xref: /freebsd/sys/dev/e1000/if_em.c (revision 56e53cb8ef000c3ef72337a4095987a932cdedef)
1 /*-
2  * SPDX-License-Identifier: BSD-2-Clause
3  *
4  * Copyright (c) 2016 Matthew Macy <mmacy@mattmacy.io>
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 #define M_TSO_LEN			66
414 
415 #define MAX_INTS_PER_SEC	8000
416 #define DEFAULT_ITR		(1000000000/(MAX_INTS_PER_SEC * 256))
417 
418 /* Allow common code without TSO */
419 #ifndef CSUM_TSO
420 #define CSUM_TSO	0
421 #endif
422 
423 #define TSO_WORKAROUND	4
424 
425 static SYSCTL_NODE(_hw, OID_AUTO, em, CTLFLAG_RD, 0, "EM driver parameters");
426 
427 static int em_disable_crc_stripping = 0;
428 SYSCTL_INT(_hw_em, OID_AUTO, disable_crc_stripping, CTLFLAG_RDTUN,
429     &em_disable_crc_stripping, 0, "Disable CRC Stripping");
430 
431 static int em_tx_int_delay_dflt = EM_TICKS_TO_USECS(EM_TIDV);
432 static int em_rx_int_delay_dflt = EM_TICKS_TO_USECS(EM_RDTR);
433 SYSCTL_INT(_hw_em, OID_AUTO, tx_int_delay, CTLFLAG_RDTUN, &em_tx_int_delay_dflt,
434     0, "Default transmit interrupt delay in usecs");
435 SYSCTL_INT(_hw_em, OID_AUTO, rx_int_delay, CTLFLAG_RDTUN, &em_rx_int_delay_dflt,
436     0, "Default receive interrupt delay in usecs");
437 
438 static int em_tx_abs_int_delay_dflt = EM_TICKS_TO_USECS(EM_TADV);
439 static int em_rx_abs_int_delay_dflt = EM_TICKS_TO_USECS(EM_RADV);
440 SYSCTL_INT(_hw_em, OID_AUTO, tx_abs_int_delay, CTLFLAG_RDTUN,
441     &em_tx_abs_int_delay_dflt, 0,
442     "Default transmit interrupt delay limit in usecs");
443 SYSCTL_INT(_hw_em, OID_AUTO, rx_abs_int_delay, CTLFLAG_RDTUN,
444     &em_rx_abs_int_delay_dflt, 0,
445     "Default receive interrupt delay limit in usecs");
446 
447 static int em_smart_pwr_down = FALSE;
448 SYSCTL_INT(_hw_em, OID_AUTO, smart_pwr_down, CTLFLAG_RDTUN, &em_smart_pwr_down,
449     0, "Set to true to leave smart power down enabled on newer adapters");
450 
451 /* Controls whether promiscuous also shows bad packets */
452 static int em_debug_sbp = TRUE;
453 SYSCTL_INT(_hw_em, OID_AUTO, sbp, CTLFLAG_RDTUN, &em_debug_sbp, 0,
454     "Show bad packets in promiscuous mode");
455 
456 /* How many packets rxeof tries to clean at a time */
457 static int em_rx_process_limit = 100;
458 SYSCTL_INT(_hw_em, OID_AUTO, rx_process_limit, CTLFLAG_RDTUN,
459     &em_rx_process_limit, 0,
460     "Maximum number of received packets to process "
461     "at a time, -1 means unlimited");
462 
463 /* Energy efficient ethernet - default to OFF */
464 static int eee_setting = 1;
465 SYSCTL_INT(_hw_em, OID_AUTO, eee_setting, CTLFLAG_RDTUN, &eee_setting, 0,
466     "Enable Energy Efficient Ethernet");
467 
468 /*
469 ** Tuneable Interrupt rate
470 */
471 static int em_max_interrupt_rate = 8000;
472 SYSCTL_INT(_hw_em, OID_AUTO, max_interrupt_rate, CTLFLAG_RDTUN,
473     &em_max_interrupt_rate, 0, "Maximum interrupts per second");
474 
475 
476 
477 /* Global used in WOL setup with multiport cards */
478 static int global_quad_port_a = 0;
479 
480 extern struct if_txrx igb_txrx;
481 extern struct if_txrx em_txrx;
482 extern struct if_txrx lem_txrx;
483 
484 static struct if_shared_ctx em_sctx_init = {
485 	.isc_magic = IFLIB_MAGIC,
486 	.isc_q_align = PAGE_SIZE,
487 	.isc_tx_maxsize = EM_TSO_SIZE,
488 	.isc_tx_maxsegsize = PAGE_SIZE,
489 	.isc_rx_maxsize = MJUM9BYTES,
490 	.isc_rx_nsegments = 1,
491 	.isc_rx_maxsegsize = MJUM9BYTES,
492 	.isc_nfl = 1,
493 	.isc_nrxqs = 1,
494 	.isc_ntxqs = 1,
495 	.isc_admin_intrcnt = 1,
496 	.isc_vendor_info = em_vendor_info_array,
497 	.isc_driver_version = em_driver_version,
498 	.isc_driver = &em_if_driver,
499 	.isc_flags = IFLIB_NEED_SCRATCH | IFLIB_TSO_INIT_IP | IFLIB_NEED_ZERO_CSUM,
500 
501 	.isc_nrxd_min = {EM_MIN_RXD},
502 	.isc_ntxd_min = {EM_MIN_TXD},
503 	.isc_nrxd_max = {EM_MAX_RXD},
504 	.isc_ntxd_max = {EM_MAX_TXD},
505 	.isc_nrxd_default = {EM_DEFAULT_RXD},
506 	.isc_ntxd_default = {EM_DEFAULT_TXD},
507 };
508 
509 if_shared_ctx_t em_sctx = &em_sctx_init;
510 
511 
512 static struct if_shared_ctx igb_sctx_init = {
513 	.isc_magic = IFLIB_MAGIC,
514 	.isc_q_align = PAGE_SIZE,
515 	.isc_tx_maxsize = EM_TSO_SIZE,
516 	.isc_tx_maxsegsize = PAGE_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 
701 #define EM_CAPS \
702 	IFCAP_TSO4 | IFCAP_TXCSUM | IFCAP_LRO | IFCAP_RXCSUM | IFCAP_VLAN_HWFILTER | IFCAP_WOL_MAGIC | \
703 	IFCAP_WOL_MCAST | IFCAP_WOL | IFCAP_VLAN_HWTSO | IFCAP_HWCSUM | IFCAP_VLAN_HWTAGGING | \
704 	IFCAP_VLAN_HWCSUM | IFCAP_VLAN_HWTSO | IFCAP_VLAN_MTU;
705 
706 #define IGB_CAPS \
707 	IFCAP_TSO4 | IFCAP_TXCSUM | IFCAP_LRO | IFCAP_RXCSUM | IFCAP_VLAN_HWFILTER | IFCAP_WOL_MAGIC | \
708 	IFCAP_WOL_MCAST | IFCAP_WOL | IFCAP_VLAN_HWTSO | IFCAP_HWCSUM | IFCAP_VLAN_HWTAGGING | IFCAP_VLAN_HWCSUM | \
709 	IFCAP_VLAN_HWTSO | IFCAP_VLAN_MTU | IFCAP_TXCSUM_IPV6 | IFCAP_HWCSUM_IPV6 | IFCAP_JUMBO_MTU;
710 
711 /*********************************************************************
712  *  Device initialization routine
713  *
714  *  The attach entry point is called when the driver is being loaded.
715  *  This routine identifies the type of hardware, allocates all resources
716  *  and initializes the hardware.
717  *
718  *  return 0 on success, positive on failure
719  *********************************************************************/
720 
721 static int
722 em_if_attach_pre(if_ctx_t ctx)
723 {
724 	struct adapter *adapter;
725 	if_softc_ctx_t scctx;
726 	device_t dev;
727 	struct e1000_hw *hw;
728 	int error = 0;
729 
730 	INIT_DEBUGOUT("em_if_attach_pre begin");
731 	dev = iflib_get_dev(ctx);
732 	adapter = iflib_get_softc(ctx);
733 
734 	if (resource_disabled("em", device_get_unit(dev))) {
735 		device_printf(dev, "Disabled by device hint\n");
736 		return (ENXIO);
737 	}
738 
739 	adapter->ctx = ctx;
740 	adapter->dev = adapter->osdep.dev = dev;
741 	scctx = adapter->shared = iflib_get_softc_ctx(ctx);
742 	adapter->media = iflib_get_media(ctx);
743 	hw = &adapter->hw;
744 
745 	adapter->tx_process_limit = scctx->isc_ntxd[0];
746 
747 	/* SYSCTL stuff */
748 	SYSCTL_ADD_PROC(device_get_sysctl_ctx(dev),
749 	    SYSCTL_CHILDREN(device_get_sysctl_tree(dev)),
750 	    OID_AUTO, "nvm", CTLTYPE_INT|CTLFLAG_RW, adapter, 0,
751 	    em_sysctl_nvm_info, "I", "NVM Information");
752 
753 	SYSCTL_ADD_PROC(device_get_sysctl_ctx(dev),
754 	    SYSCTL_CHILDREN(device_get_sysctl_tree(dev)),
755 	    OID_AUTO, "debug", CTLTYPE_INT|CTLFLAG_RW, adapter, 0,
756 	    em_sysctl_debug_info, "I", "Debug Information");
757 
758 	SYSCTL_ADD_PROC(device_get_sysctl_ctx(dev),
759 	    SYSCTL_CHILDREN(device_get_sysctl_tree(dev)),
760 	    OID_AUTO, "fc", CTLTYPE_INT|CTLFLAG_RW, adapter, 0,
761 	    em_set_flowcntl, "I", "Flow Control");
762 
763 	SYSCTL_ADD_PROC(device_get_sysctl_ctx(dev),
764 	    SYSCTL_CHILDREN(device_get_sysctl_tree(dev)),
765 	    OID_AUTO, "reg_dump", CTLTYPE_STRING | CTLFLAG_RD, adapter, 0,
766 	    em_get_regs, "A", "Dump Registers");
767 
768 	SYSCTL_ADD_PROC(device_get_sysctl_ctx(dev),
769 	    SYSCTL_CHILDREN(device_get_sysctl_tree(dev)),
770 	    OID_AUTO, "rs_dump", CTLTYPE_INT | CTLFLAG_RW, adapter, 0,
771 	    em_get_rs, "I", "Dump RS indexes");
772 
773 	/* Determine hardware and mac info */
774 	em_identify_hardware(ctx);
775 
776 	/* Set isc_msix_bar */
777 	scctx->isc_msix_bar = PCIR_BAR(EM_MSIX_BAR);
778 	scctx->isc_tx_nsegments = EM_MAX_SCATTER;
779 	scctx->isc_tx_tso_segments_max = scctx->isc_tx_nsegments;
780 	scctx->isc_tx_tso_size_max = EM_TSO_SIZE;
781 	scctx->isc_tx_tso_segsize_max = EM_TSO_SEG_SIZE;
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 	scctx->isc_tx_csum_flags = CSUM_TCP | CSUM_UDP | CSUM_IP_TSO;
786 
787 
788 	if (adapter->hw.mac.type >= igb_mac_min) {
789 		int try_second_bar;
790 
791 		scctx->isc_txqsizes[0] = roundup2(scctx->isc_ntxd[0] * sizeof(union e1000_adv_tx_desc), EM_DBA_ALIGN);
792 		scctx->isc_rxqsizes[0] = roundup2(scctx->isc_nrxd[0] * sizeof(union e1000_adv_rx_desc), EM_DBA_ALIGN);
793 		scctx->isc_txd_size[0] = sizeof(union e1000_adv_tx_desc);
794 		scctx->isc_rxd_size[0] = sizeof(union e1000_adv_rx_desc);
795 		scctx->isc_txrx = &igb_txrx;
796 		scctx->isc_capenable = IGB_CAPS;
797 		scctx->isc_tx_csum_flags = CSUM_TCP | CSUM_UDP | CSUM_TSO | CSUM_IP6_TCP \
798 			| CSUM_IP6_UDP | CSUM_IP6_TCP;
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 
811 	} else if (adapter->hw.mac.type >= em_mac_min) {
812 		scctx->isc_txqsizes[0] = roundup2(scctx->isc_ntxd[0]* sizeof(struct e1000_tx_desc), EM_DBA_ALIGN);
813 		scctx->isc_rxqsizes[0] = roundup2(scctx->isc_nrxd[0] * sizeof(union e1000_rx_desc_extended), EM_DBA_ALIGN);
814 		scctx->isc_txd_size[0] = sizeof(struct e1000_tx_desc);
815 		scctx->isc_rxd_size[0] = sizeof(union e1000_rx_desc_extended);
816 		scctx->isc_txrx = &em_txrx;
817 		scctx->isc_capenable = EM_CAPS;
818 		scctx->isc_tx_csum_flags = CSUM_TCP | CSUM_UDP | CSUM_IP_TSO;
819 	} else {
820 		scctx->isc_txqsizes[0] = roundup2((scctx->isc_ntxd[0] + 1) * sizeof(struct e1000_tx_desc), EM_DBA_ALIGN);
821 		scctx->isc_rxqsizes[0] = roundup2((scctx->isc_nrxd[0] + 1) * sizeof(struct e1000_rx_desc), EM_DBA_ALIGN);
822 		scctx->isc_txd_size[0] = sizeof(struct e1000_tx_desc);
823 		scctx->isc_rxd_size[0] = sizeof(struct e1000_rx_desc);
824 		scctx->isc_tx_csum_flags = CSUM_TCP | CSUM_UDP | CSUM_IP_TSO;
825 		scctx->isc_txrx = &lem_txrx;
826 		scctx->isc_capenable = EM_CAPS;
827 		if (adapter->hw.mac.type < e1000_82543)
828 			scctx->isc_capenable &= ~(IFCAP_HWCSUM|IFCAP_VLAN_HWCSUM);
829 		scctx->isc_tx_csum_flags = CSUM_TCP | CSUM_UDP | CSUM_IP_TSO;
830 		scctx->isc_msix_bar = 0;
831 	}
832 
833 	/* Setup PCI resources */
834 	if (em_allocate_pci_resources(ctx)) {
835 		device_printf(dev, "Allocation of PCI resources failed\n");
836 		error = ENXIO;
837 		goto err_pci;
838 	}
839 
840 	/*
841 	** For ICH8 and family we need to
842 	** map the flash memory, and this
843 	** must happen after the MAC is
844 	** identified
845 	*/
846 	if ((hw->mac.type == e1000_ich8lan) ||
847 	    (hw->mac.type == e1000_ich9lan) ||
848 	    (hw->mac.type == e1000_ich10lan) ||
849 	    (hw->mac.type == e1000_pchlan) ||
850 	    (hw->mac.type == e1000_pch2lan) ||
851 	    (hw->mac.type == e1000_pch_lpt)) {
852 		int rid = EM_BAR_TYPE_FLASH;
853 		adapter->flash = bus_alloc_resource_any(dev,
854 		    SYS_RES_MEMORY, &rid, RF_ACTIVE);
855 		if (adapter->flash == NULL) {
856 			device_printf(dev, "Mapping of Flash failed\n");
857 			error = ENXIO;
858 			goto err_pci;
859 		}
860 		/* This is used in the shared code */
861 		hw->flash_address = (u8 *)adapter->flash;
862 		adapter->osdep.flash_bus_space_tag =
863 		    rman_get_bustag(adapter->flash);
864 		adapter->osdep.flash_bus_space_handle =
865 		    rman_get_bushandle(adapter->flash);
866 	}
867 	/*
868 	** In the new SPT device flash is not  a
869 	** separate BAR, rather it is also in BAR0,
870 	** so use the same tag and an offset handle for the
871 	** FLASH read/write macros in the shared code.
872 	*/
873 	else if (hw->mac.type >= e1000_pch_spt) {
874 		adapter->osdep.flash_bus_space_tag =
875 		    adapter->osdep.mem_bus_space_tag;
876 		adapter->osdep.flash_bus_space_handle =
877 		    adapter->osdep.mem_bus_space_handle
878 		    + E1000_FLASH_BASE_ADDR;
879 	}
880 
881 	/* Do Shared Code initialization */
882 	error = e1000_setup_init_funcs(hw, TRUE);
883 	if (error) {
884 		device_printf(dev, "Setup of Shared code failed, error %d\n",
885 		    error);
886 		error = ENXIO;
887 		goto err_pci;
888 	}
889 
890 	em_setup_msix(ctx);
891 	e1000_get_bus_info(hw);
892 
893 	/* Set up some sysctls for the tunable interrupt delays */
894 	em_add_int_delay_sysctl(adapter, "rx_int_delay",
895 	    "receive interrupt delay in usecs", &adapter->rx_int_delay,
896 	    E1000_REGISTER(hw, E1000_RDTR), em_rx_int_delay_dflt);
897 	em_add_int_delay_sysctl(adapter, "tx_int_delay",
898 	    "transmit interrupt delay in usecs", &adapter->tx_int_delay,
899 	    E1000_REGISTER(hw, E1000_TIDV), em_tx_int_delay_dflt);
900 	em_add_int_delay_sysctl(adapter, "rx_abs_int_delay",
901 	    "receive interrupt delay limit in usecs",
902 	    &adapter->rx_abs_int_delay,
903 	    E1000_REGISTER(hw, E1000_RADV),
904 	    em_rx_abs_int_delay_dflt);
905 	em_add_int_delay_sysctl(adapter, "tx_abs_int_delay",
906 	    "transmit interrupt delay limit in usecs",
907 	    &adapter->tx_abs_int_delay,
908 	    E1000_REGISTER(hw, E1000_TADV),
909 	    em_tx_abs_int_delay_dflt);
910 	em_add_int_delay_sysctl(adapter, "itr",
911 	    "interrupt delay limit in usecs/4",
912 	    &adapter->tx_itr,
913 	    E1000_REGISTER(hw, E1000_ITR),
914 	    DEFAULT_ITR);
915 
916 	hw->mac.autoneg = DO_AUTO_NEG;
917 	hw->phy.autoneg_wait_to_complete = FALSE;
918 	hw->phy.autoneg_advertised = AUTONEG_ADV_DEFAULT;
919 
920 	if (adapter->hw.mac.type < em_mac_min) {
921 		e1000_init_script_state_82541(&adapter->hw, TRUE);
922 		e1000_set_tbi_compatibility_82543(&adapter->hw, TRUE);
923 	}
924 	/* Copper options */
925 	if (hw->phy.media_type == e1000_media_type_copper) {
926 		hw->phy.mdix = AUTO_ALL_MODES;
927 		hw->phy.disable_polarity_correction = FALSE;
928 		hw->phy.ms_type = EM_MASTER_SLAVE;
929 	}
930 
931 	/*
932 	 * Set the frame limits assuming
933 	 * standard ethernet sized frames.
934 	 */
935 	scctx->isc_max_frame_size = adapter->hw.mac.max_frame_size =
936 	    ETHERMTU + ETHER_HDR_LEN + ETHERNET_FCS_SIZE;
937 
938 	/*
939 	 * This controls when hardware reports transmit completion
940 	 * status.
941 	 */
942 	hw->mac.report_tx_early = 1;
943 
944 	/* Allocate multicast array memory. */
945 	adapter->mta = malloc(sizeof(u8) * ETH_ADDR_LEN *
946 	    MAX_NUM_MULTICAST_ADDRESSES, M_DEVBUF, M_NOWAIT);
947 	if (adapter->mta == NULL) {
948 		device_printf(dev, "Can not allocate multicast setup array\n");
949 		error = ENOMEM;
950 		goto err_late;
951 	}
952 
953 	/* Check SOL/IDER usage */
954 	if (e1000_check_reset_block(hw))
955 		device_printf(dev, "PHY reset is blocked"
956 			      " due to SOL/IDER session.\n");
957 
958 	/* Sysctl for setting Energy Efficient Ethernet */
959 	hw->dev_spec.ich8lan.eee_disable = eee_setting;
960 	SYSCTL_ADD_PROC(device_get_sysctl_ctx(dev),
961 	    SYSCTL_CHILDREN(device_get_sysctl_tree(dev)),
962 	    OID_AUTO, "eee_control", CTLTYPE_INT|CTLFLAG_RW,
963 	    adapter, 0, em_sysctl_eee, "I",
964 	    "Disable Energy Efficient Ethernet");
965 
966 	/*
967 	** Start from a known state, this is
968 	** important in reading the nvm and
969 	** mac from that.
970 	*/
971 	e1000_reset_hw(hw);
972 
973 	/* Make sure we have a good EEPROM before we read from it */
974 	if (e1000_validate_nvm_checksum(hw) < 0) {
975 		/*
976 		** Some PCI-E parts fail the first check due to
977 		** the link being in sleep state, call it again,
978 		** if it fails a second time its a real issue.
979 		*/
980 		if (e1000_validate_nvm_checksum(hw) < 0) {
981 			device_printf(dev,
982 			    "The EEPROM Checksum Is Not Valid\n");
983 			error = EIO;
984 			goto err_late;
985 		}
986 	}
987 
988 	/* Copy the permanent MAC address out of the EEPROM */
989 	if (e1000_read_mac_addr(hw) < 0) {
990 		device_printf(dev, "EEPROM read error while reading MAC"
991 			      " address\n");
992 		error = EIO;
993 		goto err_late;
994 	}
995 
996 	if (!em_is_valid_ether_addr(hw->mac.addr)) {
997 		device_printf(dev, "Invalid MAC address\n");
998 		error = EIO;
999 		goto err_late;
1000 	}
1001 
1002 	/* Disable ULP support */
1003 	e1000_disable_ulp_lpt_lp(hw, TRUE);
1004 
1005 	/*
1006 	 * Get Wake-on-Lan and Management info for later use
1007 	 */
1008 	em_get_wakeup(ctx);
1009 
1010 	iflib_set_mac(ctx, hw->mac.addr);
1011 
1012 	return (0);
1013 
1014 err_late:
1015 	em_release_hw_control(adapter);
1016 err_pci:
1017 	em_free_pci_resources(ctx);
1018 	free(adapter->mta, M_DEVBUF);
1019 
1020 	return (error);
1021 }
1022 
1023 static int
1024 em_if_attach_post(if_ctx_t ctx)
1025 {
1026 	struct adapter *adapter = iflib_get_softc(ctx);
1027 	struct e1000_hw *hw = &adapter->hw;
1028 	int error = 0;
1029 
1030 	/* Setup OS specific network interface */
1031 	error = em_setup_interface(ctx);
1032 	if (error != 0) {
1033 		goto err_late;
1034 	}
1035 
1036 	em_reset(ctx);
1037 
1038 	/* Initialize statistics */
1039 	em_update_stats_counters(adapter);
1040 	hw->mac.get_link_status = 1;
1041 	em_if_update_admin_status(ctx);
1042 	em_add_hw_stats(adapter);
1043 
1044 	/* Non-AMT based hardware can now take control from firmware */
1045 	if (adapter->has_manage && !adapter->has_amt)
1046 		em_get_hw_control(adapter);
1047 
1048 	INIT_DEBUGOUT("em_if_attach_post: end");
1049 
1050 	return (error);
1051 
1052 err_late:
1053 	em_release_hw_control(adapter);
1054 	em_free_pci_resources(ctx);
1055 	em_if_queues_free(ctx);
1056 	free(adapter->mta, M_DEVBUF);
1057 
1058 	return (error);
1059 }
1060 
1061 /*********************************************************************
1062  *  Device removal routine
1063  *
1064  *  The detach entry point is called when the driver is being removed.
1065  *  This routine stops the adapter and deallocates all the resources
1066  *  that were allocated for driver operation.
1067  *
1068  *  return 0 on success, positive on failure
1069  *********************************************************************/
1070 
1071 static int
1072 em_if_detach(if_ctx_t ctx)
1073 {
1074 	struct adapter	*adapter = iflib_get_softc(ctx);
1075 
1076 	INIT_DEBUGOUT("em_detach: begin");
1077 
1078 	e1000_phy_hw_reset(&adapter->hw);
1079 
1080 	em_release_manageability(adapter);
1081 	em_release_hw_control(adapter);
1082 	em_free_pci_resources(ctx);
1083 
1084 	return (0);
1085 }
1086 
1087 /*********************************************************************
1088  *
1089  *  Shutdown entry point
1090  *
1091  **********************************************************************/
1092 
1093 static int
1094 em_if_shutdown(if_ctx_t ctx)
1095 {
1096 	return em_if_suspend(ctx);
1097 }
1098 
1099 /*
1100  * Suspend/resume device methods.
1101  */
1102 static int
1103 em_if_suspend(if_ctx_t ctx)
1104 {
1105 	struct adapter *adapter = iflib_get_softc(ctx);
1106 
1107 	em_release_manageability(adapter);
1108 	em_release_hw_control(adapter);
1109 	em_enable_wakeup(ctx);
1110 	return (0);
1111 }
1112 
1113 static int
1114 em_if_resume(if_ctx_t ctx)
1115 {
1116 	struct adapter *adapter = iflib_get_softc(ctx);
1117 
1118 	if (adapter->hw.mac.type == e1000_pch2lan)
1119 		e1000_resume_workarounds_pchlan(&adapter->hw);
1120 	em_if_init(ctx);
1121 	em_init_manageability(adapter);
1122 
1123 	return(0);
1124 }
1125 
1126 static int
1127 em_if_mtu_set(if_ctx_t ctx, uint32_t mtu)
1128 {
1129 	int max_frame_size;
1130 	struct adapter *adapter = iflib_get_softc(ctx);
1131 	if_softc_ctx_t scctx = iflib_get_softc_ctx(ctx);
1132 
1133 	 IOCTL_DEBUGOUT("ioctl rcv'd: SIOCSIFMTU (Set Interface MTU)");
1134 
1135 	switch (adapter->hw.mac.type) {
1136 	case e1000_82571:
1137 	case e1000_82572:
1138 	case e1000_ich9lan:
1139 	case e1000_ich10lan:
1140 	case e1000_pch2lan:
1141 	case e1000_pch_lpt:
1142 	case e1000_pch_spt:
1143 	case e1000_pch_cnp:
1144 	case e1000_82574:
1145 	case e1000_82583:
1146 	case e1000_80003es2lan:
1147 		/* 9K Jumbo Frame size */
1148 		max_frame_size = 9234;
1149 		break;
1150 	case e1000_pchlan:
1151 		max_frame_size = 4096;
1152 		break;
1153 	case e1000_82542:
1154 	case e1000_ich8lan:
1155 		/* Adapters that do not support jumbo frames */
1156 		max_frame_size = ETHER_MAX_LEN;
1157 		break;
1158 	default:
1159 		if (adapter->hw.mac.type >= igb_mac_min)
1160 			max_frame_size = 9234;
1161 		else /* lem */
1162 			max_frame_size = MAX_JUMBO_FRAME_SIZE;
1163 	}
1164 	if (mtu > max_frame_size - ETHER_HDR_LEN - ETHER_CRC_LEN) {
1165 		return (EINVAL);
1166 	}
1167 
1168 	scctx->isc_max_frame_size = adapter->hw.mac.max_frame_size =
1169 	    mtu + ETHER_HDR_LEN + ETHER_CRC_LEN;
1170 	return (0);
1171 }
1172 
1173 /*********************************************************************
1174  *  Init entry point
1175  *
1176  *  This routine is used in two ways. It is used by the stack as
1177  *  init entry point in network interface structure. It is also used
1178  *  by the driver as a hw/sw initialization routine to get to a
1179  *  consistent state.
1180  *
1181  *  return 0 on success, positive on failure
1182  **********************************************************************/
1183 
1184 static void
1185 em_if_init(if_ctx_t ctx)
1186 {
1187 	struct adapter *adapter = iflib_get_softc(ctx);
1188 	struct ifnet *ifp = iflib_get_ifp(ctx);
1189 	struct em_tx_queue *tx_que;
1190 	int i;
1191 	INIT_DEBUGOUT("em_if_init: begin");
1192 
1193 	/* Get the latest mac address, User can use a LAA */
1194 	bcopy(if_getlladdr(ifp), adapter->hw.mac.addr,
1195 	    ETHER_ADDR_LEN);
1196 
1197 	/* Put the address into the Receive Address Array */
1198 	e1000_rar_set(&adapter->hw, adapter->hw.mac.addr, 0);
1199 
1200 	/*
1201 	 * With the 82571 adapter, RAR[0] may be overwritten
1202 	 * when the other port is reset, we make a duplicate
1203 	 * in RAR[14] for that eventuality, this assures
1204 	 * the interface continues to function.
1205 	 */
1206 	if (adapter->hw.mac.type == e1000_82571) {
1207 		e1000_set_laa_state_82571(&adapter->hw, TRUE);
1208 		e1000_rar_set(&adapter->hw, adapter->hw.mac.addr,
1209 		    E1000_RAR_ENTRIES - 1);
1210 	}
1211 
1212 
1213 	/* Initialize the hardware */
1214 	em_reset(ctx);
1215 	em_if_update_admin_status(ctx);
1216 
1217 	for (i = 0, tx_que = adapter->tx_queues; i < adapter->tx_num_queues; i++, tx_que++) {
1218 		struct tx_ring *txr = &tx_que->txr;
1219 
1220 		txr->tx_rs_cidx = txr->tx_rs_pidx = txr->tx_cidx_processed = 0;
1221 	}
1222 
1223 	/* Setup VLAN support, basic and offload if available */
1224 	E1000_WRITE_REG(&adapter->hw, E1000_VET, ETHERTYPE_VLAN);
1225 
1226 	/* Clear bad data from Rx FIFOs */
1227 	if (adapter->hw.mac.type >= igb_mac_min)
1228 		e1000_rx_fifo_flush_82575(&adapter->hw);
1229 
1230 	/* Configure for OS presence */
1231 	em_init_manageability(adapter);
1232 
1233 	/* Prepare transmit descriptors and buffers */
1234 	em_initialize_transmit_unit(ctx);
1235 
1236 	/* Setup Multicast table */
1237 	em_if_multi_set(ctx);
1238 
1239 	/*
1240 	 * Figure out the desired mbuf
1241 	 * pool for doing jumbos
1242 	 */
1243 	if (adapter->hw.mac.max_frame_size <= 2048)
1244 		adapter->rx_mbuf_sz = MCLBYTES;
1245 #ifndef CONTIGMALLOC_WORKS
1246 	else
1247 		adapter->rx_mbuf_sz = MJUMPAGESIZE;
1248 #else
1249 	else if (adapter->hw.mac.max_frame_size <= 4096)
1250 		adapter->rx_mbuf_sz = MJUMPAGESIZE;
1251 	else
1252 		adapter->rx_mbuf_sz = MJUM9BYTES;
1253 #endif
1254 	em_initialize_receive_unit(ctx);
1255 
1256 	/* Use real VLAN Filter support? */
1257 	if (if_getcapenable(ifp) & IFCAP_VLAN_HWTAGGING) {
1258 		if (if_getcapenable(ifp) & IFCAP_VLAN_HWFILTER)
1259 			/* Use real VLAN Filter support */
1260 			em_setup_vlan_hw_support(adapter);
1261 		else {
1262 			u32 ctrl;
1263 			ctrl = E1000_READ_REG(&adapter->hw, E1000_CTRL);
1264 			ctrl |= E1000_CTRL_VME;
1265 			E1000_WRITE_REG(&adapter->hw, E1000_CTRL, ctrl);
1266 		}
1267 	}
1268 
1269 	/* Don't lose promiscuous settings */
1270 	em_if_set_promisc(ctx, IFF_PROMISC);
1271 	e1000_clear_hw_cntrs_base_generic(&adapter->hw);
1272 
1273 	/* MSI/X configuration for 82574 */
1274 	if (adapter->hw.mac.type == e1000_82574) {
1275 		int tmp = E1000_READ_REG(&adapter->hw, E1000_CTRL_EXT);
1276 
1277 		tmp |= E1000_CTRL_EXT_PBA_CLR;
1278 		E1000_WRITE_REG(&adapter->hw, E1000_CTRL_EXT, tmp);
1279 		/* Set the IVAR - interrupt vector routing. */
1280 		E1000_WRITE_REG(&adapter->hw, E1000_IVAR, adapter->ivars);
1281 	} else if (adapter->intr_type == IFLIB_INTR_MSIX) /* Set up queue routing */
1282 		igb_configure_queues(adapter);
1283 
1284 	/* this clears any pending interrupts */
1285 	E1000_READ_REG(&adapter->hw, E1000_ICR);
1286 	E1000_WRITE_REG(&adapter->hw, E1000_ICS, E1000_ICS_LSC);
1287 
1288 	/* AMT based hardware can now take control from firmware */
1289 	if (adapter->has_manage && adapter->has_amt)
1290 		em_get_hw_control(adapter);
1291 
1292 	/* Set Energy Efficient Ethernet */
1293 	if (adapter->hw.mac.type >= igb_mac_min &&
1294 	    adapter->hw.phy.media_type == e1000_media_type_copper) {
1295 		if (adapter->hw.mac.type == e1000_i354)
1296 			e1000_set_eee_i354(&adapter->hw, TRUE, TRUE);
1297 		else
1298 			e1000_set_eee_i350(&adapter->hw, TRUE, TRUE);
1299 	}
1300 }
1301 
1302 /*********************************************************************
1303  *
1304  *  Fast Legacy/MSI Combined Interrupt Service routine
1305  *
1306  *********************************************************************/
1307 int
1308 em_intr(void *arg)
1309 {
1310 	struct adapter *adapter = arg;
1311 	if_ctx_t ctx = adapter->ctx;
1312 	u32 reg_icr;
1313 
1314 	reg_icr = E1000_READ_REG(&adapter->hw, E1000_ICR);
1315 
1316 	if (adapter->intr_type != IFLIB_INTR_LEGACY)
1317 		goto skip_stray;
1318 	/* Hot eject? */
1319 	if (reg_icr == 0xffffffff)
1320 		return FILTER_STRAY;
1321 
1322 	/* Definitely not our interrupt. */
1323 	if (reg_icr == 0x0)
1324 		return FILTER_STRAY;
1325 
1326 	/*
1327 	 * Starting with the 82571 chip, bit 31 should be used to
1328 	 * determine whether the interrupt belongs to us.
1329 	 */
1330 	if (adapter->hw.mac.type >= e1000_82571 &&
1331 	    (reg_icr & E1000_ICR_INT_ASSERTED) == 0)
1332 		return FILTER_STRAY;
1333 
1334 skip_stray:
1335 	/* Link status change */
1336 	if (reg_icr & (E1000_ICR_RXSEQ | E1000_ICR_LSC)) {
1337 		adapter->hw.mac.get_link_status = 1;
1338 		iflib_admin_intr_deferred(ctx);
1339 	}
1340 
1341 	if (reg_icr & E1000_ICR_RXO)
1342 		adapter->rx_overruns++;
1343 
1344 	return (FILTER_SCHEDULE_THREAD);
1345 }
1346 
1347 static void
1348 igb_rx_enable_queue(struct adapter *adapter, struct em_rx_queue *rxq)
1349 {
1350 	E1000_WRITE_REG(&adapter->hw, E1000_EIMS, rxq->eims);
1351 }
1352 
1353 static void
1354 em_rx_enable_queue(struct adapter *adapter, struct em_rx_queue *rxq)
1355 {
1356 	E1000_WRITE_REG(&adapter->hw, E1000_IMS, rxq->eims);
1357 }
1358 
1359 static void
1360 igb_tx_enable_queue(struct adapter *adapter, struct em_tx_queue *txq)
1361 {
1362 	E1000_WRITE_REG(&adapter->hw, E1000_EIMS, txq->eims);
1363 }
1364 
1365 static void
1366 em_tx_enable_queue(struct adapter *adapter, struct em_tx_queue *txq)
1367 {
1368 	E1000_WRITE_REG(&adapter->hw, E1000_IMS, txq->eims);
1369 }
1370 
1371 static int
1372 em_if_rx_queue_intr_enable(if_ctx_t ctx, uint16_t rxqid)
1373 {
1374 	struct adapter *adapter = iflib_get_softc(ctx);
1375 	struct em_rx_queue *rxq = &adapter->rx_queues[rxqid];
1376 
1377 	if (adapter->hw.mac.type >= igb_mac_min)
1378 		igb_rx_enable_queue(adapter, rxq);
1379 	else
1380 		em_rx_enable_queue(adapter, rxq);
1381 	return (0);
1382 }
1383 
1384 static int
1385 em_if_tx_queue_intr_enable(if_ctx_t ctx, uint16_t txqid)
1386 {
1387 	struct adapter *adapter = iflib_get_softc(ctx);
1388 	struct em_tx_queue *txq = &adapter->tx_queues[txqid];
1389 
1390 	if (adapter->hw.mac.type >= igb_mac_min)
1391 		igb_tx_enable_queue(adapter, txq);
1392 	else
1393 		em_tx_enable_queue(adapter, txq);
1394 	return (0);
1395 }
1396 
1397 /*********************************************************************
1398  *
1399  *  MSIX RX Interrupt Service routine
1400  *
1401  **********************************************************************/
1402 static int
1403 em_msix_que(void *arg)
1404 {
1405 	struct em_rx_queue *que = arg;
1406 
1407 	++que->irqs;
1408 
1409 	return (FILTER_SCHEDULE_THREAD);
1410 }
1411 
1412 /*********************************************************************
1413  *
1414  *  MSIX Link Fast Interrupt Service routine
1415  *
1416  **********************************************************************/
1417 static int
1418 em_msix_link(void *arg)
1419 {
1420 	struct adapter *adapter = arg;
1421 	u32 reg_icr;
1422 
1423 	++adapter->link_irq;
1424 	MPASS(adapter->hw.back != NULL);
1425 	reg_icr = E1000_READ_REG(&adapter->hw, E1000_ICR);
1426 
1427 	if (reg_icr & E1000_ICR_RXO)
1428 		adapter->rx_overruns++;
1429 
1430 	if (reg_icr & (E1000_ICR_RXSEQ | E1000_ICR_LSC)) {
1431 		em_handle_link(adapter->ctx);
1432 	} else {
1433 		E1000_WRITE_REG(&adapter->hw, E1000_IMS,
1434 				EM_MSIX_LINK | E1000_IMS_LSC);
1435 		if (adapter->hw.mac.type >= igb_mac_min)
1436 			E1000_WRITE_REG(&adapter->hw, E1000_EIMS, adapter->link_mask);
1437 	}
1438 
1439 	/*
1440 	 * Because we must read the ICR for this interrupt
1441 	 * it may clear other causes using autoclear, for
1442 	 * this reason we simply create a soft interrupt
1443 	 * for all these vectors.
1444 	 */
1445 	if (reg_icr && adapter->hw.mac.type < igb_mac_min) {
1446 		E1000_WRITE_REG(&adapter->hw,
1447 			E1000_ICS, adapter->ims);
1448 	}
1449 
1450 	return (FILTER_HANDLED);
1451 }
1452 
1453 static void
1454 em_handle_link(void *context)
1455 {
1456 	if_ctx_t ctx = context;
1457 	struct adapter *adapter = iflib_get_softc(ctx);
1458 
1459 	adapter->hw.mac.get_link_status = 1;
1460 	iflib_admin_intr_deferred(ctx);
1461 }
1462 
1463 
1464 /*********************************************************************
1465  *
1466  *  Media Ioctl callback
1467  *
1468  *  This routine is called whenever the user queries the status of
1469  *  the interface using ifconfig.
1470  *
1471  **********************************************************************/
1472 static void
1473 em_if_media_status(if_ctx_t ctx, struct ifmediareq *ifmr)
1474 {
1475 	struct adapter *adapter = iflib_get_softc(ctx);
1476 	u_char fiber_type = IFM_1000_SX;
1477 
1478 	INIT_DEBUGOUT("em_if_media_status: begin");
1479 
1480 	iflib_admin_intr_deferred(ctx);
1481 
1482 	ifmr->ifm_status = IFM_AVALID;
1483 	ifmr->ifm_active = IFM_ETHER;
1484 
1485 	if (!adapter->link_active) {
1486 		return;
1487 	}
1488 
1489 	ifmr->ifm_status |= IFM_ACTIVE;
1490 
1491 	if ((adapter->hw.phy.media_type == e1000_media_type_fiber) ||
1492 	    (adapter->hw.phy.media_type == e1000_media_type_internal_serdes)) {
1493 		if (adapter->hw.mac.type == e1000_82545)
1494 			fiber_type = IFM_1000_LX;
1495 		ifmr->ifm_active |= fiber_type | IFM_FDX;
1496 	} else {
1497 		switch (adapter->link_speed) {
1498 		case 10:
1499 			ifmr->ifm_active |= IFM_10_T;
1500 			break;
1501 		case 100:
1502 			ifmr->ifm_active |= IFM_100_TX;
1503 			break;
1504 		case 1000:
1505 			ifmr->ifm_active |= IFM_1000_T;
1506 			break;
1507 		}
1508 		if (adapter->link_duplex == FULL_DUPLEX)
1509 			ifmr->ifm_active |= IFM_FDX;
1510 		else
1511 			ifmr->ifm_active |= IFM_HDX;
1512 	}
1513 }
1514 
1515 /*********************************************************************
1516  *
1517  *  Media Ioctl callback
1518  *
1519  *  This routine is called when the user changes speed/duplex using
1520  *  media/mediopt option with ifconfig.
1521  *
1522  **********************************************************************/
1523 static int
1524 em_if_media_change(if_ctx_t ctx)
1525 {
1526 	struct adapter *adapter = iflib_get_softc(ctx);
1527 	struct ifmedia *ifm = iflib_get_media(ctx);
1528 
1529 	INIT_DEBUGOUT("em_if_media_change: begin");
1530 
1531 	if (IFM_TYPE(ifm->ifm_media) != IFM_ETHER)
1532 		return (EINVAL);
1533 
1534 	switch (IFM_SUBTYPE(ifm->ifm_media)) {
1535 	case IFM_AUTO:
1536 		adapter->hw.mac.autoneg = DO_AUTO_NEG;
1537 		adapter->hw.phy.autoneg_advertised = AUTONEG_ADV_DEFAULT;
1538 		break;
1539 	case IFM_1000_LX:
1540 	case IFM_1000_SX:
1541 	case IFM_1000_T:
1542 		adapter->hw.mac.autoneg = DO_AUTO_NEG;
1543 		adapter->hw.phy.autoneg_advertised = ADVERTISE_1000_FULL;
1544 		break;
1545 	case IFM_100_TX:
1546 		adapter->hw.mac.autoneg = FALSE;
1547 		adapter->hw.phy.autoneg_advertised = 0;
1548 		if ((ifm->ifm_media & IFM_GMASK) == IFM_FDX)
1549 			adapter->hw.mac.forced_speed_duplex = ADVERTISE_100_FULL;
1550 		else
1551 			adapter->hw.mac.forced_speed_duplex = ADVERTISE_100_HALF;
1552 		break;
1553 	case IFM_10_T:
1554 		adapter->hw.mac.autoneg = FALSE;
1555 		adapter->hw.phy.autoneg_advertised = 0;
1556 		if ((ifm->ifm_media & IFM_GMASK) == IFM_FDX)
1557 			adapter->hw.mac.forced_speed_duplex = ADVERTISE_10_FULL;
1558 		else
1559 			adapter->hw.mac.forced_speed_duplex = ADVERTISE_10_HALF;
1560 		break;
1561 	default:
1562 		device_printf(adapter->dev, "Unsupported media type\n");
1563 	}
1564 
1565 	em_if_init(ctx);
1566 
1567 	return (0);
1568 }
1569 
1570 static int
1571 em_if_set_promisc(if_ctx_t ctx, int flags)
1572 {
1573 	struct adapter *adapter = iflib_get_softc(ctx);
1574 	u32 reg_rctl;
1575 
1576 	em_disable_promisc(ctx);
1577 
1578 	reg_rctl = E1000_READ_REG(&adapter->hw, E1000_RCTL);
1579 
1580 	if (flags & IFF_PROMISC) {
1581 		reg_rctl |= (E1000_RCTL_UPE | E1000_RCTL_MPE);
1582 		/* Turn this on if you want to see bad packets */
1583 		if (em_debug_sbp)
1584 			reg_rctl |= E1000_RCTL_SBP;
1585 		E1000_WRITE_REG(&adapter->hw, E1000_RCTL, reg_rctl);
1586 	} else if (flags & IFF_ALLMULTI) {
1587 		reg_rctl |= E1000_RCTL_MPE;
1588 		reg_rctl &= ~E1000_RCTL_UPE;
1589 		E1000_WRITE_REG(&adapter->hw, E1000_RCTL, reg_rctl);
1590 	}
1591 	return (0);
1592 }
1593 
1594 static void
1595 em_disable_promisc(if_ctx_t ctx)
1596 {
1597 	struct adapter *adapter = iflib_get_softc(ctx);
1598 	struct ifnet *ifp = iflib_get_ifp(ctx);
1599 	u32 reg_rctl;
1600 	int mcnt = 0;
1601 
1602 	reg_rctl = E1000_READ_REG(&adapter->hw, E1000_RCTL);
1603 	reg_rctl &= (~E1000_RCTL_UPE);
1604 	if (if_getflags(ifp) & IFF_ALLMULTI)
1605 		mcnt = MAX_NUM_MULTICAST_ADDRESSES;
1606 	else
1607 		mcnt = if_multiaddr_count(ifp, MAX_NUM_MULTICAST_ADDRESSES);
1608 	/* Don't disable if in MAX groups */
1609 	if (mcnt < MAX_NUM_MULTICAST_ADDRESSES)
1610 		reg_rctl &=  (~E1000_RCTL_MPE);
1611 	reg_rctl &=  (~E1000_RCTL_SBP);
1612 	E1000_WRITE_REG(&adapter->hw, E1000_RCTL, reg_rctl);
1613 }
1614 
1615 
1616 /*********************************************************************
1617  *  Multicast Update
1618  *
1619  *  This routine is called whenever multicast address list is updated.
1620  *
1621  **********************************************************************/
1622 
1623 static void
1624 em_if_multi_set(if_ctx_t ctx)
1625 {
1626 	struct adapter *adapter = iflib_get_softc(ctx);
1627 	struct ifnet *ifp = iflib_get_ifp(ctx);
1628 	u32 reg_rctl = 0;
1629 	u8  *mta; /* Multicast array memory */
1630 	int mcnt = 0;
1631 
1632 	IOCTL_DEBUGOUT("em_set_multi: begin");
1633 
1634 	mta = adapter->mta;
1635 	bzero(mta, sizeof(u8) * ETH_ADDR_LEN * MAX_NUM_MULTICAST_ADDRESSES);
1636 
1637 	if (adapter->hw.mac.type == e1000_82542 &&
1638 	    adapter->hw.revision_id == E1000_REVISION_2) {
1639 		reg_rctl = E1000_READ_REG(&adapter->hw, E1000_RCTL);
1640 		if (adapter->hw.bus.pci_cmd_word & CMD_MEM_WRT_INVALIDATE)
1641 			e1000_pci_clear_mwi(&adapter->hw);
1642 		reg_rctl |= E1000_RCTL_RST;
1643 		E1000_WRITE_REG(&adapter->hw, E1000_RCTL, reg_rctl);
1644 		msec_delay(5);
1645 	}
1646 
1647 	if_multiaddr_array(ifp, mta, &mcnt, MAX_NUM_MULTICAST_ADDRESSES);
1648 
1649 	if (mcnt >= MAX_NUM_MULTICAST_ADDRESSES) {
1650 		reg_rctl = E1000_READ_REG(&adapter->hw, E1000_RCTL);
1651 		reg_rctl |= E1000_RCTL_MPE;
1652 		E1000_WRITE_REG(&adapter->hw, E1000_RCTL, reg_rctl);
1653 	} else
1654 		e1000_update_mc_addr_list(&adapter->hw, mta, mcnt);
1655 
1656 	if (adapter->hw.mac.type == e1000_82542 &&
1657 	    adapter->hw.revision_id == E1000_REVISION_2) {
1658 		reg_rctl = E1000_READ_REG(&adapter->hw, E1000_RCTL);
1659 		reg_rctl &= ~E1000_RCTL_RST;
1660 		E1000_WRITE_REG(&adapter->hw, E1000_RCTL, reg_rctl);
1661 		msec_delay(5);
1662 		if (adapter->hw.bus.pci_cmd_word & CMD_MEM_WRT_INVALIDATE)
1663 			e1000_pci_set_mwi(&adapter->hw);
1664 	}
1665 }
1666 
1667 
1668 /*********************************************************************
1669  *  Timer routine
1670  *
1671  *  This routine checks for link status and updates statistics.
1672  *
1673  **********************************************************************/
1674 
1675 static void
1676 em_if_timer(if_ctx_t ctx, uint16_t qid)
1677 {
1678 	struct adapter *adapter = iflib_get_softc(ctx);
1679 	struct em_rx_queue *que;
1680 	int i;
1681 	int trigger = 0;
1682 
1683 	if (qid != 0)
1684 		return;
1685 
1686 	iflib_admin_intr_deferred(ctx);
1687 	/* Reset LAA into RAR[0] on 82571 */
1688 	if ((adapter->hw.mac.type == e1000_82571) &&
1689 	    e1000_get_laa_state_82571(&adapter->hw))
1690 		e1000_rar_set(&adapter->hw, adapter->hw.mac.addr, 0);
1691 
1692 	if (adapter->hw.mac.type < em_mac_min)
1693 		lem_smartspeed(adapter);
1694 
1695 	/* Mask to use in the irq trigger */
1696 	if (adapter->intr_type == IFLIB_INTR_MSIX) {
1697 		for (i = 0, que = adapter->rx_queues; i < adapter->rx_num_queues; i++, que++)
1698 			trigger |= que->eims;
1699 	} else {
1700 		trigger = E1000_ICS_RXDMT0;
1701 	}
1702 }
1703 
1704 
1705 static void
1706 em_if_update_admin_status(if_ctx_t ctx)
1707 {
1708 	struct adapter *adapter = iflib_get_softc(ctx);
1709 	struct e1000_hw *hw = &adapter->hw;
1710 	struct ifnet *ifp = iflib_get_ifp(ctx);
1711 	device_t dev = iflib_get_dev(ctx);
1712 	u32 link_check, thstat, ctrl;
1713 
1714 	link_check = thstat = ctrl = 0;
1715 	/* Get the cached link value or read phy for real */
1716 	switch (hw->phy.media_type) {
1717 	case e1000_media_type_copper:
1718 		if (hw->mac.get_link_status) {
1719 			if (hw->mac.type == e1000_pch_spt)
1720 				msec_delay(50);
1721 			/* Do the work to read phy */
1722 			e1000_check_for_link(hw);
1723 			link_check = !hw->mac.get_link_status;
1724 			if (link_check) /* ESB2 fix */
1725 				e1000_cfg_on_link_up(hw);
1726 		} else {
1727 			link_check = TRUE;
1728 		}
1729 		break;
1730 	case e1000_media_type_fiber:
1731 		e1000_check_for_link(hw);
1732 		link_check = (E1000_READ_REG(hw, E1000_STATUS) &
1733 			    E1000_STATUS_LU);
1734 		break;
1735 	case e1000_media_type_internal_serdes:
1736 		e1000_check_for_link(hw);
1737 		link_check = adapter->hw.mac.serdes_has_link;
1738 		break;
1739 	/* VF device is type_unknown */
1740 	case e1000_media_type_unknown:
1741 		e1000_check_for_link(hw);
1742 		link_check = !hw->mac.get_link_status;
1743 		/* FALLTHROUGH */
1744 	default:
1745 		break;
1746 	}
1747 
1748 	/* Check for thermal downshift or shutdown */
1749 	if (hw->mac.type == e1000_i350) {
1750 		thstat = E1000_READ_REG(hw, E1000_THSTAT);
1751 		ctrl = E1000_READ_REG(hw, E1000_CTRL_EXT);
1752 	}
1753 
1754 	/* Now check for a transition */
1755 	if (link_check && (adapter->link_active == 0)) {
1756 		e1000_get_speed_and_duplex(hw, &adapter->link_speed,
1757 		    &adapter->link_duplex);
1758 		/* Check if we must disable SPEED_MODE bit on PCI-E */
1759 		if ((adapter->link_speed != SPEED_1000) &&
1760 		    ((hw->mac.type == e1000_82571) ||
1761 		    (hw->mac.type == e1000_82572))) {
1762 			int tarc0;
1763 			tarc0 = E1000_READ_REG(hw, E1000_TARC(0));
1764 			tarc0 &= ~TARC_SPEED_MODE_BIT;
1765 			E1000_WRITE_REG(hw, E1000_TARC(0), tarc0);
1766 		}
1767 		if (bootverbose)
1768 			device_printf(dev, "Link is up %d Mbps %s\n",
1769 			    adapter->link_speed,
1770 			    ((adapter->link_duplex == FULL_DUPLEX) ?
1771 			    "Full Duplex" : "Half Duplex"));
1772 		adapter->link_active = 1;
1773 		adapter->smartspeed = 0;
1774 		if_setbaudrate(ifp, adapter->link_speed * 1000000);
1775 		if ((ctrl & E1000_CTRL_EXT_LINK_MODE_GMII) &&
1776 		    (thstat & E1000_THSTAT_LINK_THROTTLE))
1777 			device_printf(dev, "Link: thermal downshift\n");
1778 		/* Delay Link Up for Phy update */
1779 		if (((hw->mac.type == e1000_i210) ||
1780 		    (hw->mac.type == e1000_i211)) &&
1781 		    (hw->phy.id == I210_I_PHY_ID))
1782 			msec_delay(I210_LINK_DELAY);
1783 		/* Reset if the media type changed. */
1784 		if ((hw->dev_spec._82575.media_changed) &&
1785 			(adapter->hw.mac.type >= igb_mac_min)) {
1786 			hw->dev_spec._82575.media_changed = false;
1787 			adapter->flags |= IGB_MEDIA_RESET;
1788 			em_reset(ctx);
1789 		}
1790 		iflib_link_state_change(ctx, LINK_STATE_UP, ifp->if_baudrate);
1791 		printf("Link state changed to up\n");
1792 	} else if (!link_check && (adapter->link_active == 1)) {
1793 		if_setbaudrate(ifp, 0);
1794 		adapter->link_speed = 0;
1795 		adapter->link_duplex = 0;
1796 		if (bootverbose)
1797 			device_printf(dev, "Link is Down\n");
1798 		adapter->link_active = 0;
1799 		iflib_link_state_change(ctx, LINK_STATE_DOWN, ifp->if_baudrate);
1800 		printf("link state changed to down\n");
1801 	}
1802 	em_update_stats_counters(adapter);
1803 
1804 	E1000_WRITE_REG(&adapter->hw, E1000_IMS, EM_MSIX_LINK | E1000_IMS_LSC);
1805 }
1806 
1807 /*********************************************************************
1808  *
1809  *  This routine disables all traffic on the adapter by issuing a
1810  *  global reset on the MAC and deallocates TX/RX buffers.
1811  *
1812  *  This routine should always be called with BOTH the CORE
1813  *  and TX locks.
1814  **********************************************************************/
1815 
1816 static void
1817 em_if_stop(if_ctx_t ctx)
1818 {
1819 	struct adapter *adapter = iflib_get_softc(ctx);
1820 
1821 	INIT_DEBUGOUT("em_stop: begin");
1822 
1823 	e1000_reset_hw(&adapter->hw);
1824 	if (adapter->hw.mac.type >= e1000_82544)
1825 		E1000_WRITE_REG(&adapter->hw, E1000_WUFC, 0);
1826 
1827 	e1000_led_off(&adapter->hw);
1828 	e1000_cleanup_led(&adapter->hw);
1829 }
1830 
1831 
1832 /*********************************************************************
1833  *
1834  *  Determine hardware revision.
1835  *
1836  **********************************************************************/
1837 static void
1838 em_identify_hardware(if_ctx_t ctx)
1839 {
1840 	device_t dev = iflib_get_dev(ctx);
1841 	struct adapter *adapter = iflib_get_softc(ctx);
1842 
1843 	/* Make sure our PCI config space has the necessary stuff set */
1844 	adapter->hw.bus.pci_cmd_word = pci_read_config(dev, PCIR_COMMAND, 2);
1845 
1846 	/* Save off the information about this board */
1847 	adapter->hw.vendor_id = pci_get_vendor(dev);
1848 	adapter->hw.device_id = pci_get_device(dev);
1849 	adapter->hw.revision_id = pci_read_config(dev, PCIR_REVID, 1);
1850 	adapter->hw.subsystem_vendor_id =
1851 	    pci_read_config(dev, PCIR_SUBVEND_0, 2);
1852 	adapter->hw.subsystem_device_id =
1853 	    pci_read_config(dev, PCIR_SUBDEV_0, 2);
1854 
1855 	/* Do Shared Code Init and Setup */
1856 	if (e1000_set_mac_type(&adapter->hw)) {
1857 		device_printf(dev, "Setup init failure\n");
1858 		return;
1859 	}
1860 }
1861 
1862 static int
1863 em_allocate_pci_resources(if_ctx_t ctx)
1864 {
1865 	struct adapter *adapter = iflib_get_softc(ctx);
1866 	device_t dev = iflib_get_dev(ctx);
1867 	int rid, val;
1868 
1869 	rid = PCIR_BAR(0);
1870 	adapter->memory = bus_alloc_resource_any(dev, SYS_RES_MEMORY,
1871 	    &rid, RF_ACTIVE);
1872 	if (adapter->memory == NULL) {
1873 		device_printf(dev, "Unable to allocate bus resource: memory\n");
1874 		return (ENXIO);
1875 	}
1876 	adapter->osdep.mem_bus_space_tag = rman_get_bustag(adapter->memory);
1877 	adapter->osdep.mem_bus_space_handle =
1878 	    rman_get_bushandle(adapter->memory);
1879 	adapter->hw.hw_addr = (u8 *)&adapter->osdep.mem_bus_space_handle;
1880 
1881 	/* Only older adapters use IO mapping */
1882 	if (adapter->hw.mac.type < em_mac_min &&
1883 	    adapter->hw.mac.type > e1000_82543) {
1884 		/* Figure our where our IO BAR is ? */
1885 		for (rid = PCIR_BAR(0); rid < PCIR_CIS;) {
1886 			val = pci_read_config(dev, rid, 4);
1887 			if (EM_BAR_TYPE(val) == EM_BAR_TYPE_IO) {
1888 				adapter->io_rid = rid;
1889 				break;
1890 			}
1891 			rid += 4;
1892 			/* check for 64bit BAR */
1893 			if (EM_BAR_MEM_TYPE(val) == EM_BAR_MEM_TYPE_64BIT)
1894 				rid += 4;
1895 		}
1896 		if (rid >= PCIR_CIS) {
1897 			device_printf(dev, "Unable to locate IO BAR\n");
1898 			return (ENXIO);
1899 		}
1900 		adapter->ioport = bus_alloc_resource_any(dev,
1901 		    SYS_RES_IOPORT, &adapter->io_rid, RF_ACTIVE);
1902 		if (adapter->ioport == NULL) {
1903 			device_printf(dev, "Unable to allocate bus resource: "
1904 			    "ioport\n");
1905 			return (ENXIO);
1906 		}
1907 		adapter->hw.io_base = 0;
1908 		adapter->osdep.io_bus_space_tag =
1909 		    rman_get_bustag(adapter->ioport);
1910 		adapter->osdep.io_bus_space_handle =
1911 		    rman_get_bushandle(adapter->ioport);
1912 	}
1913 
1914 	adapter->hw.back = &adapter->osdep;
1915 
1916 	return (0);
1917 }
1918 
1919 /*********************************************************************
1920  *
1921  *  Setup the MSIX Interrupt handlers
1922  *
1923  **********************************************************************/
1924 static int
1925 em_if_msix_intr_assign(if_ctx_t ctx, int msix)
1926 {
1927 	struct adapter *adapter = iflib_get_softc(ctx);
1928 	struct em_rx_queue *rx_que = adapter->rx_queues;
1929 	struct em_tx_queue *tx_que = adapter->tx_queues;
1930 	int error, rid, i, vector = 0, rx_vectors;
1931 	char buf[16];
1932 
1933 	/* First set up ring resources */
1934 	for (i = 0; i < adapter->rx_num_queues; i++, rx_que++, vector++) {
1935 		rid = vector + 1;
1936 		snprintf(buf, sizeof(buf), "rxq%d", i);
1937 		error = iflib_irq_alloc_generic(ctx, &rx_que->que_irq, rid, IFLIB_INTR_RXTX, em_msix_que, rx_que, rx_que->me, buf);
1938 		if (error) {
1939 			device_printf(iflib_get_dev(ctx), "Failed to allocate que int %d err: %d", i, error);
1940 			adapter->rx_num_queues = i + 1;
1941 			goto fail;
1942 		}
1943 
1944 		rx_que->msix =  vector;
1945 
1946 		/*
1947 		 * Set the bit to enable interrupt
1948 		 * in E1000_IMS -- bits 20 and 21
1949 		 * are for RX0 and RX1, note this has
1950 		 * NOTHING to do with the MSIX vector
1951 		 */
1952 		if (adapter->hw.mac.type == e1000_82574) {
1953 			rx_que->eims = 1 << (20 + i);
1954 			adapter->ims |= rx_que->eims;
1955 			adapter->ivars |= (8 | rx_que->msix) << (i * 4);
1956 		} else if (adapter->hw.mac.type == e1000_82575)
1957 			rx_que->eims = E1000_EICR_TX_QUEUE0 << vector;
1958 		else
1959 			rx_que->eims = 1 << vector;
1960 	}
1961 	rx_vectors = vector;
1962 
1963 	vector = 0;
1964 	for (i = 0; i < adapter->tx_num_queues; i++, tx_que++, vector++) {
1965 		rid = vector + 1;
1966 		snprintf(buf, sizeof(buf), "txq%d", i);
1967 		tx_que = &adapter->tx_queues[i];
1968 		iflib_softirq_alloc_generic(ctx,
1969 		    &adapter->rx_queues[i % adapter->rx_num_queues].que_irq,
1970 		    IFLIB_INTR_TX, tx_que, tx_que->me, buf);
1971 
1972 		tx_que->msix = (vector % adapter->tx_num_queues);
1973 
1974 		/*
1975 		 * Set the bit to enable interrupt
1976 		 * in E1000_IMS -- bits 22 and 23
1977 		 * are for TX0 and TX1, note this has
1978 		 * NOTHING to do with the MSIX vector
1979 		 */
1980 		if (adapter->hw.mac.type == e1000_82574) {
1981 			tx_que->eims = 1 << (22 + i);
1982 			adapter->ims |= tx_que->eims;
1983 			adapter->ivars |= (8 | tx_que->msix) << (8 + (i * 4));
1984 		} else if (adapter->hw.mac.type == e1000_82575) {
1985 			tx_que->eims = E1000_EICR_TX_QUEUE0 << (i %  adapter->tx_num_queues);
1986 		} else {
1987 			tx_que->eims = 1 << (i %  adapter->tx_num_queues);
1988 		}
1989 	}
1990 
1991 	/* Link interrupt */
1992 	rid = rx_vectors + 1;
1993 	error = iflib_irq_alloc_generic(ctx, &adapter->irq, rid, IFLIB_INTR_ADMIN, em_msix_link, adapter, 0, "aq");
1994 
1995 	if (error) {
1996 		device_printf(iflib_get_dev(ctx), "Failed to register admin handler");
1997 		goto fail;
1998 	}
1999 	adapter->linkvec = rx_vectors;
2000 	if (adapter->hw.mac.type < igb_mac_min) {
2001 		adapter->ivars |=  (8 | rx_vectors) << 16;
2002 		adapter->ivars |= 0x80000000;
2003 	}
2004 	return (0);
2005 fail:
2006 	iflib_irq_free(ctx, &adapter->irq);
2007 	rx_que = adapter->rx_queues;
2008 	for (int i = 0; i < adapter->rx_num_queues; i++, rx_que++)
2009 		iflib_irq_free(ctx, &rx_que->que_irq);
2010 	return (error);
2011 }
2012 
2013 static void
2014 igb_configure_queues(struct adapter *adapter)
2015 {
2016 	struct e1000_hw *hw = &adapter->hw;
2017 	struct em_rx_queue *rx_que;
2018 	struct em_tx_queue *tx_que;
2019 	u32 tmp, ivar = 0, newitr = 0;
2020 
2021 	/* First turn on RSS capability */
2022 	if (adapter->hw.mac.type != e1000_82575)
2023 		E1000_WRITE_REG(hw, E1000_GPIE,
2024 		    E1000_GPIE_MSIX_MODE | E1000_GPIE_EIAME |
2025 		    E1000_GPIE_PBA | E1000_GPIE_NSICR);
2026 
2027 	/* Turn on MSIX */
2028 	switch (adapter->hw.mac.type) {
2029 	case e1000_82580:
2030 	case e1000_i350:
2031 	case e1000_i354:
2032 	case e1000_i210:
2033 	case e1000_i211:
2034 	case e1000_vfadapt:
2035 	case e1000_vfadapt_i350:
2036 		/* RX entries */
2037 		for (int i = 0; i < adapter->rx_num_queues; i++) {
2038 			u32 index = i >> 1;
2039 			ivar = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index);
2040 			rx_que = &adapter->rx_queues[i];
2041 			if (i & 1) {
2042 				ivar &= 0xFF00FFFF;
2043 				ivar |= (rx_que->msix | E1000_IVAR_VALID) << 16;
2044 			} else {
2045 				ivar &= 0xFFFFFF00;
2046 				ivar |= rx_que->msix | E1000_IVAR_VALID;
2047 			}
2048 			E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, ivar);
2049 		}
2050 		/* TX entries */
2051 		for (int i = 0; i < adapter->tx_num_queues; i++) {
2052 			u32 index = i >> 1;
2053 			ivar = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index);
2054 			tx_que = &adapter->tx_queues[i];
2055 			if (i & 1) {
2056 				ivar &= 0x00FFFFFF;
2057 				ivar |= (tx_que->msix | E1000_IVAR_VALID) << 24;
2058 			} else {
2059 				ivar &= 0xFFFF00FF;
2060 				ivar |= (tx_que->msix | E1000_IVAR_VALID) << 8;
2061 			}
2062 			E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, ivar);
2063 			adapter->que_mask |= tx_que->eims;
2064 		}
2065 
2066 		/* And for the link interrupt */
2067 		ivar = (adapter->linkvec | E1000_IVAR_VALID) << 8;
2068 		adapter->link_mask = 1 << adapter->linkvec;
2069 		E1000_WRITE_REG(hw, E1000_IVAR_MISC, ivar);
2070 		break;
2071 	case e1000_82576:
2072 		/* RX entries */
2073 		for (int i = 0; i < adapter->rx_num_queues; i++) {
2074 			u32 index = i & 0x7; /* Each IVAR has two entries */
2075 			ivar = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index);
2076 			rx_que = &adapter->rx_queues[i];
2077 			if (i < 8) {
2078 				ivar &= 0xFFFFFF00;
2079 				ivar |= rx_que->msix | E1000_IVAR_VALID;
2080 			} else {
2081 				ivar &= 0xFF00FFFF;
2082 				ivar |= (rx_que->msix | E1000_IVAR_VALID) << 16;
2083 			}
2084 			E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, ivar);
2085 			adapter->que_mask |= rx_que->eims;
2086 		}
2087 		/* TX entries */
2088 		for (int i = 0; i < adapter->tx_num_queues; i++) {
2089 			u32 index = i & 0x7; /* Each IVAR has two entries */
2090 			ivar = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index);
2091 			tx_que = &adapter->tx_queues[i];
2092 			if (i < 8) {
2093 				ivar &= 0xFFFF00FF;
2094 				ivar |= (tx_que->msix | E1000_IVAR_VALID) << 8;
2095 			} else {
2096 				ivar &= 0x00FFFFFF;
2097 				ivar |= (tx_que->msix | E1000_IVAR_VALID) << 24;
2098 			}
2099 			E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, ivar);
2100 			adapter->que_mask |= tx_que->eims;
2101 		}
2102 
2103 		/* And for the link interrupt */
2104 		ivar = (adapter->linkvec | E1000_IVAR_VALID) << 8;
2105 		adapter->link_mask = 1 << adapter->linkvec;
2106 		E1000_WRITE_REG(hw, E1000_IVAR_MISC, ivar);
2107 		break;
2108 
2109 	case e1000_82575:
2110 		/* enable MSI-X support*/
2111 		tmp = E1000_READ_REG(hw, E1000_CTRL_EXT);
2112 		tmp |= E1000_CTRL_EXT_PBA_CLR;
2113 		/* Auto-Mask interrupts upon ICR read. */
2114 		tmp |= E1000_CTRL_EXT_EIAME;
2115 		tmp |= E1000_CTRL_EXT_IRCA;
2116 		E1000_WRITE_REG(hw, E1000_CTRL_EXT, tmp);
2117 
2118 		/* Queues */
2119 		for (int i = 0; i < adapter->rx_num_queues; i++) {
2120 			rx_que = &adapter->rx_queues[i];
2121 			tmp = E1000_EICR_RX_QUEUE0 << i;
2122 			tmp |= E1000_EICR_TX_QUEUE0 << i;
2123 			rx_que->eims = tmp;
2124 			E1000_WRITE_REG_ARRAY(hw, E1000_MSIXBM(0),
2125 			    i, rx_que->eims);
2126 			adapter->que_mask |= rx_que->eims;
2127 		}
2128 
2129 		/* Link */
2130 		E1000_WRITE_REG(hw, E1000_MSIXBM(adapter->linkvec),
2131 		    E1000_EIMS_OTHER);
2132 		adapter->link_mask |= E1000_EIMS_OTHER;
2133 	default:
2134 		break;
2135 	}
2136 
2137 	/* Set the starting interrupt rate */
2138 	if (em_max_interrupt_rate > 0)
2139 		newitr = (4000000 / em_max_interrupt_rate) & 0x7FFC;
2140 
2141 	if (hw->mac.type == e1000_82575)
2142 		newitr |= newitr << 16;
2143 	else
2144 		newitr |= E1000_EITR_CNT_IGNR;
2145 
2146 	for (int i = 0; i < adapter->rx_num_queues; i++) {
2147 		rx_que = &adapter->rx_queues[i];
2148 		E1000_WRITE_REG(hw, E1000_EITR(rx_que->msix), newitr);
2149 	}
2150 
2151 	return;
2152 }
2153 
2154 static void
2155 em_free_pci_resources(if_ctx_t ctx)
2156 {
2157 	struct adapter *adapter = iflib_get_softc(ctx);
2158 	struct em_rx_queue *que = adapter->rx_queues;
2159 	device_t dev = iflib_get_dev(ctx);
2160 
2161 	/* Release all msix queue resources */
2162 	if (adapter->intr_type == IFLIB_INTR_MSIX)
2163 		iflib_irq_free(ctx, &adapter->irq);
2164 
2165 	for (int i = 0; i < adapter->rx_num_queues; i++, que++) {
2166 		iflib_irq_free(ctx, &que->que_irq);
2167 	}
2168 
2169 	/* First release all the interrupt resources */
2170 	if (adapter->memory != NULL) {
2171 		bus_release_resource(dev, SYS_RES_MEMORY,
2172 				     PCIR_BAR(0), adapter->memory);
2173 		adapter->memory = NULL;
2174 	}
2175 
2176 	if (adapter->flash != NULL) {
2177 		bus_release_resource(dev, SYS_RES_MEMORY,
2178 				     EM_FLASH, adapter->flash);
2179 		adapter->flash = NULL;
2180 	}
2181 	if (adapter->ioport != NULL)
2182 		bus_release_resource(dev, SYS_RES_IOPORT,
2183 		    adapter->io_rid, adapter->ioport);
2184 }
2185 
2186 /* Setup MSI or MSI/X */
2187 static int
2188 em_setup_msix(if_ctx_t ctx)
2189 {
2190 	struct adapter *adapter = iflib_get_softc(ctx);
2191 
2192 	if (adapter->hw.mac.type == e1000_82574) {
2193 		em_enable_vectors_82574(ctx);
2194 	}
2195 	return (0);
2196 }
2197 
2198 /*********************************************************************
2199  *
2200  *  Initialize the hardware to a configuration
2201  *  as specified by the adapter structure.
2202  *
2203  **********************************************************************/
2204 
2205 static void
2206 lem_smartspeed(struct adapter *adapter)
2207 {
2208 	u16 phy_tmp;
2209 
2210 	if (adapter->link_active || (adapter->hw.phy.type != e1000_phy_igp) ||
2211 	    adapter->hw.mac.autoneg == 0 ||
2212 	    (adapter->hw.phy.autoneg_advertised & ADVERTISE_1000_FULL) == 0)
2213 		return;
2214 
2215 	if (adapter->smartspeed == 0) {
2216 		/* If Master/Slave config fault is asserted twice,
2217 		 * we assume back-to-back */
2218 		e1000_read_phy_reg(&adapter->hw, PHY_1000T_STATUS, &phy_tmp);
2219 		if (!(phy_tmp & SR_1000T_MS_CONFIG_FAULT))
2220 			return;
2221 		e1000_read_phy_reg(&adapter->hw, PHY_1000T_STATUS, &phy_tmp);
2222 		if (phy_tmp & SR_1000T_MS_CONFIG_FAULT) {
2223 			e1000_read_phy_reg(&adapter->hw,
2224 			    PHY_1000T_CTRL, &phy_tmp);
2225 			if(phy_tmp & CR_1000T_MS_ENABLE) {
2226 				phy_tmp &= ~CR_1000T_MS_ENABLE;
2227 				e1000_write_phy_reg(&adapter->hw,
2228 				    PHY_1000T_CTRL, phy_tmp);
2229 				adapter->smartspeed++;
2230 				if(adapter->hw.mac.autoneg &&
2231 				   !e1000_copper_link_autoneg(&adapter->hw) &&
2232 				   !e1000_read_phy_reg(&adapter->hw,
2233 				    PHY_CONTROL, &phy_tmp)) {
2234 					phy_tmp |= (MII_CR_AUTO_NEG_EN |
2235 						    MII_CR_RESTART_AUTO_NEG);
2236 					e1000_write_phy_reg(&adapter->hw,
2237 					    PHY_CONTROL, phy_tmp);
2238 				}
2239 			}
2240 		}
2241 		return;
2242 	} else if(adapter->smartspeed == EM_SMARTSPEED_DOWNSHIFT) {
2243 		/* If still no link, perhaps using 2/3 pair cable */
2244 		e1000_read_phy_reg(&adapter->hw, PHY_1000T_CTRL, &phy_tmp);
2245 		phy_tmp |= CR_1000T_MS_ENABLE;
2246 		e1000_write_phy_reg(&adapter->hw, PHY_1000T_CTRL, phy_tmp);
2247 		if(adapter->hw.mac.autoneg &&
2248 		   !e1000_copper_link_autoneg(&adapter->hw) &&
2249 		   !e1000_read_phy_reg(&adapter->hw, PHY_CONTROL, &phy_tmp)) {
2250 			phy_tmp |= (MII_CR_AUTO_NEG_EN |
2251 				    MII_CR_RESTART_AUTO_NEG);
2252 			e1000_write_phy_reg(&adapter->hw, PHY_CONTROL, phy_tmp);
2253 		}
2254 	}
2255 	/* Restart process after EM_SMARTSPEED_MAX iterations */
2256 	if(adapter->smartspeed++ == EM_SMARTSPEED_MAX)
2257 		adapter->smartspeed = 0;
2258 }
2259 
2260 /*********************************************************************
2261  *
2262  *  Initialize the DMA Coalescing feature
2263  *
2264  **********************************************************************/
2265 static void
2266 igb_init_dmac(struct adapter *adapter, u32 pba)
2267 {
2268 	device_t	dev = adapter->dev;
2269 	struct e1000_hw *hw = &adapter->hw;
2270 	u32 		dmac, reg = ~E1000_DMACR_DMAC_EN;
2271 	u16		hwm;
2272 	u16		max_frame_size;
2273 
2274 	if (hw->mac.type == e1000_i211)
2275 		return;
2276 
2277 	max_frame_size = adapter->shared->isc_max_frame_size;
2278 	if (hw->mac.type > e1000_82580) {
2279 
2280 		if (adapter->dmac == 0) { /* Disabling it */
2281 			E1000_WRITE_REG(hw, E1000_DMACR, reg);
2282 			return;
2283 		} else
2284 			device_printf(dev, "DMA Coalescing enabled\n");
2285 
2286 		/* Set starting threshold */
2287 		E1000_WRITE_REG(hw, E1000_DMCTXTH, 0);
2288 
2289 		hwm = 64 * pba - max_frame_size / 16;
2290 		if (hwm < 64 * (pba - 6))
2291 			hwm = 64 * (pba - 6);
2292 		reg = E1000_READ_REG(hw, E1000_FCRTC);
2293 		reg &= ~E1000_FCRTC_RTH_COAL_MASK;
2294 		reg |= ((hwm << E1000_FCRTC_RTH_COAL_SHIFT)
2295 		    & E1000_FCRTC_RTH_COAL_MASK);
2296 		E1000_WRITE_REG(hw, E1000_FCRTC, reg);
2297 
2298 
2299 		dmac = pba - max_frame_size / 512;
2300 		if (dmac < pba - 10)
2301 			dmac = pba - 10;
2302 		reg = E1000_READ_REG(hw, E1000_DMACR);
2303 		reg &= ~E1000_DMACR_DMACTHR_MASK;
2304 		reg = ((dmac << E1000_DMACR_DMACTHR_SHIFT)
2305 		    & E1000_DMACR_DMACTHR_MASK);
2306 
2307 		/* transition to L0x or L1 if available..*/
2308 		reg |= (E1000_DMACR_DMAC_EN | E1000_DMACR_DMAC_LX_MASK);
2309 
2310 		/* Check if status is 2.5Gb backplane connection
2311 		* before configuration of watchdog timer, which is
2312 		* in msec values in 12.8usec intervals
2313 		* watchdog timer= msec values in 32usec intervals
2314 		* for non 2.5Gb connection
2315 		*/
2316 		if (hw->mac.type == e1000_i354) {
2317 			int status = E1000_READ_REG(hw, E1000_STATUS);
2318 			if ((status & E1000_STATUS_2P5_SKU) &&
2319 			    (!(status & E1000_STATUS_2P5_SKU_OVER)))
2320 				reg |= ((adapter->dmac * 5) >> 6);
2321 			else
2322 				reg |= (adapter->dmac >> 5);
2323 		} else {
2324 			reg |= (adapter->dmac >> 5);
2325 		}
2326 
2327 		E1000_WRITE_REG(hw, E1000_DMACR, reg);
2328 
2329 		E1000_WRITE_REG(hw, E1000_DMCRTRH, 0);
2330 
2331 		/* Set the interval before transition */
2332 		reg = E1000_READ_REG(hw, E1000_DMCTLX);
2333 		if (hw->mac.type == e1000_i350)
2334 			reg |= IGB_DMCTLX_DCFLUSH_DIS;
2335 		/*
2336 		** in 2.5Gb connection, TTLX unit is 0.4 usec
2337 		** which is 0x4*2 = 0xA. But delay is still 4 usec
2338 		*/
2339 		if (hw->mac.type == e1000_i354) {
2340 			int status = E1000_READ_REG(hw, E1000_STATUS);
2341 			if ((status & E1000_STATUS_2P5_SKU) &&
2342 			    (!(status & E1000_STATUS_2P5_SKU_OVER)))
2343 				reg |= 0xA;
2344 			else
2345 				reg |= 0x4;
2346 		} else {
2347 			reg |= 0x4;
2348 		}
2349 
2350 		E1000_WRITE_REG(hw, E1000_DMCTLX, reg);
2351 
2352 		/* free space in tx packet buffer to wake from DMA coal */
2353 		E1000_WRITE_REG(hw, E1000_DMCTXTH, (IGB_TXPBSIZE -
2354 		    (2 * max_frame_size)) >> 6);
2355 
2356 		/* make low power state decision controlled by DMA coal */
2357 		reg = E1000_READ_REG(hw, E1000_PCIEMISC);
2358 		reg &= ~E1000_PCIEMISC_LX_DECISION;
2359 		E1000_WRITE_REG(hw, E1000_PCIEMISC, reg);
2360 
2361 	} else if (hw->mac.type == e1000_82580) {
2362 		u32 reg = E1000_READ_REG(hw, E1000_PCIEMISC);
2363 		E1000_WRITE_REG(hw, E1000_PCIEMISC,
2364 		    reg & ~E1000_PCIEMISC_LX_DECISION);
2365 		E1000_WRITE_REG(hw, E1000_DMACR, 0);
2366 	}
2367 }
2368 
2369 static void
2370 em_reset(if_ctx_t ctx)
2371 {
2372 	device_t dev = iflib_get_dev(ctx);
2373 	struct adapter *adapter = iflib_get_softc(ctx);
2374 	struct ifnet *ifp = iflib_get_ifp(ctx);
2375 	struct e1000_hw *hw = &adapter->hw;
2376 	u16 rx_buffer_size;
2377 	u32 pba;
2378 
2379 	INIT_DEBUGOUT("em_reset: begin");
2380 	/* Let the firmware know the OS is in control */
2381 	em_get_hw_control(adapter);
2382 
2383 	/* Set up smart power down as default off on newer adapters. */
2384 	if (!em_smart_pwr_down && (hw->mac.type == e1000_82571 ||
2385 	    hw->mac.type == e1000_82572)) {
2386 		u16 phy_tmp = 0;
2387 
2388 		/* Speed up time to link by disabling smart power down. */
2389 		e1000_read_phy_reg(hw, IGP02E1000_PHY_POWER_MGMT, &phy_tmp);
2390 		phy_tmp &= ~IGP02E1000_PM_SPD;
2391 		e1000_write_phy_reg(hw, IGP02E1000_PHY_POWER_MGMT, phy_tmp);
2392 	}
2393 
2394 	/*
2395 	 * Packet Buffer Allocation (PBA)
2396 	 * Writing PBA sets the receive portion of the buffer
2397 	 * the remainder is used for the transmit buffer.
2398 	 */
2399 	switch (hw->mac.type) {
2400 	/* Total Packet Buffer on these is 48K */
2401 	case e1000_82571:
2402 	case e1000_82572:
2403 	case e1000_80003es2lan:
2404 			pba = E1000_PBA_32K; /* 32K for Rx, 16K for Tx */
2405 		break;
2406 	case e1000_82573: /* 82573: Total Packet Buffer is 32K */
2407 			pba = E1000_PBA_12K; /* 12K for Rx, 20K for Tx */
2408 		break;
2409 	case e1000_82574:
2410 	case e1000_82583:
2411 			pba = E1000_PBA_20K; /* 20K for Rx, 20K for Tx */
2412 		break;
2413 	case e1000_ich8lan:
2414 		pba = E1000_PBA_8K;
2415 		break;
2416 	case e1000_ich9lan:
2417 	case e1000_ich10lan:
2418 		/* Boost Receive side for jumbo frames */
2419 		if (adapter->hw.mac.max_frame_size > 4096)
2420 			pba = E1000_PBA_14K;
2421 		else
2422 			pba = E1000_PBA_10K;
2423 		break;
2424 	case e1000_pchlan:
2425 	case e1000_pch2lan:
2426 	case e1000_pch_lpt:
2427 	case e1000_pch_spt:
2428 	case e1000_pch_cnp:
2429 		pba = E1000_PBA_26K;
2430 		break;
2431 	case e1000_82575:
2432 		pba = E1000_PBA_32K;
2433 		break;
2434 	case e1000_82576:
2435 	case e1000_vfadapt:
2436 		pba = E1000_READ_REG(hw, E1000_RXPBS);
2437 		pba &= E1000_RXPBS_SIZE_MASK_82576;
2438 		break;
2439 	case e1000_82580:
2440 	case e1000_i350:
2441 	case e1000_i354:
2442 	case e1000_vfadapt_i350:
2443 		pba = E1000_READ_REG(hw, E1000_RXPBS);
2444 		pba = e1000_rxpbs_adjust_82580(pba);
2445 		break;
2446 	case e1000_i210:
2447 	case e1000_i211:
2448 		pba = E1000_PBA_34K;
2449 		break;
2450 	default:
2451 		if (adapter->hw.mac.max_frame_size > 8192)
2452 			pba = E1000_PBA_40K; /* 40K for Rx, 24K for Tx */
2453 		else
2454 			pba = E1000_PBA_48K; /* 48K for Rx, 16K for Tx */
2455 	}
2456 
2457 	/* Special needs in case of Jumbo frames */
2458 	if ((hw->mac.type == e1000_82575) && (ifp->if_mtu > ETHERMTU)) {
2459 		u32 tx_space, min_tx, min_rx;
2460 		pba = E1000_READ_REG(hw, E1000_PBA);
2461 		tx_space = pba >> 16;
2462 		pba &= 0xffff;
2463 		min_tx = (adapter->hw.mac.max_frame_size +
2464 		    sizeof(struct e1000_tx_desc) - ETHERNET_FCS_SIZE) * 2;
2465 		min_tx = roundup2(min_tx, 1024);
2466 		min_tx >>= 10;
2467 		min_rx = adapter->hw.mac.max_frame_size;
2468 		min_rx = roundup2(min_rx, 1024);
2469 		min_rx >>= 10;
2470 		if (tx_space < min_tx &&
2471 		    ((min_tx - tx_space) < pba)) {
2472 			pba = pba - (min_tx - tx_space);
2473 			/*
2474 			 * if short on rx space, rx wins
2475 			 * and must trump tx adjustment
2476 			 */
2477 			if (pba < min_rx)
2478 				pba = min_rx;
2479 		}
2480 		E1000_WRITE_REG(hw, E1000_PBA, pba);
2481 	}
2482 
2483 	if (hw->mac.type < igb_mac_min)
2484 		E1000_WRITE_REG(&adapter->hw, E1000_PBA, pba);
2485 
2486 	INIT_DEBUGOUT1("em_reset: pba=%dK",pba);
2487 
2488 	/*
2489 	 * These parameters control the automatic generation (Tx) and
2490 	 * response (Rx) to Ethernet PAUSE frames.
2491 	 * - High water mark should allow for at least two frames to be
2492 	 *   received after sending an XOFF.
2493 	 * - Low water mark works best when it is very near the high water mark.
2494 	 *   This allows the receiver to restart by sending XON when it has
2495 	 *   drained a bit. Here we use an arbitrary value of 1500 which will
2496 	 *   restart after one full frame is pulled from the buffer. There
2497 	 *   could be several smaller frames in the buffer and if so they will
2498 	 *   not trigger the XON until their total number reduces the buffer
2499 	 *   by 1500.
2500 	 * - The pause time is fairly large at 1000 x 512ns = 512 usec.
2501 	 */
2502 	rx_buffer_size = (pba & 0xffff) << 10;
2503 	hw->fc.high_water = rx_buffer_size -
2504 	    roundup2(adapter->hw.mac.max_frame_size, 1024);
2505 	hw->fc.low_water = hw->fc.high_water - 1500;
2506 
2507 	if (adapter->fc) /* locally set flow control value? */
2508 		hw->fc.requested_mode = adapter->fc;
2509 	else
2510 		hw->fc.requested_mode = e1000_fc_full;
2511 
2512 	if (hw->mac.type == e1000_80003es2lan)
2513 		hw->fc.pause_time = 0xFFFF;
2514 	else
2515 		hw->fc.pause_time = EM_FC_PAUSE_TIME;
2516 
2517 	hw->fc.send_xon = TRUE;
2518 
2519 	/* Device specific overrides/settings */
2520 	switch (hw->mac.type) {
2521 	case e1000_pchlan:
2522 		/* Workaround: no TX flow ctrl for PCH */
2523 		hw->fc.requested_mode = e1000_fc_rx_pause;
2524 		hw->fc.pause_time = 0xFFFF; /* override */
2525 		if (if_getmtu(ifp) > ETHERMTU) {
2526 			hw->fc.high_water = 0x3500;
2527 			hw->fc.low_water = 0x1500;
2528 		} else {
2529 			hw->fc.high_water = 0x5000;
2530 			hw->fc.low_water = 0x3000;
2531 		}
2532 		hw->fc.refresh_time = 0x1000;
2533 		break;
2534 	case e1000_pch2lan:
2535 	case e1000_pch_lpt:
2536 	case e1000_pch_spt:
2537 	case e1000_pch_cnp:
2538 		hw->fc.high_water = 0x5C20;
2539 		hw->fc.low_water = 0x5048;
2540 		hw->fc.pause_time = 0x0650;
2541 		hw->fc.refresh_time = 0x0400;
2542 		/* Jumbos need adjusted PBA */
2543 		if (if_getmtu(ifp) > ETHERMTU)
2544 			E1000_WRITE_REG(hw, E1000_PBA, 12);
2545 		else
2546 			E1000_WRITE_REG(hw, E1000_PBA, 26);
2547 		break;
2548 	case e1000_82575:
2549 	case e1000_82576:
2550 		/* 8-byte granularity */
2551 		hw->fc.low_water = hw->fc.high_water - 8;
2552 		break;
2553 	case e1000_82580:
2554 	case e1000_i350:
2555 	case e1000_i354:
2556 	case e1000_i210:
2557 	case e1000_i211:
2558 	case e1000_vfadapt:
2559 	case e1000_vfadapt_i350:
2560 		/* 16-byte granularity */
2561 		hw->fc.low_water = hw->fc.high_water - 16;
2562 		break;
2563 	case e1000_ich9lan:
2564 	case e1000_ich10lan:
2565 		if (if_getmtu(ifp) > ETHERMTU) {
2566 			hw->fc.high_water = 0x2800;
2567 			hw->fc.low_water = hw->fc.high_water - 8;
2568 			break;
2569 		}
2570 		/* FALLTHROUGH */
2571 	default:
2572 		if (hw->mac.type == e1000_80003es2lan)
2573 			hw->fc.pause_time = 0xFFFF;
2574 		break;
2575 	}
2576 
2577 	/* Issue a global reset */
2578 	e1000_reset_hw(hw);
2579 	if (adapter->hw.mac.type >= igb_mac_min) {
2580 		E1000_WRITE_REG(hw, E1000_WUC, 0);
2581 	} else {
2582 		E1000_WRITE_REG(hw, E1000_WUFC, 0);
2583 		em_disable_aspm(adapter);
2584 	}
2585 	if (adapter->flags & IGB_MEDIA_RESET) {
2586 		e1000_setup_init_funcs(hw, TRUE);
2587 		e1000_get_bus_info(hw);
2588 		adapter->flags &= ~IGB_MEDIA_RESET;
2589 	}
2590 	/* and a re-init */
2591 	if (e1000_init_hw(hw) < 0) {
2592 		device_printf(dev, "Hardware Initialization Failed\n");
2593 		return;
2594 	}
2595 	if (adapter->hw.mac.type >= igb_mac_min)
2596 		igb_init_dmac(adapter, pba);
2597 
2598 	E1000_WRITE_REG(hw, E1000_VET, ETHERTYPE_VLAN);
2599 	e1000_get_phy_info(hw);
2600 	e1000_check_for_link(hw);
2601 }
2602 
2603 #define RSSKEYLEN 10
2604 static void
2605 em_initialize_rss_mapping(struct adapter *adapter)
2606 {
2607 	uint8_t  rss_key[4 * RSSKEYLEN];
2608 	uint32_t reta = 0;
2609 	struct e1000_hw	*hw = &adapter->hw;
2610 	int i;
2611 
2612 	/*
2613 	 * Configure RSS key
2614 	 */
2615 	arc4rand(rss_key, sizeof(rss_key), 0);
2616 	for (i = 0; i < RSSKEYLEN; ++i) {
2617 		uint32_t rssrk = 0;
2618 
2619 		rssrk = EM_RSSRK_VAL(rss_key, i);
2620 		E1000_WRITE_REG(hw,E1000_RSSRK(i), rssrk);
2621 	}
2622 
2623 	/*
2624 	 * Configure RSS redirect table in following fashion:
2625 	 * (hash & ring_cnt_mask) == rdr_table[(hash & rdr_table_mask)]
2626 	 */
2627 	for (i = 0; i < sizeof(reta); ++i) {
2628 		uint32_t q;
2629 
2630 		q = (i % adapter->rx_num_queues) << 7;
2631 		reta |= q << (8 * i);
2632 	}
2633 
2634 	for (i = 0; i < 32; ++i)
2635 		E1000_WRITE_REG(hw, E1000_RETA(i), reta);
2636 
2637 	E1000_WRITE_REG(hw, E1000_MRQC, E1000_MRQC_RSS_ENABLE_2Q |
2638 			E1000_MRQC_RSS_FIELD_IPV4_TCP |
2639 			E1000_MRQC_RSS_FIELD_IPV4 |
2640 			E1000_MRQC_RSS_FIELD_IPV6_TCP_EX |
2641 			E1000_MRQC_RSS_FIELD_IPV6_EX |
2642 			E1000_MRQC_RSS_FIELD_IPV6);
2643 
2644 }
2645 
2646 static void
2647 igb_initialize_rss_mapping(struct adapter *adapter)
2648 {
2649 	struct e1000_hw *hw = &adapter->hw;
2650 	int i;
2651 	int queue_id;
2652 	u32 reta;
2653 	u32 rss_key[10], mrqc, shift = 0;
2654 
2655 	/* XXX? */
2656 	if (adapter->hw.mac.type == e1000_82575)
2657 		shift = 6;
2658 
2659 	/*
2660 	 * The redirection table controls which destination
2661 	 * queue each bucket redirects traffic to.
2662 	 * Each DWORD represents four queues, with the LSB
2663 	 * being the first queue in the DWORD.
2664 	 *
2665 	 * This just allocates buckets to queues using round-robin
2666 	 * allocation.
2667 	 *
2668 	 * NOTE: It Just Happens to line up with the default
2669 	 * RSS allocation method.
2670 	 */
2671 
2672 	/* Warning FM follows */
2673 	reta = 0;
2674 	for (i = 0; i < 128; i++) {
2675 #ifdef RSS
2676 		queue_id = rss_get_indirection_to_bucket(i);
2677 		/*
2678 		 * If we have more queues than buckets, we'll
2679 		 * end up mapping buckets to a subset of the
2680 		 * queues.
2681 		 *
2682 		 * If we have more buckets than queues, we'll
2683 		 * end up instead assigning multiple buckets
2684 		 * to queues.
2685 		 *
2686 		 * Both are suboptimal, but we need to handle
2687 		 * the case so we don't go out of bounds
2688 		 * indexing arrays and such.
2689 		 */
2690 		queue_id = queue_id % adapter->rx_num_queues;
2691 #else
2692 		queue_id = (i % adapter->rx_num_queues);
2693 #endif
2694 		/* Adjust if required */
2695 		queue_id = queue_id << shift;
2696 
2697 		/*
2698 		 * The low 8 bits are for hash value (n+0);
2699 		 * The next 8 bits are for hash value (n+1), etc.
2700 		 */
2701 		reta = reta >> 8;
2702 		reta = reta | ( ((uint32_t) queue_id) << 24);
2703 		if ((i & 3) == 3) {
2704 			E1000_WRITE_REG(hw, E1000_RETA(i >> 2), reta);
2705 			reta = 0;
2706 		}
2707 	}
2708 
2709 	/* Now fill in hash table */
2710 
2711 	/*
2712 	 * MRQC: Multiple Receive Queues Command
2713 	 * Set queuing to RSS control, number depends on the device.
2714 	 */
2715 	mrqc = E1000_MRQC_ENABLE_RSS_8Q;
2716 
2717 #ifdef RSS
2718 	/* XXX ew typecasting */
2719 	rss_getkey((uint8_t *) &rss_key);
2720 #else
2721 	arc4rand(&rss_key, sizeof(rss_key), 0);
2722 #endif
2723 	for (i = 0; i < 10; i++)
2724 		E1000_WRITE_REG_ARRAY(hw, E1000_RSSRK(0), i, rss_key[i]);
2725 
2726 	/*
2727 	 * Configure the RSS fields to hash upon.
2728 	 */
2729 	mrqc |= (E1000_MRQC_RSS_FIELD_IPV4 |
2730 	    E1000_MRQC_RSS_FIELD_IPV4_TCP);
2731 	mrqc |= (E1000_MRQC_RSS_FIELD_IPV6 |
2732 	    E1000_MRQC_RSS_FIELD_IPV6_TCP);
2733 	mrqc |=( E1000_MRQC_RSS_FIELD_IPV4_UDP |
2734 	    E1000_MRQC_RSS_FIELD_IPV6_UDP);
2735 	mrqc |=( E1000_MRQC_RSS_FIELD_IPV6_UDP_EX |
2736 	    E1000_MRQC_RSS_FIELD_IPV6_TCP_EX);
2737 
2738 	E1000_WRITE_REG(hw, E1000_MRQC, mrqc);
2739 }
2740 
2741 /*********************************************************************
2742  *
2743  *  Setup networking device structure and register an interface.
2744  *
2745  **********************************************************************/
2746 static int
2747 em_setup_interface(if_ctx_t ctx)
2748 {
2749 	struct ifnet *ifp = iflib_get_ifp(ctx);
2750 	struct adapter *adapter = iflib_get_softc(ctx);
2751 	if_softc_ctx_t scctx = adapter->shared;
2752 	uint64_t cap = 0;
2753 
2754 	INIT_DEBUGOUT("em_setup_interface: begin");
2755 
2756 	/* TSO parameters */
2757 	if_sethwtsomax(ifp, IP_MAXPACKET);
2758 	/* Take m_pullup(9)'s in em_xmit() w/ TSO into acount. */
2759 	if_sethwtsomaxsegcount(ifp, EM_MAX_SCATTER - 5);
2760 	if_sethwtsomaxsegsize(ifp, EM_TSO_SEG_SIZE);
2761 
2762 	/* Single Queue */
2763 	if (adapter->tx_num_queues == 1) {
2764 		if_setsendqlen(ifp, scctx->isc_ntxd[0] - 1);
2765 		if_setsendqready(ifp);
2766 	}
2767 
2768 	cap = IFCAP_HWCSUM | IFCAP_VLAN_HWCSUM | IFCAP_TSO4;
2769 	cap |= IFCAP_VLAN_HWTAGGING | IFCAP_VLAN_HWTSO | IFCAP_VLAN_MTU;
2770 
2771 	/*
2772 	 * Tell the upper layer(s) we
2773 	 * support full VLAN capability
2774 	 */
2775 	if_setifheaderlen(ifp, sizeof(struct ether_vlan_header));
2776 	if_setcapabilitiesbit(ifp, cap, 0);
2777 
2778 	/*
2779 	 * Don't turn this on by default, if vlans are
2780 	 * created on another pseudo device (eg. lagg)
2781 	 * then vlan events are not passed thru, breaking
2782 	 * operation, but with HW FILTER off it works. If
2783 	 * using vlans directly on the em driver you can
2784 	 * enable this and get full hardware tag filtering.
2785 	 */
2786 	if_setcapabilitiesbit(ifp, IFCAP_VLAN_HWFILTER,0);
2787 
2788 	/* Enable only WOL MAGIC by default */
2789 	if (adapter->wol) {
2790 		if_setcapenablebit(ifp, IFCAP_WOL_MAGIC,
2791 			    IFCAP_WOL_MCAST| IFCAP_WOL_UCAST);
2792 	} else {
2793 		if_setcapenablebit(ifp, 0, IFCAP_WOL_MAGIC |
2794 			     IFCAP_WOL_MCAST| IFCAP_WOL_UCAST);
2795 	}
2796 
2797 	/*
2798 	 * Specify the media types supported by this adapter and register
2799 	 * callbacks to update media and link information
2800 	 */
2801 	if ((adapter->hw.phy.media_type == e1000_media_type_fiber) ||
2802 	    (adapter->hw.phy.media_type == e1000_media_type_internal_serdes)) {
2803 		u_char fiber_type = IFM_1000_SX;	/* default type */
2804 
2805 		if (adapter->hw.mac.type == e1000_82545)
2806 			fiber_type = IFM_1000_LX;
2807 		ifmedia_add(adapter->media, IFM_ETHER | fiber_type | IFM_FDX, 0, NULL);
2808 		ifmedia_add(adapter->media, IFM_ETHER | fiber_type, 0, NULL);
2809 	} else {
2810 		ifmedia_add(adapter->media, IFM_ETHER | IFM_10_T, 0, NULL);
2811 		ifmedia_add(adapter->media, IFM_ETHER | IFM_10_T | IFM_FDX, 0, NULL);
2812 		ifmedia_add(adapter->media, IFM_ETHER | IFM_100_TX, 0, NULL);
2813 		ifmedia_add(adapter->media, IFM_ETHER | IFM_100_TX | IFM_FDX, 0, NULL);
2814 		if (adapter->hw.phy.type != e1000_phy_ife) {
2815 			ifmedia_add(adapter->media, IFM_ETHER | IFM_1000_T | IFM_FDX, 0, NULL);
2816 			ifmedia_add(adapter->media, IFM_ETHER | IFM_1000_T, 0, NULL);
2817 		}
2818 	}
2819 	ifmedia_add(adapter->media, IFM_ETHER | IFM_AUTO, 0, NULL);
2820 	ifmedia_set(adapter->media, IFM_ETHER | IFM_AUTO);
2821 	return (0);
2822 }
2823 
2824 static int
2825 em_if_tx_queues_alloc(if_ctx_t ctx, caddr_t *vaddrs, uint64_t *paddrs, int ntxqs, int ntxqsets)
2826 {
2827 	struct adapter *adapter = iflib_get_softc(ctx);
2828 	if_softc_ctx_t scctx = adapter->shared;
2829 	int error = E1000_SUCCESS;
2830 	struct em_tx_queue *que;
2831 	int i, j;
2832 
2833 	MPASS(adapter->tx_num_queues > 0);
2834 	MPASS(adapter->tx_num_queues == ntxqsets);
2835 
2836 	/* First allocate the top level queue structs */
2837 	if (!(adapter->tx_queues =
2838 	    (struct em_tx_queue *) malloc(sizeof(struct em_tx_queue) *
2839 	    adapter->tx_num_queues, M_DEVBUF, M_NOWAIT | M_ZERO))) {
2840 		device_printf(iflib_get_dev(ctx), "Unable to allocate queue memory\n");
2841 		return(ENOMEM);
2842 	}
2843 
2844 	for (i = 0, que = adapter->tx_queues; i < adapter->tx_num_queues; i++, que++) {
2845 		/* Set up some basics */
2846 
2847 		struct tx_ring *txr = &que->txr;
2848 		txr->adapter = que->adapter = adapter;
2849 		que->me = txr->me =  i;
2850 
2851 		/* Allocate report status array */
2852 		if (!(txr->tx_rsq = (qidx_t *) malloc(sizeof(qidx_t) * scctx->isc_ntxd[0], M_DEVBUF, M_NOWAIT | M_ZERO))) {
2853 			device_printf(iflib_get_dev(ctx), "failed to allocate rs_idxs memory\n");
2854 			error = ENOMEM;
2855 			goto fail;
2856 		}
2857 		for (j = 0; j < scctx->isc_ntxd[0]; j++)
2858 			txr->tx_rsq[j] = QIDX_INVALID;
2859 		/* get the virtual and physical address of the hardware queues */
2860 		txr->tx_base = (struct e1000_tx_desc *)vaddrs[i*ntxqs];
2861 		txr->tx_paddr = paddrs[i*ntxqs];
2862 	}
2863 
2864 	device_printf(iflib_get_dev(ctx), "allocated for %d tx_queues\n", adapter->tx_num_queues);
2865 	return (0);
2866 fail:
2867 	em_if_queues_free(ctx);
2868 	return (error);
2869 }
2870 
2871 static int
2872 em_if_rx_queues_alloc(if_ctx_t ctx, caddr_t *vaddrs, uint64_t *paddrs, int nrxqs, int nrxqsets)
2873 {
2874 	struct adapter *adapter = iflib_get_softc(ctx);
2875 	int error = E1000_SUCCESS;
2876 	struct em_rx_queue *que;
2877 	int i;
2878 
2879 	MPASS(adapter->rx_num_queues > 0);
2880 	MPASS(adapter->rx_num_queues == nrxqsets);
2881 
2882 	/* First allocate the top level queue structs */
2883 	if (!(adapter->rx_queues =
2884 	    (struct em_rx_queue *) malloc(sizeof(struct em_rx_queue) *
2885 	    adapter->rx_num_queues, M_DEVBUF, M_NOWAIT | M_ZERO))) {
2886 		device_printf(iflib_get_dev(ctx), "Unable to allocate queue memory\n");
2887 		error = ENOMEM;
2888 		goto fail;
2889 	}
2890 
2891 	for (i = 0, que = adapter->rx_queues; i < nrxqsets; i++, que++) {
2892 		/* Set up some basics */
2893 		struct rx_ring *rxr = &que->rxr;
2894 		rxr->adapter = que->adapter = adapter;
2895 		rxr->que = que;
2896 		que->me = rxr->me =  i;
2897 
2898 		/* get the virtual and physical address of the hardware queues */
2899 		rxr->rx_base = (union e1000_rx_desc_extended *)vaddrs[i*nrxqs];
2900 		rxr->rx_paddr = paddrs[i*nrxqs];
2901 	}
2902 
2903 	device_printf(iflib_get_dev(ctx), "allocated for %d rx_queues\n", adapter->rx_num_queues);
2904 
2905 	return (0);
2906 fail:
2907 	em_if_queues_free(ctx);
2908 	return (error);
2909 }
2910 
2911 static void
2912 em_if_queues_free(if_ctx_t ctx)
2913 {
2914 	struct adapter *adapter = iflib_get_softc(ctx);
2915 	struct em_tx_queue *tx_que = adapter->tx_queues;
2916 	struct em_rx_queue *rx_que = adapter->rx_queues;
2917 
2918 	if (tx_que != NULL) {
2919 		for (int i = 0; i < adapter->tx_num_queues; i++, tx_que++) {
2920 			struct tx_ring *txr = &tx_que->txr;
2921 			if (txr->tx_rsq == NULL)
2922 				break;
2923 
2924 			free(txr->tx_rsq, M_DEVBUF);
2925 			txr->tx_rsq = NULL;
2926 		}
2927 		free(adapter->tx_queues, M_DEVBUF);
2928 		adapter->tx_queues = NULL;
2929 	}
2930 
2931 	if (rx_que != NULL) {
2932 		free(adapter->rx_queues, M_DEVBUF);
2933 		adapter->rx_queues = NULL;
2934 	}
2935 
2936 	em_release_hw_control(adapter);
2937 
2938 	if (adapter->mta != NULL) {
2939 		free(adapter->mta, M_DEVBUF);
2940 	}
2941 }
2942 
2943 /*********************************************************************
2944  *
2945  *  Enable transmit unit.
2946  *
2947  **********************************************************************/
2948 static void
2949 em_initialize_transmit_unit(if_ctx_t ctx)
2950 {
2951 	struct adapter *adapter = iflib_get_softc(ctx);
2952 	if_softc_ctx_t scctx = adapter->shared;
2953 	struct em_tx_queue *que;
2954 	struct tx_ring	*txr;
2955 	struct e1000_hw	*hw = &adapter->hw;
2956 	u32 tctl, txdctl = 0, tarc, tipg = 0;
2957 
2958 	INIT_DEBUGOUT("em_initialize_transmit_unit: begin");
2959 
2960 	for (int i = 0; i < adapter->tx_num_queues; i++, txr++) {
2961 		u64 bus_addr;
2962 		caddr_t offp, endp;
2963 
2964 		que = &adapter->tx_queues[i];
2965 		txr = &que->txr;
2966 		bus_addr = txr->tx_paddr;
2967 
2968 		/* Clear checksum offload context. */
2969 		offp = (caddr_t)&txr->csum_flags;
2970 		endp = (caddr_t)(txr + 1);
2971 		bzero(offp, endp - offp);
2972 
2973 		/* Base and Len of TX Ring */
2974 		E1000_WRITE_REG(hw, E1000_TDLEN(i),
2975 		    scctx->isc_ntxd[0] * sizeof(struct e1000_tx_desc));
2976 		E1000_WRITE_REG(hw, E1000_TDBAH(i),
2977 		    (u32)(bus_addr >> 32));
2978 		E1000_WRITE_REG(hw, E1000_TDBAL(i),
2979 		    (u32)bus_addr);
2980 		/* Init the HEAD/TAIL indices */
2981 		E1000_WRITE_REG(hw, E1000_TDT(i), 0);
2982 		E1000_WRITE_REG(hw, E1000_TDH(i), 0);
2983 
2984 		HW_DEBUGOUT2("Base = %x, Length = %x\n",
2985 		    E1000_READ_REG(&adapter->hw, E1000_TDBAL(i)),
2986 		    E1000_READ_REG(&adapter->hw, E1000_TDLEN(i)));
2987 
2988 		txdctl = 0; /* clear txdctl */
2989 		txdctl |= 0x1f; /* PTHRESH */
2990 		txdctl |= 1 << 8; /* HTHRESH */
2991 		txdctl |= 1 << 16;/* WTHRESH */
2992 		txdctl |= 1 << 22; /* Reserved bit 22 must always be 1 */
2993 		txdctl |= E1000_TXDCTL_GRAN;
2994 		txdctl |= 1 << 25; /* LWTHRESH */
2995 
2996 		E1000_WRITE_REG(hw, E1000_TXDCTL(i), txdctl);
2997 	}
2998 
2999 	/* Set the default values for the Tx Inter Packet Gap timer */
3000 	switch (adapter->hw.mac.type) {
3001 	case e1000_80003es2lan:
3002 		tipg = DEFAULT_82543_TIPG_IPGR1;
3003 		tipg |= DEFAULT_80003ES2LAN_TIPG_IPGR2 <<
3004 		    E1000_TIPG_IPGR2_SHIFT;
3005 		break;
3006 	case e1000_82542:
3007 		tipg = DEFAULT_82542_TIPG_IPGT;
3008 		tipg |= DEFAULT_82542_TIPG_IPGR1 << E1000_TIPG_IPGR1_SHIFT;
3009 		tipg |= DEFAULT_82542_TIPG_IPGR2 << E1000_TIPG_IPGR2_SHIFT;
3010 		break;
3011 	default:
3012 		if ((adapter->hw.phy.media_type == e1000_media_type_fiber) ||
3013 		    (adapter->hw.phy.media_type ==
3014 		    e1000_media_type_internal_serdes))
3015 			tipg = DEFAULT_82543_TIPG_IPGT_FIBER;
3016 		else
3017 			tipg = DEFAULT_82543_TIPG_IPGT_COPPER;
3018 		tipg |= DEFAULT_82543_TIPG_IPGR1 << E1000_TIPG_IPGR1_SHIFT;
3019 		tipg |= DEFAULT_82543_TIPG_IPGR2 << E1000_TIPG_IPGR2_SHIFT;
3020 	}
3021 
3022 	E1000_WRITE_REG(&adapter->hw, E1000_TIPG, tipg);
3023 	E1000_WRITE_REG(&adapter->hw, E1000_TIDV, adapter->tx_int_delay.value);
3024 
3025 	if(adapter->hw.mac.type >= e1000_82540)
3026 		E1000_WRITE_REG(&adapter->hw, E1000_TADV,
3027 		    adapter->tx_abs_int_delay.value);
3028 
3029 	if ((adapter->hw.mac.type == e1000_82571) ||
3030 	    (adapter->hw.mac.type == e1000_82572)) {
3031 		tarc = E1000_READ_REG(&adapter->hw, E1000_TARC(0));
3032 		tarc |= TARC_SPEED_MODE_BIT;
3033 		E1000_WRITE_REG(&adapter->hw, E1000_TARC(0), tarc);
3034 	} else if (adapter->hw.mac.type == e1000_80003es2lan) {
3035 		/* errata: program both queues to unweighted RR */
3036 		tarc = E1000_READ_REG(&adapter->hw, E1000_TARC(0));
3037 		tarc |= 1;
3038 		E1000_WRITE_REG(&adapter->hw, E1000_TARC(0), tarc);
3039 		tarc = E1000_READ_REG(&adapter->hw, E1000_TARC(1));
3040 		tarc |= 1;
3041 		E1000_WRITE_REG(&adapter->hw, E1000_TARC(1), tarc);
3042 	} else if (adapter->hw.mac.type == e1000_82574) {
3043 		tarc = E1000_READ_REG(&adapter->hw, E1000_TARC(0));
3044 		tarc |= TARC_ERRATA_BIT;
3045 		if ( adapter->tx_num_queues > 1) {
3046 			tarc |= (TARC_COMPENSATION_MODE | TARC_MQ_FIX);
3047 			E1000_WRITE_REG(&adapter->hw, E1000_TARC(0), tarc);
3048 			E1000_WRITE_REG(&adapter->hw, E1000_TARC(1), tarc);
3049 		} else
3050 			E1000_WRITE_REG(&adapter->hw, E1000_TARC(0), tarc);
3051 	}
3052 
3053 	if (adapter->tx_int_delay.value > 0)
3054 		adapter->txd_cmd |= E1000_TXD_CMD_IDE;
3055 
3056 	/* Program the Transmit Control Register */
3057 	tctl = E1000_READ_REG(&adapter->hw, E1000_TCTL);
3058 	tctl &= ~E1000_TCTL_CT;
3059 	tctl |= (E1000_TCTL_PSP | E1000_TCTL_RTLC | E1000_TCTL_EN |
3060 		   (E1000_COLLISION_THRESHOLD << E1000_CT_SHIFT));
3061 
3062 	if (adapter->hw.mac.type >= e1000_82571)
3063 		tctl |= E1000_TCTL_MULR;
3064 
3065 	/* This write will effectively turn on the transmit unit. */
3066 	E1000_WRITE_REG(&adapter->hw, E1000_TCTL, tctl);
3067 
3068 	/* SPT and KBL errata workarounds */
3069 	if (hw->mac.type == e1000_pch_spt) {
3070 		u32 reg;
3071 		reg = E1000_READ_REG(hw, E1000_IOSFPC);
3072 		reg |= E1000_RCTL_RDMTS_HEX;
3073 		E1000_WRITE_REG(hw, E1000_IOSFPC, reg);
3074 		/* i218-i219 Specification Update 1.5.4.5 */
3075 		reg = E1000_READ_REG(hw, E1000_TARC(0));
3076 		reg &= ~E1000_TARC0_CB_MULTIQ_3_REQ;
3077 		reg |= E1000_TARC0_CB_MULTIQ_2_REQ;
3078 		E1000_WRITE_REG(hw, E1000_TARC(0), reg);
3079 	}
3080 }
3081 
3082 /*********************************************************************
3083  *
3084  *  Enable receive unit.
3085  *
3086  **********************************************************************/
3087 
3088 static void
3089 em_initialize_receive_unit(if_ctx_t ctx)
3090 {
3091 	struct adapter *adapter = iflib_get_softc(ctx);
3092 	if_softc_ctx_t scctx = adapter->shared;
3093 	struct ifnet *ifp = iflib_get_ifp(ctx);
3094 	struct e1000_hw	*hw = &adapter->hw;
3095 	struct em_rx_queue *que;
3096 	int i;
3097 	u32 rctl, rxcsum, rfctl;
3098 
3099 	INIT_DEBUGOUT("em_initialize_receive_units: begin");
3100 
3101 	/*
3102 	 * Make sure receives are disabled while setting
3103 	 * up the descriptor ring
3104 	 */
3105 	rctl = E1000_READ_REG(hw, E1000_RCTL);
3106 	/* Do not disable if ever enabled on this hardware */
3107 	if ((hw->mac.type != e1000_82574) && (hw->mac.type != e1000_82583))
3108 		E1000_WRITE_REG(hw, E1000_RCTL, rctl & ~E1000_RCTL_EN);
3109 
3110 	/* Setup the Receive Control Register */
3111 	rctl &= ~(3 << E1000_RCTL_MO_SHIFT);
3112 	rctl |= E1000_RCTL_EN | E1000_RCTL_BAM |
3113 	    E1000_RCTL_LBM_NO | E1000_RCTL_RDMTS_HALF |
3114 	    (hw->mac.mc_filter_type << E1000_RCTL_MO_SHIFT);
3115 
3116 	/* Do not store bad packets */
3117 	rctl &= ~E1000_RCTL_SBP;
3118 
3119 	/* Enable Long Packet receive */
3120 	if (if_getmtu(ifp) > ETHERMTU)
3121 		rctl |= E1000_RCTL_LPE;
3122 	else
3123 		rctl &= ~E1000_RCTL_LPE;
3124 
3125 	/* Strip the CRC */
3126 	if (!em_disable_crc_stripping)
3127 		rctl |= E1000_RCTL_SECRC;
3128 
3129 	if (adapter->hw.mac.type >= e1000_82540) {
3130 		E1000_WRITE_REG(&adapter->hw, E1000_RADV,
3131 			    adapter->rx_abs_int_delay.value);
3132 
3133 		/*
3134 		 * Set the interrupt throttling rate. Value is calculated
3135 		 * as DEFAULT_ITR = 1/(MAX_INTS_PER_SEC * 256ns)
3136 		 */
3137 		E1000_WRITE_REG(hw, E1000_ITR, DEFAULT_ITR);
3138 	}
3139 	E1000_WRITE_REG(&adapter->hw, E1000_RDTR,
3140 	    adapter->rx_int_delay.value);
3141 
3142 	/* Use extended rx descriptor formats */
3143 	rfctl = E1000_READ_REG(hw, E1000_RFCTL);
3144 	rfctl |= E1000_RFCTL_EXTEN;
3145 	/*
3146 	 * When using MSIX interrupts we need to throttle
3147 	 * using the EITR register (82574 only)
3148 	 */
3149 	if (hw->mac.type == e1000_82574) {
3150 		for (int i = 0; i < 4; i++)
3151 			E1000_WRITE_REG(hw, E1000_EITR_82574(i),
3152 			    DEFAULT_ITR);
3153 		/* Disable accelerated acknowledge */
3154 		rfctl |= E1000_RFCTL_ACK_DIS;
3155 	}
3156 	E1000_WRITE_REG(hw, E1000_RFCTL, rfctl);
3157 
3158 	rxcsum = E1000_READ_REG(hw, E1000_RXCSUM);
3159 	if (if_getcapenable(ifp) & IFCAP_RXCSUM &&
3160 	    adapter->hw.mac.type >= e1000_82543) {
3161 		if (adapter->tx_num_queues > 1) {
3162 			if (adapter->hw.mac.type >= igb_mac_min) {
3163 				rxcsum |= E1000_RXCSUM_PCSD;
3164 				if (hw->mac.type != e1000_82575)
3165 					rxcsum |= E1000_RXCSUM_CRCOFL;
3166 			} else
3167 				rxcsum |= E1000_RXCSUM_TUOFL |
3168 					E1000_RXCSUM_IPOFL |
3169 					E1000_RXCSUM_PCSD;
3170 		} else {
3171 			if (adapter->hw.mac.type >= igb_mac_min)
3172 				rxcsum |= E1000_RXCSUM_IPPCSE;
3173 			else
3174 				rxcsum |= E1000_RXCSUM_TUOFL | E1000_RXCSUM_IPOFL;
3175 			if (adapter->hw.mac.type > e1000_82575)
3176 				rxcsum |= E1000_RXCSUM_CRCOFL;
3177 		}
3178 	} else
3179 		rxcsum &= ~E1000_RXCSUM_TUOFL;
3180 
3181 	E1000_WRITE_REG(hw, E1000_RXCSUM, rxcsum);
3182 
3183 	if (adapter->rx_num_queues > 1) {
3184 		if (adapter->hw.mac.type >= igb_mac_min)
3185 			igb_initialize_rss_mapping(adapter);
3186 		else
3187 			em_initialize_rss_mapping(adapter);
3188 	}
3189 
3190 	/*
3191 	 * XXX TEMPORARY WORKAROUND: on some systems with 82573
3192 	 * long latencies are observed, like Lenovo X60. This
3193 	 * change eliminates the problem, but since having positive
3194 	 * values in RDTR is a known source of problems on other
3195 	 * platforms another solution is being sought.
3196 	 */
3197 	if (hw->mac.type == e1000_82573)
3198 		E1000_WRITE_REG(hw, E1000_RDTR, 0x20);
3199 
3200 	for (i = 0, que = adapter->rx_queues; i < adapter->rx_num_queues; i++, que++) {
3201 		struct rx_ring *rxr = &que->rxr;
3202 		/* Setup the Base and Length of the Rx Descriptor Ring */
3203 		u64 bus_addr = rxr->rx_paddr;
3204 #if 0
3205 		u32 rdt = adapter->rx_num_queues -1;  /* default */
3206 #endif
3207 
3208 		E1000_WRITE_REG(hw, E1000_RDLEN(i),
3209 		    scctx->isc_nrxd[0] * sizeof(union e1000_rx_desc_extended));
3210 		E1000_WRITE_REG(hw, E1000_RDBAH(i), (u32)(bus_addr >> 32));
3211 		E1000_WRITE_REG(hw, E1000_RDBAL(i), (u32)bus_addr);
3212 		/* Setup the Head and Tail Descriptor Pointers */
3213 		E1000_WRITE_REG(hw, E1000_RDH(i), 0);
3214 		E1000_WRITE_REG(hw, E1000_RDT(i), 0);
3215 	}
3216 
3217 	/*
3218 	 * Set PTHRESH for improved jumbo performance
3219 	 * According to 10.2.5.11 of Intel 82574 Datasheet,
3220 	 * RXDCTL(1) is written whenever RXDCTL(0) is written.
3221 	 * Only write to RXDCTL(1) if there is a need for different
3222 	 * settings.
3223 	 */
3224 
3225 	if (((adapter->hw.mac.type == e1000_ich9lan) ||
3226 	    (adapter->hw.mac.type == e1000_pch2lan) ||
3227 	    (adapter->hw.mac.type == e1000_ich10lan)) &&
3228 	    (if_getmtu(ifp) > ETHERMTU)) {
3229 		u32 rxdctl = E1000_READ_REG(hw, E1000_RXDCTL(0));
3230 		E1000_WRITE_REG(hw, E1000_RXDCTL(0), rxdctl | 3);
3231 	} else if (adapter->hw.mac.type == e1000_82574) {
3232 		for (int i = 0; i < adapter->rx_num_queues; i++) {
3233 			u32 rxdctl = E1000_READ_REG(hw, E1000_RXDCTL(i));
3234 			rxdctl |= 0x20; /* PTHRESH */
3235 			rxdctl |= 4 << 8; /* HTHRESH */
3236 			rxdctl |= 4 << 16;/* WTHRESH */
3237 			rxdctl |= 1 << 24; /* Switch to granularity */
3238 			E1000_WRITE_REG(hw, E1000_RXDCTL(i), rxdctl);
3239 		}
3240 	} else if (adapter->hw.mac.type >= igb_mac_min) {
3241 		u32 psize, srrctl = 0;
3242 
3243 		if (if_getmtu(ifp) > ETHERMTU) {
3244 			/* Set maximum packet len */
3245 			if (adapter->rx_mbuf_sz <= 4096) {
3246 				srrctl |= 4096 >> E1000_SRRCTL_BSIZEPKT_SHIFT;
3247 				rctl |= E1000_RCTL_SZ_4096 | E1000_RCTL_BSEX;
3248 			} else if (adapter->rx_mbuf_sz > 4096) {
3249 				srrctl |= 8192 >> E1000_SRRCTL_BSIZEPKT_SHIFT;
3250 				rctl |= E1000_RCTL_SZ_8192 | E1000_RCTL_BSEX;
3251 			}
3252 			psize = scctx->isc_max_frame_size;
3253 			/* are we on a vlan? */
3254 			if (ifp->if_vlantrunk != NULL)
3255 				psize += VLAN_TAG_SIZE;
3256 			E1000_WRITE_REG(&adapter->hw, E1000_RLPML, psize);
3257 		} else {
3258 			srrctl |= 2048 >> E1000_SRRCTL_BSIZEPKT_SHIFT;
3259 			rctl |= E1000_RCTL_SZ_2048;
3260 		}
3261 
3262 		/*
3263 		 * If TX flow control is disabled and there's >1 queue defined,
3264 		 * enable DROP.
3265 		 *
3266 		 * This drops frames rather than hanging the RX MAC for all queues.
3267 		 */
3268 		if ((adapter->rx_num_queues > 1) &&
3269 		    (adapter->fc == e1000_fc_none ||
3270 		     adapter->fc == e1000_fc_rx_pause)) {
3271 			srrctl |= E1000_SRRCTL_DROP_EN;
3272 		}
3273 			/* Setup the Base and Length of the Rx Descriptor Rings */
3274 		for (i = 0, que = adapter->rx_queues; i < adapter->rx_num_queues; i++, que++) {
3275 			struct rx_ring *rxr = &que->rxr;
3276 			u64 bus_addr = rxr->rx_paddr;
3277 			u32 rxdctl;
3278 
3279 #ifdef notyet
3280 			/* Configure for header split? -- ignore for now */
3281 			rxr->hdr_split = igb_header_split;
3282 #else
3283 			srrctl |= E1000_SRRCTL_DESCTYPE_ADV_ONEBUF;
3284 #endif
3285 
3286 			E1000_WRITE_REG(hw, E1000_RDLEN(i),
3287 					scctx->isc_nrxd[0] * sizeof(struct e1000_rx_desc));
3288 			E1000_WRITE_REG(hw, E1000_RDBAH(i),
3289 					(uint32_t)(bus_addr >> 32));
3290 			E1000_WRITE_REG(hw, E1000_RDBAL(i),
3291 					(uint32_t)bus_addr);
3292 			E1000_WRITE_REG(hw, E1000_SRRCTL(i), srrctl);
3293 			/* Enable this Queue */
3294 			rxdctl = E1000_READ_REG(hw, E1000_RXDCTL(i));
3295 			rxdctl |= E1000_RXDCTL_QUEUE_ENABLE;
3296 			rxdctl &= 0xFFF00000;
3297 			rxdctl |= IGB_RX_PTHRESH;
3298 			rxdctl |= IGB_RX_HTHRESH << 8;
3299 			rxdctl |= IGB_RX_WTHRESH << 16;
3300 			E1000_WRITE_REG(hw, E1000_RXDCTL(i), rxdctl);
3301 		}
3302 	} else if (adapter->hw.mac.type >= e1000_pch2lan) {
3303 		if (if_getmtu(ifp) > ETHERMTU)
3304 			e1000_lv_jumbo_workaround_ich8lan(hw, TRUE);
3305 		else
3306 			e1000_lv_jumbo_workaround_ich8lan(hw, FALSE);
3307 	}
3308 
3309 	/* Make sure VLAN Filters are off */
3310 	rctl &= ~E1000_RCTL_VFE;
3311 
3312 	if (adapter->hw.mac.type < igb_mac_min) {
3313 		if (adapter->rx_mbuf_sz == MCLBYTES)
3314 			rctl |= E1000_RCTL_SZ_2048;
3315 		else if (adapter->rx_mbuf_sz == MJUMPAGESIZE)
3316 			rctl |= E1000_RCTL_SZ_4096 | E1000_RCTL_BSEX;
3317 		else if (adapter->rx_mbuf_sz > MJUMPAGESIZE)
3318 			rctl |= E1000_RCTL_SZ_8192 | E1000_RCTL_BSEX;
3319 
3320 		/* ensure we clear use DTYPE of 00 here */
3321 		rctl &= ~0x00000C00;
3322 	}
3323 
3324 	/* Write out the settings */
3325 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
3326 
3327 	return;
3328 }
3329 
3330 static void
3331 em_if_vlan_register(if_ctx_t ctx, u16 vtag)
3332 {
3333 	struct adapter *adapter = iflib_get_softc(ctx);
3334 	u32 index, bit;
3335 
3336 	index = (vtag >> 5) & 0x7F;
3337 	bit = vtag & 0x1F;
3338 	adapter->shadow_vfta[index] |= (1 << bit);
3339 	++adapter->num_vlans;
3340 }
3341 
3342 static void
3343 em_if_vlan_unregister(if_ctx_t ctx, u16 vtag)
3344 {
3345 	struct adapter *adapter = iflib_get_softc(ctx);
3346 	u32 index, bit;
3347 
3348 	index = (vtag >> 5) & 0x7F;
3349 	bit = vtag & 0x1F;
3350 	adapter->shadow_vfta[index] &= ~(1 << bit);
3351 	--adapter->num_vlans;
3352 }
3353 
3354 static void
3355 em_setup_vlan_hw_support(struct adapter *adapter)
3356 {
3357 	struct e1000_hw *hw = &adapter->hw;
3358 	u32 reg;
3359 
3360 	/*
3361 	 * We get here thru init_locked, meaning
3362 	 * a soft reset, this has already cleared
3363 	 * the VFTA and other state, so if there
3364 	 * have been no vlan's registered do nothing.
3365 	 */
3366 	if (adapter->num_vlans == 0)
3367 		return;
3368 
3369 	/*
3370 	 * A soft reset zero's out the VFTA, so
3371 	 * we need to repopulate it now.
3372 	 */
3373 	for (int i = 0; i < EM_VFTA_SIZE; i++)
3374 		if (adapter->shadow_vfta[i] != 0)
3375 			E1000_WRITE_REG_ARRAY(hw, E1000_VFTA,
3376 			    i, adapter->shadow_vfta[i]);
3377 
3378 	reg = E1000_READ_REG(hw, E1000_CTRL);
3379 	reg |= E1000_CTRL_VME;
3380 	E1000_WRITE_REG(hw, E1000_CTRL, reg);
3381 
3382 	/* Enable the Filter Table */
3383 	reg = E1000_READ_REG(hw, E1000_RCTL);
3384 	reg &= ~E1000_RCTL_CFIEN;
3385 	reg |= E1000_RCTL_VFE;
3386 	E1000_WRITE_REG(hw, E1000_RCTL, reg);
3387 }
3388 
3389 static void
3390 em_if_enable_intr(if_ctx_t ctx)
3391 {
3392 	struct adapter *adapter = iflib_get_softc(ctx);
3393 	struct e1000_hw *hw = &adapter->hw;
3394 	u32 ims_mask = IMS_ENABLE_MASK;
3395 
3396 	if (hw->mac.type == e1000_82574) {
3397 		E1000_WRITE_REG(hw, EM_EIAC, EM_MSIX_MASK);
3398 		ims_mask |= adapter->ims;
3399 	} else if (adapter->intr_type == IFLIB_INTR_MSIX && hw->mac.type >= igb_mac_min)  {
3400 		u32 mask = (adapter->que_mask | adapter->link_mask);
3401 
3402 		E1000_WRITE_REG(&adapter->hw, E1000_EIAC, mask);
3403 		E1000_WRITE_REG(&adapter->hw, E1000_EIAM, mask);
3404 		E1000_WRITE_REG(&adapter->hw, E1000_EIMS, mask);
3405 		ims_mask = E1000_IMS_LSC;
3406 	}
3407 
3408 	E1000_WRITE_REG(hw, E1000_IMS, ims_mask);
3409 }
3410 
3411 static void
3412 em_if_disable_intr(if_ctx_t ctx)
3413 {
3414 	struct adapter *adapter = iflib_get_softc(ctx);
3415 	struct e1000_hw *hw = &adapter->hw;
3416 
3417 	if (adapter->intr_type == IFLIB_INTR_MSIX) {
3418 		if (hw->mac.type >= igb_mac_min)
3419 			E1000_WRITE_REG(&adapter->hw, E1000_EIMC, ~0);
3420 		E1000_WRITE_REG(&adapter->hw, E1000_EIAC, 0);
3421 	}
3422 	E1000_WRITE_REG(&adapter->hw, E1000_IMC, 0xffffffff);
3423 }
3424 
3425 /*
3426  * Bit of a misnomer, what this really means is
3427  * to enable OS management of the system... aka
3428  * to disable special hardware management features
3429  */
3430 static void
3431 em_init_manageability(struct adapter *adapter)
3432 {
3433 	/* A shared code workaround */
3434 #define E1000_82542_MANC2H E1000_MANC2H
3435 	if (adapter->has_manage) {
3436 		int manc2h = E1000_READ_REG(&adapter->hw, E1000_MANC2H);
3437 		int manc = E1000_READ_REG(&adapter->hw, E1000_MANC);
3438 
3439 		/* disable hardware interception of ARP */
3440 		manc &= ~(E1000_MANC_ARP_EN);
3441 
3442 		/* enable receiving management packets to the host */
3443 		manc |= E1000_MANC_EN_MNG2HOST;
3444 #define E1000_MNG2HOST_PORT_623 (1 << 5)
3445 #define E1000_MNG2HOST_PORT_664 (1 << 6)
3446 		manc2h |= E1000_MNG2HOST_PORT_623;
3447 		manc2h |= E1000_MNG2HOST_PORT_664;
3448 		E1000_WRITE_REG(&adapter->hw, E1000_MANC2H, manc2h);
3449 		E1000_WRITE_REG(&adapter->hw, E1000_MANC, manc);
3450 	}
3451 }
3452 
3453 /*
3454  * Give control back to hardware management
3455  * controller if there is one.
3456  */
3457 static void
3458 em_release_manageability(struct adapter *adapter)
3459 {
3460 	if (adapter->has_manage) {
3461 		int manc = E1000_READ_REG(&adapter->hw, E1000_MANC);
3462 
3463 		/* re-enable hardware interception of ARP */
3464 		manc |= E1000_MANC_ARP_EN;
3465 		manc &= ~E1000_MANC_EN_MNG2HOST;
3466 
3467 		E1000_WRITE_REG(&adapter->hw, E1000_MANC, manc);
3468 	}
3469 }
3470 
3471 /*
3472  * em_get_hw_control sets the {CTRL_EXT|FWSM}:DRV_LOAD bit.
3473  * For ASF and Pass Through versions of f/w this means
3474  * that the driver is loaded. For AMT version type f/w
3475  * this means that the network i/f is open.
3476  */
3477 static void
3478 em_get_hw_control(struct adapter *adapter)
3479 {
3480 	u32 ctrl_ext, swsm;
3481 
3482 	if (adapter->vf_ifp)
3483 		return;
3484 
3485 	if (adapter->hw.mac.type == e1000_82573) {
3486 		swsm = E1000_READ_REG(&adapter->hw, E1000_SWSM);
3487 		E1000_WRITE_REG(&adapter->hw, E1000_SWSM,
3488 		    swsm | E1000_SWSM_DRV_LOAD);
3489 		return;
3490 	}
3491 	/* else */
3492 	ctrl_ext = E1000_READ_REG(&adapter->hw, E1000_CTRL_EXT);
3493 	E1000_WRITE_REG(&adapter->hw, E1000_CTRL_EXT,
3494 	    ctrl_ext | E1000_CTRL_EXT_DRV_LOAD);
3495 }
3496 
3497 /*
3498  * em_release_hw_control resets {CTRL_EXT|FWSM}:DRV_LOAD bit.
3499  * For ASF and Pass Through versions of f/w this means that
3500  * the driver is no longer loaded. For AMT versions of the
3501  * f/w this means that the network i/f is closed.
3502  */
3503 static void
3504 em_release_hw_control(struct adapter *adapter)
3505 {
3506 	u32 ctrl_ext, swsm;
3507 
3508 	if (!adapter->has_manage)
3509 		return;
3510 
3511 	if (adapter->hw.mac.type == e1000_82573) {
3512 		swsm = E1000_READ_REG(&adapter->hw, E1000_SWSM);
3513 		E1000_WRITE_REG(&adapter->hw, E1000_SWSM,
3514 		    swsm & ~E1000_SWSM_DRV_LOAD);
3515 		return;
3516 	}
3517 	/* else */
3518 	ctrl_ext = E1000_READ_REG(&adapter->hw, E1000_CTRL_EXT);
3519 	E1000_WRITE_REG(&adapter->hw, E1000_CTRL_EXT,
3520 	    ctrl_ext & ~E1000_CTRL_EXT_DRV_LOAD);
3521 	return;
3522 }
3523 
3524 static int
3525 em_is_valid_ether_addr(u8 *addr)
3526 {
3527 	char zero_addr[6] = { 0, 0, 0, 0, 0, 0 };
3528 
3529 	if ((addr[0] & 1) || (!bcmp(addr, zero_addr, ETHER_ADDR_LEN))) {
3530 		return (FALSE);
3531 	}
3532 
3533 	return (TRUE);
3534 }
3535 
3536 /*
3537 ** Parse the interface capabilities with regard
3538 ** to both system management and wake-on-lan for
3539 ** later use.
3540 */
3541 static void
3542 em_get_wakeup(if_ctx_t ctx)
3543 {
3544 	struct adapter *adapter = iflib_get_softc(ctx);
3545 	device_t dev = iflib_get_dev(ctx);
3546 	u16 eeprom_data = 0, device_id, apme_mask;
3547 
3548 	adapter->has_manage = e1000_enable_mng_pass_thru(&adapter->hw);
3549 	apme_mask = EM_EEPROM_APME;
3550 
3551 	switch (adapter->hw.mac.type) {
3552 	case e1000_82542:
3553 	case e1000_82543:
3554 		break;
3555 	case e1000_82544:
3556 		e1000_read_nvm(&adapter->hw,
3557 		    NVM_INIT_CONTROL2_REG, 1, &eeprom_data);
3558 		apme_mask = EM_82544_APME;
3559 		break;
3560 	case e1000_82546:
3561 	case e1000_82546_rev_3:
3562 		if (adapter->hw.bus.func == 1) {
3563 			e1000_read_nvm(&adapter->hw,
3564 			    NVM_INIT_CONTROL3_PORT_B, 1, &eeprom_data);
3565 			break;
3566 		} else
3567 			e1000_read_nvm(&adapter->hw,
3568 			    NVM_INIT_CONTROL3_PORT_A, 1, &eeprom_data);
3569 		break;
3570 	case e1000_82573:
3571 	case e1000_82583:
3572 		adapter->has_amt = TRUE;
3573 		/* FALLTHROUGH */
3574 	case e1000_82571:
3575 	case e1000_82572:
3576 	case e1000_80003es2lan:
3577 		if (adapter->hw.bus.func == 1) {
3578 			e1000_read_nvm(&adapter->hw,
3579 			    NVM_INIT_CONTROL3_PORT_B, 1, &eeprom_data);
3580 			break;
3581 		} else
3582 			e1000_read_nvm(&adapter->hw,
3583 			    NVM_INIT_CONTROL3_PORT_A, 1, &eeprom_data);
3584 		break;
3585 	case e1000_ich8lan:
3586 	case e1000_ich9lan:
3587 	case e1000_ich10lan:
3588 	case e1000_pchlan:
3589 	case e1000_pch2lan:
3590 	case e1000_pch_lpt:
3591 	case e1000_pch_spt:
3592 	case e1000_82575:	/* listing all igb devices */
3593 	case e1000_82576:
3594 	case e1000_82580:
3595 	case e1000_i350:
3596 	case e1000_i354:
3597 	case e1000_i210:
3598 	case e1000_i211:
3599 	case e1000_vfadapt:
3600 	case e1000_vfadapt_i350:
3601 		apme_mask = E1000_WUC_APME;
3602 		adapter->has_amt = TRUE;
3603 		eeprom_data = E1000_READ_REG(&adapter->hw, E1000_WUC);
3604 		break;
3605 	default:
3606 		e1000_read_nvm(&adapter->hw,
3607 		    NVM_INIT_CONTROL3_PORT_A, 1, &eeprom_data);
3608 		break;
3609 	}
3610 	if (eeprom_data & apme_mask)
3611 		adapter->wol = (E1000_WUFC_MAG | E1000_WUFC_MC);
3612 	/*
3613 	 * We have the eeprom settings, now apply the special cases
3614 	 * where the eeprom may be wrong or the board won't support
3615 	 * wake on lan on a particular port
3616 	 */
3617 	device_id = pci_get_device(dev);
3618 	switch (device_id) {
3619 	case E1000_DEV_ID_82546GB_PCIE:
3620 		adapter->wol = 0;
3621 		break;
3622 	case E1000_DEV_ID_82546EB_FIBER:
3623 	case E1000_DEV_ID_82546GB_FIBER:
3624 		/* Wake events only supported on port A for dual fiber
3625 		 * regardless of eeprom setting */
3626 		if (E1000_READ_REG(&adapter->hw, E1000_STATUS) &
3627 		    E1000_STATUS_FUNC_1)
3628 			adapter->wol = 0;
3629 		break;
3630 	case E1000_DEV_ID_82546GB_QUAD_COPPER_KSP3:
3631 		/* if quad port adapter, disable WoL on all but port A */
3632 		if (global_quad_port_a != 0)
3633 			adapter->wol = 0;
3634 		/* Reset for multiple quad port adapters */
3635 		if (++global_quad_port_a == 4)
3636 			global_quad_port_a = 0;
3637 		break;
3638 	case E1000_DEV_ID_82571EB_FIBER:
3639 		/* Wake events only supported on port A for dual fiber
3640 		 * regardless of eeprom setting */
3641 		if (E1000_READ_REG(&adapter->hw, E1000_STATUS) &
3642 		    E1000_STATUS_FUNC_1)
3643 			adapter->wol = 0;
3644 		break;
3645 	case E1000_DEV_ID_82571EB_QUAD_COPPER:
3646 	case E1000_DEV_ID_82571EB_QUAD_FIBER:
3647 	case E1000_DEV_ID_82571EB_QUAD_COPPER_LP:
3648 		/* if quad port adapter, disable WoL on all but port A */
3649 		if (global_quad_port_a != 0)
3650 			adapter->wol = 0;
3651 		/* Reset for multiple quad port adapters */
3652 		if (++global_quad_port_a == 4)
3653 			global_quad_port_a = 0;
3654 		break;
3655 	}
3656 	return;
3657 }
3658 
3659 
3660 /*
3661  * Enable PCI Wake On Lan capability
3662  */
3663 static void
3664 em_enable_wakeup(if_ctx_t ctx)
3665 {
3666 	struct adapter *adapter = iflib_get_softc(ctx);
3667 	device_t dev = iflib_get_dev(ctx);
3668 	if_t ifp = iflib_get_ifp(ctx);
3669 	int error = 0;
3670 	u32 pmc, ctrl, ctrl_ext, rctl;
3671 	u16 status;
3672 
3673 	if (pci_find_cap(dev, PCIY_PMG, &pmc) != 0)
3674 		return;
3675 
3676 	/*
3677 	 * Determine type of Wakeup: note that wol
3678 	 * is set with all bits on by default.
3679 	 */
3680 	if ((if_getcapenable(ifp) & IFCAP_WOL_MAGIC) == 0)
3681 		adapter->wol &= ~E1000_WUFC_MAG;
3682 
3683 	if ((if_getcapenable(ifp) & IFCAP_WOL_UCAST) == 0)
3684 		adapter->wol &= ~E1000_WUFC_EX;
3685 
3686 	if ((if_getcapenable(ifp) & IFCAP_WOL_MCAST) == 0)
3687 		adapter->wol &= ~E1000_WUFC_MC;
3688 	else {
3689 		rctl = E1000_READ_REG(&adapter->hw, E1000_RCTL);
3690 		rctl |= E1000_RCTL_MPE;
3691 		E1000_WRITE_REG(&adapter->hw, E1000_RCTL, rctl);
3692 	}
3693 
3694 	if (!(adapter->wol & (E1000_WUFC_EX | E1000_WUFC_MAG | E1000_WUFC_MC)))
3695 		goto pme;
3696 
3697 	/* Advertise the wakeup capability */
3698 	ctrl = E1000_READ_REG(&adapter->hw, E1000_CTRL);
3699 	ctrl |= (E1000_CTRL_SWDPIN2 | E1000_CTRL_SWDPIN3);
3700 	E1000_WRITE_REG(&adapter->hw, E1000_CTRL, ctrl);
3701 
3702 	/* Keep the laser running on Fiber adapters */
3703 	if (adapter->hw.phy.media_type == e1000_media_type_fiber ||
3704 	    adapter->hw.phy.media_type == e1000_media_type_internal_serdes) {
3705 		ctrl_ext = E1000_READ_REG(&adapter->hw, E1000_CTRL_EXT);
3706 		ctrl_ext |= E1000_CTRL_EXT_SDP3_DATA;
3707 		E1000_WRITE_REG(&adapter->hw, E1000_CTRL_EXT, ctrl_ext);
3708 	}
3709 
3710 	if ((adapter->hw.mac.type == e1000_ich8lan) ||
3711 	    (adapter->hw.mac.type == e1000_pchlan) ||
3712 	    (adapter->hw.mac.type == e1000_ich9lan) ||
3713 	    (adapter->hw.mac.type == e1000_ich10lan))
3714 		e1000_suspend_workarounds_ich8lan(&adapter->hw);
3715 
3716 	if ( adapter->hw.mac.type >= e1000_pchlan) {
3717 		error = em_enable_phy_wakeup(adapter);
3718 		if (error)
3719 			goto pme;
3720 	} else {
3721 		/* Enable wakeup by the MAC */
3722 		E1000_WRITE_REG(&adapter->hw, E1000_WUC, E1000_WUC_PME_EN);
3723 		E1000_WRITE_REG(&adapter->hw, E1000_WUFC, adapter->wol);
3724 	}
3725 
3726 	if (adapter->hw.phy.type == e1000_phy_igp_3)
3727 		e1000_igp3_phy_powerdown_workaround_ich8lan(&adapter->hw);
3728 
3729 pme:
3730 	status = pci_read_config(dev, pmc + PCIR_POWER_STATUS, 2);
3731 	status &= ~(PCIM_PSTAT_PME | PCIM_PSTAT_PMEENABLE);
3732 	if (!error && (if_getcapenable(ifp) & IFCAP_WOL))
3733 		status |= PCIM_PSTAT_PME | PCIM_PSTAT_PMEENABLE;
3734 	pci_write_config(dev, pmc + PCIR_POWER_STATUS, status, 2);
3735 
3736 	return;
3737 }
3738 
3739 /*
3740  * WOL in the newer chipset interfaces (pchlan)
3741  * require thing to be copied into the phy
3742  */
3743 static int
3744 em_enable_phy_wakeup(struct adapter *adapter)
3745 {
3746 	struct e1000_hw *hw = &adapter->hw;
3747 	u32 mreg, ret = 0;
3748 	u16 preg;
3749 
3750 	/* copy MAC RARs to PHY RARs */
3751 	e1000_copy_rx_addrs_to_phy_ich8lan(hw);
3752 
3753 	/* copy MAC MTA to PHY MTA */
3754 	for (int i = 0; i < adapter->hw.mac.mta_reg_count; i++) {
3755 		mreg = E1000_READ_REG_ARRAY(hw, E1000_MTA, i);
3756 		e1000_write_phy_reg(hw, BM_MTA(i), (u16)(mreg & 0xFFFF));
3757 		e1000_write_phy_reg(hw, BM_MTA(i) + 1,
3758 		    (u16)((mreg >> 16) & 0xFFFF));
3759 	}
3760 
3761 	/* configure PHY Rx Control register */
3762 	e1000_read_phy_reg(&adapter->hw, BM_RCTL, &preg);
3763 	mreg = E1000_READ_REG(hw, E1000_RCTL);
3764 	if (mreg & E1000_RCTL_UPE)
3765 		preg |= BM_RCTL_UPE;
3766 	if (mreg & E1000_RCTL_MPE)
3767 		preg |= BM_RCTL_MPE;
3768 	preg &= ~(BM_RCTL_MO_MASK);
3769 	if (mreg & E1000_RCTL_MO_3)
3770 		preg |= (((mreg & E1000_RCTL_MO_3) >> E1000_RCTL_MO_SHIFT)
3771 				<< BM_RCTL_MO_SHIFT);
3772 	if (mreg & E1000_RCTL_BAM)
3773 		preg |= BM_RCTL_BAM;
3774 	if (mreg & E1000_RCTL_PMCF)
3775 		preg |= BM_RCTL_PMCF;
3776 	mreg = E1000_READ_REG(hw, E1000_CTRL);
3777 	if (mreg & E1000_CTRL_RFCE)
3778 		preg |= BM_RCTL_RFCE;
3779 	e1000_write_phy_reg(&adapter->hw, BM_RCTL, preg);
3780 
3781 	/* enable PHY wakeup in MAC register */
3782 	E1000_WRITE_REG(hw, E1000_WUC,
3783 	    E1000_WUC_PHY_WAKE | E1000_WUC_PME_EN | E1000_WUC_APME);
3784 	E1000_WRITE_REG(hw, E1000_WUFC, adapter->wol);
3785 
3786 	/* configure and enable PHY wakeup in PHY registers */
3787 	e1000_write_phy_reg(&adapter->hw, BM_WUFC, adapter->wol);
3788 	e1000_write_phy_reg(&adapter->hw, BM_WUC, E1000_WUC_PME_EN);
3789 
3790 	/* activate PHY wakeup */
3791 	ret = hw->phy.ops.acquire(hw);
3792 	if (ret) {
3793 		printf("Could not acquire PHY\n");
3794 		return ret;
3795 	}
3796 	e1000_write_phy_reg_mdic(hw, IGP01E1000_PHY_PAGE_SELECT,
3797 	                         (BM_WUC_ENABLE_PAGE << IGP_PAGE_SHIFT));
3798 	ret = e1000_read_phy_reg_mdic(hw, BM_WUC_ENABLE_REG, &preg);
3799 	if (ret) {
3800 		printf("Could not read PHY page 769\n");
3801 		goto out;
3802 	}
3803 	preg |= BM_WUC_ENABLE_BIT | BM_WUC_HOST_WU_BIT;
3804 	ret = e1000_write_phy_reg_mdic(hw, BM_WUC_ENABLE_REG, preg);
3805 	if (ret)
3806 		printf("Could not set PHY Host Wakeup bit\n");
3807 out:
3808 	hw->phy.ops.release(hw);
3809 
3810 	return ret;
3811 }
3812 
3813 static void
3814 em_if_led_func(if_ctx_t ctx, int onoff)
3815 {
3816 	struct adapter *adapter = iflib_get_softc(ctx);
3817 
3818 	if (onoff) {
3819 		e1000_setup_led(&adapter->hw);
3820 		e1000_led_on(&adapter->hw);
3821 	} else {
3822 		e1000_led_off(&adapter->hw);
3823 		e1000_cleanup_led(&adapter->hw);
3824 	}
3825 }
3826 
3827 /*
3828  * Disable the L0S and L1 LINK states
3829  */
3830 static void
3831 em_disable_aspm(struct adapter *adapter)
3832 {
3833 	int base, reg;
3834 	u16 link_cap,link_ctrl;
3835 	device_t dev = adapter->dev;
3836 
3837 	switch (adapter->hw.mac.type) {
3838 	case e1000_82573:
3839 	case e1000_82574:
3840 	case e1000_82583:
3841 		break;
3842 	default:
3843 		return;
3844 	}
3845 	if (pci_find_cap(dev, PCIY_EXPRESS, &base) != 0)
3846 		return;
3847 	reg = base + PCIER_LINK_CAP;
3848 	link_cap = pci_read_config(dev, reg, 2);
3849 	if ((link_cap & PCIEM_LINK_CAP_ASPM) == 0)
3850 		return;
3851 	reg = base + PCIER_LINK_CTL;
3852 	link_ctrl = pci_read_config(dev, reg, 2);
3853 	link_ctrl &= ~PCIEM_LINK_CTL_ASPMC;
3854 	pci_write_config(dev, reg, link_ctrl, 2);
3855 	return;
3856 }
3857 
3858 /**********************************************************************
3859  *
3860  *  Update the board statistics counters.
3861  *
3862  **********************************************************************/
3863 static void
3864 em_update_stats_counters(struct adapter *adapter)
3865 {
3866 
3867 	if(adapter->hw.phy.media_type == e1000_media_type_copper ||
3868 	   (E1000_READ_REG(&adapter->hw, E1000_STATUS) & E1000_STATUS_LU)) {
3869 		adapter->stats.symerrs += E1000_READ_REG(&adapter->hw, E1000_SYMERRS);
3870 		adapter->stats.sec += E1000_READ_REG(&adapter->hw, E1000_SEC);
3871 	}
3872 	adapter->stats.crcerrs += E1000_READ_REG(&adapter->hw, E1000_CRCERRS);
3873 	adapter->stats.mpc += E1000_READ_REG(&adapter->hw, E1000_MPC);
3874 	adapter->stats.scc += E1000_READ_REG(&adapter->hw, E1000_SCC);
3875 	adapter->stats.ecol += E1000_READ_REG(&adapter->hw, E1000_ECOL);
3876 
3877 	adapter->stats.mcc += E1000_READ_REG(&adapter->hw, E1000_MCC);
3878 	adapter->stats.latecol += E1000_READ_REG(&adapter->hw, E1000_LATECOL);
3879 	adapter->stats.colc += E1000_READ_REG(&adapter->hw, E1000_COLC);
3880 	adapter->stats.dc += E1000_READ_REG(&adapter->hw, E1000_DC);
3881 	adapter->stats.rlec += E1000_READ_REG(&adapter->hw, E1000_RLEC);
3882 	adapter->stats.xonrxc += E1000_READ_REG(&adapter->hw, E1000_XONRXC);
3883 	adapter->stats.xontxc += E1000_READ_REG(&adapter->hw, E1000_XONTXC);
3884 	adapter->stats.xoffrxc += E1000_READ_REG(&adapter->hw, E1000_XOFFRXC);
3885 	/*
3886 	 ** For watchdog management we need to know if we have been
3887 	 ** paused during the last interval, so capture that here.
3888 	*/
3889 	adapter->shared->isc_pause_frames = adapter->stats.xoffrxc;
3890 	adapter->stats.xofftxc += E1000_READ_REG(&adapter->hw, E1000_XOFFTXC);
3891 	adapter->stats.fcruc += E1000_READ_REG(&adapter->hw, E1000_FCRUC);
3892 	adapter->stats.prc64 += E1000_READ_REG(&adapter->hw, E1000_PRC64);
3893 	adapter->stats.prc127 += E1000_READ_REG(&adapter->hw, E1000_PRC127);
3894 	adapter->stats.prc255 += E1000_READ_REG(&adapter->hw, E1000_PRC255);
3895 	adapter->stats.prc511 += E1000_READ_REG(&adapter->hw, E1000_PRC511);
3896 	adapter->stats.prc1023 += E1000_READ_REG(&adapter->hw, E1000_PRC1023);
3897 	adapter->stats.prc1522 += E1000_READ_REG(&adapter->hw, E1000_PRC1522);
3898 	adapter->stats.gprc += E1000_READ_REG(&adapter->hw, E1000_GPRC);
3899 	adapter->stats.bprc += E1000_READ_REG(&adapter->hw, E1000_BPRC);
3900 	adapter->stats.mprc += E1000_READ_REG(&adapter->hw, E1000_MPRC);
3901 	adapter->stats.gptc += E1000_READ_REG(&adapter->hw, E1000_GPTC);
3902 
3903 	/* For the 64-bit byte counters the low dword must be read first. */
3904 	/* Both registers clear on the read of the high dword */
3905 
3906 	adapter->stats.gorc += E1000_READ_REG(&adapter->hw, E1000_GORCL) +
3907 	    ((u64)E1000_READ_REG(&adapter->hw, E1000_GORCH) << 32);
3908 	adapter->stats.gotc += E1000_READ_REG(&adapter->hw, E1000_GOTCL) +
3909 	    ((u64)E1000_READ_REG(&adapter->hw, E1000_GOTCH) << 32);
3910 
3911 	adapter->stats.rnbc += E1000_READ_REG(&adapter->hw, E1000_RNBC);
3912 	adapter->stats.ruc += E1000_READ_REG(&adapter->hw, E1000_RUC);
3913 	adapter->stats.rfc += E1000_READ_REG(&adapter->hw, E1000_RFC);
3914 	adapter->stats.roc += E1000_READ_REG(&adapter->hw, E1000_ROC);
3915 	adapter->stats.rjc += E1000_READ_REG(&adapter->hw, E1000_RJC);
3916 
3917 	adapter->stats.tor += E1000_READ_REG(&adapter->hw, E1000_TORH);
3918 	adapter->stats.tot += E1000_READ_REG(&adapter->hw, E1000_TOTH);
3919 
3920 	adapter->stats.tpr += E1000_READ_REG(&adapter->hw, E1000_TPR);
3921 	adapter->stats.tpt += E1000_READ_REG(&adapter->hw, E1000_TPT);
3922 	adapter->stats.ptc64 += E1000_READ_REG(&adapter->hw, E1000_PTC64);
3923 	adapter->stats.ptc127 += E1000_READ_REG(&adapter->hw, E1000_PTC127);
3924 	adapter->stats.ptc255 += E1000_READ_REG(&adapter->hw, E1000_PTC255);
3925 	adapter->stats.ptc511 += E1000_READ_REG(&adapter->hw, E1000_PTC511);
3926 	adapter->stats.ptc1023 += E1000_READ_REG(&adapter->hw, E1000_PTC1023);
3927 	adapter->stats.ptc1522 += E1000_READ_REG(&adapter->hw, E1000_PTC1522);
3928 	adapter->stats.mptc += E1000_READ_REG(&adapter->hw, E1000_MPTC);
3929 	adapter->stats.bptc += E1000_READ_REG(&adapter->hw, E1000_BPTC);
3930 
3931 	/* Interrupt Counts */
3932 
3933 	adapter->stats.iac += E1000_READ_REG(&adapter->hw, E1000_IAC);
3934 	adapter->stats.icrxptc += E1000_READ_REG(&adapter->hw, E1000_ICRXPTC);
3935 	adapter->stats.icrxatc += E1000_READ_REG(&adapter->hw, E1000_ICRXATC);
3936 	adapter->stats.ictxptc += E1000_READ_REG(&adapter->hw, E1000_ICTXPTC);
3937 	adapter->stats.ictxatc += E1000_READ_REG(&adapter->hw, E1000_ICTXATC);
3938 	adapter->stats.ictxqec += E1000_READ_REG(&adapter->hw, E1000_ICTXQEC);
3939 	adapter->stats.ictxqmtc += E1000_READ_REG(&adapter->hw, E1000_ICTXQMTC);
3940 	adapter->stats.icrxdmtc += E1000_READ_REG(&adapter->hw, E1000_ICRXDMTC);
3941 	adapter->stats.icrxoc += E1000_READ_REG(&adapter->hw, E1000_ICRXOC);
3942 
3943 	if (adapter->hw.mac.type >= e1000_82543) {
3944 		adapter->stats.algnerrc +=
3945 		E1000_READ_REG(&adapter->hw, E1000_ALGNERRC);
3946 		adapter->stats.rxerrc +=
3947 		E1000_READ_REG(&adapter->hw, E1000_RXERRC);
3948 		adapter->stats.tncrs +=
3949 		E1000_READ_REG(&adapter->hw, E1000_TNCRS);
3950 		adapter->stats.cexterr +=
3951 		E1000_READ_REG(&adapter->hw, E1000_CEXTERR);
3952 		adapter->stats.tsctc +=
3953 		E1000_READ_REG(&adapter->hw, E1000_TSCTC);
3954 		adapter->stats.tsctfc +=
3955 		E1000_READ_REG(&adapter->hw, E1000_TSCTFC);
3956 	}
3957 }
3958 
3959 static uint64_t
3960 em_if_get_counter(if_ctx_t ctx, ift_counter cnt)
3961 {
3962 	struct adapter *adapter = iflib_get_softc(ctx);
3963 	struct ifnet *ifp = iflib_get_ifp(ctx);
3964 
3965 	switch (cnt) {
3966 	case IFCOUNTER_COLLISIONS:
3967 		return (adapter->stats.colc);
3968 	case IFCOUNTER_IERRORS:
3969 		return (adapter->dropped_pkts + adapter->stats.rxerrc +
3970 		    adapter->stats.crcerrs + adapter->stats.algnerrc +
3971 		    adapter->stats.ruc + adapter->stats.roc +
3972 		    adapter->stats.mpc + adapter->stats.cexterr);
3973 	case IFCOUNTER_OERRORS:
3974 		return (adapter->stats.ecol + adapter->stats.latecol +
3975 		    adapter->watchdog_events);
3976 	default:
3977 		return (if_get_counter_default(ifp, cnt));
3978 	}
3979 }
3980 
3981 /* Export a single 32-bit register via a read-only sysctl. */
3982 static int
3983 em_sysctl_reg_handler(SYSCTL_HANDLER_ARGS)
3984 {
3985 	struct adapter *adapter;
3986 	u_int val;
3987 
3988 	adapter = oidp->oid_arg1;
3989 	val = E1000_READ_REG(&adapter->hw, oidp->oid_arg2);
3990 	return (sysctl_handle_int(oidp, &val, 0, req));
3991 }
3992 
3993 /*
3994  * Add sysctl variables, one per statistic, to the system.
3995  */
3996 static void
3997 em_add_hw_stats(struct adapter *adapter)
3998 {
3999 	device_t dev = iflib_get_dev(adapter->ctx);
4000 	struct em_tx_queue *tx_que = adapter->tx_queues;
4001 	struct em_rx_queue *rx_que = adapter->rx_queues;
4002 
4003 	struct sysctl_ctx_list *ctx = device_get_sysctl_ctx(dev);
4004 	struct sysctl_oid *tree = device_get_sysctl_tree(dev);
4005 	struct sysctl_oid_list *child = SYSCTL_CHILDREN(tree);
4006 	struct e1000_hw_stats *stats = &adapter->stats;
4007 
4008 	struct sysctl_oid *stat_node, *queue_node, *int_node;
4009 	struct sysctl_oid_list *stat_list, *queue_list, *int_list;
4010 
4011 #define QUEUE_NAME_LEN 32
4012 	char namebuf[QUEUE_NAME_LEN];
4013 
4014 	/* Driver Statistics */
4015 	SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "dropped",
4016 			CTLFLAG_RD, &adapter->dropped_pkts,
4017 			"Driver dropped packets");
4018 	SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "link_irq",
4019 			CTLFLAG_RD, &adapter->link_irq,
4020 			"Link MSIX IRQ Handled");
4021 	SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "mbuf_defrag_fail",
4022 			 CTLFLAG_RD, &adapter->mbuf_defrag_failed,
4023 			 "Defragmenting mbuf chain failed");
4024 	SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "tx_dma_fail",
4025 			CTLFLAG_RD, &adapter->no_tx_dma_setup,
4026 			"Driver tx dma failure in xmit");
4027 	SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "rx_overruns",
4028 			CTLFLAG_RD, &adapter->rx_overruns,
4029 			"RX overruns");
4030 	SYSCTL_ADD_ULONG(ctx, child, OID_AUTO, "watchdog_timeouts",
4031 			CTLFLAG_RD, &adapter->watchdog_events,
4032 			"Watchdog timeouts");
4033 	SYSCTL_ADD_PROC(ctx, child, OID_AUTO, "device_control",
4034 			CTLTYPE_UINT | CTLFLAG_RD, adapter, E1000_CTRL,
4035 			em_sysctl_reg_handler, "IU",
4036 			"Device Control Register");
4037 	SYSCTL_ADD_PROC(ctx, child, OID_AUTO, "rx_control",
4038 			CTLTYPE_UINT | CTLFLAG_RD, adapter, E1000_RCTL,
4039 			em_sysctl_reg_handler, "IU",
4040 			"Receiver Control Register");
4041 	SYSCTL_ADD_UINT(ctx, child, OID_AUTO, "fc_high_water",
4042 			CTLFLAG_RD, &adapter->hw.fc.high_water, 0,
4043 			"Flow Control High Watermark");
4044 	SYSCTL_ADD_UINT(ctx, child, OID_AUTO, "fc_low_water",
4045 			CTLFLAG_RD, &adapter->hw.fc.low_water, 0,
4046 			"Flow Control Low Watermark");
4047 
4048 	for (int i = 0; i < adapter->tx_num_queues; i++, tx_que++) {
4049 		struct tx_ring *txr = &tx_que->txr;
4050 		snprintf(namebuf, QUEUE_NAME_LEN, "queue_tx_%d", i);
4051 		queue_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, namebuf,
4052 					    CTLFLAG_RD, NULL, "TX Queue Name");
4053 		queue_list = SYSCTL_CHILDREN(queue_node);
4054 
4055 		SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "txd_head",
4056 				CTLTYPE_UINT | CTLFLAG_RD, adapter,
4057 				E1000_TDH(txr->me),
4058 				em_sysctl_reg_handler, "IU",
4059 				"Transmit Descriptor Head");
4060 		SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "txd_tail",
4061 				CTLTYPE_UINT | CTLFLAG_RD, adapter,
4062 				E1000_TDT(txr->me),
4063 				em_sysctl_reg_handler, "IU",
4064 				"Transmit Descriptor Tail");
4065 		SYSCTL_ADD_ULONG(ctx, queue_list, OID_AUTO, "tx_irq",
4066 				CTLFLAG_RD, &txr->tx_irq,
4067 				"Queue MSI-X Transmit Interrupts");
4068 	}
4069 
4070 	for (int j = 0; j < adapter->rx_num_queues; j++, rx_que++) {
4071 		struct rx_ring *rxr = &rx_que->rxr;
4072 		snprintf(namebuf, QUEUE_NAME_LEN, "queue_rx_%d", j);
4073 		queue_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, namebuf,
4074 					    CTLFLAG_RD, NULL, "RX Queue Name");
4075 		queue_list = SYSCTL_CHILDREN(queue_node);
4076 
4077 		SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "rxd_head",
4078 				CTLTYPE_UINT | CTLFLAG_RD, adapter,
4079 				E1000_RDH(rxr->me),
4080 				em_sysctl_reg_handler, "IU",
4081 				"Receive Descriptor Head");
4082 		SYSCTL_ADD_PROC(ctx, queue_list, OID_AUTO, "rxd_tail",
4083 				CTLTYPE_UINT | CTLFLAG_RD, adapter,
4084 				E1000_RDT(rxr->me),
4085 				em_sysctl_reg_handler, "IU",
4086 				"Receive Descriptor Tail");
4087 		SYSCTL_ADD_ULONG(ctx, queue_list, OID_AUTO, "rx_irq",
4088 				CTLFLAG_RD, &rxr->rx_irq,
4089 				"Queue MSI-X Receive Interrupts");
4090 	}
4091 
4092 	/* MAC stats get their own sub node */
4093 
4094 	stat_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, "mac_stats",
4095 				    CTLFLAG_RD, NULL, "Statistics");
4096 	stat_list = SYSCTL_CHILDREN(stat_node);
4097 
4098 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "excess_coll",
4099 			CTLFLAG_RD, &stats->ecol,
4100 			"Excessive collisions");
4101 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "single_coll",
4102 			CTLFLAG_RD, &stats->scc,
4103 			"Single collisions");
4104 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "multiple_coll",
4105 			CTLFLAG_RD, &stats->mcc,
4106 			"Multiple collisions");
4107 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "late_coll",
4108 			CTLFLAG_RD, &stats->latecol,
4109 			"Late collisions");
4110 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "collision_count",
4111 			CTLFLAG_RD, &stats->colc,
4112 			"Collision Count");
4113 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "symbol_errors",
4114 			CTLFLAG_RD, &adapter->stats.symerrs,
4115 			"Symbol Errors");
4116 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "sequence_errors",
4117 			CTLFLAG_RD, &adapter->stats.sec,
4118 			"Sequence Errors");
4119 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "defer_count",
4120 			CTLFLAG_RD, &adapter->stats.dc,
4121 			"Defer Count");
4122 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "missed_packets",
4123 			CTLFLAG_RD, &adapter->stats.mpc,
4124 			"Missed Packets");
4125 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_no_buff",
4126 			CTLFLAG_RD, &adapter->stats.rnbc,
4127 			"Receive No Buffers");
4128 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_undersize",
4129 			CTLFLAG_RD, &adapter->stats.ruc,
4130 			"Receive Undersize");
4131 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_fragmented",
4132 			CTLFLAG_RD, &adapter->stats.rfc,
4133 			"Fragmented Packets Received ");
4134 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_oversize",
4135 			CTLFLAG_RD, &adapter->stats.roc,
4136 			"Oversized Packets Received");
4137 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_jabber",
4138 			CTLFLAG_RD, &adapter->stats.rjc,
4139 			"Recevied Jabber");
4140 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "recv_errs",
4141 			CTLFLAG_RD, &adapter->stats.rxerrc,
4142 			"Receive Errors");
4143 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "crc_errs",
4144 			CTLFLAG_RD, &adapter->stats.crcerrs,
4145 			"CRC errors");
4146 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "alignment_errs",
4147 			CTLFLAG_RD, &adapter->stats.algnerrc,
4148 			"Alignment Errors");
4149 	/* On 82575 these are collision counts */
4150 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "coll_ext_errs",
4151 			CTLFLAG_RD, &adapter->stats.cexterr,
4152 			"Collision/Carrier extension errors");
4153 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xon_recvd",
4154 			CTLFLAG_RD, &adapter->stats.xonrxc,
4155 			"XON Received");
4156 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xon_txd",
4157 			CTLFLAG_RD, &adapter->stats.xontxc,
4158 			"XON Transmitted");
4159 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xoff_recvd",
4160 			CTLFLAG_RD, &adapter->stats.xoffrxc,
4161 			"XOFF Received");
4162 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "xoff_txd",
4163 			CTLFLAG_RD, &adapter->stats.xofftxc,
4164 			"XOFF Transmitted");
4165 
4166 	/* Packet Reception Stats */
4167 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "total_pkts_recvd",
4168 			CTLFLAG_RD, &adapter->stats.tpr,
4169 			"Total Packets Received ");
4170 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_pkts_recvd",
4171 			CTLFLAG_RD, &adapter->stats.gprc,
4172 			"Good Packets Received");
4173 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "bcast_pkts_recvd",
4174 			CTLFLAG_RD, &adapter->stats.bprc,
4175 			"Broadcast Packets Received");
4176 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "mcast_pkts_recvd",
4177 			CTLFLAG_RD, &adapter->stats.mprc,
4178 			"Multicast Packets Received");
4179 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_64",
4180 			CTLFLAG_RD, &adapter->stats.prc64,
4181 			"64 byte frames received ");
4182 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_65_127",
4183 			CTLFLAG_RD, &adapter->stats.prc127,
4184 			"65-127 byte frames received");
4185 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_128_255",
4186 			CTLFLAG_RD, &adapter->stats.prc255,
4187 			"128-255 byte frames received");
4188 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_256_511",
4189 			CTLFLAG_RD, &adapter->stats.prc511,
4190 			"256-511 byte frames received");
4191 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_512_1023",
4192 			CTLFLAG_RD, &adapter->stats.prc1023,
4193 			"512-1023 byte frames received");
4194 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "rx_frames_1024_1522",
4195 			CTLFLAG_RD, &adapter->stats.prc1522,
4196 			"1023-1522 byte frames received");
4197 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_octets_recvd",
4198 			CTLFLAG_RD, &adapter->stats.gorc,
4199 			"Good Octets Received");
4200 
4201 	/* Packet Transmission Stats */
4202 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_octets_txd",
4203 			CTLFLAG_RD, &adapter->stats.gotc,
4204 			"Good Octets Transmitted");
4205 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "total_pkts_txd",
4206 			CTLFLAG_RD, &adapter->stats.tpt,
4207 			"Total Packets Transmitted");
4208 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "good_pkts_txd",
4209 			CTLFLAG_RD, &adapter->stats.gptc,
4210 			"Good Packets Transmitted");
4211 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "bcast_pkts_txd",
4212 			CTLFLAG_RD, &adapter->stats.bptc,
4213 			"Broadcast Packets Transmitted");
4214 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "mcast_pkts_txd",
4215 			CTLFLAG_RD, &adapter->stats.mptc,
4216 			"Multicast Packets Transmitted");
4217 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_64",
4218 			CTLFLAG_RD, &adapter->stats.ptc64,
4219 			"64 byte frames transmitted ");
4220 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_65_127",
4221 			CTLFLAG_RD, &adapter->stats.ptc127,
4222 			"65-127 byte frames transmitted");
4223 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_128_255",
4224 			CTLFLAG_RD, &adapter->stats.ptc255,
4225 			"128-255 byte frames transmitted");
4226 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_256_511",
4227 			CTLFLAG_RD, &adapter->stats.ptc511,
4228 			"256-511 byte frames transmitted");
4229 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_512_1023",
4230 			CTLFLAG_RD, &adapter->stats.ptc1023,
4231 			"512-1023 byte frames transmitted");
4232 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tx_frames_1024_1522",
4233 			CTLFLAG_RD, &adapter->stats.ptc1522,
4234 			"1024-1522 byte frames transmitted");
4235 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tso_txd",
4236 			CTLFLAG_RD, &adapter->stats.tsctc,
4237 			"TSO Contexts Transmitted");
4238 	SYSCTL_ADD_UQUAD(ctx, stat_list, OID_AUTO, "tso_ctx_fail",
4239 			CTLFLAG_RD, &adapter->stats.tsctfc,
4240 			"TSO Contexts Failed");
4241 
4242 
4243 	/* Interrupt Stats */
4244 
4245 	int_node = SYSCTL_ADD_NODE(ctx, child, OID_AUTO, "interrupts",
4246 				    CTLFLAG_RD, NULL, "Interrupt Statistics");
4247 	int_list = SYSCTL_CHILDREN(int_node);
4248 
4249 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "asserts",
4250 			CTLFLAG_RD, &adapter->stats.iac,
4251 			"Interrupt Assertion Count");
4252 
4253 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "rx_pkt_timer",
4254 			CTLFLAG_RD, &adapter->stats.icrxptc,
4255 			"Interrupt Cause Rx Pkt Timer Expire Count");
4256 
4257 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "rx_abs_timer",
4258 			CTLFLAG_RD, &adapter->stats.icrxatc,
4259 			"Interrupt Cause Rx Abs Timer Expire Count");
4260 
4261 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "tx_pkt_timer",
4262 			CTLFLAG_RD, &adapter->stats.ictxptc,
4263 			"Interrupt Cause Tx Pkt Timer Expire Count");
4264 
4265 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "tx_abs_timer",
4266 			CTLFLAG_RD, &adapter->stats.ictxatc,
4267 			"Interrupt Cause Tx Abs Timer Expire Count");
4268 
4269 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "tx_queue_empty",
4270 			CTLFLAG_RD, &adapter->stats.ictxqec,
4271 			"Interrupt Cause Tx Queue Empty Count");
4272 
4273 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "tx_queue_min_thresh",
4274 			CTLFLAG_RD, &adapter->stats.ictxqmtc,
4275 			"Interrupt Cause Tx Queue Min Thresh Count");
4276 
4277 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "rx_desc_min_thresh",
4278 			CTLFLAG_RD, &adapter->stats.icrxdmtc,
4279 			"Interrupt Cause Rx Desc Min Thresh Count");
4280 
4281 	SYSCTL_ADD_UQUAD(ctx, int_list, OID_AUTO, "rx_overrun",
4282 			CTLFLAG_RD, &adapter->stats.icrxoc,
4283 			"Interrupt Cause Receiver Overrun Count");
4284 }
4285 
4286 /**********************************************************************
4287  *
4288  *  This routine provides a way to dump out the adapter eeprom,
4289  *  often a useful debug/service tool. This only dumps the first
4290  *  32 words, stuff that matters is in that extent.
4291  *
4292  **********************************************************************/
4293 static int
4294 em_sysctl_nvm_info(SYSCTL_HANDLER_ARGS)
4295 {
4296 	struct adapter *adapter = (struct adapter *)arg1;
4297 	int error;
4298 	int result;
4299 
4300 	result = -1;
4301 	error = sysctl_handle_int(oidp, &result, 0, req);
4302 
4303 	if (error || !req->newptr)
4304 		return (error);
4305 
4306 	/*
4307 	 * This value will cause a hex dump of the
4308 	 * first 32 16-bit words of the EEPROM to
4309 	 * the screen.
4310 	 */
4311 	if (result == 1)
4312 		em_print_nvm_info(adapter);
4313 
4314 	return (error);
4315 }
4316 
4317 static void
4318 em_print_nvm_info(struct adapter *adapter)
4319 {
4320 	u16 eeprom_data;
4321 	int i, j, row = 0;
4322 
4323 	/* Its a bit crude, but it gets the job done */
4324 	printf("\nInterface EEPROM Dump:\n");
4325 	printf("Offset\n0x0000  ");
4326 	for (i = 0, j = 0; i < 32; i++, j++) {
4327 		if (j == 8) { /* Make the offset block */
4328 			j = 0; ++row;
4329 			printf("\n0x00%x0  ",row);
4330 		}
4331 		e1000_read_nvm(&adapter->hw, i, 1, &eeprom_data);
4332 		printf("%04x ", eeprom_data);
4333 	}
4334 	printf("\n");
4335 }
4336 
4337 static int
4338 em_sysctl_int_delay(SYSCTL_HANDLER_ARGS)
4339 {
4340 	struct em_int_delay_info *info;
4341 	struct adapter *adapter;
4342 	u32 regval;
4343 	int error, usecs, ticks;
4344 
4345 	info = (struct em_int_delay_info *) arg1;
4346 	usecs = info->value;
4347 	error = sysctl_handle_int(oidp, &usecs, 0, req);
4348 	if (error != 0 || req->newptr == NULL)
4349 		return (error);
4350 	if (usecs < 0 || usecs > EM_TICKS_TO_USECS(65535))
4351 		return (EINVAL);
4352 	info->value = usecs;
4353 	ticks = EM_USECS_TO_TICKS(usecs);
4354 	if (info->offset == E1000_ITR)	/* units are 256ns here */
4355 		ticks *= 4;
4356 
4357 	adapter = info->adapter;
4358 
4359 	regval = E1000_READ_OFFSET(&adapter->hw, info->offset);
4360 	regval = (regval & ~0xffff) | (ticks & 0xffff);
4361 	/* Handle a few special cases. */
4362 	switch (info->offset) {
4363 	case E1000_RDTR:
4364 		break;
4365 	case E1000_TIDV:
4366 		if (ticks == 0) {
4367 			adapter->txd_cmd &= ~E1000_TXD_CMD_IDE;
4368 			/* Don't write 0 into the TIDV register. */
4369 			regval++;
4370 		} else
4371 			adapter->txd_cmd |= E1000_TXD_CMD_IDE;
4372 		break;
4373 	}
4374 	E1000_WRITE_OFFSET(&adapter->hw, info->offset, regval);
4375 	return (0);
4376 }
4377 
4378 static void
4379 em_add_int_delay_sysctl(struct adapter *adapter, const char *name,
4380 	const char *description, struct em_int_delay_info *info,
4381 	int offset, int value)
4382 {
4383 	info->adapter = adapter;
4384 	info->offset = offset;
4385 	info->value = value;
4386 	SYSCTL_ADD_PROC(device_get_sysctl_ctx(adapter->dev),
4387 	    SYSCTL_CHILDREN(device_get_sysctl_tree(adapter->dev)),
4388 	    OID_AUTO, name, CTLTYPE_INT|CTLFLAG_RW,
4389 	    info, 0, em_sysctl_int_delay, "I", description);
4390 }
4391 
4392 /*
4393  * Set flow control using sysctl:
4394  * Flow control values:
4395  *      0 - off
4396  *      1 - rx pause
4397  *      2 - tx pause
4398  *      3 - full
4399  */
4400 static int
4401 em_set_flowcntl(SYSCTL_HANDLER_ARGS)
4402 {
4403 	int error;
4404 	static int input = 3; /* default is full */
4405 	struct adapter	*adapter = (struct adapter *) arg1;
4406 
4407 	error = sysctl_handle_int(oidp, &input, 0, req);
4408 
4409 	if ((error) || (req->newptr == NULL))
4410 		return (error);
4411 
4412 	if (input == adapter->fc) /* no change? */
4413 		return (error);
4414 
4415 	switch (input) {
4416 	case e1000_fc_rx_pause:
4417 	case e1000_fc_tx_pause:
4418 	case e1000_fc_full:
4419 	case e1000_fc_none:
4420 		adapter->hw.fc.requested_mode = input;
4421 		adapter->fc = input;
4422 		break;
4423 	default:
4424 		/* Do nothing */
4425 		return (error);
4426 	}
4427 
4428 	adapter->hw.fc.current_mode = adapter->hw.fc.requested_mode;
4429 	e1000_force_mac_fc(&adapter->hw);
4430 	return (error);
4431 }
4432 
4433 /*
4434  * Manage Energy Efficient Ethernet:
4435  * Control values:
4436  *     0/1 - enabled/disabled
4437  */
4438 static int
4439 em_sysctl_eee(SYSCTL_HANDLER_ARGS)
4440 {
4441 	struct adapter *adapter = (struct adapter *) arg1;
4442 	int error, value;
4443 
4444 	value = adapter->hw.dev_spec.ich8lan.eee_disable;
4445 	error = sysctl_handle_int(oidp, &value, 0, req);
4446 	if (error || req->newptr == NULL)
4447 		return (error);
4448 	adapter->hw.dev_spec.ich8lan.eee_disable = (value != 0);
4449 	em_if_init(adapter->ctx);
4450 
4451 	return (0);
4452 }
4453 
4454 static int
4455 em_sysctl_debug_info(SYSCTL_HANDLER_ARGS)
4456 {
4457 	struct adapter *adapter;
4458 	int error;
4459 	int result;
4460 
4461 	result = -1;
4462 	error = sysctl_handle_int(oidp, &result, 0, req);
4463 
4464 	if (error || !req->newptr)
4465 		return (error);
4466 
4467 	if (result == 1) {
4468 		adapter = (struct adapter *) arg1;
4469 		em_print_debug_info(adapter);
4470 	}
4471 
4472 	return (error);
4473 }
4474 
4475 static int
4476 em_get_rs(SYSCTL_HANDLER_ARGS)
4477 {
4478 	struct adapter *adapter = (struct adapter *) arg1;
4479 	int error;
4480 	int result;
4481 
4482 	result = 0;
4483 	error = sysctl_handle_int(oidp, &result, 0, req);
4484 
4485 	if (error || !req->newptr || result != 1)
4486 		return (error);
4487 	em_dump_rs(adapter);
4488 
4489 	return (error);
4490 }
4491 
4492 static void
4493 em_if_debug(if_ctx_t ctx)
4494 {
4495 	em_dump_rs(iflib_get_softc(ctx));
4496 }
4497 
4498 /*
4499  * This routine is meant to be fluid, add whatever is
4500  * needed for debugging a problem.  -jfv
4501  */
4502 static void
4503 em_print_debug_info(struct adapter *adapter)
4504 {
4505 	device_t dev = iflib_get_dev(adapter->ctx);
4506 	struct ifnet *ifp = iflib_get_ifp(adapter->ctx);
4507 	struct tx_ring *txr = &adapter->tx_queues->txr;
4508 	struct rx_ring *rxr = &adapter->rx_queues->rxr;
4509 
4510 	if (if_getdrvflags(ifp) & IFF_DRV_RUNNING)
4511 		printf("Interface is RUNNING ");
4512 	else
4513 		printf("Interface is NOT RUNNING\n");
4514 
4515 	if (if_getdrvflags(ifp) & IFF_DRV_OACTIVE)
4516 		printf("and INACTIVE\n");
4517 	else
4518 		printf("and ACTIVE\n");
4519 
4520 	for (int i = 0; i < adapter->tx_num_queues; i++, txr++) {
4521 		device_printf(dev, "TX Queue %d ------\n", i);
4522 		device_printf(dev, "hw tdh = %d, hw tdt = %d\n",
4523 			E1000_READ_REG(&adapter->hw, E1000_TDH(i)),
4524 			E1000_READ_REG(&adapter->hw, E1000_TDT(i)));
4525 
4526 	}
4527 	for (int j=0; j < adapter->rx_num_queues; j++, rxr++) {
4528 		device_printf(dev, "RX Queue %d ------\n", j);
4529 		device_printf(dev, "hw rdh = %d, hw rdt = %d\n",
4530 			E1000_READ_REG(&adapter->hw, E1000_RDH(j)),
4531 			E1000_READ_REG(&adapter->hw, E1000_RDT(j)));
4532 	}
4533 }
4534 
4535 /*
4536  * 82574 only:
4537  * Write a new value to the EEPROM increasing the number of MSIX
4538  * vectors from 3 to 5, for proper multiqueue support.
4539  */
4540 static void
4541 em_enable_vectors_82574(if_ctx_t ctx)
4542 {
4543 	struct adapter *adapter = iflib_get_softc(ctx);
4544 	struct e1000_hw *hw = &adapter->hw;
4545 	device_t dev = iflib_get_dev(ctx);
4546 	u16 edata;
4547 
4548 	e1000_read_nvm(hw, EM_NVM_PCIE_CTRL, 1, &edata);
4549 	printf("Current cap: %#06x\n", edata);
4550 	if (((edata & EM_NVM_MSIX_N_MASK) >> EM_NVM_MSIX_N_SHIFT) != 4) {
4551 		device_printf(dev, "Writing to eeprom: increasing "
4552 		    "reported MSIX vectors from 3 to 5...\n");
4553 		edata &= ~(EM_NVM_MSIX_N_MASK);
4554 		edata |= 4 << EM_NVM_MSIX_N_SHIFT;
4555 		e1000_write_nvm(hw, EM_NVM_PCIE_CTRL, 1, &edata);
4556 		e1000_update_nvm_checksum(hw);
4557 		device_printf(dev, "Writing to eeprom: done\n");
4558 	}
4559 }
4560