not allocate sbuf for output.\n");
4797 		return (ENOMEM);
4798 	}
4799 
4800 	em_sbuf_fw_version(&sc->fw_ver, buf);
4801 
4802 	error = sbuf_finish(buf);
4803 	if (error)
4804 		device_printf(dev, "Error finishing sbuf: %d\n", error);
4805 
4806 	sbuf_delete(buf);
4807 
4808 	return (0);
4809 }
4810 
4811 /**********************************************************************
4812  *
4813  *  This routine provides a way to dump out the adapter eeprom,
4814  *  often a useful debug/service tool. This only dumps the first
4815  *  32 words, stuff that matters is in that extent.
4816  *
4817  **********************************************************************/
4818 static int
4819 em_sysctl_nvm_info(SYSCTL_HANDLER_ARGS)
4820 {
4821 	struct e1000_softc *sc = (struct e1000_softc *)arg1;
4822 	int error;
4823 	int result;
4824 
4825 	result = -1;
4826 	error = sysctl_handle_int(oidp, &result, 0, req);
4827 
4828 	if (error || !req->newptr)
4829 		return (error);
4830 
4831 	/*
4832 	 * This value will cause a hex dump of the
4833 	 * first 32 16-bit words of the EEPROM to
4834 	 * the screen.
4835 	 */
4836 	if (result == 1)
4837 		em_print_nvm_info(sc);
4838 
4839 	return (error);
4840 }
4841 
4842 static void
4843 em_print_nvm_info(struct e1000_softc *sc)
4844 {
4845 	struct e1000_hw *hw = &sc->hw;
4846 	struct sx *iflib_ctx_lock = iflib_ctx_lock_get(sc->ctx);
4847 	u16 eeprom_data;
4848 	int i, j, row = 0;
4849 
4850 	/* Its a bit crude, but it gets the job done */
4851 	printf("\nInterface EEPROM Dump:\n");
4852 	printf("Offset\n0x0000  ");
4853 
4854 	/* We rely on the IFLIB_CTX_LOCK as part of NVM locking model */
4855 	sx_xlock(iflib_ctx_lock);
4856 	ASSERT_CTX_LOCK_HELD(hw);
4857 	for (i = 0, j = 0; i < 32; i++, j++) {
4858 		if (j == 8) { /* Make the offset block */
4859 			j = 0; ++row;
4860 			printf("\n0x00%x0  ",row);
4861 		}
4862 		e1000_read_nvm(hw, i, 1, &eeprom_data);
4863 		printf("%04x ", eeprom_data);
4864 	}
4865 	sx_xunlock(iflib_ctx_lock);
4866 	printf("\n");
4867 }
4868 
4869 static int
4870 em_sysctl_int_delay(SYSCTL_HANDLER_ARGS)
4871 {
4872 	struct em_int_delay_info *info;
4873 	struct e1000_softc *sc;
4874 	u32 regval;
4875 	int error, usecs, ticks;
4876 
4877 	info = (struct em_int_delay_info *) arg1;
4878 	usecs = info->value;
4879 	error = sysctl_handle_int(oidp, &usecs, 0, req);
4880 	if (error != 0 || req->newptr == NULL)
4881 		return (error);
4882 	if (usecs < 0 || usecs > EM_TICKS_TO_USECS(65535))
4883 		return (EINVAL);
4884 	info->value = usecs;
4885 	ticks = EM_USECS_TO_TICKS(usecs);
4886 	if (info->offset == E1000_ITR)	/* units are 256ns here */
4887 		ticks *= 4;
4888 
4889 	sc = info->sc;
4890 
4891 	regval = E1000_READ_OFFSET(&sc->hw, info->offset);
4892 	regval = (regval & ~0xffff) | (ticks & 0xffff);
4893 	/* Handle a few special cases. */
4894 	switch (info->offset) {
4895 	case E1000_RDTR:
4896 		break;
4897 	case E1000_TIDV:
4898 		if (ticks == 0) {
4899 			sc->txd_cmd &= ~E1000_TXD_CMD_IDE;
4900 			/* Don't write 0 into the TIDV register. */
4901 			regval++;
4902 		} else
4903 			sc->txd_cmd |= E1000_TXD_CMD_IDE;
4904 		break;
4905 	}
4906 	E1000_WRITE_OFFSET(&sc->hw, info->offset, regval);
4907 	return (0);
4908 }
4909 
4910 static void
4911 em_add_int_delay_sysctl(struct e1000_softc *sc, const char *name,
4912 	const char *description, struct em_int_delay_info *info,
4913 	int offset, int value)
4914 {
4915 	info->sc = sc;
4916 	info->offset = offset;
4917 	info->value = value;
4918 	SYSCTL_ADD_PROC(device_get_sysctl_ctx(sc->dev),
4919 	    SYSCTL_CHILDREN(device_get_sysctl_tree(sc->dev)),
4920 	    OID_AUTO, name, CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_NEEDGIANT,
4921 	    info, 0, em_sysctl_int_delay, "I", description);
4922 }
4923 
4924 /*
4925  * Set flow control using sysctl:
4926  * Flow control values:
4927  *      0 - off
4928  *      1 - rx pause
4929  *      2 - tx pause
4930  *      3 - full
4931  */
4932 static int
4933 em_set_flowcntl(SYSCTL_HANDLER_ARGS)
4934 {
4935 	int error;
4936 	static int input = 3; /* default is full */
4937 	struct e1000_softc	*sc = (struct e1000_softc *) arg1;
4938 
4939 	error = sysctl_handle_int(oidp, &input, 0, req);
4940 
4941 	if ((error) || (req->newptr == NULL))
4942 		return (error);
4943 
4944 	if (input == sc->fc) /* no change? */
4945 		return (error);
4946 
4947 	switch (input) {
4948 	case e1000_fc_rx_pause:
4949 	case e1000_fc_tx_pause:
4950 	case e1000_fc_full:
4951 	case e1000_fc_none:
4952 		sc->hw.fc.requested_mode = input;
4953 		sc->fc = input;
4954 		break;
4955 	default:
4956 		/* Do nothing */
4957 		return (error);
4958 	}
4959 
4960 	sc->hw.fc.current_mode = sc->hw.fc.requested_mode;
4961 	e1000_force_mac_fc(&sc->hw);
4962 	return (error);
4963 }
4964 
4965 /*
4966  * Manage Energy Efficient Ethernet:
4967  * Control values:
4968  *     0/1 - enabled/disabled
4969  */
4970 static int
4971 em_sysctl_eee(SYSCTL_HANDLER_ARGS)
4972 {
4973 	struct e1000_softc *sc = (struct e1000_softc *) arg1;
4974 	int error, value;
4975 
4976 	value = sc->hw.dev_spec.ich8lan.eee_disable;
4977 	error = sysctl_handle_int(oidp, &value, 0, req);
4978 	if (error || req->newptr == NULL)
4979 		return (error);
4980 	sc->hw.dev_spec.ich8lan.eee_disable = (value != 0);
4981 	em_if_init(sc->ctx);
4982 
4983 	return (0);
4984 }
4985 
4986 static int
4987 em_sysctl_debug_info(SYSCTL_HANDLER_ARGS)
4988 {
4989 	struct e1000_softc *sc;
4990 	int error;
4991 	int result;
4992 
4993 	result = -1;
4994 	error = sysctl_handle_int(oidp, &result, 0, req);
4995 
4996 	if (error || !req->newptr)
4997 		return (error);
4998 
4999 	if (result == 1) {
5000 		sc = (struct e1000_softc *) arg1;
5001 		em_print_debug_info(sc);
5002 	}
5003 
5004 	return (error);
5005 }
5006 
5007 static int
5008 em_get_rs(SYSCTL_HANDLER_ARGS)
5009 {
5010 	struct e1000_softc *sc = (struct e1000_softc *) arg1;
5011 	int error;
5012 	int result;
5013 
5014 	result = 0;
5015 	error = sysctl_handle_int(oidp, &result, 0, req);
5016 
5017 	if (error || !req->newptr || result != 1)
5018 		return (error);
5019 	em_dump_rs(sc);
5020 
5021 	return (error);
5022 }
5023 
5024 static void
5025 em_if_debug(if_ctx_t ctx)
5026 {
5027 	em_dump_rs(iflib_get_softc(ctx));
5028 }
5029 
5030 /*
5031  * This routine is meant to be fluid, add whatever is
5032  * needed for debugging a problem.  -jfv
5033  */
5034 static void
5035 em_print_debug_info(struct e1000_softc *sc)
5036 {
5037 	device_t dev = iflib_get_dev(sc->ctx);
5038 	if_t ifp = iflib_get_ifp(sc->ctx);
5039 	struct tx_ring *txr = &sc->tx_queues->txr;
5040 	struct rx_ring *rxr = &sc->rx_queues->rxr;
5041 
5042 	if (if_getdrvflags(ifp) & IFF_DRV_RUNNING)
5043 		printf("Interface is RUNNING ");
5044 	else
5045 		printf("Interface is NOT RUNNING\n");
5046 
5047 	if (if_getdrvflags(ifp) & IFF_DRV_OACTIVE)
5048 		printf("and INACTIVE\n");
5049 	else
5050 		printf("and ACTIVE\n");
5051 
5052 	for (int i = 0; i < sc->tx_num_queues; i++, txr++) {
5053 		device_printf(dev, "TX Queue %d ------\n", i);
5054 		device_printf(dev, "hw tdh = %d, hw tdt = %d\n",
5055 			E1000_READ_REG(&sc->hw, E1000_TDH(i)),
5056 			E1000_READ_REG(&sc->hw, E1000_TDT(i)));
5057 
5058 	}
5059 	for (int j=0; j < sc->rx_num_queues; j++, rxr++) {
5060 		device_printf(dev, "RX Queue %d ------\n", j);
5061 		device_printf(dev, "hw rdh = %d, hw rdt = %d\n",
5062 			E1000_READ_REG(&sc->hw, E1000_RDH(j)),
5063 			E1000_READ_REG(&sc->hw, E1000_RDT(j)));
5064 	}
5065 }
5066 
5067 /*
5068  * 82574 only:
5069  * Write a new value to the EEPROM increasing the number of MSI-X
5070  * vectors from 3 to 5, for proper multiqueue support.
5071  */
5072 static void
5073 em_enable_vectors_82574(if_ctx_t ctx)
5074 {
5075 	struct e1000_softc *sc = iflib_get_softc(ctx);
5076 	struct e1000_hw *hw = &sc->hw;
5077 	device_t dev = iflib_get_dev(ctx);
5078 	u16 edata;
5079 
5080 	e1000_read_nvm(hw, EM_NVM_PCIE_CTRL, 1, &edata);
5081 	if (bootverbose)
5082 		device_printf(dev, "EM_NVM_PCIE_CTRL = %#06x\n", edata);
5083 	if (((edata & EM_NVM_MSIX_N_MASK) >> EM_NVM_MSIX_N_SHIFT) != 4) {
5084 		device_printf(dev, "Writing to eeprom: increasing "
5085 		    "reported MSI-X vectors from 3 to 5...\n");
5086 		edata &= ~(EM_NVM_MSIX_N_MASK);
5087 		edata |= 4 << EM_NVM_MSIX_N_SHIFT;
5088 		e1000_write_nvm(hw, EM_NVM_PCIE_CTRL, 1, &edata);
5089 		e1000_update_nvm_checksum(hw);
5090 		device_printf(dev, "Writing to eeprom: done\n");
5091 	}
5092 }
5093