xref: /freebsd/sys/dev/e1000/if_em.c (revision 162ae9c834f6d9f9cb443bd62cceb23e0b5fef48)
1 /*-
2  * SPDX-License-Identifier: BSD-2-Clause
3  *
4  * Copyright (c) 2016 Nicole Graziano <nicole@nextbsd.org>
5  * All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  *
16  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
17  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
18  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
19  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
20  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
21  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
22  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
23  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
24  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
25  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
26  * SUCH DAMAGE.
27  */
28 
29 /* $FreeBSD$ */
30 #include "if_em.h"
31 #include <sys/sbuf.h>
32 #include <machine/_inttypes.h>
33 
34 #define em_mac_min e1000_82547
35 #define igb_mac_min e1000_82575
36 
37 /*********************************************************************
38  *  Driver version:
39  *********************************************************************/
40 char em_driver_version[] = "7.6.1-k";
41 
42 /*********************************************************************
43  *  PCI Device ID Table
44  *
45  *  Used by probe to select devices to load on
46  *  Last field stores an index into e1000_strings
47  *  Last entry must be all 0s
48  *
49  *  { Vendor ID, Device ID, SubVendor ID, SubDevice ID, String Index }
50  *********************************************************************/
51 
52 static pci_vendor_info_t em_vendor_info_array[] =
53 {
54 	/* Intel(R) PRO/1000 Network Connection - Legacy em*/
55 	PVID(0x8086, E1000_DEV_ID_82540EM, "Intel(R) PRO/1000 Network Connection"),
56 	PVID(0x8086, E1000_DEV_ID_82540EM_LOM, "Intel(R) PRO/1000 Network Connection"),
57 	PVID(0x8086, E1000_DEV_ID_82540EP, "Intel(R) PRO/1000 Network Connection"),
58 	PVID(0x8086, E1000_DEV_ID_82540EP_LOM, "Intel(R) PRO/1000 Network Connection"),
59 	PVID(0x8086, E1000_DEV_ID_82540EP_LP, "Intel(R) PRO/1000 Network Connection"),
60 
61 	PVID(0x8086, E1000_DEV_ID_82541EI, "Intel(R) PRO/1000 Network Connection"),
62 	PVID(0x8086, E1000_DEV_ID_82541ER, "Intel(R) PRO/1000 Network Connection"),
63 	PVID(0x8086, E1000_DEV_ID_82541ER_LOM, "Intel(R) PRO/1000 Network Connection"),
64 	PVID(0x8086, E1000_DEV_ID_82541EI_MOBILE, "Intel(R) PRO/1000 Network Connection"),
65 	PVID(0x8086, E1000_DEV_ID_82541GI, "Intel(R) PRO/1000 Network Connection"),
66 	PVID(0x8086, E1000_DEV_ID_82541GI_LF, "Intel(R) PRO/1000 Network Connection"),
67 	PVID(0x8086, E1000_DEV_ID_82541GI_MOBILE, "Intel(R) PRO/1000 Network Connection"),
68 
69 	PVID(0x8086, E1000_DEV_ID_82542, "Intel(R) PRO/1000 Network Connection"),
70 
71 	PVID(0x8086, E1000_DEV_ID_82543GC_FIBER, "Intel(R) PRO/1000 Network Connection"),
72 	PVID(0x8086, E1000_DEV_ID_82543GC_COPPER, "Intel(R) PRO/1000 Network Connection"),
73 
74 	PVID(0x8086, E1000_DEV_ID_82544EI_COPPER, "Intel(R) PRO/1000 Network Connection"),
75 	PVID(0x8086, E1000_DEV_ID_82544EI_FIBER, "Intel(R) PRO/1000 Network Connection"),
76 	PVID(0x8086, E1000_DEV_ID_82544GC_COPPER, "Intel(R) PRO/1000 Network Connection"),
77 	PVID(0x8086, E1000_DEV_ID_82544GC_LOM, "Intel(R) PRO/1000 Network Connection"),
78 
79 	PVID(0x8086, E1000_DEV_ID_82545EM_COPPER, "Intel(R) PRO/1000 Network Connection"),
80 	PVID(0x8086, E1000_DEV_ID_82545EM_FIBER, "Intel(R) PRO/1000 Network Connection"),
81 	PVID(0x8086, E1000_DEV_ID_82545GM_COPPER, "Intel(R) PRO/1000 Network Connection"),
82 	PVID(0x8086, E1000_DEV_ID_82545GM_FIBER, "Intel(R) PRO/1000 Network Connection"),
83 	PVID(0x8086, E1000_DEV_ID_82545GM_SERDES, "Intel(R) PRO/1000 Network Connection"),
84 
85 	PVID(0x8086, E1000_DEV_ID_82546EB_COPPER, "Intel(R) PRO/1000 Network Connection"),
86 	PVID(0x8086, E1000_DEV_ID_82546EB_FIBER, "Intel(R) PRO/1000 Network Connection"),
87 	PVID(0x8086, E1000_DEV_ID_82546EB_QUAD_COPPER, "Intel(R) PRO/1000 Network Connection"),
88 	PVID(0x8086, E1000_DEV_ID_82546GB_COPPER, "Intel(R) PRO/1000 Network Connection"),
89 	PVID(0x8086, E1000_DEV_ID_82546GB_FIBER, "Intel(R) PRO/1000 Network Connection"),
90 	PVID(0x8086, E1000_DEV_ID_82546GB_SERDES, "Intel(R) PRO/1000 Network Connection"),
91 	PVID(0x8086, E1000_DEV_ID_82546GB_PCIE, "Intel(R) PRO/1000 Network Connection"),
92 	PVID(0x8086, E1000_DEV_ID_82546GB_QUAD_COPPER, "Intel(R) PRO/1000 Network Connection"),
93 	PVID(0x8086, E1000_DEV_ID_82546GB_QUAD_COPPER_KSP3, "Intel(R) PRO/1000 Network Connection"),
94 
95 	PVID(0x8086, E1000_DEV_ID_82547EI, "Intel(R) PRO/1000 Network Connection"),
96 	PVID(0x8086, E1000_DEV_ID_82547EI_MOBILE, "Intel(R) PRO/1000 Network Connection"),
97 	PVID(0x8086, E1000_DEV_ID_82547GI, "Intel(R) PRO/1000 Network Connection"),
98 
99 	/* Intel(R) PRO/1000 Network Connection - em */
100 	PVID(0x8086, E1000_DEV_ID_82571EB_COPPER, "Intel(R) PRO/1000 Network Connection"),
101 	PVID(0x8086, E1000_DEV_ID_82571EB_FIBER, "Intel(R) PRO/1000 Network Connection"),
102 	PVID(0x8086, E1000_DEV_ID_82571EB_SERDES, "Intel(R) PRO/1000 Network Connection"),
103 	PVID(0x8086, E1000_DEV_ID_82571EB_SERDES_DUAL, "Intel(R) PRO/1000 Network Connection"),
104 	PVID(0x8086, E1000_DEV_ID_82571EB_SERDES_QUAD, "Intel(R) PRO/1000 Network Connection"),
105 	PVID(0x8086, E1000_DEV_ID_82571EB_QUAD_COPPER, "Intel(R) PRO/1000 Network Connection"),
106 	PVID(0x8086, E1000_DEV_ID_82571EB_QUAD_COPPER_LP, "Intel(R) PRO/1000 Network Connection"),
107 	PVID(0x8086, E1000_DEV_ID_82571EB_QUAD_FIBER, "Intel(R) PRO/1000 Network Connection"),
108 	PVID(0x8086, E1000_DEV_ID_82571PT_QUAD_COPPER, "Intel(R) PRO/1000 Network Connection"),
109 	PVID(0x8086, E1000_DEV_ID_82572EI, "Intel(R) PRO/1000 Network Connection"),
110 	PVID(0x8086, E1000_DEV_ID_82572EI_COPPER, "Intel(R) PRO/1000 Network Connection"),
111 	PVID(0x8086, E1000_DEV_ID_82572EI_FIBER, "Intel(R) PRO/1000 Network Connection"),
112 	PVID(0x8086, E1000_DEV_ID_82572EI_SERDES, "Intel(R) PRO/1000 Network Connection"),
113 	PVID(0x8086, E1000_DEV_ID_82573E, "Intel(R) PRO/1000 Network Connection"),
114 	PVID(0x8086, E1000_DEV_ID_82573E_IAMT, "Intel(R) PRO/1000 Network Connection"),
115 	PVID(0x8086, E1000_DEV_ID_82573L, "Intel(R) PRO/1000 Network Connection"),
116 	PVID(0x8086, E1000_DEV_ID_82583V, "Intel(R) PRO/1000 Network Connection"),
117 	PVID(0x8086, E1000_DEV_ID_80003ES2LAN_COPPER_SPT, "Intel(R) PRO/1000 Network Connection"),
118 	PVID(0x8086, E1000_DEV_ID_80003ES2LAN_SERDES_SPT, "Intel(R) PRO/1000 Network Connection"),
119 	PVID(0x8086, E1000_DEV_ID_80003ES2LAN_COPPER_DPT, "Intel(R) PRO/1000 Network Connection"),
120 	PVID(0x8086, E1000_DEV_ID_80003ES2LAN_SERDES_DPT, "Intel(R) PRO/1000 Network Connection"),
121 	PVID(0x8086, E1000_DEV_ID_ICH8_IGP_M_AMT, "Intel(R) PRO/1000 Network Connection"),
122 	PVID(0x8086, E1000_DEV_ID_ICH8_IGP_AMT, "Intel(R) PRO/1000 Network Connection"),
123 	PVID(0x8086, E1000_DEV_ID_ICH8_IGP_C, "Intel(R) PRO/1000 Network Connection"),
124 	PVID(0x8086, E1000_DEV_ID_ICH8_IFE, "Intel(R) PRO/1000 Network Connection"),
125 	PVID(0x8086, E1000_DEV_ID_ICH8_IFE_GT, "Intel(R) PRO/1000 Network Connection"),
126 	PVID(0x8086, E1000_DEV_ID_ICH8_IFE_G, "Intel(R) PRO/1000 Network Connection"),
127 	PVID(0x8086, E1000_DEV_ID_ICH8_IGP_M, "Intel(R) PRO/1000 Network Connection"),
128 	PVID(0x8086, E1000_DEV_ID_ICH8_82567V_3, "Intel(R) PRO/1000 Network Connection"),
129 	PVID(0x8086, E1000_DEV_ID_ICH9_IGP_M_AMT, "Intel(R) PRO/1000 Network Connection"),
130 	PVID(0x8086, E1000_DEV_ID_ICH9_IGP_AMT, "Intel(R) PRO/1000 Network Connection"),
131 	PVID(0x8086, E1000_DEV_ID_ICH9_IGP_C, "Intel(R) PRO/1000 Network Connection"),
132 	PVID(0x8086, E1000_DEV_ID_ICH9_IGP_M, "Intel(R) PRO/1000 Network Connection"),
133 	PVID(0x8086, E1000_DEV_ID_ICH9_IGP_M_V, "Intel(R) PRO/1000 Network Connection"),
134 	PVID(0x8086, E1000_DEV_ID_ICH9_IFE, "Intel(R) PRO/1000 Network Connection"),
135 	PVID(0x8086, E1000_DEV_ID_ICH9_IFE_GT, "Intel(R) PRO/1000 Network Connection"),
136 	PVID(0x8086, E1000_DEV_ID_ICH9_IFE_G, "Intel(R) PRO/1000 Network Connection"),
137 	PVID(0x8086, E1000_DEV_ID_ICH9_BM, "Intel(R) PRO/1000 Network Connection"),
138 	PVID(0x8086, E1000_DEV_ID_82574L, "Intel(R) PRO/1000 Network Connection"),
139 	PVID(0x8086, E1000_DEV_ID_82574LA, "Intel(R) PRO/1000 Network Connection"),
140 	PVID(0x8086, E1000_DEV_ID_ICH10_R_BM_LM, "Intel(R) PRO/1000 Network Connection"),
141 	PVID(0x8086, E1000_DEV_ID_ICH10_R_BM_LF, "Intel(R) PRO/1000 Network Connection"),
142 	PVID(0x8086, E1000_DEV_ID_ICH10_R_BM_V, "Intel(R) PRO/1000 Network Connection"),
143 	PVID(0x8086, E1000_DEV_ID_ICH10_D_BM_LM, "Intel(R) PRO/1000 Network Connection"),
144 	PVID(0x8086, E1000_DEV_ID_ICH10_D_BM_LF, "Intel(R) PRO/1000 Network Connection"),
145 	PVID(0x8086, E1000_DEV_ID_ICH10_D_BM_V, "Intel(R) PRO/1000 Network Connection"),
146 	PVID(0x8086, E1000_DEV_ID_PCH_M_HV_LM, "Intel(R) PRO/1000 Network Connection"),
147 	PVID(0x8086, E1000_DEV_ID_PCH_M_HV_LC, "Intel(R) PRO/1000 Network Connection"),
148 	PVID(0x8086, E1000_DEV_ID_PCH_D_HV_DM, "Intel(R) PRO/1000 Network Connection"),
149 	PVID(0x8086, E1000_DEV_ID_PCH_D_HV_DC, "Intel(R) PRO/1000 Network Connection"),
150 	PVID(0x8086, E1000_DEV_ID_PCH2_LV_LM, "Intel(R) PRO/1000 Network Connection"),
151 	PVID(0x8086, E1000_DEV_ID_PCH2_LV_V, "Intel(R) PRO/1000 Network Connection"),
152 	PVID(0x8086, E1000_DEV_ID_PCH_LPT_I217_LM, "Intel(R) PRO/1000 Network Connection"),
153 	PVID(0x8086, E1000_DEV_ID_PCH_LPT_I217_V, "Intel(R) PRO/1000 Network Connection"),
154 	PVID(0x8086, E1000_DEV_ID_PCH_LPTLP_I218_LM, "Intel(R) PRO/1000 Network Connection"),
155 	PVID(0x8086, E1000_DEV_ID_PCH_LPTLP_I218_V, "Intel(R) PRO/1000 Network Connection"),
156 	PVID(0x8086, E1000_DEV_ID_PCH_I218_LM2, "Intel(R) PRO/1000 Network Connection"),
157 	PVID(0x8086, E1000_DEV_ID_PCH_I218_V2, "Intel(R) PRO/1000 Network Connection"),
158 	PVID(0x8086, E1000_DEV_ID_PCH_I218_LM3, "Intel(R) PRO/1000 Network Connection"),
159 	PVID(0x8086, E1000_DEV_ID_PCH_I218_V3, "Intel(R) PRO/1000 Network Connection"),
160 	PVID(0x8086, E1000_DEV_ID_PCH_SPT_I219_LM, "Intel(R) PRO/1000 Network Connection"),
161 	PVID(0x8086, E1000_DEV_ID_PCH_SPT_I219_V, "Intel(R) PRO/1000 Network Connection"),
162 	PVID(0x8086, E1000_DEV_ID_PCH_SPT_I219_LM2, "Intel(R) PRO/1000 Network Connection"),
163 	PVID(0x8086, E1000_DEV_ID_PCH_SPT_I219_V2, "Intel(R) PRO/1000 Network Connection"),
164 	PVID(0x8086, E1000_DEV_ID_PCH_LBG_I219_LM3, "Intel(R) PRO/1000 Network Connection"),
165 	PVID(0x8086, E1000_DEV_ID_PCH_SPT_I219_LM4, "Intel(R) PRO/1000 Network Connection"),
166 	PVID(0x8086, E1000_DEV_ID_PCH_SPT_I219_V4, "Intel(R) PRO/1000 Network Connection"),
167 	PVID(0x8086, E1000_DEV_ID_PCH_SPT_I219_LM5, "Intel(R) PRO/1000 Network Connection"),
168 	PVID(0x8086, E1000_DEV_ID_PCH_SPT_I219_V5, "Intel(R) PRO/1000 Network Connection"),
169 	PVID(0x8086, E1000_DEV_ID_PCH_CNP_I219_LM6, "Intel(R) PRO/1000 Network Connection"),
170 	PVID(0x8086, E1000_DEV_ID_PCH_CNP_I219_V6, "Intel(R) PRO/1000 Network Connection"),
171 	PVID(0x8086, E1000_DEV_ID_PCH_CNP_I219_LM7, "Intel(R) PRO/1000 Network Connection"),
172 	PVID(0x8086, E1000_DEV_ID_PCH_CNP_I219_V7, "Intel(R) PRO/1000 Network Connection"),
173 	PVID(0x8086, E1000_DEV_ID_PCH_ICP_I219_LM8, "Intel(R) PRO/1000 Network Connection"),
174 	PVID(0x8086, E1000_DEV_ID_PCH_ICP_I219_V8, "Intel(R) PRO/1000 Network Connection"),
175 	PVID(0x8086, E1000_DEV_ID_PCH_ICP_I219_LM9, "Intel(R) PRO/1000 Network Connection"),
176 	PVID(0x8086, E1000_DEV_ID_PCH_ICP_I219_V9, "Intel(R) PRO/1000 Network Connection"),
177 	PVID(0x8086, E1000_DEV_ID_PCH_ICP_I219_V10, "Intel(R) PRO/1000 Network Connection"),
178 	/* required last entry */
179 	PVID_END
180 };
181 
182 static pci_vendor_info_t igb_vendor_info_array[] =
183 {
184 	/* Intel(R) PRO/1000 Network Connection - igb */
185 	PVID(0x8086, E1000_DEV_ID_82575EB_COPPER, "Intel(R) PRO/1000 PCI-Express Network Driver"),
186 	PVID(0x8086, E1000_DEV_ID_82575EB_FIBER_SERDES, "Intel(R) PRO/1000 PCI-Express Network Driver"),
187 	PVID(0x8086, E1000_DEV_ID_82575GB_QUAD_COPPER, "Intel(R) PRO/1000 PCI-Express Network Driver"),
188 	PVID(0x8086, E1000_DEV_ID_82576, "Intel(R) PRO/1000 PCI-Express Network Driver"),
189 	PVID(0x8086, E1000_DEV_ID_82576_NS, "Intel(R) PRO/1000 PCI-Express Network Driver"),
190 	PVID(0x8086, E1000_DEV_ID_82576_NS_SERDES, "Intel(R) PRO/1000 PCI-Express Network Driver"),
191 	PVID(0x8086, E1000_DEV_ID_82576_FIBER, "Intel(R) PRO/1000 PCI-Express Network Driver"),
192 	PVID(0x8086, E1000_DEV_ID_82576_SERDES, "Intel(R) PRO/1000 PCI-Express Network Driver"),
193 	PVID(0x8086, E1000_DEV_ID_82576_SERDES_QUAD, "Intel(R) PRO/1000 PCI-Express Network Driver"),
194 	PVID(0x8086, E1000_DEV_ID_82576_QUAD_COPPER, "Intel(R) PRO/1000 PCI-Express Network Driver"),
195 	PVID(0x8086, E1000_DEV_ID_82576_QUAD_COPPER_ET2, "Intel(R) PRO/1000 PCI-Express Network Driver"),
196 	PVID(0x8086, E1000_DEV_ID_82576_VF, "Intel(R) PRO/1000 PCI-Express Network Driver"),
197 	PVID(0x8086, E1000_DEV_ID_82580_COPPER, "Intel(R) PRO/1000 PCI-Express Network Driver"),
198 	PVID(0x8086, E1000_DEV_ID_82580_FIBER, "Intel(R) PRO/1000 PCI-Express Network Driver"),
199 	PVID(0x8086, E1000_DEV_ID_82580_SERDES, "Intel(R) PRO/1000 PCI-Express Network Driver"),
200 	PVID(0x8086, E1000_DEV_ID_82580_SGMII, "Intel(R) PRO/1000 PCI-Express Network Driver"),
201 	PVID(0x8086, E1000_DEV_ID_82580_COPPER_DUAL, "Intel(R) PRO/1000 PCI-Express Network Driver"),
202 	PVID(0x8086, E1000_DEV_ID_82580_QUAD_FIBER, "Intel(R) PRO/1000 PCI-Express Network Driver"),
203 	PVID(0x8086, E1000_DEV_ID_DH89XXCC_SERDES, "Intel(R) PRO/1000 PCI-Express Network Driver"),
204 	PVID(0x8086, E1000_DEV_ID_DH89XXCC_SGMII, "Intel(R) PRO/1000 PCI-Express Network Driver"),
205 	PVID(0x8086, E1000_DEV_ID_DH89XXCC_SFP, "Intel(R) PRO/1000 PCI-Express Network Driver"),
206 	PVID(0x8086, E1000_DEV_ID_DH89XXCC_BACKPLANE, "Intel(R) PRO/1000 PCI-Express Network Driver"),
207 	PVID(0x8086, E1000_DEV_ID_I350_COPPER, "Intel(R) PRO/1000 PCI-Express Network Driver"),
208 	PVID(0x8086, E1000_DEV_ID_I350_FIBER, "Intel(R) PRO/1000 PCI-Express Network Driver"),
209 	PVID(0x8086, E1000_DEV_ID_I350_SERDES, "Intel(R) PRO/1000 PCI-Express Network Driver"),
210 	PVID(0x8086, E1000_DEV_ID_I350_SGMII, "Intel(R) PRO/1000 PCI-Express Network Driver"),
211 	PVID(0x8086, E1000_DEV_ID_I350_VF, "Intel(R) PRO/1000 PCI-Express Network Driver"),
212 	PVID(0x8086, E1000_DEV_ID_I210_COPPER, "Intel(R) PRO/1000 PCI-Express Network Driver"),
213 	PVID(0x8086, E1000_DEV_ID_I210_COPPER_IT, "Intel(R) PRO/1000 PCI-Express Network Driver"),
214 	PVID(0x8086, E1000_DEV_ID_I210_COPPER_OEM1, "Intel(R) PRO/1000 PCI-Express Network Driver"),
215 	PVID(0x8086, E1000_DEV_ID_I210_COPPER_FLASHLESS, "Intel(R) PRO/1000 PCI-Express Network Driver"),
216 	PVID(0x8086, E1000_DEV_ID_I210_SERDES_FLASHLESS, "Intel(R) PRO/1000 PCI-Express Network Driver"),
217 	PVID(0x8086, E1000_DEV_ID_I210_FIBER, "Intel(R) PRO/1000 PCI-Express Network Driver"),
218 	PVID(0x8086, E1000_DEV_ID_I210_SERDES, "Intel(R) PRO/1000 PCI-Express Network Driver"),
219 	PVID(0x8086, E1000_DEV_ID_I210_SGMII, "Intel(R) PRO/1000 PCI-Express Network Driver"),
220 	PVID(0x8086, E1000_DEV_ID_I211_COPPER, "Intel(R) PRO/1000 PCI-Express Network Driver"),
221 	PVID(0x8086, E1000_DEV_ID_I354_BACKPLANE_1GBPS, "Intel(R) PRO/1000 PCI-Express Network Driver"),
222 	PVID(0x8086, E1000_DEV_ID_I354_BACKPLANE_2_5GBPS, "Intel(R) PRO/1000 PCI-Express Network Driver"),
223 	PVID(0x8086, E1000_DEV_ID_I354_SGMII, "Intel(R) PRO/1000 PCI-Express Network Driver"),
224 	/* required last entry */
225 	PVID_END
226 };
227 
228 /*********************************************************************
229  *  Function prototypes
230  *********************************************************************/
231 static void	*em_register(device_t dev);
232 static void	*igb_register(device_t dev);
233 static int	em_if_attach_pre(if_ctx_t ctx);
234 static int	em_if_attach_post(if_ctx_t ctx);
235 static int	em_if_detach(if_ctx_t ctx);
236 static int	em_if_shutdown(if_ctx_t ctx);
237 static int	em_if_suspend(if_ctx_t ctx);
238 static int	em_if_resume(if_ctx_t ctx);
239 
240 static int	em_if_tx_queues_alloc(if_ctx_t ctx, caddr_t *vaddrs, uint64_t *paddrs, int ntxqs, int ntxqsets);
241 static int	em_if_rx_queues_alloc(if_ctx_t ctx, caddr_t *vaddrs, uint64_t *paddrs, int nrxqs, int nrxqsets);
242 static void	em_if_queues_free(if_ctx_t ctx);
243 
244 static uint64_t	em_if_get_counter(if_ctx_t, ift_counter);
245 static void	em_if_init(if_ctx_t ctx);
246 static void	em_if_stop(if_ctx_t ctx);
247 static void	em_if_media_status(if_ctx_t, struct ifmediareq *);
248 static int	em_if_media_change(if_ctx_t ctx);
249 static int	em_if_mtu_set(if_ctx_t ctx, uint32_t mtu);
250 static void	em_if_timer(if_ctx_t ctx, uint16_t qid);
251 static void	em_if_vlan_register(if_ctx_t ctx, u16 vtag);
252 static void	em_if_vlan_unregister(if_ctx_t ctx, u16 vtag);
253 static void	em_if_watchdog_reset(if_ctx_t ctx);
254 
255 static void	em_identify_hardware(if_ctx_t ctx);
256 static int	em_allocate_pci_resources(if_ctx_t ctx);
257 static void	em_free_pci_resources(if_ctx_t ctx);
258 static void	em_reset(if_ctx_t ctx);
259 static int	em_setup_interface(if_ctx_t ctx);
260 static int	em_setup_msix(if_ctx_t ctx);
261 
262 static void	em_initialize_transmit_unit(if_ctx_t ctx);
263 static void	em_initialize_receive_unit(if_ctx_t ctx);
264 
265 static void	em_if_intr_enable(if_ctx_t ctx);
266 static void	em_if_intr_disable(if_ctx_t ctx);
267 static void	igb_if_intr_enable(if_ctx_t ctx);
268 static void	igb_if_intr_disable(if_ctx_t ctx);
269 static int	em_if_rx_queue_intr_enable(if_ctx_t ctx, uint16_t rxqid);
270 static int	em_if_tx_queue_intr_enable(if_ctx_t ctx, uint16_t txqid);
271 static int	igb_if_rx_queue_intr_enable(if_ctx_t ctx, uint16_t rxqid);
272 static int	igb_if_tx_queue_intr_enable(if_ctx_t ctx, uint16_t txqid);
273 static void	em_if_multi_set(if_ctx_t ctx);
274 static void	em_if_update_admin_status(if_ctx_t ctx);
275 static void	em_if_debug(if_ctx_t ctx);
276 static void	em_update_stats_counters(struct adapter *);
277 static void	em_add_hw_stats(struct adapter *adapter);
278 static int	em_if_set_promisc(if_ctx_t ctx, int flags);
279 static void	em_setup_vlan_hw_support(struct adapter *);
280 static int	em_sysctl_nvm_info(SYSCTL_HANDLER_ARGS);
281 static void	em_print_nvm_info(struct adapter *);
282 static int	em_sysctl_debug_info(SYSCTL_HANDLER_ARGS);
283 static int	em_get_rs(SYSCTL_HANDLER_ARGS);
284 static void	em_print_debug_info(struct adapter *);
285 static int 	em_is_valid_ether_addr(u8 *);
286 static int	em_sysctl_int_delay(SYSCTL_HANDLER_ARGS);
287 static void	em_add_int_delay_sysctl(struct adapter *, const char *,
288 		    const char *, struct em_int_delay_info *, int, int);
289 /* Management and WOL Support */
290 static void	em_init_manageability(struct adapter *);
291 static void	em_release_manageability(struct adapter *);
292 static void	em_get_hw_control(struct adapter *);
293 static void	em_release_hw_control(struct adapter *);
294 static void	em_get_wakeup(if_ctx_t ctx);
295 static void	em_enable_wakeup(if_ctx_t ctx);
296 static int	em_enable_phy_wakeup(struct adapter *);
297 static void	em_disable_aspm(struct adapter *);
298 
299 int		em_intr(void *arg);
300 static void	em_disable_promisc(if_ctx_t ctx);
301 
302 /* MSI-X handlers */
303 static int	em_if_msix_intr_assign(if_ctx_t, int);
304 static int	em_msix_link(void *);
305 static void	em_handle_link(void *context);
306 
307 static void	em_enable_vectors_82574(if_ctx_t);
308 
309 static int	em_set_flowcntl(SYSCTL_HANDLER_ARGS);
310 static int	em_sysctl_eee(SYSCTL_HANDLER_ARGS);
311 static void	em_if_led_func(if_ctx_t ctx, int onoff);
312 
313 static int	em_get_regs(SYSCTL_HANDLER_ARGS);
314 
315 static void	lem_smartspeed(struct adapter *adapter);
316 static void	igb_configure_queues(struct adapter *adapter);
317 
318 
319 /*********************************************************************
320  *  FreeBSD Device Interface Entry Points
321  *********************************************************************/
322 static device_method_t em_methods[] = {
323 	/* Device interface */
324 	DEVMETHOD(device_register, em_register),
325 	DEVMETHOD(device_probe, iflib_device_probe),
326 	DEVMETHOD(device_attach, iflib_device_attach),
327 	DEVMETHOD(device_detach, iflib_device_detach),
328 	DEVMETHOD(device_shutdown, iflib_device_shutdown),
329 	DEVMETHOD(device_suspend, iflib_device_suspend),
330 	DEVMETHOD(device_resume, iflib_device_resume),
331 	DEVMETHOD_END
332 };
333 
334 static device_method_t igb_methods[] = {
335 	/* Device interface */
336 	DEVMETHOD(device_register, igb_register),
337 	DEVMETHOD(device_probe, iflib_device_probe),
338 	DEVMETHOD(device_attach, iflib_device_attach),
339 	DEVMETHOD(device_detach, iflib_device_detach),
340 	DEVMETHOD(device_shutdown, iflib_device_shutdown),
341 	DEVMETHOD(device_suspend, iflib_device_suspend),
342 	DEVMETHOD(device_resume, iflib_device_resume),
343 	DEVMETHOD_END
344 };
345 
346 
347 static driver_t em_driver = {
348 	"em", em_methods, sizeof(struct adapter),
349 };
350 
351 static devclass_t em_devclass;
352 DRIVER_MODULE(em, pci, em_driver, em_devclass, 0, 0);
353 
354 MODULE_DEPEND(em, pci, 1, 1, 1);
355 MODULE_DEPEND(em, ether, 1, 1, 1);
356 MODULE_DEPEND(em, iflib, 1, 1, 1);
357 
358 IFLIB_PNP_INFO(pci, em, em_vendor_info_array);
359 
360 static driver_t igb_driver = {
361 	"igb", igb_methods, sizeof(struct adapter),
362 };
363 
364 static devclass_t igb_devclass;
365 DRIVER_MODULE(igb, pci, igb_driver, igb_devclass, 0, 0);
366 
367 MODULE_DEPEND(igb, pci, 1, 1, 1);
368 MODULE_DEPEND(igb, ether, 1, 1, 1);
369 MODULE_DEPEND(igb, iflib, 1, 1, 1);
370 
371 IFLIB_PNP_INFO(pci, igb, igb_vendor_info_array);
372 
373 static device_method_t em_if_methods[] = {
374 	DEVMETHOD(ifdi_attach_pre, em_if_attach_pre),
375 	DEVMETHOD(ifdi_attach_post, em_if_attach_post),
376 	DEVMETHOD(ifdi_detach, em_if_detach),
377 	DEVMETHOD(ifdi_shutdown, em_if_shutdown),
378 	DEVMETHOD(ifdi_suspend, em_if_suspend),
379 	DEVMETHOD(ifdi_resume, em_if_resume),
380 	DEVMETHOD(ifdi_init, em_if_init),
381 	DEVMETHOD(ifdi_stop, em_if_stop),
382 	DEVMETHOD(ifdi_msix_intr_assign, em_if_msix_intr_assign),
383 	DEVMETHOD(ifdi_intr_enable, em_if_intr_enable),
384 	DEVMETHOD(ifdi_intr_disable, em_if_intr_disable),
385 	DEVMETHOD(ifdi_tx_queues_alloc, em_if_tx_queues_alloc),
386 	DEVMETHOD(ifdi_rx_queues_alloc, em_if_rx_queues_alloc),
387 	DEVMETHOD(ifdi_queues_free, em_if_queues_free),
388 	DEVMETHOD(ifdi_update_admin_status, em_if_update_admin_status),
389 	DEVMETHOD(ifdi_multi_set, em_if_multi_set),
390 	DEVMETHOD(ifdi_media_status, em_if_media_status),
391 	DEVMETHOD(ifdi_media_change, em_if_media_change),
392 	DEVMETHOD(ifdi_mtu_set, em_if_mtu_set),
393 	DEVMETHOD(ifdi_promisc_set, em_if_set_promisc),
394 	DEVMETHOD(ifdi_timer, em_if_timer),
395 	DEVMETHOD(ifdi_watchdog_reset, em_if_watchdog_reset),
396 	DEVMETHOD(ifdi_vlan_register, em_if_vlan_register),
397 	DEVMETHOD(ifdi_vlan_unregister, em_if_vlan_unregister),
398 	DEVMETHOD(ifdi_get_counter, em_if_get_counter),
399 	DEVMETHOD(ifdi_led_func, em_if_led_func),
400 	DEVMETHOD(ifdi_rx_queue_intr_enable, em_if_rx_queue_intr_enable),
401 	DEVMETHOD(ifdi_tx_queue_intr_enable, em_if_tx_queue_intr_enable),
402 	DEVMETHOD(ifdi_debug, em_if_debug),
403 	DEVMETHOD_END
404 };
405 
406 static driver_t em_if_driver = {
407 	"em_if", em_if_methods, sizeof(struct adapter)
408 };
409 
410 static device_method_t igb_if_methods[] = {
411 	DEVMETHOD(ifdi_attach_pre, em_if_attach_pre),
412 	DEVMETHOD(ifdi_attach_post, em_if_attach_post),
413 	DEVMETHOD(ifdi_detach, em_if_detach),
414 	DEVMETHOD(ifdi_shutdown, em_if_shutdown),
415 	DEVMETHOD(ifdi_suspend, em_if_suspend),
416 	DEVMETHOD(ifdi_resume, em_if_resume),
417 	DEVMETHOD(ifdi_init, em_if_init),
418 	DEVMETHOD(ifdi_stop, em_if_stop),
419 	DEVMETHOD(ifdi_msix_intr_assign, em_if_msix_intr_assign),
420 	DEVMETHOD(ifdi_intr_enable, igb_if_intr_enable),
421 	DEVMETHOD(ifdi_intr_disable, igb_if_intr_disable),
422 	DEVMETHOD(ifdi_tx_queues_alloc, em_if_tx_queues_alloc),
423 	DEVMETHOD(ifdi_rx_queues_alloc, em_if_rx_queues_alloc),
424 	DEVMETHOD(ifdi_queues_free, em_if_queues_free),
425 	DEVMETHOD(ifdi_update_admin_status, em_if_update_admin_status),
426 	DEVMETHOD(ifdi_multi_set, em_if_multi_set),
427 	DEVMETHOD(ifdi_media_status, em_if_media_status),
428 	DEVMETHOD(ifdi_media_change, em_if_media_change),
429 	DEVMETHOD(ifdi_mtu_set, em_if_mtu_set),
430 	DEVMETHOD(ifdi_promisc_set, em_if_set_promisc),
431 	DEVMETHOD(ifdi_timer, em_if_timer),
432 	DEVMETHOD(ifdi_watchdog_reset, em_if_watchdog_reset),
433 	DEVMETHOD(ifdi_vlan_register, em_if_vlan_register),
434 	DEVMETHOD(ifdi_vlan_unregister, em_if_vlan_unregister),
435 	DEVMETHOD(ifdi_get_counter, em_if_get_counter),
436 	DEVMETHOD(ifdi_led_func, em_if_led_func),
437 	DEVMETHOD(ifdi_rx_queue_intr_enable, igb_if_rx_queue_intr_enable),
438 	DEVMETHOD(ifdi_tx_queue_intr_enable, igb_if_tx_queue_intr_enable),
439 	DEVMETHOD(ifdi_debug, em_if_debug),
440 	DEVMETHOD_END
441 };
442 
443 static driver_t igb_if_driver = {
444 	"igb_if", igb_if_methods, sizeof(struct adapter)
445 };
446 
447 /*********************************************************************
448  *  Tunable default values.
449  *********************************************************************/
450 
451 #define EM_TICKS_TO_USECS(ticks)	((1024 * (ticks) + 500) / 1000)
452 #define EM_USECS_TO_TICKS(usecs)	((1000 * (usecs) + 512) / 1024)
453 
454 #define MAX_INTS_PER_SEC	8000
455 #define DEFAULT_ITR		(1000000000/(MAX_INTS_PER_SEC * 256))
456 
457 /* Allow common code without TSO */
458 #ifndef CSUM_TSO
459 #define CSUM_TSO	0
460 #endif
461 
462 static SYSCTL_NODE(_hw, OID_AUTO, em, CTLFLAG_RD, 0, "EM driver parameters");
463 
464 static int em_disable_crc_stripping = 0;
465 SYSCTL_INT(_hw_em, OID_AUTO, disable_crc_stripping, CTLFLAG_RDTUN,
466     &em_disable_crc_stripping, 0, "Disable CRC Stripping");
467 
468 static int em_tx_int_delay_dflt = EM_TICKS_TO_USECS(EM_TIDV);
469 static int em_rx_int_delay_dflt = EM_TICKS_TO_USECS(EM_RDTR);
470 SYSCTL_INT(_hw_em, OID_AUTO, tx_int_delay, CTLFLAG_RDTUN, &em_tx_int_delay_dflt,
471     0, "Default transmit interrupt delay in usecs");
472 SYSCTL_INT(_hw_em, OID_AUTO, rx_int_delay, CTLFLAG_RDTUN, &em_rx_int_delay_dflt,
473     0, "Default receive interrupt delay in usecs");
474 
475 static int em_tx_abs_int_delay_dflt = EM_TICKS_TO_USECS(EM_TADV);
476 static int em_rx_abs_int_delay_dflt = EM_TICKS_TO_USECS(EM_RADV);
477 SYSCTL_INT(_hw_em, OID_AUTO, tx_abs_int_delay, CTLFLAG_RDTUN,
478     &em_tx_abs_int_delay_dflt, 0,
479     "Default transmit interrupt delay limit in usecs");
480 SYSCTL_INT(_hw_em, OID_AUTO, rx_abs_int_delay, CTLFLAG_RDTUN,
481     &em_rx_abs_int_delay_dflt, 0,
482     "Default receive interrupt delay limit in usecs");
483 
484 static int em_smart_pwr_down = FALSE;
485 SYSCTL_INT(_hw_em, OID_AUTO, smart_pwr_down, CTLFLAG_RDTUN, &em_smart_pwr_down,
486     0, "Set to true to leave smart power down enabled on newer adapters");
487 
488 /* Controls whether promiscuous also shows bad packets */
489 static int em_debug_sbp = TRUE;
490 SYSCTL_INT(_hw_em, OID_AUTO, sbp, CTLFLAG_RDTUN, &em_debug_sbp, 0,
491     "Show bad packets in promiscuous mode");
492 
493 /* How many packets rxeof tries to clean at a time */
494 static int em_rx_process_limit = 100;
495 SYSCTL_INT(_hw_em, OID_AUTO, rx_process_limit, CTLFLAG_RDTUN,
496     &em_rx_process_limit, 0,
497     "Maximum number of received packets to process "
498     "at a time, -1 means unlimited");
499 
500 /* Energy efficient ethernet - default to OFF */
501 static int eee_setting = 1;
502 SYSCTL_INT(_hw_em, OID_AUTO, eee_setting, CTLFLAG_RDTUN, &eee_setting, 0,
503     "Enable Energy Efficient Ethernet");
504 
505 /*
506 ** Tuneable Interrupt rate
507 */
508 static int em_max_interrupt_rate = 8000;
509 SYSCTL_INT(_hw_em, OID_AUTO, max_interrupt_rate, CTLFLAG_RDTUN,
510     &em_max_interrupt_rate, 0, "Maximum interrupts per second");
511 
512 
513 
514 /* Global used in WOL setup with multiport cards */
515 static int global_quad_port_a = 0;
516 
517 extern struct if_txrx igb_txrx;
518 extern struct if_txrx em_txrx;
519 extern struct if_txrx lem_txrx;
520 
521 static struct if_shared_ctx em_sctx_init = {
522 	.isc_magic = IFLIB_MAGIC,
523 	.isc_q_align = PAGE_SIZE,
524 	.isc_tx_maxsize = EM_TSO_SIZE + sizeof(struct ether_vlan_header),
525 	.isc_tx_maxsegsize = PAGE_SIZE,
526 	.isc_tso_maxsize = EM_TSO_SIZE + sizeof(struct ether_vlan_header),
527 	.isc_tso_maxsegsize = EM_TSO_SEG_SIZE,
528 	.isc_rx_maxsize = MJUM9BYTES,
529 	.isc_rx_nsegments = 1,
530 	.isc_rx_maxsegsize = MJUM9BYTES,
531 	.isc_nfl = 1,
532 	.isc_nrxqs = 1,
533 	.isc_ntxqs = 1,
534 	.isc_admin_intrcnt = 1,
535 	.isc_vendor_info = em_vendor_info_array,
536 	.isc_driver_version = em_driver_version,
537 	.isc_driver = &em_if_driver,
538 	.isc_flags = IFLIB_NEED_SCRATCH | IFLIB_TSO_INIT_IP | IFLIB_NEED_ZERO_CSUM,
539 
540 	.isc_nrxd_min = {EM_MIN_RXD},
541 	.isc_ntxd_min = {EM_MIN_TXD},
542 	.isc_nrxd_max = {EM_MAX_RXD},
543 	.isc_ntxd_max = {EM_MAX_TXD},
544 	.isc_nrxd_default = {EM_DEFAULT_RXD},
545 	.isc_ntxd_default = {EM_DEFAULT_TXD},
546 };
547 
548 if_shared_ctx_t em_sctx = &em_sctx_init;
549 
550 static struct if_shared_ctx igb_sctx_init = {
551 	.isc_magic = IFLIB_MAGIC,
552 	.isc_q_align = PAGE_SIZE,
553 	.isc_tx_maxsize = EM_TSO_SIZE + sizeof(struct ether_vlan_header),
554 	.isc_tx_maxsegsize = PAGE_SIZE,
555 	.isc_tso_maxsize = EM_TSO_SIZE + sizeof(struct ether_vlan_header),
556 	.isc_tso_maxsegsize = EM_TSO_SEG_SIZE,
557 	.isc_rx_maxsize = MJUM9BYTES,
558 	.isc_rx_nsegments = 1,
559 	.isc_rx_maxsegsize = MJUM9BYTES,
560 	.isc_nfl = 1,
561 	.isc_nrxqs = 1,
562 	.isc_ntxqs = 1,
563 	.isc_admin_intrcnt = 1,
564 	.isc_vendor_info = igb_vendor_info_array,
565 	.isc_driver_version = em_driver_version,
566 	.isc_driver = &igb_if_driver,
567 	.isc_flags = IFLIB_NEED_SCRATCH | IFLIB_TSO_INIT_IP | IFLIB_NEED_ZERO_CSUM,
568 
569 	.isc_nrxd_min = {EM_MIN_RXD},
570 	.isc_ntxd_min = {EM_MIN_TXD},
571 	.isc_nrxd_max = {IGB_MAX_RXD},
572 	.isc_ntxd_max = {IGB_MAX_TXD},
573 	.isc_nrxd_default = {EM_DEFAULT_RXD},
574 	.isc_ntxd_default = {EM_DEFAULT_TXD},
575 };
576 
577 if_shared_ctx_t igb_sctx = &igb_sctx_init;
578 
579 /*****************************************************************
580  *
581  * Dump Registers
582  *
583  ****************************************************************/
584 #define IGB_REGS_LEN 739
585 
586 static int em_get_regs(SYSCTL_HANDLER_ARGS)
587 {
588 	struct adapter *adapter = (struct adapter *)arg1;
589 	struct e1000_hw *hw = &adapter->hw;
590 	struct sbuf *sb;
591 	u32 *regs_buff;
592 	int rc;
593 
594 	regs_buff = malloc(sizeof(u32) * IGB_REGS_LEN, M_DEVBUF, M_WAITOK);
595 	memset(regs_buff, 0, IGB_REGS_LEN * sizeof(u32));
596 
597 	rc = sysctl_wire_old_buffer(req, 0);
598 	MPASS(rc == 0);
599 	if (rc != 0) {
600 		free(regs_buff, M_DEVBUF);
601 		return (rc);
602 	}
603 
604 	sb = sbuf_new_for_sysctl(NULL, NULL, 32*400, req);
605 	MPASS(sb != NULL);
606 	if (sb == NULL) {
607 		free(regs_buff, M_DEVBUF);
608 		return (ENOMEM);
609 	}
610 
611 	/* General Registers */
612 	regs_buff[0] = E1000_READ_REG(hw, E1000_CTRL);
613 	regs_buff[1] = E1000_READ_REG(hw, E1000_STATUS);
614 	regs_buff[2] = E1000_READ_REG(hw, E1000_CTRL_EXT);
615 	regs_buff[3] = E1000_READ_REG(hw, E1000_ICR);
616 	regs_buff[4] = E1000_READ_REG(hw, E1000_RCTL);
617 	regs_buff[5] = E1000_READ_REG(hw, E1000_RDLEN(0));
618 	regs_buff[6] = E1000_READ_REG(hw, E1000_RDH(0));
619 	regs_buff[7] = E1000_READ_REG(hw, E1000_RDT(0));
620 	regs_buff[8] = E1000_READ_REG(hw, E1000_RXDCTL(0));
621 	regs_buff[9] = E1000_READ_REG(hw, E1000_RDBAL(0));
622 	regs_buff[10] = E1000_READ_REG(hw, E1000_RDBAH(0));
623 	regs_buff[11] = E1000_READ_REG(hw, E1000_TCTL);
624 	regs_buff[12] = E1000_READ_REG(hw, E1000_TDBAL(0));
625 	regs_buff[13] = E1000_READ_REG(hw, E1000_TDBAH(0));
626 	regs_buff[14] = E1000_READ_REG(hw, E1000_TDLEN(0));
627 	regs_buff[15] = E1000_READ_REG(hw, E1000_TDH(0));
628 	regs_buff[16] = E1000_READ_REG(hw, E1000_TDT(0));
629 	regs_buff[17] = E1000_READ_REG(hw, E1000_TXDCTL(0));
630 	regs_buff[18] = E1000_READ_REG(hw, E1000_TDFH);
631 	regs_buff[19] = E1000_READ_REG(hw, E1000_TDFT);
632 	regs_buff[20] = E1000_READ_REG(hw, E1000_TDFHS);
633 	regs_buff[21] = E1000_READ_REG(hw, E1000_TDFPC);
634 
635 	sbuf_printf(sb, "General Registers\n");
636 	sbuf_printf(sb, "\tCTRL\t %08x\n", regs_buff[0]);
637 	sbuf_printf(sb, "\tSTATUS\t %08x\n", regs_buff[1]);
638 	sbuf_printf(sb, "\tCTRL_EXIT\t %08x\n\n", regs_buff[2]);
639 
640 	sbuf_printf(sb, "Interrupt Registers\n");
641 	sbuf_printf(sb, "\tICR\t %08x\n\n", regs_buff[3]);
642 
643 	sbuf_printf(sb, "RX Registers\n");
644 	sbuf_printf(sb, "\tRCTL\t %08x\n", regs_buff[4]);
645 	sbuf_printf(sb, "\tRDLEN\t %08x\n", regs_buff[5]);
646 	sbuf_printf(sb, "\tRDH\t %08x\n", regs_buff[6]);
647 	sbuf_printf(sb, "\tRDT\t %08x\n", regs_buff[7]);
648 	sbuf_printf(sb, "\tRXDCTL\t %08x\n", regs_buff[8]);
649 	sbuf_printf(sb, "\tRDBAL\t %08x\n", regs_buff[9]);
650 	sbuf_printf(sb, "\tRDBAH\t %08x\n\n", regs_buff[10]);
651 
652 	sbuf_printf(sb, "TX Registers\n");
653 	sbuf_printf(sb, "\tTCTL\t %08x\n", regs_buff[11]);
654 	sbuf_printf(sb, "\tTDBAL\t %08x\n", regs_buff[12]);
655 	sbuf_printf(sb, "\tTDBAH\t %08x\n", regs_buff[13]);
656 	sbuf_printf(sb, "\tTDLEN\t %08x\n", regs_buff[14]);
657 	sbuf_printf(sb, "\tTDH\t %08x\n", regs_buff[15]);
658 	sbuf_printf(sb, "\tTDT\t %08x\n", regs_buff[16]);
659 	sbuf_printf(sb, "\tTXDCTL\t %08x\n", regs_buff[17]);
660 	sbuf_printf(sb, "\tTDFH\t %08x\n", regs_buff[18]);
661 	sbuf_printf(sb, "\tTDFT\t %08x\n", regs_buff[19]);
662 	sbuf_printf(sb, "\tTDFHS\t %08x\n", regs_buff[20]);
663 	sbuf_printf(sb, "\tTDFPC\t %08x\n\n", regs_buff[21]);
664 
665 	free(regs_buff, M_DEVBUF);
666 
667 #ifdef DUMP_DESCS
668 	{
669 		if_softc_ctx_t scctx = adapter->shared;
670 		struct rx_ring *rxr = &rx_que->rxr;
671 		struct tx_ring *txr = &tx_que->txr;
672 		int ntxd = scctx->isc_ntxd[0];
673 		int nrxd = scctx->isc_nrxd[0];
674 		int j;
675 
676 	for (j = 0; j < nrxd; j++) {
677 		u32 staterr = le32toh(rxr->rx_base[j].wb.upper.status_error);
678 		u32 length =  le32toh(rxr->rx_base[j].wb.upper.length);
679 		sbuf_printf(sb, "\tReceive Descriptor Address %d: %08" PRIx64 "  Error:%d  Length:%d\n", j, rxr->rx_base[j].read.buffer_addr, staterr, length);
680 	}
681 
682 	for (j = 0; j < min(ntxd, 256); j++) {
683 		unsigned int *ptr = (unsigned int *)&txr->tx_base[j];
684 
685 		sbuf_printf(sb, "\tTXD[%03d] [0]: %08x [1]: %08x [2]: %08x [3]: %08x  eop: %d DD=%d\n",
686 			    j, ptr[0], ptr[1], ptr[2], ptr[3], buf->eop,
687 			    buf->eop != -1 ? txr->tx_base[buf->eop].upper.fields.status & E1000_TXD_STAT_DD : 0);
688 
689 	}
690 	}
691 #endif
692 
693 	rc = sbuf_finish(sb);
694 	sbuf_delete(sb);
695 	return(rc);
696 }
697 
698 static void *
699 em_register(device_t dev)
700 {
701 	return (em_sctx);
702 }
703 
704 static void *
705 igb_register(device_t dev)
706 {
707 	return (igb_sctx);
708 }
709 
710 static int
711 em_set_num_queues(if_ctx_t ctx)
712 {
713 	struct adapter *adapter = iflib_get_softc(ctx);
714 	int maxqueues;
715 
716 	/* Sanity check based on HW */
717 	switch (adapter->hw.mac.type) {
718 	case e1000_82576:
719 	case e1000_82580:
720 	case e1000_i350:
721 	case e1000_i354:
722 		maxqueues = 8;
723 		break;
724 	case e1000_i210:
725 	case e1000_82575:
726 		maxqueues = 4;
727 		break;
728 	case e1000_i211:
729 	case e1000_82574:
730 		maxqueues = 2;
731 		break;
732 	default:
733 		maxqueues = 1;
734 		break;
735 	}
736 
737 	return (maxqueues);
738 }
739 
740 #define	LEM_CAPS							\
741     IFCAP_HWCSUM | IFCAP_VLAN_MTU | IFCAP_VLAN_HWTAGGING |		\
742     IFCAP_VLAN_HWCSUM | IFCAP_WOL | IFCAP_VLAN_HWFILTER
743 
744 #define	EM_CAPS								\
745     IFCAP_HWCSUM | IFCAP_VLAN_MTU | IFCAP_VLAN_HWTAGGING |		\
746     IFCAP_VLAN_HWCSUM | IFCAP_WOL | IFCAP_VLAN_HWFILTER | IFCAP_TSO4 |	\
747     IFCAP_LRO | IFCAP_VLAN_HWTSO
748 
749 #define	IGB_CAPS							\
750     IFCAP_HWCSUM | IFCAP_VLAN_MTU | IFCAP_VLAN_HWTAGGING |		\
751     IFCAP_VLAN_HWCSUM | IFCAP_WOL | IFCAP_VLAN_HWFILTER | IFCAP_TSO4 |	\
752     IFCAP_LRO | IFCAP_VLAN_HWTSO | IFCAP_JUMBO_MTU | IFCAP_HWCSUM_IPV6 |\
753     IFCAP_TSO6
754 
755 /*********************************************************************
756  *  Device initialization routine
757  *
758  *  The attach entry point is called when the driver is being loaded.
759  *  This routine identifies the type of hardware, allocates all resources
760  *  and initializes the hardware.
761  *
762  *  return 0 on success, positive on failure
763  *********************************************************************/
764 static int
765 em_if_attach_pre(if_ctx_t ctx)
766 {
767 	struct adapter *adapter;
768 	if_softc_ctx_t scctx;
769 	device_t dev;
770 	struct e1000_hw *hw;
771 	int error = 0;
772 
773 	INIT_DEBUGOUT("em_if_attach_pre: begin");
774 	dev = iflib_get_dev(ctx);
775 	adapter = iflib_get_softc(ctx);
776 
777 	adapter->ctx = adapter->osdep.ctx = ctx;
778 	adapter->dev = adapter->osdep.dev = dev;
779 	scctx = adapter->shared = iflib_get_softc_ctx(ctx);
780 	adapter->media = iflib_get_media(ctx);
781 	hw = &adapter->hw;
782 
783 	adapter->tx_process_limit = scctx->isc_ntxd[0];
784 
785 	/* SYSCTL stuff */
786 	SYSCTL_ADD_PROC(device_get_sysctl_ctx(dev),
787 	    SYSCTL_CHILDREN(device_get_sysctl_tree(dev)),
788 	    OID_AUTO, "nvm", CTLTYPE_INT|CTLFLAG_RW, adapter, 0,
789 	    em_sysctl_nvm_info, "I", "NVM Information");
790 
791 	SYSCTL_ADD_PROC(device_get_sysctl_ctx(dev),
792 	    SYSCTL_CHILDREN(device_get_sysctl_tree(dev)),
793 	    OID_AUTO, "debug", CTLTYPE_INT|CTLFLAG_RW, adapter, 0,
794 	    em_sysctl_debug_info, "I", "Debug Information");
795 
796 	SYSCTL_ADD_PROC(device_get_sysctl_ctx(dev),
797 	    SYSCTL_CHILDREN(device_get_sysctl_tree(dev)),
798 	    OID_AUTO, "fc", CTLTYPE_INT|CTLFLAG_RW, adapter, 0,
799 	    em_set_flowcntl, "I", "Flow Control");
800 
801 	SYSCTL_ADD_PROC(device_get_sysctl_ctx(dev),
802 	    SYSCTL_CHILDREN(device_get_sysctl_tree(dev)),
803 	    OID_AUTO, "reg_dump", CTLTYPE_STRING | CTLFLAG_RD, adapter, 0,
804 	    em_get_regs, "A", "Dump Registers");
805 
806 	SYSCTL_ADD_PROC(device_get_sysctl_ctx(dev),
807 	    SYSCTL_CHILDREN(device_get_sysctl_tree(dev)),
808 	    OID_AUTO, "rs_dump", CTLTYPE_INT | CTLFLAG_RW, adapter, 0,
809 	    em_get_rs, "I", "Dump RS indexes");
810 
811 	/* Determine hardware and mac info */
812 	em_identify_hardware(ctx);
813 
814 	scctx->isc_tx_nsegments = EM_MAX_SCATTER;
815 	scctx->isc_nrxqsets_max = scctx->isc_ntxqsets_max = em_set_num_queues(ctx);
816 	if (bootverbose)
817 		device_printf(dev, "attach_pre capping queues at %d\n",
818 		    scctx->isc_ntxqsets_max);
819 
820 	if (adapter->hw.mac.type >= igb_mac_min) {
821 		scctx->isc_txqsizes[0] = roundup2(scctx->isc_ntxd[0] * sizeof(union e1000_adv_tx_desc), EM_DBA_ALIGN);
822 		scctx->isc_rxqsizes[0] = roundup2(scctx->isc_nrxd[0] * sizeof(union e1000_adv_rx_desc), EM_DBA_ALIGN);
823 		scctx->isc_txd_size[0] = sizeof(union e1000_adv_tx_desc);
824 		scctx->isc_rxd_size[0] = sizeof(union e1000_adv_rx_desc);
825 		scctx->isc_txrx = &igb_txrx;
826 		scctx->isc_tx_tso_segments_max = EM_MAX_SCATTER;
827 		scctx->isc_tx_tso_size_max = EM_TSO_SIZE;
828 		scctx->isc_tx_tso_segsize_max = EM_TSO_SEG_SIZE;
829 		scctx->isc_capabilities = scctx->isc_capenable = IGB_CAPS;
830 		scctx->isc_tx_csum_flags = CSUM_TCP | CSUM_UDP | CSUM_TSO |
831 		     CSUM_IP6_TCP | CSUM_IP6_UDP;
832 		if (adapter->hw.mac.type != e1000_82575)
833 			scctx->isc_tx_csum_flags |= CSUM_SCTP | CSUM_IP6_SCTP;
834 		/*
835 		** Some new devices, as with ixgbe, now may
836 		** use a different BAR, so we need to keep
837 		** track of which is used.
838 		*/
839 		scctx->isc_msix_bar = PCIR_BAR(EM_MSIX_BAR);
840 		if (pci_read_config(dev, scctx->isc_msix_bar, 4) == 0)
841 			scctx->isc_msix_bar += 4;
842 	} else if (adapter->hw.mac.type >= em_mac_min) {
843 		scctx->isc_txqsizes[0] = roundup2(scctx->isc_ntxd[0]* sizeof(struct e1000_tx_desc), EM_DBA_ALIGN);
844 		scctx->isc_rxqsizes[0] = roundup2(scctx->isc_nrxd[0] * sizeof(union e1000_rx_desc_extended), EM_DBA_ALIGN);
845 		scctx->isc_txd_size[0] = sizeof(struct e1000_tx_desc);
846 		scctx->isc_rxd_size[0] = sizeof(union e1000_rx_desc_extended);
847 		scctx->isc_txrx = &em_txrx;
848 		scctx->isc_tx_tso_segments_max = EM_MAX_SCATTER;
849 		scctx->isc_tx_tso_size_max = EM_TSO_SIZE;
850 		scctx->isc_tx_tso_segsize_max = EM_TSO_SEG_SIZE;
851 		scctx->isc_capabilities = scctx->isc_capenable = EM_CAPS;
852 		/*
853 		 * For EM-class devices, don't enable IFCAP_{TSO4,VLAN_HWTSO}
854 		 * by default as we don't have workarounds for all associated
855 		 * silicon errata.  E. g., with several MACs such as 82573E,
856 		 * TSO only works at Gigabit speed and otherwise can cause the
857 		 * hardware to hang (which also would be next to impossible to
858 		 * work around given that already queued TSO-using descriptors
859 		 * would need to be flushed and vlan(4) reconfigured at runtime
860 		 * in case of a link speed change).  Moreover, MACs like 82579
861 		 * still can hang at Gigabit even with all publicly documented
862 		 * TSO workarounds implemented.  Generally, the penality of
863 		 * these workarounds is rather high and may involve copying
864 		 * mbuf data around so advantages of TSO lapse.  Still, TSO may
865 		 * work for a few MACs of this class - at least when sticking
866 		 * with Gigabit - in which case users may enable TSO manually.
867 		 */
868 		scctx->isc_capenable &= ~(IFCAP_TSO4 | IFCAP_VLAN_HWTSO);
869 		scctx->isc_tx_csum_flags = CSUM_TCP | CSUM_UDP | CSUM_IP_TSO;
870 		/*
871 		 * We support MSI-X with 82574 only, but indicate to iflib(4)
872 		 * that it shall give MSI at least a try with other devices.
873 		 */
874 		if (adapter->hw.mac.type == e1000_82574) {
875 			scctx->isc_msix_bar = PCIR_BAR(EM_MSIX_BAR);
876 		} else {
877 			scctx->isc_msix_bar = -1;
878 			scctx->isc_disable_msix = 1;
879 		}
880 	} else {
881 		scctx->isc_txqsizes[0] = roundup2((scctx->isc_ntxd[0] + 1) * sizeof(struct e1000_tx_desc), EM_DBA_ALIGN);
882 		scctx->isc_rxqsizes[0] = roundup2((scctx->isc_nrxd[0] + 1) * sizeof(struct e1000_rx_desc), EM_DBA_ALIGN);
883 		scctx->isc_txd_size[0] = sizeof(struct e1000_tx_desc);
884 		scctx->isc_rxd_size[0] = sizeof(struct e1000_rx_desc);
885 		scctx->isc_tx_csum_flags = CSUM_TCP | CSUM_UDP;
886 		scctx->isc_txrx = &lem_txrx;
887 		scctx->isc_capabilities = scctx->isc_capenable = LEM_CAPS;
888 		if (adapter->hw.mac.type < e1000_82543)
889 			scctx->isc_capenable &= ~(IFCAP_HWCSUM|IFCAP_VLAN_HWCSUM);
890 		/* INTx only */
891 		scctx->isc_msix_bar = 0;
892 	}
893 
894 	/* Setup PCI resources */
895 	if (em_allocate_pci_resources(ctx)) {
896 		device_printf(dev, "Allocation of PCI resources failed\n");
897 		error = ENXIO;
898 		goto err_pci;
899 	}
900 
901 	/*
902 	** For ICH8 and family we need to
903 	** map the flash memory, and this
904 	** must happen after the MAC is
905 	** identified
906 	*/
907 	if ((hw->mac.type == e1000_ich8lan) ||
908 	    (hw->mac.type == e1000_ich9lan) ||
909 	    (hw->mac.type == e1000_ich10lan) ||
910 	    (hw->mac.type == e1000_pchlan) ||
911 	    (hw->mac.type == e1000_pch2lan) ||
912 	    (hw->mac.type == e1000_pch_lpt)) {
913 		int rid = EM_BAR_TYPE_FLASH;
914 		adapter->flash = bus_alloc_resource_any(dev,
915 		    SYS_RES_MEMORY, &rid, RF_ACTIVE);
916 		if (adapter->flash == NULL) {
917 			device_printf(dev, "Mapping of Flash failed\n");
918 			error = ENXIO;
919 			goto err_pci;
920 		}
921 		/* This is used in the shared code */
922 		hw->flash_address = (u8 *)adapter->flash;
923 		adapter->osdep.flash_bus_space_tag =
924 		    rman_get_bustag(adapter->flash);
925 		adapter->osdep.flash_bus_space_handle =
926 		    rman_get_bushandle(adapter->flash);
927 	}
928 	/*
929 	** In the new SPT device flash is not  a
930 	** separate BAR, rather it is also in BAR0,
931 	** so use the same tag and an offset handle for the
932 	** FLASH read/write macros in the shared code.
933 	*/
934 	else if (hw->mac.type >= e1000_pch_spt) {
935 		adapter->osdep.flash_bus_space_tag =
936 		    adapter->osdep.mem_bus_space_tag;
937 		adapter->osdep.flash_bus_space_handle =
938 		    adapter->osdep.mem_bus_space_handle
939 		    + E1000_FLASH_BASE_ADDR;
940 	}
941 
942 	/* Do Shared Code initialization */
943 	error = e1000_setup_init_funcs(hw, TRUE);
944 	if (error) {
945 		device_printf(dev, "Setup of Shared code failed, error %d\n",
946 		    error);
947 		error = ENXIO;
948 		goto err_pci;
949 	}
950 
951 	em_setup_msix(ctx);
952 	e1000_get_bus_info(hw);
953 
954 	/* Set up some sysctls for the tunable interrupt delays */
955 	em_add_int_delay_sysctl(adapter, "rx_int_delay",
956 	    "receive interrupt delay in usecs", &adapter->rx_int_delay,
957 	    E1000_REGISTER(hw, E1000_RDTR), em_rx_int_delay_dflt);
958 	em_add_int_delay_sysctl(adapter, "tx_int_delay",
959 	    "transmit interrupt delay in usecs", &adapter->tx_int_delay,
960 	    E1000_REGISTER(hw, E1000_TIDV), em_tx_int_delay_dflt);
961 	em_add_int_delay_sysctl(adapter, "rx_abs_int_delay",
962 	    "receive interrupt delay limit in usecs",
963 	    &adapter->rx_abs_int_delay,
964 	    E1000_REGISTER(hw, E1000_RADV),
965 	    em_rx_abs_int_delay_dflt);
966 	em_add_int_delay_sysctl(adapter, "tx_abs_int_delay",
967 	    "transmit interrupt delay limit in usecs",
968 	    &adapter->tx_abs_int_delay,
969 	    E1000_REGISTER(hw, E1000_TADV),
970 	    em_tx_abs_int_delay_dflt);
971 	em_add_int_delay_sysctl(adapter, "itr",
972 	    "interrupt delay limit in usecs/4",
973 	    &adapter->tx_itr,
974 	    E1000_REGISTER(hw, E1000_ITR),
975 	    DEFAULT_ITR);
976 
977 	hw->mac.autoneg = DO_AUTO_NEG;
978 	hw->phy.autoneg_wait_to_complete = FALSE;
979 	hw->phy.autoneg_advertised = AUTONEG_ADV_DEFAULT;
980 
981 	if (adapter->hw.mac.type < em_mac_min) {
982 		e1000_init_script_state_82541(&adapter->hw, TRUE);
983 		e1000_set_tbi_compatibility_82543(&adapter->hw, TRUE);
984 	}
985 	/* Copper options */
986 	if (hw->phy.media_type == e1000_media_type_copper) {
987 		hw->phy.mdix = AUTO_ALL_MODES;
988 		hw->phy.disable_polarity_correction = FALSE;
989 		hw->phy.ms_type = EM_MASTER_SLAVE;
990 	}
991 
992 	/*
993 	 * Set the frame limits assuming
994 	 * standard ethernet sized frames.
995 	 */
996 	scctx->isc_max_frame_size = adapter->hw.mac.max_frame_size =
997 	    ETHERMTU + ETHER_HDR_LEN + ETHERNET_FCS_SIZE;
998 
999 	/*
1000 	 * This controls when hardware reports transmit completion
1001 	 * status.
1002 	 */
1003 	hw->mac.report_tx_early = 1;
1004 
1005 	/* Allocate multicast array memory. */
1006 	adapter->mta = malloc(sizeof(u8) * ETHER_ADDR_LEN *
1007 	    MAX_NUM_MULTICAST_ADDRESSES, M_DEVBUF, M_NOWAIT);
1008 	if (adapter->mta == NULL) {
1009 		device_printf(dev, "Can not allocate multicast setup array\n");
1010 		error = ENOMEM;
1011 		goto err_late;
1012 	}
1013 
1014 	/* Check SOL/IDER usage */
1015 	if (e1000_check_reset_block(hw))
1016 		device_printf(dev, "PHY reset is blocked"
1017 			      " due to SOL/IDER session.\n");
1018 
1019 	/* Sysctl for setting Energy Efficient Ethernet */
1020 	hw->dev_spec.ich8lan.eee_disable = eee_setting;
1021 	SYSCTL_ADD_PROC(device_get_sysctl_ctx(dev),
1022 	    SYSCTL_CHILDREN(device_get_sysctl_tree(dev)),
1023 	    OID_AUTO, "eee_control", CTLTYPE_INT|CTLFLAG_RW,
1024 	    adapter, 0, em_sysctl_eee, "I",
1025 	    "Disable Energy Efficient Ethernet");
1026 
1027 	/*
1028 	** Start from a known state, this is
1029 	** important in reading the nvm and
1030 	** mac from that.
1031 	*/
1032 	e1000_reset_hw(hw);
1033 
1034 	/* Make sure we have a good EEPROM before we read from it */
1035 	if (e1000_validate_nvm_checksum(hw) < 0) {
1036 		/*
1037 		** Some PCI-E parts fail the first check due to
1038 		** the link being in sleep state, call it again,
1039 		** if it fails a second time its a real issue.
1040 		*/
1041 		if (e1000_validate_nvm_checksum(hw) < 0) {
1042 			device_printf(dev,
1043 			    "The EEPROM Checksum Is Not Valid\n");
1044 			error = EIO;
1045 			goto err_late;
1046 		}
1047 	}
1048 
1049 	/* Copy the permanent MAC address out of the EEPROM */
1050 	if (e1000_read_mac_addr(hw) < 0) {
1051 		device_printf(dev, "EEPROM read error while reading MAC"
1052 			      " address\n");
1053 		error = EIO;
1054 		goto err_late;
1055 	}
1056 
1057 	if (!em_is_valid_ether_addr(hw->mac.addr)) {
1058 		device_printf(dev, "Invalid MAC address\n");
1059 		error = EIO;
1060 		goto err_late;
1061 	}
1062 
1063 	/* Disable ULP support */
1064 	e1000_disable_ulp_lpt_lp(hw, TRUE);
1065 
1066 	/*
1067 	 * Get Wake-on-Lan and Management info for later use
1068 	 */
1069 	em_get_wakeup(ctx);
1070 
1071 	/* Enable only WOL MAGIC by default */
1072 	scctx->isc_capenable &= ~IFCAP_WOL;
1073 	if (adapter->wol != 0)
1074 		scctx->isc_capenable |= IFCAP_WOL_MAGIC;
1075 
1076 	iflib_set_mac(ctx, hw->mac.addr);
1077 
1078 	return (0);
1079 
1080 err_late:
1081 	em_release_hw_control(adapter);
1082 err_pci:
1083 	em_free_pci_resources(ctx);
1084 	free(adapter->mta, M_DEVBUF);
1085 
1086 	return (error);
1087 }
1088 
1089 static int
1090 em_if_attach_post(if_ctx_t ctx)
1091 {
1092 	struct adapter *adapter = iflib_get_softc(ctx);
1093 	struct e1000_hw *hw = &adapter->hw;
1094 	int error = 0;
1095 
1096 	/* Setup OS specific network interface */
1097 	error = em_setup_interface(ctx);
1098 	if (error != 0) {
1099 		goto err_late;
1100 	}
1101 
1102 	em_reset(ctx);
1103 
1104 	/* Initialize statistics */
1105 	em_update_stats_counters(adapter);
1106 	hw->mac.get_link_status = 1;
1107 	em_if_update_admin_status(ctx);
1108 	em_add_hw_stats(adapter);
1109 
1110 	/* Non-AMT based hardware can now take control from firmware */
1111 	if (adapter->has_manage && !adapter->has_amt)
1112 		em_get_hw_control(adapter);
1113 
1114 	INIT_DEBUGOUT("em_if_attach_post: end");
1115 
1116 	return (error);
1117 
1118 err_late:
1119 	em_release_hw_control(adapter);
1120 	em_free_pci_resources(ctx);
1121 	em_if_queues_free(ctx);
1122 	free(adapter->mta, M_DEVBUF);
1123 
1124 	return (error);
1125 }
1126 
1127 /*********************************************************************
1128  *  Device removal routine
1129  *
1130  *  The detach entry point is called when the driver is being removed.
1131  *  This routine stops the adapter and deallocates all the resources
1132  *  that were allocated for driver operation.
1133  *
1134  *  return 0 on success, positive on failure
1135  *********************************************************************/
1136 static int
1137 em_if_detach(if_ctx_t ctx)
1138 {
1139 	struct adapter	*adapter = iflib_get_softc(ctx);
1140 
1141 	INIT_DEBUGOUT("em_if_detach: begin");
1142 
1143 	e1000_phy_hw_reset(&adapter->hw);
1144 
1145 	em_release_manageability(adapter);
1146 	em_release_hw_control(adapter);
1147 	em_free_pci_resources(ctx);
1148 
1149 	return (0);
1150 }
1151 
1152 /*********************************************************************
1153  *
1154  *  Shutdown entry point
1155  *
1156  **********************************************************************/
1157 
1158 static int
1159 em_if_shutdown(if_ctx_t ctx)
1160 {
1161 	return em_if_suspend(ctx);
1162 }
1163 
1164 /*
1165  * Suspend/resume device methods.
1166  */
1167 static int
1168 em_if_suspend(if_ctx_t ctx)
1169 {
1170 	struct adapter *adapter = iflib_get_softc(ctx);
1171 
1172 	em_release_manageability(adapter);
1173 	em_release_hw_control(adapter);
1174 	em_enable_wakeup(ctx);
1175 	return (0);
1176 }
1177 
1178 static int
1179 em_if_resume(if_ctx_t ctx)
1180 {
1181 	struct adapter *adapter = iflib_get_softc(ctx);
1182 
1183 	if (adapter->hw.mac.type == e1000_pch2lan)
1184 		e1000_resume_workarounds_pchlan(&adapter->hw);
1185 	em_if_init(ctx);
1186 	em_init_manageability(adapter);
1187 
1188 	return(0);
1189 }
1190 
1191 static int
1192 em_if_mtu_set(if_ctx_t ctx, uint32_t mtu)
1193 {
1194 	int max_frame_size;
1195 	struct adapter *adapter = iflib_get_softc(ctx);
1196 	if_softc_ctx_t scctx = iflib_get_softc_ctx(ctx);
1197 
1198 	 IOCTL_DEBUGOUT("ioctl rcv'd: SIOCSIFMTU (Set Interface MTU)");
1199 
1200 	switch (adapter->hw.mac.type) {
1201 	case e1000_82571:
1202 	case e1000_82572:
1203 	case e1000_ich9lan:
1204 	case e1000_ich10lan:
1205 	case e1000_pch2lan:
1206 	case e1000_pch_lpt:
1207 	case e1000_pch_spt:
1208 	case e1000_pch_cnp:
1209 	case e1000_82574:
1210 	case e1000_82583:
1211 	case e1000_80003es2lan:
1212 		/* 9K Jumbo Frame size */
1213 		max_frame_size = 9234;
1214 		break;
1215 	case e1000_pchlan:
1216 		max_frame_size = 4096;
1217 		break;
1218 	case e1000_82542:
1219 	case e1000_ich8lan:
1220 		/* Adapters that do not support jumbo frames */
1221 		max_frame_size = ETHER_MAX_LEN;
1222 		break;
1223 	default:
1224 		if (adapter->hw.mac.type >= igb_mac_min)
1225 			max_frame_size = 9234;
1226 		else /* lem */
1227 			max_frame_size = MAX_JUMBO_FRAME_SIZE;
1228 	}
1229 	if (mtu > max_frame_size - ETHER_HDR_LEN - ETHER_CRC_LEN) {
1230 		return (EINVAL);
1231 	}
1232 
1233 	scctx->isc_max_frame_size = adapter->hw.mac.max_frame_size =
1234 	    mtu + ETHER_HDR_LEN + ETHER_CRC_LEN;
1235 	return (0);
1236 }
1237 
1238 /*********************************************************************
1239  *  Init entry point
1240  *
1241  *  This routine is used in two ways. It is used by the stack as
1242  *  init entry point in network interface structure. It is also used
1243  *  by the driver as a hw/sw initialization routine to get to a
1244  *  consistent state.
1245  *
1246  **********************************************************************/
1247 static void
1248 em_if_init(if_ctx_t ctx)
1249 {
1250 	struct adapter *adapter = iflib_get_softc(ctx);
1251 	if_softc_ctx_t scctx = adapter->shared;
1252 	struct ifnet *ifp = iflib_get_ifp(ctx);
1253 	struct em_tx_queue *tx_que;
1254 	int i;
1255 
1256 	INIT_DEBUGOUT("em_if_init: begin");
1257 
1258 	/* Get the latest mac address, User can use a LAA */
1259 	bcopy(if_getlladdr(ifp), adapter->hw.mac.addr,
1260 	    ETHER_ADDR_LEN);
1261 
1262 	/* Put the address into the Receive Address Array */
1263 	e1000_rar_set(&adapter->hw, adapter->hw.mac.addr, 0);
1264 
1265 	/*
1266 	 * With the 82571 adapter, RAR[0] may be overwritten
1267 	 * when the other port is reset, we make a duplicate
1268 	 * in RAR[14] for that eventuality, this assures
1269 	 * the interface continues to function.
1270 	 */
1271 	if (adapter->hw.mac.type == e1000_82571) {
1272 		e1000_set_laa_state_82571(&adapter->hw, TRUE);
1273 		e1000_rar_set(&adapter->hw, adapter->hw.mac.addr,
1274 		    E1000_RAR_ENTRIES - 1);
1275 	}
1276 
1277 
1278 	/* Initialize the hardware */
1279 	em_reset(ctx);
1280 	em_if_update_admin_status(ctx);
1281 
1282 	for (i = 0, tx_que = adapter->tx_queues; i < adapter->tx_num_queues; i++, tx_que++) {
1283 		struct tx_ring *txr = &tx_que->txr;
1284 
1285 		txr->tx_rs_cidx = txr->tx_rs_pidx;
1286 
1287 		/* Initialize the last processed descriptor to be the end of
1288 		 * the ring, rather than the start, so that we avoid an
1289 		 * off-by-one error when calculating how many descriptors are
1290 		 * done in the credits_update function.
1291 		 */
1292 		txr->tx_cidx_processed = scctx->isc_ntxd[0] - 1;
1293 	}
1294 
1295 	/* Setup VLAN support, basic and offload if available */
1296 	E1000_WRITE_REG(&adapter->hw, E1000_VET, ETHERTYPE_VLAN);
1297 
1298 	/* Clear bad data from Rx FIFOs */
1299 	if (adapter->hw.mac.type >= igb_mac_min)
1300 		e1000_rx_fifo_flush_82575(&adapter->hw);
1301 
1302 	/* Configure for OS presence */
1303 	em_init_manageability(adapter);
1304 
1305 	/* Prepare transmit descriptors and buffers */
1306 	em_initialize_transmit_unit(ctx);
1307 
1308 	/* Setup Multicast table */
1309 	em_if_multi_set(ctx);
1310 
1311 	adapter->rx_mbuf_sz = iflib_get_rx_mbuf_sz(ctx);
1312 	em_initialize_receive_unit(ctx);
1313 
1314 	/* Use real VLAN Filter support? */
1315 	if (if_getcapenable(ifp) & IFCAP_VLAN_HWTAGGING) {
1316 		if (if_getcapenable(ifp) & IFCAP_VLAN_HWFILTER)
1317 			/* Use real VLAN Filter support */
1318 			em_setup_vlan_hw_support(adapter);
1319 		else {
1320 			u32 ctrl;
1321 			ctrl = E1000_READ_REG(&adapter->hw, E1000_CTRL);
1322 			ctrl |= E1000_CTRL_VME;
1323 			E1000_WRITE_REG(&adapter->hw, E1000_CTRL, ctrl);
1324 		}
1325 	}
1326 
1327 	/* Don't lose promiscuous settings */
1328 	em_if_set_promisc(ctx, IFF_PROMISC);
1329 	e1000_clear_hw_cntrs_base_generic(&adapter->hw);
1330 
1331 	/* MSI-X configuration for 82574 */
1332 	if (adapter->hw.mac.type == e1000_82574) {
1333 		int tmp = E1000_READ_REG(&adapter->hw, E1000_CTRL_EXT);
1334 
1335 		tmp |= E1000_CTRL_EXT_PBA_CLR;
1336 		E1000_WRITE_REG(&adapter->hw, E1000_CTRL_EXT, tmp);
1337 		/* Set the IVAR - interrupt vector routing. */
1338 		E1000_WRITE_REG(&adapter->hw, E1000_IVAR, adapter->ivars);
1339 	} else if (adapter->intr_type == IFLIB_INTR_MSIX) /* Set up queue routing */
1340 		igb_configure_queues(adapter);
1341 
1342 	/* this clears any pending interrupts */
1343 	E1000_READ_REG(&adapter->hw, E1000_ICR);
1344 	E1000_WRITE_REG(&adapter->hw, E1000_ICS, E1000_ICS_LSC);
1345 
1346 	/* AMT based hardware can now take control from firmware */
1347 	if (adapter->has_manage && adapter->has_amt)
1348 		em_get_hw_control(adapter);
1349 
1350 	/* Set Energy Efficient Ethernet */
1351 	if (adapter->hw.mac.type >= igb_mac_min &&
1352 	    adapter->hw.phy.media_type == e1000_media_type_copper) {
1353 		if (adapter->hw.mac.type == e1000_i354)
1354 			e1000_set_eee_i354(&adapter->hw, TRUE, TRUE);
1355 		else
1356 			e1000_set_eee_i350(&adapter->hw, TRUE, TRUE);
1357 	}
1358 }
1359 
1360 /*********************************************************************
1361  *
1362  *  Fast Legacy/MSI Combined Interrupt Service routine
1363  *
1364  *********************************************************************/
1365 int
1366 em_intr(void *arg)
1367 {
1368 	struct adapter *adapter = arg;
1369 	if_ctx_t ctx = adapter->ctx;
1370 	u32 reg_icr;
1371 
1372 	reg_icr = E1000_READ_REG(&adapter->hw, E1000_ICR);
1373 
1374 	/* Hot eject? */
1375 	if (reg_icr == 0xffffffff)
1376 		return FILTER_STRAY;
1377 
1378 	/* Definitely not our interrupt. */
1379 	if (reg_icr == 0x0)
1380 		return FILTER_STRAY;
1381 
1382 	/*
1383 	 * Starting with the 82571 chip, bit 31 should be used to
1384 	 * determine whether the interrupt belongs to us.
1385 	 */
1386 	if (adapter->hw.mac.type >= e1000_82571 &&
1387 	    (reg_icr & E1000_ICR_INT_ASSERTED) == 0)
1388 		return FILTER_STRAY;
1389 
1390 	/*
1391 	 * Only MSI-X interrupts have one-shot behavior by taking advantage
1392 	 * of the EIAC register.  Thus, explicitly disable interrupts.  This
1393 	 * also works around the MSI message reordering errata on certain
1394 	 * systems.
1395 	 */
1396 	IFDI_INTR_DISABLE(ctx);
1397 
1398 	/* Link status change */
1399 	if (reg_icr & (E1000_ICR_RXSEQ | E1000_ICR_LSC))
1400 		em_handle_link(ctx);
1401 
1402 	if (reg_icr & E1000_ICR_RXO)
1403 		adapter->rx_overruns++;
1404 
1405 	return (FILTER_SCHEDULE_THREAD);
1406 }
1407 
1408 static int
1409 em_if_rx_queue_intr_enable(if_ctx_t ctx, uint16_t rxqid)
1410 {
1411 	struct adapter *adapter = iflib_get_softc(ctx);
1412 	struct em_rx_queue *rxq = &adapter->rx_queues[rxqid];
1413 
1414 	E1000_WRITE_REG(&adapter->hw, E1000_IMS, rxq->eims);
1415 	return (0);
1416 }
1417 
1418 static int
1419 em_if_tx_queue_intr_enable(if_ctx_t ctx, uint16_t txqid)
1420 {
1421 	struct adapter *adapter = iflib_get_softc(ctx);
1422 	struct em_tx_queue *txq = &adapter->tx_queues[txqid];
1423 
1424 	E1000_WRITE_REG(&adapter->hw, E1000_IMS, txq->eims);
1425 	return (0);
1426 }
1427 
1428 static int
1429 igb_if_rx_queue_intr_enable(if_ctx_t ctx, uint16_t rxqid)
1430 {
1431 	struct adapter *adapter = iflib_get_softc(ctx);
1432 	struct em_rx_queue *rxq = &adapter->rx_queues[rxqid];
1433 
1434 	E1000_WRITE_REG(&adapter->hw, E1000_EIMS, rxq->eims);
1435 	return (0);
1436 }
1437 
1438 static int
1439 igb_if_tx_queue_intr_enable(if_ctx_t ctx, uint16_t txqid)
1440 {
1441 	struct adapter *adapter = iflib_get_softc(ctx);
1442 	struct em_tx_queue *txq = &adapter->tx_queues[txqid];
1443 
1444 	E1000_WRITE_REG(&adapter->hw, E1000_EIMS, txq->eims);
1445 	return (0);
1446 }
1447 
1448 /*********************************************************************
1449  *
1450  *  MSI-X RX Interrupt Service routine
1451  *
1452  **********************************************************************/
1453 static int
1454 em_msix_que(void *arg)
1455 {
1456 	struct em_rx_queue *que = arg;
1457 
1458 	++que->irqs;
1459 
1460 	return (FILTER_SCHEDULE_THREAD);
1461 }
1462 
1463 /*********************************************************************
1464  *
1465  *  MSI-X Link Fast Interrupt Service routine
1466  *
1467  **********************************************************************/
1468 static int
1469 em_msix_link(void *arg)
1470 {
1471 	struct adapter *adapter = arg;
1472 	u32 reg_icr;
1473 
1474 	++adapter->link_irq;
1475 	MPASS(adapter->hw.back != NULL);
1476 	reg_icr = E1000_READ_REG(&adapter->hw, E1000_ICR);
1477 
1478 	if (reg_icr & E1000_ICR_RXO)
1479 		adapter->rx_overruns++;
1480 
1481 	if (reg_icr & (E1000_ICR_RXSEQ | E1000_ICR_LSC)) {
1482 		em_handle_link(adapter->ctx);
1483 	} else if (adapter->hw.mac.type == e1000_82574) {
1484 		/* Only re-arm 82574 if em_if_update_admin_status() won't. */
1485 		E1000_WRITE_REG(&adapter->hw, E1000_IMS, EM_MSIX_LINK |
1486 		    E1000_IMS_LSC);
1487 	}
1488 
1489 	if (adapter->hw.mac.type == e1000_82574) {
1490 		/*
1491 		 * Because we must read the ICR for this interrupt it may
1492 		 * clear other causes using autoclear, for this reason we
1493 		 * simply create a soft interrupt for all these vectors.
1494 		 */
1495 		if (reg_icr)
1496 			E1000_WRITE_REG(&adapter->hw, E1000_ICS, adapter->ims);
1497 	} else {
1498 		/* Re-arm unconditionally */
1499 		E1000_WRITE_REG(&adapter->hw, E1000_IMS, E1000_IMS_LSC);
1500 		E1000_WRITE_REG(&adapter->hw, E1000_EIMS, adapter->link_mask);
1501 	}
1502 
1503 	return (FILTER_HANDLED);
1504 }
1505 
1506 static void
1507 em_handle_link(void *context)
1508 {
1509 	if_ctx_t ctx = context;
1510 	struct adapter *adapter = iflib_get_softc(ctx);
1511 
1512 	adapter->hw.mac.get_link_status = 1;
1513 	iflib_admin_intr_deferred(ctx);
1514 }
1515 
1516 /*********************************************************************
1517  *
1518  *  Media Ioctl callback
1519  *
1520  *  This routine is called whenever the user queries the status of
1521  *  the interface using ifconfig.
1522  *
1523  **********************************************************************/
1524 static void
1525 em_if_media_status(if_ctx_t ctx, struct ifmediareq *ifmr)
1526 {
1527 	struct adapter *adapter = iflib_get_softc(ctx);
1528 	u_char fiber_type = IFM_1000_SX;
1529 
1530 	INIT_DEBUGOUT("em_if_media_status: begin");
1531 
1532 	iflib_admin_intr_deferred(ctx);
1533 
1534 	ifmr->ifm_status = IFM_AVALID;
1535 	ifmr->ifm_active = IFM_ETHER;
1536 
1537 	if (!adapter->link_active) {
1538 		return;
1539 	}
1540 
1541 	ifmr->ifm_status |= IFM_ACTIVE;
1542 
1543 	if ((adapter->hw.phy.media_type == e1000_media_type_fiber) ||
1544 	    (adapter->hw.phy.media_type == e1000_media_type_internal_serdes)) {
1545 		if (adapter->hw.mac.type == e1000_82545)
1546 			fiber_type = IFM_1000_LX;
1547 		ifmr->ifm_active |= fiber_type | IFM_FDX;
1548 	} else {
1549 		switch (adapter->link_speed) {
1550 		case 10:
1551 			ifmr->ifm_active |= IFM_10_T;
1552 			break;
1553 		case 100:
1554 			ifmr->ifm_active |= IFM_100_TX;
1555 			break;
1556 		case 1000:
1557 			ifmr->ifm_active |= IFM_1000_T;
1558 			break;
1559 		}
1560 		if (adapter->link_duplex == FULL_DUPLEX)
1561 			ifmr->ifm_active |= IFM_FDX;
1562 		else
1563 			ifmr->ifm_active |= IFM_HDX;
1564 	}
1565 }
1566 
1567 /*********************************************************************
1568  *
1569  *  Media Ioctl callback
1570  *
1571  *  This routine is called when the user changes speed/duplex using
1572  *  media/mediopt option with ifconfig.
1573  *
1574  **********************************************************************/
1575 static int
1576 em_if_media_change(if_ctx_t ctx)
1577 {
1578 	struct adapter *adapter = iflib_get_softc(ctx);
1579 	struct ifmedia *ifm = iflib_get_media(ctx);
1580 
1581 	INIT_DEBUGOUT("em_if_media_change: begin");
1582 
1583 	if (IFM_TYPE(ifm->ifm_media) != IFM_ETHER)
1584 		return (EINVAL);
1585 
1586 	switch (IFM_SUBTYPE(ifm->ifm_media)) {
1587 	case IFM_AUTO:
1588 		adapter->hw.mac.autoneg = DO_AUTO_NEG;
1589 		adapter->hw.phy.autoneg_advertised = AUTONEG_ADV_DEFAULT;
1590 		break;
1591 	case IFM_1000_LX:
1592 	case IFM_1000_SX:
1593 	case IFM_1000_T:
1594 		adapter->hw.mac.autoneg = DO_AUTO_NEG;
1595 		adapter->hw.phy.autoneg_advertised = ADVERTISE_1000_FULL;
1596 		break;
1597 	case IFM_100_TX:
1598 		adapter->hw.mac.autoneg = FALSE;
1599 		adapter->hw.phy.autoneg_advertised = 0;
1600 		if ((ifm->ifm_media & IFM_GMASK) == IFM_FDX)
1601 			adapter->hw.mac.forced_speed_duplex = ADVERTISE_100_FULL;
1602 		else
1603 			adapter->hw.mac.forced_speed_duplex = ADVERTISE_100_HALF;
1604 		break;
1605 	case IFM_10_T:
1606 		adapter->hw.mac.autoneg = FALSE;
1607 		adapter->hw.phy.autoneg_advertised = 0;
1608 		if ((ifm->ifm_media & IFM_GMASK) == IFM_FDX)
1609 			adapter->hw.mac.forced_speed_duplex = ADVERTISE_10_FULL;
1610 		else
1611 			adapter->hw.mac.forced_speed_duplex = ADVERTISE_10_HALF;
1612 		break;
1613 	default:
1614 		device_printf(adapter->dev, "Unsupported media type\n");
1615 	}
1616 
1617 	em_if_init(ctx);
1618 
1619 	return (0);
1620 }
1621 
1622 static int
1623 em_if_set_promisc(if_ctx_t ctx, int flags)
1624 {
1625 	struct adapter *adapter = iflib_get_softc(ctx);
1626 	u32 reg_rctl;
1627 
1628 	em_disable_promisc(ctx);
1629 
1630 	reg_rctl = E1000_READ_REG(&adapter->hw, E1000_RCTL);
1631 
1632 	if (flags & IFF_PROMISC) {
1633 		reg_rctl |= (E1000_RCTL_UPE | E1000_RCTL_MPE);
1634 		/* Turn this on if you want to see bad packets */
1635 		if (em_debug_sbp)
1636 			reg_rctl |= E1000_RCTL_SBP;
1637 		E1000_WRITE_REG(&adapter->hw, E1000_RCTL, reg_rctl);
1638 	} else if (flags & IFF_ALLMULTI) {
1639 		reg_rctl |= E1000_RCTL_MPE;
1640 		reg_rctl &= ~E1000_RCTL_UPE;
1641 		E1000_WRITE_REG(&adapter->hw, E1000_RCTL, reg_rctl);
1642 	}
1643 	return (0);
1644 }
1645 
1646 static void
1647 em_disable_promisc(if_ctx_t ctx)
1648 {
1649 	struct adapter *adapter = iflib_get_softc(ctx);
1650 	struct ifnet *ifp = iflib_get_ifp(ctx);
1651 	u32 reg_rctl;
1652 	int mcnt = 0;
1653 
1654 	reg_rctl = E1000_READ_REG(&adapter->hw, E1000_RCTL);
1655 	reg_rctl &= (~E1000_RCTL_UPE);
1656 	if (if_getflags(ifp) & IFF_ALLMULTI)
1657 		mcnt = MAX_NUM_MULTICAST_ADDRESSES;
1658 	else
1659 		mcnt = if_llmaddr_count(ifp);
1660 	/* Don't disable if in MAX groups */
1661 	if (mcnt < MAX_NUM_MULTICAST_ADDRESSES)
1662 		reg_rctl &=  (~E1000_RCTL_MPE);
1663 	reg_rctl &=  (~E1000_RCTL_SBP);
1664 	E1000_WRITE_REG(&adapter->hw, E1000_RCTL, reg_rctl);
1665 }
1666 
1667 
1668 static u_int
1669 em_copy_maddr(void *arg, struct sockaddr_dl *sdl, u_int cnt)
1670 {
1671 	u8 *mta = arg;
1672 
1673 	if (cnt == MAX_NUM_MULTICAST_ADDRESSES)
1674 		return (1);
1675 
1676 	bcopy(LLADDR(sdl), &mta[cnt * ETHER_ADDR_LEN], ETHER_ADDR_LEN);
1677 
1678 	return (1);
1679 }
1680 
1681 /*********************************************************************
1682  *  Multicast Update
1683  *
1684  *  This routine is called whenever multicast address list is updated.
1685  *
1686  **********************************************************************/
1687 
1688 static void
1689 em_if_multi_set(if_ctx_t ctx)
1690 {
1691 	struct adapter *adapter = iflib_get_softc(ctx);
1692 	struct ifnet *ifp = iflib_get_ifp(ctx);
1693 	u32 reg_rctl = 0;
1694 	u8  *mta; /* Multicast array memory */
1695 	int mcnt = 0;
1696 
1697 	IOCTL_DEBUGOUT("em_set_multi: begin");
1698 
1699 	mta = adapter->mta;
1700 	bzero(mta, sizeof(u8) * ETHER_ADDR_LEN * MAX_NUM_MULTICAST_ADDRESSES);
1701 
1702 	if (adapter->hw.mac.type == e1000_82542 &&
1703 	    adapter->hw.revision_id == E1000_REVISION_2) {
1704 		reg_rctl = E1000_READ_REG(&adapter->hw, E1000_RCTL);
1705 		if (adapter->hw.bus.pci_cmd_word & CMD_MEM_WRT_INVALIDATE)
1706 			e1000_pci_clear_mwi(&adapter->hw);
1707 		reg_rctl |= E1000_RCTL_RST;
1708 		E1000_WRITE_REG(&adapter->hw, E1000_RCTL, reg_rctl);
1709 		msec_delay(5);
1710 	}
1711 
1712 	mcnt = if_foreach_llmaddr(ifp, em_copy_maddr, mta);
1713 
1714 	if (mcnt >= MAX_NUM_MULTICAST_ADDRESSES) {
1715 		reg_rctl = E1000_READ_REG(&adapter->hw, E1000_RCTL);
1716 		reg_rctl |= E1000_RCTL_MPE;
1717 		E1000_WRITE_REG(&adapter->hw, E1000_RCTL, reg_rctl);
1718 	} else
1719 		e1000_update_mc_addr_list(&adapter->hw, mta, mcnt);
1720 
1721 	if (adapter->hw.mac.type == e1000_82542 &&
1722 	    adapter->hw.revision_id == E1000_REVISION_2) {
1723 		reg_rctl = E1000_READ_REG(&adapter->hw, E1000_RCTL);
1724 		reg_rctl &= ~E1000_RCTL_RST;
1725 		E1000_WRITE_REG(&adapter->hw, E1000_RCTL, reg_rctl);
1726 		msec_delay(5);
1727 		if (adapter->hw.bus.pci_cmd_word & CMD_MEM_WRT_INVALIDATE)
1728 			e1000_pci_set_mwi(&adapter->hw);
1729 	}
1730 }
1731 
1732 /*********************************************************************
1733  *  Timer routine
1734  *
1735  *  This routine schedules em_if_update_admin_status() to check for
1736  *  link status and to gather statistics as well as to perform some
1737  *  controller-specific hardware patting.
1738  *
1739  **********************************************************************/
1740 static void
1741 em_if_timer(if_ctx_t ctx, uint16_t qid)
1742 {
1743 
1744 	if (qid != 0)
1745 		return;
1746 
1747 	iflib_admin_intr_deferred(ctx);
1748 }
1749 
1750 static void
1751 em_if_update_admin_status(if_ctx_t ctx)
1752 {
1753 	struct adapter *adapter = iflib_get_softc(ctx);
1754 	struct e1000_hw *hw = &adapter->hw;
1755 	device_t dev = iflib_get_dev(ctx);
1756 	u32 link_check, thstat, ctrl;
1757 
1758 	link_check = thstat = ctrl = 0;
1759 	/* Get the cached link value or read phy for real */
1760 	switch (hw->phy.media_type) {
1761 	case e1000_media_type_copper:
1762 		if (hw->mac.get_link_status) {
1763 			if (hw->mac.type == e1000_pch_spt)
1764 				msec_delay(50);
1765 			/* Do the work to read phy */
1766 			e1000_check_for_link(hw);
1767 			link_check = !hw->mac.get_link_status;
1768 			if (link_check) /* ESB2 fix */
1769 				e1000_cfg_on_link_up(hw);
1770 		} else {
1771 			link_check = TRUE;
1772 		}
1773 		break;
1774 	case e1000_media_type_fiber:
1775 		e1000_check_for_link(hw);
1776 		link_check = (E1000_READ_REG(hw, E1000_STATUS) &
1777 			    E1000_STATUS_LU);
1778 		break;
1779 	case e1000_media_type_internal_serdes:
1780 		e1000_check_for_link(hw);
1781 		link_check = adapter->hw.mac.serdes_has_link;
1782 		break;
1783 	/* VF device is type_unknown */
1784 	case e1000_media_type_unknown:
1785 		e1000_check_for_link(hw);
1786 		link_check = !hw->mac.get_link_status;
1787 		/* FALLTHROUGH */
1788 	default:
1789 		break;
1790 	}
1791 
1792 	/* Check for thermal downshift or shutdown */
1793 	if (hw->mac.type == e1000_i350) {
1794 		thstat = E1000_READ_REG(hw, E1000_THSTAT);
1795 		ctrl = E1000_READ_REG(hw, E1000_CTRL_EXT);
1796 	}
1797 
1798 	/* Now check for a transition */
1799 	if (link_check && (adapter->link_active == 0)) {
1800 		e1000_get_speed_and_duplex(hw, &adapter->link_speed,
1801 		    &adapter->link_duplex);
1802 		/* Check if we must disable SPEED_MODE bit on PCI-E */
1803 		if ((adapter->link_speed != SPEED_1000) &&
1804 		    ((hw->mac.type == e1000_82571) ||
1805 		    (hw->mac.type == e1000_82572))) {
1806 			int tarc0;
1807 			tarc0 = E1000_READ_REG(hw, E1000_TARC(0));
1808 			tarc0 &= ~TARC_SPEED_MODE_BIT;
1809 			E1000_WRITE_REG(hw, E1000_TARC(0), tarc0);
1810 		}
1811 		if (bootverbose)
1812 			device_printf(dev, "Link is up %d Mbps %s\n",
1813 			    adapter->link_speed,
1814 			    ((adapter->link_duplex == FULL_DUPLEX) ?
1815 			    "Full Duplex" : "Half Duplex"));
1816 		adapter->link_active = 1;
1817 		adapter->smartspeed = 0;
1818 		if ((ctrl & E1000_CTRL_EXT_LINK_MODE_MASK) ==
1819 		    E1000_CTRL_EXT_LINK_MODE_GMII &&
1820 		    (thstat & E1000_THSTAT_LINK_THROTTLE))
1821 			device_printf(dev, "Link: thermal downshift\n");
1822 		/* Delay Link Up for Phy update */
1823 		if (((hw->mac.type == e1000_i210) ||
1824 		    (hw->mac.type == e1000_i211)) &&
1825 		    (hw->phy.id == I210_I_PHY_ID))
1826 			msec_delay(I210_LINK_DELAY);
1827 		/* Reset if the media type changed. */
1828 		if ((hw->dev_spec._82575.media_changed) &&
1829 			(adapter->hw.mac.type >= igb_mac_min)) {
1830 			hw->dev_spec._82575.media_changed = false;
1831 			adapter->flags |= IGB_MEDIA_RESET;
1832 			em_reset(ctx);
1833 		}
1834 		iflib_link_state_change(ctx, LINK_STATE_UP,
1835 		    IF_Mbps(adapter->link_speed));
1836 	} else if (!link_check && (adapter->link_active == 1)) {
1837 		adapter->link_speed = 0;
1838 		adapter->link_duplex = 0;
1839 		adapter->link_active = 0;
1840 		iflib_link_state_change(ctx, LINK_STATE_DOWN, 0);
1841 	}
1842 	em_update_stats_counters(adapter);
1843 
1844 	/* Reset LAA into RAR[0] on 82571 */
1845 	if (hw->mac.type == e1000_82571 && e1000_get_laa_state_82571(hw))
1846 		e1000_rar_set(hw, hw->mac.addr, 0);
1847 
1848 	if (hw->mac.type < em_mac_min)
1849 		lem_smartspeed(adapter);
1850 	else if (hw->mac.type == e1000_82574 &&
1851 	    adapter->intr_type == IFLIB_INTR_MSIX)
1852 		E1000_WRITE_REG(&adapter->hw, E1000_IMS, EM_MSIX_LINK |
1853 		    E1000_IMS_LSC);
1854 }
1855 
1856 static void
1857 em_if_watchdog_reset(if_ctx_t ctx)
1858 {
1859 	struct adapter *adapter = iflib_get_softc(ctx);
1860 
1861 	/*
1862 	 * Just count the event; iflib(4) will already trigger a
1863 	 * sufficient reset of the controller.
1864 	 */
1865 	adapter->watchdog_events++;
1866 }
1867 
1868 /*********************************************************************
1869  *
1870  *  This routine disables all traffic on the adapter by issuing a
1871  *  global reset on the MAC.
1872  *
1873  **********************************************************************/
1874 static void
1875 em_if_stop(if_ctx_t ctx)
1876 {
1877 	struct adapter *adapter = iflib_get_softc(ctx);
1878 
1879 	INIT_DEBUGOUT("em_if_stop: begin");
1880 
1881 	e1000_reset_hw(&adapter->hw);
1882 	if (adapter->hw.mac.type >= e1000_82544)
1883 		E1000_WRITE_REG(&adapter->hw, E1000_WUFC, 0);
1884 
1885 	e1000_led_off(&adapter->hw);
1886 	e1000_cleanup_led(&adapter->hw);
1887 }
1888 
1889 /*********************************************************************
1890  *
1891  *  Determine hardware revision.
1892  *
1893  **********************************************************************/
1894 static void
1895 em_identify_hardware(if_ctx_t ctx)
1896 {
1897 	device_t dev = iflib_get_dev(ctx);
1898 	struct adapter *adapter = iflib_get_softc(ctx);
1899 
1900 	/* Make sure our PCI config space has the necessary stuff set */
1901 	adapter->hw.bus.pci_cmd_word = pci_read_config(dev, PCIR_COMMAND, 2);
1902 
1903 	/* Save off the information about this board */
1904 	adapter->hw.vendor_id = pci_get_vendor(dev);
1905 	adapter->hw.device_id = pci_get_device(dev);
1906 	adapter->hw.revision_id = pci_read_config(dev, PCIR_REVID, 1);
1907 	adapter->hw.subsystem_vendor_id =
1908 	    pci_read_config(dev, PCIR_SUBVEND_0, 2);
1909 	adapter->hw.subsystem_device_id =
1910 	    pci_read_config(dev, PCIR_SUBDEV_0, 2);
1911 
1912 	/* Do Shared Code Init and Setup */
1913 	if (e1000_set_mac_type(&adapter->hw)) {
1914 		device_printf(dev, "Setup init failure\n");
1915 		return;
1916 	}
1917 }
1918 
1919 static int
1920 em_allocate_pci_resources(if_ctx_t ctx)
1921 {
1922 	struct adapter *adapter = iflib_get_softc(ctx);
1923 	device_t dev = iflib_get_dev(ctx);
1924 	int rid, val;
1925 
1926 	rid = PCIR_BAR(0);
1927 	adapter->memory = bus_alloc_resource_any(dev, SYS_RES_MEMORY,
1928 	    &rid, RF_ACTIVE);
1929 	if (adapter->memory == NULL) {
1930 		device_printf(dev, "Unable to allocate bus resource: memory\n");
1931 		return (ENXIO);
1932 	}
1933 	adapter->osdep.mem_bus_space_tag = rman_get_bustag(adapter->memory);
1934 	adapter->osdep.mem_bus_space_handle =
1935 	    rman_get_bushandle(adapter->memory);
1936 	adapter->hw.hw_addr = (u8 *)&adapter->osdep.mem_bus_space_handle;
1937 
1938 	/* Only older adapters use IO mapping */
1939 	if (adapter->hw.mac.type < em_mac_min &&
1940 	    adapter->hw.mac.type > e1000_82543) {
1941 		/* Figure our where our IO BAR is ? */
1942 		for (rid = PCIR_BAR(0); rid < PCIR_CIS;) {
1943 			val = pci_read_config(dev, rid, 4);
1944 			if (EM_BAR_TYPE(val) == EM_BAR_TYPE_IO) {
1945 				break;
1946 			}
1947 			rid += 4;
1948 			/* check for 64bit BAR */
1949 			if (EM_BAR_MEM_TYPE(val) == EM_BAR_MEM_TYPE_64BIT)
1950 				rid += 4;
1951 		}
1952 		if (rid >= PCIR_CIS) {
1953 			device_printf(dev, "Unable to locate IO BAR\n");
1954 			return (ENXIO);
1955 		}
1956 		adapter->ioport = bus_alloc_resource_any(dev, SYS_RES_IOPORT,
1957 		    &rid, RF_ACTIVE);
1958 		if (adapter->ioport == NULL) {
1959 			device_printf(dev, "Unable to allocate bus resource: "
1960 			    "ioport\n");
1961 			return (ENXIO);
1962 		}
1963 		adapter->hw.io_base = 0;
1964 		adapter->osdep.io_bus_space_tag =
1965 		    rman_get_bustag(adapter->ioport);
1966 		adapter->osdep.io_bus_space_handle =
1967 		    rman_get_bushandle(adapter->ioport);
1968 	}
1969 
1970 	adapter->hw.back = &adapter->osdep;
1971 
1972 	return (0);
1973 }
1974 
1975 /*********************************************************************
1976  *
1977  *  Set up the MSI-X Interrupt handlers
1978  *
1979  **********************************************************************/
1980 static int
1981 em_if_msix_intr_assign(if_ctx_t ctx, int msix)
1982 {
1983 	struct adapter *adapter = iflib_get_softc(ctx);
1984 	struct em_rx_queue *rx_que = adapter->rx_queues;
1985 	struct em_tx_queue *tx_que = adapter->tx_queues;
1986 	int error, rid, i, vector = 0, rx_vectors;
1987 	char buf[16];
1988 
1989 	/* First set up ring resources */
1990 	for (i = 0; i < adapter->rx_num_queues; i++, rx_que++, vector++) {
1991 		rid = vector + 1;
1992 		snprintf(buf, sizeof(buf), "rxq%d", i);
1993 		error = iflib_irq_alloc_generic(ctx, &rx_que->que_irq, rid, IFLIB_INTR_RXTX, em_msix_que, rx_que, rx_que->me, buf);
1994 		if (error) {
1995 			device_printf(iflib_get_dev(ctx), "Failed to allocate que int %d err: %d", i, error);
1996 			adapter->rx_num_queues = i + 1;
1997 			goto fail;
1998 		}
1999 
2000 		rx_que->msix =  vector;
2001 
2002 		/*
2003 		 * Set the bit to enable interrupt
2004 		 * in E1000_IMS -- bits 20 and 21
2005 		 * are for RX0 and RX1, note this has
2006 		 * NOTHING to do with the MSI-X vector
2007 		 */
2008 		if (adapter->hw.mac.type == e1000_82574) {
2009 			rx_que->eims = 1 << (20 + i);
2010 			adapter->ims |= rx_que->eims;
2011 			adapter->ivars |= (8 | rx_que->msix) << (i * 4);
2012 		} else if (adapter->hw.mac.type == e1000_82575)
2013 			rx_que->eims = E1000_EICR_TX_QUEUE0 << vector;
2014 		else
2015 			rx_que->eims = 1 << vector;
2016 	}
2017 	rx_vectors = vector;
2018 
2019 	vector = 0;
2020 	for (i = 0; i < adapter->tx_num_queues; i++, tx_que++, vector++) {
2021 		snprintf(buf, sizeof(buf), "txq%d", i);
2022 		tx_que = &adapter->tx_queues[i];
2023 		iflib_softirq_alloc_generic(ctx,
2024 		    &adapter->rx_queues[i % adapter->rx_num_queues].que_irq,
2025 		    IFLIB_INTR_TX, tx_que, tx_que->me, buf);
2026 
2027 		tx_que->msix = (vector % adapter->rx_num_queues);
2028 
2029 		/*
2030 		 * Set the bit to enable interrupt
2031 		 * in E1000_IMS -- bits 22 and 23
2032 		 * are for TX0 and TX1, note this has
2033 		 * NOTHING to do with the MSI-X vector
2034 		 */
2035 		if (adapter->hw.mac.type == e1000_82574) {
2036 			tx_que->eims = 1 << (22 + i);
2037 			adapter->ims |= tx_que->eims;
2038 			adapter->ivars |= (8 | tx_que->msix) << (8 + (i * 4));
2039 		} else if (adapter->hw.mac.type == e1000_82575) {
2040 			tx_que->eims = E1000_EICR_TX_QUEUE0 << i;
2041 		} else {
2042 			tx_que->eims = 1 << i;
2043 		}
2044 	}
2045 
2046 	/* Link interrupt */
2047 	rid = rx_vectors + 1;
2048 	error = iflib_irq_alloc_generic(ctx, &adapter->irq, rid, IFLIB_INTR_ADMIN, em_msix_link, adapter, 0, "aq");
2049 
2050 	if (error) {
2051 		device_printf(iflib_get_dev(ctx), "Failed to register admin handler");
2052 		goto fail;
2053 	}
2054 	adapter->linkvec = rx_vectors;
2055 	if (adapter->hw.mac.type < igb_mac_min) {
2056 		adapter->ivars |=  (8 | rx_vectors) << 16;
2057 		adapter->ivars |= 0x80000000;
2058 	}
2059 	return (0);
2060 fail:
2061 	iflib_irq_free(ctx, &adapter->irq);
2062 	rx_que = adapter->rx_queues;
2063 	for (int i = 0; i < adapter->rx_num_queues; i++, rx_que++)
2064 		iflib_irq_free(ctx, &rx_que->que_irq);
2065 	return (error);
2066 }
2067 
2068 static void
2069 igb_configure_queues(struct adapter *adapter)
2070 {
2071 	struct e1000_hw *hw = &adapter->hw;
2072 	struct em_rx_queue *rx_que;
2073 	struct em_tx_queue *tx_que;
2074 	u32 tmp, ivar = 0, newitr = 0;
2075 
2076 	/* First turn on RSS capability */
2077 	if (adapter->hw.mac.type != e1000_82575)
2078 		E1000_WRITE_REG(hw, E1000_GPIE,
2079 		    E1000_GPIE_MSIX_MODE | E1000_GPIE_EIAME |
2080 		    E1000_GPIE_PBA | E1000_GPIE_NSICR);
2081 
2082 	/* Turn on MSI-X */
2083 	switch (adapter->hw.mac.type) {
2084 	case e1000_82580:
2085 	case e1000_i350:
2086 	case e1000_i354:
2087 	case e1000_i210:
2088 	case e1000_i211:
2089 	case e1000_vfadapt:
2090 	case e1000_vfadapt_i350:
2091 		/* RX entries */
2092 		for (int i = 0; i < adapter->rx_num_queues; i++) {
2093 			u32 index = i >> 1;
2094 			ivar = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index);
2095 			rx_que = &adapter->rx_queues[i];
2096 			if (i & 1) {
2097 				ivar &= 0xFF00FFFF;
2098 				ivar |= (rx_que->msix | E1000_IVAR_VALID) << 16;
2099 			} else {
2100 				ivar &= 0xFFFFFF00;
2101 				ivar |= rx_que->msix | E1000_IVAR_VALID;
2102 			}
2103 			E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, ivar);
2104 		}
2105 		/* TX entries */
2106 		for (int i = 0; i < adapter->tx_num_queues; i++) {
2107 			u32 index = i >> 1;
2108 			ivar = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index);
2109 			tx_que = &adapter->tx_queues[i];
2110 			if (i & 1) {
2111 				ivar &= 0x00FFFFFF;
2112 				ivar |= (tx_que->msix | E1000_IVAR_VALID) << 24;
2113 			} else {
2114 				ivar &= 0xFFFF00FF;
2115 				ivar |= (tx_que->msix | E1000_IVAR_VALID) << 8;
2116 			}
2117 			E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, ivar);
2118 			adapter->que_mask |= tx_que->eims;
2119 		}
2120 
2121 		/* And for the link interrupt */
2122 		ivar = (adapter->linkvec | E1000_IVAR_VALID) << 8;
2123 		adapter->link_mask = 1 << adapter->linkvec;
2124 		E1000_WRITE_REG(hw, E1000_IVAR_MISC, ivar);
2125 		break;
2126 	case e1000_82576:
2127 		/* RX entries */
2128 		for (int i = 0; i < adapter->rx_num_queues; i++) {
2129 			u32 index = i & 0x7; /* Each IVAR has two entries */
2130 			ivar = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index);
2131 			rx_que = &adapter->rx_queues[i];
2132 			if (i < 8) {
2133 				ivar &= 0xFFFFFF00;
2134 				ivar |= rx_que->msix | E1000_IVAR_VALID;
2135 			} else {
2136 				ivar &= 0xFF00FFFF;
2137 				ivar |= (rx_que->msix | E1000_IVAR_VALID) << 16;
2138 			}
2139 			E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, ivar);
2140 			adapter->que_mask |= rx_que->eims;
2141 		}
2142 		/* TX entries */
2143 		for (int i = 0; i < adapter->tx_num_queues; i++) {
2144 			u32 index = i & 0x7; /* Each IVAR has two entries */
2145 			ivar = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index);
2146 			tx_que = &adapter->tx_queues[i];
2147 			if (i < 8) {
2148 				ivar &= 0xFFFF00FF;
2149 				ivar |= (tx_que->msix | E1000_IVAR_VALID) << 8;
2150 			} else {
2151 				ivar &= 0x00FFFFFF;
2152 				ivar |= (tx_que->msix | E1000_IVAR_VALID) << 24;
2153 			}
2154 			E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, ivar);
2155 			adapter->que_mask |= tx_que->eims;
2156 		}
2157 
2158 		/* And for the link interrupt */
2159 		ivar = (adapter->linkvec | E1000_IVAR_VALID) << 8;
2160 		adapter->link_mask = 1 << adapter->linkvec;
2161 		E1000_WRITE_REG(hw, E1000_IVAR_MISC, ivar);
2162 		break;
2163 
2164 	case e1000_82575:
2165 		/* enable MSI-X support*/
2166 		tmp = E1000_READ_REG(hw, E1000_CTRL_EXT);
2167 		tmp |= E1000_CTRL_EXT_PBA_CLR;
2168 		/* Auto-Mask interrupts upon ICR read. */
2169 		tmp |= E1000_CTRL_EXT_EIAME;
2170 		tmp |= E1000_CTRL_EXT_IRCA;
2171 		E1000_WRITE_REG(hw, E1000_CTRL_EXT, tmp);
2172 
2173 		/* Queues */
2174 		for (int i = 0; i < adapter->rx_num_queues; i++) {
2175 			rx_que = &adapter->rx_queues[i];
2176 			tmp = E1000_EICR_RX_QUEUE0 << i;
2177 			tmp |= E1000_EICR_TX_QUEUE0 << i;
2178 			rx_que->eims = tmp;
2179 			E1000_WRITE_REG_ARRAY(hw, E1000_MSIXBM(0),
2180 			    i, rx_que->eims);
2181 			adapter->que_mask |= rx_que->eims;
2182 		}
2183 
2184 		/* Link */
2185 		E1000_WRITE_REG(hw, E1000_MSIXBM(adapter->linkvec),
2186 		    E1000_EIMS_OTHER);
2187 		adapter->link_mask |= E1000_EIMS_OTHER;
2188 	default:
2189 		break;
2190 	}
2191 
2192 	/* Set the starting interrupt rate */
2193 	if (em_max_interrupt_rate > 0)
2194 		newitr = (4000000 / em_max_interrupt_rate) & 0x7FFC;
2195 
2196 	if (hw->mac.type == e1000_82575)
2197 		newitr |= newitr << 16;
2198 	else
2199 		newitr |= E1000_EITR_CNT_IGNR;
2200 
2201 	for (int i = 0; i < adapter->rx_num_queues; i++) {
2202 		rx_que = &adapter->rx_queues[i];
2203 		E1000_WRITE_REG(hw, E1000_EITR(rx_que->msix), newitr);
2204 	}
2205 
2206 	return;
2207 }
2208 
2209 static void
2210 em_free_pci_resources(if_ctx_t ctx)
2211 {
2212 	struct adapter *adapter = iflib_get_softc(ctx);
2213 	struct em_rx_queue *que = adapter->rx_queues;
2214 	device_t dev = iflib_get_dev(ctx);
2215 
2216 	/* Release all MSI-X queue resources */
2217 	if (adapter->intr_type == IFLIB_INTR_MSIX)
2218 		iflib_irq_free(ctx, &adapter->irq);
2219 
2220 	for (int i = 0; i < adapter->rx_num_queues; i++, que++) {
2221 		iflib_irq_free(ctx, &que->que_irq);
2222 	}
2223 
2224 	if (adapter->memory != NULL) {
2225 		bus_release_resource(dev, SYS_RES_MEMORY,
2226 		    rman_get_rid(adapter->memory), adapter->memory);
2227 		adapter->memory = NULL;
2228 	}
2229 
2230 	if (adapter->flash != NULL) {
2231 		bus_release_resource(dev, SYS_RES_MEMORY,
2232 		    rman_get_rid(adapter->flash), adapter->flash);
2233 		adapter->flash = NULL;
2234 	}
2235 
2236 	if (adapter->ioport != NULL) {
2237 		bus_release_resource(dev, SYS_RES_IOPORT,
2238 		    rman_get_rid(adapter->ioport), adapter->ioport);
2239 		adapter->ioport = NULL;
2240 	}
2241 }
2242 
2243 /* Set up MSI or MSI-X */
2244 static int
2245 em_setup_msix(if_ctx_t ctx)
2246 {
2247 	struct adapter *adapter = iflib_get_softc(ctx);
2248 
2249 	if (adapter->hw.mac.type == e1000_82574) {
2250 		em_enable_vectors_82574(ctx);
2251 	}
2252 	return (0);
2253 }
2254 
2255 /*********************************************************************
2256  *
2257  *  Workaround for SmartSpeed on 82541 and 82547 controllers
2258  *
2259  **********************************************************************/
2260 static void
2261 lem_smartspeed(struct adapter *adapter)
2262 {
2263 	u16 phy_tmp;
2264 
2265 	if (adapter->link_active || (adapter->hw.phy.type != e1000_phy_igp) ||
2266 	    adapter->hw.mac.autoneg == 0 ||
2267 	    (adapter->hw.phy.autoneg_advertised & ADVERTISE_1000_FULL) == 0)
2268 		return;
2269 
2270 	if (adapter->smartspeed == 0) {
2271 		/* If Master/Slave config fault is asserted twice,
2272 		 * we assume back-to-back */
2273 		e1000_read_phy_reg(&adapter->hw, PHY_1000T_STATUS, &phy_tmp);
2274 		if (!(phy_tmp & SR_1000T_MS_CONFIG_FAULT))
2275 			return;
2276 		e1000_read_phy_reg(&adapter->hw, PHY_1000T_STATUS, &phy_tmp);
2277 		if (phy_tmp & SR_1000T_MS_CONFIG_FAULT) {
2278 			e1000_read_phy_reg(&adapter->hw,
2279 			    PHY_1000T_CTRL, &phy_tmp);
2280 			if(phy_tmp & CR_1000T_MS_ENABLE) {
2281 				phy_tmp &= ~CR_1000T_MS_ENABLE;
2282 				e1000_write_phy_reg(&adapter->hw,
2283 				    PHY_1000T_CTRL, phy_tmp);
2284 				adapter->smartspeed++;
2285 				if(adapter->hw.mac.autoneg &&
2286 				   !e1000_copper_link_autoneg(&adapter->hw) &&
2287 				   !e1000_read_phy_reg(&adapter->hw,
2288 				    PHY_CONTROL, &phy_tmp)) {
2289 					phy_tmp |= (MII_CR_AUTO_NEG_EN |
2290 						    MII_CR_RESTART_AUTO_NEG);
2291 					e1000_write_phy_reg(&adapter->hw,
2292 					    PHY_CONTROL, phy_tmp);
2293 				}
2294 			}
2295 		}
2296 		return;
2297 	} else if(adapter->smartspeed == EM_SMARTSPEED_DOWNSHIFT) {
2298 		/* If still no link, perhaps using 2/3 pair cable */
2299 		e1000_read_phy_reg(&adapter->hw, PHY_1000T_CTRL, &phy_tmp);
2300 		phy_tmp |= CR_1000T_MS_ENABLE;
2301 		e1000_write_phy_reg(&adapter->hw, PHY_1000T_CTRL, phy_tmp);
2302 		if(adapter->hw.mac.autoneg &&
2303 		   !e1000_copper_link_autoneg(&adapter->hw) &&
2304 		   !e1000_read_phy_reg(&adapter->hw, PHY_CONTROL, &phy_tmp)) {
2305 			phy_tmp |= (MII_CR_AUTO_NEG_EN |
2306 				    MII_CR_RESTART_AUTO_NEG);
2307 			e1000_write_phy_reg(&adapter->hw, PHY_CONTROL, phy_tmp);
2308 		}
2309 	}
2310 	/* Restart process after EM_SMARTSPEED_MAX iterations */
2311 	if(adapter->smartspeed++ == EM_SMARTSPEED_MAX)
2312 		adapter->smartspeed = 0;
2313 }
2314 
2315 /*********************************************************************
2316  *
2317  *  Initialize the DMA Coalescing feature
2318  *
2319  **********************************************************************/
2320 static void
2321 igb_init_dmac(struct adapter *adapter, u32 pba)
2322 {
2323 	device_t	dev = adapter->dev;
2324 	struct e1000_hw *hw = &adapter->hw;
2325 	u32 		dmac, reg = ~E1000_DMACR_DMAC_EN;
2326 	u16		hwm;
2327 	u16		max_frame_size;
2328 
2329 	if (hw->mac.type == e1000_i211)
2330 		return;
2331 
2332 	max_frame_size = adapter->shared->isc_max_frame_size;
2333 	if (hw->mac.type > e1000_82580) {
2334 
2335 		if (adapter->dmac == 0) { /* Disabling it */
2336 			E1000_WRITE_REG(hw, E1000_DMACR, reg);
2337 			return;
2338 		} else
2339 			device_printf(dev, "DMA Coalescing enabled\n");
2340 
2341 		/* Set starting threshold */
2342 		E1000_WRITE_REG(hw, E1000_DMCTXTH, 0);
2343 
2344 		hwm = 64 * pba - max_frame_size / 16;
2345 		if (hwm < 64 * (pba - 6))
2346 			hwm = 64 * (pba - 6);
2347 		reg = E1000_READ_REG(hw, E1000_FCRTC);
2348 		reg &= ~E1000_FCRTC_RTH_COAL_MASK;
2349 		reg |= ((hwm << E1000_FCRTC_RTH_COAL_SHIFT)
2350 		    & E1000_FCRTC_RTH_COAL_MASK);
2351 		E1000_WRITE_REG(hw, E1000_FCRTC, reg);
2352 
2353 
2354 		dmac = pba - max_frame_size / 512;
2355 		if (dmac < pba - 10)
2356 			dmac = pba - 10;
2357 		reg = E1000_READ_REG(hw, E1000_DMACR);
2358 		reg &= ~E1000_DMACR_DMACTHR_MASK;
2359 		reg |= ((dmac << E1000_DMACR_DMACTHR_SHIFT)
2360 		    & E1000_DMACR_DMACTHR_MASK);
2361 
2362 		/* transition to L0x or L1 if available..*/
2363 		reg |= (E1000_DMACR_DMAC_EN | E1000_DMACR_DMAC_LX_MASK);
2364 
2365 		/* Check if status is 2.5Gb backplane connection
2366 		* before configuration of watchdog timer, which is
2367 		* in msec values in 12.8usec intervals
2368 		* watchdog timer= msec values in 32usec intervals
2369 		* for non 2.5Gb connection
2370 		*/
2371 		if (hw->mac.type == e1000_i354) {
2372 			int status = E1000_READ_REG(hw, E1000_STATUS);
2373 			if ((status & E1000_STATUS_2P5_SKU) &&
2374 			    (!(status & E1000_STATUS_2P5_SKU_OVER)))
2375 				reg |= ((adapter->dmac * 5) >> 6);
2376 			else
2377 				reg |= (adapter->dmac >> 5);
2378 		} else {
2379 			reg |= (adapter->dmac >> 5);
2380 		}
2381 
2382 		E1000_WRITE_REG(hw, E1000_DMACR, reg);
2383 
2384 		E1000_WRITE_REG(hw, E1000_DMCRTRH, 0);
2385 
2386 		/* Set the interval before transition */
2387 		reg = E1000_READ_REG(hw, E1000_DMCTLX);
2388 		if (hw->mac.type == e1000_i350)
2389 			reg |= IGB_DMCTLX_DCFLUSH_DIS;
2390 		/*
2391 		** in 2.5Gb connection, TTLX unit is 0.4 usec
2392 		** which is 0x4*2 = 0xA. But delay is still 4 usec
2393 		*/
2394 		if (hw->mac.type == e1000_i354) {
2395 			int status = E1000_READ_REG(hw, E1000_STATUS);
2396 			if ((status & E1000_STATUS_2P5_SKU) &&
2397 			    (!(status & E1000_STATUS_2P5_SKU_OVER)))
2398 				reg |= 0xA;
2399 			else
2400 				reg |= 0x4;
2401 		} else {
2402 			reg |= 0x4;
2403 		}
2404 
2405 		E1000_WRITE_REG(hw, E1000_DMCTLX, reg);
2406 
2407 		/* free space in tx packet buffer to wake from DMA coal */
2408 		E1000_WRITE_REG(hw, E1000_DMCTXTH, (IGB_TXPBSIZE -
2409 		    (2 * max_frame_size)) >> 6);
2410 
2411 		/* make low power state decision controlled by DMA coal */
2412 		reg = E1000_READ_REG(hw, E1000_PCIEMISC);
2413 		reg &= ~E1000_PCIEMISC_LX_DECISION;
2414 		E1000_WRITE_REG(hw, E1000_PCIEMISC, reg);
2415 
2416 	} else if (hw->mac.type == e1000_82580) {
2417 		u32 reg = E1000_READ_REG(hw, E1000_PCIEMISC);
2418 		E1000_WRITE_REG(hw, E1000_PCIEMISC,
2419 		    reg & ~E1000_PCIEMISC_LX_DECISION);
2420 		E1000_WRITE_REG(hw, E1000_DMACR, 0);
2421 	}
2422 }
2423 
2424 /*********************************************************************
2425  *
2426  *  Initialize the hardware to a configuration as specified by the
2427  *  adapter structure.
2428  *
2429  **********************************************************************/
2430 static void
2431 em_reset(if_ctx_t ctx)
2432 {
2433 	device_t dev = iflib_get_dev(ctx);
2434 	struct adapter *adapter = iflib_get_softc(ctx);
2435 	struct ifnet *ifp = iflib_get_ifp(ctx);
2436 	struct e1000_hw *hw = &adapter->hw;
2437 	u16 rx_buffer_size;
2438 	u32 pba;
2439 
2440 	INIT_DEBUGOUT("em_reset: begin");
2441 	/* Let the firmware know the OS is in control */
2442 	em_get_hw_control(adapter);
2443 
2444 	/* Set up smart power down as default off on newer adapters. */
2445 	if (!em_smart_pwr_down && (hw->mac.type == e1000_82571 ||
2446 	    hw->mac.type == e1000_82572)) {
2447 		u16 phy_tmp = 0;
2448 
2449 		/* Speed up time to link by disabling smart power down. */
2450 		e1000_read_phy_reg(hw, IGP02E1000_PHY_POWER_MGMT, &phy_tmp);
2451 		phy_tmp &= ~IGP02E1000_PM_SPD;
2452 		e1000_write_phy_reg(hw, IGP02E1000_PHY_POWER_MGMT, phy_tmp);
2453 	}
2454 
2455 	/*
2456 	 * Packet Buffer Allocation (PBA)
2457 	 * Writing PBA sets the receive portion of the buffer
2458 	 * the remainder is used for the transmit buffer.
2459 	 */
2460 	switch (hw->mac.type) {
2461 	/* Total Packet Buffer on these is 48K */
2462 	case e1000_82571:
2463 	case e1000_82572:
2464 	case e1000_80003es2lan:
2465 			pba = E1000_PBA_32K; /* 32K for Rx, 16K for Tx */
2466 		break;
2467 	case e1000_82573: /* 82573: Total Packet Buffer is 32K */
2468 			pba = E1000_PBA_12K; /* 12K for Rx, 20K for Tx */
2469 		break;
2470 	case e1000_82574:
2471 	case e1000_82583:
2472 			pba = E1000_PBA_20K; /* 20K for Rx, 20K for Tx */
2473 		break;
2474 	case e1000_ich8lan:
2475 		pba = E1000_PBA_8K;
2476 		break;
2477 	case e1000_ich9lan:
2478 	case e1000_ich10lan:
2479 		/* Boost Receive side for jumbo frames */
2480 		if (adapter->hw.mac.max_frame_size > 4096)
2481 			pba = E1000_PBA_14K;
2482 		else
2483 			pba = E1000_PBA_10K;
2484 		break;
2485 	case e1000_pchlan:
2486 	case e1000_pch2lan:
2487 	case e1000_pch_lpt:
2488 	case e1000_pch_spt:
2489 	case e1000_pch_cnp:
2490 		pba = E1000_PBA_26K;
2491 		break;
2492 	case e1000_82575:
2493 		pba = E1000_PBA_32K;
2494 		break;
2495 	case e1000_82576:
2496 	case e1000_vfadapt:
2497 		pba = E1000_READ_REG(hw, E1000_RXPBS);
2498 		pba &= E1000_RXPBS_SIZE_MASK_82576;
2499 		break;
2500 	case e1000_82580:
2501 	case e1000_i350:
2502 	case e1000_i354:
2503 	case e1000_vfadapt_i350:
2504 		pba = E1000_READ_REG(hw, E1000_RXPBS);
2505 		pba = e1000_rxpbs_adjust_82580(pba);
2506 		break;
2507 	case e1000_i210:
2508 	case e1000_i211:
2509 		pba = E1000_PBA_34K;
2510 		break;
2511 	default:
2512 		if (adapter->hw.mac.max_frame_size > 8192)
2513 			pba = E1000_PBA_40K; /* 40K for Rx, 24K for Tx */
2514 		else
2515 			pba = E1000_PBA_48K; /* 48K for Rx, 16K for Tx */
2516 	}
2517 
2518 	/* Special needs in case of Jumbo frames */
2519 	if ((hw->mac.type == e1000_82575) && (ifp->if_mtu > ETHERMTU)) {
2520 		u32 tx_space, min_tx, min_rx;
2521 		pba = E1000_READ_REG(hw, E1000_PBA);
2522 		tx_space = pba >> 16;
2523 		pba &= 0xffff;
2524 		min_tx = (adapter->hw.mac.max_frame_size +
2525 		    sizeof(struct e1000_tx_desc) - ETHERNET_FCS_SIZE) * 2;
2526 		min_tx = roundup2(min_tx, 1024);
2527 		min_tx >>= 10;
2528 		min_rx = adapter->hw.mac.max_frame_size;
2529 		min_rx = roundup2(min_rx, 1024);
2530 		min_rx >>= 10;
2531 		if (tx_space < min_tx &&
2532 		    ((min_tx - tx_space) < pba)) {
2533 			pba = pba - (min_tx - tx_space);
2534 			/*
2535 			 * if short on rx space, rx wins
2536 			 * and must trump tx adjustment
2537 			 */
2538 			if (pba < min_rx)
2539 				pba = min_rx;
2540 		}
2541 		E1000_WRITE_REG(hw, E1000_PBA, pba);
2542 	}
2543 
2544 	if (hw->mac.type < igb_mac_min)
2545 		E1000_WRITE_REG(&adapter->hw, E1000_PBA, pba);
2546 
2547 	INIT_DEBUGOUT1("em_reset: pba=%dK",pba);
2548 
2549 	/*
2550 	 * These parameters control the automatic generation (Tx) and
2551 	 * response (Rx) to Ethernet PAUSE frames.
2552 	 * - High water mark should allow for at least two frames to be
2553 	 *   received after sending an XOFF.
2554 	 * - Low water mark works best when it is very near the high water mark.
2555 	 *   This allows the receiver to restart by sending XON when it has
2556 	 *   drained a bit. Here we use an arbitrary value of 1500 which will
2557 	 *   restart after one full frame is pulled from the buffer. There
2558 	 *   could be several smaller frames in the buffer and if so they will
2559 	 *   not trigger the XON until their total number reduces the buffer
2560 	 *   by 1500.
2561 	 * - The pause time is fairly large at 1000 x 512ns = 512 usec.
2562 	 */
2563 	rx_buffer_size = (pba & 0xffff) << 10;
2564 	hw->fc.high_water = rx_buffer_size -
2565 	    roundup2(adapter->hw.mac.max_frame_size, 1024);
2566 	hw->fc.low_water = hw->fc.high_water - 1500;
2567 
2568 	if (adapter->fc) /* locally set flow control value? */
2569 		hw->fc.requested_mode = adapter->fc;
2570 	else
2571 		hw->fc.requested_mode = e1000_fc_full;
2572 
2573 	if (hw->mac.type == e1000_80003es2lan)
2574 		hw->fc.pause_time = 0xFFFF;
2575 	else
2576 		hw->fc.pause_time = EM_FC_PAUSE_TIME;
2577 
2578 	hw->fc.send_xon = TRUE;
2579 
2580 	/* Device specific overrides/settings */
2581 	switch (hw->mac.type) {
2582 	case e1000_pchlan:
2583 		/* Workaround: no TX flow ctrl for PCH */
2584 		hw->fc.requested_mode = e1000_fc_rx_pause;
2585 		hw->fc.pause_time = 0xFFFF; /* override */
2586 		if (if_getmtu(ifp) > ETHERMTU) {
2587 			hw->fc.high_water = 0x3500;
2588 			hw->fc.low_water = 0x1500;
2589 		} else {
2590 			hw->fc.high_water = 0x5000;
2591 			hw->fc.low_water = 0x3000;
2592 		}
2593 		hw->fc.refresh_time = 0x1000;
2594 		break;
2595 	case e1000_pch2lan:
2596 	case e1000_pch_lpt:
2597 	case e1000_pch_spt:
2598 	case e1000_pch_cnp:
2599 		hw->fc.high_water = 0x5C20;
2600 		hw->fc.low_water = 0x5048;
2601 		hw->fc.pause_time = 0x0650;
2602 		hw->fc.refresh_time = 0x0400;
2603 		/* Jumbos need adjusted PBA */
2604 		if (if_getmtu(ifp) > ETHERMTU)
2605 			E1000_WRITE_REG(hw, E1000_PBA, 12);
2606 		else
2607 			E1000_WRITE_REG(hw, E1000_PBA, 26);
2608 		break;
2609 	case e1000_82575:
2610 	case e1000_82576:
2611 		/* 8-byte granularity */
2612 		hw->fc.low_water = hw->fc.high_water - 8;
2613 		break;
2614 	case e1000_82580:
2615 	case e1000_i350:
2616 	case e1000_i354:
2617 	case e1000_i210:
2618 	case e1000_i211:
2619 	case e1000_vfadapt:
2620 	case e1000_vfadapt_i350:
2621 		/* 16-byte granularity */
2622 		hw->fc.low_water = hw->fc.high_water - 16;
2623 		break;
2624 	case e1000_ich9lan:
2625 	case e1000_ich10lan:
2626 		if (if_getmtu(ifp) > ETHERMTU) {
2627 			hw->fc.high_water = 0x2800;
2628 			hw->fc.low_water = hw->fc.high_water - 8;
2629 			break;
2630 		}
2631 		/* FALLTHROUGH */
2632 	default:
2633 		if (hw->mac.type == e1000_80003es2lan)
2634 			hw->fc.pause_time = 0xFFFF;
2635 		break;
2636 	}
2637 
2638 	/* Issue a global reset */
2639 	e1000_reset_hw(hw);
2640 	if (adapter->hw.mac.type >= igb_mac_min) {
2641 		E1000_WRITE_REG(hw, E1000_WUC, 0);
2642 	} else {
2643 		E1000_WRITE_REG(hw, E1000_WUFC, 0);
2644 		em_disable_aspm(adapter);
2645 	}
2646 	if (adapter->flags & IGB_MEDIA_RESET) {
2647 		e1000_setup_init_funcs(hw, TRUE);
2648 		e1000_get_bus_info(hw);
2649 		adapter->flags &= ~IGB_MEDIA_RESET;
2650 	}
2651 	/* and a re-init */
2652 	if (e1000_init_hw(hw) < 0) {
2653 		device_printf(dev, "Hardware Initialization Failed\n");
2654 		return;
2655 	}
2656 	if (adapter->hw.mac.type >= igb_mac_min)
2657 		igb_init_dmac(adapter, pba);
2658 
2659 	E1000_WRITE_REG(hw, E1000_VET, ETHERTYPE_VLAN);
2660 	e1000_get_phy_info(hw);
2661 	e1000_check_for_link(hw);
2662 }
2663 
2664 /*
2665  * Initialise the RSS mapping for NICs that support multiple transmit/
2666  * receive rings.
2667  */
2668 
2669 #define RSSKEYLEN 10
2670 static void
2671 em_initialize_rss_mapping(struct adapter *adapter)
2672 {
2673 	uint8_t  rss_key[4 * RSSKEYLEN];
2674 	uint32_t reta = 0;
2675 	struct e1000_hw	*hw = &adapter->hw;
2676 	int i;
2677 
2678 	/*
2679 	 * Configure RSS key
2680 	 */
2681 	arc4rand(rss_key, sizeof(rss_key), 0);
2682 	for (i = 0; i < RSSKEYLEN; ++i) {
2683 		uint32_t rssrk = 0;
2684 
2685 		rssrk = EM_RSSRK_VAL(rss_key, i);
2686 		E1000_WRITE_REG(hw,E1000_RSSRK(i), rssrk);
2687 	}
2688 
2689 	/*
2690 	 * Configure RSS redirect table in following fashion:
2691 	 * (hash & ring_cnt_mask) == rdr_table[(hash & rdr_table_mask)]
2692 	 */
2693 	for (i = 0; i < sizeof(reta); ++i) {
2694 		uint32_t q;
2695 
2696 		q = (i % adapter->rx_num_queues) << 7;
2697 		reta |= q << (8 * i);
2698 	}
2699 
2700 	for (i = 0; i < 32; ++i)
2701 		E1000_WRITE_REG(hw, E1000_RETA(i), reta);
2702 
2703 	E1000_WRITE_REG(hw, E1000_MRQC, E1000_MRQC_RSS_ENABLE_2Q |
2704 			E1000_MRQC_RSS_FIELD_IPV4_TCP |
2705 			E1000_MRQC_RSS_FIELD_IPV4 |
2706 			E1000_MRQC_RSS_FIELD_IPV6_TCP_EX |
2707 			E1000_MRQC_RSS_FIELD_IPV6_EX |
2708 			E1000_MRQC_RSS_FIELD_IPV6);
2709 }
2710 
2711 static void
2712 igb_initialize_rss_mapping(struct adapter *adapter)
2713 {
2714 	struct e1000_hw *hw = &adapter->hw;
2715 	int i;
2716 	int queue_id;
2717 	u32 reta;
2718 	u32 rss_key[10], mrqc, shift = 0;
2719 
2720 	/* XXX? */
2721 	if (adapter->hw.mac.type == e1000_82575)
2722 		shift = 6;
2723 
2724 	/*
2725 	 * The redirection table controls which destination
2726 	 * queue each bucket redirects traffic to.
2727 	 * Each DWORD represents four queues, with the LSB
2728 	 * being the first queue in the DWORD.
2729 	 *
2730 	 * This just allocates buckets to queues using round-robin
2731 	 * allocation.
2732 	 *
2733 	 * NOTE: It Just Happens to line up with the default
2734 	 * RSS allocation method.
2735 	 */
2736 
2737 	/* Warning FM follows */
2738 	reta = 0;
2739 	for (i = 0; i < 128; i++) {
2740 #ifdef RSS
2741 		queue_id = rss_get_indirection_to_bucket(i);
2742 		/*
2743 		 * If we have more queues than buckets, we'll
2744 		 * end up mapping buckets to a subset of the
2745 		 * queues.
2746 		 *
2747 		 * If we have more buckets than queues, we'll
2748 		 * end up instead assigning multiple buckets
2749 		 * to queues.
2750 		 *
2751 		 * Both are suboptimal, but we need to handle
2752 		 * the case so we don't go out of bounds
2753 		 * indexing arrays and such.
2754 		 */
2755 		queue_id = queue_id % adapter->rx_num_queues;
2756 #else
2757 		queue_id = (i % adapter->rx_num_queues);
2758 #endif
2759 		/* Adjust if required */
2760 		queue_id = queue_id << shift;
2761 
2762 		/*
2763 		 * The low 8 bits are for hash value (n+0);
2764 		 * The next 8 bits are for hash value (n+1), etc.
2765 		 */
2766 		reta = reta >> 8;
2767 		reta = reta | ( ((uint32_t) queue_id) << 24);
2768 		if ((i & 3) == 3) {
2769 			E1000_WRITE_REG(hw, E1000_RETA(i >> 2), reta);
2770 			reta = 0;
2771 		}
2772 	}
2773 
2774 	/* Now fill in hash table */
2775 
2776 	/*
2777 	 * MRQC: Multiple Receive Queues Command
2778 	 * Set queuing to RSS control, number depends on the device.
2779 	 */
2780 	mrqc = E1000_MRQC_ENABLE_RSS_8Q;
2781 
2782 #ifdef RSS
2783 	/* XXX ew typecasting */
2784 	rss_getkey((uint8_t *) &rss_key);
2785 #else
2786 	arc4rand(&rss_key, sizeof(rss_key), 0);
2787 #endif
2788 	for (i = 0; i < 10; i++)
2789 		E1000_WRITE_REG_ARRAY(hw, E1000_RSSRK(0), i, rss_key[i]);
2790 
2791 	/*
2792 	 * Configure the RSS fields to hash upon.
2793 	 */
2794 	mrqc |= (E1000_MRQC_RSS_FIELD_IPV4 |
2795 	    E1000_MRQC_RSS_FIELD_IPV4_TCP);
2796 	mrqc |= (E1000_MRQC_RSS_FIELD_IPV6 |
2797 	    E1000_MRQC_RSS_FIELD_IPV6_TCP);
2798 	mrqc |=( E1000_MRQC_RSS_FIELD_IPV4_UDP |
2799 	    E1000_MRQC_RSS_FIELD_IPV6_UDP);
2800 	mrqc |=( E1000_MRQC_RSS_FIELD_IPV6_UDP_EX |
2801 	    E1000_MRQC_RSS_FIELD_IPV6_TCP_EX);
2802 
2803 	E1000_WRITE_REG(hw, E1000_MRQC, mrqc);
2804 }
2805 
2806 /*********************************************************************
2807  *
2808  *  Setup networking device structure and register interface media.
2809  *
2810  **********************************************************************/
2811 static int
2812 em_setup_interface(if_ctx_t ctx)
2813 {
2814 	struct ifnet *ifp = iflib_get_ifp(ctx);
2815 	struct adapter *adapter = iflib_get_softc(ctx);
2816 	if_softc_ctx_t scctx = adapter->shared;
2817 
2818 	INIT_DEBUGOUT("em_setup_interface: begin");
2819 
2820 	/* Single Queue */
2821 	if (adapter->tx_num_queues == 1) {
2822 		if_setsendqlen(ifp, scctx->isc_ntxd[0] - 1);
2823 		if_setsendqready(ifp);
2824 	}
2825 
2826 	/*
2827 	 * Specify the media types supported by this adapter and register
2828 	 * callbacks to update media and link information
2829 	 */
2830 	if ((adapter->hw.phy.media_type == e1000_media_type_fiber) ||
2831 	    (adapter->hw.phy.media_type == e1000_media_type_internal_serdes)) {
2832 		u_char fiber_type = IFM_1000_SX;	/* default type */
2833 
2834 		if (adapter->hw.mac.type == e1000_82545)
2835 			fiber_type = IFM_1000_LX;
2836 		ifmedia_add(adapter->media, IFM_ETHER | fiber_type | IFM_FDX, 0, NULL);
2837 		ifmedia_add(adapter->media, IFM_ETHER | fiber_type, 0, NULL);
2838 	} else {
2839 		ifmedia_add(adapter->media, IFM_ETHER | IFM_10_T, 0, NULL);
2840 		ifmedia_add(adapter->media, IFM_ETHER | IFM_10_T | IFM_FDX, 0, NULL);
2841 		ifmedia_add(adapter->media, IFM_ETHER | IFM_100_TX, 0, NULL);
2842 		ifmedia_add(adapter->media, IFM_ETHER | IFM_100_TX | IFM_FDX, 0, NULL);
2843 		if (adapter->hw.phy.type != e1000_phy_ife) {
2844 			ifmedia_add(adapter->media, IFM_ETHER | IFM_1000_T | IFM_FDX, 0, NULL);
2845 			ifmedia_add(adapter->media, IFM_ETHER | IFM_1000_T, 0, NULL);
2846 		}
2847 	}
2848 	ifmedia_add(adapter->media, IFM_ETHER | IFM_AUTO, 0, NULL);
2849 	ifmedia_set(adapter->media, IFM_ETHER | IFM_AUTO);
2850 	return (0);
2851 }
2852 
2853 static int
2854 em_if_tx_queues_alloc(if_ctx_t ctx, caddr_t *vaddrs, uint64_t *paddrs, int ntxqs, int ntxqsets)
2855 {
2856 	struct adapter *adapter = iflib_get_softc(ctx);
2857 	if_softc_ctx_t scctx = adapter->shared;
2858 	int error = E1000_SUCCESS;
2859 	struct em_tx_queue *que;
2860 	int i, j;
2861 
2862 	MPASS(adapter->tx_num_queues > 0);
2863 	MPASS(adapter->tx_num_queues == ntxqsets);
2864 
2865 	/* First allocate the top level queue structs */
2866 	if (!(adapter->tx_queues =
2867 	    (struct em_tx_queue *) malloc(sizeof(struct em_tx_queue) *
2868 	    adapter->tx_num_queues, M_DEVBUF, M_NOWAIT | M_ZERO))) {
2869 		device_printf(iflib_get_dev(ctx), "Unable to allocate queue memory\n");
2870 		return(ENOMEM);
2871 	}
2872 
2873 	for (i = 0, que = adapter->tx_queues; i < adapter->tx_num_queues; i++, que++) {
2874 		/* Set up some basics */
2875 
2876 		struct tx_ring *txr = &que->txr;
2877 		txr->adapter = que->adapter = adapter;
2878 		que->me = txr->me =  i;
2879 
2880 		/* Allocate report status array */
2881 		if (!(txr->tx_rsq = (qidx_t *) malloc(sizeof(qidx_t) * scctx->isc_ntxd[0], M_DEVBUF, M_NOWAIT | M_ZERO))) {
2882 			device_printf(iflib_get_dev(ctx), "failed to allocate rs_idxs memory\n");
2883 			error = ENOMEM;
2884 			goto fail;
2885 		}
2886 		for (j = 0; j < scctx->isc_ntxd[0]; j++)
2887 			txr->tx_rsq[j] = QIDX_INVALID;
2888 		/* get the virtual and physical address of the hardware queues */
2889 		txr->tx_base = (struct e1000_tx_desc *)vaddrs[i*ntxqs];
2890 		txr->tx_paddr = paddrs[i*ntxqs];
2891 	}
2892 
2893 	if (bootverbose)
2894 		device_printf(iflib_get_dev(ctx),
2895 		    "allocated for %d tx_queues\n", adapter->tx_num_queues);
2896 	return (0);
2897 fail:
2898 	em_if_queues_free(ctx);
2899 	return (error);
2900 }
2901 
2902 static int
2903 em_if_rx_queues_alloc(if_ctx_t ctx, caddr_t *vaddrs, uint64_t *paddrs, int nrxqs, int nrxqsets)
2904 {
2905 	struct adapter *adapter = iflib_get_softc(ctx);
2906 	int error = E1000_SUCCESS;
2907 	struct em_rx_queue *que;
2908 	int i;
2909 
2910 	MPASS(adapter->rx_num_queues > 0);
2911 	MPASS(adapter->rx_num_queues == nrxqsets);
2912 
2913 	/* First allocate the top level queue structs */
2914 	if (!(adapter->rx_queues =
2915 	    (struct em_rx_queue *) malloc(sizeof(struct em_rx_queue) *
2916 	    adapter->rx_num_queues, M_DEVBUF, M_NOWAIT | M_ZERO))) {
2917 		device_printf(iflib_get_dev(ctx), "Unable to allocate queue memory\n");
2918 		error = ENOMEM;
2919 		goto fail;
2920 	}
2921 
2922 	for (i = 0, que = adapter->rx_queues; i < nrxqsets; i++, que++) {
2923 		/* Set up some basics */
2924 		struct rx_ring *rxr = &que->rxr;
2925 		rxr->adapter = que->adapter = adapter;
2926 		rxr->que = que;
2927 		que->me = rxr->me =  i;
2928 
2929 		/* get the virtual and physical address of the hardware queues */
2930 		rxr->rx_base = (union e1000_rx_desc_extended *)vaddrs[i*nrxqs];
2931 		rxr->rx_paddr = paddrs[i*nrxqs];
2932 	}
2933 
2934 	if (bootverbose)
2935 		device_printf(iflib_get_dev(ctx),
2936 		    "allocated for %d rx_queues\n", adapter->rx_num_queues);
2937 
2938 	return (0);
2939 fail:
2940 	em_if_queues_free(ctx);
2941 	return (error);
2942 }
2943 
2944 static void
2945 em_if_queues_free(if_ctx_t ctx)
2946 {
2947 	struct adapter *adapter = iflib_get_softc(ctx);
2948 	struct em_tx_queue *tx_que = adapter->tx_queues;
2949 	struct em_rx_queue *rx_que = adapter->rx_queues;
2950 
2951 	if (tx_que != NULL) {
2952 		for (int i = 0; i < adapter->tx_num_queues; i++, tx_que++) {
2953 			struct tx_ring *txr = &tx_que->txr;
2954 			if (txr->tx_rsq == NULL)
2955 				break;
2956 
2957 			free(txr->tx_rsq, M_DEVBUF);
2958 			txr->tx_rsq = NULL;
2959 		}
2960 		free(adapter->tx_queues, M_DEVBUF);
2961 		adapter->tx_queues = NULL;
2962 	}
2963 
2964 	if (rx_que != NULL) {
2965 		free(adapter->rx_queues, M_DEVBUF);
2966 		adapter->rx_queues = NULL;
2967 	}
2968 
2969 	em_release_hw_control(adapter);
2970 
2971 	if (adapter->mta != NULL) {
2972 		free(adapter->mta, M_DEVBUF);
2973 	}
2974 }
2975 
2976 /*********************************************************************
2977  *
2978  *  Enable transmit unit.
2979  *
2980  **********************************************************************/
2981 static void
2982 em_initialize_transmit_unit(if_ctx_t ctx)
2983 {
2984 	struct adapter *adapter = iflib_get_softc(ctx);
2985 	if_softc_ctx_t scctx = adapter->shared;
2986 	struct em_tx_queue *que;
2987 	struct tx_ring	*txr;
2988 	struct e1000_hw	*hw = &adapter->hw;
2989 	u32 tctl, txdctl = 0, tarc, tipg = 0;
2990 
2991 	INIT_DEBUGOUT("em_initialize_transmit_unit: begin");
2992 
2993 	for (int i = 0; i < adapter->tx_num_queues; i++, txr++) {
2994 		u64 bus_addr;
2995 		caddr_t offp, endp;
2996 
2997 		que = &adapter->tx_queues[i];
2998 		txr = &que->txr;
2999 		bus_addr = txr->tx_paddr;
3000 
3001 		/* Clear checksum offload context. */
3002 		offp = (caddr_t)&txr->csum_flags;
3003 		endp = (caddr_t)(txr + 1);
3004 		bzero(offp, endp - offp);
3005 
3006 		/* Base and Len of TX Ring */
3007 		E1000_WRITE_REG(hw, E1000_TDLEN(i),
3008 		    scctx->isc_ntxd[0] * sizeof(struct e1000_tx_desc));
3009 		E1000_WRITE_REG(hw, E1000_TDBAH(i),
3010 		    (u32)(bus_addr >> 32));
3011 		E1000_WRITE_REG(hw, E1000_TDBAL(i),
3012 		    (u32)bus_addr);
3013 		/* Init the HEAD/TAIL indices */
3014 		E1000_WRITE_REG(hw, E1000_TDT(i), 0);
3015 		E1000_WRITE_REG(hw, E1000_TDH(i), 0);
3016 
3017 		HW_DEBUGOUT2("Base = %x, Length = %x\n",
3018 		    E1000_READ_REG(&adapter->hw, E1000_TDBAL(i)),
3019 		    E1000_READ_REG(&adapter->hw, E1000_TDLEN(i)));
3020 
3021 		txdctl = 0; /* clear txdctl */
3022 		txdctl |= 0x1f; /* PTHRESH */
3023 		txdctl |= 1 << 8; /* HTHRESH */
3024 		txdctl |= 1 << 16;/* WTHRESH */
3025 		txdctl |= 1 << 22; /* Reserved bit 22 must always be 1 */
3026 		txdctl |= E1000_TXDCTL_GRAN;
3027 		txdctl |= 1 << 25; /* LWTHRESH */
3028 
3029 		E1000_WRITE_REG(hw, E1000_TXDCTL(i), txdctl);
3030 	}
3031 
3032 	/* Set the default values for the Tx Inter Packet Gap timer */
3033 	switch (adapter->hw.mac.type) {
3034 	case e1000_80003es2lan:
3035 		tipg = DEFAULT_82543_TIPG_IPGR1;
3036 		tipg |= DEFAULT_80003ES2LAN_TIPG_IPGR2 <<
3037 		    E1000_TIPG_IPGR2_SHIFT;
3038 		break;
3039 	case e1000_82542:
3040 		tipg = DEFAULT_82542_TIPG_IPGT;
3041 		tipg |= DEFAULT_82542_TIPG_IPGR1 << E1000_TIPG_IPGR1_SHIFT;
3042 		tipg |= DEFAULT_82542_TIPG_IPGR2 << E1000_TIPG_IPGR2_SHIFT;
3043 		break;
3044 	default:
3045 		if ((adapter->hw.phy.media_type == e1000_media_type_fiber) ||
3046 		    (adapter->hw.phy.media_type ==
3047 		    e1000_media_type_internal_serdes))
3048 			tipg = DEFAULT_82543_TIPG_IPGT_FIBER;
3049 		else
3050 			tipg = DEFAULT_82543_TIPG_IPGT_COPPER;
3051 		tipg |= DEFAULT_82543_TIPG_IPGR1 << E1000_TIPG_IPGR1_SHIFT;
3052 		tipg |= DEFAULT_82543_TIPG_IPGR2 << E1000_TIPG_IPGR2_SHIFT;
3053 	}
3054 
3055 	E1000_WRITE_REG(&adapter->hw, E1000_TIPG, tipg);
3056 	E1000_WRITE_REG(&adapter->hw, E1000_TIDV, adapter->tx_int_delay.value);
3057 
3058 	if(adapter->hw.mac.type >= e1000_82540)
3059 		E1000_WRITE_REG(&adapter->hw, E1000_TADV,
3060 		    adapter->tx_abs_int_delay.value);
3061 
3062 	if ((adapter->hw.mac.type == e1000_82571) ||
3063 	    (adapter->hw.mac.type == e1000_82572)) {
3064 		tarc = E1000_READ_REG(&adapter->hw, E1000_TARC(0));
3065 		tarc |= TARC_SPEED_MODE_BIT;
3066 		E1000_WRITE_REG(&adapter->hw, E1000_TARC(0), tarc);
3067 	} else if (adapter->hw.mac.type == e1000_80003es2lan) {
3068 		/* errata: program both queues to unweighted RR */
3069 		tarc = E1000_READ_REG(&adapter->hw, E1000_TARC(0));
3070 		tarc |= 1;
3071 		E1000_WRITE_REG(&adapter->hw, E1000_TARC(0), tarc);
3072 		tarc = E1000_READ_REG(&adapter->hw, E1000_TARC(1));
3073 		tarc |= 1;
3074 		E1000_WRITE_REG(&adapter->hw, E1000_TARC(1), tarc);
3075 	} else if (adapter->hw.mac.type == e1000_82574) {
3076 		tarc = E1000_READ_REG(&adapter->hw, E1000_TARC(0));
3077 		tarc |= TARC_ERRATA_BIT;
3078 		if ( adapter->tx_num_queues > 1) {
3079 			tarc |= (TARC_COMPENSATION_MODE | TARC_MQ_FIX);
3080 			E1000_WRITE_REG(&adapter->hw, E1000_TARC(0), tarc);
3081 			E1000_WRITE_REG(&adapter->hw, E1000_TARC(1), tarc);
3082 		} else
3083 			E1000_WRITE_REG(&adapter->hw, E1000_TARC(0), tarc);
3084 	}
3085 
3086 	if (adapter->tx_int_delay.value > 0)
3087 		adapter->txd_cmd |= E1000_TXD_CMD_IDE;
3088 
3089 	/* Program the Transmit Control Register */
3090 	tctl = E1000_READ_REG(&adapter->hw, E1000_TCTL);
3091 	tctl &= ~E1000_TCTL_CT;
3092 	tctl |= (E1000_TCTL_PSP | E1000_TCTL_RTLC | E1000_TCTL_EN |
3093 		   (E1000_COLLISION_THRESHOLD << E1000_CT_SHIFT));
3094 
3095 	if (adapter->hw.mac.type >= e1000_82571)
3096 		tctl |= E1000_TCTL_MULR;
3097 
3098 	/* This write will effectively turn on the transmit unit. */
3099 	E1000_WRITE_REG(&adapter->hw, E1000_TCTL, tctl);
3100 
3101 	/* SPT and KBL errata workarounds */
3102 	if (hw->mac.type == e1000_pch_spt) {
3103 		u32 reg;
3104 		reg = E1000_READ_REG(hw, E1000_IOSFPC);
3105 		reg |= E1000_RCTL_RDMTS_HEX;
3106 		E1000_WRITE_REG(hw, E1000_IOSFPC, reg);
3107 		/* i218-i219 Specification Update 1.5.4.5 */
3108 		reg = E1000_READ_REG(hw, E1000_TARC(0));
3109 		reg &= ~E1000_TARC0_CB_MULTIQ_3_REQ;
3110 		reg |= E1000_TARC0_CB_MULTIQ_2_REQ;
3111 		E1000_WRITE_REG(hw, E1000_TARC(0), reg);
3112 	}
3113 }
3114 
3115 /*********************************************************************
3116  *
3117  *  Enable receive unit.
3118  *
3119  **********************************************************************/
3120 
3121 static void
3122 em_initialize_receive_unit(if_ctx_t ctx)
3123 {
3124 	struct adapter *adapter = iflib_get_softc(ctx);
3125 	if_softc_ctx_t scctx = adapter->shared;
3126 	struct ifnet *ifp = iflib_get_ifp(ctx);
3127 	struct e1000_hw	*hw = &adapter->hw;
3128 	struct em_rx_queue *que;
3129 	int i;
3130 	u32 rctl, rxcsum, rfctl;
3131 
3132 	INIT_DEBUGOUT("em_initialize_receive_units: begin");
3133 
3134 	/*
3135 	 * Make sure receives are disabled while setting
3136 	 * up the descriptor ring
3137 	 */
3138 	rctl = E1000_READ_REG(hw, E1000_RCTL);
3139 	/* Do not disable if ever enabled on this hardware */
3140 	if ((hw->mac.type != e1000_82574) && (hw->mac.type != e1000_82583))
3141 		E1000_WRITE_REG(hw, E1000_RCTL, rctl & ~E1000_RCTL_EN);
3142 
3143 	/* Setup the Receive Control Register */
3144 	rctl &= ~(3 << E1000_RCTL_MO_SHIFT);
3145 	rctl |= E1000_RCTL_EN | E1000_RCTL_BAM |
3146 	    E1000_RCTL_LBM_NO | E1000_RCTL_RDMTS_HALF |
3147 	    (hw->mac.mc_filter_type << E1000_RCTL_MO_SHIFT);
3148 
3149 	/* Do not store bad packets */
3150 	rctl &= ~E1000_RCTL_SBP;
3151 
3152 	/* Enable Long Packet receive */
3153 	if (if_getmtu(ifp) > ETHERMTU)
3154 		rctl |= E1000_RCTL_LPE;
3155 	else
3156 		rctl &= ~E1000_RCTL_LPE;
3157 
3158 	/* Strip the CRC */
3159 	if (!em_disable_crc_stripping)
3160 		rctl |= E1000_RCTL_SECRC;
3161 
3162 	if (adapter->hw.mac.type >= e1000_82540) {
3163 		E1000_WRITE_REG(&adapter->hw, E1000_RADV,
3164 			    adapter->rx_abs_int_delay.value);
3165 
3166 		/*
3167 		 * Set the interrupt throttling rate. Value is calculated
3168 		 * as DEFAULT_ITR = 1/(MAX_INTS_PER_SEC * 256ns)
3169 		 */
3170 		E1000_WRITE_REG(hw, E1000_ITR, DEFAULT_ITR);
3171 	}
3172 	E1000_WRITE_REG(&adapter->hw, E1000_RDTR,
3173 	    adapter->rx_int_delay.value);
3174 
3175 	/* Use extended rx descriptor formats */
3176 	rfctl = E1000_READ_REG(hw, E1000_RFCTL);
3177 	rfctl |= E1000_RFCTL_EXTEN;
3178 	/*
3179 	 * When using MSI-X interrupts we need to throttle
3180 	 * using the EITR register (82574 only)
3181 	 */
3182 	if (hw->mac.type == e1000_82574) {
3183 		for (int i = 0; i < 4; i++)
3184 			E1000_WRITE_REG(hw, E1000_EITR_82574(i),
3185 			    DEFAULT_ITR);
3186 		/* Disable accelerated acknowledge */
3187 		rfctl |= E1000_RFCTL_ACK_DIS;
3188 	}
3189 	E1000_WRITE_REG(hw, E1000_RFCTL, rfctl);
3190 
3191 	rxcsum = E1000_READ_REG(hw, E1000_RXCSUM);
3192 	if (if_getcapenable(ifp) & IFCAP_RXCSUM &&
3193 	    adapter->hw.mac.type >= e1000_82543) {
3194 		if (adapter->tx_num_queues > 1) {
3195 			if (adapter->hw.mac.type >= igb_mac_min) {
3196 				rxcsum |= E1000_RXCSUM_PCSD;
3197 				if (hw->mac.type != e1000_82575)
3198 					rxcsum |= E1000_RXCSUM_CRCOFL;
3199 			} else
3200 				rxcsum |= E1000_RXCSUM_TUOFL |
3201 					E1000_RXCSUM_IPOFL |
3202 					E1000_RXCSUM_PCSD;
3203 		} else {
3204 			if (adapter->hw.mac.type >= igb_mac_min)
3205 				rxcsum |= E1000_RXCSUM_IPPCSE;
3206 			else
3207 				rxcsum |= E1000_RXCSUM_TUOFL | E1000_RXCSUM_IPOFL;
3208 			if (adapter->hw.mac.type > e1000_82575)
3209 				rxcsum |= E1000_RXCSUM_CRCOFL;
3210 		}
3211 	} else
3212 		rxcsum &= ~E1000_RXCSUM_TUOFL;
3213 
3214 	E1000_WRITE_REG(hw, E1000_RXCSUM, rxcsum);
3215 
3216 	if (adapter->rx_num_queues > 1) {
3217 		if (adapter->hw.mac.type >= igb_mac_min)
3218 			igb_initialize_rss_mapping(adapter);
3219 		else
3220 			em_initialize_rss_mapping(adapter);
3221 	}
3222 
3223 	/*
3224 	 * XXX TEMPORARY WORKAROUND: on some systems with 82573
3225 	 * long latencies are observed, like Lenovo X60. This
3226 	 * change eliminates the problem, but since having positive
3227 	 * values in RDTR is a known source of problems on other
3228 	 * platforms another solution is being sought.
3229 	 */
3230 	if (hw->mac.type == e1000_82573)
3231 		E1000_WRITE_REG(hw, E1000_RDTR, 0x20);
3232 
3233 	for (i = 0, que = adapter->rx_queues; i < adapter->rx_num_queues; i++, que++) {
3234 		struct rx_ring *rxr = &que->rxr;
3235 		/* Setup the Base and Length of the Rx Descriptor Ring */
3236 		u64 bus_addr = rxr->rx_paddr;
3237 #if 0
3238 		u32 rdt = adapter->rx_num_queues -1;  /* default */
3239 #endif
3240 
3241 		E1000_WRITE_REG(hw, E1000_RDLEN(i),
3242 		    scctx->isc_nrxd[0] * sizeof(union e1000_rx_desc_extended));
3243 		E1000_WRITE_REG(hw, E1000_RDBAH(i), (u32)(bus_addr >> 32));
3244 		E1000_WRITE_REG(hw, E1000_RDBAL(i), (u32)bus_addr);
3245 		/* Setup the Head and Tail Descriptor Pointers */
3246 		E1000_WRITE_REG(hw, E1000_RDH(i), 0);
3247 		E1000_WRITE_REG(hw, E1000_RDT(i), 0);
3248 	}
3249 
3250 	/*
3251 	 * Set PTHRESH for improved jumbo performance
3252 	 * According to 10.2.5.11 of Intel 82574 Datasheet,
3253 	 * RXDCTL(1) is written whenever RXDCTL(0) is written.
3254 	 * Only write to RXDCTL(1) if there is a need for different
3255 	 * settings.
3256 	 */
3257 
3258 	if (((adapter->hw.mac.type == e1000_ich9lan) ||
3259 	    (adapter->hw.mac.type == e1000_pch2lan) ||
3260 	    (adapter->hw.mac.type == e1000_ich10lan)) &&
3261 	    (if_getmtu(ifp) > ETHERMTU)) {
3262 		u32 rxdctl = E1000_READ_REG(hw, E1000_RXDCTL(0));
3263 		E1000_WRITE_REG(hw, E1000_RXDCTL(0), rxdctl | 3);
3264 	} else if (adapter->hw.mac.type == e1000_82574) {
3265 		for (int i = 0; i < adapter->rx_num_queues; i++) {
3266 			u32 rxdctl = E1000_READ_REG(hw, E1000_RXDCTL(i));
3267 			rxdctl |= 0x20; /* PTHRESH */
3268 			rxdctl |= 4 << 8; /* HTHRESH */
3269 			rxdctl |= 4 << 16;/* WTHRESH */
3270 			rxdctl |= 1 << 24; /* Switch to granularity */
3271 			E1000_WRITE_REG(hw, E1000_RXDCTL(i), rxdctl);
3272 		}
3273 	} else if (adapter->hw.mac.type >= igb_mac_min) {
3274 		u32 psize, srrctl = 0;
3275 
3276 		if (if_getmtu(ifp) > ETHERMTU) {
3277 			/* Set maximum packet len */
3278 			if (adapter->rx_mbuf_sz <= 4096) {
3279 				srrctl |= 4096 >> E1000_SRRCTL_BSIZEPKT_SHIFT;
3280 				rctl |= E1000_RCTL_SZ_4096 | E1000_RCTL_BSEX;
3281 			} else if (adapter->rx_mbuf_sz > 4096) {
3282 				srrctl |= 8192 >> E1000_SRRCTL_BSIZEPKT_SHIFT;
3283 				rctl |= E1000_RCTL_SZ_8192 | E1000_RCTL_BSEX;
3284 			}
3285 			psize = scctx->isc_max_frame_size;
3286 			/* are we on a vlan? */
3287 			if (ifp->if_vlantrunk != NULL)
3288 				psize += VLAN_TAG_SIZE;
3289 			E1000_WRITE_REG(&adapter->hw, E1000_RLPML, psize);
3290 		} else {
3291 			srrctl |= 2048 >> E1000_SRRCTL_BSIZEPKT_SHIFT;
3292 			rctl |= E1000_RCTL_SZ_2048;
3293 		}
3294 
3295 		/*
3296 		 * If TX flow control is disabled and there's >1 queue defined,
3297 		 * enable DROP.
3298 		 *
3299 		 * This drops frames rather than hanging the RX MAC for all queues.
3300 		 */
3301 		if ((adapter->rx_num_queues > 1) &&
3302 		    (adapter->fc == e1000_fc_none ||
3303 		     adapter->fc == e1000_fc_rx_pause)) {
3304 			srrctl |= E1000_SRRCTL_DROP_EN;
3305 		}
3306 			/* Setup the Base and Length of the Rx Descriptor Rings */
3307 		for (i = 0, que = adapter->rx_queues; i < adapter->rx_num_queues; i++, que++) {
3308 			struct rx_ring *rxr = &que->rxr;
3309 			u64 bus_addr = rxr->rx_paddr;
3310 			u32 rxdctl;
3311 
3312 #ifdef notyet
3313 			/* Configure for header split? -- ignore for now */
3314 			rxr->hdr_split = igb_header_split;
3315 #else
3316 			srrctl |= E1000_SRRCTL_DESCTYPE_ADV_ONEBUF;
3317 #endif
3318 
3319 			E1000_WRITE_REG(hw, E1000_RDLEN(i),
3320 					scctx->isc_nrxd[0] * sizeof(struct e1000_rx_desc));
3321 			E1000_WRITE_REG(hw, E1000_RDBAH(i),
3322 					(uint32_t)(bus_addr >> 32));
3323 			E1000_WRITE_REG(hw, E1000_RDBAL(i),
3324 					(uint32_t)bus_addr);
3325 			E1000_WRITE_REG(hw, E1000_SRRCTL(i), srrctl);
3326 			/* Enable this Queue */
3327 			rxdctl = E1000_READ_REG(hw, E1000_RXDCTL(i));
3328 			rxdctl |= E1000_RXDCTL_QUEUE_ENABLE;
3329 			rxdctl &= 0xFFF00000;
3330 			rxdctl |= IGB_RX_PTHRESH;
3331 			rxdctl |= IGB_RX_HTHRESH << 8;
3332 			rxdctl |= IGB_RX_WTHRESH << 16;
3333 			E1000_WRITE_REG(hw, E1000_RXDCTL(i), rxdctl);
3334 		}
3335 	} else if (adapter->hw.mac.type >= e1000_pch2lan) {
3336 		if (if_getmtu(ifp) > ETHERMTU)
3337 			e1000_lv_jumbo_workaround_ich8lan(hw, TRUE);
3338 		else
3339 			e1000_lv_jumbo_workaround_ich8lan(hw, FALSE);
3340 	}
3341 
3342 	/* Make sure VLAN Filters are off */
3343 	rctl &= ~E1000_RCTL_VFE;
3344 
3345 	if (adapter->hw.mac.type < igb_mac_min) {
3346 		if (adapter->rx_mbuf_sz == MCLBYTES)
3347 			rctl |= E1000_RCTL_SZ_2048;
3348 		else if (adapter->rx_mbuf_sz == MJUMPAGESIZE)
3349 			rctl |= E1000_RCTL_SZ_4096 | E1000_RCTL_BSEX;
3350 		else if (adapter->rx_mbuf_sz > MJUMPAGESIZE)
3351 			rctl |= E1000_RCTL_SZ_8192 | E1000_RCTL_BSEX;
3352 
3353 		/* ensure we clear use DTYPE of 00 here */
3354 		rctl &= ~0x00000C00;
3355 	}
3356 
3357 	/* Write out the settings */
3358 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
3359 
3360 	return;
3361 }
3362 
3363 static void
3364 em_if_vlan_register(if_ctx_t ctx, u16 vtag)
3365 {
3366 	struct adapter *adapter = iflib_get_softc(ctx);
3367 	u32 index, bit;
3368 
3369 	index = (vtag >> 5) & 0x7F;
3370 	bit = vtag & 0x1F;
3371 	adapter->shadow_vfta[index] |= (1 << bit);
3372 	++adapter->num_vlans;
3373 }
3374 
3375 static void
3376 em_if_vlan_unregister(if_ctx_t ctx, u16 vtag)
3377 {
3378 	struct adapter *adapter = iflib_get_softc(ctx);
3379 	u32 index, bit;
3380 
3381 	index = (vtag >> 5) & 0x7F;
3382 	bit = vtag & 0x1F;
3383 	adapter->shadow_vfta[index] &= ~(1 << bit);
3384 	--adapter->num_vlans;
3385 }
3386 
3387 static void
3388 em_setup_vlan_hw_support(struct adapter *adapter)
3389 {
3390 	struct e1000_hw *hw = &adapter->hw;
3391 	u32 reg;
3392 
3393 	/*
3394 	 * We get here thru init_locked, meaning
3395 	 * a soft reset, this has already cleared
3396 	 * the VFTA and other state, so if there
3397 	 * have been no vlan's registered do nothing.
3398 	 */
3399 	if (adapter->num_vlans == 0)
3400 		return;
3401 
3402 	/*
3403 	 * A soft reset zero's out the VFTA, so
3404 	 * we need to repopulate it now.
3405 	 */
3406 	for (int i = 0; i < EM_VFTA_SIZE; i++)
3407 		if (adapter->shadow_vfta[i] != 0)
3408 			E1000_WRITE_REG_ARRAY(hw, E1000_VFTA,
3409 			    i, adapter->shadow_vfta[i]);
3410 
3411 	reg = E1000_READ_REG(hw, E1000_CTRL);
3412 	reg |= E1000_CTRL_VME;
3413 	E1000_WRITE_REG(hw, E1000_CTRL, reg);
3414 
3415 	/* Enable the Filter Table */
3416 	reg = E1000_READ_REG(hw, E1000_RCTL);
3417 	reg &= ~E1000_RCTL_CFIEN;
3418 	reg |= E1000_RCTL_VFE;
3419 	E1000_WRITE_REG(hw, E1000_RCTL, reg);
3420 }
3421 
3422 static void
3423 em_if_intr_enable(if_ctx_t ctx)
3424 {
3425 	struct adapter *adapter = iflib_get_softc(ctx);
3426 	struct e1000_hw *hw = &adapter->hw;
3427 	u32 ims_mask = IMS_ENABLE_MASK;
3428 
3429 	if (hw->mac.type == e1000_82574) {
3430 		E1000_WRITE_REG(hw, EM_EIAC, EM_MSIX_MASK);
3431 		ims_mask |= adapter->ims;
3432 	}
3433 	E1000_WRITE_REG(hw, E1000_IMS, ims_mask);
3434 }
3435 
3436 static void
3437 em_if_intr_disable(if_ctx_t ctx)
3438 {
3439 	struct adapter *adapter = iflib_get_softc(ctx);
3440 	struct e1000_hw *hw = &adapter->hw;
3441 
3442 	if (hw->mac.type == e1000_82574)
3443 		E1000_WRITE_REG(hw, EM_EIAC, 0);
3444 	E1000_WRITE_REG(hw, E1000_IMC, 0xffffffff);
3445 }
3446 
3447 static void
3448 igb_if_intr_enable(if_ctx_t ctx)
3449 {
3450 	struct adapter *adapter = iflib_get_softc(ctx);
3451 	struct e1000_hw *hw = &adapter->hw;
3452 	u32 mask;
3453 
3454 	if (__predict_true(adapter->intr_type == IFLIB_INTR_MSIX)) {
3455 		mask = (adapter->que_mask | adapter->link_mask);
3456 		E1000_WRITE_REG(hw, E1000_EIAC, mask);
3457 		E1000_WRITE_REG(hw, E1000_EIAM, mask);
3458 		E1000_WRITE_REG(hw, E1000_EIMS, mask);
3459 		E1000_WRITE_REG(hw, E1000_IMS, E1000_IMS_LSC);
3460 	} else
3461 		E1000_WRITE_REG(hw, E1000_IMS, IMS_ENABLE_MASK);
3462 	E1000_WRITE_FLUSH(hw);
3463 }
3464 
3465 static void
3466 igb_if_intr_disable(if_ctx_t ctx)
3467 {
3468 	struct adapter *adapter = iflib_get_softc(ctx);
3469 	struct e1000_hw *hw = &adapter->hw;
3470 
3471 	if (__predict_true(adapter->intr_type == IFLIB_INTR_MSIX)) {
3472 		E1000_WRITE_REG(hw, E1000_EIMC, 0xffffffff);
3473 		E1000_WRITE_REG(hw, E1000_EIAC, 0);
3474 	}
3475 	E1000_WRITE_REG(hw, E1000_IMC, 0xffffffff);
3476 	E1000_WRITE_FLUSH(hw);
3477 }
3478 
3479 /*
3480  * Bit of a misnomer, what this really means is
3481  * to enable OS management of the system... aka
3482  * to disable special hardware management features
3483  */
3484 static void
3485 em_init_manageability(struct adapter *adapter)
3486 {
3487 	/* A shared code workaround */
3488 #define E1000_82542_MANC2H E1000_MANC2H
3489 	if (adapter->has_manage) {
3490 		int manc2h = E1000_READ_REG(&adapter->hw, E1000_MANC2H);
3491 		int manc = E1000_READ_REG(&adapter->hw, E1000_MANC);
3492 
3493 		/* disable hardware interception of ARP */
3494 		manc &= ~(E1000_MANC_ARP_EN);
3495 
3496 		/* enable receiving management packets to the host */
3497 		manc |= E1000_MANC_EN_MNG2HOST;
3498 #define E1000_MNG2HOST_PORT_623 (1 << 5)
3499 #define E1000_MNG2HOST_PORT_664 (1 << 6)
3500 		manc2h |= E1000_MNG2HOST_PORT_623;
3501 		manc2h |= E1000_MNG2HOST_PORT_664;
3502 		E1000_WRITE_REG(&adapter->hw, E1000_MANC2H, manc2h);
3503 		E1000_WRITE_REG(&adapter->hw, E1000_MANC, manc);
3504 	}
3505 }
3506 
3507 /*
3508  * Give control back to hardware management
3509  * controller if there is one.
3510  */
3511 static void
3512 em_release_manageability(struct adapter *adapter)
3513 {
3514 	if (adapter->has_manage) {
3515 		int manc = E1000_READ_REG(&adapter->hw, E1000_MANC);
3516 
3517 		/* re-enable hardware interception of ARP */
3518 		manc |= E1000_MANC_ARP_EN;
3519 		manc &= ~E1000_MANC_EN_MNG2HOST;
3520 
3521 		E1000_WRITE_REG(&adapter->hw, E1000_MANC, manc);
3522 	}
3523 }
3524 
3525 /*
3526  * em_get_hw_control sets the {CTRL_EXT|FWSM}:DRV_LOAD bit.
3527  * For ASF and Pass Through versions of f/w this means
3528  * that the driver is loaded. For AMT version type f/w
3529  * this means that the network i/f is open.
3530  */
3531 static void
3532 em_get_hw_control(struct adapter *adapter)
3533 {
3534 	u32 ctrl_ext, swsm;
3535 
3536 	if (adapter->vf_ifp)
3537 		return;
3538 
3539 	if (adapter->hw.mac.type == e1000_82573) {
3540 		swsm = E1000_READ_REG(&adapter->hw, E1000_SWSM);
3541 		E1000_WRITE_REG(&adapter->hw, E1000_SWSM,
3542 		    swsm | E1000_SWSM_DRV_LOAD);
3543 		return;
3544 	}
3545 	/* else */
3546 	ctrl_ext = E1000_READ_REG(&adapter->hw, E1000_CTRL_EXT);
3547 	E1000_WRITE_REG(&adapter->hw, E1000_CTRL_EXT,
3548 	    ctrl_ext | E1000_CTRL_EXT_DRV_LOAD);
3549 }
3550 
3551 /*
3552  * em_release_hw_control resets {CTRL_EXT|FWSM}:DRV_LOAD bit.
3553  * For ASF and Pass Through versions of f/w this means that
3554  * the driver is no longer loaded. For AMT versions of the
3555  * f/w this means that the network i/f is closed.
3556  */
3557 static void
3558 em_release_hw_control(struct adapter *adapter)
3559 {
3560 	u32 ctrl_ext, swsm;
3561 
3562 	if (!adapter->has_manage)
3563 		return;
3564 
3565 	if (adapter->hw.mac.type == e1000_82573) {
3566 		swsm = E1000_READ_REG(&adapter->hw, E1000_SWSM);
3567 		E1000_WRITE_REG(&adapter->hw, E1000_SWSM,
3568 		    swsm & ~E1000_SWSM_DRV_LOAD);
3569 		return;
3570 	}
3571 	/* else */
3572 	ctrl_ext = E1000_READ_REG(&adapter->hw, E1000_CTRL_EXT);
3573 	E1000_WRITE_REG(&adapter->hw, E1000_CTRL_EXT,
3574 	    ctrl_ext & ~E1000_CTRL_EXT_DRV_LOAD);
3575 	return;
3576 }
3577 
3578 static int
3579 em_is_valid_ether_addr(u8 *addr)
3580 {
3581 	char zero_addr[6] = { 0, 0, 0, 0, 0, 0 };
3582 
3583 	if ((addr[0] & 1) || (!bcmp(addr, zero_addr, ETHER_ADDR_LEN))) {
3584 		return (FALSE);
3585 	}
3586 
3587 	return (TRUE);
3588 }
3589 
3590 /*
3591 ** Parse the interface capabilities with regard
3592 ** to both system management and wake-on-lan for
3593 ** later use.
3594 */
3595 static void
3596 em_get_wakeup(if_ctx_t ctx)
3597 {
3598 	struct adapter *adapter = iflib_get_softc(ctx);
3599 	device_t dev = iflib_get_dev(ctx);
3600 	u16 eeprom_data = 0, device_id, apme_mask;
3601 
3602 	adapter->has_manage = e1000_enable_mng_pass_thru(&adapter->hw);
3603 	apme_mask = EM_EEPROM_APME;
3604 
3605 	switch (adapter->hw.mac.type) {
3606 	case e1000_82542:
3607 	case e1000_82543:
3608 		break;
3609 	case e1000_82544:
3610 		e1000_read_nvm(&adapter->hw,
3611 		    NVM_INIT_CONTROL2_REG, 1, &eeprom_data);
3612 		apme_mask = EM_82544_APME;
3613 		break;
3614 	case e1000_82546:
3615 	case e1000_82546_rev_3:
3616 		if (adapter->hw.bus.func == 1) {
3617 			e1000_read_nvm(&adapter->hw,
3618 			    NVM_INIT_CONTROL3_PORT_B, 1, &eeprom_data);
3619 			break;
3620 		} else
3621 			e1000_read_nvm(&adapter->hw,
3622 			    NVM_INIT_CONTROL3_PORT_A, 1, &eeprom_data);
3623 		break;
3624 	case e1000_82573:
3625 	case e1000_82583:
3626 		adapter->has_amt = TRUE;
3627 		/* FALLTHROUGH */
3628 	case e1000_82571:
3629 	case e1000_82572:
3630 	case e1000_80003es2lan:
3631 		if (adapter->hw.bus.func == 1) {
3632 			e1000_read_nvm(&adapter->hw,
3633 			    NVM_INIT_CONTROL3_PORT_B, 1, &eeprom_data);
3634 			break;
3635 		} else
3636 			e1000_read_nvm(&adapter->hw,
3637 			    NVM_INIT_CONTROL3_PORT_A, 1, &eeprom_data);
3638 		break;
3639 	case e1000_ich8lan:
3640 	case e1000_ich9lan:
3641 	case e1000_ich10lan:
3642 	case e1000_pchlan:
3643 	case e1000_pch2lan:
3644 	case e1000_pch_lpt:
3645 	case e1000_pch_spt:
3646 	case e1000_82575:	/* listing all igb devices */
3647 	case e1000_82576:
3648 	case e1000_82580:
3649 	case e1000_i350:
3650 	case e1000_i354:
3651 	case e1000_i210:
3652 	case e1000_i211:
3653 	case e1000_vfadapt:
3654 	case e1000_vfadapt_i350:
3655 		apme_mask = E1000_WUC_APME;
3656 		adapter->has_amt = TRUE;
3657 		eeprom_data = E1000_READ_REG(&adapter->hw, E1000_WUC);
3658 		break;
3659 	default:
3660 		e1000_read_nvm(&adapter->hw,
3661 		    NVM_INIT_CONTROL3_PORT_A, 1, &eeprom_data);
3662 		break;
3663 	}
3664 	if (eeprom_data & apme_mask)
3665 		adapter->wol = (E1000_WUFC_MAG | E1000_WUFC_MC);
3666 	/*
3667 	 * We have the eeprom settings, now apply the special cases
3668 	 * where the eeprom may be wrong or the board won't support
3669 	 * wake on lan on a particular port
3670 	 */
3671 	device_id = pci_get_device(dev);
3672 	switch (device_id) {
3673 	case E1000_DEV_ID_82546GB_PCIE:
3674 		adapter->wol = 0;
3675 		break;
3676 	case E1000_DEV_ID_82546EB_FIBER:
3677 	case E1000_DEV_ID_82546GB_FIBER:
3678 		/* Wake events only supported on port A for dual fiber
3679 		 * regardless of eeprom setting */
3680 		if (E1000_READ_REG(&adapter->hw, E1000_STATUS) &
3681 		    E1000_STATUS_FUNC_1)
3682 			adapter->wol = 0;
3683 		break;
3684 	case E1000_DEV_ID_82546GB_QUAD_COPPER_KSP3:
3685 		/* if quad port adapter, disable WoL on all but port A */
3686 		if (global_quad_port_a != 0)
3687 			adapter->wol = 0;
3688 		/* Reset for multiple quad port adapters */
3689 		if (++global_quad_port_a == 4)
3690 			global_quad_port_a = 0;
3691 		break;
3692 	case E1000_DEV_ID_82571EB_FIBER:
3693 		/* Wake events only supported on port A for dual fiber
3694 		 * regardless of eeprom setting */
3695 		if (E1000_READ_REG(&adapter->hw, E1000_STATUS) &
3696 		    E1000_STATUS_FUNC_1)
3697 			adapter->wol = 0;
3698 		break;
3699 	case E1000_DEV_ID_82571EB_QUAD_COPPER:
3700 	case E1000_DEV_ID_82571EB_QUAD_FIBER:
3701 	case E1000_DEV_ID_82571EB_QUAD_COPPER_LP:
3702 		/* if quad port adapter, disable WoL on all but port A */
3703 		if (global_quad_port_a != 0)
3704 			adapter->wol = 0;
3705 		/* Reset for multiple quad port adapters */
3706 		if (++global_quad_port_a == 4)
3707 			global_quad_port_a = 0;
3708 		break;
3709 	}
3710 	return;
3711 }
3712 
3713 
3714 /*
3715  * Enable PCI Wake On Lan capability
3716  */
3717 static void
3718 em_enable_wakeup(if_ctx_t ctx)
3719 {
3720 	struct adapter *adapter = iflib_get_softc(ctx);
3721 	device_t dev = iflib_get_dev(ctx);
3722 	if_t ifp = iflib_get_ifp(ctx);
3723 	int error = 0;
3724 	u32 pmc, ctrl, ctrl_ext, rctl;
3725 	u16 status;
3726 
3727 	if (pci_find_cap(dev, PCIY_PMG, &pmc) != 0)
3728 		return;
3729 
3730 	/*
3731 	 * Determine type of Wakeup: note that wol
3732 	 * is set with all bits on by default.
3733 	 */
3734 	if ((if_getcapenable(ifp) & IFCAP_WOL_MAGIC) == 0)
3735 		adapter->wol &= ~E1000_WUFC_MAG;
3736 
3737 	if ((if_getcapenable(ifp) & IFCAP_WOL_UCAST) == 0)
3738 		adapter->wol &= ~E1000_WUFC_EX;
3739 
3740 	if ((if_getcapenable(ifp) & IFCAP_WOL_MCAST) == 0)
3741 		adapter->wol &= ~E1000_WUFC_MC;
3742 	else {
3743 		rctl = E1000_READ_REG(&adapter->hw, E1000_RCTL);
3744 		rctl |= E1000_RCTL_MPE;
3745 		E1000_WRITE_REG(&adapter->hw, E1000_RCTL, rctl);
3746 	}
3747 
3748 	if (!(adapter->wol & (E1000_WUFC_EX | E1000_WUFC_MAG | E1000_WUFC_MC)))
3749 		goto pme;
3750 
3751 	/* Advertise the wakeup capability */
3752 	ctrl = E1000_READ_REG(&adapter->hw, E1000_CTRL);
3753 	ctrl |= (E1000_CTRL_SWDPIN2 | E1000_CTRL_SWDPIN3);
3754 	E1000_WRITE_REG(&adapter->hw, E1000_CTRL, ctrl);
3755 
3756 	/* Keep the laser running on Fiber adapters */
3757 	if (adapter->hw.phy.media_type == e1000_media_type_fiber ||
3758 	    adapter->hw.phy.media_type == e1000_media_type_internal_serdes) {
3759 		ctrl_ext = E1000_READ_REG(&adapter->hw, E1000_CTRL_EXT);
3760 		ctrl_ext |= E1000_CTRL_EXT_SDP3_DATA;
3761 		E1000_WRITE_REG(&adapter->hw, E1000_CTRL_EXT, ctrl_ext);
3762 	}
3763 
3764 	if ((adapter->hw.mac.type == e1000_ich8lan) ||
3765 	    (adapter->hw.mac.type == e1000_pchlan) ||
3766 	    (adapter->hw.mac.type == e1000_ich9lan) ||
3767 	    (adapter->hw.mac.type == e1000_ich10lan))
3768 		e1000_suspend_workarounds_ich8lan(&adapter->hw);
3769 
3770 	if ( adapter->hw.mac.type >= e1000_pchlan) {
3771 		error = em_enable_phy_wakeup(adapter);
3772 		if (error)
3773 			goto pme;
3774 	} else {
3775 		/* Enable wakeup by the MAC */
3776 		E1000_WRITE_REG(&adapter->hw, E1000_WUC, E1000_WUC_PME_EN);
3777 		E1000_WRITE_REG(&adapter->hw, E1000_WUFC, adapter->wol);
3778 	}
3779 
3780 	if (adapter->hw.phy.type == e1000_phy_igp_3)
3781 		e1000_igp3_phy_powerdown_workaround_ich8lan(&adapter->hw);
3782 
3783 pme:
3784 	status = pci_read_config(dev, pmc + PCIR_POWER_STATUS, 2);
3785 	status &= ~(PCIM_PSTAT_PME | PCIM_PSTAT_PMEENABLE);
3786 	if (!error && (if_getcapenable(ifp) & IFCAP_WOL))
3787 		status |= PCIM_PSTAT_PME | PCIM_PSTAT_PMEENABLE;
3788 	pci_write_config(dev, pmc + PCIR_POWER_STATUS, status, 2);
3789 
3790 	return;
3791 }
3792 
3793 /*
3794  * WOL in the newer chipset interfaces (pchlan)
3795  * require thing to be copied into the phy
3796  */
3797 static int
3798 em_enable_phy_wakeup(struct adapter *adapter)
3799 {
3800 	struct e1000_hw *hw = &adapter->hw;
3801 	u32 mreg, ret = 0;
3802 	u16 preg;
3803 
3804 	/* copy MAC RARs to PHY RARs */
3805 	e1000_copy_rx_addrs_to_phy_ich8lan(hw);
3806 
3807 	/* copy MAC MTA to PHY MTA */
3808 	for (int i = 0; i < adapter->hw.mac.mta_reg_count; i++) {
3809 		mreg = E1000_READ_REG_ARRAY(hw, E1000_MTA, i);
3810 		e1000_write_phy_reg(hw, BM_MTA(i), (u16)(mreg & 0xFFFF));
3811 		e1000_write_phy_reg(hw, BM_MTA(i) + 1,
3812 		    (u16)((mreg >> 16) & 0xFFFF));
3813 	}
3814 
3815 	/* configure PHY Rx Control register */
3816 	e1000_read_phy_reg(&adapter->hw, BM_RCTL, &preg);
3817 	mreg = E1000_READ_REG(hw, E1000_RCTL);
3818 	if (mreg & E1000_RCTL_UPE)
3819 		preg |= BM_RCTL_UPE;
3820 	if (mreg & E1000_RCTL_MPE)
3821 		preg |= BM_RCTL_MPE;
3822 	preg &= ~(BM_RCTL_MO_MASK);
3823 	if (mreg & E1000_RCTL_MO_3)
3824 		preg |= (((mreg & E1000_RCTL_MO_3) >> E1000_RCTL_MO_SHIFT)
3825 				<< BM_RCTL_MO_SHIFT);
3826 	if (mreg & E1000_RCTL_BAM)
3827 		preg |= BM_RCTL_BAM;
3828 	if (mreg & E1000_RCTL_PMCF)
3829 		preg |= BM_RCTL_PMCF;
3830 	mreg = E1000_READ_REG(hw, E1000_CTRL);
3831 	if (mreg & E1000_CTRL_RFCE)
3832 		preg |= BM_RCTL_RFCE;
3833 	e1000_write_phy_reg(&adapter->hw, BM_RCTL, preg);
3834 
3835 	/* enable PHY wakeup in MAC register */
3836 	E1000_WRITE_REG(hw, E1000_WUC,
3837 	    E1000_WUC_PHY_WAKE | E1000_WUC_PME_EN | E1000_WUC_APME);
3838 	E1000_WRITE_REG(hw, E1000_WUFC, adapter->wol);
3839 
3840 	/* configure and enable PHY wakeup in PHY registers */
3841 	e1000_write_phy_reg(&adapter->hw, BM_WUFC, adapter->wol);
3842 	e1000_write_phy_reg(&adapter->hw, BM_WUC, E1000_WUC_PME_EN);
3843 
3844 	/* activate PHY wakeup */
3845 	ret = hw->phy.ops.acquire(hw);
3846 	if (ret) {
3847 		printf("Could not acquire PHY\n");
3848 		return ret;
3849 	}
3850 	e1000_write_phy_reg_mdic(hw, IGP01E1000_PHY_PAGE_SELECT,
3851 	                         (BM_WUC_ENABLE_PAGE << IGP_PAGE_SHIFT));
3852 	ret = e1000_read_phy_reg_mdic(hw, BM_WUC_ENABLE_REG, &preg);
3853 	if (ret) {
3854 		printf("Could not read PHY page 769\n");
3855 		goto out;
3856 	}
3857 	preg |= BM_WUC_ENABLE_BIT | BM_WUC_HOST_WU_BIT;
3858 	ret = e1000_write_phy_reg_mdic(hw, BM_WUC_ENABLE_REG, preg);
3859 	if (ret)
3860 		printf("Could not set PHY Host Wakeup bit\n");
3861 out:
3862 	hw->phy.ops.release(hw);
3863 
3864 	return ret;
3865 }
3866 
3867 static void
3868 em_if_led_func(if_ctx_t ctx, int onoff)
3869 {
3870 	struct adapter *adapter = iflib_get_softc(ctx);
3871 
3872 	if (onoff) {
3873 		e1000_setup_led(&adapter->hw);
3874 		e1000_led_on(&adapter->hw);
3875 	} else {
3876 		e1000_led_off(&adapter->hw);
3877 		e1000_cleanup_led(&adapter->hw);
3878 	}
3879 }
3880 
3881 /*
3882  * Disable the L0S and L1 LINK states
3883  */
3884 static void
3885 em_disable_aspm(struct adapter *adapter)
3886 {
3887 	int base, reg;
3888 	u16 link_cap,link_ctrl;
3889 	device_t dev = adapter->dev;
3890 
3891 	switch (adapter->hw.mac.type) {
3892 	case e1000_82573:
3893 	case e1000_82574:
3894 	case e1000_82583:
3895 		break;
3896 	default:
3897 		return;
3898 	}
3899 	if (pci_find_cap(dev, PCIY_EXPRESS, &base) != 0)
3900 		return;
3901 	reg = base + PCIER_LINK_CAP;
3902 	link_cap = pci_read_config(dev, reg, 2);
3903 	if ((link_cap & PCIEM_LINK_CAP_ASPM) == 0)
3904 		return;
3905 	reg = base + PCIER_LINK_CTL;
3906 	link_ctrl = pci_read_config(dev, reg, 2);
3907 	link_ctrl &= ~PCIEM_LINK_CTL_ASPMC;
3908 	pci_write_config(dev, reg, link_ctrl, 2);
3909 	return;
3910 }
3911 
3912 /**********************************************************************
3913  *
3914  *  Update the board statistics counters.
3915  *
3916  **********************************************************************/
3917 static void
3918 em_update_stats_counters(struct adapter *adapter)
3919 {
3920 	u64 prev_xoffrxc = adapter->stats.xoffrxc;
3921 
3922 	if(adapter->hw.phy.media_type == e1000_media_type_copper ||
3923 	   (E1000_READ_REG(&adapter->hw, E1000_STATUS) & E1000_STATUS_LU)) {
3924 		adapter->stats.symerrs += E1000_READ_REG(&adapter->hw, E1000_SYMERRS);
3925 		adapter->stats.sec += E1000_READ_REG(&adapter->hw, E1000_SEC);
3926 	}
3927 	adapter->stats.crcerrs += E1000_READ_REG(&adapter->hw, E1000_CRCERRS);
3928 	adapter->stats.mpc += E1000_READ_REG(&adapter->hw, E1000_MPC);
3929 	adapter->stats.scc += E1000_READ_REG(&adapter->hw, E1000_SCC);
3930 	adapter->stats.ecol += E1000_READ_REG(&adapter->hw, E1000_ECOL);
3931 
3932 	adapter->stats.mcc += E1000_READ_REG(&adapter->hw, E1000_MCC);
3933 	adapter->stats.latecol += E1000_READ_REG(&adapter->hw, E1000_LATECOL);
3934 	adapter->stats.colc += E1000_READ_REG(&adapter->hw, E1000_COLC);
3935 	adapter->stats.dc += E1000_READ_REG(&adapter->hw, E1000_DC);
3936 	adapter->stats.rlec += E1000_READ_REG(&adapter->hw, E1000_RLEC);
3937 	adapter->stats.xonrxc += E1000_READ_REG(&adapter->hw, E1000_XONRXC);
3938 	adapter->stats.xontxc += E1000_READ_REG(&adapter->hw, E1000_XONTXC);
3939 	adapter->stats.xoffrxc += E1000_READ_REG(&adapter->hw, E1000_XOFFRXC);
3940 	/*
3941 	 ** For watchdog management we need to know if we have been
3942 	 ** paused during the last interval, so capture that here.
3943 	*/
3944 	if (adapter->stats.xoffrxc != prev_xoffrxc)
3945 		adapter->shared->isc_pause_frames = 1;
3946 	adapter->stats.xofftxc += E1000_READ_REG(&adapter->hw, E1000_XOFFTXC);
3947 	adapter->stats.fcruc += E1000_READ_REG(&adapter->hw, E1000_FCRUC);
3948 	adapter->stats.prc64 += E1000_READ_REG(&adapter->hw, E1000_PRC64);
3949 	adapter->stats.prc127 += E1000_READ_REG(&adapter->hw, E1000_PRC127);
3950 	adapter->stats.prc255 += E1000_READ_REG(&adapter->hw, E1000_PRC255);
3951 	adapter->stats.prc511 += E1000_READ_REG(&adapter->hw, E1000_PRC511);
3952 	adapter->stats.prc1023 += E1000_READ_REG(&adapter->hw, E1000_PRC1023);
3953 	adapter->stats.prc1522 += E1000_READ_REG(&adapter->hw, E1000_PRC1522);
3954 	adapter->stats.gprc += E1000_READ_REG(&adapter->hw, E1000_GPRC);
3955 	adapter->stats.bprc += E1000_READ_REG(&adapter->hw, E1000_BPRC);
3956 	adapter->stats.mprc += E1000_READ_REG(&adapter->hw, E1000_MPRC);
3957 	adapter->stats.gptc += E1000_READ_REG(&adapter->hw, E1000_GPTC);
3958 
3959 	/* For the 64-bit byte counters the low dword must be read first. */
3960 	/* Both registers clear on the read of the high dword */
3961 
3962 	adapter->stats.gorc += E1000_READ_REG(&adapter->hw, E1000_GORCL) +
3963 	    ((u64)E1000_READ_REG(&adapter->hw, E1000_GORCH) << 32);
3964 	adapter->stats.gotc += E1000_READ_REG(&adapter->hw, E1000_GOTCL) +
3965 	    ((u64)E1000_READ_REG(&adapter->hw, E1000_GOTCH) << 32);
3966 
3967 	adapter->stats.rnbc += E1000_READ_REG(&adapter->hw, E1000_RNBC);
3968 	adapter->stats.ruc += E1000_READ_REG(&adapter->hw, E1000_RUC);
3969 	adapter->stats.rfc += E1000_READ_REG(&adapter->hw, E1000_RFC);
3970 	adapter->stats.roc += E1000_READ_REG(&adapter->hw, E1000_ROC);
3971 	adapter->stats.rjc += E1000_READ_REG(&adapter->hw, E1000_RJC);
3972 
3973 	adapter->stats.tor += E1000_READ_REG(&adapter->hw, E1000_TORH);
3974 	adapter->stats.tot += E1000_READ_REG(&adapter->hw, E1000_TOTH);
3975 
3976 	adapter->stats.tpr += E1000_READ_REG(&adapter->hw, E1000_TPR);
3977 	adapter->stats.tpt += E1000_READ_REG(&adapter->hw, E1000_TPT);
3978 	adapter->stats.ptc64 += E1000_READ_REG(&adapter->hw, E1000_PTC64);
3979 	adapter->stats.ptc127 += E1000_READ_REG(&adapter->hw, E1000_PTC127);
3980 	adapter->stats.ptc255 += E1000_READ_REG(&adapter->hw, E1000_PTC255);
3981 	adapter->stats.ptc511 += E1000_READ_REG(&adapter->hw, E1000_PTC511);
3982 	adapter->stats.ptc1023 += E1000_READ_REG(&adapter->hw, E1000_PTC1023);
3983 	adapter->stats.ptc1522 += E1000_READ_REG(&adapter->hw, E1000_PTC1522);
3984 	adapter->stats.mptc += E1000_READ_REG(&adapter->hw, E1000_MPTC);
3985 	adapter->stats.bptc += E1000_READ_REG(&adapter->hw, E1000_BPTC);
3986 
3987 	/* Interrupt Counts */
3988 
3989 	adapter->stats.iac += E1000_READ_REG(&adapter->hw, E1000_IAC);
3990 	adapter->stats.icrxptc += E1000_READ_REG(&adapter->hw, E1000_ICRXPTC);
3991 	adapter->stats.icrxatc += E1000_READ_REG(&adapter->hw, E1000_ICRXATC);
3992 	adapter->stats.ictxptc += E1000_READ_REG(&adapter->hw, E1000_ICTXPTC);
3993 	adapter->stats.ictxatc += E1000_READ_REG(&adapter->hw, E1000_ICTXATC);
3994 	adapter->stats.ictxqec += E1000_READ_REG(&adapter->hw, E1000_ICTXQEC);
3995 	adapter->stats.ictxqmtc += E1000_READ_REG(&adapter->hw, E1000_ICTXQMTC);
3996 	adapter->stats.icrxdmtc += E1000_READ_REG(&adapter->hw, E1000_ICRXDMTC);
3997 	adapter->stats.icrxoc += E1000_READ_REG(&adapter->hw, E1000_ICRXOC);
3998 
3999 	if (adapter->hw.mac.type >= e1000_82543) {
4000 		adapter->stats.algnerrc +=
4001 		E1000_READ_REG(&adapter->hw, E1000_ALGNERRC);
4002 		adapter->stats.rxerrc +=
4003 		E1000_READ_REG(&adapter->hw, E1000_RXERRC);
4004 		adapter->stats.tncrs +=
4005 		E1000_READ_REG(&adapter->hw, E1000_TNCRS);
4006 		adapter->stats.cexterr +=
4007 		E1000_READ_REG(&adapter->hw, E1000_CEXTERR);
4008 		adapter->stats.tsctc +=
4009 		E1000_READ_REG(&adapter->hw, E1000_TSCTC);
4010 		adapter->stats.tsctfc +=
4011 		E1000_READ_REG(&adapter->hw, E1000_TSCTFC);
4012 	}
4013 }
4014 
4015 static uint64_t
4016 em_if_get_counter(if_ctx_t ctx, ift_counter cnt)
4017 {
4018 	struct adapter *adapter = iflib_get_softc(ctx);
4019 	struct ifnet *ifp = iflib_get_ifp(ctx);
4020 
4021 	switch (cnt) {
4022 	case IFCOUNTER_COLLISIONS:
4023 		return (adapter->stats.colc);
4024 	case IFCOUNTER_IERRORS:
4025 		return (adapter->dropped_pkts + adapter->stats.rxerrc +
4026 		    adapter->stats.crcerrs + adapter->stats.algnerrc +
4027 		    adapter->stats.ruc + adapter->stats.roc +
4028 		    adapter->stats.mpc + adapter->stats.cexterr);
4029 	case IFCOUNTER_OERRORS:
4030 		return (adapter->stats.ecol + adapter->stats.latecol +
4031 		    adapter->watchdog_events);
4032 	default:
4033 		return (if_get_counter_default(ifp, cnt));
4034 	}
4035 }
4036 
4037 /* Export a single 32-bit register via a read-only sysctl. */
4038 static int
4039 em_sysctl_reg_handler(SYSCTL_HANDLER_ARGS)
4040 {
4041 	struct adapter *adapter;
4042 	u_int val;
4043 
4044 	adapter = oidp->oid_arg1;
4045 	val = E1000_READ_REG(&adapter->hw, oidp->oid_arg2);
4046 	return (sysctl_handle_int(oidp, &val, 0, req));
4047 }
4048 
4049 /*
4050  * Add sysctl variables, one per statistic, to the system.
4051  */
4052 static void
4053 em_add_hw_stats(struct adapter *adapter)
4054 {
4055 	device_t dev = iflib_get_dev(adapter->ctx);
4056 	struct em_tx_queue *tx_que = adapter->tx_queues;
4057 	struct em_rx_queue *rx_que = adapter->rx_queues;
4058 
4059 	struct sysctl_ctx_list *ctx = device_get_sysctl_ctx(dev);
4060 	struct sysctl_oid *tree = device_get_sysctl_tree(dev);
4061 	struct sysctl_oid_list *child = SYSCTL_CHILDREN(tree);
4062 	struct e1000_hw_stats *stats = &adapter->stats;
4063 
4064 	struct sysctl_oid *stat_node, *queue_node, *int_node;
4065 	struct sysctl_oid_list *stat_list, *queue_list, *int_list;
4066 
4067 #define QUEUE_NAME_LEN 32
4068 	char namebuf[QUEUE_NAME_LEN];
4069 
4070 	/* Driver Statistics */
4071 	SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "dropped",
4072 			CTLFLAG_RD, &adapter->dropped_pkts,
4073 			"Driver dropped packets");
4074 	SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "link_irq",
4075 			CTLFLAG_RD, &adapter->link_irq,
4076 			"Link MSI-X IRQ Handled");
4077 	SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "rx_overruns",
4078 			CTLFLAG_RD, &adapter->rx_overruns,
4079 			"RX overruns");
4080 	SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "watchdog_timeouts",
4081 			CTLFLAG_RD, &adapter->watchdog_events,
4082 			"Watchdog timeouts");
4083 	SYSCTL_ADD_PROC(ctx, child, OID_AUTO, "device_control",
4084 			CTLTYPE_UINT | CTLFLAG_RD, adapter, E1000_CTRL,
4085 			em_sysctl_reg_handler, "IU",
4086 			"Device Control Register");
4087 	SYSCTL_ADD_PROC(ctx, child, OID_AUTO, "rx_control",
4088 			CTLTYPE_UINT | CTLFLAG_RD, adapter, E1000_RCTL,
4089 			em_sysctl_reg_handler, "IU",
4090 			"Receiver Control Register");
4091 	SYSCTL_ADD_UINT(ctx, child, OID_AUTO, "fc_high_water",
4092 			CTLFLAG_RD, &adapter->hw.fc.high_water, 0,
4093 			"Flow Control High Watermark");
4094 	SYSCTL_ADD_UINT(ctx, child, OID_AUTO, "fc_low_water",
4095 			CTLFLAG_RD, &adapter->hw.fc.low_water, 0,
4096 			"Flow Control Low Watermark");
4097 
4098 	for (int i = 0; i < adapter->tx_num_queues; i++, tx_que++) {
4099 		struct tx_ring *txr = &tx_que->txr;
4100 		snprintf(namebuf, QUEUE_NAME_LEN, "queue_tx_%d", i);
4101 		queue_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, namebuf,
4102 					    CTLFLAG_RD, NULL, "TX Queue Name");
4103 		queue_list = SYSCTL_CHILDREN(queue_node);
4104 
4105 		SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "txd_head",
4106 				CTLTYPE_UINT | CTLFLAG_RD, adapter,
4107 				E1000_TDH(txr->me),
4108 				em_sysctl_reg_handler, "IU",
4109 				"Transmit Descriptor Head");
4110 		SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "txd_tail",
4111 				CTLTYPE_UINT | CTLFLAG_RD, adapter,
4112 				E1000_TDT(txr->me),
4113 				em_sysctl_reg_handler, "IU",
4114 				"Transmit Descriptor Tail");
4115 		SYSCTL_ADD_ULONG(ctx, queue_list, OID_AUTO, "tx_irq",
4116 				CTLFLAG_RD, &txr->tx_irq,
4117 				"Queue MSI-X Transmit Interrupts");
4118 	}
4119 
4120 	for (int j = 0; j < adapter->rx_num_queues; j++, rx_que++) {
4121 		struct rx_ring *rxr = &rx_que->rxr;
4122 		snprintf(namebuf, QUEUE_NAME_LEN, "queue_rx_%d", j);
4123 		queue_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, namebuf,
4124 					    CTLFLAG_RD, NULL, "RX Queue Name");
4125 		queue_list = SYSCTL_CHILDREN(queue_node);
4126 
4127 		SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "rxd_head",
4128 				CTLTYPE_UINT | CTLFLAG_RD, adapter,
4129 				E1000_RDH(rxr->me),
4130 				em_sysctl_reg_handler, "IU",
4131 				"Receive Descriptor Head");
4132 		SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "rxd_tail",
4133 				CTLTYPE_UINT | CTLFLAG_RD, adapter,
4134 				E1000_RDT(rxr->me),
4135 				em_sysctl_reg_handler, "IU",
4136 				"Receive Descriptor Tail");
4137 		SYSCTL_ADD_ULONG(ctx, queue_list, OID_AUTO, "rx_irq",
4138 				CTLFLAG_RD, &rxr->rx_irq,
4139 				"Queue MSI-X Receive Interrupts");
4140 	}
4141 
4142 	/* MAC stats get their own sub node */
4143 
4144 	stat_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, "mac_stats",
4145 				    CTLFLAG_RD, NULL, "Statistics");
4146 	stat_list = SYSCTL_CHILDREN(stat_node);
4147 
4148 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "excess_coll",
4149 			CTLFLAG_RD, &stats->ecol,
4150 			"Excessive collisions");
4151 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "single_coll",
4152 			CTLFLAG_RD, &stats->scc,
4153 			"Single collisions");
4154 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "multiple_coll",
4155 			CTLFLAG_RD, &stats->mcc,
4156 			"Multiple collisions");
4157 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "late_coll",
4158 			CTLFLAG_RD, &stats->latecol,
4159 			"Late collisions");
4160 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "collision_count",
4161 			CTLFLAG_RD, &stats->colc,
4162 			"Collision Count");
4163 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "symbol_errors",
4164 			CTLFLAG_RD, &adapter->stats.symerrs,
4165 			"Symbol Errors");
4166 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "sequence_errors",
4167 			CTLFLAG_RD, &adapter->stats.sec,
4168 			"Sequence Errors");
4169 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "defer_count",
4170 			CTLFLAG_RD, &adapter->stats.dc,
4171 			"Defer Count");
4172 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "missed_packets",
4173 			CTLFLAG_RD, &adapter->stats.mpc,
4174 			"Missed Packets");
4175 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_no_buff",
4176 			CTLFLAG_RD, &adapter->stats.rnbc,
4177 			"Receive No Buffers");
4178 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_undersize",
4179 			CTLFLAG_RD, &adapter->stats.ruc,
4180 			"Receive Undersize");
4181 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_fragmented",
4182 			CTLFLAG_RD, &adapter->stats.rfc,
4183 			"Fragmented Packets Received ");
4184 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_oversize",
4185 			CTLFLAG_RD, &adapter->stats.roc,
4186 			"Oversized Packets Received");
4187 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_jabber",
4188 			CTLFLAG_RD, &adapter->stats.rjc,
4189 			"Recevied Jabber");
4190 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_errs",
4191 			CTLFLAG_RD, &adapter->stats.rxerrc,
4192 			"Receive Errors");
4193 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "crc_errs",
4194 			CTLFLAG_RD, &adapter->stats.crcerrs,
4195 			"CRC errors");
4196 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "alignment_errs",
4197 			CTLFLAG_RD, &adapter->stats.algnerrc,
4198 			"Alignment Errors");
4199 	/* On 82575 these are collision counts */
4200 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "coll_ext_errs",
4201 			CTLFLAG_RD, &adapter->stats.cexterr,
4202 			"Collision/Carrier extension errors");
4203 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xon_recvd",
4204 			CTLFLAG_RD, &adapter->stats.xonrxc,
4205 			"XON Received");
4206 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xon_txd",
4207 			CTLFLAG_RD, &adapter->stats.xontxc,
4208 			"XON Transmitted");
4209 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xoff_recvd",
4210 			CTLFLAG_RD, &adapter->stats.xoffrxc,
4211 			"XOFF Received");
4212 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xoff_txd",
4213 			CTLFLAG_RD, &adapter->stats.xofftxc,
4214 			"XOFF Transmitted");
4215 
4216 	/* Packet Reception Stats */
4217 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "total_pkts_recvd",
4218 			CTLFLAG_RD, &adapter->stats.tpr,
4219 			"Total Packets Received ");
4220 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_pkts_recvd",
4221 			CTLFLAG_RD, &adapter->stats.gprc,
4222 			"Good Packets Received");
4223 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "bcast_pkts_recvd",
4224 			CTLFLAG_RD, &adapter->stats.bprc,
4225 			"Broadcast Packets Received");
4226 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "mcast_pkts_recvd",
4227 			CTLFLAG_RD, &adapter->stats.mprc,
4228 			"Multicast Packets Received");
4229 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_64",
4230 			CTLFLAG_RD, &adapter->stats.prc64,
4231 			"64 byte frames received ");
4232 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_65_127",
4233 			CTLFLAG_RD, &adapter->stats.prc127,
4234 			"65-127 byte frames received");
4235 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_128_255",
4236 			CTLFLAG_RD, &adapter->stats.prc255,
4237 			"128-255 byte frames received");
4238 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_256_511",
4239 			CTLFLAG_RD, &adapter->stats.prc511,
4240 			"256-511 byte frames received");
4241 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_512_1023",
4242 			CTLFLAG_RD, &adapter->stats.prc1023,
4243 			"512-1023 byte frames received");
4244 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_1024_1522",
4245 			CTLFLAG_RD, &adapter->stats.prc1522,
4246 			"1023-1522 byte frames received");
4247 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_octets_recvd",
4248 			CTLFLAG_RD, &adapter->stats.gorc,
4249 			"Good Octets Received");
4250 
4251 	/* Packet Transmission Stats */
4252 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_octets_txd",
4253 			CTLFLAG_RD, &adapter->stats.gotc,
4254 			"Good Octets Transmitted");
4255 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "total_pkts_txd",
4256 			CTLFLAG_RD, &adapter->stats.tpt,
4257 			"Total Packets Transmitted");
4258 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_pkts_txd",
4259 			CTLFLAG_RD, &adapter->stats.gptc,
4260 			"Good Packets Transmitted");
4261 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "bcast_pkts_txd",
4262 			CTLFLAG_RD, &adapter->stats.bptc,
4263 			"Broadcast Packets Transmitted");
4264 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "mcast_pkts_txd",
4265 			CTLFLAG_RD, &adapter->stats.mptc,
4266 			"Multicast Packets Transmitted");
4267 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_64",
4268 			CTLFLAG_RD, &adapter->stats.ptc64,
4269 			"64 byte frames transmitted ");
4270 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_65_127",
4271 			CTLFLAG_RD, &adapter->stats.ptc127,
4272 			"65-127 byte frames transmitted");
4273 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_128_255",
4274 			CTLFLAG_RD, &adapter->stats.ptc255,
4275 			"128-255 byte frames transmitted");
4276 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_256_511",
4277 			CTLFLAG_RD, &adapter->stats.ptc511,
4278 			"256-511 byte frames transmitted");
4279 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_512_1023",
4280 			CTLFLAG_RD, &adapter->stats.ptc1023,
4281 			"512-1023 byte frames transmitted");
4282 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_1024_1522",
4283 			CTLFLAG_RD, &adapter->stats.ptc1522,
4284 			"1024-1522 byte frames transmitted");
4285 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tso_txd",
4286 			CTLFLAG_RD, &adapter->stats.tsctc,
4287 			"TSO Contexts Transmitted");
4288 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tso_ctx_fail",
4289 			CTLFLAG_RD, &adapter->stats.tsctfc,
4290 			"TSO Contexts Failed");
4291 
4292 
4293 	/* Interrupt Stats */
4294 
4295 	int_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, "interrupts",
4296 				    CTLFLAG_RD, NULL, "Interrupt Statistics");
4297 	int_list = SYSCTL_CHILDREN(int_node);
4298 
4299 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "asserts",
4300 			CTLFLAG_RD, &adapter->stats.iac,
4301 			"Interrupt Assertion Count");
4302 
4303 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "rx_pkt_timer",
4304 			CTLFLAG_RD, &adapter->stats.icrxptc,
4305 			"Interrupt Cause Rx Pkt Timer Expire Count");
4306 
4307 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "rx_abs_timer",
4308 			CTLFLAG_RD, &adapter->stats.icrxatc,
4309 			"Interrupt Cause Rx Abs Timer Expire Count");
4310 
4311 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "tx_pkt_timer",
4312 			CTLFLAG_RD, &adapter->stats.ictxptc,
4313 			"Interrupt Cause Tx Pkt Timer Expire Count");
4314 
4315 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "tx_abs_timer",
4316 			CTLFLAG_RD, &adapter->stats.ictxatc,
4317 			"Interrupt Cause Tx Abs Timer Expire Count");
4318 
4319 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "tx_queue_empty",
4320 			CTLFLAG_RD, &adapter->stats.ictxqec,
4321 			"Interrupt Cause Tx Queue Empty Count");
4322 
4323 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "tx_queue_min_thresh",
4324 			CTLFLAG_RD, &adapter->stats.ictxqmtc,
4325 			"Interrupt Cause Tx Queue Min Thresh Count");
4326 
4327 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "rx_desc_min_thresh",
4328 			CTLFLAG_RD, &adapter->stats.icrxdmtc,
4329 			"Interrupt Cause Rx Desc Min Thresh Count");
4330 
4331 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "rx_overrun",
4332 			CTLFLAG_RD, &adapter->stats.icrxoc,
4333 			"Interrupt Cause Receiver Overrun Count");
4334 }
4335 
4336 /**********************************************************************
4337  *
4338  *  This routine provides a way to dump out the adapter eeprom,
4339  *  often a useful debug/service tool. This only dumps the first
4340  *  32 words, stuff that matters is in that extent.
4341  *
4342  **********************************************************************/
4343 static int
4344 em_sysctl_nvm_info(SYSCTL_HANDLER_ARGS)
4345 {
4346 	struct adapter *adapter = (struct adapter *)arg1;
4347 	int error;
4348 	int result;
4349 
4350 	result = -1;
4351 	error = sysctl_handle_int(oidp, &result, 0, req);
4352 
4353 	if (error || !req->newptr)
4354 		return (error);
4355 
4356 	/*
4357 	 * This value will cause a hex dump of the
4358 	 * first 32 16-bit words of the EEPROM to
4359 	 * the screen.
4360 	 */
4361 	if (result == 1)
4362 		em_print_nvm_info(adapter);
4363 
4364 	return (error);
4365 }
4366 
4367 static void
4368 em_print_nvm_info(struct adapter *adapter)
4369 {
4370 	u16 eeprom_data;
4371 	int i, j, row = 0;
4372 
4373 	/* Its a bit crude, but it gets the job done */
4374 	printf("\nInterface EEPROM Dump:\n");
4375 	printf("Offset\n0x0000  ");
4376 	for (i = 0, j = 0; i < 32; i++, j++) {
4377 		if (j == 8) { /* Make the offset block */
4378 			j = 0; ++row;
4379 			printf("\n0x00%x0  ",row);
4380 		}
4381 		e1000_read_nvm(&adapter->hw, i, 1, &eeprom_data);
4382 		printf("%04x ", eeprom_data);
4383 	}
4384 	printf("\n");
4385 }
4386 
4387 static int
4388 em_sysctl_int_delay(SYSCTL_HANDLER_ARGS)
4389 {
4390 	struct em_int_delay_info *info;
4391 	struct adapter *adapter;
4392 	u32 regval;
4393 	int error, usecs, ticks;
4394 
4395 	info = (struct em_int_delay_info *) arg1;
4396 	usecs = info->value;
4397 	error = sysctl_handle_int(oidp, &usecs, 0, req);
4398 	if (error != 0 || req->newptr == NULL)
4399 		return (error);
4400 	if (usecs < 0 || usecs > EM_TICKS_TO_USECS(65535))
4401 		return (EINVAL);
4402 	info->value = usecs;
4403 	ticks = EM_USECS_TO_TICKS(usecs);
4404 	if (info->offset == E1000_ITR)	/* units are 256ns here */
4405 		ticks *= 4;
4406 
4407 	adapter = info->adapter;
4408 
4409 	regval = E1000_READ_OFFSET(&adapter->hw, info->offset);
4410 	regval = (regval & ~0xffff) | (ticks & 0xffff);
4411 	/* Handle a few special cases. */
4412 	switch (info->offset) {
4413 	case E1000_RDTR:
4414 		break;
4415 	case E1000_TIDV:
4416 		if (ticks == 0) {
4417 			adapter->txd_cmd &= ~E1000_TXD_CMD_IDE;
4418 			/* Don't write 0 into the TIDV register. */
4419 			regval++;
4420 		} else
4421 			adapter->txd_cmd |= E1000_TXD_CMD_IDE;
4422 		break;
4423 	}
4424 	E1000_WRITE_OFFSET(&adapter->hw, info->offset, regval);
4425 	return (0);
4426 }
4427 
4428 static void
4429 em_add_int_delay_sysctl(struct adapter *adapter, const char *name,
4430 	const char *description, struct em_int_delay_info *info,
4431 	int offset, int value)
4432 {
4433 	info->adapter = adapter;
4434 	info->offset = offset;
4435 	info->value = value;
4436 	SYSCTL_ADD_PROC(device_get_sysctl_ctx(adapter->dev),
4437 	    SYSCTL_CHILDREN(device_get_sysctl_tree(adapter->dev)),
4438 	    OID_AUTO, name, CTLTYPE_INT|CTLFLAG_RW,
4439 	    info, 0, em_sysctl_int_delay, "I", description);
4440 }
4441 
4442 /*
4443  * Set flow control using sysctl:
4444  * Flow control values:
4445  *      0 - off
4446  *      1 - rx pause
4447  *      2 - tx pause
4448  *      3 - full
4449  */
4450 static int
4451 em_set_flowcntl(SYSCTL_HANDLER_ARGS)
4452 {
4453 	int error;
4454 	static int input = 3; /* default is full */
4455 	struct adapter	*adapter = (struct adapter *) arg1;
4456 
4457 	error = sysctl_handle_int(oidp, &input, 0, req);
4458 
4459 	if ((error) || (req->newptr == NULL))
4460 		return (error);
4461 
4462 	if (input == adapter->fc) /* no change? */
4463 		return (error);
4464 
4465 	switch (input) {
4466 	case e1000_fc_rx_pause:
4467 	case e1000_fc_tx_pause:
4468 	case e1000_fc_full:
4469 	case e1000_fc_none:
4470 		adapter->hw.fc.requested_mode = input;
4471 		adapter->fc = input;
4472 		break;
4473 	default:
4474 		/* Do nothing */
4475 		return (error);
4476 	}
4477 
4478 	adapter->hw.fc.current_mode = adapter->hw.fc.requested_mode;
4479 	e1000_force_mac_fc(&adapter->hw);
4480 	return (error);
4481 }
4482 
4483 /*
4484  * Manage Energy Efficient Ethernet:
4485  * Control values:
4486  *     0/1 - enabled/disabled
4487  */
4488 static int
4489 em_sysctl_eee(SYSCTL_HANDLER_ARGS)
4490 {
4491 	struct adapter *adapter = (struct adapter *) arg1;
4492 	int error, value;
4493 
4494 	value = adapter->hw.dev_spec.ich8lan.eee_disable;
4495 	error = sysctl_handle_int(oidp, &value, 0, req);
4496 	if (error || req->newptr == NULL)
4497 		return (error);
4498 	adapter->hw.dev_spec.ich8lan.eee_disable = (value != 0);
4499 	em_if_init(adapter->ctx);
4500 
4501 	return (0);
4502 }
4503 
4504 static int
4505 em_sysctl_debug_info(SYSCTL_HANDLER_ARGS)
4506 {
4507 	struct adapter *adapter;
4508 	int error;
4509 	int result;
4510 
4511 	result = -1;
4512 	error = sysctl_handle_int(oidp, &result, 0, req);
4513 
4514 	if (error || !req->newptr)
4515 		return (error);
4516 
4517 	if (result == 1) {
4518 		adapter = (struct adapter *) arg1;
4519 		em_print_debug_info(adapter);
4520 	}
4521 
4522 	return (error);
4523 }
4524 
4525 static int
4526 em_get_rs(SYSCTL_HANDLER_ARGS)
4527 {
4528 	struct adapter *adapter = (struct adapter *) arg1;
4529 	int error;
4530 	int result;
4531 
4532 	result = 0;
4533 	error = sysctl_handle_int(oidp, &result, 0, req);
4534 
4535 	if (error || !req->newptr || result != 1)
4536 		return (error);
4537 	em_dump_rs(adapter);
4538 
4539 	return (error);
4540 }
4541 
4542 static void
4543 em_if_debug(if_ctx_t ctx)
4544 {
4545 	em_dump_rs(iflib_get_softc(ctx));
4546 }
4547 
4548 /*
4549  * This routine is meant to be fluid, add whatever is
4550  * needed for debugging a problem.  -jfv
4551  */
4552 static void
4553 em_print_debug_info(struct adapter *adapter)
4554 {
4555 	device_t dev = iflib_get_dev(adapter->ctx);
4556 	struct ifnet *ifp = iflib_get_ifp(adapter->ctx);
4557 	struct tx_ring *txr = &adapter->tx_queues->txr;
4558 	struct rx_ring *rxr = &adapter->rx_queues->rxr;
4559 
4560 	if (if_getdrvflags(ifp) & IFF_DRV_RUNNING)
4561 		printf("Interface is RUNNING ");
4562 	else
4563 		printf("Interface is NOT RUNNING\n");
4564 
4565 	if (if_getdrvflags(ifp) & IFF_DRV_OACTIVE)
4566 		printf("and INACTIVE\n");
4567 	else
4568 		printf("and ACTIVE\n");
4569 
4570 	for (int i = 0; i < adapter->tx_num_queues; i++, txr++) {
4571 		device_printf(dev, "TX Queue %d ------\n", i);
4572 		device_printf(dev, "hw tdh = %d, hw tdt = %d\n",
4573 			E1000_READ_REG(&adapter->hw, E1000_TDH(i)),
4574 			E1000_READ_REG(&adapter->hw, E1000_TDT(i)));
4575 
4576 	}
4577 	for (int j=0; j < adapter->rx_num_queues; j++, rxr++) {
4578 		device_printf(dev, "RX Queue %d ------\n", j);
4579 		device_printf(dev, "hw rdh = %d, hw rdt = %d\n",
4580 			E1000_READ_REG(&adapter->hw, E1000_RDH(j)),
4581 			E1000_READ_REG(&adapter->hw, E1000_RDT(j)));
4582 	}
4583 }
4584 
4585 /*
4586  * 82574 only:
4587  * Write a new value to the EEPROM increasing the number of MSI-X
4588  * vectors from 3 to 5, for proper multiqueue support.
4589  */
4590 static void
4591 em_enable_vectors_82574(if_ctx_t ctx)
4592 {
4593 	struct adapter *adapter = iflib_get_softc(ctx);
4594 	struct e1000_hw *hw = &adapter->hw;
4595 	device_t dev = iflib_get_dev(ctx);
4596 	u16 edata;
4597 
4598 	e1000_read_nvm(hw, EM_NVM_PCIE_CTRL, 1, &edata);
4599 	if (bootverbose)
4600 		device_printf(dev, "EM_NVM_PCIE_CTRL = %#06x\n", edata);
4601 	if (((edata & EM_NVM_MSIX_N_MASK) >> EM_NVM_MSIX_N_SHIFT) != 4) {
4602 		device_printf(dev, "Writing to eeprom: increasing "
4603 		    "reported MSI-X vectors from 3 to 5...\n");
4604 		edata &= ~(EM_NVM_MSIX_N_MASK);
4605 		edata |= 4 << EM_NVM_MSIX_N_SHIFT;
4606 		e1000_write_nvm(hw, EM_NVM_PCIE_CTRL, 1, &edata);
4607 		e1000_update_nvm_checksum(hw);
4608 		device_printf(dev, "Writing to eeprom: done\n");
4609 	}
4610 }
4611