xref: /freebsd/sys/dev/e1000/if_em.c (revision 7815283df299be63807225a9fe9b6e54406eae28)
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 
253 static void	em_identify_hardware(if_ctx_t ctx);
254 static int	em_allocate_pci_resources(if_ctx_t ctx);
255 static void	em_free_pci_resources(if_ctx_t ctx);
256 static void	em_reset(if_ctx_t ctx);
257 static int	em_setup_interface(if_ctx_t ctx);
258 static int	em_setup_msix(if_ctx_t ctx);
259 
260 static void	em_initialize_transmit_unit(if_ctx_t ctx);
261 static void	em_initialize_receive_unit(if_ctx_t ctx);
262 
263 static void	em_if_enable_intr(if_ctx_t ctx);
264 static void	em_if_disable_intr(if_ctx_t ctx);
265 static int	em_if_rx_queue_intr_enable(if_ctx_t ctx, uint16_t rxqid);
266 static int	em_if_tx_queue_intr_enable(if_ctx_t ctx, uint16_t txqid);
267 static void	em_if_multi_set(if_ctx_t ctx);
268 static void	em_if_update_admin_status(if_ctx_t ctx);
269 static void	em_if_debug(if_ctx_t ctx);
270 static void	em_update_stats_counters(struct adapter *);
271 static void	em_add_hw_stats(struct adapter *adapter);
272 static int	em_if_set_promisc(if_ctx_t ctx, int flags);
273 static void	em_setup_vlan_hw_support(struct adapter *);
274 static int	em_sysctl_nvm_info(SYSCTL_HANDLER_ARGS);
275 static void	em_print_nvm_info(struct adapter *);
276 static int	em_sysctl_debug_info(SYSCTL_HANDLER_ARGS);
277 static int	em_get_rs(SYSCTL_HANDLER_ARGS);
278 static void	em_print_debug_info(struct adapter *);
279 static int 	em_is_valid_ether_addr(u8 *);
280 static int	em_sysctl_int_delay(SYSCTL_HANDLER_ARGS);
281 static void	em_add_int_delay_sysctl(struct adapter *, const char *,
282 		    const char *, struct em_int_delay_info *, int, int);
283 /* Management and WOL Support */
284 static void	em_init_manageability(struct adapter *);
285 static void	em_release_manageability(struct adapter *);
286 static void	em_get_hw_control(struct adapter *);
287 static void	em_release_hw_control(struct adapter *);
288 static void	em_get_wakeup(if_ctx_t ctx);
289 static void	em_enable_wakeup(if_ctx_t ctx);
290 static int	em_enable_phy_wakeup(struct adapter *);
291 static void	em_disable_aspm(struct adapter *);
292 
293 int		em_intr(void *arg);
294 static void	em_disable_promisc(if_ctx_t ctx);
295 
296 /* MSIX handlers */
297 static int	em_if_msix_intr_assign(if_ctx_t, int);
298 static int	em_msix_link(void *);
299 static void	em_handle_link(void *context);
300 
301 static void	em_enable_vectors_82574(if_ctx_t);
302 
303 static int	em_set_flowcntl(SYSCTL_HANDLER_ARGS);
304 static int	em_sysctl_eee(SYSCTL_HANDLER_ARGS);
305 static void	em_if_led_func(if_ctx_t ctx, int onoff);
306 
307 static int	em_get_regs(SYSCTL_HANDLER_ARGS);
308 
309 static void	lem_smartspeed(struct adapter *adapter);
310 static void	igb_configure_queues(struct adapter *adapter);
311 
312 
313 /*********************************************************************
314  *  FreeBSD Device Interface Entry Points
315  *********************************************************************/
316 static device_method_t em_methods[] = {
317 	/* Device interface */
318 	DEVMETHOD(device_register, em_register),
319 	DEVMETHOD(device_probe, iflib_device_probe),
320 	DEVMETHOD(device_attach, iflib_device_attach),
321 	DEVMETHOD(device_detach, iflib_device_detach),
322 	DEVMETHOD(device_shutdown, iflib_device_shutdown),
323 	DEVMETHOD(device_suspend, iflib_device_suspend),
324 	DEVMETHOD(device_resume, iflib_device_resume),
325 	DEVMETHOD_END
326 };
327 
328 static device_method_t igb_methods[] = {
329 	/* Device interface */
330 	DEVMETHOD(device_register, igb_register),
331 	DEVMETHOD(device_probe, iflib_device_probe),
332 	DEVMETHOD(device_attach, iflib_device_attach),
333 	DEVMETHOD(device_detach, iflib_device_detach),
334 	DEVMETHOD(device_shutdown, iflib_device_shutdown),
335 	DEVMETHOD(device_suspend, iflib_device_suspend),
336 	DEVMETHOD(device_resume, iflib_device_resume),
337 	DEVMETHOD_END
338 };
339 
340 
341 static driver_t em_driver = {
342 	"em", em_methods, sizeof(struct adapter),
343 };
344 
345 static devclass_t em_devclass;
346 DRIVER_MODULE(em, pci, em_driver, em_devclass, 0, 0);
347 
348 MODULE_DEPEND(em, pci, 1, 1, 1);
349 MODULE_DEPEND(em, ether, 1, 1, 1);
350 MODULE_DEPEND(em, iflib, 1, 1, 1);
351 
352 IFLIB_PNP_INFO(pci, em, em_vendor_info_array);
353 
354 static driver_t igb_driver = {
355 	"igb", igb_methods, sizeof(struct adapter),
356 };
357 
358 static devclass_t igb_devclass;
359 DRIVER_MODULE(igb, pci, igb_driver, igb_devclass, 0, 0);
360 
361 MODULE_DEPEND(igb, pci, 1, 1, 1);
362 MODULE_DEPEND(igb, ether, 1, 1, 1);
363 MODULE_DEPEND(igb, iflib, 1, 1, 1);
364 
365 IFLIB_PNP_INFO(pci, igb, igb_vendor_info_array);
366 
367 static device_method_t em_if_methods[] = {
368 	DEVMETHOD(ifdi_attach_pre, em_if_attach_pre),
369 	DEVMETHOD(ifdi_attach_post, em_if_attach_post),
370 	DEVMETHOD(ifdi_detach, em_if_detach),
371 	DEVMETHOD(ifdi_shutdown, em_if_shutdown),
372 	DEVMETHOD(ifdi_suspend, em_if_suspend),
373 	DEVMETHOD(ifdi_resume, em_if_resume),
374 	DEVMETHOD(ifdi_init, em_if_init),
375 	DEVMETHOD(ifdi_stop, em_if_stop),
376 	DEVMETHOD(ifdi_msix_intr_assign, em_if_msix_intr_assign),
377 	DEVMETHOD(ifdi_intr_enable, em_if_enable_intr),
378 	DEVMETHOD(ifdi_intr_disable, em_if_disable_intr),
379 	DEVMETHOD(ifdi_tx_queues_alloc, em_if_tx_queues_alloc),
380 	DEVMETHOD(ifdi_rx_queues_alloc, em_if_rx_queues_alloc),
381 	DEVMETHOD(ifdi_queues_free, em_if_queues_free),
382 	DEVMETHOD(ifdi_update_admin_status, em_if_update_admin_status),
383 	DEVMETHOD(ifdi_multi_set, em_if_multi_set),
384 	DEVMETHOD(ifdi_media_status, em_if_media_status),
385 	DEVMETHOD(ifdi_media_change, em_if_media_change),
386 	DEVMETHOD(ifdi_mtu_set, em_if_mtu_set),
387 	DEVMETHOD(ifdi_promisc_set, em_if_set_promisc),
388 	DEVMETHOD(ifdi_timer, em_if_timer),
389 	DEVMETHOD(ifdi_vlan_register, em_if_vlan_register),
390 	DEVMETHOD(ifdi_vlan_unregister, em_if_vlan_unregister),
391 	DEVMETHOD(ifdi_get_counter, em_if_get_counter),
392 	DEVMETHOD(ifdi_led_func, em_if_led_func),
393 	DEVMETHOD(ifdi_rx_queue_intr_enable, em_if_rx_queue_intr_enable),
394 	DEVMETHOD(ifdi_tx_queue_intr_enable, em_if_tx_queue_intr_enable),
395 	DEVMETHOD(ifdi_debug, em_if_debug),
396 	DEVMETHOD_END
397 };
398 
399 /*
400  * note that if (adapter->msix_mem) is replaced by:
401  * if (adapter->intr_type == IFLIB_INTR_MSIX)
402  */
403 static driver_t em_if_driver = {
404 	"em_if", em_if_methods, sizeof(struct adapter)
405 };
406 
407 /*********************************************************************
408  *  Tunable default values.
409  *********************************************************************/
410 
411 #define EM_TICKS_TO_USECS(ticks)	((1024 * (ticks) + 500) / 1000)
412 #define EM_USECS_TO_TICKS(usecs)	((1000 * (usecs) + 512) / 1024)
413 
414 #define MAX_INTS_PER_SEC	8000
415 #define DEFAULT_ITR		(1000000000/(MAX_INTS_PER_SEC * 256))
416 
417 /* Allow common code without TSO */
418 #ifndef CSUM_TSO
419 #define CSUM_TSO	0
420 #endif
421 
422 static SYSCTL_NODE(_hw, OID_AUTO, em, CTLFLAG_RD, 0, "EM driver parameters");
423 
424 static int em_disable_crc_stripping = 0;
425 SYSCTL_INT(_hw_em, OID_AUTO, disable_crc_stripping, CTLFLAG_RDTUN,
426     &em_disable_crc_stripping, 0, "Disable CRC Stripping");
427 
428 static int em_tx_int_delay_dflt = EM_TICKS_TO_USECS(EM_TIDV);
429 static int em_rx_int_delay_dflt = EM_TICKS_TO_USECS(EM_RDTR);
430 SYSCTL_INT(_hw_em, OID_AUTO, tx_int_delay, CTLFLAG_RDTUN, &em_tx_int_delay_dflt,
431     0, "Default transmit interrupt delay in usecs");
432 SYSCTL_INT(_hw_em, OID_AUTO, rx_int_delay, CTLFLAG_RDTUN, &em_rx_int_delay_dflt,
433     0, "Default receive interrupt delay in usecs");
434 
435 static int em_tx_abs_int_delay_dflt = EM_TICKS_TO_USECS(EM_TADV);
436 static int em_rx_abs_int_delay_dflt = EM_TICKS_TO_USECS(EM_RADV);
437 SYSCTL_INT(_hw_em, OID_AUTO, tx_abs_int_delay, CTLFLAG_RDTUN,
438     &em_tx_abs_int_delay_dflt, 0,
439     "Default transmit interrupt delay limit in usecs");
440 SYSCTL_INT(_hw_em, OID_AUTO, rx_abs_int_delay, CTLFLAG_RDTUN,
441     &em_rx_abs_int_delay_dflt, 0,
442     "Default receive interrupt delay limit in usecs");
443 
444 static int em_smart_pwr_down = FALSE;
445 SYSCTL_INT(_hw_em, OID_AUTO, smart_pwr_down, CTLFLAG_RDTUN, &em_smart_pwr_down,
446     0, "Set to true to leave smart power down enabled on newer adapters");
447 
448 /* Controls whether promiscuous also shows bad packets */
449 static int em_debug_sbp = TRUE;
450 SYSCTL_INT(_hw_em, OID_AUTO, sbp, CTLFLAG_RDTUN, &em_debug_sbp, 0,
451     "Show bad packets in promiscuous mode");
452 
453 /* How many packets rxeof tries to clean at a time */
454 static int em_rx_process_limit = 100;
455 SYSCTL_INT(_hw_em, OID_AUTO, rx_process_limit, CTLFLAG_RDTUN,
456     &em_rx_process_limit, 0,
457     "Maximum number of received packets to process "
458     "at a time, -1 means unlimited");
459 
460 /* Energy efficient ethernet - default to OFF */
461 static int eee_setting = 1;
462 SYSCTL_INT(_hw_em, OID_AUTO, eee_setting, CTLFLAG_RDTUN, &eee_setting, 0,
463     "Enable Energy Efficient Ethernet");
464 
465 /*
466 ** Tuneable Interrupt rate
467 */
468 static int em_max_interrupt_rate = 8000;
469 SYSCTL_INT(_hw_em, OID_AUTO, max_interrupt_rate, CTLFLAG_RDTUN,
470     &em_max_interrupt_rate, 0, "Maximum interrupts per second");
471 
472 
473 
474 /* Global used in WOL setup with multiport cards */
475 static int global_quad_port_a = 0;
476 
477 extern struct if_txrx igb_txrx;
478 extern struct if_txrx em_txrx;
479 extern struct if_txrx lem_txrx;
480 
481 static struct if_shared_ctx em_sctx_init = {
482 	.isc_magic = IFLIB_MAGIC,
483 	.isc_q_align = PAGE_SIZE,
484 	.isc_tx_maxsize = EM_TSO_SIZE + sizeof(struct ether_vlan_header),
485 	.isc_tx_maxsegsize = PAGE_SIZE,
486 	.isc_tso_maxsize = EM_TSO_SIZE + sizeof(struct ether_vlan_header),
487 	.isc_tso_maxsegsize = EM_TSO_SEG_SIZE,
488 	.isc_rx_maxsize = MJUM9BYTES,
489 	.isc_rx_nsegments = 1,
490 	.isc_rx_maxsegsize = MJUM9BYTES,
491 	.isc_nfl = 1,
492 	.isc_nrxqs = 1,
493 	.isc_ntxqs = 1,
494 	.isc_admin_intrcnt = 1,
495 	.isc_vendor_info = em_vendor_info_array,
496 	.isc_driver_version = em_driver_version,
497 	.isc_driver = &em_if_driver,
498 	.isc_flags = IFLIB_NEED_SCRATCH | IFLIB_TSO_INIT_IP | IFLIB_NEED_ZERO_CSUM,
499 
500 	.isc_nrxd_min = {EM_MIN_RXD},
501 	.isc_ntxd_min = {EM_MIN_TXD},
502 	.isc_nrxd_max = {EM_MAX_RXD},
503 	.isc_ntxd_max = {EM_MAX_TXD},
504 	.isc_nrxd_default = {EM_DEFAULT_RXD},
505 	.isc_ntxd_default = {EM_DEFAULT_TXD},
506 };
507 
508 if_shared_ctx_t em_sctx = &em_sctx_init;
509 
510 static struct if_shared_ctx igb_sctx_init = {
511 	.isc_magic = IFLIB_MAGIC,
512 	.isc_q_align = PAGE_SIZE,
513 	.isc_tx_maxsize = EM_TSO_SIZE + sizeof(struct ether_vlan_header),
514 	.isc_tx_maxsegsize = PAGE_SIZE,
515 	.isc_tso_maxsize = EM_TSO_SIZE + sizeof(struct ether_vlan_header),
516 	.isc_tso_maxsegsize = EM_TSO_SEG_SIZE,
517 	.isc_rx_maxsize = MJUM9BYTES,
518 	.isc_rx_nsegments = 1,
519 	.isc_rx_maxsegsize = MJUM9BYTES,
520 	.isc_nfl = 1,
521 	.isc_nrxqs = 1,
522 	.isc_ntxqs = 1,
523 	.isc_admin_intrcnt = 1,
524 	.isc_vendor_info = igb_vendor_info_array,
525 	.isc_driver_version = em_driver_version,
526 	.isc_driver = &em_if_driver,
527 	.isc_flags = IFLIB_NEED_SCRATCH | IFLIB_TSO_INIT_IP | IFLIB_NEED_ZERO_CSUM,
528 
529 	.isc_nrxd_min = {EM_MIN_RXD},
530 	.isc_ntxd_min = {EM_MIN_TXD},
531 	.isc_nrxd_max = {IGB_MAX_RXD},
532 	.isc_ntxd_max = {IGB_MAX_TXD},
533 	.isc_nrxd_default = {EM_DEFAULT_RXD},
534 	.isc_ntxd_default = {EM_DEFAULT_TXD},
535 };
536 
537 if_shared_ctx_t igb_sctx = &igb_sctx_init;
538 
539 /*****************************************************************
540  *
541  * Dump Registers
542  *
543  ****************************************************************/
544 #define IGB_REGS_LEN 739
545 
546 static int em_get_regs(SYSCTL_HANDLER_ARGS)
547 {
548 	struct adapter *adapter = (struct adapter *)arg1;
549 	struct e1000_hw *hw = &adapter->hw;
550 	struct sbuf *sb;
551 	u32 *regs_buff;
552 	int rc;
553 
554 	regs_buff = malloc(sizeof(u32) * IGB_REGS_LEN, M_DEVBUF, M_WAITOK);
555 	memset(regs_buff, 0, IGB_REGS_LEN * sizeof(u32));
556 
557 	rc = sysctl_wire_old_buffer(req, 0);
558 	MPASS(rc == 0);
559 	if (rc != 0) {
560 		free(regs_buff, M_DEVBUF);
561 		return (rc);
562 	}
563 
564 	sb = sbuf_new_for_sysctl(NULL, NULL, 32*400, req);
565 	MPASS(sb != NULL);
566 	if (sb == NULL) {
567 		free(regs_buff, M_DEVBUF);
568 		return (ENOMEM);
569 	}
570 
571 	/* General Registers */
572 	regs_buff[0] = E1000_READ_REG(hw, E1000_CTRL);
573 	regs_buff[1] = E1000_READ_REG(hw, E1000_STATUS);
574 	regs_buff[2] = E1000_READ_REG(hw, E1000_CTRL_EXT);
575 	regs_buff[3] = E1000_READ_REG(hw, E1000_ICR);
576 	regs_buff[4] = E1000_READ_REG(hw, E1000_RCTL);
577 	regs_buff[5] = E1000_READ_REG(hw, E1000_RDLEN(0));
578 	regs_buff[6] = E1000_READ_REG(hw, E1000_RDH(0));
579 	regs_buff[7] = E1000_READ_REG(hw, E1000_RDT(0));
580 	regs_buff[8] = E1000_READ_REG(hw, E1000_RXDCTL(0));
581 	regs_buff[9] = E1000_READ_REG(hw, E1000_RDBAL(0));
582 	regs_buff[10] = E1000_READ_REG(hw, E1000_RDBAH(0));
583 	regs_buff[11] = E1000_READ_REG(hw, E1000_TCTL);
584 	regs_buff[12] = E1000_READ_REG(hw, E1000_TDBAL(0));
585 	regs_buff[13] = E1000_READ_REG(hw, E1000_TDBAH(0));
586 	regs_buff[14] = E1000_READ_REG(hw, E1000_TDLEN(0));
587 	regs_buff[15] = E1000_READ_REG(hw, E1000_TDH(0));
588 	regs_buff[16] = E1000_READ_REG(hw, E1000_TDT(0));
589 	regs_buff[17] = E1000_READ_REG(hw, E1000_TXDCTL(0));
590 	regs_buff[18] = E1000_READ_REG(hw, E1000_TDFH);
591 	regs_buff[19] = E1000_READ_REG(hw, E1000_TDFT);
592 	regs_buff[20] = E1000_READ_REG(hw, E1000_TDFHS);
593 	regs_buff[21] = E1000_READ_REG(hw, E1000_TDFPC);
594 
595 	sbuf_printf(sb, "General Registers\n");
596 	sbuf_printf(sb, "\tCTRL\t %08x\n", regs_buff[0]);
597 	sbuf_printf(sb, "\tSTATUS\t %08x\n", regs_buff[1]);
598 	sbuf_printf(sb, "\tCTRL_EXIT\t %08x\n\n", regs_buff[2]);
599 
600 	sbuf_printf(sb, "Interrupt Registers\n");
601 	sbuf_printf(sb, "\tICR\t %08x\n\n", regs_buff[3]);
602 
603 	sbuf_printf(sb, "RX Registers\n");
604 	sbuf_printf(sb, "\tRCTL\t %08x\n", regs_buff[4]);
605 	sbuf_printf(sb, "\tRDLEN\t %08x\n", regs_buff[5]);
606 	sbuf_printf(sb, "\tRDH\t %08x\n", regs_buff[6]);
607 	sbuf_printf(sb, "\tRDT\t %08x\n", regs_buff[7]);
608 	sbuf_printf(sb, "\tRXDCTL\t %08x\n", regs_buff[8]);
609 	sbuf_printf(sb, "\tRDBAL\t %08x\n", regs_buff[9]);
610 	sbuf_printf(sb, "\tRDBAH\t %08x\n\n", regs_buff[10]);
611 
612 	sbuf_printf(sb, "TX Registers\n");
613 	sbuf_printf(sb, "\tTCTL\t %08x\n", regs_buff[11]);
614 	sbuf_printf(sb, "\tTDBAL\t %08x\n", regs_buff[12]);
615 	sbuf_printf(sb, "\tTDBAH\t %08x\n", regs_buff[13]);
616 	sbuf_printf(sb, "\tTDLEN\t %08x\n", regs_buff[14]);
617 	sbuf_printf(sb, "\tTDH\t %08x\n", regs_buff[15]);
618 	sbuf_printf(sb, "\tTDT\t %08x\n", regs_buff[16]);
619 	sbuf_printf(sb, "\tTXDCTL\t %08x\n", regs_buff[17]);
620 	sbuf_printf(sb, "\tTDFH\t %08x\n", regs_buff[18]);
621 	sbuf_printf(sb, "\tTDFT\t %08x\n", regs_buff[19]);
622 	sbuf_printf(sb, "\tTDFHS\t %08x\n", regs_buff[20]);
623 	sbuf_printf(sb, "\tTDFPC\t %08x\n\n", regs_buff[21]);
624 
625 	free(regs_buff, M_DEVBUF);
626 
627 #ifdef DUMP_DESCS
628 	{
629 		if_softc_ctx_t scctx = adapter->shared;
630 		struct rx_ring *rxr = &rx_que->rxr;
631 		struct tx_ring *txr = &tx_que->txr;
632 		int ntxd = scctx->isc_ntxd[0];
633 		int nrxd = scctx->isc_nrxd[0];
634 		int j;
635 
636 	for (j = 0; j < nrxd; j++) {
637 		u32 staterr = le32toh(rxr->rx_base[j].wb.upper.status_error);
638 		u32 length =  le32toh(rxr->rx_base[j].wb.upper.length);
639 		sbuf_printf(sb, "\tReceive Descriptor Address %d: %08" PRIx64 "  Error:%d  Length:%d\n", j, rxr->rx_base[j].read.buffer_addr, staterr, length);
640 	}
641 
642 	for (j = 0; j < min(ntxd, 256); j++) {
643 		unsigned int *ptr = (unsigned int *)&txr->tx_base[j];
644 
645 		sbuf_printf(sb, "\tTXD[%03d] [0]: %08x [1]: %08x [2]: %08x [3]: %08x  eop: %d DD=%d\n",
646 			    j, ptr[0], ptr[1], ptr[2], ptr[3], buf->eop,
647 			    buf->eop != -1 ? txr->tx_base[buf->eop].upper.fields.status & E1000_TXD_STAT_DD : 0);
648 
649 	}
650 	}
651 #endif
652 
653 	rc = sbuf_finish(sb);
654 	sbuf_delete(sb);
655 	return(rc);
656 }
657 
658 static void *
659 em_register(device_t dev)
660 {
661 	return (em_sctx);
662 }
663 
664 static void *
665 igb_register(device_t dev)
666 {
667 	return (igb_sctx);
668 }
669 
670 static int
671 em_set_num_queues(if_ctx_t ctx)
672 {
673 	struct adapter *adapter = iflib_get_softc(ctx);
674 	int maxqueues;
675 
676 	/* Sanity check based on HW */
677 	switch (adapter->hw.mac.type) {
678 	case e1000_82576:
679 	case e1000_82580:
680 	case e1000_i350:
681 	case e1000_i354:
682 		maxqueues = 8;
683 		break;
684 	case e1000_i210:
685 	case e1000_82575:
686 		maxqueues = 4;
687 		break;
688 	case e1000_i211:
689 	case e1000_82574:
690 		maxqueues = 2;
691 		break;
692 	default:
693 		maxqueues = 1;
694 		break;
695 	}
696 
697 	return (maxqueues);
698 }
699 
700 #define	LEM_CAPS							\
701     IFCAP_HWCSUM | IFCAP_VLAN_MTU | IFCAP_VLAN_HWTAGGING |		\
702     IFCAP_VLAN_HWCSUM | IFCAP_WOL | IFCAP_VLAN_HWFILTER
703 
704 #define	EM_CAPS								\
705     IFCAP_HWCSUM | IFCAP_VLAN_MTU | IFCAP_VLAN_HWTAGGING |		\
706     IFCAP_VLAN_HWCSUM | IFCAP_WOL | IFCAP_VLAN_HWFILTER | IFCAP_TSO4 |	\
707     IFCAP_LRO | IFCAP_VLAN_HWTSO
708 
709 #define	IGB_CAPS							\
710     IFCAP_HWCSUM | IFCAP_VLAN_MTU | IFCAP_VLAN_HWTAGGING |		\
711     IFCAP_VLAN_HWCSUM | IFCAP_WOL | IFCAP_VLAN_HWFILTER | IFCAP_TSO4 |	\
712     IFCAP_LRO | IFCAP_VLAN_HWTSO | IFCAP_JUMBO_MTU | IFCAP_HWCSUM_IPV6 |\
713     IFCAP_TSO6
714 
715 /*********************************************************************
716  *  Device initialization routine
717  *
718  *  The attach entry point is called when the driver is being loaded.
719  *  This routine identifies the type of hardware, allocates all resources
720  *  and initializes the hardware.
721  *
722  *  return 0 on success, positive on failure
723  *********************************************************************/
724 
725 static int
726 em_if_attach_pre(if_ctx_t ctx)
727 {
728 	struct adapter *adapter;
729 	if_softc_ctx_t scctx;
730 	device_t dev;
731 	struct e1000_hw *hw;
732 	int error = 0;
733 
734 	INIT_DEBUGOUT("em_if_attach_pre begin");
735 	dev = iflib_get_dev(ctx);
736 	adapter = iflib_get_softc(ctx);
737 
738 	if (resource_disabled("em", device_get_unit(dev))) {
739 		device_printf(dev, "Disabled by device hint\n");
740 		return (ENXIO);
741 	}
742 
743 	adapter->ctx = adapter->osdep.ctx = ctx;
744 	adapter->dev = adapter->osdep.dev = dev;
745 	scctx = adapter->shared = iflib_get_softc_ctx(ctx);
746 	adapter->media = iflib_get_media(ctx);
747 	hw = &adapter->hw;
748 
749 	adapter->tx_process_limit = scctx->isc_ntxd[0];
750 
751 	/* SYSCTL stuff */
752 	SYSCTL_ADD_PROC(device_get_sysctl_ctx(dev),
753 	    SYSCTL_CHILDREN(device_get_sysctl_tree(dev)),
754 	    OID_AUTO, "nvm", CTLTYPE_INT|CTLFLAG_RW, adapter, 0,
755 	    em_sysctl_nvm_info, "I", "NVM Information");
756 
757 	SYSCTL_ADD_PROC(device_get_sysctl_ctx(dev),
758 	    SYSCTL_CHILDREN(device_get_sysctl_tree(dev)),
759 	    OID_AUTO, "debug", CTLTYPE_INT|CTLFLAG_RW, adapter, 0,
760 	    em_sysctl_debug_info, "I", "Debug Information");
761 
762 	SYSCTL_ADD_PROC(device_get_sysctl_ctx(dev),
763 	    SYSCTL_CHILDREN(device_get_sysctl_tree(dev)),
764 	    OID_AUTO, "fc", CTLTYPE_INT|CTLFLAG_RW, adapter, 0,
765 	    em_set_flowcntl, "I", "Flow Control");
766 
767 	SYSCTL_ADD_PROC(device_get_sysctl_ctx(dev),
768 	    SYSCTL_CHILDREN(device_get_sysctl_tree(dev)),
769 	    OID_AUTO, "reg_dump", CTLTYPE_STRING | CTLFLAG_RD, adapter, 0,
770 	    em_get_regs, "A", "Dump Registers");
771 
772 	SYSCTL_ADD_PROC(device_get_sysctl_ctx(dev),
773 	    SYSCTL_CHILDREN(device_get_sysctl_tree(dev)),
774 	    OID_AUTO, "rs_dump", CTLTYPE_INT | CTLFLAG_RW, adapter, 0,
775 	    em_get_rs, "I", "Dump RS indexes");
776 
777 	/* Determine hardware and mac info */
778 	em_identify_hardware(ctx);
779 
780 	scctx->isc_msix_bar = PCIR_BAR(EM_MSIX_BAR);
781 	scctx->isc_tx_nsegments = EM_MAX_SCATTER;
782 	scctx->isc_nrxqsets_max = scctx->isc_ntxqsets_max = em_set_num_queues(ctx);
783 	device_printf(dev, "attach_pre capping queues at %d\n", scctx->isc_ntxqsets_max);
784 
785 	if (adapter->hw.mac.type >= igb_mac_min) {
786 		int try_second_bar;
787 
788 		scctx->isc_txqsizes[0] = roundup2(scctx->isc_ntxd[0] * sizeof(union e1000_adv_tx_desc), EM_DBA_ALIGN);
789 		scctx->isc_rxqsizes[0] = roundup2(scctx->isc_nrxd[0] * sizeof(union e1000_adv_rx_desc), EM_DBA_ALIGN);
790 		scctx->isc_txd_size[0] = sizeof(union e1000_adv_tx_desc);
791 		scctx->isc_rxd_size[0] = sizeof(union e1000_adv_rx_desc);
792 		scctx->isc_txrx = &igb_txrx;
793 		scctx->isc_tx_tso_segments_max = EM_MAX_SCATTER;
794 		scctx->isc_tx_tso_size_max = EM_TSO_SIZE;
795 		scctx->isc_tx_tso_segsize_max = EM_TSO_SEG_SIZE;
796 		scctx->isc_capabilities = scctx->isc_capenable = IGB_CAPS;
797 		scctx->isc_tx_csum_flags = CSUM_TCP | CSUM_UDP | CSUM_TSO |
798 		     CSUM_IP6_TCP | CSUM_IP6_UDP;
799 		if (adapter->hw.mac.type != e1000_82575)
800 			scctx->isc_tx_csum_flags |= CSUM_SCTP | CSUM_IP6_SCTP;
801 
802 		/*
803 		** Some new devices, as with ixgbe, now may
804 		** use a different BAR, so we need to keep
805 		** track of which is used.
806 		*/
807 		try_second_bar = pci_read_config(dev, scctx->isc_msix_bar, 4);
808 		if (try_second_bar == 0)
809 			scctx->isc_msix_bar += 4;
810 	} else if (adapter->hw.mac.type >= em_mac_min) {
811 		scctx->isc_txqsizes[0] = roundup2(scctx->isc_ntxd[0]* sizeof(struct e1000_tx_desc), EM_DBA_ALIGN);
812 		scctx->isc_rxqsizes[0] = roundup2(scctx->isc_nrxd[0] * sizeof(union e1000_rx_desc_extended), EM_DBA_ALIGN);
813 		scctx->isc_txd_size[0] = sizeof(struct e1000_tx_desc);
814 		scctx->isc_rxd_size[0] = sizeof(union e1000_rx_desc_extended);
815 		scctx->isc_txrx = &em_txrx;
816 		scctx->isc_tx_tso_segments_max = EM_MAX_SCATTER;
817 		scctx->isc_tx_tso_size_max = EM_TSO_SIZE;
818 		scctx->isc_tx_tso_segsize_max = EM_TSO_SEG_SIZE;
819 		scctx->isc_capabilities = scctx->isc_capenable = EM_CAPS;
820 		/*
821 		 * For EM-class devices, don't enable IFCAP_{TSO4,VLAN_HWTSO}
822 		 * by default as we don't have workarounds for all associated
823 		 * silicon errata.  E. g., with several MACs such as 82573E,
824 		 * TSO only works at Gigabit speed and otherwise can cause the
825 		 * hardware to hang (which also would be next to impossible to
826 		 * work around given that already queued TSO-using descriptors
827 		 * would need to be flushed and vlan(4) reconfigured at runtime
828 		 * in case of a link speed change).  Moreover, MACs like 82579
829 		 * still can hang at Gigabit even with all publicly documented
830 		 * TSO workarounds implemented.  Generally, the penality of
831 		 * these workarounds is rather high and may involve copying
832 		 * mbuf data around so advantages of TSO lapse.  Still, TSO may
833 		 * work for a few MACs of this class - at least when sticking
834 		 * with Gigabit - in which case users may enable TSO manually.
835 		 */
836 		scctx->isc_capenable &= ~(IFCAP_TSO4 | IFCAP_VLAN_HWTSO);
837 		scctx->isc_tx_csum_flags = CSUM_TCP | CSUM_UDP | CSUM_IP_TSO;
838 	} else {
839 		scctx->isc_txqsizes[0] = roundup2((scctx->isc_ntxd[0] + 1) * sizeof(struct e1000_tx_desc), EM_DBA_ALIGN);
840 		scctx->isc_rxqsizes[0] = roundup2((scctx->isc_nrxd[0] + 1) * sizeof(struct e1000_rx_desc), EM_DBA_ALIGN);
841 		scctx->isc_txd_size[0] = sizeof(struct e1000_tx_desc);
842 		scctx->isc_rxd_size[0] = sizeof(struct e1000_rx_desc);
843 		scctx->isc_tx_csum_flags = CSUM_TCP | CSUM_UDP;
844 		scctx->isc_txrx = &lem_txrx;
845 		scctx->isc_capabilities = scctx->isc_capenable = LEM_CAPS;
846 		if (adapter->hw.mac.type < e1000_82543)
847 			scctx->isc_capenable &= ~(IFCAP_HWCSUM|IFCAP_VLAN_HWCSUM);
848 		scctx->isc_msix_bar = 0;
849 	}
850 
851 	/* Setup PCI resources */
852 	if (em_allocate_pci_resources(ctx)) {
853 		device_printf(dev, "Allocation of PCI resources failed\n");
854 		error = ENXIO;
855 		goto err_pci;
856 	}
857 
858 	/*
859 	** For ICH8 and family we need to
860 	** map the flash memory, and this
861 	** must happen after the MAC is
862 	** identified
863 	*/
864 	if ((hw->mac.type == e1000_ich8lan) ||
865 	    (hw->mac.type == e1000_ich9lan) ||
866 	    (hw->mac.type == e1000_ich10lan) ||
867 	    (hw->mac.type == e1000_pchlan) ||
868 	    (hw->mac.type == e1000_pch2lan) ||
869 	    (hw->mac.type == e1000_pch_lpt)) {
870 		int rid = EM_BAR_TYPE_FLASH;
871 		adapter->flash = bus_alloc_resource_any(dev,
872 		    SYS_RES_MEMORY, &rid, RF_ACTIVE);
873 		if (adapter->flash == NULL) {
874 			device_printf(dev, "Mapping of Flash failed\n");
875 			error = ENXIO;
876 			goto err_pci;
877 		}
878 		/* This is used in the shared code */
879 		hw->flash_address = (u8 *)adapter->flash;
880 		adapter->osdep.flash_bus_space_tag =
881 		    rman_get_bustag(adapter->flash);
882 		adapter->osdep.flash_bus_space_handle =
883 		    rman_get_bushandle(adapter->flash);
884 	}
885 	/*
886 	** In the new SPT device flash is not  a
887 	** separate BAR, rather it is also in BAR0,
888 	** so use the same tag and an offset handle for the
889 	** FLASH read/write macros in the shared code.
890 	*/
891 	else if (hw->mac.type >= e1000_pch_spt) {
892 		adapter->osdep.flash_bus_space_tag =
893 		    adapter->osdep.mem_bus_space_tag;
894 		adapter->osdep.flash_bus_space_handle =
895 		    adapter->osdep.mem_bus_space_handle
896 		    + E1000_FLASH_BASE_ADDR;
897 	}
898 
899 	/* Do Shared Code initialization */
900 	error = e1000_setup_init_funcs(hw, TRUE);
901 	if (error) {
902 		device_printf(dev, "Setup of Shared code failed, error %d\n",
903 		    error);
904 		error = ENXIO;
905 		goto err_pci;
906 	}
907 
908 	em_setup_msix(ctx);
909 	e1000_get_bus_info(hw);
910 
911 	/* Set up some sysctls for the tunable interrupt delays */
912 	em_add_int_delay_sysctl(adapter, "rx_int_delay",
913 	    "receive interrupt delay in usecs", &adapter->rx_int_delay,
914 	    E1000_REGISTER(hw, E1000_RDTR), em_rx_int_delay_dflt);
915 	em_add_int_delay_sysctl(adapter, "tx_int_delay",
916 	    "transmit interrupt delay in usecs", &adapter->tx_int_delay,
917 	    E1000_REGISTER(hw, E1000_TIDV), em_tx_int_delay_dflt);
918 	em_add_int_delay_sysctl(adapter, "rx_abs_int_delay",
919 	    "receive interrupt delay limit in usecs",
920 	    &adapter->rx_abs_int_delay,
921 	    E1000_REGISTER(hw, E1000_RADV),
922 	    em_rx_abs_int_delay_dflt);
923 	em_add_int_delay_sysctl(adapter, "tx_abs_int_delay",
924 	    "transmit interrupt delay limit in usecs",
925 	    &adapter->tx_abs_int_delay,
926 	    E1000_REGISTER(hw, E1000_TADV),
927 	    em_tx_abs_int_delay_dflt);
928 	em_add_int_delay_sysctl(adapter, "itr",
929 	    "interrupt delay limit in usecs/4",
930 	    &adapter->tx_itr,
931 	    E1000_REGISTER(hw, E1000_ITR),
932 	    DEFAULT_ITR);
933 
934 	hw->mac.autoneg = DO_AUTO_NEG;
935 	hw->phy.autoneg_wait_to_complete = FALSE;
936 	hw->phy.autoneg_advertised = AUTONEG_ADV_DEFAULT;
937 
938 	if (adapter->hw.mac.type < em_mac_min) {
939 		e1000_init_script_state_82541(&adapter->hw, TRUE);
940 		e1000_set_tbi_compatibility_82543(&adapter->hw, TRUE);
941 	}
942 	/* Copper options */
943 	if (hw->phy.media_type == e1000_media_type_copper) {
944 		hw->phy.mdix = AUTO_ALL_MODES;
945 		hw->phy.disable_polarity_correction = FALSE;
946 		hw->phy.ms_type = EM_MASTER_SLAVE;
947 	}
948 
949 	/*
950 	 * Set the frame limits assuming
951 	 * standard ethernet sized frames.
952 	 */
953 	scctx->isc_max_frame_size = adapter->hw.mac.max_frame_size =
954 	    ETHERMTU + ETHER_HDR_LEN + ETHERNET_FCS_SIZE;
955 
956 	/*
957 	 * This controls when hardware reports transmit completion
958 	 * status.
959 	 */
960 	hw->mac.report_tx_early = 1;
961 
962 	/* Allocate multicast array memory. */
963 	adapter->mta = malloc(sizeof(u8) * ETH_ADDR_LEN *
964 	    MAX_NUM_MULTICAST_ADDRESSES, M_DEVBUF, M_NOWAIT);
965 	if (adapter->mta == NULL) {
966 		device_printf(dev, "Can not allocate multicast setup array\n");
967 		error = ENOMEM;
968 		goto err_late;
969 	}
970 
971 	/* Check SOL/IDER usage */
972 	if (e1000_check_reset_block(hw))
973 		device_printf(dev, "PHY reset is blocked"
974 			      " due to SOL/IDER session.\n");
975 
976 	/* Sysctl for setting Energy Efficient Ethernet */
977 	hw->dev_spec.ich8lan.eee_disable = eee_setting;
978 	SYSCTL_ADD_PROC(device_get_sysctl_ctx(dev),
979 	    SYSCTL_CHILDREN(device_get_sysctl_tree(dev)),
980 	    OID_AUTO, "eee_control", CTLTYPE_INT|CTLFLAG_RW,
981 	    adapter, 0, em_sysctl_eee, "I",
982 	    "Disable Energy Efficient Ethernet");
983 
984 	/*
985 	** Start from a known state, this is
986 	** important in reading the nvm and
987 	** mac from that.
988 	*/
989 	e1000_reset_hw(hw);
990 
991 	/* Make sure we have a good EEPROM before we read from it */
992 	if (e1000_validate_nvm_checksum(hw) < 0) {
993 		/*
994 		** Some PCI-E parts fail the first check due to
995 		** the link being in sleep state, call it again,
996 		** if it fails a second time its a real issue.
997 		*/
998 		if (e1000_validate_nvm_checksum(hw) < 0) {
999 			device_printf(dev,
1000 			    "The EEPROM Checksum Is Not Valid\n");
1001 			error = EIO;
1002 			goto err_late;
1003 		}
1004 	}
1005 
1006 	/* Copy the permanent MAC address out of the EEPROM */
1007 	if (e1000_read_mac_addr(hw) < 0) {
1008 		device_printf(dev, "EEPROM read error while reading MAC"
1009 			      " address\n");
1010 		error = EIO;
1011 		goto err_late;
1012 	}
1013 
1014 	if (!em_is_valid_ether_addr(hw->mac.addr)) {
1015 		device_printf(dev, "Invalid MAC address\n");
1016 		error = EIO;
1017 		goto err_late;
1018 	}
1019 
1020 	/* Disable ULP support */
1021 	e1000_disable_ulp_lpt_lp(hw, TRUE);
1022 
1023 	/*
1024 	 * Get Wake-on-Lan and Management info for later use
1025 	 */
1026 	em_get_wakeup(ctx);
1027 
1028 	/* Enable only WOL MAGIC by default */
1029 	scctx->isc_capenable &= ~IFCAP_WOL;
1030 	if (adapter->wol != 0)
1031 		scctx->isc_capenable |= IFCAP_WOL_MAGIC;
1032 
1033 	iflib_set_mac(ctx, hw->mac.addr);
1034 
1035 	return (0);
1036 
1037 err_late:
1038 	em_release_hw_control(adapter);
1039 err_pci:
1040 	em_free_pci_resources(ctx);
1041 	free(adapter->mta, M_DEVBUF);
1042 
1043 	return (error);
1044 }
1045 
1046 static int
1047 em_if_attach_post(if_ctx_t ctx)
1048 {
1049 	struct adapter *adapter = iflib_get_softc(ctx);
1050 	struct e1000_hw *hw = &adapter->hw;
1051 	int error = 0;
1052 
1053 	/* Setup OS specific network interface */
1054 	error = em_setup_interface(ctx);
1055 	if (error != 0) {
1056 		goto err_late;
1057 	}
1058 
1059 	em_reset(ctx);
1060 
1061 	/* Initialize statistics */
1062 	em_update_stats_counters(adapter);
1063 	hw->mac.get_link_status = 1;
1064 	em_if_update_admin_status(ctx);
1065 	em_add_hw_stats(adapter);
1066 
1067 	/* Non-AMT based hardware can now take control from firmware */
1068 	if (adapter->has_manage && !adapter->has_amt)
1069 		em_get_hw_control(adapter);
1070 
1071 	INIT_DEBUGOUT("em_if_attach_post: end");
1072 
1073 	return (error);
1074 
1075 err_late:
1076 	em_release_hw_control(adapter);
1077 	em_free_pci_resources(ctx);
1078 	em_if_queues_free(ctx);
1079 	free(adapter->mta, M_DEVBUF);
1080 
1081 	return (error);
1082 }
1083 
1084 /*********************************************************************
1085  *  Device removal routine
1086  *
1087  *  The detach entry point is called when the driver is being removed.
1088  *  This routine stops the adapter and deallocates all the resources
1089  *  that were allocated for driver operation.
1090  *
1091  *  return 0 on success, positive on failure
1092  *********************************************************************/
1093 
1094 static int
1095 em_if_detach(if_ctx_t ctx)
1096 {
1097 	struct adapter	*adapter = iflib_get_softc(ctx);
1098 
1099 	INIT_DEBUGOUT("em_detach: begin");
1100 
1101 	e1000_phy_hw_reset(&adapter->hw);
1102 
1103 	em_release_manageability(adapter);
1104 	em_release_hw_control(adapter);
1105 	em_free_pci_resources(ctx);
1106 
1107 	return (0);
1108 }
1109 
1110 /*********************************************************************
1111  *
1112  *  Shutdown entry point
1113  *
1114  **********************************************************************/
1115 
1116 static int
1117 em_if_shutdown(if_ctx_t ctx)
1118 {
1119 	return em_if_suspend(ctx);
1120 }
1121 
1122 /*
1123  * Suspend/resume device methods.
1124  */
1125 static int
1126 em_if_suspend(if_ctx_t ctx)
1127 {
1128 	struct adapter *adapter = iflib_get_softc(ctx);
1129 
1130 	em_release_manageability(adapter);
1131 	em_release_hw_control(adapter);
1132 	em_enable_wakeup(ctx);
1133 	return (0);
1134 }
1135 
1136 static int
1137 em_if_resume(if_ctx_t ctx)
1138 {
1139 	struct adapter *adapter = iflib_get_softc(ctx);
1140 
1141 	if (adapter->hw.mac.type == e1000_pch2lan)
1142 		e1000_resume_workarounds_pchlan(&adapter->hw);
1143 	em_if_init(ctx);
1144 	em_init_manageability(adapter);
1145 
1146 	return(0);
1147 }
1148 
1149 static int
1150 em_if_mtu_set(if_ctx_t ctx, uint32_t mtu)
1151 {
1152 	int max_frame_size;
1153 	struct adapter *adapter = iflib_get_softc(ctx);
1154 	if_softc_ctx_t scctx = iflib_get_softc_ctx(ctx);
1155 
1156 	 IOCTL_DEBUGOUT("ioctl rcv'd: SIOCSIFMTU (Set Interface MTU)");
1157 
1158 	switch (adapter->hw.mac.type) {
1159 	case e1000_82571:
1160 	case e1000_82572:
1161 	case e1000_ich9lan:
1162 	case e1000_ich10lan:
1163 	case e1000_pch2lan:
1164 	case e1000_pch_lpt:
1165 	case e1000_pch_spt:
1166 	case e1000_pch_cnp:
1167 	case e1000_82574:
1168 	case e1000_82583:
1169 	case e1000_80003es2lan:
1170 		/* 9K Jumbo Frame size */
1171 		max_frame_size = 9234;
1172 		break;
1173 	case e1000_pchlan:
1174 		max_frame_size = 4096;
1175 		break;
1176 	case e1000_82542:
1177 	case e1000_ich8lan:
1178 		/* Adapters that do not support jumbo frames */
1179 		max_frame_size = ETHER_MAX_LEN;
1180 		break;
1181 	default:
1182 		if (adapter->hw.mac.type >= igb_mac_min)
1183 			max_frame_size = 9234;
1184 		else /* lem */
1185 			max_frame_size = MAX_JUMBO_FRAME_SIZE;
1186 	}
1187 	if (mtu > max_frame_size - ETHER_HDR_LEN - ETHER_CRC_LEN) {
1188 		return (EINVAL);
1189 	}
1190 
1191 	scctx->isc_max_frame_size = adapter->hw.mac.max_frame_size =
1192 	    mtu + ETHER_HDR_LEN + ETHER_CRC_LEN;
1193 	return (0);
1194 }
1195 
1196 /*********************************************************************
1197  *  Init entry point
1198  *
1199  *  This routine is used in two ways. It is used by the stack as
1200  *  init entry point in network interface structure. It is also used
1201  *  by the driver as a hw/sw initialization routine to get to a
1202  *  consistent state.
1203  *
1204  *  return 0 on success, positive on failure
1205  **********************************************************************/
1206 
1207 static void
1208 em_if_init(if_ctx_t ctx)
1209 {
1210 	struct adapter *adapter = iflib_get_softc(ctx);
1211 	struct ifnet *ifp = iflib_get_ifp(ctx);
1212 	struct em_tx_queue *tx_que;
1213 	int i;
1214 	INIT_DEBUGOUT("em_if_init: begin");
1215 
1216 	/* Get the latest mac address, User can use a LAA */
1217 	bcopy(if_getlladdr(ifp), adapter->hw.mac.addr,
1218 	    ETHER_ADDR_LEN);
1219 
1220 	/* Put the address into the Receive Address Array */
1221 	e1000_rar_set(&adapter->hw, adapter->hw.mac.addr, 0);
1222 
1223 	/*
1224 	 * With the 82571 adapter, RAR[0] may be overwritten
1225 	 * when the other port is reset, we make a duplicate
1226 	 * in RAR[14] for that eventuality, this assures
1227 	 * the interface continues to function.
1228 	 */
1229 	if (adapter->hw.mac.type == e1000_82571) {
1230 		e1000_set_laa_state_82571(&adapter->hw, TRUE);
1231 		e1000_rar_set(&adapter->hw, adapter->hw.mac.addr,
1232 		    E1000_RAR_ENTRIES - 1);
1233 	}
1234 
1235 
1236 	/* Initialize the hardware */
1237 	em_reset(ctx);
1238 	em_if_update_admin_status(ctx);
1239 
1240 	for (i = 0, tx_que = adapter->tx_queues; i < adapter->tx_num_queues; i++, tx_que++) {
1241 		struct tx_ring *txr = &tx_que->txr;
1242 
1243 		txr->tx_rs_cidx = txr->tx_rs_pidx = txr->tx_cidx_processed = 0;
1244 	}
1245 
1246 	/* Setup VLAN support, basic and offload if available */
1247 	E1000_WRITE_REG(&adapter->hw, E1000_VET, ETHERTYPE_VLAN);
1248 
1249 	/* Clear bad data from Rx FIFOs */
1250 	if (adapter->hw.mac.type >= igb_mac_min)
1251 		e1000_rx_fifo_flush_82575(&adapter->hw);
1252 
1253 	/* Configure for OS presence */
1254 	em_init_manageability(adapter);
1255 
1256 	/* Prepare transmit descriptors and buffers */
1257 	em_initialize_transmit_unit(ctx);
1258 
1259 	/* Setup Multicast table */
1260 	em_if_multi_set(ctx);
1261 
1262 	/*
1263 	 * Figure out the desired mbuf
1264 	 * pool for doing jumbos
1265 	 */
1266 	if (adapter->hw.mac.max_frame_size <= 2048)
1267 		adapter->rx_mbuf_sz = MCLBYTES;
1268 #ifndef CONTIGMALLOC_WORKS
1269 	else
1270 		adapter->rx_mbuf_sz = MJUMPAGESIZE;
1271 #else
1272 	else if (adapter->hw.mac.max_frame_size <= 4096)
1273 		adapter->rx_mbuf_sz = MJUMPAGESIZE;
1274 	else
1275 		adapter->rx_mbuf_sz = MJUM9BYTES;
1276 #endif
1277 	em_initialize_receive_unit(ctx);
1278 
1279 	/* Use real VLAN Filter support? */
1280 	if (if_getcapenable(ifp) & IFCAP_VLAN_HWTAGGING) {
1281 		if (if_getcapenable(ifp) & IFCAP_VLAN_HWFILTER)
1282 			/* Use real VLAN Filter support */
1283 			em_setup_vlan_hw_support(adapter);
1284 		else {
1285 			u32 ctrl;
1286 			ctrl = E1000_READ_REG(&adapter->hw, E1000_CTRL);
1287 			ctrl |= E1000_CTRL_VME;
1288 			E1000_WRITE_REG(&adapter->hw, E1000_CTRL, ctrl);
1289 		}
1290 	}
1291 
1292 	/* Don't lose promiscuous settings */
1293 	em_if_set_promisc(ctx, IFF_PROMISC);
1294 	e1000_clear_hw_cntrs_base_generic(&adapter->hw);
1295 
1296 	/* MSI/X configuration for 82574 */
1297 	if (adapter->hw.mac.type == e1000_82574) {
1298 		int tmp = E1000_READ_REG(&adapter->hw, E1000_CTRL_EXT);
1299 
1300 		tmp |= E1000_CTRL_EXT_PBA_CLR;
1301 		E1000_WRITE_REG(&adapter->hw, E1000_CTRL_EXT, tmp);
1302 		/* Set the IVAR - interrupt vector routing. */
1303 		E1000_WRITE_REG(&adapter->hw, E1000_IVAR, adapter->ivars);
1304 	} else if (adapter->intr_type == IFLIB_INTR_MSIX) /* Set up queue routing */
1305 		igb_configure_queues(adapter);
1306 
1307 	/* this clears any pending interrupts */
1308 	E1000_READ_REG(&adapter->hw, E1000_ICR);
1309 	E1000_WRITE_REG(&adapter->hw, E1000_ICS, E1000_ICS_LSC);
1310 
1311 	/* AMT based hardware can now take control from firmware */
1312 	if (adapter->has_manage && adapter->has_amt)
1313 		em_get_hw_control(adapter);
1314 
1315 	/* Set Energy Efficient Ethernet */
1316 	if (adapter->hw.mac.type >= igb_mac_min &&
1317 	    adapter->hw.phy.media_type == e1000_media_type_copper) {
1318 		if (adapter->hw.mac.type == e1000_i354)
1319 			e1000_set_eee_i354(&adapter->hw, TRUE, TRUE);
1320 		else
1321 			e1000_set_eee_i350(&adapter->hw, TRUE, TRUE);
1322 	}
1323 }
1324 
1325 /*********************************************************************
1326  *
1327  *  Fast Legacy/MSI Combined Interrupt Service routine
1328  *
1329  *********************************************************************/
1330 int
1331 em_intr(void *arg)
1332 {
1333 	struct adapter *adapter = arg;
1334 	if_ctx_t ctx = adapter->ctx;
1335 	u32 reg_icr;
1336 
1337 	reg_icr = E1000_READ_REG(&adapter->hw, E1000_ICR);
1338 
1339 	if (adapter->intr_type != IFLIB_INTR_LEGACY)
1340 		goto skip_stray;
1341 	/* Hot eject? */
1342 	if (reg_icr == 0xffffffff)
1343 		return FILTER_STRAY;
1344 
1345 	/* Definitely not our interrupt. */
1346 	if (reg_icr == 0x0)
1347 		return FILTER_STRAY;
1348 
1349 	/*
1350 	 * Starting with the 82571 chip, bit 31 should be used to
1351 	 * determine whether the interrupt belongs to us.
1352 	 */
1353 	if (adapter->hw.mac.type >= e1000_82571 &&
1354 	    (reg_icr & E1000_ICR_INT_ASSERTED) == 0)
1355 		return FILTER_STRAY;
1356 
1357 skip_stray:
1358 	/* Link status change */
1359 	if (reg_icr & (E1000_ICR_RXSEQ | E1000_ICR_LSC)) {
1360 		adapter->hw.mac.get_link_status = 1;
1361 		iflib_admin_intr_deferred(ctx);
1362 	}
1363 
1364 	if (reg_icr & E1000_ICR_RXO)
1365 		adapter->rx_overruns++;
1366 
1367 	return (FILTER_SCHEDULE_THREAD);
1368 }
1369 
1370 static void
1371 igb_rx_enable_queue(struct adapter *adapter, struct em_rx_queue *rxq)
1372 {
1373 	E1000_WRITE_REG(&adapter->hw, E1000_EIMS, rxq->eims);
1374 }
1375 
1376 static void
1377 em_rx_enable_queue(struct adapter *adapter, struct em_rx_queue *rxq)
1378 {
1379 	E1000_WRITE_REG(&adapter->hw, E1000_IMS, rxq->eims);
1380 }
1381 
1382 static void
1383 igb_tx_enable_queue(struct adapter *adapter, struct em_tx_queue *txq)
1384 {
1385 	E1000_WRITE_REG(&adapter->hw, E1000_EIMS, txq->eims);
1386 }
1387 
1388 static void
1389 em_tx_enable_queue(struct adapter *adapter, struct em_tx_queue *txq)
1390 {
1391 	E1000_WRITE_REG(&adapter->hw, E1000_IMS, txq->eims);
1392 }
1393 
1394 static int
1395 em_if_rx_queue_intr_enable(if_ctx_t ctx, uint16_t rxqid)
1396 {
1397 	struct adapter *adapter = iflib_get_softc(ctx);
1398 	struct em_rx_queue *rxq = &adapter->rx_queues[rxqid];
1399 
1400 	if (adapter->hw.mac.type >= igb_mac_min)
1401 		igb_rx_enable_queue(adapter, rxq);
1402 	else
1403 		em_rx_enable_queue(adapter, rxq);
1404 	return (0);
1405 }
1406 
1407 static int
1408 em_if_tx_queue_intr_enable(if_ctx_t ctx, uint16_t txqid)
1409 {
1410 	struct adapter *adapter = iflib_get_softc(ctx);
1411 	struct em_tx_queue *txq = &adapter->tx_queues[txqid];
1412 
1413 	if (adapter->hw.mac.type >= igb_mac_min)
1414 		igb_tx_enable_queue(adapter, txq);
1415 	else
1416 		em_tx_enable_queue(adapter, txq);
1417 	return (0);
1418 }
1419 
1420 /*********************************************************************
1421  *
1422  *  MSIX RX Interrupt Service routine
1423  *
1424  **********************************************************************/
1425 static int
1426 em_msix_que(void *arg)
1427 {
1428 	struct em_rx_queue *que = arg;
1429 
1430 	++que->irqs;
1431 
1432 	return (FILTER_SCHEDULE_THREAD);
1433 }
1434 
1435 /*********************************************************************
1436  *
1437  *  MSIX Link Fast Interrupt Service routine
1438  *
1439  **********************************************************************/
1440 static int
1441 em_msix_link(void *arg)
1442 {
1443 	struct adapter *adapter = arg;
1444 	u32 reg_icr;
1445 
1446 	++adapter->link_irq;
1447 	MPASS(adapter->hw.back != NULL);
1448 	reg_icr = E1000_READ_REG(&adapter->hw, E1000_ICR);
1449 
1450 	if (reg_icr & E1000_ICR_RXO)
1451 		adapter->rx_overruns++;
1452 
1453 	if (reg_icr & (E1000_ICR_RXSEQ | E1000_ICR_LSC)) {
1454 		em_handle_link(adapter->ctx);
1455 	} else {
1456 		E1000_WRITE_REG(&adapter->hw, E1000_IMS,
1457 				EM_MSIX_LINK | E1000_IMS_LSC);
1458 		if (adapter->hw.mac.type >= igb_mac_min)
1459 			E1000_WRITE_REG(&adapter->hw, E1000_EIMS, adapter->link_mask);
1460 	}
1461 
1462 	/*
1463 	 * Because we must read the ICR for this interrupt
1464 	 * it may clear other causes using autoclear, for
1465 	 * this reason we simply create a soft interrupt
1466 	 * for all these vectors.
1467 	 */
1468 	if (reg_icr && adapter->hw.mac.type < igb_mac_min) {
1469 		E1000_WRITE_REG(&adapter->hw,
1470 			E1000_ICS, adapter->ims);
1471 	}
1472 
1473 	return (FILTER_HANDLED);
1474 }
1475 
1476 static void
1477 em_handle_link(void *context)
1478 {
1479 	if_ctx_t ctx = context;
1480 	struct adapter *adapter = iflib_get_softc(ctx);
1481 
1482 	adapter->hw.mac.get_link_status = 1;
1483 	iflib_admin_intr_deferred(ctx);
1484 }
1485 
1486 
1487 /*********************************************************************
1488  *
1489  *  Media Ioctl callback
1490  *
1491  *  This routine is called whenever the user queries the status of
1492  *  the interface using ifconfig.
1493  *
1494  **********************************************************************/
1495 static void
1496 em_if_media_status(if_ctx_t ctx, struct ifmediareq *ifmr)
1497 {
1498 	struct adapter *adapter = iflib_get_softc(ctx);
1499 	u_char fiber_type = IFM_1000_SX;
1500 
1501 	INIT_DEBUGOUT("em_if_media_status: begin");
1502 
1503 	iflib_admin_intr_deferred(ctx);
1504 
1505 	ifmr->ifm_status = IFM_AVALID;
1506 	ifmr->ifm_active = IFM_ETHER;
1507 
1508 	if (!adapter->link_active) {
1509 		return;
1510 	}
1511 
1512 	ifmr->ifm_status |= IFM_ACTIVE;
1513 
1514 	if ((adapter->hw.phy.media_type == e1000_media_type_fiber) ||
1515 	    (adapter->hw.phy.media_type == e1000_media_type_internal_serdes)) {
1516 		if (adapter->hw.mac.type == e1000_82545)
1517 			fiber_type = IFM_1000_LX;
1518 		ifmr->ifm_active |= fiber_type | IFM_FDX;
1519 	} else {
1520 		switch (adapter->link_speed) {
1521 		case 10:
1522 			ifmr->ifm_active |= IFM_10_T;
1523 			break;
1524 		case 100:
1525 			ifmr->ifm_active |= IFM_100_TX;
1526 			break;
1527 		case 1000:
1528 			ifmr->ifm_active |= IFM_1000_T;
1529 			break;
1530 		}
1531 		if (adapter->link_duplex == FULL_DUPLEX)
1532 			ifmr->ifm_active |= IFM_FDX;
1533 		else
1534 			ifmr->ifm_active |= IFM_HDX;
1535 	}
1536 }
1537 
1538 /*********************************************************************
1539  *
1540  *  Media Ioctl callback
1541  *
1542  *  This routine is called when the user changes speed/duplex using
1543  *  media/mediopt option with ifconfig.
1544  *
1545  **********************************************************************/
1546 static int
1547 em_if_media_change(if_ctx_t ctx)
1548 {
1549 	struct adapter *adapter = iflib_get_softc(ctx);
1550 	struct ifmedia *ifm = iflib_get_media(ctx);
1551 
1552 	INIT_DEBUGOUT("em_if_media_change: begin");
1553 
1554 	if (IFM_TYPE(ifm->ifm_media) != IFM_ETHER)
1555 		return (EINVAL);
1556 
1557 	switch (IFM_SUBTYPE(ifm->ifm_media)) {
1558 	case IFM_AUTO:
1559 		adapter->hw.mac.autoneg = DO_AUTO_NEG;
1560 		adapter->hw.phy.autoneg_advertised = AUTONEG_ADV_DEFAULT;
1561 		break;
1562 	case IFM_1000_LX:
1563 	case IFM_1000_SX:
1564 	case IFM_1000_T:
1565 		adapter->hw.mac.autoneg = DO_AUTO_NEG;
1566 		adapter->hw.phy.autoneg_advertised = ADVERTISE_1000_FULL;
1567 		break;
1568 	case IFM_100_TX:
1569 		adapter->hw.mac.autoneg = FALSE;
1570 		adapter->hw.phy.autoneg_advertised = 0;
1571 		if ((ifm->ifm_media & IFM_GMASK) == IFM_FDX)
1572 			adapter->hw.mac.forced_speed_duplex = ADVERTISE_100_FULL;
1573 		else
1574 			adapter->hw.mac.forced_speed_duplex = ADVERTISE_100_HALF;
1575 		break;
1576 	case IFM_10_T:
1577 		adapter->hw.mac.autoneg = FALSE;
1578 		adapter->hw.phy.autoneg_advertised = 0;
1579 		if ((ifm->ifm_media & IFM_GMASK) == IFM_FDX)
1580 			adapter->hw.mac.forced_speed_duplex = ADVERTISE_10_FULL;
1581 		else
1582 			adapter->hw.mac.forced_speed_duplex = ADVERTISE_10_HALF;
1583 		break;
1584 	default:
1585 		device_printf(adapter->dev, "Unsupported media type\n");
1586 	}
1587 
1588 	em_if_init(ctx);
1589 
1590 	return (0);
1591 }
1592 
1593 static int
1594 em_if_set_promisc(if_ctx_t ctx, int flags)
1595 {
1596 	struct adapter *adapter = iflib_get_softc(ctx);
1597 	u32 reg_rctl;
1598 
1599 	em_disable_promisc(ctx);
1600 
1601 	reg_rctl = E1000_READ_REG(&adapter->hw, E1000_RCTL);
1602 
1603 	if (flags & IFF_PROMISC) {
1604 		reg_rctl |= (E1000_RCTL_UPE | E1000_RCTL_MPE);
1605 		/* Turn this on if you want to see bad packets */
1606 		if (em_debug_sbp)
1607 			reg_rctl |= E1000_RCTL_SBP;
1608 		E1000_WRITE_REG(&adapter->hw, E1000_RCTL, reg_rctl);
1609 	} else if (flags & IFF_ALLMULTI) {
1610 		reg_rctl |= E1000_RCTL_MPE;
1611 		reg_rctl &= ~E1000_RCTL_UPE;
1612 		E1000_WRITE_REG(&adapter->hw, E1000_RCTL, reg_rctl);
1613 	}
1614 	return (0);
1615 }
1616 
1617 static void
1618 em_disable_promisc(if_ctx_t ctx)
1619 {
1620 	struct adapter *adapter = iflib_get_softc(ctx);
1621 	struct ifnet *ifp = iflib_get_ifp(ctx);
1622 	u32 reg_rctl;
1623 	int mcnt = 0;
1624 
1625 	reg_rctl = E1000_READ_REG(&adapter->hw, E1000_RCTL);
1626 	reg_rctl &= (~E1000_RCTL_UPE);
1627 	if (if_getflags(ifp) & IFF_ALLMULTI)
1628 		mcnt = MAX_NUM_MULTICAST_ADDRESSES;
1629 	else
1630 		mcnt = if_multiaddr_count(ifp, MAX_NUM_MULTICAST_ADDRESSES);
1631 	/* Don't disable if in MAX groups */
1632 	if (mcnt < MAX_NUM_MULTICAST_ADDRESSES)
1633 		reg_rctl &=  (~E1000_RCTL_MPE);
1634 	reg_rctl &=  (~E1000_RCTL_SBP);
1635 	E1000_WRITE_REG(&adapter->hw, E1000_RCTL, reg_rctl);
1636 }
1637 
1638 
1639 /*********************************************************************
1640  *  Multicast Update
1641  *
1642  *  This routine is called whenever multicast address list is updated.
1643  *
1644  **********************************************************************/
1645 
1646 static void
1647 em_if_multi_set(if_ctx_t ctx)
1648 {
1649 	struct adapter *adapter = iflib_get_softc(ctx);
1650 	struct ifnet *ifp = iflib_get_ifp(ctx);
1651 	u32 reg_rctl = 0;
1652 	u8  *mta; /* Multicast array memory */
1653 	int mcnt = 0;
1654 
1655 	IOCTL_DEBUGOUT("em_set_multi: begin");
1656 
1657 	mta = adapter->mta;
1658 	bzero(mta, sizeof(u8) * ETH_ADDR_LEN * MAX_NUM_MULTICAST_ADDRESSES);
1659 
1660 	if (adapter->hw.mac.type == e1000_82542 &&
1661 	    adapter->hw.revision_id == E1000_REVISION_2) {
1662 		reg_rctl = E1000_READ_REG(&adapter->hw, E1000_RCTL);
1663 		if (adapter->hw.bus.pci_cmd_word & CMD_MEM_WRT_INVALIDATE)
1664 			e1000_pci_clear_mwi(&adapter->hw);
1665 		reg_rctl |= E1000_RCTL_RST;
1666 		E1000_WRITE_REG(&adapter->hw, E1000_RCTL, reg_rctl);
1667 		msec_delay(5);
1668 	}
1669 
1670 	if_multiaddr_array(ifp, mta, &mcnt, MAX_NUM_MULTICAST_ADDRESSES);
1671 
1672 	if (mcnt >= MAX_NUM_MULTICAST_ADDRESSES) {
1673 		reg_rctl = E1000_READ_REG(&adapter->hw, E1000_RCTL);
1674 		reg_rctl |= E1000_RCTL_MPE;
1675 		E1000_WRITE_REG(&adapter->hw, E1000_RCTL, reg_rctl);
1676 	} else
1677 		e1000_update_mc_addr_list(&adapter->hw, mta, mcnt);
1678 
1679 	if (adapter->hw.mac.type == e1000_82542 &&
1680 	    adapter->hw.revision_id == E1000_REVISION_2) {
1681 		reg_rctl = E1000_READ_REG(&adapter->hw, E1000_RCTL);
1682 		reg_rctl &= ~E1000_RCTL_RST;
1683 		E1000_WRITE_REG(&adapter->hw, E1000_RCTL, reg_rctl);
1684 		msec_delay(5);
1685 		if (adapter->hw.bus.pci_cmd_word & CMD_MEM_WRT_INVALIDATE)
1686 			e1000_pci_set_mwi(&adapter->hw);
1687 	}
1688 }
1689 
1690 
1691 /*********************************************************************
1692  *  Timer routine
1693  *
1694  *  This routine checks for link status and updates statistics.
1695  *
1696  **********************************************************************/
1697 
1698 static void
1699 em_if_timer(if_ctx_t ctx, uint16_t qid)
1700 {
1701 	struct adapter *adapter = iflib_get_softc(ctx);
1702 	struct em_rx_queue *que;
1703 	int i;
1704 	int trigger = 0;
1705 
1706 	if (qid != 0)
1707 		return;
1708 
1709 	iflib_admin_intr_deferred(ctx);
1710 
1711 	/* Mask to use in the irq trigger */
1712 	if (adapter->intr_type == IFLIB_INTR_MSIX) {
1713 		for (i = 0, que = adapter->rx_queues; i < adapter->rx_num_queues; i++, que++)
1714 			trigger |= que->eims;
1715 	} else {
1716 		trigger = E1000_ICS_RXDMT0;
1717 	}
1718 }
1719 
1720 
1721 static void
1722 em_if_update_admin_status(if_ctx_t ctx)
1723 {
1724 	struct adapter *adapter = iflib_get_softc(ctx);
1725 	struct e1000_hw *hw = &adapter->hw;
1726 	device_t dev = iflib_get_dev(ctx);
1727 	u32 link_check, thstat, ctrl;
1728 
1729 	link_check = thstat = ctrl = 0;
1730 	/* Get the cached link value or read phy for real */
1731 	switch (hw->phy.media_type) {
1732 	case e1000_media_type_copper:
1733 		if (hw->mac.get_link_status) {
1734 			if (hw->mac.type == e1000_pch_spt)
1735 				msec_delay(50);
1736 			/* Do the work to read phy */
1737 			e1000_check_for_link(hw);
1738 			link_check = !hw->mac.get_link_status;
1739 			if (link_check) /* ESB2 fix */
1740 				e1000_cfg_on_link_up(hw);
1741 		} else {
1742 			link_check = TRUE;
1743 		}
1744 		break;
1745 	case e1000_media_type_fiber:
1746 		e1000_check_for_link(hw);
1747 		link_check = (E1000_READ_REG(hw, E1000_STATUS) &
1748 			    E1000_STATUS_LU);
1749 		break;
1750 	case e1000_media_type_internal_serdes:
1751 		e1000_check_for_link(hw);
1752 		link_check = adapter->hw.mac.serdes_has_link;
1753 		break;
1754 	/* VF device is type_unknown */
1755 	case e1000_media_type_unknown:
1756 		e1000_check_for_link(hw);
1757 		link_check = !hw->mac.get_link_status;
1758 		/* FALLTHROUGH */
1759 	default:
1760 		break;
1761 	}
1762 
1763 	/* Check for thermal downshift or shutdown */
1764 	if (hw->mac.type == e1000_i350) {
1765 		thstat = E1000_READ_REG(hw, E1000_THSTAT);
1766 		ctrl = E1000_READ_REG(hw, E1000_CTRL_EXT);
1767 	}
1768 
1769 	/* Now check for a transition */
1770 	if (link_check && (adapter->link_active == 0)) {
1771 		e1000_get_speed_and_duplex(hw, &adapter->link_speed,
1772 		    &adapter->link_duplex);
1773 		/* Check if we must disable SPEED_MODE bit on PCI-E */
1774 		if ((adapter->link_speed != SPEED_1000) &&
1775 		    ((hw->mac.type == e1000_82571) ||
1776 		    (hw->mac.type == e1000_82572))) {
1777 			int tarc0;
1778 			tarc0 = E1000_READ_REG(hw, E1000_TARC(0));
1779 			tarc0 &= ~TARC_SPEED_MODE_BIT;
1780 			E1000_WRITE_REG(hw, E1000_TARC(0), tarc0);
1781 		}
1782 		if (bootverbose)
1783 			device_printf(dev, "Link is up %d Mbps %s\n",
1784 			    adapter->link_speed,
1785 			    ((adapter->link_duplex == FULL_DUPLEX) ?
1786 			    "Full Duplex" : "Half Duplex"));
1787 		adapter->link_active = 1;
1788 		adapter->smartspeed = 0;
1789 		if ((ctrl & E1000_CTRL_EXT_LINK_MODE_MASK) ==
1790 		    E1000_CTRL_EXT_LINK_MODE_GMII &&
1791 		    (thstat & E1000_THSTAT_LINK_THROTTLE))
1792 			device_printf(dev, "Link: thermal downshift\n");
1793 		/* Delay Link Up for Phy update */
1794 		if (((hw->mac.type == e1000_i210) ||
1795 		    (hw->mac.type == e1000_i211)) &&
1796 		    (hw->phy.id == I210_I_PHY_ID))
1797 			msec_delay(I210_LINK_DELAY);
1798 		/* Reset if the media type changed. */
1799 		if ((hw->dev_spec._82575.media_changed) &&
1800 			(adapter->hw.mac.type >= igb_mac_min)) {
1801 			hw->dev_spec._82575.media_changed = false;
1802 			adapter->flags |= IGB_MEDIA_RESET;
1803 			em_reset(ctx);
1804 		}
1805 		iflib_link_state_change(ctx, LINK_STATE_UP,
1806 		    IF_Mbps(adapter->link_speed));
1807 	} else if (!link_check && (adapter->link_active == 1)) {
1808 		adapter->link_speed = 0;
1809 		adapter->link_duplex = 0;
1810 		adapter->link_active = 0;
1811 		iflib_link_state_change(ctx, LINK_STATE_DOWN, 0);
1812 	}
1813 	em_update_stats_counters(adapter);
1814 
1815 	/* Reset LAA into RAR[0] on 82571 */
1816 	if ((adapter->hw.mac.type == e1000_82571) &&
1817 	    e1000_get_laa_state_82571(&adapter->hw))
1818 		e1000_rar_set(&adapter->hw, adapter->hw.mac.addr, 0);
1819 
1820 	if (adapter->hw.mac.type < em_mac_min)
1821 		lem_smartspeed(adapter);
1822 
1823 	E1000_WRITE_REG(&adapter->hw, E1000_IMS, EM_MSIX_LINK | E1000_IMS_LSC);
1824 }
1825 
1826 /*********************************************************************
1827  *
1828  *  This routine disables all traffic on the adapter by issuing a
1829  *  global reset on the MAC and deallocates TX/RX buffers.
1830  *
1831  *  This routine should always be called with BOTH the CORE
1832  *  and TX locks.
1833  **********************************************************************/
1834 
1835 static void
1836 em_if_stop(if_ctx_t ctx)
1837 {
1838 	struct adapter *adapter = iflib_get_softc(ctx);
1839 
1840 	INIT_DEBUGOUT("em_stop: begin");
1841 
1842 	e1000_reset_hw(&adapter->hw);
1843 	if (adapter->hw.mac.type >= e1000_82544)
1844 		E1000_WRITE_REG(&adapter->hw, E1000_WUFC, 0);
1845 
1846 	e1000_led_off(&adapter->hw);
1847 	e1000_cleanup_led(&adapter->hw);
1848 }
1849 
1850 
1851 /*********************************************************************
1852  *
1853  *  Determine hardware revision.
1854  *
1855  **********************************************************************/
1856 static void
1857 em_identify_hardware(if_ctx_t ctx)
1858 {
1859 	device_t dev = iflib_get_dev(ctx);
1860 	struct adapter *adapter = iflib_get_softc(ctx);
1861 
1862 	/* Make sure our PCI config space has the necessary stuff set */
1863 	adapter->hw.bus.pci_cmd_word = pci_read_config(dev, PCIR_COMMAND, 2);
1864 
1865 	/* Save off the information about this board */
1866 	adapter->hw.vendor_id = pci_get_vendor(dev);
1867 	adapter->hw.device_id = pci_get_device(dev);
1868 	adapter->hw.revision_id = pci_read_config(dev, PCIR_REVID, 1);
1869 	adapter->hw.subsystem_vendor_id =
1870 	    pci_read_config(dev, PCIR_SUBVEND_0, 2);
1871 	adapter->hw.subsystem_device_id =
1872 	    pci_read_config(dev, PCIR_SUBDEV_0, 2);
1873 
1874 	/* Do Shared Code Init and Setup */
1875 	if (e1000_set_mac_type(&adapter->hw)) {
1876 		device_printf(dev, "Setup init failure\n");
1877 		return;
1878 	}
1879 }
1880 
1881 static int
1882 em_allocate_pci_resources(if_ctx_t ctx)
1883 {
1884 	struct adapter *adapter = iflib_get_softc(ctx);
1885 	device_t dev = iflib_get_dev(ctx);
1886 	int rid, val;
1887 
1888 	rid = PCIR_BAR(0);
1889 	adapter->memory = bus_alloc_resource_any(dev, SYS_RES_MEMORY,
1890 	    &rid, RF_ACTIVE);
1891 	if (adapter->memory == NULL) {
1892 		device_printf(dev, "Unable to allocate bus resource: memory\n");
1893 		return (ENXIO);
1894 	}
1895 	adapter->osdep.mem_bus_space_tag = rman_get_bustag(adapter->memory);
1896 	adapter->osdep.mem_bus_space_handle =
1897 	    rman_get_bushandle(adapter->memory);
1898 	adapter->hw.hw_addr = (u8 *)&adapter->osdep.mem_bus_space_handle;
1899 
1900 	/* Only older adapters use IO mapping */
1901 	if (adapter->hw.mac.type < em_mac_min &&
1902 	    adapter->hw.mac.type > e1000_82543) {
1903 		/* Figure our where our IO BAR is ? */
1904 		for (rid = PCIR_BAR(0); rid < PCIR_CIS;) {
1905 			val = pci_read_config(dev, rid, 4);
1906 			if (EM_BAR_TYPE(val) == EM_BAR_TYPE_IO) {
1907 				adapter->io_rid = rid;
1908 				break;
1909 			}
1910 			rid += 4;
1911 			/* check for 64bit BAR */
1912 			if (EM_BAR_MEM_TYPE(val) == EM_BAR_MEM_TYPE_64BIT)
1913 				rid += 4;
1914 		}
1915 		if (rid >= PCIR_CIS) {
1916 			device_printf(dev, "Unable to locate IO BAR\n");
1917 			return (ENXIO);
1918 		}
1919 		adapter->ioport = bus_alloc_resource_any(dev,
1920 		    SYS_RES_IOPORT, &adapter->io_rid, RF_ACTIVE);
1921 		if (adapter->ioport == NULL) {
1922 			device_printf(dev, "Unable to allocate bus resource: "
1923 			    "ioport\n");
1924 			return (ENXIO);
1925 		}
1926 		adapter->hw.io_base = 0;
1927 		adapter->osdep.io_bus_space_tag =
1928 		    rman_get_bustag(adapter->ioport);
1929 		adapter->osdep.io_bus_space_handle =
1930 		    rman_get_bushandle(adapter->ioport);
1931 	}
1932 
1933 	adapter->hw.back = &adapter->osdep;
1934 
1935 	return (0);
1936 }
1937 
1938 /*********************************************************************
1939  *
1940  *  Setup the MSIX Interrupt handlers
1941  *
1942  **********************************************************************/
1943 static int
1944 em_if_msix_intr_assign(if_ctx_t ctx, int msix)
1945 {
1946 	struct adapter *adapter = iflib_get_softc(ctx);
1947 	struct em_rx_queue *rx_que = adapter->rx_queues;
1948 	struct em_tx_queue *tx_que = adapter->tx_queues;
1949 	int error, rid, i, vector = 0, rx_vectors;
1950 	char buf[16];
1951 
1952 	/* First set up ring resources */
1953 	for (i = 0; i < adapter->rx_num_queues; i++, rx_que++, vector++) {
1954 		rid = vector + 1;
1955 		snprintf(buf, sizeof(buf), "rxq%d", i);
1956 		error = iflib_irq_alloc_generic(ctx, &rx_que->que_irq, rid, IFLIB_INTR_RXTX, em_msix_que, rx_que, rx_que->me, buf);
1957 		if (error) {
1958 			device_printf(iflib_get_dev(ctx), "Failed to allocate que int %d err: %d", i, error);
1959 			adapter->rx_num_queues = i + 1;
1960 			goto fail;
1961 		}
1962 
1963 		rx_que->msix =  vector;
1964 
1965 		/*
1966 		 * Set the bit to enable interrupt
1967 		 * in E1000_IMS -- bits 20 and 21
1968 		 * are for RX0 and RX1, note this has
1969 		 * NOTHING to do with the MSIX vector
1970 		 */
1971 		if (adapter->hw.mac.type == e1000_82574) {
1972 			rx_que->eims = 1 << (20 + i);
1973 			adapter->ims |= rx_que->eims;
1974 			adapter->ivars |= (8 | rx_que->msix) << (i * 4);
1975 		} else if (adapter->hw.mac.type == e1000_82575)
1976 			rx_que->eims = E1000_EICR_TX_QUEUE0 << vector;
1977 		else
1978 			rx_que->eims = 1 << vector;
1979 	}
1980 	rx_vectors = vector;
1981 
1982 	vector = 0;
1983 	for (i = 0; i < adapter->tx_num_queues; i++, tx_que++, vector++) {
1984 		snprintf(buf, sizeof(buf), "txq%d", i);
1985 		tx_que = &adapter->tx_queues[i];
1986 		iflib_softirq_alloc_generic(ctx,
1987 		    &adapter->rx_queues[i % adapter->rx_num_queues].que_irq,
1988 		    IFLIB_INTR_TX, tx_que, tx_que->me, buf);
1989 
1990 		tx_que->msix = (vector % adapter->tx_num_queues);
1991 
1992 		/*
1993 		 * Set the bit to enable interrupt
1994 		 * in E1000_IMS -- bits 22 and 23
1995 		 * are for TX0 and TX1, note this has
1996 		 * NOTHING to do with the MSIX vector
1997 		 */
1998 		if (adapter->hw.mac.type == e1000_82574) {
1999 			tx_que->eims = 1 << (22 + i);
2000 			adapter->ims |= tx_que->eims;
2001 			adapter->ivars |= (8 | tx_que->msix) << (8 + (i * 4));
2002 		} else if (adapter->hw.mac.type == e1000_82575) {
2003 			tx_que->eims = E1000_EICR_TX_QUEUE0 << (i %  adapter->tx_num_queues);
2004 		} else {
2005 			tx_que->eims = 1 << (i %  adapter->tx_num_queues);
2006 		}
2007 	}
2008 
2009 	/* Link interrupt */
2010 	rid = rx_vectors + 1;
2011 	error = iflib_irq_alloc_generic(ctx, &adapter->irq, rid, IFLIB_INTR_ADMIN, em_msix_link, adapter, 0, "aq");
2012 
2013 	if (error) {
2014 		device_printf(iflib_get_dev(ctx), "Failed to register admin handler");
2015 		goto fail;
2016 	}
2017 	adapter->linkvec = rx_vectors;
2018 	if (adapter->hw.mac.type < igb_mac_min) {
2019 		adapter->ivars |=  (8 | rx_vectors) << 16;
2020 		adapter->ivars |= 0x80000000;
2021 	}
2022 	return (0);
2023 fail:
2024 	iflib_irq_free(ctx, &adapter->irq);
2025 	rx_que = adapter->rx_queues;
2026 	for (int i = 0; i < adapter->rx_num_queues; i++, rx_que++)
2027 		iflib_irq_free(ctx, &rx_que->que_irq);
2028 	return (error);
2029 }
2030 
2031 static void
2032 igb_configure_queues(struct adapter *adapter)
2033 {
2034 	struct e1000_hw *hw = &adapter->hw;
2035 	struct em_rx_queue *rx_que;
2036 	struct em_tx_queue *tx_que;
2037 	u32 tmp, ivar = 0, newitr = 0;
2038 
2039 	/* First turn on RSS capability */
2040 	if (adapter->hw.mac.type != e1000_82575)
2041 		E1000_WRITE_REG(hw, E1000_GPIE,
2042 		    E1000_GPIE_MSIX_MODE | E1000_GPIE_EIAME |
2043 		    E1000_GPIE_PBA | E1000_GPIE_NSICR);
2044 
2045 	/* Turn on MSIX */
2046 	switch (adapter->hw.mac.type) {
2047 	case e1000_82580:
2048 	case e1000_i350:
2049 	case e1000_i354:
2050 	case e1000_i210:
2051 	case e1000_i211:
2052 	case e1000_vfadapt:
2053 	case e1000_vfadapt_i350:
2054 		/* RX entries */
2055 		for (int i = 0; i < adapter->rx_num_queues; i++) {
2056 			u32 index = i >> 1;
2057 			ivar = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index);
2058 			rx_que = &adapter->rx_queues[i];
2059 			if (i & 1) {
2060 				ivar &= 0xFF00FFFF;
2061 				ivar |= (rx_que->msix | E1000_IVAR_VALID) << 16;
2062 			} else {
2063 				ivar &= 0xFFFFFF00;
2064 				ivar |= rx_que->msix | E1000_IVAR_VALID;
2065 			}
2066 			E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, ivar);
2067 		}
2068 		/* TX entries */
2069 		for (int i = 0; i < adapter->tx_num_queues; i++) {
2070 			u32 index = i >> 1;
2071 			ivar = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index);
2072 			tx_que = &adapter->tx_queues[i];
2073 			if (i & 1) {
2074 				ivar &= 0x00FFFFFF;
2075 				ivar |= (tx_que->msix | E1000_IVAR_VALID) << 24;
2076 			} else {
2077 				ivar &= 0xFFFF00FF;
2078 				ivar |= (tx_que->msix | E1000_IVAR_VALID) << 8;
2079 			}
2080 			E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, ivar);
2081 			adapter->que_mask |= tx_que->eims;
2082 		}
2083 
2084 		/* And for the link interrupt */
2085 		ivar = (adapter->linkvec | E1000_IVAR_VALID) << 8;
2086 		adapter->link_mask = 1 << adapter->linkvec;
2087 		E1000_WRITE_REG(hw, E1000_IVAR_MISC, ivar);
2088 		break;
2089 	case e1000_82576:
2090 		/* RX entries */
2091 		for (int i = 0; i < adapter->rx_num_queues; i++) {
2092 			u32 index = i & 0x7; /* Each IVAR has two entries */
2093 			ivar = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index);
2094 			rx_que = &adapter->rx_queues[i];
2095 			if (i < 8) {
2096 				ivar &= 0xFFFFFF00;
2097 				ivar |= rx_que->msix | E1000_IVAR_VALID;
2098 			} else {
2099 				ivar &= 0xFF00FFFF;
2100 				ivar |= (rx_que->msix | E1000_IVAR_VALID) << 16;
2101 			}
2102 			E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, ivar);
2103 			adapter->que_mask |= rx_que->eims;
2104 		}
2105 		/* TX entries */
2106 		for (int i = 0; i < adapter->tx_num_queues; i++) {
2107 			u32 index = i & 0x7; /* Each IVAR has two entries */
2108 			ivar = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index);
2109 			tx_que = &adapter->tx_queues[i];
2110 			if (i < 8) {
2111 				ivar &= 0xFFFF00FF;
2112 				ivar |= (tx_que->msix | E1000_IVAR_VALID) << 8;
2113 			} else {
2114 				ivar &= 0x00FFFFFF;
2115 				ivar |= (tx_que->msix | E1000_IVAR_VALID) << 24;
2116 			}
2117 			E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, ivar);
2118 			adapter->que_mask |= tx_que->eims;
2119 		}
2120 
2121 		/* And for the link interrupt */
2122 		ivar = (adapter->linkvec | E1000_IVAR_VALID) << 8;
2123 		adapter->link_mask = 1 << adapter->linkvec;
2124 		E1000_WRITE_REG(hw, E1000_IVAR_MISC, ivar);
2125 		break;
2126 
2127 	case e1000_82575:
2128 		/* enable MSI-X support*/
2129 		tmp = E1000_READ_REG(hw, E1000_CTRL_EXT);
2130 		tmp |= E1000_CTRL_EXT_PBA_CLR;
2131 		/* Auto-Mask interrupts upon ICR read. */
2132 		tmp |= E1000_CTRL_EXT_EIAME;
2133 		tmp |= E1000_CTRL_EXT_IRCA;
2134 		E1000_WRITE_REG(hw, E1000_CTRL_EXT, tmp);
2135 
2136 		/* Queues */
2137 		for (int i = 0; i < adapter->rx_num_queues; i++) {
2138 			rx_que = &adapter->rx_queues[i];
2139 			tmp = E1000_EICR_RX_QUEUE0 << i;
2140 			tmp |= E1000_EICR_TX_QUEUE0 << i;
2141 			rx_que->eims = tmp;
2142 			E1000_WRITE_REG_ARRAY(hw, E1000_MSIXBM(0),
2143 			    i, rx_que->eims);
2144 			adapter->que_mask |= rx_que->eims;
2145 		}
2146 
2147 		/* Link */
2148 		E1000_WRITE_REG(hw, E1000_MSIXBM(adapter->linkvec),
2149 		    E1000_EIMS_OTHER);
2150 		adapter->link_mask |= E1000_EIMS_OTHER;
2151 	default:
2152 		break;
2153 	}
2154 
2155 	/* Set the starting interrupt rate */
2156 	if (em_max_interrupt_rate > 0)
2157 		newitr = (4000000 / em_max_interrupt_rate) & 0x7FFC;
2158 
2159 	if (hw->mac.type == e1000_82575)
2160 		newitr |= newitr << 16;
2161 	else
2162 		newitr |= E1000_EITR_CNT_IGNR;
2163 
2164 	for (int i = 0; i < adapter->rx_num_queues; i++) {
2165 		rx_que = &adapter->rx_queues[i];
2166 		E1000_WRITE_REG(hw, E1000_EITR(rx_que->msix), newitr);
2167 	}
2168 
2169 	return;
2170 }
2171 
2172 static void
2173 em_free_pci_resources(if_ctx_t ctx)
2174 {
2175 	struct adapter *adapter = iflib_get_softc(ctx);
2176 	struct em_rx_queue *que = adapter->rx_queues;
2177 	device_t dev = iflib_get_dev(ctx);
2178 
2179 	/* Release all msix queue resources */
2180 	if (adapter->intr_type == IFLIB_INTR_MSIX)
2181 		iflib_irq_free(ctx, &adapter->irq);
2182 
2183 	for (int i = 0; i < adapter->rx_num_queues; i++, que++) {
2184 		iflib_irq_free(ctx, &que->que_irq);
2185 	}
2186 
2187 	/* First release all the interrupt resources */
2188 	if (adapter->memory != NULL) {
2189 		bus_release_resource(dev, SYS_RES_MEMORY,
2190 				     PCIR_BAR(0), adapter->memory);
2191 		adapter->memory = NULL;
2192 	}
2193 
2194 	if (adapter->flash != NULL) {
2195 		bus_release_resource(dev, SYS_RES_MEMORY,
2196 				     EM_FLASH, adapter->flash);
2197 		adapter->flash = NULL;
2198 	}
2199 	if (adapter->ioport != NULL)
2200 		bus_release_resource(dev, SYS_RES_IOPORT,
2201 		    adapter->io_rid, adapter->ioport);
2202 }
2203 
2204 /* Setup MSI or MSI/X */
2205 static int
2206 em_setup_msix(if_ctx_t ctx)
2207 {
2208 	struct adapter *adapter = iflib_get_softc(ctx);
2209 
2210 	if (adapter->hw.mac.type == e1000_82574) {
2211 		em_enable_vectors_82574(ctx);
2212 	}
2213 	return (0);
2214 }
2215 
2216 /*********************************************************************
2217  *
2218  *  Initialize the hardware to a configuration
2219  *  as specified by the adapter structure.
2220  *
2221  **********************************************************************/
2222 
2223 static void
2224 lem_smartspeed(struct adapter *adapter)
2225 {
2226 	u16 phy_tmp;
2227 
2228 	if (adapter->link_active || (adapter->hw.phy.type != e1000_phy_igp) ||
2229 	    adapter->hw.mac.autoneg == 0 ||
2230 	    (adapter->hw.phy.autoneg_advertised & ADVERTISE_1000_FULL) == 0)
2231 		return;
2232 
2233 	if (adapter->smartspeed == 0) {
2234 		/* If Master/Slave config fault is asserted twice,
2235 		 * we assume back-to-back */
2236 		e1000_read_phy_reg(&adapter->hw, PHY_1000T_STATUS, &phy_tmp);
2237 		if (!(phy_tmp & SR_1000T_MS_CONFIG_FAULT))
2238 			return;
2239 		e1000_read_phy_reg(&adapter->hw, PHY_1000T_STATUS, &phy_tmp);
2240 		if (phy_tmp & SR_1000T_MS_CONFIG_FAULT) {
2241 			e1000_read_phy_reg(&adapter->hw,
2242 			    PHY_1000T_CTRL, &phy_tmp);
2243 			if(phy_tmp & CR_1000T_MS_ENABLE) {
2244 				phy_tmp &= ~CR_1000T_MS_ENABLE;
2245 				e1000_write_phy_reg(&adapter->hw,
2246 				    PHY_1000T_CTRL, phy_tmp);
2247 				adapter->smartspeed++;
2248 				if(adapter->hw.mac.autoneg &&
2249 				   !e1000_copper_link_autoneg(&adapter->hw) &&
2250 				   !e1000_read_phy_reg(&adapter->hw,
2251 				    PHY_CONTROL, &phy_tmp)) {
2252 					phy_tmp |= (MII_CR_AUTO_NEG_EN |
2253 						    MII_CR_RESTART_AUTO_NEG);
2254 					e1000_write_phy_reg(&adapter->hw,
2255 					    PHY_CONTROL, phy_tmp);
2256 				}
2257 			}
2258 		}
2259 		return;
2260 	} else if(adapter->smartspeed == EM_SMARTSPEED_DOWNSHIFT) {
2261 		/* If still no link, perhaps using 2/3 pair cable */
2262 		e1000_read_phy_reg(&adapter->hw, PHY_1000T_CTRL, &phy_tmp);
2263 		phy_tmp |= CR_1000T_MS_ENABLE;
2264 		e1000_write_phy_reg(&adapter->hw, PHY_1000T_CTRL, phy_tmp);
2265 		if(adapter->hw.mac.autoneg &&
2266 		   !e1000_copper_link_autoneg(&adapter->hw) &&
2267 		   !e1000_read_phy_reg(&adapter->hw, PHY_CONTROL, &phy_tmp)) {
2268 			phy_tmp |= (MII_CR_AUTO_NEG_EN |
2269 				    MII_CR_RESTART_AUTO_NEG);
2270 			e1000_write_phy_reg(&adapter->hw, PHY_CONTROL, phy_tmp);
2271 		}
2272 	}
2273 	/* Restart process after EM_SMARTSPEED_MAX iterations */
2274 	if(adapter->smartspeed++ == EM_SMARTSPEED_MAX)
2275 		adapter->smartspeed = 0;
2276 }
2277 
2278 /*********************************************************************
2279  *
2280  *  Initialize the DMA Coalescing feature
2281  *
2282  **********************************************************************/
2283 static void
2284 igb_init_dmac(struct adapter *adapter, u32 pba)
2285 {
2286 	device_t	dev = adapter->dev;
2287 	struct e1000_hw *hw = &adapter->hw;
2288 	u32 		dmac, reg = ~E1000_DMACR_DMAC_EN;
2289 	u16		hwm;
2290 	u16		max_frame_size;
2291 
2292 	if (hw->mac.type == e1000_i211)
2293 		return;
2294 
2295 	max_frame_size = adapter->shared->isc_max_frame_size;
2296 	if (hw->mac.type > e1000_82580) {
2297 
2298 		if (adapter->dmac == 0) { /* Disabling it */
2299 			E1000_WRITE_REG(hw, E1000_DMACR, reg);
2300 			return;
2301 		} else
2302 			device_printf(dev, "DMA Coalescing enabled\n");
2303 
2304 		/* Set starting threshold */
2305 		E1000_WRITE_REG(hw, E1000_DMCTXTH, 0);
2306 
2307 		hwm = 64 * pba - max_frame_size / 16;
2308 		if (hwm < 64 * (pba - 6))
2309 			hwm = 64 * (pba - 6);
2310 		reg = E1000_READ_REG(hw, E1000_FCRTC);
2311 		reg &= ~E1000_FCRTC_RTH_COAL_MASK;
2312 		reg |= ((hwm << E1000_FCRTC_RTH_COAL_SHIFT)
2313 		    & E1000_FCRTC_RTH_COAL_MASK);
2314 		E1000_WRITE_REG(hw, E1000_FCRTC, reg);
2315 
2316 
2317 		dmac = pba - max_frame_size / 512;
2318 		if (dmac < pba - 10)
2319 			dmac = pba - 10;
2320 		reg = E1000_READ_REG(hw, E1000_DMACR);
2321 		reg &= ~E1000_DMACR_DMACTHR_MASK;
2322 		reg |= ((dmac << E1000_DMACR_DMACTHR_SHIFT)
2323 		    & E1000_DMACR_DMACTHR_MASK);
2324 
2325 		/* transition to L0x or L1 if available..*/
2326 		reg |= (E1000_DMACR_DMAC_EN | E1000_DMACR_DMAC_LX_MASK);
2327 
2328 		/* Check if status is 2.5Gb backplane connection
2329 		* before configuration of watchdog timer, which is
2330 		* in msec values in 12.8usec intervals
2331 		* watchdog timer= msec values in 32usec intervals
2332 		* for non 2.5Gb connection
2333 		*/
2334 		if (hw->mac.type == e1000_i354) {
2335 			int status = E1000_READ_REG(hw, E1000_STATUS);
2336 			if ((status & E1000_STATUS_2P5_SKU) &&
2337 			    (!(status & E1000_STATUS_2P5_SKU_OVER)))
2338 				reg |= ((adapter->dmac * 5) >> 6);
2339 			else
2340 				reg |= (adapter->dmac >> 5);
2341 		} else {
2342 			reg |= (adapter->dmac >> 5);
2343 		}
2344 
2345 		E1000_WRITE_REG(hw, E1000_DMACR, reg);
2346 
2347 		E1000_WRITE_REG(hw, E1000_DMCRTRH, 0);
2348 
2349 		/* Set the interval before transition */
2350 		reg = E1000_READ_REG(hw, E1000_DMCTLX);
2351 		if (hw->mac.type == e1000_i350)
2352 			reg |= IGB_DMCTLX_DCFLUSH_DIS;
2353 		/*
2354 		** in 2.5Gb connection, TTLX unit is 0.4 usec
2355 		** which is 0x4*2 = 0xA. But delay is still 4 usec
2356 		*/
2357 		if (hw->mac.type == e1000_i354) {
2358 			int status = E1000_READ_REG(hw, E1000_STATUS);
2359 			if ((status & E1000_STATUS_2P5_SKU) &&
2360 			    (!(status & E1000_STATUS_2P5_SKU_OVER)))
2361 				reg |= 0xA;
2362 			else
2363 				reg |= 0x4;
2364 		} else {
2365 			reg |= 0x4;
2366 		}
2367 
2368 		E1000_WRITE_REG(hw, E1000_DMCTLX, reg);
2369 
2370 		/* free space in tx packet buffer to wake from DMA coal */
2371 		E1000_WRITE_REG(hw, E1000_DMCTXTH, (IGB_TXPBSIZE -
2372 		    (2 * max_frame_size)) >> 6);
2373 
2374 		/* make low power state decision controlled by DMA coal */
2375 		reg = E1000_READ_REG(hw, E1000_PCIEMISC);
2376 		reg &= ~E1000_PCIEMISC_LX_DECISION;
2377 		E1000_WRITE_REG(hw, E1000_PCIEMISC, reg);
2378 
2379 	} else if (hw->mac.type == e1000_82580) {
2380 		u32 reg = E1000_READ_REG(hw, E1000_PCIEMISC);
2381 		E1000_WRITE_REG(hw, E1000_PCIEMISC,
2382 		    reg & ~E1000_PCIEMISC_LX_DECISION);
2383 		E1000_WRITE_REG(hw, E1000_DMACR, 0);
2384 	}
2385 }
2386 
2387 static void
2388 em_reset(if_ctx_t ctx)
2389 {
2390 	device_t dev = iflib_get_dev(ctx);
2391 	struct adapter *adapter = iflib_get_softc(ctx);
2392 	struct ifnet *ifp = iflib_get_ifp(ctx);
2393 	struct e1000_hw *hw = &adapter->hw;
2394 	u16 rx_buffer_size;
2395 	u32 pba;
2396 
2397 	INIT_DEBUGOUT("em_reset: begin");
2398 	/* Let the firmware know the OS is in control */
2399 	em_get_hw_control(adapter);
2400 
2401 	/* Set up smart power down as default off on newer adapters. */
2402 	if (!em_smart_pwr_down && (hw->mac.type == e1000_82571 ||
2403 	    hw->mac.type == e1000_82572)) {
2404 		u16 phy_tmp = 0;
2405 
2406 		/* Speed up time to link by disabling smart power down. */
2407 		e1000_read_phy_reg(hw, IGP02E1000_PHY_POWER_MGMT, &phy_tmp);
2408 		phy_tmp &= ~IGP02E1000_PM_SPD;
2409 		e1000_write_phy_reg(hw, IGP02E1000_PHY_POWER_MGMT, phy_tmp);
2410 	}
2411 
2412 	/*
2413 	 * Packet Buffer Allocation (PBA)
2414 	 * Writing PBA sets the receive portion of the buffer
2415 	 * the remainder is used for the transmit buffer.
2416 	 */
2417 	switch (hw->mac.type) {
2418 	/* Total Packet Buffer on these is 48K */
2419 	case e1000_82571:
2420 	case e1000_82572:
2421 	case e1000_80003es2lan:
2422 			pba = E1000_PBA_32K; /* 32K for Rx, 16K for Tx */
2423 		break;
2424 	case e1000_82573: /* 82573: Total Packet Buffer is 32K */
2425 			pba = E1000_PBA_12K; /* 12K for Rx, 20K for Tx */
2426 		break;
2427 	case e1000_82574:
2428 	case e1000_82583:
2429 			pba = E1000_PBA_20K; /* 20K for Rx, 20K for Tx */
2430 		break;
2431 	case e1000_ich8lan:
2432 		pba = E1000_PBA_8K;
2433 		break;
2434 	case e1000_ich9lan:
2435 	case e1000_ich10lan:
2436 		/* Boost Receive side for jumbo frames */
2437 		if (adapter->hw.mac.max_frame_size > 4096)
2438 			pba = E1000_PBA_14K;
2439 		else
2440 			pba = E1000_PBA_10K;
2441 		break;
2442 	case e1000_pchlan:
2443 	case e1000_pch2lan:
2444 	case e1000_pch_lpt:
2445 	case e1000_pch_spt:
2446 	case e1000_pch_cnp:
2447 		pba = E1000_PBA_26K;
2448 		break;
2449 	case e1000_82575:
2450 		pba = E1000_PBA_32K;
2451 		break;
2452 	case e1000_82576:
2453 	case e1000_vfadapt:
2454 		pba = E1000_READ_REG(hw, E1000_RXPBS);
2455 		pba &= E1000_RXPBS_SIZE_MASK_82576;
2456 		break;
2457 	case e1000_82580:
2458 	case e1000_i350:
2459 	case e1000_i354:
2460 	case e1000_vfadapt_i350:
2461 		pba = E1000_READ_REG(hw, E1000_RXPBS);
2462 		pba = e1000_rxpbs_adjust_82580(pba);
2463 		break;
2464 	case e1000_i210:
2465 	case e1000_i211:
2466 		pba = E1000_PBA_34K;
2467 		break;
2468 	default:
2469 		if (adapter->hw.mac.max_frame_size > 8192)
2470 			pba = E1000_PBA_40K; /* 40K for Rx, 24K for Tx */
2471 		else
2472 			pba = E1000_PBA_48K; /* 48K for Rx, 16K for Tx */
2473 	}
2474 
2475 	/* Special needs in case of Jumbo frames */
2476 	if ((hw->mac.type == e1000_82575) && (ifp->if_mtu > ETHERMTU)) {
2477 		u32 tx_space, min_tx, min_rx;
2478 		pba = E1000_READ_REG(hw, E1000_PBA);
2479 		tx_space = pba >> 16;
2480 		pba &= 0xffff;
2481 		min_tx = (adapter->hw.mac.max_frame_size +
2482 		    sizeof(struct e1000_tx_desc) - ETHERNET_FCS_SIZE) * 2;
2483 		min_tx = roundup2(min_tx, 1024);
2484 		min_tx >>= 10;
2485 		min_rx = adapter->hw.mac.max_frame_size;
2486 		min_rx = roundup2(min_rx, 1024);
2487 		min_rx >>= 10;
2488 		if (tx_space < min_tx &&
2489 		    ((min_tx - tx_space) < pba)) {
2490 			pba = pba - (min_tx - tx_space);
2491 			/*
2492 			 * if short on rx space, rx wins
2493 			 * and must trump tx adjustment
2494 			 */
2495 			if (pba < min_rx)
2496 				pba = min_rx;
2497 		}
2498 		E1000_WRITE_REG(hw, E1000_PBA, pba);
2499 	}
2500 
2501 	if (hw->mac.type < igb_mac_min)
2502 		E1000_WRITE_REG(&adapter->hw, E1000_PBA, pba);
2503 
2504 	INIT_DEBUGOUT1("em_reset: pba=%dK",pba);
2505 
2506 	/*
2507 	 * These parameters control the automatic generation (Tx) and
2508 	 * response (Rx) to Ethernet PAUSE frames.
2509 	 * - High water mark should allow for at least two frames to be
2510 	 *   received after sending an XOFF.
2511 	 * - Low water mark works best when it is very near the high water mark.
2512 	 *   This allows the receiver to restart by sending XON when it has
2513 	 *   drained a bit. Here we use an arbitrary value of 1500 which will
2514 	 *   restart after one full frame is pulled from the buffer. There
2515 	 *   could be several smaller frames in the buffer and if so they will
2516 	 *   not trigger the XON until their total number reduces the buffer
2517 	 *   by 1500.
2518 	 * - The pause time is fairly large at 1000 x 512ns = 512 usec.
2519 	 */
2520 	rx_buffer_size = (pba & 0xffff) << 10;
2521 	hw->fc.high_water = rx_buffer_size -
2522 	    roundup2(adapter->hw.mac.max_frame_size, 1024);
2523 	hw->fc.low_water = hw->fc.high_water - 1500;
2524 
2525 	if (adapter->fc) /* locally set flow control value? */
2526 		hw->fc.requested_mode = adapter->fc;
2527 	else
2528 		hw->fc.requested_mode = e1000_fc_full;
2529 
2530 	if (hw->mac.type == e1000_80003es2lan)
2531 		hw->fc.pause_time = 0xFFFF;
2532 	else
2533 		hw->fc.pause_time = EM_FC_PAUSE_TIME;
2534 
2535 	hw->fc.send_xon = TRUE;
2536 
2537 	/* Device specific overrides/settings */
2538 	switch (hw->mac.type) {
2539 	case e1000_pchlan:
2540 		/* Workaround: no TX flow ctrl for PCH */
2541 		hw->fc.requested_mode = e1000_fc_rx_pause;
2542 		hw->fc.pause_time = 0xFFFF; /* override */
2543 		if (if_getmtu(ifp) > ETHERMTU) {
2544 			hw->fc.high_water = 0x3500;
2545 			hw->fc.low_water = 0x1500;
2546 		} else {
2547 			hw->fc.high_water = 0x5000;
2548 			hw->fc.low_water = 0x3000;
2549 		}
2550 		hw->fc.refresh_time = 0x1000;
2551 		break;
2552 	case e1000_pch2lan:
2553 	case e1000_pch_lpt:
2554 	case e1000_pch_spt:
2555 	case e1000_pch_cnp:
2556 		hw->fc.high_water = 0x5C20;
2557 		hw->fc.low_water = 0x5048;
2558 		hw->fc.pause_time = 0x0650;
2559 		hw->fc.refresh_time = 0x0400;
2560 		/* Jumbos need adjusted PBA */
2561 		if (if_getmtu(ifp) > ETHERMTU)
2562 			E1000_WRITE_REG(hw, E1000_PBA, 12);
2563 		else
2564 			E1000_WRITE_REG(hw, E1000_PBA, 26);
2565 		break;
2566 	case e1000_82575:
2567 	case e1000_82576:
2568 		/* 8-byte granularity */
2569 		hw->fc.low_water = hw->fc.high_water - 8;
2570 		break;
2571 	case e1000_82580:
2572 	case e1000_i350:
2573 	case e1000_i354:
2574 	case e1000_i210:
2575 	case e1000_i211:
2576 	case e1000_vfadapt:
2577 	case e1000_vfadapt_i350:
2578 		/* 16-byte granularity */
2579 		hw->fc.low_water = hw->fc.high_water - 16;
2580 		break;
2581 	case e1000_ich9lan:
2582 	case e1000_ich10lan:
2583 		if (if_getmtu(ifp) > ETHERMTU) {
2584 			hw->fc.high_water = 0x2800;
2585 			hw->fc.low_water = hw->fc.high_water - 8;
2586 			break;
2587 		}
2588 		/* FALLTHROUGH */
2589 	default:
2590 		if (hw->mac.type == e1000_80003es2lan)
2591 			hw->fc.pause_time = 0xFFFF;
2592 		break;
2593 	}
2594 
2595 	/* Issue a global reset */
2596 	e1000_reset_hw(hw);
2597 	if (adapter->hw.mac.type >= igb_mac_min) {
2598 		E1000_WRITE_REG(hw, E1000_WUC, 0);
2599 	} else {
2600 		E1000_WRITE_REG(hw, E1000_WUFC, 0);
2601 		em_disable_aspm(adapter);
2602 	}
2603 	if (adapter->flags & IGB_MEDIA_RESET) {
2604 		e1000_setup_init_funcs(hw, TRUE);
2605 		e1000_get_bus_info(hw);
2606 		adapter->flags &= ~IGB_MEDIA_RESET;
2607 	}
2608 	/* and a re-init */
2609 	if (e1000_init_hw(hw) < 0) {
2610 		device_printf(dev, "Hardware Initialization Failed\n");
2611 		return;
2612 	}
2613 	if (adapter->hw.mac.type >= igb_mac_min)
2614 		igb_init_dmac(adapter, pba);
2615 
2616 	E1000_WRITE_REG(hw, E1000_VET, ETHERTYPE_VLAN);
2617 	e1000_get_phy_info(hw);
2618 	e1000_check_for_link(hw);
2619 }
2620 
2621 #define RSSKEYLEN 10
2622 static void
2623 em_initialize_rss_mapping(struct adapter *adapter)
2624 {
2625 	uint8_t  rss_key[4 * RSSKEYLEN];
2626 	uint32_t reta = 0;
2627 	struct e1000_hw	*hw = &adapter->hw;
2628 	int i;
2629 
2630 	/*
2631 	 * Configure RSS key
2632 	 */
2633 	arc4rand(rss_key, sizeof(rss_key), 0);
2634 	for (i = 0; i < RSSKEYLEN; ++i) {
2635 		uint32_t rssrk = 0;
2636 
2637 		rssrk = EM_RSSRK_VAL(rss_key, i);
2638 		E1000_WRITE_REG(hw,E1000_RSSRK(i), rssrk);
2639 	}
2640 
2641 	/*
2642 	 * Configure RSS redirect table in following fashion:
2643 	 * (hash & ring_cnt_mask) == rdr_table[(hash & rdr_table_mask)]
2644 	 */
2645 	for (i = 0; i < sizeof(reta); ++i) {
2646 		uint32_t q;
2647 
2648 		q = (i % adapter->rx_num_queues) << 7;
2649 		reta |= q << (8 * i);
2650 	}
2651 
2652 	for (i = 0; i < 32; ++i)
2653 		E1000_WRITE_REG(hw, E1000_RETA(i), reta);
2654 
2655 	E1000_WRITE_REG(hw, E1000_MRQC, E1000_MRQC_RSS_ENABLE_2Q |
2656 			E1000_MRQC_RSS_FIELD_IPV4_TCP |
2657 			E1000_MRQC_RSS_FIELD_IPV4 |
2658 			E1000_MRQC_RSS_FIELD_IPV6_TCP_EX |
2659 			E1000_MRQC_RSS_FIELD_IPV6_EX |
2660 			E1000_MRQC_RSS_FIELD_IPV6);
2661 
2662 }
2663 
2664 static void
2665 igb_initialize_rss_mapping(struct adapter *adapter)
2666 {
2667 	struct e1000_hw *hw = &adapter->hw;
2668 	int i;
2669 	int queue_id;
2670 	u32 reta;
2671 	u32 rss_key[10], mrqc, shift = 0;
2672 
2673 	/* XXX? */
2674 	if (adapter->hw.mac.type == e1000_82575)
2675 		shift = 6;
2676 
2677 	/*
2678 	 * The redirection table controls which destination
2679 	 * queue each bucket redirects traffic to.
2680 	 * Each DWORD represents four queues, with the LSB
2681 	 * being the first queue in the DWORD.
2682 	 *
2683 	 * This just allocates buckets to queues using round-robin
2684 	 * allocation.
2685 	 *
2686 	 * NOTE: It Just Happens to line up with the default
2687 	 * RSS allocation method.
2688 	 */
2689 
2690 	/* Warning FM follows */
2691 	reta = 0;
2692 	for (i = 0; i < 128; i++) {
2693 #ifdef RSS
2694 		queue_id = rss_get_indirection_to_bucket(i);
2695 		/*
2696 		 * If we have more queues than buckets, we'll
2697 		 * end up mapping buckets to a subset of the
2698 		 * queues.
2699 		 *
2700 		 * If we have more buckets than queues, we'll
2701 		 * end up instead assigning multiple buckets
2702 		 * to queues.
2703 		 *
2704 		 * Both are suboptimal, but we need to handle
2705 		 * the case so we don't go out of bounds
2706 		 * indexing arrays and such.
2707 		 */
2708 		queue_id = queue_id % adapter->rx_num_queues;
2709 #else
2710 		queue_id = (i % adapter->rx_num_queues);
2711 #endif
2712 		/* Adjust if required */
2713 		queue_id = queue_id << shift;
2714 
2715 		/*
2716 		 * The low 8 bits are for hash value (n+0);
2717 		 * The next 8 bits are for hash value (n+1), etc.
2718 		 */
2719 		reta = reta >> 8;
2720 		reta = reta | ( ((uint32_t) queue_id) << 24);
2721 		if ((i & 3) == 3) {
2722 			E1000_WRITE_REG(hw, E1000_RETA(i >> 2), reta);
2723 			reta = 0;
2724 		}
2725 	}
2726 
2727 	/* Now fill in hash table */
2728 
2729 	/*
2730 	 * MRQC: Multiple Receive Queues Command
2731 	 * Set queuing to RSS control, number depends on the device.
2732 	 */
2733 	mrqc = E1000_MRQC_ENABLE_RSS_8Q;
2734 
2735 #ifdef RSS
2736 	/* XXX ew typecasting */
2737 	rss_getkey((uint8_t *) &rss_key);
2738 #else
2739 	arc4rand(&rss_key, sizeof(rss_key), 0);
2740 #endif
2741 	for (i = 0; i < 10; i++)
2742 		E1000_WRITE_REG_ARRAY(hw, E1000_RSSRK(0), i, rss_key[i]);
2743 
2744 	/*
2745 	 * Configure the RSS fields to hash upon.
2746 	 */
2747 	mrqc |= (E1000_MRQC_RSS_FIELD_IPV4 |
2748 	    E1000_MRQC_RSS_FIELD_IPV4_TCP);
2749 	mrqc |= (E1000_MRQC_RSS_FIELD_IPV6 |
2750 	    E1000_MRQC_RSS_FIELD_IPV6_TCP);
2751 	mrqc |=( E1000_MRQC_RSS_FIELD_IPV4_UDP |
2752 	    E1000_MRQC_RSS_FIELD_IPV6_UDP);
2753 	mrqc |=( E1000_MRQC_RSS_FIELD_IPV6_UDP_EX |
2754 	    E1000_MRQC_RSS_FIELD_IPV6_TCP_EX);
2755 
2756 	E1000_WRITE_REG(hw, E1000_MRQC, mrqc);
2757 }
2758 
2759 /*********************************************************************
2760  *
2761  *  Setup networking device structure and register an interface.
2762  *
2763  **********************************************************************/
2764 static int
2765 em_setup_interface(if_ctx_t ctx)
2766 {
2767 	struct ifnet *ifp = iflib_get_ifp(ctx);
2768 	struct adapter *adapter = iflib_get_softc(ctx);
2769 	if_softc_ctx_t scctx = adapter->shared;
2770 
2771 	INIT_DEBUGOUT("em_setup_interface: begin");
2772 
2773 	/* Single Queue */
2774 	if (adapter->tx_num_queues == 1) {
2775 		if_setsendqlen(ifp, scctx->isc_ntxd[0] - 1);
2776 		if_setsendqready(ifp);
2777 	}
2778 
2779 	/*
2780 	 * Specify the media types supported by this adapter and register
2781 	 * callbacks to update media and link information
2782 	 */
2783 	if ((adapter->hw.phy.media_type == e1000_media_type_fiber) ||
2784 	    (adapter->hw.phy.media_type == e1000_media_type_internal_serdes)) {
2785 		u_char fiber_type = IFM_1000_SX;	/* default type */
2786 
2787 		if (adapter->hw.mac.type == e1000_82545)
2788 			fiber_type = IFM_1000_LX;
2789 		ifmedia_add(adapter->media, IFM_ETHER | fiber_type | IFM_FDX, 0, NULL);
2790 		ifmedia_add(adapter->media, IFM_ETHER | fiber_type, 0, NULL);
2791 	} else {
2792 		ifmedia_add(adapter->media, IFM_ETHER | IFM_10_T, 0, NULL);
2793 		ifmedia_add(adapter->media, IFM_ETHER | IFM_10_T | IFM_FDX, 0, NULL);
2794 		ifmedia_add(adapter->media, IFM_ETHER | IFM_100_TX, 0, NULL);
2795 		ifmedia_add(adapter->media, IFM_ETHER | IFM_100_TX | IFM_FDX, 0, NULL);
2796 		if (adapter->hw.phy.type != e1000_phy_ife) {
2797 			ifmedia_add(adapter->media, IFM_ETHER | IFM_1000_T | IFM_FDX, 0, NULL);
2798 			ifmedia_add(adapter->media, IFM_ETHER | IFM_1000_T, 0, NULL);
2799 		}
2800 	}
2801 	ifmedia_add(adapter->media, IFM_ETHER | IFM_AUTO, 0, NULL);
2802 	ifmedia_set(adapter->media, IFM_ETHER | IFM_AUTO);
2803 	return (0);
2804 }
2805 
2806 static int
2807 em_if_tx_queues_alloc(if_ctx_t ctx, caddr_t *vaddrs, uint64_t *paddrs, int ntxqs, int ntxqsets)
2808 {
2809 	struct adapter *adapter = iflib_get_softc(ctx);
2810 	if_softc_ctx_t scctx = adapter->shared;
2811 	int error = E1000_SUCCESS;
2812 	struct em_tx_queue *que;
2813 	int i, j;
2814 
2815 	MPASS(adapter->tx_num_queues > 0);
2816 	MPASS(adapter->tx_num_queues == ntxqsets);
2817 
2818 	/* First allocate the top level queue structs */
2819 	if (!(adapter->tx_queues =
2820 	    (struct em_tx_queue *) malloc(sizeof(struct em_tx_queue) *
2821 	    adapter->tx_num_queues, M_DEVBUF, M_NOWAIT | M_ZERO))) {
2822 		device_printf(iflib_get_dev(ctx), "Unable to allocate queue memory\n");
2823 		return(ENOMEM);
2824 	}
2825 
2826 	for (i = 0, que = adapter->tx_queues; i < adapter->tx_num_queues; i++, que++) {
2827 		/* Set up some basics */
2828 
2829 		struct tx_ring *txr = &que->txr;
2830 		txr->adapter = que->adapter = adapter;
2831 		que->me = txr->me =  i;
2832 
2833 		/* Allocate report status array */
2834 		if (!(txr->tx_rsq = (qidx_t *) malloc(sizeof(qidx_t) * scctx->isc_ntxd[0], M_DEVBUF, M_NOWAIT | M_ZERO))) {
2835 			device_printf(iflib_get_dev(ctx), "failed to allocate rs_idxs memory\n");
2836 			error = ENOMEM;
2837 			goto fail;
2838 		}
2839 		for (j = 0; j < scctx->isc_ntxd[0]; j++)
2840 			txr->tx_rsq[j] = QIDX_INVALID;
2841 		/* get the virtual and physical address of the hardware queues */
2842 		txr->tx_base = (struct e1000_tx_desc *)vaddrs[i*ntxqs];
2843 		txr->tx_paddr = paddrs[i*ntxqs];
2844 	}
2845 
2846 	device_printf(iflib_get_dev(ctx), "allocated for %d tx_queues\n", adapter->tx_num_queues);
2847 	return (0);
2848 fail:
2849 	em_if_queues_free(ctx);
2850 	return (error);
2851 }
2852 
2853 static int
2854 em_if_rx_queues_alloc(if_ctx_t ctx, caddr_t *vaddrs, uint64_t *paddrs, int nrxqs, int nrxqsets)
2855 {
2856 	struct adapter *adapter = iflib_get_softc(ctx);
2857 	int error = E1000_SUCCESS;
2858 	struct em_rx_queue *que;
2859 	int i;
2860 
2861 	MPASS(adapter->rx_num_queues > 0);
2862 	MPASS(adapter->rx_num_queues == nrxqsets);
2863 
2864 	/* First allocate the top level queue structs */
2865 	if (!(adapter->rx_queues =
2866 	    (struct em_rx_queue *) malloc(sizeof(struct em_rx_queue) *
2867 	    adapter->rx_num_queues, M_DEVBUF, M_NOWAIT | M_ZERO))) {
2868 		device_printf(iflib_get_dev(ctx), "Unable to allocate queue memory\n");
2869 		error = ENOMEM;
2870 		goto fail;
2871 	}
2872 
2873 	for (i = 0, que = adapter->rx_queues; i < nrxqsets; i++, que++) {
2874 		/* Set up some basics */
2875 		struct rx_ring *rxr = &que->rxr;
2876 		rxr->adapter = que->adapter = adapter;
2877 		rxr->que = que;
2878 		que->me = rxr->me =  i;
2879 
2880 		/* get the virtual and physical address of the hardware queues */
2881 		rxr->rx_base = (union e1000_rx_desc_extended *)vaddrs[i*nrxqs];
2882 		rxr->rx_paddr = paddrs[i*nrxqs];
2883 	}
2884 
2885 	device_printf(iflib_get_dev(ctx), "allocated for %d rx_queues\n", adapter->rx_num_queues);
2886 
2887 	return (0);
2888 fail:
2889 	em_if_queues_free(ctx);
2890 	return (error);
2891 }
2892 
2893 static void
2894 em_if_queues_free(if_ctx_t ctx)
2895 {
2896 	struct adapter *adapter = iflib_get_softc(ctx);
2897 	struct em_tx_queue *tx_que = adapter->tx_queues;
2898 	struct em_rx_queue *rx_que = adapter->rx_queues;
2899 
2900 	if (tx_que != NULL) {
2901 		for (int i = 0; i < adapter->tx_num_queues; i++, tx_que++) {
2902 			struct tx_ring *txr = &tx_que->txr;
2903 			if (txr->tx_rsq == NULL)
2904 				break;
2905 
2906 			free(txr->tx_rsq, M_DEVBUF);
2907 			txr->tx_rsq = NULL;
2908 		}
2909 		free(adapter->tx_queues, M_DEVBUF);
2910 		adapter->tx_queues = NULL;
2911 	}
2912 
2913 	if (rx_que != NULL) {
2914 		free(adapter->rx_queues, M_DEVBUF);
2915 		adapter->rx_queues = NULL;
2916 	}
2917 
2918 	em_release_hw_control(adapter);
2919 
2920 	if (adapter->mta != NULL) {
2921 		free(adapter->mta, M_DEVBUF);
2922 	}
2923 }
2924 
2925 /*********************************************************************
2926  *
2927  *  Enable transmit unit.
2928  *
2929  **********************************************************************/
2930 static void
2931 em_initialize_transmit_unit(if_ctx_t ctx)
2932 {
2933 	struct adapter *adapter = iflib_get_softc(ctx);
2934 	if_softc_ctx_t scctx = adapter->shared;
2935 	struct em_tx_queue *que;
2936 	struct tx_ring	*txr;
2937 	struct e1000_hw	*hw = &adapter->hw;
2938 	u32 tctl, txdctl = 0, tarc, tipg = 0;
2939 
2940 	INIT_DEBUGOUT("em_initialize_transmit_unit: begin");
2941 
2942 	for (int i = 0; i < adapter->tx_num_queues; i++, txr++) {
2943 		u64 bus_addr;
2944 		caddr_t offp, endp;
2945 
2946 		que = &adapter->tx_queues[i];
2947 		txr = &que->txr;
2948 		bus_addr = txr->tx_paddr;
2949 
2950 		/* Clear checksum offload context. */
2951 		offp = (caddr_t)&txr->csum_flags;
2952 		endp = (caddr_t)(txr + 1);
2953 		bzero(offp, endp - offp);
2954 
2955 		/* Base and Len of TX Ring */
2956 		E1000_WRITE_REG(hw, E1000_TDLEN(i),
2957 		    scctx->isc_ntxd[0] * sizeof(struct e1000_tx_desc));
2958 		E1000_WRITE_REG(hw, E1000_TDBAH(i),
2959 		    (u32)(bus_addr >> 32));
2960 		E1000_WRITE_REG(hw, E1000_TDBAL(i),
2961 		    (u32)bus_addr);
2962 		/* Init the HEAD/TAIL indices */
2963 		E1000_WRITE_REG(hw, E1000_TDT(i), 0);
2964 		E1000_WRITE_REG(hw, E1000_TDH(i), 0);
2965 
2966 		HW_DEBUGOUT2("Base = %x, Length = %x\n",
2967 		    E1000_READ_REG(&adapter->hw, E1000_TDBAL(i)),
2968 		    E1000_READ_REG(&adapter->hw, E1000_TDLEN(i)));
2969 
2970 		txdctl = 0; /* clear txdctl */
2971 		txdctl |= 0x1f; /* PTHRESH */
2972 		txdctl |= 1 << 8; /* HTHRESH */
2973 		txdctl |= 1 << 16;/* WTHRESH */
2974 		txdctl |= 1 << 22; /* Reserved bit 22 must always be 1 */
2975 		txdctl |= E1000_TXDCTL_GRAN;
2976 		txdctl |= 1 << 25; /* LWTHRESH */
2977 
2978 		E1000_WRITE_REG(hw, E1000_TXDCTL(i), txdctl);
2979 	}
2980 
2981 	/* Set the default values for the Tx Inter Packet Gap timer */
2982 	switch (adapter->hw.mac.type) {
2983 	case e1000_80003es2lan:
2984 		tipg = DEFAULT_82543_TIPG_IPGR1;
2985 		tipg |= DEFAULT_80003ES2LAN_TIPG_IPGR2 <<
2986 		    E1000_TIPG_IPGR2_SHIFT;
2987 		break;
2988 	case e1000_82542:
2989 		tipg = DEFAULT_82542_TIPG_IPGT;
2990 		tipg |= DEFAULT_82542_TIPG_IPGR1 << E1000_TIPG_IPGR1_SHIFT;
2991 		tipg |= DEFAULT_82542_TIPG_IPGR2 << E1000_TIPG_IPGR2_SHIFT;
2992 		break;
2993 	default:
2994 		if ((adapter->hw.phy.media_type == e1000_media_type_fiber) ||
2995 		    (adapter->hw.phy.media_type ==
2996 		    e1000_media_type_internal_serdes))
2997 			tipg = DEFAULT_82543_TIPG_IPGT_FIBER;
2998 		else
2999 			tipg = DEFAULT_82543_TIPG_IPGT_COPPER;
3000 		tipg |= DEFAULT_82543_TIPG_IPGR1 << E1000_TIPG_IPGR1_SHIFT;
3001 		tipg |= DEFAULT_82543_TIPG_IPGR2 << E1000_TIPG_IPGR2_SHIFT;
3002 	}
3003 
3004 	E1000_WRITE_REG(&adapter->hw, E1000_TIPG, tipg);
3005 	E1000_WRITE_REG(&adapter->hw, E1000_TIDV, adapter->tx_int_delay.value);
3006 
3007 	if(adapter->hw.mac.type >= e1000_82540)
3008 		E1000_WRITE_REG(&adapter->hw, E1000_TADV,
3009 		    adapter->tx_abs_int_delay.value);
3010 
3011 	if ((adapter->hw.mac.type == e1000_82571) ||
3012 	    (adapter->hw.mac.type == e1000_82572)) {
3013 		tarc = E1000_READ_REG(&adapter->hw, E1000_TARC(0));
3014 		tarc |= TARC_SPEED_MODE_BIT;
3015 		E1000_WRITE_REG(&adapter->hw, E1000_TARC(0), tarc);
3016 	} else if (adapter->hw.mac.type == e1000_80003es2lan) {
3017 		/* errata: program both queues to unweighted RR */
3018 		tarc = E1000_READ_REG(&adapter->hw, E1000_TARC(0));
3019 		tarc |= 1;
3020 		E1000_WRITE_REG(&adapter->hw, E1000_TARC(0), tarc);
3021 		tarc = E1000_READ_REG(&adapter->hw, E1000_TARC(1));
3022 		tarc |= 1;
3023 		E1000_WRITE_REG(&adapter->hw, E1000_TARC(1), tarc);
3024 	} else if (adapter->hw.mac.type == e1000_82574) {
3025 		tarc = E1000_READ_REG(&adapter->hw, E1000_TARC(0));
3026 		tarc |= TARC_ERRATA_BIT;
3027 		if ( adapter->tx_num_queues > 1) {
3028 			tarc |= (TARC_COMPENSATION_MODE | TARC_MQ_FIX);
3029 			E1000_WRITE_REG(&adapter->hw, E1000_TARC(0), tarc);
3030 			E1000_WRITE_REG(&adapter->hw, E1000_TARC(1), tarc);
3031 		} else
3032 			E1000_WRITE_REG(&adapter->hw, E1000_TARC(0), tarc);
3033 	}
3034 
3035 	if (adapter->tx_int_delay.value > 0)
3036 		adapter->txd_cmd |= E1000_TXD_CMD_IDE;
3037 
3038 	/* Program the Transmit Control Register */
3039 	tctl = E1000_READ_REG(&adapter->hw, E1000_TCTL);
3040 	tctl &= ~E1000_TCTL_CT;
3041 	tctl |= (E1000_TCTL_PSP | E1000_TCTL_RTLC | E1000_TCTL_EN |
3042 		   (E1000_COLLISION_THRESHOLD << E1000_CT_SHIFT));
3043 
3044 	if (adapter->hw.mac.type >= e1000_82571)
3045 		tctl |= E1000_TCTL_MULR;
3046 
3047 	/* This write will effectively turn on the transmit unit. */
3048 	E1000_WRITE_REG(&adapter->hw, E1000_TCTL, tctl);
3049 
3050 	/* SPT and KBL errata workarounds */
3051 	if (hw->mac.type == e1000_pch_spt) {
3052 		u32 reg;
3053 		reg = E1000_READ_REG(hw, E1000_IOSFPC);
3054 		reg |= E1000_RCTL_RDMTS_HEX;
3055 		E1000_WRITE_REG(hw, E1000_IOSFPC, reg);
3056 		/* i218-i219 Specification Update 1.5.4.5 */
3057 		reg = E1000_READ_REG(hw, E1000_TARC(0));
3058 		reg &= ~E1000_TARC0_CB_MULTIQ_3_REQ;
3059 		reg |= E1000_TARC0_CB_MULTIQ_2_REQ;
3060 		E1000_WRITE_REG(hw, E1000_TARC(0), reg);
3061 	}
3062 }
3063 
3064 /*********************************************************************
3065  *
3066  *  Enable receive unit.
3067  *
3068  **********************************************************************/
3069 
3070 static void
3071 em_initialize_receive_unit(if_ctx_t ctx)
3072 {
3073 	struct adapter *adapter = iflib_get_softc(ctx);
3074 	if_softc_ctx_t scctx = adapter->shared;
3075 	struct ifnet *ifp = iflib_get_ifp(ctx);
3076 	struct e1000_hw	*hw = &adapter->hw;
3077 	struct em_rx_queue *que;
3078 	int i;
3079 	u32 rctl, rxcsum, rfctl;
3080 
3081 	INIT_DEBUGOUT("em_initialize_receive_units: begin");
3082 
3083 	/*
3084 	 * Make sure receives are disabled while setting
3085 	 * up the descriptor ring
3086 	 */
3087 	rctl = E1000_READ_REG(hw, E1000_RCTL);
3088 	/* Do not disable if ever enabled on this hardware */
3089 	if ((hw->mac.type != e1000_82574) && (hw->mac.type != e1000_82583))
3090 		E1000_WRITE_REG(hw, E1000_RCTL, rctl & ~E1000_RCTL_EN);
3091 
3092 	/* Setup the Receive Control Register */
3093 	rctl &= ~(3 << E1000_RCTL_MO_SHIFT);
3094 	rctl |= E1000_RCTL_EN | E1000_RCTL_BAM |
3095 	    E1000_RCTL_LBM_NO | E1000_RCTL_RDMTS_HALF |
3096 	    (hw->mac.mc_filter_type << E1000_RCTL_MO_SHIFT);
3097 
3098 	/* Do not store bad packets */
3099 	rctl &= ~E1000_RCTL_SBP;
3100 
3101 	/* Enable Long Packet receive */
3102 	if (if_getmtu(ifp) > ETHERMTU)
3103 		rctl |= E1000_RCTL_LPE;
3104 	else
3105 		rctl &= ~E1000_RCTL_LPE;
3106 
3107 	/* Strip the CRC */
3108 	if (!em_disable_crc_stripping)
3109 		rctl |= E1000_RCTL_SECRC;
3110 
3111 	if (adapter->hw.mac.type >= e1000_82540) {
3112 		E1000_WRITE_REG(&adapter->hw, E1000_RADV,
3113 			    adapter->rx_abs_int_delay.value);
3114 
3115 		/*
3116 		 * Set the interrupt throttling rate. Value is calculated
3117 		 * as DEFAULT_ITR = 1/(MAX_INTS_PER_SEC * 256ns)
3118 		 */
3119 		E1000_WRITE_REG(hw, E1000_ITR, DEFAULT_ITR);
3120 	}
3121 	E1000_WRITE_REG(&adapter->hw, E1000_RDTR,
3122 	    adapter->rx_int_delay.value);
3123 
3124 	/* Use extended rx descriptor formats */
3125 	rfctl = E1000_READ_REG(hw, E1000_RFCTL);
3126 	rfctl |= E1000_RFCTL_EXTEN;
3127 	/*
3128 	 * When using MSIX interrupts we need to throttle
3129 	 * using the EITR register (82574 only)
3130 	 */
3131 	if (hw->mac.type == e1000_82574) {
3132 		for (int i = 0; i < 4; i++)
3133 			E1000_WRITE_REG(hw, E1000_EITR_82574(i),
3134 			    DEFAULT_ITR);
3135 		/* Disable accelerated acknowledge */
3136 		rfctl |= E1000_RFCTL_ACK_DIS;
3137 	}
3138 	E1000_WRITE_REG(hw, E1000_RFCTL, rfctl);
3139 
3140 	rxcsum = E1000_READ_REG(hw, E1000_RXCSUM);
3141 	if (if_getcapenable(ifp) & IFCAP_RXCSUM &&
3142 	    adapter->hw.mac.type >= e1000_82543) {
3143 		if (adapter->tx_num_queues > 1) {
3144 			if (adapter->hw.mac.type >= igb_mac_min) {
3145 				rxcsum |= E1000_RXCSUM_PCSD;
3146 				if (hw->mac.type != e1000_82575)
3147 					rxcsum |= E1000_RXCSUM_CRCOFL;
3148 			} else
3149 				rxcsum |= E1000_RXCSUM_TUOFL |
3150 					E1000_RXCSUM_IPOFL |
3151 					E1000_RXCSUM_PCSD;
3152 		} else {
3153 			if (adapter->hw.mac.type >= igb_mac_min)
3154 				rxcsum |= E1000_RXCSUM_IPPCSE;
3155 			else
3156 				rxcsum |= E1000_RXCSUM_TUOFL | E1000_RXCSUM_IPOFL;
3157 			if (adapter->hw.mac.type > e1000_82575)
3158 				rxcsum |= E1000_RXCSUM_CRCOFL;
3159 		}
3160 	} else
3161 		rxcsum &= ~E1000_RXCSUM_TUOFL;
3162 
3163 	E1000_WRITE_REG(hw, E1000_RXCSUM, rxcsum);
3164 
3165 	if (adapter->rx_num_queues > 1) {
3166 		if (adapter->hw.mac.type >= igb_mac_min)
3167 			igb_initialize_rss_mapping(adapter);
3168 		else
3169 			em_initialize_rss_mapping(adapter);
3170 	}
3171 
3172 	/*
3173 	 * XXX TEMPORARY WORKAROUND: on some systems with 82573
3174 	 * long latencies are observed, like Lenovo X60. This
3175 	 * change eliminates the problem, but since having positive
3176 	 * values in RDTR is a known source of problems on other
3177 	 * platforms another solution is being sought.
3178 	 */
3179 	if (hw->mac.type == e1000_82573)
3180 		E1000_WRITE_REG(hw, E1000_RDTR, 0x20);
3181 
3182 	for (i = 0, que = adapter->rx_queues; i < adapter->rx_num_queues; i++, que++) {
3183 		struct rx_ring *rxr = &que->rxr;
3184 		/* Setup the Base and Length of the Rx Descriptor Ring */
3185 		u64 bus_addr = rxr->rx_paddr;
3186 #if 0
3187 		u32 rdt = adapter->rx_num_queues -1;  /* default */
3188 #endif
3189 
3190 		E1000_WRITE_REG(hw, E1000_RDLEN(i),
3191 		    scctx->isc_nrxd[0] * sizeof(union e1000_rx_desc_extended));
3192 		E1000_WRITE_REG(hw, E1000_RDBAH(i), (u32)(bus_addr >> 32));
3193 		E1000_WRITE_REG(hw, E1000_RDBAL(i), (u32)bus_addr);
3194 		/* Setup the Head and Tail Descriptor Pointers */
3195 		E1000_WRITE_REG(hw, E1000_RDH(i), 0);
3196 		E1000_WRITE_REG(hw, E1000_RDT(i), 0);
3197 	}
3198 
3199 	/*
3200 	 * Set PTHRESH for improved jumbo performance
3201 	 * According to 10.2.5.11 of Intel 82574 Datasheet,
3202 	 * RXDCTL(1) is written whenever RXDCTL(0) is written.
3203 	 * Only write to RXDCTL(1) if there is a need for different
3204 	 * settings.
3205 	 */
3206 
3207 	if (((adapter->hw.mac.type == e1000_ich9lan) ||
3208 	    (adapter->hw.mac.type == e1000_pch2lan) ||
3209 	    (adapter->hw.mac.type == e1000_ich10lan)) &&
3210 	    (if_getmtu(ifp) > ETHERMTU)) {
3211 		u32 rxdctl = E1000_READ_REG(hw, E1000_RXDCTL(0));
3212 		E1000_WRITE_REG(hw, E1000_RXDCTL(0), rxdctl | 3);
3213 	} else if (adapter->hw.mac.type == e1000_82574) {
3214 		for (int i = 0; i < adapter->rx_num_queues; i++) {
3215 			u32 rxdctl = E1000_READ_REG(hw, E1000_RXDCTL(i));
3216 			rxdctl |= 0x20; /* PTHRESH */
3217 			rxdctl |= 4 << 8; /* HTHRESH */
3218 			rxdctl |= 4 << 16;/* WTHRESH */
3219 			rxdctl |= 1 << 24; /* Switch to granularity */
3220 			E1000_WRITE_REG(hw, E1000_RXDCTL(i), rxdctl);
3221 		}
3222 	} else if (adapter->hw.mac.type >= igb_mac_min) {
3223 		u32 psize, srrctl = 0;
3224 
3225 		if (if_getmtu(ifp) > ETHERMTU) {
3226 			/* Set maximum packet len */
3227 			if (adapter->rx_mbuf_sz <= 4096) {
3228 				srrctl |= 4096 >> E1000_SRRCTL_BSIZEPKT_SHIFT;
3229 				rctl |= E1000_RCTL_SZ_4096 | E1000_RCTL_BSEX;
3230 			} else if (adapter->rx_mbuf_sz > 4096) {
3231 				srrctl |= 8192 >> E1000_SRRCTL_BSIZEPKT_SHIFT;
3232 				rctl |= E1000_RCTL_SZ_8192 | E1000_RCTL_BSEX;
3233 			}
3234 			psize = scctx->isc_max_frame_size;
3235 			/* are we on a vlan? */
3236 			if (ifp->if_vlantrunk != NULL)
3237 				psize += VLAN_TAG_SIZE;
3238 			E1000_WRITE_REG(&adapter->hw, E1000_RLPML, psize);
3239 		} else {
3240 			srrctl |= 2048 >> E1000_SRRCTL_BSIZEPKT_SHIFT;
3241 			rctl |= E1000_RCTL_SZ_2048;
3242 		}
3243 
3244 		/*
3245 		 * If TX flow control is disabled and there's >1 queue defined,
3246 		 * enable DROP.
3247 		 *
3248 		 * This drops frames rather than hanging the RX MAC for all queues.
3249 		 */
3250 		if ((adapter->rx_num_queues > 1) &&
3251 		    (adapter->fc == e1000_fc_none ||
3252 		     adapter->fc == e1000_fc_rx_pause)) {
3253 			srrctl |= E1000_SRRCTL_DROP_EN;
3254 		}
3255 			/* Setup the Base and Length of the Rx Descriptor Rings */
3256 		for (i = 0, que = adapter->rx_queues; i < adapter->rx_num_queues; i++, que++) {
3257 			struct rx_ring *rxr = &que->rxr;
3258 			u64 bus_addr = rxr->rx_paddr;
3259 			u32 rxdctl;
3260 
3261 #ifdef notyet
3262 			/* Configure for header split? -- ignore for now */
3263 			rxr->hdr_split = igb_header_split;
3264 #else
3265 			srrctl |= E1000_SRRCTL_DESCTYPE_ADV_ONEBUF;
3266 #endif
3267 
3268 			E1000_WRITE_REG(hw, E1000_RDLEN(i),
3269 					scctx->isc_nrxd[0] * sizeof(struct e1000_rx_desc));
3270 			E1000_WRITE_REG(hw, E1000_RDBAH(i),
3271 					(uint32_t)(bus_addr >> 32));
3272 			E1000_WRITE_REG(hw, E1000_RDBAL(i),
3273 					(uint32_t)bus_addr);
3274 			E1000_WRITE_REG(hw, E1000_SRRCTL(i), srrctl);
3275 			/* Enable this Queue */
3276 			rxdctl = E1000_READ_REG(hw, E1000_RXDCTL(i));
3277 			rxdctl |= E1000_RXDCTL_QUEUE_ENABLE;
3278 			rxdctl &= 0xFFF00000;
3279 			rxdctl |= IGB_RX_PTHRESH;
3280 			rxdctl |= IGB_RX_HTHRESH << 8;
3281 			rxdctl |= IGB_RX_WTHRESH << 16;
3282 			E1000_WRITE_REG(hw, E1000_RXDCTL(i), rxdctl);
3283 		}
3284 	} else if (adapter->hw.mac.type >= e1000_pch2lan) {
3285 		if (if_getmtu(ifp) > ETHERMTU)
3286 			e1000_lv_jumbo_workaround_ich8lan(hw, TRUE);
3287 		else
3288 			e1000_lv_jumbo_workaround_ich8lan(hw, FALSE);
3289 	}
3290 
3291 	/* Make sure VLAN Filters are off */
3292 	rctl &= ~E1000_RCTL_VFE;
3293 
3294 	if (adapter->hw.mac.type < igb_mac_min) {
3295 		if (adapter->rx_mbuf_sz == MCLBYTES)
3296 			rctl |= E1000_RCTL_SZ_2048;
3297 		else if (adapter->rx_mbuf_sz == MJUMPAGESIZE)
3298 			rctl |= E1000_RCTL_SZ_4096 | E1000_RCTL_BSEX;
3299 		else if (adapter->rx_mbuf_sz > MJUMPAGESIZE)
3300 			rctl |= E1000_RCTL_SZ_8192 | E1000_RCTL_BSEX;
3301 
3302 		/* ensure we clear use DTYPE of 00 here */
3303 		rctl &= ~0x00000C00;
3304 	}
3305 
3306 	/* Write out the settings */
3307 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
3308 
3309 	return;
3310 }
3311 
3312 static void
3313 em_if_vlan_register(if_ctx_t ctx, u16 vtag)
3314 {
3315 	struct adapter *adapter = iflib_get_softc(ctx);
3316 	u32 index, bit;
3317 
3318 	index = (vtag >> 5) & 0x7F;
3319 	bit = vtag & 0x1F;
3320 	adapter->shadow_vfta[index] |= (1 << bit);
3321 	++adapter->num_vlans;
3322 }
3323 
3324 static void
3325 em_if_vlan_unregister(if_ctx_t ctx, u16 vtag)
3326 {
3327 	struct adapter *adapter = iflib_get_softc(ctx);
3328 	u32 index, bit;
3329 
3330 	index = (vtag >> 5) & 0x7F;
3331 	bit = vtag & 0x1F;
3332 	adapter->shadow_vfta[index] &= ~(1 << bit);
3333 	--adapter->num_vlans;
3334 }
3335 
3336 static void
3337 em_setup_vlan_hw_support(struct adapter *adapter)
3338 {
3339 	struct e1000_hw *hw = &adapter->hw;
3340 	u32 reg;
3341 
3342 	/*
3343 	 * We get here thru init_locked, meaning
3344 	 * a soft reset, this has already cleared
3345 	 * the VFTA and other state, so if there
3346 	 * have been no vlan's registered do nothing.
3347 	 */
3348 	if (adapter->num_vlans == 0)
3349 		return;
3350 
3351 	/*
3352 	 * A soft reset zero's out the VFTA, so
3353 	 * we need to repopulate it now.
3354 	 */
3355 	for (int i = 0; i < EM_VFTA_SIZE; i++)
3356 		if (adapter->shadow_vfta[i] != 0)
3357 			E1000_WRITE_REG_ARRAY(hw, E1000_VFTA,
3358 			    i, adapter->shadow_vfta[i]);
3359 
3360 	reg = E1000_READ_REG(hw, E1000_CTRL);
3361 	reg |= E1000_CTRL_VME;
3362 	E1000_WRITE_REG(hw, E1000_CTRL, reg);
3363 
3364 	/* Enable the Filter Table */
3365 	reg = E1000_READ_REG(hw, E1000_RCTL);
3366 	reg &= ~E1000_RCTL_CFIEN;
3367 	reg |= E1000_RCTL_VFE;
3368 	E1000_WRITE_REG(hw, E1000_RCTL, reg);
3369 }
3370 
3371 static void
3372 em_if_enable_intr(if_ctx_t ctx)
3373 {
3374 	struct adapter *adapter = iflib_get_softc(ctx);
3375 	struct e1000_hw *hw = &adapter->hw;
3376 	u32 ims_mask = IMS_ENABLE_MASK;
3377 
3378 	if (hw->mac.type == e1000_82574) {
3379 		E1000_WRITE_REG(hw, EM_EIAC, EM_MSIX_MASK);
3380 		ims_mask |= adapter->ims;
3381 	} else if (adapter->intr_type == IFLIB_INTR_MSIX && hw->mac.type >= igb_mac_min)  {
3382 		u32 mask = (adapter->que_mask | adapter->link_mask);
3383 
3384 		E1000_WRITE_REG(&adapter->hw, E1000_EIAC, mask);
3385 		E1000_WRITE_REG(&adapter->hw, E1000_EIAM, mask);
3386 		E1000_WRITE_REG(&adapter->hw, E1000_EIMS, mask);
3387 		ims_mask = E1000_IMS_LSC;
3388 	}
3389 
3390 	E1000_WRITE_REG(hw, E1000_IMS, ims_mask);
3391 }
3392 
3393 static void
3394 em_if_disable_intr(if_ctx_t ctx)
3395 {
3396 	struct adapter *adapter = iflib_get_softc(ctx);
3397 	struct e1000_hw *hw = &adapter->hw;
3398 
3399 	if (adapter->intr_type == IFLIB_INTR_MSIX) {
3400 		if (hw->mac.type >= igb_mac_min)
3401 			E1000_WRITE_REG(&adapter->hw, E1000_EIMC, ~0);
3402 		E1000_WRITE_REG(&adapter->hw, E1000_EIAC, 0);
3403 	}
3404 	E1000_WRITE_REG(&adapter->hw, E1000_IMC, 0xffffffff);
3405 }
3406 
3407 /*
3408  * Bit of a misnomer, what this really means is
3409  * to enable OS management of the system... aka
3410  * to disable special hardware management features
3411  */
3412 static void
3413 em_init_manageability(struct adapter *adapter)
3414 {
3415 	/* A shared code workaround */
3416 #define E1000_82542_MANC2H E1000_MANC2H
3417 	if (adapter->has_manage) {
3418 		int manc2h = E1000_READ_REG(&adapter->hw, E1000_MANC2H);
3419 		int manc = E1000_READ_REG(&adapter->hw, E1000_MANC);
3420 
3421 		/* disable hardware interception of ARP */
3422 		manc &= ~(E1000_MANC_ARP_EN);
3423 
3424 		/* enable receiving management packets to the host */
3425 		manc |= E1000_MANC_EN_MNG2HOST;
3426 #define E1000_MNG2HOST_PORT_623 (1 << 5)
3427 #define E1000_MNG2HOST_PORT_664 (1 << 6)
3428 		manc2h |= E1000_MNG2HOST_PORT_623;
3429 		manc2h |= E1000_MNG2HOST_PORT_664;
3430 		E1000_WRITE_REG(&adapter->hw, E1000_MANC2H, manc2h);
3431 		E1000_WRITE_REG(&adapter->hw, E1000_MANC, manc);
3432 	}
3433 }
3434 
3435 /*
3436  * Give control back to hardware management
3437  * controller if there is one.
3438  */
3439 static void
3440 em_release_manageability(struct adapter *adapter)
3441 {
3442 	if (adapter->has_manage) {
3443 		int manc = E1000_READ_REG(&adapter->hw, E1000_MANC);
3444 
3445 		/* re-enable hardware interception of ARP */
3446 		manc |= E1000_MANC_ARP_EN;
3447 		manc &= ~E1000_MANC_EN_MNG2HOST;
3448 
3449 		E1000_WRITE_REG(&adapter->hw, E1000_MANC, manc);
3450 	}
3451 }
3452 
3453 /*
3454  * em_get_hw_control sets the {CTRL_EXT|FWSM}:DRV_LOAD bit.
3455  * For ASF and Pass Through versions of f/w this means
3456  * that the driver is loaded. For AMT version type f/w
3457  * this means that the network i/f is open.
3458  */
3459 static void
3460 em_get_hw_control(struct adapter *adapter)
3461 {
3462 	u32 ctrl_ext, swsm;
3463 
3464 	if (adapter->vf_ifp)
3465 		return;
3466 
3467 	if (adapter->hw.mac.type == e1000_82573) {
3468 		swsm = E1000_READ_REG(&adapter->hw, E1000_SWSM);
3469 		E1000_WRITE_REG(&adapter->hw, E1000_SWSM,
3470 		    swsm | E1000_SWSM_DRV_LOAD);
3471 		return;
3472 	}
3473 	/* else */
3474 	ctrl_ext = E1000_READ_REG(&adapter->hw, E1000_CTRL_EXT);
3475 	E1000_WRITE_REG(&adapter->hw, E1000_CTRL_EXT,
3476 	    ctrl_ext | E1000_CTRL_EXT_DRV_LOAD);
3477 }
3478 
3479 /*
3480  * em_release_hw_control resets {CTRL_EXT|FWSM}:DRV_LOAD bit.
3481  * For ASF and Pass Through versions of f/w this means that
3482  * the driver is no longer loaded. For AMT versions of the
3483  * f/w this means that the network i/f is closed.
3484  */
3485 static void
3486 em_release_hw_control(struct adapter *adapter)
3487 {
3488 	u32 ctrl_ext, swsm;
3489 
3490 	if (!adapter->has_manage)
3491 		return;
3492 
3493 	if (adapter->hw.mac.type == e1000_82573) {
3494 		swsm = E1000_READ_REG(&adapter->hw, E1000_SWSM);
3495 		E1000_WRITE_REG(&adapter->hw, E1000_SWSM,
3496 		    swsm & ~E1000_SWSM_DRV_LOAD);
3497 		return;
3498 	}
3499 	/* else */
3500 	ctrl_ext = E1000_READ_REG(&adapter->hw, E1000_CTRL_EXT);
3501 	E1000_WRITE_REG(&adapter->hw, E1000_CTRL_EXT,
3502 	    ctrl_ext & ~E1000_CTRL_EXT_DRV_LOAD);
3503 	return;
3504 }
3505 
3506 static int
3507 em_is_valid_ether_addr(u8 *addr)
3508 {
3509 	char zero_addr[6] = { 0, 0, 0, 0, 0, 0 };
3510 
3511 	if ((addr[0] & 1) || (!bcmp(addr, zero_addr, ETHER_ADDR_LEN))) {
3512 		return (FALSE);
3513 	}
3514 
3515 	return (TRUE);
3516 }
3517 
3518 /*
3519 ** Parse the interface capabilities with regard
3520 ** to both system management and wake-on-lan for
3521 ** later use.
3522 */
3523 static void
3524 em_get_wakeup(if_ctx_t ctx)
3525 {
3526 	struct adapter *adapter = iflib_get_softc(ctx);
3527 	device_t dev = iflib_get_dev(ctx);
3528 	u16 eeprom_data = 0, device_id, apme_mask;
3529 
3530 	adapter->has_manage = e1000_enable_mng_pass_thru(&adapter->hw);
3531 	apme_mask = EM_EEPROM_APME;
3532 
3533 	switch (adapter->hw.mac.type) {
3534 	case e1000_82542:
3535 	case e1000_82543:
3536 		break;
3537 	case e1000_82544:
3538 		e1000_read_nvm(&adapter->hw,
3539 		    NVM_INIT_CONTROL2_REG, 1, &eeprom_data);
3540 		apme_mask = EM_82544_APME;
3541 		break;
3542 	case e1000_82546:
3543 	case e1000_82546_rev_3:
3544 		if (adapter->hw.bus.func == 1) {
3545 			e1000_read_nvm(&adapter->hw,
3546 			    NVM_INIT_CONTROL3_PORT_B, 1, &eeprom_data);
3547 			break;
3548 		} else
3549 			e1000_read_nvm(&adapter->hw,
3550 			    NVM_INIT_CONTROL3_PORT_A, 1, &eeprom_data);
3551 		break;
3552 	case e1000_82573:
3553 	case e1000_82583:
3554 		adapter->has_amt = TRUE;
3555 		/* FALLTHROUGH */
3556 	case e1000_82571:
3557 	case e1000_82572:
3558 	case e1000_80003es2lan:
3559 		if (adapter->hw.bus.func == 1) {
3560 			e1000_read_nvm(&adapter->hw,
3561 			    NVM_INIT_CONTROL3_PORT_B, 1, &eeprom_data);
3562 			break;
3563 		} else
3564 			e1000_read_nvm(&adapter->hw,
3565 			    NVM_INIT_CONTROL3_PORT_A, 1, &eeprom_data);
3566 		break;
3567 	case e1000_ich8lan:
3568 	case e1000_ich9lan:
3569 	case e1000_ich10lan:
3570 	case e1000_pchlan:
3571 	case e1000_pch2lan:
3572 	case e1000_pch_lpt:
3573 	case e1000_pch_spt:
3574 	case e1000_82575:	/* listing all igb devices */
3575 	case e1000_82576:
3576 	case e1000_82580:
3577 	case e1000_i350:
3578 	case e1000_i354:
3579 	case e1000_i210:
3580 	case e1000_i211:
3581 	case e1000_vfadapt:
3582 	case e1000_vfadapt_i350:
3583 		apme_mask = E1000_WUC_APME;
3584 		adapter->has_amt = TRUE;
3585 		eeprom_data = E1000_READ_REG(&adapter->hw, E1000_WUC);
3586 		break;
3587 	default:
3588 		e1000_read_nvm(&adapter->hw,
3589 		    NVM_INIT_CONTROL3_PORT_A, 1, &eeprom_data);
3590 		break;
3591 	}
3592 	if (eeprom_data & apme_mask)
3593 		adapter->wol = (E1000_WUFC_MAG | E1000_WUFC_MC);
3594 	/*
3595 	 * We have the eeprom settings, now apply the special cases
3596 	 * where the eeprom may be wrong or the board won't support
3597 	 * wake on lan on a particular port
3598 	 */
3599 	device_id = pci_get_device(dev);
3600 	switch (device_id) {
3601 	case E1000_DEV_ID_82546GB_PCIE:
3602 		adapter->wol = 0;
3603 		break;
3604 	case E1000_DEV_ID_82546EB_FIBER:
3605 	case E1000_DEV_ID_82546GB_FIBER:
3606 		/* Wake events only supported on port A for dual fiber
3607 		 * regardless of eeprom setting */
3608 		if (E1000_READ_REG(&adapter->hw, E1000_STATUS) &
3609 		    E1000_STATUS_FUNC_1)
3610 			adapter->wol = 0;
3611 		break;
3612 	case E1000_DEV_ID_82546GB_QUAD_COPPER_KSP3:
3613 		/* if quad port adapter, disable WoL on all but port A */
3614 		if (global_quad_port_a != 0)
3615 			adapter->wol = 0;
3616 		/* Reset for multiple quad port adapters */
3617 		if (++global_quad_port_a == 4)
3618 			global_quad_port_a = 0;
3619 		break;
3620 	case E1000_DEV_ID_82571EB_FIBER:
3621 		/* Wake events only supported on port A for dual fiber
3622 		 * regardless of eeprom setting */
3623 		if (E1000_READ_REG(&adapter->hw, E1000_STATUS) &
3624 		    E1000_STATUS_FUNC_1)
3625 			adapter->wol = 0;
3626 		break;
3627 	case E1000_DEV_ID_82571EB_QUAD_COPPER:
3628 	case E1000_DEV_ID_82571EB_QUAD_FIBER:
3629 	case E1000_DEV_ID_82571EB_QUAD_COPPER_LP:
3630 		/* if quad port adapter, disable WoL on all but port A */
3631 		if (global_quad_port_a != 0)
3632 			adapter->wol = 0;
3633 		/* Reset for multiple quad port adapters */
3634 		if (++global_quad_port_a == 4)
3635 			global_quad_port_a = 0;
3636 		break;
3637 	}
3638 	return;
3639 }
3640 
3641 
3642 /*
3643  * Enable PCI Wake On Lan capability
3644  */
3645 static void
3646 em_enable_wakeup(if_ctx_t ctx)
3647 {
3648 	struct adapter *adapter = iflib_get_softc(ctx);
3649 	device_t dev = iflib_get_dev(ctx);
3650 	if_t ifp = iflib_get_ifp(ctx);
3651 	int error = 0;
3652 	u32 pmc, ctrl, ctrl_ext, rctl;
3653 	u16 status;
3654 
3655 	if (pci_find_cap(dev, PCIY_PMG, &pmc) != 0)
3656 		return;
3657 
3658 	/*
3659 	 * Determine type of Wakeup: note that wol
3660 	 * is set with all bits on by default.
3661 	 */
3662 	if ((if_getcapenable(ifp) & IFCAP_WOL_MAGIC) == 0)
3663 		adapter->wol &= ~E1000_WUFC_MAG;
3664 
3665 	if ((if_getcapenable(ifp) & IFCAP_WOL_UCAST) == 0)
3666 		adapter->wol &= ~E1000_WUFC_EX;
3667 
3668 	if ((if_getcapenable(ifp) & IFCAP_WOL_MCAST) == 0)
3669 		adapter->wol &= ~E1000_WUFC_MC;
3670 	else {
3671 		rctl = E1000_READ_REG(&adapter->hw, E1000_RCTL);
3672 		rctl |= E1000_RCTL_MPE;
3673 		E1000_WRITE_REG(&adapter->hw, E1000_RCTL, rctl);
3674 	}
3675 
3676 	if (!(adapter->wol & (E1000_WUFC_EX | E1000_WUFC_MAG | E1000_WUFC_MC)))
3677 		goto pme;
3678 
3679 	/* Advertise the wakeup capability */
3680 	ctrl = E1000_READ_REG(&adapter->hw, E1000_CTRL);
3681 	ctrl |= (E1000_CTRL_SWDPIN2 | E1000_CTRL_SWDPIN3);
3682 	E1000_WRITE_REG(&adapter->hw, E1000_CTRL, ctrl);
3683 
3684 	/* Keep the laser running on Fiber adapters */
3685 	if (adapter->hw.phy.media_type == e1000_media_type_fiber ||
3686 	    adapter->hw.phy.media_type == e1000_media_type_internal_serdes) {
3687 		ctrl_ext = E1000_READ_REG(&adapter->hw, E1000_CTRL_EXT);
3688 		ctrl_ext |= E1000_CTRL_EXT_SDP3_DATA;
3689 		E1000_WRITE_REG(&adapter->hw, E1000_CTRL_EXT, ctrl_ext);
3690 	}
3691 
3692 	if ((adapter->hw.mac.type == e1000_ich8lan) ||
3693 	    (adapter->hw.mac.type == e1000_pchlan) ||
3694 	    (adapter->hw.mac.type == e1000_ich9lan) ||
3695 	    (adapter->hw.mac.type == e1000_ich10lan))
3696 		e1000_suspend_workarounds_ich8lan(&adapter->hw);
3697 
3698 	if ( adapter->hw.mac.type >= e1000_pchlan) {
3699 		error = em_enable_phy_wakeup(adapter);
3700 		if (error)
3701 			goto pme;
3702 	} else {
3703 		/* Enable wakeup by the MAC */
3704 		E1000_WRITE_REG(&adapter->hw, E1000_WUC, E1000_WUC_PME_EN);
3705 		E1000_WRITE_REG(&adapter->hw, E1000_WUFC, adapter->wol);
3706 	}
3707 
3708 	if (adapter->hw.phy.type == e1000_phy_igp_3)
3709 		e1000_igp3_phy_powerdown_workaround_ich8lan(&adapter->hw);
3710 
3711 pme:
3712 	status = pci_read_config(dev, pmc + PCIR_POWER_STATUS, 2);
3713 	status &= ~(PCIM_PSTAT_PME | PCIM_PSTAT_PMEENABLE);
3714 	if (!error && (if_getcapenable(ifp) & IFCAP_WOL))
3715 		status |= PCIM_PSTAT_PME | PCIM_PSTAT_PMEENABLE;
3716 	pci_write_config(dev, pmc + PCIR_POWER_STATUS, status, 2);
3717 
3718 	return;
3719 }
3720 
3721 /*
3722  * WOL in the newer chipset interfaces (pchlan)
3723  * require thing to be copied into the phy
3724  */
3725 static int
3726 em_enable_phy_wakeup(struct adapter *adapter)
3727 {
3728 	struct e1000_hw *hw = &adapter->hw;
3729 	u32 mreg, ret = 0;
3730 	u16 preg;
3731 
3732 	/* copy MAC RARs to PHY RARs */
3733 	e1000_copy_rx_addrs_to_phy_ich8lan(hw);
3734 
3735 	/* copy MAC MTA to PHY MTA */
3736 	for (int i = 0; i < adapter->hw.mac.mta_reg_count; i++) {
3737 		mreg = E1000_READ_REG_ARRAY(hw, E1000_MTA, i);
3738 		e1000_write_phy_reg(hw, BM_MTA(i), (u16)(mreg & 0xFFFF));
3739 		e1000_write_phy_reg(hw, BM_MTA(i) + 1,
3740 		    (u16)((mreg >> 16) & 0xFFFF));
3741 	}
3742 
3743 	/* configure PHY Rx Control register */
3744 	e1000_read_phy_reg(&adapter->hw, BM_RCTL, &preg);
3745 	mreg = E1000_READ_REG(hw, E1000_RCTL);
3746 	if (mreg & E1000_RCTL_UPE)
3747 		preg |= BM_RCTL_UPE;
3748 	if (mreg & E1000_RCTL_MPE)
3749 		preg |= BM_RCTL_MPE;
3750 	preg &= ~(BM_RCTL_MO_MASK);
3751 	if (mreg & E1000_RCTL_MO_3)
3752 		preg |= (((mreg & E1000_RCTL_MO_3) >> E1000_RCTL_MO_SHIFT)
3753 				<< BM_RCTL_MO_SHIFT);
3754 	if (mreg & E1000_RCTL_BAM)
3755 		preg |= BM_RCTL_BAM;
3756 	if (mreg & E1000_RCTL_PMCF)
3757 		preg |= BM_RCTL_PMCF;
3758 	mreg = E1000_READ_REG(hw, E1000_CTRL);
3759 	if (mreg & E1000_CTRL_RFCE)
3760 		preg |= BM_RCTL_RFCE;
3761 	e1000_write_phy_reg(&adapter->hw, BM_RCTL, preg);
3762 
3763 	/* enable PHY wakeup in MAC register */
3764 	E1000_WRITE_REG(hw, E1000_WUC,
3765 	    E1000_WUC_PHY_WAKE | E1000_WUC_PME_EN | E1000_WUC_APME);
3766 	E1000_WRITE_REG(hw, E1000_WUFC, adapter->wol);
3767 
3768 	/* configure and enable PHY wakeup in PHY registers */
3769 	e1000_write_phy_reg(&adapter->hw, BM_WUFC, adapter->wol);
3770 	e1000_write_phy_reg(&adapter->hw, BM_WUC, E1000_WUC_PME_EN);
3771 
3772 	/* activate PHY wakeup */
3773 	ret = hw->phy.ops.acquire(hw);
3774 	if (ret) {
3775 		printf("Could not acquire PHY\n");
3776 		return ret;
3777 	}
3778 	e1000_write_phy_reg_mdic(hw, IGP01E1000_PHY_PAGE_SELECT,
3779 	                         (BM_WUC_ENABLE_PAGE << IGP_PAGE_SHIFT));
3780 	ret = e1000_read_phy_reg_mdic(hw, BM_WUC_ENABLE_REG, &preg);
3781 	if (ret) {
3782 		printf("Could not read PHY page 769\n");
3783 		goto out;
3784 	}
3785 	preg |= BM_WUC_ENABLE_BIT | BM_WUC_HOST_WU_BIT;
3786 	ret = e1000_write_phy_reg_mdic(hw, BM_WUC_ENABLE_REG, preg);
3787 	if (ret)
3788 		printf("Could not set PHY Host Wakeup bit\n");
3789 out:
3790 	hw->phy.ops.release(hw);
3791 
3792 	return ret;
3793 }
3794 
3795 static void
3796 em_if_led_func(if_ctx_t ctx, int onoff)
3797 {
3798 	struct adapter *adapter = iflib_get_softc(ctx);
3799 
3800 	if (onoff) {
3801 		e1000_setup_led(&adapter->hw);
3802 		e1000_led_on(&adapter->hw);
3803 	} else {
3804 		e1000_led_off(&adapter->hw);
3805 		e1000_cleanup_led(&adapter->hw);
3806 	}
3807 }
3808 
3809 /*
3810  * Disable the L0S and L1 LINK states
3811  */
3812 static void
3813 em_disable_aspm(struct adapter *adapter)
3814 {
3815 	int base, reg;
3816 	u16 link_cap,link_ctrl;
3817 	device_t dev = adapter->dev;
3818 
3819 	switch (adapter->hw.mac.type) {
3820 	case e1000_82573:
3821 	case e1000_82574:
3822 	case e1000_82583:
3823 		break;
3824 	default:
3825 		return;
3826 	}
3827 	if (pci_find_cap(dev, PCIY_EXPRESS, &base) != 0)
3828 		return;
3829 	reg = base + PCIER_LINK_CAP;
3830 	link_cap = pci_read_config(dev, reg, 2);
3831 	if ((link_cap & PCIEM_LINK_CAP_ASPM) == 0)
3832 		return;
3833 	reg = base + PCIER_LINK_CTL;
3834 	link_ctrl = pci_read_config(dev, reg, 2);
3835 	link_ctrl &= ~PCIEM_LINK_CTL_ASPMC;
3836 	pci_write_config(dev, reg, link_ctrl, 2);
3837 	return;
3838 }
3839 
3840 /**********************************************************************
3841  *
3842  *  Update the board statistics counters.
3843  *
3844  **********************************************************************/
3845 static void
3846 em_update_stats_counters(struct adapter *adapter)
3847 {
3848 
3849 	if(adapter->hw.phy.media_type == e1000_media_type_copper ||
3850 	   (E1000_READ_REG(&adapter->hw, E1000_STATUS) & E1000_STATUS_LU)) {
3851 		adapter->stats.symerrs += E1000_READ_REG(&adapter->hw, E1000_SYMERRS);
3852 		adapter->stats.sec += E1000_READ_REG(&adapter->hw, E1000_SEC);
3853 	}
3854 	adapter->stats.crcerrs += E1000_READ_REG(&adapter->hw, E1000_CRCERRS);
3855 	adapter->stats.mpc += E1000_READ_REG(&adapter->hw, E1000_MPC);
3856 	adapter->stats.scc += E1000_READ_REG(&adapter->hw, E1000_SCC);
3857 	adapter->stats.ecol += E1000_READ_REG(&adapter->hw, E1000_ECOL);
3858 
3859 	adapter->stats.mcc += E1000_READ_REG(&adapter->hw, E1000_MCC);
3860 	adapter->stats.latecol += E1000_READ_REG(&adapter->hw, E1000_LATECOL);
3861 	adapter->stats.colc += E1000_READ_REG(&adapter->hw, E1000_COLC);
3862 	adapter->stats.dc += E1000_READ_REG(&adapter->hw, E1000_DC);
3863 	adapter->stats.rlec += E1000_READ_REG(&adapter->hw, E1000_RLEC);
3864 	adapter->stats.xonrxc += E1000_READ_REG(&adapter->hw, E1000_XONRXC);
3865 	adapter->stats.xontxc += E1000_READ_REG(&adapter->hw, E1000_XONTXC);
3866 	adapter->stats.xoffrxc += E1000_READ_REG(&adapter->hw, E1000_XOFFRXC);
3867 	/*
3868 	 ** For watchdog management we need to know if we have been
3869 	 ** paused during the last interval, so capture that here.
3870 	*/
3871 	adapter->shared->isc_pause_frames = adapter->stats.xoffrxc;
3872 	adapter->stats.xofftxc += E1000_READ_REG(&adapter->hw, E1000_XOFFTXC);
3873 	adapter->stats.fcruc += E1000_READ_REG(&adapter->hw, E1000_FCRUC);
3874 	adapter->stats.prc64 += E1000_READ_REG(&adapter->hw, E1000_PRC64);
3875 	adapter->stats.prc127 += E1000_READ_REG(&adapter->hw, E1000_PRC127);
3876 	adapter->stats.prc255 += E1000_READ_REG(&adapter->hw, E1000_PRC255);
3877 	adapter->stats.prc511 += E1000_READ_REG(&adapter->hw, E1000_PRC511);
3878 	adapter->stats.prc1023 += E1000_READ_REG(&adapter->hw, E1000_PRC1023);
3879 	adapter->stats.prc1522 += E1000_READ_REG(&adapter->hw, E1000_PRC1522);
3880 	adapter->stats.gprc += E1000_READ_REG(&adapter->hw, E1000_GPRC);
3881 	adapter->stats.bprc += E1000_READ_REG(&adapter->hw, E1000_BPRC);
3882 	adapter->stats.mprc += E1000_READ_REG(&adapter->hw, E1000_MPRC);
3883 	adapter->stats.gptc += E1000_READ_REG(&adapter->hw, E1000_GPTC);
3884 
3885 	/* For the 64-bit byte counters the low dword must be read first. */
3886 	/* Both registers clear on the read of the high dword */
3887 
3888 	adapter->stats.gorc += E1000_READ_REG(&adapter->hw, E1000_GORCL) +
3889 	    ((u64)E1000_READ_REG(&adapter->hw, E1000_GORCH) << 32);
3890 	adapter->stats.gotc += E1000_READ_REG(&adapter->hw, E1000_GOTCL) +
3891 	    ((u64)E1000_READ_REG(&adapter->hw, E1000_GOTCH) << 32);
3892 
3893 	adapter->stats.rnbc += E1000_READ_REG(&adapter->hw, E1000_RNBC);
3894 	adapter->stats.ruc += E1000_READ_REG(&adapter->hw, E1000_RUC);
3895 	adapter->stats.rfc += E1000_READ_REG(&adapter->hw, E1000_RFC);
3896 	adapter->stats.roc += E1000_READ_REG(&adapter->hw, E1000_ROC);
3897 	adapter->stats.rjc += E1000_READ_REG(&adapter->hw, E1000_RJC);
3898 
3899 	adapter->stats.tor += E1000_READ_REG(&adapter->hw, E1000_TORH);
3900 	adapter->stats.tot += E1000_READ_REG(&adapter->hw, E1000_TOTH);
3901 
3902 	adapter->stats.tpr += E1000_READ_REG(&adapter->hw, E1000_TPR);
3903 	adapter->stats.tpt += E1000_READ_REG(&adapter->hw, E1000_TPT);
3904 	adapter->stats.ptc64 += E1000_READ_REG(&adapter->hw, E1000_PTC64);
3905 	adapter->stats.ptc127 += E1000_READ_REG(&adapter->hw, E1000_PTC127);
3906 	adapter->stats.ptc255 += E1000_READ_REG(&adapter->hw, E1000_PTC255);
3907 	adapter->stats.ptc511 += E1000_READ_REG(&adapter->hw, E1000_PTC511);
3908 	adapter->stats.ptc1023 += E1000_READ_REG(&adapter->hw, E1000_PTC1023);
3909 	adapter->stats.ptc1522 += E1000_READ_REG(&adapter->hw, E1000_PTC1522);
3910 	adapter->stats.mptc += E1000_READ_REG(&adapter->hw, E1000_MPTC);
3911 	adapter->stats.bptc += E1000_READ_REG(&adapter->hw, E1000_BPTC);
3912 
3913 	/* Interrupt Counts */
3914 
3915 	adapter->stats.iac += E1000_READ_REG(&adapter->hw, E1000_IAC);
3916 	adapter->stats.icrxptc += E1000_READ_REG(&adapter->hw, E1000_ICRXPTC);
3917 	adapter->stats.icrxatc += E1000_READ_REG(&adapter->hw, E1000_ICRXATC);
3918 	adapter->stats.ictxptc += E1000_READ_REG(&adapter->hw, E1000_ICTXPTC);
3919 	adapter->stats.ictxatc += E1000_READ_REG(&adapter->hw, E1000_ICTXATC);
3920 	adapter->stats.ictxqec += E1000_READ_REG(&adapter->hw, E1000_ICTXQEC);
3921 	adapter->stats.ictxqmtc += E1000_READ_REG(&adapter->hw, E1000_ICTXQMTC);
3922 	adapter->stats.icrxdmtc += E1000_READ_REG(&adapter->hw, E1000_ICRXDMTC);
3923 	adapter->stats.icrxoc += E1000_READ_REG(&adapter->hw, E1000_ICRXOC);
3924 
3925 	if (adapter->hw.mac.type >= e1000_82543) {
3926 		adapter->stats.algnerrc +=
3927 		E1000_READ_REG(&adapter->hw, E1000_ALGNERRC);
3928 		adapter->stats.rxerrc +=
3929 		E1000_READ_REG(&adapter->hw, E1000_RXERRC);
3930 		adapter->stats.tncrs +=
3931 		E1000_READ_REG(&adapter->hw, E1000_TNCRS);
3932 		adapter->stats.cexterr +=
3933 		E1000_READ_REG(&adapter->hw, E1000_CEXTERR);
3934 		adapter->stats.tsctc +=
3935 		E1000_READ_REG(&adapter->hw, E1000_TSCTC);
3936 		adapter->stats.tsctfc +=
3937 		E1000_READ_REG(&adapter->hw, E1000_TSCTFC);
3938 	}
3939 }
3940 
3941 static uint64_t
3942 em_if_get_counter(if_ctx_t ctx, ift_counter cnt)
3943 {
3944 	struct adapter *adapter = iflib_get_softc(ctx);
3945 	struct ifnet *ifp = iflib_get_ifp(ctx);
3946 
3947 	switch (cnt) {
3948 	case IFCOUNTER_COLLISIONS:
3949 		return (adapter->stats.colc);
3950 	case IFCOUNTER_IERRORS:
3951 		return (adapter->dropped_pkts + adapter->stats.rxerrc +
3952 		    adapter->stats.crcerrs + adapter->stats.algnerrc +
3953 		    adapter->stats.ruc + adapter->stats.roc +
3954 		    adapter->stats.mpc + adapter->stats.cexterr);
3955 	case IFCOUNTER_OERRORS:
3956 		return (adapter->stats.ecol + adapter->stats.latecol +
3957 		    adapter->watchdog_events);
3958 	default:
3959 		return (if_get_counter_default(ifp, cnt));
3960 	}
3961 }
3962 
3963 /* Export a single 32-bit register via a read-only sysctl. */
3964 static int
3965 em_sysctl_reg_handler(SYSCTL_HANDLER_ARGS)
3966 {
3967 	struct adapter *adapter;
3968 	u_int val;
3969 
3970 	adapter = oidp->oid_arg1;
3971 	val = E1000_READ_REG(&adapter->hw, oidp->oid_arg2);
3972 	return (sysctl_handle_int(oidp, &val, 0, req));
3973 }
3974 
3975 /*
3976  * Add sysctl variables, one per statistic, to the system.
3977  */
3978 static void
3979 em_add_hw_stats(struct adapter *adapter)
3980 {
3981 	device_t dev = iflib_get_dev(adapter->ctx);
3982 	struct em_tx_queue *tx_que = adapter->tx_queues;
3983 	struct em_rx_queue *rx_que = adapter->rx_queues;
3984 
3985 	struct sysctl_ctx_list *ctx = device_get_sysctl_ctx(dev);
3986 	struct sysctl_oid *tree = device_get_sysctl_tree(dev);
3987 	struct sysctl_oid_list *child = SYSCTL_CHILDREN(tree);
3988 	struct e1000_hw_stats *stats = &adapter->stats;
3989 
3990 	struct sysctl_oid *stat_node, *queue_node, *int_node;
3991 	struct sysctl_oid_list *stat_list, *queue_list, *int_list;
3992 
3993 #define QUEUE_NAME_LEN 32
3994 	char namebuf[QUEUE_NAME_LEN];
3995 
3996 	/* Driver Statistics */
3997 	SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "dropped",
3998 			CTLFLAG_RD, &adapter->dropped_pkts,
3999 			"Driver dropped packets");
4000 	SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "link_irq",
4001 			CTLFLAG_RD, &adapter->link_irq,
4002 			"Link MSIX IRQ Handled");
4003 	SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "mbuf_defrag_fail",
4004 			 CTLFLAG_RD, &adapter->mbuf_defrag_failed,
4005 			 "Defragmenting mbuf chain failed");
4006 	SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "tx_dma_fail",
4007 			CTLFLAG_RD, &adapter->no_tx_dma_setup,
4008 			"Driver tx dma failure in xmit");
4009 	SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "rx_overruns",
4010 			CTLFLAG_RD, &adapter->rx_overruns,
4011 			"RX overruns");
4012 	SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "watchdog_timeouts",
4013 			CTLFLAG_RD, &adapter->watchdog_events,
4014 			"Watchdog timeouts");
4015 	SYSCTL_ADD_PROC(ctx, child, OID_AUTO, "device_control",
4016 			CTLTYPE_UINT | CTLFLAG_RD, adapter, E1000_CTRL,
4017 			em_sysctl_reg_handler, "IU",
4018 			"Device Control Register");
4019 	SYSCTL_ADD_PROC(ctx, child, OID_AUTO, "rx_control",
4020 			CTLTYPE_UINT | CTLFLAG_RD, adapter, E1000_RCTL,
4021 			em_sysctl_reg_handler, "IU",
4022 			"Receiver Control Register");
4023 	SYSCTL_ADD_UINT(ctx, child, OID_AUTO, "fc_high_water",
4024 			CTLFLAG_RD, &adapter->hw.fc.high_water, 0,
4025 			"Flow Control High Watermark");
4026 	SYSCTL_ADD_UINT(ctx, child, OID_AUTO, "fc_low_water",
4027 			CTLFLAG_RD, &adapter->hw.fc.low_water, 0,
4028 			"Flow Control Low Watermark");
4029 
4030 	for (int i = 0; i < adapter->tx_num_queues; i++, tx_que++) {
4031 		struct tx_ring *txr = &tx_que->txr;
4032 		snprintf(namebuf, QUEUE_NAME_LEN, "queue_tx_%d", i);
4033 		queue_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, namebuf,
4034 					    CTLFLAG_RD, NULL, "TX Queue Name");
4035 		queue_list = SYSCTL_CHILDREN(queue_node);
4036 
4037 		SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "txd_head",
4038 				CTLTYPE_UINT | CTLFLAG_RD, adapter,
4039 				E1000_TDH(txr->me),
4040 				em_sysctl_reg_handler, "IU",
4041 				"Transmit Descriptor Head");
4042 		SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "txd_tail",
4043 				CTLTYPE_UINT | CTLFLAG_RD, adapter,
4044 				E1000_TDT(txr->me),
4045 				em_sysctl_reg_handler, "IU",
4046 				"Transmit Descriptor Tail");
4047 		SYSCTL_ADD_ULONG(ctx, queue_list, OID_AUTO, "tx_irq",
4048 				CTLFLAG_RD, &txr->tx_irq,
4049 				"Queue MSI-X Transmit Interrupts");
4050 	}
4051 
4052 	for (int j = 0; j < adapter->rx_num_queues; j++, rx_que++) {
4053 		struct rx_ring *rxr = &rx_que->rxr;
4054 		snprintf(namebuf, QUEUE_NAME_LEN, "queue_rx_%d", j);
4055 		queue_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, namebuf,
4056 					    CTLFLAG_RD, NULL, "RX Queue Name");
4057 		queue_list = SYSCTL_CHILDREN(queue_node);
4058 
4059 		SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "rxd_head",
4060 				CTLTYPE_UINT | CTLFLAG_RD, adapter,
4061 				E1000_RDH(rxr->me),
4062 				em_sysctl_reg_handler, "IU",
4063 				"Receive Descriptor Head");
4064 		SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "rxd_tail",
4065 				CTLTYPE_UINT | CTLFLAG_RD, adapter,
4066 				E1000_RDT(rxr->me),
4067 				em_sysctl_reg_handler, "IU",
4068 				"Receive Descriptor Tail");
4069 		SYSCTL_ADD_ULONG(ctx, queue_list, OID_AUTO, "rx_irq",
4070 				CTLFLAG_RD, &rxr->rx_irq,
4071 				"Queue MSI-X Receive Interrupts");
4072 	}
4073 
4074 	/* MAC stats get their own sub node */
4075 
4076 	stat_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, "mac_stats",
4077 				    CTLFLAG_RD, NULL, "Statistics");
4078 	stat_list = SYSCTL_CHILDREN(stat_node);
4079 
4080 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "excess_coll",
4081 			CTLFLAG_RD, &stats->ecol,
4082 			"Excessive collisions");
4083 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "single_coll",
4084 			CTLFLAG_RD, &stats->scc,
4085 			"Single collisions");
4086 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "multiple_coll",
4087 			CTLFLAG_RD, &stats->mcc,
4088 			"Multiple collisions");
4089 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "late_coll",
4090 			CTLFLAG_RD, &stats->latecol,
4091 			"Late collisions");
4092 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "collision_count",
4093 			CTLFLAG_RD, &stats->colc,
4094 			"Collision Count");
4095 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "symbol_errors",
4096 			CTLFLAG_RD, &adapter->stats.symerrs,
4097 			"Symbol Errors");
4098 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "sequence_errors",
4099 			CTLFLAG_RD, &adapter->stats.sec,
4100 			"Sequence Errors");
4101 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "defer_count",
4102 			CTLFLAG_RD, &adapter->stats.dc,
4103 			"Defer Count");
4104 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "missed_packets",
4105 			CTLFLAG_RD, &adapter->stats.mpc,
4106 			"Missed Packets");
4107 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_no_buff",
4108 			CTLFLAG_RD, &adapter->stats.rnbc,
4109 			"Receive No Buffers");
4110 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_undersize",
4111 			CTLFLAG_RD, &adapter->stats.ruc,
4112 			"Receive Undersize");
4113 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_fragmented",
4114 			CTLFLAG_RD, &adapter->stats.rfc,
4115 			"Fragmented Packets Received ");
4116 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_oversize",
4117 			CTLFLAG_RD, &adapter->stats.roc,
4118 			"Oversized Packets Received");
4119 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_jabber",
4120 			CTLFLAG_RD, &adapter->stats.rjc,
4121 			"Recevied Jabber");
4122 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_errs",
4123 			CTLFLAG_RD, &adapter->stats.rxerrc,
4124 			"Receive Errors");
4125 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "crc_errs",
4126 			CTLFLAG_RD, &adapter->stats.crcerrs,
4127 			"CRC errors");
4128 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "alignment_errs",
4129 			CTLFLAG_RD, &adapter->stats.algnerrc,
4130 			"Alignment Errors");
4131 	/* On 82575 these are collision counts */
4132 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "coll_ext_errs",
4133 			CTLFLAG_RD, &adapter->stats.cexterr,
4134 			"Collision/Carrier extension errors");
4135 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xon_recvd",
4136 			CTLFLAG_RD, &adapter->stats.xonrxc,
4137 			"XON Received");
4138 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xon_txd",
4139 			CTLFLAG_RD, &adapter->stats.xontxc,
4140 			"XON Transmitted");
4141 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xoff_recvd",
4142 			CTLFLAG_RD, &adapter->stats.xoffrxc,
4143 			"XOFF Received");
4144 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xoff_txd",
4145 			CTLFLAG_RD, &adapter->stats.xofftxc,
4146 			"XOFF Transmitted");
4147 
4148 	/* Packet Reception Stats */
4149 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "total_pkts_recvd",
4150 			CTLFLAG_RD, &adapter->stats.tpr,
4151 			"Total Packets Received ");
4152 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_pkts_recvd",
4153 			CTLFLAG_RD, &adapter->stats.gprc,
4154 			"Good Packets Received");
4155 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "bcast_pkts_recvd",
4156 			CTLFLAG_RD, &adapter->stats.bprc,
4157 			"Broadcast Packets Received");
4158 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "mcast_pkts_recvd",
4159 			CTLFLAG_RD, &adapter->stats.mprc,
4160 			"Multicast Packets Received");
4161 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_64",
4162 			CTLFLAG_RD, &adapter->stats.prc64,
4163 			"64 byte frames received ");
4164 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_65_127",
4165 			CTLFLAG_RD, &adapter->stats.prc127,
4166 			"65-127 byte frames received");
4167 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_128_255",
4168 			CTLFLAG_RD, &adapter->stats.prc255,
4169 			"128-255 byte frames received");
4170 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_256_511",
4171 			CTLFLAG_RD, &adapter->stats.prc511,
4172 			"256-511 byte frames received");
4173 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_512_1023",
4174 			CTLFLAG_RD, &adapter->stats.prc1023,
4175 			"512-1023 byte frames received");
4176 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_1024_1522",
4177 			CTLFLAG_RD, &adapter->stats.prc1522,
4178 			"1023-1522 byte frames received");
4179 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_octets_recvd",
4180 			CTLFLAG_RD, &adapter->stats.gorc,
4181 			"Good Octets Received");
4182 
4183 	/* Packet Transmission Stats */
4184 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_octets_txd",
4185 			CTLFLAG_RD, &adapter->stats.gotc,
4186 			"Good Octets Transmitted");
4187 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "total_pkts_txd",
4188 			CTLFLAG_RD, &adapter->stats.tpt,
4189 			"Total Packets Transmitted");
4190 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_pkts_txd",
4191 			CTLFLAG_RD, &adapter->stats.gptc,
4192 			"Good Packets Transmitted");
4193 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "bcast_pkts_txd",
4194 			CTLFLAG_RD, &adapter->stats.bptc,
4195 			"Broadcast Packets Transmitted");
4196 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "mcast_pkts_txd",
4197 			CTLFLAG_RD, &adapter->stats.mptc,
4198 			"Multicast Packets Transmitted");
4199 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_64",
4200 			CTLFLAG_RD, &adapter->stats.ptc64,
4201 			"64 byte frames transmitted ");
4202 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_65_127",
4203 			CTLFLAG_RD, &adapter->stats.ptc127,
4204 			"65-127 byte frames transmitted");
4205 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_128_255",
4206 			CTLFLAG_RD, &adapter->stats.ptc255,
4207 			"128-255 byte frames transmitted");
4208 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_256_511",
4209 			CTLFLAG_RD, &adapter->stats.ptc511,
4210 			"256-511 byte frames transmitted");
4211 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_512_1023",
4212 			CTLFLAG_RD, &adapter->stats.ptc1023,
4213 			"512-1023 byte frames transmitted");
4214 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_1024_1522",
4215 			CTLFLAG_RD, &adapter->stats.ptc1522,
4216 			"1024-1522 byte frames transmitted");
4217 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tso_txd",
4218 			CTLFLAG_RD, &adapter->stats.tsctc,
4219 			"TSO Contexts Transmitted");
4220 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tso_ctx_fail",
4221 			CTLFLAG_RD, &adapter->stats.tsctfc,
4222 			"TSO Contexts Failed");
4223 
4224 
4225 	/* Interrupt Stats */
4226 
4227 	int_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, "interrupts",
4228 				    CTLFLAG_RD, NULL, "Interrupt Statistics");
4229 	int_list = SYSCTL_CHILDREN(int_node);
4230 
4231 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "asserts",
4232 			CTLFLAG_RD, &adapter->stats.iac,
4233 			"Interrupt Assertion Count");
4234 
4235 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "rx_pkt_timer",
4236 			CTLFLAG_RD, &adapter->stats.icrxptc,
4237 			"Interrupt Cause Rx Pkt Timer Expire Count");
4238 
4239 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "rx_abs_timer",
4240 			CTLFLAG_RD, &adapter->stats.icrxatc,
4241 			"Interrupt Cause Rx Abs Timer Expire Count");
4242 
4243 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "tx_pkt_timer",
4244 			CTLFLAG_RD, &adapter->stats.ictxptc,
4245 			"Interrupt Cause Tx Pkt Timer Expire Count");
4246 
4247 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "tx_abs_timer",
4248 			CTLFLAG_RD, &adapter->stats.ictxatc,
4249 			"Interrupt Cause Tx Abs Timer Expire Count");
4250 
4251 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "tx_queue_empty",
4252 			CTLFLAG_RD, &adapter->stats.ictxqec,
4253 			"Interrupt Cause Tx Queue Empty Count");
4254 
4255 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "tx_queue_min_thresh",
4256 			CTLFLAG_RD, &adapter->stats.ictxqmtc,
4257 			"Interrupt Cause Tx Queue Min Thresh Count");
4258 
4259 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "rx_desc_min_thresh",
4260 			CTLFLAG_RD, &adapter->stats.icrxdmtc,
4261 			"Interrupt Cause Rx Desc Min Thresh Count");
4262 
4263 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "rx_overrun",
4264 			CTLFLAG_RD, &adapter->stats.icrxoc,
4265 			"Interrupt Cause Receiver Overrun Count");
4266 }
4267 
4268 /**********************************************************************
4269  *
4270  *  This routine provides a way to dump out the adapter eeprom,
4271  *  often a useful debug/service tool. This only dumps the first
4272  *  32 words, stuff that matters is in that extent.
4273  *
4274  **********************************************************************/
4275 static int
4276 em_sysctl_nvm_info(SYSCTL_HANDLER_ARGS)
4277 {
4278 	struct adapter *adapter = (struct adapter *)arg1;
4279 	int error;
4280 	int result;
4281 
4282 	result = -1;
4283 	error = sysctl_handle_int(oidp, &result, 0, req);
4284 
4285 	if (error || !req->newptr)
4286 		return (error);
4287 
4288 	/*
4289 	 * This value will cause a hex dump of the
4290 	 * first 32 16-bit words of the EEPROM to
4291 	 * the screen.
4292 	 */
4293 	if (result == 1)
4294 		em_print_nvm_info(adapter);
4295 
4296 	return (error);
4297 }
4298 
4299 static void
4300 em_print_nvm_info(struct adapter *adapter)
4301 {
4302 	u16 eeprom_data;
4303 	int i, j, row = 0;
4304 
4305 	/* Its a bit crude, but it gets the job done */
4306 	printf("\nInterface EEPROM Dump:\n");
4307 	printf("Offset\n0x0000  ");
4308 	for (i = 0, j = 0; i < 32; i++, j++) {
4309 		if (j == 8) { /* Make the offset block */
4310 			j = 0; ++row;
4311 			printf("\n0x00%x0  ",row);
4312 		}
4313 		e1000_read_nvm(&adapter->hw, i, 1, &eeprom_data);
4314 		printf("%04x ", eeprom_data);
4315 	}
4316 	printf("\n");
4317 }
4318 
4319 static int
4320 em_sysctl_int_delay(SYSCTL_HANDLER_ARGS)
4321 {
4322 	struct em_int_delay_info *info;
4323 	struct adapter *adapter;
4324 	u32 regval;
4325 	int error, usecs, ticks;
4326 
4327 	info = (struct em_int_delay_info *) arg1;
4328 	usecs = info->value;
4329 	error = sysctl_handle_int(oidp, &usecs, 0, req);
4330 	if (error != 0 || req->newptr == NULL)
4331 		return (error);
4332 	if (usecs < 0 || usecs > EM_TICKS_TO_USECS(65535))
4333 		return (EINVAL);
4334 	info->value = usecs;
4335 	ticks = EM_USECS_TO_TICKS(usecs);
4336 	if (info->offset == E1000_ITR)	/* units are 256ns here */
4337 		ticks *= 4;
4338 
4339 	adapter = info->adapter;
4340 
4341 	regval = E1000_READ_OFFSET(&adapter->hw, info->offset);
4342 	regval = (regval & ~0xffff) | (ticks & 0xffff);
4343 	/* Handle a few special cases. */
4344 	switch (info->offset) {
4345 	case E1000_RDTR:
4346 		break;
4347 	case E1000_TIDV:
4348 		if (ticks == 0) {
4349 			adapter->txd_cmd &= ~E1000_TXD_CMD_IDE;
4350 			/* Don't write 0 into the TIDV register. */
4351 			regval++;
4352 		} else
4353 			adapter->txd_cmd |= E1000_TXD_CMD_IDE;
4354 		break;
4355 	}
4356 	E1000_WRITE_OFFSET(&adapter->hw, info->offset, regval);
4357 	return (0);
4358 }
4359 
4360 static void
4361 em_add_int_delay_sysctl(struct adapter *adapter, const char *name,
4362 	const char *description, struct em_int_delay_info *info,
4363 	int offset, int value)
4364 {
4365 	info->adapter = adapter;
4366 	info->offset = offset;
4367 	info->value = value;
4368 	SYSCTL_ADD_PROC(device_get_sysctl_ctx(adapter->dev),
4369 	    SYSCTL_CHILDREN(device_get_sysctl_tree(adapter->dev)),
4370 	    OID_AUTO, name, CTLTYPE_INT|CTLFLAG_RW,
4371 	    info, 0, em_sysctl_int_delay, "I", description);
4372 }
4373 
4374 /*
4375  * Set flow control using sysctl:
4376  * Flow control values:
4377  *      0 - off
4378  *      1 - rx pause
4379  *      2 - tx pause
4380  *      3 - full
4381  */
4382 static int
4383 em_set_flowcntl(SYSCTL_HANDLER_ARGS)
4384 {
4385 	int error;
4386 	static int input = 3; /* default is full */
4387 	struct adapter	*adapter = (struct adapter *) arg1;
4388 
4389 	error = sysctl_handle_int(oidp, &input, 0, req);
4390 
4391 	if ((error) || (req->newptr == NULL))
4392 		return (error);
4393 
4394 	if (input == adapter->fc) /* no change? */
4395 		return (error);
4396 
4397 	switch (input) {
4398 	case e1000_fc_rx_pause:
4399 	case e1000_fc_tx_pause:
4400 	case e1000_fc_full:
4401 	case e1000_fc_none:
4402 		adapter->hw.fc.requested_mode = input;
4403 		adapter->fc = input;
4404 		break;
4405 	default:
4406 		/* Do nothing */
4407 		return (error);
4408 	}
4409 
4410 	adapter->hw.fc.current_mode = adapter->hw.fc.requested_mode;
4411 	e1000_force_mac_fc(&adapter->hw);
4412 	return (error);
4413 }
4414 
4415 /*
4416  * Manage Energy Efficient Ethernet:
4417  * Control values:
4418  *     0/1 - enabled/disabled
4419  */
4420 static int
4421 em_sysctl_eee(SYSCTL_HANDLER_ARGS)
4422 {
4423 	struct adapter *adapter = (struct adapter *) arg1;
4424 	int error, value;
4425 
4426 	value = adapter->hw.dev_spec.ich8lan.eee_disable;
4427 	error = sysctl_handle_int(oidp, &value, 0, req);
4428 	if (error || req->newptr == NULL)
4429 		return (error);
4430 	adapter->hw.dev_spec.ich8lan.eee_disable = (value != 0);
4431 	em_if_init(adapter->ctx);
4432 
4433 	return (0);
4434 }
4435 
4436 static int
4437 em_sysctl_debug_info(SYSCTL_HANDLER_ARGS)
4438 {
4439 	struct adapter *adapter;
4440 	int error;
4441 	int result;
4442 
4443 	result = -1;
4444 	error = sysctl_handle_int(oidp, &result, 0, req);
4445 
4446 	if (error || !req->newptr)
4447 		return (error);
4448 
4449 	if (result == 1) {
4450 		adapter = (struct adapter *) arg1;
4451 		em_print_debug_info(adapter);
4452 	}
4453 
4454 	return (error);
4455 }
4456 
4457 static int
4458 em_get_rs(SYSCTL_HANDLER_ARGS)
4459 {
4460 	struct adapter *adapter = (struct adapter *) arg1;
4461 	int error;
4462 	int result;
4463 
4464 	result = 0;
4465 	error = sysctl_handle_int(oidp, &result, 0, req);
4466 
4467 	if (error || !req->newptr || result != 1)
4468 		return (error);
4469 	em_dump_rs(adapter);
4470 
4471 	return (error);
4472 }
4473 
4474 static void
4475 em_if_debug(if_ctx_t ctx)
4476 {
4477 	em_dump_rs(iflib_get_softc(ctx));
4478 }
4479 
4480 /*
4481  * This routine is meant to be fluid, add whatever is
4482  * needed for debugging a problem.  -jfv
4483  */
4484 static void
4485 em_print_debug_info(struct adapter *adapter)
4486 {
4487 	device_t dev = iflib_get_dev(adapter->ctx);
4488 	struct ifnet *ifp = iflib_get_ifp(adapter->ctx);
4489 	struct tx_ring *txr = &adapter->tx_queues->txr;
4490 	struct rx_ring *rxr = &adapter->rx_queues->rxr;
4491 
4492 	if (if_getdrvflags(ifp) & IFF_DRV_RUNNING)
4493 		printf("Interface is RUNNING ");
4494 	else
4495 		printf("Interface is NOT RUNNING\n");
4496 
4497 	if (if_getdrvflags(ifp) & IFF_DRV_OACTIVE)
4498 		printf("and INACTIVE\n");
4499 	else
4500 		printf("and ACTIVE\n");
4501 
4502 	for (int i = 0; i < adapter->tx_num_queues; i++, txr++) {
4503 		device_printf(dev, "TX Queue %d ------\n", i);
4504 		device_printf(dev, "hw tdh = %d, hw tdt = %d\n",
4505 			E1000_READ_REG(&adapter->hw, E1000_TDH(i)),
4506 			E1000_READ_REG(&adapter->hw, E1000_TDT(i)));
4507 
4508 	}
4509 	for (int j=0; j < adapter->rx_num_queues; j++, rxr++) {
4510 		device_printf(dev, "RX Queue %d ------\n", j);
4511 		device_printf(dev, "hw rdh = %d, hw rdt = %d\n",
4512 			E1000_READ_REG(&adapter->hw, E1000_RDH(j)),
4513 			E1000_READ_REG(&adapter->hw, E1000_RDT(j)));
4514 	}
4515 }
4516 
4517 /*
4518  * 82574 only:
4519  * Write a new value to the EEPROM increasing the number of MSIX
4520  * vectors from 3 to 5, for proper multiqueue support.
4521  */
4522 static void
4523 em_enable_vectors_82574(if_ctx_t ctx)
4524 {
4525 	struct adapter *adapter = iflib_get_softc(ctx);
4526 	struct e1000_hw *hw = &adapter->hw;
4527 	device_t dev = iflib_get_dev(ctx);
4528 	u16 edata;
4529 
4530 	e1000_read_nvm(hw, EM_NVM_PCIE_CTRL, 1, &edata);
4531 	printf("Current cap: %#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 MSIX 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