xref: /freebsd/sys/dev/e1000/if_em.c (revision dbfb4063ae95b956a2b0021c37c9a8be4c2e4393)
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 	/* required last entry */
178 	PVID_END
179 };
180 
181 static pci_vendor_info_t igb_vendor_info_array[] =
182 {
183 	/* Intel(R) PRO/1000 Network Connection - igb */
184 	PVID(0x8086, E1000_DEV_ID_82575EB_COPPER, "Intel(R) PRO/1000 PCI-Express Network Driver"),
185 	PVID(0x8086, E1000_DEV_ID_82575EB_FIBER_SERDES, "Intel(R) PRO/1000 PCI-Express Network Driver"),
186 	PVID(0x8086, E1000_DEV_ID_82575GB_QUAD_COPPER, "Intel(R) PRO/1000 PCI-Express Network Driver"),
187 	PVID(0x8086, E1000_DEV_ID_82576, "Intel(R) PRO/1000 PCI-Express Network Driver"),
188 	PVID(0x8086, E1000_DEV_ID_82576_NS, "Intel(R) PRO/1000 PCI-Express Network Driver"),
189 	PVID(0x8086, E1000_DEV_ID_82576_NS_SERDES, "Intel(R) PRO/1000 PCI-Express Network Driver"),
190 	PVID(0x8086, E1000_DEV_ID_82576_FIBER, "Intel(R) PRO/1000 PCI-Express Network Driver"),
191 	PVID(0x8086, E1000_DEV_ID_82576_SERDES, "Intel(R) PRO/1000 PCI-Express Network Driver"),
192 	PVID(0x8086, E1000_DEV_ID_82576_SERDES_QUAD, "Intel(R) PRO/1000 PCI-Express Network Driver"),
193 	PVID(0x8086, E1000_DEV_ID_82576_QUAD_COPPER, "Intel(R) PRO/1000 PCI-Express Network Driver"),
194 	PVID(0x8086, E1000_DEV_ID_82576_QUAD_COPPER_ET2, "Intel(R) PRO/1000 PCI-Express Network Driver"),
195 	PVID(0x8086, E1000_DEV_ID_82576_VF, "Intel(R) PRO/1000 PCI-Express Network Driver"),
196 	PVID(0x8086, E1000_DEV_ID_82580_COPPER, "Intel(R) PRO/1000 PCI-Express Network Driver"),
197 	PVID(0x8086, E1000_DEV_ID_82580_FIBER, "Intel(R) PRO/1000 PCI-Express Network Driver"),
198 	PVID(0x8086, E1000_DEV_ID_82580_SERDES, "Intel(R) PRO/1000 PCI-Express Network Driver"),
199 	PVID(0x8086, E1000_DEV_ID_82580_SGMII, "Intel(R) PRO/1000 PCI-Express Network Driver"),
200 	PVID(0x8086, E1000_DEV_ID_82580_COPPER_DUAL, "Intel(R) PRO/1000 PCI-Express Network Driver"),
201 	PVID(0x8086, E1000_DEV_ID_82580_QUAD_FIBER, "Intel(R) PRO/1000 PCI-Express Network Driver"),
202 	PVID(0x8086, E1000_DEV_ID_DH89XXCC_SERDES, "Intel(R) PRO/1000 PCI-Express Network Driver"),
203 	PVID(0x8086, E1000_DEV_ID_DH89XXCC_SGMII, "Intel(R) PRO/1000 PCI-Express Network Driver"),
204 	PVID(0x8086, E1000_DEV_ID_DH89XXCC_SFP, "Intel(R) PRO/1000 PCI-Express Network Driver"),
205 	PVID(0x8086, E1000_DEV_ID_DH89XXCC_BACKPLANE, "Intel(R) PRO/1000 PCI-Express Network Driver"),
206 	PVID(0x8086, E1000_DEV_ID_I350_COPPER, "Intel(R) PRO/1000 PCI-Express Network Driver"),
207 	PVID(0x8086, E1000_DEV_ID_I350_FIBER, "Intel(R) PRO/1000 PCI-Express Network Driver"),
208 	PVID(0x8086, E1000_DEV_ID_I350_SERDES, "Intel(R) PRO/1000 PCI-Express Network Driver"),
209 	PVID(0x8086, E1000_DEV_ID_I350_SGMII, "Intel(R) PRO/1000 PCI-Express Network Driver"),
210 	PVID(0x8086, E1000_DEV_ID_I350_VF, "Intel(R) PRO/1000 PCI-Express Network Driver"),
211 	PVID(0x8086, E1000_DEV_ID_I210_COPPER, "Intel(R) PRO/1000 PCI-Express Network Driver"),
212 	PVID(0x8086, E1000_DEV_ID_I210_COPPER_IT, "Intel(R) PRO/1000 PCI-Express Network Driver"),
213 	PVID(0x8086, E1000_DEV_ID_I210_COPPER_OEM1, "Intel(R) PRO/1000 PCI-Express Network Driver"),
214 	PVID(0x8086, E1000_DEV_ID_I210_COPPER_FLASHLESS, "Intel(R) PRO/1000 PCI-Express Network Driver"),
215 	PVID(0x8086, E1000_DEV_ID_I210_SERDES_FLASHLESS, "Intel(R) PRO/1000 PCI-Express Network Driver"),
216 	PVID(0x8086, E1000_DEV_ID_I210_FIBER, "Intel(R) PRO/1000 PCI-Express Network Driver"),
217 	PVID(0x8086, E1000_DEV_ID_I210_SERDES, "Intel(R) PRO/1000 PCI-Express Network Driver"),
218 	PVID(0x8086, E1000_DEV_ID_I210_SGMII, "Intel(R) PRO/1000 PCI-Express Network Driver"),
219 	PVID(0x8086, E1000_DEV_ID_I211_COPPER, "Intel(R) PRO/1000 PCI-Express Network Driver"),
220 	PVID(0x8086, E1000_DEV_ID_I354_BACKPLANE_1GBPS, "Intel(R) PRO/1000 PCI-Express Network Driver"),
221 	PVID(0x8086, E1000_DEV_ID_I354_BACKPLANE_2_5GBPS, "Intel(R) PRO/1000 PCI-Express Network Driver"),
222 	PVID(0x8086, E1000_DEV_ID_I354_SGMII, "Intel(R) PRO/1000 PCI-Express Network Driver"),
223 	/* required last entry */
224 	PVID_END
225 };
226 
227 /*********************************************************************
228  *  Function prototypes
229  *********************************************************************/
230 static void	*em_register(device_t dev);
231 static void	*igb_register(device_t dev);
232 static int	em_if_attach_pre(if_ctx_t ctx);
233 static int	em_if_attach_post(if_ctx_t ctx);
234 static int	em_if_detach(if_ctx_t ctx);
235 static int	em_if_shutdown(if_ctx_t ctx);
236 static int	em_if_suspend(if_ctx_t ctx);
237 static int	em_if_resume(if_ctx_t ctx);
238 
239 static int	em_if_tx_queues_alloc(if_ctx_t ctx, caddr_t *vaddrs, uint64_t *paddrs, int ntxqs, int ntxqsets);
240 static int	em_if_rx_queues_alloc(if_ctx_t ctx, caddr_t *vaddrs, uint64_t *paddrs, int nrxqs, int nrxqsets);
241 static void	em_if_queues_free(if_ctx_t ctx);
242 
243 static uint64_t	em_if_get_counter(if_ctx_t, ift_counter);
244 static void	em_if_init(if_ctx_t ctx);
245 static void	em_if_stop(if_ctx_t ctx);
246 static void	em_if_media_status(if_ctx_t, struct ifmediareq *);
247 static int	em_if_media_change(if_ctx_t ctx);
248 static int	em_if_mtu_set(if_ctx_t ctx, uint32_t mtu);
249 static void	em_if_timer(if_ctx_t ctx, uint16_t qid);
250 static void	em_if_vlan_register(if_ctx_t ctx, u16 vtag);
251 static void	em_if_vlan_unregister(if_ctx_t ctx, u16 vtag);
252 static void	em_if_watchdog_reset(if_ctx_t ctx);
253 
254 static void	em_identify_hardware(if_ctx_t ctx);
255 static int	em_allocate_pci_resources(if_ctx_t ctx);
256 static void	em_free_pci_resources(if_ctx_t ctx);
257 static void	em_reset(if_ctx_t ctx);
258 static int	em_setup_interface(if_ctx_t ctx);
259 static int	em_setup_msix(if_ctx_t ctx);
260 
261 static void	em_initialize_transmit_unit(if_ctx_t ctx);
262 static void	em_initialize_receive_unit(if_ctx_t ctx);
263 
264 static void	em_if_enable_intr(if_ctx_t ctx);
265 static void	em_if_disable_intr(if_ctx_t ctx);
266 static int	em_if_rx_queue_intr_enable(if_ctx_t ctx, uint16_t rxqid);
267 static int	em_if_tx_queue_intr_enable(if_ctx_t ctx, uint16_t txqid);
268 static void	em_if_multi_set(if_ctx_t ctx);
269 static void	em_if_update_admin_status(if_ctx_t ctx);
270 static void	em_if_debug(if_ctx_t ctx);
271 static void	em_update_stats_counters(struct adapter *);
272 static void	em_add_hw_stats(struct adapter *adapter);
273 static int	em_if_set_promisc(if_ctx_t ctx, int flags);
274 static void	em_setup_vlan_hw_support(struct adapter *);
275 static int	em_sysctl_nvm_info(SYSCTL_HANDLER_ARGS);
276 static void	em_print_nvm_info(struct adapter *);
277 static int	em_sysctl_debug_info(SYSCTL_HANDLER_ARGS);
278 static int	em_get_rs(SYSCTL_HANDLER_ARGS);
279 static void	em_print_debug_info(struct adapter *);
280 static int 	em_is_valid_ether_addr(u8 *);
281 static int	em_sysctl_int_delay(SYSCTL_HANDLER_ARGS);
282 static void	em_add_int_delay_sysctl(struct adapter *, const char *,
283 		    const char *, struct em_int_delay_info *, int, int);
284 /* Management and WOL Support */
285 static void	em_init_manageability(struct adapter *);
286 static void	em_release_manageability(struct adapter *);
287 static void	em_get_hw_control(struct adapter *);
288 static void	em_release_hw_control(struct adapter *);
289 static void	em_get_wakeup(if_ctx_t ctx);
290 static void	em_enable_wakeup(if_ctx_t ctx);
291 static int	em_enable_phy_wakeup(struct adapter *);
292 static void	em_disable_aspm(struct adapter *);
293 
294 int		em_intr(void *arg);
295 static void	em_disable_promisc(if_ctx_t ctx);
296 
297 /* MSI-X handlers */
298 static int	em_if_msix_intr_assign(if_ctx_t, int);
299 static int	em_msix_link(void *);
300 static void	em_handle_link(void *context);
301 
302 static void	em_enable_vectors_82574(if_ctx_t);
303 
304 static int	em_set_flowcntl(SYSCTL_HANDLER_ARGS);
305 static int	em_sysctl_eee(SYSCTL_HANDLER_ARGS);
306 static void	em_if_led_func(if_ctx_t ctx, int onoff);
307 
308 static int	em_get_regs(SYSCTL_HANDLER_ARGS);
309 
310 static void	lem_smartspeed(struct adapter *adapter);
311 static void	igb_configure_queues(struct adapter *adapter);
312 
313 
314 /*********************************************************************
315  *  FreeBSD Device Interface Entry Points
316  *********************************************************************/
317 static device_method_t em_methods[] = {
318 	/* Device interface */
319 	DEVMETHOD(device_register, em_register),
320 	DEVMETHOD(device_probe, iflib_device_probe),
321 	DEVMETHOD(device_attach, iflib_device_attach),
322 	DEVMETHOD(device_detach, iflib_device_detach),
323 	DEVMETHOD(device_shutdown, iflib_device_shutdown),
324 	DEVMETHOD(device_suspend, iflib_device_suspend),
325 	DEVMETHOD(device_resume, iflib_device_resume),
326 	DEVMETHOD_END
327 };
328 
329 static device_method_t igb_methods[] = {
330 	/* Device interface */
331 	DEVMETHOD(device_register, igb_register),
332 	DEVMETHOD(device_probe, iflib_device_probe),
333 	DEVMETHOD(device_attach, iflib_device_attach),
334 	DEVMETHOD(device_detach, iflib_device_detach),
335 	DEVMETHOD(device_shutdown, iflib_device_shutdown),
336 	DEVMETHOD(device_suspend, iflib_device_suspend),
337 	DEVMETHOD(device_resume, iflib_device_resume),
338 	DEVMETHOD_END
339 };
340 
341 
342 static driver_t em_driver = {
343 	"em", em_methods, sizeof(struct adapter),
344 };
345 
346 static devclass_t em_devclass;
347 DRIVER_MODULE(em, pci, em_driver, em_devclass, 0, 0);
348 
349 MODULE_DEPEND(em, pci, 1, 1, 1);
350 MODULE_DEPEND(em, ether, 1, 1, 1);
351 MODULE_DEPEND(em, iflib, 1, 1, 1);
352 
353 IFLIB_PNP_INFO(pci, em, em_vendor_info_array);
354 
355 static driver_t igb_driver = {
356 	"igb", igb_methods, sizeof(struct adapter),
357 };
358 
359 static devclass_t igb_devclass;
360 DRIVER_MODULE(igb, pci, igb_driver, igb_devclass, 0, 0);
361 
362 MODULE_DEPEND(igb, pci, 1, 1, 1);
363 MODULE_DEPEND(igb, ether, 1, 1, 1);
364 MODULE_DEPEND(igb, iflib, 1, 1, 1);
365 
366 IFLIB_PNP_INFO(pci, igb, igb_vendor_info_array);
367 
368 static device_method_t em_if_methods[] = {
369 	DEVMETHOD(ifdi_attach_pre, em_if_attach_pre),
370 	DEVMETHOD(ifdi_attach_post, em_if_attach_post),
371 	DEVMETHOD(ifdi_detach, em_if_detach),
372 	DEVMETHOD(ifdi_shutdown, em_if_shutdown),
373 	DEVMETHOD(ifdi_suspend, em_if_suspend),
374 	DEVMETHOD(ifdi_resume, em_if_resume),
375 	DEVMETHOD(ifdi_init, em_if_init),
376 	DEVMETHOD(ifdi_stop, em_if_stop),
377 	DEVMETHOD(ifdi_msix_intr_assign, em_if_msix_intr_assign),
378 	DEVMETHOD(ifdi_intr_enable, em_if_enable_intr),
379 	DEVMETHOD(ifdi_intr_disable, em_if_disable_intr),
380 	DEVMETHOD(ifdi_tx_queues_alloc, em_if_tx_queues_alloc),
381 	DEVMETHOD(ifdi_rx_queues_alloc, em_if_rx_queues_alloc),
382 	DEVMETHOD(ifdi_queues_free, em_if_queues_free),
383 	DEVMETHOD(ifdi_update_admin_status, em_if_update_admin_status),
384 	DEVMETHOD(ifdi_multi_set, em_if_multi_set),
385 	DEVMETHOD(ifdi_media_status, em_if_media_status),
386 	DEVMETHOD(ifdi_media_change, em_if_media_change),
387 	DEVMETHOD(ifdi_mtu_set, em_if_mtu_set),
388 	DEVMETHOD(ifdi_promisc_set, em_if_set_promisc),
389 	DEVMETHOD(ifdi_timer, em_if_timer),
390 	DEVMETHOD(ifdi_watchdog_reset, em_if_watchdog_reset),
391 	DEVMETHOD(ifdi_vlan_register, em_if_vlan_register),
392 	DEVMETHOD(ifdi_vlan_unregister, em_if_vlan_unregister),
393 	DEVMETHOD(ifdi_get_counter, em_if_get_counter),
394 	DEVMETHOD(ifdi_led_func, em_if_led_func),
395 	DEVMETHOD(ifdi_rx_queue_intr_enable, em_if_rx_queue_intr_enable),
396 	DEVMETHOD(ifdi_tx_queue_intr_enable, em_if_tx_queue_intr_enable),
397 	DEVMETHOD(ifdi_debug, em_if_debug),
398 	DEVMETHOD_END
399 };
400 
401 /*
402  * note that if (adapter->msix_mem) is replaced by:
403  * if (adapter->intr_type == IFLIB_INTR_MSIX)
404  */
405 static driver_t em_if_driver = {
406 	"em_if", em_if_methods, sizeof(struct adapter)
407 };
408 
409 /*********************************************************************
410  *  Tunable default values.
411  *********************************************************************/
412 
413 #define EM_TICKS_TO_USECS(ticks)	((1024 * (ticks) + 500) / 1000)
414 #define EM_USECS_TO_TICKS(usecs)	((1000 * (usecs) + 512) / 1024)
415 
416 #define MAX_INTS_PER_SEC	8000
417 #define DEFAULT_ITR		(1000000000/(MAX_INTS_PER_SEC * 256))
418 
419 /* Allow common code without TSO */
420 #ifndef CSUM_TSO
421 #define CSUM_TSO	0
422 #endif
423 
424 static SYSCTL_NODE(_hw, OID_AUTO, em, CTLFLAG_RD, 0, "EM driver parameters");
425 
426 static int em_disable_crc_stripping = 0;
427 SYSCTL_INT(_hw_em, OID_AUTO, disable_crc_stripping, CTLFLAG_RDTUN,
428     &em_disable_crc_stripping, 0, "Disable CRC Stripping");
429 
430 static int em_tx_int_delay_dflt = EM_TICKS_TO_USECS(EM_TIDV);
431 static int em_rx_int_delay_dflt = EM_TICKS_TO_USECS(EM_RDTR);
432 SYSCTL_INT(_hw_em, OID_AUTO, tx_int_delay, CTLFLAG_RDTUN, &em_tx_int_delay_dflt,
433     0, "Default transmit interrupt delay in usecs");
434 SYSCTL_INT(_hw_em, OID_AUTO, rx_int_delay, CTLFLAG_RDTUN, &em_rx_int_delay_dflt,
435     0, "Default receive interrupt delay in usecs");
436 
437 static int em_tx_abs_int_delay_dflt = EM_TICKS_TO_USECS(EM_TADV);
438 static int em_rx_abs_int_delay_dflt = EM_TICKS_TO_USECS(EM_RADV);
439 SYSCTL_INT(_hw_em, OID_AUTO, tx_abs_int_delay, CTLFLAG_RDTUN,
440     &em_tx_abs_int_delay_dflt, 0,
441     "Default transmit interrupt delay limit in usecs");
442 SYSCTL_INT(_hw_em, OID_AUTO, rx_abs_int_delay, CTLFLAG_RDTUN,
443     &em_rx_abs_int_delay_dflt, 0,
444     "Default receive interrupt delay limit in usecs");
445 
446 static int em_smart_pwr_down = FALSE;
447 SYSCTL_INT(_hw_em, OID_AUTO, smart_pwr_down, CTLFLAG_RDTUN, &em_smart_pwr_down,
448     0, "Set to true to leave smart power down enabled on newer adapters");
449 
450 /* Controls whether promiscuous also shows bad packets */
451 static int em_debug_sbp = TRUE;
452 SYSCTL_INT(_hw_em, OID_AUTO, sbp, CTLFLAG_RDTUN, &em_debug_sbp, 0,
453     "Show bad packets in promiscuous mode");
454 
455 /* How many packets rxeof tries to clean at a time */
456 static int em_rx_process_limit = 100;
457 SYSCTL_INT(_hw_em, OID_AUTO, rx_process_limit, CTLFLAG_RDTUN,
458     &em_rx_process_limit, 0,
459     "Maximum number of received packets to process "
460     "at a time, -1 means unlimited");
461 
462 /* Energy efficient ethernet - default to OFF */
463 static int eee_setting = 1;
464 SYSCTL_INT(_hw_em, OID_AUTO, eee_setting, CTLFLAG_RDTUN, &eee_setting, 0,
465     "Enable Energy Efficient Ethernet");
466 
467 /*
468 ** Tuneable Interrupt rate
469 */
470 static int em_max_interrupt_rate = 8000;
471 SYSCTL_INT(_hw_em, OID_AUTO, max_interrupt_rate, CTLFLAG_RDTUN,
472     &em_max_interrupt_rate, 0, "Maximum interrupts per second");
473 
474 
475 
476 /* Global used in WOL setup with multiport cards */
477 static int global_quad_port_a = 0;
478 
479 extern struct if_txrx igb_txrx;
480 extern struct if_txrx em_txrx;
481 extern struct if_txrx lem_txrx;
482 
483 static struct if_shared_ctx em_sctx_init = {
484 	.isc_magic = IFLIB_MAGIC,
485 	.isc_q_align = PAGE_SIZE,
486 	.isc_tx_maxsize = EM_TSO_SIZE + sizeof(struct ether_vlan_header),
487 	.isc_tx_maxsegsize = PAGE_SIZE,
488 	.isc_tso_maxsize = EM_TSO_SIZE + sizeof(struct ether_vlan_header),
489 	.isc_tso_maxsegsize = EM_TSO_SEG_SIZE,
490 	.isc_rx_maxsize = MJUM9BYTES,
491 	.isc_rx_nsegments = 1,
492 	.isc_rx_maxsegsize = MJUM9BYTES,
493 	.isc_nfl = 1,
494 	.isc_nrxqs = 1,
495 	.isc_ntxqs = 1,
496 	.isc_admin_intrcnt = 1,
497 	.isc_vendor_info = em_vendor_info_array,
498 	.isc_driver_version = em_driver_version,
499 	.isc_driver = &em_if_driver,
500 	.isc_flags = IFLIB_NEED_SCRATCH | IFLIB_TSO_INIT_IP | IFLIB_NEED_ZERO_CSUM,
501 
502 	.isc_nrxd_min = {EM_MIN_RXD},
503 	.isc_ntxd_min = {EM_MIN_TXD},
504 	.isc_nrxd_max = {EM_MAX_RXD},
505 	.isc_ntxd_max = {EM_MAX_TXD},
506 	.isc_nrxd_default = {EM_DEFAULT_RXD},
507 	.isc_ntxd_default = {EM_DEFAULT_TXD},
508 };
509 
510 if_shared_ctx_t em_sctx = &em_sctx_init;
511 
512 static struct if_shared_ctx igb_sctx_init = {
513 	.isc_magic = IFLIB_MAGIC,
514 	.isc_q_align = PAGE_SIZE,
515 	.isc_tx_maxsize = EM_TSO_SIZE + sizeof(struct ether_vlan_header),
516 	.isc_tx_maxsegsize = PAGE_SIZE,
517 	.isc_tso_maxsize = EM_TSO_SIZE + sizeof(struct ether_vlan_header),
518 	.isc_tso_maxsegsize = EM_TSO_SEG_SIZE,
519 	.isc_rx_maxsize = MJUM9BYTES,
520 	.isc_rx_nsegments = 1,
521 	.isc_rx_maxsegsize = MJUM9BYTES,
522 	.isc_nfl = 1,
523 	.isc_nrxqs = 1,
524 	.isc_ntxqs = 1,
525 	.isc_admin_intrcnt = 1,
526 	.isc_vendor_info = igb_vendor_info_array,
527 	.isc_driver_version = em_driver_version,
528 	.isc_driver = &em_if_driver,
529 	.isc_flags = IFLIB_NEED_SCRATCH | IFLIB_TSO_INIT_IP | IFLIB_NEED_ZERO_CSUM,
530 
531 	.isc_nrxd_min = {EM_MIN_RXD},
532 	.isc_ntxd_min = {EM_MIN_TXD},
533 	.isc_nrxd_max = {IGB_MAX_RXD},
534 	.isc_ntxd_max = {IGB_MAX_TXD},
535 	.isc_nrxd_default = {EM_DEFAULT_RXD},
536 	.isc_ntxd_default = {EM_DEFAULT_TXD},
537 };
538 
539 if_shared_ctx_t igb_sctx = &igb_sctx_init;
540 
541 /*****************************************************************
542  *
543  * Dump Registers
544  *
545  ****************************************************************/
546 #define IGB_REGS_LEN 739
547 
548 static int em_get_regs(SYSCTL_HANDLER_ARGS)
549 {
550 	struct adapter *adapter = (struct adapter *)arg1;
551 	struct e1000_hw *hw = &adapter->hw;
552 	struct sbuf *sb;
553 	u32 *regs_buff;
554 	int rc;
555 
556 	regs_buff = malloc(sizeof(u32) * IGB_REGS_LEN, M_DEVBUF, M_WAITOK);
557 	memset(regs_buff, 0, IGB_REGS_LEN * sizeof(u32));
558 
559 	rc = sysctl_wire_old_buffer(req, 0);
560 	MPASS(rc == 0);
561 	if (rc != 0) {
562 		free(regs_buff, M_DEVBUF);
563 		return (rc);
564 	}
565 
566 	sb = sbuf_new_for_sysctl(NULL, NULL, 32*400, req);
567 	MPASS(sb != NULL);
568 	if (sb == NULL) {
569 		free(regs_buff, M_DEVBUF);
570 		return (ENOMEM);
571 	}
572 
573 	/* General Registers */
574 	regs_buff[0] = E1000_READ_REG(hw, E1000_CTRL);
575 	regs_buff[1] = E1000_READ_REG(hw, E1000_STATUS);
576 	regs_buff[2] = E1000_READ_REG(hw, E1000_CTRL_EXT);
577 	regs_buff[3] = E1000_READ_REG(hw, E1000_ICR);
578 	regs_buff[4] = E1000_READ_REG(hw, E1000_RCTL);
579 	regs_buff[5] = E1000_READ_REG(hw, E1000_RDLEN(0));
580 	regs_buff[6] = E1000_READ_REG(hw, E1000_RDH(0));
581 	regs_buff[7] = E1000_READ_REG(hw, E1000_RDT(0));
582 	regs_buff[8] = E1000_READ_REG(hw, E1000_RXDCTL(0));
583 	regs_buff[9] = E1000_READ_REG(hw, E1000_RDBAL(0));
584 	regs_buff[10] = E1000_READ_REG(hw, E1000_RDBAH(0));
585 	regs_buff[11] = E1000_READ_REG(hw, E1000_TCTL);
586 	regs_buff[12] = E1000_READ_REG(hw, E1000_TDBAL(0));
587 	regs_buff[13] = E1000_READ_REG(hw, E1000_TDBAH(0));
588 	regs_buff[14] = E1000_READ_REG(hw, E1000_TDLEN(0));
589 	regs_buff[15] = E1000_READ_REG(hw, E1000_TDH(0));
590 	regs_buff[16] = E1000_READ_REG(hw, E1000_TDT(0));
591 	regs_buff[17] = E1000_READ_REG(hw, E1000_TXDCTL(0));
592 	regs_buff[18] = E1000_READ_REG(hw, E1000_TDFH);
593 	regs_buff[19] = E1000_READ_REG(hw, E1000_TDFT);
594 	regs_buff[20] = E1000_READ_REG(hw, E1000_TDFHS);
595 	regs_buff[21] = E1000_READ_REG(hw, E1000_TDFPC);
596 
597 	sbuf_printf(sb, "General Registers\n");
598 	sbuf_printf(sb, "\tCTRL\t %08x\n", regs_buff[0]);
599 	sbuf_printf(sb, "\tSTATUS\t %08x\n", regs_buff[1]);
600 	sbuf_printf(sb, "\tCTRL_EXIT\t %08x\n\n", regs_buff[2]);
601 
602 	sbuf_printf(sb, "Interrupt Registers\n");
603 	sbuf_printf(sb, "\tICR\t %08x\n\n", regs_buff[3]);
604 
605 	sbuf_printf(sb, "RX Registers\n");
606 	sbuf_printf(sb, "\tRCTL\t %08x\n", regs_buff[4]);
607 	sbuf_printf(sb, "\tRDLEN\t %08x\n", regs_buff[5]);
608 	sbuf_printf(sb, "\tRDH\t %08x\n", regs_buff[6]);
609 	sbuf_printf(sb, "\tRDT\t %08x\n", regs_buff[7]);
610 	sbuf_printf(sb, "\tRXDCTL\t %08x\n", regs_buff[8]);
611 	sbuf_printf(sb, "\tRDBAL\t %08x\n", regs_buff[9]);
612 	sbuf_printf(sb, "\tRDBAH\t %08x\n\n", regs_buff[10]);
613 
614 	sbuf_printf(sb, "TX Registers\n");
615 	sbuf_printf(sb, "\tTCTL\t %08x\n", regs_buff[11]);
616 	sbuf_printf(sb, "\tTDBAL\t %08x\n", regs_buff[12]);
617 	sbuf_printf(sb, "\tTDBAH\t %08x\n", regs_buff[13]);
618 	sbuf_printf(sb, "\tTDLEN\t %08x\n", regs_buff[14]);
619 	sbuf_printf(sb, "\tTDH\t %08x\n", regs_buff[15]);
620 	sbuf_printf(sb, "\tTDT\t %08x\n", regs_buff[16]);
621 	sbuf_printf(sb, "\tTXDCTL\t %08x\n", regs_buff[17]);
622 	sbuf_printf(sb, "\tTDFH\t %08x\n", regs_buff[18]);
623 	sbuf_printf(sb, "\tTDFT\t %08x\n", regs_buff[19]);
624 	sbuf_printf(sb, "\tTDFHS\t %08x\n", regs_buff[20]);
625 	sbuf_printf(sb, "\tTDFPC\t %08x\n\n", regs_buff[21]);
626 
627 	free(regs_buff, M_DEVBUF);
628 
629 #ifdef DUMP_DESCS
630 	{
631 		if_softc_ctx_t scctx = adapter->shared;
632 		struct rx_ring *rxr = &rx_que->rxr;
633 		struct tx_ring *txr = &tx_que->txr;
634 		int ntxd = scctx->isc_ntxd[0];
635 		int nrxd = scctx->isc_nrxd[0];
636 		int j;
637 
638 	for (j = 0; j < nrxd; j++) {
639 		u32 staterr = le32toh(rxr->rx_base[j].wb.upper.status_error);
640 		u32 length =  le32toh(rxr->rx_base[j].wb.upper.length);
641 		sbuf_printf(sb, "\tReceive Descriptor Address %d: %08" PRIx64 "  Error:%d  Length:%d\n", j, rxr->rx_base[j].read.buffer_addr, staterr, length);
642 	}
643 
644 	for (j = 0; j < min(ntxd, 256); j++) {
645 		unsigned int *ptr = (unsigned int *)&txr->tx_base[j];
646 
647 		sbuf_printf(sb, "\tTXD[%03d] [0]: %08x [1]: %08x [2]: %08x [3]: %08x  eop: %d DD=%d\n",
648 			    j, ptr[0], ptr[1], ptr[2], ptr[3], buf->eop,
649 			    buf->eop != -1 ? txr->tx_base[buf->eop].upper.fields.status & E1000_TXD_STAT_DD : 0);
650 
651 	}
652 	}
653 #endif
654 
655 	rc = sbuf_finish(sb);
656 	sbuf_delete(sb);
657 	return(rc);
658 }
659 
660 static void *
661 em_register(device_t dev)
662 {
663 	return (em_sctx);
664 }
665 
666 static void *
667 igb_register(device_t dev)
668 {
669 	return (igb_sctx);
670 }
671 
672 static int
673 em_set_num_queues(if_ctx_t ctx)
674 {
675 	struct adapter *adapter = iflib_get_softc(ctx);
676 	int maxqueues;
677 
678 	/* Sanity check based on HW */
679 	switch (adapter->hw.mac.type) {
680 	case e1000_82576:
681 	case e1000_82580:
682 	case e1000_i350:
683 	case e1000_i354:
684 		maxqueues = 8;
685 		break;
686 	case e1000_i210:
687 	case e1000_82575:
688 		maxqueues = 4;
689 		break;
690 	case e1000_i211:
691 	case e1000_82574:
692 		maxqueues = 2;
693 		break;
694 	default:
695 		maxqueues = 1;
696 		break;
697 	}
698 
699 	return (maxqueues);
700 }
701 
702 #define	LEM_CAPS							\
703     IFCAP_HWCSUM | IFCAP_VLAN_MTU | IFCAP_VLAN_HWTAGGING |		\
704     IFCAP_VLAN_HWCSUM | IFCAP_WOL | IFCAP_VLAN_HWFILTER
705 
706 #define	EM_CAPS								\
707     IFCAP_HWCSUM | IFCAP_VLAN_MTU | IFCAP_VLAN_HWTAGGING |		\
708     IFCAP_VLAN_HWCSUM | IFCAP_WOL | IFCAP_VLAN_HWFILTER | IFCAP_TSO4 |	\
709     IFCAP_LRO | IFCAP_VLAN_HWTSO
710 
711 #define	IGB_CAPS							\
712     IFCAP_HWCSUM | IFCAP_VLAN_MTU | IFCAP_VLAN_HWTAGGING |		\
713     IFCAP_VLAN_HWCSUM | IFCAP_WOL | IFCAP_VLAN_HWFILTER | IFCAP_TSO4 |	\
714     IFCAP_LRO | IFCAP_VLAN_HWTSO | IFCAP_JUMBO_MTU | IFCAP_HWCSUM_IPV6 |\
715     IFCAP_TSO6
716 
717 /*********************************************************************
718  *  Device initialization routine
719  *
720  *  The attach entry point is called when the driver is being loaded.
721  *  This routine identifies the type of hardware, allocates all resources
722  *  and initializes the hardware.
723  *
724  *  return 0 on success, positive on failure
725  *********************************************************************/
726 static int
727 em_if_attach_pre(if_ctx_t ctx)
728 {
729 	struct adapter *adapter;
730 	if_softc_ctx_t scctx;
731 	device_t dev;
732 	struct e1000_hw *hw;
733 	int error = 0;
734 
735 	INIT_DEBUGOUT("em_if_attach_pre: begin");
736 	dev = iflib_get_dev(ctx);
737 	adapter = iflib_get_softc(ctx);
738 
739 	adapter->ctx = adapter->osdep.ctx = ctx;
740 	adapter->dev = adapter->osdep.dev = dev;
741 	scctx = adapter->shared = iflib_get_softc_ctx(ctx);
742 	adapter->media = iflib_get_media(ctx);
743 	hw = &adapter->hw;
744 
745 	adapter->tx_process_limit = scctx->isc_ntxd[0];
746 
747 	/* SYSCTL stuff */
748 	SYSCTL_ADD_PROC(device_get_sysctl_ctx(dev),
749 	    SYSCTL_CHILDREN(device_get_sysctl_tree(dev)),
750 	    OID_AUTO, "nvm", CTLTYPE_INT|CTLFLAG_RW, adapter, 0,
751 	    em_sysctl_nvm_info, "I", "NVM Information");
752 
753 	SYSCTL_ADD_PROC(device_get_sysctl_ctx(dev),
754 	    SYSCTL_CHILDREN(device_get_sysctl_tree(dev)),
755 	    OID_AUTO, "debug", CTLTYPE_INT|CTLFLAG_RW, adapter, 0,
756 	    em_sysctl_debug_info, "I", "Debug Information");
757 
758 	SYSCTL_ADD_PROC(device_get_sysctl_ctx(dev),
759 	    SYSCTL_CHILDREN(device_get_sysctl_tree(dev)),
760 	    OID_AUTO, "fc", CTLTYPE_INT|CTLFLAG_RW, adapter, 0,
761 	    em_set_flowcntl, "I", "Flow Control");
762 
763 	SYSCTL_ADD_PROC(device_get_sysctl_ctx(dev),
764 	    SYSCTL_CHILDREN(device_get_sysctl_tree(dev)),
765 	    OID_AUTO, "reg_dump", CTLTYPE_STRING | CTLFLAG_RD, adapter, 0,
766 	    em_get_regs, "A", "Dump Registers");
767 
768 	SYSCTL_ADD_PROC(device_get_sysctl_ctx(dev),
769 	    SYSCTL_CHILDREN(device_get_sysctl_tree(dev)),
770 	    OID_AUTO, "rs_dump", CTLTYPE_INT | CTLFLAG_RW, adapter, 0,
771 	    em_get_rs, "I", "Dump RS indexes");
772 
773 	/* Determine hardware and mac info */
774 	em_identify_hardware(ctx);
775 
776 	scctx->isc_tx_nsegments = EM_MAX_SCATTER;
777 	scctx->isc_nrxqsets_max = scctx->isc_ntxqsets_max = em_set_num_queues(ctx);
778 	if (bootverbose)
779 		device_printf(dev, "attach_pre capping queues at %d\n",
780 		    scctx->isc_ntxqsets_max);
781 
782 	if (adapter->hw.mac.type >= igb_mac_min) {
783 		scctx->isc_txqsizes[0] = roundup2(scctx->isc_ntxd[0] * sizeof(union e1000_adv_tx_desc), EM_DBA_ALIGN);
784 		scctx->isc_rxqsizes[0] = roundup2(scctx->isc_nrxd[0] * sizeof(union e1000_adv_rx_desc), EM_DBA_ALIGN);
785 		scctx->isc_txd_size[0] = sizeof(union e1000_adv_tx_desc);
786 		scctx->isc_rxd_size[0] = sizeof(union e1000_adv_rx_desc);
787 		scctx->isc_txrx = &igb_txrx;
788 		scctx->isc_tx_tso_segments_max = EM_MAX_SCATTER;
789 		scctx->isc_tx_tso_size_max = EM_TSO_SIZE;
790 		scctx->isc_tx_tso_segsize_max = EM_TSO_SEG_SIZE;
791 		scctx->isc_capabilities = scctx->isc_capenable = IGB_CAPS;
792 		scctx->isc_tx_csum_flags = CSUM_TCP | CSUM_UDP | CSUM_TSO |
793 		     CSUM_IP6_TCP | CSUM_IP6_UDP;
794 		if (adapter->hw.mac.type != e1000_82575)
795 			scctx->isc_tx_csum_flags |= CSUM_SCTP | CSUM_IP6_SCTP;
796 		/*
797 		** Some new devices, as with ixgbe, now may
798 		** use a different BAR, so we need to keep
799 		** track of which is used.
800 		*/
801 		scctx->isc_msix_bar = PCIR_BAR(EM_MSIX_BAR);
802 		if (pci_read_config(dev, scctx->isc_msix_bar, 4) == 0)
803 			scctx->isc_msix_bar += 4;
804 	} else if (adapter->hw.mac.type >= em_mac_min) {
805 		scctx->isc_txqsizes[0] = roundup2(scctx->isc_ntxd[0]* sizeof(struct e1000_tx_desc), EM_DBA_ALIGN);
806 		scctx->isc_rxqsizes[0] = roundup2(scctx->isc_nrxd[0] * sizeof(union e1000_rx_desc_extended), EM_DBA_ALIGN);
807 		scctx->isc_txd_size[0] = sizeof(struct e1000_tx_desc);
808 		scctx->isc_rxd_size[0] = sizeof(union e1000_rx_desc_extended);
809 		scctx->isc_txrx = &em_txrx;
810 		scctx->isc_tx_tso_segments_max = EM_MAX_SCATTER;
811 		scctx->isc_tx_tso_size_max = EM_TSO_SIZE;
812 		scctx->isc_tx_tso_segsize_max = EM_TSO_SEG_SIZE;
813 		scctx->isc_capabilities = scctx->isc_capenable = EM_CAPS;
814 		/*
815 		 * For EM-class devices, don't enable IFCAP_{TSO4,VLAN_HWTSO}
816 		 * by default as we don't have workarounds for all associated
817 		 * silicon errata.  E. g., with several MACs such as 82573E,
818 		 * TSO only works at Gigabit speed and otherwise can cause the
819 		 * hardware to hang (which also would be next to impossible to
820 		 * work around given that already queued TSO-using descriptors
821 		 * would need to be flushed and vlan(4) reconfigured at runtime
822 		 * in case of a link speed change).  Moreover, MACs like 82579
823 		 * still can hang at Gigabit even with all publicly documented
824 		 * TSO workarounds implemented.  Generally, the penality of
825 		 * these workarounds is rather high and may involve copying
826 		 * mbuf data around so advantages of TSO lapse.  Still, TSO may
827 		 * work for a few MACs of this class - at least when sticking
828 		 * with Gigabit - in which case users may enable TSO manually.
829 		 */
830 		scctx->isc_capenable &= ~(IFCAP_TSO4 | IFCAP_VLAN_HWTSO);
831 		scctx->isc_tx_csum_flags = CSUM_TCP | CSUM_UDP | CSUM_IP_TSO;
832 		/*
833 		 * We support MSI-X with 82574 only, but indicate to iflib(4)
834 		 * that it shall give MSI at least a try with other devices.
835 		 */
836 		if (adapter->hw.mac.type == e1000_82574) {
837 			scctx->isc_msix_bar = PCIR_BAR(EM_MSIX_BAR);
838 		} else {
839 			scctx->isc_msix_bar = -1;
840 			scctx->isc_disable_msix = 1;
841 		}
842 	} else {
843 		scctx->isc_txqsizes[0] = roundup2((scctx->isc_ntxd[0] + 1) * sizeof(struct e1000_tx_desc), EM_DBA_ALIGN);
844 		scctx->isc_rxqsizes[0] = roundup2((scctx->isc_nrxd[0] + 1) * sizeof(struct e1000_rx_desc), EM_DBA_ALIGN);
845 		scctx->isc_txd_size[0] = sizeof(struct e1000_tx_desc);
846 		scctx->isc_rxd_size[0] = sizeof(struct e1000_rx_desc);
847 		scctx->isc_tx_csum_flags = CSUM_TCP | CSUM_UDP;
848 		scctx->isc_txrx = &lem_txrx;
849 		scctx->isc_capabilities = scctx->isc_capenable = LEM_CAPS;
850 		if (adapter->hw.mac.type < e1000_82543)
851 			scctx->isc_capenable &= ~(IFCAP_HWCSUM|IFCAP_VLAN_HWCSUM);
852 		/* INTx only */
853 		scctx->isc_msix_bar = 0;
854 	}
855 
856 	/* Setup PCI resources */
857 	if (em_allocate_pci_resources(ctx)) {
858 		device_printf(dev, "Allocation of PCI resources failed\n");
859 		error = ENXIO;
860 		goto err_pci;
861 	}
862 
863 	/*
864 	** For ICH8 and family we need to
865 	** map the flash memory, and this
866 	** must happen after the MAC is
867 	** identified
868 	*/
869 	if ((hw->mac.type == e1000_ich8lan) ||
870 	    (hw->mac.type == e1000_ich9lan) ||
871 	    (hw->mac.type == e1000_ich10lan) ||
872 	    (hw->mac.type == e1000_pchlan) ||
873 	    (hw->mac.type == e1000_pch2lan) ||
874 	    (hw->mac.type == e1000_pch_lpt)) {
875 		int rid = EM_BAR_TYPE_FLASH;
876 		adapter->flash = bus_alloc_resource_any(dev,
877 		    SYS_RES_MEMORY, &rid, RF_ACTIVE);
878 		if (adapter->flash == NULL) {
879 			device_printf(dev, "Mapping of Flash failed\n");
880 			error = ENXIO;
881 			goto err_pci;
882 		}
883 		/* This is used in the shared code */
884 		hw->flash_address = (u8 *)adapter->flash;
885 		adapter->osdep.flash_bus_space_tag =
886 		    rman_get_bustag(adapter->flash);
887 		adapter->osdep.flash_bus_space_handle =
888 		    rman_get_bushandle(adapter->flash);
889 	}
890 	/*
891 	** In the new SPT device flash is not  a
892 	** separate BAR, rather it is also in BAR0,
893 	** so use the same tag and an offset handle for the
894 	** FLASH read/write macros in the shared code.
895 	*/
896 	else if (hw->mac.type >= e1000_pch_spt) {
897 		adapter->osdep.flash_bus_space_tag =
898 		    adapter->osdep.mem_bus_space_tag;
899 		adapter->osdep.flash_bus_space_handle =
900 		    adapter->osdep.mem_bus_space_handle
901 		    + E1000_FLASH_BASE_ADDR;
902 	}
903 
904 	/* Do Shared Code initialization */
905 	error = e1000_setup_init_funcs(hw, TRUE);
906 	if (error) {
907 		device_printf(dev, "Setup of Shared code failed, error %d\n",
908 		    error);
909 		error = ENXIO;
910 		goto err_pci;
911 	}
912 
913 	em_setup_msix(ctx);
914 	e1000_get_bus_info(hw);
915 
916 	/* Set up some sysctls for the tunable interrupt delays */
917 	em_add_int_delay_sysctl(adapter, "rx_int_delay",
918 	    "receive interrupt delay in usecs", &adapter->rx_int_delay,
919 	    E1000_REGISTER(hw, E1000_RDTR), em_rx_int_delay_dflt);
920 	em_add_int_delay_sysctl(adapter, "tx_int_delay",
921 	    "transmit interrupt delay in usecs", &adapter->tx_int_delay,
922 	    E1000_REGISTER(hw, E1000_TIDV), em_tx_int_delay_dflt);
923 	em_add_int_delay_sysctl(adapter, "rx_abs_int_delay",
924 	    "receive interrupt delay limit in usecs",
925 	    &adapter->rx_abs_int_delay,
926 	    E1000_REGISTER(hw, E1000_RADV),
927 	    em_rx_abs_int_delay_dflt);
928 	em_add_int_delay_sysctl(adapter, "tx_abs_int_delay",
929 	    "transmit interrupt delay limit in usecs",
930 	    &adapter->tx_abs_int_delay,
931 	    E1000_REGISTER(hw, E1000_TADV),
932 	    em_tx_abs_int_delay_dflt);
933 	em_add_int_delay_sysctl(adapter, "itr",
934 	    "interrupt delay limit in usecs/4",
935 	    &adapter->tx_itr,
936 	    E1000_REGISTER(hw, E1000_ITR),
937 	    DEFAULT_ITR);
938 
939 	hw->mac.autoneg = DO_AUTO_NEG;
940 	hw->phy.autoneg_wait_to_complete = FALSE;
941 	hw->phy.autoneg_advertised = AUTONEG_ADV_DEFAULT;
942 
943 	if (adapter->hw.mac.type < em_mac_min) {
944 		e1000_init_script_state_82541(&adapter->hw, TRUE);
945 		e1000_set_tbi_compatibility_82543(&adapter->hw, TRUE);
946 	}
947 	/* Copper options */
948 	if (hw->phy.media_type == e1000_media_type_copper) {
949 		hw->phy.mdix = AUTO_ALL_MODES;
950 		hw->phy.disable_polarity_correction = FALSE;
951 		hw->phy.ms_type = EM_MASTER_SLAVE;
952 	}
953 
954 	/*
955 	 * Set the frame limits assuming
956 	 * standard ethernet sized frames.
957 	 */
958 	scctx->isc_max_frame_size = adapter->hw.mac.max_frame_size =
959 	    ETHERMTU + ETHER_HDR_LEN + ETHERNET_FCS_SIZE;
960 
961 	/*
962 	 * This controls when hardware reports transmit completion
963 	 * status.
964 	 */
965 	hw->mac.report_tx_early = 1;
966 
967 	/* Allocate multicast array memory. */
968 	adapter->mta = malloc(sizeof(u8) * ETH_ADDR_LEN *
969 	    MAX_NUM_MULTICAST_ADDRESSES, M_DEVBUF, M_NOWAIT);
970 	if (adapter->mta == NULL) {
971 		device_printf(dev, "Can not allocate multicast setup array\n");
972 		error = ENOMEM;
973 		goto err_late;
974 	}
975 
976 	/* Check SOL/IDER usage */
977 	if (e1000_check_reset_block(hw))
978 		device_printf(dev, "PHY reset is blocked"
979 			      " due to SOL/IDER session.\n");
980 
981 	/* Sysctl for setting Energy Efficient Ethernet */
982 	hw->dev_spec.ich8lan.eee_disable = eee_setting;
983 	SYSCTL_ADD_PROC(device_get_sysctl_ctx(dev),
984 	    SYSCTL_CHILDREN(device_get_sysctl_tree(dev)),
985 	    OID_AUTO, "eee_control", CTLTYPE_INT|CTLFLAG_RW,
986 	    adapter, 0, em_sysctl_eee, "I",
987 	    "Disable Energy Efficient Ethernet");
988 
989 	/*
990 	** Start from a known state, this is
991 	** important in reading the nvm and
992 	** mac from that.
993 	*/
994 	e1000_reset_hw(hw);
995 
996 	/* Make sure we have a good EEPROM before we read from it */
997 	if (e1000_validate_nvm_checksum(hw) < 0) {
998 		/*
999 		** Some PCI-E parts fail the first check due to
1000 		** the link being in sleep state, call it again,
1001 		** if it fails a second time its a real issue.
1002 		*/
1003 		if (e1000_validate_nvm_checksum(hw) < 0) {
1004 			device_printf(dev,
1005 			    "The EEPROM Checksum Is Not Valid\n");
1006 			error = EIO;
1007 			goto err_late;
1008 		}
1009 	}
1010 
1011 	/* Copy the permanent MAC address out of the EEPROM */
1012 	if (e1000_read_mac_addr(hw) < 0) {
1013 		device_printf(dev, "EEPROM read error while reading MAC"
1014 			      " address\n");
1015 		error = EIO;
1016 		goto err_late;
1017 	}
1018 
1019 	if (!em_is_valid_ether_addr(hw->mac.addr)) {
1020 		device_printf(dev, "Invalid MAC address\n");
1021 		error = EIO;
1022 		goto err_late;
1023 	}
1024 
1025 	/* Disable ULP support */
1026 	e1000_disable_ulp_lpt_lp(hw, TRUE);
1027 
1028 	/*
1029 	 * Get Wake-on-Lan and Management info for later use
1030 	 */
1031 	em_get_wakeup(ctx);
1032 
1033 	/* Enable only WOL MAGIC by default */
1034 	scctx->isc_capenable &= ~IFCAP_WOL;
1035 	if (adapter->wol != 0)
1036 		scctx->isc_capenable |= IFCAP_WOL_MAGIC;
1037 
1038 	iflib_set_mac(ctx, hw->mac.addr);
1039 
1040 	return (0);
1041 
1042 err_late:
1043 	em_release_hw_control(adapter);
1044 err_pci:
1045 	em_free_pci_resources(ctx);
1046 	free(adapter->mta, M_DEVBUF);
1047 
1048 	return (error);
1049 }
1050 
1051 static int
1052 em_if_attach_post(if_ctx_t ctx)
1053 {
1054 	struct adapter *adapter = iflib_get_softc(ctx);
1055 	struct e1000_hw *hw = &adapter->hw;
1056 	int error = 0;
1057 
1058 	/* Setup OS specific network interface */
1059 	error = em_setup_interface(ctx);
1060 	if (error != 0) {
1061 		goto err_late;
1062 	}
1063 
1064 	em_reset(ctx);
1065 
1066 	/* Initialize statistics */
1067 	em_update_stats_counters(adapter);
1068 	hw->mac.get_link_status = 1;
1069 	em_if_update_admin_status(ctx);
1070 	em_add_hw_stats(adapter);
1071 
1072 	/* Non-AMT based hardware can now take control from firmware */
1073 	if (adapter->has_manage && !adapter->has_amt)
1074 		em_get_hw_control(adapter);
1075 
1076 	INIT_DEBUGOUT("em_if_attach_post: end");
1077 
1078 	return (error);
1079 
1080 err_late:
1081 	em_release_hw_control(adapter);
1082 	em_free_pci_resources(ctx);
1083 	em_if_queues_free(ctx);
1084 	free(adapter->mta, M_DEVBUF);
1085 
1086 	return (error);
1087 }
1088 
1089 /*********************************************************************
1090  *  Device removal routine
1091  *
1092  *  The detach entry point is called when the driver is being removed.
1093  *  This routine stops the adapter and deallocates all the resources
1094  *  that were allocated for driver operation.
1095  *
1096  *  return 0 on success, positive on failure
1097  *********************************************************************/
1098 static int
1099 em_if_detach(if_ctx_t ctx)
1100 {
1101 	struct adapter	*adapter = iflib_get_softc(ctx);
1102 
1103 	INIT_DEBUGOUT("em_if_detach: begin");
1104 
1105 	e1000_phy_hw_reset(&adapter->hw);
1106 
1107 	em_release_manageability(adapter);
1108 	em_release_hw_control(adapter);
1109 	em_free_pci_resources(ctx);
1110 
1111 	return (0);
1112 }
1113 
1114 /*********************************************************************
1115  *
1116  *  Shutdown entry point
1117  *
1118  **********************************************************************/
1119 
1120 static int
1121 em_if_shutdown(if_ctx_t ctx)
1122 {
1123 	return em_if_suspend(ctx);
1124 }
1125 
1126 /*
1127  * Suspend/resume device methods.
1128  */
1129 static int
1130 em_if_suspend(if_ctx_t ctx)
1131 {
1132 	struct adapter *adapter = iflib_get_softc(ctx);
1133 
1134 	em_release_manageability(adapter);
1135 	em_release_hw_control(adapter);
1136 	em_enable_wakeup(ctx);
1137 	return (0);
1138 }
1139 
1140 static int
1141 em_if_resume(if_ctx_t ctx)
1142 {
1143 	struct adapter *adapter = iflib_get_softc(ctx);
1144 
1145 	if (adapter->hw.mac.type == e1000_pch2lan)
1146 		e1000_resume_workarounds_pchlan(&adapter->hw);
1147 	em_if_init(ctx);
1148 	em_init_manageability(adapter);
1149 
1150 	return(0);
1151 }
1152 
1153 static int
1154 em_if_mtu_set(if_ctx_t ctx, uint32_t mtu)
1155 {
1156 	int max_frame_size;
1157 	struct adapter *adapter = iflib_get_softc(ctx);
1158 	if_softc_ctx_t scctx = iflib_get_softc_ctx(ctx);
1159 
1160 	 IOCTL_DEBUGOUT("ioctl rcv'd: SIOCSIFMTU (Set Interface MTU)");
1161 
1162 	switch (adapter->hw.mac.type) {
1163 	case e1000_82571:
1164 	case e1000_82572:
1165 	case e1000_ich9lan:
1166 	case e1000_ich10lan:
1167 	case e1000_pch2lan:
1168 	case e1000_pch_lpt:
1169 	case e1000_pch_spt:
1170 	case e1000_pch_cnp:
1171 	case e1000_82574:
1172 	case e1000_82583:
1173 	case e1000_80003es2lan:
1174 		/* 9K Jumbo Frame size */
1175 		max_frame_size = 9234;
1176 		break;
1177 	case e1000_pchlan:
1178 		max_frame_size = 4096;
1179 		break;
1180 	case e1000_82542:
1181 	case e1000_ich8lan:
1182 		/* Adapters that do not support jumbo frames */
1183 		max_frame_size = ETHER_MAX_LEN;
1184 		break;
1185 	default:
1186 		if (adapter->hw.mac.type >= igb_mac_min)
1187 			max_frame_size = 9234;
1188 		else /* lem */
1189 			max_frame_size = MAX_JUMBO_FRAME_SIZE;
1190 	}
1191 	if (mtu > max_frame_size - ETHER_HDR_LEN - ETHER_CRC_LEN) {
1192 		return (EINVAL);
1193 	}
1194 
1195 	scctx->isc_max_frame_size = adapter->hw.mac.max_frame_size =
1196 	    mtu + ETHER_HDR_LEN + ETHER_CRC_LEN;
1197 	return (0);
1198 }
1199 
1200 /*********************************************************************
1201  *  Init entry point
1202  *
1203  *  This routine is used in two ways. It is used by the stack as
1204  *  init entry point in network interface structure. It is also used
1205  *  by the driver as a hw/sw initialization routine to get to a
1206  *  consistent state.
1207  *
1208  **********************************************************************/
1209 static void
1210 em_if_init(if_ctx_t ctx)
1211 {
1212 	struct adapter *adapter = iflib_get_softc(ctx);
1213 	if_softc_ctx_t scctx = adapter->shared;
1214 	struct ifnet *ifp = iflib_get_ifp(ctx);
1215 	struct em_tx_queue *tx_que;
1216 	int i;
1217 
1218 	INIT_DEBUGOUT("em_if_init: begin");
1219 
1220 	/* Get the latest mac address, User can use a LAA */
1221 	bcopy(if_getlladdr(ifp), adapter->hw.mac.addr,
1222 	    ETHER_ADDR_LEN);
1223 
1224 	/* Put the address into the Receive Address Array */
1225 	e1000_rar_set(&adapter->hw, adapter->hw.mac.addr, 0);
1226 
1227 	/*
1228 	 * With the 82571 adapter, RAR[0] may be overwritten
1229 	 * when the other port is reset, we make a duplicate
1230 	 * in RAR[14] for that eventuality, this assures
1231 	 * the interface continues to function.
1232 	 */
1233 	if (adapter->hw.mac.type == e1000_82571) {
1234 		e1000_set_laa_state_82571(&adapter->hw, TRUE);
1235 		e1000_rar_set(&adapter->hw, adapter->hw.mac.addr,
1236 		    E1000_RAR_ENTRIES - 1);
1237 	}
1238 
1239 
1240 	/* Initialize the hardware */
1241 	em_reset(ctx);
1242 	em_if_update_admin_status(ctx);
1243 
1244 	for (i = 0, tx_que = adapter->tx_queues; i < adapter->tx_num_queues; i++, tx_que++) {
1245 		struct tx_ring *txr = &tx_que->txr;
1246 
1247 		txr->tx_rs_cidx = txr->tx_rs_pidx;
1248 
1249 		/* Initialize the last processed descriptor to be the end of
1250 		 * the ring, rather than the start, so that we avoid an
1251 		 * off-by-one error when calculating how many descriptors are
1252 		 * done in the credits_update function.
1253 		 */
1254 		txr->tx_cidx_processed = scctx->isc_ntxd[0] - 1;
1255 	}
1256 
1257 	/* Setup VLAN support, basic and offload if available */
1258 	E1000_WRITE_REG(&adapter->hw, E1000_VET, ETHERTYPE_VLAN);
1259 
1260 	/* Clear bad data from Rx FIFOs */
1261 	if (adapter->hw.mac.type >= igb_mac_min)
1262 		e1000_rx_fifo_flush_82575(&adapter->hw);
1263 
1264 	/* Configure for OS presence */
1265 	em_init_manageability(adapter);
1266 
1267 	/* Prepare transmit descriptors and buffers */
1268 	em_initialize_transmit_unit(ctx);
1269 
1270 	/* Setup Multicast table */
1271 	em_if_multi_set(ctx);
1272 
1273 	adapter->rx_mbuf_sz = iflib_get_rx_mbuf_sz(ctx);
1274 	em_initialize_receive_unit(ctx);
1275 
1276 	/* Use real VLAN Filter support? */
1277 	if (if_getcapenable(ifp) & IFCAP_VLAN_HWTAGGING) {
1278 		if (if_getcapenable(ifp) & IFCAP_VLAN_HWFILTER)
1279 			/* Use real VLAN Filter support */
1280 			em_setup_vlan_hw_support(adapter);
1281 		else {
1282 			u32 ctrl;
1283 			ctrl = E1000_READ_REG(&adapter->hw, E1000_CTRL);
1284 			ctrl |= E1000_CTRL_VME;
1285 			E1000_WRITE_REG(&adapter->hw, E1000_CTRL, ctrl);
1286 		}
1287 	}
1288 
1289 	/* Don't lose promiscuous settings */
1290 	em_if_set_promisc(ctx, IFF_PROMISC);
1291 	e1000_clear_hw_cntrs_base_generic(&adapter->hw);
1292 
1293 	/* MSI-X configuration for 82574 */
1294 	if (adapter->hw.mac.type == e1000_82574) {
1295 		int tmp = E1000_READ_REG(&adapter->hw, E1000_CTRL_EXT);
1296 
1297 		tmp |= E1000_CTRL_EXT_PBA_CLR;
1298 		E1000_WRITE_REG(&adapter->hw, E1000_CTRL_EXT, tmp);
1299 		/* Set the IVAR - interrupt vector routing. */
1300 		E1000_WRITE_REG(&adapter->hw, E1000_IVAR, adapter->ivars);
1301 	} else if (adapter->intr_type == IFLIB_INTR_MSIX) /* Set up queue routing */
1302 		igb_configure_queues(adapter);
1303 
1304 	/* this clears any pending interrupts */
1305 	E1000_READ_REG(&adapter->hw, E1000_ICR);
1306 	E1000_WRITE_REG(&adapter->hw, E1000_ICS, E1000_ICS_LSC);
1307 
1308 	/* AMT based hardware can now take control from firmware */
1309 	if (adapter->has_manage && adapter->has_amt)
1310 		em_get_hw_control(adapter);
1311 
1312 	/* Set Energy Efficient Ethernet */
1313 	if (adapter->hw.mac.type >= igb_mac_min &&
1314 	    adapter->hw.phy.media_type == e1000_media_type_copper) {
1315 		if (adapter->hw.mac.type == e1000_i354)
1316 			e1000_set_eee_i354(&adapter->hw, TRUE, TRUE);
1317 		else
1318 			e1000_set_eee_i350(&adapter->hw, TRUE, TRUE);
1319 	}
1320 }
1321 
1322 /*********************************************************************
1323  *
1324  *  Fast Legacy/MSI Combined Interrupt Service routine
1325  *
1326  *********************************************************************/
1327 int
1328 em_intr(void *arg)
1329 {
1330 	struct adapter *adapter = arg;
1331 	if_ctx_t ctx = adapter->ctx;
1332 	u32 reg_icr;
1333 
1334 	reg_icr = E1000_READ_REG(&adapter->hw, E1000_ICR);
1335 
1336 	if (adapter->intr_type != IFLIB_INTR_LEGACY)
1337 		goto skip_stray;
1338 	/* Hot eject? */
1339 	if (reg_icr == 0xffffffff)
1340 		return FILTER_STRAY;
1341 
1342 	/* Definitely not our interrupt. */
1343 	if (reg_icr == 0x0)
1344 		return FILTER_STRAY;
1345 
1346 	/*
1347 	 * Starting with the 82571 chip, bit 31 should be used to
1348 	 * determine whether the interrupt belongs to us.
1349 	 */
1350 	if (adapter->hw.mac.type >= e1000_82571 &&
1351 	    (reg_icr & E1000_ICR_INT_ASSERTED) == 0)
1352 		return FILTER_STRAY;
1353 
1354 skip_stray:
1355 	/* Link status change */
1356 	if (reg_icr & (E1000_ICR_RXSEQ | E1000_ICR_LSC)) {
1357 		adapter->hw.mac.get_link_status = 1;
1358 		iflib_admin_intr_deferred(ctx);
1359 	}
1360 
1361 	if (reg_icr & E1000_ICR_RXO)
1362 		adapter->rx_overruns++;
1363 
1364 	return (FILTER_SCHEDULE_THREAD);
1365 }
1366 
1367 static void
1368 igb_rx_enable_queue(struct adapter *adapter, struct em_rx_queue *rxq)
1369 {
1370 	E1000_WRITE_REG(&adapter->hw, E1000_EIMS, rxq->eims);
1371 }
1372 
1373 static void
1374 em_rx_enable_queue(struct adapter *adapter, struct em_rx_queue *rxq)
1375 {
1376 	E1000_WRITE_REG(&adapter->hw, E1000_IMS, rxq->eims);
1377 }
1378 
1379 static void
1380 igb_tx_enable_queue(struct adapter *adapter, struct em_tx_queue *txq)
1381 {
1382 	E1000_WRITE_REG(&adapter->hw, E1000_EIMS, txq->eims);
1383 }
1384 
1385 static void
1386 em_tx_enable_queue(struct adapter *adapter, struct em_tx_queue *txq)
1387 {
1388 	E1000_WRITE_REG(&adapter->hw, E1000_IMS, txq->eims);
1389 }
1390 
1391 static int
1392 em_if_rx_queue_intr_enable(if_ctx_t ctx, uint16_t rxqid)
1393 {
1394 	struct adapter *adapter = iflib_get_softc(ctx);
1395 	struct em_rx_queue *rxq = &adapter->rx_queues[rxqid];
1396 
1397 	if (adapter->hw.mac.type >= igb_mac_min)
1398 		igb_rx_enable_queue(adapter, rxq);
1399 	else
1400 		em_rx_enable_queue(adapter, rxq);
1401 	return (0);
1402 }
1403 
1404 static int
1405 em_if_tx_queue_intr_enable(if_ctx_t ctx, uint16_t txqid)
1406 {
1407 	struct adapter *adapter = iflib_get_softc(ctx);
1408 	struct em_tx_queue *txq = &adapter->tx_queues[txqid];
1409 
1410 	if (adapter->hw.mac.type >= igb_mac_min)
1411 		igb_tx_enable_queue(adapter, txq);
1412 	else
1413 		em_tx_enable_queue(adapter, txq);
1414 	return (0);
1415 }
1416 
1417 /*********************************************************************
1418  *
1419  *  MSI-X RX Interrupt Service routine
1420  *
1421  **********************************************************************/
1422 static int
1423 em_msix_que(void *arg)
1424 {
1425 	struct em_rx_queue *que = arg;
1426 
1427 	++que->irqs;
1428 
1429 	return (FILTER_SCHEDULE_THREAD);
1430 }
1431 
1432 /*********************************************************************
1433  *
1434  *  MSI-X Link Fast Interrupt Service routine
1435  *
1436  **********************************************************************/
1437 static int
1438 em_msix_link(void *arg)
1439 {
1440 	struct adapter *adapter = arg;
1441 	u32 reg_icr;
1442 
1443 	++adapter->link_irq;
1444 	MPASS(adapter->hw.back != NULL);
1445 	reg_icr = E1000_READ_REG(&adapter->hw, E1000_ICR);
1446 
1447 	if (reg_icr & E1000_ICR_RXO)
1448 		adapter->rx_overruns++;
1449 
1450 	if (reg_icr & (E1000_ICR_RXSEQ | E1000_ICR_LSC)) {
1451 		em_handle_link(adapter->ctx);
1452 	} else {
1453 		E1000_WRITE_REG(&adapter->hw, E1000_IMS,
1454 				EM_MSIX_LINK | E1000_IMS_LSC);
1455 		if (adapter->hw.mac.type >= igb_mac_min)
1456 			E1000_WRITE_REG(&adapter->hw, E1000_EIMS, adapter->link_mask);
1457 	}
1458 
1459 	/*
1460 	 * Because we must read the ICR for this interrupt
1461 	 * it may clear other causes using autoclear, for
1462 	 * this reason we simply create a soft interrupt
1463 	 * for all these vectors.
1464 	 */
1465 	if (reg_icr && adapter->hw.mac.type < igb_mac_min) {
1466 		E1000_WRITE_REG(&adapter->hw,
1467 			E1000_ICS, adapter->ims);
1468 	}
1469 
1470 	return (FILTER_HANDLED);
1471 }
1472 
1473 static void
1474 em_handle_link(void *context)
1475 {
1476 	if_ctx_t ctx = context;
1477 	struct adapter *adapter = iflib_get_softc(ctx);
1478 
1479 	adapter->hw.mac.get_link_status = 1;
1480 	iflib_admin_intr_deferred(ctx);
1481 }
1482 
1483 
1484 /*********************************************************************
1485  *
1486  *  Media Ioctl callback
1487  *
1488  *  This routine is called whenever the user queries the status of
1489  *  the interface using ifconfig.
1490  *
1491  **********************************************************************/
1492 static void
1493 em_if_media_status(if_ctx_t ctx, struct ifmediareq *ifmr)
1494 {
1495 	struct adapter *adapter = iflib_get_softc(ctx);
1496 	u_char fiber_type = IFM_1000_SX;
1497 
1498 	INIT_DEBUGOUT("em_if_media_status: begin");
1499 
1500 	iflib_admin_intr_deferred(ctx);
1501 
1502 	ifmr->ifm_status = IFM_AVALID;
1503 	ifmr->ifm_active = IFM_ETHER;
1504 
1505 	if (!adapter->link_active) {
1506 		return;
1507 	}
1508 
1509 	ifmr->ifm_status |= IFM_ACTIVE;
1510 
1511 	if ((adapter->hw.phy.media_type == e1000_media_type_fiber) ||
1512 	    (adapter->hw.phy.media_type == e1000_media_type_internal_serdes)) {
1513 		if (adapter->hw.mac.type == e1000_82545)
1514 			fiber_type = IFM_1000_LX;
1515 		ifmr->ifm_active |= fiber_type | IFM_FDX;
1516 	} else {
1517 		switch (adapter->link_speed) {
1518 		case 10:
1519 			ifmr->ifm_active |= IFM_10_T;
1520 			break;
1521 		case 100:
1522 			ifmr->ifm_active |= IFM_100_TX;
1523 			break;
1524 		case 1000:
1525 			ifmr->ifm_active |= IFM_1000_T;
1526 			break;
1527 		}
1528 		if (adapter->link_duplex == FULL_DUPLEX)
1529 			ifmr->ifm_active |= IFM_FDX;
1530 		else
1531 			ifmr->ifm_active |= IFM_HDX;
1532 	}
1533 }
1534 
1535 /*********************************************************************
1536  *
1537  *  Media Ioctl callback
1538  *
1539  *  This routine is called when the user changes speed/duplex using
1540  *  media/mediopt option with ifconfig.
1541  *
1542  **********************************************************************/
1543 static int
1544 em_if_media_change(if_ctx_t ctx)
1545 {
1546 	struct adapter *adapter = iflib_get_softc(ctx);
1547 	struct ifmedia *ifm = iflib_get_media(ctx);
1548 
1549 	INIT_DEBUGOUT("em_if_media_change: begin");
1550 
1551 	if (IFM_TYPE(ifm->ifm_media) != IFM_ETHER)
1552 		return (EINVAL);
1553 
1554 	switch (IFM_SUBTYPE(ifm->ifm_media)) {
1555 	case IFM_AUTO:
1556 		adapter->hw.mac.autoneg = DO_AUTO_NEG;
1557 		adapter->hw.phy.autoneg_advertised = AUTONEG_ADV_DEFAULT;
1558 		break;
1559 	case IFM_1000_LX:
1560 	case IFM_1000_SX:
1561 	case IFM_1000_T:
1562 		adapter->hw.mac.autoneg = DO_AUTO_NEG;
1563 		adapter->hw.phy.autoneg_advertised = ADVERTISE_1000_FULL;
1564 		break;
1565 	case IFM_100_TX:
1566 		adapter->hw.mac.autoneg = FALSE;
1567 		adapter->hw.phy.autoneg_advertised = 0;
1568 		if ((ifm->ifm_media & IFM_GMASK) == IFM_FDX)
1569 			adapter->hw.mac.forced_speed_duplex = ADVERTISE_100_FULL;
1570 		else
1571 			adapter->hw.mac.forced_speed_duplex = ADVERTISE_100_HALF;
1572 		break;
1573 	case IFM_10_T:
1574 		adapter->hw.mac.autoneg = FALSE;
1575 		adapter->hw.phy.autoneg_advertised = 0;
1576 		if ((ifm->ifm_media & IFM_GMASK) == IFM_FDX)
1577 			adapter->hw.mac.forced_speed_duplex = ADVERTISE_10_FULL;
1578 		else
1579 			adapter->hw.mac.forced_speed_duplex = ADVERTISE_10_HALF;
1580 		break;
1581 	default:
1582 		device_printf(adapter->dev, "Unsupported media type\n");
1583 	}
1584 
1585 	em_if_init(ctx);
1586 
1587 	return (0);
1588 }
1589 
1590 static int
1591 em_if_set_promisc(if_ctx_t ctx, int flags)
1592 {
1593 	struct adapter *adapter = iflib_get_softc(ctx);
1594 	u32 reg_rctl;
1595 
1596 	em_disable_promisc(ctx);
1597 
1598 	reg_rctl = E1000_READ_REG(&adapter->hw, E1000_RCTL);
1599 
1600 	if (flags & IFF_PROMISC) {
1601 		reg_rctl |= (E1000_RCTL_UPE | E1000_RCTL_MPE);
1602 		/* Turn this on if you want to see bad packets */
1603 		if (em_debug_sbp)
1604 			reg_rctl |= E1000_RCTL_SBP;
1605 		E1000_WRITE_REG(&adapter->hw, E1000_RCTL, reg_rctl);
1606 	} else if (flags & IFF_ALLMULTI) {
1607 		reg_rctl |= E1000_RCTL_MPE;
1608 		reg_rctl &= ~E1000_RCTL_UPE;
1609 		E1000_WRITE_REG(&adapter->hw, E1000_RCTL, reg_rctl);
1610 	}
1611 	return (0);
1612 }
1613 
1614 static void
1615 em_disable_promisc(if_ctx_t ctx)
1616 {
1617 	struct adapter *adapter = iflib_get_softc(ctx);
1618 	struct ifnet *ifp = iflib_get_ifp(ctx);
1619 	u32 reg_rctl;
1620 	int mcnt = 0;
1621 
1622 	reg_rctl = E1000_READ_REG(&adapter->hw, E1000_RCTL);
1623 	reg_rctl &= (~E1000_RCTL_UPE);
1624 	if (if_getflags(ifp) & IFF_ALLMULTI)
1625 		mcnt = MAX_NUM_MULTICAST_ADDRESSES;
1626 	else
1627 		mcnt = if_multiaddr_count(ifp, MAX_NUM_MULTICAST_ADDRESSES);
1628 	/* Don't disable if in MAX groups */
1629 	if (mcnt < MAX_NUM_MULTICAST_ADDRESSES)
1630 		reg_rctl &=  (~E1000_RCTL_MPE);
1631 	reg_rctl &=  (~E1000_RCTL_SBP);
1632 	E1000_WRITE_REG(&adapter->hw, E1000_RCTL, reg_rctl);
1633 }
1634 
1635 
1636 /*********************************************************************
1637  *  Multicast Update
1638  *
1639  *  This routine is called whenever multicast address list is updated.
1640  *
1641  **********************************************************************/
1642 
1643 static void
1644 em_if_multi_set(if_ctx_t ctx)
1645 {
1646 	struct adapter *adapter = iflib_get_softc(ctx);
1647 	struct ifnet *ifp = iflib_get_ifp(ctx);
1648 	u32 reg_rctl = 0;
1649 	u8  *mta; /* Multicast array memory */
1650 	int mcnt = 0;
1651 
1652 	IOCTL_DEBUGOUT("em_set_multi: begin");
1653 
1654 	mta = adapter->mta;
1655 	bzero(mta, sizeof(u8) * ETH_ADDR_LEN * MAX_NUM_MULTICAST_ADDRESSES);
1656 
1657 	if (adapter->hw.mac.type == e1000_82542 &&
1658 	    adapter->hw.revision_id == E1000_REVISION_2) {
1659 		reg_rctl = E1000_READ_REG(&adapter->hw, E1000_RCTL);
1660 		if (adapter->hw.bus.pci_cmd_word & CMD_MEM_WRT_INVALIDATE)
1661 			e1000_pci_clear_mwi(&adapter->hw);
1662 		reg_rctl |= E1000_RCTL_RST;
1663 		E1000_WRITE_REG(&adapter->hw, E1000_RCTL, reg_rctl);
1664 		msec_delay(5);
1665 	}
1666 
1667 	if_multiaddr_array(ifp, mta, &mcnt, MAX_NUM_MULTICAST_ADDRESSES);
1668 
1669 	if (mcnt >= MAX_NUM_MULTICAST_ADDRESSES) {
1670 		reg_rctl = E1000_READ_REG(&adapter->hw, E1000_RCTL);
1671 		reg_rctl |= E1000_RCTL_MPE;
1672 		E1000_WRITE_REG(&adapter->hw, E1000_RCTL, reg_rctl);
1673 	} else
1674 		e1000_update_mc_addr_list(&adapter->hw, mta, mcnt);
1675 
1676 	if (adapter->hw.mac.type == e1000_82542 &&
1677 	    adapter->hw.revision_id == E1000_REVISION_2) {
1678 		reg_rctl = E1000_READ_REG(&adapter->hw, E1000_RCTL);
1679 		reg_rctl &= ~E1000_RCTL_RST;
1680 		E1000_WRITE_REG(&adapter->hw, E1000_RCTL, reg_rctl);
1681 		msec_delay(5);
1682 		if (adapter->hw.bus.pci_cmd_word & CMD_MEM_WRT_INVALIDATE)
1683 			e1000_pci_set_mwi(&adapter->hw);
1684 	}
1685 }
1686 
1687 /*********************************************************************
1688  *  Timer routine
1689  *
1690  *  This routine schedules em_if_update_admin_status() to check for
1691  *  link status and to gather statistics as well as to perform some
1692  *  controller-specific hardware patting.
1693  *
1694  **********************************************************************/
1695 static void
1696 em_if_timer(if_ctx_t ctx, uint16_t qid)
1697 {
1698 
1699 	if (qid != 0)
1700 		return;
1701 
1702 	iflib_admin_intr_deferred(ctx);
1703 }
1704 
1705 static void
1706 em_if_update_admin_status(if_ctx_t ctx)
1707 {
1708 	struct adapter *adapter = iflib_get_softc(ctx);
1709 	struct e1000_hw *hw = &adapter->hw;
1710 	device_t dev = iflib_get_dev(ctx);
1711 	u32 link_check, thstat, ctrl;
1712 
1713 	link_check = thstat = ctrl = 0;
1714 	/* Get the cached link value or read phy for real */
1715 	switch (hw->phy.media_type) {
1716 	case e1000_media_type_copper:
1717 		if (hw->mac.get_link_status) {
1718 			if (hw->mac.type == e1000_pch_spt)
1719 				msec_delay(50);
1720 			/* Do the work to read phy */
1721 			e1000_check_for_link(hw);
1722 			link_check = !hw->mac.get_link_status;
1723 			if (link_check) /* ESB2 fix */
1724 				e1000_cfg_on_link_up(hw);
1725 		} else {
1726 			link_check = TRUE;
1727 		}
1728 		break;
1729 	case e1000_media_type_fiber:
1730 		e1000_check_for_link(hw);
1731 		link_check = (E1000_READ_REG(hw, E1000_STATUS) &
1732 			    E1000_STATUS_LU);
1733 		break;
1734 	case e1000_media_type_internal_serdes:
1735 		e1000_check_for_link(hw);
1736 		link_check = adapter->hw.mac.serdes_has_link;
1737 		break;
1738 	/* VF device is type_unknown */
1739 	case e1000_media_type_unknown:
1740 		e1000_check_for_link(hw);
1741 		link_check = !hw->mac.get_link_status;
1742 		/* FALLTHROUGH */
1743 	default:
1744 		break;
1745 	}
1746 
1747 	/* Check for thermal downshift or shutdown */
1748 	if (hw->mac.type == e1000_i350) {
1749 		thstat = E1000_READ_REG(hw, E1000_THSTAT);
1750 		ctrl = E1000_READ_REG(hw, E1000_CTRL_EXT);
1751 	}
1752 
1753 	/* Now check for a transition */
1754 	if (link_check && (adapter->link_active == 0)) {
1755 		e1000_get_speed_and_duplex(hw, &adapter->link_speed,
1756 		    &adapter->link_duplex);
1757 		/* Check if we must disable SPEED_MODE bit on PCI-E */
1758 		if ((adapter->link_speed != SPEED_1000) &&
1759 		    ((hw->mac.type == e1000_82571) ||
1760 		    (hw->mac.type == e1000_82572))) {
1761 			int tarc0;
1762 			tarc0 = E1000_READ_REG(hw, E1000_TARC(0));
1763 			tarc0 &= ~TARC_SPEED_MODE_BIT;
1764 			E1000_WRITE_REG(hw, E1000_TARC(0), tarc0);
1765 		}
1766 		if (bootverbose)
1767 			device_printf(dev, "Link is up %d Mbps %s\n",
1768 			    adapter->link_speed,
1769 			    ((adapter->link_duplex == FULL_DUPLEX) ?
1770 			    "Full Duplex" : "Half Duplex"));
1771 		adapter->link_active = 1;
1772 		adapter->smartspeed = 0;
1773 		if ((ctrl & E1000_CTRL_EXT_LINK_MODE_MASK) ==
1774 		    E1000_CTRL_EXT_LINK_MODE_GMII &&
1775 		    (thstat & E1000_THSTAT_LINK_THROTTLE))
1776 			device_printf(dev, "Link: thermal downshift\n");
1777 		/* Delay Link Up for Phy update */
1778 		if (((hw->mac.type == e1000_i210) ||
1779 		    (hw->mac.type == e1000_i211)) &&
1780 		    (hw->phy.id == I210_I_PHY_ID))
1781 			msec_delay(I210_LINK_DELAY);
1782 		/* Reset if the media type changed. */
1783 		if ((hw->dev_spec._82575.media_changed) &&
1784 			(adapter->hw.mac.type >= igb_mac_min)) {
1785 			hw->dev_spec._82575.media_changed = false;
1786 			adapter->flags |= IGB_MEDIA_RESET;
1787 			em_reset(ctx);
1788 		}
1789 		iflib_link_state_change(ctx, LINK_STATE_UP,
1790 		    IF_Mbps(adapter->link_speed));
1791 	} else if (!link_check && (adapter->link_active == 1)) {
1792 		adapter->link_speed = 0;
1793 		adapter->link_duplex = 0;
1794 		adapter->link_active = 0;
1795 		iflib_link_state_change(ctx, LINK_STATE_DOWN, 0);
1796 	}
1797 	em_update_stats_counters(adapter);
1798 
1799 	/* Reset LAA into RAR[0] on 82571 */
1800 	if ((adapter->hw.mac.type == e1000_82571) &&
1801 	    e1000_get_laa_state_82571(&adapter->hw))
1802 		e1000_rar_set(&adapter->hw, adapter->hw.mac.addr, 0);
1803 
1804 	if (adapter->hw.mac.type < em_mac_min)
1805 		lem_smartspeed(adapter);
1806 
1807 	E1000_WRITE_REG(&adapter->hw, E1000_IMS, EM_MSIX_LINK | E1000_IMS_LSC);
1808 }
1809 
1810 static void
1811 em_if_watchdog_reset(if_ctx_t ctx)
1812 {
1813 	struct adapter *adapter = iflib_get_softc(ctx);
1814 
1815 	/*
1816 	 * Just count the event; iflib(4) will already trigger a
1817 	 * sufficient reset of the controller.
1818 	 */
1819 	adapter->watchdog_events++;
1820 }
1821 
1822 /*********************************************************************
1823  *
1824  *  This routine disables all traffic on the adapter by issuing a
1825  *  global reset on the MAC.
1826  *
1827  **********************************************************************/
1828 static void
1829 em_if_stop(if_ctx_t ctx)
1830 {
1831 	struct adapter *adapter = iflib_get_softc(ctx);
1832 
1833 	INIT_DEBUGOUT("em_if_stop: begin");
1834 
1835 	e1000_reset_hw(&adapter->hw);
1836 	if (adapter->hw.mac.type >= e1000_82544)
1837 		E1000_WRITE_REG(&adapter->hw, E1000_WUFC, 0);
1838 
1839 	e1000_led_off(&adapter->hw);
1840 	e1000_cleanup_led(&adapter->hw);
1841 }
1842 
1843 /*********************************************************************
1844  *
1845  *  Determine hardware revision.
1846  *
1847  **********************************************************************/
1848 static void
1849 em_identify_hardware(if_ctx_t ctx)
1850 {
1851 	device_t dev = iflib_get_dev(ctx);
1852 	struct adapter *adapter = iflib_get_softc(ctx);
1853 
1854 	/* Make sure our PCI config space has the necessary stuff set */
1855 	adapter->hw.bus.pci_cmd_word = pci_read_config(dev, PCIR_COMMAND, 2);
1856 
1857 	/* Save off the information about this board */
1858 	adapter->hw.vendor_id = pci_get_vendor(dev);
1859 	adapter->hw.device_id = pci_get_device(dev);
1860 	adapter->hw.revision_id = pci_read_config(dev, PCIR_REVID, 1);
1861 	adapter->hw.subsystem_vendor_id =
1862 	    pci_read_config(dev, PCIR_SUBVEND_0, 2);
1863 	adapter->hw.subsystem_device_id =
1864 	    pci_read_config(dev, PCIR_SUBDEV_0, 2);
1865 
1866 	/* Do Shared Code Init and Setup */
1867 	if (e1000_set_mac_type(&adapter->hw)) {
1868 		device_printf(dev, "Setup init failure\n");
1869 		return;
1870 	}
1871 }
1872 
1873 static int
1874 em_allocate_pci_resources(if_ctx_t ctx)
1875 {
1876 	struct adapter *adapter = iflib_get_softc(ctx);
1877 	device_t dev = iflib_get_dev(ctx);
1878 	int rid, val;
1879 
1880 	rid = PCIR_BAR(0);
1881 	adapter->memory = bus_alloc_resource_any(dev, SYS_RES_MEMORY,
1882 	    &rid, RF_ACTIVE);
1883 	if (adapter->memory == NULL) {
1884 		device_printf(dev, "Unable to allocate bus resource: memory\n");
1885 		return (ENXIO);
1886 	}
1887 	adapter->osdep.mem_bus_space_tag = rman_get_bustag(adapter->memory);
1888 	adapter->osdep.mem_bus_space_handle =
1889 	    rman_get_bushandle(adapter->memory);
1890 	adapter->hw.hw_addr = (u8 *)&adapter->osdep.mem_bus_space_handle;
1891 
1892 	/* Only older adapters use IO mapping */
1893 	if (adapter->hw.mac.type < em_mac_min &&
1894 	    adapter->hw.mac.type > e1000_82543) {
1895 		/* Figure our where our IO BAR is ? */
1896 		for (rid = PCIR_BAR(0); rid < PCIR_CIS;) {
1897 			val = pci_read_config(dev, rid, 4);
1898 			if (EM_BAR_TYPE(val) == EM_BAR_TYPE_IO) {
1899 				break;
1900 			}
1901 			rid += 4;
1902 			/* check for 64bit BAR */
1903 			if (EM_BAR_MEM_TYPE(val) == EM_BAR_MEM_TYPE_64BIT)
1904 				rid += 4;
1905 		}
1906 		if (rid >= PCIR_CIS) {
1907 			device_printf(dev, "Unable to locate IO BAR\n");
1908 			return (ENXIO);
1909 		}
1910 		adapter->ioport = bus_alloc_resource_any(dev, SYS_RES_IOPORT,
1911 		    &rid, RF_ACTIVE);
1912 		if (adapter->ioport == NULL) {
1913 			device_printf(dev, "Unable to allocate bus resource: "
1914 			    "ioport\n");
1915 			return (ENXIO);
1916 		}
1917 		adapter->hw.io_base = 0;
1918 		adapter->osdep.io_bus_space_tag =
1919 		    rman_get_bustag(adapter->ioport);
1920 		adapter->osdep.io_bus_space_handle =
1921 		    rman_get_bushandle(adapter->ioport);
1922 	}
1923 
1924 	adapter->hw.back = &adapter->osdep;
1925 
1926 	return (0);
1927 }
1928 
1929 /*********************************************************************
1930  *
1931  *  Set up the MSI-X Interrupt handlers
1932  *
1933  **********************************************************************/
1934 static int
1935 em_if_msix_intr_assign(if_ctx_t ctx, int msix)
1936 {
1937 	struct adapter *adapter = iflib_get_softc(ctx);
1938 	struct em_rx_queue *rx_que = adapter->rx_queues;
1939 	struct em_tx_queue *tx_que = adapter->tx_queues;
1940 	int error, rid, i, vector = 0, rx_vectors;
1941 	char buf[16];
1942 
1943 	/* First set up ring resources */
1944 	for (i = 0; i < adapter->rx_num_queues; i++, rx_que++, vector++) {
1945 		rid = vector + 1;
1946 		snprintf(buf, sizeof(buf), "rxq%d", i);
1947 		error = iflib_irq_alloc_generic(ctx, &rx_que->que_irq, rid, IFLIB_INTR_RXTX, em_msix_que, rx_que, rx_que->me, buf);
1948 		if (error) {
1949 			device_printf(iflib_get_dev(ctx), "Failed to allocate que int %d err: %d", i, error);
1950 			adapter->rx_num_queues = i + 1;
1951 			goto fail;
1952 		}
1953 
1954 		rx_que->msix =  vector;
1955 
1956 		/*
1957 		 * Set the bit to enable interrupt
1958 		 * in E1000_IMS -- bits 20 and 21
1959 		 * are for RX0 and RX1, note this has
1960 		 * NOTHING to do with the MSI-X vector
1961 		 */
1962 		if (adapter->hw.mac.type == e1000_82574) {
1963 			rx_que->eims = 1 << (20 + i);
1964 			adapter->ims |= rx_que->eims;
1965 			adapter->ivars |= (8 | rx_que->msix) << (i * 4);
1966 		} else if (adapter->hw.mac.type == e1000_82575)
1967 			rx_que->eims = E1000_EICR_TX_QUEUE0 << vector;
1968 		else
1969 			rx_que->eims = 1 << vector;
1970 	}
1971 	rx_vectors = vector;
1972 
1973 	vector = 0;
1974 	for (i = 0; i < adapter->tx_num_queues; i++, tx_que++, vector++) {
1975 		snprintf(buf, sizeof(buf), "txq%d", i);
1976 		tx_que = &adapter->tx_queues[i];
1977 		iflib_softirq_alloc_generic(ctx,
1978 		    &adapter->rx_queues[i % adapter->rx_num_queues].que_irq,
1979 		    IFLIB_INTR_TX, tx_que, tx_que->me, buf);
1980 
1981 		tx_que->msix = (vector % adapter->rx_num_queues);
1982 
1983 		/*
1984 		 * Set the bit to enable interrupt
1985 		 * in E1000_IMS -- bits 22 and 23
1986 		 * are for TX0 and TX1, note this has
1987 		 * NOTHING to do with the MSI-X vector
1988 		 */
1989 		if (adapter->hw.mac.type == e1000_82574) {
1990 			tx_que->eims = 1 << (22 + i);
1991 			adapter->ims |= tx_que->eims;
1992 			adapter->ivars |= (8 | tx_que->msix) << (8 + (i * 4));
1993 		} else if (adapter->hw.mac.type == e1000_82575) {
1994 			tx_que->eims = E1000_EICR_TX_QUEUE0 << i;
1995 		} else {
1996 			tx_que->eims = 1 << i;
1997 		}
1998 	}
1999 
2000 	/* Link interrupt */
2001 	rid = rx_vectors + 1;
2002 	error = iflib_irq_alloc_generic(ctx, &adapter->irq, rid, IFLIB_INTR_ADMIN, em_msix_link, adapter, 0, "aq");
2003 
2004 	if (error) {
2005 		device_printf(iflib_get_dev(ctx), "Failed to register admin handler");
2006 		goto fail;
2007 	}
2008 	adapter->linkvec = rx_vectors;
2009 	if (adapter->hw.mac.type < igb_mac_min) {
2010 		adapter->ivars |=  (8 | rx_vectors) << 16;
2011 		adapter->ivars |= 0x80000000;
2012 	}
2013 	return (0);
2014 fail:
2015 	iflib_irq_free(ctx, &adapter->irq);
2016 	rx_que = adapter->rx_queues;
2017 	for (int i = 0; i < adapter->rx_num_queues; i++, rx_que++)
2018 		iflib_irq_free(ctx, &rx_que->que_irq);
2019 	return (error);
2020 }
2021 
2022 static void
2023 igb_configure_queues(struct adapter *adapter)
2024 {
2025 	struct e1000_hw *hw = &adapter->hw;
2026 	struct em_rx_queue *rx_que;
2027 	struct em_tx_queue *tx_que;
2028 	u32 tmp, ivar = 0, newitr = 0;
2029 
2030 	/* First turn on RSS capability */
2031 	if (adapter->hw.mac.type != e1000_82575)
2032 		E1000_WRITE_REG(hw, E1000_GPIE,
2033 		    E1000_GPIE_MSIX_MODE | E1000_GPIE_EIAME |
2034 		    E1000_GPIE_PBA | E1000_GPIE_NSICR);
2035 
2036 	/* Turn on MSI-X */
2037 	switch (adapter->hw.mac.type) {
2038 	case e1000_82580:
2039 	case e1000_i350:
2040 	case e1000_i354:
2041 	case e1000_i210:
2042 	case e1000_i211:
2043 	case e1000_vfadapt:
2044 	case e1000_vfadapt_i350:
2045 		/* RX entries */
2046 		for (int i = 0; i < adapter->rx_num_queues; i++) {
2047 			u32 index = i >> 1;
2048 			ivar = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index);
2049 			rx_que = &adapter->rx_queues[i];
2050 			if (i & 1) {
2051 				ivar &= 0xFF00FFFF;
2052 				ivar |= (rx_que->msix | E1000_IVAR_VALID) << 16;
2053 			} else {
2054 				ivar &= 0xFFFFFF00;
2055 				ivar |= rx_que->msix | E1000_IVAR_VALID;
2056 			}
2057 			E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, ivar);
2058 		}
2059 		/* TX entries */
2060 		for (int i = 0; i < adapter->tx_num_queues; i++) {
2061 			u32 index = i >> 1;
2062 			ivar = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index);
2063 			tx_que = &adapter->tx_queues[i];
2064 			if (i & 1) {
2065 				ivar &= 0x00FFFFFF;
2066 				ivar |= (tx_que->msix | E1000_IVAR_VALID) << 24;
2067 			} else {
2068 				ivar &= 0xFFFF00FF;
2069 				ivar |= (tx_que->msix | E1000_IVAR_VALID) << 8;
2070 			}
2071 			E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, ivar);
2072 			adapter->que_mask |= tx_que->eims;
2073 		}
2074 
2075 		/* And for the link interrupt */
2076 		ivar = (adapter->linkvec | E1000_IVAR_VALID) << 8;
2077 		adapter->link_mask = 1 << adapter->linkvec;
2078 		E1000_WRITE_REG(hw, E1000_IVAR_MISC, ivar);
2079 		break;
2080 	case e1000_82576:
2081 		/* RX entries */
2082 		for (int i = 0; i < adapter->rx_num_queues; i++) {
2083 			u32 index = i & 0x7; /* Each IVAR has two entries */
2084 			ivar = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index);
2085 			rx_que = &adapter->rx_queues[i];
2086 			if (i < 8) {
2087 				ivar &= 0xFFFFFF00;
2088 				ivar |= rx_que->msix | E1000_IVAR_VALID;
2089 			} else {
2090 				ivar &= 0xFF00FFFF;
2091 				ivar |= (rx_que->msix | E1000_IVAR_VALID) << 16;
2092 			}
2093 			E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, ivar);
2094 			adapter->que_mask |= rx_que->eims;
2095 		}
2096 		/* TX entries */
2097 		for (int i = 0; i < adapter->tx_num_queues; i++) {
2098 			u32 index = i & 0x7; /* Each IVAR has two entries */
2099 			ivar = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index);
2100 			tx_que = &adapter->tx_queues[i];
2101 			if (i < 8) {
2102 				ivar &= 0xFFFF00FF;
2103 				ivar |= (tx_que->msix | E1000_IVAR_VALID) << 8;
2104 			} else {
2105 				ivar &= 0x00FFFFFF;
2106 				ivar |= (tx_que->msix | E1000_IVAR_VALID) << 24;
2107 			}
2108 			E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, ivar);
2109 			adapter->que_mask |= tx_que->eims;
2110 		}
2111 
2112 		/* And for the link interrupt */
2113 		ivar = (adapter->linkvec | E1000_IVAR_VALID) << 8;
2114 		adapter->link_mask = 1 << adapter->linkvec;
2115 		E1000_WRITE_REG(hw, E1000_IVAR_MISC, ivar);
2116 		break;
2117 
2118 	case e1000_82575:
2119 		/* enable MSI-X support*/
2120 		tmp = E1000_READ_REG(hw, E1000_CTRL_EXT);
2121 		tmp |= E1000_CTRL_EXT_PBA_CLR;
2122 		/* Auto-Mask interrupts upon ICR read. */
2123 		tmp |= E1000_CTRL_EXT_EIAME;
2124 		tmp |= E1000_CTRL_EXT_IRCA;
2125 		E1000_WRITE_REG(hw, E1000_CTRL_EXT, tmp);
2126 
2127 		/* Queues */
2128 		for (int i = 0; i < adapter->rx_num_queues; i++) {
2129 			rx_que = &adapter->rx_queues[i];
2130 			tmp = E1000_EICR_RX_QUEUE0 << i;
2131 			tmp |= E1000_EICR_TX_QUEUE0 << i;
2132 			rx_que->eims = tmp;
2133 			E1000_WRITE_REG_ARRAY(hw, E1000_MSIXBM(0),
2134 			    i, rx_que->eims);
2135 			adapter->que_mask |= rx_que->eims;
2136 		}
2137 
2138 		/* Link */
2139 		E1000_WRITE_REG(hw, E1000_MSIXBM(adapter->linkvec),
2140 		    E1000_EIMS_OTHER);
2141 		adapter->link_mask |= E1000_EIMS_OTHER;
2142 	default:
2143 		break;
2144 	}
2145 
2146 	/* Set the starting interrupt rate */
2147 	if (em_max_interrupt_rate > 0)
2148 		newitr = (4000000 / em_max_interrupt_rate) & 0x7FFC;
2149 
2150 	if (hw->mac.type == e1000_82575)
2151 		newitr |= newitr << 16;
2152 	else
2153 		newitr |= E1000_EITR_CNT_IGNR;
2154 
2155 	for (int i = 0; i < adapter->rx_num_queues; i++) {
2156 		rx_que = &adapter->rx_queues[i];
2157 		E1000_WRITE_REG(hw, E1000_EITR(rx_que->msix), newitr);
2158 	}
2159 
2160 	return;
2161 }
2162 
2163 static void
2164 em_free_pci_resources(if_ctx_t ctx)
2165 {
2166 	struct adapter *adapter = iflib_get_softc(ctx);
2167 	struct em_rx_queue *que = adapter->rx_queues;
2168 	device_t dev = iflib_get_dev(ctx);
2169 
2170 	/* Release all MSI-X queue resources */
2171 	if (adapter->intr_type == IFLIB_INTR_MSIX)
2172 		iflib_irq_free(ctx, &adapter->irq);
2173 
2174 	for (int i = 0; i < adapter->rx_num_queues; i++, que++) {
2175 		iflib_irq_free(ctx, &que->que_irq);
2176 	}
2177 
2178 	if (adapter->memory != NULL) {
2179 		bus_release_resource(dev, SYS_RES_MEMORY,
2180 		    rman_get_rid(adapter->memory), adapter->memory);
2181 		adapter->memory = NULL;
2182 	}
2183 
2184 	if (adapter->flash != NULL) {
2185 		bus_release_resource(dev, SYS_RES_MEMORY,
2186 		    rman_get_rid(adapter->flash), adapter->flash);
2187 		adapter->flash = NULL;
2188 	}
2189 
2190 	if (adapter->ioport != NULL) {
2191 		bus_release_resource(dev, SYS_RES_IOPORT,
2192 		    rman_get_rid(adapter->ioport), adapter->ioport);
2193 		adapter->ioport = NULL;
2194 	}
2195 }
2196 
2197 /* Set up MSI or MSI-X */
2198 static int
2199 em_setup_msix(if_ctx_t ctx)
2200 {
2201 	struct adapter *adapter = iflib_get_softc(ctx);
2202 
2203 	if (adapter->hw.mac.type == e1000_82574) {
2204 		em_enable_vectors_82574(ctx);
2205 	}
2206 	return (0);
2207 }
2208 
2209 /*********************************************************************
2210  *
2211  *  Workaround for SmartSpeed on 82541 and 82547 controllers
2212  *
2213  **********************************************************************/
2214 static void
2215 lem_smartspeed(struct adapter *adapter)
2216 {
2217 	u16 phy_tmp;
2218 
2219 	if (adapter->link_active || (adapter->hw.phy.type != e1000_phy_igp) ||
2220 	    adapter->hw.mac.autoneg == 0 ||
2221 	    (adapter->hw.phy.autoneg_advertised & ADVERTISE_1000_FULL) == 0)
2222 		return;
2223 
2224 	if (adapter->smartspeed == 0) {
2225 		/* If Master/Slave config fault is asserted twice,
2226 		 * we assume back-to-back */
2227 		e1000_read_phy_reg(&adapter->hw, PHY_1000T_STATUS, &phy_tmp);
2228 		if (!(phy_tmp & SR_1000T_MS_CONFIG_FAULT))
2229 			return;
2230 		e1000_read_phy_reg(&adapter->hw, PHY_1000T_STATUS, &phy_tmp);
2231 		if (phy_tmp & SR_1000T_MS_CONFIG_FAULT) {
2232 			e1000_read_phy_reg(&adapter->hw,
2233 			    PHY_1000T_CTRL, &phy_tmp);
2234 			if(phy_tmp & CR_1000T_MS_ENABLE) {
2235 				phy_tmp &= ~CR_1000T_MS_ENABLE;
2236 				e1000_write_phy_reg(&adapter->hw,
2237 				    PHY_1000T_CTRL, phy_tmp);
2238 				adapter->smartspeed++;
2239 				if(adapter->hw.mac.autoneg &&
2240 				   !e1000_copper_link_autoneg(&adapter->hw) &&
2241 				   !e1000_read_phy_reg(&adapter->hw,
2242 				    PHY_CONTROL, &phy_tmp)) {
2243 					phy_tmp |= (MII_CR_AUTO_NEG_EN |
2244 						    MII_CR_RESTART_AUTO_NEG);
2245 					e1000_write_phy_reg(&adapter->hw,
2246 					    PHY_CONTROL, phy_tmp);
2247 				}
2248 			}
2249 		}
2250 		return;
2251 	} else if(adapter->smartspeed == EM_SMARTSPEED_DOWNSHIFT) {
2252 		/* If still no link, perhaps using 2/3 pair cable */
2253 		e1000_read_phy_reg(&adapter->hw, PHY_1000T_CTRL, &phy_tmp);
2254 		phy_tmp |= CR_1000T_MS_ENABLE;
2255 		e1000_write_phy_reg(&adapter->hw, PHY_1000T_CTRL, phy_tmp);
2256 		if(adapter->hw.mac.autoneg &&
2257 		   !e1000_copper_link_autoneg(&adapter->hw) &&
2258 		   !e1000_read_phy_reg(&adapter->hw, PHY_CONTROL, &phy_tmp)) {
2259 			phy_tmp |= (MII_CR_AUTO_NEG_EN |
2260 				    MII_CR_RESTART_AUTO_NEG);
2261 			e1000_write_phy_reg(&adapter->hw, PHY_CONTROL, phy_tmp);
2262 		}
2263 	}
2264 	/* Restart process after EM_SMARTSPEED_MAX iterations */
2265 	if(adapter->smartspeed++ == EM_SMARTSPEED_MAX)
2266 		adapter->smartspeed = 0;
2267 }
2268 
2269 /*********************************************************************
2270  *
2271  *  Initialize the DMA Coalescing feature
2272  *
2273  **********************************************************************/
2274 static void
2275 igb_init_dmac(struct adapter *adapter, u32 pba)
2276 {
2277 	device_t	dev = adapter->dev;
2278 	struct e1000_hw *hw = &adapter->hw;
2279 	u32 		dmac, reg = ~E1000_DMACR_DMAC_EN;
2280 	u16		hwm;
2281 	u16		max_frame_size;
2282 
2283 	if (hw->mac.type == e1000_i211)
2284 		return;
2285 
2286 	max_frame_size = adapter->shared->isc_max_frame_size;
2287 	if (hw->mac.type > e1000_82580) {
2288 
2289 		if (adapter->dmac == 0) { /* Disabling it */
2290 			E1000_WRITE_REG(hw, E1000_DMACR, reg);
2291 			return;
2292 		} else
2293 			device_printf(dev, "DMA Coalescing enabled\n");
2294 
2295 		/* Set starting threshold */
2296 		E1000_WRITE_REG(hw, E1000_DMCTXTH, 0);
2297 
2298 		hwm = 64 * pba - max_frame_size / 16;
2299 		if (hwm < 64 * (pba - 6))
2300 			hwm = 64 * (pba - 6);
2301 		reg = E1000_READ_REG(hw, E1000_FCRTC);
2302 		reg &= ~E1000_FCRTC_RTH_COAL_MASK;
2303 		reg |= ((hwm << E1000_FCRTC_RTH_COAL_SHIFT)
2304 		    & E1000_FCRTC_RTH_COAL_MASK);
2305 		E1000_WRITE_REG(hw, E1000_FCRTC, reg);
2306 
2307 
2308 		dmac = pba - max_frame_size / 512;
2309 		if (dmac < pba - 10)
2310 			dmac = pba - 10;
2311 		reg = E1000_READ_REG(hw, E1000_DMACR);
2312 		reg &= ~E1000_DMACR_DMACTHR_MASK;
2313 		reg |= ((dmac << E1000_DMACR_DMACTHR_SHIFT)
2314 		    & E1000_DMACR_DMACTHR_MASK);
2315 
2316 		/* transition to L0x or L1 if available..*/
2317 		reg |= (E1000_DMACR_DMAC_EN | E1000_DMACR_DMAC_LX_MASK);
2318 
2319 		/* Check if status is 2.5Gb backplane connection
2320 		* before configuration of watchdog timer, which is
2321 		* in msec values in 12.8usec intervals
2322 		* watchdog timer= msec values in 32usec intervals
2323 		* for non 2.5Gb connection
2324 		*/
2325 		if (hw->mac.type == e1000_i354) {
2326 			int status = E1000_READ_REG(hw, E1000_STATUS);
2327 			if ((status & E1000_STATUS_2P5_SKU) &&
2328 			    (!(status & E1000_STATUS_2P5_SKU_OVER)))
2329 				reg |= ((adapter->dmac * 5) >> 6);
2330 			else
2331 				reg |= (adapter->dmac >> 5);
2332 		} else {
2333 			reg |= (adapter->dmac >> 5);
2334 		}
2335 
2336 		E1000_WRITE_REG(hw, E1000_DMACR, reg);
2337 
2338 		E1000_WRITE_REG(hw, E1000_DMCRTRH, 0);
2339 
2340 		/* Set the interval before transition */
2341 		reg = E1000_READ_REG(hw, E1000_DMCTLX);
2342 		if (hw->mac.type == e1000_i350)
2343 			reg |= IGB_DMCTLX_DCFLUSH_DIS;
2344 		/*
2345 		** in 2.5Gb connection, TTLX unit is 0.4 usec
2346 		** which is 0x4*2 = 0xA. But delay is still 4 usec
2347 		*/
2348 		if (hw->mac.type == e1000_i354) {
2349 			int status = E1000_READ_REG(hw, E1000_STATUS);
2350 			if ((status & E1000_STATUS_2P5_SKU) &&
2351 			    (!(status & E1000_STATUS_2P5_SKU_OVER)))
2352 				reg |= 0xA;
2353 			else
2354 				reg |= 0x4;
2355 		} else {
2356 			reg |= 0x4;
2357 		}
2358 
2359 		E1000_WRITE_REG(hw, E1000_DMCTLX, reg);
2360 
2361 		/* free space in tx packet buffer to wake from DMA coal */
2362 		E1000_WRITE_REG(hw, E1000_DMCTXTH, (IGB_TXPBSIZE -
2363 		    (2 * max_frame_size)) >> 6);
2364 
2365 		/* make low power state decision controlled by DMA coal */
2366 		reg = E1000_READ_REG(hw, E1000_PCIEMISC);
2367 		reg &= ~E1000_PCIEMISC_LX_DECISION;
2368 		E1000_WRITE_REG(hw, E1000_PCIEMISC, reg);
2369 
2370 	} else if (hw->mac.type == e1000_82580) {
2371 		u32 reg = E1000_READ_REG(hw, E1000_PCIEMISC);
2372 		E1000_WRITE_REG(hw, E1000_PCIEMISC,
2373 		    reg & ~E1000_PCIEMISC_LX_DECISION);
2374 		E1000_WRITE_REG(hw, E1000_DMACR, 0);
2375 	}
2376 }
2377 
2378 /*********************************************************************
2379  *
2380  *  Initialize the hardware to a configuration as specified by the
2381  *  adapter structure.
2382  *
2383  **********************************************************************/
2384 static void
2385 em_reset(if_ctx_t ctx)
2386 {
2387 	device_t dev = iflib_get_dev(ctx);
2388 	struct adapter *adapter = iflib_get_softc(ctx);
2389 	struct ifnet *ifp = iflib_get_ifp(ctx);
2390 	struct e1000_hw *hw = &adapter->hw;
2391 	u16 rx_buffer_size;
2392 	u32 pba;
2393 
2394 	INIT_DEBUGOUT("em_reset: begin");
2395 	/* Let the firmware know the OS is in control */
2396 	em_get_hw_control(adapter);
2397 
2398 	/* Set up smart power down as default off on newer adapters. */
2399 	if (!em_smart_pwr_down && (hw->mac.type == e1000_82571 ||
2400 	    hw->mac.type == e1000_82572)) {
2401 		u16 phy_tmp = 0;
2402 
2403 		/* Speed up time to link by disabling smart power down. */
2404 		e1000_read_phy_reg(hw, IGP02E1000_PHY_POWER_MGMT, &phy_tmp);
2405 		phy_tmp &= ~IGP02E1000_PM_SPD;
2406 		e1000_write_phy_reg(hw, IGP02E1000_PHY_POWER_MGMT, phy_tmp);
2407 	}
2408 
2409 	/*
2410 	 * Packet Buffer Allocation (PBA)
2411 	 * Writing PBA sets the receive portion of the buffer
2412 	 * the remainder is used for the transmit buffer.
2413 	 */
2414 	switch (hw->mac.type) {
2415 	/* Total Packet Buffer on these is 48K */
2416 	case e1000_82571:
2417 	case e1000_82572:
2418 	case e1000_80003es2lan:
2419 			pba = E1000_PBA_32K; /* 32K for Rx, 16K for Tx */
2420 		break;
2421 	case e1000_82573: /* 82573: Total Packet Buffer is 32K */
2422 			pba = E1000_PBA_12K; /* 12K for Rx, 20K for Tx */
2423 		break;
2424 	case e1000_82574:
2425 	case e1000_82583:
2426 			pba = E1000_PBA_20K; /* 20K for Rx, 20K for Tx */
2427 		break;
2428 	case e1000_ich8lan:
2429 		pba = E1000_PBA_8K;
2430 		break;
2431 	case e1000_ich9lan:
2432 	case e1000_ich10lan:
2433 		/* Boost Receive side for jumbo frames */
2434 		if (adapter->hw.mac.max_frame_size > 4096)
2435 			pba = E1000_PBA_14K;
2436 		else
2437 			pba = E1000_PBA_10K;
2438 		break;
2439 	case e1000_pchlan:
2440 	case e1000_pch2lan:
2441 	case e1000_pch_lpt:
2442 	case e1000_pch_spt:
2443 	case e1000_pch_cnp:
2444 		pba = E1000_PBA_26K;
2445 		break;
2446 	case e1000_82575:
2447 		pba = E1000_PBA_32K;
2448 		break;
2449 	case e1000_82576:
2450 	case e1000_vfadapt:
2451 		pba = E1000_READ_REG(hw, E1000_RXPBS);
2452 		pba &= E1000_RXPBS_SIZE_MASK_82576;
2453 		break;
2454 	case e1000_82580:
2455 	case e1000_i350:
2456 	case e1000_i354:
2457 	case e1000_vfadapt_i350:
2458 		pba = E1000_READ_REG(hw, E1000_RXPBS);
2459 		pba = e1000_rxpbs_adjust_82580(pba);
2460 		break;
2461 	case e1000_i210:
2462 	case e1000_i211:
2463 		pba = E1000_PBA_34K;
2464 		break;
2465 	default:
2466 		if (adapter->hw.mac.max_frame_size > 8192)
2467 			pba = E1000_PBA_40K; /* 40K for Rx, 24K for Tx */
2468 		else
2469 			pba = E1000_PBA_48K; /* 48K for Rx, 16K for Tx */
2470 	}
2471 
2472 	/* Special needs in case of Jumbo frames */
2473 	if ((hw->mac.type == e1000_82575) && (ifp->if_mtu > ETHERMTU)) {
2474 		u32 tx_space, min_tx, min_rx;
2475 		pba = E1000_READ_REG(hw, E1000_PBA);
2476 		tx_space = pba >> 16;
2477 		pba &= 0xffff;
2478 		min_tx = (adapter->hw.mac.max_frame_size +
2479 		    sizeof(struct e1000_tx_desc) - ETHERNET_FCS_SIZE) * 2;
2480 		min_tx = roundup2(min_tx, 1024);
2481 		min_tx >>= 10;
2482 		min_rx = adapter->hw.mac.max_frame_size;
2483 		min_rx = roundup2(min_rx, 1024);
2484 		min_rx >>= 10;
2485 		if (tx_space < min_tx &&
2486 		    ((min_tx - tx_space) < pba)) {
2487 			pba = pba - (min_tx - tx_space);
2488 			/*
2489 			 * if short on rx space, rx wins
2490 			 * and must trump tx adjustment
2491 			 */
2492 			if (pba < min_rx)
2493 				pba = min_rx;
2494 		}
2495 		E1000_WRITE_REG(hw, E1000_PBA, pba);
2496 	}
2497 
2498 	if (hw->mac.type < igb_mac_min)
2499 		E1000_WRITE_REG(&adapter->hw, E1000_PBA, pba);
2500 
2501 	INIT_DEBUGOUT1("em_reset: pba=%dK",pba);
2502 
2503 	/*
2504 	 * These parameters control the automatic generation (Tx) and
2505 	 * response (Rx) to Ethernet PAUSE frames.
2506 	 * - High water mark should allow for at least two frames to be
2507 	 *   received after sending an XOFF.
2508 	 * - Low water mark works best when it is very near the high water mark.
2509 	 *   This allows the receiver to restart by sending XON when it has
2510 	 *   drained a bit. Here we use an arbitrary value of 1500 which will
2511 	 *   restart after one full frame is pulled from the buffer. There
2512 	 *   could be several smaller frames in the buffer and if so they will
2513 	 *   not trigger the XON until their total number reduces the buffer
2514 	 *   by 1500.
2515 	 * - The pause time is fairly large at 1000 x 512ns = 512 usec.
2516 	 */
2517 	rx_buffer_size = (pba & 0xffff) << 10;
2518 	hw->fc.high_water = rx_buffer_size -
2519 	    roundup2(adapter->hw.mac.max_frame_size, 1024);
2520 	hw->fc.low_water = hw->fc.high_water - 1500;
2521 
2522 	if (adapter->fc) /* locally set flow control value? */
2523 		hw->fc.requested_mode = adapter->fc;
2524 	else
2525 		hw->fc.requested_mode = e1000_fc_full;
2526 
2527 	if (hw->mac.type == e1000_80003es2lan)
2528 		hw->fc.pause_time = 0xFFFF;
2529 	else
2530 		hw->fc.pause_time = EM_FC_PAUSE_TIME;
2531 
2532 	hw->fc.send_xon = TRUE;
2533 
2534 	/* Device specific overrides/settings */
2535 	switch (hw->mac.type) {
2536 	case e1000_pchlan:
2537 		/* Workaround: no TX flow ctrl for PCH */
2538 		hw->fc.requested_mode = e1000_fc_rx_pause;
2539 		hw->fc.pause_time = 0xFFFF; /* override */
2540 		if (if_getmtu(ifp) > ETHERMTU) {
2541 			hw->fc.high_water = 0x3500;
2542 			hw->fc.low_water = 0x1500;
2543 		} else {
2544 			hw->fc.high_water = 0x5000;
2545 			hw->fc.low_water = 0x3000;
2546 		}
2547 		hw->fc.refresh_time = 0x1000;
2548 		break;
2549 	case e1000_pch2lan:
2550 	case e1000_pch_lpt:
2551 	case e1000_pch_spt:
2552 	case e1000_pch_cnp:
2553 		hw->fc.high_water = 0x5C20;
2554 		hw->fc.low_water = 0x5048;
2555 		hw->fc.pause_time = 0x0650;
2556 		hw->fc.refresh_time = 0x0400;
2557 		/* Jumbos need adjusted PBA */
2558 		if (if_getmtu(ifp) > ETHERMTU)
2559 			E1000_WRITE_REG(hw, E1000_PBA, 12);
2560 		else
2561 			E1000_WRITE_REG(hw, E1000_PBA, 26);
2562 		break;
2563 	case e1000_82575:
2564 	case e1000_82576:
2565 		/* 8-byte granularity */
2566 		hw->fc.low_water = hw->fc.high_water - 8;
2567 		break;
2568 	case e1000_82580:
2569 	case e1000_i350:
2570 	case e1000_i354:
2571 	case e1000_i210:
2572 	case e1000_i211:
2573 	case e1000_vfadapt:
2574 	case e1000_vfadapt_i350:
2575 		/* 16-byte granularity */
2576 		hw->fc.low_water = hw->fc.high_water - 16;
2577 		break;
2578 	case e1000_ich9lan:
2579 	case e1000_ich10lan:
2580 		if (if_getmtu(ifp) > ETHERMTU) {
2581 			hw->fc.high_water = 0x2800;
2582 			hw->fc.low_water = hw->fc.high_water - 8;
2583 			break;
2584 		}
2585 		/* FALLTHROUGH */
2586 	default:
2587 		if (hw->mac.type == e1000_80003es2lan)
2588 			hw->fc.pause_time = 0xFFFF;
2589 		break;
2590 	}
2591 
2592 	/* Issue a global reset */
2593 	e1000_reset_hw(hw);
2594 	if (adapter->hw.mac.type >= igb_mac_min) {
2595 		E1000_WRITE_REG(hw, E1000_WUC, 0);
2596 	} else {
2597 		E1000_WRITE_REG(hw, E1000_WUFC, 0);
2598 		em_disable_aspm(adapter);
2599 	}
2600 	if (adapter->flags & IGB_MEDIA_RESET) {
2601 		e1000_setup_init_funcs(hw, TRUE);
2602 		e1000_get_bus_info(hw);
2603 		adapter->flags &= ~IGB_MEDIA_RESET;
2604 	}
2605 	/* and a re-init */
2606 	if (e1000_init_hw(hw) < 0) {
2607 		device_printf(dev, "Hardware Initialization Failed\n");
2608 		return;
2609 	}
2610 	if (adapter->hw.mac.type >= igb_mac_min)
2611 		igb_init_dmac(adapter, pba);
2612 
2613 	E1000_WRITE_REG(hw, E1000_VET, ETHERTYPE_VLAN);
2614 	e1000_get_phy_info(hw);
2615 	e1000_check_for_link(hw);
2616 }
2617 
2618 /*
2619  * Initialise the RSS mapping for NICs that support multiple transmit/
2620  * receive rings.
2621  */
2622 
2623 #define RSSKEYLEN 10
2624 static void
2625 em_initialize_rss_mapping(struct adapter *adapter)
2626 {
2627 	uint8_t  rss_key[4 * RSSKEYLEN];
2628 	uint32_t reta = 0;
2629 	struct e1000_hw	*hw = &adapter->hw;
2630 	int i;
2631 
2632 	/*
2633 	 * Configure RSS key
2634 	 */
2635 	arc4rand(rss_key, sizeof(rss_key), 0);
2636 	for (i = 0; i < RSSKEYLEN; ++i) {
2637 		uint32_t rssrk = 0;
2638 
2639 		rssrk = EM_RSSRK_VAL(rss_key, i);
2640 		E1000_WRITE_REG(hw,E1000_RSSRK(i), rssrk);
2641 	}
2642 
2643 	/*
2644 	 * Configure RSS redirect table in following fashion:
2645 	 * (hash & ring_cnt_mask) == rdr_table[(hash & rdr_table_mask)]
2646 	 */
2647 	for (i = 0; i < sizeof(reta); ++i) {
2648 		uint32_t q;
2649 
2650 		q = (i % adapter->rx_num_queues) << 7;
2651 		reta |= q << (8 * i);
2652 	}
2653 
2654 	for (i = 0; i < 32; ++i)
2655 		E1000_WRITE_REG(hw, E1000_RETA(i), reta);
2656 
2657 	E1000_WRITE_REG(hw, E1000_MRQC, E1000_MRQC_RSS_ENABLE_2Q |
2658 			E1000_MRQC_RSS_FIELD_IPV4_TCP |
2659 			E1000_MRQC_RSS_FIELD_IPV4 |
2660 			E1000_MRQC_RSS_FIELD_IPV6_TCP_EX |
2661 			E1000_MRQC_RSS_FIELD_IPV6_EX |
2662 			E1000_MRQC_RSS_FIELD_IPV6);
2663 }
2664 
2665 static void
2666 igb_initialize_rss_mapping(struct adapter *adapter)
2667 {
2668 	struct e1000_hw *hw = &adapter->hw;
2669 	int i;
2670 	int queue_id;
2671 	u32 reta;
2672 	u32 rss_key[10], mrqc, shift = 0;
2673 
2674 	/* XXX? */
2675 	if (adapter->hw.mac.type == e1000_82575)
2676 		shift = 6;
2677 
2678 	/*
2679 	 * The redirection table controls which destination
2680 	 * queue each bucket redirects traffic to.
2681 	 * Each DWORD represents four queues, with the LSB
2682 	 * being the first queue in the DWORD.
2683 	 *
2684 	 * This just allocates buckets to queues using round-robin
2685 	 * allocation.
2686 	 *
2687 	 * NOTE: It Just Happens to line up with the default
2688 	 * RSS allocation method.
2689 	 */
2690 
2691 	/* Warning FM follows */
2692 	reta = 0;
2693 	for (i = 0; i < 128; i++) {
2694 #ifdef RSS
2695 		queue_id = rss_get_indirection_to_bucket(i);
2696 		/*
2697 		 * If we have more queues than buckets, we'll
2698 		 * end up mapping buckets to a subset of the
2699 		 * queues.
2700 		 *
2701 		 * If we have more buckets than queues, we'll
2702 		 * end up instead assigning multiple buckets
2703 		 * to queues.
2704 		 *
2705 		 * Both are suboptimal, but we need to handle
2706 		 * the case so we don't go out of bounds
2707 		 * indexing arrays and such.
2708 		 */
2709 		queue_id = queue_id % adapter->rx_num_queues;
2710 #else
2711 		queue_id = (i % adapter->rx_num_queues);
2712 #endif
2713 		/* Adjust if required */
2714 		queue_id = queue_id << shift;
2715 
2716 		/*
2717 		 * The low 8 bits are for hash value (n+0);
2718 		 * The next 8 bits are for hash value (n+1), etc.
2719 		 */
2720 		reta = reta >> 8;
2721 		reta = reta | ( ((uint32_t) queue_id) << 24);
2722 		if ((i & 3) == 3) {
2723 			E1000_WRITE_REG(hw, E1000_RETA(i >> 2), reta);
2724 			reta = 0;
2725 		}
2726 	}
2727 
2728 	/* Now fill in hash table */
2729 
2730 	/*
2731 	 * MRQC: Multiple Receive Queues Command
2732 	 * Set queuing to RSS control, number depends on the device.
2733 	 */
2734 	mrqc = E1000_MRQC_ENABLE_RSS_8Q;
2735 
2736 #ifdef RSS
2737 	/* XXX ew typecasting */
2738 	rss_getkey((uint8_t *) &rss_key);
2739 #else
2740 	arc4rand(&rss_key, sizeof(rss_key), 0);
2741 #endif
2742 	for (i = 0; i < 10; i++)
2743 		E1000_WRITE_REG_ARRAY(hw, E1000_RSSRK(0), i, rss_key[i]);
2744 
2745 	/*
2746 	 * Configure the RSS fields to hash upon.
2747 	 */
2748 	mrqc |= (E1000_MRQC_RSS_FIELD_IPV4 |
2749 	    E1000_MRQC_RSS_FIELD_IPV4_TCP);
2750 	mrqc |= (E1000_MRQC_RSS_FIELD_IPV6 |
2751 	    E1000_MRQC_RSS_FIELD_IPV6_TCP);
2752 	mrqc |=( E1000_MRQC_RSS_FIELD_IPV4_UDP |
2753 	    E1000_MRQC_RSS_FIELD_IPV6_UDP);
2754 	mrqc |=( E1000_MRQC_RSS_FIELD_IPV6_UDP_EX |
2755 	    E1000_MRQC_RSS_FIELD_IPV6_TCP_EX);
2756 
2757 	E1000_WRITE_REG(hw, E1000_MRQC, mrqc);
2758 }
2759 
2760 /*********************************************************************
2761  *
2762  *  Setup networking device structure and register interface media.
2763  *
2764  **********************************************************************/
2765 static int
2766 em_setup_interface(if_ctx_t ctx)
2767 {
2768 	struct ifnet *ifp = iflib_get_ifp(ctx);
2769 	struct adapter *adapter = iflib_get_softc(ctx);
2770 	if_softc_ctx_t scctx = adapter->shared;
2771 
2772 	INIT_DEBUGOUT("em_setup_interface: begin");
2773 
2774 	/* Single Queue */
2775 	if (adapter->tx_num_queues == 1) {
2776 		if_setsendqlen(ifp, scctx->isc_ntxd[0] - 1);
2777 		if_setsendqready(ifp);
2778 	}
2779 
2780 	/*
2781 	 * Specify the media types supported by this adapter and register
2782 	 * callbacks to update media and link information
2783 	 */
2784 	if ((adapter->hw.phy.media_type == e1000_media_type_fiber) ||
2785 	    (adapter->hw.phy.media_type == e1000_media_type_internal_serdes)) {
2786 		u_char fiber_type = IFM_1000_SX;	/* default type */
2787 
2788 		if (adapter->hw.mac.type == e1000_82545)
2789 			fiber_type = IFM_1000_LX;
2790 		ifmedia_add(adapter->media, IFM_ETHER | fiber_type | IFM_FDX, 0, NULL);
2791 		ifmedia_add(adapter->media, IFM_ETHER | fiber_type, 0, NULL);
2792 	} else {
2793 		ifmedia_add(adapter->media, IFM_ETHER | IFM_10_T, 0, NULL);
2794 		ifmedia_add(adapter->media, IFM_ETHER | IFM_10_T | IFM_FDX, 0, NULL);
2795 		ifmedia_add(adapter->media, IFM_ETHER | IFM_100_TX, 0, NULL);
2796 		ifmedia_add(adapter->media, IFM_ETHER | IFM_100_TX | IFM_FDX, 0, NULL);
2797 		if (adapter->hw.phy.type != e1000_phy_ife) {
2798 			ifmedia_add(adapter->media, IFM_ETHER | IFM_1000_T | IFM_FDX, 0, NULL);
2799 			ifmedia_add(adapter->media, IFM_ETHER | IFM_1000_T, 0, NULL);
2800 		}
2801 	}
2802 	ifmedia_add(adapter->media, IFM_ETHER | IFM_AUTO, 0, NULL);
2803 	ifmedia_set(adapter->media, IFM_ETHER | IFM_AUTO);
2804 	return (0);
2805 }
2806 
2807 static int
2808 em_if_tx_queues_alloc(if_ctx_t ctx, caddr_t *vaddrs, uint64_t *paddrs, int ntxqs, int ntxqsets)
2809 {
2810 	struct adapter *adapter = iflib_get_softc(ctx);
2811 	if_softc_ctx_t scctx = adapter->shared;
2812 	int error = E1000_SUCCESS;
2813 	struct em_tx_queue *que;
2814 	int i, j;
2815 
2816 	MPASS(adapter->tx_num_queues > 0);
2817 	MPASS(adapter->tx_num_queues == ntxqsets);
2818 
2819 	/* First allocate the top level queue structs */
2820 	if (!(adapter->tx_queues =
2821 	    (struct em_tx_queue *) malloc(sizeof(struct em_tx_queue) *
2822 	    adapter->tx_num_queues, M_DEVBUF, M_NOWAIT | M_ZERO))) {
2823 		device_printf(iflib_get_dev(ctx), "Unable to allocate queue memory\n");
2824 		return(ENOMEM);
2825 	}
2826 
2827 	for (i = 0, que = adapter->tx_queues; i < adapter->tx_num_queues; i++, que++) {
2828 		/* Set up some basics */
2829 
2830 		struct tx_ring *txr = &que->txr;
2831 		txr->adapter = que->adapter = adapter;
2832 		que->me = txr->me =  i;
2833 
2834 		/* Allocate report status array */
2835 		if (!(txr->tx_rsq = (qidx_t *) malloc(sizeof(qidx_t) * scctx->isc_ntxd[0], M_DEVBUF, M_NOWAIT | M_ZERO))) {
2836 			device_printf(iflib_get_dev(ctx), "failed to allocate rs_idxs memory\n");
2837 			error = ENOMEM;
2838 			goto fail;
2839 		}
2840 		for (j = 0; j < scctx->isc_ntxd[0]; j++)
2841 			txr->tx_rsq[j] = QIDX_INVALID;
2842 		/* get the virtual and physical address of the hardware queues */
2843 		txr->tx_base = (struct e1000_tx_desc *)vaddrs[i*ntxqs];
2844 		txr->tx_paddr = paddrs[i*ntxqs];
2845 	}
2846 
2847 	if (bootverbose)
2848 		device_printf(iflib_get_dev(ctx),
2849 		    "allocated for %d tx_queues\n", adapter->tx_num_queues);
2850 	return (0);
2851 fail:
2852 	em_if_queues_free(ctx);
2853 	return (error);
2854 }
2855 
2856 static int
2857 em_if_rx_queues_alloc(if_ctx_t ctx, caddr_t *vaddrs, uint64_t *paddrs, int nrxqs, int nrxqsets)
2858 {
2859 	struct adapter *adapter = iflib_get_softc(ctx);
2860 	int error = E1000_SUCCESS;
2861 	struct em_rx_queue *que;
2862 	int i;
2863 
2864 	MPASS(adapter->rx_num_queues > 0);
2865 	MPASS(adapter->rx_num_queues == nrxqsets);
2866 
2867 	/* First allocate the top level queue structs */
2868 	if (!(adapter->rx_queues =
2869 	    (struct em_rx_queue *) malloc(sizeof(struct em_rx_queue) *
2870 	    adapter->rx_num_queues, M_DEVBUF, M_NOWAIT | M_ZERO))) {
2871 		device_printf(iflib_get_dev(ctx), "Unable to allocate queue memory\n");
2872 		error = ENOMEM;
2873 		goto fail;
2874 	}
2875 
2876 	for (i = 0, que = adapter->rx_queues; i < nrxqsets; i++, que++) {
2877 		/* Set up some basics */
2878 		struct rx_ring *rxr = &que->rxr;
2879 		rxr->adapter = que->adapter = adapter;
2880 		rxr->que = que;
2881 		que->me = rxr->me =  i;
2882 
2883 		/* get the virtual and physical address of the hardware queues */
2884 		rxr->rx_base = (union e1000_rx_desc_extended *)vaddrs[i*nrxqs];
2885 		rxr->rx_paddr = paddrs[i*nrxqs];
2886 	}
2887 
2888 	if (bootverbose)
2889 		device_printf(iflib_get_dev(ctx),
2890 		    "allocated for %d rx_queues\n", adapter->rx_num_queues);
2891 
2892 	return (0);
2893 fail:
2894 	em_if_queues_free(ctx);
2895 	return (error);
2896 }
2897 
2898 static void
2899 em_if_queues_free(if_ctx_t ctx)
2900 {
2901 	struct adapter *adapter = iflib_get_softc(ctx);
2902 	struct em_tx_queue *tx_que = adapter->tx_queues;
2903 	struct em_rx_queue *rx_que = adapter->rx_queues;
2904 
2905 	if (tx_que != NULL) {
2906 		for (int i = 0; i < adapter->tx_num_queues; i++, tx_que++) {
2907 			struct tx_ring *txr = &tx_que->txr;
2908 			if (txr->tx_rsq == NULL)
2909 				break;
2910 
2911 			free(txr->tx_rsq, M_DEVBUF);
2912 			txr->tx_rsq = NULL;
2913 		}
2914 		free(adapter->tx_queues, M_DEVBUF);
2915 		adapter->tx_queues = NULL;
2916 	}
2917 
2918 	if (rx_que != NULL) {
2919 		free(adapter->rx_queues, M_DEVBUF);
2920 		adapter->rx_queues = NULL;
2921 	}
2922 
2923 	em_release_hw_control(adapter);
2924 
2925 	if (adapter->mta != NULL) {
2926 		free(adapter->mta, M_DEVBUF);
2927 	}
2928 }
2929 
2930 /*********************************************************************
2931  *
2932  *  Enable transmit unit.
2933  *
2934  **********************************************************************/
2935 static void
2936 em_initialize_transmit_unit(if_ctx_t ctx)
2937 {
2938 	struct adapter *adapter = iflib_get_softc(ctx);
2939 	if_softc_ctx_t scctx = adapter->shared;
2940 	struct em_tx_queue *que;
2941 	struct tx_ring	*txr;
2942 	struct e1000_hw	*hw = &adapter->hw;
2943 	u32 tctl, txdctl = 0, tarc, tipg = 0;
2944 
2945 	INIT_DEBUGOUT("em_initialize_transmit_unit: begin");
2946 
2947 	for (int i = 0; i < adapter->tx_num_queues; i++, txr++) {
2948 		u64 bus_addr;
2949 		caddr_t offp, endp;
2950 
2951 		que = &adapter->tx_queues[i];
2952 		txr = &que->txr;
2953 		bus_addr = txr->tx_paddr;
2954 
2955 		/* Clear checksum offload context. */
2956 		offp = (caddr_t)&txr->csum_flags;
2957 		endp = (caddr_t)(txr + 1);
2958 		bzero(offp, endp - offp);
2959 
2960 		/* Base and Len of TX Ring */
2961 		E1000_WRITE_REG(hw, E1000_TDLEN(i),
2962 		    scctx->isc_ntxd[0] * sizeof(struct e1000_tx_desc));
2963 		E1000_WRITE_REG(hw, E1000_TDBAH(i),
2964 		    (u32)(bus_addr >> 32));
2965 		E1000_WRITE_REG(hw, E1000_TDBAL(i),
2966 		    (u32)bus_addr);
2967 		/* Init the HEAD/TAIL indices */
2968 		E1000_WRITE_REG(hw, E1000_TDT(i), 0);
2969 		E1000_WRITE_REG(hw, E1000_TDH(i), 0);
2970 
2971 		HW_DEBUGOUT2("Base = %x, Length = %x\n",
2972 		    E1000_READ_REG(&adapter->hw, E1000_TDBAL(i)),
2973 		    E1000_READ_REG(&adapter->hw, E1000_TDLEN(i)));
2974 
2975 		txdctl = 0; /* clear txdctl */
2976 		txdctl |= 0x1f; /* PTHRESH */
2977 		txdctl |= 1 << 8; /* HTHRESH */
2978 		txdctl |= 1 << 16;/* WTHRESH */
2979 		txdctl |= 1 << 22; /* Reserved bit 22 must always be 1 */
2980 		txdctl |= E1000_TXDCTL_GRAN;
2981 		txdctl |= 1 << 25; /* LWTHRESH */
2982 
2983 		E1000_WRITE_REG(hw, E1000_TXDCTL(i), txdctl);
2984 	}
2985 
2986 	/* Set the default values for the Tx Inter Packet Gap timer */
2987 	switch (adapter->hw.mac.type) {
2988 	case e1000_80003es2lan:
2989 		tipg = DEFAULT_82543_TIPG_IPGR1;
2990 		tipg |= DEFAULT_80003ES2LAN_TIPG_IPGR2 <<
2991 		    E1000_TIPG_IPGR2_SHIFT;
2992 		break;
2993 	case e1000_82542:
2994 		tipg = DEFAULT_82542_TIPG_IPGT;
2995 		tipg |= DEFAULT_82542_TIPG_IPGR1 << E1000_TIPG_IPGR1_SHIFT;
2996 		tipg |= DEFAULT_82542_TIPG_IPGR2 << E1000_TIPG_IPGR2_SHIFT;
2997 		break;
2998 	default:
2999 		if ((adapter->hw.phy.media_type == e1000_media_type_fiber) ||
3000 		    (adapter->hw.phy.media_type ==
3001 		    e1000_media_type_internal_serdes))
3002 			tipg = DEFAULT_82543_TIPG_IPGT_FIBER;
3003 		else
3004 			tipg = DEFAULT_82543_TIPG_IPGT_COPPER;
3005 		tipg |= DEFAULT_82543_TIPG_IPGR1 << E1000_TIPG_IPGR1_SHIFT;
3006 		tipg |= DEFAULT_82543_TIPG_IPGR2 << E1000_TIPG_IPGR2_SHIFT;
3007 	}
3008 
3009 	E1000_WRITE_REG(&adapter->hw, E1000_TIPG, tipg);
3010 	E1000_WRITE_REG(&adapter->hw, E1000_TIDV, adapter->tx_int_delay.value);
3011 
3012 	if(adapter->hw.mac.type >= e1000_82540)
3013 		E1000_WRITE_REG(&adapter->hw, E1000_TADV,
3014 		    adapter->tx_abs_int_delay.value);
3015 
3016 	if ((adapter->hw.mac.type == e1000_82571) ||
3017 	    (adapter->hw.mac.type == e1000_82572)) {
3018 		tarc = E1000_READ_REG(&adapter->hw, E1000_TARC(0));
3019 		tarc |= TARC_SPEED_MODE_BIT;
3020 		E1000_WRITE_REG(&adapter->hw, E1000_TARC(0), tarc);
3021 	} else if (adapter->hw.mac.type == e1000_80003es2lan) {
3022 		/* errata: program both queues to unweighted RR */
3023 		tarc = E1000_READ_REG(&adapter->hw, E1000_TARC(0));
3024 		tarc |= 1;
3025 		E1000_WRITE_REG(&adapter->hw, E1000_TARC(0), tarc);
3026 		tarc = E1000_READ_REG(&adapter->hw, E1000_TARC(1));
3027 		tarc |= 1;
3028 		E1000_WRITE_REG(&adapter->hw, E1000_TARC(1), tarc);
3029 	} else if (adapter->hw.mac.type == e1000_82574) {
3030 		tarc = E1000_READ_REG(&adapter->hw, E1000_TARC(0));
3031 		tarc |= TARC_ERRATA_BIT;
3032 		if ( adapter->tx_num_queues > 1) {
3033 			tarc |= (TARC_COMPENSATION_MODE | TARC_MQ_FIX);
3034 			E1000_WRITE_REG(&adapter->hw, E1000_TARC(0), tarc);
3035 			E1000_WRITE_REG(&adapter->hw, E1000_TARC(1), tarc);
3036 		} else
3037 			E1000_WRITE_REG(&adapter->hw, E1000_TARC(0), tarc);
3038 	}
3039 
3040 	if (adapter->tx_int_delay.value > 0)
3041 		adapter->txd_cmd |= E1000_TXD_CMD_IDE;
3042 
3043 	/* Program the Transmit Control Register */
3044 	tctl = E1000_READ_REG(&adapter->hw, E1000_TCTL);
3045 	tctl &= ~E1000_TCTL_CT;
3046 	tctl |= (E1000_TCTL_PSP | E1000_TCTL_RTLC | E1000_TCTL_EN |
3047 		   (E1000_COLLISION_THRESHOLD << E1000_CT_SHIFT));
3048 
3049 	if (adapter->hw.mac.type >= e1000_82571)
3050 		tctl |= E1000_TCTL_MULR;
3051 
3052 	/* This write will effectively turn on the transmit unit. */
3053 	E1000_WRITE_REG(&adapter->hw, E1000_TCTL, tctl);
3054 
3055 	/* SPT and KBL errata workarounds */
3056 	if (hw->mac.type == e1000_pch_spt) {
3057 		u32 reg;
3058 		reg = E1000_READ_REG(hw, E1000_IOSFPC);
3059 		reg |= E1000_RCTL_RDMTS_HEX;
3060 		E1000_WRITE_REG(hw, E1000_IOSFPC, reg);
3061 		/* i218-i219 Specification Update 1.5.4.5 */
3062 		reg = E1000_READ_REG(hw, E1000_TARC(0));
3063 		reg &= ~E1000_TARC0_CB_MULTIQ_3_REQ;
3064 		reg |= E1000_TARC0_CB_MULTIQ_2_REQ;
3065 		E1000_WRITE_REG(hw, E1000_TARC(0), reg);
3066 	}
3067 }
3068 
3069 /*********************************************************************
3070  *
3071  *  Enable receive unit.
3072  *
3073  **********************************************************************/
3074 
3075 static void
3076 em_initialize_receive_unit(if_ctx_t ctx)
3077 {
3078 	struct adapter *adapter = iflib_get_softc(ctx);
3079 	if_softc_ctx_t scctx = adapter->shared;
3080 	struct ifnet *ifp = iflib_get_ifp(ctx);
3081 	struct e1000_hw	*hw = &adapter->hw;
3082 	struct em_rx_queue *que;
3083 	int i;
3084 	u32 rctl, rxcsum, rfctl;
3085 
3086 	INIT_DEBUGOUT("em_initialize_receive_units: begin");
3087 
3088 	/*
3089 	 * Make sure receives are disabled while setting
3090 	 * up the descriptor ring
3091 	 */
3092 	rctl = E1000_READ_REG(hw, E1000_RCTL);
3093 	/* Do not disable if ever enabled on this hardware */
3094 	if ((hw->mac.type != e1000_82574) && (hw->mac.type != e1000_82583))
3095 		E1000_WRITE_REG(hw, E1000_RCTL, rctl & ~E1000_RCTL_EN);
3096 
3097 	/* Setup the Receive Control Register */
3098 	rctl &= ~(3 << E1000_RCTL_MO_SHIFT);
3099 	rctl |= E1000_RCTL_EN | E1000_RCTL_BAM |
3100 	    E1000_RCTL_LBM_NO | E1000_RCTL_RDMTS_HALF |
3101 	    (hw->mac.mc_filter_type << E1000_RCTL_MO_SHIFT);
3102 
3103 	/* Do not store bad packets */
3104 	rctl &= ~E1000_RCTL_SBP;
3105 
3106 	/* Enable Long Packet receive */
3107 	if (if_getmtu(ifp) > ETHERMTU)
3108 		rctl |= E1000_RCTL_LPE;
3109 	else
3110 		rctl &= ~E1000_RCTL_LPE;
3111 
3112 	/* Strip the CRC */
3113 	if (!em_disable_crc_stripping)
3114 		rctl |= E1000_RCTL_SECRC;
3115 
3116 	if (adapter->hw.mac.type >= e1000_82540) {
3117 		E1000_WRITE_REG(&adapter->hw, E1000_RADV,
3118 			    adapter->rx_abs_int_delay.value);
3119 
3120 		/*
3121 		 * Set the interrupt throttling rate. Value is calculated
3122 		 * as DEFAULT_ITR = 1/(MAX_INTS_PER_SEC * 256ns)
3123 		 */
3124 		E1000_WRITE_REG(hw, E1000_ITR, DEFAULT_ITR);
3125 	}
3126 	E1000_WRITE_REG(&adapter->hw, E1000_RDTR,
3127 	    adapter->rx_int_delay.value);
3128 
3129 	/* Use extended rx descriptor formats */
3130 	rfctl = E1000_READ_REG(hw, E1000_RFCTL);
3131 	rfctl |= E1000_RFCTL_EXTEN;
3132 	/*
3133 	 * When using MSI-X interrupts we need to throttle
3134 	 * using the EITR register (82574 only)
3135 	 */
3136 	if (hw->mac.type == e1000_82574) {
3137 		for (int i = 0; i < 4; i++)
3138 			E1000_WRITE_REG(hw, E1000_EITR_82574(i),
3139 			    DEFAULT_ITR);
3140 		/* Disable accelerated acknowledge */
3141 		rfctl |= E1000_RFCTL_ACK_DIS;
3142 	}
3143 	E1000_WRITE_REG(hw, E1000_RFCTL, rfctl);
3144 
3145 	rxcsum = E1000_READ_REG(hw, E1000_RXCSUM);
3146 	if (if_getcapenable(ifp) & IFCAP_RXCSUM &&
3147 	    adapter->hw.mac.type >= e1000_82543) {
3148 		if (adapter->tx_num_queues > 1) {
3149 			if (adapter->hw.mac.type >= igb_mac_min) {
3150 				rxcsum |= E1000_RXCSUM_PCSD;
3151 				if (hw->mac.type != e1000_82575)
3152 					rxcsum |= E1000_RXCSUM_CRCOFL;
3153 			} else
3154 				rxcsum |= E1000_RXCSUM_TUOFL |
3155 					E1000_RXCSUM_IPOFL |
3156 					E1000_RXCSUM_PCSD;
3157 		} else {
3158 			if (adapter->hw.mac.type >= igb_mac_min)
3159 				rxcsum |= E1000_RXCSUM_IPPCSE;
3160 			else
3161 				rxcsum |= E1000_RXCSUM_TUOFL | E1000_RXCSUM_IPOFL;
3162 			if (adapter->hw.mac.type > e1000_82575)
3163 				rxcsum |= E1000_RXCSUM_CRCOFL;
3164 		}
3165 	} else
3166 		rxcsum &= ~E1000_RXCSUM_TUOFL;
3167 
3168 	E1000_WRITE_REG(hw, E1000_RXCSUM, rxcsum);
3169 
3170 	if (adapter->rx_num_queues > 1) {
3171 		if (adapter->hw.mac.type >= igb_mac_min)
3172 			igb_initialize_rss_mapping(adapter);
3173 		else
3174 			em_initialize_rss_mapping(adapter);
3175 	}
3176 
3177 	/*
3178 	 * XXX TEMPORARY WORKAROUND: on some systems with 82573
3179 	 * long latencies are observed, like Lenovo X60. This
3180 	 * change eliminates the problem, but since having positive
3181 	 * values in RDTR is a known source of problems on other
3182 	 * platforms another solution is being sought.
3183 	 */
3184 	if (hw->mac.type == e1000_82573)
3185 		E1000_WRITE_REG(hw, E1000_RDTR, 0x20);
3186 
3187 	for (i = 0, que = adapter->rx_queues; i < adapter->rx_num_queues; i++, que++) {
3188 		struct rx_ring *rxr = &que->rxr;
3189 		/* Setup the Base and Length of the Rx Descriptor Ring */
3190 		u64 bus_addr = rxr->rx_paddr;
3191 #if 0
3192 		u32 rdt = adapter->rx_num_queues -1;  /* default */
3193 #endif
3194 
3195 		E1000_WRITE_REG(hw, E1000_RDLEN(i),
3196 		    scctx->isc_nrxd[0] * sizeof(union e1000_rx_desc_extended));
3197 		E1000_WRITE_REG(hw, E1000_RDBAH(i), (u32)(bus_addr >> 32));
3198 		E1000_WRITE_REG(hw, E1000_RDBAL(i), (u32)bus_addr);
3199 		/* Setup the Head and Tail Descriptor Pointers */
3200 		E1000_WRITE_REG(hw, E1000_RDH(i), 0);
3201 		E1000_WRITE_REG(hw, E1000_RDT(i), 0);
3202 	}
3203 
3204 	/*
3205 	 * Set PTHRESH for improved jumbo performance
3206 	 * According to 10.2.5.11 of Intel 82574 Datasheet,
3207 	 * RXDCTL(1) is written whenever RXDCTL(0) is written.
3208 	 * Only write to RXDCTL(1) if there is a need for different
3209 	 * settings.
3210 	 */
3211 
3212 	if (((adapter->hw.mac.type == e1000_ich9lan) ||
3213 	    (adapter->hw.mac.type == e1000_pch2lan) ||
3214 	    (adapter->hw.mac.type == e1000_ich10lan)) &&
3215 	    (if_getmtu(ifp) > ETHERMTU)) {
3216 		u32 rxdctl = E1000_READ_REG(hw, E1000_RXDCTL(0));
3217 		E1000_WRITE_REG(hw, E1000_RXDCTL(0), rxdctl | 3);
3218 	} else if (adapter->hw.mac.type == e1000_82574) {
3219 		for (int i = 0; i < adapter->rx_num_queues; i++) {
3220 			u32 rxdctl = E1000_READ_REG(hw, E1000_RXDCTL(i));
3221 			rxdctl |= 0x20; /* PTHRESH */
3222 			rxdctl |= 4 << 8; /* HTHRESH */
3223 			rxdctl |= 4 << 16;/* WTHRESH */
3224 			rxdctl |= 1 << 24; /* Switch to granularity */
3225 			E1000_WRITE_REG(hw, E1000_RXDCTL(i), rxdctl);
3226 		}
3227 	} else if (adapter->hw.mac.type >= igb_mac_min) {
3228 		u32 psize, srrctl = 0;
3229 
3230 		if (if_getmtu(ifp) > ETHERMTU) {
3231 			/* Set maximum packet len */
3232 			if (adapter->rx_mbuf_sz <= 4096) {
3233 				srrctl |= 4096 >> E1000_SRRCTL_BSIZEPKT_SHIFT;
3234 				rctl |= E1000_RCTL_SZ_4096 | E1000_RCTL_BSEX;
3235 			} else if (adapter->rx_mbuf_sz > 4096) {
3236 				srrctl |= 8192 >> E1000_SRRCTL_BSIZEPKT_SHIFT;
3237 				rctl |= E1000_RCTL_SZ_8192 | E1000_RCTL_BSEX;
3238 			}
3239 			psize = scctx->isc_max_frame_size;
3240 			/* are we on a vlan? */
3241 			if (ifp->if_vlantrunk != NULL)
3242 				psize += VLAN_TAG_SIZE;
3243 			E1000_WRITE_REG(&adapter->hw, E1000_RLPML, psize);
3244 		} else {
3245 			srrctl |= 2048 >> E1000_SRRCTL_BSIZEPKT_SHIFT;
3246 			rctl |= E1000_RCTL_SZ_2048;
3247 		}
3248 
3249 		/*
3250 		 * If TX flow control is disabled and there's >1 queue defined,
3251 		 * enable DROP.
3252 		 *
3253 		 * This drops frames rather than hanging the RX MAC for all queues.
3254 		 */
3255 		if ((adapter->rx_num_queues > 1) &&
3256 		    (adapter->fc == e1000_fc_none ||
3257 		     adapter->fc == e1000_fc_rx_pause)) {
3258 			srrctl |= E1000_SRRCTL_DROP_EN;
3259 		}
3260 			/* Setup the Base and Length of the Rx Descriptor Rings */
3261 		for (i = 0, que = adapter->rx_queues; i < adapter->rx_num_queues; i++, que++) {
3262 			struct rx_ring *rxr = &que->rxr;
3263 			u64 bus_addr = rxr->rx_paddr;
3264 			u32 rxdctl;
3265 
3266 #ifdef notyet
3267 			/* Configure for header split? -- ignore for now */
3268 			rxr->hdr_split = igb_header_split;
3269 #else
3270 			srrctl |= E1000_SRRCTL_DESCTYPE_ADV_ONEBUF;
3271 #endif
3272 
3273 			E1000_WRITE_REG(hw, E1000_RDLEN(i),
3274 					scctx->isc_nrxd[0] * sizeof(struct e1000_rx_desc));
3275 			E1000_WRITE_REG(hw, E1000_RDBAH(i),
3276 					(uint32_t)(bus_addr >> 32));
3277 			E1000_WRITE_REG(hw, E1000_RDBAL(i),
3278 					(uint32_t)bus_addr);
3279 			E1000_WRITE_REG(hw, E1000_SRRCTL(i), srrctl);
3280 			/* Enable this Queue */
3281 			rxdctl = E1000_READ_REG(hw, E1000_RXDCTL(i));
3282 			rxdctl |= E1000_RXDCTL_QUEUE_ENABLE;
3283 			rxdctl &= 0xFFF00000;
3284 			rxdctl |= IGB_RX_PTHRESH;
3285 			rxdctl |= IGB_RX_HTHRESH << 8;
3286 			rxdctl |= IGB_RX_WTHRESH << 16;
3287 			E1000_WRITE_REG(hw, E1000_RXDCTL(i), rxdctl);
3288 		}
3289 	} else if (adapter->hw.mac.type >= e1000_pch2lan) {
3290 		if (if_getmtu(ifp) > ETHERMTU)
3291 			e1000_lv_jumbo_workaround_ich8lan(hw, TRUE);
3292 		else
3293 			e1000_lv_jumbo_workaround_ich8lan(hw, FALSE);
3294 	}
3295 
3296 	/* Make sure VLAN Filters are off */
3297 	rctl &= ~E1000_RCTL_VFE;
3298 
3299 	if (adapter->hw.mac.type < igb_mac_min) {
3300 		if (adapter->rx_mbuf_sz == MCLBYTES)
3301 			rctl |= E1000_RCTL_SZ_2048;
3302 		else if (adapter->rx_mbuf_sz == MJUMPAGESIZE)
3303 			rctl |= E1000_RCTL_SZ_4096 | E1000_RCTL_BSEX;
3304 		else if (adapter->rx_mbuf_sz > MJUMPAGESIZE)
3305 			rctl |= E1000_RCTL_SZ_8192 | E1000_RCTL_BSEX;
3306 
3307 		/* ensure we clear use DTYPE of 00 here */
3308 		rctl &= ~0x00000C00;
3309 	}
3310 
3311 	/* Write out the settings */
3312 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
3313 
3314 	return;
3315 }
3316 
3317 static void
3318 em_if_vlan_register(if_ctx_t ctx, u16 vtag)
3319 {
3320 	struct adapter *adapter = iflib_get_softc(ctx);
3321 	u32 index, bit;
3322 
3323 	index = (vtag >> 5) & 0x7F;
3324 	bit = vtag & 0x1F;
3325 	adapter->shadow_vfta[index] |= (1 << bit);
3326 	++adapter->num_vlans;
3327 }
3328 
3329 static void
3330 em_if_vlan_unregister(if_ctx_t ctx, u16 vtag)
3331 {
3332 	struct adapter *adapter = iflib_get_softc(ctx);
3333 	u32 index, bit;
3334 
3335 	index = (vtag >> 5) & 0x7F;
3336 	bit = vtag & 0x1F;
3337 	adapter->shadow_vfta[index] &= ~(1 << bit);
3338 	--adapter->num_vlans;
3339 }
3340 
3341 static void
3342 em_setup_vlan_hw_support(struct adapter *adapter)
3343 {
3344 	struct e1000_hw *hw = &adapter->hw;
3345 	u32 reg;
3346 
3347 	/*
3348 	 * We get here thru init_locked, meaning
3349 	 * a soft reset, this has already cleared
3350 	 * the VFTA and other state, so if there
3351 	 * have been no vlan's registered do nothing.
3352 	 */
3353 	if (adapter->num_vlans == 0)
3354 		return;
3355 
3356 	/*
3357 	 * A soft reset zero's out the VFTA, so
3358 	 * we need to repopulate it now.
3359 	 */
3360 	for (int i = 0; i < EM_VFTA_SIZE; i++)
3361 		if (adapter->shadow_vfta[i] != 0)
3362 			E1000_WRITE_REG_ARRAY(hw, E1000_VFTA,
3363 			    i, adapter->shadow_vfta[i]);
3364 
3365 	reg = E1000_READ_REG(hw, E1000_CTRL);
3366 	reg |= E1000_CTRL_VME;
3367 	E1000_WRITE_REG(hw, E1000_CTRL, reg);
3368 
3369 	/* Enable the Filter Table */
3370 	reg = E1000_READ_REG(hw, E1000_RCTL);
3371 	reg &= ~E1000_RCTL_CFIEN;
3372 	reg |= E1000_RCTL_VFE;
3373 	E1000_WRITE_REG(hw, E1000_RCTL, reg);
3374 }
3375 
3376 static void
3377 em_if_enable_intr(if_ctx_t ctx)
3378 {
3379 	struct adapter *adapter = iflib_get_softc(ctx);
3380 	struct e1000_hw *hw = &adapter->hw;
3381 	u32 ims_mask = IMS_ENABLE_MASK;
3382 
3383 	if (hw->mac.type == e1000_82574) {
3384 		E1000_WRITE_REG(hw, EM_EIAC, EM_MSIX_MASK);
3385 		ims_mask |= adapter->ims;
3386 	} else if (adapter->intr_type == IFLIB_INTR_MSIX && hw->mac.type >= igb_mac_min)  {
3387 		u32 mask = (adapter->que_mask | adapter->link_mask);
3388 
3389 		E1000_WRITE_REG(&adapter->hw, E1000_EIAC, mask);
3390 		E1000_WRITE_REG(&adapter->hw, E1000_EIAM, mask);
3391 		E1000_WRITE_REG(&adapter->hw, E1000_EIMS, mask);
3392 		ims_mask = E1000_IMS_LSC;
3393 	}
3394 
3395 	E1000_WRITE_REG(hw, E1000_IMS, ims_mask);
3396 }
3397 
3398 static void
3399 em_if_disable_intr(if_ctx_t ctx)
3400 {
3401 	struct adapter *adapter = iflib_get_softc(ctx);
3402 	struct e1000_hw *hw = &adapter->hw;
3403 
3404 	if (adapter->intr_type == IFLIB_INTR_MSIX) {
3405 		if (hw->mac.type >= igb_mac_min)
3406 			E1000_WRITE_REG(&adapter->hw, E1000_EIMC, ~0);
3407 		E1000_WRITE_REG(&adapter->hw, E1000_EIAC, 0);
3408 	}
3409 	E1000_WRITE_REG(&adapter->hw, E1000_IMC, 0xffffffff);
3410 }
3411 
3412 /*
3413  * Bit of a misnomer, what this really means is
3414  * to enable OS management of the system... aka
3415  * to disable special hardware management features
3416  */
3417 static void
3418 em_init_manageability(struct adapter *adapter)
3419 {
3420 	/* A shared code workaround */
3421 #define E1000_82542_MANC2H E1000_MANC2H
3422 	if (adapter->has_manage) {
3423 		int manc2h = E1000_READ_REG(&adapter->hw, E1000_MANC2H);
3424 		int manc = E1000_READ_REG(&adapter->hw, E1000_MANC);
3425 
3426 		/* disable hardware interception of ARP */
3427 		manc &= ~(E1000_MANC_ARP_EN);
3428 
3429 		/* enable receiving management packets to the host */
3430 		manc |= E1000_MANC_EN_MNG2HOST;
3431 #define E1000_MNG2HOST_PORT_623 (1 << 5)
3432 #define E1000_MNG2HOST_PORT_664 (1 << 6)
3433 		manc2h |= E1000_MNG2HOST_PORT_623;
3434 		manc2h |= E1000_MNG2HOST_PORT_664;
3435 		E1000_WRITE_REG(&adapter->hw, E1000_MANC2H, manc2h);
3436 		E1000_WRITE_REG(&adapter->hw, E1000_MANC, manc);
3437 	}
3438 }
3439 
3440 /*
3441  * Give control back to hardware management
3442  * controller if there is one.
3443  */
3444 static void
3445 em_release_manageability(struct adapter *adapter)
3446 {
3447 	if (adapter->has_manage) {
3448 		int manc = E1000_READ_REG(&adapter->hw, E1000_MANC);
3449 
3450 		/* re-enable hardware interception of ARP */
3451 		manc |= E1000_MANC_ARP_EN;
3452 		manc &= ~E1000_MANC_EN_MNG2HOST;
3453 
3454 		E1000_WRITE_REG(&adapter->hw, E1000_MANC, manc);
3455 	}
3456 }
3457 
3458 /*
3459  * em_get_hw_control sets the {CTRL_EXT|FWSM}:DRV_LOAD bit.
3460  * For ASF and Pass Through versions of f/w this means
3461  * that the driver is loaded. For AMT version type f/w
3462  * this means that the network i/f is open.
3463  */
3464 static void
3465 em_get_hw_control(struct adapter *adapter)
3466 {
3467 	u32 ctrl_ext, swsm;
3468 
3469 	if (adapter->vf_ifp)
3470 		return;
3471 
3472 	if (adapter->hw.mac.type == e1000_82573) {
3473 		swsm = E1000_READ_REG(&adapter->hw, E1000_SWSM);
3474 		E1000_WRITE_REG(&adapter->hw, E1000_SWSM,
3475 		    swsm | E1000_SWSM_DRV_LOAD);
3476 		return;
3477 	}
3478 	/* else */
3479 	ctrl_ext = E1000_READ_REG(&adapter->hw, E1000_CTRL_EXT);
3480 	E1000_WRITE_REG(&adapter->hw, E1000_CTRL_EXT,
3481 	    ctrl_ext | E1000_CTRL_EXT_DRV_LOAD);
3482 }
3483 
3484 /*
3485  * em_release_hw_control resets {CTRL_EXT|FWSM}:DRV_LOAD bit.
3486  * For ASF and Pass Through versions of f/w this means that
3487  * the driver is no longer loaded. For AMT versions of the
3488  * f/w this means that the network i/f is closed.
3489  */
3490 static void
3491 em_release_hw_control(struct adapter *adapter)
3492 {
3493 	u32 ctrl_ext, swsm;
3494 
3495 	if (!adapter->has_manage)
3496 		return;
3497 
3498 	if (adapter->hw.mac.type == e1000_82573) {
3499 		swsm = E1000_READ_REG(&adapter->hw, E1000_SWSM);
3500 		E1000_WRITE_REG(&adapter->hw, E1000_SWSM,
3501 		    swsm & ~E1000_SWSM_DRV_LOAD);
3502 		return;
3503 	}
3504 	/* else */
3505 	ctrl_ext = E1000_READ_REG(&adapter->hw, E1000_CTRL_EXT);
3506 	E1000_WRITE_REG(&adapter->hw, E1000_CTRL_EXT,
3507 	    ctrl_ext & ~E1000_CTRL_EXT_DRV_LOAD);
3508 	return;
3509 }
3510 
3511 static int
3512 em_is_valid_ether_addr(u8 *addr)
3513 {
3514 	char zero_addr[6] = { 0, 0, 0, 0, 0, 0 };
3515 
3516 	if ((addr[0] & 1) || (!bcmp(addr, zero_addr, ETHER_ADDR_LEN))) {
3517 		return (FALSE);
3518 	}
3519 
3520 	return (TRUE);
3521 }
3522 
3523 /*
3524 ** Parse the interface capabilities with regard
3525 ** to both system management and wake-on-lan for
3526 ** later use.
3527 */
3528 static void
3529 em_get_wakeup(if_ctx_t ctx)
3530 {
3531 	struct adapter *adapter = iflib_get_softc(ctx);
3532 	device_t dev = iflib_get_dev(ctx);
3533 	u16 eeprom_data = 0, device_id, apme_mask;
3534 
3535 	adapter->has_manage = e1000_enable_mng_pass_thru(&adapter->hw);
3536 	apme_mask = EM_EEPROM_APME;
3537 
3538 	switch (adapter->hw.mac.type) {
3539 	case e1000_82542:
3540 	case e1000_82543:
3541 		break;
3542 	case e1000_82544:
3543 		e1000_read_nvm(&adapter->hw,
3544 		    NVM_INIT_CONTROL2_REG, 1, &eeprom_data);
3545 		apme_mask = EM_82544_APME;
3546 		break;
3547 	case e1000_82546:
3548 	case e1000_82546_rev_3:
3549 		if (adapter->hw.bus.func == 1) {
3550 			e1000_read_nvm(&adapter->hw,
3551 			    NVM_INIT_CONTROL3_PORT_B, 1, &eeprom_data);
3552 			break;
3553 		} else
3554 			e1000_read_nvm(&adapter->hw,
3555 			    NVM_INIT_CONTROL3_PORT_A, 1, &eeprom_data);
3556 		break;
3557 	case e1000_82573:
3558 	case e1000_82583:
3559 		adapter->has_amt = TRUE;
3560 		/* FALLTHROUGH */
3561 	case e1000_82571:
3562 	case e1000_82572:
3563 	case e1000_80003es2lan:
3564 		if (adapter->hw.bus.func == 1) {
3565 			e1000_read_nvm(&adapter->hw,
3566 			    NVM_INIT_CONTROL3_PORT_B, 1, &eeprom_data);
3567 			break;
3568 		} else
3569 			e1000_read_nvm(&adapter->hw,
3570 			    NVM_INIT_CONTROL3_PORT_A, 1, &eeprom_data);
3571 		break;
3572 	case e1000_ich8lan:
3573 	case e1000_ich9lan:
3574 	case e1000_ich10lan:
3575 	case e1000_pchlan:
3576 	case e1000_pch2lan:
3577 	case e1000_pch_lpt:
3578 	case e1000_pch_spt:
3579 	case e1000_82575:	/* listing all igb devices */
3580 	case e1000_82576:
3581 	case e1000_82580:
3582 	case e1000_i350:
3583 	case e1000_i354:
3584 	case e1000_i210:
3585 	case e1000_i211:
3586 	case e1000_vfadapt:
3587 	case e1000_vfadapt_i350:
3588 		apme_mask = E1000_WUC_APME;
3589 		adapter->has_amt = TRUE;
3590 		eeprom_data = E1000_READ_REG(&adapter->hw, E1000_WUC);
3591 		break;
3592 	default:
3593 		e1000_read_nvm(&adapter->hw,
3594 		    NVM_INIT_CONTROL3_PORT_A, 1, &eeprom_data);
3595 		break;
3596 	}
3597 	if (eeprom_data & apme_mask)
3598 		adapter->wol = (E1000_WUFC_MAG | E1000_WUFC_MC);
3599 	/*
3600 	 * We have the eeprom settings, now apply the special cases
3601 	 * where the eeprom may be wrong or the board won't support
3602 	 * wake on lan on a particular port
3603 	 */
3604 	device_id = pci_get_device(dev);
3605 	switch (device_id) {
3606 	case E1000_DEV_ID_82546GB_PCIE:
3607 		adapter->wol = 0;
3608 		break;
3609 	case E1000_DEV_ID_82546EB_FIBER:
3610 	case E1000_DEV_ID_82546GB_FIBER:
3611 		/* Wake events only supported on port A for dual fiber
3612 		 * regardless of eeprom setting */
3613 		if (E1000_READ_REG(&adapter->hw, E1000_STATUS) &
3614 		    E1000_STATUS_FUNC_1)
3615 			adapter->wol = 0;
3616 		break;
3617 	case E1000_DEV_ID_82546GB_QUAD_COPPER_KSP3:
3618 		/* if quad port adapter, disable WoL on all but port A */
3619 		if (global_quad_port_a != 0)
3620 			adapter->wol = 0;
3621 		/* Reset for multiple quad port adapters */
3622 		if (++global_quad_port_a == 4)
3623 			global_quad_port_a = 0;
3624 		break;
3625 	case E1000_DEV_ID_82571EB_FIBER:
3626 		/* Wake events only supported on port A for dual fiber
3627 		 * regardless of eeprom setting */
3628 		if (E1000_READ_REG(&adapter->hw, E1000_STATUS) &
3629 		    E1000_STATUS_FUNC_1)
3630 			adapter->wol = 0;
3631 		break;
3632 	case E1000_DEV_ID_82571EB_QUAD_COPPER:
3633 	case E1000_DEV_ID_82571EB_QUAD_FIBER:
3634 	case E1000_DEV_ID_82571EB_QUAD_COPPER_LP:
3635 		/* if quad port adapter, disable WoL on all but port A */
3636 		if (global_quad_port_a != 0)
3637 			adapter->wol = 0;
3638 		/* Reset for multiple quad port adapters */
3639 		if (++global_quad_port_a == 4)
3640 			global_quad_port_a = 0;
3641 		break;
3642 	}
3643 	return;
3644 }
3645 
3646 
3647 /*
3648  * Enable PCI Wake On Lan capability
3649  */
3650 static void
3651 em_enable_wakeup(if_ctx_t ctx)
3652 {
3653 	struct adapter *adapter = iflib_get_softc(ctx);
3654 	device_t dev = iflib_get_dev(ctx);
3655 	if_t ifp = iflib_get_ifp(ctx);
3656 	int error = 0;
3657 	u32 pmc, ctrl, ctrl_ext, rctl;
3658 	u16 status;
3659 
3660 	if (pci_find_cap(dev, PCIY_PMG, &pmc) != 0)
3661 		return;
3662 
3663 	/*
3664 	 * Determine type of Wakeup: note that wol
3665 	 * is set with all bits on by default.
3666 	 */
3667 	if ((if_getcapenable(ifp) & IFCAP_WOL_MAGIC) == 0)
3668 		adapter->wol &= ~E1000_WUFC_MAG;
3669 
3670 	if ((if_getcapenable(ifp) & IFCAP_WOL_UCAST) == 0)
3671 		adapter->wol &= ~E1000_WUFC_EX;
3672 
3673 	if ((if_getcapenable(ifp) & IFCAP_WOL_MCAST) == 0)
3674 		adapter->wol &= ~E1000_WUFC_MC;
3675 	else {
3676 		rctl = E1000_READ_REG(&adapter->hw, E1000_RCTL);
3677 		rctl |= E1000_RCTL_MPE;
3678 		E1000_WRITE_REG(&adapter->hw, E1000_RCTL, rctl);
3679 	}
3680 
3681 	if (!(adapter->wol & (E1000_WUFC_EX | E1000_WUFC_MAG | E1000_WUFC_MC)))
3682 		goto pme;
3683 
3684 	/* Advertise the wakeup capability */
3685 	ctrl = E1000_READ_REG(&adapter->hw, E1000_CTRL);
3686 	ctrl |= (E1000_CTRL_SWDPIN2 | E1000_CTRL_SWDPIN3);
3687 	E1000_WRITE_REG(&adapter->hw, E1000_CTRL, ctrl);
3688 
3689 	/* Keep the laser running on Fiber adapters */
3690 	if (adapter->hw.phy.media_type == e1000_media_type_fiber ||
3691 	    adapter->hw.phy.media_type == e1000_media_type_internal_serdes) {
3692 		ctrl_ext = E1000_READ_REG(&adapter->hw, E1000_CTRL_EXT);
3693 		ctrl_ext |= E1000_CTRL_EXT_SDP3_DATA;
3694 		E1000_WRITE_REG(&adapter->hw, E1000_CTRL_EXT, ctrl_ext);
3695 	}
3696 
3697 	if ((adapter->hw.mac.type == e1000_ich8lan) ||
3698 	    (adapter->hw.mac.type == e1000_pchlan) ||
3699 	    (adapter->hw.mac.type == e1000_ich9lan) ||
3700 	    (adapter->hw.mac.type == e1000_ich10lan))
3701 		e1000_suspend_workarounds_ich8lan(&adapter->hw);
3702 
3703 	if ( adapter->hw.mac.type >= e1000_pchlan) {
3704 		error = em_enable_phy_wakeup(adapter);
3705 		if (error)
3706 			goto pme;
3707 	} else {
3708 		/* Enable wakeup by the MAC */
3709 		E1000_WRITE_REG(&adapter->hw, E1000_WUC, E1000_WUC_PME_EN);
3710 		E1000_WRITE_REG(&adapter->hw, E1000_WUFC, adapter->wol);
3711 	}
3712 
3713 	if (adapter->hw.phy.type == e1000_phy_igp_3)
3714 		e1000_igp3_phy_powerdown_workaround_ich8lan(&adapter->hw);
3715 
3716 pme:
3717 	status = pci_read_config(dev, pmc + PCIR_POWER_STATUS, 2);
3718 	status &= ~(PCIM_PSTAT_PME | PCIM_PSTAT_PMEENABLE);
3719 	if (!error && (if_getcapenable(ifp) & IFCAP_WOL))
3720 		status |= PCIM_PSTAT_PME | PCIM_PSTAT_PMEENABLE;
3721 	pci_write_config(dev, pmc + PCIR_POWER_STATUS, status, 2);
3722 
3723 	return;
3724 }
3725 
3726 /*
3727  * WOL in the newer chipset interfaces (pchlan)
3728  * require thing to be copied into the phy
3729  */
3730 static int
3731 em_enable_phy_wakeup(struct adapter *adapter)
3732 {
3733 	struct e1000_hw *hw = &adapter->hw;
3734 	u32 mreg, ret = 0;
3735 	u16 preg;
3736 
3737 	/* copy MAC RARs to PHY RARs */
3738 	e1000_copy_rx_addrs_to_phy_ich8lan(hw);
3739 
3740 	/* copy MAC MTA to PHY MTA */
3741 	for (int i = 0; i < adapter->hw.mac.mta_reg_count; i++) {
3742 		mreg = E1000_READ_REG_ARRAY(hw, E1000_MTA, i);
3743 		e1000_write_phy_reg(hw, BM_MTA(i), (u16)(mreg & 0xFFFF));
3744 		e1000_write_phy_reg(hw, BM_MTA(i) + 1,
3745 		    (u16)((mreg >> 16) & 0xFFFF));
3746 	}
3747 
3748 	/* configure PHY Rx Control register */
3749 	e1000_read_phy_reg(&adapter->hw, BM_RCTL, &preg);
3750 	mreg = E1000_READ_REG(hw, E1000_RCTL);
3751 	if (mreg & E1000_RCTL_UPE)
3752 		preg |= BM_RCTL_UPE;
3753 	if (mreg & E1000_RCTL_MPE)
3754 		preg |= BM_RCTL_MPE;
3755 	preg &= ~(BM_RCTL_MO_MASK);
3756 	if (mreg & E1000_RCTL_MO_3)
3757 		preg |= (((mreg & E1000_RCTL_MO_3) >> E1000_RCTL_MO_SHIFT)
3758 				<< BM_RCTL_MO_SHIFT);
3759 	if (mreg & E1000_RCTL_BAM)
3760 		preg |= BM_RCTL_BAM;
3761 	if (mreg & E1000_RCTL_PMCF)
3762 		preg |= BM_RCTL_PMCF;
3763 	mreg = E1000_READ_REG(hw, E1000_CTRL);
3764 	if (mreg & E1000_CTRL_RFCE)
3765 		preg |= BM_RCTL_RFCE;
3766 	e1000_write_phy_reg(&adapter->hw, BM_RCTL, preg);
3767 
3768 	/* enable PHY wakeup in MAC register */
3769 	E1000_WRITE_REG(hw, E1000_WUC,
3770 	    E1000_WUC_PHY_WAKE | E1000_WUC_PME_EN | E1000_WUC_APME);
3771 	E1000_WRITE_REG(hw, E1000_WUFC, adapter->wol);
3772 
3773 	/* configure and enable PHY wakeup in PHY registers */
3774 	e1000_write_phy_reg(&adapter->hw, BM_WUFC, adapter->wol);
3775 	e1000_write_phy_reg(&adapter->hw, BM_WUC, E1000_WUC_PME_EN);
3776 
3777 	/* activate PHY wakeup */
3778 	ret = hw->phy.ops.acquire(hw);
3779 	if (ret) {
3780 		printf("Could not acquire PHY\n");
3781 		return ret;
3782 	}
3783 	e1000_write_phy_reg_mdic(hw, IGP01E1000_PHY_PAGE_SELECT,
3784 	                         (BM_WUC_ENABLE_PAGE << IGP_PAGE_SHIFT));
3785 	ret = e1000_read_phy_reg_mdic(hw, BM_WUC_ENABLE_REG, &preg);
3786 	if (ret) {
3787 		printf("Could not read PHY page 769\n");
3788 		goto out;
3789 	}
3790 	preg |= BM_WUC_ENABLE_BIT | BM_WUC_HOST_WU_BIT;
3791 	ret = e1000_write_phy_reg_mdic(hw, BM_WUC_ENABLE_REG, preg);
3792 	if (ret)
3793 		printf("Could not set PHY Host Wakeup bit\n");
3794 out:
3795 	hw->phy.ops.release(hw);
3796 
3797 	return ret;
3798 }
3799 
3800 static void
3801 em_if_led_func(if_ctx_t ctx, int onoff)
3802 {
3803 	struct adapter *adapter = iflib_get_softc(ctx);
3804 
3805 	if (onoff) {
3806 		e1000_setup_led(&adapter->hw);
3807 		e1000_led_on(&adapter->hw);
3808 	} else {
3809 		e1000_led_off(&adapter->hw);
3810 		e1000_cleanup_led(&adapter->hw);
3811 	}
3812 }
3813 
3814 /*
3815  * Disable the L0S and L1 LINK states
3816  */
3817 static void
3818 em_disable_aspm(struct adapter *adapter)
3819 {
3820 	int base, reg;
3821 	u16 link_cap,link_ctrl;
3822 	device_t dev = adapter->dev;
3823 
3824 	switch (adapter->hw.mac.type) {
3825 	case e1000_82573:
3826 	case e1000_82574:
3827 	case e1000_82583:
3828 		break;
3829 	default:
3830 		return;
3831 	}
3832 	if (pci_find_cap(dev, PCIY_EXPRESS, &base) != 0)
3833 		return;
3834 	reg = base + PCIER_LINK_CAP;
3835 	link_cap = pci_read_config(dev, reg, 2);
3836 	if ((link_cap & PCIEM_LINK_CAP_ASPM) == 0)
3837 		return;
3838 	reg = base + PCIER_LINK_CTL;
3839 	link_ctrl = pci_read_config(dev, reg, 2);
3840 	link_ctrl &= ~PCIEM_LINK_CTL_ASPMC;
3841 	pci_write_config(dev, reg, link_ctrl, 2);
3842 	return;
3843 }
3844 
3845 /**********************************************************************
3846  *
3847  *  Update the board statistics counters.
3848  *
3849  **********************************************************************/
3850 static void
3851 em_update_stats_counters(struct adapter *adapter)
3852 {
3853 
3854 	if(adapter->hw.phy.media_type == e1000_media_type_copper ||
3855 	   (E1000_READ_REG(&adapter->hw, E1000_STATUS) & E1000_STATUS_LU)) {
3856 		adapter->stats.symerrs += E1000_READ_REG(&adapter->hw, E1000_SYMERRS);
3857 		adapter->stats.sec += E1000_READ_REG(&adapter->hw, E1000_SEC);
3858 	}
3859 	adapter->stats.crcerrs += E1000_READ_REG(&adapter->hw, E1000_CRCERRS);
3860 	adapter->stats.mpc += E1000_READ_REG(&adapter->hw, E1000_MPC);
3861 	adapter->stats.scc += E1000_READ_REG(&adapter->hw, E1000_SCC);
3862 	adapter->stats.ecol += E1000_READ_REG(&adapter->hw, E1000_ECOL);
3863 
3864 	adapter->stats.mcc += E1000_READ_REG(&adapter->hw, E1000_MCC);
3865 	adapter->stats.latecol += E1000_READ_REG(&adapter->hw, E1000_LATECOL);
3866 	adapter->stats.colc += E1000_READ_REG(&adapter->hw, E1000_COLC);
3867 	adapter->stats.dc += E1000_READ_REG(&adapter->hw, E1000_DC);
3868 	adapter->stats.rlec += E1000_READ_REG(&adapter->hw, E1000_RLEC);
3869 	adapter->stats.xonrxc += E1000_READ_REG(&adapter->hw, E1000_XONRXC);
3870 	adapter->stats.xontxc += E1000_READ_REG(&adapter->hw, E1000_XONTXC);
3871 	adapter->stats.xoffrxc += E1000_READ_REG(&adapter->hw, E1000_XOFFRXC);
3872 	/*
3873 	 ** For watchdog management we need to know if we have been
3874 	 ** paused during the last interval, so capture that here.
3875 	*/
3876 	adapter->shared->isc_pause_frames = adapter->stats.xoffrxc;
3877 	adapter->stats.xofftxc += E1000_READ_REG(&adapter->hw, E1000_XOFFTXC);
3878 	adapter->stats.fcruc += E1000_READ_REG(&adapter->hw, E1000_FCRUC);
3879 	adapter->stats.prc64 += E1000_READ_REG(&adapter->hw, E1000_PRC64);
3880 	adapter->stats.prc127 += E1000_READ_REG(&adapter->hw, E1000_PRC127);
3881 	adapter->stats.prc255 += E1000_READ_REG(&adapter->hw, E1000_PRC255);
3882 	adapter->stats.prc511 += E1000_READ_REG(&adapter->hw, E1000_PRC511);
3883 	adapter->stats.prc1023 += E1000_READ_REG(&adapter->hw, E1000_PRC1023);
3884 	adapter->stats.prc1522 += E1000_READ_REG(&adapter->hw, E1000_PRC1522);
3885 	adapter->stats.gprc += E1000_READ_REG(&adapter->hw, E1000_GPRC);
3886 	adapter->stats.bprc += E1000_READ_REG(&adapter->hw, E1000_BPRC);
3887 	adapter->stats.mprc += E1000_READ_REG(&adapter->hw, E1000_MPRC);
3888 	adapter->stats.gptc += E1000_READ_REG(&adapter->hw, E1000_GPTC);
3889 
3890 	/* For the 64-bit byte counters the low dword must be read first. */
3891 	/* Both registers clear on the read of the high dword */
3892 
3893 	adapter->stats.gorc += E1000_READ_REG(&adapter->hw, E1000_GORCL) +
3894 	    ((u64)E1000_READ_REG(&adapter->hw, E1000_GORCH) << 32);
3895 	adapter->stats.gotc += E1000_READ_REG(&adapter->hw, E1000_GOTCL) +
3896 	    ((u64)E1000_READ_REG(&adapter->hw, E1000_GOTCH) << 32);
3897 
3898 	adapter->stats.rnbc += E1000_READ_REG(&adapter->hw, E1000_RNBC);
3899 	adapter->stats.ruc += E1000_READ_REG(&adapter->hw, E1000_RUC);
3900 	adapter->stats.rfc += E1000_READ_REG(&adapter->hw, E1000_RFC);
3901 	adapter->stats.roc += E1000_READ_REG(&adapter->hw, E1000_ROC);
3902 	adapter->stats.rjc += E1000_READ_REG(&adapter->hw, E1000_RJC);
3903 
3904 	adapter->stats.tor += E1000_READ_REG(&adapter->hw, E1000_TORH);
3905 	adapter->stats.tot += E1000_READ_REG(&adapter->hw, E1000_TOTH);
3906 
3907 	adapter->stats.tpr += E1000_READ_REG(&adapter->hw, E1000_TPR);
3908 	adapter->stats.tpt += E1000_READ_REG(&adapter->hw, E1000_TPT);
3909 	adapter->stats.ptc64 += E1000_READ_REG(&adapter->hw, E1000_PTC64);
3910 	adapter->stats.ptc127 += E1000_READ_REG(&adapter->hw, E1000_PTC127);
3911 	adapter->stats.ptc255 += E1000_READ_REG(&adapter->hw, E1000_PTC255);
3912 	adapter->stats.ptc511 += E1000_READ_REG(&adapter->hw, E1000_PTC511);
3913 	adapter->stats.ptc1023 += E1000_READ_REG(&adapter->hw, E1000_PTC1023);
3914 	adapter->stats.ptc1522 += E1000_READ_REG(&adapter->hw, E1000_PTC1522);
3915 	adapter->stats.mptc += E1000_READ_REG(&adapter->hw, E1000_MPTC);
3916 	adapter->stats.bptc += E1000_READ_REG(&adapter->hw, E1000_BPTC);
3917 
3918 	/* Interrupt Counts */
3919 
3920 	adapter->stats.iac += E1000_READ_REG(&adapter->hw, E1000_IAC);
3921 	adapter->stats.icrxptc += E1000_READ_REG(&adapter->hw, E1000_ICRXPTC);
3922 	adapter->stats.icrxatc += E1000_READ_REG(&adapter->hw, E1000_ICRXATC);
3923 	adapter->stats.ictxptc += E1000_READ_REG(&adapter->hw, E1000_ICTXPTC);
3924 	adapter->stats.ictxatc += E1000_READ_REG(&adapter->hw, E1000_ICTXATC);
3925 	adapter->stats.ictxqec += E1000_READ_REG(&adapter->hw, E1000_ICTXQEC);
3926 	adapter->stats.ictxqmtc += E1000_READ_REG(&adapter->hw, E1000_ICTXQMTC);
3927 	adapter->stats.icrxdmtc += E1000_READ_REG(&adapter->hw, E1000_ICRXDMTC);
3928 	adapter->stats.icrxoc += E1000_READ_REG(&adapter->hw, E1000_ICRXOC);
3929 
3930 	if (adapter->hw.mac.type >= e1000_82543) {
3931 		adapter->stats.algnerrc +=
3932 		E1000_READ_REG(&adapter->hw, E1000_ALGNERRC);
3933 		adapter->stats.rxerrc +=
3934 		E1000_READ_REG(&adapter->hw, E1000_RXERRC);
3935 		adapter->stats.tncrs +=
3936 		E1000_READ_REG(&adapter->hw, E1000_TNCRS);
3937 		adapter->stats.cexterr +=
3938 		E1000_READ_REG(&adapter->hw, E1000_CEXTERR);
3939 		adapter->stats.tsctc +=
3940 		E1000_READ_REG(&adapter->hw, E1000_TSCTC);
3941 		adapter->stats.tsctfc +=
3942 		E1000_READ_REG(&adapter->hw, E1000_TSCTFC);
3943 	}
3944 }
3945 
3946 static uint64_t
3947 em_if_get_counter(if_ctx_t ctx, ift_counter cnt)
3948 {
3949 	struct adapter *adapter = iflib_get_softc(ctx);
3950 	struct ifnet *ifp = iflib_get_ifp(ctx);
3951 
3952 	switch (cnt) {
3953 	case IFCOUNTER_COLLISIONS:
3954 		return (adapter->stats.colc);
3955 	case IFCOUNTER_IERRORS:
3956 		return (adapter->dropped_pkts + adapter->stats.rxerrc +
3957 		    adapter->stats.crcerrs + adapter->stats.algnerrc +
3958 		    adapter->stats.ruc + adapter->stats.roc +
3959 		    adapter->stats.mpc + adapter->stats.cexterr);
3960 	case IFCOUNTER_OERRORS:
3961 		return (adapter->stats.ecol + adapter->stats.latecol +
3962 		    adapter->watchdog_events);
3963 	default:
3964 		return (if_get_counter_default(ifp, cnt));
3965 	}
3966 }
3967 
3968 /* Export a single 32-bit register via a read-only sysctl. */
3969 static int
3970 em_sysctl_reg_handler(SYSCTL_HANDLER_ARGS)
3971 {
3972 	struct adapter *adapter;
3973 	u_int val;
3974 
3975 	adapter = oidp->oid_arg1;
3976 	val = E1000_READ_REG(&adapter->hw, oidp->oid_arg2);
3977 	return (sysctl_handle_int(oidp, &val, 0, req));
3978 }
3979 
3980 /*
3981  * Add sysctl variables, one per statistic, to the system.
3982  */
3983 static void
3984 em_add_hw_stats(struct adapter *adapter)
3985 {
3986 	device_t dev = iflib_get_dev(adapter->ctx);
3987 	struct em_tx_queue *tx_que = adapter->tx_queues;
3988 	struct em_rx_queue *rx_que = adapter->rx_queues;
3989 
3990 	struct sysctl_ctx_list *ctx = device_get_sysctl_ctx(dev);
3991 	struct sysctl_oid *tree = device_get_sysctl_tree(dev);
3992 	struct sysctl_oid_list *child = SYSCTL_CHILDREN(tree);
3993 	struct e1000_hw_stats *stats = &adapter->stats;
3994 
3995 	struct sysctl_oid *stat_node, *queue_node, *int_node;
3996 	struct sysctl_oid_list *stat_list, *queue_list, *int_list;
3997 
3998 #define QUEUE_NAME_LEN 32
3999 	char namebuf[QUEUE_NAME_LEN];
4000 
4001 	/* Driver Statistics */
4002 	SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "dropped",
4003 			CTLFLAG_RD, &adapter->dropped_pkts,
4004 			"Driver dropped packets");
4005 	SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "link_irq",
4006 			CTLFLAG_RD, &adapter->link_irq,
4007 			"Link MSI-X IRQ Handled");
4008 	SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "rx_overruns",
4009 			CTLFLAG_RD, &adapter->rx_overruns,
4010 			"RX overruns");
4011 	SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "watchdog_timeouts",
4012 			CTLFLAG_RD, &adapter->watchdog_events,
4013 			"Watchdog timeouts");
4014 	SYSCTL_ADD_PROC(ctx, child, OID_AUTO, "device_control",
4015 			CTLTYPE_UINT | CTLFLAG_RD, adapter, E1000_CTRL,
4016 			em_sysctl_reg_handler, "IU",
4017 			"Device Control Register");
4018 	SYSCTL_ADD_PROC(ctx, child, OID_AUTO, "rx_control",
4019 			CTLTYPE_UINT | CTLFLAG_RD, adapter, E1000_RCTL,
4020 			em_sysctl_reg_handler, "IU",
4021 			"Receiver Control Register");
4022 	SYSCTL_ADD_UINT(ctx, child, OID_AUTO, "fc_high_water",
4023 			CTLFLAG_RD, &adapter->hw.fc.high_water, 0,
4024 			"Flow Control High Watermark");
4025 	SYSCTL_ADD_UINT(ctx, child, OID_AUTO, "fc_low_water",
4026 			CTLFLAG_RD, &adapter->hw.fc.low_water, 0,
4027 			"Flow Control Low Watermark");
4028 
4029 	for (int i = 0; i < adapter->tx_num_queues; i++, tx_que++) {
4030 		struct tx_ring *txr = &tx_que->txr;
4031 		snprintf(namebuf, QUEUE_NAME_LEN, "queue_tx_%d", i);
4032 		queue_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, namebuf,
4033 					    CTLFLAG_RD, NULL, "TX Queue Name");
4034 		queue_list = SYSCTL_CHILDREN(queue_node);
4035 
4036 		SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "txd_head",
4037 				CTLTYPE_UINT | CTLFLAG_RD, adapter,
4038 				E1000_TDH(txr->me),
4039 				em_sysctl_reg_handler, "IU",
4040 				"Transmit Descriptor Head");
4041 		SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "txd_tail",
4042 				CTLTYPE_UINT | CTLFLAG_RD, adapter,
4043 				E1000_TDT(txr->me),
4044 				em_sysctl_reg_handler, "IU",
4045 				"Transmit Descriptor Tail");
4046 		SYSCTL_ADD_ULONG(ctx, queue_list, OID_AUTO, "tx_irq",
4047 				CTLFLAG_RD, &txr->tx_irq,
4048 				"Queue MSI-X Transmit Interrupts");
4049 	}
4050 
4051 	for (int j = 0; j < adapter->rx_num_queues; j++, rx_que++) {
4052 		struct rx_ring *rxr = &rx_que->rxr;
4053 		snprintf(namebuf, QUEUE_NAME_LEN, "queue_rx_%d", j);
4054 		queue_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, namebuf,
4055 					    CTLFLAG_RD, NULL, "RX Queue Name");
4056 		queue_list = SYSCTL_CHILDREN(queue_node);
4057 
4058 		SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "rxd_head",
4059 				CTLTYPE_UINT | CTLFLAG_RD, adapter,
4060 				E1000_RDH(rxr->me),
4061 				em_sysctl_reg_handler, "IU",
4062 				"Receive Descriptor Head");
4063 		SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "rxd_tail",
4064 				CTLTYPE_UINT | CTLFLAG_RD, adapter,
4065 				E1000_RDT(rxr->me),
4066 				em_sysctl_reg_handler, "IU",
4067 				"Receive Descriptor Tail");
4068 		SYSCTL_ADD_ULONG(ctx, queue_list, OID_AUTO, "rx_irq",
4069 				CTLFLAG_RD, &rxr->rx_irq,
4070 				"Queue MSI-X Receive Interrupts");
4071 	}
4072 
4073 	/* MAC stats get their own sub node */
4074 
4075 	stat_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, "mac_stats",
4076 				    CTLFLAG_RD, NULL, "Statistics");
4077 	stat_list = SYSCTL_CHILDREN(stat_node);
4078 
4079 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "excess_coll",
4080 			CTLFLAG_RD, &stats->ecol,
4081 			"Excessive collisions");
4082 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "single_coll",
4083 			CTLFLAG_RD, &stats->scc,
4084 			"Single collisions");
4085 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "multiple_coll",
4086 			CTLFLAG_RD, &stats->mcc,
4087 			"Multiple collisions");
4088 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "late_coll",
4089 			CTLFLAG_RD, &stats->latecol,
4090 			"Late collisions");
4091 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "collision_count",
4092 			CTLFLAG_RD, &stats->colc,
4093 			"Collision Count");
4094 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "symbol_errors",
4095 			CTLFLAG_RD, &adapter->stats.symerrs,
4096 			"Symbol Errors");
4097 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "sequence_errors",
4098 			CTLFLAG_RD, &adapter->stats.sec,
4099 			"Sequence Errors");
4100 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "defer_count",
4101 			CTLFLAG_RD, &adapter->stats.dc,
4102 			"Defer Count");
4103 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "missed_packets",
4104 			CTLFLAG_RD, &adapter->stats.mpc,
4105 			"Missed Packets");
4106 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_no_buff",
4107 			CTLFLAG_RD, &adapter->stats.rnbc,
4108 			"Receive No Buffers");
4109 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_undersize",
4110 			CTLFLAG_RD, &adapter->stats.ruc,
4111 			"Receive Undersize");
4112 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_fragmented",
4113 			CTLFLAG_RD, &adapter->stats.rfc,
4114 			"Fragmented Packets Received ");
4115 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_oversize",
4116 			CTLFLAG_RD, &adapter->stats.roc,
4117 			"Oversized Packets Received");
4118 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_jabber",
4119 			CTLFLAG_RD, &adapter->stats.rjc,
4120 			"Recevied Jabber");
4121 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_errs",
4122 			CTLFLAG_RD, &adapter->stats.rxerrc,
4123 			"Receive Errors");
4124 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "crc_errs",
4125 			CTLFLAG_RD, &adapter->stats.crcerrs,
4126 			"CRC errors");
4127 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "alignment_errs",
4128 			CTLFLAG_RD, &adapter->stats.algnerrc,
4129 			"Alignment Errors");
4130 	/* On 82575 these are collision counts */
4131 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "coll_ext_errs",
4132 			CTLFLAG_RD, &adapter->stats.cexterr,
4133 			"Collision/Carrier extension errors");
4134 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xon_recvd",
4135 			CTLFLAG_RD, &adapter->stats.xonrxc,
4136 			"XON Received");
4137 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xon_txd",
4138 			CTLFLAG_RD, &adapter->stats.xontxc,
4139 			"XON Transmitted");
4140 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xoff_recvd",
4141 			CTLFLAG_RD, &adapter->stats.xoffrxc,
4142 			"XOFF Received");
4143 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xoff_txd",
4144 			CTLFLAG_RD, &adapter->stats.xofftxc,
4145 			"XOFF Transmitted");
4146 
4147 	/* Packet Reception Stats */
4148 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "total_pkts_recvd",
4149 			CTLFLAG_RD, &adapter->stats.tpr,
4150 			"Total Packets Received ");
4151 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_pkts_recvd",
4152 			CTLFLAG_RD, &adapter->stats.gprc,
4153 			"Good Packets Received");
4154 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "bcast_pkts_recvd",
4155 			CTLFLAG_RD, &adapter->stats.bprc,
4156 			"Broadcast Packets Received");
4157 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "mcast_pkts_recvd",
4158 			CTLFLAG_RD, &adapter->stats.mprc,
4159 			"Multicast Packets Received");
4160 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_64",
4161 			CTLFLAG_RD, &adapter->stats.prc64,
4162 			"64 byte frames received ");
4163 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_65_127",
4164 			CTLFLAG_RD, &adapter->stats.prc127,
4165 			"65-127 byte frames received");
4166 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_128_255",
4167 			CTLFLAG_RD, &adapter->stats.prc255,
4168 			"128-255 byte frames received");
4169 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_256_511",
4170 			CTLFLAG_RD, &adapter->stats.prc511,
4171 			"256-511 byte frames received");
4172 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_512_1023",
4173 			CTLFLAG_RD, &adapter->stats.prc1023,
4174 			"512-1023 byte frames received");
4175 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_1024_1522",
4176 			CTLFLAG_RD, &adapter->stats.prc1522,
4177 			"1023-1522 byte frames received");
4178 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_octets_recvd",
4179 			CTLFLAG_RD, &adapter->stats.gorc,
4180 			"Good Octets Received");
4181 
4182 	/* Packet Transmission Stats */
4183 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_octets_txd",
4184 			CTLFLAG_RD, &adapter->stats.gotc,
4185 			"Good Octets Transmitted");
4186 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "total_pkts_txd",
4187 			CTLFLAG_RD, &adapter->stats.tpt,
4188 			"Total Packets Transmitted");
4189 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_pkts_txd",
4190 			CTLFLAG_RD, &adapter->stats.gptc,
4191 			"Good Packets Transmitted");
4192 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "bcast_pkts_txd",
4193 			CTLFLAG_RD, &adapter->stats.bptc,
4194 			"Broadcast Packets Transmitted");
4195 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "mcast_pkts_txd",
4196 			CTLFLAG_RD, &adapter->stats.mptc,
4197 			"Multicast Packets Transmitted");
4198 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_64",
4199 			CTLFLAG_RD, &adapter->stats.ptc64,
4200 			"64 byte frames transmitted ");
4201 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_65_127",
4202 			CTLFLAG_RD, &adapter->stats.ptc127,
4203 			"65-127 byte frames transmitted");
4204 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_128_255",
4205 			CTLFLAG_RD, &adapter->stats.ptc255,
4206 			"128-255 byte frames transmitted");
4207 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_256_511",
4208 			CTLFLAG_RD, &adapter->stats.ptc511,
4209 			"256-511 byte frames transmitted");
4210 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_512_1023",
4211 			CTLFLAG_RD, &adapter->stats.ptc1023,
4212 			"512-1023 byte frames transmitted");
4213 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_1024_1522",
4214 			CTLFLAG_RD, &adapter->stats.ptc1522,
4215 			"1024-1522 byte frames transmitted");
4216 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tso_txd",
4217 			CTLFLAG_RD, &adapter->stats.tsctc,
4218 			"TSO Contexts Transmitted");
4219 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tso_ctx_fail",
4220 			CTLFLAG_RD, &adapter->stats.tsctfc,
4221 			"TSO Contexts Failed");
4222 
4223 
4224 	/* Interrupt Stats */
4225 
4226 	int_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, "interrupts",
4227 				    CTLFLAG_RD, NULL, "Interrupt Statistics");
4228 	int_list = SYSCTL_CHILDREN(int_node);
4229 
4230 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "asserts",
4231 			CTLFLAG_RD, &adapter->stats.iac,
4232 			"Interrupt Assertion Count");
4233 
4234 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "rx_pkt_timer",
4235 			CTLFLAG_RD, &adapter->stats.icrxptc,
4236 			"Interrupt Cause Rx Pkt Timer Expire Count");
4237 
4238 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "rx_abs_timer",
4239 			CTLFLAG_RD, &adapter->stats.icrxatc,
4240 			"Interrupt Cause Rx Abs Timer Expire Count");
4241 
4242 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "tx_pkt_timer",
4243 			CTLFLAG_RD, &adapter->stats.ictxptc,
4244 			"Interrupt Cause Tx Pkt Timer Expire Count");
4245 
4246 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "tx_abs_timer",
4247 			CTLFLAG_RD, &adapter->stats.ictxatc,
4248 			"Interrupt Cause Tx Abs Timer Expire Count");
4249 
4250 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "tx_queue_empty",
4251 			CTLFLAG_RD, &adapter->stats.ictxqec,
4252 			"Interrupt Cause Tx Queue Empty Count");
4253 
4254 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "tx_queue_min_thresh",
4255 			CTLFLAG_RD, &adapter->stats.ictxqmtc,
4256 			"Interrupt Cause Tx Queue Min Thresh Count");
4257 
4258 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "rx_desc_min_thresh",
4259 			CTLFLAG_RD, &adapter->stats.icrxdmtc,
4260 			"Interrupt Cause Rx Desc Min Thresh Count");
4261 
4262 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "rx_overrun",
4263 			CTLFLAG_RD, &adapter->stats.icrxoc,
4264 			"Interrupt Cause Receiver Overrun Count");
4265 }
4266 
4267 /**********************************************************************
4268  *
4269  *  This routine provides a way to dump out the adapter eeprom,
4270  *  often a useful debug/service tool. This only dumps the first
4271  *  32 words, stuff that matters is in that extent.
4272  *
4273  **********************************************************************/
4274 static int
4275 em_sysctl_nvm_info(SYSCTL_HANDLER_ARGS)
4276 {
4277 	struct adapter *adapter = (struct adapter *)arg1;
4278 	int error;
4279 	int result;
4280 
4281 	result = -1;
4282 	error = sysctl_handle_int(oidp, &result, 0, req);
4283 
4284 	if (error || !req->newptr)
4285 		return (error);
4286 
4287 	/*
4288 	 * This value will cause a hex dump of the
4289 	 * first 32 16-bit words of the EEPROM to
4290 	 * the screen.
4291 	 */
4292 	if (result == 1)
4293 		em_print_nvm_info(adapter);
4294 
4295 	return (error);
4296 }
4297 
4298 static void
4299 em_print_nvm_info(struct adapter *adapter)
4300 {
4301 	u16 eeprom_data;
4302 	int i, j, row = 0;
4303 
4304 	/* Its a bit crude, but it gets the job done */
4305 	printf("\nInterface EEPROM Dump:\n");
4306 	printf("Offset\n0x0000  ");
4307 	for (i = 0, j = 0; i < 32; i++, j++) {
4308 		if (j == 8) { /* Make the offset block */
4309 			j = 0; ++row;
4310 			printf("\n0x00%x0  ",row);
4311 		}
4312 		e1000_read_nvm(&adapter->hw, i, 1, &eeprom_data);
4313 		printf("%04x ", eeprom_data);
4314 	}
4315 	printf("\n");
4316 }
4317 
4318 static int
4319 em_sysctl_int_delay(SYSCTL_HANDLER_ARGS)
4320 {
4321 	struct em_int_delay_info *info;
4322 	struct adapter *adapter;
4323 	u32 regval;
4324 	int error, usecs, ticks;
4325 
4326 	info = (struct em_int_delay_info *) arg1;
4327 	usecs = info->value;
4328 	error = sysctl_handle_int(oidp, &usecs, 0, req);
4329 	if (error != 0 || req->newptr == NULL)
4330 		return (error);
4331 	if (usecs < 0 || usecs > EM_TICKS_TO_USECS(65535))
4332 		return (EINVAL);
4333 	info->value = usecs;
4334 	ticks = EM_USECS_TO_TICKS(usecs);
4335 	if (info->offset == E1000_ITR)	/* units are 256ns here */
4336 		ticks *= 4;
4337 
4338 	adapter = info->adapter;
4339 
4340 	regval = E1000_READ_OFFSET(&adapter->hw, info->offset);
4341 	regval = (regval & ~0xffff) | (ticks & 0xffff);
4342 	/* Handle a few special cases. */
4343 	switch (info->offset) {
4344 	case E1000_RDTR:
4345 		break;
4346 	case E1000_TIDV:
4347 		if (ticks == 0) {
4348 			adapter->txd_cmd &= ~E1000_TXD_CMD_IDE;
4349 			/* Don't write 0 into the TIDV register. */
4350 			regval++;
4351 		} else
4352 			adapter->txd_cmd |= E1000_TXD_CMD_IDE;
4353 		break;
4354 	}
4355 	E1000_WRITE_OFFSET(&adapter->hw, info->offset, regval);
4356 	return (0);
4357 }
4358 
4359 static void
4360 em_add_int_delay_sysctl(struct adapter *adapter, const char *name,
4361 	const char *description, struct em_int_delay_info *info,
4362 	int offset, int value)
4363 {
4364 	info->adapter = adapter;
4365 	info->offset = offset;
4366 	info->value = value;
4367 	SYSCTL_ADD_PROC(device_get_sysctl_ctx(adapter->dev),
4368 	    SYSCTL_CHILDREN(device_get_sysctl_tree(adapter->dev)),
4369 	    OID_AUTO, name, CTLTYPE_INT|CTLFLAG_RW,
4370 	    info, 0, em_sysctl_int_delay, "I", description);
4371 }
4372 
4373 /*
4374  * Set flow control using sysctl:
4375  * Flow control values:
4376  *      0 - off
4377  *      1 - rx pause
4378  *      2 - tx pause
4379  *      3 - full
4380  */
4381 static int
4382 em_set_flowcntl(SYSCTL_HANDLER_ARGS)
4383 {
4384 	int error;
4385 	static int input = 3; /* default is full */
4386 	struct adapter	*adapter = (struct adapter *) arg1;
4387 
4388 	error = sysctl_handle_int(oidp, &input, 0, req);
4389 
4390 	if ((error) || (req->newptr == NULL))
4391 		return (error);
4392 
4393 	if (input == adapter->fc) /* no change? */
4394 		return (error);
4395 
4396 	switch (input) {
4397 	case e1000_fc_rx_pause:
4398 	case e1000_fc_tx_pause:
4399 	case e1000_fc_full:
4400 	case e1000_fc_none:
4401 		adapter->hw.fc.requested_mode = input;
4402 		adapter->fc = input;
4403 		break;
4404 	default:
4405 		/* Do nothing */
4406 		return (error);
4407 	}
4408 
4409 	adapter->hw.fc.current_mode = adapter->hw.fc.requested_mode;
4410 	e1000_force_mac_fc(&adapter->hw);
4411 	return (error);
4412 }
4413 
4414 /*
4415  * Manage Energy Efficient Ethernet:
4416  * Control values:
4417  *     0/1 - enabled/disabled
4418  */
4419 static int
4420 em_sysctl_eee(SYSCTL_HANDLER_ARGS)
4421 {
4422 	struct adapter *adapter = (struct adapter *) arg1;
4423 	int error, value;
4424 
4425 	value = adapter->hw.dev_spec.ich8lan.eee_disable;
4426 	error = sysctl_handle_int(oidp, &value, 0, req);
4427 	if (error || req->newptr == NULL)
4428 		return (error);
4429 	adapter->hw.dev_spec.ich8lan.eee_disable = (value != 0);
4430 	em_if_init(adapter->ctx);
4431 
4432 	return (0);
4433 }
4434 
4435 static int
4436 em_sysctl_debug_info(SYSCTL_HANDLER_ARGS)
4437 {
4438 	struct adapter *adapter;
4439 	int error;
4440 	int result;
4441 
4442 	result = -1;
4443 	error = sysctl_handle_int(oidp, &result, 0, req);
4444 
4445 	if (error || !req->newptr)
4446 		return (error);
4447 
4448 	if (result == 1) {
4449 		adapter = (struct adapter *) arg1;
4450 		em_print_debug_info(adapter);
4451 	}
4452 
4453 	return (error);
4454 }
4455 
4456 static int
4457 em_get_rs(SYSCTL_HANDLER_ARGS)
4458 {
4459 	struct adapter *adapter = (struct adapter *) arg1;
4460 	int error;
4461 	int result;
4462 
4463 	result = 0;
4464 	error = sysctl_handle_int(oidp, &result, 0, req);
4465 
4466 	if (error || !req->newptr || result != 1)
4467 		return (error);
4468 	em_dump_rs(adapter);
4469 
4470 	return (error);
4471 }
4472 
4473 static void
4474 em_if_debug(if_ctx_t ctx)
4475 {
4476 	em_dump_rs(iflib_get_softc(ctx));
4477 }
4478 
4479 /*
4480  * This routine is meant to be fluid, add whatever is
4481  * needed for debugging a problem.  -jfv
4482  */
4483 static void
4484 em_print_debug_info(struct adapter *adapter)
4485 {
4486 	device_t dev = iflib_get_dev(adapter->ctx);
4487 	struct ifnet *ifp = iflib_get_ifp(adapter->ctx);
4488 	struct tx_ring *txr = &adapter->tx_queues->txr;
4489 	struct rx_ring *rxr = &adapter->rx_queues->rxr;
4490 
4491 	if (if_getdrvflags(ifp) & IFF_DRV_RUNNING)
4492 		printf("Interface is RUNNING ");
4493 	else
4494 		printf("Interface is NOT RUNNING\n");
4495 
4496 	if (if_getdrvflags(ifp) & IFF_DRV_OACTIVE)
4497 		printf("and INACTIVE\n");
4498 	else
4499 		printf("and ACTIVE\n");
4500 
4501 	for (int i = 0; i < adapter->tx_num_queues; i++, txr++) {
4502 		device_printf(dev, "TX Queue %d ------\n", i);
4503 		device_printf(dev, "hw tdh = %d, hw tdt = %d\n",
4504 			E1000_READ_REG(&adapter->hw, E1000_TDH(i)),
4505 			E1000_READ_REG(&adapter->hw, E1000_TDT(i)));
4506 
4507 	}
4508 	for (int j=0; j < adapter->rx_num_queues; j++, rxr++) {
4509 		device_printf(dev, "RX Queue %d ------\n", j);
4510 		device_printf(dev, "hw rdh = %d, hw rdt = %d\n",
4511 			E1000_READ_REG(&adapter->hw, E1000_RDH(j)),
4512 			E1000_READ_REG(&adapter->hw, E1000_RDT(j)));
4513 	}
4514 }
4515 
4516 /*
4517  * 82574 only:
4518  * Write a new value to the EEPROM increasing the number of MSI-X
4519  * vectors from 3 to 5, for proper multiqueue support.
4520  */
4521 static void
4522 em_enable_vectors_82574(if_ctx_t ctx)
4523 {
4524 	struct adapter *adapter = iflib_get_softc(ctx);
4525 	struct e1000_hw *hw = &adapter->hw;
4526 	device_t dev = iflib_get_dev(ctx);
4527 	u16 edata;
4528 
4529 	e1000_read_nvm(hw, EM_NVM_PCIE_CTRL, 1, &edata);
4530 	if (bootverbose)
4531 		device_printf(dev, "EM_NVM_PCIE_CTRL = %#06x\n", edata);
4532 	if (((edata & EM_NVM_MSIX_N_MASK) >> EM_NVM_MSIX_N_SHIFT) != 4) {
4533 		device_printf(dev, "Writing to eeprom: increasing "
4534 		    "reported MSI-X vectors from 3 to 5...\n");
4535 		edata &= ~(EM_NVM_MSIX_N_MASK);
4536 		edata |= 4 << EM_NVM_MSIX_N_SHIFT;
4537 		e1000_write_nvm(hw, EM_NVM_PCIE_CTRL, 1, &edata);
4538 		e1000_update_nvm_checksum(hw);
4539 		device_printf(dev, "Writing to eeprom: done\n");
4540 	}
4541 }
4542